Top Banner
석사 학위논문 Master's Thesis 복합재료 손상의 무기저 진단 정량화 Reference-free Damage Detection, Localization and Quantification in Composite (Lim, Hyung Jin) 건설 환경공학과 Department of Civil and Environmental Engineering KAIST 2011
52

복합재료 손상의 무기저 진단 및 정량화 - KAISTssslab.kaist.ac.kr/article/pdf/Thesis_2.pdf · 1 Declaration of Ethical Conduct in Research: I, as a graduate student

Jun 28, 2020

Download

Documents

dariahiddleston
Welcome message from author
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Page 1: 복합재료 손상의 무기저 진단 및 정량화 - KAISTssslab.kaist.ac.kr/article/pdf/Thesis_2.pdf · 1 Declaration of Ethical Conduct in Research: I, as a graduate student

석사 학위논문 Masters Thesis

복합재료 손상의 무기저 진단 및 정량화

Reference-free Damage Detection Localization

and Quantification in Composite

임 형 진 (林 亨 眞 Lim Hyung Jin)

건설 및 환경공학과

Department of Civil and Environmental Engineering

KAIST

2011

복합재료 손상의 무기저 진단 및 정량화

Reference-free Damage Detection Localization

and Quantification in Composite

Reference-free Damage Detection Localization

and Quantification in Composite

Advisor Professor Hoon Sohn by

Hyung Jin Lim

Department of Civil and Environmental Engineering

KAIST

A thesis submitted to the faculty of KAIST in partial fulfillment of the re-quirements for the degree of Master of Science and Engineering The study was conducted in accordance with Code of Research Ethics1

2011 6 20

Approved by

Professor Hoon Sohn

[Major Advisor]

1 Declaration of Ethical Conduct in Research I as a graduate student of KAIST hereby declare that I have not committed any acts that may damage the credibility of my research These include but are not limited to falsi-fication thesis written by someone else distortion of research findings or plagiarism I affirm that my thesis contains honest conclusions based on my own careful research under the guidance of my thesis advisor

복합재료 손상의 무기저 진단 및 정량화

임 형 진

위 논문은 한국과학기술원 석사학위논문으로

학위논문심사위원회에서 심사 통과하였음

2011 년 5 월 30 일

심사위원장

심사위원

심사위원

손 훈 (인)

이 행 기 (인)

김 천 곤 (인)

MCE

20094258

임 형 진 Lim Hyung Jin Reference-free Damage Detection Localization and

Quantification in Composite 복합재료 손상의 무기저 진단 및 정량화 Department

of Civil and Environmental Engineering 2011 41 p Advisor Prof Sohn Hoon

ABSTRACT

A reference-free damage detection localization and quantification technique in a multi

layered composite plate is proposed It is reported that when A0 Lamb wave mode propa-

gates the damage time delay attenuation and multiple reflections are occurred in the dam-

age area In this thesis additional reflected waves from the damage due to multiple reflec-

tions are used as a feature for characterizing the damage First the damage is identified by

detecting the additional reflected waves from the damage area Then the location of the

damage is estimated using combined pitch-catch and pulse-echo methods Finally the dam-

age is quantified using a damage index defined as the ratio between the physical length of

the damage and the wavelength of the A0 Lamb wave mode The matching pursuit method

is utilized for analyzing the obtained signal by decomposing overlapped waves and estimat-

ing the parameters of each wave such as arrival time driving frequency amplitude and so

on The proposed technique has an advantage that the damage can be characterized without

reference data obtained from pristine condition of the monitored structure within a single

path The feasibility of the proposed technique is verified through numerical simulations

and experimental tests In numerical simulation different damage models (length material

properties and location) are investigated In experimental tests a Teflon tape inserted and

impact-induced damage are characterized under temperature variations The results are

suggesting a reliable damage characterization in composites

Keywords Reference-free PZT transducers Composites Finite element method Matching

pursuit method

i

TABLE OF CONTENTS

ABSTRACT middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot i

TABLE OF CONTENTS middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot iii

LIST OF TABLES middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot iv

LIST OF FIGURESmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot v

CHAPTER 1 INTRODUCTION

11 Motivation middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 1

12 Literature review middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 1

13 Objective and uniqueness middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 3

14 Organization middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 4

CHAPTER 2 INTERACTION OF LAMB WAVE WITH DAMAGE

21 A0 Lamb wave mode delay and attenuation in the time domain middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 5

22 Multiple reflections in the damage area middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 6

CHAPTER 3 THEORETICAL DEVELOPMENT

31 Overview middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 7

32 Damage index (m-value) middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 8

33 Damage detection algorithm (Level 1) middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 10

34 Damage quantification algorithm (Level 3) middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 11

35 Damage localization algorithm (Level 2amp3) middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 11

36 Matching pursuit method considering the dispersion effect of Lamb wave

middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 13

CHAPTER 4 NUMERICAL SIMULATION

41 Composite modeling methodmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 15

42 Damage modeling method middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 16

43 Numerical simulation description middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 17

44 Numerical simulation results middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 19

ii

CHAPTER 5 EXPERIMENTAL VERIFICATION

51 Overview middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 21

511 Specimen middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 22

512 Data acquisition system and PZT middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 22

513 Impact test middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 23

514 Temperature effect test middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 24

52 Teflon inserted damage middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 25

53 Impact-induced damagemiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 26

54 Temperature effect test middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 27

541 Teflon inserted damage middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 27

542 Impact-induced damage middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 29

CHAPTER 6 CONCLUDING REMARKS

61 Executive summary middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 31

62 Suggestion for further study middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 32

REFERCENCES middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 33

SUMMARY (IN KOREAN) middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 37

ACKNOWLEDGEMENTS (IN KOREAN)

CURRICULUM VITAE

iii

LIST OF TABLES

I Material property of simulation model and test specimen middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 17

II Damage cases for numerical simulation middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 18

III Damage characterization for the numerical simulation models middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 20

IV Damage characterization for the Teflon inserted damage specimen middotmiddotmiddotmiddotmiddotmiddotmiddot 25

V Damage characterization for the impact-induced damage specimen middotmiddotmiddotmiddotmiddotmiddotmiddot 27

VI Temperature effect test for the Teflon inserted damage specimen middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 28

VII Temperature effect test for the impact induced damage specimen middotmiddotmiddotmiddotmiddotmiddotmiddot 30

iv

LIST OF FIGURES

1 Lamb wave propagation characteristics (a) A0 Lamb wave mode time delay and

attenuation after passing through a damage (b) The time delay and amplitude varia-

tion of A0 Lamb wave mode under temperature variation in pristine condition of

structure Because of this conventional damage detection techniques in composites

have limitations for real structure application middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 5

2 Multiple reflections in the damage area The multiple reflections of A0 Lamb

waves in the damage area are utilize as a feature for characterization of the damage

in this study middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 6

3 Parameters and notations of the proposed damage characterization technique

Damage is detected by identifying the additional reflected wave from the damage due

to multiple reflections (dashed lines) Localization and quantification of the damage

is carried out by measuring the arrival time of transmitted and reflected waves from

the damage middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 7

4 Two different FEM models which have different 119889119889 and 119907119907119889119889 (or G) This repre-

sents that the damage cannot be characterized uniquely based on the measured data

available middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 9

5 (a) Example of the overlapped signal with comparing approximated signal using

the matching pursuit method (b) Decomposed signal using the matching pursuit

method The matching pursuit method provides a conventional way to decompose

and identify each wave middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 9

6 Description of the numerical simulation model middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 17

7 Comparison of A0 Lamb wave mode signals from numerical simulation be-

tween intact and damage cases Time delay attenuation and multiple reflections due

to damage simulated successfully using the damage modeling method described in

v

the previous section middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 19

8 Test specimens for the experimental verification In the Teflon inserted damage

Teflon tapes with 25mm diameter were inserted at the center of the specimen The

impact-induced damage was introduced to the specimen using the impact test equip-

ment middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 21

9 Overall experimental setup (a) Data acquisition (DAQ) system and (b) Installed

circular PZT transducer on the specimen with its dimension middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 23

10 (a) Impact test equipment for introducing impact-induced damage and (b) Tip

of the impact tester with 10mm diametermiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 23

11 Test specimen in temperature chamber middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 24

12 A0 Lamb wave mode obtained using pitch-catch method from the Teflon in-

serted damage specimen (a) The reflected wave from damage is shown (b) The Tef-

lon inserted damage is detected by capturing the reflected wave middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 25

13 The signal obtained from the test specimen subjected to impact The impact

was introduced at the same location of the specimen and the signal was obtained after

each impacts As the number of impact is increased the transmitted A0 Lamb wave

mode is delayed and attenuated middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 26

14 False alarm test for the proposed damage detection algorithm in pristine condi-

tion of the specimen Reflected wave from damage is not captured false alarm is not

occurred middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 26

15 Signals obtained from the Teflon inserted damage specimen under temperature

variation Time delay and amplitude variation are occurred middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 27

16 False alarm test under temperature variations It is shown that proposed dam-

vi

age characterization algorithm is robust under temperature variationsmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 29

17 Signals obtained from impact-induced damage specimen (Impact 3) under

temperature variation middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 29

vii

CHAPTER 1 INTRODUCTION

11 Motivation

Composite materials has been increasingly demands in modern structures ranging from the

latest aircraft to retrofitted bridge deck due to its light weight superior specific strength and

stiffness properties However the occurrence of damage due to abrupt impact or sequential

accumulation of defect may considerably degrade structural strength or stiffness and reduce

the safety and reliability of the system Recently Lamb waves based damage detection tech-

niques using piezoelectric (PZT) transducers have emerged as one of the promising structural

health monitoring (SHM) technology for detecting hidden damage because of the possibility

of long range inspection and high sensitivity [1-4]

12 Literature review

This section will discuss researches that have been previously carried out on topics of

guided wave based SHM techniques using PZT transducers for damage in composite materi-

als The early investigations of interaction between Lamb wave and damage in composite can

be found in Rose et al and Guo et al [5-6] From further studies reflected signals from dam-

age were shown to be an effective method for damage diagnosis and localization using PZTs

[7-9] The reflected signals from damage were collected using one PZT as an actuator and the

other PZT as a sensor However the techniques cannot be used to determine the extent of

- 1 -

damage with reasonable precision

Kesseler et al proposed a technique monitoring the wavelet coefficients of transmitted ul-

trasonic signal amplitude as a potential way to damage detection [10] One drawback of this

technique is that the variability in coupling and the sensitivity of installed PZT transducers

can induce changes in the amplitude and energy of the signal that are cause false-alarm Oth-

er wavelet based delamination detection and localization technique developed using PZT

sensor network and outlier analysis [11]

Damage detection and quantification in composite plate was also investigated by group de-

lay measurement due to damage [12] However it is assumed that the group velocity in dam-

age area as the group velocity propagates half thickness of the structure in dispersion curve in

the quantification procedure Thus the damage quantification was investigated only Teflon

inserted damage case

Sohn et al proposed a damage detection technique which does not rely on any past base-

line signals to assess damage in composite plates by using an enhanced time reversal method

[13] The damage is assessed using the violation of time reversibility when nonlinearity is

caused by the damage along a direct wave path Thus it allows instantaneous identification

of delamination without requiring the baseline signal for comparison

- 2 -

13 Objective and uniqueness

In this study a reference-free damage characterization technique including detection lo-

calization and quantification of impact-induced damage in composite materials is developed

The general damage characterization can be categorized into the following levels [14]

Level 0 Define damage

Level 1 Determine the presence of damage within the structure

Level 2 Locate the regions of damage

Level 3 Quantify the damage

Level 4 Predict the remaining service life of the structure

Through the proposed technique from level 1 to level 3 can be achieved

Most of common approaches of damage characterization techniques described in the pre-

vious section use prior obtained baseline data However these approaches can cause false

alarm due to the effect of environmental variations such as temperature and ambient vibra-

tions [15]

The uniqueness of this study is

1 Damage is characterized without using prior measured baseline data

2 Robustness of the proposed technique under temperature variation is investigated

3 Additional waves reflected from inside damage have been used in addition to transmit-

ted and reflected waves

- 3 -

14 Organization

The organization of this thesis is summarized as follows

Chaper1 A general introduction including motivation literature review objective and

uniqueness of this study is described

Chaper2 Interactions between Lamb wave mode and damage in composites such as time

delay attenuation and multiple reflections are summarized Damage index (m-value) for

damage characterization is defined

Chaper3 Algorithms for the reference-free damage characterization technique using m-

value are developed Modified matching pursuit for this study is also described

Chaper4 Verification of the algorithm using numerical simulation is presented Various

modeled damage cases are characterized

Chaper5 Experimental verifications are described Two damage cases (Teflon inserted and

impact-induced damage) are investigated for the feasibility of the proposed damage charac-

terization technique Temperature effect test using temperature chamber is also carried out

Chaper6 Conclusion and summary of the present study is reported In advance suggestions

and future work recommendations are represented

- 4 -

CHAPTER 2 INTERACTION OF LAMB WAVE WITH DAMAGE

21 A0 Lamb wave mode delay and attenuation in the time domain

Time delay and attenuation of the A0 mode Lamb wave after passing through the damage

area due to effective thickness reduction were reported in previous researches [812] Moreo-

ver the attenuation of the A0 Lamb wave mode is occurred by degrading of shear modulus of

in the damage area However the time delay and amplitude variation of the A0 Lamb wave

mode also occurred even when the structure is health condition by the material property

changing due to temperature variation of the structure (Fig 1) Thus conventional damage

characterization methods using the time delay and attenuation of A0 Lamb wave mode as a

feature can cause false-alarm in real application

11 115 12 125 13-10

-5

0

5

10

Time (ms)

Vol

tage

(mV

)

IntactDamage

Reflected wave from Damage

(a)

11 115 12 125 13-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

0 oC20 oC50 oC

(b)

Fig 1 Lamb wave propagation characteristics (a) A0 Lamb wave mode time delay and atten-uation after passing through a damage (b) The time delay and amplitude variation of A0 Lamb wave mode under temperature variation in pristine condition of structure Because of this conventional damage detection techniques in composites have limitations for real struc-ture application

- 5 -

22 Multiple reflections in the damage area

Standing wave patterns due to multiple reflections of A0 Lamb wave mode in the damage

area were observed through numerical simulations and ultrasonic wave-field imaging using a

scanning laser vibrometer [16] The formation of standing waves in damage is illustrated

through Fig 2 When the incident waves enter the damage area a significant portion of the

waves are reflected back from the exit Then the reflected waves travel through the damage

and undergo reflection again at the original entrance As a consequence of multiple reflec-

tions at the entrance and the exit a considerable amount of ultrasonic energy is trapped inside

the damage area [17-19]

Multiple Reflections

Incident Wave Transmitted Wave

Damage

Fig 2 Multiple reflections in the damage area The multiple reflections of A0 Lamb waves in the damage area are utilize as a feature for characterization of the damage in this study

- 6 -

CHAPTER 3 THEORETICAL DEVELOPMENT

31 Overview

Damage

l = l1 + d + l2

d = vdtd l1 = vit1 l2 = vit2

t = t1 + td + t2

Excitation and Sensing PZT Sensing PZT

l

Transmitted Wave

Reflected Waves

Incident Wave

Fig 3 Parameters and notations of the proposed damage characterization technique Damage is detected by identifying the additional reflected wave from the damage due to multiple re-flections (dashed lines) Localization and quantification of the damage is carried out by measuring the arrival time of transmitted and reflected waves from the damage

In this section the reference-free damage characterization algorithm is proposed Damage

diagnosis and quantification are achieved by identifying the additional reflected waves due to

multiple reflections in the damage area (dashed lines in Fig 3) Moreover damage is local-

ized using additional information from pulse-echo method implement For the localization

and the quantification of damage the variables such as the path length and A0 Lamb wave

mode velocity in damage area are defined (Fig 3) For example 119897119897 is expressed as

119897119897 = 1198971198971 + 119889119889 + 1198971198972 (1)

where 1198971198971 119889119889 and 1198971198972 are the distance from the excitation PZTs to the entrance of damage the

damage length and the distance from the exit of damage to the sensing PZT respectively

- 7 -

Moreover 119907119907119894119894 and 119907119907119889119889 are also defined as the A0 Lamb wave mode velocity in pristine area

and damage area respectively Note that the whole procedure of the damage characterization

is accomplished based solely on the signals obtained from the current state of the structure

32 Damage index (m-value)

The proposed reference-free damage characterization technique utilizes the waves reflect-

ed additionally from the damage area as a feature The arrival time difference of the transmit-

ted and reflected A0 Lamb wave modes reflected from the exit and the entrance of the delam-

ination area one times each (∆119905119905) contains the information of the physical length of the dam-

age (119889119889) and A0 Lamb wave mode velocity in the damage area (119907119907119889119889) The wave velocity in the

damage area which is related to the shear modulus (G) represents the severity of the damage

∆119905119905 = 2119905119905119889119889 =2119889119889119907119907119889119889

(2)

However based on the measured data available the 119889119889 and 119907119907119889119889 (or G) cannot be uniquely

determined Fig 4 shows two FEM models which have different physical length and wave

velocity (G) of the damage Detail explanation of delamination modeling technique will be

covered in section 4 The similar signal in the time domain and ∆119905119905 are obtained from the

two different damage cases

- 8 -

Damage 1

Damage 2

20

Unit mm15

90 G value reduction

943 G value reduction

m-value = 248

m-value = 248

015 02 025 03 035

-2

-1

0

1

2

Time (ms)

Ampl

itude

Damage 1Damage 2

Fig 4 Two different FEM models which have different 119889119889 and 119907119907119889119889 (or G) This represents that the damage cannot be characterized uniquely based on the measured data available

Thus in this study the damage index (m-value) which can be uniquely defined based on

the available data is introduced The m-value is defined as the ratio of the physical length of

the damage and wavelength of the A0 Lamb wave mode in the damage area (120582120582119889119889)

119898119898 =119889119889120582120582119889119889

=119889119889119889119889119907119907119889119889

(3)

In many cases the transmitted and reflected waves are overlapped and it is hard to decom-

pose and calculate the arrival times of them Matching pursuit method is an alternative way

to accurately decompose of the overlapped waves and estimate the parameters of each wave

such as arrival time (Fig 5) Detail explanation of the matching pursuit method is covered in

section 36

015 02 025 03-2

-1

0

1

2

Time (ms)

Am

plitu

de

OriginalApproximated

(a)

015 02 025 03-2

-1

0

1

2

Time (ms)

Ampl

itude

Wave 1Wave 2

(b)

Fig 5 (a) Example of the overlapped signal with comparing approximated signal using the matching pursuit method (b) Decomposed signal using the matching pursuit method The matching pursuit method provides a conventional way to decompose and identify each wave

- 9 -

33 Damage detection algorithm (Level 1)

The damage diagnosis is accomplished by extracting the additional reflected A0 Lamb

wave modes from damage due to multiple reflections If A0 Lamb wave modes are decom-

posed perfectly it is trivial that one transmitted A0 Lamb wave mode is always captured in

the response signal If damage exists other waves in the response signal behind the transmit-

ted A0 Lamb wave mode are presented Since this approach does not rely on previously col-

lected data the diagnosis is independent from environmental variation such as temperature

changes

It is reported that the dispersion of Lamb waves cause wave-packets to spread out in space

and time as they propagate through a structure [20] Thus the reflected waves from the dam-

age have longer pulse duration than the transmitted wave due to their longer propagating dis-

tance In this study the matching pursuit method is modified to capture more spread wave

than the transmitted A0 Lamb wave mode Detail explanation of the modified matching pur-

suit method is described in latter section

Let 1199041199041 and 1199041199042 as the pulse duration (scale) of the transmitted and additional reflected A0

Lamb wave mode respectively

ldquoIf 1199041199042 gt 1199041199041 damage exists

Otherwise there is no damagerdquo

- 10 -

34 Damage quantification algorithm (Level 3)

As mentioned in the previous section damage is quantized by estimating 119889119889 and 119907119907119889119889 using

pitch-catch method The procedure to estimate 119889119889 and 119907119907119889119889 is

(i) Measure 119897119897 119905119905 and 119905119905119889119889

(ii) Assume the range of 119889119889 (0 lt 119889119889 lt 119897119897)

(iii) Calculate 119907119907119894119894 and 119907119907119889119889 with a constraints (0 lt 119907119907119889119889 lt 119907119907119894119894) (Eq 4 and 5)

(iv) Estimate m-value and find a feasible range of 119889119889 and 119907119907119889119889

119907119907119894119894 =119897119897 minus 119889119889119905119905 minus 119905119905119889119889

(4)

119907119907119889119889 =119889119889119905119905119889119889

(5)

35 Damage localization algorithm (Level 2amp3)

For damage localization 1198971198971 is estimated by measuring the arrival time from the excitation

PZT to the entrance of delamination (1199051199051) with additional pulse-echo method implementation

The procedure is similar with the previous damage quantification algorithm

(i) Measure 119897119897 119905119905 119905119905119889119889 and 1199051199051

(ii) Assume the range of 1198971198971 (0 lt 1198971198971 lt 119897119897)

(ii) Compute possible combinations of 1198971198972 119889119889 and 119907119907119889119889 with constraints

(0 lt 119907119907119889119889 lt 119907119907119894119894 119889119889 gt 0) (Eq 5 6 and 7)

- 11 -

1198971198972 = 1198971198971119905119905 minus 1199051199051 minus 119905119905119889119889

1199051199051 (6)

119889119889 = 119897119897 minus 1198971198971 minus 1198971198972 (7)

Since a PZT for pulse-echo method acts as both an actuator and a sensor the obtained sig-

nal contains both the excitation and the response signal Thus the reflected wave from the

entrance of damage can be buried into the tail of excitation signal and cause error on arrival

time of reflected wave from damage Even in this case damage can be localized and quan-

tized using the constraints When first wave packet reflected from the entrance of damage is

buried into the excitation signal the next wave packet reflected from the exit of damage can

be utilized alternatively If the algorithm regards the second reflected wave as the first re-

flected wave there are no combinations of estimated variable meeting the constraints Then

the algorithm repeats the process by regarding the wave as the second reflected wave the

constraints can select proper combination of the estimated variables

The estimated values have range bounded by the constraints When additional information

of 119907119907119894119894 is utilized the damage can be characterized more accurately In this study three meth-

ods are proposed and verified as information level of 119907119907119894119894

Method 1 Solely based on the current data measurement (Reference-free)

Method 2 Limit the possible range of 119907119907119894119894 (Prior knowledge of the system)

Method 3 Measure 119907119907119894119894 (Additional data need to be collected)

Method 1 described in this section can be utilized without any information of 119907119907119894119894 Meth-

od 2 produces more accurate estimation results using prior knowledge of the system such as

- 12 -

119907119907119894119894 variation under temperature changes Method 3 gives unique estimated value using meas-

ured 119907119907119894119894 from other path in the same time In section 4 and 5 damage is characterized using

these three methods and the results are compared each another

36 Matching pursuit method considering the dispersion effect of Lamb

wave

Matching pursuit method has been developed for signal compression approximation and

characterization of obtained signal and function [21-23] The key concept of the matching

pursuit method is that any signal (function) can be decomposed into a linear expansion of

signals selected from a pool of basis signals (functions) called lsquodictionaryrsquo

Let 119891119891 be a function to be analyzed then 119891119891 can be approximated a set of functions select-

ed from the dictionary 119863119863 in order to best match This is done by successive approximations

with orthogonal projections Let1198921198921205741205740 isin 119863119863 the function 119891119891 can be decomposed into

119891119891 = lang1198911198911198921198921205741205740rang1198921198921205741205740 + 119877119877119891119891 (8)

where 119877119877119891119891 is the residual function after approximating 119891119891 using 1198921198921205741205740 1198921198921205741205740 is orthogonal to

119877119877119891119891 hence

1198911198912 = lang1198911198911198921198921205741205740rang2

+ 1198771198771198911198912 (9)

To minimize 119877119877119891119891 1198921198921205741205740 is chosen to maximize lang1198911198911198921198921205741205740rang The matching pursuit method is

an iterative algorithm that decomposes the residue 119877119877119891119891 by projecting it on a function of 119863119863

that matches 119877119877119891119891 almost at best as it was done for 119891119891 The residue 119877119877119899119899119891119891 is decomposed into

119877119877119899119899119891119891 = lang119877119877119899119899119891119891119892119892120574120574119899119899rang119892119892120574120574119899119899 + 119877119877119899119899+1119891119891 (10)

- 13 -

the decomposition is carried up to the order 119898119898 119891119891 can be expressed as

119891119891 = lang119877119877119899119899119891119891119892119892120574120574119899119899rang119892119892120574120574119899119899

119898119898minus1

119899119899=0

+ 119877119877119898119898119891119891 (11)

The original function 119891119891 is decomposed into a sum of dictionary elements that are chosen to

best match its residues

In this study the matching pursuit method is utilized to accurately decompose of the over-

lapped Gaussian pulses and estimate the parameters of the pulse which used for damage

characterization The 119894119894119905119905ℎ Gaussian pulse in the dictionary for approximating the obtained

signal is expressed as

119892119892120574120574119894119894 = 119891119891(119860119860119894119894 119906119906119894119894 119904119904119894119894 119889119889119894119894120601120601119894119894) (12)

where 119860119860119894119894119906119906119894119894119904119904119894119894119889119889119894119894 and 120601120601119894119894 represent amplitude arrival time scale (pulse duration) excita-

tion(central) frequency and phase of the Gaussian pulse respectively The parameters of the

pulse are estimated using a nonlinear regression based optimization method (Gauss-Newton

method) based on an initial guess of the parameters Detail procedure for estimating the pa-

rameters of the pulse is expressed in Hong et al [23]

Among the estimated parameters of the pulses pulse duration (scale) represents the disper-

sion effect If a wave-packet is spread out the estimated pulse duration is increased Using

this feature the matching pursuit method is modified to capture waves which have longer

pulse duration than transmitted A0 Lamb wave mode If the estimate pulse duration of a wave

is less than the transmitted wave the modified matching pursuit method regards it as error

- 14 -

CHAPTER 4 NUMERICAL SIMULATION

41 Composite modeling method

Composite materials can be modeled from nano to macro level depends on its engineering

purpose In general four methods are utilized for modeling of composite materials First

composite is modeled as a homogenous material expressed in terms of effective material

properties [24] The effective material properties are estimated using the modulus and the

volume fraction of each constituent theoretically (Rule of mixture) and measured experimen-

tally Second a model is established that each laminate which have own material properties

are totally bonded together In this model the stacking sequence and direction of each lami-

nates are considered additionally [25-27] Third method for composite modeling in micro

level called micromechanics [28-30] This model considers the interaction between fibers and

matrix in laminate Last multiscale modeling which the scope of modeling is extended to

molecular or quantum level of constituents [31-33] This model starts from molecular me-

chanics or molecular dynamics then the scale of model leads to the nano micro and macro

level by incorporating small scale models As the scale of models is smaller it provides more

logical way than other large scale of models while the computational cost is more expensive

The scope of this study is focused on the dynamic behavior of composite structure the beam

or plate model was utilized for the numerical simulation

- 15 -

42 Damage modeling method

Generally three damage models have been developed and utilized in the real applications

First the stiffness reduction model called as the equivalent element model was introduced

[17 34-35] This model is simply realized by replacing the region around damage area by a

region with adjusted material properties (Youngrsquos modulus shear modulus density etc)

globally Next the duplicated node model is done by disconnecting the element connectivity

assuming there is no interaction among the delaminated surface This model can be utilized

when the location and the extension of damage is important issue [34-36] Last the contact

mechanics model using the principles of contact mechanics the two interfacial forces (con-

tact resistance and sliding resistance) in the delaminated surface [37-38] In this model the

delaminated surfaces are modeled as nonlinear functions of the interlaminar normal strain

and the interlaminar shear strain This model is computationally expensive while provides

more logical way than other models There is however no universally accepted model for

general impact-induced damage in composites In the case of impact-induced damage it is

impossible to predict or estimate the location and number of delaminated layers Thus from

the uncertainty of impact-induced damage the equivalent element model can be considered

as the effective damage model

- 16 -

43 Numerical simulation description

TABLE I Material property of simulation model and test specimen

Tensile Strength 657 MPa Tensile Modulus 59 GPa Shear Strength 804 MPa Shear Modulus 24 Gpa Specific Gravity 16

PZT A

PZT C

PZT BDamage

3

250 125 125

500 Unit mm

Fig 6 Description of the numerical simulation model

The feasibility of the proposed damage characterization technique is verified by 2-D FEM

simulation using MSC PATRAN and NASTRAN Fig 6 presents the dimension PZTs and

damage location of the simulation model In the model 05mm x 05mm square shell element

was used for modeling of the A0 mode whose flexural wave characteristic The element size

was determined by considering the wavelength of the propagation Lamb wave modes The

material properties of the simulation model and damage cases are presented in Table I and

Table II respectively

PZT A and C act as both actuator and sensor for collecting pulse-echo signal They were

installed on the plate which two of them are both-side collocated for the A0 Lamb wave

- 17 -

mode decomposition [39-40] PZT B was only used as a sensor for collecting pitch-catch sig-

nal For simulating the excitation and sensing of PZTs the pin-force model was adopted that

an input force apples to the end nodes of excitation PZT and averaging element strains corre-

sponding sensing area A Gaussian pulse (tone-burst) 50 kHz central frequency was excited

as an input signal

TABLE II Damage cases for numerical simulation

119949119949120783120783 (mm) 119941119941 (mm) G value reduction ()

Case 1 650 200 90 darr

Case 2 675 150 90 darr

Case 3 675 150 93 darr

Case 4 900 150 90 darr

- 18 -

44 Numerical simulation results

015 02 025 03 035

-2

0

2

Time (ms)

Am

plitu

de

IntactDamage

Reflected wave from damage

(a) Pitch-catch signal

015 02 025 03 035

-2

0

2

Time (ms)

Am

plitu

de

IntactDamage

Reflected waves from damage

(b) Pulse-echo signal Fig 7 Comparison of A0 Lamb wave mode signals from numerical simulation between intact and damage cases Time delay attenuation and multiple reflections due to damage simulated successfully using the damage modeling method described in the previous section

Fig 7 shows that comparison of A0 Lamb wave mode signals between intact and one of

damage cases in the simulation The features of A0 Lamb wave mode passing through dam-

age area described in section 2 are successfully simulated Thus the equivalent element

method is proper for modeling damage in composites In the damage cases the additional

waves reflected from damage is observed in both pitch-catch and pulse echo signal

Table III presents the result of the proposed damage characterization algorithm using the

numerical simulation cases shown in Table II In lsquoMethod 2rsquo 119907119907119894119894 variation is assumed to plusmn

5 of the normal value (1190~1315ms) The normal value of 119907119907119894119894 for lsquoMethod 3rsquo is meas-

ured as 1253 ms In each case the damage is characterized using the proposed technique

Note that as the information of 119907119907119894119894 is specified the estimated value is closer to the exact

value

- 19 -

TABLE III Damage characterization for the numerical simulation models

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

Case 1

Exact 650 200

Method 1 248 50~70 79~414 160~834

Method 2 248 603~667 135~241 271~486

Method 3 248 635 188 379

Case 2

Exact 675 150

Method 1 193 60~75 01~251 37~651

Method 2 193 618~682 114~222 295~575

Method 3 193 65 168 435

Case 3

Exact 675 150

Method 1 233 60~75 66~302 140~649

Method 2 233 668~738 85~196 182~419

Method 3 233 703 140 301

Case 4

Exact 900 150

Method 1 204 75~100 48~349 118~856

Method 2 204 860~951 107~216 264~530

Method 3 204 906 162 397

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1253 ms 1190~1315 ms Measured 119907119907119894119894 1253 ms

- 20 -

CHAPTER 5 EXPERIMENTAL VERIFICATION

51 Overview

In this section experimental study is investigated so that the proposed damage characteri-

zation technique Overall experimental setup and test specimen are shown in Fig 8 and 9

The experimental setup consists of test specimens data acquisition (DAQ) system and self-

sensing circuit Self-sensing circuit was installed for pulse-echo measurement using PZTs

[40-41]

PZT A(PZT C) PZT B

500

500

140

75

250

Teflon Inserted Damage (d =25)

Unit mm

Thickness = 3

PZT A(PZT C) PZT B

500

500

150

90

250

Impact Location

Unit mm

Thickness = 3

(a) Teflon inserted damage specimen

(b) Impact-induced damage specimen

Fig 8 Test specimens for the experimental verification In the Teflon inserted damage Teflon tapes with 25mm diameter were inserted at the center of the specimen The impact-induced damage was introduced to the specimen using the impact test equipment

- 21 -

511 Specimen

Two carbon composite plates with a dimension of 500times500times3mm3 were manufactured by

Nexcomes Inc by stacking 12 layers of CF3327 3K prepreg (woven fabrics type) manufac-

tured by Hankuk Fiber Inc The material properties of the specimen are presented in Table I

One of the specimens Teflon tapes with 25mm diameter was inserted in every 3 layers at the

center of the specimen to mimic the damage The other specimen is utilized for the character-

ization of real impact-induced damage

512 Data acquisition (DAQ) system and PZT

The data acquisition (DAQ) system consists of an arbitrary waveform generator (AWG) a

high speed signal digitizer (DIG) and two multiplexers (MUX) Using the 16-bits AWG

tone-burst input signal with a plusmn10 peak-to-peak voltage were generated The response signal

was measured by the 14-bit DIG For the signal-to-noise ratio reduction both pitch-catch and

pulse-echo signals were measured 150 times and averaged in the time domain 3 identical

circular PZTs of 10mm diameter manufactured by APC International Ltd were installed on

the specimen A Gaussian pulse (tone-burst) 50 kHz central frequency was excited as an in-

put signal The excitation frequency was chosen in order to avoid overlapping of the S0 and

A0 Lamb wave modes and generating of the higher modes (S1 A1hellip)

- 22 -

PZT Actuator

PZT Sensor

AWG

DIG

DAQ System Specimen

SensingActuatingMUX

1

23

4

1 Sending signal to MUX

4 Sending signal to DIG

Self-SensingCircuit

18 Unit mm

10

(a)

(b)

Fig 9 Overall experimental setup (a) Data acquisition (DAQ) system and (b) Installed circu-lar PZT transducer on the specimen with its dimension

513 Impact test

The impact-induced damage was introduced to the specimen using the impact test equip-

ment presented in Fig 10 A 5kg mass with 10mm tip was dropped from 015m height onto

the same location of the specimen The signals were obtained at each number of impacts

Since the reflected wave is not detected at the first impact case the data obtained from the

second and third impact were utilized for the analysis

10

Unit mm (a)

(b)

Fig 10 (a) Impact test equipment for introducing impact-induced damage and (b) Tip of the impact tester with 10mm diameter

- 23 -

514 Temperature effect test

In reality structures are subjected to changing environmental conditions such as tempera-

ture that affect measured signal and cause false-alarm Thus the robustness of the proposed

damage characterization technique under temperature variation in investigated (Fig 11) The

specimens were placed inside a temperature chamber and a thermocouple was installed on

the specimen to measure temperature of surface of the specimen The signal was obtained

under three temperature cases (0 20 and 50degC) and the humidity was kept at 30

Temperature Chamber

Test Specimen

Fig 11 Test specimen in temperature chamber

- 24 -

52 Teflon inserted damage

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Reflected wave from damage

(a)

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Transmitted WaveReflected Wave

Detected reflected wave

(b)

Fig 12 A0 Lamb wave mode obtained using pitch-catch method from the Teflon inserted damage specimen (a) The reflected wave from damage is shown (b) The Teflon inserted damage is detected by capturing the reflected wave

Fig 12 shows the obtained pitch-catch signal from the Teflon inserted damage specimen

and damage diagnosis result The Teflon inserted damage is detected successfully by captur-

ing the reflected wave Table IV presents the result of the Teflon inserted damage characteri-

zation 119907119907119894119894 is measured as 1213ms and the variation is assumed to plusmn 5 of the normal val-

ue (1152~1273ms) The results indicate that the proposed damage characterization approach

works well in real application

TABLE IV Damage characterization for the Teflon inserted damage specimen

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

Exact 635 25

Method 1 123 60~75 73~338 298~1220

Method 2 123 619~662 229~310 934~1188

Method 3 123 631 284 1161

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1213 ms

- 25 -

53 Impact-induced damage

11 115 12 125 13-4

-2

0

2

4

Time (ms)

Vol

tage

(mV

)

IntactImpact2Impact3

Reflected wave from damage

(a) Pitch-catch signal

11 115 12 125 13 135

-05

0

05

1

Time (ms)

Volta

ge (m

V)

IntactImpact2Impact3

Boundary reflection

Reflected wave from damage

(b) Pulse-echo signal

Fig 13 The signal obtained from the test specimen subjected to impact The impact was in-troduced at the same location of the specimen and the signal was obtained after each impacts As the number of impact is increased the transmitted A0 Lamb wave mode is delayed and attenuated

The obtained signals from impact-induced damage specimen are presented in Fig 13 It is

observed that as the number of impact is increased the transmitted A0 Lamb wave mode is

more delayed because the wave velocity in the damage area is reduced Fig 14 shows the

false-alarm test result whether the proposed damage detection algorithm works even in pris-

tine condition of the test specimen It is shown that no false alarm is occurred because no ad-

ditional reflected wave is captured

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Transmitted Wave

(a) Transmitted wave

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Reflected Wave

No reflected wave

(b) Reflected wave

Fig 14 False alarm test for the proposed damage detection algorithm in pristine condition of the specimen Reflected wave from damage is not captured false alarm is not occurred

- 26 -

TABLE V Damage characterization for the impact-induced damage specimen

of Impacts m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

2

(1470 J)

Method 1 080 65~90 75~246 470~1157

Method 2 080 801~885 114~245 711~1134

Method 3 080 843 179 1122

3

(2205 J)

Method 1 148 50~100 06~380 21~1134

Method 2 148 758~838 249~368 843~1125

Method 3 148 798 308 1044

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1213 ms

Table V presents the characterization results of the impact-induced damage case Here the

measured 119907119907119894119894 and the assumed variation are same as the artificial damage case It is observed

that as the number of impact is increased 119889119889 is increased while 119907119907119889119889 is reduced Thus it is

validated that the proposed damage characterization approach estimates the increment of the

damage well

54 Temperature effect test

541 Teflon inserted damage

11 115 12 125 13-10

-5

0

5

10

Time (ms)

Vol

tage

(mV

)

0 oC20 oC50 oC

(a) Pitch-catch signal

11 115 12 125 13 135

-1

0

1

2

3

Time(ms)

Volta

ge (m

V)

0 oC20 oC50 oC

(b) Pulse-echo signal

Fig 15 Signals obtained from the Teflon inserted damage specimen under temperature varia-tion Time delay and amplitude variation are occurred

Fig 15 shows the obtained signals from the Teflon inserted damage specimen subjected to

- 27 -

temperature variation Time delay and amplitude variations are occurred there is no addition-

al damage is introduced The analyzed results of the obtained signals are reported in Table VI

The measured 119907119907119894119894 is 1247 ms at 0 degC 1213 ms at 20 degC and 1168 ms at 50 degC respectively

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms (1152~1273 ms)

The estimated results are also changed under temperature variations the fluctuation of esti-

mated range and exact value is small

TABLE VI Temperature effect test for the Teflon inserted damage specimen

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

0 degC

Exact 635 25

Method 1 113 60~70 79~268 352~1291

Method 2 113 560~619 233~344 1034~1172

Method 3 113 606 257 1141

20 degC

Exact 635 25

Method 1 123 60~75 73~338 298~1220

Method 2 123 619~662 229~310 934~1188

Method 3 123 631 284 1161

50 degC

Exact 635 25

Method 1 137 60~75 72~328 263~1228

Method 2 137 622~687 183~299 666~1086

Method 3 137 642 264 960

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1247 ms at 0 degC 1213 ms at 20 degC and 1168 ms at 50 degC

- 28 -

542 Impact-induced damage

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Vol

tage

(mV

)

Transmitted WaveReflected Wave

No reflected wave

(a) 0 degC

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Transmitted WaveReflected Wave

No reflected wave

(b) 50 degC

Fig 16 False alarm test under temperature variations It is shown that proposed damage char-acterization algorithm is robust under temperature variations

The false alarm test in pristine condition of the impact-induced damage specimen under

temperature variation was investigated (Fig 16) It is validated that the proposed damage di-

agnosis algorithm is robust under temperature variations Fig 17 shows the obtained signals

from the impact-induced damage (impacted 3 times) specimen under temperature variation

The analyzed results of the obtained signals are reported in Table VII The measured 119907119907119894119894 in

each temperature and the assumed variation are same as the artificial damage case

11 115 12 125 13-5

0

5

Time (ms)

Volta

ge (m

V)

0 oC20 oC50 oC

(a) Pitch-catch signal

11 115 12 125 13 135

-1

0

1

2

3

Time(ms)

Vol

tage

(mV

)

0 oC20 oC50 oC

(b) Pulse-echo signal

Fig 17 Signals obtained from impact-induced damage specimen (Impact 3) under tempera-ture variation

- 29 -

TABLE VII Temperature effect test for the impact induced damage specimen (Impact 3 case)

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

0 degC

Method 1 135 75~90 72~310 267~1148

Method 2 135 705~779 264~381 978~1126

Method 3 135 763 289 1071

20 degC

Method 1 148 50~100 06~380 21~1134

Method 2 148 758~838 249~368 843~1125

Method 3 148 798 308 1044

50 degC

Method 1 153 75~100 53~415 174~1227

Method 2 153 804~889 214~336 702~1103

Method 3 153 815 320 1050

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1247 ms at 0 degC 1213 ms at 20 degC and 1168 ms at 50 degC

- 30 -

CHAPTER 6 CONCLUDING REMARKS

61 Executive summary

In this study the reference-free damage detection localization and quantification tech-

nique in a multi layered composite plate is proposed Additional reflected A0 Lamb wave

mode from the damage due to multiple reflections in damage area is used as a feature for

characterizing the damage First damage is characterized in terms of a damage index (m-

value) defined as the ratio of the physical length of the damage to the wavelength of the A0

Lamb wave mode Then the range of possible damage location length and severity are esti-

mated with constraints This procedure can be accomplished without prior baseline data To

realize the proposed technique the matching pursuit technique is modified for reduction of

the error by considering the dispersive characteristic of Lamb wave The matching pursuit

method estimates the arrival time of each wave packets accurately even two wave packets are

overlapped The numerical simulations and experimental tests including Teflon tape inserted

and impact-induced damage are investigated for the feasibility of the proposed technique

Moreover the effectiveness of the proposed technique under changing temperature is exper-

imentally verified

- 31 -

62 Suggestion for further study

The limitations of this study and the possible improvement and further studies are listed as

follow

1 Measurement of impact-induced damage length using conventional NDT technique

Additional experimental verification for impact-induced damage characterization is

needed by measuring the real length of damage using conventional NDT technique

such as ultrasonic scan or infrared camera

2 Limitation of both side collocated PZT installation In this study excitation PZTs are

installed both side and collocated for decomposing A0 Lamb wave mode However it

can produce decomposition error due to imperfection of PZT location and their installa-

tion Moreover in many cases assess to both surface of structures is not allowed such

as aircraft applications To improve this limitation a reference-free damage characteri-

zation using single side PZTs will be considered

3 Complex structure application In this study damage is characterized in simple plate-

like structure When the monitored structure becomes complex the response signals al-

so becomes complex due to its geometry and boundary Thus the investigations for

complex structure application such as stiffener spar and curvature are needed

- 32 -

REFERENCES

[1] Cawley P and Alleyne D (1996) ldquoThe use of Lamb waves for the long range inspection of large structuresrdquo Ultrasonics 34 pp287-290

[2] Ghosh T Kundu T and Karpur P (1998) ldquoEfficient use of Lamb waves for detecting defects in large platesrdquo Ultrasonics 36 pp791-801

[3] Guirguitui V (2003) ldquoLamb wave generation with piezoelectric wafer active sensors for structural health monitoringrdquo SPIE Conference on Smart Structures and Materials San Diego CA 5391 pp111-422

[4] Ragvahan A and Cesnik CE S (2004) ldquoModeling of piezoelectric-based Lamb wave generation and sensing for structural health monitoringrdquo SPIE Conference on Smart Struc-tures and Materials San Diego CA 5391 pp419-430

[5] Rose JL Pilarski A and Dirit JJ (1993) ldquoAn approach to guided wave mode selection for inspection of laminated platerdquo Journal of Reinforced Plastics and Composites 12(5) pp536-544

[6] Guo N and Cawley P (1993) ldquoThe interaction of Lamb waves with delaminations in composite laminatesrdquo Journal of Acoustical Society of America 94(4) pp 2240-2246

[7] Valdez SHD and Soutis C (2002) ldquoReal-time nondestructive evaluation of fiber com-posite laminates using low-frequency Lamb wavesrdquo Journal of Acoustical Society of Amer-ica 111(5) pp 2026-2033

[8] Yang M and Qiao P (2005) ldquoModeling and experimental detection of damage in virous materials using the pulse-echo method and piezoelectric sensoractuatorsrdquo Smart Materials and Structures 14 pp1083-1100

[9] Lestari W and Qiao P (2005) ldquoApplication of wave propagation analysis for damage identification in composite laminated beamsrdquo Journal of Composite Materials 39(22) pp1967-1984

[10] Kessler S S Spearing S M and Soutis C (2002) ldquoDamage detection in composite ma-terials using Lamb wave methodsrdquo Smart Materials and Structures 11 pp269-278

[11] Sohn H Park G Wail J R Limback N P and Farrar C R (2004) ldquoWavelet-based active sensing for delamination detection in composite structuresrdquo Smart Materials and Structures 13 pp 153-160

[12] Petculescu G Krishnaswamy S and Achenback J D (2008) ldquoGroup delay measure-ments using modally selective Lamb wave transducers for detection and sizing of delamina-tions in compositesrdquo Smart Materials and Structures 17

[13] Sohn H Park H W Law K H and Farrar C R (2007) ldquoDamage detection in compo-

- 33 -

site plates by using an enhanced time reversal methodrdquo Journal of Aerospace Engineering (ASCE) pp 141-151

[14] Sohn H (2007) ldquoEffects of environmental and operational variability on structural health monitoringrdquo Philosophical Transactions of the Royal Society Ardquo 365 pp539-560

[15] Rytter A (1993) ldquoVibration based inspection of civil engineering structuresrdquo PhD The-sis Department of Building Technology and Structural Engineering University of Aalborg Aalborg Denmark

[16] Hayashi T and Kawashima K (2002) ldquoMultiple reflections of Lamb waves at a delami-nationrdquo Ultrasonics 40 pp193-197

[17] Wang C H and Rose L R F (2003) ldquoWave reflection and transmission in beams con-taining delamination and inhomogeneityrdquo Journal of Sound and Vibration 264 pp 851-872

[18] Ramadas C Balasubramaniam K Joshi M and Krishnamurthy C V (2009) ldquoInterac-tion of the primary anti-symmetric Lamb mode (Ao) with symmetric delaminations numer-ical and experimental studiesrdquo Smart Materials and Structures 18

[19] Ramadas C Balasubramaniam K Joshi M and Krishnamurthy C V (2010) ldquoInterac-tion of guided Lamb waves with an asymmetrically located delamination in a laminated composite platerdquo Smart Materials and Structures 19

[20] Wilcox P Lowe M and Cawely P (2001) ldquoThe effect of dispersion on long-range in-spection using ultrasonic guided wavesrdquo NDTampE International 34 pp1-9

[21] Mallat S G and Zhang Z (1993) ldquoMatching pursuits with time-frequency dictionariesrdquo IEEE Tansactions on Signal Processing 41(12) pp3397-3415

[22] Zhang G Zhang S and Wang Y (2000) ldquoApplication of adaptive time-frequency de-composition in ultrasonic NDE of highly scattering materialsrdquo Ultrasonics 38 pp 961-964

[23] Hong J C Sun K H and Kim Y Y (2005) ldquoThe matching pursuit approach based on the modulated Gaussian pulse for efficient guided-wave damage inspectionrdquo Smart Mate-rials and Structures 14 pp548-560

[24] Herakovich C T (1997) Mechanics of Fibrous Composites Wiley 480 pages

[25] Reddy J N (1985) ldquoA review of the literature on finite-element modeling of laminated composite platesrdquo The Shock and Vibration Digest 17(4) pp 3-8

[26] Reddy J N (2003) Mechanics of Laminated Composite Plates and Shells Theory and Analysis CRC Press 856 pages

[27] Palacz M Krawczuk M and Ostachowicz W (2005) ldquoThe spectral finite element mod-el for analysis of flexure-shear coupled wave propagation Part 1 laminated multilayer

- 34 -

composite beamrdquo Composite Structures 68 pp37-44

[28] Zhang Y C and Harding J (1990) ldquoA numerical micromechanics analysis of the me-chanical properties of a plain weave compositerdquo Computer amp Structures 36(5) pp 839-844

[29] Nemat-Nasser S and Hori M (1999) Micromechanics Overall Properties of Heteroge-neous Materials Elsevier 810 pages

[30] Scida D Aboura Z Benzeggagh M L and Bocherens E (1999) ldquoA micromechanics model for 3D elasticity and failure of woven-fibre composite materialrdquo Composites Science and Technology 59 pp 505-517

[31] Thomas Y and Wu X H (1997) ldquoA multiscale finite element method for elliptic prob-lems in composite materials and porous mediardquo Journal of Computational Physics 134 pp 169-189

[32] Lee K H Moorthy S and Ghosh S (1999) ldquoMultiple scale computational model for damage in composite materialsrdquo Computer Methods in Applied Mechanics and Engineer-ing 172 pp 175-201

[33] Namilae S and Chandra N (2005) ldquoMultiscale model to study the effect of interfaces in carbon nanotube-based compositesrdquo Journal of Engineering Materials and Technology (ASME) 127 pp222-232

[34] Adams D E (2007) Health monitoring of Structural Materials and Components Wiley 460 pages

[35] Gopalakrishnan S (2009) ldquoModeling Aspects in Finite Elementrdquo Encyclopedia of Struc-tural Health Monitoring 2 pp791-809

[36] Palacz M Krawczuk M and Ostachowicz W (2005) ldquoThe spectral finite element mod-el for analysis of flexure-shear coupled wave propagation Part 2 Deminated multilayer composite beamrdquo Composite Structures 68 pp45-51

[37] Baik J M and Thopson R B (1984) ldquoUltrasonic scattering from imperfect interfaces A quasi-static modelrdquo Journal of Nondestructive Evaluation 4(3) pp177-196

[38] Luo H and Hanagud S (2000) ldquoDynamics of delamincated beamsrdquo International Jour-nal of Solids and Structures 37 pp1501-1519

[39] Kim S B and Sohn H (2007) ldquoInstantaneous reference-free crack detection based on polarization characteristics of piezoelectric materialsrdquo Smart Materials and Structures 16 pp2375-2387

[40] An Y K and Sohn H (2010) ldquoInstantaneous crack detection under varying temperature and static loading conditionsrdquo The Special Issue of Structural Control and Health Monitor-ing 17(7) pp 730-741

- 35 -

[41] Lee S J and Sohn H (2006) ldquoActive self-sensing scheme development for structural health monitoringrdquo Smart Materials and Structures 15(6) pp 1734-1746

- 36 -

요 약 문

복합재료 손상의 무기저 진단 및 정량화

복합재료는 재료의 경량성 내구성 고강도의 특성으로 인하여 항공기에서부터

교량에까지 널리 사용되고 있다 하지만 외부의 충격으로 인한 복합재료 내부의

손상은 재료의 강도를 감소시켜 구조물의 안정성을 저하시키는 요인이 된다 하지만

복합재료의 손상은 재료 내부에서 발생하여 육안으로는 식별하기도 힘들며 표면에서

보이는 것만으로 내부의 손상 정도를 판단하기는 불가능하다는 단점을 가지고 있다

본 연구에서는 손상 내부에서 반사로 인해 발생하는 파를 이용하여 기저자료를

사용하지 않고 특정 시점에서 얻은 자료만으로 복합재료의 손상을 진단하고 정량화

할 수 있는 기법을 제안한다 제안된 기법의 적용성을 검증하기 위하여 유한요소법을

이용한 수치해석을 수행하였고 Teflon tape 이 삽입된 시편과 실제 충격하중으로

인한 손상을 입은 시편을 이용한 실험을 진행하였다 수치해석 및 실험 결과 제안된

기법이 복합재료의 손상을 성공적으로 진단 및 정량화할 수 있음을 확인하였다 또한

다양한 외부 온도 환경에서 제안된 기법의 환경영향 실험을 수행하였고 다양한

환경에서 손상의 진단 및 정량화가 성공적으로 이루어짐을 검증하였다

핵심어 무기저 손상진단 및 정량화 압전센서 복합재료 유한요소법 Matching pursuit

method

- 37 -

감 사 의 글

기대와 두려움을 안고 시작한 석사과정이 어느덧 2 년 이라는 세월 속에서 추억으로 자리잡아

가고 있습니다 학부생 기간을 포함한 지난 3 년 동안 SHM 이라는 새로운 분야를 접하고 연구를

수행 할 수 있도록 도와주신 고마운 분들이 있어 그 감사를 표하고자 합니다

우선 저의 많은 부족한 점을 일깨워 주시고 연구에 대한 뜻을 더욱 확고히 세울 수 있게 해

주신 지도교수님이신 손 훈 교수님께 깊은 감사의 뜻을 전합니다 짧은 기간이었지만 교수님의

도움으로 많은 것을 배우고 느끼고 깨달을 수 있었습니다 또한 본 연구를 수행할 수 있도록 지

원해주신 국토해양부 (U-City 석 박사과정 지원사업)에 깊은 감사 드립니다

때로 힘들고 지칠 때 항상 저를 믿고 응원해 주시는 가족이 있었기에 지금의 글을 쓸 수 있었

습니다 또한 지금 연구실에서 함께 생활하고 있는 윤규형 민구형 철민이형 현석이형 진열이형

호민이형 병진이 선혜누나와 성대에 계신 창길이형 미국에 계신 의재형과 Debaditiya 에게도 감

사의 말씀을 전합니다 그리고 모두 나열할 수는 없지만 카이스트 생활의 활력소가 되어준 카이

스트 건설 및 환경공학과 선후배들 직원 분들과 SOS 멤버들에게도 감사를 표합니다

앞서 거론한 많은 이들의 도움이 없었다면 저 또한 지금 이 자리에 있을 수 없었을 것입니다

많이 부족하기에 더 노력하고 배우고자 했던 초심을 다시 한번 생각하며 더 큰 결실을 맺기 위해

앞으로 나아가겠습니다 도움을 주신 모든 분들께 이 감사의 글을 바칩니다

2011년 6월 카이스트에서 임형진 올림

- 38 -

CURRICULUM VITAE

Hyung Jin Lim

Research Assistant

The Civil and Environmental Engineering Department

Korea Advanced Institute of Science and Technology (KAIST)

291 Daehak-Ro Guseong-Dong Yuseong-Gu

Daejeon Republic of Korea 305-701

Phone (82)+42-350-3665 Fax (82)+42-350-3610

Email limhj87gmailcom

RESEARCH INTERESTS

Damage Detection Localization and Quantification on Composite and Metallic Structures Pattern Recognition Finite Element Modeling (FEM)

EDUCATION

2009 ndash 2011 MS Civil and Environmental Eng Korea Advanced Institute of Science and Technology (KAIST) Korea Rep

2006 ndash 2009 BS Civil and Environmental Eng Korea Advanced Institute of Science and Technology (KAIST) Korea Rep

JOURNAL PUBLICATIONS The corresponding author is underlined 1 Hyung Jin Lim Hoon Sohn ldquoReference-free damage detection localization and quantification

using in compositerdquo in preparation for Journal of the Acoustical Society of America 2011

2 Chul Min Yeum Hoon Sohn Jeong Bum Ihn Hyung Jin Lim ldquoInstantaneous delamination detection in a composite using a dual piezoelectric transducer networkrdquo Submitted to Compo-sites Science and Technology 2011

3 Yun-Kyu An Hyung Jin Lim Min Koo Kim Jin Yeol Yang Hoon Sohn Chang Guen Lee ldquoInvestigation of applicability of local reference-free damage detection techniques to in-situ

- 39 -

bridgesrdquo Submitted to ASCE Journal of Structural Engineering 2011

4 Hyung Jin Lim Min Koo Kim Hoon Sohn Chan Yik Park ldquoImpedance based damage detec-tion under varying temperature and loading conditionsrdquo Submitted to NDTampE International 2011

5 Hyung Jin Lim Hoon Sohn ldquoQuantitative estimation of transmitted and reflected lamb waves at discontinuityrdquo Journal of the Korean Society for Nondestructive Testing Vol30 No 4 pp1-8 2010

CONFERENCE PROCEEDINGS

The corresponding author is underlined 1 Hyung Jin Lim and Hoon Sohn ldquoReference-free damage detection localization and quantifi-

cation in compositesrdquo ASCE Engineering Mechanics Institute Conference Boston June 2-4 2011

2 Hyung Jin Lim and Hoon Sohn ldquoReference-free delamination quantification within multi-layer composite materialsrdquo KSNT Spring Conference May 26-27 Gyeong-Ju Korea 2011

3 Min Koo Kim Hyun Jin Lim Hoon Sohn Chan Yik Park ldquoImpedance-based bolt loosening detection under varying temperature and loadingrdquo Asian Pacific Workshop on Structural Health Monitoring Tokyo Japan November 30-December 2 2010

4 Hyung Jin Lim Chul Min Yeum and Hoon Sohn ldquoModeling of Impact-induced Delamination in a Multilayer Composite Platerdquo The 23rd KKCNN Symposium on Civil Engineering Taipei (Taiwan) November 13-15 2010

5 Yun-Kyu An Hyung Jin Lim and Hoon Sohn ldquoTemperature independent Delamination Detec-tion using data normalizationrdquo The 3rd International Conference on Multi-Functional Materi-als and Structures Chonbuk National Univ Jeonju Korea September 14-18 2010

6 Hoon Sohn Yun-Kyu An Chang Gil Lee Min Koo Kim Hyung Jin Lim ldquoApplication of online NDT techniques to decommissioned bridge testingrdquo the 5th European Workshop on Structural Health Monitoring Sorrento Italy June 29-July 02 2010

7 Min Koo Kim Hyung Jin Lim Hoon Sohn Chan Yik Park ldquoImpedance based damage diag-nosis of a complex composite aircraft wing under changing loading and temperature condi-tionsrdquo the 5th European Workshop on Structural Health Monitoring Sorrento Italy June 29-July 02 2010

8 Yun-Kyu An Min Koo Kim Hyung Jin Lim and Hoon Sohn ldquoValidations of reference-free crack detection techniques through a decommissioned bridge testrdquo 2010 COSEIK Annual Con-ference Jeju Korea April 8-9 2010

- 40 -

PARTICIPANT PROJECTS

1 Development of Smart Steel Members for Online Steel Structure Monitoring POSCO Corporation (Funded 41000000 KRW (41000 USD) for 010511 to 123111)

2 Development of Enabling Technologies for Structural Health Monitoring Korean Agency for Defense Development (Funded 240000000 KRW (240000 USD) out of total 3200000000 KRW (3200000 USD) for 060108-053111)

3 Development of On-board SHM Technologies for Composite Air Vehicles The Boeing Company (Funded 400000 USD (400000000 KRW) for 080108 to 073111)

EDUCATIONAL EXPERIENCES

Fall 2009

Teaching Assistant CE207 ldquoElementary Structural Engineering and Laboratoryrdquo KAIST 3 unit undergraduate course

EXPERIMENTAL EXPERIENTCES

2010 Monitoring of Real Aircraft Composite Wing Segment Korea Agency for Defense De-velopment (ADD) Daejeon Korea Real time monitoring of composite air vehicles fund-ed by Korea Agency for Defense Development (ADD)

2009-2011

Field Monitoring of Real Bridge Structures Gimpo (Ramp G bridge) amp Icheon (Sam-seung bridge) Korea US-Korea collaborative research for bridge monitoring testbeds

HONORS amp AWARDS

Scholarships for outstanding students granted KAIST (2007-2009) The three prominent students in the civil and environmental engineering department of KAIST were selected as recipients This scholarship provided 2000$ per year for three years

- 41 -

Page 2: 복합재료 손상의 무기저 진단 및 정량화 - KAISTssslab.kaist.ac.kr/article/pdf/Thesis_2.pdf · 1 Declaration of Ethical Conduct in Research: I, as a graduate student

복합재료 손상의 무기저 진단 및 정량화

Reference-free Damage Detection Localization

and Quantification in Composite

Reference-free Damage Detection Localization

and Quantification in Composite

Advisor Professor Hoon Sohn by

Hyung Jin Lim

Department of Civil and Environmental Engineering

KAIST

A thesis submitted to the faculty of KAIST in partial fulfillment of the re-quirements for the degree of Master of Science and Engineering The study was conducted in accordance with Code of Research Ethics1

2011 6 20

Approved by

Professor Hoon Sohn

[Major Advisor]

1 Declaration of Ethical Conduct in Research I as a graduate student of KAIST hereby declare that I have not committed any acts that may damage the credibility of my research These include but are not limited to falsi-fication thesis written by someone else distortion of research findings or plagiarism I affirm that my thesis contains honest conclusions based on my own careful research under the guidance of my thesis advisor

복합재료 손상의 무기저 진단 및 정량화

임 형 진

위 논문은 한국과학기술원 석사학위논문으로

학위논문심사위원회에서 심사 통과하였음

2011 년 5 월 30 일

심사위원장

심사위원

심사위원

손 훈 (인)

이 행 기 (인)

김 천 곤 (인)

MCE

20094258

임 형 진 Lim Hyung Jin Reference-free Damage Detection Localization and

Quantification in Composite 복합재료 손상의 무기저 진단 및 정량화 Department

of Civil and Environmental Engineering 2011 41 p Advisor Prof Sohn Hoon

ABSTRACT

A reference-free damage detection localization and quantification technique in a multi

layered composite plate is proposed It is reported that when A0 Lamb wave mode propa-

gates the damage time delay attenuation and multiple reflections are occurred in the dam-

age area In this thesis additional reflected waves from the damage due to multiple reflec-

tions are used as a feature for characterizing the damage First the damage is identified by

detecting the additional reflected waves from the damage area Then the location of the

damage is estimated using combined pitch-catch and pulse-echo methods Finally the dam-

age is quantified using a damage index defined as the ratio between the physical length of

the damage and the wavelength of the A0 Lamb wave mode The matching pursuit method

is utilized for analyzing the obtained signal by decomposing overlapped waves and estimat-

ing the parameters of each wave such as arrival time driving frequency amplitude and so

on The proposed technique has an advantage that the damage can be characterized without

reference data obtained from pristine condition of the monitored structure within a single

path The feasibility of the proposed technique is verified through numerical simulations

and experimental tests In numerical simulation different damage models (length material

properties and location) are investigated In experimental tests a Teflon tape inserted and

impact-induced damage are characterized under temperature variations The results are

suggesting a reliable damage characterization in composites

Keywords Reference-free PZT transducers Composites Finite element method Matching

pursuit method

i

TABLE OF CONTENTS

ABSTRACT middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot i

TABLE OF CONTENTS middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot iii

LIST OF TABLES middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot iv

LIST OF FIGURESmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot v

CHAPTER 1 INTRODUCTION

11 Motivation middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 1

12 Literature review middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 1

13 Objective and uniqueness middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 3

14 Organization middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 4

CHAPTER 2 INTERACTION OF LAMB WAVE WITH DAMAGE

21 A0 Lamb wave mode delay and attenuation in the time domain middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 5

22 Multiple reflections in the damage area middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 6

CHAPTER 3 THEORETICAL DEVELOPMENT

31 Overview middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 7

32 Damage index (m-value) middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 8

33 Damage detection algorithm (Level 1) middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 10

34 Damage quantification algorithm (Level 3) middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 11

35 Damage localization algorithm (Level 2amp3) middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 11

36 Matching pursuit method considering the dispersion effect of Lamb wave

middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 13

CHAPTER 4 NUMERICAL SIMULATION

41 Composite modeling methodmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 15

42 Damage modeling method middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 16

43 Numerical simulation description middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 17

44 Numerical simulation results middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 19

ii

CHAPTER 5 EXPERIMENTAL VERIFICATION

51 Overview middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 21

511 Specimen middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 22

512 Data acquisition system and PZT middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 22

513 Impact test middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 23

514 Temperature effect test middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 24

52 Teflon inserted damage middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 25

53 Impact-induced damagemiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 26

54 Temperature effect test middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 27

541 Teflon inserted damage middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 27

542 Impact-induced damage middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 29

CHAPTER 6 CONCLUDING REMARKS

61 Executive summary middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 31

62 Suggestion for further study middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 32

REFERCENCES middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 33

SUMMARY (IN KOREAN) middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 37

ACKNOWLEDGEMENTS (IN KOREAN)

CURRICULUM VITAE

iii

LIST OF TABLES

I Material property of simulation model and test specimen middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 17

II Damage cases for numerical simulation middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 18

III Damage characterization for the numerical simulation models middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 20

IV Damage characterization for the Teflon inserted damage specimen middotmiddotmiddotmiddotmiddotmiddotmiddot 25

V Damage characterization for the impact-induced damage specimen middotmiddotmiddotmiddotmiddotmiddotmiddot 27

VI Temperature effect test for the Teflon inserted damage specimen middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 28

VII Temperature effect test for the impact induced damage specimen middotmiddotmiddotmiddotmiddotmiddotmiddot 30

iv

LIST OF FIGURES

1 Lamb wave propagation characteristics (a) A0 Lamb wave mode time delay and

attenuation after passing through a damage (b) The time delay and amplitude varia-

tion of A0 Lamb wave mode under temperature variation in pristine condition of

structure Because of this conventional damage detection techniques in composites

have limitations for real structure application middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 5

2 Multiple reflections in the damage area The multiple reflections of A0 Lamb

waves in the damage area are utilize as a feature for characterization of the damage

in this study middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 6

3 Parameters and notations of the proposed damage characterization technique

Damage is detected by identifying the additional reflected wave from the damage due

to multiple reflections (dashed lines) Localization and quantification of the damage

is carried out by measuring the arrival time of transmitted and reflected waves from

the damage middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 7

4 Two different FEM models which have different 119889119889 and 119907119907119889119889 (or G) This repre-

sents that the damage cannot be characterized uniquely based on the measured data

available middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 9

5 (a) Example of the overlapped signal with comparing approximated signal using

the matching pursuit method (b) Decomposed signal using the matching pursuit

method The matching pursuit method provides a conventional way to decompose

and identify each wave middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 9

6 Description of the numerical simulation model middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 17

7 Comparison of A0 Lamb wave mode signals from numerical simulation be-

tween intact and damage cases Time delay attenuation and multiple reflections due

to damage simulated successfully using the damage modeling method described in

v

the previous section middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 19

8 Test specimens for the experimental verification In the Teflon inserted damage

Teflon tapes with 25mm diameter were inserted at the center of the specimen The

impact-induced damage was introduced to the specimen using the impact test equip-

ment middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 21

9 Overall experimental setup (a) Data acquisition (DAQ) system and (b) Installed

circular PZT transducer on the specimen with its dimension middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 23

10 (a) Impact test equipment for introducing impact-induced damage and (b) Tip

of the impact tester with 10mm diametermiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 23

11 Test specimen in temperature chamber middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 24

12 A0 Lamb wave mode obtained using pitch-catch method from the Teflon in-

serted damage specimen (a) The reflected wave from damage is shown (b) The Tef-

lon inserted damage is detected by capturing the reflected wave middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 25

13 The signal obtained from the test specimen subjected to impact The impact

was introduced at the same location of the specimen and the signal was obtained after

each impacts As the number of impact is increased the transmitted A0 Lamb wave

mode is delayed and attenuated middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 26

14 False alarm test for the proposed damage detection algorithm in pristine condi-

tion of the specimen Reflected wave from damage is not captured false alarm is not

occurred middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 26

15 Signals obtained from the Teflon inserted damage specimen under temperature

variation Time delay and amplitude variation are occurred middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 27

16 False alarm test under temperature variations It is shown that proposed dam-

vi

age characterization algorithm is robust under temperature variationsmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 29

17 Signals obtained from impact-induced damage specimen (Impact 3) under

temperature variation middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 29

vii

CHAPTER 1 INTRODUCTION

11 Motivation

Composite materials has been increasingly demands in modern structures ranging from the

latest aircraft to retrofitted bridge deck due to its light weight superior specific strength and

stiffness properties However the occurrence of damage due to abrupt impact or sequential

accumulation of defect may considerably degrade structural strength or stiffness and reduce

the safety and reliability of the system Recently Lamb waves based damage detection tech-

niques using piezoelectric (PZT) transducers have emerged as one of the promising structural

health monitoring (SHM) technology for detecting hidden damage because of the possibility

of long range inspection and high sensitivity [1-4]

12 Literature review

This section will discuss researches that have been previously carried out on topics of

guided wave based SHM techniques using PZT transducers for damage in composite materi-

als The early investigations of interaction between Lamb wave and damage in composite can

be found in Rose et al and Guo et al [5-6] From further studies reflected signals from dam-

age were shown to be an effective method for damage diagnosis and localization using PZTs

[7-9] The reflected signals from damage were collected using one PZT as an actuator and the

other PZT as a sensor However the techniques cannot be used to determine the extent of

- 1 -

damage with reasonable precision

Kesseler et al proposed a technique monitoring the wavelet coefficients of transmitted ul-

trasonic signal amplitude as a potential way to damage detection [10] One drawback of this

technique is that the variability in coupling and the sensitivity of installed PZT transducers

can induce changes in the amplitude and energy of the signal that are cause false-alarm Oth-

er wavelet based delamination detection and localization technique developed using PZT

sensor network and outlier analysis [11]

Damage detection and quantification in composite plate was also investigated by group de-

lay measurement due to damage [12] However it is assumed that the group velocity in dam-

age area as the group velocity propagates half thickness of the structure in dispersion curve in

the quantification procedure Thus the damage quantification was investigated only Teflon

inserted damage case

Sohn et al proposed a damage detection technique which does not rely on any past base-

line signals to assess damage in composite plates by using an enhanced time reversal method

[13] The damage is assessed using the violation of time reversibility when nonlinearity is

caused by the damage along a direct wave path Thus it allows instantaneous identification

of delamination without requiring the baseline signal for comparison

- 2 -

13 Objective and uniqueness

In this study a reference-free damage characterization technique including detection lo-

calization and quantification of impact-induced damage in composite materials is developed

The general damage characterization can be categorized into the following levels [14]

Level 0 Define damage

Level 1 Determine the presence of damage within the structure

Level 2 Locate the regions of damage

Level 3 Quantify the damage

Level 4 Predict the remaining service life of the structure

Through the proposed technique from level 1 to level 3 can be achieved

Most of common approaches of damage characterization techniques described in the pre-

vious section use prior obtained baseline data However these approaches can cause false

alarm due to the effect of environmental variations such as temperature and ambient vibra-

tions [15]

The uniqueness of this study is

1 Damage is characterized without using prior measured baseline data

2 Robustness of the proposed technique under temperature variation is investigated

3 Additional waves reflected from inside damage have been used in addition to transmit-

ted and reflected waves

- 3 -

14 Organization

The organization of this thesis is summarized as follows

Chaper1 A general introduction including motivation literature review objective and

uniqueness of this study is described

Chaper2 Interactions between Lamb wave mode and damage in composites such as time

delay attenuation and multiple reflections are summarized Damage index (m-value) for

damage characterization is defined

Chaper3 Algorithms for the reference-free damage characterization technique using m-

value are developed Modified matching pursuit for this study is also described

Chaper4 Verification of the algorithm using numerical simulation is presented Various

modeled damage cases are characterized

Chaper5 Experimental verifications are described Two damage cases (Teflon inserted and

impact-induced damage) are investigated for the feasibility of the proposed damage charac-

terization technique Temperature effect test using temperature chamber is also carried out

Chaper6 Conclusion and summary of the present study is reported In advance suggestions

and future work recommendations are represented

- 4 -

CHAPTER 2 INTERACTION OF LAMB WAVE WITH DAMAGE

21 A0 Lamb wave mode delay and attenuation in the time domain

Time delay and attenuation of the A0 mode Lamb wave after passing through the damage

area due to effective thickness reduction were reported in previous researches [812] Moreo-

ver the attenuation of the A0 Lamb wave mode is occurred by degrading of shear modulus of

in the damage area However the time delay and amplitude variation of the A0 Lamb wave

mode also occurred even when the structure is health condition by the material property

changing due to temperature variation of the structure (Fig 1) Thus conventional damage

characterization methods using the time delay and attenuation of A0 Lamb wave mode as a

feature can cause false-alarm in real application

11 115 12 125 13-10

-5

0

5

10

Time (ms)

Vol

tage

(mV

)

IntactDamage

Reflected wave from Damage

(a)

11 115 12 125 13-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

0 oC20 oC50 oC

(b)

Fig 1 Lamb wave propagation characteristics (a) A0 Lamb wave mode time delay and atten-uation after passing through a damage (b) The time delay and amplitude variation of A0 Lamb wave mode under temperature variation in pristine condition of structure Because of this conventional damage detection techniques in composites have limitations for real struc-ture application

- 5 -

22 Multiple reflections in the damage area

Standing wave patterns due to multiple reflections of A0 Lamb wave mode in the damage

area were observed through numerical simulations and ultrasonic wave-field imaging using a

scanning laser vibrometer [16] The formation of standing waves in damage is illustrated

through Fig 2 When the incident waves enter the damage area a significant portion of the

waves are reflected back from the exit Then the reflected waves travel through the damage

and undergo reflection again at the original entrance As a consequence of multiple reflec-

tions at the entrance and the exit a considerable amount of ultrasonic energy is trapped inside

the damage area [17-19]

Multiple Reflections

Incident Wave Transmitted Wave

Damage

Fig 2 Multiple reflections in the damage area The multiple reflections of A0 Lamb waves in the damage area are utilize as a feature for characterization of the damage in this study

- 6 -

CHAPTER 3 THEORETICAL DEVELOPMENT

31 Overview

Damage

l = l1 + d + l2

d = vdtd l1 = vit1 l2 = vit2

t = t1 + td + t2

Excitation and Sensing PZT Sensing PZT

l

Transmitted Wave

Reflected Waves

Incident Wave

Fig 3 Parameters and notations of the proposed damage characterization technique Damage is detected by identifying the additional reflected wave from the damage due to multiple re-flections (dashed lines) Localization and quantification of the damage is carried out by measuring the arrival time of transmitted and reflected waves from the damage

In this section the reference-free damage characterization algorithm is proposed Damage

diagnosis and quantification are achieved by identifying the additional reflected waves due to

multiple reflections in the damage area (dashed lines in Fig 3) Moreover damage is local-

ized using additional information from pulse-echo method implement For the localization

and the quantification of damage the variables such as the path length and A0 Lamb wave

mode velocity in damage area are defined (Fig 3) For example 119897119897 is expressed as

119897119897 = 1198971198971 + 119889119889 + 1198971198972 (1)

where 1198971198971 119889119889 and 1198971198972 are the distance from the excitation PZTs to the entrance of damage the

damage length and the distance from the exit of damage to the sensing PZT respectively

- 7 -

Moreover 119907119907119894119894 and 119907119907119889119889 are also defined as the A0 Lamb wave mode velocity in pristine area

and damage area respectively Note that the whole procedure of the damage characterization

is accomplished based solely on the signals obtained from the current state of the structure

32 Damage index (m-value)

The proposed reference-free damage characterization technique utilizes the waves reflect-

ed additionally from the damage area as a feature The arrival time difference of the transmit-

ted and reflected A0 Lamb wave modes reflected from the exit and the entrance of the delam-

ination area one times each (∆119905119905) contains the information of the physical length of the dam-

age (119889119889) and A0 Lamb wave mode velocity in the damage area (119907119907119889119889) The wave velocity in the

damage area which is related to the shear modulus (G) represents the severity of the damage

∆119905119905 = 2119905119905119889119889 =2119889119889119907119907119889119889

(2)

However based on the measured data available the 119889119889 and 119907119907119889119889 (or G) cannot be uniquely

determined Fig 4 shows two FEM models which have different physical length and wave

velocity (G) of the damage Detail explanation of delamination modeling technique will be

covered in section 4 The similar signal in the time domain and ∆119905119905 are obtained from the

two different damage cases

- 8 -

Damage 1

Damage 2

20

Unit mm15

90 G value reduction

943 G value reduction

m-value = 248

m-value = 248

015 02 025 03 035

-2

-1

0

1

2

Time (ms)

Ampl

itude

Damage 1Damage 2

Fig 4 Two different FEM models which have different 119889119889 and 119907119907119889119889 (or G) This represents that the damage cannot be characterized uniquely based on the measured data available

Thus in this study the damage index (m-value) which can be uniquely defined based on

the available data is introduced The m-value is defined as the ratio of the physical length of

the damage and wavelength of the A0 Lamb wave mode in the damage area (120582120582119889119889)

119898119898 =119889119889120582120582119889119889

=119889119889119889119889119907119907119889119889

(3)

In many cases the transmitted and reflected waves are overlapped and it is hard to decom-

pose and calculate the arrival times of them Matching pursuit method is an alternative way

to accurately decompose of the overlapped waves and estimate the parameters of each wave

such as arrival time (Fig 5) Detail explanation of the matching pursuit method is covered in

section 36

015 02 025 03-2

-1

0

1

2

Time (ms)

Am

plitu

de

OriginalApproximated

(a)

015 02 025 03-2

-1

0

1

2

Time (ms)

Ampl

itude

Wave 1Wave 2

(b)

Fig 5 (a) Example of the overlapped signal with comparing approximated signal using the matching pursuit method (b) Decomposed signal using the matching pursuit method The matching pursuit method provides a conventional way to decompose and identify each wave

- 9 -

33 Damage detection algorithm (Level 1)

The damage diagnosis is accomplished by extracting the additional reflected A0 Lamb

wave modes from damage due to multiple reflections If A0 Lamb wave modes are decom-

posed perfectly it is trivial that one transmitted A0 Lamb wave mode is always captured in

the response signal If damage exists other waves in the response signal behind the transmit-

ted A0 Lamb wave mode are presented Since this approach does not rely on previously col-

lected data the diagnosis is independent from environmental variation such as temperature

changes

It is reported that the dispersion of Lamb waves cause wave-packets to spread out in space

and time as they propagate through a structure [20] Thus the reflected waves from the dam-

age have longer pulse duration than the transmitted wave due to their longer propagating dis-

tance In this study the matching pursuit method is modified to capture more spread wave

than the transmitted A0 Lamb wave mode Detail explanation of the modified matching pur-

suit method is described in latter section

Let 1199041199041 and 1199041199042 as the pulse duration (scale) of the transmitted and additional reflected A0

Lamb wave mode respectively

ldquoIf 1199041199042 gt 1199041199041 damage exists

Otherwise there is no damagerdquo

- 10 -

34 Damage quantification algorithm (Level 3)

As mentioned in the previous section damage is quantized by estimating 119889119889 and 119907119907119889119889 using

pitch-catch method The procedure to estimate 119889119889 and 119907119907119889119889 is

(i) Measure 119897119897 119905119905 and 119905119905119889119889

(ii) Assume the range of 119889119889 (0 lt 119889119889 lt 119897119897)

(iii) Calculate 119907119907119894119894 and 119907119907119889119889 with a constraints (0 lt 119907119907119889119889 lt 119907119907119894119894) (Eq 4 and 5)

(iv) Estimate m-value and find a feasible range of 119889119889 and 119907119907119889119889

119907119907119894119894 =119897119897 minus 119889119889119905119905 minus 119905119905119889119889

(4)

119907119907119889119889 =119889119889119905119905119889119889

(5)

35 Damage localization algorithm (Level 2amp3)

For damage localization 1198971198971 is estimated by measuring the arrival time from the excitation

PZT to the entrance of delamination (1199051199051) with additional pulse-echo method implementation

The procedure is similar with the previous damage quantification algorithm

(i) Measure 119897119897 119905119905 119905119905119889119889 and 1199051199051

(ii) Assume the range of 1198971198971 (0 lt 1198971198971 lt 119897119897)

(ii) Compute possible combinations of 1198971198972 119889119889 and 119907119907119889119889 with constraints

(0 lt 119907119907119889119889 lt 119907119907119894119894 119889119889 gt 0) (Eq 5 6 and 7)

- 11 -

1198971198972 = 1198971198971119905119905 minus 1199051199051 minus 119905119905119889119889

1199051199051 (6)

119889119889 = 119897119897 minus 1198971198971 minus 1198971198972 (7)

Since a PZT for pulse-echo method acts as both an actuator and a sensor the obtained sig-

nal contains both the excitation and the response signal Thus the reflected wave from the

entrance of damage can be buried into the tail of excitation signal and cause error on arrival

time of reflected wave from damage Even in this case damage can be localized and quan-

tized using the constraints When first wave packet reflected from the entrance of damage is

buried into the excitation signal the next wave packet reflected from the exit of damage can

be utilized alternatively If the algorithm regards the second reflected wave as the first re-

flected wave there are no combinations of estimated variable meeting the constraints Then

the algorithm repeats the process by regarding the wave as the second reflected wave the

constraints can select proper combination of the estimated variables

The estimated values have range bounded by the constraints When additional information

of 119907119907119894119894 is utilized the damage can be characterized more accurately In this study three meth-

ods are proposed and verified as information level of 119907119907119894119894

Method 1 Solely based on the current data measurement (Reference-free)

Method 2 Limit the possible range of 119907119907119894119894 (Prior knowledge of the system)

Method 3 Measure 119907119907119894119894 (Additional data need to be collected)

Method 1 described in this section can be utilized without any information of 119907119907119894119894 Meth-

od 2 produces more accurate estimation results using prior knowledge of the system such as

- 12 -

119907119907119894119894 variation under temperature changes Method 3 gives unique estimated value using meas-

ured 119907119907119894119894 from other path in the same time In section 4 and 5 damage is characterized using

these three methods and the results are compared each another

36 Matching pursuit method considering the dispersion effect of Lamb

wave

Matching pursuit method has been developed for signal compression approximation and

characterization of obtained signal and function [21-23] The key concept of the matching

pursuit method is that any signal (function) can be decomposed into a linear expansion of

signals selected from a pool of basis signals (functions) called lsquodictionaryrsquo

Let 119891119891 be a function to be analyzed then 119891119891 can be approximated a set of functions select-

ed from the dictionary 119863119863 in order to best match This is done by successive approximations

with orthogonal projections Let1198921198921205741205740 isin 119863119863 the function 119891119891 can be decomposed into

119891119891 = lang1198911198911198921198921205741205740rang1198921198921205741205740 + 119877119877119891119891 (8)

where 119877119877119891119891 is the residual function after approximating 119891119891 using 1198921198921205741205740 1198921198921205741205740 is orthogonal to

119877119877119891119891 hence

1198911198912 = lang1198911198911198921198921205741205740rang2

+ 1198771198771198911198912 (9)

To minimize 119877119877119891119891 1198921198921205741205740 is chosen to maximize lang1198911198911198921198921205741205740rang The matching pursuit method is

an iterative algorithm that decomposes the residue 119877119877119891119891 by projecting it on a function of 119863119863

that matches 119877119877119891119891 almost at best as it was done for 119891119891 The residue 119877119877119899119899119891119891 is decomposed into

119877119877119899119899119891119891 = lang119877119877119899119899119891119891119892119892120574120574119899119899rang119892119892120574120574119899119899 + 119877119877119899119899+1119891119891 (10)

- 13 -

the decomposition is carried up to the order 119898119898 119891119891 can be expressed as

119891119891 = lang119877119877119899119899119891119891119892119892120574120574119899119899rang119892119892120574120574119899119899

119898119898minus1

119899119899=0

+ 119877119877119898119898119891119891 (11)

The original function 119891119891 is decomposed into a sum of dictionary elements that are chosen to

best match its residues

In this study the matching pursuit method is utilized to accurately decompose of the over-

lapped Gaussian pulses and estimate the parameters of the pulse which used for damage

characterization The 119894119894119905119905ℎ Gaussian pulse in the dictionary for approximating the obtained

signal is expressed as

119892119892120574120574119894119894 = 119891119891(119860119860119894119894 119906119906119894119894 119904119904119894119894 119889119889119894119894120601120601119894119894) (12)

where 119860119860119894119894119906119906119894119894119904119904119894119894119889119889119894119894 and 120601120601119894119894 represent amplitude arrival time scale (pulse duration) excita-

tion(central) frequency and phase of the Gaussian pulse respectively The parameters of the

pulse are estimated using a nonlinear regression based optimization method (Gauss-Newton

method) based on an initial guess of the parameters Detail procedure for estimating the pa-

rameters of the pulse is expressed in Hong et al [23]

Among the estimated parameters of the pulses pulse duration (scale) represents the disper-

sion effect If a wave-packet is spread out the estimated pulse duration is increased Using

this feature the matching pursuit method is modified to capture waves which have longer

pulse duration than transmitted A0 Lamb wave mode If the estimate pulse duration of a wave

is less than the transmitted wave the modified matching pursuit method regards it as error

- 14 -

CHAPTER 4 NUMERICAL SIMULATION

41 Composite modeling method

Composite materials can be modeled from nano to macro level depends on its engineering

purpose In general four methods are utilized for modeling of composite materials First

composite is modeled as a homogenous material expressed in terms of effective material

properties [24] The effective material properties are estimated using the modulus and the

volume fraction of each constituent theoretically (Rule of mixture) and measured experimen-

tally Second a model is established that each laminate which have own material properties

are totally bonded together In this model the stacking sequence and direction of each lami-

nates are considered additionally [25-27] Third method for composite modeling in micro

level called micromechanics [28-30] This model considers the interaction between fibers and

matrix in laminate Last multiscale modeling which the scope of modeling is extended to

molecular or quantum level of constituents [31-33] This model starts from molecular me-

chanics or molecular dynamics then the scale of model leads to the nano micro and macro

level by incorporating small scale models As the scale of models is smaller it provides more

logical way than other large scale of models while the computational cost is more expensive

The scope of this study is focused on the dynamic behavior of composite structure the beam

or plate model was utilized for the numerical simulation

- 15 -

42 Damage modeling method

Generally three damage models have been developed and utilized in the real applications

First the stiffness reduction model called as the equivalent element model was introduced

[17 34-35] This model is simply realized by replacing the region around damage area by a

region with adjusted material properties (Youngrsquos modulus shear modulus density etc)

globally Next the duplicated node model is done by disconnecting the element connectivity

assuming there is no interaction among the delaminated surface This model can be utilized

when the location and the extension of damage is important issue [34-36] Last the contact

mechanics model using the principles of contact mechanics the two interfacial forces (con-

tact resistance and sliding resistance) in the delaminated surface [37-38] In this model the

delaminated surfaces are modeled as nonlinear functions of the interlaminar normal strain

and the interlaminar shear strain This model is computationally expensive while provides

more logical way than other models There is however no universally accepted model for

general impact-induced damage in composites In the case of impact-induced damage it is

impossible to predict or estimate the location and number of delaminated layers Thus from

the uncertainty of impact-induced damage the equivalent element model can be considered

as the effective damage model

- 16 -

43 Numerical simulation description

TABLE I Material property of simulation model and test specimen

Tensile Strength 657 MPa Tensile Modulus 59 GPa Shear Strength 804 MPa Shear Modulus 24 Gpa Specific Gravity 16

PZT A

PZT C

PZT BDamage

3

250 125 125

500 Unit mm

Fig 6 Description of the numerical simulation model

The feasibility of the proposed damage characterization technique is verified by 2-D FEM

simulation using MSC PATRAN and NASTRAN Fig 6 presents the dimension PZTs and

damage location of the simulation model In the model 05mm x 05mm square shell element

was used for modeling of the A0 mode whose flexural wave characteristic The element size

was determined by considering the wavelength of the propagation Lamb wave modes The

material properties of the simulation model and damage cases are presented in Table I and

Table II respectively

PZT A and C act as both actuator and sensor for collecting pulse-echo signal They were

installed on the plate which two of them are both-side collocated for the A0 Lamb wave

- 17 -

mode decomposition [39-40] PZT B was only used as a sensor for collecting pitch-catch sig-

nal For simulating the excitation and sensing of PZTs the pin-force model was adopted that

an input force apples to the end nodes of excitation PZT and averaging element strains corre-

sponding sensing area A Gaussian pulse (tone-burst) 50 kHz central frequency was excited

as an input signal

TABLE II Damage cases for numerical simulation

119949119949120783120783 (mm) 119941119941 (mm) G value reduction ()

Case 1 650 200 90 darr

Case 2 675 150 90 darr

Case 3 675 150 93 darr

Case 4 900 150 90 darr

- 18 -

44 Numerical simulation results

015 02 025 03 035

-2

0

2

Time (ms)

Am

plitu

de

IntactDamage

Reflected wave from damage

(a) Pitch-catch signal

015 02 025 03 035

-2

0

2

Time (ms)

Am

plitu

de

IntactDamage

Reflected waves from damage

(b) Pulse-echo signal Fig 7 Comparison of A0 Lamb wave mode signals from numerical simulation between intact and damage cases Time delay attenuation and multiple reflections due to damage simulated successfully using the damage modeling method described in the previous section

Fig 7 shows that comparison of A0 Lamb wave mode signals between intact and one of

damage cases in the simulation The features of A0 Lamb wave mode passing through dam-

age area described in section 2 are successfully simulated Thus the equivalent element

method is proper for modeling damage in composites In the damage cases the additional

waves reflected from damage is observed in both pitch-catch and pulse echo signal

Table III presents the result of the proposed damage characterization algorithm using the

numerical simulation cases shown in Table II In lsquoMethod 2rsquo 119907119907119894119894 variation is assumed to plusmn

5 of the normal value (1190~1315ms) The normal value of 119907119907119894119894 for lsquoMethod 3rsquo is meas-

ured as 1253 ms In each case the damage is characterized using the proposed technique

Note that as the information of 119907119907119894119894 is specified the estimated value is closer to the exact

value

- 19 -

TABLE III Damage characterization for the numerical simulation models

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

Case 1

Exact 650 200

Method 1 248 50~70 79~414 160~834

Method 2 248 603~667 135~241 271~486

Method 3 248 635 188 379

Case 2

Exact 675 150

Method 1 193 60~75 01~251 37~651

Method 2 193 618~682 114~222 295~575

Method 3 193 65 168 435

Case 3

Exact 675 150

Method 1 233 60~75 66~302 140~649

Method 2 233 668~738 85~196 182~419

Method 3 233 703 140 301

Case 4

Exact 900 150

Method 1 204 75~100 48~349 118~856

Method 2 204 860~951 107~216 264~530

Method 3 204 906 162 397

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1253 ms 1190~1315 ms Measured 119907119907119894119894 1253 ms

- 20 -

CHAPTER 5 EXPERIMENTAL VERIFICATION

51 Overview

In this section experimental study is investigated so that the proposed damage characteri-

zation technique Overall experimental setup and test specimen are shown in Fig 8 and 9

The experimental setup consists of test specimens data acquisition (DAQ) system and self-

sensing circuit Self-sensing circuit was installed for pulse-echo measurement using PZTs

[40-41]

PZT A(PZT C) PZT B

500

500

140

75

250

Teflon Inserted Damage (d =25)

Unit mm

Thickness = 3

PZT A(PZT C) PZT B

500

500

150

90

250

Impact Location

Unit mm

Thickness = 3

(a) Teflon inserted damage specimen

(b) Impact-induced damage specimen

Fig 8 Test specimens for the experimental verification In the Teflon inserted damage Teflon tapes with 25mm diameter were inserted at the center of the specimen The impact-induced damage was introduced to the specimen using the impact test equipment

- 21 -

511 Specimen

Two carbon composite plates with a dimension of 500times500times3mm3 were manufactured by

Nexcomes Inc by stacking 12 layers of CF3327 3K prepreg (woven fabrics type) manufac-

tured by Hankuk Fiber Inc The material properties of the specimen are presented in Table I

One of the specimens Teflon tapes with 25mm diameter was inserted in every 3 layers at the

center of the specimen to mimic the damage The other specimen is utilized for the character-

ization of real impact-induced damage

512 Data acquisition (DAQ) system and PZT

The data acquisition (DAQ) system consists of an arbitrary waveform generator (AWG) a

high speed signal digitizer (DIG) and two multiplexers (MUX) Using the 16-bits AWG

tone-burst input signal with a plusmn10 peak-to-peak voltage were generated The response signal

was measured by the 14-bit DIG For the signal-to-noise ratio reduction both pitch-catch and

pulse-echo signals were measured 150 times and averaged in the time domain 3 identical

circular PZTs of 10mm diameter manufactured by APC International Ltd were installed on

the specimen A Gaussian pulse (tone-burst) 50 kHz central frequency was excited as an in-

put signal The excitation frequency was chosen in order to avoid overlapping of the S0 and

A0 Lamb wave modes and generating of the higher modes (S1 A1hellip)

- 22 -

PZT Actuator

PZT Sensor

AWG

DIG

DAQ System Specimen

SensingActuatingMUX

1

23

4

1 Sending signal to MUX

4 Sending signal to DIG

Self-SensingCircuit

18 Unit mm

10

(a)

(b)

Fig 9 Overall experimental setup (a) Data acquisition (DAQ) system and (b) Installed circu-lar PZT transducer on the specimen with its dimension

513 Impact test

The impact-induced damage was introduced to the specimen using the impact test equip-

ment presented in Fig 10 A 5kg mass with 10mm tip was dropped from 015m height onto

the same location of the specimen The signals were obtained at each number of impacts

Since the reflected wave is not detected at the first impact case the data obtained from the

second and third impact were utilized for the analysis

10

Unit mm (a)

(b)

Fig 10 (a) Impact test equipment for introducing impact-induced damage and (b) Tip of the impact tester with 10mm diameter

- 23 -

514 Temperature effect test

In reality structures are subjected to changing environmental conditions such as tempera-

ture that affect measured signal and cause false-alarm Thus the robustness of the proposed

damage characterization technique under temperature variation in investigated (Fig 11) The

specimens were placed inside a temperature chamber and a thermocouple was installed on

the specimen to measure temperature of surface of the specimen The signal was obtained

under three temperature cases (0 20 and 50degC) and the humidity was kept at 30

Temperature Chamber

Test Specimen

Fig 11 Test specimen in temperature chamber

- 24 -

52 Teflon inserted damage

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Reflected wave from damage

(a)

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Transmitted WaveReflected Wave

Detected reflected wave

(b)

Fig 12 A0 Lamb wave mode obtained using pitch-catch method from the Teflon inserted damage specimen (a) The reflected wave from damage is shown (b) The Teflon inserted damage is detected by capturing the reflected wave

Fig 12 shows the obtained pitch-catch signal from the Teflon inserted damage specimen

and damage diagnosis result The Teflon inserted damage is detected successfully by captur-

ing the reflected wave Table IV presents the result of the Teflon inserted damage characteri-

zation 119907119907119894119894 is measured as 1213ms and the variation is assumed to plusmn 5 of the normal val-

ue (1152~1273ms) The results indicate that the proposed damage characterization approach

works well in real application

TABLE IV Damage characterization for the Teflon inserted damage specimen

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

Exact 635 25

Method 1 123 60~75 73~338 298~1220

Method 2 123 619~662 229~310 934~1188

Method 3 123 631 284 1161

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1213 ms

- 25 -

53 Impact-induced damage

11 115 12 125 13-4

-2

0

2

4

Time (ms)

Vol

tage

(mV

)

IntactImpact2Impact3

Reflected wave from damage

(a) Pitch-catch signal

11 115 12 125 13 135

-05

0

05

1

Time (ms)

Volta

ge (m

V)

IntactImpact2Impact3

Boundary reflection

Reflected wave from damage

(b) Pulse-echo signal

Fig 13 The signal obtained from the test specimen subjected to impact The impact was in-troduced at the same location of the specimen and the signal was obtained after each impacts As the number of impact is increased the transmitted A0 Lamb wave mode is delayed and attenuated

The obtained signals from impact-induced damage specimen are presented in Fig 13 It is

observed that as the number of impact is increased the transmitted A0 Lamb wave mode is

more delayed because the wave velocity in the damage area is reduced Fig 14 shows the

false-alarm test result whether the proposed damage detection algorithm works even in pris-

tine condition of the test specimen It is shown that no false alarm is occurred because no ad-

ditional reflected wave is captured

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Transmitted Wave

(a) Transmitted wave

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Reflected Wave

No reflected wave

(b) Reflected wave

Fig 14 False alarm test for the proposed damage detection algorithm in pristine condition of the specimen Reflected wave from damage is not captured false alarm is not occurred

- 26 -

TABLE V Damage characterization for the impact-induced damage specimen

of Impacts m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

2

(1470 J)

Method 1 080 65~90 75~246 470~1157

Method 2 080 801~885 114~245 711~1134

Method 3 080 843 179 1122

3

(2205 J)

Method 1 148 50~100 06~380 21~1134

Method 2 148 758~838 249~368 843~1125

Method 3 148 798 308 1044

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1213 ms

Table V presents the characterization results of the impact-induced damage case Here the

measured 119907119907119894119894 and the assumed variation are same as the artificial damage case It is observed

that as the number of impact is increased 119889119889 is increased while 119907119907119889119889 is reduced Thus it is

validated that the proposed damage characterization approach estimates the increment of the

damage well

54 Temperature effect test

541 Teflon inserted damage

11 115 12 125 13-10

-5

0

5

10

Time (ms)

Vol

tage

(mV

)

0 oC20 oC50 oC

(a) Pitch-catch signal

11 115 12 125 13 135

-1

0

1

2

3

Time(ms)

Volta

ge (m

V)

0 oC20 oC50 oC

(b) Pulse-echo signal

Fig 15 Signals obtained from the Teflon inserted damage specimen under temperature varia-tion Time delay and amplitude variation are occurred

Fig 15 shows the obtained signals from the Teflon inserted damage specimen subjected to

- 27 -

temperature variation Time delay and amplitude variations are occurred there is no addition-

al damage is introduced The analyzed results of the obtained signals are reported in Table VI

The measured 119907119907119894119894 is 1247 ms at 0 degC 1213 ms at 20 degC and 1168 ms at 50 degC respectively

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms (1152~1273 ms)

The estimated results are also changed under temperature variations the fluctuation of esti-

mated range and exact value is small

TABLE VI Temperature effect test for the Teflon inserted damage specimen

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

0 degC

Exact 635 25

Method 1 113 60~70 79~268 352~1291

Method 2 113 560~619 233~344 1034~1172

Method 3 113 606 257 1141

20 degC

Exact 635 25

Method 1 123 60~75 73~338 298~1220

Method 2 123 619~662 229~310 934~1188

Method 3 123 631 284 1161

50 degC

Exact 635 25

Method 1 137 60~75 72~328 263~1228

Method 2 137 622~687 183~299 666~1086

Method 3 137 642 264 960

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1247 ms at 0 degC 1213 ms at 20 degC and 1168 ms at 50 degC

- 28 -

542 Impact-induced damage

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Vol

tage

(mV

)

Transmitted WaveReflected Wave

No reflected wave

(a) 0 degC

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Transmitted WaveReflected Wave

No reflected wave

(b) 50 degC

Fig 16 False alarm test under temperature variations It is shown that proposed damage char-acterization algorithm is robust under temperature variations

The false alarm test in pristine condition of the impact-induced damage specimen under

temperature variation was investigated (Fig 16) It is validated that the proposed damage di-

agnosis algorithm is robust under temperature variations Fig 17 shows the obtained signals

from the impact-induced damage (impacted 3 times) specimen under temperature variation

The analyzed results of the obtained signals are reported in Table VII The measured 119907119907119894119894 in

each temperature and the assumed variation are same as the artificial damage case

11 115 12 125 13-5

0

5

Time (ms)

Volta

ge (m

V)

0 oC20 oC50 oC

(a) Pitch-catch signal

11 115 12 125 13 135

-1

0

1

2

3

Time(ms)

Vol

tage

(mV

)

0 oC20 oC50 oC

(b) Pulse-echo signal

Fig 17 Signals obtained from impact-induced damage specimen (Impact 3) under tempera-ture variation

- 29 -

TABLE VII Temperature effect test for the impact induced damage specimen (Impact 3 case)

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

0 degC

Method 1 135 75~90 72~310 267~1148

Method 2 135 705~779 264~381 978~1126

Method 3 135 763 289 1071

20 degC

Method 1 148 50~100 06~380 21~1134

Method 2 148 758~838 249~368 843~1125

Method 3 148 798 308 1044

50 degC

Method 1 153 75~100 53~415 174~1227

Method 2 153 804~889 214~336 702~1103

Method 3 153 815 320 1050

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1247 ms at 0 degC 1213 ms at 20 degC and 1168 ms at 50 degC

- 30 -

CHAPTER 6 CONCLUDING REMARKS

61 Executive summary

In this study the reference-free damage detection localization and quantification tech-

nique in a multi layered composite plate is proposed Additional reflected A0 Lamb wave

mode from the damage due to multiple reflections in damage area is used as a feature for

characterizing the damage First damage is characterized in terms of a damage index (m-

value) defined as the ratio of the physical length of the damage to the wavelength of the A0

Lamb wave mode Then the range of possible damage location length and severity are esti-

mated with constraints This procedure can be accomplished without prior baseline data To

realize the proposed technique the matching pursuit technique is modified for reduction of

the error by considering the dispersive characteristic of Lamb wave The matching pursuit

method estimates the arrival time of each wave packets accurately even two wave packets are

overlapped The numerical simulations and experimental tests including Teflon tape inserted

and impact-induced damage are investigated for the feasibility of the proposed technique

Moreover the effectiveness of the proposed technique under changing temperature is exper-

imentally verified

- 31 -

62 Suggestion for further study

The limitations of this study and the possible improvement and further studies are listed as

follow

1 Measurement of impact-induced damage length using conventional NDT technique

Additional experimental verification for impact-induced damage characterization is

needed by measuring the real length of damage using conventional NDT technique

such as ultrasonic scan or infrared camera

2 Limitation of both side collocated PZT installation In this study excitation PZTs are

installed both side and collocated for decomposing A0 Lamb wave mode However it

can produce decomposition error due to imperfection of PZT location and their installa-

tion Moreover in many cases assess to both surface of structures is not allowed such

as aircraft applications To improve this limitation a reference-free damage characteri-

zation using single side PZTs will be considered

3 Complex structure application In this study damage is characterized in simple plate-

like structure When the monitored structure becomes complex the response signals al-

so becomes complex due to its geometry and boundary Thus the investigations for

complex structure application such as stiffener spar and curvature are needed

- 32 -

REFERENCES

[1] Cawley P and Alleyne D (1996) ldquoThe use of Lamb waves for the long range inspection of large structuresrdquo Ultrasonics 34 pp287-290

[2] Ghosh T Kundu T and Karpur P (1998) ldquoEfficient use of Lamb waves for detecting defects in large platesrdquo Ultrasonics 36 pp791-801

[3] Guirguitui V (2003) ldquoLamb wave generation with piezoelectric wafer active sensors for structural health monitoringrdquo SPIE Conference on Smart Structures and Materials San Diego CA 5391 pp111-422

[4] Ragvahan A and Cesnik CE S (2004) ldquoModeling of piezoelectric-based Lamb wave generation and sensing for structural health monitoringrdquo SPIE Conference on Smart Struc-tures and Materials San Diego CA 5391 pp419-430

[5] Rose JL Pilarski A and Dirit JJ (1993) ldquoAn approach to guided wave mode selection for inspection of laminated platerdquo Journal of Reinforced Plastics and Composites 12(5) pp536-544

[6] Guo N and Cawley P (1993) ldquoThe interaction of Lamb waves with delaminations in composite laminatesrdquo Journal of Acoustical Society of America 94(4) pp 2240-2246

[7] Valdez SHD and Soutis C (2002) ldquoReal-time nondestructive evaluation of fiber com-posite laminates using low-frequency Lamb wavesrdquo Journal of Acoustical Society of Amer-ica 111(5) pp 2026-2033

[8] Yang M and Qiao P (2005) ldquoModeling and experimental detection of damage in virous materials using the pulse-echo method and piezoelectric sensoractuatorsrdquo Smart Materials and Structures 14 pp1083-1100

[9] Lestari W and Qiao P (2005) ldquoApplication of wave propagation analysis for damage identification in composite laminated beamsrdquo Journal of Composite Materials 39(22) pp1967-1984

[10] Kessler S S Spearing S M and Soutis C (2002) ldquoDamage detection in composite ma-terials using Lamb wave methodsrdquo Smart Materials and Structures 11 pp269-278

[11] Sohn H Park G Wail J R Limback N P and Farrar C R (2004) ldquoWavelet-based active sensing for delamination detection in composite structuresrdquo Smart Materials and Structures 13 pp 153-160

[12] Petculescu G Krishnaswamy S and Achenback J D (2008) ldquoGroup delay measure-ments using modally selective Lamb wave transducers for detection and sizing of delamina-tions in compositesrdquo Smart Materials and Structures 17

[13] Sohn H Park H W Law K H and Farrar C R (2007) ldquoDamage detection in compo-

- 33 -

site plates by using an enhanced time reversal methodrdquo Journal of Aerospace Engineering (ASCE) pp 141-151

[14] Sohn H (2007) ldquoEffects of environmental and operational variability on structural health monitoringrdquo Philosophical Transactions of the Royal Society Ardquo 365 pp539-560

[15] Rytter A (1993) ldquoVibration based inspection of civil engineering structuresrdquo PhD The-sis Department of Building Technology and Structural Engineering University of Aalborg Aalborg Denmark

[16] Hayashi T and Kawashima K (2002) ldquoMultiple reflections of Lamb waves at a delami-nationrdquo Ultrasonics 40 pp193-197

[17] Wang C H and Rose L R F (2003) ldquoWave reflection and transmission in beams con-taining delamination and inhomogeneityrdquo Journal of Sound and Vibration 264 pp 851-872

[18] Ramadas C Balasubramaniam K Joshi M and Krishnamurthy C V (2009) ldquoInterac-tion of the primary anti-symmetric Lamb mode (Ao) with symmetric delaminations numer-ical and experimental studiesrdquo Smart Materials and Structures 18

[19] Ramadas C Balasubramaniam K Joshi M and Krishnamurthy C V (2010) ldquoInterac-tion of guided Lamb waves with an asymmetrically located delamination in a laminated composite platerdquo Smart Materials and Structures 19

[20] Wilcox P Lowe M and Cawely P (2001) ldquoThe effect of dispersion on long-range in-spection using ultrasonic guided wavesrdquo NDTampE International 34 pp1-9

[21] Mallat S G and Zhang Z (1993) ldquoMatching pursuits with time-frequency dictionariesrdquo IEEE Tansactions on Signal Processing 41(12) pp3397-3415

[22] Zhang G Zhang S and Wang Y (2000) ldquoApplication of adaptive time-frequency de-composition in ultrasonic NDE of highly scattering materialsrdquo Ultrasonics 38 pp 961-964

[23] Hong J C Sun K H and Kim Y Y (2005) ldquoThe matching pursuit approach based on the modulated Gaussian pulse for efficient guided-wave damage inspectionrdquo Smart Mate-rials and Structures 14 pp548-560

[24] Herakovich C T (1997) Mechanics of Fibrous Composites Wiley 480 pages

[25] Reddy J N (1985) ldquoA review of the literature on finite-element modeling of laminated composite platesrdquo The Shock and Vibration Digest 17(4) pp 3-8

[26] Reddy J N (2003) Mechanics of Laminated Composite Plates and Shells Theory and Analysis CRC Press 856 pages

[27] Palacz M Krawczuk M and Ostachowicz W (2005) ldquoThe spectral finite element mod-el for analysis of flexure-shear coupled wave propagation Part 1 laminated multilayer

- 34 -

composite beamrdquo Composite Structures 68 pp37-44

[28] Zhang Y C and Harding J (1990) ldquoA numerical micromechanics analysis of the me-chanical properties of a plain weave compositerdquo Computer amp Structures 36(5) pp 839-844

[29] Nemat-Nasser S and Hori M (1999) Micromechanics Overall Properties of Heteroge-neous Materials Elsevier 810 pages

[30] Scida D Aboura Z Benzeggagh M L and Bocherens E (1999) ldquoA micromechanics model for 3D elasticity and failure of woven-fibre composite materialrdquo Composites Science and Technology 59 pp 505-517

[31] Thomas Y and Wu X H (1997) ldquoA multiscale finite element method for elliptic prob-lems in composite materials and porous mediardquo Journal of Computational Physics 134 pp 169-189

[32] Lee K H Moorthy S and Ghosh S (1999) ldquoMultiple scale computational model for damage in composite materialsrdquo Computer Methods in Applied Mechanics and Engineer-ing 172 pp 175-201

[33] Namilae S and Chandra N (2005) ldquoMultiscale model to study the effect of interfaces in carbon nanotube-based compositesrdquo Journal of Engineering Materials and Technology (ASME) 127 pp222-232

[34] Adams D E (2007) Health monitoring of Structural Materials and Components Wiley 460 pages

[35] Gopalakrishnan S (2009) ldquoModeling Aspects in Finite Elementrdquo Encyclopedia of Struc-tural Health Monitoring 2 pp791-809

[36] Palacz M Krawczuk M and Ostachowicz W (2005) ldquoThe spectral finite element mod-el for analysis of flexure-shear coupled wave propagation Part 2 Deminated multilayer composite beamrdquo Composite Structures 68 pp45-51

[37] Baik J M and Thopson R B (1984) ldquoUltrasonic scattering from imperfect interfaces A quasi-static modelrdquo Journal of Nondestructive Evaluation 4(3) pp177-196

[38] Luo H and Hanagud S (2000) ldquoDynamics of delamincated beamsrdquo International Jour-nal of Solids and Structures 37 pp1501-1519

[39] Kim S B and Sohn H (2007) ldquoInstantaneous reference-free crack detection based on polarization characteristics of piezoelectric materialsrdquo Smart Materials and Structures 16 pp2375-2387

[40] An Y K and Sohn H (2010) ldquoInstantaneous crack detection under varying temperature and static loading conditionsrdquo The Special Issue of Structural Control and Health Monitor-ing 17(7) pp 730-741

- 35 -

[41] Lee S J and Sohn H (2006) ldquoActive self-sensing scheme development for structural health monitoringrdquo Smart Materials and Structures 15(6) pp 1734-1746

- 36 -

요 약 문

복합재료 손상의 무기저 진단 및 정량화

복합재료는 재료의 경량성 내구성 고강도의 특성으로 인하여 항공기에서부터

교량에까지 널리 사용되고 있다 하지만 외부의 충격으로 인한 복합재료 내부의

손상은 재료의 강도를 감소시켜 구조물의 안정성을 저하시키는 요인이 된다 하지만

복합재료의 손상은 재료 내부에서 발생하여 육안으로는 식별하기도 힘들며 표면에서

보이는 것만으로 내부의 손상 정도를 판단하기는 불가능하다는 단점을 가지고 있다

본 연구에서는 손상 내부에서 반사로 인해 발생하는 파를 이용하여 기저자료를

사용하지 않고 특정 시점에서 얻은 자료만으로 복합재료의 손상을 진단하고 정량화

할 수 있는 기법을 제안한다 제안된 기법의 적용성을 검증하기 위하여 유한요소법을

이용한 수치해석을 수행하였고 Teflon tape 이 삽입된 시편과 실제 충격하중으로

인한 손상을 입은 시편을 이용한 실험을 진행하였다 수치해석 및 실험 결과 제안된

기법이 복합재료의 손상을 성공적으로 진단 및 정량화할 수 있음을 확인하였다 또한

다양한 외부 온도 환경에서 제안된 기법의 환경영향 실험을 수행하였고 다양한

환경에서 손상의 진단 및 정량화가 성공적으로 이루어짐을 검증하였다

핵심어 무기저 손상진단 및 정량화 압전센서 복합재료 유한요소법 Matching pursuit

method

- 37 -

감 사 의 글

기대와 두려움을 안고 시작한 석사과정이 어느덧 2 년 이라는 세월 속에서 추억으로 자리잡아

가고 있습니다 학부생 기간을 포함한 지난 3 년 동안 SHM 이라는 새로운 분야를 접하고 연구를

수행 할 수 있도록 도와주신 고마운 분들이 있어 그 감사를 표하고자 합니다

우선 저의 많은 부족한 점을 일깨워 주시고 연구에 대한 뜻을 더욱 확고히 세울 수 있게 해

주신 지도교수님이신 손 훈 교수님께 깊은 감사의 뜻을 전합니다 짧은 기간이었지만 교수님의

도움으로 많은 것을 배우고 느끼고 깨달을 수 있었습니다 또한 본 연구를 수행할 수 있도록 지

원해주신 국토해양부 (U-City 석 박사과정 지원사업)에 깊은 감사 드립니다

때로 힘들고 지칠 때 항상 저를 믿고 응원해 주시는 가족이 있었기에 지금의 글을 쓸 수 있었

습니다 또한 지금 연구실에서 함께 생활하고 있는 윤규형 민구형 철민이형 현석이형 진열이형

호민이형 병진이 선혜누나와 성대에 계신 창길이형 미국에 계신 의재형과 Debaditiya 에게도 감

사의 말씀을 전합니다 그리고 모두 나열할 수는 없지만 카이스트 생활의 활력소가 되어준 카이

스트 건설 및 환경공학과 선후배들 직원 분들과 SOS 멤버들에게도 감사를 표합니다

앞서 거론한 많은 이들의 도움이 없었다면 저 또한 지금 이 자리에 있을 수 없었을 것입니다

많이 부족하기에 더 노력하고 배우고자 했던 초심을 다시 한번 생각하며 더 큰 결실을 맺기 위해

앞으로 나아가겠습니다 도움을 주신 모든 분들께 이 감사의 글을 바칩니다

2011년 6월 카이스트에서 임형진 올림

- 38 -

CURRICULUM VITAE

Hyung Jin Lim

Research Assistant

The Civil and Environmental Engineering Department

Korea Advanced Institute of Science and Technology (KAIST)

291 Daehak-Ro Guseong-Dong Yuseong-Gu

Daejeon Republic of Korea 305-701

Phone (82)+42-350-3665 Fax (82)+42-350-3610

Email limhj87gmailcom

RESEARCH INTERESTS

Damage Detection Localization and Quantification on Composite and Metallic Structures Pattern Recognition Finite Element Modeling (FEM)

EDUCATION

2009 ndash 2011 MS Civil and Environmental Eng Korea Advanced Institute of Science and Technology (KAIST) Korea Rep

2006 ndash 2009 BS Civil and Environmental Eng Korea Advanced Institute of Science and Technology (KAIST) Korea Rep

JOURNAL PUBLICATIONS The corresponding author is underlined 1 Hyung Jin Lim Hoon Sohn ldquoReference-free damage detection localization and quantification

using in compositerdquo in preparation for Journal of the Acoustical Society of America 2011

2 Chul Min Yeum Hoon Sohn Jeong Bum Ihn Hyung Jin Lim ldquoInstantaneous delamination detection in a composite using a dual piezoelectric transducer networkrdquo Submitted to Compo-sites Science and Technology 2011

3 Yun-Kyu An Hyung Jin Lim Min Koo Kim Jin Yeol Yang Hoon Sohn Chang Guen Lee ldquoInvestigation of applicability of local reference-free damage detection techniques to in-situ

- 39 -

bridgesrdquo Submitted to ASCE Journal of Structural Engineering 2011

4 Hyung Jin Lim Min Koo Kim Hoon Sohn Chan Yik Park ldquoImpedance based damage detec-tion under varying temperature and loading conditionsrdquo Submitted to NDTampE International 2011

5 Hyung Jin Lim Hoon Sohn ldquoQuantitative estimation of transmitted and reflected lamb waves at discontinuityrdquo Journal of the Korean Society for Nondestructive Testing Vol30 No 4 pp1-8 2010

CONFERENCE PROCEEDINGS

The corresponding author is underlined 1 Hyung Jin Lim and Hoon Sohn ldquoReference-free damage detection localization and quantifi-

cation in compositesrdquo ASCE Engineering Mechanics Institute Conference Boston June 2-4 2011

2 Hyung Jin Lim and Hoon Sohn ldquoReference-free delamination quantification within multi-layer composite materialsrdquo KSNT Spring Conference May 26-27 Gyeong-Ju Korea 2011

3 Min Koo Kim Hyun Jin Lim Hoon Sohn Chan Yik Park ldquoImpedance-based bolt loosening detection under varying temperature and loadingrdquo Asian Pacific Workshop on Structural Health Monitoring Tokyo Japan November 30-December 2 2010

4 Hyung Jin Lim Chul Min Yeum and Hoon Sohn ldquoModeling of Impact-induced Delamination in a Multilayer Composite Platerdquo The 23rd KKCNN Symposium on Civil Engineering Taipei (Taiwan) November 13-15 2010

5 Yun-Kyu An Hyung Jin Lim and Hoon Sohn ldquoTemperature independent Delamination Detec-tion using data normalizationrdquo The 3rd International Conference on Multi-Functional Materi-als and Structures Chonbuk National Univ Jeonju Korea September 14-18 2010

6 Hoon Sohn Yun-Kyu An Chang Gil Lee Min Koo Kim Hyung Jin Lim ldquoApplication of online NDT techniques to decommissioned bridge testingrdquo the 5th European Workshop on Structural Health Monitoring Sorrento Italy June 29-July 02 2010

7 Min Koo Kim Hyung Jin Lim Hoon Sohn Chan Yik Park ldquoImpedance based damage diag-nosis of a complex composite aircraft wing under changing loading and temperature condi-tionsrdquo the 5th European Workshop on Structural Health Monitoring Sorrento Italy June 29-July 02 2010

8 Yun-Kyu An Min Koo Kim Hyung Jin Lim and Hoon Sohn ldquoValidations of reference-free crack detection techniques through a decommissioned bridge testrdquo 2010 COSEIK Annual Con-ference Jeju Korea April 8-9 2010

- 40 -

PARTICIPANT PROJECTS

1 Development of Smart Steel Members for Online Steel Structure Monitoring POSCO Corporation (Funded 41000000 KRW (41000 USD) for 010511 to 123111)

2 Development of Enabling Technologies for Structural Health Monitoring Korean Agency for Defense Development (Funded 240000000 KRW (240000 USD) out of total 3200000000 KRW (3200000 USD) for 060108-053111)

3 Development of On-board SHM Technologies for Composite Air Vehicles The Boeing Company (Funded 400000 USD (400000000 KRW) for 080108 to 073111)

EDUCATIONAL EXPERIENCES

Fall 2009

Teaching Assistant CE207 ldquoElementary Structural Engineering and Laboratoryrdquo KAIST 3 unit undergraduate course

EXPERIMENTAL EXPERIENTCES

2010 Monitoring of Real Aircraft Composite Wing Segment Korea Agency for Defense De-velopment (ADD) Daejeon Korea Real time monitoring of composite air vehicles fund-ed by Korea Agency for Defense Development (ADD)

2009-2011

Field Monitoring of Real Bridge Structures Gimpo (Ramp G bridge) amp Icheon (Sam-seung bridge) Korea US-Korea collaborative research for bridge monitoring testbeds

HONORS amp AWARDS

Scholarships for outstanding students granted KAIST (2007-2009) The three prominent students in the civil and environmental engineering department of KAIST were selected as recipients This scholarship provided 2000$ per year for three years

- 41 -

Page 3: 복합재료 손상의 무기저 진단 및 정량화 - KAISTssslab.kaist.ac.kr/article/pdf/Thesis_2.pdf · 1 Declaration of Ethical Conduct in Research: I, as a graduate student

Reference-free Damage Detection Localization

and Quantification in Composite

Advisor Professor Hoon Sohn by

Hyung Jin Lim

Department of Civil and Environmental Engineering

KAIST

A thesis submitted to the faculty of KAIST in partial fulfillment of the re-quirements for the degree of Master of Science and Engineering The study was conducted in accordance with Code of Research Ethics1

2011 6 20

Approved by

Professor Hoon Sohn

[Major Advisor]

1 Declaration of Ethical Conduct in Research I as a graduate student of KAIST hereby declare that I have not committed any acts that may damage the credibility of my research These include but are not limited to falsi-fication thesis written by someone else distortion of research findings or plagiarism I affirm that my thesis contains honest conclusions based on my own careful research under the guidance of my thesis advisor

복합재료 손상의 무기저 진단 및 정량화

임 형 진

위 논문은 한국과학기술원 석사학위논문으로

학위논문심사위원회에서 심사 통과하였음

2011 년 5 월 30 일

심사위원장

심사위원

심사위원

손 훈 (인)

이 행 기 (인)

김 천 곤 (인)

MCE

20094258

임 형 진 Lim Hyung Jin Reference-free Damage Detection Localization and

Quantification in Composite 복합재료 손상의 무기저 진단 및 정량화 Department

of Civil and Environmental Engineering 2011 41 p Advisor Prof Sohn Hoon

ABSTRACT

A reference-free damage detection localization and quantification technique in a multi

layered composite plate is proposed It is reported that when A0 Lamb wave mode propa-

gates the damage time delay attenuation and multiple reflections are occurred in the dam-

age area In this thesis additional reflected waves from the damage due to multiple reflec-

tions are used as a feature for characterizing the damage First the damage is identified by

detecting the additional reflected waves from the damage area Then the location of the

damage is estimated using combined pitch-catch and pulse-echo methods Finally the dam-

age is quantified using a damage index defined as the ratio between the physical length of

the damage and the wavelength of the A0 Lamb wave mode The matching pursuit method

is utilized for analyzing the obtained signal by decomposing overlapped waves and estimat-

ing the parameters of each wave such as arrival time driving frequency amplitude and so

on The proposed technique has an advantage that the damage can be characterized without

reference data obtained from pristine condition of the monitored structure within a single

path The feasibility of the proposed technique is verified through numerical simulations

and experimental tests In numerical simulation different damage models (length material

properties and location) are investigated In experimental tests a Teflon tape inserted and

impact-induced damage are characterized under temperature variations The results are

suggesting a reliable damage characterization in composites

Keywords Reference-free PZT transducers Composites Finite element method Matching

pursuit method

i

TABLE OF CONTENTS

ABSTRACT middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot i

TABLE OF CONTENTS middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot iii

LIST OF TABLES middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot iv

LIST OF FIGURESmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot v

CHAPTER 1 INTRODUCTION

11 Motivation middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 1

12 Literature review middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 1

13 Objective and uniqueness middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 3

14 Organization middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 4

CHAPTER 2 INTERACTION OF LAMB WAVE WITH DAMAGE

21 A0 Lamb wave mode delay and attenuation in the time domain middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 5

22 Multiple reflections in the damage area middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 6

CHAPTER 3 THEORETICAL DEVELOPMENT

31 Overview middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 7

32 Damage index (m-value) middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 8

33 Damage detection algorithm (Level 1) middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 10

34 Damage quantification algorithm (Level 3) middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 11

35 Damage localization algorithm (Level 2amp3) middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 11

36 Matching pursuit method considering the dispersion effect of Lamb wave

middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 13

CHAPTER 4 NUMERICAL SIMULATION

41 Composite modeling methodmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 15

42 Damage modeling method middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 16

43 Numerical simulation description middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 17

44 Numerical simulation results middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 19

ii

CHAPTER 5 EXPERIMENTAL VERIFICATION

51 Overview middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 21

511 Specimen middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 22

512 Data acquisition system and PZT middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 22

513 Impact test middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 23

514 Temperature effect test middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 24

52 Teflon inserted damage middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 25

53 Impact-induced damagemiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 26

54 Temperature effect test middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 27

541 Teflon inserted damage middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 27

542 Impact-induced damage middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 29

CHAPTER 6 CONCLUDING REMARKS

61 Executive summary middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 31

62 Suggestion for further study middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 32

REFERCENCES middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 33

SUMMARY (IN KOREAN) middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 37

ACKNOWLEDGEMENTS (IN KOREAN)

CURRICULUM VITAE

iii

LIST OF TABLES

I Material property of simulation model and test specimen middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 17

II Damage cases for numerical simulation middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 18

III Damage characterization for the numerical simulation models middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 20

IV Damage characterization for the Teflon inserted damage specimen middotmiddotmiddotmiddotmiddotmiddotmiddot 25

V Damage characterization for the impact-induced damage specimen middotmiddotmiddotmiddotmiddotmiddotmiddot 27

VI Temperature effect test for the Teflon inserted damage specimen middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 28

VII Temperature effect test for the impact induced damage specimen middotmiddotmiddotmiddotmiddotmiddotmiddot 30

iv

LIST OF FIGURES

1 Lamb wave propagation characteristics (a) A0 Lamb wave mode time delay and

attenuation after passing through a damage (b) The time delay and amplitude varia-

tion of A0 Lamb wave mode under temperature variation in pristine condition of

structure Because of this conventional damage detection techniques in composites

have limitations for real structure application middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 5

2 Multiple reflections in the damage area The multiple reflections of A0 Lamb

waves in the damage area are utilize as a feature for characterization of the damage

in this study middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 6

3 Parameters and notations of the proposed damage characterization technique

Damage is detected by identifying the additional reflected wave from the damage due

to multiple reflections (dashed lines) Localization and quantification of the damage

is carried out by measuring the arrival time of transmitted and reflected waves from

the damage middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 7

4 Two different FEM models which have different 119889119889 and 119907119907119889119889 (or G) This repre-

sents that the damage cannot be characterized uniquely based on the measured data

available middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 9

5 (a) Example of the overlapped signal with comparing approximated signal using

the matching pursuit method (b) Decomposed signal using the matching pursuit

method The matching pursuit method provides a conventional way to decompose

and identify each wave middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 9

6 Description of the numerical simulation model middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 17

7 Comparison of A0 Lamb wave mode signals from numerical simulation be-

tween intact and damage cases Time delay attenuation and multiple reflections due

to damage simulated successfully using the damage modeling method described in

v

the previous section middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 19

8 Test specimens for the experimental verification In the Teflon inserted damage

Teflon tapes with 25mm diameter were inserted at the center of the specimen The

impact-induced damage was introduced to the specimen using the impact test equip-

ment middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 21

9 Overall experimental setup (a) Data acquisition (DAQ) system and (b) Installed

circular PZT transducer on the specimen with its dimension middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 23

10 (a) Impact test equipment for introducing impact-induced damage and (b) Tip

of the impact tester with 10mm diametermiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 23

11 Test specimen in temperature chamber middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 24

12 A0 Lamb wave mode obtained using pitch-catch method from the Teflon in-

serted damage specimen (a) The reflected wave from damage is shown (b) The Tef-

lon inserted damage is detected by capturing the reflected wave middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 25

13 The signal obtained from the test specimen subjected to impact The impact

was introduced at the same location of the specimen and the signal was obtained after

each impacts As the number of impact is increased the transmitted A0 Lamb wave

mode is delayed and attenuated middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 26

14 False alarm test for the proposed damage detection algorithm in pristine condi-

tion of the specimen Reflected wave from damage is not captured false alarm is not

occurred middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 26

15 Signals obtained from the Teflon inserted damage specimen under temperature

variation Time delay and amplitude variation are occurred middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 27

16 False alarm test under temperature variations It is shown that proposed dam-

vi

age characterization algorithm is robust under temperature variationsmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 29

17 Signals obtained from impact-induced damage specimen (Impact 3) under

temperature variation middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 29

vii

CHAPTER 1 INTRODUCTION

11 Motivation

Composite materials has been increasingly demands in modern structures ranging from the

latest aircraft to retrofitted bridge deck due to its light weight superior specific strength and

stiffness properties However the occurrence of damage due to abrupt impact or sequential

accumulation of defect may considerably degrade structural strength or stiffness and reduce

the safety and reliability of the system Recently Lamb waves based damage detection tech-

niques using piezoelectric (PZT) transducers have emerged as one of the promising structural

health monitoring (SHM) technology for detecting hidden damage because of the possibility

of long range inspection and high sensitivity [1-4]

12 Literature review

This section will discuss researches that have been previously carried out on topics of

guided wave based SHM techniques using PZT transducers for damage in composite materi-

als The early investigations of interaction between Lamb wave and damage in composite can

be found in Rose et al and Guo et al [5-6] From further studies reflected signals from dam-

age were shown to be an effective method for damage diagnosis and localization using PZTs

[7-9] The reflected signals from damage were collected using one PZT as an actuator and the

other PZT as a sensor However the techniques cannot be used to determine the extent of

- 1 -

damage with reasonable precision

Kesseler et al proposed a technique monitoring the wavelet coefficients of transmitted ul-

trasonic signal amplitude as a potential way to damage detection [10] One drawback of this

technique is that the variability in coupling and the sensitivity of installed PZT transducers

can induce changes in the amplitude and energy of the signal that are cause false-alarm Oth-

er wavelet based delamination detection and localization technique developed using PZT

sensor network and outlier analysis [11]

Damage detection and quantification in composite plate was also investigated by group de-

lay measurement due to damage [12] However it is assumed that the group velocity in dam-

age area as the group velocity propagates half thickness of the structure in dispersion curve in

the quantification procedure Thus the damage quantification was investigated only Teflon

inserted damage case

Sohn et al proposed a damage detection technique which does not rely on any past base-

line signals to assess damage in composite plates by using an enhanced time reversal method

[13] The damage is assessed using the violation of time reversibility when nonlinearity is

caused by the damage along a direct wave path Thus it allows instantaneous identification

of delamination without requiring the baseline signal for comparison

- 2 -

13 Objective and uniqueness

In this study a reference-free damage characterization technique including detection lo-

calization and quantification of impact-induced damage in composite materials is developed

The general damage characterization can be categorized into the following levels [14]

Level 0 Define damage

Level 1 Determine the presence of damage within the structure

Level 2 Locate the regions of damage

Level 3 Quantify the damage

Level 4 Predict the remaining service life of the structure

Through the proposed technique from level 1 to level 3 can be achieved

Most of common approaches of damage characterization techniques described in the pre-

vious section use prior obtained baseline data However these approaches can cause false

alarm due to the effect of environmental variations such as temperature and ambient vibra-

tions [15]

The uniqueness of this study is

1 Damage is characterized without using prior measured baseline data

2 Robustness of the proposed technique under temperature variation is investigated

3 Additional waves reflected from inside damage have been used in addition to transmit-

ted and reflected waves

- 3 -

14 Organization

The organization of this thesis is summarized as follows

Chaper1 A general introduction including motivation literature review objective and

uniqueness of this study is described

Chaper2 Interactions between Lamb wave mode and damage in composites such as time

delay attenuation and multiple reflections are summarized Damage index (m-value) for

damage characterization is defined

Chaper3 Algorithms for the reference-free damage characterization technique using m-

value are developed Modified matching pursuit for this study is also described

Chaper4 Verification of the algorithm using numerical simulation is presented Various

modeled damage cases are characterized

Chaper5 Experimental verifications are described Two damage cases (Teflon inserted and

impact-induced damage) are investigated for the feasibility of the proposed damage charac-

terization technique Temperature effect test using temperature chamber is also carried out

Chaper6 Conclusion and summary of the present study is reported In advance suggestions

and future work recommendations are represented

- 4 -

CHAPTER 2 INTERACTION OF LAMB WAVE WITH DAMAGE

21 A0 Lamb wave mode delay and attenuation in the time domain

Time delay and attenuation of the A0 mode Lamb wave after passing through the damage

area due to effective thickness reduction were reported in previous researches [812] Moreo-

ver the attenuation of the A0 Lamb wave mode is occurred by degrading of shear modulus of

in the damage area However the time delay and amplitude variation of the A0 Lamb wave

mode also occurred even when the structure is health condition by the material property

changing due to temperature variation of the structure (Fig 1) Thus conventional damage

characterization methods using the time delay and attenuation of A0 Lamb wave mode as a

feature can cause false-alarm in real application

11 115 12 125 13-10

-5

0

5

10

Time (ms)

Vol

tage

(mV

)

IntactDamage

Reflected wave from Damage

(a)

11 115 12 125 13-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

0 oC20 oC50 oC

(b)

Fig 1 Lamb wave propagation characteristics (a) A0 Lamb wave mode time delay and atten-uation after passing through a damage (b) The time delay and amplitude variation of A0 Lamb wave mode under temperature variation in pristine condition of structure Because of this conventional damage detection techniques in composites have limitations for real struc-ture application

- 5 -

22 Multiple reflections in the damage area

Standing wave patterns due to multiple reflections of A0 Lamb wave mode in the damage

area were observed through numerical simulations and ultrasonic wave-field imaging using a

scanning laser vibrometer [16] The formation of standing waves in damage is illustrated

through Fig 2 When the incident waves enter the damage area a significant portion of the

waves are reflected back from the exit Then the reflected waves travel through the damage

and undergo reflection again at the original entrance As a consequence of multiple reflec-

tions at the entrance and the exit a considerable amount of ultrasonic energy is trapped inside

the damage area [17-19]

Multiple Reflections

Incident Wave Transmitted Wave

Damage

Fig 2 Multiple reflections in the damage area The multiple reflections of A0 Lamb waves in the damage area are utilize as a feature for characterization of the damage in this study

- 6 -

CHAPTER 3 THEORETICAL DEVELOPMENT

31 Overview

Damage

l = l1 + d + l2

d = vdtd l1 = vit1 l2 = vit2

t = t1 + td + t2

Excitation and Sensing PZT Sensing PZT

l

Transmitted Wave

Reflected Waves

Incident Wave

Fig 3 Parameters and notations of the proposed damage characterization technique Damage is detected by identifying the additional reflected wave from the damage due to multiple re-flections (dashed lines) Localization and quantification of the damage is carried out by measuring the arrival time of transmitted and reflected waves from the damage

In this section the reference-free damage characterization algorithm is proposed Damage

diagnosis and quantification are achieved by identifying the additional reflected waves due to

multiple reflections in the damage area (dashed lines in Fig 3) Moreover damage is local-

ized using additional information from pulse-echo method implement For the localization

and the quantification of damage the variables such as the path length and A0 Lamb wave

mode velocity in damage area are defined (Fig 3) For example 119897119897 is expressed as

119897119897 = 1198971198971 + 119889119889 + 1198971198972 (1)

where 1198971198971 119889119889 and 1198971198972 are the distance from the excitation PZTs to the entrance of damage the

damage length and the distance from the exit of damage to the sensing PZT respectively

- 7 -

Moreover 119907119907119894119894 and 119907119907119889119889 are also defined as the A0 Lamb wave mode velocity in pristine area

and damage area respectively Note that the whole procedure of the damage characterization

is accomplished based solely on the signals obtained from the current state of the structure

32 Damage index (m-value)

The proposed reference-free damage characterization technique utilizes the waves reflect-

ed additionally from the damage area as a feature The arrival time difference of the transmit-

ted and reflected A0 Lamb wave modes reflected from the exit and the entrance of the delam-

ination area one times each (∆119905119905) contains the information of the physical length of the dam-

age (119889119889) and A0 Lamb wave mode velocity in the damage area (119907119907119889119889) The wave velocity in the

damage area which is related to the shear modulus (G) represents the severity of the damage

∆119905119905 = 2119905119905119889119889 =2119889119889119907119907119889119889

(2)

However based on the measured data available the 119889119889 and 119907119907119889119889 (or G) cannot be uniquely

determined Fig 4 shows two FEM models which have different physical length and wave

velocity (G) of the damage Detail explanation of delamination modeling technique will be

covered in section 4 The similar signal in the time domain and ∆119905119905 are obtained from the

two different damage cases

- 8 -

Damage 1

Damage 2

20

Unit mm15

90 G value reduction

943 G value reduction

m-value = 248

m-value = 248

015 02 025 03 035

-2

-1

0

1

2

Time (ms)

Ampl

itude

Damage 1Damage 2

Fig 4 Two different FEM models which have different 119889119889 and 119907119907119889119889 (or G) This represents that the damage cannot be characterized uniquely based on the measured data available

Thus in this study the damage index (m-value) which can be uniquely defined based on

the available data is introduced The m-value is defined as the ratio of the physical length of

the damage and wavelength of the A0 Lamb wave mode in the damage area (120582120582119889119889)

119898119898 =119889119889120582120582119889119889

=119889119889119889119889119907119907119889119889

(3)

In many cases the transmitted and reflected waves are overlapped and it is hard to decom-

pose and calculate the arrival times of them Matching pursuit method is an alternative way

to accurately decompose of the overlapped waves and estimate the parameters of each wave

such as arrival time (Fig 5) Detail explanation of the matching pursuit method is covered in

section 36

015 02 025 03-2

-1

0

1

2

Time (ms)

Am

plitu

de

OriginalApproximated

(a)

015 02 025 03-2

-1

0

1

2

Time (ms)

Ampl

itude

Wave 1Wave 2

(b)

Fig 5 (a) Example of the overlapped signal with comparing approximated signal using the matching pursuit method (b) Decomposed signal using the matching pursuit method The matching pursuit method provides a conventional way to decompose and identify each wave

- 9 -

33 Damage detection algorithm (Level 1)

The damage diagnosis is accomplished by extracting the additional reflected A0 Lamb

wave modes from damage due to multiple reflections If A0 Lamb wave modes are decom-

posed perfectly it is trivial that one transmitted A0 Lamb wave mode is always captured in

the response signal If damage exists other waves in the response signal behind the transmit-

ted A0 Lamb wave mode are presented Since this approach does not rely on previously col-

lected data the diagnosis is independent from environmental variation such as temperature

changes

It is reported that the dispersion of Lamb waves cause wave-packets to spread out in space

and time as they propagate through a structure [20] Thus the reflected waves from the dam-

age have longer pulse duration than the transmitted wave due to their longer propagating dis-

tance In this study the matching pursuit method is modified to capture more spread wave

than the transmitted A0 Lamb wave mode Detail explanation of the modified matching pur-

suit method is described in latter section

Let 1199041199041 and 1199041199042 as the pulse duration (scale) of the transmitted and additional reflected A0

Lamb wave mode respectively

ldquoIf 1199041199042 gt 1199041199041 damage exists

Otherwise there is no damagerdquo

- 10 -

34 Damage quantification algorithm (Level 3)

As mentioned in the previous section damage is quantized by estimating 119889119889 and 119907119907119889119889 using

pitch-catch method The procedure to estimate 119889119889 and 119907119907119889119889 is

(i) Measure 119897119897 119905119905 and 119905119905119889119889

(ii) Assume the range of 119889119889 (0 lt 119889119889 lt 119897119897)

(iii) Calculate 119907119907119894119894 and 119907119907119889119889 with a constraints (0 lt 119907119907119889119889 lt 119907119907119894119894) (Eq 4 and 5)

(iv) Estimate m-value and find a feasible range of 119889119889 and 119907119907119889119889

119907119907119894119894 =119897119897 minus 119889119889119905119905 minus 119905119905119889119889

(4)

119907119907119889119889 =119889119889119905119905119889119889

(5)

35 Damage localization algorithm (Level 2amp3)

For damage localization 1198971198971 is estimated by measuring the arrival time from the excitation

PZT to the entrance of delamination (1199051199051) with additional pulse-echo method implementation

The procedure is similar with the previous damage quantification algorithm

(i) Measure 119897119897 119905119905 119905119905119889119889 and 1199051199051

(ii) Assume the range of 1198971198971 (0 lt 1198971198971 lt 119897119897)

(ii) Compute possible combinations of 1198971198972 119889119889 and 119907119907119889119889 with constraints

(0 lt 119907119907119889119889 lt 119907119907119894119894 119889119889 gt 0) (Eq 5 6 and 7)

- 11 -

1198971198972 = 1198971198971119905119905 minus 1199051199051 minus 119905119905119889119889

1199051199051 (6)

119889119889 = 119897119897 minus 1198971198971 minus 1198971198972 (7)

Since a PZT for pulse-echo method acts as both an actuator and a sensor the obtained sig-

nal contains both the excitation and the response signal Thus the reflected wave from the

entrance of damage can be buried into the tail of excitation signal and cause error on arrival

time of reflected wave from damage Even in this case damage can be localized and quan-

tized using the constraints When first wave packet reflected from the entrance of damage is

buried into the excitation signal the next wave packet reflected from the exit of damage can

be utilized alternatively If the algorithm regards the second reflected wave as the first re-

flected wave there are no combinations of estimated variable meeting the constraints Then

the algorithm repeats the process by regarding the wave as the second reflected wave the

constraints can select proper combination of the estimated variables

The estimated values have range bounded by the constraints When additional information

of 119907119907119894119894 is utilized the damage can be characterized more accurately In this study three meth-

ods are proposed and verified as information level of 119907119907119894119894

Method 1 Solely based on the current data measurement (Reference-free)

Method 2 Limit the possible range of 119907119907119894119894 (Prior knowledge of the system)

Method 3 Measure 119907119907119894119894 (Additional data need to be collected)

Method 1 described in this section can be utilized without any information of 119907119907119894119894 Meth-

od 2 produces more accurate estimation results using prior knowledge of the system such as

- 12 -

119907119907119894119894 variation under temperature changes Method 3 gives unique estimated value using meas-

ured 119907119907119894119894 from other path in the same time In section 4 and 5 damage is characterized using

these three methods and the results are compared each another

36 Matching pursuit method considering the dispersion effect of Lamb

wave

Matching pursuit method has been developed for signal compression approximation and

characterization of obtained signal and function [21-23] The key concept of the matching

pursuit method is that any signal (function) can be decomposed into a linear expansion of

signals selected from a pool of basis signals (functions) called lsquodictionaryrsquo

Let 119891119891 be a function to be analyzed then 119891119891 can be approximated a set of functions select-

ed from the dictionary 119863119863 in order to best match This is done by successive approximations

with orthogonal projections Let1198921198921205741205740 isin 119863119863 the function 119891119891 can be decomposed into

119891119891 = lang1198911198911198921198921205741205740rang1198921198921205741205740 + 119877119877119891119891 (8)

where 119877119877119891119891 is the residual function after approximating 119891119891 using 1198921198921205741205740 1198921198921205741205740 is orthogonal to

119877119877119891119891 hence

1198911198912 = lang1198911198911198921198921205741205740rang2

+ 1198771198771198911198912 (9)

To minimize 119877119877119891119891 1198921198921205741205740 is chosen to maximize lang1198911198911198921198921205741205740rang The matching pursuit method is

an iterative algorithm that decomposes the residue 119877119877119891119891 by projecting it on a function of 119863119863

that matches 119877119877119891119891 almost at best as it was done for 119891119891 The residue 119877119877119899119899119891119891 is decomposed into

119877119877119899119899119891119891 = lang119877119877119899119899119891119891119892119892120574120574119899119899rang119892119892120574120574119899119899 + 119877119877119899119899+1119891119891 (10)

- 13 -

the decomposition is carried up to the order 119898119898 119891119891 can be expressed as

119891119891 = lang119877119877119899119899119891119891119892119892120574120574119899119899rang119892119892120574120574119899119899

119898119898minus1

119899119899=0

+ 119877119877119898119898119891119891 (11)

The original function 119891119891 is decomposed into a sum of dictionary elements that are chosen to

best match its residues

In this study the matching pursuit method is utilized to accurately decompose of the over-

lapped Gaussian pulses and estimate the parameters of the pulse which used for damage

characterization The 119894119894119905119905ℎ Gaussian pulse in the dictionary for approximating the obtained

signal is expressed as

119892119892120574120574119894119894 = 119891119891(119860119860119894119894 119906119906119894119894 119904119904119894119894 119889119889119894119894120601120601119894119894) (12)

where 119860119860119894119894119906119906119894119894119904119904119894119894119889119889119894119894 and 120601120601119894119894 represent amplitude arrival time scale (pulse duration) excita-

tion(central) frequency and phase of the Gaussian pulse respectively The parameters of the

pulse are estimated using a nonlinear regression based optimization method (Gauss-Newton

method) based on an initial guess of the parameters Detail procedure for estimating the pa-

rameters of the pulse is expressed in Hong et al [23]

Among the estimated parameters of the pulses pulse duration (scale) represents the disper-

sion effect If a wave-packet is spread out the estimated pulse duration is increased Using

this feature the matching pursuit method is modified to capture waves which have longer

pulse duration than transmitted A0 Lamb wave mode If the estimate pulse duration of a wave

is less than the transmitted wave the modified matching pursuit method regards it as error

- 14 -

CHAPTER 4 NUMERICAL SIMULATION

41 Composite modeling method

Composite materials can be modeled from nano to macro level depends on its engineering

purpose In general four methods are utilized for modeling of composite materials First

composite is modeled as a homogenous material expressed in terms of effective material

properties [24] The effective material properties are estimated using the modulus and the

volume fraction of each constituent theoretically (Rule of mixture) and measured experimen-

tally Second a model is established that each laminate which have own material properties

are totally bonded together In this model the stacking sequence and direction of each lami-

nates are considered additionally [25-27] Third method for composite modeling in micro

level called micromechanics [28-30] This model considers the interaction between fibers and

matrix in laminate Last multiscale modeling which the scope of modeling is extended to

molecular or quantum level of constituents [31-33] This model starts from molecular me-

chanics or molecular dynamics then the scale of model leads to the nano micro and macro

level by incorporating small scale models As the scale of models is smaller it provides more

logical way than other large scale of models while the computational cost is more expensive

The scope of this study is focused on the dynamic behavior of composite structure the beam

or plate model was utilized for the numerical simulation

- 15 -

42 Damage modeling method

Generally three damage models have been developed and utilized in the real applications

First the stiffness reduction model called as the equivalent element model was introduced

[17 34-35] This model is simply realized by replacing the region around damage area by a

region with adjusted material properties (Youngrsquos modulus shear modulus density etc)

globally Next the duplicated node model is done by disconnecting the element connectivity

assuming there is no interaction among the delaminated surface This model can be utilized

when the location and the extension of damage is important issue [34-36] Last the contact

mechanics model using the principles of contact mechanics the two interfacial forces (con-

tact resistance and sliding resistance) in the delaminated surface [37-38] In this model the

delaminated surfaces are modeled as nonlinear functions of the interlaminar normal strain

and the interlaminar shear strain This model is computationally expensive while provides

more logical way than other models There is however no universally accepted model for

general impact-induced damage in composites In the case of impact-induced damage it is

impossible to predict or estimate the location and number of delaminated layers Thus from

the uncertainty of impact-induced damage the equivalent element model can be considered

as the effective damage model

- 16 -

43 Numerical simulation description

TABLE I Material property of simulation model and test specimen

Tensile Strength 657 MPa Tensile Modulus 59 GPa Shear Strength 804 MPa Shear Modulus 24 Gpa Specific Gravity 16

PZT A

PZT C

PZT BDamage

3

250 125 125

500 Unit mm

Fig 6 Description of the numerical simulation model

The feasibility of the proposed damage characterization technique is verified by 2-D FEM

simulation using MSC PATRAN and NASTRAN Fig 6 presents the dimension PZTs and

damage location of the simulation model In the model 05mm x 05mm square shell element

was used for modeling of the A0 mode whose flexural wave characteristic The element size

was determined by considering the wavelength of the propagation Lamb wave modes The

material properties of the simulation model and damage cases are presented in Table I and

Table II respectively

PZT A and C act as both actuator and sensor for collecting pulse-echo signal They were

installed on the plate which two of them are both-side collocated for the A0 Lamb wave

- 17 -

mode decomposition [39-40] PZT B was only used as a sensor for collecting pitch-catch sig-

nal For simulating the excitation and sensing of PZTs the pin-force model was adopted that

an input force apples to the end nodes of excitation PZT and averaging element strains corre-

sponding sensing area A Gaussian pulse (tone-burst) 50 kHz central frequency was excited

as an input signal

TABLE II Damage cases for numerical simulation

119949119949120783120783 (mm) 119941119941 (mm) G value reduction ()

Case 1 650 200 90 darr

Case 2 675 150 90 darr

Case 3 675 150 93 darr

Case 4 900 150 90 darr

- 18 -

44 Numerical simulation results

015 02 025 03 035

-2

0

2

Time (ms)

Am

plitu

de

IntactDamage

Reflected wave from damage

(a) Pitch-catch signal

015 02 025 03 035

-2

0

2

Time (ms)

Am

plitu

de

IntactDamage

Reflected waves from damage

(b) Pulse-echo signal Fig 7 Comparison of A0 Lamb wave mode signals from numerical simulation between intact and damage cases Time delay attenuation and multiple reflections due to damage simulated successfully using the damage modeling method described in the previous section

Fig 7 shows that comparison of A0 Lamb wave mode signals between intact and one of

damage cases in the simulation The features of A0 Lamb wave mode passing through dam-

age area described in section 2 are successfully simulated Thus the equivalent element

method is proper for modeling damage in composites In the damage cases the additional

waves reflected from damage is observed in both pitch-catch and pulse echo signal

Table III presents the result of the proposed damage characterization algorithm using the

numerical simulation cases shown in Table II In lsquoMethod 2rsquo 119907119907119894119894 variation is assumed to plusmn

5 of the normal value (1190~1315ms) The normal value of 119907119907119894119894 for lsquoMethod 3rsquo is meas-

ured as 1253 ms In each case the damage is characterized using the proposed technique

Note that as the information of 119907119907119894119894 is specified the estimated value is closer to the exact

value

- 19 -

TABLE III Damage characterization for the numerical simulation models

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

Case 1

Exact 650 200

Method 1 248 50~70 79~414 160~834

Method 2 248 603~667 135~241 271~486

Method 3 248 635 188 379

Case 2

Exact 675 150

Method 1 193 60~75 01~251 37~651

Method 2 193 618~682 114~222 295~575

Method 3 193 65 168 435

Case 3

Exact 675 150

Method 1 233 60~75 66~302 140~649

Method 2 233 668~738 85~196 182~419

Method 3 233 703 140 301

Case 4

Exact 900 150

Method 1 204 75~100 48~349 118~856

Method 2 204 860~951 107~216 264~530

Method 3 204 906 162 397

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1253 ms 1190~1315 ms Measured 119907119907119894119894 1253 ms

- 20 -

CHAPTER 5 EXPERIMENTAL VERIFICATION

51 Overview

In this section experimental study is investigated so that the proposed damage characteri-

zation technique Overall experimental setup and test specimen are shown in Fig 8 and 9

The experimental setup consists of test specimens data acquisition (DAQ) system and self-

sensing circuit Self-sensing circuit was installed for pulse-echo measurement using PZTs

[40-41]

PZT A(PZT C) PZT B

500

500

140

75

250

Teflon Inserted Damage (d =25)

Unit mm

Thickness = 3

PZT A(PZT C) PZT B

500

500

150

90

250

Impact Location

Unit mm

Thickness = 3

(a) Teflon inserted damage specimen

(b) Impact-induced damage specimen

Fig 8 Test specimens for the experimental verification In the Teflon inserted damage Teflon tapes with 25mm diameter were inserted at the center of the specimen The impact-induced damage was introduced to the specimen using the impact test equipment

- 21 -

511 Specimen

Two carbon composite plates with a dimension of 500times500times3mm3 were manufactured by

Nexcomes Inc by stacking 12 layers of CF3327 3K prepreg (woven fabrics type) manufac-

tured by Hankuk Fiber Inc The material properties of the specimen are presented in Table I

One of the specimens Teflon tapes with 25mm diameter was inserted in every 3 layers at the

center of the specimen to mimic the damage The other specimen is utilized for the character-

ization of real impact-induced damage

512 Data acquisition (DAQ) system and PZT

The data acquisition (DAQ) system consists of an arbitrary waveform generator (AWG) a

high speed signal digitizer (DIG) and two multiplexers (MUX) Using the 16-bits AWG

tone-burst input signal with a plusmn10 peak-to-peak voltage were generated The response signal

was measured by the 14-bit DIG For the signal-to-noise ratio reduction both pitch-catch and

pulse-echo signals were measured 150 times and averaged in the time domain 3 identical

circular PZTs of 10mm diameter manufactured by APC International Ltd were installed on

the specimen A Gaussian pulse (tone-burst) 50 kHz central frequency was excited as an in-

put signal The excitation frequency was chosen in order to avoid overlapping of the S0 and

A0 Lamb wave modes and generating of the higher modes (S1 A1hellip)

- 22 -

PZT Actuator

PZT Sensor

AWG

DIG

DAQ System Specimen

SensingActuatingMUX

1

23

4

1 Sending signal to MUX

4 Sending signal to DIG

Self-SensingCircuit

18 Unit mm

10

(a)

(b)

Fig 9 Overall experimental setup (a) Data acquisition (DAQ) system and (b) Installed circu-lar PZT transducer on the specimen with its dimension

513 Impact test

The impact-induced damage was introduced to the specimen using the impact test equip-

ment presented in Fig 10 A 5kg mass with 10mm tip was dropped from 015m height onto

the same location of the specimen The signals were obtained at each number of impacts

Since the reflected wave is not detected at the first impact case the data obtained from the

second and third impact were utilized for the analysis

10

Unit mm (a)

(b)

Fig 10 (a) Impact test equipment for introducing impact-induced damage and (b) Tip of the impact tester with 10mm diameter

- 23 -

514 Temperature effect test

In reality structures are subjected to changing environmental conditions such as tempera-

ture that affect measured signal and cause false-alarm Thus the robustness of the proposed

damage characterization technique under temperature variation in investigated (Fig 11) The

specimens were placed inside a temperature chamber and a thermocouple was installed on

the specimen to measure temperature of surface of the specimen The signal was obtained

under three temperature cases (0 20 and 50degC) and the humidity was kept at 30

Temperature Chamber

Test Specimen

Fig 11 Test specimen in temperature chamber

- 24 -

52 Teflon inserted damage

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Reflected wave from damage

(a)

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Transmitted WaveReflected Wave

Detected reflected wave

(b)

Fig 12 A0 Lamb wave mode obtained using pitch-catch method from the Teflon inserted damage specimen (a) The reflected wave from damage is shown (b) The Teflon inserted damage is detected by capturing the reflected wave

Fig 12 shows the obtained pitch-catch signal from the Teflon inserted damage specimen

and damage diagnosis result The Teflon inserted damage is detected successfully by captur-

ing the reflected wave Table IV presents the result of the Teflon inserted damage characteri-

zation 119907119907119894119894 is measured as 1213ms and the variation is assumed to plusmn 5 of the normal val-

ue (1152~1273ms) The results indicate that the proposed damage characterization approach

works well in real application

TABLE IV Damage characterization for the Teflon inserted damage specimen

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

Exact 635 25

Method 1 123 60~75 73~338 298~1220

Method 2 123 619~662 229~310 934~1188

Method 3 123 631 284 1161

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1213 ms

- 25 -

53 Impact-induced damage

11 115 12 125 13-4

-2

0

2

4

Time (ms)

Vol

tage

(mV

)

IntactImpact2Impact3

Reflected wave from damage

(a) Pitch-catch signal

11 115 12 125 13 135

-05

0

05

1

Time (ms)

Volta

ge (m

V)

IntactImpact2Impact3

Boundary reflection

Reflected wave from damage

(b) Pulse-echo signal

Fig 13 The signal obtained from the test specimen subjected to impact The impact was in-troduced at the same location of the specimen and the signal was obtained after each impacts As the number of impact is increased the transmitted A0 Lamb wave mode is delayed and attenuated

The obtained signals from impact-induced damage specimen are presented in Fig 13 It is

observed that as the number of impact is increased the transmitted A0 Lamb wave mode is

more delayed because the wave velocity in the damage area is reduced Fig 14 shows the

false-alarm test result whether the proposed damage detection algorithm works even in pris-

tine condition of the test specimen It is shown that no false alarm is occurred because no ad-

ditional reflected wave is captured

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Transmitted Wave

(a) Transmitted wave

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Reflected Wave

No reflected wave

(b) Reflected wave

Fig 14 False alarm test for the proposed damage detection algorithm in pristine condition of the specimen Reflected wave from damage is not captured false alarm is not occurred

- 26 -

TABLE V Damage characterization for the impact-induced damage specimen

of Impacts m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

2

(1470 J)

Method 1 080 65~90 75~246 470~1157

Method 2 080 801~885 114~245 711~1134

Method 3 080 843 179 1122

3

(2205 J)

Method 1 148 50~100 06~380 21~1134

Method 2 148 758~838 249~368 843~1125

Method 3 148 798 308 1044

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1213 ms

Table V presents the characterization results of the impact-induced damage case Here the

measured 119907119907119894119894 and the assumed variation are same as the artificial damage case It is observed

that as the number of impact is increased 119889119889 is increased while 119907119907119889119889 is reduced Thus it is

validated that the proposed damage characterization approach estimates the increment of the

damage well

54 Temperature effect test

541 Teflon inserted damage

11 115 12 125 13-10

-5

0

5

10

Time (ms)

Vol

tage

(mV

)

0 oC20 oC50 oC

(a) Pitch-catch signal

11 115 12 125 13 135

-1

0

1

2

3

Time(ms)

Volta

ge (m

V)

0 oC20 oC50 oC

(b) Pulse-echo signal

Fig 15 Signals obtained from the Teflon inserted damage specimen under temperature varia-tion Time delay and amplitude variation are occurred

Fig 15 shows the obtained signals from the Teflon inserted damage specimen subjected to

- 27 -

temperature variation Time delay and amplitude variations are occurred there is no addition-

al damage is introduced The analyzed results of the obtained signals are reported in Table VI

The measured 119907119907119894119894 is 1247 ms at 0 degC 1213 ms at 20 degC and 1168 ms at 50 degC respectively

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms (1152~1273 ms)

The estimated results are also changed under temperature variations the fluctuation of esti-

mated range and exact value is small

TABLE VI Temperature effect test for the Teflon inserted damage specimen

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

0 degC

Exact 635 25

Method 1 113 60~70 79~268 352~1291

Method 2 113 560~619 233~344 1034~1172

Method 3 113 606 257 1141

20 degC

Exact 635 25

Method 1 123 60~75 73~338 298~1220

Method 2 123 619~662 229~310 934~1188

Method 3 123 631 284 1161

50 degC

Exact 635 25

Method 1 137 60~75 72~328 263~1228

Method 2 137 622~687 183~299 666~1086

Method 3 137 642 264 960

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1247 ms at 0 degC 1213 ms at 20 degC and 1168 ms at 50 degC

- 28 -

542 Impact-induced damage

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Vol

tage

(mV

)

Transmitted WaveReflected Wave

No reflected wave

(a) 0 degC

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Transmitted WaveReflected Wave

No reflected wave

(b) 50 degC

Fig 16 False alarm test under temperature variations It is shown that proposed damage char-acterization algorithm is robust under temperature variations

The false alarm test in pristine condition of the impact-induced damage specimen under

temperature variation was investigated (Fig 16) It is validated that the proposed damage di-

agnosis algorithm is robust under temperature variations Fig 17 shows the obtained signals

from the impact-induced damage (impacted 3 times) specimen under temperature variation

The analyzed results of the obtained signals are reported in Table VII The measured 119907119907119894119894 in

each temperature and the assumed variation are same as the artificial damage case

11 115 12 125 13-5

0

5

Time (ms)

Volta

ge (m

V)

0 oC20 oC50 oC

(a) Pitch-catch signal

11 115 12 125 13 135

-1

0

1

2

3

Time(ms)

Vol

tage

(mV

)

0 oC20 oC50 oC

(b) Pulse-echo signal

Fig 17 Signals obtained from impact-induced damage specimen (Impact 3) under tempera-ture variation

- 29 -

TABLE VII Temperature effect test for the impact induced damage specimen (Impact 3 case)

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

0 degC

Method 1 135 75~90 72~310 267~1148

Method 2 135 705~779 264~381 978~1126

Method 3 135 763 289 1071

20 degC

Method 1 148 50~100 06~380 21~1134

Method 2 148 758~838 249~368 843~1125

Method 3 148 798 308 1044

50 degC

Method 1 153 75~100 53~415 174~1227

Method 2 153 804~889 214~336 702~1103

Method 3 153 815 320 1050

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1247 ms at 0 degC 1213 ms at 20 degC and 1168 ms at 50 degC

- 30 -

CHAPTER 6 CONCLUDING REMARKS

61 Executive summary

In this study the reference-free damage detection localization and quantification tech-

nique in a multi layered composite plate is proposed Additional reflected A0 Lamb wave

mode from the damage due to multiple reflections in damage area is used as a feature for

characterizing the damage First damage is characterized in terms of a damage index (m-

value) defined as the ratio of the physical length of the damage to the wavelength of the A0

Lamb wave mode Then the range of possible damage location length and severity are esti-

mated with constraints This procedure can be accomplished without prior baseline data To

realize the proposed technique the matching pursuit technique is modified for reduction of

the error by considering the dispersive characteristic of Lamb wave The matching pursuit

method estimates the arrival time of each wave packets accurately even two wave packets are

overlapped The numerical simulations and experimental tests including Teflon tape inserted

and impact-induced damage are investigated for the feasibility of the proposed technique

Moreover the effectiveness of the proposed technique under changing temperature is exper-

imentally verified

- 31 -

62 Suggestion for further study

The limitations of this study and the possible improvement and further studies are listed as

follow

1 Measurement of impact-induced damage length using conventional NDT technique

Additional experimental verification for impact-induced damage characterization is

needed by measuring the real length of damage using conventional NDT technique

such as ultrasonic scan or infrared camera

2 Limitation of both side collocated PZT installation In this study excitation PZTs are

installed both side and collocated for decomposing A0 Lamb wave mode However it

can produce decomposition error due to imperfection of PZT location and their installa-

tion Moreover in many cases assess to both surface of structures is not allowed such

as aircraft applications To improve this limitation a reference-free damage characteri-

zation using single side PZTs will be considered

3 Complex structure application In this study damage is characterized in simple plate-

like structure When the monitored structure becomes complex the response signals al-

so becomes complex due to its geometry and boundary Thus the investigations for

complex structure application such as stiffener spar and curvature are needed

- 32 -

REFERENCES

[1] Cawley P and Alleyne D (1996) ldquoThe use of Lamb waves for the long range inspection of large structuresrdquo Ultrasonics 34 pp287-290

[2] Ghosh T Kundu T and Karpur P (1998) ldquoEfficient use of Lamb waves for detecting defects in large platesrdquo Ultrasonics 36 pp791-801

[3] Guirguitui V (2003) ldquoLamb wave generation with piezoelectric wafer active sensors for structural health monitoringrdquo SPIE Conference on Smart Structures and Materials San Diego CA 5391 pp111-422

[4] Ragvahan A and Cesnik CE S (2004) ldquoModeling of piezoelectric-based Lamb wave generation and sensing for structural health monitoringrdquo SPIE Conference on Smart Struc-tures and Materials San Diego CA 5391 pp419-430

[5] Rose JL Pilarski A and Dirit JJ (1993) ldquoAn approach to guided wave mode selection for inspection of laminated platerdquo Journal of Reinforced Plastics and Composites 12(5) pp536-544

[6] Guo N and Cawley P (1993) ldquoThe interaction of Lamb waves with delaminations in composite laminatesrdquo Journal of Acoustical Society of America 94(4) pp 2240-2246

[7] Valdez SHD and Soutis C (2002) ldquoReal-time nondestructive evaluation of fiber com-posite laminates using low-frequency Lamb wavesrdquo Journal of Acoustical Society of Amer-ica 111(5) pp 2026-2033

[8] Yang M and Qiao P (2005) ldquoModeling and experimental detection of damage in virous materials using the pulse-echo method and piezoelectric sensoractuatorsrdquo Smart Materials and Structures 14 pp1083-1100

[9] Lestari W and Qiao P (2005) ldquoApplication of wave propagation analysis for damage identification in composite laminated beamsrdquo Journal of Composite Materials 39(22) pp1967-1984

[10] Kessler S S Spearing S M and Soutis C (2002) ldquoDamage detection in composite ma-terials using Lamb wave methodsrdquo Smart Materials and Structures 11 pp269-278

[11] Sohn H Park G Wail J R Limback N P and Farrar C R (2004) ldquoWavelet-based active sensing for delamination detection in composite structuresrdquo Smart Materials and Structures 13 pp 153-160

[12] Petculescu G Krishnaswamy S and Achenback J D (2008) ldquoGroup delay measure-ments using modally selective Lamb wave transducers for detection and sizing of delamina-tions in compositesrdquo Smart Materials and Structures 17

[13] Sohn H Park H W Law K H and Farrar C R (2007) ldquoDamage detection in compo-

- 33 -

site plates by using an enhanced time reversal methodrdquo Journal of Aerospace Engineering (ASCE) pp 141-151

[14] Sohn H (2007) ldquoEffects of environmental and operational variability on structural health monitoringrdquo Philosophical Transactions of the Royal Society Ardquo 365 pp539-560

[15] Rytter A (1993) ldquoVibration based inspection of civil engineering structuresrdquo PhD The-sis Department of Building Technology and Structural Engineering University of Aalborg Aalborg Denmark

[16] Hayashi T and Kawashima K (2002) ldquoMultiple reflections of Lamb waves at a delami-nationrdquo Ultrasonics 40 pp193-197

[17] Wang C H and Rose L R F (2003) ldquoWave reflection and transmission in beams con-taining delamination and inhomogeneityrdquo Journal of Sound and Vibration 264 pp 851-872

[18] Ramadas C Balasubramaniam K Joshi M and Krishnamurthy C V (2009) ldquoInterac-tion of the primary anti-symmetric Lamb mode (Ao) with symmetric delaminations numer-ical and experimental studiesrdquo Smart Materials and Structures 18

[19] Ramadas C Balasubramaniam K Joshi M and Krishnamurthy C V (2010) ldquoInterac-tion of guided Lamb waves with an asymmetrically located delamination in a laminated composite platerdquo Smart Materials and Structures 19

[20] Wilcox P Lowe M and Cawely P (2001) ldquoThe effect of dispersion on long-range in-spection using ultrasonic guided wavesrdquo NDTampE International 34 pp1-9

[21] Mallat S G and Zhang Z (1993) ldquoMatching pursuits with time-frequency dictionariesrdquo IEEE Tansactions on Signal Processing 41(12) pp3397-3415

[22] Zhang G Zhang S and Wang Y (2000) ldquoApplication of adaptive time-frequency de-composition in ultrasonic NDE of highly scattering materialsrdquo Ultrasonics 38 pp 961-964

[23] Hong J C Sun K H and Kim Y Y (2005) ldquoThe matching pursuit approach based on the modulated Gaussian pulse for efficient guided-wave damage inspectionrdquo Smart Mate-rials and Structures 14 pp548-560

[24] Herakovich C T (1997) Mechanics of Fibrous Composites Wiley 480 pages

[25] Reddy J N (1985) ldquoA review of the literature on finite-element modeling of laminated composite platesrdquo The Shock and Vibration Digest 17(4) pp 3-8

[26] Reddy J N (2003) Mechanics of Laminated Composite Plates and Shells Theory and Analysis CRC Press 856 pages

[27] Palacz M Krawczuk M and Ostachowicz W (2005) ldquoThe spectral finite element mod-el for analysis of flexure-shear coupled wave propagation Part 1 laminated multilayer

- 34 -

composite beamrdquo Composite Structures 68 pp37-44

[28] Zhang Y C and Harding J (1990) ldquoA numerical micromechanics analysis of the me-chanical properties of a plain weave compositerdquo Computer amp Structures 36(5) pp 839-844

[29] Nemat-Nasser S and Hori M (1999) Micromechanics Overall Properties of Heteroge-neous Materials Elsevier 810 pages

[30] Scida D Aboura Z Benzeggagh M L and Bocherens E (1999) ldquoA micromechanics model for 3D elasticity and failure of woven-fibre composite materialrdquo Composites Science and Technology 59 pp 505-517

[31] Thomas Y and Wu X H (1997) ldquoA multiscale finite element method for elliptic prob-lems in composite materials and porous mediardquo Journal of Computational Physics 134 pp 169-189

[32] Lee K H Moorthy S and Ghosh S (1999) ldquoMultiple scale computational model for damage in composite materialsrdquo Computer Methods in Applied Mechanics and Engineer-ing 172 pp 175-201

[33] Namilae S and Chandra N (2005) ldquoMultiscale model to study the effect of interfaces in carbon nanotube-based compositesrdquo Journal of Engineering Materials and Technology (ASME) 127 pp222-232

[34] Adams D E (2007) Health monitoring of Structural Materials and Components Wiley 460 pages

[35] Gopalakrishnan S (2009) ldquoModeling Aspects in Finite Elementrdquo Encyclopedia of Struc-tural Health Monitoring 2 pp791-809

[36] Palacz M Krawczuk M and Ostachowicz W (2005) ldquoThe spectral finite element mod-el for analysis of flexure-shear coupled wave propagation Part 2 Deminated multilayer composite beamrdquo Composite Structures 68 pp45-51

[37] Baik J M and Thopson R B (1984) ldquoUltrasonic scattering from imperfect interfaces A quasi-static modelrdquo Journal of Nondestructive Evaluation 4(3) pp177-196

[38] Luo H and Hanagud S (2000) ldquoDynamics of delamincated beamsrdquo International Jour-nal of Solids and Structures 37 pp1501-1519

[39] Kim S B and Sohn H (2007) ldquoInstantaneous reference-free crack detection based on polarization characteristics of piezoelectric materialsrdquo Smart Materials and Structures 16 pp2375-2387

[40] An Y K and Sohn H (2010) ldquoInstantaneous crack detection under varying temperature and static loading conditionsrdquo The Special Issue of Structural Control and Health Monitor-ing 17(7) pp 730-741

- 35 -

[41] Lee S J and Sohn H (2006) ldquoActive self-sensing scheme development for structural health monitoringrdquo Smart Materials and Structures 15(6) pp 1734-1746

- 36 -

요 약 문

복합재료 손상의 무기저 진단 및 정량화

복합재료는 재료의 경량성 내구성 고강도의 특성으로 인하여 항공기에서부터

교량에까지 널리 사용되고 있다 하지만 외부의 충격으로 인한 복합재료 내부의

손상은 재료의 강도를 감소시켜 구조물의 안정성을 저하시키는 요인이 된다 하지만

복합재료의 손상은 재료 내부에서 발생하여 육안으로는 식별하기도 힘들며 표면에서

보이는 것만으로 내부의 손상 정도를 판단하기는 불가능하다는 단점을 가지고 있다

본 연구에서는 손상 내부에서 반사로 인해 발생하는 파를 이용하여 기저자료를

사용하지 않고 특정 시점에서 얻은 자료만으로 복합재료의 손상을 진단하고 정량화

할 수 있는 기법을 제안한다 제안된 기법의 적용성을 검증하기 위하여 유한요소법을

이용한 수치해석을 수행하였고 Teflon tape 이 삽입된 시편과 실제 충격하중으로

인한 손상을 입은 시편을 이용한 실험을 진행하였다 수치해석 및 실험 결과 제안된

기법이 복합재료의 손상을 성공적으로 진단 및 정량화할 수 있음을 확인하였다 또한

다양한 외부 온도 환경에서 제안된 기법의 환경영향 실험을 수행하였고 다양한

환경에서 손상의 진단 및 정량화가 성공적으로 이루어짐을 검증하였다

핵심어 무기저 손상진단 및 정량화 압전센서 복합재료 유한요소법 Matching pursuit

method

- 37 -

감 사 의 글

기대와 두려움을 안고 시작한 석사과정이 어느덧 2 년 이라는 세월 속에서 추억으로 자리잡아

가고 있습니다 학부생 기간을 포함한 지난 3 년 동안 SHM 이라는 새로운 분야를 접하고 연구를

수행 할 수 있도록 도와주신 고마운 분들이 있어 그 감사를 표하고자 합니다

우선 저의 많은 부족한 점을 일깨워 주시고 연구에 대한 뜻을 더욱 확고히 세울 수 있게 해

주신 지도교수님이신 손 훈 교수님께 깊은 감사의 뜻을 전합니다 짧은 기간이었지만 교수님의

도움으로 많은 것을 배우고 느끼고 깨달을 수 있었습니다 또한 본 연구를 수행할 수 있도록 지

원해주신 국토해양부 (U-City 석 박사과정 지원사업)에 깊은 감사 드립니다

때로 힘들고 지칠 때 항상 저를 믿고 응원해 주시는 가족이 있었기에 지금의 글을 쓸 수 있었

습니다 또한 지금 연구실에서 함께 생활하고 있는 윤규형 민구형 철민이형 현석이형 진열이형

호민이형 병진이 선혜누나와 성대에 계신 창길이형 미국에 계신 의재형과 Debaditiya 에게도 감

사의 말씀을 전합니다 그리고 모두 나열할 수는 없지만 카이스트 생활의 활력소가 되어준 카이

스트 건설 및 환경공학과 선후배들 직원 분들과 SOS 멤버들에게도 감사를 표합니다

앞서 거론한 많은 이들의 도움이 없었다면 저 또한 지금 이 자리에 있을 수 없었을 것입니다

많이 부족하기에 더 노력하고 배우고자 했던 초심을 다시 한번 생각하며 더 큰 결실을 맺기 위해

앞으로 나아가겠습니다 도움을 주신 모든 분들께 이 감사의 글을 바칩니다

2011년 6월 카이스트에서 임형진 올림

- 38 -

CURRICULUM VITAE

Hyung Jin Lim

Research Assistant

The Civil and Environmental Engineering Department

Korea Advanced Institute of Science and Technology (KAIST)

291 Daehak-Ro Guseong-Dong Yuseong-Gu

Daejeon Republic of Korea 305-701

Phone (82)+42-350-3665 Fax (82)+42-350-3610

Email limhj87gmailcom

RESEARCH INTERESTS

Damage Detection Localization and Quantification on Composite and Metallic Structures Pattern Recognition Finite Element Modeling (FEM)

EDUCATION

2009 ndash 2011 MS Civil and Environmental Eng Korea Advanced Institute of Science and Technology (KAIST) Korea Rep

2006 ndash 2009 BS Civil and Environmental Eng Korea Advanced Institute of Science and Technology (KAIST) Korea Rep

JOURNAL PUBLICATIONS The corresponding author is underlined 1 Hyung Jin Lim Hoon Sohn ldquoReference-free damage detection localization and quantification

using in compositerdquo in preparation for Journal of the Acoustical Society of America 2011

2 Chul Min Yeum Hoon Sohn Jeong Bum Ihn Hyung Jin Lim ldquoInstantaneous delamination detection in a composite using a dual piezoelectric transducer networkrdquo Submitted to Compo-sites Science and Technology 2011

3 Yun-Kyu An Hyung Jin Lim Min Koo Kim Jin Yeol Yang Hoon Sohn Chang Guen Lee ldquoInvestigation of applicability of local reference-free damage detection techniques to in-situ

- 39 -

bridgesrdquo Submitted to ASCE Journal of Structural Engineering 2011

4 Hyung Jin Lim Min Koo Kim Hoon Sohn Chan Yik Park ldquoImpedance based damage detec-tion under varying temperature and loading conditionsrdquo Submitted to NDTampE International 2011

5 Hyung Jin Lim Hoon Sohn ldquoQuantitative estimation of transmitted and reflected lamb waves at discontinuityrdquo Journal of the Korean Society for Nondestructive Testing Vol30 No 4 pp1-8 2010

CONFERENCE PROCEEDINGS

The corresponding author is underlined 1 Hyung Jin Lim and Hoon Sohn ldquoReference-free damage detection localization and quantifi-

cation in compositesrdquo ASCE Engineering Mechanics Institute Conference Boston June 2-4 2011

2 Hyung Jin Lim and Hoon Sohn ldquoReference-free delamination quantification within multi-layer composite materialsrdquo KSNT Spring Conference May 26-27 Gyeong-Ju Korea 2011

3 Min Koo Kim Hyun Jin Lim Hoon Sohn Chan Yik Park ldquoImpedance-based bolt loosening detection under varying temperature and loadingrdquo Asian Pacific Workshop on Structural Health Monitoring Tokyo Japan November 30-December 2 2010

4 Hyung Jin Lim Chul Min Yeum and Hoon Sohn ldquoModeling of Impact-induced Delamination in a Multilayer Composite Platerdquo The 23rd KKCNN Symposium on Civil Engineering Taipei (Taiwan) November 13-15 2010

5 Yun-Kyu An Hyung Jin Lim and Hoon Sohn ldquoTemperature independent Delamination Detec-tion using data normalizationrdquo The 3rd International Conference on Multi-Functional Materi-als and Structures Chonbuk National Univ Jeonju Korea September 14-18 2010

6 Hoon Sohn Yun-Kyu An Chang Gil Lee Min Koo Kim Hyung Jin Lim ldquoApplication of online NDT techniques to decommissioned bridge testingrdquo the 5th European Workshop on Structural Health Monitoring Sorrento Italy June 29-July 02 2010

7 Min Koo Kim Hyung Jin Lim Hoon Sohn Chan Yik Park ldquoImpedance based damage diag-nosis of a complex composite aircraft wing under changing loading and temperature condi-tionsrdquo the 5th European Workshop on Structural Health Monitoring Sorrento Italy June 29-July 02 2010

8 Yun-Kyu An Min Koo Kim Hyung Jin Lim and Hoon Sohn ldquoValidations of reference-free crack detection techniques through a decommissioned bridge testrdquo 2010 COSEIK Annual Con-ference Jeju Korea April 8-9 2010

- 40 -

PARTICIPANT PROJECTS

1 Development of Smart Steel Members for Online Steel Structure Monitoring POSCO Corporation (Funded 41000000 KRW (41000 USD) for 010511 to 123111)

2 Development of Enabling Technologies for Structural Health Monitoring Korean Agency for Defense Development (Funded 240000000 KRW (240000 USD) out of total 3200000000 KRW (3200000 USD) for 060108-053111)

3 Development of On-board SHM Technologies for Composite Air Vehicles The Boeing Company (Funded 400000 USD (400000000 KRW) for 080108 to 073111)

EDUCATIONAL EXPERIENCES

Fall 2009

Teaching Assistant CE207 ldquoElementary Structural Engineering and Laboratoryrdquo KAIST 3 unit undergraduate course

EXPERIMENTAL EXPERIENTCES

2010 Monitoring of Real Aircraft Composite Wing Segment Korea Agency for Defense De-velopment (ADD) Daejeon Korea Real time monitoring of composite air vehicles fund-ed by Korea Agency for Defense Development (ADD)

2009-2011

Field Monitoring of Real Bridge Structures Gimpo (Ramp G bridge) amp Icheon (Sam-seung bridge) Korea US-Korea collaborative research for bridge monitoring testbeds

HONORS amp AWARDS

Scholarships for outstanding students granted KAIST (2007-2009) The three prominent students in the civil and environmental engineering department of KAIST were selected as recipients This scholarship provided 2000$ per year for three years

- 41 -

Page 4: 복합재료 손상의 무기저 진단 및 정량화 - KAISTssslab.kaist.ac.kr/article/pdf/Thesis_2.pdf · 1 Declaration of Ethical Conduct in Research: I, as a graduate student

복합재료 손상의 무기저 진단 및 정량화

임 형 진

위 논문은 한국과학기술원 석사학위논문으로

학위논문심사위원회에서 심사 통과하였음

2011 년 5 월 30 일

심사위원장

심사위원

심사위원

손 훈 (인)

이 행 기 (인)

김 천 곤 (인)

MCE

20094258

임 형 진 Lim Hyung Jin Reference-free Damage Detection Localization and

Quantification in Composite 복합재료 손상의 무기저 진단 및 정량화 Department

of Civil and Environmental Engineering 2011 41 p Advisor Prof Sohn Hoon

ABSTRACT

A reference-free damage detection localization and quantification technique in a multi

layered composite plate is proposed It is reported that when A0 Lamb wave mode propa-

gates the damage time delay attenuation and multiple reflections are occurred in the dam-

age area In this thesis additional reflected waves from the damage due to multiple reflec-

tions are used as a feature for characterizing the damage First the damage is identified by

detecting the additional reflected waves from the damage area Then the location of the

damage is estimated using combined pitch-catch and pulse-echo methods Finally the dam-

age is quantified using a damage index defined as the ratio between the physical length of

the damage and the wavelength of the A0 Lamb wave mode The matching pursuit method

is utilized for analyzing the obtained signal by decomposing overlapped waves and estimat-

ing the parameters of each wave such as arrival time driving frequency amplitude and so

on The proposed technique has an advantage that the damage can be characterized without

reference data obtained from pristine condition of the monitored structure within a single

path The feasibility of the proposed technique is verified through numerical simulations

and experimental tests In numerical simulation different damage models (length material

properties and location) are investigated In experimental tests a Teflon tape inserted and

impact-induced damage are characterized under temperature variations The results are

suggesting a reliable damage characterization in composites

Keywords Reference-free PZT transducers Composites Finite element method Matching

pursuit method

i

TABLE OF CONTENTS

ABSTRACT middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot i

TABLE OF CONTENTS middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot iii

LIST OF TABLES middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot iv

LIST OF FIGURESmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot v

CHAPTER 1 INTRODUCTION

11 Motivation middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 1

12 Literature review middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 1

13 Objective and uniqueness middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 3

14 Organization middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 4

CHAPTER 2 INTERACTION OF LAMB WAVE WITH DAMAGE

21 A0 Lamb wave mode delay and attenuation in the time domain middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 5

22 Multiple reflections in the damage area middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 6

CHAPTER 3 THEORETICAL DEVELOPMENT

31 Overview middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 7

32 Damage index (m-value) middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 8

33 Damage detection algorithm (Level 1) middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 10

34 Damage quantification algorithm (Level 3) middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 11

35 Damage localization algorithm (Level 2amp3) middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 11

36 Matching pursuit method considering the dispersion effect of Lamb wave

middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 13

CHAPTER 4 NUMERICAL SIMULATION

41 Composite modeling methodmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 15

42 Damage modeling method middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 16

43 Numerical simulation description middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 17

44 Numerical simulation results middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 19

ii

CHAPTER 5 EXPERIMENTAL VERIFICATION

51 Overview middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 21

511 Specimen middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 22

512 Data acquisition system and PZT middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 22

513 Impact test middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 23

514 Temperature effect test middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 24

52 Teflon inserted damage middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 25

53 Impact-induced damagemiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 26

54 Temperature effect test middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 27

541 Teflon inserted damage middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 27

542 Impact-induced damage middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 29

CHAPTER 6 CONCLUDING REMARKS

61 Executive summary middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 31

62 Suggestion for further study middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 32

REFERCENCES middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 33

SUMMARY (IN KOREAN) middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 37

ACKNOWLEDGEMENTS (IN KOREAN)

CURRICULUM VITAE

iii

LIST OF TABLES

I Material property of simulation model and test specimen middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 17

II Damage cases for numerical simulation middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 18

III Damage characterization for the numerical simulation models middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 20

IV Damage characterization for the Teflon inserted damage specimen middotmiddotmiddotmiddotmiddotmiddotmiddot 25

V Damage characterization for the impact-induced damage specimen middotmiddotmiddotmiddotmiddotmiddotmiddot 27

VI Temperature effect test for the Teflon inserted damage specimen middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 28

VII Temperature effect test for the impact induced damage specimen middotmiddotmiddotmiddotmiddotmiddotmiddot 30

iv

LIST OF FIGURES

1 Lamb wave propagation characteristics (a) A0 Lamb wave mode time delay and

attenuation after passing through a damage (b) The time delay and amplitude varia-

tion of A0 Lamb wave mode under temperature variation in pristine condition of

structure Because of this conventional damage detection techniques in composites

have limitations for real structure application middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 5

2 Multiple reflections in the damage area The multiple reflections of A0 Lamb

waves in the damage area are utilize as a feature for characterization of the damage

in this study middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 6

3 Parameters and notations of the proposed damage characterization technique

Damage is detected by identifying the additional reflected wave from the damage due

to multiple reflections (dashed lines) Localization and quantification of the damage

is carried out by measuring the arrival time of transmitted and reflected waves from

the damage middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 7

4 Two different FEM models which have different 119889119889 and 119907119907119889119889 (or G) This repre-

sents that the damage cannot be characterized uniquely based on the measured data

available middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 9

5 (a) Example of the overlapped signal with comparing approximated signal using

the matching pursuit method (b) Decomposed signal using the matching pursuit

method The matching pursuit method provides a conventional way to decompose

and identify each wave middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 9

6 Description of the numerical simulation model middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 17

7 Comparison of A0 Lamb wave mode signals from numerical simulation be-

tween intact and damage cases Time delay attenuation and multiple reflections due

to damage simulated successfully using the damage modeling method described in

v

the previous section middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 19

8 Test specimens for the experimental verification In the Teflon inserted damage

Teflon tapes with 25mm diameter were inserted at the center of the specimen The

impact-induced damage was introduced to the specimen using the impact test equip-

ment middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 21

9 Overall experimental setup (a) Data acquisition (DAQ) system and (b) Installed

circular PZT transducer on the specimen with its dimension middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 23

10 (a) Impact test equipment for introducing impact-induced damage and (b) Tip

of the impact tester with 10mm diametermiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 23

11 Test specimen in temperature chamber middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 24

12 A0 Lamb wave mode obtained using pitch-catch method from the Teflon in-

serted damage specimen (a) The reflected wave from damage is shown (b) The Tef-

lon inserted damage is detected by capturing the reflected wave middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 25

13 The signal obtained from the test specimen subjected to impact The impact

was introduced at the same location of the specimen and the signal was obtained after

each impacts As the number of impact is increased the transmitted A0 Lamb wave

mode is delayed and attenuated middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 26

14 False alarm test for the proposed damage detection algorithm in pristine condi-

tion of the specimen Reflected wave from damage is not captured false alarm is not

occurred middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 26

15 Signals obtained from the Teflon inserted damage specimen under temperature

variation Time delay and amplitude variation are occurred middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 27

16 False alarm test under temperature variations It is shown that proposed dam-

vi

age characterization algorithm is robust under temperature variationsmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 29

17 Signals obtained from impact-induced damage specimen (Impact 3) under

temperature variation middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 29

vii

CHAPTER 1 INTRODUCTION

11 Motivation

Composite materials has been increasingly demands in modern structures ranging from the

latest aircraft to retrofitted bridge deck due to its light weight superior specific strength and

stiffness properties However the occurrence of damage due to abrupt impact or sequential

accumulation of defect may considerably degrade structural strength or stiffness and reduce

the safety and reliability of the system Recently Lamb waves based damage detection tech-

niques using piezoelectric (PZT) transducers have emerged as one of the promising structural

health monitoring (SHM) technology for detecting hidden damage because of the possibility

of long range inspection and high sensitivity [1-4]

12 Literature review

This section will discuss researches that have been previously carried out on topics of

guided wave based SHM techniques using PZT transducers for damage in composite materi-

als The early investigations of interaction between Lamb wave and damage in composite can

be found in Rose et al and Guo et al [5-6] From further studies reflected signals from dam-

age were shown to be an effective method for damage diagnosis and localization using PZTs

[7-9] The reflected signals from damage were collected using one PZT as an actuator and the

other PZT as a sensor However the techniques cannot be used to determine the extent of

- 1 -

damage with reasonable precision

Kesseler et al proposed a technique monitoring the wavelet coefficients of transmitted ul-

trasonic signal amplitude as a potential way to damage detection [10] One drawback of this

technique is that the variability in coupling and the sensitivity of installed PZT transducers

can induce changes in the amplitude and energy of the signal that are cause false-alarm Oth-

er wavelet based delamination detection and localization technique developed using PZT

sensor network and outlier analysis [11]

Damage detection and quantification in composite plate was also investigated by group de-

lay measurement due to damage [12] However it is assumed that the group velocity in dam-

age area as the group velocity propagates half thickness of the structure in dispersion curve in

the quantification procedure Thus the damage quantification was investigated only Teflon

inserted damage case

Sohn et al proposed a damage detection technique which does not rely on any past base-

line signals to assess damage in composite plates by using an enhanced time reversal method

[13] The damage is assessed using the violation of time reversibility when nonlinearity is

caused by the damage along a direct wave path Thus it allows instantaneous identification

of delamination without requiring the baseline signal for comparison

- 2 -

13 Objective and uniqueness

In this study a reference-free damage characterization technique including detection lo-

calization and quantification of impact-induced damage in composite materials is developed

The general damage characterization can be categorized into the following levels [14]

Level 0 Define damage

Level 1 Determine the presence of damage within the structure

Level 2 Locate the regions of damage

Level 3 Quantify the damage

Level 4 Predict the remaining service life of the structure

Through the proposed technique from level 1 to level 3 can be achieved

Most of common approaches of damage characterization techniques described in the pre-

vious section use prior obtained baseline data However these approaches can cause false

alarm due to the effect of environmental variations such as temperature and ambient vibra-

tions [15]

The uniqueness of this study is

1 Damage is characterized without using prior measured baseline data

2 Robustness of the proposed technique under temperature variation is investigated

3 Additional waves reflected from inside damage have been used in addition to transmit-

ted and reflected waves

- 3 -

14 Organization

The organization of this thesis is summarized as follows

Chaper1 A general introduction including motivation literature review objective and

uniqueness of this study is described

Chaper2 Interactions between Lamb wave mode and damage in composites such as time

delay attenuation and multiple reflections are summarized Damage index (m-value) for

damage characterization is defined

Chaper3 Algorithms for the reference-free damage characterization technique using m-

value are developed Modified matching pursuit for this study is also described

Chaper4 Verification of the algorithm using numerical simulation is presented Various

modeled damage cases are characterized

Chaper5 Experimental verifications are described Two damage cases (Teflon inserted and

impact-induced damage) are investigated for the feasibility of the proposed damage charac-

terization technique Temperature effect test using temperature chamber is also carried out

Chaper6 Conclusion and summary of the present study is reported In advance suggestions

and future work recommendations are represented

- 4 -

CHAPTER 2 INTERACTION OF LAMB WAVE WITH DAMAGE

21 A0 Lamb wave mode delay and attenuation in the time domain

Time delay and attenuation of the A0 mode Lamb wave after passing through the damage

area due to effective thickness reduction were reported in previous researches [812] Moreo-

ver the attenuation of the A0 Lamb wave mode is occurred by degrading of shear modulus of

in the damage area However the time delay and amplitude variation of the A0 Lamb wave

mode also occurred even when the structure is health condition by the material property

changing due to temperature variation of the structure (Fig 1) Thus conventional damage

characterization methods using the time delay and attenuation of A0 Lamb wave mode as a

feature can cause false-alarm in real application

11 115 12 125 13-10

-5

0

5

10

Time (ms)

Vol

tage

(mV

)

IntactDamage

Reflected wave from Damage

(a)

11 115 12 125 13-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

0 oC20 oC50 oC

(b)

Fig 1 Lamb wave propagation characteristics (a) A0 Lamb wave mode time delay and atten-uation after passing through a damage (b) The time delay and amplitude variation of A0 Lamb wave mode under temperature variation in pristine condition of structure Because of this conventional damage detection techniques in composites have limitations for real struc-ture application

- 5 -

22 Multiple reflections in the damage area

Standing wave patterns due to multiple reflections of A0 Lamb wave mode in the damage

area were observed through numerical simulations and ultrasonic wave-field imaging using a

scanning laser vibrometer [16] The formation of standing waves in damage is illustrated

through Fig 2 When the incident waves enter the damage area a significant portion of the

waves are reflected back from the exit Then the reflected waves travel through the damage

and undergo reflection again at the original entrance As a consequence of multiple reflec-

tions at the entrance and the exit a considerable amount of ultrasonic energy is trapped inside

the damage area [17-19]

Multiple Reflections

Incident Wave Transmitted Wave

Damage

Fig 2 Multiple reflections in the damage area The multiple reflections of A0 Lamb waves in the damage area are utilize as a feature for characterization of the damage in this study

- 6 -

CHAPTER 3 THEORETICAL DEVELOPMENT

31 Overview

Damage

l = l1 + d + l2

d = vdtd l1 = vit1 l2 = vit2

t = t1 + td + t2

Excitation and Sensing PZT Sensing PZT

l

Transmitted Wave

Reflected Waves

Incident Wave

Fig 3 Parameters and notations of the proposed damage characterization technique Damage is detected by identifying the additional reflected wave from the damage due to multiple re-flections (dashed lines) Localization and quantification of the damage is carried out by measuring the arrival time of transmitted and reflected waves from the damage

In this section the reference-free damage characterization algorithm is proposed Damage

diagnosis and quantification are achieved by identifying the additional reflected waves due to

multiple reflections in the damage area (dashed lines in Fig 3) Moreover damage is local-

ized using additional information from pulse-echo method implement For the localization

and the quantification of damage the variables such as the path length and A0 Lamb wave

mode velocity in damage area are defined (Fig 3) For example 119897119897 is expressed as

119897119897 = 1198971198971 + 119889119889 + 1198971198972 (1)

where 1198971198971 119889119889 and 1198971198972 are the distance from the excitation PZTs to the entrance of damage the

damage length and the distance from the exit of damage to the sensing PZT respectively

- 7 -

Moreover 119907119907119894119894 and 119907119907119889119889 are also defined as the A0 Lamb wave mode velocity in pristine area

and damage area respectively Note that the whole procedure of the damage characterization

is accomplished based solely on the signals obtained from the current state of the structure

32 Damage index (m-value)

The proposed reference-free damage characterization technique utilizes the waves reflect-

ed additionally from the damage area as a feature The arrival time difference of the transmit-

ted and reflected A0 Lamb wave modes reflected from the exit and the entrance of the delam-

ination area one times each (∆119905119905) contains the information of the physical length of the dam-

age (119889119889) and A0 Lamb wave mode velocity in the damage area (119907119907119889119889) The wave velocity in the

damage area which is related to the shear modulus (G) represents the severity of the damage

∆119905119905 = 2119905119905119889119889 =2119889119889119907119907119889119889

(2)

However based on the measured data available the 119889119889 and 119907119907119889119889 (or G) cannot be uniquely

determined Fig 4 shows two FEM models which have different physical length and wave

velocity (G) of the damage Detail explanation of delamination modeling technique will be

covered in section 4 The similar signal in the time domain and ∆119905119905 are obtained from the

two different damage cases

- 8 -

Damage 1

Damage 2

20

Unit mm15

90 G value reduction

943 G value reduction

m-value = 248

m-value = 248

015 02 025 03 035

-2

-1

0

1

2

Time (ms)

Ampl

itude

Damage 1Damage 2

Fig 4 Two different FEM models which have different 119889119889 and 119907119907119889119889 (or G) This represents that the damage cannot be characterized uniquely based on the measured data available

Thus in this study the damage index (m-value) which can be uniquely defined based on

the available data is introduced The m-value is defined as the ratio of the physical length of

the damage and wavelength of the A0 Lamb wave mode in the damage area (120582120582119889119889)

119898119898 =119889119889120582120582119889119889

=119889119889119889119889119907119907119889119889

(3)

In many cases the transmitted and reflected waves are overlapped and it is hard to decom-

pose and calculate the arrival times of them Matching pursuit method is an alternative way

to accurately decompose of the overlapped waves and estimate the parameters of each wave

such as arrival time (Fig 5) Detail explanation of the matching pursuit method is covered in

section 36

015 02 025 03-2

-1

0

1

2

Time (ms)

Am

plitu

de

OriginalApproximated

(a)

015 02 025 03-2

-1

0

1

2

Time (ms)

Ampl

itude

Wave 1Wave 2

(b)

Fig 5 (a) Example of the overlapped signal with comparing approximated signal using the matching pursuit method (b) Decomposed signal using the matching pursuit method The matching pursuit method provides a conventional way to decompose and identify each wave

- 9 -

33 Damage detection algorithm (Level 1)

The damage diagnosis is accomplished by extracting the additional reflected A0 Lamb

wave modes from damage due to multiple reflections If A0 Lamb wave modes are decom-

posed perfectly it is trivial that one transmitted A0 Lamb wave mode is always captured in

the response signal If damage exists other waves in the response signal behind the transmit-

ted A0 Lamb wave mode are presented Since this approach does not rely on previously col-

lected data the diagnosis is independent from environmental variation such as temperature

changes

It is reported that the dispersion of Lamb waves cause wave-packets to spread out in space

and time as they propagate through a structure [20] Thus the reflected waves from the dam-

age have longer pulse duration than the transmitted wave due to their longer propagating dis-

tance In this study the matching pursuit method is modified to capture more spread wave

than the transmitted A0 Lamb wave mode Detail explanation of the modified matching pur-

suit method is described in latter section

Let 1199041199041 and 1199041199042 as the pulse duration (scale) of the transmitted and additional reflected A0

Lamb wave mode respectively

ldquoIf 1199041199042 gt 1199041199041 damage exists

Otherwise there is no damagerdquo

- 10 -

34 Damage quantification algorithm (Level 3)

As mentioned in the previous section damage is quantized by estimating 119889119889 and 119907119907119889119889 using

pitch-catch method The procedure to estimate 119889119889 and 119907119907119889119889 is

(i) Measure 119897119897 119905119905 and 119905119905119889119889

(ii) Assume the range of 119889119889 (0 lt 119889119889 lt 119897119897)

(iii) Calculate 119907119907119894119894 and 119907119907119889119889 with a constraints (0 lt 119907119907119889119889 lt 119907119907119894119894) (Eq 4 and 5)

(iv) Estimate m-value and find a feasible range of 119889119889 and 119907119907119889119889

119907119907119894119894 =119897119897 minus 119889119889119905119905 minus 119905119905119889119889

(4)

119907119907119889119889 =119889119889119905119905119889119889

(5)

35 Damage localization algorithm (Level 2amp3)

For damage localization 1198971198971 is estimated by measuring the arrival time from the excitation

PZT to the entrance of delamination (1199051199051) with additional pulse-echo method implementation

The procedure is similar with the previous damage quantification algorithm

(i) Measure 119897119897 119905119905 119905119905119889119889 and 1199051199051

(ii) Assume the range of 1198971198971 (0 lt 1198971198971 lt 119897119897)

(ii) Compute possible combinations of 1198971198972 119889119889 and 119907119907119889119889 with constraints

(0 lt 119907119907119889119889 lt 119907119907119894119894 119889119889 gt 0) (Eq 5 6 and 7)

- 11 -

1198971198972 = 1198971198971119905119905 minus 1199051199051 minus 119905119905119889119889

1199051199051 (6)

119889119889 = 119897119897 minus 1198971198971 minus 1198971198972 (7)

Since a PZT for pulse-echo method acts as both an actuator and a sensor the obtained sig-

nal contains both the excitation and the response signal Thus the reflected wave from the

entrance of damage can be buried into the tail of excitation signal and cause error on arrival

time of reflected wave from damage Even in this case damage can be localized and quan-

tized using the constraints When first wave packet reflected from the entrance of damage is

buried into the excitation signal the next wave packet reflected from the exit of damage can

be utilized alternatively If the algorithm regards the second reflected wave as the first re-

flected wave there are no combinations of estimated variable meeting the constraints Then

the algorithm repeats the process by regarding the wave as the second reflected wave the

constraints can select proper combination of the estimated variables

The estimated values have range bounded by the constraints When additional information

of 119907119907119894119894 is utilized the damage can be characterized more accurately In this study three meth-

ods are proposed and verified as information level of 119907119907119894119894

Method 1 Solely based on the current data measurement (Reference-free)

Method 2 Limit the possible range of 119907119907119894119894 (Prior knowledge of the system)

Method 3 Measure 119907119907119894119894 (Additional data need to be collected)

Method 1 described in this section can be utilized without any information of 119907119907119894119894 Meth-

od 2 produces more accurate estimation results using prior knowledge of the system such as

- 12 -

119907119907119894119894 variation under temperature changes Method 3 gives unique estimated value using meas-

ured 119907119907119894119894 from other path in the same time In section 4 and 5 damage is characterized using

these three methods and the results are compared each another

36 Matching pursuit method considering the dispersion effect of Lamb

wave

Matching pursuit method has been developed for signal compression approximation and

characterization of obtained signal and function [21-23] The key concept of the matching

pursuit method is that any signal (function) can be decomposed into a linear expansion of

signals selected from a pool of basis signals (functions) called lsquodictionaryrsquo

Let 119891119891 be a function to be analyzed then 119891119891 can be approximated a set of functions select-

ed from the dictionary 119863119863 in order to best match This is done by successive approximations

with orthogonal projections Let1198921198921205741205740 isin 119863119863 the function 119891119891 can be decomposed into

119891119891 = lang1198911198911198921198921205741205740rang1198921198921205741205740 + 119877119877119891119891 (8)

where 119877119877119891119891 is the residual function after approximating 119891119891 using 1198921198921205741205740 1198921198921205741205740 is orthogonal to

119877119877119891119891 hence

1198911198912 = lang1198911198911198921198921205741205740rang2

+ 1198771198771198911198912 (9)

To minimize 119877119877119891119891 1198921198921205741205740 is chosen to maximize lang1198911198911198921198921205741205740rang The matching pursuit method is

an iterative algorithm that decomposes the residue 119877119877119891119891 by projecting it on a function of 119863119863

that matches 119877119877119891119891 almost at best as it was done for 119891119891 The residue 119877119877119899119899119891119891 is decomposed into

119877119877119899119899119891119891 = lang119877119877119899119899119891119891119892119892120574120574119899119899rang119892119892120574120574119899119899 + 119877119877119899119899+1119891119891 (10)

- 13 -

the decomposition is carried up to the order 119898119898 119891119891 can be expressed as

119891119891 = lang119877119877119899119899119891119891119892119892120574120574119899119899rang119892119892120574120574119899119899

119898119898minus1

119899119899=0

+ 119877119877119898119898119891119891 (11)

The original function 119891119891 is decomposed into a sum of dictionary elements that are chosen to

best match its residues

In this study the matching pursuit method is utilized to accurately decompose of the over-

lapped Gaussian pulses and estimate the parameters of the pulse which used for damage

characterization The 119894119894119905119905ℎ Gaussian pulse in the dictionary for approximating the obtained

signal is expressed as

119892119892120574120574119894119894 = 119891119891(119860119860119894119894 119906119906119894119894 119904119904119894119894 119889119889119894119894120601120601119894119894) (12)

where 119860119860119894119894119906119906119894119894119904119904119894119894119889119889119894119894 and 120601120601119894119894 represent amplitude arrival time scale (pulse duration) excita-

tion(central) frequency and phase of the Gaussian pulse respectively The parameters of the

pulse are estimated using a nonlinear regression based optimization method (Gauss-Newton

method) based on an initial guess of the parameters Detail procedure for estimating the pa-

rameters of the pulse is expressed in Hong et al [23]

Among the estimated parameters of the pulses pulse duration (scale) represents the disper-

sion effect If a wave-packet is spread out the estimated pulse duration is increased Using

this feature the matching pursuit method is modified to capture waves which have longer

pulse duration than transmitted A0 Lamb wave mode If the estimate pulse duration of a wave

is less than the transmitted wave the modified matching pursuit method regards it as error

- 14 -

CHAPTER 4 NUMERICAL SIMULATION

41 Composite modeling method

Composite materials can be modeled from nano to macro level depends on its engineering

purpose In general four methods are utilized for modeling of composite materials First

composite is modeled as a homogenous material expressed in terms of effective material

properties [24] The effective material properties are estimated using the modulus and the

volume fraction of each constituent theoretically (Rule of mixture) and measured experimen-

tally Second a model is established that each laminate which have own material properties

are totally bonded together In this model the stacking sequence and direction of each lami-

nates are considered additionally [25-27] Third method for composite modeling in micro

level called micromechanics [28-30] This model considers the interaction between fibers and

matrix in laminate Last multiscale modeling which the scope of modeling is extended to

molecular or quantum level of constituents [31-33] This model starts from molecular me-

chanics or molecular dynamics then the scale of model leads to the nano micro and macro

level by incorporating small scale models As the scale of models is smaller it provides more

logical way than other large scale of models while the computational cost is more expensive

The scope of this study is focused on the dynamic behavior of composite structure the beam

or plate model was utilized for the numerical simulation

- 15 -

42 Damage modeling method

Generally three damage models have been developed and utilized in the real applications

First the stiffness reduction model called as the equivalent element model was introduced

[17 34-35] This model is simply realized by replacing the region around damage area by a

region with adjusted material properties (Youngrsquos modulus shear modulus density etc)

globally Next the duplicated node model is done by disconnecting the element connectivity

assuming there is no interaction among the delaminated surface This model can be utilized

when the location and the extension of damage is important issue [34-36] Last the contact

mechanics model using the principles of contact mechanics the two interfacial forces (con-

tact resistance and sliding resistance) in the delaminated surface [37-38] In this model the

delaminated surfaces are modeled as nonlinear functions of the interlaminar normal strain

and the interlaminar shear strain This model is computationally expensive while provides

more logical way than other models There is however no universally accepted model for

general impact-induced damage in composites In the case of impact-induced damage it is

impossible to predict or estimate the location and number of delaminated layers Thus from

the uncertainty of impact-induced damage the equivalent element model can be considered

as the effective damage model

- 16 -

43 Numerical simulation description

TABLE I Material property of simulation model and test specimen

Tensile Strength 657 MPa Tensile Modulus 59 GPa Shear Strength 804 MPa Shear Modulus 24 Gpa Specific Gravity 16

PZT A

PZT C

PZT BDamage

3

250 125 125

500 Unit mm

Fig 6 Description of the numerical simulation model

The feasibility of the proposed damage characterization technique is verified by 2-D FEM

simulation using MSC PATRAN and NASTRAN Fig 6 presents the dimension PZTs and

damage location of the simulation model In the model 05mm x 05mm square shell element

was used for modeling of the A0 mode whose flexural wave characteristic The element size

was determined by considering the wavelength of the propagation Lamb wave modes The

material properties of the simulation model and damage cases are presented in Table I and

Table II respectively

PZT A and C act as both actuator and sensor for collecting pulse-echo signal They were

installed on the plate which two of them are both-side collocated for the A0 Lamb wave

- 17 -

mode decomposition [39-40] PZT B was only used as a sensor for collecting pitch-catch sig-

nal For simulating the excitation and sensing of PZTs the pin-force model was adopted that

an input force apples to the end nodes of excitation PZT and averaging element strains corre-

sponding sensing area A Gaussian pulse (tone-burst) 50 kHz central frequency was excited

as an input signal

TABLE II Damage cases for numerical simulation

119949119949120783120783 (mm) 119941119941 (mm) G value reduction ()

Case 1 650 200 90 darr

Case 2 675 150 90 darr

Case 3 675 150 93 darr

Case 4 900 150 90 darr

- 18 -

44 Numerical simulation results

015 02 025 03 035

-2

0

2

Time (ms)

Am

plitu

de

IntactDamage

Reflected wave from damage

(a) Pitch-catch signal

015 02 025 03 035

-2

0

2

Time (ms)

Am

plitu

de

IntactDamage

Reflected waves from damage

(b) Pulse-echo signal Fig 7 Comparison of A0 Lamb wave mode signals from numerical simulation between intact and damage cases Time delay attenuation and multiple reflections due to damage simulated successfully using the damage modeling method described in the previous section

Fig 7 shows that comparison of A0 Lamb wave mode signals between intact and one of

damage cases in the simulation The features of A0 Lamb wave mode passing through dam-

age area described in section 2 are successfully simulated Thus the equivalent element

method is proper for modeling damage in composites In the damage cases the additional

waves reflected from damage is observed in both pitch-catch and pulse echo signal

Table III presents the result of the proposed damage characterization algorithm using the

numerical simulation cases shown in Table II In lsquoMethod 2rsquo 119907119907119894119894 variation is assumed to plusmn

5 of the normal value (1190~1315ms) The normal value of 119907119907119894119894 for lsquoMethod 3rsquo is meas-

ured as 1253 ms In each case the damage is characterized using the proposed technique

Note that as the information of 119907119907119894119894 is specified the estimated value is closer to the exact

value

- 19 -

TABLE III Damage characterization for the numerical simulation models

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

Case 1

Exact 650 200

Method 1 248 50~70 79~414 160~834

Method 2 248 603~667 135~241 271~486

Method 3 248 635 188 379

Case 2

Exact 675 150

Method 1 193 60~75 01~251 37~651

Method 2 193 618~682 114~222 295~575

Method 3 193 65 168 435

Case 3

Exact 675 150

Method 1 233 60~75 66~302 140~649

Method 2 233 668~738 85~196 182~419

Method 3 233 703 140 301

Case 4

Exact 900 150

Method 1 204 75~100 48~349 118~856

Method 2 204 860~951 107~216 264~530

Method 3 204 906 162 397

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1253 ms 1190~1315 ms Measured 119907119907119894119894 1253 ms

- 20 -

CHAPTER 5 EXPERIMENTAL VERIFICATION

51 Overview

In this section experimental study is investigated so that the proposed damage characteri-

zation technique Overall experimental setup and test specimen are shown in Fig 8 and 9

The experimental setup consists of test specimens data acquisition (DAQ) system and self-

sensing circuit Self-sensing circuit was installed for pulse-echo measurement using PZTs

[40-41]

PZT A(PZT C) PZT B

500

500

140

75

250

Teflon Inserted Damage (d =25)

Unit mm

Thickness = 3

PZT A(PZT C) PZT B

500

500

150

90

250

Impact Location

Unit mm

Thickness = 3

(a) Teflon inserted damage specimen

(b) Impact-induced damage specimen

Fig 8 Test specimens for the experimental verification In the Teflon inserted damage Teflon tapes with 25mm diameter were inserted at the center of the specimen The impact-induced damage was introduced to the specimen using the impact test equipment

- 21 -

511 Specimen

Two carbon composite plates with a dimension of 500times500times3mm3 were manufactured by

Nexcomes Inc by stacking 12 layers of CF3327 3K prepreg (woven fabrics type) manufac-

tured by Hankuk Fiber Inc The material properties of the specimen are presented in Table I

One of the specimens Teflon tapes with 25mm diameter was inserted in every 3 layers at the

center of the specimen to mimic the damage The other specimen is utilized for the character-

ization of real impact-induced damage

512 Data acquisition (DAQ) system and PZT

The data acquisition (DAQ) system consists of an arbitrary waveform generator (AWG) a

high speed signal digitizer (DIG) and two multiplexers (MUX) Using the 16-bits AWG

tone-burst input signal with a plusmn10 peak-to-peak voltage were generated The response signal

was measured by the 14-bit DIG For the signal-to-noise ratio reduction both pitch-catch and

pulse-echo signals were measured 150 times and averaged in the time domain 3 identical

circular PZTs of 10mm diameter manufactured by APC International Ltd were installed on

the specimen A Gaussian pulse (tone-burst) 50 kHz central frequency was excited as an in-

put signal The excitation frequency was chosen in order to avoid overlapping of the S0 and

A0 Lamb wave modes and generating of the higher modes (S1 A1hellip)

- 22 -

PZT Actuator

PZT Sensor

AWG

DIG

DAQ System Specimen

SensingActuatingMUX

1

23

4

1 Sending signal to MUX

4 Sending signal to DIG

Self-SensingCircuit

18 Unit mm

10

(a)

(b)

Fig 9 Overall experimental setup (a) Data acquisition (DAQ) system and (b) Installed circu-lar PZT transducer on the specimen with its dimension

513 Impact test

The impact-induced damage was introduced to the specimen using the impact test equip-

ment presented in Fig 10 A 5kg mass with 10mm tip was dropped from 015m height onto

the same location of the specimen The signals were obtained at each number of impacts

Since the reflected wave is not detected at the first impact case the data obtained from the

second and third impact were utilized for the analysis

10

Unit mm (a)

(b)

Fig 10 (a) Impact test equipment for introducing impact-induced damage and (b) Tip of the impact tester with 10mm diameter

- 23 -

514 Temperature effect test

In reality structures are subjected to changing environmental conditions such as tempera-

ture that affect measured signal and cause false-alarm Thus the robustness of the proposed

damage characterization technique under temperature variation in investigated (Fig 11) The

specimens were placed inside a temperature chamber and a thermocouple was installed on

the specimen to measure temperature of surface of the specimen The signal was obtained

under three temperature cases (0 20 and 50degC) and the humidity was kept at 30

Temperature Chamber

Test Specimen

Fig 11 Test specimen in temperature chamber

- 24 -

52 Teflon inserted damage

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Reflected wave from damage

(a)

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Transmitted WaveReflected Wave

Detected reflected wave

(b)

Fig 12 A0 Lamb wave mode obtained using pitch-catch method from the Teflon inserted damage specimen (a) The reflected wave from damage is shown (b) The Teflon inserted damage is detected by capturing the reflected wave

Fig 12 shows the obtained pitch-catch signal from the Teflon inserted damage specimen

and damage diagnosis result The Teflon inserted damage is detected successfully by captur-

ing the reflected wave Table IV presents the result of the Teflon inserted damage characteri-

zation 119907119907119894119894 is measured as 1213ms and the variation is assumed to plusmn 5 of the normal val-

ue (1152~1273ms) The results indicate that the proposed damage characterization approach

works well in real application

TABLE IV Damage characterization for the Teflon inserted damage specimen

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

Exact 635 25

Method 1 123 60~75 73~338 298~1220

Method 2 123 619~662 229~310 934~1188

Method 3 123 631 284 1161

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1213 ms

- 25 -

53 Impact-induced damage

11 115 12 125 13-4

-2

0

2

4

Time (ms)

Vol

tage

(mV

)

IntactImpact2Impact3

Reflected wave from damage

(a) Pitch-catch signal

11 115 12 125 13 135

-05

0

05

1

Time (ms)

Volta

ge (m

V)

IntactImpact2Impact3

Boundary reflection

Reflected wave from damage

(b) Pulse-echo signal

Fig 13 The signal obtained from the test specimen subjected to impact The impact was in-troduced at the same location of the specimen and the signal was obtained after each impacts As the number of impact is increased the transmitted A0 Lamb wave mode is delayed and attenuated

The obtained signals from impact-induced damage specimen are presented in Fig 13 It is

observed that as the number of impact is increased the transmitted A0 Lamb wave mode is

more delayed because the wave velocity in the damage area is reduced Fig 14 shows the

false-alarm test result whether the proposed damage detection algorithm works even in pris-

tine condition of the test specimen It is shown that no false alarm is occurred because no ad-

ditional reflected wave is captured

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Transmitted Wave

(a) Transmitted wave

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Reflected Wave

No reflected wave

(b) Reflected wave

Fig 14 False alarm test for the proposed damage detection algorithm in pristine condition of the specimen Reflected wave from damage is not captured false alarm is not occurred

- 26 -

TABLE V Damage characterization for the impact-induced damage specimen

of Impacts m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

2

(1470 J)

Method 1 080 65~90 75~246 470~1157

Method 2 080 801~885 114~245 711~1134

Method 3 080 843 179 1122

3

(2205 J)

Method 1 148 50~100 06~380 21~1134

Method 2 148 758~838 249~368 843~1125

Method 3 148 798 308 1044

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1213 ms

Table V presents the characterization results of the impact-induced damage case Here the

measured 119907119907119894119894 and the assumed variation are same as the artificial damage case It is observed

that as the number of impact is increased 119889119889 is increased while 119907119907119889119889 is reduced Thus it is

validated that the proposed damage characterization approach estimates the increment of the

damage well

54 Temperature effect test

541 Teflon inserted damage

11 115 12 125 13-10

-5

0

5

10

Time (ms)

Vol

tage

(mV

)

0 oC20 oC50 oC

(a) Pitch-catch signal

11 115 12 125 13 135

-1

0

1

2

3

Time(ms)

Volta

ge (m

V)

0 oC20 oC50 oC

(b) Pulse-echo signal

Fig 15 Signals obtained from the Teflon inserted damage specimen under temperature varia-tion Time delay and amplitude variation are occurred

Fig 15 shows the obtained signals from the Teflon inserted damage specimen subjected to

- 27 -

temperature variation Time delay and amplitude variations are occurred there is no addition-

al damage is introduced The analyzed results of the obtained signals are reported in Table VI

The measured 119907119907119894119894 is 1247 ms at 0 degC 1213 ms at 20 degC and 1168 ms at 50 degC respectively

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms (1152~1273 ms)

The estimated results are also changed under temperature variations the fluctuation of esti-

mated range and exact value is small

TABLE VI Temperature effect test for the Teflon inserted damage specimen

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

0 degC

Exact 635 25

Method 1 113 60~70 79~268 352~1291

Method 2 113 560~619 233~344 1034~1172

Method 3 113 606 257 1141

20 degC

Exact 635 25

Method 1 123 60~75 73~338 298~1220

Method 2 123 619~662 229~310 934~1188

Method 3 123 631 284 1161

50 degC

Exact 635 25

Method 1 137 60~75 72~328 263~1228

Method 2 137 622~687 183~299 666~1086

Method 3 137 642 264 960

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1247 ms at 0 degC 1213 ms at 20 degC and 1168 ms at 50 degC

- 28 -

542 Impact-induced damage

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Vol

tage

(mV

)

Transmitted WaveReflected Wave

No reflected wave

(a) 0 degC

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Transmitted WaveReflected Wave

No reflected wave

(b) 50 degC

Fig 16 False alarm test under temperature variations It is shown that proposed damage char-acterization algorithm is robust under temperature variations

The false alarm test in pristine condition of the impact-induced damage specimen under

temperature variation was investigated (Fig 16) It is validated that the proposed damage di-

agnosis algorithm is robust under temperature variations Fig 17 shows the obtained signals

from the impact-induced damage (impacted 3 times) specimen under temperature variation

The analyzed results of the obtained signals are reported in Table VII The measured 119907119907119894119894 in

each temperature and the assumed variation are same as the artificial damage case

11 115 12 125 13-5

0

5

Time (ms)

Volta

ge (m

V)

0 oC20 oC50 oC

(a) Pitch-catch signal

11 115 12 125 13 135

-1

0

1

2

3

Time(ms)

Vol

tage

(mV

)

0 oC20 oC50 oC

(b) Pulse-echo signal

Fig 17 Signals obtained from impact-induced damage specimen (Impact 3) under tempera-ture variation

- 29 -

TABLE VII Temperature effect test for the impact induced damage specimen (Impact 3 case)

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

0 degC

Method 1 135 75~90 72~310 267~1148

Method 2 135 705~779 264~381 978~1126

Method 3 135 763 289 1071

20 degC

Method 1 148 50~100 06~380 21~1134

Method 2 148 758~838 249~368 843~1125

Method 3 148 798 308 1044

50 degC

Method 1 153 75~100 53~415 174~1227

Method 2 153 804~889 214~336 702~1103

Method 3 153 815 320 1050

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1247 ms at 0 degC 1213 ms at 20 degC and 1168 ms at 50 degC

- 30 -

CHAPTER 6 CONCLUDING REMARKS

61 Executive summary

In this study the reference-free damage detection localization and quantification tech-

nique in a multi layered composite plate is proposed Additional reflected A0 Lamb wave

mode from the damage due to multiple reflections in damage area is used as a feature for

characterizing the damage First damage is characterized in terms of a damage index (m-

value) defined as the ratio of the physical length of the damage to the wavelength of the A0

Lamb wave mode Then the range of possible damage location length and severity are esti-

mated with constraints This procedure can be accomplished without prior baseline data To

realize the proposed technique the matching pursuit technique is modified for reduction of

the error by considering the dispersive characteristic of Lamb wave The matching pursuit

method estimates the arrival time of each wave packets accurately even two wave packets are

overlapped The numerical simulations and experimental tests including Teflon tape inserted

and impact-induced damage are investigated for the feasibility of the proposed technique

Moreover the effectiveness of the proposed technique under changing temperature is exper-

imentally verified

- 31 -

62 Suggestion for further study

The limitations of this study and the possible improvement and further studies are listed as

follow

1 Measurement of impact-induced damage length using conventional NDT technique

Additional experimental verification for impact-induced damage characterization is

needed by measuring the real length of damage using conventional NDT technique

such as ultrasonic scan or infrared camera

2 Limitation of both side collocated PZT installation In this study excitation PZTs are

installed both side and collocated for decomposing A0 Lamb wave mode However it

can produce decomposition error due to imperfection of PZT location and their installa-

tion Moreover in many cases assess to both surface of structures is not allowed such

as aircraft applications To improve this limitation a reference-free damage characteri-

zation using single side PZTs will be considered

3 Complex structure application In this study damage is characterized in simple plate-

like structure When the monitored structure becomes complex the response signals al-

so becomes complex due to its geometry and boundary Thus the investigations for

complex structure application such as stiffener spar and curvature are needed

- 32 -

REFERENCES

[1] Cawley P and Alleyne D (1996) ldquoThe use of Lamb waves for the long range inspection of large structuresrdquo Ultrasonics 34 pp287-290

[2] Ghosh T Kundu T and Karpur P (1998) ldquoEfficient use of Lamb waves for detecting defects in large platesrdquo Ultrasonics 36 pp791-801

[3] Guirguitui V (2003) ldquoLamb wave generation with piezoelectric wafer active sensors for structural health monitoringrdquo SPIE Conference on Smart Structures and Materials San Diego CA 5391 pp111-422

[4] Ragvahan A and Cesnik CE S (2004) ldquoModeling of piezoelectric-based Lamb wave generation and sensing for structural health monitoringrdquo SPIE Conference on Smart Struc-tures and Materials San Diego CA 5391 pp419-430

[5] Rose JL Pilarski A and Dirit JJ (1993) ldquoAn approach to guided wave mode selection for inspection of laminated platerdquo Journal of Reinforced Plastics and Composites 12(5) pp536-544

[6] Guo N and Cawley P (1993) ldquoThe interaction of Lamb waves with delaminations in composite laminatesrdquo Journal of Acoustical Society of America 94(4) pp 2240-2246

[7] Valdez SHD and Soutis C (2002) ldquoReal-time nondestructive evaluation of fiber com-posite laminates using low-frequency Lamb wavesrdquo Journal of Acoustical Society of Amer-ica 111(5) pp 2026-2033

[8] Yang M and Qiao P (2005) ldquoModeling and experimental detection of damage in virous materials using the pulse-echo method and piezoelectric sensoractuatorsrdquo Smart Materials and Structures 14 pp1083-1100

[9] Lestari W and Qiao P (2005) ldquoApplication of wave propagation analysis for damage identification in composite laminated beamsrdquo Journal of Composite Materials 39(22) pp1967-1984

[10] Kessler S S Spearing S M and Soutis C (2002) ldquoDamage detection in composite ma-terials using Lamb wave methodsrdquo Smart Materials and Structures 11 pp269-278

[11] Sohn H Park G Wail J R Limback N P and Farrar C R (2004) ldquoWavelet-based active sensing for delamination detection in composite structuresrdquo Smart Materials and Structures 13 pp 153-160

[12] Petculescu G Krishnaswamy S and Achenback J D (2008) ldquoGroup delay measure-ments using modally selective Lamb wave transducers for detection and sizing of delamina-tions in compositesrdquo Smart Materials and Structures 17

[13] Sohn H Park H W Law K H and Farrar C R (2007) ldquoDamage detection in compo-

- 33 -

site plates by using an enhanced time reversal methodrdquo Journal of Aerospace Engineering (ASCE) pp 141-151

[14] Sohn H (2007) ldquoEffects of environmental and operational variability on structural health monitoringrdquo Philosophical Transactions of the Royal Society Ardquo 365 pp539-560

[15] Rytter A (1993) ldquoVibration based inspection of civil engineering structuresrdquo PhD The-sis Department of Building Technology and Structural Engineering University of Aalborg Aalborg Denmark

[16] Hayashi T and Kawashima K (2002) ldquoMultiple reflections of Lamb waves at a delami-nationrdquo Ultrasonics 40 pp193-197

[17] Wang C H and Rose L R F (2003) ldquoWave reflection and transmission in beams con-taining delamination and inhomogeneityrdquo Journal of Sound and Vibration 264 pp 851-872

[18] Ramadas C Balasubramaniam K Joshi M and Krishnamurthy C V (2009) ldquoInterac-tion of the primary anti-symmetric Lamb mode (Ao) with symmetric delaminations numer-ical and experimental studiesrdquo Smart Materials and Structures 18

[19] Ramadas C Balasubramaniam K Joshi M and Krishnamurthy C V (2010) ldquoInterac-tion of guided Lamb waves with an asymmetrically located delamination in a laminated composite platerdquo Smart Materials and Structures 19

[20] Wilcox P Lowe M and Cawely P (2001) ldquoThe effect of dispersion on long-range in-spection using ultrasonic guided wavesrdquo NDTampE International 34 pp1-9

[21] Mallat S G and Zhang Z (1993) ldquoMatching pursuits with time-frequency dictionariesrdquo IEEE Tansactions on Signal Processing 41(12) pp3397-3415

[22] Zhang G Zhang S and Wang Y (2000) ldquoApplication of adaptive time-frequency de-composition in ultrasonic NDE of highly scattering materialsrdquo Ultrasonics 38 pp 961-964

[23] Hong J C Sun K H and Kim Y Y (2005) ldquoThe matching pursuit approach based on the modulated Gaussian pulse for efficient guided-wave damage inspectionrdquo Smart Mate-rials and Structures 14 pp548-560

[24] Herakovich C T (1997) Mechanics of Fibrous Composites Wiley 480 pages

[25] Reddy J N (1985) ldquoA review of the literature on finite-element modeling of laminated composite platesrdquo The Shock and Vibration Digest 17(4) pp 3-8

[26] Reddy J N (2003) Mechanics of Laminated Composite Plates and Shells Theory and Analysis CRC Press 856 pages

[27] Palacz M Krawczuk M and Ostachowicz W (2005) ldquoThe spectral finite element mod-el for analysis of flexure-shear coupled wave propagation Part 1 laminated multilayer

- 34 -

composite beamrdquo Composite Structures 68 pp37-44

[28] Zhang Y C and Harding J (1990) ldquoA numerical micromechanics analysis of the me-chanical properties of a plain weave compositerdquo Computer amp Structures 36(5) pp 839-844

[29] Nemat-Nasser S and Hori M (1999) Micromechanics Overall Properties of Heteroge-neous Materials Elsevier 810 pages

[30] Scida D Aboura Z Benzeggagh M L and Bocherens E (1999) ldquoA micromechanics model for 3D elasticity and failure of woven-fibre composite materialrdquo Composites Science and Technology 59 pp 505-517

[31] Thomas Y and Wu X H (1997) ldquoA multiscale finite element method for elliptic prob-lems in composite materials and porous mediardquo Journal of Computational Physics 134 pp 169-189

[32] Lee K H Moorthy S and Ghosh S (1999) ldquoMultiple scale computational model for damage in composite materialsrdquo Computer Methods in Applied Mechanics and Engineer-ing 172 pp 175-201

[33] Namilae S and Chandra N (2005) ldquoMultiscale model to study the effect of interfaces in carbon nanotube-based compositesrdquo Journal of Engineering Materials and Technology (ASME) 127 pp222-232

[34] Adams D E (2007) Health monitoring of Structural Materials and Components Wiley 460 pages

[35] Gopalakrishnan S (2009) ldquoModeling Aspects in Finite Elementrdquo Encyclopedia of Struc-tural Health Monitoring 2 pp791-809

[36] Palacz M Krawczuk M and Ostachowicz W (2005) ldquoThe spectral finite element mod-el for analysis of flexure-shear coupled wave propagation Part 2 Deminated multilayer composite beamrdquo Composite Structures 68 pp45-51

[37] Baik J M and Thopson R B (1984) ldquoUltrasonic scattering from imperfect interfaces A quasi-static modelrdquo Journal of Nondestructive Evaluation 4(3) pp177-196

[38] Luo H and Hanagud S (2000) ldquoDynamics of delamincated beamsrdquo International Jour-nal of Solids and Structures 37 pp1501-1519

[39] Kim S B and Sohn H (2007) ldquoInstantaneous reference-free crack detection based on polarization characteristics of piezoelectric materialsrdquo Smart Materials and Structures 16 pp2375-2387

[40] An Y K and Sohn H (2010) ldquoInstantaneous crack detection under varying temperature and static loading conditionsrdquo The Special Issue of Structural Control and Health Monitor-ing 17(7) pp 730-741

- 35 -

[41] Lee S J and Sohn H (2006) ldquoActive self-sensing scheme development for structural health monitoringrdquo Smart Materials and Structures 15(6) pp 1734-1746

- 36 -

요 약 문

복합재료 손상의 무기저 진단 및 정량화

복합재료는 재료의 경량성 내구성 고강도의 특성으로 인하여 항공기에서부터

교량에까지 널리 사용되고 있다 하지만 외부의 충격으로 인한 복합재료 내부의

손상은 재료의 강도를 감소시켜 구조물의 안정성을 저하시키는 요인이 된다 하지만

복합재료의 손상은 재료 내부에서 발생하여 육안으로는 식별하기도 힘들며 표면에서

보이는 것만으로 내부의 손상 정도를 판단하기는 불가능하다는 단점을 가지고 있다

본 연구에서는 손상 내부에서 반사로 인해 발생하는 파를 이용하여 기저자료를

사용하지 않고 특정 시점에서 얻은 자료만으로 복합재료의 손상을 진단하고 정량화

할 수 있는 기법을 제안한다 제안된 기법의 적용성을 검증하기 위하여 유한요소법을

이용한 수치해석을 수행하였고 Teflon tape 이 삽입된 시편과 실제 충격하중으로

인한 손상을 입은 시편을 이용한 실험을 진행하였다 수치해석 및 실험 결과 제안된

기법이 복합재료의 손상을 성공적으로 진단 및 정량화할 수 있음을 확인하였다 또한

다양한 외부 온도 환경에서 제안된 기법의 환경영향 실험을 수행하였고 다양한

환경에서 손상의 진단 및 정량화가 성공적으로 이루어짐을 검증하였다

핵심어 무기저 손상진단 및 정량화 압전센서 복합재료 유한요소법 Matching pursuit

method

- 37 -

감 사 의 글

기대와 두려움을 안고 시작한 석사과정이 어느덧 2 년 이라는 세월 속에서 추억으로 자리잡아

가고 있습니다 학부생 기간을 포함한 지난 3 년 동안 SHM 이라는 새로운 분야를 접하고 연구를

수행 할 수 있도록 도와주신 고마운 분들이 있어 그 감사를 표하고자 합니다

우선 저의 많은 부족한 점을 일깨워 주시고 연구에 대한 뜻을 더욱 확고히 세울 수 있게 해

주신 지도교수님이신 손 훈 교수님께 깊은 감사의 뜻을 전합니다 짧은 기간이었지만 교수님의

도움으로 많은 것을 배우고 느끼고 깨달을 수 있었습니다 또한 본 연구를 수행할 수 있도록 지

원해주신 국토해양부 (U-City 석 박사과정 지원사업)에 깊은 감사 드립니다

때로 힘들고 지칠 때 항상 저를 믿고 응원해 주시는 가족이 있었기에 지금의 글을 쓸 수 있었

습니다 또한 지금 연구실에서 함께 생활하고 있는 윤규형 민구형 철민이형 현석이형 진열이형

호민이형 병진이 선혜누나와 성대에 계신 창길이형 미국에 계신 의재형과 Debaditiya 에게도 감

사의 말씀을 전합니다 그리고 모두 나열할 수는 없지만 카이스트 생활의 활력소가 되어준 카이

스트 건설 및 환경공학과 선후배들 직원 분들과 SOS 멤버들에게도 감사를 표합니다

앞서 거론한 많은 이들의 도움이 없었다면 저 또한 지금 이 자리에 있을 수 없었을 것입니다

많이 부족하기에 더 노력하고 배우고자 했던 초심을 다시 한번 생각하며 더 큰 결실을 맺기 위해

앞으로 나아가겠습니다 도움을 주신 모든 분들께 이 감사의 글을 바칩니다

2011년 6월 카이스트에서 임형진 올림

- 38 -

CURRICULUM VITAE

Hyung Jin Lim

Research Assistant

The Civil and Environmental Engineering Department

Korea Advanced Institute of Science and Technology (KAIST)

291 Daehak-Ro Guseong-Dong Yuseong-Gu

Daejeon Republic of Korea 305-701

Phone (82)+42-350-3665 Fax (82)+42-350-3610

Email limhj87gmailcom

RESEARCH INTERESTS

Damage Detection Localization and Quantification on Composite and Metallic Structures Pattern Recognition Finite Element Modeling (FEM)

EDUCATION

2009 ndash 2011 MS Civil and Environmental Eng Korea Advanced Institute of Science and Technology (KAIST) Korea Rep

2006 ndash 2009 BS Civil and Environmental Eng Korea Advanced Institute of Science and Technology (KAIST) Korea Rep

JOURNAL PUBLICATIONS The corresponding author is underlined 1 Hyung Jin Lim Hoon Sohn ldquoReference-free damage detection localization and quantification

using in compositerdquo in preparation for Journal of the Acoustical Society of America 2011

2 Chul Min Yeum Hoon Sohn Jeong Bum Ihn Hyung Jin Lim ldquoInstantaneous delamination detection in a composite using a dual piezoelectric transducer networkrdquo Submitted to Compo-sites Science and Technology 2011

3 Yun-Kyu An Hyung Jin Lim Min Koo Kim Jin Yeol Yang Hoon Sohn Chang Guen Lee ldquoInvestigation of applicability of local reference-free damage detection techniques to in-situ

- 39 -

bridgesrdquo Submitted to ASCE Journal of Structural Engineering 2011

4 Hyung Jin Lim Min Koo Kim Hoon Sohn Chan Yik Park ldquoImpedance based damage detec-tion under varying temperature and loading conditionsrdquo Submitted to NDTampE International 2011

5 Hyung Jin Lim Hoon Sohn ldquoQuantitative estimation of transmitted and reflected lamb waves at discontinuityrdquo Journal of the Korean Society for Nondestructive Testing Vol30 No 4 pp1-8 2010

CONFERENCE PROCEEDINGS

The corresponding author is underlined 1 Hyung Jin Lim and Hoon Sohn ldquoReference-free damage detection localization and quantifi-

cation in compositesrdquo ASCE Engineering Mechanics Institute Conference Boston June 2-4 2011

2 Hyung Jin Lim and Hoon Sohn ldquoReference-free delamination quantification within multi-layer composite materialsrdquo KSNT Spring Conference May 26-27 Gyeong-Ju Korea 2011

3 Min Koo Kim Hyun Jin Lim Hoon Sohn Chan Yik Park ldquoImpedance-based bolt loosening detection under varying temperature and loadingrdquo Asian Pacific Workshop on Structural Health Monitoring Tokyo Japan November 30-December 2 2010

4 Hyung Jin Lim Chul Min Yeum and Hoon Sohn ldquoModeling of Impact-induced Delamination in a Multilayer Composite Platerdquo The 23rd KKCNN Symposium on Civil Engineering Taipei (Taiwan) November 13-15 2010

5 Yun-Kyu An Hyung Jin Lim and Hoon Sohn ldquoTemperature independent Delamination Detec-tion using data normalizationrdquo The 3rd International Conference on Multi-Functional Materi-als and Structures Chonbuk National Univ Jeonju Korea September 14-18 2010

6 Hoon Sohn Yun-Kyu An Chang Gil Lee Min Koo Kim Hyung Jin Lim ldquoApplication of online NDT techniques to decommissioned bridge testingrdquo the 5th European Workshop on Structural Health Monitoring Sorrento Italy June 29-July 02 2010

7 Min Koo Kim Hyung Jin Lim Hoon Sohn Chan Yik Park ldquoImpedance based damage diag-nosis of a complex composite aircraft wing under changing loading and temperature condi-tionsrdquo the 5th European Workshop on Structural Health Monitoring Sorrento Italy June 29-July 02 2010

8 Yun-Kyu An Min Koo Kim Hyung Jin Lim and Hoon Sohn ldquoValidations of reference-free crack detection techniques through a decommissioned bridge testrdquo 2010 COSEIK Annual Con-ference Jeju Korea April 8-9 2010

- 40 -

PARTICIPANT PROJECTS

1 Development of Smart Steel Members for Online Steel Structure Monitoring POSCO Corporation (Funded 41000000 KRW (41000 USD) for 010511 to 123111)

2 Development of Enabling Technologies for Structural Health Monitoring Korean Agency for Defense Development (Funded 240000000 KRW (240000 USD) out of total 3200000000 KRW (3200000 USD) for 060108-053111)

3 Development of On-board SHM Technologies for Composite Air Vehicles The Boeing Company (Funded 400000 USD (400000000 KRW) for 080108 to 073111)

EDUCATIONAL EXPERIENCES

Fall 2009

Teaching Assistant CE207 ldquoElementary Structural Engineering and Laboratoryrdquo KAIST 3 unit undergraduate course

EXPERIMENTAL EXPERIENTCES

2010 Monitoring of Real Aircraft Composite Wing Segment Korea Agency for Defense De-velopment (ADD) Daejeon Korea Real time monitoring of composite air vehicles fund-ed by Korea Agency for Defense Development (ADD)

2009-2011

Field Monitoring of Real Bridge Structures Gimpo (Ramp G bridge) amp Icheon (Sam-seung bridge) Korea US-Korea collaborative research for bridge monitoring testbeds

HONORS amp AWARDS

Scholarships for outstanding students granted KAIST (2007-2009) The three prominent students in the civil and environmental engineering department of KAIST were selected as recipients This scholarship provided 2000$ per year for three years

- 41 -

Page 5: 복합재료 손상의 무기저 진단 및 정량화 - KAISTssslab.kaist.ac.kr/article/pdf/Thesis_2.pdf · 1 Declaration of Ethical Conduct in Research: I, as a graduate student

MCE

20094258

임 형 진 Lim Hyung Jin Reference-free Damage Detection Localization and

Quantification in Composite 복합재료 손상의 무기저 진단 및 정량화 Department

of Civil and Environmental Engineering 2011 41 p Advisor Prof Sohn Hoon

ABSTRACT

A reference-free damage detection localization and quantification technique in a multi

layered composite plate is proposed It is reported that when A0 Lamb wave mode propa-

gates the damage time delay attenuation and multiple reflections are occurred in the dam-

age area In this thesis additional reflected waves from the damage due to multiple reflec-

tions are used as a feature for characterizing the damage First the damage is identified by

detecting the additional reflected waves from the damage area Then the location of the

damage is estimated using combined pitch-catch and pulse-echo methods Finally the dam-

age is quantified using a damage index defined as the ratio between the physical length of

the damage and the wavelength of the A0 Lamb wave mode The matching pursuit method

is utilized for analyzing the obtained signal by decomposing overlapped waves and estimat-

ing the parameters of each wave such as arrival time driving frequency amplitude and so

on The proposed technique has an advantage that the damage can be characterized without

reference data obtained from pristine condition of the monitored structure within a single

path The feasibility of the proposed technique is verified through numerical simulations

and experimental tests In numerical simulation different damage models (length material

properties and location) are investigated In experimental tests a Teflon tape inserted and

impact-induced damage are characterized under temperature variations The results are

suggesting a reliable damage characterization in composites

Keywords Reference-free PZT transducers Composites Finite element method Matching

pursuit method

i

TABLE OF CONTENTS

ABSTRACT middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot i

TABLE OF CONTENTS middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot iii

LIST OF TABLES middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot iv

LIST OF FIGURESmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot v

CHAPTER 1 INTRODUCTION

11 Motivation middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 1

12 Literature review middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 1

13 Objective and uniqueness middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 3

14 Organization middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 4

CHAPTER 2 INTERACTION OF LAMB WAVE WITH DAMAGE

21 A0 Lamb wave mode delay and attenuation in the time domain middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 5

22 Multiple reflections in the damage area middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 6

CHAPTER 3 THEORETICAL DEVELOPMENT

31 Overview middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 7

32 Damage index (m-value) middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 8

33 Damage detection algorithm (Level 1) middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 10

34 Damage quantification algorithm (Level 3) middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 11

35 Damage localization algorithm (Level 2amp3) middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 11

36 Matching pursuit method considering the dispersion effect of Lamb wave

middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 13

CHAPTER 4 NUMERICAL SIMULATION

41 Composite modeling methodmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 15

42 Damage modeling method middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 16

43 Numerical simulation description middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 17

44 Numerical simulation results middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 19

ii

CHAPTER 5 EXPERIMENTAL VERIFICATION

51 Overview middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 21

511 Specimen middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 22

512 Data acquisition system and PZT middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 22

513 Impact test middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 23

514 Temperature effect test middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 24

52 Teflon inserted damage middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 25

53 Impact-induced damagemiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 26

54 Temperature effect test middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 27

541 Teflon inserted damage middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 27

542 Impact-induced damage middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 29

CHAPTER 6 CONCLUDING REMARKS

61 Executive summary middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 31

62 Suggestion for further study middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 32

REFERCENCES middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 33

SUMMARY (IN KOREAN) middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 37

ACKNOWLEDGEMENTS (IN KOREAN)

CURRICULUM VITAE

iii

LIST OF TABLES

I Material property of simulation model and test specimen middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 17

II Damage cases for numerical simulation middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 18

III Damage characterization for the numerical simulation models middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 20

IV Damage characterization for the Teflon inserted damage specimen middotmiddotmiddotmiddotmiddotmiddotmiddot 25

V Damage characterization for the impact-induced damage specimen middotmiddotmiddotmiddotmiddotmiddotmiddot 27

VI Temperature effect test for the Teflon inserted damage specimen middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 28

VII Temperature effect test for the impact induced damage specimen middotmiddotmiddotmiddotmiddotmiddotmiddot 30

iv

LIST OF FIGURES

1 Lamb wave propagation characteristics (a) A0 Lamb wave mode time delay and

attenuation after passing through a damage (b) The time delay and amplitude varia-

tion of A0 Lamb wave mode under temperature variation in pristine condition of

structure Because of this conventional damage detection techniques in composites

have limitations for real structure application middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 5

2 Multiple reflections in the damage area The multiple reflections of A0 Lamb

waves in the damage area are utilize as a feature for characterization of the damage

in this study middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 6

3 Parameters and notations of the proposed damage characterization technique

Damage is detected by identifying the additional reflected wave from the damage due

to multiple reflections (dashed lines) Localization and quantification of the damage

is carried out by measuring the arrival time of transmitted and reflected waves from

the damage middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 7

4 Two different FEM models which have different 119889119889 and 119907119907119889119889 (or G) This repre-

sents that the damage cannot be characterized uniquely based on the measured data

available middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 9

5 (a) Example of the overlapped signal with comparing approximated signal using

the matching pursuit method (b) Decomposed signal using the matching pursuit

method The matching pursuit method provides a conventional way to decompose

and identify each wave middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 9

6 Description of the numerical simulation model middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 17

7 Comparison of A0 Lamb wave mode signals from numerical simulation be-

tween intact and damage cases Time delay attenuation and multiple reflections due

to damage simulated successfully using the damage modeling method described in

v

the previous section middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 19

8 Test specimens for the experimental verification In the Teflon inserted damage

Teflon tapes with 25mm diameter were inserted at the center of the specimen The

impact-induced damage was introduced to the specimen using the impact test equip-

ment middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 21

9 Overall experimental setup (a) Data acquisition (DAQ) system and (b) Installed

circular PZT transducer on the specimen with its dimension middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 23

10 (a) Impact test equipment for introducing impact-induced damage and (b) Tip

of the impact tester with 10mm diametermiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 23

11 Test specimen in temperature chamber middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 24

12 A0 Lamb wave mode obtained using pitch-catch method from the Teflon in-

serted damage specimen (a) The reflected wave from damage is shown (b) The Tef-

lon inserted damage is detected by capturing the reflected wave middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 25

13 The signal obtained from the test specimen subjected to impact The impact

was introduced at the same location of the specimen and the signal was obtained after

each impacts As the number of impact is increased the transmitted A0 Lamb wave

mode is delayed and attenuated middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 26

14 False alarm test for the proposed damage detection algorithm in pristine condi-

tion of the specimen Reflected wave from damage is not captured false alarm is not

occurred middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 26

15 Signals obtained from the Teflon inserted damage specimen under temperature

variation Time delay and amplitude variation are occurred middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 27

16 False alarm test under temperature variations It is shown that proposed dam-

vi

age characterization algorithm is robust under temperature variationsmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 29

17 Signals obtained from impact-induced damage specimen (Impact 3) under

temperature variation middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 29

vii

CHAPTER 1 INTRODUCTION

11 Motivation

Composite materials has been increasingly demands in modern structures ranging from the

latest aircraft to retrofitted bridge deck due to its light weight superior specific strength and

stiffness properties However the occurrence of damage due to abrupt impact or sequential

accumulation of defect may considerably degrade structural strength or stiffness and reduce

the safety and reliability of the system Recently Lamb waves based damage detection tech-

niques using piezoelectric (PZT) transducers have emerged as one of the promising structural

health monitoring (SHM) technology for detecting hidden damage because of the possibility

of long range inspection and high sensitivity [1-4]

12 Literature review

This section will discuss researches that have been previously carried out on topics of

guided wave based SHM techniques using PZT transducers for damage in composite materi-

als The early investigations of interaction between Lamb wave and damage in composite can

be found in Rose et al and Guo et al [5-6] From further studies reflected signals from dam-

age were shown to be an effective method for damage diagnosis and localization using PZTs

[7-9] The reflected signals from damage were collected using one PZT as an actuator and the

other PZT as a sensor However the techniques cannot be used to determine the extent of

- 1 -

damage with reasonable precision

Kesseler et al proposed a technique monitoring the wavelet coefficients of transmitted ul-

trasonic signal amplitude as a potential way to damage detection [10] One drawback of this

technique is that the variability in coupling and the sensitivity of installed PZT transducers

can induce changes in the amplitude and energy of the signal that are cause false-alarm Oth-

er wavelet based delamination detection and localization technique developed using PZT

sensor network and outlier analysis [11]

Damage detection and quantification in composite plate was also investigated by group de-

lay measurement due to damage [12] However it is assumed that the group velocity in dam-

age area as the group velocity propagates half thickness of the structure in dispersion curve in

the quantification procedure Thus the damage quantification was investigated only Teflon

inserted damage case

Sohn et al proposed a damage detection technique which does not rely on any past base-

line signals to assess damage in composite plates by using an enhanced time reversal method

[13] The damage is assessed using the violation of time reversibility when nonlinearity is

caused by the damage along a direct wave path Thus it allows instantaneous identification

of delamination without requiring the baseline signal for comparison

- 2 -

13 Objective and uniqueness

In this study a reference-free damage characterization technique including detection lo-

calization and quantification of impact-induced damage in composite materials is developed

The general damage characterization can be categorized into the following levels [14]

Level 0 Define damage

Level 1 Determine the presence of damage within the structure

Level 2 Locate the regions of damage

Level 3 Quantify the damage

Level 4 Predict the remaining service life of the structure

Through the proposed technique from level 1 to level 3 can be achieved

Most of common approaches of damage characterization techniques described in the pre-

vious section use prior obtained baseline data However these approaches can cause false

alarm due to the effect of environmental variations such as temperature and ambient vibra-

tions [15]

The uniqueness of this study is

1 Damage is characterized without using prior measured baseline data

2 Robustness of the proposed technique under temperature variation is investigated

3 Additional waves reflected from inside damage have been used in addition to transmit-

ted and reflected waves

- 3 -

14 Organization

The organization of this thesis is summarized as follows

Chaper1 A general introduction including motivation literature review objective and

uniqueness of this study is described

Chaper2 Interactions between Lamb wave mode and damage in composites such as time

delay attenuation and multiple reflections are summarized Damage index (m-value) for

damage characterization is defined

Chaper3 Algorithms for the reference-free damage characterization technique using m-

value are developed Modified matching pursuit for this study is also described

Chaper4 Verification of the algorithm using numerical simulation is presented Various

modeled damage cases are characterized

Chaper5 Experimental verifications are described Two damage cases (Teflon inserted and

impact-induced damage) are investigated for the feasibility of the proposed damage charac-

terization technique Temperature effect test using temperature chamber is also carried out

Chaper6 Conclusion and summary of the present study is reported In advance suggestions

and future work recommendations are represented

- 4 -

CHAPTER 2 INTERACTION OF LAMB WAVE WITH DAMAGE

21 A0 Lamb wave mode delay and attenuation in the time domain

Time delay and attenuation of the A0 mode Lamb wave after passing through the damage

area due to effective thickness reduction were reported in previous researches [812] Moreo-

ver the attenuation of the A0 Lamb wave mode is occurred by degrading of shear modulus of

in the damage area However the time delay and amplitude variation of the A0 Lamb wave

mode also occurred even when the structure is health condition by the material property

changing due to temperature variation of the structure (Fig 1) Thus conventional damage

characterization methods using the time delay and attenuation of A0 Lamb wave mode as a

feature can cause false-alarm in real application

11 115 12 125 13-10

-5

0

5

10

Time (ms)

Vol

tage

(mV

)

IntactDamage

Reflected wave from Damage

(a)

11 115 12 125 13-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

0 oC20 oC50 oC

(b)

Fig 1 Lamb wave propagation characteristics (a) A0 Lamb wave mode time delay and atten-uation after passing through a damage (b) The time delay and amplitude variation of A0 Lamb wave mode under temperature variation in pristine condition of structure Because of this conventional damage detection techniques in composites have limitations for real struc-ture application

- 5 -

22 Multiple reflections in the damage area

Standing wave patterns due to multiple reflections of A0 Lamb wave mode in the damage

area were observed through numerical simulations and ultrasonic wave-field imaging using a

scanning laser vibrometer [16] The formation of standing waves in damage is illustrated

through Fig 2 When the incident waves enter the damage area a significant portion of the

waves are reflected back from the exit Then the reflected waves travel through the damage

and undergo reflection again at the original entrance As a consequence of multiple reflec-

tions at the entrance and the exit a considerable amount of ultrasonic energy is trapped inside

the damage area [17-19]

Multiple Reflections

Incident Wave Transmitted Wave

Damage

Fig 2 Multiple reflections in the damage area The multiple reflections of A0 Lamb waves in the damage area are utilize as a feature for characterization of the damage in this study

- 6 -

CHAPTER 3 THEORETICAL DEVELOPMENT

31 Overview

Damage

l = l1 + d + l2

d = vdtd l1 = vit1 l2 = vit2

t = t1 + td + t2

Excitation and Sensing PZT Sensing PZT

l

Transmitted Wave

Reflected Waves

Incident Wave

Fig 3 Parameters and notations of the proposed damage characterization technique Damage is detected by identifying the additional reflected wave from the damage due to multiple re-flections (dashed lines) Localization and quantification of the damage is carried out by measuring the arrival time of transmitted and reflected waves from the damage

In this section the reference-free damage characterization algorithm is proposed Damage

diagnosis and quantification are achieved by identifying the additional reflected waves due to

multiple reflections in the damage area (dashed lines in Fig 3) Moreover damage is local-

ized using additional information from pulse-echo method implement For the localization

and the quantification of damage the variables such as the path length and A0 Lamb wave

mode velocity in damage area are defined (Fig 3) For example 119897119897 is expressed as

119897119897 = 1198971198971 + 119889119889 + 1198971198972 (1)

where 1198971198971 119889119889 and 1198971198972 are the distance from the excitation PZTs to the entrance of damage the

damage length and the distance from the exit of damage to the sensing PZT respectively

- 7 -

Moreover 119907119907119894119894 and 119907119907119889119889 are also defined as the A0 Lamb wave mode velocity in pristine area

and damage area respectively Note that the whole procedure of the damage characterization

is accomplished based solely on the signals obtained from the current state of the structure

32 Damage index (m-value)

The proposed reference-free damage characterization technique utilizes the waves reflect-

ed additionally from the damage area as a feature The arrival time difference of the transmit-

ted and reflected A0 Lamb wave modes reflected from the exit and the entrance of the delam-

ination area one times each (∆119905119905) contains the information of the physical length of the dam-

age (119889119889) and A0 Lamb wave mode velocity in the damage area (119907119907119889119889) The wave velocity in the

damage area which is related to the shear modulus (G) represents the severity of the damage

∆119905119905 = 2119905119905119889119889 =2119889119889119907119907119889119889

(2)

However based on the measured data available the 119889119889 and 119907119907119889119889 (or G) cannot be uniquely

determined Fig 4 shows two FEM models which have different physical length and wave

velocity (G) of the damage Detail explanation of delamination modeling technique will be

covered in section 4 The similar signal in the time domain and ∆119905119905 are obtained from the

two different damage cases

- 8 -

Damage 1

Damage 2

20

Unit mm15

90 G value reduction

943 G value reduction

m-value = 248

m-value = 248

015 02 025 03 035

-2

-1

0

1

2

Time (ms)

Ampl

itude

Damage 1Damage 2

Fig 4 Two different FEM models which have different 119889119889 and 119907119907119889119889 (or G) This represents that the damage cannot be characterized uniquely based on the measured data available

Thus in this study the damage index (m-value) which can be uniquely defined based on

the available data is introduced The m-value is defined as the ratio of the physical length of

the damage and wavelength of the A0 Lamb wave mode in the damage area (120582120582119889119889)

119898119898 =119889119889120582120582119889119889

=119889119889119889119889119907119907119889119889

(3)

In many cases the transmitted and reflected waves are overlapped and it is hard to decom-

pose and calculate the arrival times of them Matching pursuit method is an alternative way

to accurately decompose of the overlapped waves and estimate the parameters of each wave

such as arrival time (Fig 5) Detail explanation of the matching pursuit method is covered in

section 36

015 02 025 03-2

-1

0

1

2

Time (ms)

Am

plitu

de

OriginalApproximated

(a)

015 02 025 03-2

-1

0

1

2

Time (ms)

Ampl

itude

Wave 1Wave 2

(b)

Fig 5 (a) Example of the overlapped signal with comparing approximated signal using the matching pursuit method (b) Decomposed signal using the matching pursuit method The matching pursuit method provides a conventional way to decompose and identify each wave

- 9 -

33 Damage detection algorithm (Level 1)

The damage diagnosis is accomplished by extracting the additional reflected A0 Lamb

wave modes from damage due to multiple reflections If A0 Lamb wave modes are decom-

posed perfectly it is trivial that one transmitted A0 Lamb wave mode is always captured in

the response signal If damage exists other waves in the response signal behind the transmit-

ted A0 Lamb wave mode are presented Since this approach does not rely on previously col-

lected data the diagnosis is independent from environmental variation such as temperature

changes

It is reported that the dispersion of Lamb waves cause wave-packets to spread out in space

and time as they propagate through a structure [20] Thus the reflected waves from the dam-

age have longer pulse duration than the transmitted wave due to their longer propagating dis-

tance In this study the matching pursuit method is modified to capture more spread wave

than the transmitted A0 Lamb wave mode Detail explanation of the modified matching pur-

suit method is described in latter section

Let 1199041199041 and 1199041199042 as the pulse duration (scale) of the transmitted and additional reflected A0

Lamb wave mode respectively

ldquoIf 1199041199042 gt 1199041199041 damage exists

Otherwise there is no damagerdquo

- 10 -

34 Damage quantification algorithm (Level 3)

As mentioned in the previous section damage is quantized by estimating 119889119889 and 119907119907119889119889 using

pitch-catch method The procedure to estimate 119889119889 and 119907119907119889119889 is

(i) Measure 119897119897 119905119905 and 119905119905119889119889

(ii) Assume the range of 119889119889 (0 lt 119889119889 lt 119897119897)

(iii) Calculate 119907119907119894119894 and 119907119907119889119889 with a constraints (0 lt 119907119907119889119889 lt 119907119907119894119894) (Eq 4 and 5)

(iv) Estimate m-value and find a feasible range of 119889119889 and 119907119907119889119889

119907119907119894119894 =119897119897 minus 119889119889119905119905 minus 119905119905119889119889

(4)

119907119907119889119889 =119889119889119905119905119889119889

(5)

35 Damage localization algorithm (Level 2amp3)

For damage localization 1198971198971 is estimated by measuring the arrival time from the excitation

PZT to the entrance of delamination (1199051199051) with additional pulse-echo method implementation

The procedure is similar with the previous damage quantification algorithm

(i) Measure 119897119897 119905119905 119905119905119889119889 and 1199051199051

(ii) Assume the range of 1198971198971 (0 lt 1198971198971 lt 119897119897)

(ii) Compute possible combinations of 1198971198972 119889119889 and 119907119907119889119889 with constraints

(0 lt 119907119907119889119889 lt 119907119907119894119894 119889119889 gt 0) (Eq 5 6 and 7)

- 11 -

1198971198972 = 1198971198971119905119905 minus 1199051199051 minus 119905119905119889119889

1199051199051 (6)

119889119889 = 119897119897 minus 1198971198971 minus 1198971198972 (7)

Since a PZT for pulse-echo method acts as both an actuator and a sensor the obtained sig-

nal contains both the excitation and the response signal Thus the reflected wave from the

entrance of damage can be buried into the tail of excitation signal and cause error on arrival

time of reflected wave from damage Even in this case damage can be localized and quan-

tized using the constraints When first wave packet reflected from the entrance of damage is

buried into the excitation signal the next wave packet reflected from the exit of damage can

be utilized alternatively If the algorithm regards the second reflected wave as the first re-

flected wave there are no combinations of estimated variable meeting the constraints Then

the algorithm repeats the process by regarding the wave as the second reflected wave the

constraints can select proper combination of the estimated variables

The estimated values have range bounded by the constraints When additional information

of 119907119907119894119894 is utilized the damage can be characterized more accurately In this study three meth-

ods are proposed and verified as information level of 119907119907119894119894

Method 1 Solely based on the current data measurement (Reference-free)

Method 2 Limit the possible range of 119907119907119894119894 (Prior knowledge of the system)

Method 3 Measure 119907119907119894119894 (Additional data need to be collected)

Method 1 described in this section can be utilized without any information of 119907119907119894119894 Meth-

od 2 produces more accurate estimation results using prior knowledge of the system such as

- 12 -

119907119907119894119894 variation under temperature changes Method 3 gives unique estimated value using meas-

ured 119907119907119894119894 from other path in the same time In section 4 and 5 damage is characterized using

these three methods and the results are compared each another

36 Matching pursuit method considering the dispersion effect of Lamb

wave

Matching pursuit method has been developed for signal compression approximation and

characterization of obtained signal and function [21-23] The key concept of the matching

pursuit method is that any signal (function) can be decomposed into a linear expansion of

signals selected from a pool of basis signals (functions) called lsquodictionaryrsquo

Let 119891119891 be a function to be analyzed then 119891119891 can be approximated a set of functions select-

ed from the dictionary 119863119863 in order to best match This is done by successive approximations

with orthogonal projections Let1198921198921205741205740 isin 119863119863 the function 119891119891 can be decomposed into

119891119891 = lang1198911198911198921198921205741205740rang1198921198921205741205740 + 119877119877119891119891 (8)

where 119877119877119891119891 is the residual function after approximating 119891119891 using 1198921198921205741205740 1198921198921205741205740 is orthogonal to

119877119877119891119891 hence

1198911198912 = lang1198911198911198921198921205741205740rang2

+ 1198771198771198911198912 (9)

To minimize 119877119877119891119891 1198921198921205741205740 is chosen to maximize lang1198911198911198921198921205741205740rang The matching pursuit method is

an iterative algorithm that decomposes the residue 119877119877119891119891 by projecting it on a function of 119863119863

that matches 119877119877119891119891 almost at best as it was done for 119891119891 The residue 119877119877119899119899119891119891 is decomposed into

119877119877119899119899119891119891 = lang119877119877119899119899119891119891119892119892120574120574119899119899rang119892119892120574120574119899119899 + 119877119877119899119899+1119891119891 (10)

- 13 -

the decomposition is carried up to the order 119898119898 119891119891 can be expressed as

119891119891 = lang119877119877119899119899119891119891119892119892120574120574119899119899rang119892119892120574120574119899119899

119898119898minus1

119899119899=0

+ 119877119877119898119898119891119891 (11)

The original function 119891119891 is decomposed into a sum of dictionary elements that are chosen to

best match its residues

In this study the matching pursuit method is utilized to accurately decompose of the over-

lapped Gaussian pulses and estimate the parameters of the pulse which used for damage

characterization The 119894119894119905119905ℎ Gaussian pulse in the dictionary for approximating the obtained

signal is expressed as

119892119892120574120574119894119894 = 119891119891(119860119860119894119894 119906119906119894119894 119904119904119894119894 119889119889119894119894120601120601119894119894) (12)

where 119860119860119894119894119906119906119894119894119904119904119894119894119889119889119894119894 and 120601120601119894119894 represent amplitude arrival time scale (pulse duration) excita-

tion(central) frequency and phase of the Gaussian pulse respectively The parameters of the

pulse are estimated using a nonlinear regression based optimization method (Gauss-Newton

method) based on an initial guess of the parameters Detail procedure for estimating the pa-

rameters of the pulse is expressed in Hong et al [23]

Among the estimated parameters of the pulses pulse duration (scale) represents the disper-

sion effect If a wave-packet is spread out the estimated pulse duration is increased Using

this feature the matching pursuit method is modified to capture waves which have longer

pulse duration than transmitted A0 Lamb wave mode If the estimate pulse duration of a wave

is less than the transmitted wave the modified matching pursuit method regards it as error

- 14 -

CHAPTER 4 NUMERICAL SIMULATION

41 Composite modeling method

Composite materials can be modeled from nano to macro level depends on its engineering

purpose In general four methods are utilized for modeling of composite materials First

composite is modeled as a homogenous material expressed in terms of effective material

properties [24] The effective material properties are estimated using the modulus and the

volume fraction of each constituent theoretically (Rule of mixture) and measured experimen-

tally Second a model is established that each laminate which have own material properties

are totally bonded together In this model the stacking sequence and direction of each lami-

nates are considered additionally [25-27] Third method for composite modeling in micro

level called micromechanics [28-30] This model considers the interaction between fibers and

matrix in laminate Last multiscale modeling which the scope of modeling is extended to

molecular or quantum level of constituents [31-33] This model starts from molecular me-

chanics or molecular dynamics then the scale of model leads to the nano micro and macro

level by incorporating small scale models As the scale of models is smaller it provides more

logical way than other large scale of models while the computational cost is more expensive

The scope of this study is focused on the dynamic behavior of composite structure the beam

or plate model was utilized for the numerical simulation

- 15 -

42 Damage modeling method

Generally three damage models have been developed and utilized in the real applications

First the stiffness reduction model called as the equivalent element model was introduced

[17 34-35] This model is simply realized by replacing the region around damage area by a

region with adjusted material properties (Youngrsquos modulus shear modulus density etc)

globally Next the duplicated node model is done by disconnecting the element connectivity

assuming there is no interaction among the delaminated surface This model can be utilized

when the location and the extension of damage is important issue [34-36] Last the contact

mechanics model using the principles of contact mechanics the two interfacial forces (con-

tact resistance and sliding resistance) in the delaminated surface [37-38] In this model the

delaminated surfaces are modeled as nonlinear functions of the interlaminar normal strain

and the interlaminar shear strain This model is computationally expensive while provides

more logical way than other models There is however no universally accepted model for

general impact-induced damage in composites In the case of impact-induced damage it is

impossible to predict or estimate the location and number of delaminated layers Thus from

the uncertainty of impact-induced damage the equivalent element model can be considered

as the effective damage model

- 16 -

43 Numerical simulation description

TABLE I Material property of simulation model and test specimen

Tensile Strength 657 MPa Tensile Modulus 59 GPa Shear Strength 804 MPa Shear Modulus 24 Gpa Specific Gravity 16

PZT A

PZT C

PZT BDamage

3

250 125 125

500 Unit mm

Fig 6 Description of the numerical simulation model

The feasibility of the proposed damage characterization technique is verified by 2-D FEM

simulation using MSC PATRAN and NASTRAN Fig 6 presents the dimension PZTs and

damage location of the simulation model In the model 05mm x 05mm square shell element

was used for modeling of the A0 mode whose flexural wave characteristic The element size

was determined by considering the wavelength of the propagation Lamb wave modes The

material properties of the simulation model and damage cases are presented in Table I and

Table II respectively

PZT A and C act as both actuator and sensor for collecting pulse-echo signal They were

installed on the plate which two of them are both-side collocated for the A0 Lamb wave

- 17 -

mode decomposition [39-40] PZT B was only used as a sensor for collecting pitch-catch sig-

nal For simulating the excitation and sensing of PZTs the pin-force model was adopted that

an input force apples to the end nodes of excitation PZT and averaging element strains corre-

sponding sensing area A Gaussian pulse (tone-burst) 50 kHz central frequency was excited

as an input signal

TABLE II Damage cases for numerical simulation

119949119949120783120783 (mm) 119941119941 (mm) G value reduction ()

Case 1 650 200 90 darr

Case 2 675 150 90 darr

Case 3 675 150 93 darr

Case 4 900 150 90 darr

- 18 -

44 Numerical simulation results

015 02 025 03 035

-2

0

2

Time (ms)

Am

plitu

de

IntactDamage

Reflected wave from damage

(a) Pitch-catch signal

015 02 025 03 035

-2

0

2

Time (ms)

Am

plitu

de

IntactDamage

Reflected waves from damage

(b) Pulse-echo signal Fig 7 Comparison of A0 Lamb wave mode signals from numerical simulation between intact and damage cases Time delay attenuation and multiple reflections due to damage simulated successfully using the damage modeling method described in the previous section

Fig 7 shows that comparison of A0 Lamb wave mode signals between intact and one of

damage cases in the simulation The features of A0 Lamb wave mode passing through dam-

age area described in section 2 are successfully simulated Thus the equivalent element

method is proper for modeling damage in composites In the damage cases the additional

waves reflected from damage is observed in both pitch-catch and pulse echo signal

Table III presents the result of the proposed damage characterization algorithm using the

numerical simulation cases shown in Table II In lsquoMethod 2rsquo 119907119907119894119894 variation is assumed to plusmn

5 of the normal value (1190~1315ms) The normal value of 119907119907119894119894 for lsquoMethod 3rsquo is meas-

ured as 1253 ms In each case the damage is characterized using the proposed technique

Note that as the information of 119907119907119894119894 is specified the estimated value is closer to the exact

value

- 19 -

TABLE III Damage characterization for the numerical simulation models

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

Case 1

Exact 650 200

Method 1 248 50~70 79~414 160~834

Method 2 248 603~667 135~241 271~486

Method 3 248 635 188 379

Case 2

Exact 675 150

Method 1 193 60~75 01~251 37~651

Method 2 193 618~682 114~222 295~575

Method 3 193 65 168 435

Case 3

Exact 675 150

Method 1 233 60~75 66~302 140~649

Method 2 233 668~738 85~196 182~419

Method 3 233 703 140 301

Case 4

Exact 900 150

Method 1 204 75~100 48~349 118~856

Method 2 204 860~951 107~216 264~530

Method 3 204 906 162 397

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1253 ms 1190~1315 ms Measured 119907119907119894119894 1253 ms

- 20 -

CHAPTER 5 EXPERIMENTAL VERIFICATION

51 Overview

In this section experimental study is investigated so that the proposed damage characteri-

zation technique Overall experimental setup and test specimen are shown in Fig 8 and 9

The experimental setup consists of test specimens data acquisition (DAQ) system and self-

sensing circuit Self-sensing circuit was installed for pulse-echo measurement using PZTs

[40-41]

PZT A(PZT C) PZT B

500

500

140

75

250

Teflon Inserted Damage (d =25)

Unit mm

Thickness = 3

PZT A(PZT C) PZT B

500

500

150

90

250

Impact Location

Unit mm

Thickness = 3

(a) Teflon inserted damage specimen

(b) Impact-induced damage specimen

Fig 8 Test specimens for the experimental verification In the Teflon inserted damage Teflon tapes with 25mm diameter were inserted at the center of the specimen The impact-induced damage was introduced to the specimen using the impact test equipment

- 21 -

511 Specimen

Two carbon composite plates with a dimension of 500times500times3mm3 were manufactured by

Nexcomes Inc by stacking 12 layers of CF3327 3K prepreg (woven fabrics type) manufac-

tured by Hankuk Fiber Inc The material properties of the specimen are presented in Table I

One of the specimens Teflon tapes with 25mm diameter was inserted in every 3 layers at the

center of the specimen to mimic the damage The other specimen is utilized for the character-

ization of real impact-induced damage

512 Data acquisition (DAQ) system and PZT

The data acquisition (DAQ) system consists of an arbitrary waveform generator (AWG) a

high speed signal digitizer (DIG) and two multiplexers (MUX) Using the 16-bits AWG

tone-burst input signal with a plusmn10 peak-to-peak voltage were generated The response signal

was measured by the 14-bit DIG For the signal-to-noise ratio reduction both pitch-catch and

pulse-echo signals were measured 150 times and averaged in the time domain 3 identical

circular PZTs of 10mm diameter manufactured by APC International Ltd were installed on

the specimen A Gaussian pulse (tone-burst) 50 kHz central frequency was excited as an in-

put signal The excitation frequency was chosen in order to avoid overlapping of the S0 and

A0 Lamb wave modes and generating of the higher modes (S1 A1hellip)

- 22 -

PZT Actuator

PZT Sensor

AWG

DIG

DAQ System Specimen

SensingActuatingMUX

1

23

4

1 Sending signal to MUX

4 Sending signal to DIG

Self-SensingCircuit

18 Unit mm

10

(a)

(b)

Fig 9 Overall experimental setup (a) Data acquisition (DAQ) system and (b) Installed circu-lar PZT transducer on the specimen with its dimension

513 Impact test

The impact-induced damage was introduced to the specimen using the impact test equip-

ment presented in Fig 10 A 5kg mass with 10mm tip was dropped from 015m height onto

the same location of the specimen The signals were obtained at each number of impacts

Since the reflected wave is not detected at the first impact case the data obtained from the

second and third impact were utilized for the analysis

10

Unit mm (a)

(b)

Fig 10 (a) Impact test equipment for introducing impact-induced damage and (b) Tip of the impact tester with 10mm diameter

- 23 -

514 Temperature effect test

In reality structures are subjected to changing environmental conditions such as tempera-

ture that affect measured signal and cause false-alarm Thus the robustness of the proposed

damage characterization technique under temperature variation in investigated (Fig 11) The

specimens were placed inside a temperature chamber and a thermocouple was installed on

the specimen to measure temperature of surface of the specimen The signal was obtained

under three temperature cases (0 20 and 50degC) and the humidity was kept at 30

Temperature Chamber

Test Specimen

Fig 11 Test specimen in temperature chamber

- 24 -

52 Teflon inserted damage

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Reflected wave from damage

(a)

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Transmitted WaveReflected Wave

Detected reflected wave

(b)

Fig 12 A0 Lamb wave mode obtained using pitch-catch method from the Teflon inserted damage specimen (a) The reflected wave from damage is shown (b) The Teflon inserted damage is detected by capturing the reflected wave

Fig 12 shows the obtained pitch-catch signal from the Teflon inserted damage specimen

and damage diagnosis result The Teflon inserted damage is detected successfully by captur-

ing the reflected wave Table IV presents the result of the Teflon inserted damage characteri-

zation 119907119907119894119894 is measured as 1213ms and the variation is assumed to plusmn 5 of the normal val-

ue (1152~1273ms) The results indicate that the proposed damage characterization approach

works well in real application

TABLE IV Damage characterization for the Teflon inserted damage specimen

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

Exact 635 25

Method 1 123 60~75 73~338 298~1220

Method 2 123 619~662 229~310 934~1188

Method 3 123 631 284 1161

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1213 ms

- 25 -

53 Impact-induced damage

11 115 12 125 13-4

-2

0

2

4

Time (ms)

Vol

tage

(mV

)

IntactImpact2Impact3

Reflected wave from damage

(a) Pitch-catch signal

11 115 12 125 13 135

-05

0

05

1

Time (ms)

Volta

ge (m

V)

IntactImpact2Impact3

Boundary reflection

Reflected wave from damage

(b) Pulse-echo signal

Fig 13 The signal obtained from the test specimen subjected to impact The impact was in-troduced at the same location of the specimen and the signal was obtained after each impacts As the number of impact is increased the transmitted A0 Lamb wave mode is delayed and attenuated

The obtained signals from impact-induced damage specimen are presented in Fig 13 It is

observed that as the number of impact is increased the transmitted A0 Lamb wave mode is

more delayed because the wave velocity in the damage area is reduced Fig 14 shows the

false-alarm test result whether the proposed damage detection algorithm works even in pris-

tine condition of the test specimen It is shown that no false alarm is occurred because no ad-

ditional reflected wave is captured

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Transmitted Wave

(a) Transmitted wave

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Reflected Wave

No reflected wave

(b) Reflected wave

Fig 14 False alarm test for the proposed damage detection algorithm in pristine condition of the specimen Reflected wave from damage is not captured false alarm is not occurred

- 26 -

TABLE V Damage characterization for the impact-induced damage specimen

of Impacts m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

2

(1470 J)

Method 1 080 65~90 75~246 470~1157

Method 2 080 801~885 114~245 711~1134

Method 3 080 843 179 1122

3

(2205 J)

Method 1 148 50~100 06~380 21~1134

Method 2 148 758~838 249~368 843~1125

Method 3 148 798 308 1044

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1213 ms

Table V presents the characterization results of the impact-induced damage case Here the

measured 119907119907119894119894 and the assumed variation are same as the artificial damage case It is observed

that as the number of impact is increased 119889119889 is increased while 119907119907119889119889 is reduced Thus it is

validated that the proposed damage characterization approach estimates the increment of the

damage well

54 Temperature effect test

541 Teflon inserted damage

11 115 12 125 13-10

-5

0

5

10

Time (ms)

Vol

tage

(mV

)

0 oC20 oC50 oC

(a) Pitch-catch signal

11 115 12 125 13 135

-1

0

1

2

3

Time(ms)

Volta

ge (m

V)

0 oC20 oC50 oC

(b) Pulse-echo signal

Fig 15 Signals obtained from the Teflon inserted damage specimen under temperature varia-tion Time delay and amplitude variation are occurred

Fig 15 shows the obtained signals from the Teflon inserted damage specimen subjected to

- 27 -

temperature variation Time delay and amplitude variations are occurred there is no addition-

al damage is introduced The analyzed results of the obtained signals are reported in Table VI

The measured 119907119907119894119894 is 1247 ms at 0 degC 1213 ms at 20 degC and 1168 ms at 50 degC respectively

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms (1152~1273 ms)

The estimated results are also changed under temperature variations the fluctuation of esti-

mated range and exact value is small

TABLE VI Temperature effect test for the Teflon inserted damage specimen

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

0 degC

Exact 635 25

Method 1 113 60~70 79~268 352~1291

Method 2 113 560~619 233~344 1034~1172

Method 3 113 606 257 1141

20 degC

Exact 635 25

Method 1 123 60~75 73~338 298~1220

Method 2 123 619~662 229~310 934~1188

Method 3 123 631 284 1161

50 degC

Exact 635 25

Method 1 137 60~75 72~328 263~1228

Method 2 137 622~687 183~299 666~1086

Method 3 137 642 264 960

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1247 ms at 0 degC 1213 ms at 20 degC and 1168 ms at 50 degC

- 28 -

542 Impact-induced damage

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Vol

tage

(mV

)

Transmitted WaveReflected Wave

No reflected wave

(a) 0 degC

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Transmitted WaveReflected Wave

No reflected wave

(b) 50 degC

Fig 16 False alarm test under temperature variations It is shown that proposed damage char-acterization algorithm is robust under temperature variations

The false alarm test in pristine condition of the impact-induced damage specimen under

temperature variation was investigated (Fig 16) It is validated that the proposed damage di-

agnosis algorithm is robust under temperature variations Fig 17 shows the obtained signals

from the impact-induced damage (impacted 3 times) specimen under temperature variation

The analyzed results of the obtained signals are reported in Table VII The measured 119907119907119894119894 in

each temperature and the assumed variation are same as the artificial damage case

11 115 12 125 13-5

0

5

Time (ms)

Volta

ge (m

V)

0 oC20 oC50 oC

(a) Pitch-catch signal

11 115 12 125 13 135

-1

0

1

2

3

Time(ms)

Vol

tage

(mV

)

0 oC20 oC50 oC

(b) Pulse-echo signal

Fig 17 Signals obtained from impact-induced damage specimen (Impact 3) under tempera-ture variation

- 29 -

TABLE VII Temperature effect test for the impact induced damage specimen (Impact 3 case)

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

0 degC

Method 1 135 75~90 72~310 267~1148

Method 2 135 705~779 264~381 978~1126

Method 3 135 763 289 1071

20 degC

Method 1 148 50~100 06~380 21~1134

Method 2 148 758~838 249~368 843~1125

Method 3 148 798 308 1044

50 degC

Method 1 153 75~100 53~415 174~1227

Method 2 153 804~889 214~336 702~1103

Method 3 153 815 320 1050

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1247 ms at 0 degC 1213 ms at 20 degC and 1168 ms at 50 degC

- 30 -

CHAPTER 6 CONCLUDING REMARKS

61 Executive summary

In this study the reference-free damage detection localization and quantification tech-

nique in a multi layered composite plate is proposed Additional reflected A0 Lamb wave

mode from the damage due to multiple reflections in damage area is used as a feature for

characterizing the damage First damage is characterized in terms of a damage index (m-

value) defined as the ratio of the physical length of the damage to the wavelength of the A0

Lamb wave mode Then the range of possible damage location length and severity are esti-

mated with constraints This procedure can be accomplished without prior baseline data To

realize the proposed technique the matching pursuit technique is modified for reduction of

the error by considering the dispersive characteristic of Lamb wave The matching pursuit

method estimates the arrival time of each wave packets accurately even two wave packets are

overlapped The numerical simulations and experimental tests including Teflon tape inserted

and impact-induced damage are investigated for the feasibility of the proposed technique

Moreover the effectiveness of the proposed technique under changing temperature is exper-

imentally verified

- 31 -

62 Suggestion for further study

The limitations of this study and the possible improvement and further studies are listed as

follow

1 Measurement of impact-induced damage length using conventional NDT technique

Additional experimental verification for impact-induced damage characterization is

needed by measuring the real length of damage using conventional NDT technique

such as ultrasonic scan or infrared camera

2 Limitation of both side collocated PZT installation In this study excitation PZTs are

installed both side and collocated for decomposing A0 Lamb wave mode However it

can produce decomposition error due to imperfection of PZT location and their installa-

tion Moreover in many cases assess to both surface of structures is not allowed such

as aircraft applications To improve this limitation a reference-free damage characteri-

zation using single side PZTs will be considered

3 Complex structure application In this study damage is characterized in simple plate-

like structure When the monitored structure becomes complex the response signals al-

so becomes complex due to its geometry and boundary Thus the investigations for

complex structure application such as stiffener spar and curvature are needed

- 32 -

REFERENCES

[1] Cawley P and Alleyne D (1996) ldquoThe use of Lamb waves for the long range inspection of large structuresrdquo Ultrasonics 34 pp287-290

[2] Ghosh T Kundu T and Karpur P (1998) ldquoEfficient use of Lamb waves for detecting defects in large platesrdquo Ultrasonics 36 pp791-801

[3] Guirguitui V (2003) ldquoLamb wave generation with piezoelectric wafer active sensors for structural health monitoringrdquo SPIE Conference on Smart Structures and Materials San Diego CA 5391 pp111-422

[4] Ragvahan A and Cesnik CE S (2004) ldquoModeling of piezoelectric-based Lamb wave generation and sensing for structural health monitoringrdquo SPIE Conference on Smart Struc-tures and Materials San Diego CA 5391 pp419-430

[5] Rose JL Pilarski A and Dirit JJ (1993) ldquoAn approach to guided wave mode selection for inspection of laminated platerdquo Journal of Reinforced Plastics and Composites 12(5) pp536-544

[6] Guo N and Cawley P (1993) ldquoThe interaction of Lamb waves with delaminations in composite laminatesrdquo Journal of Acoustical Society of America 94(4) pp 2240-2246

[7] Valdez SHD and Soutis C (2002) ldquoReal-time nondestructive evaluation of fiber com-posite laminates using low-frequency Lamb wavesrdquo Journal of Acoustical Society of Amer-ica 111(5) pp 2026-2033

[8] Yang M and Qiao P (2005) ldquoModeling and experimental detection of damage in virous materials using the pulse-echo method and piezoelectric sensoractuatorsrdquo Smart Materials and Structures 14 pp1083-1100

[9] Lestari W and Qiao P (2005) ldquoApplication of wave propagation analysis for damage identification in composite laminated beamsrdquo Journal of Composite Materials 39(22) pp1967-1984

[10] Kessler S S Spearing S M and Soutis C (2002) ldquoDamage detection in composite ma-terials using Lamb wave methodsrdquo Smart Materials and Structures 11 pp269-278

[11] Sohn H Park G Wail J R Limback N P and Farrar C R (2004) ldquoWavelet-based active sensing for delamination detection in composite structuresrdquo Smart Materials and Structures 13 pp 153-160

[12] Petculescu G Krishnaswamy S and Achenback J D (2008) ldquoGroup delay measure-ments using modally selective Lamb wave transducers for detection and sizing of delamina-tions in compositesrdquo Smart Materials and Structures 17

[13] Sohn H Park H W Law K H and Farrar C R (2007) ldquoDamage detection in compo-

- 33 -

site plates by using an enhanced time reversal methodrdquo Journal of Aerospace Engineering (ASCE) pp 141-151

[14] Sohn H (2007) ldquoEffects of environmental and operational variability on structural health monitoringrdquo Philosophical Transactions of the Royal Society Ardquo 365 pp539-560

[15] Rytter A (1993) ldquoVibration based inspection of civil engineering structuresrdquo PhD The-sis Department of Building Technology and Structural Engineering University of Aalborg Aalborg Denmark

[16] Hayashi T and Kawashima K (2002) ldquoMultiple reflections of Lamb waves at a delami-nationrdquo Ultrasonics 40 pp193-197

[17] Wang C H and Rose L R F (2003) ldquoWave reflection and transmission in beams con-taining delamination and inhomogeneityrdquo Journal of Sound and Vibration 264 pp 851-872

[18] Ramadas C Balasubramaniam K Joshi M and Krishnamurthy C V (2009) ldquoInterac-tion of the primary anti-symmetric Lamb mode (Ao) with symmetric delaminations numer-ical and experimental studiesrdquo Smart Materials and Structures 18

[19] Ramadas C Balasubramaniam K Joshi M and Krishnamurthy C V (2010) ldquoInterac-tion of guided Lamb waves with an asymmetrically located delamination in a laminated composite platerdquo Smart Materials and Structures 19

[20] Wilcox P Lowe M and Cawely P (2001) ldquoThe effect of dispersion on long-range in-spection using ultrasonic guided wavesrdquo NDTampE International 34 pp1-9

[21] Mallat S G and Zhang Z (1993) ldquoMatching pursuits with time-frequency dictionariesrdquo IEEE Tansactions on Signal Processing 41(12) pp3397-3415

[22] Zhang G Zhang S and Wang Y (2000) ldquoApplication of adaptive time-frequency de-composition in ultrasonic NDE of highly scattering materialsrdquo Ultrasonics 38 pp 961-964

[23] Hong J C Sun K H and Kim Y Y (2005) ldquoThe matching pursuit approach based on the modulated Gaussian pulse for efficient guided-wave damage inspectionrdquo Smart Mate-rials and Structures 14 pp548-560

[24] Herakovich C T (1997) Mechanics of Fibrous Composites Wiley 480 pages

[25] Reddy J N (1985) ldquoA review of the literature on finite-element modeling of laminated composite platesrdquo The Shock and Vibration Digest 17(4) pp 3-8

[26] Reddy J N (2003) Mechanics of Laminated Composite Plates and Shells Theory and Analysis CRC Press 856 pages

[27] Palacz M Krawczuk M and Ostachowicz W (2005) ldquoThe spectral finite element mod-el for analysis of flexure-shear coupled wave propagation Part 1 laminated multilayer

- 34 -

composite beamrdquo Composite Structures 68 pp37-44

[28] Zhang Y C and Harding J (1990) ldquoA numerical micromechanics analysis of the me-chanical properties of a plain weave compositerdquo Computer amp Structures 36(5) pp 839-844

[29] Nemat-Nasser S and Hori M (1999) Micromechanics Overall Properties of Heteroge-neous Materials Elsevier 810 pages

[30] Scida D Aboura Z Benzeggagh M L and Bocherens E (1999) ldquoA micromechanics model for 3D elasticity and failure of woven-fibre composite materialrdquo Composites Science and Technology 59 pp 505-517

[31] Thomas Y and Wu X H (1997) ldquoA multiscale finite element method for elliptic prob-lems in composite materials and porous mediardquo Journal of Computational Physics 134 pp 169-189

[32] Lee K H Moorthy S and Ghosh S (1999) ldquoMultiple scale computational model for damage in composite materialsrdquo Computer Methods in Applied Mechanics and Engineer-ing 172 pp 175-201

[33] Namilae S and Chandra N (2005) ldquoMultiscale model to study the effect of interfaces in carbon nanotube-based compositesrdquo Journal of Engineering Materials and Technology (ASME) 127 pp222-232

[34] Adams D E (2007) Health monitoring of Structural Materials and Components Wiley 460 pages

[35] Gopalakrishnan S (2009) ldquoModeling Aspects in Finite Elementrdquo Encyclopedia of Struc-tural Health Monitoring 2 pp791-809

[36] Palacz M Krawczuk M and Ostachowicz W (2005) ldquoThe spectral finite element mod-el for analysis of flexure-shear coupled wave propagation Part 2 Deminated multilayer composite beamrdquo Composite Structures 68 pp45-51

[37] Baik J M and Thopson R B (1984) ldquoUltrasonic scattering from imperfect interfaces A quasi-static modelrdquo Journal of Nondestructive Evaluation 4(3) pp177-196

[38] Luo H and Hanagud S (2000) ldquoDynamics of delamincated beamsrdquo International Jour-nal of Solids and Structures 37 pp1501-1519

[39] Kim S B and Sohn H (2007) ldquoInstantaneous reference-free crack detection based on polarization characteristics of piezoelectric materialsrdquo Smart Materials and Structures 16 pp2375-2387

[40] An Y K and Sohn H (2010) ldquoInstantaneous crack detection under varying temperature and static loading conditionsrdquo The Special Issue of Structural Control and Health Monitor-ing 17(7) pp 730-741

- 35 -

[41] Lee S J and Sohn H (2006) ldquoActive self-sensing scheme development for structural health monitoringrdquo Smart Materials and Structures 15(6) pp 1734-1746

- 36 -

요 약 문

복합재료 손상의 무기저 진단 및 정량화

복합재료는 재료의 경량성 내구성 고강도의 특성으로 인하여 항공기에서부터

교량에까지 널리 사용되고 있다 하지만 외부의 충격으로 인한 복합재료 내부의

손상은 재료의 강도를 감소시켜 구조물의 안정성을 저하시키는 요인이 된다 하지만

복합재료의 손상은 재료 내부에서 발생하여 육안으로는 식별하기도 힘들며 표면에서

보이는 것만으로 내부의 손상 정도를 판단하기는 불가능하다는 단점을 가지고 있다

본 연구에서는 손상 내부에서 반사로 인해 발생하는 파를 이용하여 기저자료를

사용하지 않고 특정 시점에서 얻은 자료만으로 복합재료의 손상을 진단하고 정량화

할 수 있는 기법을 제안한다 제안된 기법의 적용성을 검증하기 위하여 유한요소법을

이용한 수치해석을 수행하였고 Teflon tape 이 삽입된 시편과 실제 충격하중으로

인한 손상을 입은 시편을 이용한 실험을 진행하였다 수치해석 및 실험 결과 제안된

기법이 복합재료의 손상을 성공적으로 진단 및 정량화할 수 있음을 확인하였다 또한

다양한 외부 온도 환경에서 제안된 기법의 환경영향 실험을 수행하였고 다양한

환경에서 손상의 진단 및 정량화가 성공적으로 이루어짐을 검증하였다

핵심어 무기저 손상진단 및 정량화 압전센서 복합재료 유한요소법 Matching pursuit

method

- 37 -

감 사 의 글

기대와 두려움을 안고 시작한 석사과정이 어느덧 2 년 이라는 세월 속에서 추억으로 자리잡아

가고 있습니다 학부생 기간을 포함한 지난 3 년 동안 SHM 이라는 새로운 분야를 접하고 연구를

수행 할 수 있도록 도와주신 고마운 분들이 있어 그 감사를 표하고자 합니다

우선 저의 많은 부족한 점을 일깨워 주시고 연구에 대한 뜻을 더욱 확고히 세울 수 있게 해

주신 지도교수님이신 손 훈 교수님께 깊은 감사의 뜻을 전합니다 짧은 기간이었지만 교수님의

도움으로 많은 것을 배우고 느끼고 깨달을 수 있었습니다 또한 본 연구를 수행할 수 있도록 지

원해주신 국토해양부 (U-City 석 박사과정 지원사업)에 깊은 감사 드립니다

때로 힘들고 지칠 때 항상 저를 믿고 응원해 주시는 가족이 있었기에 지금의 글을 쓸 수 있었

습니다 또한 지금 연구실에서 함께 생활하고 있는 윤규형 민구형 철민이형 현석이형 진열이형

호민이형 병진이 선혜누나와 성대에 계신 창길이형 미국에 계신 의재형과 Debaditiya 에게도 감

사의 말씀을 전합니다 그리고 모두 나열할 수는 없지만 카이스트 생활의 활력소가 되어준 카이

스트 건설 및 환경공학과 선후배들 직원 분들과 SOS 멤버들에게도 감사를 표합니다

앞서 거론한 많은 이들의 도움이 없었다면 저 또한 지금 이 자리에 있을 수 없었을 것입니다

많이 부족하기에 더 노력하고 배우고자 했던 초심을 다시 한번 생각하며 더 큰 결실을 맺기 위해

앞으로 나아가겠습니다 도움을 주신 모든 분들께 이 감사의 글을 바칩니다

2011년 6월 카이스트에서 임형진 올림

- 38 -

CURRICULUM VITAE

Hyung Jin Lim

Research Assistant

The Civil and Environmental Engineering Department

Korea Advanced Institute of Science and Technology (KAIST)

291 Daehak-Ro Guseong-Dong Yuseong-Gu

Daejeon Republic of Korea 305-701

Phone (82)+42-350-3665 Fax (82)+42-350-3610

Email limhj87gmailcom

RESEARCH INTERESTS

Damage Detection Localization and Quantification on Composite and Metallic Structures Pattern Recognition Finite Element Modeling (FEM)

EDUCATION

2009 ndash 2011 MS Civil and Environmental Eng Korea Advanced Institute of Science and Technology (KAIST) Korea Rep

2006 ndash 2009 BS Civil and Environmental Eng Korea Advanced Institute of Science and Technology (KAIST) Korea Rep

JOURNAL PUBLICATIONS The corresponding author is underlined 1 Hyung Jin Lim Hoon Sohn ldquoReference-free damage detection localization and quantification

using in compositerdquo in preparation for Journal of the Acoustical Society of America 2011

2 Chul Min Yeum Hoon Sohn Jeong Bum Ihn Hyung Jin Lim ldquoInstantaneous delamination detection in a composite using a dual piezoelectric transducer networkrdquo Submitted to Compo-sites Science and Technology 2011

3 Yun-Kyu An Hyung Jin Lim Min Koo Kim Jin Yeol Yang Hoon Sohn Chang Guen Lee ldquoInvestigation of applicability of local reference-free damage detection techniques to in-situ

- 39 -

bridgesrdquo Submitted to ASCE Journal of Structural Engineering 2011

4 Hyung Jin Lim Min Koo Kim Hoon Sohn Chan Yik Park ldquoImpedance based damage detec-tion under varying temperature and loading conditionsrdquo Submitted to NDTampE International 2011

5 Hyung Jin Lim Hoon Sohn ldquoQuantitative estimation of transmitted and reflected lamb waves at discontinuityrdquo Journal of the Korean Society for Nondestructive Testing Vol30 No 4 pp1-8 2010

CONFERENCE PROCEEDINGS

The corresponding author is underlined 1 Hyung Jin Lim and Hoon Sohn ldquoReference-free damage detection localization and quantifi-

cation in compositesrdquo ASCE Engineering Mechanics Institute Conference Boston June 2-4 2011

2 Hyung Jin Lim and Hoon Sohn ldquoReference-free delamination quantification within multi-layer composite materialsrdquo KSNT Spring Conference May 26-27 Gyeong-Ju Korea 2011

3 Min Koo Kim Hyun Jin Lim Hoon Sohn Chan Yik Park ldquoImpedance-based bolt loosening detection under varying temperature and loadingrdquo Asian Pacific Workshop on Structural Health Monitoring Tokyo Japan November 30-December 2 2010

4 Hyung Jin Lim Chul Min Yeum and Hoon Sohn ldquoModeling of Impact-induced Delamination in a Multilayer Composite Platerdquo The 23rd KKCNN Symposium on Civil Engineering Taipei (Taiwan) November 13-15 2010

5 Yun-Kyu An Hyung Jin Lim and Hoon Sohn ldquoTemperature independent Delamination Detec-tion using data normalizationrdquo The 3rd International Conference on Multi-Functional Materi-als and Structures Chonbuk National Univ Jeonju Korea September 14-18 2010

6 Hoon Sohn Yun-Kyu An Chang Gil Lee Min Koo Kim Hyung Jin Lim ldquoApplication of online NDT techniques to decommissioned bridge testingrdquo the 5th European Workshop on Structural Health Monitoring Sorrento Italy June 29-July 02 2010

7 Min Koo Kim Hyung Jin Lim Hoon Sohn Chan Yik Park ldquoImpedance based damage diag-nosis of a complex composite aircraft wing under changing loading and temperature condi-tionsrdquo the 5th European Workshop on Structural Health Monitoring Sorrento Italy June 29-July 02 2010

8 Yun-Kyu An Min Koo Kim Hyung Jin Lim and Hoon Sohn ldquoValidations of reference-free crack detection techniques through a decommissioned bridge testrdquo 2010 COSEIK Annual Con-ference Jeju Korea April 8-9 2010

- 40 -

PARTICIPANT PROJECTS

1 Development of Smart Steel Members for Online Steel Structure Monitoring POSCO Corporation (Funded 41000000 KRW (41000 USD) for 010511 to 123111)

2 Development of Enabling Technologies for Structural Health Monitoring Korean Agency for Defense Development (Funded 240000000 KRW (240000 USD) out of total 3200000000 KRW (3200000 USD) for 060108-053111)

3 Development of On-board SHM Technologies for Composite Air Vehicles The Boeing Company (Funded 400000 USD (400000000 KRW) for 080108 to 073111)

EDUCATIONAL EXPERIENCES

Fall 2009

Teaching Assistant CE207 ldquoElementary Structural Engineering and Laboratoryrdquo KAIST 3 unit undergraduate course

EXPERIMENTAL EXPERIENTCES

2010 Monitoring of Real Aircraft Composite Wing Segment Korea Agency for Defense De-velopment (ADD) Daejeon Korea Real time monitoring of composite air vehicles fund-ed by Korea Agency for Defense Development (ADD)

2009-2011

Field Monitoring of Real Bridge Structures Gimpo (Ramp G bridge) amp Icheon (Sam-seung bridge) Korea US-Korea collaborative research for bridge monitoring testbeds

HONORS amp AWARDS

Scholarships for outstanding students granted KAIST (2007-2009) The three prominent students in the civil and environmental engineering department of KAIST were selected as recipients This scholarship provided 2000$ per year for three years

- 41 -

Page 6: 복합재료 손상의 무기저 진단 및 정량화 - KAISTssslab.kaist.ac.kr/article/pdf/Thesis_2.pdf · 1 Declaration of Ethical Conduct in Research: I, as a graduate student

TABLE OF CONTENTS

ABSTRACT middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot i

TABLE OF CONTENTS middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot iii

LIST OF TABLES middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot iv

LIST OF FIGURESmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot v

CHAPTER 1 INTRODUCTION

11 Motivation middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 1

12 Literature review middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 1

13 Objective and uniqueness middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 3

14 Organization middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 4

CHAPTER 2 INTERACTION OF LAMB WAVE WITH DAMAGE

21 A0 Lamb wave mode delay and attenuation in the time domain middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 5

22 Multiple reflections in the damage area middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 6

CHAPTER 3 THEORETICAL DEVELOPMENT

31 Overview middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 7

32 Damage index (m-value) middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 8

33 Damage detection algorithm (Level 1) middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 10

34 Damage quantification algorithm (Level 3) middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 11

35 Damage localization algorithm (Level 2amp3) middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 11

36 Matching pursuit method considering the dispersion effect of Lamb wave

middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 13

CHAPTER 4 NUMERICAL SIMULATION

41 Composite modeling methodmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 15

42 Damage modeling method middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 16

43 Numerical simulation description middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 17

44 Numerical simulation results middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 19

ii

CHAPTER 5 EXPERIMENTAL VERIFICATION

51 Overview middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 21

511 Specimen middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 22

512 Data acquisition system and PZT middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 22

513 Impact test middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 23

514 Temperature effect test middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 24

52 Teflon inserted damage middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 25

53 Impact-induced damagemiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 26

54 Temperature effect test middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 27

541 Teflon inserted damage middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 27

542 Impact-induced damage middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 29

CHAPTER 6 CONCLUDING REMARKS

61 Executive summary middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 31

62 Suggestion for further study middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 32

REFERCENCES middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 33

SUMMARY (IN KOREAN) middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 37

ACKNOWLEDGEMENTS (IN KOREAN)

CURRICULUM VITAE

iii

LIST OF TABLES

I Material property of simulation model and test specimen middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 17

II Damage cases for numerical simulation middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 18

III Damage characterization for the numerical simulation models middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 20

IV Damage characterization for the Teflon inserted damage specimen middotmiddotmiddotmiddotmiddotmiddotmiddot 25

V Damage characterization for the impact-induced damage specimen middotmiddotmiddotmiddotmiddotmiddotmiddot 27

VI Temperature effect test for the Teflon inserted damage specimen middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 28

VII Temperature effect test for the impact induced damage specimen middotmiddotmiddotmiddotmiddotmiddotmiddot 30

iv

LIST OF FIGURES

1 Lamb wave propagation characteristics (a) A0 Lamb wave mode time delay and

attenuation after passing through a damage (b) The time delay and amplitude varia-

tion of A0 Lamb wave mode under temperature variation in pristine condition of

structure Because of this conventional damage detection techniques in composites

have limitations for real structure application middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 5

2 Multiple reflections in the damage area The multiple reflections of A0 Lamb

waves in the damage area are utilize as a feature for characterization of the damage

in this study middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 6

3 Parameters and notations of the proposed damage characterization technique

Damage is detected by identifying the additional reflected wave from the damage due

to multiple reflections (dashed lines) Localization and quantification of the damage

is carried out by measuring the arrival time of transmitted and reflected waves from

the damage middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 7

4 Two different FEM models which have different 119889119889 and 119907119907119889119889 (or G) This repre-

sents that the damage cannot be characterized uniquely based on the measured data

available middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 9

5 (a) Example of the overlapped signal with comparing approximated signal using

the matching pursuit method (b) Decomposed signal using the matching pursuit

method The matching pursuit method provides a conventional way to decompose

and identify each wave middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 9

6 Description of the numerical simulation model middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 17

7 Comparison of A0 Lamb wave mode signals from numerical simulation be-

tween intact and damage cases Time delay attenuation and multiple reflections due

to damage simulated successfully using the damage modeling method described in

v

the previous section middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 19

8 Test specimens for the experimental verification In the Teflon inserted damage

Teflon tapes with 25mm diameter were inserted at the center of the specimen The

impact-induced damage was introduced to the specimen using the impact test equip-

ment middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 21

9 Overall experimental setup (a) Data acquisition (DAQ) system and (b) Installed

circular PZT transducer on the specimen with its dimension middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 23

10 (a) Impact test equipment for introducing impact-induced damage and (b) Tip

of the impact tester with 10mm diametermiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 23

11 Test specimen in temperature chamber middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 24

12 A0 Lamb wave mode obtained using pitch-catch method from the Teflon in-

serted damage specimen (a) The reflected wave from damage is shown (b) The Tef-

lon inserted damage is detected by capturing the reflected wave middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 25

13 The signal obtained from the test specimen subjected to impact The impact

was introduced at the same location of the specimen and the signal was obtained after

each impacts As the number of impact is increased the transmitted A0 Lamb wave

mode is delayed and attenuated middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 26

14 False alarm test for the proposed damage detection algorithm in pristine condi-

tion of the specimen Reflected wave from damage is not captured false alarm is not

occurred middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 26

15 Signals obtained from the Teflon inserted damage specimen under temperature

variation Time delay and amplitude variation are occurred middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 27

16 False alarm test under temperature variations It is shown that proposed dam-

vi

age characterization algorithm is robust under temperature variationsmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 29

17 Signals obtained from impact-induced damage specimen (Impact 3) under

temperature variation middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 29

vii

CHAPTER 1 INTRODUCTION

11 Motivation

Composite materials has been increasingly demands in modern structures ranging from the

latest aircraft to retrofitted bridge deck due to its light weight superior specific strength and

stiffness properties However the occurrence of damage due to abrupt impact or sequential

accumulation of defect may considerably degrade structural strength or stiffness and reduce

the safety and reliability of the system Recently Lamb waves based damage detection tech-

niques using piezoelectric (PZT) transducers have emerged as one of the promising structural

health monitoring (SHM) technology for detecting hidden damage because of the possibility

of long range inspection and high sensitivity [1-4]

12 Literature review

This section will discuss researches that have been previously carried out on topics of

guided wave based SHM techniques using PZT transducers for damage in composite materi-

als The early investigations of interaction between Lamb wave and damage in composite can

be found in Rose et al and Guo et al [5-6] From further studies reflected signals from dam-

age were shown to be an effective method for damage diagnosis and localization using PZTs

[7-9] The reflected signals from damage were collected using one PZT as an actuator and the

other PZT as a sensor However the techniques cannot be used to determine the extent of

- 1 -

damage with reasonable precision

Kesseler et al proposed a technique monitoring the wavelet coefficients of transmitted ul-

trasonic signal amplitude as a potential way to damage detection [10] One drawback of this

technique is that the variability in coupling and the sensitivity of installed PZT transducers

can induce changes in the amplitude and energy of the signal that are cause false-alarm Oth-

er wavelet based delamination detection and localization technique developed using PZT

sensor network and outlier analysis [11]

Damage detection and quantification in composite plate was also investigated by group de-

lay measurement due to damage [12] However it is assumed that the group velocity in dam-

age area as the group velocity propagates half thickness of the structure in dispersion curve in

the quantification procedure Thus the damage quantification was investigated only Teflon

inserted damage case

Sohn et al proposed a damage detection technique which does not rely on any past base-

line signals to assess damage in composite plates by using an enhanced time reversal method

[13] The damage is assessed using the violation of time reversibility when nonlinearity is

caused by the damage along a direct wave path Thus it allows instantaneous identification

of delamination without requiring the baseline signal for comparison

- 2 -

13 Objective and uniqueness

In this study a reference-free damage characterization technique including detection lo-

calization and quantification of impact-induced damage in composite materials is developed

The general damage characterization can be categorized into the following levels [14]

Level 0 Define damage

Level 1 Determine the presence of damage within the structure

Level 2 Locate the regions of damage

Level 3 Quantify the damage

Level 4 Predict the remaining service life of the structure

Through the proposed technique from level 1 to level 3 can be achieved

Most of common approaches of damage characterization techniques described in the pre-

vious section use prior obtained baseline data However these approaches can cause false

alarm due to the effect of environmental variations such as temperature and ambient vibra-

tions [15]

The uniqueness of this study is

1 Damage is characterized without using prior measured baseline data

2 Robustness of the proposed technique under temperature variation is investigated

3 Additional waves reflected from inside damage have been used in addition to transmit-

ted and reflected waves

- 3 -

14 Organization

The organization of this thesis is summarized as follows

Chaper1 A general introduction including motivation literature review objective and

uniqueness of this study is described

Chaper2 Interactions between Lamb wave mode and damage in composites such as time

delay attenuation and multiple reflections are summarized Damage index (m-value) for

damage characterization is defined

Chaper3 Algorithms for the reference-free damage characterization technique using m-

value are developed Modified matching pursuit for this study is also described

Chaper4 Verification of the algorithm using numerical simulation is presented Various

modeled damage cases are characterized

Chaper5 Experimental verifications are described Two damage cases (Teflon inserted and

impact-induced damage) are investigated for the feasibility of the proposed damage charac-

terization technique Temperature effect test using temperature chamber is also carried out

Chaper6 Conclusion and summary of the present study is reported In advance suggestions

and future work recommendations are represented

- 4 -

CHAPTER 2 INTERACTION OF LAMB WAVE WITH DAMAGE

21 A0 Lamb wave mode delay and attenuation in the time domain

Time delay and attenuation of the A0 mode Lamb wave after passing through the damage

area due to effective thickness reduction were reported in previous researches [812] Moreo-

ver the attenuation of the A0 Lamb wave mode is occurred by degrading of shear modulus of

in the damage area However the time delay and amplitude variation of the A0 Lamb wave

mode also occurred even when the structure is health condition by the material property

changing due to temperature variation of the structure (Fig 1) Thus conventional damage

characterization methods using the time delay and attenuation of A0 Lamb wave mode as a

feature can cause false-alarm in real application

11 115 12 125 13-10

-5

0

5

10

Time (ms)

Vol

tage

(mV

)

IntactDamage

Reflected wave from Damage

(a)

11 115 12 125 13-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

0 oC20 oC50 oC

(b)

Fig 1 Lamb wave propagation characteristics (a) A0 Lamb wave mode time delay and atten-uation after passing through a damage (b) The time delay and amplitude variation of A0 Lamb wave mode under temperature variation in pristine condition of structure Because of this conventional damage detection techniques in composites have limitations for real struc-ture application

- 5 -

22 Multiple reflections in the damage area

Standing wave patterns due to multiple reflections of A0 Lamb wave mode in the damage

area were observed through numerical simulations and ultrasonic wave-field imaging using a

scanning laser vibrometer [16] The formation of standing waves in damage is illustrated

through Fig 2 When the incident waves enter the damage area a significant portion of the

waves are reflected back from the exit Then the reflected waves travel through the damage

and undergo reflection again at the original entrance As a consequence of multiple reflec-

tions at the entrance and the exit a considerable amount of ultrasonic energy is trapped inside

the damage area [17-19]

Multiple Reflections

Incident Wave Transmitted Wave

Damage

Fig 2 Multiple reflections in the damage area The multiple reflections of A0 Lamb waves in the damage area are utilize as a feature for characterization of the damage in this study

- 6 -

CHAPTER 3 THEORETICAL DEVELOPMENT

31 Overview

Damage

l = l1 + d + l2

d = vdtd l1 = vit1 l2 = vit2

t = t1 + td + t2

Excitation and Sensing PZT Sensing PZT

l

Transmitted Wave

Reflected Waves

Incident Wave

Fig 3 Parameters and notations of the proposed damage characterization technique Damage is detected by identifying the additional reflected wave from the damage due to multiple re-flections (dashed lines) Localization and quantification of the damage is carried out by measuring the arrival time of transmitted and reflected waves from the damage

In this section the reference-free damage characterization algorithm is proposed Damage

diagnosis and quantification are achieved by identifying the additional reflected waves due to

multiple reflections in the damage area (dashed lines in Fig 3) Moreover damage is local-

ized using additional information from pulse-echo method implement For the localization

and the quantification of damage the variables such as the path length and A0 Lamb wave

mode velocity in damage area are defined (Fig 3) For example 119897119897 is expressed as

119897119897 = 1198971198971 + 119889119889 + 1198971198972 (1)

where 1198971198971 119889119889 and 1198971198972 are the distance from the excitation PZTs to the entrance of damage the

damage length and the distance from the exit of damage to the sensing PZT respectively

- 7 -

Moreover 119907119907119894119894 and 119907119907119889119889 are also defined as the A0 Lamb wave mode velocity in pristine area

and damage area respectively Note that the whole procedure of the damage characterization

is accomplished based solely on the signals obtained from the current state of the structure

32 Damage index (m-value)

The proposed reference-free damage characterization technique utilizes the waves reflect-

ed additionally from the damage area as a feature The arrival time difference of the transmit-

ted and reflected A0 Lamb wave modes reflected from the exit and the entrance of the delam-

ination area one times each (∆119905119905) contains the information of the physical length of the dam-

age (119889119889) and A0 Lamb wave mode velocity in the damage area (119907119907119889119889) The wave velocity in the

damage area which is related to the shear modulus (G) represents the severity of the damage

∆119905119905 = 2119905119905119889119889 =2119889119889119907119907119889119889

(2)

However based on the measured data available the 119889119889 and 119907119907119889119889 (or G) cannot be uniquely

determined Fig 4 shows two FEM models which have different physical length and wave

velocity (G) of the damage Detail explanation of delamination modeling technique will be

covered in section 4 The similar signal in the time domain and ∆119905119905 are obtained from the

two different damage cases

- 8 -

Damage 1

Damage 2

20

Unit mm15

90 G value reduction

943 G value reduction

m-value = 248

m-value = 248

015 02 025 03 035

-2

-1

0

1

2

Time (ms)

Ampl

itude

Damage 1Damage 2

Fig 4 Two different FEM models which have different 119889119889 and 119907119907119889119889 (or G) This represents that the damage cannot be characterized uniquely based on the measured data available

Thus in this study the damage index (m-value) which can be uniquely defined based on

the available data is introduced The m-value is defined as the ratio of the physical length of

the damage and wavelength of the A0 Lamb wave mode in the damage area (120582120582119889119889)

119898119898 =119889119889120582120582119889119889

=119889119889119889119889119907119907119889119889

(3)

In many cases the transmitted and reflected waves are overlapped and it is hard to decom-

pose and calculate the arrival times of them Matching pursuit method is an alternative way

to accurately decompose of the overlapped waves and estimate the parameters of each wave

such as arrival time (Fig 5) Detail explanation of the matching pursuit method is covered in

section 36

015 02 025 03-2

-1

0

1

2

Time (ms)

Am

plitu

de

OriginalApproximated

(a)

015 02 025 03-2

-1

0

1

2

Time (ms)

Ampl

itude

Wave 1Wave 2

(b)

Fig 5 (a) Example of the overlapped signal with comparing approximated signal using the matching pursuit method (b) Decomposed signal using the matching pursuit method The matching pursuit method provides a conventional way to decompose and identify each wave

- 9 -

33 Damage detection algorithm (Level 1)

The damage diagnosis is accomplished by extracting the additional reflected A0 Lamb

wave modes from damage due to multiple reflections If A0 Lamb wave modes are decom-

posed perfectly it is trivial that one transmitted A0 Lamb wave mode is always captured in

the response signal If damage exists other waves in the response signal behind the transmit-

ted A0 Lamb wave mode are presented Since this approach does not rely on previously col-

lected data the diagnosis is independent from environmental variation such as temperature

changes

It is reported that the dispersion of Lamb waves cause wave-packets to spread out in space

and time as they propagate through a structure [20] Thus the reflected waves from the dam-

age have longer pulse duration than the transmitted wave due to their longer propagating dis-

tance In this study the matching pursuit method is modified to capture more spread wave

than the transmitted A0 Lamb wave mode Detail explanation of the modified matching pur-

suit method is described in latter section

Let 1199041199041 and 1199041199042 as the pulse duration (scale) of the transmitted and additional reflected A0

Lamb wave mode respectively

ldquoIf 1199041199042 gt 1199041199041 damage exists

Otherwise there is no damagerdquo

- 10 -

34 Damage quantification algorithm (Level 3)

As mentioned in the previous section damage is quantized by estimating 119889119889 and 119907119907119889119889 using

pitch-catch method The procedure to estimate 119889119889 and 119907119907119889119889 is

(i) Measure 119897119897 119905119905 and 119905119905119889119889

(ii) Assume the range of 119889119889 (0 lt 119889119889 lt 119897119897)

(iii) Calculate 119907119907119894119894 and 119907119907119889119889 with a constraints (0 lt 119907119907119889119889 lt 119907119907119894119894) (Eq 4 and 5)

(iv) Estimate m-value and find a feasible range of 119889119889 and 119907119907119889119889

119907119907119894119894 =119897119897 minus 119889119889119905119905 minus 119905119905119889119889

(4)

119907119907119889119889 =119889119889119905119905119889119889

(5)

35 Damage localization algorithm (Level 2amp3)

For damage localization 1198971198971 is estimated by measuring the arrival time from the excitation

PZT to the entrance of delamination (1199051199051) with additional pulse-echo method implementation

The procedure is similar with the previous damage quantification algorithm

(i) Measure 119897119897 119905119905 119905119905119889119889 and 1199051199051

(ii) Assume the range of 1198971198971 (0 lt 1198971198971 lt 119897119897)

(ii) Compute possible combinations of 1198971198972 119889119889 and 119907119907119889119889 with constraints

(0 lt 119907119907119889119889 lt 119907119907119894119894 119889119889 gt 0) (Eq 5 6 and 7)

- 11 -

1198971198972 = 1198971198971119905119905 minus 1199051199051 minus 119905119905119889119889

1199051199051 (6)

119889119889 = 119897119897 minus 1198971198971 minus 1198971198972 (7)

Since a PZT for pulse-echo method acts as both an actuator and a sensor the obtained sig-

nal contains both the excitation and the response signal Thus the reflected wave from the

entrance of damage can be buried into the tail of excitation signal and cause error on arrival

time of reflected wave from damage Even in this case damage can be localized and quan-

tized using the constraints When first wave packet reflected from the entrance of damage is

buried into the excitation signal the next wave packet reflected from the exit of damage can

be utilized alternatively If the algorithm regards the second reflected wave as the first re-

flected wave there are no combinations of estimated variable meeting the constraints Then

the algorithm repeats the process by regarding the wave as the second reflected wave the

constraints can select proper combination of the estimated variables

The estimated values have range bounded by the constraints When additional information

of 119907119907119894119894 is utilized the damage can be characterized more accurately In this study three meth-

ods are proposed and verified as information level of 119907119907119894119894

Method 1 Solely based on the current data measurement (Reference-free)

Method 2 Limit the possible range of 119907119907119894119894 (Prior knowledge of the system)

Method 3 Measure 119907119907119894119894 (Additional data need to be collected)

Method 1 described in this section can be utilized without any information of 119907119907119894119894 Meth-

od 2 produces more accurate estimation results using prior knowledge of the system such as

- 12 -

119907119907119894119894 variation under temperature changes Method 3 gives unique estimated value using meas-

ured 119907119907119894119894 from other path in the same time In section 4 and 5 damage is characterized using

these three methods and the results are compared each another

36 Matching pursuit method considering the dispersion effect of Lamb

wave

Matching pursuit method has been developed for signal compression approximation and

characterization of obtained signal and function [21-23] The key concept of the matching

pursuit method is that any signal (function) can be decomposed into a linear expansion of

signals selected from a pool of basis signals (functions) called lsquodictionaryrsquo

Let 119891119891 be a function to be analyzed then 119891119891 can be approximated a set of functions select-

ed from the dictionary 119863119863 in order to best match This is done by successive approximations

with orthogonal projections Let1198921198921205741205740 isin 119863119863 the function 119891119891 can be decomposed into

119891119891 = lang1198911198911198921198921205741205740rang1198921198921205741205740 + 119877119877119891119891 (8)

where 119877119877119891119891 is the residual function after approximating 119891119891 using 1198921198921205741205740 1198921198921205741205740 is orthogonal to

119877119877119891119891 hence

1198911198912 = lang1198911198911198921198921205741205740rang2

+ 1198771198771198911198912 (9)

To minimize 119877119877119891119891 1198921198921205741205740 is chosen to maximize lang1198911198911198921198921205741205740rang The matching pursuit method is

an iterative algorithm that decomposes the residue 119877119877119891119891 by projecting it on a function of 119863119863

that matches 119877119877119891119891 almost at best as it was done for 119891119891 The residue 119877119877119899119899119891119891 is decomposed into

119877119877119899119899119891119891 = lang119877119877119899119899119891119891119892119892120574120574119899119899rang119892119892120574120574119899119899 + 119877119877119899119899+1119891119891 (10)

- 13 -

the decomposition is carried up to the order 119898119898 119891119891 can be expressed as

119891119891 = lang119877119877119899119899119891119891119892119892120574120574119899119899rang119892119892120574120574119899119899

119898119898minus1

119899119899=0

+ 119877119877119898119898119891119891 (11)

The original function 119891119891 is decomposed into a sum of dictionary elements that are chosen to

best match its residues

In this study the matching pursuit method is utilized to accurately decompose of the over-

lapped Gaussian pulses and estimate the parameters of the pulse which used for damage

characterization The 119894119894119905119905ℎ Gaussian pulse in the dictionary for approximating the obtained

signal is expressed as

119892119892120574120574119894119894 = 119891119891(119860119860119894119894 119906119906119894119894 119904119904119894119894 119889119889119894119894120601120601119894119894) (12)

where 119860119860119894119894119906119906119894119894119904119904119894119894119889119889119894119894 and 120601120601119894119894 represent amplitude arrival time scale (pulse duration) excita-

tion(central) frequency and phase of the Gaussian pulse respectively The parameters of the

pulse are estimated using a nonlinear regression based optimization method (Gauss-Newton

method) based on an initial guess of the parameters Detail procedure for estimating the pa-

rameters of the pulse is expressed in Hong et al [23]

Among the estimated parameters of the pulses pulse duration (scale) represents the disper-

sion effect If a wave-packet is spread out the estimated pulse duration is increased Using

this feature the matching pursuit method is modified to capture waves which have longer

pulse duration than transmitted A0 Lamb wave mode If the estimate pulse duration of a wave

is less than the transmitted wave the modified matching pursuit method regards it as error

- 14 -

CHAPTER 4 NUMERICAL SIMULATION

41 Composite modeling method

Composite materials can be modeled from nano to macro level depends on its engineering

purpose In general four methods are utilized for modeling of composite materials First

composite is modeled as a homogenous material expressed in terms of effective material

properties [24] The effective material properties are estimated using the modulus and the

volume fraction of each constituent theoretically (Rule of mixture) and measured experimen-

tally Second a model is established that each laminate which have own material properties

are totally bonded together In this model the stacking sequence and direction of each lami-

nates are considered additionally [25-27] Third method for composite modeling in micro

level called micromechanics [28-30] This model considers the interaction between fibers and

matrix in laminate Last multiscale modeling which the scope of modeling is extended to

molecular or quantum level of constituents [31-33] This model starts from molecular me-

chanics or molecular dynamics then the scale of model leads to the nano micro and macro

level by incorporating small scale models As the scale of models is smaller it provides more

logical way than other large scale of models while the computational cost is more expensive

The scope of this study is focused on the dynamic behavior of composite structure the beam

or plate model was utilized for the numerical simulation

- 15 -

42 Damage modeling method

Generally three damage models have been developed and utilized in the real applications

First the stiffness reduction model called as the equivalent element model was introduced

[17 34-35] This model is simply realized by replacing the region around damage area by a

region with adjusted material properties (Youngrsquos modulus shear modulus density etc)

globally Next the duplicated node model is done by disconnecting the element connectivity

assuming there is no interaction among the delaminated surface This model can be utilized

when the location and the extension of damage is important issue [34-36] Last the contact

mechanics model using the principles of contact mechanics the two interfacial forces (con-

tact resistance and sliding resistance) in the delaminated surface [37-38] In this model the

delaminated surfaces are modeled as nonlinear functions of the interlaminar normal strain

and the interlaminar shear strain This model is computationally expensive while provides

more logical way than other models There is however no universally accepted model for

general impact-induced damage in composites In the case of impact-induced damage it is

impossible to predict or estimate the location and number of delaminated layers Thus from

the uncertainty of impact-induced damage the equivalent element model can be considered

as the effective damage model

- 16 -

43 Numerical simulation description

TABLE I Material property of simulation model and test specimen

Tensile Strength 657 MPa Tensile Modulus 59 GPa Shear Strength 804 MPa Shear Modulus 24 Gpa Specific Gravity 16

PZT A

PZT C

PZT BDamage

3

250 125 125

500 Unit mm

Fig 6 Description of the numerical simulation model

The feasibility of the proposed damage characterization technique is verified by 2-D FEM

simulation using MSC PATRAN and NASTRAN Fig 6 presents the dimension PZTs and

damage location of the simulation model In the model 05mm x 05mm square shell element

was used for modeling of the A0 mode whose flexural wave characteristic The element size

was determined by considering the wavelength of the propagation Lamb wave modes The

material properties of the simulation model and damage cases are presented in Table I and

Table II respectively

PZT A and C act as both actuator and sensor for collecting pulse-echo signal They were

installed on the plate which two of them are both-side collocated for the A0 Lamb wave

- 17 -

mode decomposition [39-40] PZT B was only used as a sensor for collecting pitch-catch sig-

nal For simulating the excitation and sensing of PZTs the pin-force model was adopted that

an input force apples to the end nodes of excitation PZT and averaging element strains corre-

sponding sensing area A Gaussian pulse (tone-burst) 50 kHz central frequency was excited

as an input signal

TABLE II Damage cases for numerical simulation

119949119949120783120783 (mm) 119941119941 (mm) G value reduction ()

Case 1 650 200 90 darr

Case 2 675 150 90 darr

Case 3 675 150 93 darr

Case 4 900 150 90 darr

- 18 -

44 Numerical simulation results

015 02 025 03 035

-2

0

2

Time (ms)

Am

plitu

de

IntactDamage

Reflected wave from damage

(a) Pitch-catch signal

015 02 025 03 035

-2

0

2

Time (ms)

Am

plitu

de

IntactDamage

Reflected waves from damage

(b) Pulse-echo signal Fig 7 Comparison of A0 Lamb wave mode signals from numerical simulation between intact and damage cases Time delay attenuation and multiple reflections due to damage simulated successfully using the damage modeling method described in the previous section

Fig 7 shows that comparison of A0 Lamb wave mode signals between intact and one of

damage cases in the simulation The features of A0 Lamb wave mode passing through dam-

age area described in section 2 are successfully simulated Thus the equivalent element

method is proper for modeling damage in composites In the damage cases the additional

waves reflected from damage is observed in both pitch-catch and pulse echo signal

Table III presents the result of the proposed damage characterization algorithm using the

numerical simulation cases shown in Table II In lsquoMethod 2rsquo 119907119907119894119894 variation is assumed to plusmn

5 of the normal value (1190~1315ms) The normal value of 119907119907119894119894 for lsquoMethod 3rsquo is meas-

ured as 1253 ms In each case the damage is characterized using the proposed technique

Note that as the information of 119907119907119894119894 is specified the estimated value is closer to the exact

value

- 19 -

TABLE III Damage characterization for the numerical simulation models

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

Case 1

Exact 650 200

Method 1 248 50~70 79~414 160~834

Method 2 248 603~667 135~241 271~486

Method 3 248 635 188 379

Case 2

Exact 675 150

Method 1 193 60~75 01~251 37~651

Method 2 193 618~682 114~222 295~575

Method 3 193 65 168 435

Case 3

Exact 675 150

Method 1 233 60~75 66~302 140~649

Method 2 233 668~738 85~196 182~419

Method 3 233 703 140 301

Case 4

Exact 900 150

Method 1 204 75~100 48~349 118~856

Method 2 204 860~951 107~216 264~530

Method 3 204 906 162 397

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1253 ms 1190~1315 ms Measured 119907119907119894119894 1253 ms

- 20 -

CHAPTER 5 EXPERIMENTAL VERIFICATION

51 Overview

In this section experimental study is investigated so that the proposed damage characteri-

zation technique Overall experimental setup and test specimen are shown in Fig 8 and 9

The experimental setup consists of test specimens data acquisition (DAQ) system and self-

sensing circuit Self-sensing circuit was installed for pulse-echo measurement using PZTs

[40-41]

PZT A(PZT C) PZT B

500

500

140

75

250

Teflon Inserted Damage (d =25)

Unit mm

Thickness = 3

PZT A(PZT C) PZT B

500

500

150

90

250

Impact Location

Unit mm

Thickness = 3

(a) Teflon inserted damage specimen

(b) Impact-induced damage specimen

Fig 8 Test specimens for the experimental verification In the Teflon inserted damage Teflon tapes with 25mm diameter were inserted at the center of the specimen The impact-induced damage was introduced to the specimen using the impact test equipment

- 21 -

511 Specimen

Two carbon composite plates with a dimension of 500times500times3mm3 were manufactured by

Nexcomes Inc by stacking 12 layers of CF3327 3K prepreg (woven fabrics type) manufac-

tured by Hankuk Fiber Inc The material properties of the specimen are presented in Table I

One of the specimens Teflon tapes with 25mm diameter was inserted in every 3 layers at the

center of the specimen to mimic the damage The other specimen is utilized for the character-

ization of real impact-induced damage

512 Data acquisition (DAQ) system and PZT

The data acquisition (DAQ) system consists of an arbitrary waveform generator (AWG) a

high speed signal digitizer (DIG) and two multiplexers (MUX) Using the 16-bits AWG

tone-burst input signal with a plusmn10 peak-to-peak voltage were generated The response signal

was measured by the 14-bit DIG For the signal-to-noise ratio reduction both pitch-catch and

pulse-echo signals were measured 150 times and averaged in the time domain 3 identical

circular PZTs of 10mm diameter manufactured by APC International Ltd were installed on

the specimen A Gaussian pulse (tone-burst) 50 kHz central frequency was excited as an in-

put signal The excitation frequency was chosen in order to avoid overlapping of the S0 and

A0 Lamb wave modes and generating of the higher modes (S1 A1hellip)

- 22 -

PZT Actuator

PZT Sensor

AWG

DIG

DAQ System Specimen

SensingActuatingMUX

1

23

4

1 Sending signal to MUX

4 Sending signal to DIG

Self-SensingCircuit

18 Unit mm

10

(a)

(b)

Fig 9 Overall experimental setup (a) Data acquisition (DAQ) system and (b) Installed circu-lar PZT transducer on the specimen with its dimension

513 Impact test

The impact-induced damage was introduced to the specimen using the impact test equip-

ment presented in Fig 10 A 5kg mass with 10mm tip was dropped from 015m height onto

the same location of the specimen The signals were obtained at each number of impacts

Since the reflected wave is not detected at the first impact case the data obtained from the

second and third impact were utilized for the analysis

10

Unit mm (a)

(b)

Fig 10 (a) Impact test equipment for introducing impact-induced damage and (b) Tip of the impact tester with 10mm diameter

- 23 -

514 Temperature effect test

In reality structures are subjected to changing environmental conditions such as tempera-

ture that affect measured signal and cause false-alarm Thus the robustness of the proposed

damage characterization technique under temperature variation in investigated (Fig 11) The

specimens were placed inside a temperature chamber and a thermocouple was installed on

the specimen to measure temperature of surface of the specimen The signal was obtained

under three temperature cases (0 20 and 50degC) and the humidity was kept at 30

Temperature Chamber

Test Specimen

Fig 11 Test specimen in temperature chamber

- 24 -

52 Teflon inserted damage

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Reflected wave from damage

(a)

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Transmitted WaveReflected Wave

Detected reflected wave

(b)

Fig 12 A0 Lamb wave mode obtained using pitch-catch method from the Teflon inserted damage specimen (a) The reflected wave from damage is shown (b) The Teflon inserted damage is detected by capturing the reflected wave

Fig 12 shows the obtained pitch-catch signal from the Teflon inserted damage specimen

and damage diagnosis result The Teflon inserted damage is detected successfully by captur-

ing the reflected wave Table IV presents the result of the Teflon inserted damage characteri-

zation 119907119907119894119894 is measured as 1213ms and the variation is assumed to plusmn 5 of the normal val-

ue (1152~1273ms) The results indicate that the proposed damage characterization approach

works well in real application

TABLE IV Damage characterization for the Teflon inserted damage specimen

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

Exact 635 25

Method 1 123 60~75 73~338 298~1220

Method 2 123 619~662 229~310 934~1188

Method 3 123 631 284 1161

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1213 ms

- 25 -

53 Impact-induced damage

11 115 12 125 13-4

-2

0

2

4

Time (ms)

Vol

tage

(mV

)

IntactImpact2Impact3

Reflected wave from damage

(a) Pitch-catch signal

11 115 12 125 13 135

-05

0

05

1

Time (ms)

Volta

ge (m

V)

IntactImpact2Impact3

Boundary reflection

Reflected wave from damage

(b) Pulse-echo signal

Fig 13 The signal obtained from the test specimen subjected to impact The impact was in-troduced at the same location of the specimen and the signal was obtained after each impacts As the number of impact is increased the transmitted A0 Lamb wave mode is delayed and attenuated

The obtained signals from impact-induced damage specimen are presented in Fig 13 It is

observed that as the number of impact is increased the transmitted A0 Lamb wave mode is

more delayed because the wave velocity in the damage area is reduced Fig 14 shows the

false-alarm test result whether the proposed damage detection algorithm works even in pris-

tine condition of the test specimen It is shown that no false alarm is occurred because no ad-

ditional reflected wave is captured

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Transmitted Wave

(a) Transmitted wave

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Reflected Wave

No reflected wave

(b) Reflected wave

Fig 14 False alarm test for the proposed damage detection algorithm in pristine condition of the specimen Reflected wave from damage is not captured false alarm is not occurred

- 26 -

TABLE V Damage characterization for the impact-induced damage specimen

of Impacts m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

2

(1470 J)

Method 1 080 65~90 75~246 470~1157

Method 2 080 801~885 114~245 711~1134

Method 3 080 843 179 1122

3

(2205 J)

Method 1 148 50~100 06~380 21~1134

Method 2 148 758~838 249~368 843~1125

Method 3 148 798 308 1044

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1213 ms

Table V presents the characterization results of the impact-induced damage case Here the

measured 119907119907119894119894 and the assumed variation are same as the artificial damage case It is observed

that as the number of impact is increased 119889119889 is increased while 119907119907119889119889 is reduced Thus it is

validated that the proposed damage characterization approach estimates the increment of the

damage well

54 Temperature effect test

541 Teflon inserted damage

11 115 12 125 13-10

-5

0

5

10

Time (ms)

Vol

tage

(mV

)

0 oC20 oC50 oC

(a) Pitch-catch signal

11 115 12 125 13 135

-1

0

1

2

3

Time(ms)

Volta

ge (m

V)

0 oC20 oC50 oC

(b) Pulse-echo signal

Fig 15 Signals obtained from the Teflon inserted damage specimen under temperature varia-tion Time delay and amplitude variation are occurred

Fig 15 shows the obtained signals from the Teflon inserted damage specimen subjected to

- 27 -

temperature variation Time delay and amplitude variations are occurred there is no addition-

al damage is introduced The analyzed results of the obtained signals are reported in Table VI

The measured 119907119907119894119894 is 1247 ms at 0 degC 1213 ms at 20 degC and 1168 ms at 50 degC respectively

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms (1152~1273 ms)

The estimated results are also changed under temperature variations the fluctuation of esti-

mated range and exact value is small

TABLE VI Temperature effect test for the Teflon inserted damage specimen

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

0 degC

Exact 635 25

Method 1 113 60~70 79~268 352~1291

Method 2 113 560~619 233~344 1034~1172

Method 3 113 606 257 1141

20 degC

Exact 635 25

Method 1 123 60~75 73~338 298~1220

Method 2 123 619~662 229~310 934~1188

Method 3 123 631 284 1161

50 degC

Exact 635 25

Method 1 137 60~75 72~328 263~1228

Method 2 137 622~687 183~299 666~1086

Method 3 137 642 264 960

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1247 ms at 0 degC 1213 ms at 20 degC and 1168 ms at 50 degC

- 28 -

542 Impact-induced damage

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Vol

tage

(mV

)

Transmitted WaveReflected Wave

No reflected wave

(a) 0 degC

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Transmitted WaveReflected Wave

No reflected wave

(b) 50 degC

Fig 16 False alarm test under temperature variations It is shown that proposed damage char-acterization algorithm is robust under temperature variations

The false alarm test in pristine condition of the impact-induced damage specimen under

temperature variation was investigated (Fig 16) It is validated that the proposed damage di-

agnosis algorithm is robust under temperature variations Fig 17 shows the obtained signals

from the impact-induced damage (impacted 3 times) specimen under temperature variation

The analyzed results of the obtained signals are reported in Table VII The measured 119907119907119894119894 in

each temperature and the assumed variation are same as the artificial damage case

11 115 12 125 13-5

0

5

Time (ms)

Volta

ge (m

V)

0 oC20 oC50 oC

(a) Pitch-catch signal

11 115 12 125 13 135

-1

0

1

2

3

Time(ms)

Vol

tage

(mV

)

0 oC20 oC50 oC

(b) Pulse-echo signal

Fig 17 Signals obtained from impact-induced damage specimen (Impact 3) under tempera-ture variation

- 29 -

TABLE VII Temperature effect test for the impact induced damage specimen (Impact 3 case)

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

0 degC

Method 1 135 75~90 72~310 267~1148

Method 2 135 705~779 264~381 978~1126

Method 3 135 763 289 1071

20 degC

Method 1 148 50~100 06~380 21~1134

Method 2 148 758~838 249~368 843~1125

Method 3 148 798 308 1044

50 degC

Method 1 153 75~100 53~415 174~1227

Method 2 153 804~889 214~336 702~1103

Method 3 153 815 320 1050

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1247 ms at 0 degC 1213 ms at 20 degC and 1168 ms at 50 degC

- 30 -

CHAPTER 6 CONCLUDING REMARKS

61 Executive summary

In this study the reference-free damage detection localization and quantification tech-

nique in a multi layered composite plate is proposed Additional reflected A0 Lamb wave

mode from the damage due to multiple reflections in damage area is used as a feature for

characterizing the damage First damage is characterized in terms of a damage index (m-

value) defined as the ratio of the physical length of the damage to the wavelength of the A0

Lamb wave mode Then the range of possible damage location length and severity are esti-

mated with constraints This procedure can be accomplished without prior baseline data To

realize the proposed technique the matching pursuit technique is modified for reduction of

the error by considering the dispersive characteristic of Lamb wave The matching pursuit

method estimates the arrival time of each wave packets accurately even two wave packets are

overlapped The numerical simulations and experimental tests including Teflon tape inserted

and impact-induced damage are investigated for the feasibility of the proposed technique

Moreover the effectiveness of the proposed technique under changing temperature is exper-

imentally verified

- 31 -

62 Suggestion for further study

The limitations of this study and the possible improvement and further studies are listed as

follow

1 Measurement of impact-induced damage length using conventional NDT technique

Additional experimental verification for impact-induced damage characterization is

needed by measuring the real length of damage using conventional NDT technique

such as ultrasonic scan or infrared camera

2 Limitation of both side collocated PZT installation In this study excitation PZTs are

installed both side and collocated for decomposing A0 Lamb wave mode However it

can produce decomposition error due to imperfection of PZT location and their installa-

tion Moreover in many cases assess to both surface of structures is not allowed such

as aircraft applications To improve this limitation a reference-free damage characteri-

zation using single side PZTs will be considered

3 Complex structure application In this study damage is characterized in simple plate-

like structure When the monitored structure becomes complex the response signals al-

so becomes complex due to its geometry and boundary Thus the investigations for

complex structure application such as stiffener spar and curvature are needed

- 32 -

REFERENCES

[1] Cawley P and Alleyne D (1996) ldquoThe use of Lamb waves for the long range inspection of large structuresrdquo Ultrasonics 34 pp287-290

[2] Ghosh T Kundu T and Karpur P (1998) ldquoEfficient use of Lamb waves for detecting defects in large platesrdquo Ultrasonics 36 pp791-801

[3] Guirguitui V (2003) ldquoLamb wave generation with piezoelectric wafer active sensors for structural health monitoringrdquo SPIE Conference on Smart Structures and Materials San Diego CA 5391 pp111-422

[4] Ragvahan A and Cesnik CE S (2004) ldquoModeling of piezoelectric-based Lamb wave generation and sensing for structural health monitoringrdquo SPIE Conference on Smart Struc-tures and Materials San Diego CA 5391 pp419-430

[5] Rose JL Pilarski A and Dirit JJ (1993) ldquoAn approach to guided wave mode selection for inspection of laminated platerdquo Journal of Reinforced Plastics and Composites 12(5) pp536-544

[6] Guo N and Cawley P (1993) ldquoThe interaction of Lamb waves with delaminations in composite laminatesrdquo Journal of Acoustical Society of America 94(4) pp 2240-2246

[7] Valdez SHD and Soutis C (2002) ldquoReal-time nondestructive evaluation of fiber com-posite laminates using low-frequency Lamb wavesrdquo Journal of Acoustical Society of Amer-ica 111(5) pp 2026-2033

[8] Yang M and Qiao P (2005) ldquoModeling and experimental detection of damage in virous materials using the pulse-echo method and piezoelectric sensoractuatorsrdquo Smart Materials and Structures 14 pp1083-1100

[9] Lestari W and Qiao P (2005) ldquoApplication of wave propagation analysis for damage identification in composite laminated beamsrdquo Journal of Composite Materials 39(22) pp1967-1984

[10] Kessler S S Spearing S M and Soutis C (2002) ldquoDamage detection in composite ma-terials using Lamb wave methodsrdquo Smart Materials and Structures 11 pp269-278

[11] Sohn H Park G Wail J R Limback N P and Farrar C R (2004) ldquoWavelet-based active sensing for delamination detection in composite structuresrdquo Smart Materials and Structures 13 pp 153-160

[12] Petculescu G Krishnaswamy S and Achenback J D (2008) ldquoGroup delay measure-ments using modally selective Lamb wave transducers for detection and sizing of delamina-tions in compositesrdquo Smart Materials and Structures 17

[13] Sohn H Park H W Law K H and Farrar C R (2007) ldquoDamage detection in compo-

- 33 -

site plates by using an enhanced time reversal methodrdquo Journal of Aerospace Engineering (ASCE) pp 141-151

[14] Sohn H (2007) ldquoEffects of environmental and operational variability on structural health monitoringrdquo Philosophical Transactions of the Royal Society Ardquo 365 pp539-560

[15] Rytter A (1993) ldquoVibration based inspection of civil engineering structuresrdquo PhD The-sis Department of Building Technology and Structural Engineering University of Aalborg Aalborg Denmark

[16] Hayashi T and Kawashima K (2002) ldquoMultiple reflections of Lamb waves at a delami-nationrdquo Ultrasonics 40 pp193-197

[17] Wang C H and Rose L R F (2003) ldquoWave reflection and transmission in beams con-taining delamination and inhomogeneityrdquo Journal of Sound and Vibration 264 pp 851-872

[18] Ramadas C Balasubramaniam K Joshi M and Krishnamurthy C V (2009) ldquoInterac-tion of the primary anti-symmetric Lamb mode (Ao) with symmetric delaminations numer-ical and experimental studiesrdquo Smart Materials and Structures 18

[19] Ramadas C Balasubramaniam K Joshi M and Krishnamurthy C V (2010) ldquoInterac-tion of guided Lamb waves with an asymmetrically located delamination in a laminated composite platerdquo Smart Materials and Structures 19

[20] Wilcox P Lowe M and Cawely P (2001) ldquoThe effect of dispersion on long-range in-spection using ultrasonic guided wavesrdquo NDTampE International 34 pp1-9

[21] Mallat S G and Zhang Z (1993) ldquoMatching pursuits with time-frequency dictionariesrdquo IEEE Tansactions on Signal Processing 41(12) pp3397-3415

[22] Zhang G Zhang S and Wang Y (2000) ldquoApplication of adaptive time-frequency de-composition in ultrasonic NDE of highly scattering materialsrdquo Ultrasonics 38 pp 961-964

[23] Hong J C Sun K H and Kim Y Y (2005) ldquoThe matching pursuit approach based on the modulated Gaussian pulse for efficient guided-wave damage inspectionrdquo Smart Mate-rials and Structures 14 pp548-560

[24] Herakovich C T (1997) Mechanics of Fibrous Composites Wiley 480 pages

[25] Reddy J N (1985) ldquoA review of the literature on finite-element modeling of laminated composite platesrdquo The Shock and Vibration Digest 17(4) pp 3-8

[26] Reddy J N (2003) Mechanics of Laminated Composite Plates and Shells Theory and Analysis CRC Press 856 pages

[27] Palacz M Krawczuk M and Ostachowicz W (2005) ldquoThe spectral finite element mod-el for analysis of flexure-shear coupled wave propagation Part 1 laminated multilayer

- 34 -

composite beamrdquo Composite Structures 68 pp37-44

[28] Zhang Y C and Harding J (1990) ldquoA numerical micromechanics analysis of the me-chanical properties of a plain weave compositerdquo Computer amp Structures 36(5) pp 839-844

[29] Nemat-Nasser S and Hori M (1999) Micromechanics Overall Properties of Heteroge-neous Materials Elsevier 810 pages

[30] Scida D Aboura Z Benzeggagh M L and Bocherens E (1999) ldquoA micromechanics model for 3D elasticity and failure of woven-fibre composite materialrdquo Composites Science and Technology 59 pp 505-517

[31] Thomas Y and Wu X H (1997) ldquoA multiscale finite element method for elliptic prob-lems in composite materials and porous mediardquo Journal of Computational Physics 134 pp 169-189

[32] Lee K H Moorthy S and Ghosh S (1999) ldquoMultiple scale computational model for damage in composite materialsrdquo Computer Methods in Applied Mechanics and Engineer-ing 172 pp 175-201

[33] Namilae S and Chandra N (2005) ldquoMultiscale model to study the effect of interfaces in carbon nanotube-based compositesrdquo Journal of Engineering Materials and Technology (ASME) 127 pp222-232

[34] Adams D E (2007) Health monitoring of Structural Materials and Components Wiley 460 pages

[35] Gopalakrishnan S (2009) ldquoModeling Aspects in Finite Elementrdquo Encyclopedia of Struc-tural Health Monitoring 2 pp791-809

[36] Palacz M Krawczuk M and Ostachowicz W (2005) ldquoThe spectral finite element mod-el for analysis of flexure-shear coupled wave propagation Part 2 Deminated multilayer composite beamrdquo Composite Structures 68 pp45-51

[37] Baik J M and Thopson R B (1984) ldquoUltrasonic scattering from imperfect interfaces A quasi-static modelrdquo Journal of Nondestructive Evaluation 4(3) pp177-196

[38] Luo H and Hanagud S (2000) ldquoDynamics of delamincated beamsrdquo International Jour-nal of Solids and Structures 37 pp1501-1519

[39] Kim S B and Sohn H (2007) ldquoInstantaneous reference-free crack detection based on polarization characteristics of piezoelectric materialsrdquo Smart Materials and Structures 16 pp2375-2387

[40] An Y K and Sohn H (2010) ldquoInstantaneous crack detection under varying temperature and static loading conditionsrdquo The Special Issue of Structural Control and Health Monitor-ing 17(7) pp 730-741

- 35 -

[41] Lee S J and Sohn H (2006) ldquoActive self-sensing scheme development for structural health monitoringrdquo Smart Materials and Structures 15(6) pp 1734-1746

- 36 -

요 약 문

복합재료 손상의 무기저 진단 및 정량화

복합재료는 재료의 경량성 내구성 고강도의 특성으로 인하여 항공기에서부터

교량에까지 널리 사용되고 있다 하지만 외부의 충격으로 인한 복합재료 내부의

손상은 재료의 강도를 감소시켜 구조물의 안정성을 저하시키는 요인이 된다 하지만

복합재료의 손상은 재료 내부에서 발생하여 육안으로는 식별하기도 힘들며 표면에서

보이는 것만으로 내부의 손상 정도를 판단하기는 불가능하다는 단점을 가지고 있다

본 연구에서는 손상 내부에서 반사로 인해 발생하는 파를 이용하여 기저자료를

사용하지 않고 특정 시점에서 얻은 자료만으로 복합재료의 손상을 진단하고 정량화

할 수 있는 기법을 제안한다 제안된 기법의 적용성을 검증하기 위하여 유한요소법을

이용한 수치해석을 수행하였고 Teflon tape 이 삽입된 시편과 실제 충격하중으로

인한 손상을 입은 시편을 이용한 실험을 진행하였다 수치해석 및 실험 결과 제안된

기법이 복합재료의 손상을 성공적으로 진단 및 정량화할 수 있음을 확인하였다 또한

다양한 외부 온도 환경에서 제안된 기법의 환경영향 실험을 수행하였고 다양한

환경에서 손상의 진단 및 정량화가 성공적으로 이루어짐을 검증하였다

핵심어 무기저 손상진단 및 정량화 압전센서 복합재료 유한요소법 Matching pursuit

method

- 37 -

감 사 의 글

기대와 두려움을 안고 시작한 석사과정이 어느덧 2 년 이라는 세월 속에서 추억으로 자리잡아

가고 있습니다 학부생 기간을 포함한 지난 3 년 동안 SHM 이라는 새로운 분야를 접하고 연구를

수행 할 수 있도록 도와주신 고마운 분들이 있어 그 감사를 표하고자 합니다

우선 저의 많은 부족한 점을 일깨워 주시고 연구에 대한 뜻을 더욱 확고히 세울 수 있게 해

주신 지도교수님이신 손 훈 교수님께 깊은 감사의 뜻을 전합니다 짧은 기간이었지만 교수님의

도움으로 많은 것을 배우고 느끼고 깨달을 수 있었습니다 또한 본 연구를 수행할 수 있도록 지

원해주신 국토해양부 (U-City 석 박사과정 지원사업)에 깊은 감사 드립니다

때로 힘들고 지칠 때 항상 저를 믿고 응원해 주시는 가족이 있었기에 지금의 글을 쓸 수 있었

습니다 또한 지금 연구실에서 함께 생활하고 있는 윤규형 민구형 철민이형 현석이형 진열이형

호민이형 병진이 선혜누나와 성대에 계신 창길이형 미국에 계신 의재형과 Debaditiya 에게도 감

사의 말씀을 전합니다 그리고 모두 나열할 수는 없지만 카이스트 생활의 활력소가 되어준 카이

스트 건설 및 환경공학과 선후배들 직원 분들과 SOS 멤버들에게도 감사를 표합니다

앞서 거론한 많은 이들의 도움이 없었다면 저 또한 지금 이 자리에 있을 수 없었을 것입니다

많이 부족하기에 더 노력하고 배우고자 했던 초심을 다시 한번 생각하며 더 큰 결실을 맺기 위해

앞으로 나아가겠습니다 도움을 주신 모든 분들께 이 감사의 글을 바칩니다

2011년 6월 카이스트에서 임형진 올림

- 38 -

CURRICULUM VITAE

Hyung Jin Lim

Research Assistant

The Civil and Environmental Engineering Department

Korea Advanced Institute of Science and Technology (KAIST)

291 Daehak-Ro Guseong-Dong Yuseong-Gu

Daejeon Republic of Korea 305-701

Phone (82)+42-350-3665 Fax (82)+42-350-3610

Email limhj87gmailcom

RESEARCH INTERESTS

Damage Detection Localization and Quantification on Composite and Metallic Structures Pattern Recognition Finite Element Modeling (FEM)

EDUCATION

2009 ndash 2011 MS Civil and Environmental Eng Korea Advanced Institute of Science and Technology (KAIST) Korea Rep

2006 ndash 2009 BS Civil and Environmental Eng Korea Advanced Institute of Science and Technology (KAIST) Korea Rep

JOURNAL PUBLICATIONS The corresponding author is underlined 1 Hyung Jin Lim Hoon Sohn ldquoReference-free damage detection localization and quantification

using in compositerdquo in preparation for Journal of the Acoustical Society of America 2011

2 Chul Min Yeum Hoon Sohn Jeong Bum Ihn Hyung Jin Lim ldquoInstantaneous delamination detection in a composite using a dual piezoelectric transducer networkrdquo Submitted to Compo-sites Science and Technology 2011

3 Yun-Kyu An Hyung Jin Lim Min Koo Kim Jin Yeol Yang Hoon Sohn Chang Guen Lee ldquoInvestigation of applicability of local reference-free damage detection techniques to in-situ

- 39 -

bridgesrdquo Submitted to ASCE Journal of Structural Engineering 2011

4 Hyung Jin Lim Min Koo Kim Hoon Sohn Chan Yik Park ldquoImpedance based damage detec-tion under varying temperature and loading conditionsrdquo Submitted to NDTampE International 2011

5 Hyung Jin Lim Hoon Sohn ldquoQuantitative estimation of transmitted and reflected lamb waves at discontinuityrdquo Journal of the Korean Society for Nondestructive Testing Vol30 No 4 pp1-8 2010

CONFERENCE PROCEEDINGS

The corresponding author is underlined 1 Hyung Jin Lim and Hoon Sohn ldquoReference-free damage detection localization and quantifi-

cation in compositesrdquo ASCE Engineering Mechanics Institute Conference Boston June 2-4 2011

2 Hyung Jin Lim and Hoon Sohn ldquoReference-free delamination quantification within multi-layer composite materialsrdquo KSNT Spring Conference May 26-27 Gyeong-Ju Korea 2011

3 Min Koo Kim Hyun Jin Lim Hoon Sohn Chan Yik Park ldquoImpedance-based bolt loosening detection under varying temperature and loadingrdquo Asian Pacific Workshop on Structural Health Monitoring Tokyo Japan November 30-December 2 2010

4 Hyung Jin Lim Chul Min Yeum and Hoon Sohn ldquoModeling of Impact-induced Delamination in a Multilayer Composite Platerdquo The 23rd KKCNN Symposium on Civil Engineering Taipei (Taiwan) November 13-15 2010

5 Yun-Kyu An Hyung Jin Lim and Hoon Sohn ldquoTemperature independent Delamination Detec-tion using data normalizationrdquo The 3rd International Conference on Multi-Functional Materi-als and Structures Chonbuk National Univ Jeonju Korea September 14-18 2010

6 Hoon Sohn Yun-Kyu An Chang Gil Lee Min Koo Kim Hyung Jin Lim ldquoApplication of online NDT techniques to decommissioned bridge testingrdquo the 5th European Workshop on Structural Health Monitoring Sorrento Italy June 29-July 02 2010

7 Min Koo Kim Hyung Jin Lim Hoon Sohn Chan Yik Park ldquoImpedance based damage diag-nosis of a complex composite aircraft wing under changing loading and temperature condi-tionsrdquo the 5th European Workshop on Structural Health Monitoring Sorrento Italy June 29-July 02 2010

8 Yun-Kyu An Min Koo Kim Hyung Jin Lim and Hoon Sohn ldquoValidations of reference-free crack detection techniques through a decommissioned bridge testrdquo 2010 COSEIK Annual Con-ference Jeju Korea April 8-9 2010

- 40 -

PARTICIPANT PROJECTS

1 Development of Smart Steel Members for Online Steel Structure Monitoring POSCO Corporation (Funded 41000000 KRW (41000 USD) for 010511 to 123111)

2 Development of Enabling Technologies for Structural Health Monitoring Korean Agency for Defense Development (Funded 240000000 KRW (240000 USD) out of total 3200000000 KRW (3200000 USD) for 060108-053111)

3 Development of On-board SHM Technologies for Composite Air Vehicles The Boeing Company (Funded 400000 USD (400000000 KRW) for 080108 to 073111)

EDUCATIONAL EXPERIENCES

Fall 2009

Teaching Assistant CE207 ldquoElementary Structural Engineering and Laboratoryrdquo KAIST 3 unit undergraduate course

EXPERIMENTAL EXPERIENTCES

2010 Monitoring of Real Aircraft Composite Wing Segment Korea Agency for Defense De-velopment (ADD) Daejeon Korea Real time monitoring of composite air vehicles fund-ed by Korea Agency for Defense Development (ADD)

2009-2011

Field Monitoring of Real Bridge Structures Gimpo (Ramp G bridge) amp Icheon (Sam-seung bridge) Korea US-Korea collaborative research for bridge monitoring testbeds

HONORS amp AWARDS

Scholarships for outstanding students granted KAIST (2007-2009) The three prominent students in the civil and environmental engineering department of KAIST were selected as recipients This scholarship provided 2000$ per year for three years

- 41 -

Page 7: 복합재료 손상의 무기저 진단 및 정량화 - KAISTssslab.kaist.ac.kr/article/pdf/Thesis_2.pdf · 1 Declaration of Ethical Conduct in Research: I, as a graduate student

CHAPTER 5 EXPERIMENTAL VERIFICATION

51 Overview middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 21

511 Specimen middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 22

512 Data acquisition system and PZT middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 22

513 Impact test middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 23

514 Temperature effect test middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 24

52 Teflon inserted damage middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 25

53 Impact-induced damagemiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 26

54 Temperature effect test middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 27

541 Teflon inserted damage middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 27

542 Impact-induced damage middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 29

CHAPTER 6 CONCLUDING REMARKS

61 Executive summary middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 31

62 Suggestion for further study middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 32

REFERCENCES middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 33

SUMMARY (IN KOREAN) middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 37

ACKNOWLEDGEMENTS (IN KOREAN)

CURRICULUM VITAE

iii

LIST OF TABLES

I Material property of simulation model and test specimen middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 17

II Damage cases for numerical simulation middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 18

III Damage characterization for the numerical simulation models middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 20

IV Damage characterization for the Teflon inserted damage specimen middotmiddotmiddotmiddotmiddotmiddotmiddot 25

V Damage characterization for the impact-induced damage specimen middotmiddotmiddotmiddotmiddotmiddotmiddot 27

VI Temperature effect test for the Teflon inserted damage specimen middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 28

VII Temperature effect test for the impact induced damage specimen middotmiddotmiddotmiddotmiddotmiddotmiddot 30

iv

LIST OF FIGURES

1 Lamb wave propagation characteristics (a) A0 Lamb wave mode time delay and

attenuation after passing through a damage (b) The time delay and amplitude varia-

tion of A0 Lamb wave mode under temperature variation in pristine condition of

structure Because of this conventional damage detection techniques in composites

have limitations for real structure application middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 5

2 Multiple reflections in the damage area The multiple reflections of A0 Lamb

waves in the damage area are utilize as a feature for characterization of the damage

in this study middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 6

3 Parameters and notations of the proposed damage characterization technique

Damage is detected by identifying the additional reflected wave from the damage due

to multiple reflections (dashed lines) Localization and quantification of the damage

is carried out by measuring the arrival time of transmitted and reflected waves from

the damage middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 7

4 Two different FEM models which have different 119889119889 and 119907119907119889119889 (or G) This repre-

sents that the damage cannot be characterized uniquely based on the measured data

available middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 9

5 (a) Example of the overlapped signal with comparing approximated signal using

the matching pursuit method (b) Decomposed signal using the matching pursuit

method The matching pursuit method provides a conventional way to decompose

and identify each wave middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 9

6 Description of the numerical simulation model middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 17

7 Comparison of A0 Lamb wave mode signals from numerical simulation be-

tween intact and damage cases Time delay attenuation and multiple reflections due

to damage simulated successfully using the damage modeling method described in

v

the previous section middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 19

8 Test specimens for the experimental verification In the Teflon inserted damage

Teflon tapes with 25mm diameter were inserted at the center of the specimen The

impact-induced damage was introduced to the specimen using the impact test equip-

ment middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 21

9 Overall experimental setup (a) Data acquisition (DAQ) system and (b) Installed

circular PZT transducer on the specimen with its dimension middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 23

10 (a) Impact test equipment for introducing impact-induced damage and (b) Tip

of the impact tester with 10mm diametermiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 23

11 Test specimen in temperature chamber middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 24

12 A0 Lamb wave mode obtained using pitch-catch method from the Teflon in-

serted damage specimen (a) The reflected wave from damage is shown (b) The Tef-

lon inserted damage is detected by capturing the reflected wave middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 25

13 The signal obtained from the test specimen subjected to impact The impact

was introduced at the same location of the specimen and the signal was obtained after

each impacts As the number of impact is increased the transmitted A0 Lamb wave

mode is delayed and attenuated middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 26

14 False alarm test for the proposed damage detection algorithm in pristine condi-

tion of the specimen Reflected wave from damage is not captured false alarm is not

occurred middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 26

15 Signals obtained from the Teflon inserted damage specimen under temperature

variation Time delay and amplitude variation are occurred middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 27

16 False alarm test under temperature variations It is shown that proposed dam-

vi

age characterization algorithm is robust under temperature variationsmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 29

17 Signals obtained from impact-induced damage specimen (Impact 3) under

temperature variation middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 29

vii

CHAPTER 1 INTRODUCTION

11 Motivation

Composite materials has been increasingly demands in modern structures ranging from the

latest aircraft to retrofitted bridge deck due to its light weight superior specific strength and

stiffness properties However the occurrence of damage due to abrupt impact or sequential

accumulation of defect may considerably degrade structural strength or stiffness and reduce

the safety and reliability of the system Recently Lamb waves based damage detection tech-

niques using piezoelectric (PZT) transducers have emerged as one of the promising structural

health monitoring (SHM) technology for detecting hidden damage because of the possibility

of long range inspection and high sensitivity [1-4]

12 Literature review

This section will discuss researches that have been previously carried out on topics of

guided wave based SHM techniques using PZT transducers for damage in composite materi-

als The early investigations of interaction between Lamb wave and damage in composite can

be found in Rose et al and Guo et al [5-6] From further studies reflected signals from dam-

age were shown to be an effective method for damage diagnosis and localization using PZTs

[7-9] The reflected signals from damage were collected using one PZT as an actuator and the

other PZT as a sensor However the techniques cannot be used to determine the extent of

- 1 -

damage with reasonable precision

Kesseler et al proposed a technique monitoring the wavelet coefficients of transmitted ul-

trasonic signal amplitude as a potential way to damage detection [10] One drawback of this

technique is that the variability in coupling and the sensitivity of installed PZT transducers

can induce changes in the amplitude and energy of the signal that are cause false-alarm Oth-

er wavelet based delamination detection and localization technique developed using PZT

sensor network and outlier analysis [11]

Damage detection and quantification in composite plate was also investigated by group de-

lay measurement due to damage [12] However it is assumed that the group velocity in dam-

age area as the group velocity propagates half thickness of the structure in dispersion curve in

the quantification procedure Thus the damage quantification was investigated only Teflon

inserted damage case

Sohn et al proposed a damage detection technique which does not rely on any past base-

line signals to assess damage in composite plates by using an enhanced time reversal method

[13] The damage is assessed using the violation of time reversibility when nonlinearity is

caused by the damage along a direct wave path Thus it allows instantaneous identification

of delamination without requiring the baseline signal for comparison

- 2 -

13 Objective and uniqueness

In this study a reference-free damage characterization technique including detection lo-

calization and quantification of impact-induced damage in composite materials is developed

The general damage characterization can be categorized into the following levels [14]

Level 0 Define damage

Level 1 Determine the presence of damage within the structure

Level 2 Locate the regions of damage

Level 3 Quantify the damage

Level 4 Predict the remaining service life of the structure

Through the proposed technique from level 1 to level 3 can be achieved

Most of common approaches of damage characterization techniques described in the pre-

vious section use prior obtained baseline data However these approaches can cause false

alarm due to the effect of environmental variations such as temperature and ambient vibra-

tions [15]

The uniqueness of this study is

1 Damage is characterized without using prior measured baseline data

2 Robustness of the proposed technique under temperature variation is investigated

3 Additional waves reflected from inside damage have been used in addition to transmit-

ted and reflected waves

- 3 -

14 Organization

The organization of this thesis is summarized as follows

Chaper1 A general introduction including motivation literature review objective and

uniqueness of this study is described

Chaper2 Interactions between Lamb wave mode and damage in composites such as time

delay attenuation and multiple reflections are summarized Damage index (m-value) for

damage characterization is defined

Chaper3 Algorithms for the reference-free damage characterization technique using m-

value are developed Modified matching pursuit for this study is also described

Chaper4 Verification of the algorithm using numerical simulation is presented Various

modeled damage cases are characterized

Chaper5 Experimental verifications are described Two damage cases (Teflon inserted and

impact-induced damage) are investigated for the feasibility of the proposed damage charac-

terization technique Temperature effect test using temperature chamber is also carried out

Chaper6 Conclusion and summary of the present study is reported In advance suggestions

and future work recommendations are represented

- 4 -

CHAPTER 2 INTERACTION OF LAMB WAVE WITH DAMAGE

21 A0 Lamb wave mode delay and attenuation in the time domain

Time delay and attenuation of the A0 mode Lamb wave after passing through the damage

area due to effective thickness reduction were reported in previous researches [812] Moreo-

ver the attenuation of the A0 Lamb wave mode is occurred by degrading of shear modulus of

in the damage area However the time delay and amplitude variation of the A0 Lamb wave

mode also occurred even when the structure is health condition by the material property

changing due to temperature variation of the structure (Fig 1) Thus conventional damage

characterization methods using the time delay and attenuation of A0 Lamb wave mode as a

feature can cause false-alarm in real application

11 115 12 125 13-10

-5

0

5

10

Time (ms)

Vol

tage

(mV

)

IntactDamage

Reflected wave from Damage

(a)

11 115 12 125 13-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

0 oC20 oC50 oC

(b)

Fig 1 Lamb wave propagation characteristics (a) A0 Lamb wave mode time delay and atten-uation after passing through a damage (b) The time delay and amplitude variation of A0 Lamb wave mode under temperature variation in pristine condition of structure Because of this conventional damage detection techniques in composites have limitations for real struc-ture application

- 5 -

22 Multiple reflections in the damage area

Standing wave patterns due to multiple reflections of A0 Lamb wave mode in the damage

area were observed through numerical simulations and ultrasonic wave-field imaging using a

scanning laser vibrometer [16] The formation of standing waves in damage is illustrated

through Fig 2 When the incident waves enter the damage area a significant portion of the

waves are reflected back from the exit Then the reflected waves travel through the damage

and undergo reflection again at the original entrance As a consequence of multiple reflec-

tions at the entrance and the exit a considerable amount of ultrasonic energy is trapped inside

the damage area [17-19]

Multiple Reflections

Incident Wave Transmitted Wave

Damage

Fig 2 Multiple reflections in the damage area The multiple reflections of A0 Lamb waves in the damage area are utilize as a feature for characterization of the damage in this study

- 6 -

CHAPTER 3 THEORETICAL DEVELOPMENT

31 Overview

Damage

l = l1 + d + l2

d = vdtd l1 = vit1 l2 = vit2

t = t1 + td + t2

Excitation and Sensing PZT Sensing PZT

l

Transmitted Wave

Reflected Waves

Incident Wave

Fig 3 Parameters and notations of the proposed damage characterization technique Damage is detected by identifying the additional reflected wave from the damage due to multiple re-flections (dashed lines) Localization and quantification of the damage is carried out by measuring the arrival time of transmitted and reflected waves from the damage

In this section the reference-free damage characterization algorithm is proposed Damage

diagnosis and quantification are achieved by identifying the additional reflected waves due to

multiple reflections in the damage area (dashed lines in Fig 3) Moreover damage is local-

ized using additional information from pulse-echo method implement For the localization

and the quantification of damage the variables such as the path length and A0 Lamb wave

mode velocity in damage area are defined (Fig 3) For example 119897119897 is expressed as

119897119897 = 1198971198971 + 119889119889 + 1198971198972 (1)

where 1198971198971 119889119889 and 1198971198972 are the distance from the excitation PZTs to the entrance of damage the

damage length and the distance from the exit of damage to the sensing PZT respectively

- 7 -

Moreover 119907119907119894119894 and 119907119907119889119889 are also defined as the A0 Lamb wave mode velocity in pristine area

and damage area respectively Note that the whole procedure of the damage characterization

is accomplished based solely on the signals obtained from the current state of the structure

32 Damage index (m-value)

The proposed reference-free damage characterization technique utilizes the waves reflect-

ed additionally from the damage area as a feature The arrival time difference of the transmit-

ted and reflected A0 Lamb wave modes reflected from the exit and the entrance of the delam-

ination area one times each (∆119905119905) contains the information of the physical length of the dam-

age (119889119889) and A0 Lamb wave mode velocity in the damage area (119907119907119889119889) The wave velocity in the

damage area which is related to the shear modulus (G) represents the severity of the damage

∆119905119905 = 2119905119905119889119889 =2119889119889119907119907119889119889

(2)

However based on the measured data available the 119889119889 and 119907119907119889119889 (or G) cannot be uniquely

determined Fig 4 shows two FEM models which have different physical length and wave

velocity (G) of the damage Detail explanation of delamination modeling technique will be

covered in section 4 The similar signal in the time domain and ∆119905119905 are obtained from the

two different damage cases

- 8 -

Damage 1

Damage 2

20

Unit mm15

90 G value reduction

943 G value reduction

m-value = 248

m-value = 248

015 02 025 03 035

-2

-1

0

1

2

Time (ms)

Ampl

itude

Damage 1Damage 2

Fig 4 Two different FEM models which have different 119889119889 and 119907119907119889119889 (or G) This represents that the damage cannot be characterized uniquely based on the measured data available

Thus in this study the damage index (m-value) which can be uniquely defined based on

the available data is introduced The m-value is defined as the ratio of the physical length of

the damage and wavelength of the A0 Lamb wave mode in the damage area (120582120582119889119889)

119898119898 =119889119889120582120582119889119889

=119889119889119889119889119907119907119889119889

(3)

In many cases the transmitted and reflected waves are overlapped and it is hard to decom-

pose and calculate the arrival times of them Matching pursuit method is an alternative way

to accurately decompose of the overlapped waves and estimate the parameters of each wave

such as arrival time (Fig 5) Detail explanation of the matching pursuit method is covered in

section 36

015 02 025 03-2

-1

0

1

2

Time (ms)

Am

plitu

de

OriginalApproximated

(a)

015 02 025 03-2

-1

0

1

2

Time (ms)

Ampl

itude

Wave 1Wave 2

(b)

Fig 5 (a) Example of the overlapped signal with comparing approximated signal using the matching pursuit method (b) Decomposed signal using the matching pursuit method The matching pursuit method provides a conventional way to decompose and identify each wave

- 9 -

33 Damage detection algorithm (Level 1)

The damage diagnosis is accomplished by extracting the additional reflected A0 Lamb

wave modes from damage due to multiple reflections If A0 Lamb wave modes are decom-

posed perfectly it is trivial that one transmitted A0 Lamb wave mode is always captured in

the response signal If damage exists other waves in the response signal behind the transmit-

ted A0 Lamb wave mode are presented Since this approach does not rely on previously col-

lected data the diagnosis is independent from environmental variation such as temperature

changes

It is reported that the dispersion of Lamb waves cause wave-packets to spread out in space

and time as they propagate through a structure [20] Thus the reflected waves from the dam-

age have longer pulse duration than the transmitted wave due to their longer propagating dis-

tance In this study the matching pursuit method is modified to capture more spread wave

than the transmitted A0 Lamb wave mode Detail explanation of the modified matching pur-

suit method is described in latter section

Let 1199041199041 and 1199041199042 as the pulse duration (scale) of the transmitted and additional reflected A0

Lamb wave mode respectively

ldquoIf 1199041199042 gt 1199041199041 damage exists

Otherwise there is no damagerdquo

- 10 -

34 Damage quantification algorithm (Level 3)

As mentioned in the previous section damage is quantized by estimating 119889119889 and 119907119907119889119889 using

pitch-catch method The procedure to estimate 119889119889 and 119907119907119889119889 is

(i) Measure 119897119897 119905119905 and 119905119905119889119889

(ii) Assume the range of 119889119889 (0 lt 119889119889 lt 119897119897)

(iii) Calculate 119907119907119894119894 and 119907119907119889119889 with a constraints (0 lt 119907119907119889119889 lt 119907119907119894119894) (Eq 4 and 5)

(iv) Estimate m-value and find a feasible range of 119889119889 and 119907119907119889119889

119907119907119894119894 =119897119897 minus 119889119889119905119905 minus 119905119905119889119889

(4)

119907119907119889119889 =119889119889119905119905119889119889

(5)

35 Damage localization algorithm (Level 2amp3)

For damage localization 1198971198971 is estimated by measuring the arrival time from the excitation

PZT to the entrance of delamination (1199051199051) with additional pulse-echo method implementation

The procedure is similar with the previous damage quantification algorithm

(i) Measure 119897119897 119905119905 119905119905119889119889 and 1199051199051

(ii) Assume the range of 1198971198971 (0 lt 1198971198971 lt 119897119897)

(ii) Compute possible combinations of 1198971198972 119889119889 and 119907119907119889119889 with constraints

(0 lt 119907119907119889119889 lt 119907119907119894119894 119889119889 gt 0) (Eq 5 6 and 7)

- 11 -

1198971198972 = 1198971198971119905119905 minus 1199051199051 minus 119905119905119889119889

1199051199051 (6)

119889119889 = 119897119897 minus 1198971198971 minus 1198971198972 (7)

Since a PZT for pulse-echo method acts as both an actuator and a sensor the obtained sig-

nal contains both the excitation and the response signal Thus the reflected wave from the

entrance of damage can be buried into the tail of excitation signal and cause error on arrival

time of reflected wave from damage Even in this case damage can be localized and quan-

tized using the constraints When first wave packet reflected from the entrance of damage is

buried into the excitation signal the next wave packet reflected from the exit of damage can

be utilized alternatively If the algorithm regards the second reflected wave as the first re-

flected wave there are no combinations of estimated variable meeting the constraints Then

the algorithm repeats the process by regarding the wave as the second reflected wave the

constraints can select proper combination of the estimated variables

The estimated values have range bounded by the constraints When additional information

of 119907119907119894119894 is utilized the damage can be characterized more accurately In this study three meth-

ods are proposed and verified as information level of 119907119907119894119894

Method 1 Solely based on the current data measurement (Reference-free)

Method 2 Limit the possible range of 119907119907119894119894 (Prior knowledge of the system)

Method 3 Measure 119907119907119894119894 (Additional data need to be collected)

Method 1 described in this section can be utilized without any information of 119907119907119894119894 Meth-

od 2 produces more accurate estimation results using prior knowledge of the system such as

- 12 -

119907119907119894119894 variation under temperature changes Method 3 gives unique estimated value using meas-

ured 119907119907119894119894 from other path in the same time In section 4 and 5 damage is characterized using

these three methods and the results are compared each another

36 Matching pursuit method considering the dispersion effect of Lamb

wave

Matching pursuit method has been developed for signal compression approximation and

characterization of obtained signal and function [21-23] The key concept of the matching

pursuit method is that any signal (function) can be decomposed into a linear expansion of

signals selected from a pool of basis signals (functions) called lsquodictionaryrsquo

Let 119891119891 be a function to be analyzed then 119891119891 can be approximated a set of functions select-

ed from the dictionary 119863119863 in order to best match This is done by successive approximations

with orthogonal projections Let1198921198921205741205740 isin 119863119863 the function 119891119891 can be decomposed into

119891119891 = lang1198911198911198921198921205741205740rang1198921198921205741205740 + 119877119877119891119891 (8)

where 119877119877119891119891 is the residual function after approximating 119891119891 using 1198921198921205741205740 1198921198921205741205740 is orthogonal to

119877119877119891119891 hence

1198911198912 = lang1198911198911198921198921205741205740rang2

+ 1198771198771198911198912 (9)

To minimize 119877119877119891119891 1198921198921205741205740 is chosen to maximize lang1198911198911198921198921205741205740rang The matching pursuit method is

an iterative algorithm that decomposes the residue 119877119877119891119891 by projecting it on a function of 119863119863

that matches 119877119877119891119891 almost at best as it was done for 119891119891 The residue 119877119877119899119899119891119891 is decomposed into

119877119877119899119899119891119891 = lang119877119877119899119899119891119891119892119892120574120574119899119899rang119892119892120574120574119899119899 + 119877119877119899119899+1119891119891 (10)

- 13 -

the decomposition is carried up to the order 119898119898 119891119891 can be expressed as

119891119891 = lang119877119877119899119899119891119891119892119892120574120574119899119899rang119892119892120574120574119899119899

119898119898minus1

119899119899=0

+ 119877119877119898119898119891119891 (11)

The original function 119891119891 is decomposed into a sum of dictionary elements that are chosen to

best match its residues

In this study the matching pursuit method is utilized to accurately decompose of the over-

lapped Gaussian pulses and estimate the parameters of the pulse which used for damage

characterization The 119894119894119905119905ℎ Gaussian pulse in the dictionary for approximating the obtained

signal is expressed as

119892119892120574120574119894119894 = 119891119891(119860119860119894119894 119906119906119894119894 119904119904119894119894 119889119889119894119894120601120601119894119894) (12)

where 119860119860119894119894119906119906119894119894119904119904119894119894119889119889119894119894 and 120601120601119894119894 represent amplitude arrival time scale (pulse duration) excita-

tion(central) frequency and phase of the Gaussian pulse respectively The parameters of the

pulse are estimated using a nonlinear regression based optimization method (Gauss-Newton

method) based on an initial guess of the parameters Detail procedure for estimating the pa-

rameters of the pulse is expressed in Hong et al [23]

Among the estimated parameters of the pulses pulse duration (scale) represents the disper-

sion effect If a wave-packet is spread out the estimated pulse duration is increased Using

this feature the matching pursuit method is modified to capture waves which have longer

pulse duration than transmitted A0 Lamb wave mode If the estimate pulse duration of a wave

is less than the transmitted wave the modified matching pursuit method regards it as error

- 14 -

CHAPTER 4 NUMERICAL SIMULATION

41 Composite modeling method

Composite materials can be modeled from nano to macro level depends on its engineering

purpose In general four methods are utilized for modeling of composite materials First

composite is modeled as a homogenous material expressed in terms of effective material

properties [24] The effective material properties are estimated using the modulus and the

volume fraction of each constituent theoretically (Rule of mixture) and measured experimen-

tally Second a model is established that each laminate which have own material properties

are totally bonded together In this model the stacking sequence and direction of each lami-

nates are considered additionally [25-27] Third method for composite modeling in micro

level called micromechanics [28-30] This model considers the interaction between fibers and

matrix in laminate Last multiscale modeling which the scope of modeling is extended to

molecular or quantum level of constituents [31-33] This model starts from molecular me-

chanics or molecular dynamics then the scale of model leads to the nano micro and macro

level by incorporating small scale models As the scale of models is smaller it provides more

logical way than other large scale of models while the computational cost is more expensive

The scope of this study is focused on the dynamic behavior of composite structure the beam

or plate model was utilized for the numerical simulation

- 15 -

42 Damage modeling method

Generally three damage models have been developed and utilized in the real applications

First the stiffness reduction model called as the equivalent element model was introduced

[17 34-35] This model is simply realized by replacing the region around damage area by a

region with adjusted material properties (Youngrsquos modulus shear modulus density etc)

globally Next the duplicated node model is done by disconnecting the element connectivity

assuming there is no interaction among the delaminated surface This model can be utilized

when the location and the extension of damage is important issue [34-36] Last the contact

mechanics model using the principles of contact mechanics the two interfacial forces (con-

tact resistance and sliding resistance) in the delaminated surface [37-38] In this model the

delaminated surfaces are modeled as nonlinear functions of the interlaminar normal strain

and the interlaminar shear strain This model is computationally expensive while provides

more logical way than other models There is however no universally accepted model for

general impact-induced damage in composites In the case of impact-induced damage it is

impossible to predict or estimate the location and number of delaminated layers Thus from

the uncertainty of impact-induced damage the equivalent element model can be considered

as the effective damage model

- 16 -

43 Numerical simulation description

TABLE I Material property of simulation model and test specimen

Tensile Strength 657 MPa Tensile Modulus 59 GPa Shear Strength 804 MPa Shear Modulus 24 Gpa Specific Gravity 16

PZT A

PZT C

PZT BDamage

3

250 125 125

500 Unit mm

Fig 6 Description of the numerical simulation model

The feasibility of the proposed damage characterization technique is verified by 2-D FEM

simulation using MSC PATRAN and NASTRAN Fig 6 presents the dimension PZTs and

damage location of the simulation model In the model 05mm x 05mm square shell element

was used for modeling of the A0 mode whose flexural wave characteristic The element size

was determined by considering the wavelength of the propagation Lamb wave modes The

material properties of the simulation model and damage cases are presented in Table I and

Table II respectively

PZT A and C act as both actuator and sensor for collecting pulse-echo signal They were

installed on the plate which two of them are both-side collocated for the A0 Lamb wave

- 17 -

mode decomposition [39-40] PZT B was only used as a sensor for collecting pitch-catch sig-

nal For simulating the excitation and sensing of PZTs the pin-force model was adopted that

an input force apples to the end nodes of excitation PZT and averaging element strains corre-

sponding sensing area A Gaussian pulse (tone-burst) 50 kHz central frequency was excited

as an input signal

TABLE II Damage cases for numerical simulation

119949119949120783120783 (mm) 119941119941 (mm) G value reduction ()

Case 1 650 200 90 darr

Case 2 675 150 90 darr

Case 3 675 150 93 darr

Case 4 900 150 90 darr

- 18 -

44 Numerical simulation results

015 02 025 03 035

-2

0

2

Time (ms)

Am

plitu

de

IntactDamage

Reflected wave from damage

(a) Pitch-catch signal

015 02 025 03 035

-2

0

2

Time (ms)

Am

plitu

de

IntactDamage

Reflected waves from damage

(b) Pulse-echo signal Fig 7 Comparison of A0 Lamb wave mode signals from numerical simulation between intact and damage cases Time delay attenuation and multiple reflections due to damage simulated successfully using the damage modeling method described in the previous section

Fig 7 shows that comparison of A0 Lamb wave mode signals between intact and one of

damage cases in the simulation The features of A0 Lamb wave mode passing through dam-

age area described in section 2 are successfully simulated Thus the equivalent element

method is proper for modeling damage in composites In the damage cases the additional

waves reflected from damage is observed in both pitch-catch and pulse echo signal

Table III presents the result of the proposed damage characterization algorithm using the

numerical simulation cases shown in Table II In lsquoMethod 2rsquo 119907119907119894119894 variation is assumed to plusmn

5 of the normal value (1190~1315ms) The normal value of 119907119907119894119894 for lsquoMethod 3rsquo is meas-

ured as 1253 ms In each case the damage is characterized using the proposed technique

Note that as the information of 119907119907119894119894 is specified the estimated value is closer to the exact

value

- 19 -

TABLE III Damage characterization for the numerical simulation models

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

Case 1

Exact 650 200

Method 1 248 50~70 79~414 160~834

Method 2 248 603~667 135~241 271~486

Method 3 248 635 188 379

Case 2

Exact 675 150

Method 1 193 60~75 01~251 37~651

Method 2 193 618~682 114~222 295~575

Method 3 193 65 168 435

Case 3

Exact 675 150

Method 1 233 60~75 66~302 140~649

Method 2 233 668~738 85~196 182~419

Method 3 233 703 140 301

Case 4

Exact 900 150

Method 1 204 75~100 48~349 118~856

Method 2 204 860~951 107~216 264~530

Method 3 204 906 162 397

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1253 ms 1190~1315 ms Measured 119907119907119894119894 1253 ms

- 20 -

CHAPTER 5 EXPERIMENTAL VERIFICATION

51 Overview

In this section experimental study is investigated so that the proposed damage characteri-

zation technique Overall experimental setup and test specimen are shown in Fig 8 and 9

The experimental setup consists of test specimens data acquisition (DAQ) system and self-

sensing circuit Self-sensing circuit was installed for pulse-echo measurement using PZTs

[40-41]

PZT A(PZT C) PZT B

500

500

140

75

250

Teflon Inserted Damage (d =25)

Unit mm

Thickness = 3

PZT A(PZT C) PZT B

500

500

150

90

250

Impact Location

Unit mm

Thickness = 3

(a) Teflon inserted damage specimen

(b) Impact-induced damage specimen

Fig 8 Test specimens for the experimental verification In the Teflon inserted damage Teflon tapes with 25mm diameter were inserted at the center of the specimen The impact-induced damage was introduced to the specimen using the impact test equipment

- 21 -

511 Specimen

Two carbon composite plates with a dimension of 500times500times3mm3 were manufactured by

Nexcomes Inc by stacking 12 layers of CF3327 3K prepreg (woven fabrics type) manufac-

tured by Hankuk Fiber Inc The material properties of the specimen are presented in Table I

One of the specimens Teflon tapes with 25mm diameter was inserted in every 3 layers at the

center of the specimen to mimic the damage The other specimen is utilized for the character-

ization of real impact-induced damage

512 Data acquisition (DAQ) system and PZT

The data acquisition (DAQ) system consists of an arbitrary waveform generator (AWG) a

high speed signal digitizer (DIG) and two multiplexers (MUX) Using the 16-bits AWG

tone-burst input signal with a plusmn10 peak-to-peak voltage were generated The response signal

was measured by the 14-bit DIG For the signal-to-noise ratio reduction both pitch-catch and

pulse-echo signals were measured 150 times and averaged in the time domain 3 identical

circular PZTs of 10mm diameter manufactured by APC International Ltd were installed on

the specimen A Gaussian pulse (tone-burst) 50 kHz central frequency was excited as an in-

put signal The excitation frequency was chosen in order to avoid overlapping of the S0 and

A0 Lamb wave modes and generating of the higher modes (S1 A1hellip)

- 22 -

PZT Actuator

PZT Sensor

AWG

DIG

DAQ System Specimen

SensingActuatingMUX

1

23

4

1 Sending signal to MUX

4 Sending signal to DIG

Self-SensingCircuit

18 Unit mm

10

(a)

(b)

Fig 9 Overall experimental setup (a) Data acquisition (DAQ) system and (b) Installed circu-lar PZT transducer on the specimen with its dimension

513 Impact test

The impact-induced damage was introduced to the specimen using the impact test equip-

ment presented in Fig 10 A 5kg mass with 10mm tip was dropped from 015m height onto

the same location of the specimen The signals were obtained at each number of impacts

Since the reflected wave is not detected at the first impact case the data obtained from the

second and third impact were utilized for the analysis

10

Unit mm (a)

(b)

Fig 10 (a) Impact test equipment for introducing impact-induced damage and (b) Tip of the impact tester with 10mm diameter

- 23 -

514 Temperature effect test

In reality structures are subjected to changing environmental conditions such as tempera-

ture that affect measured signal and cause false-alarm Thus the robustness of the proposed

damage characterization technique under temperature variation in investigated (Fig 11) The

specimens were placed inside a temperature chamber and a thermocouple was installed on

the specimen to measure temperature of surface of the specimen The signal was obtained

under three temperature cases (0 20 and 50degC) and the humidity was kept at 30

Temperature Chamber

Test Specimen

Fig 11 Test specimen in temperature chamber

- 24 -

52 Teflon inserted damage

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Reflected wave from damage

(a)

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Transmitted WaveReflected Wave

Detected reflected wave

(b)

Fig 12 A0 Lamb wave mode obtained using pitch-catch method from the Teflon inserted damage specimen (a) The reflected wave from damage is shown (b) The Teflon inserted damage is detected by capturing the reflected wave

Fig 12 shows the obtained pitch-catch signal from the Teflon inserted damage specimen

and damage diagnosis result The Teflon inserted damage is detected successfully by captur-

ing the reflected wave Table IV presents the result of the Teflon inserted damage characteri-

zation 119907119907119894119894 is measured as 1213ms and the variation is assumed to plusmn 5 of the normal val-

ue (1152~1273ms) The results indicate that the proposed damage characterization approach

works well in real application

TABLE IV Damage characterization for the Teflon inserted damage specimen

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

Exact 635 25

Method 1 123 60~75 73~338 298~1220

Method 2 123 619~662 229~310 934~1188

Method 3 123 631 284 1161

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1213 ms

- 25 -

53 Impact-induced damage

11 115 12 125 13-4

-2

0

2

4

Time (ms)

Vol

tage

(mV

)

IntactImpact2Impact3

Reflected wave from damage

(a) Pitch-catch signal

11 115 12 125 13 135

-05

0

05

1

Time (ms)

Volta

ge (m

V)

IntactImpact2Impact3

Boundary reflection

Reflected wave from damage

(b) Pulse-echo signal

Fig 13 The signal obtained from the test specimen subjected to impact The impact was in-troduced at the same location of the specimen and the signal was obtained after each impacts As the number of impact is increased the transmitted A0 Lamb wave mode is delayed and attenuated

The obtained signals from impact-induced damage specimen are presented in Fig 13 It is

observed that as the number of impact is increased the transmitted A0 Lamb wave mode is

more delayed because the wave velocity in the damage area is reduced Fig 14 shows the

false-alarm test result whether the proposed damage detection algorithm works even in pris-

tine condition of the test specimen It is shown that no false alarm is occurred because no ad-

ditional reflected wave is captured

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Transmitted Wave

(a) Transmitted wave

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Reflected Wave

No reflected wave

(b) Reflected wave

Fig 14 False alarm test for the proposed damage detection algorithm in pristine condition of the specimen Reflected wave from damage is not captured false alarm is not occurred

- 26 -

TABLE V Damage characterization for the impact-induced damage specimen

of Impacts m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

2

(1470 J)

Method 1 080 65~90 75~246 470~1157

Method 2 080 801~885 114~245 711~1134

Method 3 080 843 179 1122

3

(2205 J)

Method 1 148 50~100 06~380 21~1134

Method 2 148 758~838 249~368 843~1125

Method 3 148 798 308 1044

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1213 ms

Table V presents the characterization results of the impact-induced damage case Here the

measured 119907119907119894119894 and the assumed variation are same as the artificial damage case It is observed

that as the number of impact is increased 119889119889 is increased while 119907119907119889119889 is reduced Thus it is

validated that the proposed damage characterization approach estimates the increment of the

damage well

54 Temperature effect test

541 Teflon inserted damage

11 115 12 125 13-10

-5

0

5

10

Time (ms)

Vol

tage

(mV

)

0 oC20 oC50 oC

(a) Pitch-catch signal

11 115 12 125 13 135

-1

0

1

2

3

Time(ms)

Volta

ge (m

V)

0 oC20 oC50 oC

(b) Pulse-echo signal

Fig 15 Signals obtained from the Teflon inserted damage specimen under temperature varia-tion Time delay and amplitude variation are occurred

Fig 15 shows the obtained signals from the Teflon inserted damage specimen subjected to

- 27 -

temperature variation Time delay and amplitude variations are occurred there is no addition-

al damage is introduced The analyzed results of the obtained signals are reported in Table VI

The measured 119907119907119894119894 is 1247 ms at 0 degC 1213 ms at 20 degC and 1168 ms at 50 degC respectively

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms (1152~1273 ms)

The estimated results are also changed under temperature variations the fluctuation of esti-

mated range and exact value is small

TABLE VI Temperature effect test for the Teflon inserted damage specimen

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

0 degC

Exact 635 25

Method 1 113 60~70 79~268 352~1291

Method 2 113 560~619 233~344 1034~1172

Method 3 113 606 257 1141

20 degC

Exact 635 25

Method 1 123 60~75 73~338 298~1220

Method 2 123 619~662 229~310 934~1188

Method 3 123 631 284 1161

50 degC

Exact 635 25

Method 1 137 60~75 72~328 263~1228

Method 2 137 622~687 183~299 666~1086

Method 3 137 642 264 960

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1247 ms at 0 degC 1213 ms at 20 degC and 1168 ms at 50 degC

- 28 -

542 Impact-induced damage

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Vol

tage

(mV

)

Transmitted WaveReflected Wave

No reflected wave

(a) 0 degC

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Transmitted WaveReflected Wave

No reflected wave

(b) 50 degC

Fig 16 False alarm test under temperature variations It is shown that proposed damage char-acterization algorithm is robust under temperature variations

The false alarm test in pristine condition of the impact-induced damage specimen under

temperature variation was investigated (Fig 16) It is validated that the proposed damage di-

agnosis algorithm is robust under temperature variations Fig 17 shows the obtained signals

from the impact-induced damage (impacted 3 times) specimen under temperature variation

The analyzed results of the obtained signals are reported in Table VII The measured 119907119907119894119894 in

each temperature and the assumed variation are same as the artificial damage case

11 115 12 125 13-5

0

5

Time (ms)

Volta

ge (m

V)

0 oC20 oC50 oC

(a) Pitch-catch signal

11 115 12 125 13 135

-1

0

1

2

3

Time(ms)

Vol

tage

(mV

)

0 oC20 oC50 oC

(b) Pulse-echo signal

Fig 17 Signals obtained from impact-induced damage specimen (Impact 3) under tempera-ture variation

- 29 -

TABLE VII Temperature effect test for the impact induced damage specimen (Impact 3 case)

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

0 degC

Method 1 135 75~90 72~310 267~1148

Method 2 135 705~779 264~381 978~1126

Method 3 135 763 289 1071

20 degC

Method 1 148 50~100 06~380 21~1134

Method 2 148 758~838 249~368 843~1125

Method 3 148 798 308 1044

50 degC

Method 1 153 75~100 53~415 174~1227

Method 2 153 804~889 214~336 702~1103

Method 3 153 815 320 1050

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1247 ms at 0 degC 1213 ms at 20 degC and 1168 ms at 50 degC

- 30 -

CHAPTER 6 CONCLUDING REMARKS

61 Executive summary

In this study the reference-free damage detection localization and quantification tech-

nique in a multi layered composite plate is proposed Additional reflected A0 Lamb wave

mode from the damage due to multiple reflections in damage area is used as a feature for

characterizing the damage First damage is characterized in terms of a damage index (m-

value) defined as the ratio of the physical length of the damage to the wavelength of the A0

Lamb wave mode Then the range of possible damage location length and severity are esti-

mated with constraints This procedure can be accomplished without prior baseline data To

realize the proposed technique the matching pursuit technique is modified for reduction of

the error by considering the dispersive characteristic of Lamb wave The matching pursuit

method estimates the arrival time of each wave packets accurately even two wave packets are

overlapped The numerical simulations and experimental tests including Teflon tape inserted

and impact-induced damage are investigated for the feasibility of the proposed technique

Moreover the effectiveness of the proposed technique under changing temperature is exper-

imentally verified

- 31 -

62 Suggestion for further study

The limitations of this study and the possible improvement and further studies are listed as

follow

1 Measurement of impact-induced damage length using conventional NDT technique

Additional experimental verification for impact-induced damage characterization is

needed by measuring the real length of damage using conventional NDT technique

such as ultrasonic scan or infrared camera

2 Limitation of both side collocated PZT installation In this study excitation PZTs are

installed both side and collocated for decomposing A0 Lamb wave mode However it

can produce decomposition error due to imperfection of PZT location and their installa-

tion Moreover in many cases assess to both surface of structures is not allowed such

as aircraft applications To improve this limitation a reference-free damage characteri-

zation using single side PZTs will be considered

3 Complex structure application In this study damage is characterized in simple plate-

like structure When the monitored structure becomes complex the response signals al-

so becomes complex due to its geometry and boundary Thus the investigations for

complex structure application such as stiffener spar and curvature are needed

- 32 -

REFERENCES

[1] Cawley P and Alleyne D (1996) ldquoThe use of Lamb waves for the long range inspection of large structuresrdquo Ultrasonics 34 pp287-290

[2] Ghosh T Kundu T and Karpur P (1998) ldquoEfficient use of Lamb waves for detecting defects in large platesrdquo Ultrasonics 36 pp791-801

[3] Guirguitui V (2003) ldquoLamb wave generation with piezoelectric wafer active sensors for structural health monitoringrdquo SPIE Conference on Smart Structures and Materials San Diego CA 5391 pp111-422

[4] Ragvahan A and Cesnik CE S (2004) ldquoModeling of piezoelectric-based Lamb wave generation and sensing for structural health monitoringrdquo SPIE Conference on Smart Struc-tures and Materials San Diego CA 5391 pp419-430

[5] Rose JL Pilarski A and Dirit JJ (1993) ldquoAn approach to guided wave mode selection for inspection of laminated platerdquo Journal of Reinforced Plastics and Composites 12(5) pp536-544

[6] Guo N and Cawley P (1993) ldquoThe interaction of Lamb waves with delaminations in composite laminatesrdquo Journal of Acoustical Society of America 94(4) pp 2240-2246

[7] Valdez SHD and Soutis C (2002) ldquoReal-time nondestructive evaluation of fiber com-posite laminates using low-frequency Lamb wavesrdquo Journal of Acoustical Society of Amer-ica 111(5) pp 2026-2033

[8] Yang M and Qiao P (2005) ldquoModeling and experimental detection of damage in virous materials using the pulse-echo method and piezoelectric sensoractuatorsrdquo Smart Materials and Structures 14 pp1083-1100

[9] Lestari W and Qiao P (2005) ldquoApplication of wave propagation analysis for damage identification in composite laminated beamsrdquo Journal of Composite Materials 39(22) pp1967-1984

[10] Kessler S S Spearing S M and Soutis C (2002) ldquoDamage detection in composite ma-terials using Lamb wave methodsrdquo Smart Materials and Structures 11 pp269-278

[11] Sohn H Park G Wail J R Limback N P and Farrar C R (2004) ldquoWavelet-based active sensing for delamination detection in composite structuresrdquo Smart Materials and Structures 13 pp 153-160

[12] Petculescu G Krishnaswamy S and Achenback J D (2008) ldquoGroup delay measure-ments using modally selective Lamb wave transducers for detection and sizing of delamina-tions in compositesrdquo Smart Materials and Structures 17

[13] Sohn H Park H W Law K H and Farrar C R (2007) ldquoDamage detection in compo-

- 33 -

site plates by using an enhanced time reversal methodrdquo Journal of Aerospace Engineering (ASCE) pp 141-151

[14] Sohn H (2007) ldquoEffects of environmental and operational variability on structural health monitoringrdquo Philosophical Transactions of the Royal Society Ardquo 365 pp539-560

[15] Rytter A (1993) ldquoVibration based inspection of civil engineering structuresrdquo PhD The-sis Department of Building Technology and Structural Engineering University of Aalborg Aalborg Denmark

[16] Hayashi T and Kawashima K (2002) ldquoMultiple reflections of Lamb waves at a delami-nationrdquo Ultrasonics 40 pp193-197

[17] Wang C H and Rose L R F (2003) ldquoWave reflection and transmission in beams con-taining delamination and inhomogeneityrdquo Journal of Sound and Vibration 264 pp 851-872

[18] Ramadas C Balasubramaniam K Joshi M and Krishnamurthy C V (2009) ldquoInterac-tion of the primary anti-symmetric Lamb mode (Ao) with symmetric delaminations numer-ical and experimental studiesrdquo Smart Materials and Structures 18

[19] Ramadas C Balasubramaniam K Joshi M and Krishnamurthy C V (2010) ldquoInterac-tion of guided Lamb waves with an asymmetrically located delamination in a laminated composite platerdquo Smart Materials and Structures 19

[20] Wilcox P Lowe M and Cawely P (2001) ldquoThe effect of dispersion on long-range in-spection using ultrasonic guided wavesrdquo NDTampE International 34 pp1-9

[21] Mallat S G and Zhang Z (1993) ldquoMatching pursuits with time-frequency dictionariesrdquo IEEE Tansactions on Signal Processing 41(12) pp3397-3415

[22] Zhang G Zhang S and Wang Y (2000) ldquoApplication of adaptive time-frequency de-composition in ultrasonic NDE of highly scattering materialsrdquo Ultrasonics 38 pp 961-964

[23] Hong J C Sun K H and Kim Y Y (2005) ldquoThe matching pursuit approach based on the modulated Gaussian pulse for efficient guided-wave damage inspectionrdquo Smart Mate-rials and Structures 14 pp548-560

[24] Herakovich C T (1997) Mechanics of Fibrous Composites Wiley 480 pages

[25] Reddy J N (1985) ldquoA review of the literature on finite-element modeling of laminated composite platesrdquo The Shock and Vibration Digest 17(4) pp 3-8

[26] Reddy J N (2003) Mechanics of Laminated Composite Plates and Shells Theory and Analysis CRC Press 856 pages

[27] Palacz M Krawczuk M and Ostachowicz W (2005) ldquoThe spectral finite element mod-el for analysis of flexure-shear coupled wave propagation Part 1 laminated multilayer

- 34 -

composite beamrdquo Composite Structures 68 pp37-44

[28] Zhang Y C and Harding J (1990) ldquoA numerical micromechanics analysis of the me-chanical properties of a plain weave compositerdquo Computer amp Structures 36(5) pp 839-844

[29] Nemat-Nasser S and Hori M (1999) Micromechanics Overall Properties of Heteroge-neous Materials Elsevier 810 pages

[30] Scida D Aboura Z Benzeggagh M L and Bocherens E (1999) ldquoA micromechanics model for 3D elasticity and failure of woven-fibre composite materialrdquo Composites Science and Technology 59 pp 505-517

[31] Thomas Y and Wu X H (1997) ldquoA multiscale finite element method for elliptic prob-lems in composite materials and porous mediardquo Journal of Computational Physics 134 pp 169-189

[32] Lee K H Moorthy S and Ghosh S (1999) ldquoMultiple scale computational model for damage in composite materialsrdquo Computer Methods in Applied Mechanics and Engineer-ing 172 pp 175-201

[33] Namilae S and Chandra N (2005) ldquoMultiscale model to study the effect of interfaces in carbon nanotube-based compositesrdquo Journal of Engineering Materials and Technology (ASME) 127 pp222-232

[34] Adams D E (2007) Health monitoring of Structural Materials and Components Wiley 460 pages

[35] Gopalakrishnan S (2009) ldquoModeling Aspects in Finite Elementrdquo Encyclopedia of Struc-tural Health Monitoring 2 pp791-809

[36] Palacz M Krawczuk M and Ostachowicz W (2005) ldquoThe spectral finite element mod-el for analysis of flexure-shear coupled wave propagation Part 2 Deminated multilayer composite beamrdquo Composite Structures 68 pp45-51

[37] Baik J M and Thopson R B (1984) ldquoUltrasonic scattering from imperfect interfaces A quasi-static modelrdquo Journal of Nondestructive Evaluation 4(3) pp177-196

[38] Luo H and Hanagud S (2000) ldquoDynamics of delamincated beamsrdquo International Jour-nal of Solids and Structures 37 pp1501-1519

[39] Kim S B and Sohn H (2007) ldquoInstantaneous reference-free crack detection based on polarization characteristics of piezoelectric materialsrdquo Smart Materials and Structures 16 pp2375-2387

[40] An Y K and Sohn H (2010) ldquoInstantaneous crack detection under varying temperature and static loading conditionsrdquo The Special Issue of Structural Control and Health Monitor-ing 17(7) pp 730-741

- 35 -

[41] Lee S J and Sohn H (2006) ldquoActive self-sensing scheme development for structural health monitoringrdquo Smart Materials and Structures 15(6) pp 1734-1746

- 36 -

요 약 문

복합재료 손상의 무기저 진단 및 정량화

복합재료는 재료의 경량성 내구성 고강도의 특성으로 인하여 항공기에서부터

교량에까지 널리 사용되고 있다 하지만 외부의 충격으로 인한 복합재료 내부의

손상은 재료의 강도를 감소시켜 구조물의 안정성을 저하시키는 요인이 된다 하지만

복합재료의 손상은 재료 내부에서 발생하여 육안으로는 식별하기도 힘들며 표면에서

보이는 것만으로 내부의 손상 정도를 판단하기는 불가능하다는 단점을 가지고 있다

본 연구에서는 손상 내부에서 반사로 인해 발생하는 파를 이용하여 기저자료를

사용하지 않고 특정 시점에서 얻은 자료만으로 복합재료의 손상을 진단하고 정량화

할 수 있는 기법을 제안한다 제안된 기법의 적용성을 검증하기 위하여 유한요소법을

이용한 수치해석을 수행하였고 Teflon tape 이 삽입된 시편과 실제 충격하중으로

인한 손상을 입은 시편을 이용한 실험을 진행하였다 수치해석 및 실험 결과 제안된

기법이 복합재료의 손상을 성공적으로 진단 및 정량화할 수 있음을 확인하였다 또한

다양한 외부 온도 환경에서 제안된 기법의 환경영향 실험을 수행하였고 다양한

환경에서 손상의 진단 및 정량화가 성공적으로 이루어짐을 검증하였다

핵심어 무기저 손상진단 및 정량화 압전센서 복합재료 유한요소법 Matching pursuit

method

- 37 -

감 사 의 글

기대와 두려움을 안고 시작한 석사과정이 어느덧 2 년 이라는 세월 속에서 추억으로 자리잡아

가고 있습니다 학부생 기간을 포함한 지난 3 년 동안 SHM 이라는 새로운 분야를 접하고 연구를

수행 할 수 있도록 도와주신 고마운 분들이 있어 그 감사를 표하고자 합니다

우선 저의 많은 부족한 점을 일깨워 주시고 연구에 대한 뜻을 더욱 확고히 세울 수 있게 해

주신 지도교수님이신 손 훈 교수님께 깊은 감사의 뜻을 전합니다 짧은 기간이었지만 교수님의

도움으로 많은 것을 배우고 느끼고 깨달을 수 있었습니다 또한 본 연구를 수행할 수 있도록 지

원해주신 국토해양부 (U-City 석 박사과정 지원사업)에 깊은 감사 드립니다

때로 힘들고 지칠 때 항상 저를 믿고 응원해 주시는 가족이 있었기에 지금의 글을 쓸 수 있었

습니다 또한 지금 연구실에서 함께 생활하고 있는 윤규형 민구형 철민이형 현석이형 진열이형

호민이형 병진이 선혜누나와 성대에 계신 창길이형 미국에 계신 의재형과 Debaditiya 에게도 감

사의 말씀을 전합니다 그리고 모두 나열할 수는 없지만 카이스트 생활의 활력소가 되어준 카이

스트 건설 및 환경공학과 선후배들 직원 분들과 SOS 멤버들에게도 감사를 표합니다

앞서 거론한 많은 이들의 도움이 없었다면 저 또한 지금 이 자리에 있을 수 없었을 것입니다

많이 부족하기에 더 노력하고 배우고자 했던 초심을 다시 한번 생각하며 더 큰 결실을 맺기 위해

앞으로 나아가겠습니다 도움을 주신 모든 분들께 이 감사의 글을 바칩니다

2011년 6월 카이스트에서 임형진 올림

- 38 -

CURRICULUM VITAE

Hyung Jin Lim

Research Assistant

The Civil and Environmental Engineering Department

Korea Advanced Institute of Science and Technology (KAIST)

291 Daehak-Ro Guseong-Dong Yuseong-Gu

Daejeon Republic of Korea 305-701

Phone (82)+42-350-3665 Fax (82)+42-350-3610

Email limhj87gmailcom

RESEARCH INTERESTS

Damage Detection Localization and Quantification on Composite and Metallic Structures Pattern Recognition Finite Element Modeling (FEM)

EDUCATION

2009 ndash 2011 MS Civil and Environmental Eng Korea Advanced Institute of Science and Technology (KAIST) Korea Rep

2006 ndash 2009 BS Civil and Environmental Eng Korea Advanced Institute of Science and Technology (KAIST) Korea Rep

JOURNAL PUBLICATIONS The corresponding author is underlined 1 Hyung Jin Lim Hoon Sohn ldquoReference-free damage detection localization and quantification

using in compositerdquo in preparation for Journal of the Acoustical Society of America 2011

2 Chul Min Yeum Hoon Sohn Jeong Bum Ihn Hyung Jin Lim ldquoInstantaneous delamination detection in a composite using a dual piezoelectric transducer networkrdquo Submitted to Compo-sites Science and Technology 2011

3 Yun-Kyu An Hyung Jin Lim Min Koo Kim Jin Yeol Yang Hoon Sohn Chang Guen Lee ldquoInvestigation of applicability of local reference-free damage detection techniques to in-situ

- 39 -

bridgesrdquo Submitted to ASCE Journal of Structural Engineering 2011

4 Hyung Jin Lim Min Koo Kim Hoon Sohn Chan Yik Park ldquoImpedance based damage detec-tion under varying temperature and loading conditionsrdquo Submitted to NDTampE International 2011

5 Hyung Jin Lim Hoon Sohn ldquoQuantitative estimation of transmitted and reflected lamb waves at discontinuityrdquo Journal of the Korean Society for Nondestructive Testing Vol30 No 4 pp1-8 2010

CONFERENCE PROCEEDINGS

The corresponding author is underlined 1 Hyung Jin Lim and Hoon Sohn ldquoReference-free damage detection localization and quantifi-

cation in compositesrdquo ASCE Engineering Mechanics Institute Conference Boston June 2-4 2011

2 Hyung Jin Lim and Hoon Sohn ldquoReference-free delamination quantification within multi-layer composite materialsrdquo KSNT Spring Conference May 26-27 Gyeong-Ju Korea 2011

3 Min Koo Kim Hyun Jin Lim Hoon Sohn Chan Yik Park ldquoImpedance-based bolt loosening detection under varying temperature and loadingrdquo Asian Pacific Workshop on Structural Health Monitoring Tokyo Japan November 30-December 2 2010

4 Hyung Jin Lim Chul Min Yeum and Hoon Sohn ldquoModeling of Impact-induced Delamination in a Multilayer Composite Platerdquo The 23rd KKCNN Symposium on Civil Engineering Taipei (Taiwan) November 13-15 2010

5 Yun-Kyu An Hyung Jin Lim and Hoon Sohn ldquoTemperature independent Delamination Detec-tion using data normalizationrdquo The 3rd International Conference on Multi-Functional Materi-als and Structures Chonbuk National Univ Jeonju Korea September 14-18 2010

6 Hoon Sohn Yun-Kyu An Chang Gil Lee Min Koo Kim Hyung Jin Lim ldquoApplication of online NDT techniques to decommissioned bridge testingrdquo the 5th European Workshop on Structural Health Monitoring Sorrento Italy June 29-July 02 2010

7 Min Koo Kim Hyung Jin Lim Hoon Sohn Chan Yik Park ldquoImpedance based damage diag-nosis of a complex composite aircraft wing under changing loading and temperature condi-tionsrdquo the 5th European Workshop on Structural Health Monitoring Sorrento Italy June 29-July 02 2010

8 Yun-Kyu An Min Koo Kim Hyung Jin Lim and Hoon Sohn ldquoValidations of reference-free crack detection techniques through a decommissioned bridge testrdquo 2010 COSEIK Annual Con-ference Jeju Korea April 8-9 2010

- 40 -

PARTICIPANT PROJECTS

1 Development of Smart Steel Members for Online Steel Structure Monitoring POSCO Corporation (Funded 41000000 KRW (41000 USD) for 010511 to 123111)

2 Development of Enabling Technologies for Structural Health Monitoring Korean Agency for Defense Development (Funded 240000000 KRW (240000 USD) out of total 3200000000 KRW (3200000 USD) for 060108-053111)

3 Development of On-board SHM Technologies for Composite Air Vehicles The Boeing Company (Funded 400000 USD (400000000 KRW) for 080108 to 073111)

EDUCATIONAL EXPERIENCES

Fall 2009

Teaching Assistant CE207 ldquoElementary Structural Engineering and Laboratoryrdquo KAIST 3 unit undergraduate course

EXPERIMENTAL EXPERIENTCES

2010 Monitoring of Real Aircraft Composite Wing Segment Korea Agency for Defense De-velopment (ADD) Daejeon Korea Real time monitoring of composite air vehicles fund-ed by Korea Agency for Defense Development (ADD)

2009-2011

Field Monitoring of Real Bridge Structures Gimpo (Ramp G bridge) amp Icheon (Sam-seung bridge) Korea US-Korea collaborative research for bridge monitoring testbeds

HONORS amp AWARDS

Scholarships for outstanding students granted KAIST (2007-2009) The three prominent students in the civil and environmental engineering department of KAIST were selected as recipients This scholarship provided 2000$ per year for three years

- 41 -

Page 8: 복합재료 손상의 무기저 진단 및 정량화 - KAISTssslab.kaist.ac.kr/article/pdf/Thesis_2.pdf · 1 Declaration of Ethical Conduct in Research: I, as a graduate student

LIST OF TABLES

I Material property of simulation model and test specimen middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 17

II Damage cases for numerical simulation middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 18

III Damage characterization for the numerical simulation models middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 20

IV Damage characterization for the Teflon inserted damage specimen middotmiddotmiddotmiddotmiddotmiddotmiddot 25

V Damage characterization for the impact-induced damage specimen middotmiddotmiddotmiddotmiddotmiddotmiddot 27

VI Temperature effect test for the Teflon inserted damage specimen middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 28

VII Temperature effect test for the impact induced damage specimen middotmiddotmiddotmiddotmiddotmiddotmiddot 30

iv

LIST OF FIGURES

1 Lamb wave propagation characteristics (a) A0 Lamb wave mode time delay and

attenuation after passing through a damage (b) The time delay and amplitude varia-

tion of A0 Lamb wave mode under temperature variation in pristine condition of

structure Because of this conventional damage detection techniques in composites

have limitations for real structure application middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 5

2 Multiple reflections in the damage area The multiple reflections of A0 Lamb

waves in the damage area are utilize as a feature for characterization of the damage

in this study middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 6

3 Parameters and notations of the proposed damage characterization technique

Damage is detected by identifying the additional reflected wave from the damage due

to multiple reflections (dashed lines) Localization and quantification of the damage

is carried out by measuring the arrival time of transmitted and reflected waves from

the damage middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 7

4 Two different FEM models which have different 119889119889 and 119907119907119889119889 (or G) This repre-

sents that the damage cannot be characterized uniquely based on the measured data

available middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 9

5 (a) Example of the overlapped signal with comparing approximated signal using

the matching pursuit method (b) Decomposed signal using the matching pursuit

method The matching pursuit method provides a conventional way to decompose

and identify each wave middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 9

6 Description of the numerical simulation model middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 17

7 Comparison of A0 Lamb wave mode signals from numerical simulation be-

tween intact and damage cases Time delay attenuation and multiple reflections due

to damage simulated successfully using the damage modeling method described in

v

the previous section middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 19

8 Test specimens for the experimental verification In the Teflon inserted damage

Teflon tapes with 25mm diameter were inserted at the center of the specimen The

impact-induced damage was introduced to the specimen using the impact test equip-

ment middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 21

9 Overall experimental setup (a) Data acquisition (DAQ) system and (b) Installed

circular PZT transducer on the specimen with its dimension middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 23

10 (a) Impact test equipment for introducing impact-induced damage and (b) Tip

of the impact tester with 10mm diametermiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 23

11 Test specimen in temperature chamber middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 24

12 A0 Lamb wave mode obtained using pitch-catch method from the Teflon in-

serted damage specimen (a) The reflected wave from damage is shown (b) The Tef-

lon inserted damage is detected by capturing the reflected wave middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 25

13 The signal obtained from the test specimen subjected to impact The impact

was introduced at the same location of the specimen and the signal was obtained after

each impacts As the number of impact is increased the transmitted A0 Lamb wave

mode is delayed and attenuated middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 26

14 False alarm test for the proposed damage detection algorithm in pristine condi-

tion of the specimen Reflected wave from damage is not captured false alarm is not

occurred middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 26

15 Signals obtained from the Teflon inserted damage specimen under temperature

variation Time delay and amplitude variation are occurred middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 27

16 False alarm test under temperature variations It is shown that proposed dam-

vi

age characterization algorithm is robust under temperature variationsmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 29

17 Signals obtained from impact-induced damage specimen (Impact 3) under

temperature variation middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 29

vii

CHAPTER 1 INTRODUCTION

11 Motivation

Composite materials has been increasingly demands in modern structures ranging from the

latest aircraft to retrofitted bridge deck due to its light weight superior specific strength and

stiffness properties However the occurrence of damage due to abrupt impact or sequential

accumulation of defect may considerably degrade structural strength or stiffness and reduce

the safety and reliability of the system Recently Lamb waves based damage detection tech-

niques using piezoelectric (PZT) transducers have emerged as one of the promising structural

health monitoring (SHM) technology for detecting hidden damage because of the possibility

of long range inspection and high sensitivity [1-4]

12 Literature review

This section will discuss researches that have been previously carried out on topics of

guided wave based SHM techniques using PZT transducers for damage in composite materi-

als The early investigations of interaction between Lamb wave and damage in composite can

be found in Rose et al and Guo et al [5-6] From further studies reflected signals from dam-

age were shown to be an effective method for damage diagnosis and localization using PZTs

[7-9] The reflected signals from damage were collected using one PZT as an actuator and the

other PZT as a sensor However the techniques cannot be used to determine the extent of

- 1 -

damage with reasonable precision

Kesseler et al proposed a technique monitoring the wavelet coefficients of transmitted ul-

trasonic signal amplitude as a potential way to damage detection [10] One drawback of this

technique is that the variability in coupling and the sensitivity of installed PZT transducers

can induce changes in the amplitude and energy of the signal that are cause false-alarm Oth-

er wavelet based delamination detection and localization technique developed using PZT

sensor network and outlier analysis [11]

Damage detection and quantification in composite plate was also investigated by group de-

lay measurement due to damage [12] However it is assumed that the group velocity in dam-

age area as the group velocity propagates half thickness of the structure in dispersion curve in

the quantification procedure Thus the damage quantification was investigated only Teflon

inserted damage case

Sohn et al proposed a damage detection technique which does not rely on any past base-

line signals to assess damage in composite plates by using an enhanced time reversal method

[13] The damage is assessed using the violation of time reversibility when nonlinearity is

caused by the damage along a direct wave path Thus it allows instantaneous identification

of delamination without requiring the baseline signal for comparison

- 2 -

13 Objective and uniqueness

In this study a reference-free damage characterization technique including detection lo-

calization and quantification of impact-induced damage in composite materials is developed

The general damage characterization can be categorized into the following levels [14]

Level 0 Define damage

Level 1 Determine the presence of damage within the structure

Level 2 Locate the regions of damage

Level 3 Quantify the damage

Level 4 Predict the remaining service life of the structure

Through the proposed technique from level 1 to level 3 can be achieved

Most of common approaches of damage characterization techniques described in the pre-

vious section use prior obtained baseline data However these approaches can cause false

alarm due to the effect of environmental variations such as temperature and ambient vibra-

tions [15]

The uniqueness of this study is

1 Damage is characterized without using prior measured baseline data

2 Robustness of the proposed technique under temperature variation is investigated

3 Additional waves reflected from inside damage have been used in addition to transmit-

ted and reflected waves

- 3 -

14 Organization

The organization of this thesis is summarized as follows

Chaper1 A general introduction including motivation literature review objective and

uniqueness of this study is described

Chaper2 Interactions between Lamb wave mode and damage in composites such as time

delay attenuation and multiple reflections are summarized Damage index (m-value) for

damage characterization is defined

Chaper3 Algorithms for the reference-free damage characterization technique using m-

value are developed Modified matching pursuit for this study is also described

Chaper4 Verification of the algorithm using numerical simulation is presented Various

modeled damage cases are characterized

Chaper5 Experimental verifications are described Two damage cases (Teflon inserted and

impact-induced damage) are investigated for the feasibility of the proposed damage charac-

terization technique Temperature effect test using temperature chamber is also carried out

Chaper6 Conclusion and summary of the present study is reported In advance suggestions

and future work recommendations are represented

- 4 -

CHAPTER 2 INTERACTION OF LAMB WAVE WITH DAMAGE

21 A0 Lamb wave mode delay and attenuation in the time domain

Time delay and attenuation of the A0 mode Lamb wave after passing through the damage

area due to effective thickness reduction were reported in previous researches [812] Moreo-

ver the attenuation of the A0 Lamb wave mode is occurred by degrading of shear modulus of

in the damage area However the time delay and amplitude variation of the A0 Lamb wave

mode also occurred even when the structure is health condition by the material property

changing due to temperature variation of the structure (Fig 1) Thus conventional damage

characterization methods using the time delay and attenuation of A0 Lamb wave mode as a

feature can cause false-alarm in real application

11 115 12 125 13-10

-5

0

5

10

Time (ms)

Vol

tage

(mV

)

IntactDamage

Reflected wave from Damage

(a)

11 115 12 125 13-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

0 oC20 oC50 oC

(b)

Fig 1 Lamb wave propagation characteristics (a) A0 Lamb wave mode time delay and atten-uation after passing through a damage (b) The time delay and amplitude variation of A0 Lamb wave mode under temperature variation in pristine condition of structure Because of this conventional damage detection techniques in composites have limitations for real struc-ture application

- 5 -

22 Multiple reflections in the damage area

Standing wave patterns due to multiple reflections of A0 Lamb wave mode in the damage

area were observed through numerical simulations and ultrasonic wave-field imaging using a

scanning laser vibrometer [16] The formation of standing waves in damage is illustrated

through Fig 2 When the incident waves enter the damage area a significant portion of the

waves are reflected back from the exit Then the reflected waves travel through the damage

and undergo reflection again at the original entrance As a consequence of multiple reflec-

tions at the entrance and the exit a considerable amount of ultrasonic energy is trapped inside

the damage area [17-19]

Multiple Reflections

Incident Wave Transmitted Wave

Damage

Fig 2 Multiple reflections in the damage area The multiple reflections of A0 Lamb waves in the damage area are utilize as a feature for characterization of the damage in this study

- 6 -

CHAPTER 3 THEORETICAL DEVELOPMENT

31 Overview

Damage

l = l1 + d + l2

d = vdtd l1 = vit1 l2 = vit2

t = t1 + td + t2

Excitation and Sensing PZT Sensing PZT

l

Transmitted Wave

Reflected Waves

Incident Wave

Fig 3 Parameters and notations of the proposed damage characterization technique Damage is detected by identifying the additional reflected wave from the damage due to multiple re-flections (dashed lines) Localization and quantification of the damage is carried out by measuring the arrival time of transmitted and reflected waves from the damage

In this section the reference-free damage characterization algorithm is proposed Damage

diagnosis and quantification are achieved by identifying the additional reflected waves due to

multiple reflections in the damage area (dashed lines in Fig 3) Moreover damage is local-

ized using additional information from pulse-echo method implement For the localization

and the quantification of damage the variables such as the path length and A0 Lamb wave

mode velocity in damage area are defined (Fig 3) For example 119897119897 is expressed as

119897119897 = 1198971198971 + 119889119889 + 1198971198972 (1)

where 1198971198971 119889119889 and 1198971198972 are the distance from the excitation PZTs to the entrance of damage the

damage length and the distance from the exit of damage to the sensing PZT respectively

- 7 -

Moreover 119907119907119894119894 and 119907119907119889119889 are also defined as the A0 Lamb wave mode velocity in pristine area

and damage area respectively Note that the whole procedure of the damage characterization

is accomplished based solely on the signals obtained from the current state of the structure

32 Damage index (m-value)

The proposed reference-free damage characterization technique utilizes the waves reflect-

ed additionally from the damage area as a feature The arrival time difference of the transmit-

ted and reflected A0 Lamb wave modes reflected from the exit and the entrance of the delam-

ination area one times each (∆119905119905) contains the information of the physical length of the dam-

age (119889119889) and A0 Lamb wave mode velocity in the damage area (119907119907119889119889) The wave velocity in the

damage area which is related to the shear modulus (G) represents the severity of the damage

∆119905119905 = 2119905119905119889119889 =2119889119889119907119907119889119889

(2)

However based on the measured data available the 119889119889 and 119907119907119889119889 (or G) cannot be uniquely

determined Fig 4 shows two FEM models which have different physical length and wave

velocity (G) of the damage Detail explanation of delamination modeling technique will be

covered in section 4 The similar signal in the time domain and ∆119905119905 are obtained from the

two different damage cases

- 8 -

Damage 1

Damage 2

20

Unit mm15

90 G value reduction

943 G value reduction

m-value = 248

m-value = 248

015 02 025 03 035

-2

-1

0

1

2

Time (ms)

Ampl

itude

Damage 1Damage 2

Fig 4 Two different FEM models which have different 119889119889 and 119907119907119889119889 (or G) This represents that the damage cannot be characterized uniquely based on the measured data available

Thus in this study the damage index (m-value) which can be uniquely defined based on

the available data is introduced The m-value is defined as the ratio of the physical length of

the damage and wavelength of the A0 Lamb wave mode in the damage area (120582120582119889119889)

119898119898 =119889119889120582120582119889119889

=119889119889119889119889119907119907119889119889

(3)

In many cases the transmitted and reflected waves are overlapped and it is hard to decom-

pose and calculate the arrival times of them Matching pursuit method is an alternative way

to accurately decompose of the overlapped waves and estimate the parameters of each wave

such as arrival time (Fig 5) Detail explanation of the matching pursuit method is covered in

section 36

015 02 025 03-2

-1

0

1

2

Time (ms)

Am

plitu

de

OriginalApproximated

(a)

015 02 025 03-2

-1

0

1

2

Time (ms)

Ampl

itude

Wave 1Wave 2

(b)

Fig 5 (a) Example of the overlapped signal with comparing approximated signal using the matching pursuit method (b) Decomposed signal using the matching pursuit method The matching pursuit method provides a conventional way to decompose and identify each wave

- 9 -

33 Damage detection algorithm (Level 1)

The damage diagnosis is accomplished by extracting the additional reflected A0 Lamb

wave modes from damage due to multiple reflections If A0 Lamb wave modes are decom-

posed perfectly it is trivial that one transmitted A0 Lamb wave mode is always captured in

the response signal If damage exists other waves in the response signal behind the transmit-

ted A0 Lamb wave mode are presented Since this approach does not rely on previously col-

lected data the diagnosis is independent from environmental variation such as temperature

changes

It is reported that the dispersion of Lamb waves cause wave-packets to spread out in space

and time as they propagate through a structure [20] Thus the reflected waves from the dam-

age have longer pulse duration than the transmitted wave due to their longer propagating dis-

tance In this study the matching pursuit method is modified to capture more spread wave

than the transmitted A0 Lamb wave mode Detail explanation of the modified matching pur-

suit method is described in latter section

Let 1199041199041 and 1199041199042 as the pulse duration (scale) of the transmitted and additional reflected A0

Lamb wave mode respectively

ldquoIf 1199041199042 gt 1199041199041 damage exists

Otherwise there is no damagerdquo

- 10 -

34 Damage quantification algorithm (Level 3)

As mentioned in the previous section damage is quantized by estimating 119889119889 and 119907119907119889119889 using

pitch-catch method The procedure to estimate 119889119889 and 119907119907119889119889 is

(i) Measure 119897119897 119905119905 and 119905119905119889119889

(ii) Assume the range of 119889119889 (0 lt 119889119889 lt 119897119897)

(iii) Calculate 119907119907119894119894 and 119907119907119889119889 with a constraints (0 lt 119907119907119889119889 lt 119907119907119894119894) (Eq 4 and 5)

(iv) Estimate m-value and find a feasible range of 119889119889 and 119907119907119889119889

119907119907119894119894 =119897119897 minus 119889119889119905119905 minus 119905119905119889119889

(4)

119907119907119889119889 =119889119889119905119905119889119889

(5)

35 Damage localization algorithm (Level 2amp3)

For damage localization 1198971198971 is estimated by measuring the arrival time from the excitation

PZT to the entrance of delamination (1199051199051) with additional pulse-echo method implementation

The procedure is similar with the previous damage quantification algorithm

(i) Measure 119897119897 119905119905 119905119905119889119889 and 1199051199051

(ii) Assume the range of 1198971198971 (0 lt 1198971198971 lt 119897119897)

(ii) Compute possible combinations of 1198971198972 119889119889 and 119907119907119889119889 with constraints

(0 lt 119907119907119889119889 lt 119907119907119894119894 119889119889 gt 0) (Eq 5 6 and 7)

- 11 -

1198971198972 = 1198971198971119905119905 minus 1199051199051 minus 119905119905119889119889

1199051199051 (6)

119889119889 = 119897119897 minus 1198971198971 minus 1198971198972 (7)

Since a PZT for pulse-echo method acts as both an actuator and a sensor the obtained sig-

nal contains both the excitation and the response signal Thus the reflected wave from the

entrance of damage can be buried into the tail of excitation signal and cause error on arrival

time of reflected wave from damage Even in this case damage can be localized and quan-

tized using the constraints When first wave packet reflected from the entrance of damage is

buried into the excitation signal the next wave packet reflected from the exit of damage can

be utilized alternatively If the algorithm regards the second reflected wave as the first re-

flected wave there are no combinations of estimated variable meeting the constraints Then

the algorithm repeats the process by regarding the wave as the second reflected wave the

constraints can select proper combination of the estimated variables

The estimated values have range bounded by the constraints When additional information

of 119907119907119894119894 is utilized the damage can be characterized more accurately In this study three meth-

ods are proposed and verified as information level of 119907119907119894119894

Method 1 Solely based on the current data measurement (Reference-free)

Method 2 Limit the possible range of 119907119907119894119894 (Prior knowledge of the system)

Method 3 Measure 119907119907119894119894 (Additional data need to be collected)

Method 1 described in this section can be utilized without any information of 119907119907119894119894 Meth-

od 2 produces more accurate estimation results using prior knowledge of the system such as

- 12 -

119907119907119894119894 variation under temperature changes Method 3 gives unique estimated value using meas-

ured 119907119907119894119894 from other path in the same time In section 4 and 5 damage is characterized using

these three methods and the results are compared each another

36 Matching pursuit method considering the dispersion effect of Lamb

wave

Matching pursuit method has been developed for signal compression approximation and

characterization of obtained signal and function [21-23] The key concept of the matching

pursuit method is that any signal (function) can be decomposed into a linear expansion of

signals selected from a pool of basis signals (functions) called lsquodictionaryrsquo

Let 119891119891 be a function to be analyzed then 119891119891 can be approximated a set of functions select-

ed from the dictionary 119863119863 in order to best match This is done by successive approximations

with orthogonal projections Let1198921198921205741205740 isin 119863119863 the function 119891119891 can be decomposed into

119891119891 = lang1198911198911198921198921205741205740rang1198921198921205741205740 + 119877119877119891119891 (8)

where 119877119877119891119891 is the residual function after approximating 119891119891 using 1198921198921205741205740 1198921198921205741205740 is orthogonal to

119877119877119891119891 hence

1198911198912 = lang1198911198911198921198921205741205740rang2

+ 1198771198771198911198912 (9)

To minimize 119877119877119891119891 1198921198921205741205740 is chosen to maximize lang1198911198911198921198921205741205740rang The matching pursuit method is

an iterative algorithm that decomposes the residue 119877119877119891119891 by projecting it on a function of 119863119863

that matches 119877119877119891119891 almost at best as it was done for 119891119891 The residue 119877119877119899119899119891119891 is decomposed into

119877119877119899119899119891119891 = lang119877119877119899119899119891119891119892119892120574120574119899119899rang119892119892120574120574119899119899 + 119877119877119899119899+1119891119891 (10)

- 13 -

the decomposition is carried up to the order 119898119898 119891119891 can be expressed as

119891119891 = lang119877119877119899119899119891119891119892119892120574120574119899119899rang119892119892120574120574119899119899

119898119898minus1

119899119899=0

+ 119877119877119898119898119891119891 (11)

The original function 119891119891 is decomposed into a sum of dictionary elements that are chosen to

best match its residues

In this study the matching pursuit method is utilized to accurately decompose of the over-

lapped Gaussian pulses and estimate the parameters of the pulse which used for damage

characterization The 119894119894119905119905ℎ Gaussian pulse in the dictionary for approximating the obtained

signal is expressed as

119892119892120574120574119894119894 = 119891119891(119860119860119894119894 119906119906119894119894 119904119904119894119894 119889119889119894119894120601120601119894119894) (12)

where 119860119860119894119894119906119906119894119894119904119904119894119894119889119889119894119894 and 120601120601119894119894 represent amplitude arrival time scale (pulse duration) excita-

tion(central) frequency and phase of the Gaussian pulse respectively The parameters of the

pulse are estimated using a nonlinear regression based optimization method (Gauss-Newton

method) based on an initial guess of the parameters Detail procedure for estimating the pa-

rameters of the pulse is expressed in Hong et al [23]

Among the estimated parameters of the pulses pulse duration (scale) represents the disper-

sion effect If a wave-packet is spread out the estimated pulse duration is increased Using

this feature the matching pursuit method is modified to capture waves which have longer

pulse duration than transmitted A0 Lamb wave mode If the estimate pulse duration of a wave

is less than the transmitted wave the modified matching pursuit method regards it as error

- 14 -

CHAPTER 4 NUMERICAL SIMULATION

41 Composite modeling method

Composite materials can be modeled from nano to macro level depends on its engineering

purpose In general four methods are utilized for modeling of composite materials First

composite is modeled as a homogenous material expressed in terms of effective material

properties [24] The effective material properties are estimated using the modulus and the

volume fraction of each constituent theoretically (Rule of mixture) and measured experimen-

tally Second a model is established that each laminate which have own material properties

are totally bonded together In this model the stacking sequence and direction of each lami-

nates are considered additionally [25-27] Third method for composite modeling in micro

level called micromechanics [28-30] This model considers the interaction between fibers and

matrix in laminate Last multiscale modeling which the scope of modeling is extended to

molecular or quantum level of constituents [31-33] This model starts from molecular me-

chanics or molecular dynamics then the scale of model leads to the nano micro and macro

level by incorporating small scale models As the scale of models is smaller it provides more

logical way than other large scale of models while the computational cost is more expensive

The scope of this study is focused on the dynamic behavior of composite structure the beam

or plate model was utilized for the numerical simulation

- 15 -

42 Damage modeling method

Generally three damage models have been developed and utilized in the real applications

First the stiffness reduction model called as the equivalent element model was introduced

[17 34-35] This model is simply realized by replacing the region around damage area by a

region with adjusted material properties (Youngrsquos modulus shear modulus density etc)

globally Next the duplicated node model is done by disconnecting the element connectivity

assuming there is no interaction among the delaminated surface This model can be utilized

when the location and the extension of damage is important issue [34-36] Last the contact

mechanics model using the principles of contact mechanics the two interfacial forces (con-

tact resistance and sliding resistance) in the delaminated surface [37-38] In this model the

delaminated surfaces are modeled as nonlinear functions of the interlaminar normal strain

and the interlaminar shear strain This model is computationally expensive while provides

more logical way than other models There is however no universally accepted model for

general impact-induced damage in composites In the case of impact-induced damage it is

impossible to predict or estimate the location and number of delaminated layers Thus from

the uncertainty of impact-induced damage the equivalent element model can be considered

as the effective damage model

- 16 -

43 Numerical simulation description

TABLE I Material property of simulation model and test specimen

Tensile Strength 657 MPa Tensile Modulus 59 GPa Shear Strength 804 MPa Shear Modulus 24 Gpa Specific Gravity 16

PZT A

PZT C

PZT BDamage

3

250 125 125

500 Unit mm

Fig 6 Description of the numerical simulation model

The feasibility of the proposed damage characterization technique is verified by 2-D FEM

simulation using MSC PATRAN and NASTRAN Fig 6 presents the dimension PZTs and

damage location of the simulation model In the model 05mm x 05mm square shell element

was used for modeling of the A0 mode whose flexural wave characteristic The element size

was determined by considering the wavelength of the propagation Lamb wave modes The

material properties of the simulation model and damage cases are presented in Table I and

Table II respectively

PZT A and C act as both actuator and sensor for collecting pulse-echo signal They were

installed on the plate which two of them are both-side collocated for the A0 Lamb wave

- 17 -

mode decomposition [39-40] PZT B was only used as a sensor for collecting pitch-catch sig-

nal For simulating the excitation and sensing of PZTs the pin-force model was adopted that

an input force apples to the end nodes of excitation PZT and averaging element strains corre-

sponding sensing area A Gaussian pulse (tone-burst) 50 kHz central frequency was excited

as an input signal

TABLE II Damage cases for numerical simulation

119949119949120783120783 (mm) 119941119941 (mm) G value reduction ()

Case 1 650 200 90 darr

Case 2 675 150 90 darr

Case 3 675 150 93 darr

Case 4 900 150 90 darr

- 18 -

44 Numerical simulation results

015 02 025 03 035

-2

0

2

Time (ms)

Am

plitu

de

IntactDamage

Reflected wave from damage

(a) Pitch-catch signal

015 02 025 03 035

-2

0

2

Time (ms)

Am

plitu

de

IntactDamage

Reflected waves from damage

(b) Pulse-echo signal Fig 7 Comparison of A0 Lamb wave mode signals from numerical simulation between intact and damage cases Time delay attenuation and multiple reflections due to damage simulated successfully using the damage modeling method described in the previous section

Fig 7 shows that comparison of A0 Lamb wave mode signals between intact and one of

damage cases in the simulation The features of A0 Lamb wave mode passing through dam-

age area described in section 2 are successfully simulated Thus the equivalent element

method is proper for modeling damage in composites In the damage cases the additional

waves reflected from damage is observed in both pitch-catch and pulse echo signal

Table III presents the result of the proposed damage characterization algorithm using the

numerical simulation cases shown in Table II In lsquoMethod 2rsquo 119907119907119894119894 variation is assumed to plusmn

5 of the normal value (1190~1315ms) The normal value of 119907119907119894119894 for lsquoMethod 3rsquo is meas-

ured as 1253 ms In each case the damage is characterized using the proposed technique

Note that as the information of 119907119907119894119894 is specified the estimated value is closer to the exact

value

- 19 -

TABLE III Damage characterization for the numerical simulation models

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

Case 1

Exact 650 200

Method 1 248 50~70 79~414 160~834

Method 2 248 603~667 135~241 271~486

Method 3 248 635 188 379

Case 2

Exact 675 150

Method 1 193 60~75 01~251 37~651

Method 2 193 618~682 114~222 295~575

Method 3 193 65 168 435

Case 3

Exact 675 150

Method 1 233 60~75 66~302 140~649

Method 2 233 668~738 85~196 182~419

Method 3 233 703 140 301

Case 4

Exact 900 150

Method 1 204 75~100 48~349 118~856

Method 2 204 860~951 107~216 264~530

Method 3 204 906 162 397

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1253 ms 1190~1315 ms Measured 119907119907119894119894 1253 ms

- 20 -

CHAPTER 5 EXPERIMENTAL VERIFICATION

51 Overview

In this section experimental study is investigated so that the proposed damage characteri-

zation technique Overall experimental setup and test specimen are shown in Fig 8 and 9

The experimental setup consists of test specimens data acquisition (DAQ) system and self-

sensing circuit Self-sensing circuit was installed for pulse-echo measurement using PZTs

[40-41]

PZT A(PZT C) PZT B

500

500

140

75

250

Teflon Inserted Damage (d =25)

Unit mm

Thickness = 3

PZT A(PZT C) PZT B

500

500

150

90

250

Impact Location

Unit mm

Thickness = 3

(a) Teflon inserted damage specimen

(b) Impact-induced damage specimen

Fig 8 Test specimens for the experimental verification In the Teflon inserted damage Teflon tapes with 25mm diameter were inserted at the center of the specimen The impact-induced damage was introduced to the specimen using the impact test equipment

- 21 -

511 Specimen

Two carbon composite plates with a dimension of 500times500times3mm3 were manufactured by

Nexcomes Inc by stacking 12 layers of CF3327 3K prepreg (woven fabrics type) manufac-

tured by Hankuk Fiber Inc The material properties of the specimen are presented in Table I

One of the specimens Teflon tapes with 25mm diameter was inserted in every 3 layers at the

center of the specimen to mimic the damage The other specimen is utilized for the character-

ization of real impact-induced damage

512 Data acquisition (DAQ) system and PZT

The data acquisition (DAQ) system consists of an arbitrary waveform generator (AWG) a

high speed signal digitizer (DIG) and two multiplexers (MUX) Using the 16-bits AWG

tone-burst input signal with a plusmn10 peak-to-peak voltage were generated The response signal

was measured by the 14-bit DIG For the signal-to-noise ratio reduction both pitch-catch and

pulse-echo signals were measured 150 times and averaged in the time domain 3 identical

circular PZTs of 10mm diameter manufactured by APC International Ltd were installed on

the specimen A Gaussian pulse (tone-burst) 50 kHz central frequency was excited as an in-

put signal The excitation frequency was chosen in order to avoid overlapping of the S0 and

A0 Lamb wave modes and generating of the higher modes (S1 A1hellip)

- 22 -

PZT Actuator

PZT Sensor

AWG

DIG

DAQ System Specimen

SensingActuatingMUX

1

23

4

1 Sending signal to MUX

4 Sending signal to DIG

Self-SensingCircuit

18 Unit mm

10

(a)

(b)

Fig 9 Overall experimental setup (a) Data acquisition (DAQ) system and (b) Installed circu-lar PZT transducer on the specimen with its dimension

513 Impact test

The impact-induced damage was introduced to the specimen using the impact test equip-

ment presented in Fig 10 A 5kg mass with 10mm tip was dropped from 015m height onto

the same location of the specimen The signals were obtained at each number of impacts

Since the reflected wave is not detected at the first impact case the data obtained from the

second and third impact were utilized for the analysis

10

Unit mm (a)

(b)

Fig 10 (a) Impact test equipment for introducing impact-induced damage and (b) Tip of the impact tester with 10mm diameter

- 23 -

514 Temperature effect test

In reality structures are subjected to changing environmental conditions such as tempera-

ture that affect measured signal and cause false-alarm Thus the robustness of the proposed

damage characterization technique under temperature variation in investigated (Fig 11) The

specimens were placed inside a temperature chamber and a thermocouple was installed on

the specimen to measure temperature of surface of the specimen The signal was obtained

under three temperature cases (0 20 and 50degC) and the humidity was kept at 30

Temperature Chamber

Test Specimen

Fig 11 Test specimen in temperature chamber

- 24 -

52 Teflon inserted damage

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Reflected wave from damage

(a)

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Transmitted WaveReflected Wave

Detected reflected wave

(b)

Fig 12 A0 Lamb wave mode obtained using pitch-catch method from the Teflon inserted damage specimen (a) The reflected wave from damage is shown (b) The Teflon inserted damage is detected by capturing the reflected wave

Fig 12 shows the obtained pitch-catch signal from the Teflon inserted damage specimen

and damage diagnosis result The Teflon inserted damage is detected successfully by captur-

ing the reflected wave Table IV presents the result of the Teflon inserted damage characteri-

zation 119907119907119894119894 is measured as 1213ms and the variation is assumed to plusmn 5 of the normal val-

ue (1152~1273ms) The results indicate that the proposed damage characterization approach

works well in real application

TABLE IV Damage characterization for the Teflon inserted damage specimen

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

Exact 635 25

Method 1 123 60~75 73~338 298~1220

Method 2 123 619~662 229~310 934~1188

Method 3 123 631 284 1161

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1213 ms

- 25 -

53 Impact-induced damage

11 115 12 125 13-4

-2

0

2

4

Time (ms)

Vol

tage

(mV

)

IntactImpact2Impact3

Reflected wave from damage

(a) Pitch-catch signal

11 115 12 125 13 135

-05

0

05

1

Time (ms)

Volta

ge (m

V)

IntactImpact2Impact3

Boundary reflection

Reflected wave from damage

(b) Pulse-echo signal

Fig 13 The signal obtained from the test specimen subjected to impact The impact was in-troduced at the same location of the specimen and the signal was obtained after each impacts As the number of impact is increased the transmitted A0 Lamb wave mode is delayed and attenuated

The obtained signals from impact-induced damage specimen are presented in Fig 13 It is

observed that as the number of impact is increased the transmitted A0 Lamb wave mode is

more delayed because the wave velocity in the damage area is reduced Fig 14 shows the

false-alarm test result whether the proposed damage detection algorithm works even in pris-

tine condition of the test specimen It is shown that no false alarm is occurred because no ad-

ditional reflected wave is captured

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Transmitted Wave

(a) Transmitted wave

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Reflected Wave

No reflected wave

(b) Reflected wave

Fig 14 False alarm test for the proposed damage detection algorithm in pristine condition of the specimen Reflected wave from damage is not captured false alarm is not occurred

- 26 -

TABLE V Damage characterization for the impact-induced damage specimen

of Impacts m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

2

(1470 J)

Method 1 080 65~90 75~246 470~1157

Method 2 080 801~885 114~245 711~1134

Method 3 080 843 179 1122

3

(2205 J)

Method 1 148 50~100 06~380 21~1134

Method 2 148 758~838 249~368 843~1125

Method 3 148 798 308 1044

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1213 ms

Table V presents the characterization results of the impact-induced damage case Here the

measured 119907119907119894119894 and the assumed variation are same as the artificial damage case It is observed

that as the number of impact is increased 119889119889 is increased while 119907119907119889119889 is reduced Thus it is

validated that the proposed damage characterization approach estimates the increment of the

damage well

54 Temperature effect test

541 Teflon inserted damage

11 115 12 125 13-10

-5

0

5

10

Time (ms)

Vol

tage

(mV

)

0 oC20 oC50 oC

(a) Pitch-catch signal

11 115 12 125 13 135

-1

0

1

2

3

Time(ms)

Volta

ge (m

V)

0 oC20 oC50 oC

(b) Pulse-echo signal

Fig 15 Signals obtained from the Teflon inserted damage specimen under temperature varia-tion Time delay and amplitude variation are occurred

Fig 15 shows the obtained signals from the Teflon inserted damage specimen subjected to

- 27 -

temperature variation Time delay and amplitude variations are occurred there is no addition-

al damage is introduced The analyzed results of the obtained signals are reported in Table VI

The measured 119907119907119894119894 is 1247 ms at 0 degC 1213 ms at 20 degC and 1168 ms at 50 degC respectively

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms (1152~1273 ms)

The estimated results are also changed under temperature variations the fluctuation of esti-

mated range and exact value is small

TABLE VI Temperature effect test for the Teflon inserted damage specimen

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

0 degC

Exact 635 25

Method 1 113 60~70 79~268 352~1291

Method 2 113 560~619 233~344 1034~1172

Method 3 113 606 257 1141

20 degC

Exact 635 25

Method 1 123 60~75 73~338 298~1220

Method 2 123 619~662 229~310 934~1188

Method 3 123 631 284 1161

50 degC

Exact 635 25

Method 1 137 60~75 72~328 263~1228

Method 2 137 622~687 183~299 666~1086

Method 3 137 642 264 960

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1247 ms at 0 degC 1213 ms at 20 degC and 1168 ms at 50 degC

- 28 -

542 Impact-induced damage

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Vol

tage

(mV

)

Transmitted WaveReflected Wave

No reflected wave

(a) 0 degC

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Transmitted WaveReflected Wave

No reflected wave

(b) 50 degC

Fig 16 False alarm test under temperature variations It is shown that proposed damage char-acterization algorithm is robust under temperature variations

The false alarm test in pristine condition of the impact-induced damage specimen under

temperature variation was investigated (Fig 16) It is validated that the proposed damage di-

agnosis algorithm is robust under temperature variations Fig 17 shows the obtained signals

from the impact-induced damage (impacted 3 times) specimen under temperature variation

The analyzed results of the obtained signals are reported in Table VII The measured 119907119907119894119894 in

each temperature and the assumed variation are same as the artificial damage case

11 115 12 125 13-5

0

5

Time (ms)

Volta

ge (m

V)

0 oC20 oC50 oC

(a) Pitch-catch signal

11 115 12 125 13 135

-1

0

1

2

3

Time(ms)

Vol

tage

(mV

)

0 oC20 oC50 oC

(b) Pulse-echo signal

Fig 17 Signals obtained from impact-induced damage specimen (Impact 3) under tempera-ture variation

- 29 -

TABLE VII Temperature effect test for the impact induced damage specimen (Impact 3 case)

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

0 degC

Method 1 135 75~90 72~310 267~1148

Method 2 135 705~779 264~381 978~1126

Method 3 135 763 289 1071

20 degC

Method 1 148 50~100 06~380 21~1134

Method 2 148 758~838 249~368 843~1125

Method 3 148 798 308 1044

50 degC

Method 1 153 75~100 53~415 174~1227

Method 2 153 804~889 214~336 702~1103

Method 3 153 815 320 1050

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1247 ms at 0 degC 1213 ms at 20 degC and 1168 ms at 50 degC

- 30 -

CHAPTER 6 CONCLUDING REMARKS

61 Executive summary

In this study the reference-free damage detection localization and quantification tech-

nique in a multi layered composite plate is proposed Additional reflected A0 Lamb wave

mode from the damage due to multiple reflections in damage area is used as a feature for

characterizing the damage First damage is characterized in terms of a damage index (m-

value) defined as the ratio of the physical length of the damage to the wavelength of the A0

Lamb wave mode Then the range of possible damage location length and severity are esti-

mated with constraints This procedure can be accomplished without prior baseline data To

realize the proposed technique the matching pursuit technique is modified for reduction of

the error by considering the dispersive characteristic of Lamb wave The matching pursuit

method estimates the arrival time of each wave packets accurately even two wave packets are

overlapped The numerical simulations and experimental tests including Teflon tape inserted

and impact-induced damage are investigated for the feasibility of the proposed technique

Moreover the effectiveness of the proposed technique under changing temperature is exper-

imentally verified

- 31 -

62 Suggestion for further study

The limitations of this study and the possible improvement and further studies are listed as

follow

1 Measurement of impact-induced damage length using conventional NDT technique

Additional experimental verification for impact-induced damage characterization is

needed by measuring the real length of damage using conventional NDT technique

such as ultrasonic scan or infrared camera

2 Limitation of both side collocated PZT installation In this study excitation PZTs are

installed both side and collocated for decomposing A0 Lamb wave mode However it

can produce decomposition error due to imperfection of PZT location and their installa-

tion Moreover in many cases assess to both surface of structures is not allowed such

as aircraft applications To improve this limitation a reference-free damage characteri-

zation using single side PZTs will be considered

3 Complex structure application In this study damage is characterized in simple plate-

like structure When the monitored structure becomes complex the response signals al-

so becomes complex due to its geometry and boundary Thus the investigations for

complex structure application such as stiffener spar and curvature are needed

- 32 -

REFERENCES

[1] Cawley P and Alleyne D (1996) ldquoThe use of Lamb waves for the long range inspection of large structuresrdquo Ultrasonics 34 pp287-290

[2] Ghosh T Kundu T and Karpur P (1998) ldquoEfficient use of Lamb waves for detecting defects in large platesrdquo Ultrasonics 36 pp791-801

[3] Guirguitui V (2003) ldquoLamb wave generation with piezoelectric wafer active sensors for structural health monitoringrdquo SPIE Conference on Smart Structures and Materials San Diego CA 5391 pp111-422

[4] Ragvahan A and Cesnik CE S (2004) ldquoModeling of piezoelectric-based Lamb wave generation and sensing for structural health monitoringrdquo SPIE Conference on Smart Struc-tures and Materials San Diego CA 5391 pp419-430

[5] Rose JL Pilarski A and Dirit JJ (1993) ldquoAn approach to guided wave mode selection for inspection of laminated platerdquo Journal of Reinforced Plastics and Composites 12(5) pp536-544

[6] Guo N and Cawley P (1993) ldquoThe interaction of Lamb waves with delaminations in composite laminatesrdquo Journal of Acoustical Society of America 94(4) pp 2240-2246

[7] Valdez SHD and Soutis C (2002) ldquoReal-time nondestructive evaluation of fiber com-posite laminates using low-frequency Lamb wavesrdquo Journal of Acoustical Society of Amer-ica 111(5) pp 2026-2033

[8] Yang M and Qiao P (2005) ldquoModeling and experimental detection of damage in virous materials using the pulse-echo method and piezoelectric sensoractuatorsrdquo Smart Materials and Structures 14 pp1083-1100

[9] Lestari W and Qiao P (2005) ldquoApplication of wave propagation analysis for damage identification in composite laminated beamsrdquo Journal of Composite Materials 39(22) pp1967-1984

[10] Kessler S S Spearing S M and Soutis C (2002) ldquoDamage detection in composite ma-terials using Lamb wave methodsrdquo Smart Materials and Structures 11 pp269-278

[11] Sohn H Park G Wail J R Limback N P and Farrar C R (2004) ldquoWavelet-based active sensing for delamination detection in composite structuresrdquo Smart Materials and Structures 13 pp 153-160

[12] Petculescu G Krishnaswamy S and Achenback J D (2008) ldquoGroup delay measure-ments using modally selective Lamb wave transducers for detection and sizing of delamina-tions in compositesrdquo Smart Materials and Structures 17

[13] Sohn H Park H W Law K H and Farrar C R (2007) ldquoDamage detection in compo-

- 33 -

site plates by using an enhanced time reversal methodrdquo Journal of Aerospace Engineering (ASCE) pp 141-151

[14] Sohn H (2007) ldquoEffects of environmental and operational variability on structural health monitoringrdquo Philosophical Transactions of the Royal Society Ardquo 365 pp539-560

[15] Rytter A (1993) ldquoVibration based inspection of civil engineering structuresrdquo PhD The-sis Department of Building Technology and Structural Engineering University of Aalborg Aalborg Denmark

[16] Hayashi T and Kawashima K (2002) ldquoMultiple reflections of Lamb waves at a delami-nationrdquo Ultrasonics 40 pp193-197

[17] Wang C H and Rose L R F (2003) ldquoWave reflection and transmission in beams con-taining delamination and inhomogeneityrdquo Journal of Sound and Vibration 264 pp 851-872

[18] Ramadas C Balasubramaniam K Joshi M and Krishnamurthy C V (2009) ldquoInterac-tion of the primary anti-symmetric Lamb mode (Ao) with symmetric delaminations numer-ical and experimental studiesrdquo Smart Materials and Structures 18

[19] Ramadas C Balasubramaniam K Joshi M and Krishnamurthy C V (2010) ldquoInterac-tion of guided Lamb waves with an asymmetrically located delamination in a laminated composite platerdquo Smart Materials and Structures 19

[20] Wilcox P Lowe M and Cawely P (2001) ldquoThe effect of dispersion on long-range in-spection using ultrasonic guided wavesrdquo NDTampE International 34 pp1-9

[21] Mallat S G and Zhang Z (1993) ldquoMatching pursuits with time-frequency dictionariesrdquo IEEE Tansactions on Signal Processing 41(12) pp3397-3415

[22] Zhang G Zhang S and Wang Y (2000) ldquoApplication of adaptive time-frequency de-composition in ultrasonic NDE of highly scattering materialsrdquo Ultrasonics 38 pp 961-964

[23] Hong J C Sun K H and Kim Y Y (2005) ldquoThe matching pursuit approach based on the modulated Gaussian pulse for efficient guided-wave damage inspectionrdquo Smart Mate-rials and Structures 14 pp548-560

[24] Herakovich C T (1997) Mechanics of Fibrous Composites Wiley 480 pages

[25] Reddy J N (1985) ldquoA review of the literature on finite-element modeling of laminated composite platesrdquo The Shock and Vibration Digest 17(4) pp 3-8

[26] Reddy J N (2003) Mechanics of Laminated Composite Plates and Shells Theory and Analysis CRC Press 856 pages

[27] Palacz M Krawczuk M and Ostachowicz W (2005) ldquoThe spectral finite element mod-el for analysis of flexure-shear coupled wave propagation Part 1 laminated multilayer

- 34 -

composite beamrdquo Composite Structures 68 pp37-44

[28] Zhang Y C and Harding J (1990) ldquoA numerical micromechanics analysis of the me-chanical properties of a plain weave compositerdquo Computer amp Structures 36(5) pp 839-844

[29] Nemat-Nasser S and Hori M (1999) Micromechanics Overall Properties of Heteroge-neous Materials Elsevier 810 pages

[30] Scida D Aboura Z Benzeggagh M L and Bocherens E (1999) ldquoA micromechanics model for 3D elasticity and failure of woven-fibre composite materialrdquo Composites Science and Technology 59 pp 505-517

[31] Thomas Y and Wu X H (1997) ldquoA multiscale finite element method for elliptic prob-lems in composite materials and porous mediardquo Journal of Computational Physics 134 pp 169-189

[32] Lee K H Moorthy S and Ghosh S (1999) ldquoMultiple scale computational model for damage in composite materialsrdquo Computer Methods in Applied Mechanics and Engineer-ing 172 pp 175-201

[33] Namilae S and Chandra N (2005) ldquoMultiscale model to study the effect of interfaces in carbon nanotube-based compositesrdquo Journal of Engineering Materials and Technology (ASME) 127 pp222-232

[34] Adams D E (2007) Health monitoring of Structural Materials and Components Wiley 460 pages

[35] Gopalakrishnan S (2009) ldquoModeling Aspects in Finite Elementrdquo Encyclopedia of Struc-tural Health Monitoring 2 pp791-809

[36] Palacz M Krawczuk M and Ostachowicz W (2005) ldquoThe spectral finite element mod-el for analysis of flexure-shear coupled wave propagation Part 2 Deminated multilayer composite beamrdquo Composite Structures 68 pp45-51

[37] Baik J M and Thopson R B (1984) ldquoUltrasonic scattering from imperfect interfaces A quasi-static modelrdquo Journal of Nondestructive Evaluation 4(3) pp177-196

[38] Luo H and Hanagud S (2000) ldquoDynamics of delamincated beamsrdquo International Jour-nal of Solids and Structures 37 pp1501-1519

[39] Kim S B and Sohn H (2007) ldquoInstantaneous reference-free crack detection based on polarization characteristics of piezoelectric materialsrdquo Smart Materials and Structures 16 pp2375-2387

[40] An Y K and Sohn H (2010) ldquoInstantaneous crack detection under varying temperature and static loading conditionsrdquo The Special Issue of Structural Control and Health Monitor-ing 17(7) pp 730-741

- 35 -

[41] Lee S J and Sohn H (2006) ldquoActive self-sensing scheme development for structural health monitoringrdquo Smart Materials and Structures 15(6) pp 1734-1746

- 36 -

요 약 문

복합재료 손상의 무기저 진단 및 정량화

복합재료는 재료의 경량성 내구성 고강도의 특성으로 인하여 항공기에서부터

교량에까지 널리 사용되고 있다 하지만 외부의 충격으로 인한 복합재료 내부의

손상은 재료의 강도를 감소시켜 구조물의 안정성을 저하시키는 요인이 된다 하지만

복합재료의 손상은 재료 내부에서 발생하여 육안으로는 식별하기도 힘들며 표면에서

보이는 것만으로 내부의 손상 정도를 판단하기는 불가능하다는 단점을 가지고 있다

본 연구에서는 손상 내부에서 반사로 인해 발생하는 파를 이용하여 기저자료를

사용하지 않고 특정 시점에서 얻은 자료만으로 복합재료의 손상을 진단하고 정량화

할 수 있는 기법을 제안한다 제안된 기법의 적용성을 검증하기 위하여 유한요소법을

이용한 수치해석을 수행하였고 Teflon tape 이 삽입된 시편과 실제 충격하중으로

인한 손상을 입은 시편을 이용한 실험을 진행하였다 수치해석 및 실험 결과 제안된

기법이 복합재료의 손상을 성공적으로 진단 및 정량화할 수 있음을 확인하였다 또한

다양한 외부 온도 환경에서 제안된 기법의 환경영향 실험을 수행하였고 다양한

환경에서 손상의 진단 및 정량화가 성공적으로 이루어짐을 검증하였다

핵심어 무기저 손상진단 및 정량화 압전센서 복합재료 유한요소법 Matching pursuit

method

- 37 -

감 사 의 글

기대와 두려움을 안고 시작한 석사과정이 어느덧 2 년 이라는 세월 속에서 추억으로 자리잡아

가고 있습니다 학부생 기간을 포함한 지난 3 년 동안 SHM 이라는 새로운 분야를 접하고 연구를

수행 할 수 있도록 도와주신 고마운 분들이 있어 그 감사를 표하고자 합니다

우선 저의 많은 부족한 점을 일깨워 주시고 연구에 대한 뜻을 더욱 확고히 세울 수 있게 해

주신 지도교수님이신 손 훈 교수님께 깊은 감사의 뜻을 전합니다 짧은 기간이었지만 교수님의

도움으로 많은 것을 배우고 느끼고 깨달을 수 있었습니다 또한 본 연구를 수행할 수 있도록 지

원해주신 국토해양부 (U-City 석 박사과정 지원사업)에 깊은 감사 드립니다

때로 힘들고 지칠 때 항상 저를 믿고 응원해 주시는 가족이 있었기에 지금의 글을 쓸 수 있었

습니다 또한 지금 연구실에서 함께 생활하고 있는 윤규형 민구형 철민이형 현석이형 진열이형

호민이형 병진이 선혜누나와 성대에 계신 창길이형 미국에 계신 의재형과 Debaditiya 에게도 감

사의 말씀을 전합니다 그리고 모두 나열할 수는 없지만 카이스트 생활의 활력소가 되어준 카이

스트 건설 및 환경공학과 선후배들 직원 분들과 SOS 멤버들에게도 감사를 표합니다

앞서 거론한 많은 이들의 도움이 없었다면 저 또한 지금 이 자리에 있을 수 없었을 것입니다

많이 부족하기에 더 노력하고 배우고자 했던 초심을 다시 한번 생각하며 더 큰 결실을 맺기 위해

앞으로 나아가겠습니다 도움을 주신 모든 분들께 이 감사의 글을 바칩니다

2011년 6월 카이스트에서 임형진 올림

- 38 -

CURRICULUM VITAE

Hyung Jin Lim

Research Assistant

The Civil and Environmental Engineering Department

Korea Advanced Institute of Science and Technology (KAIST)

291 Daehak-Ro Guseong-Dong Yuseong-Gu

Daejeon Republic of Korea 305-701

Phone (82)+42-350-3665 Fax (82)+42-350-3610

Email limhj87gmailcom

RESEARCH INTERESTS

Damage Detection Localization and Quantification on Composite and Metallic Structures Pattern Recognition Finite Element Modeling (FEM)

EDUCATION

2009 ndash 2011 MS Civil and Environmental Eng Korea Advanced Institute of Science and Technology (KAIST) Korea Rep

2006 ndash 2009 BS Civil and Environmental Eng Korea Advanced Institute of Science and Technology (KAIST) Korea Rep

JOURNAL PUBLICATIONS The corresponding author is underlined 1 Hyung Jin Lim Hoon Sohn ldquoReference-free damage detection localization and quantification

using in compositerdquo in preparation for Journal of the Acoustical Society of America 2011

2 Chul Min Yeum Hoon Sohn Jeong Bum Ihn Hyung Jin Lim ldquoInstantaneous delamination detection in a composite using a dual piezoelectric transducer networkrdquo Submitted to Compo-sites Science and Technology 2011

3 Yun-Kyu An Hyung Jin Lim Min Koo Kim Jin Yeol Yang Hoon Sohn Chang Guen Lee ldquoInvestigation of applicability of local reference-free damage detection techniques to in-situ

- 39 -

bridgesrdquo Submitted to ASCE Journal of Structural Engineering 2011

4 Hyung Jin Lim Min Koo Kim Hoon Sohn Chan Yik Park ldquoImpedance based damage detec-tion under varying temperature and loading conditionsrdquo Submitted to NDTampE International 2011

5 Hyung Jin Lim Hoon Sohn ldquoQuantitative estimation of transmitted and reflected lamb waves at discontinuityrdquo Journal of the Korean Society for Nondestructive Testing Vol30 No 4 pp1-8 2010

CONFERENCE PROCEEDINGS

The corresponding author is underlined 1 Hyung Jin Lim and Hoon Sohn ldquoReference-free damage detection localization and quantifi-

cation in compositesrdquo ASCE Engineering Mechanics Institute Conference Boston June 2-4 2011

2 Hyung Jin Lim and Hoon Sohn ldquoReference-free delamination quantification within multi-layer composite materialsrdquo KSNT Spring Conference May 26-27 Gyeong-Ju Korea 2011

3 Min Koo Kim Hyun Jin Lim Hoon Sohn Chan Yik Park ldquoImpedance-based bolt loosening detection under varying temperature and loadingrdquo Asian Pacific Workshop on Structural Health Monitoring Tokyo Japan November 30-December 2 2010

4 Hyung Jin Lim Chul Min Yeum and Hoon Sohn ldquoModeling of Impact-induced Delamination in a Multilayer Composite Platerdquo The 23rd KKCNN Symposium on Civil Engineering Taipei (Taiwan) November 13-15 2010

5 Yun-Kyu An Hyung Jin Lim and Hoon Sohn ldquoTemperature independent Delamination Detec-tion using data normalizationrdquo The 3rd International Conference on Multi-Functional Materi-als and Structures Chonbuk National Univ Jeonju Korea September 14-18 2010

6 Hoon Sohn Yun-Kyu An Chang Gil Lee Min Koo Kim Hyung Jin Lim ldquoApplication of online NDT techniques to decommissioned bridge testingrdquo the 5th European Workshop on Structural Health Monitoring Sorrento Italy June 29-July 02 2010

7 Min Koo Kim Hyung Jin Lim Hoon Sohn Chan Yik Park ldquoImpedance based damage diag-nosis of a complex composite aircraft wing under changing loading and temperature condi-tionsrdquo the 5th European Workshop on Structural Health Monitoring Sorrento Italy June 29-July 02 2010

8 Yun-Kyu An Min Koo Kim Hyung Jin Lim and Hoon Sohn ldquoValidations of reference-free crack detection techniques through a decommissioned bridge testrdquo 2010 COSEIK Annual Con-ference Jeju Korea April 8-9 2010

- 40 -

PARTICIPANT PROJECTS

1 Development of Smart Steel Members for Online Steel Structure Monitoring POSCO Corporation (Funded 41000000 KRW (41000 USD) for 010511 to 123111)

2 Development of Enabling Technologies for Structural Health Monitoring Korean Agency for Defense Development (Funded 240000000 KRW (240000 USD) out of total 3200000000 KRW (3200000 USD) for 060108-053111)

3 Development of On-board SHM Technologies for Composite Air Vehicles The Boeing Company (Funded 400000 USD (400000000 KRW) for 080108 to 073111)

EDUCATIONAL EXPERIENCES

Fall 2009

Teaching Assistant CE207 ldquoElementary Structural Engineering and Laboratoryrdquo KAIST 3 unit undergraduate course

EXPERIMENTAL EXPERIENTCES

2010 Monitoring of Real Aircraft Composite Wing Segment Korea Agency for Defense De-velopment (ADD) Daejeon Korea Real time monitoring of composite air vehicles fund-ed by Korea Agency for Defense Development (ADD)

2009-2011

Field Monitoring of Real Bridge Structures Gimpo (Ramp G bridge) amp Icheon (Sam-seung bridge) Korea US-Korea collaborative research for bridge monitoring testbeds

HONORS amp AWARDS

Scholarships for outstanding students granted KAIST (2007-2009) The three prominent students in the civil and environmental engineering department of KAIST were selected as recipients This scholarship provided 2000$ per year for three years

- 41 -

Page 9: 복합재료 손상의 무기저 진단 및 정량화 - KAISTssslab.kaist.ac.kr/article/pdf/Thesis_2.pdf · 1 Declaration of Ethical Conduct in Research: I, as a graduate student

LIST OF FIGURES

1 Lamb wave propagation characteristics (a) A0 Lamb wave mode time delay and

attenuation after passing through a damage (b) The time delay and amplitude varia-

tion of A0 Lamb wave mode under temperature variation in pristine condition of

structure Because of this conventional damage detection techniques in composites

have limitations for real structure application middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 5

2 Multiple reflections in the damage area The multiple reflections of A0 Lamb

waves in the damage area are utilize as a feature for characterization of the damage

in this study middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 6

3 Parameters and notations of the proposed damage characterization technique

Damage is detected by identifying the additional reflected wave from the damage due

to multiple reflections (dashed lines) Localization and quantification of the damage

is carried out by measuring the arrival time of transmitted and reflected waves from

the damage middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 7

4 Two different FEM models which have different 119889119889 and 119907119907119889119889 (or G) This repre-

sents that the damage cannot be characterized uniquely based on the measured data

available middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 9

5 (a) Example of the overlapped signal with comparing approximated signal using

the matching pursuit method (b) Decomposed signal using the matching pursuit

method The matching pursuit method provides a conventional way to decompose

and identify each wave middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 9

6 Description of the numerical simulation model middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 17

7 Comparison of A0 Lamb wave mode signals from numerical simulation be-

tween intact and damage cases Time delay attenuation and multiple reflections due

to damage simulated successfully using the damage modeling method described in

v

the previous section middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 19

8 Test specimens for the experimental verification In the Teflon inserted damage

Teflon tapes with 25mm diameter were inserted at the center of the specimen The

impact-induced damage was introduced to the specimen using the impact test equip-

ment middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 21

9 Overall experimental setup (a) Data acquisition (DAQ) system and (b) Installed

circular PZT transducer on the specimen with its dimension middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 23

10 (a) Impact test equipment for introducing impact-induced damage and (b) Tip

of the impact tester with 10mm diametermiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 23

11 Test specimen in temperature chamber middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 24

12 A0 Lamb wave mode obtained using pitch-catch method from the Teflon in-

serted damage specimen (a) The reflected wave from damage is shown (b) The Tef-

lon inserted damage is detected by capturing the reflected wave middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 25

13 The signal obtained from the test specimen subjected to impact The impact

was introduced at the same location of the specimen and the signal was obtained after

each impacts As the number of impact is increased the transmitted A0 Lamb wave

mode is delayed and attenuated middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 26

14 False alarm test for the proposed damage detection algorithm in pristine condi-

tion of the specimen Reflected wave from damage is not captured false alarm is not

occurred middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 26

15 Signals obtained from the Teflon inserted damage specimen under temperature

variation Time delay and amplitude variation are occurred middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 27

16 False alarm test under temperature variations It is shown that proposed dam-

vi

age characterization algorithm is robust under temperature variationsmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 29

17 Signals obtained from impact-induced damage specimen (Impact 3) under

temperature variation middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 29

vii

CHAPTER 1 INTRODUCTION

11 Motivation

Composite materials has been increasingly demands in modern structures ranging from the

latest aircraft to retrofitted bridge deck due to its light weight superior specific strength and

stiffness properties However the occurrence of damage due to abrupt impact or sequential

accumulation of defect may considerably degrade structural strength or stiffness and reduce

the safety and reliability of the system Recently Lamb waves based damage detection tech-

niques using piezoelectric (PZT) transducers have emerged as one of the promising structural

health monitoring (SHM) technology for detecting hidden damage because of the possibility

of long range inspection and high sensitivity [1-4]

12 Literature review

This section will discuss researches that have been previously carried out on topics of

guided wave based SHM techniques using PZT transducers for damage in composite materi-

als The early investigations of interaction between Lamb wave and damage in composite can

be found in Rose et al and Guo et al [5-6] From further studies reflected signals from dam-

age were shown to be an effective method for damage diagnosis and localization using PZTs

[7-9] The reflected signals from damage were collected using one PZT as an actuator and the

other PZT as a sensor However the techniques cannot be used to determine the extent of

- 1 -

damage with reasonable precision

Kesseler et al proposed a technique monitoring the wavelet coefficients of transmitted ul-

trasonic signal amplitude as a potential way to damage detection [10] One drawback of this

technique is that the variability in coupling and the sensitivity of installed PZT transducers

can induce changes in the amplitude and energy of the signal that are cause false-alarm Oth-

er wavelet based delamination detection and localization technique developed using PZT

sensor network and outlier analysis [11]

Damage detection and quantification in composite plate was also investigated by group de-

lay measurement due to damage [12] However it is assumed that the group velocity in dam-

age area as the group velocity propagates half thickness of the structure in dispersion curve in

the quantification procedure Thus the damage quantification was investigated only Teflon

inserted damage case

Sohn et al proposed a damage detection technique which does not rely on any past base-

line signals to assess damage in composite plates by using an enhanced time reversal method

[13] The damage is assessed using the violation of time reversibility when nonlinearity is

caused by the damage along a direct wave path Thus it allows instantaneous identification

of delamination without requiring the baseline signal for comparison

- 2 -

13 Objective and uniqueness

In this study a reference-free damage characterization technique including detection lo-

calization and quantification of impact-induced damage in composite materials is developed

The general damage characterization can be categorized into the following levels [14]

Level 0 Define damage

Level 1 Determine the presence of damage within the structure

Level 2 Locate the regions of damage

Level 3 Quantify the damage

Level 4 Predict the remaining service life of the structure

Through the proposed technique from level 1 to level 3 can be achieved

Most of common approaches of damage characterization techniques described in the pre-

vious section use prior obtained baseline data However these approaches can cause false

alarm due to the effect of environmental variations such as temperature and ambient vibra-

tions [15]

The uniqueness of this study is

1 Damage is characterized without using prior measured baseline data

2 Robustness of the proposed technique under temperature variation is investigated

3 Additional waves reflected from inside damage have been used in addition to transmit-

ted and reflected waves

- 3 -

14 Organization

The organization of this thesis is summarized as follows

Chaper1 A general introduction including motivation literature review objective and

uniqueness of this study is described

Chaper2 Interactions between Lamb wave mode and damage in composites such as time

delay attenuation and multiple reflections are summarized Damage index (m-value) for

damage characterization is defined

Chaper3 Algorithms for the reference-free damage characterization technique using m-

value are developed Modified matching pursuit for this study is also described

Chaper4 Verification of the algorithm using numerical simulation is presented Various

modeled damage cases are characterized

Chaper5 Experimental verifications are described Two damage cases (Teflon inserted and

impact-induced damage) are investigated for the feasibility of the proposed damage charac-

terization technique Temperature effect test using temperature chamber is also carried out

Chaper6 Conclusion and summary of the present study is reported In advance suggestions

and future work recommendations are represented

- 4 -

CHAPTER 2 INTERACTION OF LAMB WAVE WITH DAMAGE

21 A0 Lamb wave mode delay and attenuation in the time domain

Time delay and attenuation of the A0 mode Lamb wave after passing through the damage

area due to effective thickness reduction were reported in previous researches [812] Moreo-

ver the attenuation of the A0 Lamb wave mode is occurred by degrading of shear modulus of

in the damage area However the time delay and amplitude variation of the A0 Lamb wave

mode also occurred even when the structure is health condition by the material property

changing due to temperature variation of the structure (Fig 1) Thus conventional damage

characterization methods using the time delay and attenuation of A0 Lamb wave mode as a

feature can cause false-alarm in real application

11 115 12 125 13-10

-5

0

5

10

Time (ms)

Vol

tage

(mV

)

IntactDamage

Reflected wave from Damage

(a)

11 115 12 125 13-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

0 oC20 oC50 oC

(b)

Fig 1 Lamb wave propagation characteristics (a) A0 Lamb wave mode time delay and atten-uation after passing through a damage (b) The time delay and amplitude variation of A0 Lamb wave mode under temperature variation in pristine condition of structure Because of this conventional damage detection techniques in composites have limitations for real struc-ture application

- 5 -

22 Multiple reflections in the damage area

Standing wave patterns due to multiple reflections of A0 Lamb wave mode in the damage

area were observed through numerical simulations and ultrasonic wave-field imaging using a

scanning laser vibrometer [16] The formation of standing waves in damage is illustrated

through Fig 2 When the incident waves enter the damage area a significant portion of the

waves are reflected back from the exit Then the reflected waves travel through the damage

and undergo reflection again at the original entrance As a consequence of multiple reflec-

tions at the entrance and the exit a considerable amount of ultrasonic energy is trapped inside

the damage area [17-19]

Multiple Reflections

Incident Wave Transmitted Wave

Damage

Fig 2 Multiple reflections in the damage area The multiple reflections of A0 Lamb waves in the damage area are utilize as a feature for characterization of the damage in this study

- 6 -

CHAPTER 3 THEORETICAL DEVELOPMENT

31 Overview

Damage

l = l1 + d + l2

d = vdtd l1 = vit1 l2 = vit2

t = t1 + td + t2

Excitation and Sensing PZT Sensing PZT

l

Transmitted Wave

Reflected Waves

Incident Wave

Fig 3 Parameters and notations of the proposed damage characterization technique Damage is detected by identifying the additional reflected wave from the damage due to multiple re-flections (dashed lines) Localization and quantification of the damage is carried out by measuring the arrival time of transmitted and reflected waves from the damage

In this section the reference-free damage characterization algorithm is proposed Damage

diagnosis and quantification are achieved by identifying the additional reflected waves due to

multiple reflections in the damage area (dashed lines in Fig 3) Moreover damage is local-

ized using additional information from pulse-echo method implement For the localization

and the quantification of damage the variables such as the path length and A0 Lamb wave

mode velocity in damage area are defined (Fig 3) For example 119897119897 is expressed as

119897119897 = 1198971198971 + 119889119889 + 1198971198972 (1)

where 1198971198971 119889119889 and 1198971198972 are the distance from the excitation PZTs to the entrance of damage the

damage length and the distance from the exit of damage to the sensing PZT respectively

- 7 -

Moreover 119907119907119894119894 and 119907119907119889119889 are also defined as the A0 Lamb wave mode velocity in pristine area

and damage area respectively Note that the whole procedure of the damage characterization

is accomplished based solely on the signals obtained from the current state of the structure

32 Damage index (m-value)

The proposed reference-free damage characterization technique utilizes the waves reflect-

ed additionally from the damage area as a feature The arrival time difference of the transmit-

ted and reflected A0 Lamb wave modes reflected from the exit and the entrance of the delam-

ination area one times each (∆119905119905) contains the information of the physical length of the dam-

age (119889119889) and A0 Lamb wave mode velocity in the damage area (119907119907119889119889) The wave velocity in the

damage area which is related to the shear modulus (G) represents the severity of the damage

∆119905119905 = 2119905119905119889119889 =2119889119889119907119907119889119889

(2)

However based on the measured data available the 119889119889 and 119907119907119889119889 (or G) cannot be uniquely

determined Fig 4 shows two FEM models which have different physical length and wave

velocity (G) of the damage Detail explanation of delamination modeling technique will be

covered in section 4 The similar signal in the time domain and ∆119905119905 are obtained from the

two different damage cases

- 8 -

Damage 1

Damage 2

20

Unit mm15

90 G value reduction

943 G value reduction

m-value = 248

m-value = 248

015 02 025 03 035

-2

-1

0

1

2

Time (ms)

Ampl

itude

Damage 1Damage 2

Fig 4 Two different FEM models which have different 119889119889 and 119907119907119889119889 (or G) This represents that the damage cannot be characterized uniquely based on the measured data available

Thus in this study the damage index (m-value) which can be uniquely defined based on

the available data is introduced The m-value is defined as the ratio of the physical length of

the damage and wavelength of the A0 Lamb wave mode in the damage area (120582120582119889119889)

119898119898 =119889119889120582120582119889119889

=119889119889119889119889119907119907119889119889

(3)

In many cases the transmitted and reflected waves are overlapped and it is hard to decom-

pose and calculate the arrival times of them Matching pursuit method is an alternative way

to accurately decompose of the overlapped waves and estimate the parameters of each wave

such as arrival time (Fig 5) Detail explanation of the matching pursuit method is covered in

section 36

015 02 025 03-2

-1

0

1

2

Time (ms)

Am

plitu

de

OriginalApproximated

(a)

015 02 025 03-2

-1

0

1

2

Time (ms)

Ampl

itude

Wave 1Wave 2

(b)

Fig 5 (a) Example of the overlapped signal with comparing approximated signal using the matching pursuit method (b) Decomposed signal using the matching pursuit method The matching pursuit method provides a conventional way to decompose and identify each wave

- 9 -

33 Damage detection algorithm (Level 1)

The damage diagnosis is accomplished by extracting the additional reflected A0 Lamb

wave modes from damage due to multiple reflections If A0 Lamb wave modes are decom-

posed perfectly it is trivial that one transmitted A0 Lamb wave mode is always captured in

the response signal If damage exists other waves in the response signal behind the transmit-

ted A0 Lamb wave mode are presented Since this approach does not rely on previously col-

lected data the diagnosis is independent from environmental variation such as temperature

changes

It is reported that the dispersion of Lamb waves cause wave-packets to spread out in space

and time as they propagate through a structure [20] Thus the reflected waves from the dam-

age have longer pulse duration than the transmitted wave due to their longer propagating dis-

tance In this study the matching pursuit method is modified to capture more spread wave

than the transmitted A0 Lamb wave mode Detail explanation of the modified matching pur-

suit method is described in latter section

Let 1199041199041 and 1199041199042 as the pulse duration (scale) of the transmitted and additional reflected A0

Lamb wave mode respectively

ldquoIf 1199041199042 gt 1199041199041 damage exists

Otherwise there is no damagerdquo

- 10 -

34 Damage quantification algorithm (Level 3)

As mentioned in the previous section damage is quantized by estimating 119889119889 and 119907119907119889119889 using

pitch-catch method The procedure to estimate 119889119889 and 119907119907119889119889 is

(i) Measure 119897119897 119905119905 and 119905119905119889119889

(ii) Assume the range of 119889119889 (0 lt 119889119889 lt 119897119897)

(iii) Calculate 119907119907119894119894 and 119907119907119889119889 with a constraints (0 lt 119907119907119889119889 lt 119907119907119894119894) (Eq 4 and 5)

(iv) Estimate m-value and find a feasible range of 119889119889 and 119907119907119889119889

119907119907119894119894 =119897119897 minus 119889119889119905119905 minus 119905119905119889119889

(4)

119907119907119889119889 =119889119889119905119905119889119889

(5)

35 Damage localization algorithm (Level 2amp3)

For damage localization 1198971198971 is estimated by measuring the arrival time from the excitation

PZT to the entrance of delamination (1199051199051) with additional pulse-echo method implementation

The procedure is similar with the previous damage quantification algorithm

(i) Measure 119897119897 119905119905 119905119905119889119889 and 1199051199051

(ii) Assume the range of 1198971198971 (0 lt 1198971198971 lt 119897119897)

(ii) Compute possible combinations of 1198971198972 119889119889 and 119907119907119889119889 with constraints

(0 lt 119907119907119889119889 lt 119907119907119894119894 119889119889 gt 0) (Eq 5 6 and 7)

- 11 -

1198971198972 = 1198971198971119905119905 minus 1199051199051 minus 119905119905119889119889

1199051199051 (6)

119889119889 = 119897119897 minus 1198971198971 minus 1198971198972 (7)

Since a PZT for pulse-echo method acts as both an actuator and a sensor the obtained sig-

nal contains both the excitation and the response signal Thus the reflected wave from the

entrance of damage can be buried into the tail of excitation signal and cause error on arrival

time of reflected wave from damage Even in this case damage can be localized and quan-

tized using the constraints When first wave packet reflected from the entrance of damage is

buried into the excitation signal the next wave packet reflected from the exit of damage can

be utilized alternatively If the algorithm regards the second reflected wave as the first re-

flected wave there are no combinations of estimated variable meeting the constraints Then

the algorithm repeats the process by regarding the wave as the second reflected wave the

constraints can select proper combination of the estimated variables

The estimated values have range bounded by the constraints When additional information

of 119907119907119894119894 is utilized the damage can be characterized more accurately In this study three meth-

ods are proposed and verified as information level of 119907119907119894119894

Method 1 Solely based on the current data measurement (Reference-free)

Method 2 Limit the possible range of 119907119907119894119894 (Prior knowledge of the system)

Method 3 Measure 119907119907119894119894 (Additional data need to be collected)

Method 1 described in this section can be utilized without any information of 119907119907119894119894 Meth-

od 2 produces more accurate estimation results using prior knowledge of the system such as

- 12 -

119907119907119894119894 variation under temperature changes Method 3 gives unique estimated value using meas-

ured 119907119907119894119894 from other path in the same time In section 4 and 5 damage is characterized using

these three methods and the results are compared each another

36 Matching pursuit method considering the dispersion effect of Lamb

wave

Matching pursuit method has been developed for signal compression approximation and

characterization of obtained signal and function [21-23] The key concept of the matching

pursuit method is that any signal (function) can be decomposed into a linear expansion of

signals selected from a pool of basis signals (functions) called lsquodictionaryrsquo

Let 119891119891 be a function to be analyzed then 119891119891 can be approximated a set of functions select-

ed from the dictionary 119863119863 in order to best match This is done by successive approximations

with orthogonal projections Let1198921198921205741205740 isin 119863119863 the function 119891119891 can be decomposed into

119891119891 = lang1198911198911198921198921205741205740rang1198921198921205741205740 + 119877119877119891119891 (8)

where 119877119877119891119891 is the residual function after approximating 119891119891 using 1198921198921205741205740 1198921198921205741205740 is orthogonal to

119877119877119891119891 hence

1198911198912 = lang1198911198911198921198921205741205740rang2

+ 1198771198771198911198912 (9)

To minimize 119877119877119891119891 1198921198921205741205740 is chosen to maximize lang1198911198911198921198921205741205740rang The matching pursuit method is

an iterative algorithm that decomposes the residue 119877119877119891119891 by projecting it on a function of 119863119863

that matches 119877119877119891119891 almost at best as it was done for 119891119891 The residue 119877119877119899119899119891119891 is decomposed into

119877119877119899119899119891119891 = lang119877119877119899119899119891119891119892119892120574120574119899119899rang119892119892120574120574119899119899 + 119877119877119899119899+1119891119891 (10)

- 13 -

the decomposition is carried up to the order 119898119898 119891119891 can be expressed as

119891119891 = lang119877119877119899119899119891119891119892119892120574120574119899119899rang119892119892120574120574119899119899

119898119898minus1

119899119899=0

+ 119877119877119898119898119891119891 (11)

The original function 119891119891 is decomposed into a sum of dictionary elements that are chosen to

best match its residues

In this study the matching pursuit method is utilized to accurately decompose of the over-

lapped Gaussian pulses and estimate the parameters of the pulse which used for damage

characterization The 119894119894119905119905ℎ Gaussian pulse in the dictionary for approximating the obtained

signal is expressed as

119892119892120574120574119894119894 = 119891119891(119860119860119894119894 119906119906119894119894 119904119904119894119894 119889119889119894119894120601120601119894119894) (12)

where 119860119860119894119894119906119906119894119894119904119904119894119894119889119889119894119894 and 120601120601119894119894 represent amplitude arrival time scale (pulse duration) excita-

tion(central) frequency and phase of the Gaussian pulse respectively The parameters of the

pulse are estimated using a nonlinear regression based optimization method (Gauss-Newton

method) based on an initial guess of the parameters Detail procedure for estimating the pa-

rameters of the pulse is expressed in Hong et al [23]

Among the estimated parameters of the pulses pulse duration (scale) represents the disper-

sion effect If a wave-packet is spread out the estimated pulse duration is increased Using

this feature the matching pursuit method is modified to capture waves which have longer

pulse duration than transmitted A0 Lamb wave mode If the estimate pulse duration of a wave

is less than the transmitted wave the modified matching pursuit method regards it as error

- 14 -

CHAPTER 4 NUMERICAL SIMULATION

41 Composite modeling method

Composite materials can be modeled from nano to macro level depends on its engineering

purpose In general four methods are utilized for modeling of composite materials First

composite is modeled as a homogenous material expressed in terms of effective material

properties [24] The effective material properties are estimated using the modulus and the

volume fraction of each constituent theoretically (Rule of mixture) and measured experimen-

tally Second a model is established that each laminate which have own material properties

are totally bonded together In this model the stacking sequence and direction of each lami-

nates are considered additionally [25-27] Third method for composite modeling in micro

level called micromechanics [28-30] This model considers the interaction between fibers and

matrix in laminate Last multiscale modeling which the scope of modeling is extended to

molecular or quantum level of constituents [31-33] This model starts from molecular me-

chanics or molecular dynamics then the scale of model leads to the nano micro and macro

level by incorporating small scale models As the scale of models is smaller it provides more

logical way than other large scale of models while the computational cost is more expensive

The scope of this study is focused on the dynamic behavior of composite structure the beam

or plate model was utilized for the numerical simulation

- 15 -

42 Damage modeling method

Generally three damage models have been developed and utilized in the real applications

First the stiffness reduction model called as the equivalent element model was introduced

[17 34-35] This model is simply realized by replacing the region around damage area by a

region with adjusted material properties (Youngrsquos modulus shear modulus density etc)

globally Next the duplicated node model is done by disconnecting the element connectivity

assuming there is no interaction among the delaminated surface This model can be utilized

when the location and the extension of damage is important issue [34-36] Last the contact

mechanics model using the principles of contact mechanics the two interfacial forces (con-

tact resistance and sliding resistance) in the delaminated surface [37-38] In this model the

delaminated surfaces are modeled as nonlinear functions of the interlaminar normal strain

and the interlaminar shear strain This model is computationally expensive while provides

more logical way than other models There is however no universally accepted model for

general impact-induced damage in composites In the case of impact-induced damage it is

impossible to predict or estimate the location and number of delaminated layers Thus from

the uncertainty of impact-induced damage the equivalent element model can be considered

as the effective damage model

- 16 -

43 Numerical simulation description

TABLE I Material property of simulation model and test specimen

Tensile Strength 657 MPa Tensile Modulus 59 GPa Shear Strength 804 MPa Shear Modulus 24 Gpa Specific Gravity 16

PZT A

PZT C

PZT BDamage

3

250 125 125

500 Unit mm

Fig 6 Description of the numerical simulation model

The feasibility of the proposed damage characterization technique is verified by 2-D FEM

simulation using MSC PATRAN and NASTRAN Fig 6 presents the dimension PZTs and

damage location of the simulation model In the model 05mm x 05mm square shell element

was used for modeling of the A0 mode whose flexural wave characteristic The element size

was determined by considering the wavelength of the propagation Lamb wave modes The

material properties of the simulation model and damage cases are presented in Table I and

Table II respectively

PZT A and C act as both actuator and sensor for collecting pulse-echo signal They were

installed on the plate which two of them are both-side collocated for the A0 Lamb wave

- 17 -

mode decomposition [39-40] PZT B was only used as a sensor for collecting pitch-catch sig-

nal For simulating the excitation and sensing of PZTs the pin-force model was adopted that

an input force apples to the end nodes of excitation PZT and averaging element strains corre-

sponding sensing area A Gaussian pulse (tone-burst) 50 kHz central frequency was excited

as an input signal

TABLE II Damage cases for numerical simulation

119949119949120783120783 (mm) 119941119941 (mm) G value reduction ()

Case 1 650 200 90 darr

Case 2 675 150 90 darr

Case 3 675 150 93 darr

Case 4 900 150 90 darr

- 18 -

44 Numerical simulation results

015 02 025 03 035

-2

0

2

Time (ms)

Am

plitu

de

IntactDamage

Reflected wave from damage

(a) Pitch-catch signal

015 02 025 03 035

-2

0

2

Time (ms)

Am

plitu

de

IntactDamage

Reflected waves from damage

(b) Pulse-echo signal Fig 7 Comparison of A0 Lamb wave mode signals from numerical simulation between intact and damage cases Time delay attenuation and multiple reflections due to damage simulated successfully using the damage modeling method described in the previous section

Fig 7 shows that comparison of A0 Lamb wave mode signals between intact and one of

damage cases in the simulation The features of A0 Lamb wave mode passing through dam-

age area described in section 2 are successfully simulated Thus the equivalent element

method is proper for modeling damage in composites In the damage cases the additional

waves reflected from damage is observed in both pitch-catch and pulse echo signal

Table III presents the result of the proposed damage characterization algorithm using the

numerical simulation cases shown in Table II In lsquoMethod 2rsquo 119907119907119894119894 variation is assumed to plusmn

5 of the normal value (1190~1315ms) The normal value of 119907119907119894119894 for lsquoMethod 3rsquo is meas-

ured as 1253 ms In each case the damage is characterized using the proposed technique

Note that as the information of 119907119907119894119894 is specified the estimated value is closer to the exact

value

- 19 -

TABLE III Damage characterization for the numerical simulation models

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

Case 1

Exact 650 200

Method 1 248 50~70 79~414 160~834

Method 2 248 603~667 135~241 271~486

Method 3 248 635 188 379

Case 2

Exact 675 150

Method 1 193 60~75 01~251 37~651

Method 2 193 618~682 114~222 295~575

Method 3 193 65 168 435

Case 3

Exact 675 150

Method 1 233 60~75 66~302 140~649

Method 2 233 668~738 85~196 182~419

Method 3 233 703 140 301

Case 4

Exact 900 150

Method 1 204 75~100 48~349 118~856

Method 2 204 860~951 107~216 264~530

Method 3 204 906 162 397

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1253 ms 1190~1315 ms Measured 119907119907119894119894 1253 ms

- 20 -

CHAPTER 5 EXPERIMENTAL VERIFICATION

51 Overview

In this section experimental study is investigated so that the proposed damage characteri-

zation technique Overall experimental setup and test specimen are shown in Fig 8 and 9

The experimental setup consists of test specimens data acquisition (DAQ) system and self-

sensing circuit Self-sensing circuit was installed for pulse-echo measurement using PZTs

[40-41]

PZT A(PZT C) PZT B

500

500

140

75

250

Teflon Inserted Damage (d =25)

Unit mm

Thickness = 3

PZT A(PZT C) PZT B

500

500

150

90

250

Impact Location

Unit mm

Thickness = 3

(a) Teflon inserted damage specimen

(b) Impact-induced damage specimen

Fig 8 Test specimens for the experimental verification In the Teflon inserted damage Teflon tapes with 25mm diameter were inserted at the center of the specimen The impact-induced damage was introduced to the specimen using the impact test equipment

- 21 -

511 Specimen

Two carbon composite plates with a dimension of 500times500times3mm3 were manufactured by

Nexcomes Inc by stacking 12 layers of CF3327 3K prepreg (woven fabrics type) manufac-

tured by Hankuk Fiber Inc The material properties of the specimen are presented in Table I

One of the specimens Teflon tapes with 25mm diameter was inserted in every 3 layers at the

center of the specimen to mimic the damage The other specimen is utilized for the character-

ization of real impact-induced damage

512 Data acquisition (DAQ) system and PZT

The data acquisition (DAQ) system consists of an arbitrary waveform generator (AWG) a

high speed signal digitizer (DIG) and two multiplexers (MUX) Using the 16-bits AWG

tone-burst input signal with a plusmn10 peak-to-peak voltage were generated The response signal

was measured by the 14-bit DIG For the signal-to-noise ratio reduction both pitch-catch and

pulse-echo signals were measured 150 times and averaged in the time domain 3 identical

circular PZTs of 10mm diameter manufactured by APC International Ltd were installed on

the specimen A Gaussian pulse (tone-burst) 50 kHz central frequency was excited as an in-

put signal The excitation frequency was chosen in order to avoid overlapping of the S0 and

A0 Lamb wave modes and generating of the higher modes (S1 A1hellip)

- 22 -

PZT Actuator

PZT Sensor

AWG

DIG

DAQ System Specimen

SensingActuatingMUX

1

23

4

1 Sending signal to MUX

4 Sending signal to DIG

Self-SensingCircuit

18 Unit mm

10

(a)

(b)

Fig 9 Overall experimental setup (a) Data acquisition (DAQ) system and (b) Installed circu-lar PZT transducer on the specimen with its dimension

513 Impact test

The impact-induced damage was introduced to the specimen using the impact test equip-

ment presented in Fig 10 A 5kg mass with 10mm tip was dropped from 015m height onto

the same location of the specimen The signals were obtained at each number of impacts

Since the reflected wave is not detected at the first impact case the data obtained from the

second and third impact were utilized for the analysis

10

Unit mm (a)

(b)

Fig 10 (a) Impact test equipment for introducing impact-induced damage and (b) Tip of the impact tester with 10mm diameter

- 23 -

514 Temperature effect test

In reality structures are subjected to changing environmental conditions such as tempera-

ture that affect measured signal and cause false-alarm Thus the robustness of the proposed

damage characterization technique under temperature variation in investigated (Fig 11) The

specimens were placed inside a temperature chamber and a thermocouple was installed on

the specimen to measure temperature of surface of the specimen The signal was obtained

under three temperature cases (0 20 and 50degC) and the humidity was kept at 30

Temperature Chamber

Test Specimen

Fig 11 Test specimen in temperature chamber

- 24 -

52 Teflon inserted damage

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Reflected wave from damage

(a)

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Transmitted WaveReflected Wave

Detected reflected wave

(b)

Fig 12 A0 Lamb wave mode obtained using pitch-catch method from the Teflon inserted damage specimen (a) The reflected wave from damage is shown (b) The Teflon inserted damage is detected by capturing the reflected wave

Fig 12 shows the obtained pitch-catch signal from the Teflon inserted damage specimen

and damage diagnosis result The Teflon inserted damage is detected successfully by captur-

ing the reflected wave Table IV presents the result of the Teflon inserted damage characteri-

zation 119907119907119894119894 is measured as 1213ms and the variation is assumed to plusmn 5 of the normal val-

ue (1152~1273ms) The results indicate that the proposed damage characterization approach

works well in real application

TABLE IV Damage characterization for the Teflon inserted damage specimen

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

Exact 635 25

Method 1 123 60~75 73~338 298~1220

Method 2 123 619~662 229~310 934~1188

Method 3 123 631 284 1161

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1213 ms

- 25 -

53 Impact-induced damage

11 115 12 125 13-4

-2

0

2

4

Time (ms)

Vol

tage

(mV

)

IntactImpact2Impact3

Reflected wave from damage

(a) Pitch-catch signal

11 115 12 125 13 135

-05

0

05

1

Time (ms)

Volta

ge (m

V)

IntactImpact2Impact3

Boundary reflection

Reflected wave from damage

(b) Pulse-echo signal

Fig 13 The signal obtained from the test specimen subjected to impact The impact was in-troduced at the same location of the specimen and the signal was obtained after each impacts As the number of impact is increased the transmitted A0 Lamb wave mode is delayed and attenuated

The obtained signals from impact-induced damage specimen are presented in Fig 13 It is

observed that as the number of impact is increased the transmitted A0 Lamb wave mode is

more delayed because the wave velocity in the damage area is reduced Fig 14 shows the

false-alarm test result whether the proposed damage detection algorithm works even in pris-

tine condition of the test specimen It is shown that no false alarm is occurred because no ad-

ditional reflected wave is captured

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Transmitted Wave

(a) Transmitted wave

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Reflected Wave

No reflected wave

(b) Reflected wave

Fig 14 False alarm test for the proposed damage detection algorithm in pristine condition of the specimen Reflected wave from damage is not captured false alarm is not occurred

- 26 -

TABLE V Damage characterization for the impact-induced damage specimen

of Impacts m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

2

(1470 J)

Method 1 080 65~90 75~246 470~1157

Method 2 080 801~885 114~245 711~1134

Method 3 080 843 179 1122

3

(2205 J)

Method 1 148 50~100 06~380 21~1134

Method 2 148 758~838 249~368 843~1125

Method 3 148 798 308 1044

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1213 ms

Table V presents the characterization results of the impact-induced damage case Here the

measured 119907119907119894119894 and the assumed variation are same as the artificial damage case It is observed

that as the number of impact is increased 119889119889 is increased while 119907119907119889119889 is reduced Thus it is

validated that the proposed damage characterization approach estimates the increment of the

damage well

54 Temperature effect test

541 Teflon inserted damage

11 115 12 125 13-10

-5

0

5

10

Time (ms)

Vol

tage

(mV

)

0 oC20 oC50 oC

(a) Pitch-catch signal

11 115 12 125 13 135

-1

0

1

2

3

Time(ms)

Volta

ge (m

V)

0 oC20 oC50 oC

(b) Pulse-echo signal

Fig 15 Signals obtained from the Teflon inserted damage specimen under temperature varia-tion Time delay and amplitude variation are occurred

Fig 15 shows the obtained signals from the Teflon inserted damage specimen subjected to

- 27 -

temperature variation Time delay and amplitude variations are occurred there is no addition-

al damage is introduced The analyzed results of the obtained signals are reported in Table VI

The measured 119907119907119894119894 is 1247 ms at 0 degC 1213 ms at 20 degC and 1168 ms at 50 degC respectively

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms (1152~1273 ms)

The estimated results are also changed under temperature variations the fluctuation of esti-

mated range and exact value is small

TABLE VI Temperature effect test for the Teflon inserted damage specimen

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

0 degC

Exact 635 25

Method 1 113 60~70 79~268 352~1291

Method 2 113 560~619 233~344 1034~1172

Method 3 113 606 257 1141

20 degC

Exact 635 25

Method 1 123 60~75 73~338 298~1220

Method 2 123 619~662 229~310 934~1188

Method 3 123 631 284 1161

50 degC

Exact 635 25

Method 1 137 60~75 72~328 263~1228

Method 2 137 622~687 183~299 666~1086

Method 3 137 642 264 960

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1247 ms at 0 degC 1213 ms at 20 degC and 1168 ms at 50 degC

- 28 -

542 Impact-induced damage

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Vol

tage

(mV

)

Transmitted WaveReflected Wave

No reflected wave

(a) 0 degC

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Transmitted WaveReflected Wave

No reflected wave

(b) 50 degC

Fig 16 False alarm test under temperature variations It is shown that proposed damage char-acterization algorithm is robust under temperature variations

The false alarm test in pristine condition of the impact-induced damage specimen under

temperature variation was investigated (Fig 16) It is validated that the proposed damage di-

agnosis algorithm is robust under temperature variations Fig 17 shows the obtained signals

from the impact-induced damage (impacted 3 times) specimen under temperature variation

The analyzed results of the obtained signals are reported in Table VII The measured 119907119907119894119894 in

each temperature and the assumed variation are same as the artificial damage case

11 115 12 125 13-5

0

5

Time (ms)

Volta

ge (m

V)

0 oC20 oC50 oC

(a) Pitch-catch signal

11 115 12 125 13 135

-1

0

1

2

3

Time(ms)

Vol

tage

(mV

)

0 oC20 oC50 oC

(b) Pulse-echo signal

Fig 17 Signals obtained from impact-induced damage specimen (Impact 3) under tempera-ture variation

- 29 -

TABLE VII Temperature effect test for the impact induced damage specimen (Impact 3 case)

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

0 degC

Method 1 135 75~90 72~310 267~1148

Method 2 135 705~779 264~381 978~1126

Method 3 135 763 289 1071

20 degC

Method 1 148 50~100 06~380 21~1134

Method 2 148 758~838 249~368 843~1125

Method 3 148 798 308 1044

50 degC

Method 1 153 75~100 53~415 174~1227

Method 2 153 804~889 214~336 702~1103

Method 3 153 815 320 1050

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1247 ms at 0 degC 1213 ms at 20 degC and 1168 ms at 50 degC

- 30 -

CHAPTER 6 CONCLUDING REMARKS

61 Executive summary

In this study the reference-free damage detection localization and quantification tech-

nique in a multi layered composite plate is proposed Additional reflected A0 Lamb wave

mode from the damage due to multiple reflections in damage area is used as a feature for

characterizing the damage First damage is characterized in terms of a damage index (m-

value) defined as the ratio of the physical length of the damage to the wavelength of the A0

Lamb wave mode Then the range of possible damage location length and severity are esti-

mated with constraints This procedure can be accomplished without prior baseline data To

realize the proposed technique the matching pursuit technique is modified for reduction of

the error by considering the dispersive characteristic of Lamb wave The matching pursuit

method estimates the arrival time of each wave packets accurately even two wave packets are

overlapped The numerical simulations and experimental tests including Teflon tape inserted

and impact-induced damage are investigated for the feasibility of the proposed technique

Moreover the effectiveness of the proposed technique under changing temperature is exper-

imentally verified

- 31 -

62 Suggestion for further study

The limitations of this study and the possible improvement and further studies are listed as

follow

1 Measurement of impact-induced damage length using conventional NDT technique

Additional experimental verification for impact-induced damage characterization is

needed by measuring the real length of damage using conventional NDT technique

such as ultrasonic scan or infrared camera

2 Limitation of both side collocated PZT installation In this study excitation PZTs are

installed both side and collocated for decomposing A0 Lamb wave mode However it

can produce decomposition error due to imperfection of PZT location and their installa-

tion Moreover in many cases assess to both surface of structures is not allowed such

as aircraft applications To improve this limitation a reference-free damage characteri-

zation using single side PZTs will be considered

3 Complex structure application In this study damage is characterized in simple plate-

like structure When the monitored structure becomes complex the response signals al-

so becomes complex due to its geometry and boundary Thus the investigations for

complex structure application such as stiffener spar and curvature are needed

- 32 -

REFERENCES

[1] Cawley P and Alleyne D (1996) ldquoThe use of Lamb waves for the long range inspection of large structuresrdquo Ultrasonics 34 pp287-290

[2] Ghosh T Kundu T and Karpur P (1998) ldquoEfficient use of Lamb waves for detecting defects in large platesrdquo Ultrasonics 36 pp791-801

[3] Guirguitui V (2003) ldquoLamb wave generation with piezoelectric wafer active sensors for structural health monitoringrdquo SPIE Conference on Smart Structures and Materials San Diego CA 5391 pp111-422

[4] Ragvahan A and Cesnik CE S (2004) ldquoModeling of piezoelectric-based Lamb wave generation and sensing for structural health monitoringrdquo SPIE Conference on Smart Struc-tures and Materials San Diego CA 5391 pp419-430

[5] Rose JL Pilarski A and Dirit JJ (1993) ldquoAn approach to guided wave mode selection for inspection of laminated platerdquo Journal of Reinforced Plastics and Composites 12(5) pp536-544

[6] Guo N and Cawley P (1993) ldquoThe interaction of Lamb waves with delaminations in composite laminatesrdquo Journal of Acoustical Society of America 94(4) pp 2240-2246

[7] Valdez SHD and Soutis C (2002) ldquoReal-time nondestructive evaluation of fiber com-posite laminates using low-frequency Lamb wavesrdquo Journal of Acoustical Society of Amer-ica 111(5) pp 2026-2033

[8] Yang M and Qiao P (2005) ldquoModeling and experimental detection of damage in virous materials using the pulse-echo method and piezoelectric sensoractuatorsrdquo Smart Materials and Structures 14 pp1083-1100

[9] Lestari W and Qiao P (2005) ldquoApplication of wave propagation analysis for damage identification in composite laminated beamsrdquo Journal of Composite Materials 39(22) pp1967-1984

[10] Kessler S S Spearing S M and Soutis C (2002) ldquoDamage detection in composite ma-terials using Lamb wave methodsrdquo Smart Materials and Structures 11 pp269-278

[11] Sohn H Park G Wail J R Limback N P and Farrar C R (2004) ldquoWavelet-based active sensing for delamination detection in composite structuresrdquo Smart Materials and Structures 13 pp 153-160

[12] Petculescu G Krishnaswamy S and Achenback J D (2008) ldquoGroup delay measure-ments using modally selective Lamb wave transducers for detection and sizing of delamina-tions in compositesrdquo Smart Materials and Structures 17

[13] Sohn H Park H W Law K H and Farrar C R (2007) ldquoDamage detection in compo-

- 33 -

site plates by using an enhanced time reversal methodrdquo Journal of Aerospace Engineering (ASCE) pp 141-151

[14] Sohn H (2007) ldquoEffects of environmental and operational variability on structural health monitoringrdquo Philosophical Transactions of the Royal Society Ardquo 365 pp539-560

[15] Rytter A (1993) ldquoVibration based inspection of civil engineering structuresrdquo PhD The-sis Department of Building Technology and Structural Engineering University of Aalborg Aalborg Denmark

[16] Hayashi T and Kawashima K (2002) ldquoMultiple reflections of Lamb waves at a delami-nationrdquo Ultrasonics 40 pp193-197

[17] Wang C H and Rose L R F (2003) ldquoWave reflection and transmission in beams con-taining delamination and inhomogeneityrdquo Journal of Sound and Vibration 264 pp 851-872

[18] Ramadas C Balasubramaniam K Joshi M and Krishnamurthy C V (2009) ldquoInterac-tion of the primary anti-symmetric Lamb mode (Ao) with symmetric delaminations numer-ical and experimental studiesrdquo Smart Materials and Structures 18

[19] Ramadas C Balasubramaniam K Joshi M and Krishnamurthy C V (2010) ldquoInterac-tion of guided Lamb waves with an asymmetrically located delamination in a laminated composite platerdquo Smart Materials and Structures 19

[20] Wilcox P Lowe M and Cawely P (2001) ldquoThe effect of dispersion on long-range in-spection using ultrasonic guided wavesrdquo NDTampE International 34 pp1-9

[21] Mallat S G and Zhang Z (1993) ldquoMatching pursuits with time-frequency dictionariesrdquo IEEE Tansactions on Signal Processing 41(12) pp3397-3415

[22] Zhang G Zhang S and Wang Y (2000) ldquoApplication of adaptive time-frequency de-composition in ultrasonic NDE of highly scattering materialsrdquo Ultrasonics 38 pp 961-964

[23] Hong J C Sun K H and Kim Y Y (2005) ldquoThe matching pursuit approach based on the modulated Gaussian pulse for efficient guided-wave damage inspectionrdquo Smart Mate-rials and Structures 14 pp548-560

[24] Herakovich C T (1997) Mechanics of Fibrous Composites Wiley 480 pages

[25] Reddy J N (1985) ldquoA review of the literature on finite-element modeling of laminated composite platesrdquo The Shock and Vibration Digest 17(4) pp 3-8

[26] Reddy J N (2003) Mechanics of Laminated Composite Plates and Shells Theory and Analysis CRC Press 856 pages

[27] Palacz M Krawczuk M and Ostachowicz W (2005) ldquoThe spectral finite element mod-el for analysis of flexure-shear coupled wave propagation Part 1 laminated multilayer

- 34 -

composite beamrdquo Composite Structures 68 pp37-44

[28] Zhang Y C and Harding J (1990) ldquoA numerical micromechanics analysis of the me-chanical properties of a plain weave compositerdquo Computer amp Structures 36(5) pp 839-844

[29] Nemat-Nasser S and Hori M (1999) Micromechanics Overall Properties of Heteroge-neous Materials Elsevier 810 pages

[30] Scida D Aboura Z Benzeggagh M L and Bocherens E (1999) ldquoA micromechanics model for 3D elasticity and failure of woven-fibre composite materialrdquo Composites Science and Technology 59 pp 505-517

[31] Thomas Y and Wu X H (1997) ldquoA multiscale finite element method for elliptic prob-lems in composite materials and porous mediardquo Journal of Computational Physics 134 pp 169-189

[32] Lee K H Moorthy S and Ghosh S (1999) ldquoMultiple scale computational model for damage in composite materialsrdquo Computer Methods in Applied Mechanics and Engineer-ing 172 pp 175-201

[33] Namilae S and Chandra N (2005) ldquoMultiscale model to study the effect of interfaces in carbon nanotube-based compositesrdquo Journal of Engineering Materials and Technology (ASME) 127 pp222-232

[34] Adams D E (2007) Health monitoring of Structural Materials and Components Wiley 460 pages

[35] Gopalakrishnan S (2009) ldquoModeling Aspects in Finite Elementrdquo Encyclopedia of Struc-tural Health Monitoring 2 pp791-809

[36] Palacz M Krawczuk M and Ostachowicz W (2005) ldquoThe spectral finite element mod-el for analysis of flexure-shear coupled wave propagation Part 2 Deminated multilayer composite beamrdquo Composite Structures 68 pp45-51

[37] Baik J M and Thopson R B (1984) ldquoUltrasonic scattering from imperfect interfaces A quasi-static modelrdquo Journal of Nondestructive Evaluation 4(3) pp177-196

[38] Luo H and Hanagud S (2000) ldquoDynamics of delamincated beamsrdquo International Jour-nal of Solids and Structures 37 pp1501-1519

[39] Kim S B and Sohn H (2007) ldquoInstantaneous reference-free crack detection based on polarization characteristics of piezoelectric materialsrdquo Smart Materials and Structures 16 pp2375-2387

[40] An Y K and Sohn H (2010) ldquoInstantaneous crack detection under varying temperature and static loading conditionsrdquo The Special Issue of Structural Control and Health Monitor-ing 17(7) pp 730-741

- 35 -

[41] Lee S J and Sohn H (2006) ldquoActive self-sensing scheme development for structural health monitoringrdquo Smart Materials and Structures 15(6) pp 1734-1746

- 36 -

요 약 문

복합재료 손상의 무기저 진단 및 정량화

복합재료는 재료의 경량성 내구성 고강도의 특성으로 인하여 항공기에서부터

교량에까지 널리 사용되고 있다 하지만 외부의 충격으로 인한 복합재료 내부의

손상은 재료의 강도를 감소시켜 구조물의 안정성을 저하시키는 요인이 된다 하지만

복합재료의 손상은 재료 내부에서 발생하여 육안으로는 식별하기도 힘들며 표면에서

보이는 것만으로 내부의 손상 정도를 판단하기는 불가능하다는 단점을 가지고 있다

본 연구에서는 손상 내부에서 반사로 인해 발생하는 파를 이용하여 기저자료를

사용하지 않고 특정 시점에서 얻은 자료만으로 복합재료의 손상을 진단하고 정량화

할 수 있는 기법을 제안한다 제안된 기법의 적용성을 검증하기 위하여 유한요소법을

이용한 수치해석을 수행하였고 Teflon tape 이 삽입된 시편과 실제 충격하중으로

인한 손상을 입은 시편을 이용한 실험을 진행하였다 수치해석 및 실험 결과 제안된

기법이 복합재료의 손상을 성공적으로 진단 및 정량화할 수 있음을 확인하였다 또한

다양한 외부 온도 환경에서 제안된 기법의 환경영향 실험을 수행하였고 다양한

환경에서 손상의 진단 및 정량화가 성공적으로 이루어짐을 검증하였다

핵심어 무기저 손상진단 및 정량화 압전센서 복합재료 유한요소법 Matching pursuit

method

- 37 -

감 사 의 글

기대와 두려움을 안고 시작한 석사과정이 어느덧 2 년 이라는 세월 속에서 추억으로 자리잡아

가고 있습니다 학부생 기간을 포함한 지난 3 년 동안 SHM 이라는 새로운 분야를 접하고 연구를

수행 할 수 있도록 도와주신 고마운 분들이 있어 그 감사를 표하고자 합니다

우선 저의 많은 부족한 점을 일깨워 주시고 연구에 대한 뜻을 더욱 확고히 세울 수 있게 해

주신 지도교수님이신 손 훈 교수님께 깊은 감사의 뜻을 전합니다 짧은 기간이었지만 교수님의

도움으로 많은 것을 배우고 느끼고 깨달을 수 있었습니다 또한 본 연구를 수행할 수 있도록 지

원해주신 국토해양부 (U-City 석 박사과정 지원사업)에 깊은 감사 드립니다

때로 힘들고 지칠 때 항상 저를 믿고 응원해 주시는 가족이 있었기에 지금의 글을 쓸 수 있었

습니다 또한 지금 연구실에서 함께 생활하고 있는 윤규형 민구형 철민이형 현석이형 진열이형

호민이형 병진이 선혜누나와 성대에 계신 창길이형 미국에 계신 의재형과 Debaditiya 에게도 감

사의 말씀을 전합니다 그리고 모두 나열할 수는 없지만 카이스트 생활의 활력소가 되어준 카이

스트 건설 및 환경공학과 선후배들 직원 분들과 SOS 멤버들에게도 감사를 표합니다

앞서 거론한 많은 이들의 도움이 없었다면 저 또한 지금 이 자리에 있을 수 없었을 것입니다

많이 부족하기에 더 노력하고 배우고자 했던 초심을 다시 한번 생각하며 더 큰 결실을 맺기 위해

앞으로 나아가겠습니다 도움을 주신 모든 분들께 이 감사의 글을 바칩니다

2011년 6월 카이스트에서 임형진 올림

- 38 -

CURRICULUM VITAE

Hyung Jin Lim

Research Assistant

The Civil and Environmental Engineering Department

Korea Advanced Institute of Science and Technology (KAIST)

291 Daehak-Ro Guseong-Dong Yuseong-Gu

Daejeon Republic of Korea 305-701

Phone (82)+42-350-3665 Fax (82)+42-350-3610

Email limhj87gmailcom

RESEARCH INTERESTS

Damage Detection Localization and Quantification on Composite and Metallic Structures Pattern Recognition Finite Element Modeling (FEM)

EDUCATION

2009 ndash 2011 MS Civil and Environmental Eng Korea Advanced Institute of Science and Technology (KAIST) Korea Rep

2006 ndash 2009 BS Civil and Environmental Eng Korea Advanced Institute of Science and Technology (KAIST) Korea Rep

JOURNAL PUBLICATIONS The corresponding author is underlined 1 Hyung Jin Lim Hoon Sohn ldquoReference-free damage detection localization and quantification

using in compositerdquo in preparation for Journal of the Acoustical Society of America 2011

2 Chul Min Yeum Hoon Sohn Jeong Bum Ihn Hyung Jin Lim ldquoInstantaneous delamination detection in a composite using a dual piezoelectric transducer networkrdquo Submitted to Compo-sites Science and Technology 2011

3 Yun-Kyu An Hyung Jin Lim Min Koo Kim Jin Yeol Yang Hoon Sohn Chang Guen Lee ldquoInvestigation of applicability of local reference-free damage detection techniques to in-situ

- 39 -

bridgesrdquo Submitted to ASCE Journal of Structural Engineering 2011

4 Hyung Jin Lim Min Koo Kim Hoon Sohn Chan Yik Park ldquoImpedance based damage detec-tion under varying temperature and loading conditionsrdquo Submitted to NDTampE International 2011

5 Hyung Jin Lim Hoon Sohn ldquoQuantitative estimation of transmitted and reflected lamb waves at discontinuityrdquo Journal of the Korean Society for Nondestructive Testing Vol30 No 4 pp1-8 2010

CONFERENCE PROCEEDINGS

The corresponding author is underlined 1 Hyung Jin Lim and Hoon Sohn ldquoReference-free damage detection localization and quantifi-

cation in compositesrdquo ASCE Engineering Mechanics Institute Conference Boston June 2-4 2011

2 Hyung Jin Lim and Hoon Sohn ldquoReference-free delamination quantification within multi-layer composite materialsrdquo KSNT Spring Conference May 26-27 Gyeong-Ju Korea 2011

3 Min Koo Kim Hyun Jin Lim Hoon Sohn Chan Yik Park ldquoImpedance-based bolt loosening detection under varying temperature and loadingrdquo Asian Pacific Workshop on Structural Health Monitoring Tokyo Japan November 30-December 2 2010

4 Hyung Jin Lim Chul Min Yeum and Hoon Sohn ldquoModeling of Impact-induced Delamination in a Multilayer Composite Platerdquo The 23rd KKCNN Symposium on Civil Engineering Taipei (Taiwan) November 13-15 2010

5 Yun-Kyu An Hyung Jin Lim and Hoon Sohn ldquoTemperature independent Delamination Detec-tion using data normalizationrdquo The 3rd International Conference on Multi-Functional Materi-als and Structures Chonbuk National Univ Jeonju Korea September 14-18 2010

6 Hoon Sohn Yun-Kyu An Chang Gil Lee Min Koo Kim Hyung Jin Lim ldquoApplication of online NDT techniques to decommissioned bridge testingrdquo the 5th European Workshop on Structural Health Monitoring Sorrento Italy June 29-July 02 2010

7 Min Koo Kim Hyung Jin Lim Hoon Sohn Chan Yik Park ldquoImpedance based damage diag-nosis of a complex composite aircraft wing under changing loading and temperature condi-tionsrdquo the 5th European Workshop on Structural Health Monitoring Sorrento Italy June 29-July 02 2010

8 Yun-Kyu An Min Koo Kim Hyung Jin Lim and Hoon Sohn ldquoValidations of reference-free crack detection techniques through a decommissioned bridge testrdquo 2010 COSEIK Annual Con-ference Jeju Korea April 8-9 2010

- 40 -

PARTICIPANT PROJECTS

1 Development of Smart Steel Members for Online Steel Structure Monitoring POSCO Corporation (Funded 41000000 KRW (41000 USD) for 010511 to 123111)

2 Development of Enabling Technologies for Structural Health Monitoring Korean Agency for Defense Development (Funded 240000000 KRW (240000 USD) out of total 3200000000 KRW (3200000 USD) for 060108-053111)

3 Development of On-board SHM Technologies for Composite Air Vehicles The Boeing Company (Funded 400000 USD (400000000 KRW) for 080108 to 073111)

EDUCATIONAL EXPERIENCES

Fall 2009

Teaching Assistant CE207 ldquoElementary Structural Engineering and Laboratoryrdquo KAIST 3 unit undergraduate course

EXPERIMENTAL EXPERIENTCES

2010 Monitoring of Real Aircraft Composite Wing Segment Korea Agency for Defense De-velopment (ADD) Daejeon Korea Real time monitoring of composite air vehicles fund-ed by Korea Agency for Defense Development (ADD)

2009-2011

Field Monitoring of Real Bridge Structures Gimpo (Ramp G bridge) amp Icheon (Sam-seung bridge) Korea US-Korea collaborative research for bridge monitoring testbeds

HONORS amp AWARDS

Scholarships for outstanding students granted KAIST (2007-2009) The three prominent students in the civil and environmental engineering department of KAIST were selected as recipients This scholarship provided 2000$ per year for three years

- 41 -

Page 10: 복합재료 손상의 무기저 진단 및 정량화 - KAISTssslab.kaist.ac.kr/article/pdf/Thesis_2.pdf · 1 Declaration of Ethical Conduct in Research: I, as a graduate student

the previous section middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 19

8 Test specimens for the experimental verification In the Teflon inserted damage

Teflon tapes with 25mm diameter were inserted at the center of the specimen The

impact-induced damage was introduced to the specimen using the impact test equip-

ment middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 21

9 Overall experimental setup (a) Data acquisition (DAQ) system and (b) Installed

circular PZT transducer on the specimen with its dimension middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 23

10 (a) Impact test equipment for introducing impact-induced damage and (b) Tip

of the impact tester with 10mm diametermiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 23

11 Test specimen in temperature chamber middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 24

12 A0 Lamb wave mode obtained using pitch-catch method from the Teflon in-

serted damage specimen (a) The reflected wave from damage is shown (b) The Tef-

lon inserted damage is detected by capturing the reflected wave middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 25

13 The signal obtained from the test specimen subjected to impact The impact

was introduced at the same location of the specimen and the signal was obtained after

each impacts As the number of impact is increased the transmitted A0 Lamb wave

mode is delayed and attenuated middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 26

14 False alarm test for the proposed damage detection algorithm in pristine condi-

tion of the specimen Reflected wave from damage is not captured false alarm is not

occurred middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 26

15 Signals obtained from the Teflon inserted damage specimen under temperature

variation Time delay and amplitude variation are occurred middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 27

16 False alarm test under temperature variations It is shown that proposed dam-

vi

age characterization algorithm is robust under temperature variationsmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 29

17 Signals obtained from impact-induced damage specimen (Impact 3) under

temperature variation middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 29

vii

CHAPTER 1 INTRODUCTION

11 Motivation

Composite materials has been increasingly demands in modern structures ranging from the

latest aircraft to retrofitted bridge deck due to its light weight superior specific strength and

stiffness properties However the occurrence of damage due to abrupt impact or sequential

accumulation of defect may considerably degrade structural strength or stiffness and reduce

the safety and reliability of the system Recently Lamb waves based damage detection tech-

niques using piezoelectric (PZT) transducers have emerged as one of the promising structural

health monitoring (SHM) technology for detecting hidden damage because of the possibility

of long range inspection and high sensitivity [1-4]

12 Literature review

This section will discuss researches that have been previously carried out on topics of

guided wave based SHM techniques using PZT transducers for damage in composite materi-

als The early investigations of interaction between Lamb wave and damage in composite can

be found in Rose et al and Guo et al [5-6] From further studies reflected signals from dam-

age were shown to be an effective method for damage diagnosis and localization using PZTs

[7-9] The reflected signals from damage were collected using one PZT as an actuator and the

other PZT as a sensor However the techniques cannot be used to determine the extent of

- 1 -

damage with reasonable precision

Kesseler et al proposed a technique monitoring the wavelet coefficients of transmitted ul-

trasonic signal amplitude as a potential way to damage detection [10] One drawback of this

technique is that the variability in coupling and the sensitivity of installed PZT transducers

can induce changes in the amplitude and energy of the signal that are cause false-alarm Oth-

er wavelet based delamination detection and localization technique developed using PZT

sensor network and outlier analysis [11]

Damage detection and quantification in composite plate was also investigated by group de-

lay measurement due to damage [12] However it is assumed that the group velocity in dam-

age area as the group velocity propagates half thickness of the structure in dispersion curve in

the quantification procedure Thus the damage quantification was investigated only Teflon

inserted damage case

Sohn et al proposed a damage detection technique which does not rely on any past base-

line signals to assess damage in composite plates by using an enhanced time reversal method

[13] The damage is assessed using the violation of time reversibility when nonlinearity is

caused by the damage along a direct wave path Thus it allows instantaneous identification

of delamination without requiring the baseline signal for comparison

- 2 -

13 Objective and uniqueness

In this study a reference-free damage characterization technique including detection lo-

calization and quantification of impact-induced damage in composite materials is developed

The general damage characterization can be categorized into the following levels [14]

Level 0 Define damage

Level 1 Determine the presence of damage within the structure

Level 2 Locate the regions of damage

Level 3 Quantify the damage

Level 4 Predict the remaining service life of the structure

Through the proposed technique from level 1 to level 3 can be achieved

Most of common approaches of damage characterization techniques described in the pre-

vious section use prior obtained baseline data However these approaches can cause false

alarm due to the effect of environmental variations such as temperature and ambient vibra-

tions [15]

The uniqueness of this study is

1 Damage is characterized without using prior measured baseline data

2 Robustness of the proposed technique under temperature variation is investigated

3 Additional waves reflected from inside damage have been used in addition to transmit-

ted and reflected waves

- 3 -

14 Organization

The organization of this thesis is summarized as follows

Chaper1 A general introduction including motivation literature review objective and

uniqueness of this study is described

Chaper2 Interactions between Lamb wave mode and damage in composites such as time

delay attenuation and multiple reflections are summarized Damage index (m-value) for

damage characterization is defined

Chaper3 Algorithms for the reference-free damage characterization technique using m-

value are developed Modified matching pursuit for this study is also described

Chaper4 Verification of the algorithm using numerical simulation is presented Various

modeled damage cases are characterized

Chaper5 Experimental verifications are described Two damage cases (Teflon inserted and

impact-induced damage) are investigated for the feasibility of the proposed damage charac-

terization technique Temperature effect test using temperature chamber is also carried out

Chaper6 Conclusion and summary of the present study is reported In advance suggestions

and future work recommendations are represented

- 4 -

CHAPTER 2 INTERACTION OF LAMB WAVE WITH DAMAGE

21 A0 Lamb wave mode delay and attenuation in the time domain

Time delay and attenuation of the A0 mode Lamb wave after passing through the damage

area due to effective thickness reduction were reported in previous researches [812] Moreo-

ver the attenuation of the A0 Lamb wave mode is occurred by degrading of shear modulus of

in the damage area However the time delay and amplitude variation of the A0 Lamb wave

mode also occurred even when the structure is health condition by the material property

changing due to temperature variation of the structure (Fig 1) Thus conventional damage

characterization methods using the time delay and attenuation of A0 Lamb wave mode as a

feature can cause false-alarm in real application

11 115 12 125 13-10

-5

0

5

10

Time (ms)

Vol

tage

(mV

)

IntactDamage

Reflected wave from Damage

(a)

11 115 12 125 13-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

0 oC20 oC50 oC

(b)

Fig 1 Lamb wave propagation characteristics (a) A0 Lamb wave mode time delay and atten-uation after passing through a damage (b) The time delay and amplitude variation of A0 Lamb wave mode under temperature variation in pristine condition of structure Because of this conventional damage detection techniques in composites have limitations for real struc-ture application

- 5 -

22 Multiple reflections in the damage area

Standing wave patterns due to multiple reflections of A0 Lamb wave mode in the damage

area were observed through numerical simulations and ultrasonic wave-field imaging using a

scanning laser vibrometer [16] The formation of standing waves in damage is illustrated

through Fig 2 When the incident waves enter the damage area a significant portion of the

waves are reflected back from the exit Then the reflected waves travel through the damage

and undergo reflection again at the original entrance As a consequence of multiple reflec-

tions at the entrance and the exit a considerable amount of ultrasonic energy is trapped inside

the damage area [17-19]

Multiple Reflections

Incident Wave Transmitted Wave

Damage

Fig 2 Multiple reflections in the damage area The multiple reflections of A0 Lamb waves in the damage area are utilize as a feature for characterization of the damage in this study

- 6 -

CHAPTER 3 THEORETICAL DEVELOPMENT

31 Overview

Damage

l = l1 + d + l2

d = vdtd l1 = vit1 l2 = vit2

t = t1 + td + t2

Excitation and Sensing PZT Sensing PZT

l

Transmitted Wave

Reflected Waves

Incident Wave

Fig 3 Parameters and notations of the proposed damage characterization technique Damage is detected by identifying the additional reflected wave from the damage due to multiple re-flections (dashed lines) Localization and quantification of the damage is carried out by measuring the arrival time of transmitted and reflected waves from the damage

In this section the reference-free damage characterization algorithm is proposed Damage

diagnosis and quantification are achieved by identifying the additional reflected waves due to

multiple reflections in the damage area (dashed lines in Fig 3) Moreover damage is local-

ized using additional information from pulse-echo method implement For the localization

and the quantification of damage the variables such as the path length and A0 Lamb wave

mode velocity in damage area are defined (Fig 3) For example 119897119897 is expressed as

119897119897 = 1198971198971 + 119889119889 + 1198971198972 (1)

where 1198971198971 119889119889 and 1198971198972 are the distance from the excitation PZTs to the entrance of damage the

damage length and the distance from the exit of damage to the sensing PZT respectively

- 7 -

Moreover 119907119907119894119894 and 119907119907119889119889 are also defined as the A0 Lamb wave mode velocity in pristine area

and damage area respectively Note that the whole procedure of the damage characterization

is accomplished based solely on the signals obtained from the current state of the structure

32 Damage index (m-value)

The proposed reference-free damage characterization technique utilizes the waves reflect-

ed additionally from the damage area as a feature The arrival time difference of the transmit-

ted and reflected A0 Lamb wave modes reflected from the exit and the entrance of the delam-

ination area one times each (∆119905119905) contains the information of the physical length of the dam-

age (119889119889) and A0 Lamb wave mode velocity in the damage area (119907119907119889119889) The wave velocity in the

damage area which is related to the shear modulus (G) represents the severity of the damage

∆119905119905 = 2119905119905119889119889 =2119889119889119907119907119889119889

(2)

However based on the measured data available the 119889119889 and 119907119907119889119889 (or G) cannot be uniquely

determined Fig 4 shows two FEM models which have different physical length and wave

velocity (G) of the damage Detail explanation of delamination modeling technique will be

covered in section 4 The similar signal in the time domain and ∆119905119905 are obtained from the

two different damage cases

- 8 -

Damage 1

Damage 2

20

Unit mm15

90 G value reduction

943 G value reduction

m-value = 248

m-value = 248

015 02 025 03 035

-2

-1

0

1

2

Time (ms)

Ampl

itude

Damage 1Damage 2

Fig 4 Two different FEM models which have different 119889119889 and 119907119907119889119889 (or G) This represents that the damage cannot be characterized uniquely based on the measured data available

Thus in this study the damage index (m-value) which can be uniquely defined based on

the available data is introduced The m-value is defined as the ratio of the physical length of

the damage and wavelength of the A0 Lamb wave mode in the damage area (120582120582119889119889)

119898119898 =119889119889120582120582119889119889

=119889119889119889119889119907119907119889119889

(3)

In many cases the transmitted and reflected waves are overlapped and it is hard to decom-

pose and calculate the arrival times of them Matching pursuit method is an alternative way

to accurately decompose of the overlapped waves and estimate the parameters of each wave

such as arrival time (Fig 5) Detail explanation of the matching pursuit method is covered in

section 36

015 02 025 03-2

-1

0

1

2

Time (ms)

Am

plitu

de

OriginalApproximated

(a)

015 02 025 03-2

-1

0

1

2

Time (ms)

Ampl

itude

Wave 1Wave 2

(b)

Fig 5 (a) Example of the overlapped signal with comparing approximated signal using the matching pursuit method (b) Decomposed signal using the matching pursuit method The matching pursuit method provides a conventional way to decompose and identify each wave

- 9 -

33 Damage detection algorithm (Level 1)

The damage diagnosis is accomplished by extracting the additional reflected A0 Lamb

wave modes from damage due to multiple reflections If A0 Lamb wave modes are decom-

posed perfectly it is trivial that one transmitted A0 Lamb wave mode is always captured in

the response signal If damage exists other waves in the response signal behind the transmit-

ted A0 Lamb wave mode are presented Since this approach does not rely on previously col-

lected data the diagnosis is independent from environmental variation such as temperature

changes

It is reported that the dispersion of Lamb waves cause wave-packets to spread out in space

and time as they propagate through a structure [20] Thus the reflected waves from the dam-

age have longer pulse duration than the transmitted wave due to their longer propagating dis-

tance In this study the matching pursuit method is modified to capture more spread wave

than the transmitted A0 Lamb wave mode Detail explanation of the modified matching pur-

suit method is described in latter section

Let 1199041199041 and 1199041199042 as the pulse duration (scale) of the transmitted and additional reflected A0

Lamb wave mode respectively

ldquoIf 1199041199042 gt 1199041199041 damage exists

Otherwise there is no damagerdquo

- 10 -

34 Damage quantification algorithm (Level 3)

As mentioned in the previous section damage is quantized by estimating 119889119889 and 119907119907119889119889 using

pitch-catch method The procedure to estimate 119889119889 and 119907119907119889119889 is

(i) Measure 119897119897 119905119905 and 119905119905119889119889

(ii) Assume the range of 119889119889 (0 lt 119889119889 lt 119897119897)

(iii) Calculate 119907119907119894119894 and 119907119907119889119889 with a constraints (0 lt 119907119907119889119889 lt 119907119907119894119894) (Eq 4 and 5)

(iv) Estimate m-value and find a feasible range of 119889119889 and 119907119907119889119889

119907119907119894119894 =119897119897 minus 119889119889119905119905 minus 119905119905119889119889

(4)

119907119907119889119889 =119889119889119905119905119889119889

(5)

35 Damage localization algorithm (Level 2amp3)

For damage localization 1198971198971 is estimated by measuring the arrival time from the excitation

PZT to the entrance of delamination (1199051199051) with additional pulse-echo method implementation

The procedure is similar with the previous damage quantification algorithm

(i) Measure 119897119897 119905119905 119905119905119889119889 and 1199051199051

(ii) Assume the range of 1198971198971 (0 lt 1198971198971 lt 119897119897)

(ii) Compute possible combinations of 1198971198972 119889119889 and 119907119907119889119889 with constraints

(0 lt 119907119907119889119889 lt 119907119907119894119894 119889119889 gt 0) (Eq 5 6 and 7)

- 11 -

1198971198972 = 1198971198971119905119905 minus 1199051199051 minus 119905119905119889119889

1199051199051 (6)

119889119889 = 119897119897 minus 1198971198971 minus 1198971198972 (7)

Since a PZT for pulse-echo method acts as both an actuator and a sensor the obtained sig-

nal contains both the excitation and the response signal Thus the reflected wave from the

entrance of damage can be buried into the tail of excitation signal and cause error on arrival

time of reflected wave from damage Even in this case damage can be localized and quan-

tized using the constraints When first wave packet reflected from the entrance of damage is

buried into the excitation signal the next wave packet reflected from the exit of damage can

be utilized alternatively If the algorithm regards the second reflected wave as the first re-

flected wave there are no combinations of estimated variable meeting the constraints Then

the algorithm repeats the process by regarding the wave as the second reflected wave the

constraints can select proper combination of the estimated variables

The estimated values have range bounded by the constraints When additional information

of 119907119907119894119894 is utilized the damage can be characterized more accurately In this study three meth-

ods are proposed and verified as information level of 119907119907119894119894

Method 1 Solely based on the current data measurement (Reference-free)

Method 2 Limit the possible range of 119907119907119894119894 (Prior knowledge of the system)

Method 3 Measure 119907119907119894119894 (Additional data need to be collected)

Method 1 described in this section can be utilized without any information of 119907119907119894119894 Meth-

od 2 produces more accurate estimation results using prior knowledge of the system such as

- 12 -

119907119907119894119894 variation under temperature changes Method 3 gives unique estimated value using meas-

ured 119907119907119894119894 from other path in the same time In section 4 and 5 damage is characterized using

these three methods and the results are compared each another

36 Matching pursuit method considering the dispersion effect of Lamb

wave

Matching pursuit method has been developed for signal compression approximation and

characterization of obtained signal and function [21-23] The key concept of the matching

pursuit method is that any signal (function) can be decomposed into a linear expansion of

signals selected from a pool of basis signals (functions) called lsquodictionaryrsquo

Let 119891119891 be a function to be analyzed then 119891119891 can be approximated a set of functions select-

ed from the dictionary 119863119863 in order to best match This is done by successive approximations

with orthogonal projections Let1198921198921205741205740 isin 119863119863 the function 119891119891 can be decomposed into

119891119891 = lang1198911198911198921198921205741205740rang1198921198921205741205740 + 119877119877119891119891 (8)

where 119877119877119891119891 is the residual function after approximating 119891119891 using 1198921198921205741205740 1198921198921205741205740 is orthogonal to

119877119877119891119891 hence

1198911198912 = lang1198911198911198921198921205741205740rang2

+ 1198771198771198911198912 (9)

To minimize 119877119877119891119891 1198921198921205741205740 is chosen to maximize lang1198911198911198921198921205741205740rang The matching pursuit method is

an iterative algorithm that decomposes the residue 119877119877119891119891 by projecting it on a function of 119863119863

that matches 119877119877119891119891 almost at best as it was done for 119891119891 The residue 119877119877119899119899119891119891 is decomposed into

119877119877119899119899119891119891 = lang119877119877119899119899119891119891119892119892120574120574119899119899rang119892119892120574120574119899119899 + 119877119877119899119899+1119891119891 (10)

- 13 -

the decomposition is carried up to the order 119898119898 119891119891 can be expressed as

119891119891 = lang119877119877119899119899119891119891119892119892120574120574119899119899rang119892119892120574120574119899119899

119898119898minus1

119899119899=0

+ 119877119877119898119898119891119891 (11)

The original function 119891119891 is decomposed into a sum of dictionary elements that are chosen to

best match its residues

In this study the matching pursuit method is utilized to accurately decompose of the over-

lapped Gaussian pulses and estimate the parameters of the pulse which used for damage

characterization The 119894119894119905119905ℎ Gaussian pulse in the dictionary for approximating the obtained

signal is expressed as

119892119892120574120574119894119894 = 119891119891(119860119860119894119894 119906119906119894119894 119904119904119894119894 119889119889119894119894120601120601119894119894) (12)

where 119860119860119894119894119906119906119894119894119904119904119894119894119889119889119894119894 and 120601120601119894119894 represent amplitude arrival time scale (pulse duration) excita-

tion(central) frequency and phase of the Gaussian pulse respectively The parameters of the

pulse are estimated using a nonlinear regression based optimization method (Gauss-Newton

method) based on an initial guess of the parameters Detail procedure for estimating the pa-

rameters of the pulse is expressed in Hong et al [23]

Among the estimated parameters of the pulses pulse duration (scale) represents the disper-

sion effect If a wave-packet is spread out the estimated pulse duration is increased Using

this feature the matching pursuit method is modified to capture waves which have longer

pulse duration than transmitted A0 Lamb wave mode If the estimate pulse duration of a wave

is less than the transmitted wave the modified matching pursuit method regards it as error

- 14 -

CHAPTER 4 NUMERICAL SIMULATION

41 Composite modeling method

Composite materials can be modeled from nano to macro level depends on its engineering

purpose In general four methods are utilized for modeling of composite materials First

composite is modeled as a homogenous material expressed in terms of effective material

properties [24] The effective material properties are estimated using the modulus and the

volume fraction of each constituent theoretically (Rule of mixture) and measured experimen-

tally Second a model is established that each laminate which have own material properties

are totally bonded together In this model the stacking sequence and direction of each lami-

nates are considered additionally [25-27] Third method for composite modeling in micro

level called micromechanics [28-30] This model considers the interaction between fibers and

matrix in laminate Last multiscale modeling which the scope of modeling is extended to

molecular or quantum level of constituents [31-33] This model starts from molecular me-

chanics or molecular dynamics then the scale of model leads to the nano micro and macro

level by incorporating small scale models As the scale of models is smaller it provides more

logical way than other large scale of models while the computational cost is more expensive

The scope of this study is focused on the dynamic behavior of composite structure the beam

or plate model was utilized for the numerical simulation

- 15 -

42 Damage modeling method

Generally three damage models have been developed and utilized in the real applications

First the stiffness reduction model called as the equivalent element model was introduced

[17 34-35] This model is simply realized by replacing the region around damage area by a

region with adjusted material properties (Youngrsquos modulus shear modulus density etc)

globally Next the duplicated node model is done by disconnecting the element connectivity

assuming there is no interaction among the delaminated surface This model can be utilized

when the location and the extension of damage is important issue [34-36] Last the contact

mechanics model using the principles of contact mechanics the two interfacial forces (con-

tact resistance and sliding resistance) in the delaminated surface [37-38] In this model the

delaminated surfaces are modeled as nonlinear functions of the interlaminar normal strain

and the interlaminar shear strain This model is computationally expensive while provides

more logical way than other models There is however no universally accepted model for

general impact-induced damage in composites In the case of impact-induced damage it is

impossible to predict or estimate the location and number of delaminated layers Thus from

the uncertainty of impact-induced damage the equivalent element model can be considered

as the effective damage model

- 16 -

43 Numerical simulation description

TABLE I Material property of simulation model and test specimen

Tensile Strength 657 MPa Tensile Modulus 59 GPa Shear Strength 804 MPa Shear Modulus 24 Gpa Specific Gravity 16

PZT A

PZT C

PZT BDamage

3

250 125 125

500 Unit mm

Fig 6 Description of the numerical simulation model

The feasibility of the proposed damage characterization technique is verified by 2-D FEM

simulation using MSC PATRAN and NASTRAN Fig 6 presents the dimension PZTs and

damage location of the simulation model In the model 05mm x 05mm square shell element

was used for modeling of the A0 mode whose flexural wave characteristic The element size

was determined by considering the wavelength of the propagation Lamb wave modes The

material properties of the simulation model and damage cases are presented in Table I and

Table II respectively

PZT A and C act as both actuator and sensor for collecting pulse-echo signal They were

installed on the plate which two of them are both-side collocated for the A0 Lamb wave

- 17 -

mode decomposition [39-40] PZT B was only used as a sensor for collecting pitch-catch sig-

nal For simulating the excitation and sensing of PZTs the pin-force model was adopted that

an input force apples to the end nodes of excitation PZT and averaging element strains corre-

sponding sensing area A Gaussian pulse (tone-burst) 50 kHz central frequency was excited

as an input signal

TABLE II Damage cases for numerical simulation

119949119949120783120783 (mm) 119941119941 (mm) G value reduction ()

Case 1 650 200 90 darr

Case 2 675 150 90 darr

Case 3 675 150 93 darr

Case 4 900 150 90 darr

- 18 -

44 Numerical simulation results

015 02 025 03 035

-2

0

2

Time (ms)

Am

plitu

de

IntactDamage

Reflected wave from damage

(a) Pitch-catch signal

015 02 025 03 035

-2

0

2

Time (ms)

Am

plitu

de

IntactDamage

Reflected waves from damage

(b) Pulse-echo signal Fig 7 Comparison of A0 Lamb wave mode signals from numerical simulation between intact and damage cases Time delay attenuation and multiple reflections due to damage simulated successfully using the damage modeling method described in the previous section

Fig 7 shows that comparison of A0 Lamb wave mode signals between intact and one of

damage cases in the simulation The features of A0 Lamb wave mode passing through dam-

age area described in section 2 are successfully simulated Thus the equivalent element

method is proper for modeling damage in composites In the damage cases the additional

waves reflected from damage is observed in both pitch-catch and pulse echo signal

Table III presents the result of the proposed damage characterization algorithm using the

numerical simulation cases shown in Table II In lsquoMethod 2rsquo 119907119907119894119894 variation is assumed to plusmn

5 of the normal value (1190~1315ms) The normal value of 119907119907119894119894 for lsquoMethod 3rsquo is meas-

ured as 1253 ms In each case the damage is characterized using the proposed technique

Note that as the information of 119907119907119894119894 is specified the estimated value is closer to the exact

value

- 19 -

TABLE III Damage characterization for the numerical simulation models

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

Case 1

Exact 650 200

Method 1 248 50~70 79~414 160~834

Method 2 248 603~667 135~241 271~486

Method 3 248 635 188 379

Case 2

Exact 675 150

Method 1 193 60~75 01~251 37~651

Method 2 193 618~682 114~222 295~575

Method 3 193 65 168 435

Case 3

Exact 675 150

Method 1 233 60~75 66~302 140~649

Method 2 233 668~738 85~196 182~419

Method 3 233 703 140 301

Case 4

Exact 900 150

Method 1 204 75~100 48~349 118~856

Method 2 204 860~951 107~216 264~530

Method 3 204 906 162 397

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1253 ms 1190~1315 ms Measured 119907119907119894119894 1253 ms

- 20 -

CHAPTER 5 EXPERIMENTAL VERIFICATION

51 Overview

In this section experimental study is investigated so that the proposed damage characteri-

zation technique Overall experimental setup and test specimen are shown in Fig 8 and 9

The experimental setup consists of test specimens data acquisition (DAQ) system and self-

sensing circuit Self-sensing circuit was installed for pulse-echo measurement using PZTs

[40-41]

PZT A(PZT C) PZT B

500

500

140

75

250

Teflon Inserted Damage (d =25)

Unit mm

Thickness = 3

PZT A(PZT C) PZT B

500

500

150

90

250

Impact Location

Unit mm

Thickness = 3

(a) Teflon inserted damage specimen

(b) Impact-induced damage specimen

Fig 8 Test specimens for the experimental verification In the Teflon inserted damage Teflon tapes with 25mm diameter were inserted at the center of the specimen The impact-induced damage was introduced to the specimen using the impact test equipment

- 21 -

511 Specimen

Two carbon composite plates with a dimension of 500times500times3mm3 were manufactured by

Nexcomes Inc by stacking 12 layers of CF3327 3K prepreg (woven fabrics type) manufac-

tured by Hankuk Fiber Inc The material properties of the specimen are presented in Table I

One of the specimens Teflon tapes with 25mm diameter was inserted in every 3 layers at the

center of the specimen to mimic the damage The other specimen is utilized for the character-

ization of real impact-induced damage

512 Data acquisition (DAQ) system and PZT

The data acquisition (DAQ) system consists of an arbitrary waveform generator (AWG) a

high speed signal digitizer (DIG) and two multiplexers (MUX) Using the 16-bits AWG

tone-burst input signal with a plusmn10 peak-to-peak voltage were generated The response signal

was measured by the 14-bit DIG For the signal-to-noise ratio reduction both pitch-catch and

pulse-echo signals were measured 150 times and averaged in the time domain 3 identical

circular PZTs of 10mm diameter manufactured by APC International Ltd were installed on

the specimen A Gaussian pulse (tone-burst) 50 kHz central frequency was excited as an in-

put signal The excitation frequency was chosen in order to avoid overlapping of the S0 and

A0 Lamb wave modes and generating of the higher modes (S1 A1hellip)

- 22 -

PZT Actuator

PZT Sensor

AWG

DIG

DAQ System Specimen

SensingActuatingMUX

1

23

4

1 Sending signal to MUX

4 Sending signal to DIG

Self-SensingCircuit

18 Unit mm

10

(a)

(b)

Fig 9 Overall experimental setup (a) Data acquisition (DAQ) system and (b) Installed circu-lar PZT transducer on the specimen with its dimension

513 Impact test

The impact-induced damage was introduced to the specimen using the impact test equip-

ment presented in Fig 10 A 5kg mass with 10mm tip was dropped from 015m height onto

the same location of the specimen The signals were obtained at each number of impacts

Since the reflected wave is not detected at the first impact case the data obtained from the

second and third impact were utilized for the analysis

10

Unit mm (a)

(b)

Fig 10 (a) Impact test equipment for introducing impact-induced damage and (b) Tip of the impact tester with 10mm diameter

- 23 -

514 Temperature effect test

In reality structures are subjected to changing environmental conditions such as tempera-

ture that affect measured signal and cause false-alarm Thus the robustness of the proposed

damage characterization technique under temperature variation in investigated (Fig 11) The

specimens were placed inside a temperature chamber and a thermocouple was installed on

the specimen to measure temperature of surface of the specimen The signal was obtained

under three temperature cases (0 20 and 50degC) and the humidity was kept at 30

Temperature Chamber

Test Specimen

Fig 11 Test specimen in temperature chamber

- 24 -

52 Teflon inserted damage

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Reflected wave from damage

(a)

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Transmitted WaveReflected Wave

Detected reflected wave

(b)

Fig 12 A0 Lamb wave mode obtained using pitch-catch method from the Teflon inserted damage specimen (a) The reflected wave from damage is shown (b) The Teflon inserted damage is detected by capturing the reflected wave

Fig 12 shows the obtained pitch-catch signal from the Teflon inserted damage specimen

and damage diagnosis result The Teflon inserted damage is detected successfully by captur-

ing the reflected wave Table IV presents the result of the Teflon inserted damage characteri-

zation 119907119907119894119894 is measured as 1213ms and the variation is assumed to plusmn 5 of the normal val-

ue (1152~1273ms) The results indicate that the proposed damage characterization approach

works well in real application

TABLE IV Damage characterization for the Teflon inserted damage specimen

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

Exact 635 25

Method 1 123 60~75 73~338 298~1220

Method 2 123 619~662 229~310 934~1188

Method 3 123 631 284 1161

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1213 ms

- 25 -

53 Impact-induced damage

11 115 12 125 13-4

-2

0

2

4

Time (ms)

Vol

tage

(mV

)

IntactImpact2Impact3

Reflected wave from damage

(a) Pitch-catch signal

11 115 12 125 13 135

-05

0

05

1

Time (ms)

Volta

ge (m

V)

IntactImpact2Impact3

Boundary reflection

Reflected wave from damage

(b) Pulse-echo signal

Fig 13 The signal obtained from the test specimen subjected to impact The impact was in-troduced at the same location of the specimen and the signal was obtained after each impacts As the number of impact is increased the transmitted A0 Lamb wave mode is delayed and attenuated

The obtained signals from impact-induced damage specimen are presented in Fig 13 It is

observed that as the number of impact is increased the transmitted A0 Lamb wave mode is

more delayed because the wave velocity in the damage area is reduced Fig 14 shows the

false-alarm test result whether the proposed damage detection algorithm works even in pris-

tine condition of the test specimen It is shown that no false alarm is occurred because no ad-

ditional reflected wave is captured

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Transmitted Wave

(a) Transmitted wave

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Reflected Wave

No reflected wave

(b) Reflected wave

Fig 14 False alarm test for the proposed damage detection algorithm in pristine condition of the specimen Reflected wave from damage is not captured false alarm is not occurred

- 26 -

TABLE V Damage characterization for the impact-induced damage specimen

of Impacts m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

2

(1470 J)

Method 1 080 65~90 75~246 470~1157

Method 2 080 801~885 114~245 711~1134

Method 3 080 843 179 1122

3

(2205 J)

Method 1 148 50~100 06~380 21~1134

Method 2 148 758~838 249~368 843~1125

Method 3 148 798 308 1044

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1213 ms

Table V presents the characterization results of the impact-induced damage case Here the

measured 119907119907119894119894 and the assumed variation are same as the artificial damage case It is observed

that as the number of impact is increased 119889119889 is increased while 119907119907119889119889 is reduced Thus it is

validated that the proposed damage characterization approach estimates the increment of the

damage well

54 Temperature effect test

541 Teflon inserted damage

11 115 12 125 13-10

-5

0

5

10

Time (ms)

Vol

tage

(mV

)

0 oC20 oC50 oC

(a) Pitch-catch signal

11 115 12 125 13 135

-1

0

1

2

3

Time(ms)

Volta

ge (m

V)

0 oC20 oC50 oC

(b) Pulse-echo signal

Fig 15 Signals obtained from the Teflon inserted damage specimen under temperature varia-tion Time delay and amplitude variation are occurred

Fig 15 shows the obtained signals from the Teflon inserted damage specimen subjected to

- 27 -

temperature variation Time delay and amplitude variations are occurred there is no addition-

al damage is introduced The analyzed results of the obtained signals are reported in Table VI

The measured 119907119907119894119894 is 1247 ms at 0 degC 1213 ms at 20 degC and 1168 ms at 50 degC respectively

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms (1152~1273 ms)

The estimated results are also changed under temperature variations the fluctuation of esti-

mated range and exact value is small

TABLE VI Temperature effect test for the Teflon inserted damage specimen

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

0 degC

Exact 635 25

Method 1 113 60~70 79~268 352~1291

Method 2 113 560~619 233~344 1034~1172

Method 3 113 606 257 1141

20 degC

Exact 635 25

Method 1 123 60~75 73~338 298~1220

Method 2 123 619~662 229~310 934~1188

Method 3 123 631 284 1161

50 degC

Exact 635 25

Method 1 137 60~75 72~328 263~1228

Method 2 137 622~687 183~299 666~1086

Method 3 137 642 264 960

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1247 ms at 0 degC 1213 ms at 20 degC and 1168 ms at 50 degC

- 28 -

542 Impact-induced damage

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Vol

tage

(mV

)

Transmitted WaveReflected Wave

No reflected wave

(a) 0 degC

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Transmitted WaveReflected Wave

No reflected wave

(b) 50 degC

Fig 16 False alarm test under temperature variations It is shown that proposed damage char-acterization algorithm is robust under temperature variations

The false alarm test in pristine condition of the impact-induced damage specimen under

temperature variation was investigated (Fig 16) It is validated that the proposed damage di-

agnosis algorithm is robust under temperature variations Fig 17 shows the obtained signals

from the impact-induced damage (impacted 3 times) specimen under temperature variation

The analyzed results of the obtained signals are reported in Table VII The measured 119907119907119894119894 in

each temperature and the assumed variation are same as the artificial damage case

11 115 12 125 13-5

0

5

Time (ms)

Volta

ge (m

V)

0 oC20 oC50 oC

(a) Pitch-catch signal

11 115 12 125 13 135

-1

0

1

2

3

Time(ms)

Vol

tage

(mV

)

0 oC20 oC50 oC

(b) Pulse-echo signal

Fig 17 Signals obtained from impact-induced damage specimen (Impact 3) under tempera-ture variation

- 29 -

TABLE VII Temperature effect test for the impact induced damage specimen (Impact 3 case)

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

0 degC

Method 1 135 75~90 72~310 267~1148

Method 2 135 705~779 264~381 978~1126

Method 3 135 763 289 1071

20 degC

Method 1 148 50~100 06~380 21~1134

Method 2 148 758~838 249~368 843~1125

Method 3 148 798 308 1044

50 degC

Method 1 153 75~100 53~415 174~1227

Method 2 153 804~889 214~336 702~1103

Method 3 153 815 320 1050

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1247 ms at 0 degC 1213 ms at 20 degC and 1168 ms at 50 degC

- 30 -

CHAPTER 6 CONCLUDING REMARKS

61 Executive summary

In this study the reference-free damage detection localization and quantification tech-

nique in a multi layered composite plate is proposed Additional reflected A0 Lamb wave

mode from the damage due to multiple reflections in damage area is used as a feature for

characterizing the damage First damage is characterized in terms of a damage index (m-

value) defined as the ratio of the physical length of the damage to the wavelength of the A0

Lamb wave mode Then the range of possible damage location length and severity are esti-

mated with constraints This procedure can be accomplished without prior baseline data To

realize the proposed technique the matching pursuit technique is modified for reduction of

the error by considering the dispersive characteristic of Lamb wave The matching pursuit

method estimates the arrival time of each wave packets accurately even two wave packets are

overlapped The numerical simulations and experimental tests including Teflon tape inserted

and impact-induced damage are investigated for the feasibility of the proposed technique

Moreover the effectiveness of the proposed technique under changing temperature is exper-

imentally verified

- 31 -

62 Suggestion for further study

The limitations of this study and the possible improvement and further studies are listed as

follow

1 Measurement of impact-induced damage length using conventional NDT technique

Additional experimental verification for impact-induced damage characterization is

needed by measuring the real length of damage using conventional NDT technique

such as ultrasonic scan or infrared camera

2 Limitation of both side collocated PZT installation In this study excitation PZTs are

installed both side and collocated for decomposing A0 Lamb wave mode However it

can produce decomposition error due to imperfection of PZT location and their installa-

tion Moreover in many cases assess to both surface of structures is not allowed such

as aircraft applications To improve this limitation a reference-free damage characteri-

zation using single side PZTs will be considered

3 Complex structure application In this study damage is characterized in simple plate-

like structure When the monitored structure becomes complex the response signals al-

so becomes complex due to its geometry and boundary Thus the investigations for

complex structure application such as stiffener spar and curvature are needed

- 32 -

REFERENCES

[1] Cawley P and Alleyne D (1996) ldquoThe use of Lamb waves for the long range inspection of large structuresrdquo Ultrasonics 34 pp287-290

[2] Ghosh T Kundu T and Karpur P (1998) ldquoEfficient use of Lamb waves for detecting defects in large platesrdquo Ultrasonics 36 pp791-801

[3] Guirguitui V (2003) ldquoLamb wave generation with piezoelectric wafer active sensors for structural health monitoringrdquo SPIE Conference on Smart Structures and Materials San Diego CA 5391 pp111-422

[4] Ragvahan A and Cesnik CE S (2004) ldquoModeling of piezoelectric-based Lamb wave generation and sensing for structural health monitoringrdquo SPIE Conference on Smart Struc-tures and Materials San Diego CA 5391 pp419-430

[5] Rose JL Pilarski A and Dirit JJ (1993) ldquoAn approach to guided wave mode selection for inspection of laminated platerdquo Journal of Reinforced Plastics and Composites 12(5) pp536-544

[6] Guo N and Cawley P (1993) ldquoThe interaction of Lamb waves with delaminations in composite laminatesrdquo Journal of Acoustical Society of America 94(4) pp 2240-2246

[7] Valdez SHD and Soutis C (2002) ldquoReal-time nondestructive evaluation of fiber com-posite laminates using low-frequency Lamb wavesrdquo Journal of Acoustical Society of Amer-ica 111(5) pp 2026-2033

[8] Yang M and Qiao P (2005) ldquoModeling and experimental detection of damage in virous materials using the pulse-echo method and piezoelectric sensoractuatorsrdquo Smart Materials and Structures 14 pp1083-1100

[9] Lestari W and Qiao P (2005) ldquoApplication of wave propagation analysis for damage identification in composite laminated beamsrdquo Journal of Composite Materials 39(22) pp1967-1984

[10] Kessler S S Spearing S M and Soutis C (2002) ldquoDamage detection in composite ma-terials using Lamb wave methodsrdquo Smart Materials and Structures 11 pp269-278

[11] Sohn H Park G Wail J R Limback N P and Farrar C R (2004) ldquoWavelet-based active sensing for delamination detection in composite structuresrdquo Smart Materials and Structures 13 pp 153-160

[12] Petculescu G Krishnaswamy S and Achenback J D (2008) ldquoGroup delay measure-ments using modally selective Lamb wave transducers for detection and sizing of delamina-tions in compositesrdquo Smart Materials and Structures 17

[13] Sohn H Park H W Law K H and Farrar C R (2007) ldquoDamage detection in compo-

- 33 -

site plates by using an enhanced time reversal methodrdquo Journal of Aerospace Engineering (ASCE) pp 141-151

[14] Sohn H (2007) ldquoEffects of environmental and operational variability on structural health monitoringrdquo Philosophical Transactions of the Royal Society Ardquo 365 pp539-560

[15] Rytter A (1993) ldquoVibration based inspection of civil engineering structuresrdquo PhD The-sis Department of Building Technology and Structural Engineering University of Aalborg Aalborg Denmark

[16] Hayashi T and Kawashima K (2002) ldquoMultiple reflections of Lamb waves at a delami-nationrdquo Ultrasonics 40 pp193-197

[17] Wang C H and Rose L R F (2003) ldquoWave reflection and transmission in beams con-taining delamination and inhomogeneityrdquo Journal of Sound and Vibration 264 pp 851-872

[18] Ramadas C Balasubramaniam K Joshi M and Krishnamurthy C V (2009) ldquoInterac-tion of the primary anti-symmetric Lamb mode (Ao) with symmetric delaminations numer-ical and experimental studiesrdquo Smart Materials and Structures 18

[19] Ramadas C Balasubramaniam K Joshi M and Krishnamurthy C V (2010) ldquoInterac-tion of guided Lamb waves with an asymmetrically located delamination in a laminated composite platerdquo Smart Materials and Structures 19

[20] Wilcox P Lowe M and Cawely P (2001) ldquoThe effect of dispersion on long-range in-spection using ultrasonic guided wavesrdquo NDTampE International 34 pp1-9

[21] Mallat S G and Zhang Z (1993) ldquoMatching pursuits with time-frequency dictionariesrdquo IEEE Tansactions on Signal Processing 41(12) pp3397-3415

[22] Zhang G Zhang S and Wang Y (2000) ldquoApplication of adaptive time-frequency de-composition in ultrasonic NDE of highly scattering materialsrdquo Ultrasonics 38 pp 961-964

[23] Hong J C Sun K H and Kim Y Y (2005) ldquoThe matching pursuit approach based on the modulated Gaussian pulse for efficient guided-wave damage inspectionrdquo Smart Mate-rials and Structures 14 pp548-560

[24] Herakovich C T (1997) Mechanics of Fibrous Composites Wiley 480 pages

[25] Reddy J N (1985) ldquoA review of the literature on finite-element modeling of laminated composite platesrdquo The Shock and Vibration Digest 17(4) pp 3-8

[26] Reddy J N (2003) Mechanics of Laminated Composite Plates and Shells Theory and Analysis CRC Press 856 pages

[27] Palacz M Krawczuk M and Ostachowicz W (2005) ldquoThe spectral finite element mod-el for analysis of flexure-shear coupled wave propagation Part 1 laminated multilayer

- 34 -

composite beamrdquo Composite Structures 68 pp37-44

[28] Zhang Y C and Harding J (1990) ldquoA numerical micromechanics analysis of the me-chanical properties of a plain weave compositerdquo Computer amp Structures 36(5) pp 839-844

[29] Nemat-Nasser S and Hori M (1999) Micromechanics Overall Properties of Heteroge-neous Materials Elsevier 810 pages

[30] Scida D Aboura Z Benzeggagh M L and Bocherens E (1999) ldquoA micromechanics model for 3D elasticity and failure of woven-fibre composite materialrdquo Composites Science and Technology 59 pp 505-517

[31] Thomas Y and Wu X H (1997) ldquoA multiscale finite element method for elliptic prob-lems in composite materials and porous mediardquo Journal of Computational Physics 134 pp 169-189

[32] Lee K H Moorthy S and Ghosh S (1999) ldquoMultiple scale computational model for damage in composite materialsrdquo Computer Methods in Applied Mechanics and Engineer-ing 172 pp 175-201

[33] Namilae S and Chandra N (2005) ldquoMultiscale model to study the effect of interfaces in carbon nanotube-based compositesrdquo Journal of Engineering Materials and Technology (ASME) 127 pp222-232

[34] Adams D E (2007) Health monitoring of Structural Materials and Components Wiley 460 pages

[35] Gopalakrishnan S (2009) ldquoModeling Aspects in Finite Elementrdquo Encyclopedia of Struc-tural Health Monitoring 2 pp791-809

[36] Palacz M Krawczuk M and Ostachowicz W (2005) ldquoThe spectral finite element mod-el for analysis of flexure-shear coupled wave propagation Part 2 Deminated multilayer composite beamrdquo Composite Structures 68 pp45-51

[37] Baik J M and Thopson R B (1984) ldquoUltrasonic scattering from imperfect interfaces A quasi-static modelrdquo Journal of Nondestructive Evaluation 4(3) pp177-196

[38] Luo H and Hanagud S (2000) ldquoDynamics of delamincated beamsrdquo International Jour-nal of Solids and Structures 37 pp1501-1519

[39] Kim S B and Sohn H (2007) ldquoInstantaneous reference-free crack detection based on polarization characteristics of piezoelectric materialsrdquo Smart Materials and Structures 16 pp2375-2387

[40] An Y K and Sohn H (2010) ldquoInstantaneous crack detection under varying temperature and static loading conditionsrdquo The Special Issue of Structural Control and Health Monitor-ing 17(7) pp 730-741

- 35 -

[41] Lee S J and Sohn H (2006) ldquoActive self-sensing scheme development for structural health monitoringrdquo Smart Materials and Structures 15(6) pp 1734-1746

- 36 -

요 약 문

복합재료 손상의 무기저 진단 및 정량화

복합재료는 재료의 경량성 내구성 고강도의 특성으로 인하여 항공기에서부터

교량에까지 널리 사용되고 있다 하지만 외부의 충격으로 인한 복합재료 내부의

손상은 재료의 강도를 감소시켜 구조물의 안정성을 저하시키는 요인이 된다 하지만

복합재료의 손상은 재료 내부에서 발생하여 육안으로는 식별하기도 힘들며 표면에서

보이는 것만으로 내부의 손상 정도를 판단하기는 불가능하다는 단점을 가지고 있다

본 연구에서는 손상 내부에서 반사로 인해 발생하는 파를 이용하여 기저자료를

사용하지 않고 특정 시점에서 얻은 자료만으로 복합재료의 손상을 진단하고 정량화

할 수 있는 기법을 제안한다 제안된 기법의 적용성을 검증하기 위하여 유한요소법을

이용한 수치해석을 수행하였고 Teflon tape 이 삽입된 시편과 실제 충격하중으로

인한 손상을 입은 시편을 이용한 실험을 진행하였다 수치해석 및 실험 결과 제안된

기법이 복합재료의 손상을 성공적으로 진단 및 정량화할 수 있음을 확인하였다 또한

다양한 외부 온도 환경에서 제안된 기법의 환경영향 실험을 수행하였고 다양한

환경에서 손상의 진단 및 정량화가 성공적으로 이루어짐을 검증하였다

핵심어 무기저 손상진단 및 정량화 압전센서 복합재료 유한요소법 Matching pursuit

method

- 37 -

감 사 의 글

기대와 두려움을 안고 시작한 석사과정이 어느덧 2 년 이라는 세월 속에서 추억으로 자리잡아

가고 있습니다 학부생 기간을 포함한 지난 3 년 동안 SHM 이라는 새로운 분야를 접하고 연구를

수행 할 수 있도록 도와주신 고마운 분들이 있어 그 감사를 표하고자 합니다

우선 저의 많은 부족한 점을 일깨워 주시고 연구에 대한 뜻을 더욱 확고히 세울 수 있게 해

주신 지도교수님이신 손 훈 교수님께 깊은 감사의 뜻을 전합니다 짧은 기간이었지만 교수님의

도움으로 많은 것을 배우고 느끼고 깨달을 수 있었습니다 또한 본 연구를 수행할 수 있도록 지

원해주신 국토해양부 (U-City 석 박사과정 지원사업)에 깊은 감사 드립니다

때로 힘들고 지칠 때 항상 저를 믿고 응원해 주시는 가족이 있었기에 지금의 글을 쓸 수 있었

습니다 또한 지금 연구실에서 함께 생활하고 있는 윤규형 민구형 철민이형 현석이형 진열이형

호민이형 병진이 선혜누나와 성대에 계신 창길이형 미국에 계신 의재형과 Debaditiya 에게도 감

사의 말씀을 전합니다 그리고 모두 나열할 수는 없지만 카이스트 생활의 활력소가 되어준 카이

스트 건설 및 환경공학과 선후배들 직원 분들과 SOS 멤버들에게도 감사를 표합니다

앞서 거론한 많은 이들의 도움이 없었다면 저 또한 지금 이 자리에 있을 수 없었을 것입니다

많이 부족하기에 더 노력하고 배우고자 했던 초심을 다시 한번 생각하며 더 큰 결실을 맺기 위해

앞으로 나아가겠습니다 도움을 주신 모든 분들께 이 감사의 글을 바칩니다

2011년 6월 카이스트에서 임형진 올림

- 38 -

CURRICULUM VITAE

Hyung Jin Lim

Research Assistant

The Civil and Environmental Engineering Department

Korea Advanced Institute of Science and Technology (KAIST)

291 Daehak-Ro Guseong-Dong Yuseong-Gu

Daejeon Republic of Korea 305-701

Phone (82)+42-350-3665 Fax (82)+42-350-3610

Email limhj87gmailcom

RESEARCH INTERESTS

Damage Detection Localization and Quantification on Composite and Metallic Structures Pattern Recognition Finite Element Modeling (FEM)

EDUCATION

2009 ndash 2011 MS Civil and Environmental Eng Korea Advanced Institute of Science and Technology (KAIST) Korea Rep

2006 ndash 2009 BS Civil and Environmental Eng Korea Advanced Institute of Science and Technology (KAIST) Korea Rep

JOURNAL PUBLICATIONS The corresponding author is underlined 1 Hyung Jin Lim Hoon Sohn ldquoReference-free damage detection localization and quantification

using in compositerdquo in preparation for Journal of the Acoustical Society of America 2011

2 Chul Min Yeum Hoon Sohn Jeong Bum Ihn Hyung Jin Lim ldquoInstantaneous delamination detection in a composite using a dual piezoelectric transducer networkrdquo Submitted to Compo-sites Science and Technology 2011

3 Yun-Kyu An Hyung Jin Lim Min Koo Kim Jin Yeol Yang Hoon Sohn Chang Guen Lee ldquoInvestigation of applicability of local reference-free damage detection techniques to in-situ

- 39 -

bridgesrdquo Submitted to ASCE Journal of Structural Engineering 2011

4 Hyung Jin Lim Min Koo Kim Hoon Sohn Chan Yik Park ldquoImpedance based damage detec-tion under varying temperature and loading conditionsrdquo Submitted to NDTampE International 2011

5 Hyung Jin Lim Hoon Sohn ldquoQuantitative estimation of transmitted and reflected lamb waves at discontinuityrdquo Journal of the Korean Society for Nondestructive Testing Vol30 No 4 pp1-8 2010

CONFERENCE PROCEEDINGS

The corresponding author is underlined 1 Hyung Jin Lim and Hoon Sohn ldquoReference-free damage detection localization and quantifi-

cation in compositesrdquo ASCE Engineering Mechanics Institute Conference Boston June 2-4 2011

2 Hyung Jin Lim and Hoon Sohn ldquoReference-free delamination quantification within multi-layer composite materialsrdquo KSNT Spring Conference May 26-27 Gyeong-Ju Korea 2011

3 Min Koo Kim Hyun Jin Lim Hoon Sohn Chan Yik Park ldquoImpedance-based bolt loosening detection under varying temperature and loadingrdquo Asian Pacific Workshop on Structural Health Monitoring Tokyo Japan November 30-December 2 2010

4 Hyung Jin Lim Chul Min Yeum and Hoon Sohn ldquoModeling of Impact-induced Delamination in a Multilayer Composite Platerdquo The 23rd KKCNN Symposium on Civil Engineering Taipei (Taiwan) November 13-15 2010

5 Yun-Kyu An Hyung Jin Lim and Hoon Sohn ldquoTemperature independent Delamination Detec-tion using data normalizationrdquo The 3rd International Conference on Multi-Functional Materi-als and Structures Chonbuk National Univ Jeonju Korea September 14-18 2010

6 Hoon Sohn Yun-Kyu An Chang Gil Lee Min Koo Kim Hyung Jin Lim ldquoApplication of online NDT techniques to decommissioned bridge testingrdquo the 5th European Workshop on Structural Health Monitoring Sorrento Italy June 29-July 02 2010

7 Min Koo Kim Hyung Jin Lim Hoon Sohn Chan Yik Park ldquoImpedance based damage diag-nosis of a complex composite aircraft wing under changing loading and temperature condi-tionsrdquo the 5th European Workshop on Structural Health Monitoring Sorrento Italy June 29-July 02 2010

8 Yun-Kyu An Min Koo Kim Hyung Jin Lim and Hoon Sohn ldquoValidations of reference-free crack detection techniques through a decommissioned bridge testrdquo 2010 COSEIK Annual Con-ference Jeju Korea April 8-9 2010

- 40 -

PARTICIPANT PROJECTS

1 Development of Smart Steel Members for Online Steel Structure Monitoring POSCO Corporation (Funded 41000000 KRW (41000 USD) for 010511 to 123111)

2 Development of Enabling Technologies for Structural Health Monitoring Korean Agency for Defense Development (Funded 240000000 KRW (240000 USD) out of total 3200000000 KRW (3200000 USD) for 060108-053111)

3 Development of On-board SHM Technologies for Composite Air Vehicles The Boeing Company (Funded 400000 USD (400000000 KRW) for 080108 to 073111)

EDUCATIONAL EXPERIENCES

Fall 2009

Teaching Assistant CE207 ldquoElementary Structural Engineering and Laboratoryrdquo KAIST 3 unit undergraduate course

EXPERIMENTAL EXPERIENTCES

2010 Monitoring of Real Aircraft Composite Wing Segment Korea Agency for Defense De-velopment (ADD) Daejeon Korea Real time monitoring of composite air vehicles fund-ed by Korea Agency for Defense Development (ADD)

2009-2011

Field Monitoring of Real Bridge Structures Gimpo (Ramp G bridge) amp Icheon (Sam-seung bridge) Korea US-Korea collaborative research for bridge monitoring testbeds

HONORS amp AWARDS

Scholarships for outstanding students granted KAIST (2007-2009) The three prominent students in the civil and environmental engineering department of KAIST were selected as recipients This scholarship provided 2000$ per year for three years

- 41 -

Page 11: 복합재료 손상의 무기저 진단 및 정량화 - KAISTssslab.kaist.ac.kr/article/pdf/Thesis_2.pdf · 1 Declaration of Ethical Conduct in Research: I, as a graduate student

age characterization algorithm is robust under temperature variationsmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 29

17 Signals obtained from impact-induced damage specimen (Impact 3) under

temperature variation middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot 29

vii

CHAPTER 1 INTRODUCTION

11 Motivation

Composite materials has been increasingly demands in modern structures ranging from the

latest aircraft to retrofitted bridge deck due to its light weight superior specific strength and

stiffness properties However the occurrence of damage due to abrupt impact or sequential

accumulation of defect may considerably degrade structural strength or stiffness and reduce

the safety and reliability of the system Recently Lamb waves based damage detection tech-

niques using piezoelectric (PZT) transducers have emerged as one of the promising structural

health monitoring (SHM) technology for detecting hidden damage because of the possibility

of long range inspection and high sensitivity [1-4]

12 Literature review

This section will discuss researches that have been previously carried out on topics of

guided wave based SHM techniques using PZT transducers for damage in composite materi-

als The early investigations of interaction between Lamb wave and damage in composite can

be found in Rose et al and Guo et al [5-6] From further studies reflected signals from dam-

age were shown to be an effective method for damage diagnosis and localization using PZTs

[7-9] The reflected signals from damage were collected using one PZT as an actuator and the

other PZT as a sensor However the techniques cannot be used to determine the extent of

- 1 -

damage with reasonable precision

Kesseler et al proposed a technique monitoring the wavelet coefficients of transmitted ul-

trasonic signal amplitude as a potential way to damage detection [10] One drawback of this

technique is that the variability in coupling and the sensitivity of installed PZT transducers

can induce changes in the amplitude and energy of the signal that are cause false-alarm Oth-

er wavelet based delamination detection and localization technique developed using PZT

sensor network and outlier analysis [11]

Damage detection and quantification in composite plate was also investigated by group de-

lay measurement due to damage [12] However it is assumed that the group velocity in dam-

age area as the group velocity propagates half thickness of the structure in dispersion curve in

the quantification procedure Thus the damage quantification was investigated only Teflon

inserted damage case

Sohn et al proposed a damage detection technique which does not rely on any past base-

line signals to assess damage in composite plates by using an enhanced time reversal method

[13] The damage is assessed using the violation of time reversibility when nonlinearity is

caused by the damage along a direct wave path Thus it allows instantaneous identification

of delamination without requiring the baseline signal for comparison

- 2 -

13 Objective and uniqueness

In this study a reference-free damage characterization technique including detection lo-

calization and quantification of impact-induced damage in composite materials is developed

The general damage characterization can be categorized into the following levels [14]

Level 0 Define damage

Level 1 Determine the presence of damage within the structure

Level 2 Locate the regions of damage

Level 3 Quantify the damage

Level 4 Predict the remaining service life of the structure

Through the proposed technique from level 1 to level 3 can be achieved

Most of common approaches of damage characterization techniques described in the pre-

vious section use prior obtained baseline data However these approaches can cause false

alarm due to the effect of environmental variations such as temperature and ambient vibra-

tions [15]

The uniqueness of this study is

1 Damage is characterized without using prior measured baseline data

2 Robustness of the proposed technique under temperature variation is investigated

3 Additional waves reflected from inside damage have been used in addition to transmit-

ted and reflected waves

- 3 -

14 Organization

The organization of this thesis is summarized as follows

Chaper1 A general introduction including motivation literature review objective and

uniqueness of this study is described

Chaper2 Interactions between Lamb wave mode and damage in composites such as time

delay attenuation and multiple reflections are summarized Damage index (m-value) for

damage characterization is defined

Chaper3 Algorithms for the reference-free damage characterization technique using m-

value are developed Modified matching pursuit for this study is also described

Chaper4 Verification of the algorithm using numerical simulation is presented Various

modeled damage cases are characterized

Chaper5 Experimental verifications are described Two damage cases (Teflon inserted and

impact-induced damage) are investigated for the feasibility of the proposed damage charac-

terization technique Temperature effect test using temperature chamber is also carried out

Chaper6 Conclusion and summary of the present study is reported In advance suggestions

and future work recommendations are represented

- 4 -

CHAPTER 2 INTERACTION OF LAMB WAVE WITH DAMAGE

21 A0 Lamb wave mode delay and attenuation in the time domain

Time delay and attenuation of the A0 mode Lamb wave after passing through the damage

area due to effective thickness reduction were reported in previous researches [812] Moreo-

ver the attenuation of the A0 Lamb wave mode is occurred by degrading of shear modulus of

in the damage area However the time delay and amplitude variation of the A0 Lamb wave

mode also occurred even when the structure is health condition by the material property

changing due to temperature variation of the structure (Fig 1) Thus conventional damage

characterization methods using the time delay and attenuation of A0 Lamb wave mode as a

feature can cause false-alarm in real application

11 115 12 125 13-10

-5

0

5

10

Time (ms)

Vol

tage

(mV

)

IntactDamage

Reflected wave from Damage

(a)

11 115 12 125 13-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

0 oC20 oC50 oC

(b)

Fig 1 Lamb wave propagation characteristics (a) A0 Lamb wave mode time delay and atten-uation after passing through a damage (b) The time delay and amplitude variation of A0 Lamb wave mode under temperature variation in pristine condition of structure Because of this conventional damage detection techniques in composites have limitations for real struc-ture application

- 5 -

22 Multiple reflections in the damage area

Standing wave patterns due to multiple reflections of A0 Lamb wave mode in the damage

area were observed through numerical simulations and ultrasonic wave-field imaging using a

scanning laser vibrometer [16] The formation of standing waves in damage is illustrated

through Fig 2 When the incident waves enter the damage area a significant portion of the

waves are reflected back from the exit Then the reflected waves travel through the damage

and undergo reflection again at the original entrance As a consequence of multiple reflec-

tions at the entrance and the exit a considerable amount of ultrasonic energy is trapped inside

the damage area [17-19]

Multiple Reflections

Incident Wave Transmitted Wave

Damage

Fig 2 Multiple reflections in the damage area The multiple reflections of A0 Lamb waves in the damage area are utilize as a feature for characterization of the damage in this study

- 6 -

CHAPTER 3 THEORETICAL DEVELOPMENT

31 Overview

Damage

l = l1 + d + l2

d = vdtd l1 = vit1 l2 = vit2

t = t1 + td + t2

Excitation and Sensing PZT Sensing PZT

l

Transmitted Wave

Reflected Waves

Incident Wave

Fig 3 Parameters and notations of the proposed damage characterization technique Damage is detected by identifying the additional reflected wave from the damage due to multiple re-flections (dashed lines) Localization and quantification of the damage is carried out by measuring the arrival time of transmitted and reflected waves from the damage

In this section the reference-free damage characterization algorithm is proposed Damage

diagnosis and quantification are achieved by identifying the additional reflected waves due to

multiple reflections in the damage area (dashed lines in Fig 3) Moreover damage is local-

ized using additional information from pulse-echo method implement For the localization

and the quantification of damage the variables such as the path length and A0 Lamb wave

mode velocity in damage area are defined (Fig 3) For example 119897119897 is expressed as

119897119897 = 1198971198971 + 119889119889 + 1198971198972 (1)

where 1198971198971 119889119889 and 1198971198972 are the distance from the excitation PZTs to the entrance of damage the

damage length and the distance from the exit of damage to the sensing PZT respectively

- 7 -

Moreover 119907119907119894119894 and 119907119907119889119889 are also defined as the A0 Lamb wave mode velocity in pristine area

and damage area respectively Note that the whole procedure of the damage characterization

is accomplished based solely on the signals obtained from the current state of the structure

32 Damage index (m-value)

The proposed reference-free damage characterization technique utilizes the waves reflect-

ed additionally from the damage area as a feature The arrival time difference of the transmit-

ted and reflected A0 Lamb wave modes reflected from the exit and the entrance of the delam-

ination area one times each (∆119905119905) contains the information of the physical length of the dam-

age (119889119889) and A0 Lamb wave mode velocity in the damage area (119907119907119889119889) The wave velocity in the

damage area which is related to the shear modulus (G) represents the severity of the damage

∆119905119905 = 2119905119905119889119889 =2119889119889119907119907119889119889

(2)

However based on the measured data available the 119889119889 and 119907119907119889119889 (or G) cannot be uniquely

determined Fig 4 shows two FEM models which have different physical length and wave

velocity (G) of the damage Detail explanation of delamination modeling technique will be

covered in section 4 The similar signal in the time domain and ∆119905119905 are obtained from the

two different damage cases

- 8 -

Damage 1

Damage 2

20

Unit mm15

90 G value reduction

943 G value reduction

m-value = 248

m-value = 248

015 02 025 03 035

-2

-1

0

1

2

Time (ms)

Ampl

itude

Damage 1Damage 2

Fig 4 Two different FEM models which have different 119889119889 and 119907119907119889119889 (or G) This represents that the damage cannot be characterized uniquely based on the measured data available

Thus in this study the damage index (m-value) which can be uniquely defined based on

the available data is introduced The m-value is defined as the ratio of the physical length of

the damage and wavelength of the A0 Lamb wave mode in the damage area (120582120582119889119889)

119898119898 =119889119889120582120582119889119889

=119889119889119889119889119907119907119889119889

(3)

In many cases the transmitted and reflected waves are overlapped and it is hard to decom-

pose and calculate the arrival times of them Matching pursuit method is an alternative way

to accurately decompose of the overlapped waves and estimate the parameters of each wave

such as arrival time (Fig 5) Detail explanation of the matching pursuit method is covered in

section 36

015 02 025 03-2

-1

0

1

2

Time (ms)

Am

plitu

de

OriginalApproximated

(a)

015 02 025 03-2

-1

0

1

2

Time (ms)

Ampl

itude

Wave 1Wave 2

(b)

Fig 5 (a) Example of the overlapped signal with comparing approximated signal using the matching pursuit method (b) Decomposed signal using the matching pursuit method The matching pursuit method provides a conventional way to decompose and identify each wave

- 9 -

33 Damage detection algorithm (Level 1)

The damage diagnosis is accomplished by extracting the additional reflected A0 Lamb

wave modes from damage due to multiple reflections If A0 Lamb wave modes are decom-

posed perfectly it is trivial that one transmitted A0 Lamb wave mode is always captured in

the response signal If damage exists other waves in the response signal behind the transmit-

ted A0 Lamb wave mode are presented Since this approach does not rely on previously col-

lected data the diagnosis is independent from environmental variation such as temperature

changes

It is reported that the dispersion of Lamb waves cause wave-packets to spread out in space

and time as they propagate through a structure [20] Thus the reflected waves from the dam-

age have longer pulse duration than the transmitted wave due to their longer propagating dis-

tance In this study the matching pursuit method is modified to capture more spread wave

than the transmitted A0 Lamb wave mode Detail explanation of the modified matching pur-

suit method is described in latter section

Let 1199041199041 and 1199041199042 as the pulse duration (scale) of the transmitted and additional reflected A0

Lamb wave mode respectively

ldquoIf 1199041199042 gt 1199041199041 damage exists

Otherwise there is no damagerdquo

- 10 -

34 Damage quantification algorithm (Level 3)

As mentioned in the previous section damage is quantized by estimating 119889119889 and 119907119907119889119889 using

pitch-catch method The procedure to estimate 119889119889 and 119907119907119889119889 is

(i) Measure 119897119897 119905119905 and 119905119905119889119889

(ii) Assume the range of 119889119889 (0 lt 119889119889 lt 119897119897)

(iii) Calculate 119907119907119894119894 and 119907119907119889119889 with a constraints (0 lt 119907119907119889119889 lt 119907119907119894119894) (Eq 4 and 5)

(iv) Estimate m-value and find a feasible range of 119889119889 and 119907119907119889119889

119907119907119894119894 =119897119897 minus 119889119889119905119905 minus 119905119905119889119889

(4)

119907119907119889119889 =119889119889119905119905119889119889

(5)

35 Damage localization algorithm (Level 2amp3)

For damage localization 1198971198971 is estimated by measuring the arrival time from the excitation

PZT to the entrance of delamination (1199051199051) with additional pulse-echo method implementation

The procedure is similar with the previous damage quantification algorithm

(i) Measure 119897119897 119905119905 119905119905119889119889 and 1199051199051

(ii) Assume the range of 1198971198971 (0 lt 1198971198971 lt 119897119897)

(ii) Compute possible combinations of 1198971198972 119889119889 and 119907119907119889119889 with constraints

(0 lt 119907119907119889119889 lt 119907119907119894119894 119889119889 gt 0) (Eq 5 6 and 7)

- 11 -

1198971198972 = 1198971198971119905119905 minus 1199051199051 minus 119905119905119889119889

1199051199051 (6)

119889119889 = 119897119897 minus 1198971198971 minus 1198971198972 (7)

Since a PZT for pulse-echo method acts as both an actuator and a sensor the obtained sig-

nal contains both the excitation and the response signal Thus the reflected wave from the

entrance of damage can be buried into the tail of excitation signal and cause error on arrival

time of reflected wave from damage Even in this case damage can be localized and quan-

tized using the constraints When first wave packet reflected from the entrance of damage is

buried into the excitation signal the next wave packet reflected from the exit of damage can

be utilized alternatively If the algorithm regards the second reflected wave as the first re-

flected wave there are no combinations of estimated variable meeting the constraints Then

the algorithm repeats the process by regarding the wave as the second reflected wave the

constraints can select proper combination of the estimated variables

The estimated values have range bounded by the constraints When additional information

of 119907119907119894119894 is utilized the damage can be characterized more accurately In this study three meth-

ods are proposed and verified as information level of 119907119907119894119894

Method 1 Solely based on the current data measurement (Reference-free)

Method 2 Limit the possible range of 119907119907119894119894 (Prior knowledge of the system)

Method 3 Measure 119907119907119894119894 (Additional data need to be collected)

Method 1 described in this section can be utilized without any information of 119907119907119894119894 Meth-

od 2 produces more accurate estimation results using prior knowledge of the system such as

- 12 -

119907119907119894119894 variation under temperature changes Method 3 gives unique estimated value using meas-

ured 119907119907119894119894 from other path in the same time In section 4 and 5 damage is characterized using

these three methods and the results are compared each another

36 Matching pursuit method considering the dispersion effect of Lamb

wave

Matching pursuit method has been developed for signal compression approximation and

characterization of obtained signal and function [21-23] The key concept of the matching

pursuit method is that any signal (function) can be decomposed into a linear expansion of

signals selected from a pool of basis signals (functions) called lsquodictionaryrsquo

Let 119891119891 be a function to be analyzed then 119891119891 can be approximated a set of functions select-

ed from the dictionary 119863119863 in order to best match This is done by successive approximations

with orthogonal projections Let1198921198921205741205740 isin 119863119863 the function 119891119891 can be decomposed into

119891119891 = lang1198911198911198921198921205741205740rang1198921198921205741205740 + 119877119877119891119891 (8)

where 119877119877119891119891 is the residual function after approximating 119891119891 using 1198921198921205741205740 1198921198921205741205740 is orthogonal to

119877119877119891119891 hence

1198911198912 = lang1198911198911198921198921205741205740rang2

+ 1198771198771198911198912 (9)

To minimize 119877119877119891119891 1198921198921205741205740 is chosen to maximize lang1198911198911198921198921205741205740rang The matching pursuit method is

an iterative algorithm that decomposes the residue 119877119877119891119891 by projecting it on a function of 119863119863

that matches 119877119877119891119891 almost at best as it was done for 119891119891 The residue 119877119877119899119899119891119891 is decomposed into

119877119877119899119899119891119891 = lang119877119877119899119899119891119891119892119892120574120574119899119899rang119892119892120574120574119899119899 + 119877119877119899119899+1119891119891 (10)

- 13 -

the decomposition is carried up to the order 119898119898 119891119891 can be expressed as

119891119891 = lang119877119877119899119899119891119891119892119892120574120574119899119899rang119892119892120574120574119899119899

119898119898minus1

119899119899=0

+ 119877119877119898119898119891119891 (11)

The original function 119891119891 is decomposed into a sum of dictionary elements that are chosen to

best match its residues

In this study the matching pursuit method is utilized to accurately decompose of the over-

lapped Gaussian pulses and estimate the parameters of the pulse which used for damage

characterization The 119894119894119905119905ℎ Gaussian pulse in the dictionary for approximating the obtained

signal is expressed as

119892119892120574120574119894119894 = 119891119891(119860119860119894119894 119906119906119894119894 119904119904119894119894 119889119889119894119894120601120601119894119894) (12)

where 119860119860119894119894119906119906119894119894119904119904119894119894119889119889119894119894 and 120601120601119894119894 represent amplitude arrival time scale (pulse duration) excita-

tion(central) frequency and phase of the Gaussian pulse respectively The parameters of the

pulse are estimated using a nonlinear regression based optimization method (Gauss-Newton

method) based on an initial guess of the parameters Detail procedure for estimating the pa-

rameters of the pulse is expressed in Hong et al [23]

Among the estimated parameters of the pulses pulse duration (scale) represents the disper-

sion effect If a wave-packet is spread out the estimated pulse duration is increased Using

this feature the matching pursuit method is modified to capture waves which have longer

pulse duration than transmitted A0 Lamb wave mode If the estimate pulse duration of a wave

is less than the transmitted wave the modified matching pursuit method regards it as error

- 14 -

CHAPTER 4 NUMERICAL SIMULATION

41 Composite modeling method

Composite materials can be modeled from nano to macro level depends on its engineering

purpose In general four methods are utilized for modeling of composite materials First

composite is modeled as a homogenous material expressed in terms of effective material

properties [24] The effective material properties are estimated using the modulus and the

volume fraction of each constituent theoretically (Rule of mixture) and measured experimen-

tally Second a model is established that each laminate which have own material properties

are totally bonded together In this model the stacking sequence and direction of each lami-

nates are considered additionally [25-27] Third method for composite modeling in micro

level called micromechanics [28-30] This model considers the interaction between fibers and

matrix in laminate Last multiscale modeling which the scope of modeling is extended to

molecular or quantum level of constituents [31-33] This model starts from molecular me-

chanics or molecular dynamics then the scale of model leads to the nano micro and macro

level by incorporating small scale models As the scale of models is smaller it provides more

logical way than other large scale of models while the computational cost is more expensive

The scope of this study is focused on the dynamic behavior of composite structure the beam

or plate model was utilized for the numerical simulation

- 15 -

42 Damage modeling method

Generally three damage models have been developed and utilized in the real applications

First the stiffness reduction model called as the equivalent element model was introduced

[17 34-35] This model is simply realized by replacing the region around damage area by a

region with adjusted material properties (Youngrsquos modulus shear modulus density etc)

globally Next the duplicated node model is done by disconnecting the element connectivity

assuming there is no interaction among the delaminated surface This model can be utilized

when the location and the extension of damage is important issue [34-36] Last the contact

mechanics model using the principles of contact mechanics the two interfacial forces (con-

tact resistance and sliding resistance) in the delaminated surface [37-38] In this model the

delaminated surfaces are modeled as nonlinear functions of the interlaminar normal strain

and the interlaminar shear strain This model is computationally expensive while provides

more logical way than other models There is however no universally accepted model for

general impact-induced damage in composites In the case of impact-induced damage it is

impossible to predict or estimate the location and number of delaminated layers Thus from

the uncertainty of impact-induced damage the equivalent element model can be considered

as the effective damage model

- 16 -

43 Numerical simulation description

TABLE I Material property of simulation model and test specimen

Tensile Strength 657 MPa Tensile Modulus 59 GPa Shear Strength 804 MPa Shear Modulus 24 Gpa Specific Gravity 16

PZT A

PZT C

PZT BDamage

3

250 125 125

500 Unit mm

Fig 6 Description of the numerical simulation model

The feasibility of the proposed damage characterization technique is verified by 2-D FEM

simulation using MSC PATRAN and NASTRAN Fig 6 presents the dimension PZTs and

damage location of the simulation model In the model 05mm x 05mm square shell element

was used for modeling of the A0 mode whose flexural wave characteristic The element size

was determined by considering the wavelength of the propagation Lamb wave modes The

material properties of the simulation model and damage cases are presented in Table I and

Table II respectively

PZT A and C act as both actuator and sensor for collecting pulse-echo signal They were

installed on the plate which two of them are both-side collocated for the A0 Lamb wave

- 17 -

mode decomposition [39-40] PZT B was only used as a sensor for collecting pitch-catch sig-

nal For simulating the excitation and sensing of PZTs the pin-force model was adopted that

an input force apples to the end nodes of excitation PZT and averaging element strains corre-

sponding sensing area A Gaussian pulse (tone-burst) 50 kHz central frequency was excited

as an input signal

TABLE II Damage cases for numerical simulation

119949119949120783120783 (mm) 119941119941 (mm) G value reduction ()

Case 1 650 200 90 darr

Case 2 675 150 90 darr

Case 3 675 150 93 darr

Case 4 900 150 90 darr

- 18 -

44 Numerical simulation results

015 02 025 03 035

-2

0

2

Time (ms)

Am

plitu

de

IntactDamage

Reflected wave from damage

(a) Pitch-catch signal

015 02 025 03 035

-2

0

2

Time (ms)

Am

plitu

de

IntactDamage

Reflected waves from damage

(b) Pulse-echo signal Fig 7 Comparison of A0 Lamb wave mode signals from numerical simulation between intact and damage cases Time delay attenuation and multiple reflections due to damage simulated successfully using the damage modeling method described in the previous section

Fig 7 shows that comparison of A0 Lamb wave mode signals between intact and one of

damage cases in the simulation The features of A0 Lamb wave mode passing through dam-

age area described in section 2 are successfully simulated Thus the equivalent element

method is proper for modeling damage in composites In the damage cases the additional

waves reflected from damage is observed in both pitch-catch and pulse echo signal

Table III presents the result of the proposed damage characterization algorithm using the

numerical simulation cases shown in Table II In lsquoMethod 2rsquo 119907119907119894119894 variation is assumed to plusmn

5 of the normal value (1190~1315ms) The normal value of 119907119907119894119894 for lsquoMethod 3rsquo is meas-

ured as 1253 ms In each case the damage is characterized using the proposed technique

Note that as the information of 119907119907119894119894 is specified the estimated value is closer to the exact

value

- 19 -

TABLE III Damage characterization for the numerical simulation models

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

Case 1

Exact 650 200

Method 1 248 50~70 79~414 160~834

Method 2 248 603~667 135~241 271~486

Method 3 248 635 188 379

Case 2

Exact 675 150

Method 1 193 60~75 01~251 37~651

Method 2 193 618~682 114~222 295~575

Method 3 193 65 168 435

Case 3

Exact 675 150

Method 1 233 60~75 66~302 140~649

Method 2 233 668~738 85~196 182~419

Method 3 233 703 140 301

Case 4

Exact 900 150

Method 1 204 75~100 48~349 118~856

Method 2 204 860~951 107~216 264~530

Method 3 204 906 162 397

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1253 ms 1190~1315 ms Measured 119907119907119894119894 1253 ms

- 20 -

CHAPTER 5 EXPERIMENTAL VERIFICATION

51 Overview

In this section experimental study is investigated so that the proposed damage characteri-

zation technique Overall experimental setup and test specimen are shown in Fig 8 and 9

The experimental setup consists of test specimens data acquisition (DAQ) system and self-

sensing circuit Self-sensing circuit was installed for pulse-echo measurement using PZTs

[40-41]

PZT A(PZT C) PZT B

500

500

140

75

250

Teflon Inserted Damage (d =25)

Unit mm

Thickness = 3

PZT A(PZT C) PZT B

500

500

150

90

250

Impact Location

Unit mm

Thickness = 3

(a) Teflon inserted damage specimen

(b) Impact-induced damage specimen

Fig 8 Test specimens for the experimental verification In the Teflon inserted damage Teflon tapes with 25mm diameter were inserted at the center of the specimen The impact-induced damage was introduced to the specimen using the impact test equipment

- 21 -

511 Specimen

Two carbon composite plates with a dimension of 500times500times3mm3 were manufactured by

Nexcomes Inc by stacking 12 layers of CF3327 3K prepreg (woven fabrics type) manufac-

tured by Hankuk Fiber Inc The material properties of the specimen are presented in Table I

One of the specimens Teflon tapes with 25mm diameter was inserted in every 3 layers at the

center of the specimen to mimic the damage The other specimen is utilized for the character-

ization of real impact-induced damage

512 Data acquisition (DAQ) system and PZT

The data acquisition (DAQ) system consists of an arbitrary waveform generator (AWG) a

high speed signal digitizer (DIG) and two multiplexers (MUX) Using the 16-bits AWG

tone-burst input signal with a plusmn10 peak-to-peak voltage were generated The response signal

was measured by the 14-bit DIG For the signal-to-noise ratio reduction both pitch-catch and

pulse-echo signals were measured 150 times and averaged in the time domain 3 identical

circular PZTs of 10mm diameter manufactured by APC International Ltd were installed on

the specimen A Gaussian pulse (tone-burst) 50 kHz central frequency was excited as an in-

put signal The excitation frequency was chosen in order to avoid overlapping of the S0 and

A0 Lamb wave modes and generating of the higher modes (S1 A1hellip)

- 22 -

PZT Actuator

PZT Sensor

AWG

DIG

DAQ System Specimen

SensingActuatingMUX

1

23

4

1 Sending signal to MUX

4 Sending signal to DIG

Self-SensingCircuit

18 Unit mm

10

(a)

(b)

Fig 9 Overall experimental setup (a) Data acquisition (DAQ) system and (b) Installed circu-lar PZT transducer on the specimen with its dimension

513 Impact test

The impact-induced damage was introduced to the specimen using the impact test equip-

ment presented in Fig 10 A 5kg mass with 10mm tip was dropped from 015m height onto

the same location of the specimen The signals were obtained at each number of impacts

Since the reflected wave is not detected at the first impact case the data obtained from the

second and third impact were utilized for the analysis

10

Unit mm (a)

(b)

Fig 10 (a) Impact test equipment for introducing impact-induced damage and (b) Tip of the impact tester with 10mm diameter

- 23 -

514 Temperature effect test

In reality structures are subjected to changing environmental conditions such as tempera-

ture that affect measured signal and cause false-alarm Thus the robustness of the proposed

damage characterization technique under temperature variation in investigated (Fig 11) The

specimens were placed inside a temperature chamber and a thermocouple was installed on

the specimen to measure temperature of surface of the specimen The signal was obtained

under three temperature cases (0 20 and 50degC) and the humidity was kept at 30

Temperature Chamber

Test Specimen

Fig 11 Test specimen in temperature chamber

- 24 -

52 Teflon inserted damage

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Reflected wave from damage

(a)

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Transmitted WaveReflected Wave

Detected reflected wave

(b)

Fig 12 A0 Lamb wave mode obtained using pitch-catch method from the Teflon inserted damage specimen (a) The reflected wave from damage is shown (b) The Teflon inserted damage is detected by capturing the reflected wave

Fig 12 shows the obtained pitch-catch signal from the Teflon inserted damage specimen

and damage diagnosis result The Teflon inserted damage is detected successfully by captur-

ing the reflected wave Table IV presents the result of the Teflon inserted damage characteri-

zation 119907119907119894119894 is measured as 1213ms and the variation is assumed to plusmn 5 of the normal val-

ue (1152~1273ms) The results indicate that the proposed damage characterization approach

works well in real application

TABLE IV Damage characterization for the Teflon inserted damage specimen

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

Exact 635 25

Method 1 123 60~75 73~338 298~1220

Method 2 123 619~662 229~310 934~1188

Method 3 123 631 284 1161

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1213 ms

- 25 -

53 Impact-induced damage

11 115 12 125 13-4

-2

0

2

4

Time (ms)

Vol

tage

(mV

)

IntactImpact2Impact3

Reflected wave from damage

(a) Pitch-catch signal

11 115 12 125 13 135

-05

0

05

1

Time (ms)

Volta

ge (m

V)

IntactImpact2Impact3

Boundary reflection

Reflected wave from damage

(b) Pulse-echo signal

Fig 13 The signal obtained from the test specimen subjected to impact The impact was in-troduced at the same location of the specimen and the signal was obtained after each impacts As the number of impact is increased the transmitted A0 Lamb wave mode is delayed and attenuated

The obtained signals from impact-induced damage specimen are presented in Fig 13 It is

observed that as the number of impact is increased the transmitted A0 Lamb wave mode is

more delayed because the wave velocity in the damage area is reduced Fig 14 shows the

false-alarm test result whether the proposed damage detection algorithm works even in pris-

tine condition of the test specimen It is shown that no false alarm is occurred because no ad-

ditional reflected wave is captured

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Transmitted Wave

(a) Transmitted wave

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Reflected Wave

No reflected wave

(b) Reflected wave

Fig 14 False alarm test for the proposed damage detection algorithm in pristine condition of the specimen Reflected wave from damage is not captured false alarm is not occurred

- 26 -

TABLE V Damage characterization for the impact-induced damage specimen

of Impacts m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

2

(1470 J)

Method 1 080 65~90 75~246 470~1157

Method 2 080 801~885 114~245 711~1134

Method 3 080 843 179 1122

3

(2205 J)

Method 1 148 50~100 06~380 21~1134

Method 2 148 758~838 249~368 843~1125

Method 3 148 798 308 1044

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1213 ms

Table V presents the characterization results of the impact-induced damage case Here the

measured 119907119907119894119894 and the assumed variation are same as the artificial damage case It is observed

that as the number of impact is increased 119889119889 is increased while 119907119907119889119889 is reduced Thus it is

validated that the proposed damage characterization approach estimates the increment of the

damage well

54 Temperature effect test

541 Teflon inserted damage

11 115 12 125 13-10

-5

0

5

10

Time (ms)

Vol

tage

(mV

)

0 oC20 oC50 oC

(a) Pitch-catch signal

11 115 12 125 13 135

-1

0

1

2

3

Time(ms)

Volta

ge (m

V)

0 oC20 oC50 oC

(b) Pulse-echo signal

Fig 15 Signals obtained from the Teflon inserted damage specimen under temperature varia-tion Time delay and amplitude variation are occurred

Fig 15 shows the obtained signals from the Teflon inserted damage specimen subjected to

- 27 -

temperature variation Time delay and amplitude variations are occurred there is no addition-

al damage is introduced The analyzed results of the obtained signals are reported in Table VI

The measured 119907119907119894119894 is 1247 ms at 0 degC 1213 ms at 20 degC and 1168 ms at 50 degC respectively

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms (1152~1273 ms)

The estimated results are also changed under temperature variations the fluctuation of esti-

mated range and exact value is small

TABLE VI Temperature effect test for the Teflon inserted damage specimen

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

0 degC

Exact 635 25

Method 1 113 60~70 79~268 352~1291

Method 2 113 560~619 233~344 1034~1172

Method 3 113 606 257 1141

20 degC

Exact 635 25

Method 1 123 60~75 73~338 298~1220

Method 2 123 619~662 229~310 934~1188

Method 3 123 631 284 1161

50 degC

Exact 635 25

Method 1 137 60~75 72~328 263~1228

Method 2 137 622~687 183~299 666~1086

Method 3 137 642 264 960

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1247 ms at 0 degC 1213 ms at 20 degC and 1168 ms at 50 degC

- 28 -

542 Impact-induced damage

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Vol

tage

(mV

)

Transmitted WaveReflected Wave

No reflected wave

(a) 0 degC

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Transmitted WaveReflected Wave

No reflected wave

(b) 50 degC

Fig 16 False alarm test under temperature variations It is shown that proposed damage char-acterization algorithm is robust under temperature variations

The false alarm test in pristine condition of the impact-induced damage specimen under

temperature variation was investigated (Fig 16) It is validated that the proposed damage di-

agnosis algorithm is robust under temperature variations Fig 17 shows the obtained signals

from the impact-induced damage (impacted 3 times) specimen under temperature variation

The analyzed results of the obtained signals are reported in Table VII The measured 119907119907119894119894 in

each temperature and the assumed variation are same as the artificial damage case

11 115 12 125 13-5

0

5

Time (ms)

Volta

ge (m

V)

0 oC20 oC50 oC

(a) Pitch-catch signal

11 115 12 125 13 135

-1

0

1

2

3

Time(ms)

Vol

tage

(mV

)

0 oC20 oC50 oC

(b) Pulse-echo signal

Fig 17 Signals obtained from impact-induced damage specimen (Impact 3) under tempera-ture variation

- 29 -

TABLE VII Temperature effect test for the impact induced damage specimen (Impact 3 case)

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

0 degC

Method 1 135 75~90 72~310 267~1148

Method 2 135 705~779 264~381 978~1126

Method 3 135 763 289 1071

20 degC

Method 1 148 50~100 06~380 21~1134

Method 2 148 758~838 249~368 843~1125

Method 3 148 798 308 1044

50 degC

Method 1 153 75~100 53~415 174~1227

Method 2 153 804~889 214~336 702~1103

Method 3 153 815 320 1050

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1247 ms at 0 degC 1213 ms at 20 degC and 1168 ms at 50 degC

- 30 -

CHAPTER 6 CONCLUDING REMARKS

61 Executive summary

In this study the reference-free damage detection localization and quantification tech-

nique in a multi layered composite plate is proposed Additional reflected A0 Lamb wave

mode from the damage due to multiple reflections in damage area is used as a feature for

characterizing the damage First damage is characterized in terms of a damage index (m-

value) defined as the ratio of the physical length of the damage to the wavelength of the A0

Lamb wave mode Then the range of possible damage location length and severity are esti-

mated with constraints This procedure can be accomplished without prior baseline data To

realize the proposed technique the matching pursuit technique is modified for reduction of

the error by considering the dispersive characteristic of Lamb wave The matching pursuit

method estimates the arrival time of each wave packets accurately even two wave packets are

overlapped The numerical simulations and experimental tests including Teflon tape inserted

and impact-induced damage are investigated for the feasibility of the proposed technique

Moreover the effectiveness of the proposed technique under changing temperature is exper-

imentally verified

- 31 -

62 Suggestion for further study

The limitations of this study and the possible improvement and further studies are listed as

follow

1 Measurement of impact-induced damage length using conventional NDT technique

Additional experimental verification for impact-induced damage characterization is

needed by measuring the real length of damage using conventional NDT technique

such as ultrasonic scan or infrared camera

2 Limitation of both side collocated PZT installation In this study excitation PZTs are

installed both side and collocated for decomposing A0 Lamb wave mode However it

can produce decomposition error due to imperfection of PZT location and their installa-

tion Moreover in many cases assess to both surface of structures is not allowed such

as aircraft applications To improve this limitation a reference-free damage characteri-

zation using single side PZTs will be considered

3 Complex structure application In this study damage is characterized in simple plate-

like structure When the monitored structure becomes complex the response signals al-

so becomes complex due to its geometry and boundary Thus the investigations for

complex structure application such as stiffener spar and curvature are needed

- 32 -

REFERENCES

[1] Cawley P and Alleyne D (1996) ldquoThe use of Lamb waves for the long range inspection of large structuresrdquo Ultrasonics 34 pp287-290

[2] Ghosh T Kundu T and Karpur P (1998) ldquoEfficient use of Lamb waves for detecting defects in large platesrdquo Ultrasonics 36 pp791-801

[3] Guirguitui V (2003) ldquoLamb wave generation with piezoelectric wafer active sensors for structural health monitoringrdquo SPIE Conference on Smart Structures and Materials San Diego CA 5391 pp111-422

[4] Ragvahan A and Cesnik CE S (2004) ldquoModeling of piezoelectric-based Lamb wave generation and sensing for structural health monitoringrdquo SPIE Conference on Smart Struc-tures and Materials San Diego CA 5391 pp419-430

[5] Rose JL Pilarski A and Dirit JJ (1993) ldquoAn approach to guided wave mode selection for inspection of laminated platerdquo Journal of Reinforced Plastics and Composites 12(5) pp536-544

[6] Guo N and Cawley P (1993) ldquoThe interaction of Lamb waves with delaminations in composite laminatesrdquo Journal of Acoustical Society of America 94(4) pp 2240-2246

[7] Valdez SHD and Soutis C (2002) ldquoReal-time nondestructive evaluation of fiber com-posite laminates using low-frequency Lamb wavesrdquo Journal of Acoustical Society of Amer-ica 111(5) pp 2026-2033

[8] Yang M and Qiao P (2005) ldquoModeling and experimental detection of damage in virous materials using the pulse-echo method and piezoelectric sensoractuatorsrdquo Smart Materials and Structures 14 pp1083-1100

[9] Lestari W and Qiao P (2005) ldquoApplication of wave propagation analysis for damage identification in composite laminated beamsrdquo Journal of Composite Materials 39(22) pp1967-1984

[10] Kessler S S Spearing S M and Soutis C (2002) ldquoDamage detection in composite ma-terials using Lamb wave methodsrdquo Smart Materials and Structures 11 pp269-278

[11] Sohn H Park G Wail J R Limback N P and Farrar C R (2004) ldquoWavelet-based active sensing for delamination detection in composite structuresrdquo Smart Materials and Structures 13 pp 153-160

[12] Petculescu G Krishnaswamy S and Achenback J D (2008) ldquoGroup delay measure-ments using modally selective Lamb wave transducers for detection and sizing of delamina-tions in compositesrdquo Smart Materials and Structures 17

[13] Sohn H Park H W Law K H and Farrar C R (2007) ldquoDamage detection in compo-

- 33 -

site plates by using an enhanced time reversal methodrdquo Journal of Aerospace Engineering (ASCE) pp 141-151

[14] Sohn H (2007) ldquoEffects of environmental and operational variability on structural health monitoringrdquo Philosophical Transactions of the Royal Society Ardquo 365 pp539-560

[15] Rytter A (1993) ldquoVibration based inspection of civil engineering structuresrdquo PhD The-sis Department of Building Technology and Structural Engineering University of Aalborg Aalborg Denmark

[16] Hayashi T and Kawashima K (2002) ldquoMultiple reflections of Lamb waves at a delami-nationrdquo Ultrasonics 40 pp193-197

[17] Wang C H and Rose L R F (2003) ldquoWave reflection and transmission in beams con-taining delamination and inhomogeneityrdquo Journal of Sound and Vibration 264 pp 851-872

[18] Ramadas C Balasubramaniam K Joshi M and Krishnamurthy C V (2009) ldquoInterac-tion of the primary anti-symmetric Lamb mode (Ao) with symmetric delaminations numer-ical and experimental studiesrdquo Smart Materials and Structures 18

[19] Ramadas C Balasubramaniam K Joshi M and Krishnamurthy C V (2010) ldquoInterac-tion of guided Lamb waves with an asymmetrically located delamination in a laminated composite platerdquo Smart Materials and Structures 19

[20] Wilcox P Lowe M and Cawely P (2001) ldquoThe effect of dispersion on long-range in-spection using ultrasonic guided wavesrdquo NDTampE International 34 pp1-9

[21] Mallat S G and Zhang Z (1993) ldquoMatching pursuits with time-frequency dictionariesrdquo IEEE Tansactions on Signal Processing 41(12) pp3397-3415

[22] Zhang G Zhang S and Wang Y (2000) ldquoApplication of adaptive time-frequency de-composition in ultrasonic NDE of highly scattering materialsrdquo Ultrasonics 38 pp 961-964

[23] Hong J C Sun K H and Kim Y Y (2005) ldquoThe matching pursuit approach based on the modulated Gaussian pulse for efficient guided-wave damage inspectionrdquo Smart Mate-rials and Structures 14 pp548-560

[24] Herakovich C T (1997) Mechanics of Fibrous Composites Wiley 480 pages

[25] Reddy J N (1985) ldquoA review of the literature on finite-element modeling of laminated composite platesrdquo The Shock and Vibration Digest 17(4) pp 3-8

[26] Reddy J N (2003) Mechanics of Laminated Composite Plates and Shells Theory and Analysis CRC Press 856 pages

[27] Palacz M Krawczuk M and Ostachowicz W (2005) ldquoThe spectral finite element mod-el for analysis of flexure-shear coupled wave propagation Part 1 laminated multilayer

- 34 -

composite beamrdquo Composite Structures 68 pp37-44

[28] Zhang Y C and Harding J (1990) ldquoA numerical micromechanics analysis of the me-chanical properties of a plain weave compositerdquo Computer amp Structures 36(5) pp 839-844

[29] Nemat-Nasser S and Hori M (1999) Micromechanics Overall Properties of Heteroge-neous Materials Elsevier 810 pages

[30] Scida D Aboura Z Benzeggagh M L and Bocherens E (1999) ldquoA micromechanics model for 3D elasticity and failure of woven-fibre composite materialrdquo Composites Science and Technology 59 pp 505-517

[31] Thomas Y and Wu X H (1997) ldquoA multiscale finite element method for elliptic prob-lems in composite materials and porous mediardquo Journal of Computational Physics 134 pp 169-189

[32] Lee K H Moorthy S and Ghosh S (1999) ldquoMultiple scale computational model for damage in composite materialsrdquo Computer Methods in Applied Mechanics and Engineer-ing 172 pp 175-201

[33] Namilae S and Chandra N (2005) ldquoMultiscale model to study the effect of interfaces in carbon nanotube-based compositesrdquo Journal of Engineering Materials and Technology (ASME) 127 pp222-232

[34] Adams D E (2007) Health monitoring of Structural Materials and Components Wiley 460 pages

[35] Gopalakrishnan S (2009) ldquoModeling Aspects in Finite Elementrdquo Encyclopedia of Struc-tural Health Monitoring 2 pp791-809

[36] Palacz M Krawczuk M and Ostachowicz W (2005) ldquoThe spectral finite element mod-el for analysis of flexure-shear coupled wave propagation Part 2 Deminated multilayer composite beamrdquo Composite Structures 68 pp45-51

[37] Baik J M and Thopson R B (1984) ldquoUltrasonic scattering from imperfect interfaces A quasi-static modelrdquo Journal of Nondestructive Evaluation 4(3) pp177-196

[38] Luo H and Hanagud S (2000) ldquoDynamics of delamincated beamsrdquo International Jour-nal of Solids and Structures 37 pp1501-1519

[39] Kim S B and Sohn H (2007) ldquoInstantaneous reference-free crack detection based on polarization characteristics of piezoelectric materialsrdquo Smart Materials and Structures 16 pp2375-2387

[40] An Y K and Sohn H (2010) ldquoInstantaneous crack detection under varying temperature and static loading conditionsrdquo The Special Issue of Structural Control and Health Monitor-ing 17(7) pp 730-741

- 35 -

[41] Lee S J and Sohn H (2006) ldquoActive self-sensing scheme development for structural health monitoringrdquo Smart Materials and Structures 15(6) pp 1734-1746

- 36 -

요 약 문

복합재료 손상의 무기저 진단 및 정량화

복합재료는 재료의 경량성 내구성 고강도의 특성으로 인하여 항공기에서부터

교량에까지 널리 사용되고 있다 하지만 외부의 충격으로 인한 복합재료 내부의

손상은 재료의 강도를 감소시켜 구조물의 안정성을 저하시키는 요인이 된다 하지만

복합재료의 손상은 재료 내부에서 발생하여 육안으로는 식별하기도 힘들며 표면에서

보이는 것만으로 내부의 손상 정도를 판단하기는 불가능하다는 단점을 가지고 있다

본 연구에서는 손상 내부에서 반사로 인해 발생하는 파를 이용하여 기저자료를

사용하지 않고 특정 시점에서 얻은 자료만으로 복합재료의 손상을 진단하고 정량화

할 수 있는 기법을 제안한다 제안된 기법의 적용성을 검증하기 위하여 유한요소법을

이용한 수치해석을 수행하였고 Teflon tape 이 삽입된 시편과 실제 충격하중으로

인한 손상을 입은 시편을 이용한 실험을 진행하였다 수치해석 및 실험 결과 제안된

기법이 복합재료의 손상을 성공적으로 진단 및 정량화할 수 있음을 확인하였다 또한

다양한 외부 온도 환경에서 제안된 기법의 환경영향 실험을 수행하였고 다양한

환경에서 손상의 진단 및 정량화가 성공적으로 이루어짐을 검증하였다

핵심어 무기저 손상진단 및 정량화 압전센서 복합재료 유한요소법 Matching pursuit

method

- 37 -

감 사 의 글

기대와 두려움을 안고 시작한 석사과정이 어느덧 2 년 이라는 세월 속에서 추억으로 자리잡아

가고 있습니다 학부생 기간을 포함한 지난 3 년 동안 SHM 이라는 새로운 분야를 접하고 연구를

수행 할 수 있도록 도와주신 고마운 분들이 있어 그 감사를 표하고자 합니다

우선 저의 많은 부족한 점을 일깨워 주시고 연구에 대한 뜻을 더욱 확고히 세울 수 있게 해

주신 지도교수님이신 손 훈 교수님께 깊은 감사의 뜻을 전합니다 짧은 기간이었지만 교수님의

도움으로 많은 것을 배우고 느끼고 깨달을 수 있었습니다 또한 본 연구를 수행할 수 있도록 지

원해주신 국토해양부 (U-City 석 박사과정 지원사업)에 깊은 감사 드립니다

때로 힘들고 지칠 때 항상 저를 믿고 응원해 주시는 가족이 있었기에 지금의 글을 쓸 수 있었

습니다 또한 지금 연구실에서 함께 생활하고 있는 윤규형 민구형 철민이형 현석이형 진열이형

호민이형 병진이 선혜누나와 성대에 계신 창길이형 미국에 계신 의재형과 Debaditiya 에게도 감

사의 말씀을 전합니다 그리고 모두 나열할 수는 없지만 카이스트 생활의 활력소가 되어준 카이

스트 건설 및 환경공학과 선후배들 직원 분들과 SOS 멤버들에게도 감사를 표합니다

앞서 거론한 많은 이들의 도움이 없었다면 저 또한 지금 이 자리에 있을 수 없었을 것입니다

많이 부족하기에 더 노력하고 배우고자 했던 초심을 다시 한번 생각하며 더 큰 결실을 맺기 위해

앞으로 나아가겠습니다 도움을 주신 모든 분들께 이 감사의 글을 바칩니다

2011년 6월 카이스트에서 임형진 올림

- 38 -

CURRICULUM VITAE

Hyung Jin Lim

Research Assistant

The Civil and Environmental Engineering Department

Korea Advanced Institute of Science and Technology (KAIST)

291 Daehak-Ro Guseong-Dong Yuseong-Gu

Daejeon Republic of Korea 305-701

Phone (82)+42-350-3665 Fax (82)+42-350-3610

Email limhj87gmailcom

RESEARCH INTERESTS

Damage Detection Localization and Quantification on Composite and Metallic Structures Pattern Recognition Finite Element Modeling (FEM)

EDUCATION

2009 ndash 2011 MS Civil and Environmental Eng Korea Advanced Institute of Science and Technology (KAIST) Korea Rep

2006 ndash 2009 BS Civil and Environmental Eng Korea Advanced Institute of Science and Technology (KAIST) Korea Rep

JOURNAL PUBLICATIONS The corresponding author is underlined 1 Hyung Jin Lim Hoon Sohn ldquoReference-free damage detection localization and quantification

using in compositerdquo in preparation for Journal of the Acoustical Society of America 2011

2 Chul Min Yeum Hoon Sohn Jeong Bum Ihn Hyung Jin Lim ldquoInstantaneous delamination detection in a composite using a dual piezoelectric transducer networkrdquo Submitted to Compo-sites Science and Technology 2011

3 Yun-Kyu An Hyung Jin Lim Min Koo Kim Jin Yeol Yang Hoon Sohn Chang Guen Lee ldquoInvestigation of applicability of local reference-free damage detection techniques to in-situ

- 39 -

bridgesrdquo Submitted to ASCE Journal of Structural Engineering 2011

4 Hyung Jin Lim Min Koo Kim Hoon Sohn Chan Yik Park ldquoImpedance based damage detec-tion under varying temperature and loading conditionsrdquo Submitted to NDTampE International 2011

5 Hyung Jin Lim Hoon Sohn ldquoQuantitative estimation of transmitted and reflected lamb waves at discontinuityrdquo Journal of the Korean Society for Nondestructive Testing Vol30 No 4 pp1-8 2010

CONFERENCE PROCEEDINGS

The corresponding author is underlined 1 Hyung Jin Lim and Hoon Sohn ldquoReference-free damage detection localization and quantifi-

cation in compositesrdquo ASCE Engineering Mechanics Institute Conference Boston June 2-4 2011

2 Hyung Jin Lim and Hoon Sohn ldquoReference-free delamination quantification within multi-layer composite materialsrdquo KSNT Spring Conference May 26-27 Gyeong-Ju Korea 2011

3 Min Koo Kim Hyun Jin Lim Hoon Sohn Chan Yik Park ldquoImpedance-based bolt loosening detection under varying temperature and loadingrdquo Asian Pacific Workshop on Structural Health Monitoring Tokyo Japan November 30-December 2 2010

4 Hyung Jin Lim Chul Min Yeum and Hoon Sohn ldquoModeling of Impact-induced Delamination in a Multilayer Composite Platerdquo The 23rd KKCNN Symposium on Civil Engineering Taipei (Taiwan) November 13-15 2010

5 Yun-Kyu An Hyung Jin Lim and Hoon Sohn ldquoTemperature independent Delamination Detec-tion using data normalizationrdquo The 3rd International Conference on Multi-Functional Materi-als and Structures Chonbuk National Univ Jeonju Korea September 14-18 2010

6 Hoon Sohn Yun-Kyu An Chang Gil Lee Min Koo Kim Hyung Jin Lim ldquoApplication of online NDT techniques to decommissioned bridge testingrdquo the 5th European Workshop on Structural Health Monitoring Sorrento Italy June 29-July 02 2010

7 Min Koo Kim Hyung Jin Lim Hoon Sohn Chan Yik Park ldquoImpedance based damage diag-nosis of a complex composite aircraft wing under changing loading and temperature condi-tionsrdquo the 5th European Workshop on Structural Health Monitoring Sorrento Italy June 29-July 02 2010

8 Yun-Kyu An Min Koo Kim Hyung Jin Lim and Hoon Sohn ldquoValidations of reference-free crack detection techniques through a decommissioned bridge testrdquo 2010 COSEIK Annual Con-ference Jeju Korea April 8-9 2010

- 40 -

PARTICIPANT PROJECTS

1 Development of Smart Steel Members for Online Steel Structure Monitoring POSCO Corporation (Funded 41000000 KRW (41000 USD) for 010511 to 123111)

2 Development of Enabling Technologies for Structural Health Monitoring Korean Agency for Defense Development (Funded 240000000 KRW (240000 USD) out of total 3200000000 KRW (3200000 USD) for 060108-053111)

3 Development of On-board SHM Technologies for Composite Air Vehicles The Boeing Company (Funded 400000 USD (400000000 KRW) for 080108 to 073111)

EDUCATIONAL EXPERIENCES

Fall 2009

Teaching Assistant CE207 ldquoElementary Structural Engineering and Laboratoryrdquo KAIST 3 unit undergraduate course

EXPERIMENTAL EXPERIENTCES

2010 Monitoring of Real Aircraft Composite Wing Segment Korea Agency for Defense De-velopment (ADD) Daejeon Korea Real time monitoring of composite air vehicles fund-ed by Korea Agency for Defense Development (ADD)

2009-2011

Field Monitoring of Real Bridge Structures Gimpo (Ramp G bridge) amp Icheon (Sam-seung bridge) Korea US-Korea collaborative research for bridge monitoring testbeds

HONORS amp AWARDS

Scholarships for outstanding students granted KAIST (2007-2009) The three prominent students in the civil and environmental engineering department of KAIST were selected as recipients This scholarship provided 2000$ per year for three years

- 41 -

Page 12: 복합재료 손상의 무기저 진단 및 정량화 - KAISTssslab.kaist.ac.kr/article/pdf/Thesis_2.pdf · 1 Declaration of Ethical Conduct in Research: I, as a graduate student

CHAPTER 1 INTRODUCTION

11 Motivation

Composite materials has been increasingly demands in modern structures ranging from the

latest aircraft to retrofitted bridge deck due to its light weight superior specific strength and

stiffness properties However the occurrence of damage due to abrupt impact or sequential

accumulation of defect may considerably degrade structural strength or stiffness and reduce

the safety and reliability of the system Recently Lamb waves based damage detection tech-

niques using piezoelectric (PZT) transducers have emerged as one of the promising structural

health monitoring (SHM) technology for detecting hidden damage because of the possibility

of long range inspection and high sensitivity [1-4]

12 Literature review

This section will discuss researches that have been previously carried out on topics of

guided wave based SHM techniques using PZT transducers for damage in composite materi-

als The early investigations of interaction between Lamb wave and damage in composite can

be found in Rose et al and Guo et al [5-6] From further studies reflected signals from dam-

age were shown to be an effective method for damage diagnosis and localization using PZTs

[7-9] The reflected signals from damage were collected using one PZT as an actuator and the

other PZT as a sensor However the techniques cannot be used to determine the extent of

- 1 -

damage with reasonable precision

Kesseler et al proposed a technique monitoring the wavelet coefficients of transmitted ul-

trasonic signal amplitude as a potential way to damage detection [10] One drawback of this

technique is that the variability in coupling and the sensitivity of installed PZT transducers

can induce changes in the amplitude and energy of the signal that are cause false-alarm Oth-

er wavelet based delamination detection and localization technique developed using PZT

sensor network and outlier analysis [11]

Damage detection and quantification in composite plate was also investigated by group de-

lay measurement due to damage [12] However it is assumed that the group velocity in dam-

age area as the group velocity propagates half thickness of the structure in dispersion curve in

the quantification procedure Thus the damage quantification was investigated only Teflon

inserted damage case

Sohn et al proposed a damage detection technique which does not rely on any past base-

line signals to assess damage in composite plates by using an enhanced time reversal method

[13] The damage is assessed using the violation of time reversibility when nonlinearity is

caused by the damage along a direct wave path Thus it allows instantaneous identification

of delamination without requiring the baseline signal for comparison

- 2 -

13 Objective and uniqueness

In this study a reference-free damage characterization technique including detection lo-

calization and quantification of impact-induced damage in composite materials is developed

The general damage characterization can be categorized into the following levels [14]

Level 0 Define damage

Level 1 Determine the presence of damage within the structure

Level 2 Locate the regions of damage

Level 3 Quantify the damage

Level 4 Predict the remaining service life of the structure

Through the proposed technique from level 1 to level 3 can be achieved

Most of common approaches of damage characterization techniques described in the pre-

vious section use prior obtained baseline data However these approaches can cause false

alarm due to the effect of environmental variations such as temperature and ambient vibra-

tions [15]

The uniqueness of this study is

1 Damage is characterized without using prior measured baseline data

2 Robustness of the proposed technique under temperature variation is investigated

3 Additional waves reflected from inside damage have been used in addition to transmit-

ted and reflected waves

- 3 -

14 Organization

The organization of this thesis is summarized as follows

Chaper1 A general introduction including motivation literature review objective and

uniqueness of this study is described

Chaper2 Interactions between Lamb wave mode and damage in composites such as time

delay attenuation and multiple reflections are summarized Damage index (m-value) for

damage characterization is defined

Chaper3 Algorithms for the reference-free damage characterization technique using m-

value are developed Modified matching pursuit for this study is also described

Chaper4 Verification of the algorithm using numerical simulation is presented Various

modeled damage cases are characterized

Chaper5 Experimental verifications are described Two damage cases (Teflon inserted and

impact-induced damage) are investigated for the feasibility of the proposed damage charac-

terization technique Temperature effect test using temperature chamber is also carried out

Chaper6 Conclusion and summary of the present study is reported In advance suggestions

and future work recommendations are represented

- 4 -

CHAPTER 2 INTERACTION OF LAMB WAVE WITH DAMAGE

21 A0 Lamb wave mode delay and attenuation in the time domain

Time delay and attenuation of the A0 mode Lamb wave after passing through the damage

area due to effective thickness reduction were reported in previous researches [812] Moreo-

ver the attenuation of the A0 Lamb wave mode is occurred by degrading of shear modulus of

in the damage area However the time delay and amplitude variation of the A0 Lamb wave

mode also occurred even when the structure is health condition by the material property

changing due to temperature variation of the structure (Fig 1) Thus conventional damage

characterization methods using the time delay and attenuation of A0 Lamb wave mode as a

feature can cause false-alarm in real application

11 115 12 125 13-10

-5

0

5

10

Time (ms)

Vol

tage

(mV

)

IntactDamage

Reflected wave from Damage

(a)

11 115 12 125 13-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

0 oC20 oC50 oC

(b)

Fig 1 Lamb wave propagation characteristics (a) A0 Lamb wave mode time delay and atten-uation after passing through a damage (b) The time delay and amplitude variation of A0 Lamb wave mode under temperature variation in pristine condition of structure Because of this conventional damage detection techniques in composites have limitations for real struc-ture application

- 5 -

22 Multiple reflections in the damage area

Standing wave patterns due to multiple reflections of A0 Lamb wave mode in the damage

area were observed through numerical simulations and ultrasonic wave-field imaging using a

scanning laser vibrometer [16] The formation of standing waves in damage is illustrated

through Fig 2 When the incident waves enter the damage area a significant portion of the

waves are reflected back from the exit Then the reflected waves travel through the damage

and undergo reflection again at the original entrance As a consequence of multiple reflec-

tions at the entrance and the exit a considerable amount of ultrasonic energy is trapped inside

the damage area [17-19]

Multiple Reflections

Incident Wave Transmitted Wave

Damage

Fig 2 Multiple reflections in the damage area The multiple reflections of A0 Lamb waves in the damage area are utilize as a feature for characterization of the damage in this study

- 6 -

CHAPTER 3 THEORETICAL DEVELOPMENT

31 Overview

Damage

l = l1 + d + l2

d = vdtd l1 = vit1 l2 = vit2

t = t1 + td + t2

Excitation and Sensing PZT Sensing PZT

l

Transmitted Wave

Reflected Waves

Incident Wave

Fig 3 Parameters and notations of the proposed damage characterization technique Damage is detected by identifying the additional reflected wave from the damage due to multiple re-flections (dashed lines) Localization and quantification of the damage is carried out by measuring the arrival time of transmitted and reflected waves from the damage

In this section the reference-free damage characterization algorithm is proposed Damage

diagnosis and quantification are achieved by identifying the additional reflected waves due to

multiple reflections in the damage area (dashed lines in Fig 3) Moreover damage is local-

ized using additional information from pulse-echo method implement For the localization

and the quantification of damage the variables such as the path length and A0 Lamb wave

mode velocity in damage area are defined (Fig 3) For example 119897119897 is expressed as

119897119897 = 1198971198971 + 119889119889 + 1198971198972 (1)

where 1198971198971 119889119889 and 1198971198972 are the distance from the excitation PZTs to the entrance of damage the

damage length and the distance from the exit of damage to the sensing PZT respectively

- 7 -

Moreover 119907119907119894119894 and 119907119907119889119889 are also defined as the A0 Lamb wave mode velocity in pristine area

and damage area respectively Note that the whole procedure of the damage characterization

is accomplished based solely on the signals obtained from the current state of the structure

32 Damage index (m-value)

The proposed reference-free damage characterization technique utilizes the waves reflect-

ed additionally from the damage area as a feature The arrival time difference of the transmit-

ted and reflected A0 Lamb wave modes reflected from the exit and the entrance of the delam-

ination area one times each (∆119905119905) contains the information of the physical length of the dam-

age (119889119889) and A0 Lamb wave mode velocity in the damage area (119907119907119889119889) The wave velocity in the

damage area which is related to the shear modulus (G) represents the severity of the damage

∆119905119905 = 2119905119905119889119889 =2119889119889119907119907119889119889

(2)

However based on the measured data available the 119889119889 and 119907119907119889119889 (or G) cannot be uniquely

determined Fig 4 shows two FEM models which have different physical length and wave

velocity (G) of the damage Detail explanation of delamination modeling technique will be

covered in section 4 The similar signal in the time domain and ∆119905119905 are obtained from the

two different damage cases

- 8 -

Damage 1

Damage 2

20

Unit mm15

90 G value reduction

943 G value reduction

m-value = 248

m-value = 248

015 02 025 03 035

-2

-1

0

1

2

Time (ms)

Ampl

itude

Damage 1Damage 2

Fig 4 Two different FEM models which have different 119889119889 and 119907119907119889119889 (or G) This represents that the damage cannot be characterized uniquely based on the measured data available

Thus in this study the damage index (m-value) which can be uniquely defined based on

the available data is introduced The m-value is defined as the ratio of the physical length of

the damage and wavelength of the A0 Lamb wave mode in the damage area (120582120582119889119889)

119898119898 =119889119889120582120582119889119889

=119889119889119889119889119907119907119889119889

(3)

In many cases the transmitted and reflected waves are overlapped and it is hard to decom-

pose and calculate the arrival times of them Matching pursuit method is an alternative way

to accurately decompose of the overlapped waves and estimate the parameters of each wave

such as arrival time (Fig 5) Detail explanation of the matching pursuit method is covered in

section 36

015 02 025 03-2

-1

0

1

2

Time (ms)

Am

plitu

de

OriginalApproximated

(a)

015 02 025 03-2

-1

0

1

2

Time (ms)

Ampl

itude

Wave 1Wave 2

(b)

Fig 5 (a) Example of the overlapped signal with comparing approximated signal using the matching pursuit method (b) Decomposed signal using the matching pursuit method The matching pursuit method provides a conventional way to decompose and identify each wave

- 9 -

33 Damage detection algorithm (Level 1)

The damage diagnosis is accomplished by extracting the additional reflected A0 Lamb

wave modes from damage due to multiple reflections If A0 Lamb wave modes are decom-

posed perfectly it is trivial that one transmitted A0 Lamb wave mode is always captured in

the response signal If damage exists other waves in the response signal behind the transmit-

ted A0 Lamb wave mode are presented Since this approach does not rely on previously col-

lected data the diagnosis is independent from environmental variation such as temperature

changes

It is reported that the dispersion of Lamb waves cause wave-packets to spread out in space

and time as they propagate through a structure [20] Thus the reflected waves from the dam-

age have longer pulse duration than the transmitted wave due to their longer propagating dis-

tance In this study the matching pursuit method is modified to capture more spread wave

than the transmitted A0 Lamb wave mode Detail explanation of the modified matching pur-

suit method is described in latter section

Let 1199041199041 and 1199041199042 as the pulse duration (scale) of the transmitted and additional reflected A0

Lamb wave mode respectively

ldquoIf 1199041199042 gt 1199041199041 damage exists

Otherwise there is no damagerdquo

- 10 -

34 Damage quantification algorithm (Level 3)

As mentioned in the previous section damage is quantized by estimating 119889119889 and 119907119907119889119889 using

pitch-catch method The procedure to estimate 119889119889 and 119907119907119889119889 is

(i) Measure 119897119897 119905119905 and 119905119905119889119889

(ii) Assume the range of 119889119889 (0 lt 119889119889 lt 119897119897)

(iii) Calculate 119907119907119894119894 and 119907119907119889119889 with a constraints (0 lt 119907119907119889119889 lt 119907119907119894119894) (Eq 4 and 5)

(iv) Estimate m-value and find a feasible range of 119889119889 and 119907119907119889119889

119907119907119894119894 =119897119897 minus 119889119889119905119905 minus 119905119905119889119889

(4)

119907119907119889119889 =119889119889119905119905119889119889

(5)

35 Damage localization algorithm (Level 2amp3)

For damage localization 1198971198971 is estimated by measuring the arrival time from the excitation

PZT to the entrance of delamination (1199051199051) with additional pulse-echo method implementation

The procedure is similar with the previous damage quantification algorithm

(i) Measure 119897119897 119905119905 119905119905119889119889 and 1199051199051

(ii) Assume the range of 1198971198971 (0 lt 1198971198971 lt 119897119897)

(ii) Compute possible combinations of 1198971198972 119889119889 and 119907119907119889119889 with constraints

(0 lt 119907119907119889119889 lt 119907119907119894119894 119889119889 gt 0) (Eq 5 6 and 7)

- 11 -

1198971198972 = 1198971198971119905119905 minus 1199051199051 minus 119905119905119889119889

1199051199051 (6)

119889119889 = 119897119897 minus 1198971198971 minus 1198971198972 (7)

Since a PZT for pulse-echo method acts as both an actuator and a sensor the obtained sig-

nal contains both the excitation and the response signal Thus the reflected wave from the

entrance of damage can be buried into the tail of excitation signal and cause error on arrival

time of reflected wave from damage Even in this case damage can be localized and quan-

tized using the constraints When first wave packet reflected from the entrance of damage is

buried into the excitation signal the next wave packet reflected from the exit of damage can

be utilized alternatively If the algorithm regards the second reflected wave as the first re-

flected wave there are no combinations of estimated variable meeting the constraints Then

the algorithm repeats the process by regarding the wave as the second reflected wave the

constraints can select proper combination of the estimated variables

The estimated values have range bounded by the constraints When additional information

of 119907119907119894119894 is utilized the damage can be characterized more accurately In this study three meth-

ods are proposed and verified as information level of 119907119907119894119894

Method 1 Solely based on the current data measurement (Reference-free)

Method 2 Limit the possible range of 119907119907119894119894 (Prior knowledge of the system)

Method 3 Measure 119907119907119894119894 (Additional data need to be collected)

Method 1 described in this section can be utilized without any information of 119907119907119894119894 Meth-

od 2 produces more accurate estimation results using prior knowledge of the system such as

- 12 -

119907119907119894119894 variation under temperature changes Method 3 gives unique estimated value using meas-

ured 119907119907119894119894 from other path in the same time In section 4 and 5 damage is characterized using

these three methods and the results are compared each another

36 Matching pursuit method considering the dispersion effect of Lamb

wave

Matching pursuit method has been developed for signal compression approximation and

characterization of obtained signal and function [21-23] The key concept of the matching

pursuit method is that any signal (function) can be decomposed into a linear expansion of

signals selected from a pool of basis signals (functions) called lsquodictionaryrsquo

Let 119891119891 be a function to be analyzed then 119891119891 can be approximated a set of functions select-

ed from the dictionary 119863119863 in order to best match This is done by successive approximations

with orthogonal projections Let1198921198921205741205740 isin 119863119863 the function 119891119891 can be decomposed into

119891119891 = lang1198911198911198921198921205741205740rang1198921198921205741205740 + 119877119877119891119891 (8)

where 119877119877119891119891 is the residual function after approximating 119891119891 using 1198921198921205741205740 1198921198921205741205740 is orthogonal to

119877119877119891119891 hence

1198911198912 = lang1198911198911198921198921205741205740rang2

+ 1198771198771198911198912 (9)

To minimize 119877119877119891119891 1198921198921205741205740 is chosen to maximize lang1198911198911198921198921205741205740rang The matching pursuit method is

an iterative algorithm that decomposes the residue 119877119877119891119891 by projecting it on a function of 119863119863

that matches 119877119877119891119891 almost at best as it was done for 119891119891 The residue 119877119877119899119899119891119891 is decomposed into

119877119877119899119899119891119891 = lang119877119877119899119899119891119891119892119892120574120574119899119899rang119892119892120574120574119899119899 + 119877119877119899119899+1119891119891 (10)

- 13 -

the decomposition is carried up to the order 119898119898 119891119891 can be expressed as

119891119891 = lang119877119877119899119899119891119891119892119892120574120574119899119899rang119892119892120574120574119899119899

119898119898minus1

119899119899=0

+ 119877119877119898119898119891119891 (11)

The original function 119891119891 is decomposed into a sum of dictionary elements that are chosen to

best match its residues

In this study the matching pursuit method is utilized to accurately decompose of the over-

lapped Gaussian pulses and estimate the parameters of the pulse which used for damage

characterization The 119894119894119905119905ℎ Gaussian pulse in the dictionary for approximating the obtained

signal is expressed as

119892119892120574120574119894119894 = 119891119891(119860119860119894119894 119906119906119894119894 119904119904119894119894 119889119889119894119894120601120601119894119894) (12)

where 119860119860119894119894119906119906119894119894119904119904119894119894119889119889119894119894 and 120601120601119894119894 represent amplitude arrival time scale (pulse duration) excita-

tion(central) frequency and phase of the Gaussian pulse respectively The parameters of the

pulse are estimated using a nonlinear regression based optimization method (Gauss-Newton

method) based on an initial guess of the parameters Detail procedure for estimating the pa-

rameters of the pulse is expressed in Hong et al [23]

Among the estimated parameters of the pulses pulse duration (scale) represents the disper-

sion effect If a wave-packet is spread out the estimated pulse duration is increased Using

this feature the matching pursuit method is modified to capture waves which have longer

pulse duration than transmitted A0 Lamb wave mode If the estimate pulse duration of a wave

is less than the transmitted wave the modified matching pursuit method regards it as error

- 14 -

CHAPTER 4 NUMERICAL SIMULATION

41 Composite modeling method

Composite materials can be modeled from nano to macro level depends on its engineering

purpose In general four methods are utilized for modeling of composite materials First

composite is modeled as a homogenous material expressed in terms of effective material

properties [24] The effective material properties are estimated using the modulus and the

volume fraction of each constituent theoretically (Rule of mixture) and measured experimen-

tally Second a model is established that each laminate which have own material properties

are totally bonded together In this model the stacking sequence and direction of each lami-

nates are considered additionally [25-27] Third method for composite modeling in micro

level called micromechanics [28-30] This model considers the interaction between fibers and

matrix in laminate Last multiscale modeling which the scope of modeling is extended to

molecular or quantum level of constituents [31-33] This model starts from molecular me-

chanics or molecular dynamics then the scale of model leads to the nano micro and macro

level by incorporating small scale models As the scale of models is smaller it provides more

logical way than other large scale of models while the computational cost is more expensive

The scope of this study is focused on the dynamic behavior of composite structure the beam

or plate model was utilized for the numerical simulation

- 15 -

42 Damage modeling method

Generally three damage models have been developed and utilized in the real applications

First the stiffness reduction model called as the equivalent element model was introduced

[17 34-35] This model is simply realized by replacing the region around damage area by a

region with adjusted material properties (Youngrsquos modulus shear modulus density etc)

globally Next the duplicated node model is done by disconnecting the element connectivity

assuming there is no interaction among the delaminated surface This model can be utilized

when the location and the extension of damage is important issue [34-36] Last the contact

mechanics model using the principles of contact mechanics the two interfacial forces (con-

tact resistance and sliding resistance) in the delaminated surface [37-38] In this model the

delaminated surfaces are modeled as nonlinear functions of the interlaminar normal strain

and the interlaminar shear strain This model is computationally expensive while provides

more logical way than other models There is however no universally accepted model for

general impact-induced damage in composites In the case of impact-induced damage it is

impossible to predict or estimate the location and number of delaminated layers Thus from

the uncertainty of impact-induced damage the equivalent element model can be considered

as the effective damage model

- 16 -

43 Numerical simulation description

TABLE I Material property of simulation model and test specimen

Tensile Strength 657 MPa Tensile Modulus 59 GPa Shear Strength 804 MPa Shear Modulus 24 Gpa Specific Gravity 16

PZT A

PZT C

PZT BDamage

3

250 125 125

500 Unit mm

Fig 6 Description of the numerical simulation model

The feasibility of the proposed damage characterization technique is verified by 2-D FEM

simulation using MSC PATRAN and NASTRAN Fig 6 presents the dimension PZTs and

damage location of the simulation model In the model 05mm x 05mm square shell element

was used for modeling of the A0 mode whose flexural wave characteristic The element size

was determined by considering the wavelength of the propagation Lamb wave modes The

material properties of the simulation model and damage cases are presented in Table I and

Table II respectively

PZT A and C act as both actuator and sensor for collecting pulse-echo signal They were

installed on the plate which two of them are both-side collocated for the A0 Lamb wave

- 17 -

mode decomposition [39-40] PZT B was only used as a sensor for collecting pitch-catch sig-

nal For simulating the excitation and sensing of PZTs the pin-force model was adopted that

an input force apples to the end nodes of excitation PZT and averaging element strains corre-

sponding sensing area A Gaussian pulse (tone-burst) 50 kHz central frequency was excited

as an input signal

TABLE II Damage cases for numerical simulation

119949119949120783120783 (mm) 119941119941 (mm) G value reduction ()

Case 1 650 200 90 darr

Case 2 675 150 90 darr

Case 3 675 150 93 darr

Case 4 900 150 90 darr

- 18 -

44 Numerical simulation results

015 02 025 03 035

-2

0

2

Time (ms)

Am

plitu

de

IntactDamage

Reflected wave from damage

(a) Pitch-catch signal

015 02 025 03 035

-2

0

2

Time (ms)

Am

plitu

de

IntactDamage

Reflected waves from damage

(b) Pulse-echo signal Fig 7 Comparison of A0 Lamb wave mode signals from numerical simulation between intact and damage cases Time delay attenuation and multiple reflections due to damage simulated successfully using the damage modeling method described in the previous section

Fig 7 shows that comparison of A0 Lamb wave mode signals between intact and one of

damage cases in the simulation The features of A0 Lamb wave mode passing through dam-

age area described in section 2 are successfully simulated Thus the equivalent element

method is proper for modeling damage in composites In the damage cases the additional

waves reflected from damage is observed in both pitch-catch and pulse echo signal

Table III presents the result of the proposed damage characterization algorithm using the

numerical simulation cases shown in Table II In lsquoMethod 2rsquo 119907119907119894119894 variation is assumed to plusmn

5 of the normal value (1190~1315ms) The normal value of 119907119907119894119894 for lsquoMethod 3rsquo is meas-

ured as 1253 ms In each case the damage is characterized using the proposed technique

Note that as the information of 119907119907119894119894 is specified the estimated value is closer to the exact

value

- 19 -

TABLE III Damage characterization for the numerical simulation models

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

Case 1

Exact 650 200

Method 1 248 50~70 79~414 160~834

Method 2 248 603~667 135~241 271~486

Method 3 248 635 188 379

Case 2

Exact 675 150

Method 1 193 60~75 01~251 37~651

Method 2 193 618~682 114~222 295~575

Method 3 193 65 168 435

Case 3

Exact 675 150

Method 1 233 60~75 66~302 140~649

Method 2 233 668~738 85~196 182~419

Method 3 233 703 140 301

Case 4

Exact 900 150

Method 1 204 75~100 48~349 118~856

Method 2 204 860~951 107~216 264~530

Method 3 204 906 162 397

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1253 ms 1190~1315 ms Measured 119907119907119894119894 1253 ms

- 20 -

CHAPTER 5 EXPERIMENTAL VERIFICATION

51 Overview

In this section experimental study is investigated so that the proposed damage characteri-

zation technique Overall experimental setup and test specimen are shown in Fig 8 and 9

The experimental setup consists of test specimens data acquisition (DAQ) system and self-

sensing circuit Self-sensing circuit was installed for pulse-echo measurement using PZTs

[40-41]

PZT A(PZT C) PZT B

500

500

140

75

250

Teflon Inserted Damage (d =25)

Unit mm

Thickness = 3

PZT A(PZT C) PZT B

500

500

150

90

250

Impact Location

Unit mm

Thickness = 3

(a) Teflon inserted damage specimen

(b) Impact-induced damage specimen

Fig 8 Test specimens for the experimental verification In the Teflon inserted damage Teflon tapes with 25mm diameter were inserted at the center of the specimen The impact-induced damage was introduced to the specimen using the impact test equipment

- 21 -

511 Specimen

Two carbon composite plates with a dimension of 500times500times3mm3 were manufactured by

Nexcomes Inc by stacking 12 layers of CF3327 3K prepreg (woven fabrics type) manufac-

tured by Hankuk Fiber Inc The material properties of the specimen are presented in Table I

One of the specimens Teflon tapes with 25mm diameter was inserted in every 3 layers at the

center of the specimen to mimic the damage The other specimen is utilized for the character-

ization of real impact-induced damage

512 Data acquisition (DAQ) system and PZT

The data acquisition (DAQ) system consists of an arbitrary waveform generator (AWG) a

high speed signal digitizer (DIG) and two multiplexers (MUX) Using the 16-bits AWG

tone-burst input signal with a plusmn10 peak-to-peak voltage were generated The response signal

was measured by the 14-bit DIG For the signal-to-noise ratio reduction both pitch-catch and

pulse-echo signals were measured 150 times and averaged in the time domain 3 identical

circular PZTs of 10mm diameter manufactured by APC International Ltd were installed on

the specimen A Gaussian pulse (tone-burst) 50 kHz central frequency was excited as an in-

put signal The excitation frequency was chosen in order to avoid overlapping of the S0 and

A0 Lamb wave modes and generating of the higher modes (S1 A1hellip)

- 22 -

PZT Actuator

PZT Sensor

AWG

DIG

DAQ System Specimen

SensingActuatingMUX

1

23

4

1 Sending signal to MUX

4 Sending signal to DIG

Self-SensingCircuit

18 Unit mm

10

(a)

(b)

Fig 9 Overall experimental setup (a) Data acquisition (DAQ) system and (b) Installed circu-lar PZT transducer on the specimen with its dimension

513 Impact test

The impact-induced damage was introduced to the specimen using the impact test equip-

ment presented in Fig 10 A 5kg mass with 10mm tip was dropped from 015m height onto

the same location of the specimen The signals were obtained at each number of impacts

Since the reflected wave is not detected at the first impact case the data obtained from the

second and third impact were utilized for the analysis

10

Unit mm (a)

(b)

Fig 10 (a) Impact test equipment for introducing impact-induced damage and (b) Tip of the impact tester with 10mm diameter

- 23 -

514 Temperature effect test

In reality structures are subjected to changing environmental conditions such as tempera-

ture that affect measured signal and cause false-alarm Thus the robustness of the proposed

damage characterization technique under temperature variation in investigated (Fig 11) The

specimens were placed inside a temperature chamber and a thermocouple was installed on

the specimen to measure temperature of surface of the specimen The signal was obtained

under three temperature cases (0 20 and 50degC) and the humidity was kept at 30

Temperature Chamber

Test Specimen

Fig 11 Test specimen in temperature chamber

- 24 -

52 Teflon inserted damage

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Reflected wave from damage

(a)

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Transmitted WaveReflected Wave

Detected reflected wave

(b)

Fig 12 A0 Lamb wave mode obtained using pitch-catch method from the Teflon inserted damage specimen (a) The reflected wave from damage is shown (b) The Teflon inserted damage is detected by capturing the reflected wave

Fig 12 shows the obtained pitch-catch signal from the Teflon inserted damage specimen

and damage diagnosis result The Teflon inserted damage is detected successfully by captur-

ing the reflected wave Table IV presents the result of the Teflon inserted damage characteri-

zation 119907119907119894119894 is measured as 1213ms and the variation is assumed to plusmn 5 of the normal val-

ue (1152~1273ms) The results indicate that the proposed damage characterization approach

works well in real application

TABLE IV Damage characterization for the Teflon inserted damage specimen

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

Exact 635 25

Method 1 123 60~75 73~338 298~1220

Method 2 123 619~662 229~310 934~1188

Method 3 123 631 284 1161

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1213 ms

- 25 -

53 Impact-induced damage

11 115 12 125 13-4

-2

0

2

4

Time (ms)

Vol

tage

(mV

)

IntactImpact2Impact3

Reflected wave from damage

(a) Pitch-catch signal

11 115 12 125 13 135

-05

0

05

1

Time (ms)

Volta

ge (m

V)

IntactImpact2Impact3

Boundary reflection

Reflected wave from damage

(b) Pulse-echo signal

Fig 13 The signal obtained from the test specimen subjected to impact The impact was in-troduced at the same location of the specimen and the signal was obtained after each impacts As the number of impact is increased the transmitted A0 Lamb wave mode is delayed and attenuated

The obtained signals from impact-induced damage specimen are presented in Fig 13 It is

observed that as the number of impact is increased the transmitted A0 Lamb wave mode is

more delayed because the wave velocity in the damage area is reduced Fig 14 shows the

false-alarm test result whether the proposed damage detection algorithm works even in pris-

tine condition of the test specimen It is shown that no false alarm is occurred because no ad-

ditional reflected wave is captured

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Transmitted Wave

(a) Transmitted wave

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Reflected Wave

No reflected wave

(b) Reflected wave

Fig 14 False alarm test for the proposed damage detection algorithm in pristine condition of the specimen Reflected wave from damage is not captured false alarm is not occurred

- 26 -

TABLE V Damage characterization for the impact-induced damage specimen

of Impacts m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

2

(1470 J)

Method 1 080 65~90 75~246 470~1157

Method 2 080 801~885 114~245 711~1134

Method 3 080 843 179 1122

3

(2205 J)

Method 1 148 50~100 06~380 21~1134

Method 2 148 758~838 249~368 843~1125

Method 3 148 798 308 1044

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1213 ms

Table V presents the characterization results of the impact-induced damage case Here the

measured 119907119907119894119894 and the assumed variation are same as the artificial damage case It is observed

that as the number of impact is increased 119889119889 is increased while 119907119907119889119889 is reduced Thus it is

validated that the proposed damage characterization approach estimates the increment of the

damage well

54 Temperature effect test

541 Teflon inserted damage

11 115 12 125 13-10

-5

0

5

10

Time (ms)

Vol

tage

(mV

)

0 oC20 oC50 oC

(a) Pitch-catch signal

11 115 12 125 13 135

-1

0

1

2

3

Time(ms)

Volta

ge (m

V)

0 oC20 oC50 oC

(b) Pulse-echo signal

Fig 15 Signals obtained from the Teflon inserted damage specimen under temperature varia-tion Time delay and amplitude variation are occurred

Fig 15 shows the obtained signals from the Teflon inserted damage specimen subjected to

- 27 -

temperature variation Time delay and amplitude variations are occurred there is no addition-

al damage is introduced The analyzed results of the obtained signals are reported in Table VI

The measured 119907119907119894119894 is 1247 ms at 0 degC 1213 ms at 20 degC and 1168 ms at 50 degC respectively

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms (1152~1273 ms)

The estimated results are also changed under temperature variations the fluctuation of esti-

mated range and exact value is small

TABLE VI Temperature effect test for the Teflon inserted damage specimen

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

0 degC

Exact 635 25

Method 1 113 60~70 79~268 352~1291

Method 2 113 560~619 233~344 1034~1172

Method 3 113 606 257 1141

20 degC

Exact 635 25

Method 1 123 60~75 73~338 298~1220

Method 2 123 619~662 229~310 934~1188

Method 3 123 631 284 1161

50 degC

Exact 635 25

Method 1 137 60~75 72~328 263~1228

Method 2 137 622~687 183~299 666~1086

Method 3 137 642 264 960

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1247 ms at 0 degC 1213 ms at 20 degC and 1168 ms at 50 degC

- 28 -

542 Impact-induced damage

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Vol

tage

(mV

)

Transmitted WaveReflected Wave

No reflected wave

(a) 0 degC

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Transmitted WaveReflected Wave

No reflected wave

(b) 50 degC

Fig 16 False alarm test under temperature variations It is shown that proposed damage char-acterization algorithm is robust under temperature variations

The false alarm test in pristine condition of the impact-induced damage specimen under

temperature variation was investigated (Fig 16) It is validated that the proposed damage di-

agnosis algorithm is robust under temperature variations Fig 17 shows the obtained signals

from the impact-induced damage (impacted 3 times) specimen under temperature variation

The analyzed results of the obtained signals are reported in Table VII The measured 119907119907119894119894 in

each temperature and the assumed variation are same as the artificial damage case

11 115 12 125 13-5

0

5

Time (ms)

Volta

ge (m

V)

0 oC20 oC50 oC

(a) Pitch-catch signal

11 115 12 125 13 135

-1

0

1

2

3

Time(ms)

Vol

tage

(mV

)

0 oC20 oC50 oC

(b) Pulse-echo signal

Fig 17 Signals obtained from impact-induced damage specimen (Impact 3) under tempera-ture variation

- 29 -

TABLE VII Temperature effect test for the impact induced damage specimen (Impact 3 case)

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

0 degC

Method 1 135 75~90 72~310 267~1148

Method 2 135 705~779 264~381 978~1126

Method 3 135 763 289 1071

20 degC

Method 1 148 50~100 06~380 21~1134

Method 2 148 758~838 249~368 843~1125

Method 3 148 798 308 1044

50 degC

Method 1 153 75~100 53~415 174~1227

Method 2 153 804~889 214~336 702~1103

Method 3 153 815 320 1050

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1247 ms at 0 degC 1213 ms at 20 degC and 1168 ms at 50 degC

- 30 -

CHAPTER 6 CONCLUDING REMARKS

61 Executive summary

In this study the reference-free damage detection localization and quantification tech-

nique in a multi layered composite plate is proposed Additional reflected A0 Lamb wave

mode from the damage due to multiple reflections in damage area is used as a feature for

characterizing the damage First damage is characterized in terms of a damage index (m-

value) defined as the ratio of the physical length of the damage to the wavelength of the A0

Lamb wave mode Then the range of possible damage location length and severity are esti-

mated with constraints This procedure can be accomplished without prior baseline data To

realize the proposed technique the matching pursuit technique is modified for reduction of

the error by considering the dispersive characteristic of Lamb wave The matching pursuit

method estimates the arrival time of each wave packets accurately even two wave packets are

overlapped The numerical simulations and experimental tests including Teflon tape inserted

and impact-induced damage are investigated for the feasibility of the proposed technique

Moreover the effectiveness of the proposed technique under changing temperature is exper-

imentally verified

- 31 -

62 Suggestion for further study

The limitations of this study and the possible improvement and further studies are listed as

follow

1 Measurement of impact-induced damage length using conventional NDT technique

Additional experimental verification for impact-induced damage characterization is

needed by measuring the real length of damage using conventional NDT technique

such as ultrasonic scan or infrared camera

2 Limitation of both side collocated PZT installation In this study excitation PZTs are

installed both side and collocated for decomposing A0 Lamb wave mode However it

can produce decomposition error due to imperfection of PZT location and their installa-

tion Moreover in many cases assess to both surface of structures is not allowed such

as aircraft applications To improve this limitation a reference-free damage characteri-

zation using single side PZTs will be considered

3 Complex structure application In this study damage is characterized in simple plate-

like structure When the monitored structure becomes complex the response signals al-

so becomes complex due to its geometry and boundary Thus the investigations for

complex structure application such as stiffener spar and curvature are needed

- 32 -

REFERENCES

[1] Cawley P and Alleyne D (1996) ldquoThe use of Lamb waves for the long range inspection of large structuresrdquo Ultrasonics 34 pp287-290

[2] Ghosh T Kundu T and Karpur P (1998) ldquoEfficient use of Lamb waves for detecting defects in large platesrdquo Ultrasonics 36 pp791-801

[3] Guirguitui V (2003) ldquoLamb wave generation with piezoelectric wafer active sensors for structural health monitoringrdquo SPIE Conference on Smart Structures and Materials San Diego CA 5391 pp111-422

[4] Ragvahan A and Cesnik CE S (2004) ldquoModeling of piezoelectric-based Lamb wave generation and sensing for structural health monitoringrdquo SPIE Conference on Smart Struc-tures and Materials San Diego CA 5391 pp419-430

[5] Rose JL Pilarski A and Dirit JJ (1993) ldquoAn approach to guided wave mode selection for inspection of laminated platerdquo Journal of Reinforced Plastics and Composites 12(5) pp536-544

[6] Guo N and Cawley P (1993) ldquoThe interaction of Lamb waves with delaminations in composite laminatesrdquo Journal of Acoustical Society of America 94(4) pp 2240-2246

[7] Valdez SHD and Soutis C (2002) ldquoReal-time nondestructive evaluation of fiber com-posite laminates using low-frequency Lamb wavesrdquo Journal of Acoustical Society of Amer-ica 111(5) pp 2026-2033

[8] Yang M and Qiao P (2005) ldquoModeling and experimental detection of damage in virous materials using the pulse-echo method and piezoelectric sensoractuatorsrdquo Smart Materials and Structures 14 pp1083-1100

[9] Lestari W and Qiao P (2005) ldquoApplication of wave propagation analysis for damage identification in composite laminated beamsrdquo Journal of Composite Materials 39(22) pp1967-1984

[10] Kessler S S Spearing S M and Soutis C (2002) ldquoDamage detection in composite ma-terials using Lamb wave methodsrdquo Smart Materials and Structures 11 pp269-278

[11] Sohn H Park G Wail J R Limback N P and Farrar C R (2004) ldquoWavelet-based active sensing for delamination detection in composite structuresrdquo Smart Materials and Structures 13 pp 153-160

[12] Petculescu G Krishnaswamy S and Achenback J D (2008) ldquoGroup delay measure-ments using modally selective Lamb wave transducers for detection and sizing of delamina-tions in compositesrdquo Smart Materials and Structures 17

[13] Sohn H Park H W Law K H and Farrar C R (2007) ldquoDamage detection in compo-

- 33 -

site plates by using an enhanced time reversal methodrdquo Journal of Aerospace Engineering (ASCE) pp 141-151

[14] Sohn H (2007) ldquoEffects of environmental and operational variability on structural health monitoringrdquo Philosophical Transactions of the Royal Society Ardquo 365 pp539-560

[15] Rytter A (1993) ldquoVibration based inspection of civil engineering structuresrdquo PhD The-sis Department of Building Technology and Structural Engineering University of Aalborg Aalborg Denmark

[16] Hayashi T and Kawashima K (2002) ldquoMultiple reflections of Lamb waves at a delami-nationrdquo Ultrasonics 40 pp193-197

[17] Wang C H and Rose L R F (2003) ldquoWave reflection and transmission in beams con-taining delamination and inhomogeneityrdquo Journal of Sound and Vibration 264 pp 851-872

[18] Ramadas C Balasubramaniam K Joshi M and Krishnamurthy C V (2009) ldquoInterac-tion of the primary anti-symmetric Lamb mode (Ao) with symmetric delaminations numer-ical and experimental studiesrdquo Smart Materials and Structures 18

[19] Ramadas C Balasubramaniam K Joshi M and Krishnamurthy C V (2010) ldquoInterac-tion of guided Lamb waves with an asymmetrically located delamination in a laminated composite platerdquo Smart Materials and Structures 19

[20] Wilcox P Lowe M and Cawely P (2001) ldquoThe effect of dispersion on long-range in-spection using ultrasonic guided wavesrdquo NDTampE International 34 pp1-9

[21] Mallat S G and Zhang Z (1993) ldquoMatching pursuits with time-frequency dictionariesrdquo IEEE Tansactions on Signal Processing 41(12) pp3397-3415

[22] Zhang G Zhang S and Wang Y (2000) ldquoApplication of adaptive time-frequency de-composition in ultrasonic NDE of highly scattering materialsrdquo Ultrasonics 38 pp 961-964

[23] Hong J C Sun K H and Kim Y Y (2005) ldquoThe matching pursuit approach based on the modulated Gaussian pulse for efficient guided-wave damage inspectionrdquo Smart Mate-rials and Structures 14 pp548-560

[24] Herakovich C T (1997) Mechanics of Fibrous Composites Wiley 480 pages

[25] Reddy J N (1985) ldquoA review of the literature on finite-element modeling of laminated composite platesrdquo The Shock and Vibration Digest 17(4) pp 3-8

[26] Reddy J N (2003) Mechanics of Laminated Composite Plates and Shells Theory and Analysis CRC Press 856 pages

[27] Palacz M Krawczuk M and Ostachowicz W (2005) ldquoThe spectral finite element mod-el for analysis of flexure-shear coupled wave propagation Part 1 laminated multilayer

- 34 -

composite beamrdquo Composite Structures 68 pp37-44

[28] Zhang Y C and Harding J (1990) ldquoA numerical micromechanics analysis of the me-chanical properties of a plain weave compositerdquo Computer amp Structures 36(5) pp 839-844

[29] Nemat-Nasser S and Hori M (1999) Micromechanics Overall Properties of Heteroge-neous Materials Elsevier 810 pages

[30] Scida D Aboura Z Benzeggagh M L and Bocherens E (1999) ldquoA micromechanics model for 3D elasticity and failure of woven-fibre composite materialrdquo Composites Science and Technology 59 pp 505-517

[31] Thomas Y and Wu X H (1997) ldquoA multiscale finite element method for elliptic prob-lems in composite materials and porous mediardquo Journal of Computational Physics 134 pp 169-189

[32] Lee K H Moorthy S and Ghosh S (1999) ldquoMultiple scale computational model for damage in composite materialsrdquo Computer Methods in Applied Mechanics and Engineer-ing 172 pp 175-201

[33] Namilae S and Chandra N (2005) ldquoMultiscale model to study the effect of interfaces in carbon nanotube-based compositesrdquo Journal of Engineering Materials and Technology (ASME) 127 pp222-232

[34] Adams D E (2007) Health monitoring of Structural Materials and Components Wiley 460 pages

[35] Gopalakrishnan S (2009) ldquoModeling Aspects in Finite Elementrdquo Encyclopedia of Struc-tural Health Monitoring 2 pp791-809

[36] Palacz M Krawczuk M and Ostachowicz W (2005) ldquoThe spectral finite element mod-el for analysis of flexure-shear coupled wave propagation Part 2 Deminated multilayer composite beamrdquo Composite Structures 68 pp45-51

[37] Baik J M and Thopson R B (1984) ldquoUltrasonic scattering from imperfect interfaces A quasi-static modelrdquo Journal of Nondestructive Evaluation 4(3) pp177-196

[38] Luo H and Hanagud S (2000) ldquoDynamics of delamincated beamsrdquo International Jour-nal of Solids and Structures 37 pp1501-1519

[39] Kim S B and Sohn H (2007) ldquoInstantaneous reference-free crack detection based on polarization characteristics of piezoelectric materialsrdquo Smart Materials and Structures 16 pp2375-2387

[40] An Y K and Sohn H (2010) ldquoInstantaneous crack detection under varying temperature and static loading conditionsrdquo The Special Issue of Structural Control and Health Monitor-ing 17(7) pp 730-741

- 35 -

[41] Lee S J and Sohn H (2006) ldquoActive self-sensing scheme development for structural health monitoringrdquo Smart Materials and Structures 15(6) pp 1734-1746

- 36 -

요 약 문

복합재료 손상의 무기저 진단 및 정량화

복합재료는 재료의 경량성 내구성 고강도의 특성으로 인하여 항공기에서부터

교량에까지 널리 사용되고 있다 하지만 외부의 충격으로 인한 복합재료 내부의

손상은 재료의 강도를 감소시켜 구조물의 안정성을 저하시키는 요인이 된다 하지만

복합재료의 손상은 재료 내부에서 발생하여 육안으로는 식별하기도 힘들며 표면에서

보이는 것만으로 내부의 손상 정도를 판단하기는 불가능하다는 단점을 가지고 있다

본 연구에서는 손상 내부에서 반사로 인해 발생하는 파를 이용하여 기저자료를

사용하지 않고 특정 시점에서 얻은 자료만으로 복합재료의 손상을 진단하고 정량화

할 수 있는 기법을 제안한다 제안된 기법의 적용성을 검증하기 위하여 유한요소법을

이용한 수치해석을 수행하였고 Teflon tape 이 삽입된 시편과 실제 충격하중으로

인한 손상을 입은 시편을 이용한 실험을 진행하였다 수치해석 및 실험 결과 제안된

기법이 복합재료의 손상을 성공적으로 진단 및 정량화할 수 있음을 확인하였다 또한

다양한 외부 온도 환경에서 제안된 기법의 환경영향 실험을 수행하였고 다양한

환경에서 손상의 진단 및 정량화가 성공적으로 이루어짐을 검증하였다

핵심어 무기저 손상진단 및 정량화 압전센서 복합재료 유한요소법 Matching pursuit

method

- 37 -

감 사 의 글

기대와 두려움을 안고 시작한 석사과정이 어느덧 2 년 이라는 세월 속에서 추억으로 자리잡아

가고 있습니다 학부생 기간을 포함한 지난 3 년 동안 SHM 이라는 새로운 분야를 접하고 연구를

수행 할 수 있도록 도와주신 고마운 분들이 있어 그 감사를 표하고자 합니다

우선 저의 많은 부족한 점을 일깨워 주시고 연구에 대한 뜻을 더욱 확고히 세울 수 있게 해

주신 지도교수님이신 손 훈 교수님께 깊은 감사의 뜻을 전합니다 짧은 기간이었지만 교수님의

도움으로 많은 것을 배우고 느끼고 깨달을 수 있었습니다 또한 본 연구를 수행할 수 있도록 지

원해주신 국토해양부 (U-City 석 박사과정 지원사업)에 깊은 감사 드립니다

때로 힘들고 지칠 때 항상 저를 믿고 응원해 주시는 가족이 있었기에 지금의 글을 쓸 수 있었

습니다 또한 지금 연구실에서 함께 생활하고 있는 윤규형 민구형 철민이형 현석이형 진열이형

호민이형 병진이 선혜누나와 성대에 계신 창길이형 미국에 계신 의재형과 Debaditiya 에게도 감

사의 말씀을 전합니다 그리고 모두 나열할 수는 없지만 카이스트 생활의 활력소가 되어준 카이

스트 건설 및 환경공학과 선후배들 직원 분들과 SOS 멤버들에게도 감사를 표합니다

앞서 거론한 많은 이들의 도움이 없었다면 저 또한 지금 이 자리에 있을 수 없었을 것입니다

많이 부족하기에 더 노력하고 배우고자 했던 초심을 다시 한번 생각하며 더 큰 결실을 맺기 위해

앞으로 나아가겠습니다 도움을 주신 모든 분들께 이 감사의 글을 바칩니다

2011년 6월 카이스트에서 임형진 올림

- 38 -

CURRICULUM VITAE

Hyung Jin Lim

Research Assistant

The Civil and Environmental Engineering Department

Korea Advanced Institute of Science and Technology (KAIST)

291 Daehak-Ro Guseong-Dong Yuseong-Gu

Daejeon Republic of Korea 305-701

Phone (82)+42-350-3665 Fax (82)+42-350-3610

Email limhj87gmailcom

RESEARCH INTERESTS

Damage Detection Localization and Quantification on Composite and Metallic Structures Pattern Recognition Finite Element Modeling (FEM)

EDUCATION

2009 ndash 2011 MS Civil and Environmental Eng Korea Advanced Institute of Science and Technology (KAIST) Korea Rep

2006 ndash 2009 BS Civil and Environmental Eng Korea Advanced Institute of Science and Technology (KAIST) Korea Rep

JOURNAL PUBLICATIONS The corresponding author is underlined 1 Hyung Jin Lim Hoon Sohn ldquoReference-free damage detection localization and quantification

using in compositerdquo in preparation for Journal of the Acoustical Society of America 2011

2 Chul Min Yeum Hoon Sohn Jeong Bum Ihn Hyung Jin Lim ldquoInstantaneous delamination detection in a composite using a dual piezoelectric transducer networkrdquo Submitted to Compo-sites Science and Technology 2011

3 Yun-Kyu An Hyung Jin Lim Min Koo Kim Jin Yeol Yang Hoon Sohn Chang Guen Lee ldquoInvestigation of applicability of local reference-free damage detection techniques to in-situ

- 39 -

bridgesrdquo Submitted to ASCE Journal of Structural Engineering 2011

4 Hyung Jin Lim Min Koo Kim Hoon Sohn Chan Yik Park ldquoImpedance based damage detec-tion under varying temperature and loading conditionsrdquo Submitted to NDTampE International 2011

5 Hyung Jin Lim Hoon Sohn ldquoQuantitative estimation of transmitted and reflected lamb waves at discontinuityrdquo Journal of the Korean Society for Nondestructive Testing Vol30 No 4 pp1-8 2010

CONFERENCE PROCEEDINGS

The corresponding author is underlined 1 Hyung Jin Lim and Hoon Sohn ldquoReference-free damage detection localization and quantifi-

cation in compositesrdquo ASCE Engineering Mechanics Institute Conference Boston June 2-4 2011

2 Hyung Jin Lim and Hoon Sohn ldquoReference-free delamination quantification within multi-layer composite materialsrdquo KSNT Spring Conference May 26-27 Gyeong-Ju Korea 2011

3 Min Koo Kim Hyun Jin Lim Hoon Sohn Chan Yik Park ldquoImpedance-based bolt loosening detection under varying temperature and loadingrdquo Asian Pacific Workshop on Structural Health Monitoring Tokyo Japan November 30-December 2 2010

4 Hyung Jin Lim Chul Min Yeum and Hoon Sohn ldquoModeling of Impact-induced Delamination in a Multilayer Composite Platerdquo The 23rd KKCNN Symposium on Civil Engineering Taipei (Taiwan) November 13-15 2010

5 Yun-Kyu An Hyung Jin Lim and Hoon Sohn ldquoTemperature independent Delamination Detec-tion using data normalizationrdquo The 3rd International Conference on Multi-Functional Materi-als and Structures Chonbuk National Univ Jeonju Korea September 14-18 2010

6 Hoon Sohn Yun-Kyu An Chang Gil Lee Min Koo Kim Hyung Jin Lim ldquoApplication of online NDT techniques to decommissioned bridge testingrdquo the 5th European Workshop on Structural Health Monitoring Sorrento Italy June 29-July 02 2010

7 Min Koo Kim Hyung Jin Lim Hoon Sohn Chan Yik Park ldquoImpedance based damage diag-nosis of a complex composite aircraft wing under changing loading and temperature condi-tionsrdquo the 5th European Workshop on Structural Health Monitoring Sorrento Italy June 29-July 02 2010

8 Yun-Kyu An Min Koo Kim Hyung Jin Lim and Hoon Sohn ldquoValidations of reference-free crack detection techniques through a decommissioned bridge testrdquo 2010 COSEIK Annual Con-ference Jeju Korea April 8-9 2010

- 40 -

PARTICIPANT PROJECTS

1 Development of Smart Steel Members for Online Steel Structure Monitoring POSCO Corporation (Funded 41000000 KRW (41000 USD) for 010511 to 123111)

2 Development of Enabling Technologies for Structural Health Monitoring Korean Agency for Defense Development (Funded 240000000 KRW (240000 USD) out of total 3200000000 KRW (3200000 USD) for 060108-053111)

3 Development of On-board SHM Technologies for Composite Air Vehicles The Boeing Company (Funded 400000 USD (400000000 KRW) for 080108 to 073111)

EDUCATIONAL EXPERIENCES

Fall 2009

Teaching Assistant CE207 ldquoElementary Structural Engineering and Laboratoryrdquo KAIST 3 unit undergraduate course

EXPERIMENTAL EXPERIENTCES

2010 Monitoring of Real Aircraft Composite Wing Segment Korea Agency for Defense De-velopment (ADD) Daejeon Korea Real time monitoring of composite air vehicles fund-ed by Korea Agency for Defense Development (ADD)

2009-2011

Field Monitoring of Real Bridge Structures Gimpo (Ramp G bridge) amp Icheon (Sam-seung bridge) Korea US-Korea collaborative research for bridge monitoring testbeds

HONORS amp AWARDS

Scholarships for outstanding students granted KAIST (2007-2009) The three prominent students in the civil and environmental engineering department of KAIST were selected as recipients This scholarship provided 2000$ per year for three years

- 41 -

Page 13: 복합재료 손상의 무기저 진단 및 정량화 - KAISTssslab.kaist.ac.kr/article/pdf/Thesis_2.pdf · 1 Declaration of Ethical Conduct in Research: I, as a graduate student

damage with reasonable precision

Kesseler et al proposed a technique monitoring the wavelet coefficients of transmitted ul-

trasonic signal amplitude as a potential way to damage detection [10] One drawback of this

technique is that the variability in coupling and the sensitivity of installed PZT transducers

can induce changes in the amplitude and energy of the signal that are cause false-alarm Oth-

er wavelet based delamination detection and localization technique developed using PZT

sensor network and outlier analysis [11]

Damage detection and quantification in composite plate was also investigated by group de-

lay measurement due to damage [12] However it is assumed that the group velocity in dam-

age area as the group velocity propagates half thickness of the structure in dispersion curve in

the quantification procedure Thus the damage quantification was investigated only Teflon

inserted damage case

Sohn et al proposed a damage detection technique which does not rely on any past base-

line signals to assess damage in composite plates by using an enhanced time reversal method

[13] The damage is assessed using the violation of time reversibility when nonlinearity is

caused by the damage along a direct wave path Thus it allows instantaneous identification

of delamination without requiring the baseline signal for comparison

- 2 -

13 Objective and uniqueness

In this study a reference-free damage characterization technique including detection lo-

calization and quantification of impact-induced damage in composite materials is developed

The general damage characterization can be categorized into the following levels [14]

Level 0 Define damage

Level 1 Determine the presence of damage within the structure

Level 2 Locate the regions of damage

Level 3 Quantify the damage

Level 4 Predict the remaining service life of the structure

Through the proposed technique from level 1 to level 3 can be achieved

Most of common approaches of damage characterization techniques described in the pre-

vious section use prior obtained baseline data However these approaches can cause false

alarm due to the effect of environmental variations such as temperature and ambient vibra-

tions [15]

The uniqueness of this study is

1 Damage is characterized without using prior measured baseline data

2 Robustness of the proposed technique under temperature variation is investigated

3 Additional waves reflected from inside damage have been used in addition to transmit-

ted and reflected waves

- 3 -

14 Organization

The organization of this thesis is summarized as follows

Chaper1 A general introduction including motivation literature review objective and

uniqueness of this study is described

Chaper2 Interactions between Lamb wave mode and damage in composites such as time

delay attenuation and multiple reflections are summarized Damage index (m-value) for

damage characterization is defined

Chaper3 Algorithms for the reference-free damage characterization technique using m-

value are developed Modified matching pursuit for this study is also described

Chaper4 Verification of the algorithm using numerical simulation is presented Various

modeled damage cases are characterized

Chaper5 Experimental verifications are described Two damage cases (Teflon inserted and

impact-induced damage) are investigated for the feasibility of the proposed damage charac-

terization technique Temperature effect test using temperature chamber is also carried out

Chaper6 Conclusion and summary of the present study is reported In advance suggestions

and future work recommendations are represented

- 4 -

CHAPTER 2 INTERACTION OF LAMB WAVE WITH DAMAGE

21 A0 Lamb wave mode delay and attenuation in the time domain

Time delay and attenuation of the A0 mode Lamb wave after passing through the damage

area due to effective thickness reduction were reported in previous researches [812] Moreo-

ver the attenuation of the A0 Lamb wave mode is occurred by degrading of shear modulus of

in the damage area However the time delay and amplitude variation of the A0 Lamb wave

mode also occurred even when the structure is health condition by the material property

changing due to temperature variation of the structure (Fig 1) Thus conventional damage

characterization methods using the time delay and attenuation of A0 Lamb wave mode as a

feature can cause false-alarm in real application

11 115 12 125 13-10

-5

0

5

10

Time (ms)

Vol

tage

(mV

)

IntactDamage

Reflected wave from Damage

(a)

11 115 12 125 13-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

0 oC20 oC50 oC

(b)

Fig 1 Lamb wave propagation characteristics (a) A0 Lamb wave mode time delay and atten-uation after passing through a damage (b) The time delay and amplitude variation of A0 Lamb wave mode under temperature variation in pristine condition of structure Because of this conventional damage detection techniques in composites have limitations for real struc-ture application

- 5 -

22 Multiple reflections in the damage area

Standing wave patterns due to multiple reflections of A0 Lamb wave mode in the damage

area were observed through numerical simulations and ultrasonic wave-field imaging using a

scanning laser vibrometer [16] The formation of standing waves in damage is illustrated

through Fig 2 When the incident waves enter the damage area a significant portion of the

waves are reflected back from the exit Then the reflected waves travel through the damage

and undergo reflection again at the original entrance As a consequence of multiple reflec-

tions at the entrance and the exit a considerable amount of ultrasonic energy is trapped inside

the damage area [17-19]

Multiple Reflections

Incident Wave Transmitted Wave

Damage

Fig 2 Multiple reflections in the damage area The multiple reflections of A0 Lamb waves in the damage area are utilize as a feature for characterization of the damage in this study

- 6 -

CHAPTER 3 THEORETICAL DEVELOPMENT

31 Overview

Damage

l = l1 + d + l2

d = vdtd l1 = vit1 l2 = vit2

t = t1 + td + t2

Excitation and Sensing PZT Sensing PZT

l

Transmitted Wave

Reflected Waves

Incident Wave

Fig 3 Parameters and notations of the proposed damage characterization technique Damage is detected by identifying the additional reflected wave from the damage due to multiple re-flections (dashed lines) Localization and quantification of the damage is carried out by measuring the arrival time of transmitted and reflected waves from the damage

In this section the reference-free damage characterization algorithm is proposed Damage

diagnosis and quantification are achieved by identifying the additional reflected waves due to

multiple reflections in the damage area (dashed lines in Fig 3) Moreover damage is local-

ized using additional information from pulse-echo method implement For the localization

and the quantification of damage the variables such as the path length and A0 Lamb wave

mode velocity in damage area are defined (Fig 3) For example 119897119897 is expressed as

119897119897 = 1198971198971 + 119889119889 + 1198971198972 (1)

where 1198971198971 119889119889 and 1198971198972 are the distance from the excitation PZTs to the entrance of damage the

damage length and the distance from the exit of damage to the sensing PZT respectively

- 7 -

Moreover 119907119907119894119894 and 119907119907119889119889 are also defined as the A0 Lamb wave mode velocity in pristine area

and damage area respectively Note that the whole procedure of the damage characterization

is accomplished based solely on the signals obtained from the current state of the structure

32 Damage index (m-value)

The proposed reference-free damage characterization technique utilizes the waves reflect-

ed additionally from the damage area as a feature The arrival time difference of the transmit-

ted and reflected A0 Lamb wave modes reflected from the exit and the entrance of the delam-

ination area one times each (∆119905119905) contains the information of the physical length of the dam-

age (119889119889) and A0 Lamb wave mode velocity in the damage area (119907119907119889119889) The wave velocity in the

damage area which is related to the shear modulus (G) represents the severity of the damage

∆119905119905 = 2119905119905119889119889 =2119889119889119907119907119889119889

(2)

However based on the measured data available the 119889119889 and 119907119907119889119889 (or G) cannot be uniquely

determined Fig 4 shows two FEM models which have different physical length and wave

velocity (G) of the damage Detail explanation of delamination modeling technique will be

covered in section 4 The similar signal in the time domain and ∆119905119905 are obtained from the

two different damage cases

- 8 -

Damage 1

Damage 2

20

Unit mm15

90 G value reduction

943 G value reduction

m-value = 248

m-value = 248

015 02 025 03 035

-2

-1

0

1

2

Time (ms)

Ampl

itude

Damage 1Damage 2

Fig 4 Two different FEM models which have different 119889119889 and 119907119907119889119889 (or G) This represents that the damage cannot be characterized uniquely based on the measured data available

Thus in this study the damage index (m-value) which can be uniquely defined based on

the available data is introduced The m-value is defined as the ratio of the physical length of

the damage and wavelength of the A0 Lamb wave mode in the damage area (120582120582119889119889)

119898119898 =119889119889120582120582119889119889

=119889119889119889119889119907119907119889119889

(3)

In many cases the transmitted and reflected waves are overlapped and it is hard to decom-

pose and calculate the arrival times of them Matching pursuit method is an alternative way

to accurately decompose of the overlapped waves and estimate the parameters of each wave

such as arrival time (Fig 5) Detail explanation of the matching pursuit method is covered in

section 36

015 02 025 03-2

-1

0

1

2

Time (ms)

Am

plitu

de

OriginalApproximated

(a)

015 02 025 03-2

-1

0

1

2

Time (ms)

Ampl

itude

Wave 1Wave 2

(b)

Fig 5 (a) Example of the overlapped signal with comparing approximated signal using the matching pursuit method (b) Decomposed signal using the matching pursuit method The matching pursuit method provides a conventional way to decompose and identify each wave

- 9 -

33 Damage detection algorithm (Level 1)

The damage diagnosis is accomplished by extracting the additional reflected A0 Lamb

wave modes from damage due to multiple reflections If A0 Lamb wave modes are decom-

posed perfectly it is trivial that one transmitted A0 Lamb wave mode is always captured in

the response signal If damage exists other waves in the response signal behind the transmit-

ted A0 Lamb wave mode are presented Since this approach does not rely on previously col-

lected data the diagnosis is independent from environmental variation such as temperature

changes

It is reported that the dispersion of Lamb waves cause wave-packets to spread out in space

and time as they propagate through a structure [20] Thus the reflected waves from the dam-

age have longer pulse duration than the transmitted wave due to their longer propagating dis-

tance In this study the matching pursuit method is modified to capture more spread wave

than the transmitted A0 Lamb wave mode Detail explanation of the modified matching pur-

suit method is described in latter section

Let 1199041199041 and 1199041199042 as the pulse duration (scale) of the transmitted and additional reflected A0

Lamb wave mode respectively

ldquoIf 1199041199042 gt 1199041199041 damage exists

Otherwise there is no damagerdquo

- 10 -

34 Damage quantification algorithm (Level 3)

As mentioned in the previous section damage is quantized by estimating 119889119889 and 119907119907119889119889 using

pitch-catch method The procedure to estimate 119889119889 and 119907119907119889119889 is

(i) Measure 119897119897 119905119905 and 119905119905119889119889

(ii) Assume the range of 119889119889 (0 lt 119889119889 lt 119897119897)

(iii) Calculate 119907119907119894119894 and 119907119907119889119889 with a constraints (0 lt 119907119907119889119889 lt 119907119907119894119894) (Eq 4 and 5)

(iv) Estimate m-value and find a feasible range of 119889119889 and 119907119907119889119889

119907119907119894119894 =119897119897 minus 119889119889119905119905 minus 119905119905119889119889

(4)

119907119907119889119889 =119889119889119905119905119889119889

(5)

35 Damage localization algorithm (Level 2amp3)

For damage localization 1198971198971 is estimated by measuring the arrival time from the excitation

PZT to the entrance of delamination (1199051199051) with additional pulse-echo method implementation

The procedure is similar with the previous damage quantification algorithm

(i) Measure 119897119897 119905119905 119905119905119889119889 and 1199051199051

(ii) Assume the range of 1198971198971 (0 lt 1198971198971 lt 119897119897)

(ii) Compute possible combinations of 1198971198972 119889119889 and 119907119907119889119889 with constraints

(0 lt 119907119907119889119889 lt 119907119907119894119894 119889119889 gt 0) (Eq 5 6 and 7)

- 11 -

1198971198972 = 1198971198971119905119905 minus 1199051199051 minus 119905119905119889119889

1199051199051 (6)

119889119889 = 119897119897 minus 1198971198971 minus 1198971198972 (7)

Since a PZT for pulse-echo method acts as both an actuator and a sensor the obtained sig-

nal contains both the excitation and the response signal Thus the reflected wave from the

entrance of damage can be buried into the tail of excitation signal and cause error on arrival

time of reflected wave from damage Even in this case damage can be localized and quan-

tized using the constraints When first wave packet reflected from the entrance of damage is

buried into the excitation signal the next wave packet reflected from the exit of damage can

be utilized alternatively If the algorithm regards the second reflected wave as the first re-

flected wave there are no combinations of estimated variable meeting the constraints Then

the algorithm repeats the process by regarding the wave as the second reflected wave the

constraints can select proper combination of the estimated variables

The estimated values have range bounded by the constraints When additional information

of 119907119907119894119894 is utilized the damage can be characterized more accurately In this study three meth-

ods are proposed and verified as information level of 119907119907119894119894

Method 1 Solely based on the current data measurement (Reference-free)

Method 2 Limit the possible range of 119907119907119894119894 (Prior knowledge of the system)

Method 3 Measure 119907119907119894119894 (Additional data need to be collected)

Method 1 described in this section can be utilized without any information of 119907119907119894119894 Meth-

od 2 produces more accurate estimation results using prior knowledge of the system such as

- 12 -

119907119907119894119894 variation under temperature changes Method 3 gives unique estimated value using meas-

ured 119907119907119894119894 from other path in the same time In section 4 and 5 damage is characterized using

these three methods and the results are compared each another

36 Matching pursuit method considering the dispersion effect of Lamb

wave

Matching pursuit method has been developed for signal compression approximation and

characterization of obtained signal and function [21-23] The key concept of the matching

pursuit method is that any signal (function) can be decomposed into a linear expansion of

signals selected from a pool of basis signals (functions) called lsquodictionaryrsquo

Let 119891119891 be a function to be analyzed then 119891119891 can be approximated a set of functions select-

ed from the dictionary 119863119863 in order to best match This is done by successive approximations

with orthogonal projections Let1198921198921205741205740 isin 119863119863 the function 119891119891 can be decomposed into

119891119891 = lang1198911198911198921198921205741205740rang1198921198921205741205740 + 119877119877119891119891 (8)

where 119877119877119891119891 is the residual function after approximating 119891119891 using 1198921198921205741205740 1198921198921205741205740 is orthogonal to

119877119877119891119891 hence

1198911198912 = lang1198911198911198921198921205741205740rang2

+ 1198771198771198911198912 (9)

To minimize 119877119877119891119891 1198921198921205741205740 is chosen to maximize lang1198911198911198921198921205741205740rang The matching pursuit method is

an iterative algorithm that decomposes the residue 119877119877119891119891 by projecting it on a function of 119863119863

that matches 119877119877119891119891 almost at best as it was done for 119891119891 The residue 119877119877119899119899119891119891 is decomposed into

119877119877119899119899119891119891 = lang119877119877119899119899119891119891119892119892120574120574119899119899rang119892119892120574120574119899119899 + 119877119877119899119899+1119891119891 (10)

- 13 -

the decomposition is carried up to the order 119898119898 119891119891 can be expressed as

119891119891 = lang119877119877119899119899119891119891119892119892120574120574119899119899rang119892119892120574120574119899119899

119898119898minus1

119899119899=0

+ 119877119877119898119898119891119891 (11)

The original function 119891119891 is decomposed into a sum of dictionary elements that are chosen to

best match its residues

In this study the matching pursuit method is utilized to accurately decompose of the over-

lapped Gaussian pulses and estimate the parameters of the pulse which used for damage

characterization The 119894119894119905119905ℎ Gaussian pulse in the dictionary for approximating the obtained

signal is expressed as

119892119892120574120574119894119894 = 119891119891(119860119860119894119894 119906119906119894119894 119904119904119894119894 119889119889119894119894120601120601119894119894) (12)

where 119860119860119894119894119906119906119894119894119904119904119894119894119889119889119894119894 and 120601120601119894119894 represent amplitude arrival time scale (pulse duration) excita-

tion(central) frequency and phase of the Gaussian pulse respectively The parameters of the

pulse are estimated using a nonlinear regression based optimization method (Gauss-Newton

method) based on an initial guess of the parameters Detail procedure for estimating the pa-

rameters of the pulse is expressed in Hong et al [23]

Among the estimated parameters of the pulses pulse duration (scale) represents the disper-

sion effect If a wave-packet is spread out the estimated pulse duration is increased Using

this feature the matching pursuit method is modified to capture waves which have longer

pulse duration than transmitted A0 Lamb wave mode If the estimate pulse duration of a wave

is less than the transmitted wave the modified matching pursuit method regards it as error

- 14 -

CHAPTER 4 NUMERICAL SIMULATION

41 Composite modeling method

Composite materials can be modeled from nano to macro level depends on its engineering

purpose In general four methods are utilized for modeling of composite materials First

composite is modeled as a homogenous material expressed in terms of effective material

properties [24] The effective material properties are estimated using the modulus and the

volume fraction of each constituent theoretically (Rule of mixture) and measured experimen-

tally Second a model is established that each laminate which have own material properties

are totally bonded together In this model the stacking sequence and direction of each lami-

nates are considered additionally [25-27] Third method for composite modeling in micro

level called micromechanics [28-30] This model considers the interaction between fibers and

matrix in laminate Last multiscale modeling which the scope of modeling is extended to

molecular or quantum level of constituents [31-33] This model starts from molecular me-

chanics or molecular dynamics then the scale of model leads to the nano micro and macro

level by incorporating small scale models As the scale of models is smaller it provides more

logical way than other large scale of models while the computational cost is more expensive

The scope of this study is focused on the dynamic behavior of composite structure the beam

or plate model was utilized for the numerical simulation

- 15 -

42 Damage modeling method

Generally three damage models have been developed and utilized in the real applications

First the stiffness reduction model called as the equivalent element model was introduced

[17 34-35] This model is simply realized by replacing the region around damage area by a

region with adjusted material properties (Youngrsquos modulus shear modulus density etc)

globally Next the duplicated node model is done by disconnecting the element connectivity

assuming there is no interaction among the delaminated surface This model can be utilized

when the location and the extension of damage is important issue [34-36] Last the contact

mechanics model using the principles of contact mechanics the two interfacial forces (con-

tact resistance and sliding resistance) in the delaminated surface [37-38] In this model the

delaminated surfaces are modeled as nonlinear functions of the interlaminar normal strain

and the interlaminar shear strain This model is computationally expensive while provides

more logical way than other models There is however no universally accepted model for

general impact-induced damage in composites In the case of impact-induced damage it is

impossible to predict or estimate the location and number of delaminated layers Thus from

the uncertainty of impact-induced damage the equivalent element model can be considered

as the effective damage model

- 16 -

43 Numerical simulation description

TABLE I Material property of simulation model and test specimen

Tensile Strength 657 MPa Tensile Modulus 59 GPa Shear Strength 804 MPa Shear Modulus 24 Gpa Specific Gravity 16

PZT A

PZT C

PZT BDamage

3

250 125 125

500 Unit mm

Fig 6 Description of the numerical simulation model

The feasibility of the proposed damage characterization technique is verified by 2-D FEM

simulation using MSC PATRAN and NASTRAN Fig 6 presents the dimension PZTs and

damage location of the simulation model In the model 05mm x 05mm square shell element

was used for modeling of the A0 mode whose flexural wave characteristic The element size

was determined by considering the wavelength of the propagation Lamb wave modes The

material properties of the simulation model and damage cases are presented in Table I and

Table II respectively

PZT A and C act as both actuator and sensor for collecting pulse-echo signal They were

installed on the plate which two of them are both-side collocated for the A0 Lamb wave

- 17 -

mode decomposition [39-40] PZT B was only used as a sensor for collecting pitch-catch sig-

nal For simulating the excitation and sensing of PZTs the pin-force model was adopted that

an input force apples to the end nodes of excitation PZT and averaging element strains corre-

sponding sensing area A Gaussian pulse (tone-burst) 50 kHz central frequency was excited

as an input signal

TABLE II Damage cases for numerical simulation

119949119949120783120783 (mm) 119941119941 (mm) G value reduction ()

Case 1 650 200 90 darr

Case 2 675 150 90 darr

Case 3 675 150 93 darr

Case 4 900 150 90 darr

- 18 -

44 Numerical simulation results

015 02 025 03 035

-2

0

2

Time (ms)

Am

plitu

de

IntactDamage

Reflected wave from damage

(a) Pitch-catch signal

015 02 025 03 035

-2

0

2

Time (ms)

Am

plitu

de

IntactDamage

Reflected waves from damage

(b) Pulse-echo signal Fig 7 Comparison of A0 Lamb wave mode signals from numerical simulation between intact and damage cases Time delay attenuation and multiple reflections due to damage simulated successfully using the damage modeling method described in the previous section

Fig 7 shows that comparison of A0 Lamb wave mode signals between intact and one of

damage cases in the simulation The features of A0 Lamb wave mode passing through dam-

age area described in section 2 are successfully simulated Thus the equivalent element

method is proper for modeling damage in composites In the damage cases the additional

waves reflected from damage is observed in both pitch-catch and pulse echo signal

Table III presents the result of the proposed damage characterization algorithm using the

numerical simulation cases shown in Table II In lsquoMethod 2rsquo 119907119907119894119894 variation is assumed to plusmn

5 of the normal value (1190~1315ms) The normal value of 119907119907119894119894 for lsquoMethod 3rsquo is meas-

ured as 1253 ms In each case the damage is characterized using the proposed technique

Note that as the information of 119907119907119894119894 is specified the estimated value is closer to the exact

value

- 19 -

TABLE III Damage characterization for the numerical simulation models

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

Case 1

Exact 650 200

Method 1 248 50~70 79~414 160~834

Method 2 248 603~667 135~241 271~486

Method 3 248 635 188 379

Case 2

Exact 675 150

Method 1 193 60~75 01~251 37~651

Method 2 193 618~682 114~222 295~575

Method 3 193 65 168 435

Case 3

Exact 675 150

Method 1 233 60~75 66~302 140~649

Method 2 233 668~738 85~196 182~419

Method 3 233 703 140 301

Case 4

Exact 900 150

Method 1 204 75~100 48~349 118~856

Method 2 204 860~951 107~216 264~530

Method 3 204 906 162 397

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1253 ms 1190~1315 ms Measured 119907119907119894119894 1253 ms

- 20 -

CHAPTER 5 EXPERIMENTAL VERIFICATION

51 Overview

In this section experimental study is investigated so that the proposed damage characteri-

zation technique Overall experimental setup and test specimen are shown in Fig 8 and 9

The experimental setup consists of test specimens data acquisition (DAQ) system and self-

sensing circuit Self-sensing circuit was installed for pulse-echo measurement using PZTs

[40-41]

PZT A(PZT C) PZT B

500

500

140

75

250

Teflon Inserted Damage (d =25)

Unit mm

Thickness = 3

PZT A(PZT C) PZT B

500

500

150

90

250

Impact Location

Unit mm

Thickness = 3

(a) Teflon inserted damage specimen

(b) Impact-induced damage specimen

Fig 8 Test specimens for the experimental verification In the Teflon inserted damage Teflon tapes with 25mm diameter were inserted at the center of the specimen The impact-induced damage was introduced to the specimen using the impact test equipment

- 21 -

511 Specimen

Two carbon composite plates with a dimension of 500times500times3mm3 were manufactured by

Nexcomes Inc by stacking 12 layers of CF3327 3K prepreg (woven fabrics type) manufac-

tured by Hankuk Fiber Inc The material properties of the specimen are presented in Table I

One of the specimens Teflon tapes with 25mm diameter was inserted in every 3 layers at the

center of the specimen to mimic the damage The other specimen is utilized for the character-

ization of real impact-induced damage

512 Data acquisition (DAQ) system and PZT

The data acquisition (DAQ) system consists of an arbitrary waveform generator (AWG) a

high speed signal digitizer (DIG) and two multiplexers (MUX) Using the 16-bits AWG

tone-burst input signal with a plusmn10 peak-to-peak voltage were generated The response signal

was measured by the 14-bit DIG For the signal-to-noise ratio reduction both pitch-catch and

pulse-echo signals were measured 150 times and averaged in the time domain 3 identical

circular PZTs of 10mm diameter manufactured by APC International Ltd were installed on

the specimen A Gaussian pulse (tone-burst) 50 kHz central frequency was excited as an in-

put signal The excitation frequency was chosen in order to avoid overlapping of the S0 and

A0 Lamb wave modes and generating of the higher modes (S1 A1hellip)

- 22 -

PZT Actuator

PZT Sensor

AWG

DIG

DAQ System Specimen

SensingActuatingMUX

1

23

4

1 Sending signal to MUX

4 Sending signal to DIG

Self-SensingCircuit

18 Unit mm

10

(a)

(b)

Fig 9 Overall experimental setup (a) Data acquisition (DAQ) system and (b) Installed circu-lar PZT transducer on the specimen with its dimension

513 Impact test

The impact-induced damage was introduced to the specimen using the impact test equip-

ment presented in Fig 10 A 5kg mass with 10mm tip was dropped from 015m height onto

the same location of the specimen The signals were obtained at each number of impacts

Since the reflected wave is not detected at the first impact case the data obtained from the

second and third impact were utilized for the analysis

10

Unit mm (a)

(b)

Fig 10 (a) Impact test equipment for introducing impact-induced damage and (b) Tip of the impact tester with 10mm diameter

- 23 -

514 Temperature effect test

In reality structures are subjected to changing environmental conditions such as tempera-

ture that affect measured signal and cause false-alarm Thus the robustness of the proposed

damage characterization technique under temperature variation in investigated (Fig 11) The

specimens were placed inside a temperature chamber and a thermocouple was installed on

the specimen to measure temperature of surface of the specimen The signal was obtained

under three temperature cases (0 20 and 50degC) and the humidity was kept at 30

Temperature Chamber

Test Specimen

Fig 11 Test specimen in temperature chamber

- 24 -

52 Teflon inserted damage

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Reflected wave from damage

(a)

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Transmitted WaveReflected Wave

Detected reflected wave

(b)

Fig 12 A0 Lamb wave mode obtained using pitch-catch method from the Teflon inserted damage specimen (a) The reflected wave from damage is shown (b) The Teflon inserted damage is detected by capturing the reflected wave

Fig 12 shows the obtained pitch-catch signal from the Teflon inserted damage specimen

and damage diagnosis result The Teflon inserted damage is detected successfully by captur-

ing the reflected wave Table IV presents the result of the Teflon inserted damage characteri-

zation 119907119907119894119894 is measured as 1213ms and the variation is assumed to plusmn 5 of the normal val-

ue (1152~1273ms) The results indicate that the proposed damage characterization approach

works well in real application

TABLE IV Damage characterization for the Teflon inserted damage specimen

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

Exact 635 25

Method 1 123 60~75 73~338 298~1220

Method 2 123 619~662 229~310 934~1188

Method 3 123 631 284 1161

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1213 ms

- 25 -

53 Impact-induced damage

11 115 12 125 13-4

-2

0

2

4

Time (ms)

Vol

tage

(mV

)

IntactImpact2Impact3

Reflected wave from damage

(a) Pitch-catch signal

11 115 12 125 13 135

-05

0

05

1

Time (ms)

Volta

ge (m

V)

IntactImpact2Impact3

Boundary reflection

Reflected wave from damage

(b) Pulse-echo signal

Fig 13 The signal obtained from the test specimen subjected to impact The impact was in-troduced at the same location of the specimen and the signal was obtained after each impacts As the number of impact is increased the transmitted A0 Lamb wave mode is delayed and attenuated

The obtained signals from impact-induced damage specimen are presented in Fig 13 It is

observed that as the number of impact is increased the transmitted A0 Lamb wave mode is

more delayed because the wave velocity in the damage area is reduced Fig 14 shows the

false-alarm test result whether the proposed damage detection algorithm works even in pris-

tine condition of the test specimen It is shown that no false alarm is occurred because no ad-

ditional reflected wave is captured

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Transmitted Wave

(a) Transmitted wave

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Reflected Wave

No reflected wave

(b) Reflected wave

Fig 14 False alarm test for the proposed damage detection algorithm in pristine condition of the specimen Reflected wave from damage is not captured false alarm is not occurred

- 26 -

TABLE V Damage characterization for the impact-induced damage specimen

of Impacts m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

2

(1470 J)

Method 1 080 65~90 75~246 470~1157

Method 2 080 801~885 114~245 711~1134

Method 3 080 843 179 1122

3

(2205 J)

Method 1 148 50~100 06~380 21~1134

Method 2 148 758~838 249~368 843~1125

Method 3 148 798 308 1044

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1213 ms

Table V presents the characterization results of the impact-induced damage case Here the

measured 119907119907119894119894 and the assumed variation are same as the artificial damage case It is observed

that as the number of impact is increased 119889119889 is increased while 119907119907119889119889 is reduced Thus it is

validated that the proposed damage characterization approach estimates the increment of the

damage well

54 Temperature effect test

541 Teflon inserted damage

11 115 12 125 13-10

-5

0

5

10

Time (ms)

Vol

tage

(mV

)

0 oC20 oC50 oC

(a) Pitch-catch signal

11 115 12 125 13 135

-1

0

1

2

3

Time(ms)

Volta

ge (m

V)

0 oC20 oC50 oC

(b) Pulse-echo signal

Fig 15 Signals obtained from the Teflon inserted damage specimen under temperature varia-tion Time delay and amplitude variation are occurred

Fig 15 shows the obtained signals from the Teflon inserted damage specimen subjected to

- 27 -

temperature variation Time delay and amplitude variations are occurred there is no addition-

al damage is introduced The analyzed results of the obtained signals are reported in Table VI

The measured 119907119907119894119894 is 1247 ms at 0 degC 1213 ms at 20 degC and 1168 ms at 50 degC respectively

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms (1152~1273 ms)

The estimated results are also changed under temperature variations the fluctuation of esti-

mated range and exact value is small

TABLE VI Temperature effect test for the Teflon inserted damage specimen

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

0 degC

Exact 635 25

Method 1 113 60~70 79~268 352~1291

Method 2 113 560~619 233~344 1034~1172

Method 3 113 606 257 1141

20 degC

Exact 635 25

Method 1 123 60~75 73~338 298~1220

Method 2 123 619~662 229~310 934~1188

Method 3 123 631 284 1161

50 degC

Exact 635 25

Method 1 137 60~75 72~328 263~1228

Method 2 137 622~687 183~299 666~1086

Method 3 137 642 264 960

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1247 ms at 0 degC 1213 ms at 20 degC and 1168 ms at 50 degC

- 28 -

542 Impact-induced damage

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Vol

tage

(mV

)

Transmitted WaveReflected Wave

No reflected wave

(a) 0 degC

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Transmitted WaveReflected Wave

No reflected wave

(b) 50 degC

Fig 16 False alarm test under temperature variations It is shown that proposed damage char-acterization algorithm is robust under temperature variations

The false alarm test in pristine condition of the impact-induced damage specimen under

temperature variation was investigated (Fig 16) It is validated that the proposed damage di-

agnosis algorithm is robust under temperature variations Fig 17 shows the obtained signals

from the impact-induced damage (impacted 3 times) specimen under temperature variation

The analyzed results of the obtained signals are reported in Table VII The measured 119907119907119894119894 in

each temperature and the assumed variation are same as the artificial damage case

11 115 12 125 13-5

0

5

Time (ms)

Volta

ge (m

V)

0 oC20 oC50 oC

(a) Pitch-catch signal

11 115 12 125 13 135

-1

0

1

2

3

Time(ms)

Vol

tage

(mV

)

0 oC20 oC50 oC

(b) Pulse-echo signal

Fig 17 Signals obtained from impact-induced damage specimen (Impact 3) under tempera-ture variation

- 29 -

TABLE VII Temperature effect test for the impact induced damage specimen (Impact 3 case)

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

0 degC

Method 1 135 75~90 72~310 267~1148

Method 2 135 705~779 264~381 978~1126

Method 3 135 763 289 1071

20 degC

Method 1 148 50~100 06~380 21~1134

Method 2 148 758~838 249~368 843~1125

Method 3 148 798 308 1044

50 degC

Method 1 153 75~100 53~415 174~1227

Method 2 153 804~889 214~336 702~1103

Method 3 153 815 320 1050

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1247 ms at 0 degC 1213 ms at 20 degC and 1168 ms at 50 degC

- 30 -

CHAPTER 6 CONCLUDING REMARKS

61 Executive summary

In this study the reference-free damage detection localization and quantification tech-

nique in a multi layered composite plate is proposed Additional reflected A0 Lamb wave

mode from the damage due to multiple reflections in damage area is used as a feature for

characterizing the damage First damage is characterized in terms of a damage index (m-

value) defined as the ratio of the physical length of the damage to the wavelength of the A0

Lamb wave mode Then the range of possible damage location length and severity are esti-

mated with constraints This procedure can be accomplished without prior baseline data To

realize the proposed technique the matching pursuit technique is modified for reduction of

the error by considering the dispersive characteristic of Lamb wave The matching pursuit

method estimates the arrival time of each wave packets accurately even two wave packets are

overlapped The numerical simulations and experimental tests including Teflon tape inserted

and impact-induced damage are investigated for the feasibility of the proposed technique

Moreover the effectiveness of the proposed technique under changing temperature is exper-

imentally verified

- 31 -

62 Suggestion for further study

The limitations of this study and the possible improvement and further studies are listed as

follow

1 Measurement of impact-induced damage length using conventional NDT technique

Additional experimental verification for impact-induced damage characterization is

needed by measuring the real length of damage using conventional NDT technique

such as ultrasonic scan or infrared camera

2 Limitation of both side collocated PZT installation In this study excitation PZTs are

installed both side and collocated for decomposing A0 Lamb wave mode However it

can produce decomposition error due to imperfection of PZT location and their installa-

tion Moreover in many cases assess to both surface of structures is not allowed such

as aircraft applications To improve this limitation a reference-free damage characteri-

zation using single side PZTs will be considered

3 Complex structure application In this study damage is characterized in simple plate-

like structure When the monitored structure becomes complex the response signals al-

so becomes complex due to its geometry and boundary Thus the investigations for

complex structure application such as stiffener spar and curvature are needed

- 32 -

REFERENCES

[1] Cawley P and Alleyne D (1996) ldquoThe use of Lamb waves for the long range inspection of large structuresrdquo Ultrasonics 34 pp287-290

[2] Ghosh T Kundu T and Karpur P (1998) ldquoEfficient use of Lamb waves for detecting defects in large platesrdquo Ultrasonics 36 pp791-801

[3] Guirguitui V (2003) ldquoLamb wave generation with piezoelectric wafer active sensors for structural health monitoringrdquo SPIE Conference on Smart Structures and Materials San Diego CA 5391 pp111-422

[4] Ragvahan A and Cesnik CE S (2004) ldquoModeling of piezoelectric-based Lamb wave generation and sensing for structural health monitoringrdquo SPIE Conference on Smart Struc-tures and Materials San Diego CA 5391 pp419-430

[5] Rose JL Pilarski A and Dirit JJ (1993) ldquoAn approach to guided wave mode selection for inspection of laminated platerdquo Journal of Reinforced Plastics and Composites 12(5) pp536-544

[6] Guo N and Cawley P (1993) ldquoThe interaction of Lamb waves with delaminations in composite laminatesrdquo Journal of Acoustical Society of America 94(4) pp 2240-2246

[7] Valdez SHD and Soutis C (2002) ldquoReal-time nondestructive evaluation of fiber com-posite laminates using low-frequency Lamb wavesrdquo Journal of Acoustical Society of Amer-ica 111(5) pp 2026-2033

[8] Yang M and Qiao P (2005) ldquoModeling and experimental detection of damage in virous materials using the pulse-echo method and piezoelectric sensoractuatorsrdquo Smart Materials and Structures 14 pp1083-1100

[9] Lestari W and Qiao P (2005) ldquoApplication of wave propagation analysis for damage identification in composite laminated beamsrdquo Journal of Composite Materials 39(22) pp1967-1984

[10] Kessler S S Spearing S M and Soutis C (2002) ldquoDamage detection in composite ma-terials using Lamb wave methodsrdquo Smart Materials and Structures 11 pp269-278

[11] Sohn H Park G Wail J R Limback N P and Farrar C R (2004) ldquoWavelet-based active sensing for delamination detection in composite structuresrdquo Smart Materials and Structures 13 pp 153-160

[12] Petculescu G Krishnaswamy S and Achenback J D (2008) ldquoGroup delay measure-ments using modally selective Lamb wave transducers for detection and sizing of delamina-tions in compositesrdquo Smart Materials and Structures 17

[13] Sohn H Park H W Law K H and Farrar C R (2007) ldquoDamage detection in compo-

- 33 -

site plates by using an enhanced time reversal methodrdquo Journal of Aerospace Engineering (ASCE) pp 141-151

[14] Sohn H (2007) ldquoEffects of environmental and operational variability on structural health monitoringrdquo Philosophical Transactions of the Royal Society Ardquo 365 pp539-560

[15] Rytter A (1993) ldquoVibration based inspection of civil engineering structuresrdquo PhD The-sis Department of Building Technology and Structural Engineering University of Aalborg Aalborg Denmark

[16] Hayashi T and Kawashima K (2002) ldquoMultiple reflections of Lamb waves at a delami-nationrdquo Ultrasonics 40 pp193-197

[17] Wang C H and Rose L R F (2003) ldquoWave reflection and transmission in beams con-taining delamination and inhomogeneityrdquo Journal of Sound and Vibration 264 pp 851-872

[18] Ramadas C Balasubramaniam K Joshi M and Krishnamurthy C V (2009) ldquoInterac-tion of the primary anti-symmetric Lamb mode (Ao) with symmetric delaminations numer-ical and experimental studiesrdquo Smart Materials and Structures 18

[19] Ramadas C Balasubramaniam K Joshi M and Krishnamurthy C V (2010) ldquoInterac-tion of guided Lamb waves with an asymmetrically located delamination in a laminated composite platerdquo Smart Materials and Structures 19

[20] Wilcox P Lowe M and Cawely P (2001) ldquoThe effect of dispersion on long-range in-spection using ultrasonic guided wavesrdquo NDTampE International 34 pp1-9

[21] Mallat S G and Zhang Z (1993) ldquoMatching pursuits with time-frequency dictionariesrdquo IEEE Tansactions on Signal Processing 41(12) pp3397-3415

[22] Zhang G Zhang S and Wang Y (2000) ldquoApplication of adaptive time-frequency de-composition in ultrasonic NDE of highly scattering materialsrdquo Ultrasonics 38 pp 961-964

[23] Hong J C Sun K H and Kim Y Y (2005) ldquoThe matching pursuit approach based on the modulated Gaussian pulse for efficient guided-wave damage inspectionrdquo Smart Mate-rials and Structures 14 pp548-560

[24] Herakovich C T (1997) Mechanics of Fibrous Composites Wiley 480 pages

[25] Reddy J N (1985) ldquoA review of the literature on finite-element modeling of laminated composite platesrdquo The Shock and Vibration Digest 17(4) pp 3-8

[26] Reddy J N (2003) Mechanics of Laminated Composite Plates and Shells Theory and Analysis CRC Press 856 pages

[27] Palacz M Krawczuk M and Ostachowicz W (2005) ldquoThe spectral finite element mod-el for analysis of flexure-shear coupled wave propagation Part 1 laminated multilayer

- 34 -

composite beamrdquo Composite Structures 68 pp37-44

[28] Zhang Y C and Harding J (1990) ldquoA numerical micromechanics analysis of the me-chanical properties of a plain weave compositerdquo Computer amp Structures 36(5) pp 839-844

[29] Nemat-Nasser S and Hori M (1999) Micromechanics Overall Properties of Heteroge-neous Materials Elsevier 810 pages

[30] Scida D Aboura Z Benzeggagh M L and Bocherens E (1999) ldquoA micromechanics model for 3D elasticity and failure of woven-fibre composite materialrdquo Composites Science and Technology 59 pp 505-517

[31] Thomas Y and Wu X H (1997) ldquoA multiscale finite element method for elliptic prob-lems in composite materials and porous mediardquo Journal of Computational Physics 134 pp 169-189

[32] Lee K H Moorthy S and Ghosh S (1999) ldquoMultiple scale computational model for damage in composite materialsrdquo Computer Methods in Applied Mechanics and Engineer-ing 172 pp 175-201

[33] Namilae S and Chandra N (2005) ldquoMultiscale model to study the effect of interfaces in carbon nanotube-based compositesrdquo Journal of Engineering Materials and Technology (ASME) 127 pp222-232

[34] Adams D E (2007) Health monitoring of Structural Materials and Components Wiley 460 pages

[35] Gopalakrishnan S (2009) ldquoModeling Aspects in Finite Elementrdquo Encyclopedia of Struc-tural Health Monitoring 2 pp791-809

[36] Palacz M Krawczuk M and Ostachowicz W (2005) ldquoThe spectral finite element mod-el for analysis of flexure-shear coupled wave propagation Part 2 Deminated multilayer composite beamrdquo Composite Structures 68 pp45-51

[37] Baik J M and Thopson R B (1984) ldquoUltrasonic scattering from imperfect interfaces A quasi-static modelrdquo Journal of Nondestructive Evaluation 4(3) pp177-196

[38] Luo H and Hanagud S (2000) ldquoDynamics of delamincated beamsrdquo International Jour-nal of Solids and Structures 37 pp1501-1519

[39] Kim S B and Sohn H (2007) ldquoInstantaneous reference-free crack detection based on polarization characteristics of piezoelectric materialsrdquo Smart Materials and Structures 16 pp2375-2387

[40] An Y K and Sohn H (2010) ldquoInstantaneous crack detection under varying temperature and static loading conditionsrdquo The Special Issue of Structural Control and Health Monitor-ing 17(7) pp 730-741

- 35 -

[41] Lee S J and Sohn H (2006) ldquoActive self-sensing scheme development for structural health monitoringrdquo Smart Materials and Structures 15(6) pp 1734-1746

- 36 -

요 약 문

복합재료 손상의 무기저 진단 및 정량화

복합재료는 재료의 경량성 내구성 고강도의 특성으로 인하여 항공기에서부터

교량에까지 널리 사용되고 있다 하지만 외부의 충격으로 인한 복합재료 내부의

손상은 재료의 강도를 감소시켜 구조물의 안정성을 저하시키는 요인이 된다 하지만

복합재료의 손상은 재료 내부에서 발생하여 육안으로는 식별하기도 힘들며 표면에서

보이는 것만으로 내부의 손상 정도를 판단하기는 불가능하다는 단점을 가지고 있다

본 연구에서는 손상 내부에서 반사로 인해 발생하는 파를 이용하여 기저자료를

사용하지 않고 특정 시점에서 얻은 자료만으로 복합재료의 손상을 진단하고 정량화

할 수 있는 기법을 제안한다 제안된 기법의 적용성을 검증하기 위하여 유한요소법을

이용한 수치해석을 수행하였고 Teflon tape 이 삽입된 시편과 실제 충격하중으로

인한 손상을 입은 시편을 이용한 실험을 진행하였다 수치해석 및 실험 결과 제안된

기법이 복합재료의 손상을 성공적으로 진단 및 정량화할 수 있음을 확인하였다 또한

다양한 외부 온도 환경에서 제안된 기법의 환경영향 실험을 수행하였고 다양한

환경에서 손상의 진단 및 정량화가 성공적으로 이루어짐을 검증하였다

핵심어 무기저 손상진단 및 정량화 압전센서 복합재료 유한요소법 Matching pursuit

method

- 37 -

감 사 의 글

기대와 두려움을 안고 시작한 석사과정이 어느덧 2 년 이라는 세월 속에서 추억으로 자리잡아

가고 있습니다 학부생 기간을 포함한 지난 3 년 동안 SHM 이라는 새로운 분야를 접하고 연구를

수행 할 수 있도록 도와주신 고마운 분들이 있어 그 감사를 표하고자 합니다

우선 저의 많은 부족한 점을 일깨워 주시고 연구에 대한 뜻을 더욱 확고히 세울 수 있게 해

주신 지도교수님이신 손 훈 교수님께 깊은 감사의 뜻을 전합니다 짧은 기간이었지만 교수님의

도움으로 많은 것을 배우고 느끼고 깨달을 수 있었습니다 또한 본 연구를 수행할 수 있도록 지

원해주신 국토해양부 (U-City 석 박사과정 지원사업)에 깊은 감사 드립니다

때로 힘들고 지칠 때 항상 저를 믿고 응원해 주시는 가족이 있었기에 지금의 글을 쓸 수 있었

습니다 또한 지금 연구실에서 함께 생활하고 있는 윤규형 민구형 철민이형 현석이형 진열이형

호민이형 병진이 선혜누나와 성대에 계신 창길이형 미국에 계신 의재형과 Debaditiya 에게도 감

사의 말씀을 전합니다 그리고 모두 나열할 수는 없지만 카이스트 생활의 활력소가 되어준 카이

스트 건설 및 환경공학과 선후배들 직원 분들과 SOS 멤버들에게도 감사를 표합니다

앞서 거론한 많은 이들의 도움이 없었다면 저 또한 지금 이 자리에 있을 수 없었을 것입니다

많이 부족하기에 더 노력하고 배우고자 했던 초심을 다시 한번 생각하며 더 큰 결실을 맺기 위해

앞으로 나아가겠습니다 도움을 주신 모든 분들께 이 감사의 글을 바칩니다

2011년 6월 카이스트에서 임형진 올림

- 38 -

CURRICULUM VITAE

Hyung Jin Lim

Research Assistant

The Civil and Environmental Engineering Department

Korea Advanced Institute of Science and Technology (KAIST)

291 Daehak-Ro Guseong-Dong Yuseong-Gu

Daejeon Republic of Korea 305-701

Phone (82)+42-350-3665 Fax (82)+42-350-3610

Email limhj87gmailcom

RESEARCH INTERESTS

Damage Detection Localization and Quantification on Composite and Metallic Structures Pattern Recognition Finite Element Modeling (FEM)

EDUCATION

2009 ndash 2011 MS Civil and Environmental Eng Korea Advanced Institute of Science and Technology (KAIST) Korea Rep

2006 ndash 2009 BS Civil and Environmental Eng Korea Advanced Institute of Science and Technology (KAIST) Korea Rep

JOURNAL PUBLICATIONS The corresponding author is underlined 1 Hyung Jin Lim Hoon Sohn ldquoReference-free damage detection localization and quantification

using in compositerdquo in preparation for Journal of the Acoustical Society of America 2011

2 Chul Min Yeum Hoon Sohn Jeong Bum Ihn Hyung Jin Lim ldquoInstantaneous delamination detection in a composite using a dual piezoelectric transducer networkrdquo Submitted to Compo-sites Science and Technology 2011

3 Yun-Kyu An Hyung Jin Lim Min Koo Kim Jin Yeol Yang Hoon Sohn Chang Guen Lee ldquoInvestigation of applicability of local reference-free damage detection techniques to in-situ

- 39 -

bridgesrdquo Submitted to ASCE Journal of Structural Engineering 2011

4 Hyung Jin Lim Min Koo Kim Hoon Sohn Chan Yik Park ldquoImpedance based damage detec-tion under varying temperature and loading conditionsrdquo Submitted to NDTampE International 2011

5 Hyung Jin Lim Hoon Sohn ldquoQuantitative estimation of transmitted and reflected lamb waves at discontinuityrdquo Journal of the Korean Society for Nondestructive Testing Vol30 No 4 pp1-8 2010

CONFERENCE PROCEEDINGS

The corresponding author is underlined 1 Hyung Jin Lim and Hoon Sohn ldquoReference-free damage detection localization and quantifi-

cation in compositesrdquo ASCE Engineering Mechanics Institute Conference Boston June 2-4 2011

2 Hyung Jin Lim and Hoon Sohn ldquoReference-free delamination quantification within multi-layer composite materialsrdquo KSNT Spring Conference May 26-27 Gyeong-Ju Korea 2011

3 Min Koo Kim Hyun Jin Lim Hoon Sohn Chan Yik Park ldquoImpedance-based bolt loosening detection under varying temperature and loadingrdquo Asian Pacific Workshop on Structural Health Monitoring Tokyo Japan November 30-December 2 2010

4 Hyung Jin Lim Chul Min Yeum and Hoon Sohn ldquoModeling of Impact-induced Delamination in a Multilayer Composite Platerdquo The 23rd KKCNN Symposium on Civil Engineering Taipei (Taiwan) November 13-15 2010

5 Yun-Kyu An Hyung Jin Lim and Hoon Sohn ldquoTemperature independent Delamination Detec-tion using data normalizationrdquo The 3rd International Conference on Multi-Functional Materi-als and Structures Chonbuk National Univ Jeonju Korea September 14-18 2010

6 Hoon Sohn Yun-Kyu An Chang Gil Lee Min Koo Kim Hyung Jin Lim ldquoApplication of online NDT techniques to decommissioned bridge testingrdquo the 5th European Workshop on Structural Health Monitoring Sorrento Italy June 29-July 02 2010

7 Min Koo Kim Hyung Jin Lim Hoon Sohn Chan Yik Park ldquoImpedance based damage diag-nosis of a complex composite aircraft wing under changing loading and temperature condi-tionsrdquo the 5th European Workshop on Structural Health Monitoring Sorrento Italy June 29-July 02 2010

8 Yun-Kyu An Min Koo Kim Hyung Jin Lim and Hoon Sohn ldquoValidations of reference-free crack detection techniques through a decommissioned bridge testrdquo 2010 COSEIK Annual Con-ference Jeju Korea April 8-9 2010

- 40 -

PARTICIPANT PROJECTS

1 Development of Smart Steel Members for Online Steel Structure Monitoring POSCO Corporation (Funded 41000000 KRW (41000 USD) for 010511 to 123111)

2 Development of Enabling Technologies for Structural Health Monitoring Korean Agency for Defense Development (Funded 240000000 KRW (240000 USD) out of total 3200000000 KRW (3200000 USD) for 060108-053111)

3 Development of On-board SHM Technologies for Composite Air Vehicles The Boeing Company (Funded 400000 USD (400000000 KRW) for 080108 to 073111)

EDUCATIONAL EXPERIENCES

Fall 2009

Teaching Assistant CE207 ldquoElementary Structural Engineering and Laboratoryrdquo KAIST 3 unit undergraduate course

EXPERIMENTAL EXPERIENTCES

2010 Monitoring of Real Aircraft Composite Wing Segment Korea Agency for Defense De-velopment (ADD) Daejeon Korea Real time monitoring of composite air vehicles fund-ed by Korea Agency for Defense Development (ADD)

2009-2011

Field Monitoring of Real Bridge Structures Gimpo (Ramp G bridge) amp Icheon (Sam-seung bridge) Korea US-Korea collaborative research for bridge monitoring testbeds

HONORS amp AWARDS

Scholarships for outstanding students granted KAIST (2007-2009) The three prominent students in the civil and environmental engineering department of KAIST were selected as recipients This scholarship provided 2000$ per year for three years

- 41 -

Page 14: 복합재료 손상의 무기저 진단 및 정량화 - KAISTssslab.kaist.ac.kr/article/pdf/Thesis_2.pdf · 1 Declaration of Ethical Conduct in Research: I, as a graduate student

13 Objective and uniqueness

In this study a reference-free damage characterization technique including detection lo-

calization and quantification of impact-induced damage in composite materials is developed

The general damage characterization can be categorized into the following levels [14]

Level 0 Define damage

Level 1 Determine the presence of damage within the structure

Level 2 Locate the regions of damage

Level 3 Quantify the damage

Level 4 Predict the remaining service life of the structure

Through the proposed technique from level 1 to level 3 can be achieved

Most of common approaches of damage characterization techniques described in the pre-

vious section use prior obtained baseline data However these approaches can cause false

alarm due to the effect of environmental variations such as temperature and ambient vibra-

tions [15]

The uniqueness of this study is

1 Damage is characterized without using prior measured baseline data

2 Robustness of the proposed technique under temperature variation is investigated

3 Additional waves reflected from inside damage have been used in addition to transmit-

ted and reflected waves

- 3 -

14 Organization

The organization of this thesis is summarized as follows

Chaper1 A general introduction including motivation literature review objective and

uniqueness of this study is described

Chaper2 Interactions between Lamb wave mode and damage in composites such as time

delay attenuation and multiple reflections are summarized Damage index (m-value) for

damage characterization is defined

Chaper3 Algorithms for the reference-free damage characterization technique using m-

value are developed Modified matching pursuit for this study is also described

Chaper4 Verification of the algorithm using numerical simulation is presented Various

modeled damage cases are characterized

Chaper5 Experimental verifications are described Two damage cases (Teflon inserted and

impact-induced damage) are investigated for the feasibility of the proposed damage charac-

terization technique Temperature effect test using temperature chamber is also carried out

Chaper6 Conclusion and summary of the present study is reported In advance suggestions

and future work recommendations are represented

- 4 -

CHAPTER 2 INTERACTION OF LAMB WAVE WITH DAMAGE

21 A0 Lamb wave mode delay and attenuation in the time domain

Time delay and attenuation of the A0 mode Lamb wave after passing through the damage

area due to effective thickness reduction were reported in previous researches [812] Moreo-

ver the attenuation of the A0 Lamb wave mode is occurred by degrading of shear modulus of

in the damage area However the time delay and amplitude variation of the A0 Lamb wave

mode also occurred even when the structure is health condition by the material property

changing due to temperature variation of the structure (Fig 1) Thus conventional damage

characterization methods using the time delay and attenuation of A0 Lamb wave mode as a

feature can cause false-alarm in real application

11 115 12 125 13-10

-5

0

5

10

Time (ms)

Vol

tage

(mV

)

IntactDamage

Reflected wave from Damage

(a)

11 115 12 125 13-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

0 oC20 oC50 oC

(b)

Fig 1 Lamb wave propagation characteristics (a) A0 Lamb wave mode time delay and atten-uation after passing through a damage (b) The time delay and amplitude variation of A0 Lamb wave mode under temperature variation in pristine condition of structure Because of this conventional damage detection techniques in composites have limitations for real struc-ture application

- 5 -

22 Multiple reflections in the damage area

Standing wave patterns due to multiple reflections of A0 Lamb wave mode in the damage

area were observed through numerical simulations and ultrasonic wave-field imaging using a

scanning laser vibrometer [16] The formation of standing waves in damage is illustrated

through Fig 2 When the incident waves enter the damage area a significant portion of the

waves are reflected back from the exit Then the reflected waves travel through the damage

and undergo reflection again at the original entrance As a consequence of multiple reflec-

tions at the entrance and the exit a considerable amount of ultrasonic energy is trapped inside

the damage area [17-19]

Multiple Reflections

Incident Wave Transmitted Wave

Damage

Fig 2 Multiple reflections in the damage area The multiple reflections of A0 Lamb waves in the damage area are utilize as a feature for characterization of the damage in this study

- 6 -

CHAPTER 3 THEORETICAL DEVELOPMENT

31 Overview

Damage

l = l1 + d + l2

d = vdtd l1 = vit1 l2 = vit2

t = t1 + td + t2

Excitation and Sensing PZT Sensing PZT

l

Transmitted Wave

Reflected Waves

Incident Wave

Fig 3 Parameters and notations of the proposed damage characterization technique Damage is detected by identifying the additional reflected wave from the damage due to multiple re-flections (dashed lines) Localization and quantification of the damage is carried out by measuring the arrival time of transmitted and reflected waves from the damage

In this section the reference-free damage characterization algorithm is proposed Damage

diagnosis and quantification are achieved by identifying the additional reflected waves due to

multiple reflections in the damage area (dashed lines in Fig 3) Moreover damage is local-

ized using additional information from pulse-echo method implement For the localization

and the quantification of damage the variables such as the path length and A0 Lamb wave

mode velocity in damage area are defined (Fig 3) For example 119897119897 is expressed as

119897119897 = 1198971198971 + 119889119889 + 1198971198972 (1)

where 1198971198971 119889119889 and 1198971198972 are the distance from the excitation PZTs to the entrance of damage the

damage length and the distance from the exit of damage to the sensing PZT respectively

- 7 -

Moreover 119907119907119894119894 and 119907119907119889119889 are also defined as the A0 Lamb wave mode velocity in pristine area

and damage area respectively Note that the whole procedure of the damage characterization

is accomplished based solely on the signals obtained from the current state of the structure

32 Damage index (m-value)

The proposed reference-free damage characterization technique utilizes the waves reflect-

ed additionally from the damage area as a feature The arrival time difference of the transmit-

ted and reflected A0 Lamb wave modes reflected from the exit and the entrance of the delam-

ination area one times each (∆119905119905) contains the information of the physical length of the dam-

age (119889119889) and A0 Lamb wave mode velocity in the damage area (119907119907119889119889) The wave velocity in the

damage area which is related to the shear modulus (G) represents the severity of the damage

∆119905119905 = 2119905119905119889119889 =2119889119889119907119907119889119889

(2)

However based on the measured data available the 119889119889 and 119907119907119889119889 (or G) cannot be uniquely

determined Fig 4 shows two FEM models which have different physical length and wave

velocity (G) of the damage Detail explanation of delamination modeling technique will be

covered in section 4 The similar signal in the time domain and ∆119905119905 are obtained from the

two different damage cases

- 8 -

Damage 1

Damage 2

20

Unit mm15

90 G value reduction

943 G value reduction

m-value = 248

m-value = 248

015 02 025 03 035

-2

-1

0

1

2

Time (ms)

Ampl

itude

Damage 1Damage 2

Fig 4 Two different FEM models which have different 119889119889 and 119907119907119889119889 (or G) This represents that the damage cannot be characterized uniquely based on the measured data available

Thus in this study the damage index (m-value) which can be uniquely defined based on

the available data is introduced The m-value is defined as the ratio of the physical length of

the damage and wavelength of the A0 Lamb wave mode in the damage area (120582120582119889119889)

119898119898 =119889119889120582120582119889119889

=119889119889119889119889119907119907119889119889

(3)

In many cases the transmitted and reflected waves are overlapped and it is hard to decom-

pose and calculate the arrival times of them Matching pursuit method is an alternative way

to accurately decompose of the overlapped waves and estimate the parameters of each wave

such as arrival time (Fig 5) Detail explanation of the matching pursuit method is covered in

section 36

015 02 025 03-2

-1

0

1

2

Time (ms)

Am

plitu

de

OriginalApproximated

(a)

015 02 025 03-2

-1

0

1

2

Time (ms)

Ampl

itude

Wave 1Wave 2

(b)

Fig 5 (a) Example of the overlapped signal with comparing approximated signal using the matching pursuit method (b) Decomposed signal using the matching pursuit method The matching pursuit method provides a conventional way to decompose and identify each wave

- 9 -

33 Damage detection algorithm (Level 1)

The damage diagnosis is accomplished by extracting the additional reflected A0 Lamb

wave modes from damage due to multiple reflections If A0 Lamb wave modes are decom-

posed perfectly it is trivial that one transmitted A0 Lamb wave mode is always captured in

the response signal If damage exists other waves in the response signal behind the transmit-

ted A0 Lamb wave mode are presented Since this approach does not rely on previously col-

lected data the diagnosis is independent from environmental variation such as temperature

changes

It is reported that the dispersion of Lamb waves cause wave-packets to spread out in space

and time as they propagate through a structure [20] Thus the reflected waves from the dam-

age have longer pulse duration than the transmitted wave due to their longer propagating dis-

tance In this study the matching pursuit method is modified to capture more spread wave

than the transmitted A0 Lamb wave mode Detail explanation of the modified matching pur-

suit method is described in latter section

Let 1199041199041 and 1199041199042 as the pulse duration (scale) of the transmitted and additional reflected A0

Lamb wave mode respectively

ldquoIf 1199041199042 gt 1199041199041 damage exists

Otherwise there is no damagerdquo

- 10 -

34 Damage quantification algorithm (Level 3)

As mentioned in the previous section damage is quantized by estimating 119889119889 and 119907119907119889119889 using

pitch-catch method The procedure to estimate 119889119889 and 119907119907119889119889 is

(i) Measure 119897119897 119905119905 and 119905119905119889119889

(ii) Assume the range of 119889119889 (0 lt 119889119889 lt 119897119897)

(iii) Calculate 119907119907119894119894 and 119907119907119889119889 with a constraints (0 lt 119907119907119889119889 lt 119907119907119894119894) (Eq 4 and 5)

(iv) Estimate m-value and find a feasible range of 119889119889 and 119907119907119889119889

119907119907119894119894 =119897119897 minus 119889119889119905119905 minus 119905119905119889119889

(4)

119907119907119889119889 =119889119889119905119905119889119889

(5)

35 Damage localization algorithm (Level 2amp3)

For damage localization 1198971198971 is estimated by measuring the arrival time from the excitation

PZT to the entrance of delamination (1199051199051) with additional pulse-echo method implementation

The procedure is similar with the previous damage quantification algorithm

(i) Measure 119897119897 119905119905 119905119905119889119889 and 1199051199051

(ii) Assume the range of 1198971198971 (0 lt 1198971198971 lt 119897119897)

(ii) Compute possible combinations of 1198971198972 119889119889 and 119907119907119889119889 with constraints

(0 lt 119907119907119889119889 lt 119907119907119894119894 119889119889 gt 0) (Eq 5 6 and 7)

- 11 -

1198971198972 = 1198971198971119905119905 minus 1199051199051 minus 119905119905119889119889

1199051199051 (6)

119889119889 = 119897119897 minus 1198971198971 minus 1198971198972 (7)

Since a PZT for pulse-echo method acts as both an actuator and a sensor the obtained sig-

nal contains both the excitation and the response signal Thus the reflected wave from the

entrance of damage can be buried into the tail of excitation signal and cause error on arrival

time of reflected wave from damage Even in this case damage can be localized and quan-

tized using the constraints When first wave packet reflected from the entrance of damage is

buried into the excitation signal the next wave packet reflected from the exit of damage can

be utilized alternatively If the algorithm regards the second reflected wave as the first re-

flected wave there are no combinations of estimated variable meeting the constraints Then

the algorithm repeats the process by regarding the wave as the second reflected wave the

constraints can select proper combination of the estimated variables

The estimated values have range bounded by the constraints When additional information

of 119907119907119894119894 is utilized the damage can be characterized more accurately In this study three meth-

ods are proposed and verified as information level of 119907119907119894119894

Method 1 Solely based on the current data measurement (Reference-free)

Method 2 Limit the possible range of 119907119907119894119894 (Prior knowledge of the system)

Method 3 Measure 119907119907119894119894 (Additional data need to be collected)

Method 1 described in this section can be utilized without any information of 119907119907119894119894 Meth-

od 2 produces more accurate estimation results using prior knowledge of the system such as

- 12 -

119907119907119894119894 variation under temperature changes Method 3 gives unique estimated value using meas-

ured 119907119907119894119894 from other path in the same time In section 4 and 5 damage is characterized using

these three methods and the results are compared each another

36 Matching pursuit method considering the dispersion effect of Lamb

wave

Matching pursuit method has been developed for signal compression approximation and

characterization of obtained signal and function [21-23] The key concept of the matching

pursuit method is that any signal (function) can be decomposed into a linear expansion of

signals selected from a pool of basis signals (functions) called lsquodictionaryrsquo

Let 119891119891 be a function to be analyzed then 119891119891 can be approximated a set of functions select-

ed from the dictionary 119863119863 in order to best match This is done by successive approximations

with orthogonal projections Let1198921198921205741205740 isin 119863119863 the function 119891119891 can be decomposed into

119891119891 = lang1198911198911198921198921205741205740rang1198921198921205741205740 + 119877119877119891119891 (8)

where 119877119877119891119891 is the residual function after approximating 119891119891 using 1198921198921205741205740 1198921198921205741205740 is orthogonal to

119877119877119891119891 hence

1198911198912 = lang1198911198911198921198921205741205740rang2

+ 1198771198771198911198912 (9)

To minimize 119877119877119891119891 1198921198921205741205740 is chosen to maximize lang1198911198911198921198921205741205740rang The matching pursuit method is

an iterative algorithm that decomposes the residue 119877119877119891119891 by projecting it on a function of 119863119863

that matches 119877119877119891119891 almost at best as it was done for 119891119891 The residue 119877119877119899119899119891119891 is decomposed into

119877119877119899119899119891119891 = lang119877119877119899119899119891119891119892119892120574120574119899119899rang119892119892120574120574119899119899 + 119877119877119899119899+1119891119891 (10)

- 13 -

the decomposition is carried up to the order 119898119898 119891119891 can be expressed as

119891119891 = lang119877119877119899119899119891119891119892119892120574120574119899119899rang119892119892120574120574119899119899

119898119898minus1

119899119899=0

+ 119877119877119898119898119891119891 (11)

The original function 119891119891 is decomposed into a sum of dictionary elements that are chosen to

best match its residues

In this study the matching pursuit method is utilized to accurately decompose of the over-

lapped Gaussian pulses and estimate the parameters of the pulse which used for damage

characterization The 119894119894119905119905ℎ Gaussian pulse in the dictionary for approximating the obtained

signal is expressed as

119892119892120574120574119894119894 = 119891119891(119860119860119894119894 119906119906119894119894 119904119904119894119894 119889119889119894119894120601120601119894119894) (12)

where 119860119860119894119894119906119906119894119894119904119904119894119894119889119889119894119894 and 120601120601119894119894 represent amplitude arrival time scale (pulse duration) excita-

tion(central) frequency and phase of the Gaussian pulse respectively The parameters of the

pulse are estimated using a nonlinear regression based optimization method (Gauss-Newton

method) based on an initial guess of the parameters Detail procedure for estimating the pa-

rameters of the pulse is expressed in Hong et al [23]

Among the estimated parameters of the pulses pulse duration (scale) represents the disper-

sion effect If a wave-packet is spread out the estimated pulse duration is increased Using

this feature the matching pursuit method is modified to capture waves which have longer

pulse duration than transmitted A0 Lamb wave mode If the estimate pulse duration of a wave

is less than the transmitted wave the modified matching pursuit method regards it as error

- 14 -

CHAPTER 4 NUMERICAL SIMULATION

41 Composite modeling method

Composite materials can be modeled from nano to macro level depends on its engineering

purpose In general four methods are utilized for modeling of composite materials First

composite is modeled as a homogenous material expressed in terms of effective material

properties [24] The effective material properties are estimated using the modulus and the

volume fraction of each constituent theoretically (Rule of mixture) and measured experimen-

tally Second a model is established that each laminate which have own material properties

are totally bonded together In this model the stacking sequence and direction of each lami-

nates are considered additionally [25-27] Third method for composite modeling in micro

level called micromechanics [28-30] This model considers the interaction between fibers and

matrix in laminate Last multiscale modeling which the scope of modeling is extended to

molecular or quantum level of constituents [31-33] This model starts from molecular me-

chanics or molecular dynamics then the scale of model leads to the nano micro and macro

level by incorporating small scale models As the scale of models is smaller it provides more

logical way than other large scale of models while the computational cost is more expensive

The scope of this study is focused on the dynamic behavior of composite structure the beam

or plate model was utilized for the numerical simulation

- 15 -

42 Damage modeling method

Generally three damage models have been developed and utilized in the real applications

First the stiffness reduction model called as the equivalent element model was introduced

[17 34-35] This model is simply realized by replacing the region around damage area by a

region with adjusted material properties (Youngrsquos modulus shear modulus density etc)

globally Next the duplicated node model is done by disconnecting the element connectivity

assuming there is no interaction among the delaminated surface This model can be utilized

when the location and the extension of damage is important issue [34-36] Last the contact

mechanics model using the principles of contact mechanics the two interfacial forces (con-

tact resistance and sliding resistance) in the delaminated surface [37-38] In this model the

delaminated surfaces are modeled as nonlinear functions of the interlaminar normal strain

and the interlaminar shear strain This model is computationally expensive while provides

more logical way than other models There is however no universally accepted model for

general impact-induced damage in composites In the case of impact-induced damage it is

impossible to predict or estimate the location and number of delaminated layers Thus from

the uncertainty of impact-induced damage the equivalent element model can be considered

as the effective damage model

- 16 -

43 Numerical simulation description

TABLE I Material property of simulation model and test specimen

Tensile Strength 657 MPa Tensile Modulus 59 GPa Shear Strength 804 MPa Shear Modulus 24 Gpa Specific Gravity 16

PZT A

PZT C

PZT BDamage

3

250 125 125

500 Unit mm

Fig 6 Description of the numerical simulation model

The feasibility of the proposed damage characterization technique is verified by 2-D FEM

simulation using MSC PATRAN and NASTRAN Fig 6 presents the dimension PZTs and

damage location of the simulation model In the model 05mm x 05mm square shell element

was used for modeling of the A0 mode whose flexural wave characteristic The element size

was determined by considering the wavelength of the propagation Lamb wave modes The

material properties of the simulation model and damage cases are presented in Table I and

Table II respectively

PZT A and C act as both actuator and sensor for collecting pulse-echo signal They were

installed on the plate which two of them are both-side collocated for the A0 Lamb wave

- 17 -

mode decomposition [39-40] PZT B was only used as a sensor for collecting pitch-catch sig-

nal For simulating the excitation and sensing of PZTs the pin-force model was adopted that

an input force apples to the end nodes of excitation PZT and averaging element strains corre-

sponding sensing area A Gaussian pulse (tone-burst) 50 kHz central frequency was excited

as an input signal

TABLE II Damage cases for numerical simulation

119949119949120783120783 (mm) 119941119941 (mm) G value reduction ()

Case 1 650 200 90 darr

Case 2 675 150 90 darr

Case 3 675 150 93 darr

Case 4 900 150 90 darr

- 18 -

44 Numerical simulation results

015 02 025 03 035

-2

0

2

Time (ms)

Am

plitu

de

IntactDamage

Reflected wave from damage

(a) Pitch-catch signal

015 02 025 03 035

-2

0

2

Time (ms)

Am

plitu

de

IntactDamage

Reflected waves from damage

(b) Pulse-echo signal Fig 7 Comparison of A0 Lamb wave mode signals from numerical simulation between intact and damage cases Time delay attenuation and multiple reflections due to damage simulated successfully using the damage modeling method described in the previous section

Fig 7 shows that comparison of A0 Lamb wave mode signals between intact and one of

damage cases in the simulation The features of A0 Lamb wave mode passing through dam-

age area described in section 2 are successfully simulated Thus the equivalent element

method is proper for modeling damage in composites In the damage cases the additional

waves reflected from damage is observed in both pitch-catch and pulse echo signal

Table III presents the result of the proposed damage characterization algorithm using the

numerical simulation cases shown in Table II In lsquoMethod 2rsquo 119907119907119894119894 variation is assumed to plusmn

5 of the normal value (1190~1315ms) The normal value of 119907119907119894119894 for lsquoMethod 3rsquo is meas-

ured as 1253 ms In each case the damage is characterized using the proposed technique

Note that as the information of 119907119907119894119894 is specified the estimated value is closer to the exact

value

- 19 -

TABLE III Damage characterization for the numerical simulation models

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

Case 1

Exact 650 200

Method 1 248 50~70 79~414 160~834

Method 2 248 603~667 135~241 271~486

Method 3 248 635 188 379

Case 2

Exact 675 150

Method 1 193 60~75 01~251 37~651

Method 2 193 618~682 114~222 295~575

Method 3 193 65 168 435

Case 3

Exact 675 150

Method 1 233 60~75 66~302 140~649

Method 2 233 668~738 85~196 182~419

Method 3 233 703 140 301

Case 4

Exact 900 150

Method 1 204 75~100 48~349 118~856

Method 2 204 860~951 107~216 264~530

Method 3 204 906 162 397

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1253 ms 1190~1315 ms Measured 119907119907119894119894 1253 ms

- 20 -

CHAPTER 5 EXPERIMENTAL VERIFICATION

51 Overview

In this section experimental study is investigated so that the proposed damage characteri-

zation technique Overall experimental setup and test specimen are shown in Fig 8 and 9

The experimental setup consists of test specimens data acquisition (DAQ) system and self-

sensing circuit Self-sensing circuit was installed for pulse-echo measurement using PZTs

[40-41]

PZT A(PZT C) PZT B

500

500

140

75

250

Teflon Inserted Damage (d =25)

Unit mm

Thickness = 3

PZT A(PZT C) PZT B

500

500

150

90

250

Impact Location

Unit mm

Thickness = 3

(a) Teflon inserted damage specimen

(b) Impact-induced damage specimen

Fig 8 Test specimens for the experimental verification In the Teflon inserted damage Teflon tapes with 25mm diameter were inserted at the center of the specimen The impact-induced damage was introduced to the specimen using the impact test equipment

- 21 -

511 Specimen

Two carbon composite plates with a dimension of 500times500times3mm3 were manufactured by

Nexcomes Inc by stacking 12 layers of CF3327 3K prepreg (woven fabrics type) manufac-

tured by Hankuk Fiber Inc The material properties of the specimen are presented in Table I

One of the specimens Teflon tapes with 25mm diameter was inserted in every 3 layers at the

center of the specimen to mimic the damage The other specimen is utilized for the character-

ization of real impact-induced damage

512 Data acquisition (DAQ) system and PZT

The data acquisition (DAQ) system consists of an arbitrary waveform generator (AWG) a

high speed signal digitizer (DIG) and two multiplexers (MUX) Using the 16-bits AWG

tone-burst input signal with a plusmn10 peak-to-peak voltage were generated The response signal

was measured by the 14-bit DIG For the signal-to-noise ratio reduction both pitch-catch and

pulse-echo signals were measured 150 times and averaged in the time domain 3 identical

circular PZTs of 10mm diameter manufactured by APC International Ltd were installed on

the specimen A Gaussian pulse (tone-burst) 50 kHz central frequency was excited as an in-

put signal The excitation frequency was chosen in order to avoid overlapping of the S0 and

A0 Lamb wave modes and generating of the higher modes (S1 A1hellip)

- 22 -

PZT Actuator

PZT Sensor

AWG

DIG

DAQ System Specimen

SensingActuatingMUX

1

23

4

1 Sending signal to MUX

4 Sending signal to DIG

Self-SensingCircuit

18 Unit mm

10

(a)

(b)

Fig 9 Overall experimental setup (a) Data acquisition (DAQ) system and (b) Installed circu-lar PZT transducer on the specimen with its dimension

513 Impact test

The impact-induced damage was introduced to the specimen using the impact test equip-

ment presented in Fig 10 A 5kg mass with 10mm tip was dropped from 015m height onto

the same location of the specimen The signals were obtained at each number of impacts

Since the reflected wave is not detected at the first impact case the data obtained from the

second and third impact were utilized for the analysis

10

Unit mm (a)

(b)

Fig 10 (a) Impact test equipment for introducing impact-induced damage and (b) Tip of the impact tester with 10mm diameter

- 23 -

514 Temperature effect test

In reality structures are subjected to changing environmental conditions such as tempera-

ture that affect measured signal and cause false-alarm Thus the robustness of the proposed

damage characterization technique under temperature variation in investigated (Fig 11) The

specimens were placed inside a temperature chamber and a thermocouple was installed on

the specimen to measure temperature of surface of the specimen The signal was obtained

under three temperature cases (0 20 and 50degC) and the humidity was kept at 30

Temperature Chamber

Test Specimen

Fig 11 Test specimen in temperature chamber

- 24 -

52 Teflon inserted damage

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Reflected wave from damage

(a)

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Transmitted WaveReflected Wave

Detected reflected wave

(b)

Fig 12 A0 Lamb wave mode obtained using pitch-catch method from the Teflon inserted damage specimen (a) The reflected wave from damage is shown (b) The Teflon inserted damage is detected by capturing the reflected wave

Fig 12 shows the obtained pitch-catch signal from the Teflon inserted damage specimen

and damage diagnosis result The Teflon inserted damage is detected successfully by captur-

ing the reflected wave Table IV presents the result of the Teflon inserted damage characteri-

zation 119907119907119894119894 is measured as 1213ms and the variation is assumed to plusmn 5 of the normal val-

ue (1152~1273ms) The results indicate that the proposed damage characterization approach

works well in real application

TABLE IV Damage characterization for the Teflon inserted damage specimen

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

Exact 635 25

Method 1 123 60~75 73~338 298~1220

Method 2 123 619~662 229~310 934~1188

Method 3 123 631 284 1161

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1213 ms

- 25 -

53 Impact-induced damage

11 115 12 125 13-4

-2

0

2

4

Time (ms)

Vol

tage

(mV

)

IntactImpact2Impact3

Reflected wave from damage

(a) Pitch-catch signal

11 115 12 125 13 135

-05

0

05

1

Time (ms)

Volta

ge (m

V)

IntactImpact2Impact3

Boundary reflection

Reflected wave from damage

(b) Pulse-echo signal

Fig 13 The signal obtained from the test specimen subjected to impact The impact was in-troduced at the same location of the specimen and the signal was obtained after each impacts As the number of impact is increased the transmitted A0 Lamb wave mode is delayed and attenuated

The obtained signals from impact-induced damage specimen are presented in Fig 13 It is

observed that as the number of impact is increased the transmitted A0 Lamb wave mode is

more delayed because the wave velocity in the damage area is reduced Fig 14 shows the

false-alarm test result whether the proposed damage detection algorithm works even in pris-

tine condition of the test specimen It is shown that no false alarm is occurred because no ad-

ditional reflected wave is captured

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Transmitted Wave

(a) Transmitted wave

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Reflected Wave

No reflected wave

(b) Reflected wave

Fig 14 False alarm test for the proposed damage detection algorithm in pristine condition of the specimen Reflected wave from damage is not captured false alarm is not occurred

- 26 -

TABLE V Damage characterization for the impact-induced damage specimen

of Impacts m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

2

(1470 J)

Method 1 080 65~90 75~246 470~1157

Method 2 080 801~885 114~245 711~1134

Method 3 080 843 179 1122

3

(2205 J)

Method 1 148 50~100 06~380 21~1134

Method 2 148 758~838 249~368 843~1125

Method 3 148 798 308 1044

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1213 ms

Table V presents the characterization results of the impact-induced damage case Here the

measured 119907119907119894119894 and the assumed variation are same as the artificial damage case It is observed

that as the number of impact is increased 119889119889 is increased while 119907119907119889119889 is reduced Thus it is

validated that the proposed damage characterization approach estimates the increment of the

damage well

54 Temperature effect test

541 Teflon inserted damage

11 115 12 125 13-10

-5

0

5

10

Time (ms)

Vol

tage

(mV

)

0 oC20 oC50 oC

(a) Pitch-catch signal

11 115 12 125 13 135

-1

0

1

2

3

Time(ms)

Volta

ge (m

V)

0 oC20 oC50 oC

(b) Pulse-echo signal

Fig 15 Signals obtained from the Teflon inserted damage specimen under temperature varia-tion Time delay and amplitude variation are occurred

Fig 15 shows the obtained signals from the Teflon inserted damage specimen subjected to

- 27 -

temperature variation Time delay and amplitude variations are occurred there is no addition-

al damage is introduced The analyzed results of the obtained signals are reported in Table VI

The measured 119907119907119894119894 is 1247 ms at 0 degC 1213 ms at 20 degC and 1168 ms at 50 degC respectively

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms (1152~1273 ms)

The estimated results are also changed under temperature variations the fluctuation of esti-

mated range and exact value is small

TABLE VI Temperature effect test for the Teflon inserted damage specimen

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

0 degC

Exact 635 25

Method 1 113 60~70 79~268 352~1291

Method 2 113 560~619 233~344 1034~1172

Method 3 113 606 257 1141

20 degC

Exact 635 25

Method 1 123 60~75 73~338 298~1220

Method 2 123 619~662 229~310 934~1188

Method 3 123 631 284 1161

50 degC

Exact 635 25

Method 1 137 60~75 72~328 263~1228

Method 2 137 622~687 183~299 666~1086

Method 3 137 642 264 960

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1247 ms at 0 degC 1213 ms at 20 degC and 1168 ms at 50 degC

- 28 -

542 Impact-induced damage

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Vol

tage

(mV

)

Transmitted WaveReflected Wave

No reflected wave

(a) 0 degC

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Transmitted WaveReflected Wave

No reflected wave

(b) 50 degC

Fig 16 False alarm test under temperature variations It is shown that proposed damage char-acterization algorithm is robust under temperature variations

The false alarm test in pristine condition of the impact-induced damage specimen under

temperature variation was investigated (Fig 16) It is validated that the proposed damage di-

agnosis algorithm is robust under temperature variations Fig 17 shows the obtained signals

from the impact-induced damage (impacted 3 times) specimen under temperature variation

The analyzed results of the obtained signals are reported in Table VII The measured 119907119907119894119894 in

each temperature and the assumed variation are same as the artificial damage case

11 115 12 125 13-5

0

5

Time (ms)

Volta

ge (m

V)

0 oC20 oC50 oC

(a) Pitch-catch signal

11 115 12 125 13 135

-1

0

1

2

3

Time(ms)

Vol

tage

(mV

)

0 oC20 oC50 oC

(b) Pulse-echo signal

Fig 17 Signals obtained from impact-induced damage specimen (Impact 3) under tempera-ture variation

- 29 -

TABLE VII Temperature effect test for the impact induced damage specimen (Impact 3 case)

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

0 degC

Method 1 135 75~90 72~310 267~1148

Method 2 135 705~779 264~381 978~1126

Method 3 135 763 289 1071

20 degC

Method 1 148 50~100 06~380 21~1134

Method 2 148 758~838 249~368 843~1125

Method 3 148 798 308 1044

50 degC

Method 1 153 75~100 53~415 174~1227

Method 2 153 804~889 214~336 702~1103

Method 3 153 815 320 1050

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1247 ms at 0 degC 1213 ms at 20 degC and 1168 ms at 50 degC

- 30 -

CHAPTER 6 CONCLUDING REMARKS

61 Executive summary

In this study the reference-free damage detection localization and quantification tech-

nique in a multi layered composite plate is proposed Additional reflected A0 Lamb wave

mode from the damage due to multiple reflections in damage area is used as a feature for

characterizing the damage First damage is characterized in terms of a damage index (m-

value) defined as the ratio of the physical length of the damage to the wavelength of the A0

Lamb wave mode Then the range of possible damage location length and severity are esti-

mated with constraints This procedure can be accomplished without prior baseline data To

realize the proposed technique the matching pursuit technique is modified for reduction of

the error by considering the dispersive characteristic of Lamb wave The matching pursuit

method estimates the arrival time of each wave packets accurately even two wave packets are

overlapped The numerical simulations and experimental tests including Teflon tape inserted

and impact-induced damage are investigated for the feasibility of the proposed technique

Moreover the effectiveness of the proposed technique under changing temperature is exper-

imentally verified

- 31 -

62 Suggestion for further study

The limitations of this study and the possible improvement and further studies are listed as

follow

1 Measurement of impact-induced damage length using conventional NDT technique

Additional experimental verification for impact-induced damage characterization is

needed by measuring the real length of damage using conventional NDT technique

such as ultrasonic scan or infrared camera

2 Limitation of both side collocated PZT installation In this study excitation PZTs are

installed both side and collocated for decomposing A0 Lamb wave mode However it

can produce decomposition error due to imperfection of PZT location and their installa-

tion Moreover in many cases assess to both surface of structures is not allowed such

as aircraft applications To improve this limitation a reference-free damage characteri-

zation using single side PZTs will be considered

3 Complex structure application In this study damage is characterized in simple plate-

like structure When the monitored structure becomes complex the response signals al-

so becomes complex due to its geometry and boundary Thus the investigations for

complex structure application such as stiffener spar and curvature are needed

- 32 -

REFERENCES

[1] Cawley P and Alleyne D (1996) ldquoThe use of Lamb waves for the long range inspection of large structuresrdquo Ultrasonics 34 pp287-290

[2] Ghosh T Kundu T and Karpur P (1998) ldquoEfficient use of Lamb waves for detecting defects in large platesrdquo Ultrasonics 36 pp791-801

[3] Guirguitui V (2003) ldquoLamb wave generation with piezoelectric wafer active sensors for structural health monitoringrdquo SPIE Conference on Smart Structures and Materials San Diego CA 5391 pp111-422

[4] Ragvahan A and Cesnik CE S (2004) ldquoModeling of piezoelectric-based Lamb wave generation and sensing for structural health monitoringrdquo SPIE Conference on Smart Struc-tures and Materials San Diego CA 5391 pp419-430

[5] Rose JL Pilarski A and Dirit JJ (1993) ldquoAn approach to guided wave mode selection for inspection of laminated platerdquo Journal of Reinforced Plastics and Composites 12(5) pp536-544

[6] Guo N and Cawley P (1993) ldquoThe interaction of Lamb waves with delaminations in composite laminatesrdquo Journal of Acoustical Society of America 94(4) pp 2240-2246

[7] Valdez SHD and Soutis C (2002) ldquoReal-time nondestructive evaluation of fiber com-posite laminates using low-frequency Lamb wavesrdquo Journal of Acoustical Society of Amer-ica 111(5) pp 2026-2033

[8] Yang M and Qiao P (2005) ldquoModeling and experimental detection of damage in virous materials using the pulse-echo method and piezoelectric sensoractuatorsrdquo Smart Materials and Structures 14 pp1083-1100

[9] Lestari W and Qiao P (2005) ldquoApplication of wave propagation analysis for damage identification in composite laminated beamsrdquo Journal of Composite Materials 39(22) pp1967-1984

[10] Kessler S S Spearing S M and Soutis C (2002) ldquoDamage detection in composite ma-terials using Lamb wave methodsrdquo Smart Materials and Structures 11 pp269-278

[11] Sohn H Park G Wail J R Limback N P and Farrar C R (2004) ldquoWavelet-based active sensing for delamination detection in composite structuresrdquo Smart Materials and Structures 13 pp 153-160

[12] Petculescu G Krishnaswamy S and Achenback J D (2008) ldquoGroup delay measure-ments using modally selective Lamb wave transducers for detection and sizing of delamina-tions in compositesrdquo Smart Materials and Structures 17

[13] Sohn H Park H W Law K H and Farrar C R (2007) ldquoDamage detection in compo-

- 33 -

site plates by using an enhanced time reversal methodrdquo Journal of Aerospace Engineering (ASCE) pp 141-151

[14] Sohn H (2007) ldquoEffects of environmental and operational variability on structural health monitoringrdquo Philosophical Transactions of the Royal Society Ardquo 365 pp539-560

[15] Rytter A (1993) ldquoVibration based inspection of civil engineering structuresrdquo PhD The-sis Department of Building Technology and Structural Engineering University of Aalborg Aalborg Denmark

[16] Hayashi T and Kawashima K (2002) ldquoMultiple reflections of Lamb waves at a delami-nationrdquo Ultrasonics 40 pp193-197

[17] Wang C H and Rose L R F (2003) ldquoWave reflection and transmission in beams con-taining delamination and inhomogeneityrdquo Journal of Sound and Vibration 264 pp 851-872

[18] Ramadas C Balasubramaniam K Joshi M and Krishnamurthy C V (2009) ldquoInterac-tion of the primary anti-symmetric Lamb mode (Ao) with symmetric delaminations numer-ical and experimental studiesrdquo Smart Materials and Structures 18

[19] Ramadas C Balasubramaniam K Joshi M and Krishnamurthy C V (2010) ldquoInterac-tion of guided Lamb waves with an asymmetrically located delamination in a laminated composite platerdquo Smart Materials and Structures 19

[20] Wilcox P Lowe M and Cawely P (2001) ldquoThe effect of dispersion on long-range in-spection using ultrasonic guided wavesrdquo NDTampE International 34 pp1-9

[21] Mallat S G and Zhang Z (1993) ldquoMatching pursuits with time-frequency dictionariesrdquo IEEE Tansactions on Signal Processing 41(12) pp3397-3415

[22] Zhang G Zhang S and Wang Y (2000) ldquoApplication of adaptive time-frequency de-composition in ultrasonic NDE of highly scattering materialsrdquo Ultrasonics 38 pp 961-964

[23] Hong J C Sun K H and Kim Y Y (2005) ldquoThe matching pursuit approach based on the modulated Gaussian pulse for efficient guided-wave damage inspectionrdquo Smart Mate-rials and Structures 14 pp548-560

[24] Herakovich C T (1997) Mechanics of Fibrous Composites Wiley 480 pages

[25] Reddy J N (1985) ldquoA review of the literature on finite-element modeling of laminated composite platesrdquo The Shock and Vibration Digest 17(4) pp 3-8

[26] Reddy J N (2003) Mechanics of Laminated Composite Plates and Shells Theory and Analysis CRC Press 856 pages

[27] Palacz M Krawczuk M and Ostachowicz W (2005) ldquoThe spectral finite element mod-el for analysis of flexure-shear coupled wave propagation Part 1 laminated multilayer

- 34 -

composite beamrdquo Composite Structures 68 pp37-44

[28] Zhang Y C and Harding J (1990) ldquoA numerical micromechanics analysis of the me-chanical properties of a plain weave compositerdquo Computer amp Structures 36(5) pp 839-844

[29] Nemat-Nasser S and Hori M (1999) Micromechanics Overall Properties of Heteroge-neous Materials Elsevier 810 pages

[30] Scida D Aboura Z Benzeggagh M L and Bocherens E (1999) ldquoA micromechanics model for 3D elasticity and failure of woven-fibre composite materialrdquo Composites Science and Technology 59 pp 505-517

[31] Thomas Y and Wu X H (1997) ldquoA multiscale finite element method for elliptic prob-lems in composite materials and porous mediardquo Journal of Computational Physics 134 pp 169-189

[32] Lee K H Moorthy S and Ghosh S (1999) ldquoMultiple scale computational model for damage in composite materialsrdquo Computer Methods in Applied Mechanics and Engineer-ing 172 pp 175-201

[33] Namilae S and Chandra N (2005) ldquoMultiscale model to study the effect of interfaces in carbon nanotube-based compositesrdquo Journal of Engineering Materials and Technology (ASME) 127 pp222-232

[34] Adams D E (2007) Health monitoring of Structural Materials and Components Wiley 460 pages

[35] Gopalakrishnan S (2009) ldquoModeling Aspects in Finite Elementrdquo Encyclopedia of Struc-tural Health Monitoring 2 pp791-809

[36] Palacz M Krawczuk M and Ostachowicz W (2005) ldquoThe spectral finite element mod-el for analysis of flexure-shear coupled wave propagation Part 2 Deminated multilayer composite beamrdquo Composite Structures 68 pp45-51

[37] Baik J M and Thopson R B (1984) ldquoUltrasonic scattering from imperfect interfaces A quasi-static modelrdquo Journal of Nondestructive Evaluation 4(3) pp177-196

[38] Luo H and Hanagud S (2000) ldquoDynamics of delamincated beamsrdquo International Jour-nal of Solids and Structures 37 pp1501-1519

[39] Kim S B and Sohn H (2007) ldquoInstantaneous reference-free crack detection based on polarization characteristics of piezoelectric materialsrdquo Smart Materials and Structures 16 pp2375-2387

[40] An Y K and Sohn H (2010) ldquoInstantaneous crack detection under varying temperature and static loading conditionsrdquo The Special Issue of Structural Control and Health Monitor-ing 17(7) pp 730-741

- 35 -

[41] Lee S J and Sohn H (2006) ldquoActive self-sensing scheme development for structural health monitoringrdquo Smart Materials and Structures 15(6) pp 1734-1746

- 36 -

요 약 문

복합재료 손상의 무기저 진단 및 정량화

복합재료는 재료의 경량성 내구성 고강도의 특성으로 인하여 항공기에서부터

교량에까지 널리 사용되고 있다 하지만 외부의 충격으로 인한 복합재료 내부의

손상은 재료의 강도를 감소시켜 구조물의 안정성을 저하시키는 요인이 된다 하지만

복합재료의 손상은 재료 내부에서 발생하여 육안으로는 식별하기도 힘들며 표면에서

보이는 것만으로 내부의 손상 정도를 판단하기는 불가능하다는 단점을 가지고 있다

본 연구에서는 손상 내부에서 반사로 인해 발생하는 파를 이용하여 기저자료를

사용하지 않고 특정 시점에서 얻은 자료만으로 복합재료의 손상을 진단하고 정량화

할 수 있는 기법을 제안한다 제안된 기법의 적용성을 검증하기 위하여 유한요소법을

이용한 수치해석을 수행하였고 Teflon tape 이 삽입된 시편과 실제 충격하중으로

인한 손상을 입은 시편을 이용한 실험을 진행하였다 수치해석 및 실험 결과 제안된

기법이 복합재료의 손상을 성공적으로 진단 및 정량화할 수 있음을 확인하였다 또한

다양한 외부 온도 환경에서 제안된 기법의 환경영향 실험을 수행하였고 다양한

환경에서 손상의 진단 및 정량화가 성공적으로 이루어짐을 검증하였다

핵심어 무기저 손상진단 및 정량화 압전센서 복합재료 유한요소법 Matching pursuit

method

- 37 -

감 사 의 글

기대와 두려움을 안고 시작한 석사과정이 어느덧 2 년 이라는 세월 속에서 추억으로 자리잡아

가고 있습니다 학부생 기간을 포함한 지난 3 년 동안 SHM 이라는 새로운 분야를 접하고 연구를

수행 할 수 있도록 도와주신 고마운 분들이 있어 그 감사를 표하고자 합니다

우선 저의 많은 부족한 점을 일깨워 주시고 연구에 대한 뜻을 더욱 확고히 세울 수 있게 해

주신 지도교수님이신 손 훈 교수님께 깊은 감사의 뜻을 전합니다 짧은 기간이었지만 교수님의

도움으로 많은 것을 배우고 느끼고 깨달을 수 있었습니다 또한 본 연구를 수행할 수 있도록 지

원해주신 국토해양부 (U-City 석 박사과정 지원사업)에 깊은 감사 드립니다

때로 힘들고 지칠 때 항상 저를 믿고 응원해 주시는 가족이 있었기에 지금의 글을 쓸 수 있었

습니다 또한 지금 연구실에서 함께 생활하고 있는 윤규형 민구형 철민이형 현석이형 진열이형

호민이형 병진이 선혜누나와 성대에 계신 창길이형 미국에 계신 의재형과 Debaditiya 에게도 감

사의 말씀을 전합니다 그리고 모두 나열할 수는 없지만 카이스트 생활의 활력소가 되어준 카이

스트 건설 및 환경공학과 선후배들 직원 분들과 SOS 멤버들에게도 감사를 표합니다

앞서 거론한 많은 이들의 도움이 없었다면 저 또한 지금 이 자리에 있을 수 없었을 것입니다

많이 부족하기에 더 노력하고 배우고자 했던 초심을 다시 한번 생각하며 더 큰 결실을 맺기 위해

앞으로 나아가겠습니다 도움을 주신 모든 분들께 이 감사의 글을 바칩니다

2011년 6월 카이스트에서 임형진 올림

- 38 -

CURRICULUM VITAE

Hyung Jin Lim

Research Assistant

The Civil and Environmental Engineering Department

Korea Advanced Institute of Science and Technology (KAIST)

291 Daehak-Ro Guseong-Dong Yuseong-Gu

Daejeon Republic of Korea 305-701

Phone (82)+42-350-3665 Fax (82)+42-350-3610

Email limhj87gmailcom

RESEARCH INTERESTS

Damage Detection Localization and Quantification on Composite and Metallic Structures Pattern Recognition Finite Element Modeling (FEM)

EDUCATION

2009 ndash 2011 MS Civil and Environmental Eng Korea Advanced Institute of Science and Technology (KAIST) Korea Rep

2006 ndash 2009 BS Civil and Environmental Eng Korea Advanced Institute of Science and Technology (KAIST) Korea Rep

JOURNAL PUBLICATIONS The corresponding author is underlined 1 Hyung Jin Lim Hoon Sohn ldquoReference-free damage detection localization and quantification

using in compositerdquo in preparation for Journal of the Acoustical Society of America 2011

2 Chul Min Yeum Hoon Sohn Jeong Bum Ihn Hyung Jin Lim ldquoInstantaneous delamination detection in a composite using a dual piezoelectric transducer networkrdquo Submitted to Compo-sites Science and Technology 2011

3 Yun-Kyu An Hyung Jin Lim Min Koo Kim Jin Yeol Yang Hoon Sohn Chang Guen Lee ldquoInvestigation of applicability of local reference-free damage detection techniques to in-situ

- 39 -

bridgesrdquo Submitted to ASCE Journal of Structural Engineering 2011

4 Hyung Jin Lim Min Koo Kim Hoon Sohn Chan Yik Park ldquoImpedance based damage detec-tion under varying temperature and loading conditionsrdquo Submitted to NDTampE International 2011

5 Hyung Jin Lim Hoon Sohn ldquoQuantitative estimation of transmitted and reflected lamb waves at discontinuityrdquo Journal of the Korean Society for Nondestructive Testing Vol30 No 4 pp1-8 2010

CONFERENCE PROCEEDINGS

The corresponding author is underlined 1 Hyung Jin Lim and Hoon Sohn ldquoReference-free damage detection localization and quantifi-

cation in compositesrdquo ASCE Engineering Mechanics Institute Conference Boston June 2-4 2011

2 Hyung Jin Lim and Hoon Sohn ldquoReference-free delamination quantification within multi-layer composite materialsrdquo KSNT Spring Conference May 26-27 Gyeong-Ju Korea 2011

3 Min Koo Kim Hyun Jin Lim Hoon Sohn Chan Yik Park ldquoImpedance-based bolt loosening detection under varying temperature and loadingrdquo Asian Pacific Workshop on Structural Health Monitoring Tokyo Japan November 30-December 2 2010

4 Hyung Jin Lim Chul Min Yeum and Hoon Sohn ldquoModeling of Impact-induced Delamination in a Multilayer Composite Platerdquo The 23rd KKCNN Symposium on Civil Engineering Taipei (Taiwan) November 13-15 2010

5 Yun-Kyu An Hyung Jin Lim and Hoon Sohn ldquoTemperature independent Delamination Detec-tion using data normalizationrdquo The 3rd International Conference on Multi-Functional Materi-als and Structures Chonbuk National Univ Jeonju Korea September 14-18 2010

6 Hoon Sohn Yun-Kyu An Chang Gil Lee Min Koo Kim Hyung Jin Lim ldquoApplication of online NDT techniques to decommissioned bridge testingrdquo the 5th European Workshop on Structural Health Monitoring Sorrento Italy June 29-July 02 2010

7 Min Koo Kim Hyung Jin Lim Hoon Sohn Chan Yik Park ldquoImpedance based damage diag-nosis of a complex composite aircraft wing under changing loading and temperature condi-tionsrdquo the 5th European Workshop on Structural Health Monitoring Sorrento Italy June 29-July 02 2010

8 Yun-Kyu An Min Koo Kim Hyung Jin Lim and Hoon Sohn ldquoValidations of reference-free crack detection techniques through a decommissioned bridge testrdquo 2010 COSEIK Annual Con-ference Jeju Korea April 8-9 2010

- 40 -

PARTICIPANT PROJECTS

1 Development of Smart Steel Members for Online Steel Structure Monitoring POSCO Corporation (Funded 41000000 KRW (41000 USD) for 010511 to 123111)

2 Development of Enabling Technologies for Structural Health Monitoring Korean Agency for Defense Development (Funded 240000000 KRW (240000 USD) out of total 3200000000 KRW (3200000 USD) for 060108-053111)

3 Development of On-board SHM Technologies for Composite Air Vehicles The Boeing Company (Funded 400000 USD (400000000 KRW) for 080108 to 073111)

EDUCATIONAL EXPERIENCES

Fall 2009

Teaching Assistant CE207 ldquoElementary Structural Engineering and Laboratoryrdquo KAIST 3 unit undergraduate course

EXPERIMENTAL EXPERIENTCES

2010 Monitoring of Real Aircraft Composite Wing Segment Korea Agency for Defense De-velopment (ADD) Daejeon Korea Real time monitoring of composite air vehicles fund-ed by Korea Agency for Defense Development (ADD)

2009-2011

Field Monitoring of Real Bridge Structures Gimpo (Ramp G bridge) amp Icheon (Sam-seung bridge) Korea US-Korea collaborative research for bridge monitoring testbeds

HONORS amp AWARDS

Scholarships for outstanding students granted KAIST (2007-2009) The three prominent students in the civil and environmental engineering department of KAIST were selected as recipients This scholarship provided 2000$ per year for three years

- 41 -

Page 15: 복합재료 손상의 무기저 진단 및 정량화 - KAISTssslab.kaist.ac.kr/article/pdf/Thesis_2.pdf · 1 Declaration of Ethical Conduct in Research: I, as a graduate student

14 Organization

The organization of this thesis is summarized as follows

Chaper1 A general introduction including motivation literature review objective and

uniqueness of this study is described

Chaper2 Interactions between Lamb wave mode and damage in composites such as time

delay attenuation and multiple reflections are summarized Damage index (m-value) for

damage characterization is defined

Chaper3 Algorithms for the reference-free damage characterization technique using m-

value are developed Modified matching pursuit for this study is also described

Chaper4 Verification of the algorithm using numerical simulation is presented Various

modeled damage cases are characterized

Chaper5 Experimental verifications are described Two damage cases (Teflon inserted and

impact-induced damage) are investigated for the feasibility of the proposed damage charac-

terization technique Temperature effect test using temperature chamber is also carried out

Chaper6 Conclusion and summary of the present study is reported In advance suggestions

and future work recommendations are represented

- 4 -

CHAPTER 2 INTERACTION OF LAMB WAVE WITH DAMAGE

21 A0 Lamb wave mode delay and attenuation in the time domain

Time delay and attenuation of the A0 mode Lamb wave after passing through the damage

area due to effective thickness reduction were reported in previous researches [812] Moreo-

ver the attenuation of the A0 Lamb wave mode is occurred by degrading of shear modulus of

in the damage area However the time delay and amplitude variation of the A0 Lamb wave

mode also occurred even when the structure is health condition by the material property

changing due to temperature variation of the structure (Fig 1) Thus conventional damage

characterization methods using the time delay and attenuation of A0 Lamb wave mode as a

feature can cause false-alarm in real application

11 115 12 125 13-10

-5

0

5

10

Time (ms)

Vol

tage

(mV

)

IntactDamage

Reflected wave from Damage

(a)

11 115 12 125 13-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

0 oC20 oC50 oC

(b)

Fig 1 Lamb wave propagation characteristics (a) A0 Lamb wave mode time delay and atten-uation after passing through a damage (b) The time delay and amplitude variation of A0 Lamb wave mode under temperature variation in pristine condition of structure Because of this conventional damage detection techniques in composites have limitations for real struc-ture application

- 5 -

22 Multiple reflections in the damage area

Standing wave patterns due to multiple reflections of A0 Lamb wave mode in the damage

area were observed through numerical simulations and ultrasonic wave-field imaging using a

scanning laser vibrometer [16] The formation of standing waves in damage is illustrated

through Fig 2 When the incident waves enter the damage area a significant portion of the

waves are reflected back from the exit Then the reflected waves travel through the damage

and undergo reflection again at the original entrance As a consequence of multiple reflec-

tions at the entrance and the exit a considerable amount of ultrasonic energy is trapped inside

the damage area [17-19]

Multiple Reflections

Incident Wave Transmitted Wave

Damage

Fig 2 Multiple reflections in the damage area The multiple reflections of A0 Lamb waves in the damage area are utilize as a feature for characterization of the damage in this study

- 6 -

CHAPTER 3 THEORETICAL DEVELOPMENT

31 Overview

Damage

l = l1 + d + l2

d = vdtd l1 = vit1 l2 = vit2

t = t1 + td + t2

Excitation and Sensing PZT Sensing PZT

l

Transmitted Wave

Reflected Waves

Incident Wave

Fig 3 Parameters and notations of the proposed damage characterization technique Damage is detected by identifying the additional reflected wave from the damage due to multiple re-flections (dashed lines) Localization and quantification of the damage is carried out by measuring the arrival time of transmitted and reflected waves from the damage

In this section the reference-free damage characterization algorithm is proposed Damage

diagnosis and quantification are achieved by identifying the additional reflected waves due to

multiple reflections in the damage area (dashed lines in Fig 3) Moreover damage is local-

ized using additional information from pulse-echo method implement For the localization

and the quantification of damage the variables such as the path length and A0 Lamb wave

mode velocity in damage area are defined (Fig 3) For example 119897119897 is expressed as

119897119897 = 1198971198971 + 119889119889 + 1198971198972 (1)

where 1198971198971 119889119889 and 1198971198972 are the distance from the excitation PZTs to the entrance of damage the

damage length and the distance from the exit of damage to the sensing PZT respectively

- 7 -

Moreover 119907119907119894119894 and 119907119907119889119889 are also defined as the A0 Lamb wave mode velocity in pristine area

and damage area respectively Note that the whole procedure of the damage characterization

is accomplished based solely on the signals obtained from the current state of the structure

32 Damage index (m-value)

The proposed reference-free damage characterization technique utilizes the waves reflect-

ed additionally from the damage area as a feature The arrival time difference of the transmit-

ted and reflected A0 Lamb wave modes reflected from the exit and the entrance of the delam-

ination area one times each (∆119905119905) contains the information of the physical length of the dam-

age (119889119889) and A0 Lamb wave mode velocity in the damage area (119907119907119889119889) The wave velocity in the

damage area which is related to the shear modulus (G) represents the severity of the damage

∆119905119905 = 2119905119905119889119889 =2119889119889119907119907119889119889

(2)

However based on the measured data available the 119889119889 and 119907119907119889119889 (or G) cannot be uniquely

determined Fig 4 shows two FEM models which have different physical length and wave

velocity (G) of the damage Detail explanation of delamination modeling technique will be

covered in section 4 The similar signal in the time domain and ∆119905119905 are obtained from the

two different damage cases

- 8 -

Damage 1

Damage 2

20

Unit mm15

90 G value reduction

943 G value reduction

m-value = 248

m-value = 248

015 02 025 03 035

-2

-1

0

1

2

Time (ms)

Ampl

itude

Damage 1Damage 2

Fig 4 Two different FEM models which have different 119889119889 and 119907119907119889119889 (or G) This represents that the damage cannot be characterized uniquely based on the measured data available

Thus in this study the damage index (m-value) which can be uniquely defined based on

the available data is introduced The m-value is defined as the ratio of the physical length of

the damage and wavelength of the A0 Lamb wave mode in the damage area (120582120582119889119889)

119898119898 =119889119889120582120582119889119889

=119889119889119889119889119907119907119889119889

(3)

In many cases the transmitted and reflected waves are overlapped and it is hard to decom-

pose and calculate the arrival times of them Matching pursuit method is an alternative way

to accurately decompose of the overlapped waves and estimate the parameters of each wave

such as arrival time (Fig 5) Detail explanation of the matching pursuit method is covered in

section 36

015 02 025 03-2

-1

0

1

2

Time (ms)

Am

plitu

de

OriginalApproximated

(a)

015 02 025 03-2

-1

0

1

2

Time (ms)

Ampl

itude

Wave 1Wave 2

(b)

Fig 5 (a) Example of the overlapped signal with comparing approximated signal using the matching pursuit method (b) Decomposed signal using the matching pursuit method The matching pursuit method provides a conventional way to decompose and identify each wave

- 9 -

33 Damage detection algorithm (Level 1)

The damage diagnosis is accomplished by extracting the additional reflected A0 Lamb

wave modes from damage due to multiple reflections If A0 Lamb wave modes are decom-

posed perfectly it is trivial that one transmitted A0 Lamb wave mode is always captured in

the response signal If damage exists other waves in the response signal behind the transmit-

ted A0 Lamb wave mode are presented Since this approach does not rely on previously col-

lected data the diagnosis is independent from environmental variation such as temperature

changes

It is reported that the dispersion of Lamb waves cause wave-packets to spread out in space

and time as they propagate through a structure [20] Thus the reflected waves from the dam-

age have longer pulse duration than the transmitted wave due to their longer propagating dis-

tance In this study the matching pursuit method is modified to capture more spread wave

than the transmitted A0 Lamb wave mode Detail explanation of the modified matching pur-

suit method is described in latter section

Let 1199041199041 and 1199041199042 as the pulse duration (scale) of the transmitted and additional reflected A0

Lamb wave mode respectively

ldquoIf 1199041199042 gt 1199041199041 damage exists

Otherwise there is no damagerdquo

- 10 -

34 Damage quantification algorithm (Level 3)

As mentioned in the previous section damage is quantized by estimating 119889119889 and 119907119907119889119889 using

pitch-catch method The procedure to estimate 119889119889 and 119907119907119889119889 is

(i) Measure 119897119897 119905119905 and 119905119905119889119889

(ii) Assume the range of 119889119889 (0 lt 119889119889 lt 119897119897)

(iii) Calculate 119907119907119894119894 and 119907119907119889119889 with a constraints (0 lt 119907119907119889119889 lt 119907119907119894119894) (Eq 4 and 5)

(iv) Estimate m-value and find a feasible range of 119889119889 and 119907119907119889119889

119907119907119894119894 =119897119897 minus 119889119889119905119905 minus 119905119905119889119889

(4)

119907119907119889119889 =119889119889119905119905119889119889

(5)

35 Damage localization algorithm (Level 2amp3)

For damage localization 1198971198971 is estimated by measuring the arrival time from the excitation

PZT to the entrance of delamination (1199051199051) with additional pulse-echo method implementation

The procedure is similar with the previous damage quantification algorithm

(i) Measure 119897119897 119905119905 119905119905119889119889 and 1199051199051

(ii) Assume the range of 1198971198971 (0 lt 1198971198971 lt 119897119897)

(ii) Compute possible combinations of 1198971198972 119889119889 and 119907119907119889119889 with constraints

(0 lt 119907119907119889119889 lt 119907119907119894119894 119889119889 gt 0) (Eq 5 6 and 7)

- 11 -

1198971198972 = 1198971198971119905119905 minus 1199051199051 minus 119905119905119889119889

1199051199051 (6)

119889119889 = 119897119897 minus 1198971198971 minus 1198971198972 (7)

Since a PZT for pulse-echo method acts as both an actuator and a sensor the obtained sig-

nal contains both the excitation and the response signal Thus the reflected wave from the

entrance of damage can be buried into the tail of excitation signal and cause error on arrival

time of reflected wave from damage Even in this case damage can be localized and quan-

tized using the constraints When first wave packet reflected from the entrance of damage is

buried into the excitation signal the next wave packet reflected from the exit of damage can

be utilized alternatively If the algorithm regards the second reflected wave as the first re-

flected wave there are no combinations of estimated variable meeting the constraints Then

the algorithm repeats the process by regarding the wave as the second reflected wave the

constraints can select proper combination of the estimated variables

The estimated values have range bounded by the constraints When additional information

of 119907119907119894119894 is utilized the damage can be characterized more accurately In this study three meth-

ods are proposed and verified as information level of 119907119907119894119894

Method 1 Solely based on the current data measurement (Reference-free)

Method 2 Limit the possible range of 119907119907119894119894 (Prior knowledge of the system)

Method 3 Measure 119907119907119894119894 (Additional data need to be collected)

Method 1 described in this section can be utilized without any information of 119907119907119894119894 Meth-

od 2 produces more accurate estimation results using prior knowledge of the system such as

- 12 -

119907119907119894119894 variation under temperature changes Method 3 gives unique estimated value using meas-

ured 119907119907119894119894 from other path in the same time In section 4 and 5 damage is characterized using

these three methods and the results are compared each another

36 Matching pursuit method considering the dispersion effect of Lamb

wave

Matching pursuit method has been developed for signal compression approximation and

characterization of obtained signal and function [21-23] The key concept of the matching

pursuit method is that any signal (function) can be decomposed into a linear expansion of

signals selected from a pool of basis signals (functions) called lsquodictionaryrsquo

Let 119891119891 be a function to be analyzed then 119891119891 can be approximated a set of functions select-

ed from the dictionary 119863119863 in order to best match This is done by successive approximations

with orthogonal projections Let1198921198921205741205740 isin 119863119863 the function 119891119891 can be decomposed into

119891119891 = lang1198911198911198921198921205741205740rang1198921198921205741205740 + 119877119877119891119891 (8)

where 119877119877119891119891 is the residual function after approximating 119891119891 using 1198921198921205741205740 1198921198921205741205740 is orthogonal to

119877119877119891119891 hence

1198911198912 = lang1198911198911198921198921205741205740rang2

+ 1198771198771198911198912 (9)

To minimize 119877119877119891119891 1198921198921205741205740 is chosen to maximize lang1198911198911198921198921205741205740rang The matching pursuit method is

an iterative algorithm that decomposes the residue 119877119877119891119891 by projecting it on a function of 119863119863

that matches 119877119877119891119891 almost at best as it was done for 119891119891 The residue 119877119877119899119899119891119891 is decomposed into

119877119877119899119899119891119891 = lang119877119877119899119899119891119891119892119892120574120574119899119899rang119892119892120574120574119899119899 + 119877119877119899119899+1119891119891 (10)

- 13 -

the decomposition is carried up to the order 119898119898 119891119891 can be expressed as

119891119891 = lang119877119877119899119899119891119891119892119892120574120574119899119899rang119892119892120574120574119899119899

119898119898minus1

119899119899=0

+ 119877119877119898119898119891119891 (11)

The original function 119891119891 is decomposed into a sum of dictionary elements that are chosen to

best match its residues

In this study the matching pursuit method is utilized to accurately decompose of the over-

lapped Gaussian pulses and estimate the parameters of the pulse which used for damage

characterization The 119894119894119905119905ℎ Gaussian pulse in the dictionary for approximating the obtained

signal is expressed as

119892119892120574120574119894119894 = 119891119891(119860119860119894119894 119906119906119894119894 119904119904119894119894 119889119889119894119894120601120601119894119894) (12)

where 119860119860119894119894119906119906119894119894119904119904119894119894119889119889119894119894 and 120601120601119894119894 represent amplitude arrival time scale (pulse duration) excita-

tion(central) frequency and phase of the Gaussian pulse respectively The parameters of the

pulse are estimated using a nonlinear regression based optimization method (Gauss-Newton

method) based on an initial guess of the parameters Detail procedure for estimating the pa-

rameters of the pulse is expressed in Hong et al [23]

Among the estimated parameters of the pulses pulse duration (scale) represents the disper-

sion effect If a wave-packet is spread out the estimated pulse duration is increased Using

this feature the matching pursuit method is modified to capture waves which have longer

pulse duration than transmitted A0 Lamb wave mode If the estimate pulse duration of a wave

is less than the transmitted wave the modified matching pursuit method regards it as error

- 14 -

CHAPTER 4 NUMERICAL SIMULATION

41 Composite modeling method

Composite materials can be modeled from nano to macro level depends on its engineering

purpose In general four methods are utilized for modeling of composite materials First

composite is modeled as a homogenous material expressed in terms of effective material

properties [24] The effective material properties are estimated using the modulus and the

volume fraction of each constituent theoretically (Rule of mixture) and measured experimen-

tally Second a model is established that each laminate which have own material properties

are totally bonded together In this model the stacking sequence and direction of each lami-

nates are considered additionally [25-27] Third method for composite modeling in micro

level called micromechanics [28-30] This model considers the interaction between fibers and

matrix in laminate Last multiscale modeling which the scope of modeling is extended to

molecular or quantum level of constituents [31-33] This model starts from molecular me-

chanics or molecular dynamics then the scale of model leads to the nano micro and macro

level by incorporating small scale models As the scale of models is smaller it provides more

logical way than other large scale of models while the computational cost is more expensive

The scope of this study is focused on the dynamic behavior of composite structure the beam

or plate model was utilized for the numerical simulation

- 15 -

42 Damage modeling method

Generally three damage models have been developed and utilized in the real applications

First the stiffness reduction model called as the equivalent element model was introduced

[17 34-35] This model is simply realized by replacing the region around damage area by a

region with adjusted material properties (Youngrsquos modulus shear modulus density etc)

globally Next the duplicated node model is done by disconnecting the element connectivity

assuming there is no interaction among the delaminated surface This model can be utilized

when the location and the extension of damage is important issue [34-36] Last the contact

mechanics model using the principles of contact mechanics the two interfacial forces (con-

tact resistance and sliding resistance) in the delaminated surface [37-38] In this model the

delaminated surfaces are modeled as nonlinear functions of the interlaminar normal strain

and the interlaminar shear strain This model is computationally expensive while provides

more logical way than other models There is however no universally accepted model for

general impact-induced damage in composites In the case of impact-induced damage it is

impossible to predict or estimate the location and number of delaminated layers Thus from

the uncertainty of impact-induced damage the equivalent element model can be considered

as the effective damage model

- 16 -

43 Numerical simulation description

TABLE I Material property of simulation model and test specimen

Tensile Strength 657 MPa Tensile Modulus 59 GPa Shear Strength 804 MPa Shear Modulus 24 Gpa Specific Gravity 16

PZT A

PZT C

PZT BDamage

3

250 125 125

500 Unit mm

Fig 6 Description of the numerical simulation model

The feasibility of the proposed damage characterization technique is verified by 2-D FEM

simulation using MSC PATRAN and NASTRAN Fig 6 presents the dimension PZTs and

damage location of the simulation model In the model 05mm x 05mm square shell element

was used for modeling of the A0 mode whose flexural wave characteristic The element size

was determined by considering the wavelength of the propagation Lamb wave modes The

material properties of the simulation model and damage cases are presented in Table I and

Table II respectively

PZT A and C act as both actuator and sensor for collecting pulse-echo signal They were

installed on the plate which two of them are both-side collocated for the A0 Lamb wave

- 17 -

mode decomposition [39-40] PZT B was only used as a sensor for collecting pitch-catch sig-

nal For simulating the excitation and sensing of PZTs the pin-force model was adopted that

an input force apples to the end nodes of excitation PZT and averaging element strains corre-

sponding sensing area A Gaussian pulse (tone-burst) 50 kHz central frequency was excited

as an input signal

TABLE II Damage cases for numerical simulation

119949119949120783120783 (mm) 119941119941 (mm) G value reduction ()

Case 1 650 200 90 darr

Case 2 675 150 90 darr

Case 3 675 150 93 darr

Case 4 900 150 90 darr

- 18 -

44 Numerical simulation results

015 02 025 03 035

-2

0

2

Time (ms)

Am

plitu

de

IntactDamage

Reflected wave from damage

(a) Pitch-catch signal

015 02 025 03 035

-2

0

2

Time (ms)

Am

plitu

de

IntactDamage

Reflected waves from damage

(b) Pulse-echo signal Fig 7 Comparison of A0 Lamb wave mode signals from numerical simulation between intact and damage cases Time delay attenuation and multiple reflections due to damage simulated successfully using the damage modeling method described in the previous section

Fig 7 shows that comparison of A0 Lamb wave mode signals between intact and one of

damage cases in the simulation The features of A0 Lamb wave mode passing through dam-

age area described in section 2 are successfully simulated Thus the equivalent element

method is proper for modeling damage in composites In the damage cases the additional

waves reflected from damage is observed in both pitch-catch and pulse echo signal

Table III presents the result of the proposed damage characterization algorithm using the

numerical simulation cases shown in Table II In lsquoMethod 2rsquo 119907119907119894119894 variation is assumed to plusmn

5 of the normal value (1190~1315ms) The normal value of 119907119907119894119894 for lsquoMethod 3rsquo is meas-

ured as 1253 ms In each case the damage is characterized using the proposed technique

Note that as the information of 119907119907119894119894 is specified the estimated value is closer to the exact

value

- 19 -

TABLE III Damage characterization for the numerical simulation models

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

Case 1

Exact 650 200

Method 1 248 50~70 79~414 160~834

Method 2 248 603~667 135~241 271~486

Method 3 248 635 188 379

Case 2

Exact 675 150

Method 1 193 60~75 01~251 37~651

Method 2 193 618~682 114~222 295~575

Method 3 193 65 168 435

Case 3

Exact 675 150

Method 1 233 60~75 66~302 140~649

Method 2 233 668~738 85~196 182~419

Method 3 233 703 140 301

Case 4

Exact 900 150

Method 1 204 75~100 48~349 118~856

Method 2 204 860~951 107~216 264~530

Method 3 204 906 162 397

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1253 ms 1190~1315 ms Measured 119907119907119894119894 1253 ms

- 20 -

CHAPTER 5 EXPERIMENTAL VERIFICATION

51 Overview

In this section experimental study is investigated so that the proposed damage characteri-

zation technique Overall experimental setup and test specimen are shown in Fig 8 and 9

The experimental setup consists of test specimens data acquisition (DAQ) system and self-

sensing circuit Self-sensing circuit was installed for pulse-echo measurement using PZTs

[40-41]

PZT A(PZT C) PZT B

500

500

140

75

250

Teflon Inserted Damage (d =25)

Unit mm

Thickness = 3

PZT A(PZT C) PZT B

500

500

150

90

250

Impact Location

Unit mm

Thickness = 3

(a) Teflon inserted damage specimen

(b) Impact-induced damage specimen

Fig 8 Test specimens for the experimental verification In the Teflon inserted damage Teflon tapes with 25mm diameter were inserted at the center of the specimen The impact-induced damage was introduced to the specimen using the impact test equipment

- 21 -

511 Specimen

Two carbon composite plates with a dimension of 500times500times3mm3 were manufactured by

Nexcomes Inc by stacking 12 layers of CF3327 3K prepreg (woven fabrics type) manufac-

tured by Hankuk Fiber Inc The material properties of the specimen are presented in Table I

One of the specimens Teflon tapes with 25mm diameter was inserted in every 3 layers at the

center of the specimen to mimic the damage The other specimen is utilized for the character-

ization of real impact-induced damage

512 Data acquisition (DAQ) system and PZT

The data acquisition (DAQ) system consists of an arbitrary waveform generator (AWG) a

high speed signal digitizer (DIG) and two multiplexers (MUX) Using the 16-bits AWG

tone-burst input signal with a plusmn10 peak-to-peak voltage were generated The response signal

was measured by the 14-bit DIG For the signal-to-noise ratio reduction both pitch-catch and

pulse-echo signals were measured 150 times and averaged in the time domain 3 identical

circular PZTs of 10mm diameter manufactured by APC International Ltd were installed on

the specimen A Gaussian pulse (tone-burst) 50 kHz central frequency was excited as an in-

put signal The excitation frequency was chosen in order to avoid overlapping of the S0 and

A0 Lamb wave modes and generating of the higher modes (S1 A1hellip)

- 22 -

PZT Actuator

PZT Sensor

AWG

DIG

DAQ System Specimen

SensingActuatingMUX

1

23

4

1 Sending signal to MUX

4 Sending signal to DIG

Self-SensingCircuit

18 Unit mm

10

(a)

(b)

Fig 9 Overall experimental setup (a) Data acquisition (DAQ) system and (b) Installed circu-lar PZT transducer on the specimen with its dimension

513 Impact test

The impact-induced damage was introduced to the specimen using the impact test equip-

ment presented in Fig 10 A 5kg mass with 10mm tip was dropped from 015m height onto

the same location of the specimen The signals were obtained at each number of impacts

Since the reflected wave is not detected at the first impact case the data obtained from the

second and third impact were utilized for the analysis

10

Unit mm (a)

(b)

Fig 10 (a) Impact test equipment for introducing impact-induced damage and (b) Tip of the impact tester with 10mm diameter

- 23 -

514 Temperature effect test

In reality structures are subjected to changing environmental conditions such as tempera-

ture that affect measured signal and cause false-alarm Thus the robustness of the proposed

damage characterization technique under temperature variation in investigated (Fig 11) The

specimens were placed inside a temperature chamber and a thermocouple was installed on

the specimen to measure temperature of surface of the specimen The signal was obtained

under three temperature cases (0 20 and 50degC) and the humidity was kept at 30

Temperature Chamber

Test Specimen

Fig 11 Test specimen in temperature chamber

- 24 -

52 Teflon inserted damage

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Reflected wave from damage

(a)

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Transmitted WaveReflected Wave

Detected reflected wave

(b)

Fig 12 A0 Lamb wave mode obtained using pitch-catch method from the Teflon inserted damage specimen (a) The reflected wave from damage is shown (b) The Teflon inserted damage is detected by capturing the reflected wave

Fig 12 shows the obtained pitch-catch signal from the Teflon inserted damage specimen

and damage diagnosis result The Teflon inserted damage is detected successfully by captur-

ing the reflected wave Table IV presents the result of the Teflon inserted damage characteri-

zation 119907119907119894119894 is measured as 1213ms and the variation is assumed to plusmn 5 of the normal val-

ue (1152~1273ms) The results indicate that the proposed damage characterization approach

works well in real application

TABLE IV Damage characterization for the Teflon inserted damage specimen

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

Exact 635 25

Method 1 123 60~75 73~338 298~1220

Method 2 123 619~662 229~310 934~1188

Method 3 123 631 284 1161

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1213 ms

- 25 -

53 Impact-induced damage

11 115 12 125 13-4

-2

0

2

4

Time (ms)

Vol

tage

(mV

)

IntactImpact2Impact3

Reflected wave from damage

(a) Pitch-catch signal

11 115 12 125 13 135

-05

0

05

1

Time (ms)

Volta

ge (m

V)

IntactImpact2Impact3

Boundary reflection

Reflected wave from damage

(b) Pulse-echo signal

Fig 13 The signal obtained from the test specimen subjected to impact The impact was in-troduced at the same location of the specimen and the signal was obtained after each impacts As the number of impact is increased the transmitted A0 Lamb wave mode is delayed and attenuated

The obtained signals from impact-induced damage specimen are presented in Fig 13 It is

observed that as the number of impact is increased the transmitted A0 Lamb wave mode is

more delayed because the wave velocity in the damage area is reduced Fig 14 shows the

false-alarm test result whether the proposed damage detection algorithm works even in pris-

tine condition of the test specimen It is shown that no false alarm is occurred because no ad-

ditional reflected wave is captured

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Transmitted Wave

(a) Transmitted wave

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Reflected Wave

No reflected wave

(b) Reflected wave

Fig 14 False alarm test for the proposed damage detection algorithm in pristine condition of the specimen Reflected wave from damage is not captured false alarm is not occurred

- 26 -

TABLE V Damage characterization for the impact-induced damage specimen

of Impacts m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

2

(1470 J)

Method 1 080 65~90 75~246 470~1157

Method 2 080 801~885 114~245 711~1134

Method 3 080 843 179 1122

3

(2205 J)

Method 1 148 50~100 06~380 21~1134

Method 2 148 758~838 249~368 843~1125

Method 3 148 798 308 1044

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1213 ms

Table V presents the characterization results of the impact-induced damage case Here the

measured 119907119907119894119894 and the assumed variation are same as the artificial damage case It is observed

that as the number of impact is increased 119889119889 is increased while 119907119907119889119889 is reduced Thus it is

validated that the proposed damage characterization approach estimates the increment of the

damage well

54 Temperature effect test

541 Teflon inserted damage

11 115 12 125 13-10

-5

0

5

10

Time (ms)

Vol

tage

(mV

)

0 oC20 oC50 oC

(a) Pitch-catch signal

11 115 12 125 13 135

-1

0

1

2

3

Time(ms)

Volta

ge (m

V)

0 oC20 oC50 oC

(b) Pulse-echo signal

Fig 15 Signals obtained from the Teflon inserted damage specimen under temperature varia-tion Time delay and amplitude variation are occurred

Fig 15 shows the obtained signals from the Teflon inserted damage specimen subjected to

- 27 -

temperature variation Time delay and amplitude variations are occurred there is no addition-

al damage is introduced The analyzed results of the obtained signals are reported in Table VI

The measured 119907119907119894119894 is 1247 ms at 0 degC 1213 ms at 20 degC and 1168 ms at 50 degC respectively

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms (1152~1273 ms)

The estimated results are also changed under temperature variations the fluctuation of esti-

mated range and exact value is small

TABLE VI Temperature effect test for the Teflon inserted damage specimen

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

0 degC

Exact 635 25

Method 1 113 60~70 79~268 352~1291

Method 2 113 560~619 233~344 1034~1172

Method 3 113 606 257 1141

20 degC

Exact 635 25

Method 1 123 60~75 73~338 298~1220

Method 2 123 619~662 229~310 934~1188

Method 3 123 631 284 1161

50 degC

Exact 635 25

Method 1 137 60~75 72~328 263~1228

Method 2 137 622~687 183~299 666~1086

Method 3 137 642 264 960

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1247 ms at 0 degC 1213 ms at 20 degC and 1168 ms at 50 degC

- 28 -

542 Impact-induced damage

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Vol

tage

(mV

)

Transmitted WaveReflected Wave

No reflected wave

(a) 0 degC

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Transmitted WaveReflected Wave

No reflected wave

(b) 50 degC

Fig 16 False alarm test under temperature variations It is shown that proposed damage char-acterization algorithm is robust under temperature variations

The false alarm test in pristine condition of the impact-induced damage specimen under

temperature variation was investigated (Fig 16) It is validated that the proposed damage di-

agnosis algorithm is robust under temperature variations Fig 17 shows the obtained signals

from the impact-induced damage (impacted 3 times) specimen under temperature variation

The analyzed results of the obtained signals are reported in Table VII The measured 119907119907119894119894 in

each temperature and the assumed variation are same as the artificial damage case

11 115 12 125 13-5

0

5

Time (ms)

Volta

ge (m

V)

0 oC20 oC50 oC

(a) Pitch-catch signal

11 115 12 125 13 135

-1

0

1

2

3

Time(ms)

Vol

tage

(mV

)

0 oC20 oC50 oC

(b) Pulse-echo signal

Fig 17 Signals obtained from impact-induced damage specimen (Impact 3) under tempera-ture variation

- 29 -

TABLE VII Temperature effect test for the impact induced damage specimen (Impact 3 case)

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

0 degC

Method 1 135 75~90 72~310 267~1148

Method 2 135 705~779 264~381 978~1126

Method 3 135 763 289 1071

20 degC

Method 1 148 50~100 06~380 21~1134

Method 2 148 758~838 249~368 843~1125

Method 3 148 798 308 1044

50 degC

Method 1 153 75~100 53~415 174~1227

Method 2 153 804~889 214~336 702~1103

Method 3 153 815 320 1050

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1247 ms at 0 degC 1213 ms at 20 degC and 1168 ms at 50 degC

- 30 -

CHAPTER 6 CONCLUDING REMARKS

61 Executive summary

In this study the reference-free damage detection localization and quantification tech-

nique in a multi layered composite plate is proposed Additional reflected A0 Lamb wave

mode from the damage due to multiple reflections in damage area is used as a feature for

characterizing the damage First damage is characterized in terms of a damage index (m-

value) defined as the ratio of the physical length of the damage to the wavelength of the A0

Lamb wave mode Then the range of possible damage location length and severity are esti-

mated with constraints This procedure can be accomplished without prior baseline data To

realize the proposed technique the matching pursuit technique is modified for reduction of

the error by considering the dispersive characteristic of Lamb wave The matching pursuit

method estimates the arrival time of each wave packets accurately even two wave packets are

overlapped The numerical simulations and experimental tests including Teflon tape inserted

and impact-induced damage are investigated for the feasibility of the proposed technique

Moreover the effectiveness of the proposed technique under changing temperature is exper-

imentally verified

- 31 -

62 Suggestion for further study

The limitations of this study and the possible improvement and further studies are listed as

follow

1 Measurement of impact-induced damage length using conventional NDT technique

Additional experimental verification for impact-induced damage characterization is

needed by measuring the real length of damage using conventional NDT technique

such as ultrasonic scan or infrared camera

2 Limitation of both side collocated PZT installation In this study excitation PZTs are

installed both side and collocated for decomposing A0 Lamb wave mode However it

can produce decomposition error due to imperfection of PZT location and their installa-

tion Moreover in many cases assess to both surface of structures is not allowed such

as aircraft applications To improve this limitation a reference-free damage characteri-

zation using single side PZTs will be considered

3 Complex structure application In this study damage is characterized in simple plate-

like structure When the monitored structure becomes complex the response signals al-

so becomes complex due to its geometry and boundary Thus the investigations for

complex structure application such as stiffener spar and curvature are needed

- 32 -

REFERENCES

[1] Cawley P and Alleyne D (1996) ldquoThe use of Lamb waves for the long range inspection of large structuresrdquo Ultrasonics 34 pp287-290

[2] Ghosh T Kundu T and Karpur P (1998) ldquoEfficient use of Lamb waves for detecting defects in large platesrdquo Ultrasonics 36 pp791-801

[3] Guirguitui V (2003) ldquoLamb wave generation with piezoelectric wafer active sensors for structural health monitoringrdquo SPIE Conference on Smart Structures and Materials San Diego CA 5391 pp111-422

[4] Ragvahan A and Cesnik CE S (2004) ldquoModeling of piezoelectric-based Lamb wave generation and sensing for structural health monitoringrdquo SPIE Conference on Smart Struc-tures and Materials San Diego CA 5391 pp419-430

[5] Rose JL Pilarski A and Dirit JJ (1993) ldquoAn approach to guided wave mode selection for inspection of laminated platerdquo Journal of Reinforced Plastics and Composites 12(5) pp536-544

[6] Guo N and Cawley P (1993) ldquoThe interaction of Lamb waves with delaminations in composite laminatesrdquo Journal of Acoustical Society of America 94(4) pp 2240-2246

[7] Valdez SHD and Soutis C (2002) ldquoReal-time nondestructive evaluation of fiber com-posite laminates using low-frequency Lamb wavesrdquo Journal of Acoustical Society of Amer-ica 111(5) pp 2026-2033

[8] Yang M and Qiao P (2005) ldquoModeling and experimental detection of damage in virous materials using the pulse-echo method and piezoelectric sensoractuatorsrdquo Smart Materials and Structures 14 pp1083-1100

[9] Lestari W and Qiao P (2005) ldquoApplication of wave propagation analysis for damage identification in composite laminated beamsrdquo Journal of Composite Materials 39(22) pp1967-1984

[10] Kessler S S Spearing S M and Soutis C (2002) ldquoDamage detection in composite ma-terials using Lamb wave methodsrdquo Smart Materials and Structures 11 pp269-278

[11] Sohn H Park G Wail J R Limback N P and Farrar C R (2004) ldquoWavelet-based active sensing for delamination detection in composite structuresrdquo Smart Materials and Structures 13 pp 153-160

[12] Petculescu G Krishnaswamy S and Achenback J D (2008) ldquoGroup delay measure-ments using modally selective Lamb wave transducers for detection and sizing of delamina-tions in compositesrdquo Smart Materials and Structures 17

[13] Sohn H Park H W Law K H and Farrar C R (2007) ldquoDamage detection in compo-

- 33 -

site plates by using an enhanced time reversal methodrdquo Journal of Aerospace Engineering (ASCE) pp 141-151

[14] Sohn H (2007) ldquoEffects of environmental and operational variability on structural health monitoringrdquo Philosophical Transactions of the Royal Society Ardquo 365 pp539-560

[15] Rytter A (1993) ldquoVibration based inspection of civil engineering structuresrdquo PhD The-sis Department of Building Technology and Structural Engineering University of Aalborg Aalborg Denmark

[16] Hayashi T and Kawashima K (2002) ldquoMultiple reflections of Lamb waves at a delami-nationrdquo Ultrasonics 40 pp193-197

[17] Wang C H and Rose L R F (2003) ldquoWave reflection and transmission in beams con-taining delamination and inhomogeneityrdquo Journal of Sound and Vibration 264 pp 851-872

[18] Ramadas C Balasubramaniam K Joshi M and Krishnamurthy C V (2009) ldquoInterac-tion of the primary anti-symmetric Lamb mode (Ao) with symmetric delaminations numer-ical and experimental studiesrdquo Smart Materials and Structures 18

[19] Ramadas C Balasubramaniam K Joshi M and Krishnamurthy C V (2010) ldquoInterac-tion of guided Lamb waves with an asymmetrically located delamination in a laminated composite platerdquo Smart Materials and Structures 19

[20] Wilcox P Lowe M and Cawely P (2001) ldquoThe effect of dispersion on long-range in-spection using ultrasonic guided wavesrdquo NDTampE International 34 pp1-9

[21] Mallat S G and Zhang Z (1993) ldquoMatching pursuits with time-frequency dictionariesrdquo IEEE Tansactions on Signal Processing 41(12) pp3397-3415

[22] Zhang G Zhang S and Wang Y (2000) ldquoApplication of adaptive time-frequency de-composition in ultrasonic NDE of highly scattering materialsrdquo Ultrasonics 38 pp 961-964

[23] Hong J C Sun K H and Kim Y Y (2005) ldquoThe matching pursuit approach based on the modulated Gaussian pulse for efficient guided-wave damage inspectionrdquo Smart Mate-rials and Structures 14 pp548-560

[24] Herakovich C T (1997) Mechanics of Fibrous Composites Wiley 480 pages

[25] Reddy J N (1985) ldquoA review of the literature on finite-element modeling of laminated composite platesrdquo The Shock and Vibration Digest 17(4) pp 3-8

[26] Reddy J N (2003) Mechanics of Laminated Composite Plates and Shells Theory and Analysis CRC Press 856 pages

[27] Palacz M Krawczuk M and Ostachowicz W (2005) ldquoThe spectral finite element mod-el for analysis of flexure-shear coupled wave propagation Part 1 laminated multilayer

- 34 -

composite beamrdquo Composite Structures 68 pp37-44

[28] Zhang Y C and Harding J (1990) ldquoA numerical micromechanics analysis of the me-chanical properties of a plain weave compositerdquo Computer amp Structures 36(5) pp 839-844

[29] Nemat-Nasser S and Hori M (1999) Micromechanics Overall Properties of Heteroge-neous Materials Elsevier 810 pages

[30] Scida D Aboura Z Benzeggagh M L and Bocherens E (1999) ldquoA micromechanics model for 3D elasticity and failure of woven-fibre composite materialrdquo Composites Science and Technology 59 pp 505-517

[31] Thomas Y and Wu X H (1997) ldquoA multiscale finite element method for elliptic prob-lems in composite materials and porous mediardquo Journal of Computational Physics 134 pp 169-189

[32] Lee K H Moorthy S and Ghosh S (1999) ldquoMultiple scale computational model for damage in composite materialsrdquo Computer Methods in Applied Mechanics and Engineer-ing 172 pp 175-201

[33] Namilae S and Chandra N (2005) ldquoMultiscale model to study the effect of interfaces in carbon nanotube-based compositesrdquo Journal of Engineering Materials and Technology (ASME) 127 pp222-232

[34] Adams D E (2007) Health monitoring of Structural Materials and Components Wiley 460 pages

[35] Gopalakrishnan S (2009) ldquoModeling Aspects in Finite Elementrdquo Encyclopedia of Struc-tural Health Monitoring 2 pp791-809

[36] Palacz M Krawczuk M and Ostachowicz W (2005) ldquoThe spectral finite element mod-el for analysis of flexure-shear coupled wave propagation Part 2 Deminated multilayer composite beamrdquo Composite Structures 68 pp45-51

[37] Baik J M and Thopson R B (1984) ldquoUltrasonic scattering from imperfect interfaces A quasi-static modelrdquo Journal of Nondestructive Evaluation 4(3) pp177-196

[38] Luo H and Hanagud S (2000) ldquoDynamics of delamincated beamsrdquo International Jour-nal of Solids and Structures 37 pp1501-1519

[39] Kim S B and Sohn H (2007) ldquoInstantaneous reference-free crack detection based on polarization characteristics of piezoelectric materialsrdquo Smart Materials and Structures 16 pp2375-2387

[40] An Y K and Sohn H (2010) ldquoInstantaneous crack detection under varying temperature and static loading conditionsrdquo The Special Issue of Structural Control and Health Monitor-ing 17(7) pp 730-741

- 35 -

[41] Lee S J and Sohn H (2006) ldquoActive self-sensing scheme development for structural health monitoringrdquo Smart Materials and Structures 15(6) pp 1734-1746

- 36 -

요 약 문

복합재료 손상의 무기저 진단 및 정량화

복합재료는 재료의 경량성 내구성 고강도의 특성으로 인하여 항공기에서부터

교량에까지 널리 사용되고 있다 하지만 외부의 충격으로 인한 복합재료 내부의

손상은 재료의 강도를 감소시켜 구조물의 안정성을 저하시키는 요인이 된다 하지만

복합재료의 손상은 재료 내부에서 발생하여 육안으로는 식별하기도 힘들며 표면에서

보이는 것만으로 내부의 손상 정도를 판단하기는 불가능하다는 단점을 가지고 있다

본 연구에서는 손상 내부에서 반사로 인해 발생하는 파를 이용하여 기저자료를

사용하지 않고 특정 시점에서 얻은 자료만으로 복합재료의 손상을 진단하고 정량화

할 수 있는 기법을 제안한다 제안된 기법의 적용성을 검증하기 위하여 유한요소법을

이용한 수치해석을 수행하였고 Teflon tape 이 삽입된 시편과 실제 충격하중으로

인한 손상을 입은 시편을 이용한 실험을 진행하였다 수치해석 및 실험 결과 제안된

기법이 복합재료의 손상을 성공적으로 진단 및 정량화할 수 있음을 확인하였다 또한

다양한 외부 온도 환경에서 제안된 기법의 환경영향 실험을 수행하였고 다양한

환경에서 손상의 진단 및 정량화가 성공적으로 이루어짐을 검증하였다

핵심어 무기저 손상진단 및 정량화 압전센서 복합재료 유한요소법 Matching pursuit

method

- 37 -

감 사 의 글

기대와 두려움을 안고 시작한 석사과정이 어느덧 2 년 이라는 세월 속에서 추억으로 자리잡아

가고 있습니다 학부생 기간을 포함한 지난 3 년 동안 SHM 이라는 새로운 분야를 접하고 연구를

수행 할 수 있도록 도와주신 고마운 분들이 있어 그 감사를 표하고자 합니다

우선 저의 많은 부족한 점을 일깨워 주시고 연구에 대한 뜻을 더욱 확고히 세울 수 있게 해

주신 지도교수님이신 손 훈 교수님께 깊은 감사의 뜻을 전합니다 짧은 기간이었지만 교수님의

도움으로 많은 것을 배우고 느끼고 깨달을 수 있었습니다 또한 본 연구를 수행할 수 있도록 지

원해주신 국토해양부 (U-City 석 박사과정 지원사업)에 깊은 감사 드립니다

때로 힘들고 지칠 때 항상 저를 믿고 응원해 주시는 가족이 있었기에 지금의 글을 쓸 수 있었

습니다 또한 지금 연구실에서 함께 생활하고 있는 윤규형 민구형 철민이형 현석이형 진열이형

호민이형 병진이 선혜누나와 성대에 계신 창길이형 미국에 계신 의재형과 Debaditiya 에게도 감

사의 말씀을 전합니다 그리고 모두 나열할 수는 없지만 카이스트 생활의 활력소가 되어준 카이

스트 건설 및 환경공학과 선후배들 직원 분들과 SOS 멤버들에게도 감사를 표합니다

앞서 거론한 많은 이들의 도움이 없었다면 저 또한 지금 이 자리에 있을 수 없었을 것입니다

많이 부족하기에 더 노력하고 배우고자 했던 초심을 다시 한번 생각하며 더 큰 결실을 맺기 위해

앞으로 나아가겠습니다 도움을 주신 모든 분들께 이 감사의 글을 바칩니다

2011년 6월 카이스트에서 임형진 올림

- 38 -

CURRICULUM VITAE

Hyung Jin Lim

Research Assistant

The Civil and Environmental Engineering Department

Korea Advanced Institute of Science and Technology (KAIST)

291 Daehak-Ro Guseong-Dong Yuseong-Gu

Daejeon Republic of Korea 305-701

Phone (82)+42-350-3665 Fax (82)+42-350-3610

Email limhj87gmailcom

RESEARCH INTERESTS

Damage Detection Localization and Quantification on Composite and Metallic Structures Pattern Recognition Finite Element Modeling (FEM)

EDUCATION

2009 ndash 2011 MS Civil and Environmental Eng Korea Advanced Institute of Science and Technology (KAIST) Korea Rep

2006 ndash 2009 BS Civil and Environmental Eng Korea Advanced Institute of Science and Technology (KAIST) Korea Rep

JOURNAL PUBLICATIONS The corresponding author is underlined 1 Hyung Jin Lim Hoon Sohn ldquoReference-free damage detection localization and quantification

using in compositerdquo in preparation for Journal of the Acoustical Society of America 2011

2 Chul Min Yeum Hoon Sohn Jeong Bum Ihn Hyung Jin Lim ldquoInstantaneous delamination detection in a composite using a dual piezoelectric transducer networkrdquo Submitted to Compo-sites Science and Technology 2011

3 Yun-Kyu An Hyung Jin Lim Min Koo Kim Jin Yeol Yang Hoon Sohn Chang Guen Lee ldquoInvestigation of applicability of local reference-free damage detection techniques to in-situ

- 39 -

bridgesrdquo Submitted to ASCE Journal of Structural Engineering 2011

4 Hyung Jin Lim Min Koo Kim Hoon Sohn Chan Yik Park ldquoImpedance based damage detec-tion under varying temperature and loading conditionsrdquo Submitted to NDTampE International 2011

5 Hyung Jin Lim Hoon Sohn ldquoQuantitative estimation of transmitted and reflected lamb waves at discontinuityrdquo Journal of the Korean Society for Nondestructive Testing Vol30 No 4 pp1-8 2010

CONFERENCE PROCEEDINGS

The corresponding author is underlined 1 Hyung Jin Lim and Hoon Sohn ldquoReference-free damage detection localization and quantifi-

cation in compositesrdquo ASCE Engineering Mechanics Institute Conference Boston June 2-4 2011

2 Hyung Jin Lim and Hoon Sohn ldquoReference-free delamination quantification within multi-layer composite materialsrdquo KSNT Spring Conference May 26-27 Gyeong-Ju Korea 2011

3 Min Koo Kim Hyun Jin Lim Hoon Sohn Chan Yik Park ldquoImpedance-based bolt loosening detection under varying temperature and loadingrdquo Asian Pacific Workshop on Structural Health Monitoring Tokyo Japan November 30-December 2 2010

4 Hyung Jin Lim Chul Min Yeum and Hoon Sohn ldquoModeling of Impact-induced Delamination in a Multilayer Composite Platerdquo The 23rd KKCNN Symposium on Civil Engineering Taipei (Taiwan) November 13-15 2010

5 Yun-Kyu An Hyung Jin Lim and Hoon Sohn ldquoTemperature independent Delamination Detec-tion using data normalizationrdquo The 3rd International Conference on Multi-Functional Materi-als and Structures Chonbuk National Univ Jeonju Korea September 14-18 2010

6 Hoon Sohn Yun-Kyu An Chang Gil Lee Min Koo Kim Hyung Jin Lim ldquoApplication of online NDT techniques to decommissioned bridge testingrdquo the 5th European Workshop on Structural Health Monitoring Sorrento Italy June 29-July 02 2010

7 Min Koo Kim Hyung Jin Lim Hoon Sohn Chan Yik Park ldquoImpedance based damage diag-nosis of a complex composite aircraft wing under changing loading and temperature condi-tionsrdquo the 5th European Workshop on Structural Health Monitoring Sorrento Italy June 29-July 02 2010

8 Yun-Kyu An Min Koo Kim Hyung Jin Lim and Hoon Sohn ldquoValidations of reference-free crack detection techniques through a decommissioned bridge testrdquo 2010 COSEIK Annual Con-ference Jeju Korea April 8-9 2010

- 40 -

PARTICIPANT PROJECTS

1 Development of Smart Steel Members for Online Steel Structure Monitoring POSCO Corporation (Funded 41000000 KRW (41000 USD) for 010511 to 123111)

2 Development of Enabling Technologies for Structural Health Monitoring Korean Agency for Defense Development (Funded 240000000 KRW (240000 USD) out of total 3200000000 KRW (3200000 USD) for 060108-053111)

3 Development of On-board SHM Technologies for Composite Air Vehicles The Boeing Company (Funded 400000 USD (400000000 KRW) for 080108 to 073111)

EDUCATIONAL EXPERIENCES

Fall 2009

Teaching Assistant CE207 ldquoElementary Structural Engineering and Laboratoryrdquo KAIST 3 unit undergraduate course

EXPERIMENTAL EXPERIENTCES

2010 Monitoring of Real Aircraft Composite Wing Segment Korea Agency for Defense De-velopment (ADD) Daejeon Korea Real time monitoring of composite air vehicles fund-ed by Korea Agency for Defense Development (ADD)

2009-2011

Field Monitoring of Real Bridge Structures Gimpo (Ramp G bridge) amp Icheon (Sam-seung bridge) Korea US-Korea collaborative research for bridge monitoring testbeds

HONORS amp AWARDS

Scholarships for outstanding students granted KAIST (2007-2009) The three prominent students in the civil and environmental engineering department of KAIST were selected as recipients This scholarship provided 2000$ per year for three years

- 41 -

Page 16: 복합재료 손상의 무기저 진단 및 정량화 - KAISTssslab.kaist.ac.kr/article/pdf/Thesis_2.pdf · 1 Declaration of Ethical Conduct in Research: I, as a graduate student

CHAPTER 2 INTERACTION OF LAMB WAVE WITH DAMAGE

21 A0 Lamb wave mode delay and attenuation in the time domain

Time delay and attenuation of the A0 mode Lamb wave after passing through the damage

area due to effective thickness reduction were reported in previous researches [812] Moreo-

ver the attenuation of the A0 Lamb wave mode is occurred by degrading of shear modulus of

in the damage area However the time delay and amplitude variation of the A0 Lamb wave

mode also occurred even when the structure is health condition by the material property

changing due to temperature variation of the structure (Fig 1) Thus conventional damage

characterization methods using the time delay and attenuation of A0 Lamb wave mode as a

feature can cause false-alarm in real application

11 115 12 125 13-10

-5

0

5

10

Time (ms)

Vol

tage

(mV

)

IntactDamage

Reflected wave from Damage

(a)

11 115 12 125 13-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

0 oC20 oC50 oC

(b)

Fig 1 Lamb wave propagation characteristics (a) A0 Lamb wave mode time delay and atten-uation after passing through a damage (b) The time delay and amplitude variation of A0 Lamb wave mode under temperature variation in pristine condition of structure Because of this conventional damage detection techniques in composites have limitations for real struc-ture application

- 5 -

22 Multiple reflections in the damage area

Standing wave patterns due to multiple reflections of A0 Lamb wave mode in the damage

area were observed through numerical simulations and ultrasonic wave-field imaging using a

scanning laser vibrometer [16] The formation of standing waves in damage is illustrated

through Fig 2 When the incident waves enter the damage area a significant portion of the

waves are reflected back from the exit Then the reflected waves travel through the damage

and undergo reflection again at the original entrance As a consequence of multiple reflec-

tions at the entrance and the exit a considerable amount of ultrasonic energy is trapped inside

the damage area [17-19]

Multiple Reflections

Incident Wave Transmitted Wave

Damage

Fig 2 Multiple reflections in the damage area The multiple reflections of A0 Lamb waves in the damage area are utilize as a feature for characterization of the damage in this study

- 6 -

CHAPTER 3 THEORETICAL DEVELOPMENT

31 Overview

Damage

l = l1 + d + l2

d = vdtd l1 = vit1 l2 = vit2

t = t1 + td + t2

Excitation and Sensing PZT Sensing PZT

l

Transmitted Wave

Reflected Waves

Incident Wave

Fig 3 Parameters and notations of the proposed damage characterization technique Damage is detected by identifying the additional reflected wave from the damage due to multiple re-flections (dashed lines) Localization and quantification of the damage is carried out by measuring the arrival time of transmitted and reflected waves from the damage

In this section the reference-free damage characterization algorithm is proposed Damage

diagnosis and quantification are achieved by identifying the additional reflected waves due to

multiple reflections in the damage area (dashed lines in Fig 3) Moreover damage is local-

ized using additional information from pulse-echo method implement For the localization

and the quantification of damage the variables such as the path length and A0 Lamb wave

mode velocity in damage area are defined (Fig 3) For example 119897119897 is expressed as

119897119897 = 1198971198971 + 119889119889 + 1198971198972 (1)

where 1198971198971 119889119889 and 1198971198972 are the distance from the excitation PZTs to the entrance of damage the

damage length and the distance from the exit of damage to the sensing PZT respectively

- 7 -

Moreover 119907119907119894119894 and 119907119907119889119889 are also defined as the A0 Lamb wave mode velocity in pristine area

and damage area respectively Note that the whole procedure of the damage characterization

is accomplished based solely on the signals obtained from the current state of the structure

32 Damage index (m-value)

The proposed reference-free damage characterization technique utilizes the waves reflect-

ed additionally from the damage area as a feature The arrival time difference of the transmit-

ted and reflected A0 Lamb wave modes reflected from the exit and the entrance of the delam-

ination area one times each (∆119905119905) contains the information of the physical length of the dam-

age (119889119889) and A0 Lamb wave mode velocity in the damage area (119907119907119889119889) The wave velocity in the

damage area which is related to the shear modulus (G) represents the severity of the damage

∆119905119905 = 2119905119905119889119889 =2119889119889119907119907119889119889

(2)

However based on the measured data available the 119889119889 and 119907119907119889119889 (or G) cannot be uniquely

determined Fig 4 shows two FEM models which have different physical length and wave

velocity (G) of the damage Detail explanation of delamination modeling technique will be

covered in section 4 The similar signal in the time domain and ∆119905119905 are obtained from the

two different damage cases

- 8 -

Damage 1

Damage 2

20

Unit mm15

90 G value reduction

943 G value reduction

m-value = 248

m-value = 248

015 02 025 03 035

-2

-1

0

1

2

Time (ms)

Ampl

itude

Damage 1Damage 2

Fig 4 Two different FEM models which have different 119889119889 and 119907119907119889119889 (or G) This represents that the damage cannot be characterized uniquely based on the measured data available

Thus in this study the damage index (m-value) which can be uniquely defined based on

the available data is introduced The m-value is defined as the ratio of the physical length of

the damage and wavelength of the A0 Lamb wave mode in the damage area (120582120582119889119889)

119898119898 =119889119889120582120582119889119889

=119889119889119889119889119907119907119889119889

(3)

In many cases the transmitted and reflected waves are overlapped and it is hard to decom-

pose and calculate the arrival times of them Matching pursuit method is an alternative way

to accurately decompose of the overlapped waves and estimate the parameters of each wave

such as arrival time (Fig 5) Detail explanation of the matching pursuit method is covered in

section 36

015 02 025 03-2

-1

0

1

2

Time (ms)

Am

plitu

de

OriginalApproximated

(a)

015 02 025 03-2

-1

0

1

2

Time (ms)

Ampl

itude

Wave 1Wave 2

(b)

Fig 5 (a) Example of the overlapped signal with comparing approximated signal using the matching pursuit method (b) Decomposed signal using the matching pursuit method The matching pursuit method provides a conventional way to decompose and identify each wave

- 9 -

33 Damage detection algorithm (Level 1)

The damage diagnosis is accomplished by extracting the additional reflected A0 Lamb

wave modes from damage due to multiple reflections If A0 Lamb wave modes are decom-

posed perfectly it is trivial that one transmitted A0 Lamb wave mode is always captured in

the response signal If damage exists other waves in the response signal behind the transmit-

ted A0 Lamb wave mode are presented Since this approach does not rely on previously col-

lected data the diagnosis is independent from environmental variation such as temperature

changes

It is reported that the dispersion of Lamb waves cause wave-packets to spread out in space

and time as they propagate through a structure [20] Thus the reflected waves from the dam-

age have longer pulse duration than the transmitted wave due to their longer propagating dis-

tance In this study the matching pursuit method is modified to capture more spread wave

than the transmitted A0 Lamb wave mode Detail explanation of the modified matching pur-

suit method is described in latter section

Let 1199041199041 and 1199041199042 as the pulse duration (scale) of the transmitted and additional reflected A0

Lamb wave mode respectively

ldquoIf 1199041199042 gt 1199041199041 damage exists

Otherwise there is no damagerdquo

- 10 -

34 Damage quantification algorithm (Level 3)

As mentioned in the previous section damage is quantized by estimating 119889119889 and 119907119907119889119889 using

pitch-catch method The procedure to estimate 119889119889 and 119907119907119889119889 is

(i) Measure 119897119897 119905119905 and 119905119905119889119889

(ii) Assume the range of 119889119889 (0 lt 119889119889 lt 119897119897)

(iii) Calculate 119907119907119894119894 and 119907119907119889119889 with a constraints (0 lt 119907119907119889119889 lt 119907119907119894119894) (Eq 4 and 5)

(iv) Estimate m-value and find a feasible range of 119889119889 and 119907119907119889119889

119907119907119894119894 =119897119897 minus 119889119889119905119905 minus 119905119905119889119889

(4)

119907119907119889119889 =119889119889119905119905119889119889

(5)

35 Damage localization algorithm (Level 2amp3)

For damage localization 1198971198971 is estimated by measuring the arrival time from the excitation

PZT to the entrance of delamination (1199051199051) with additional pulse-echo method implementation

The procedure is similar with the previous damage quantification algorithm

(i) Measure 119897119897 119905119905 119905119905119889119889 and 1199051199051

(ii) Assume the range of 1198971198971 (0 lt 1198971198971 lt 119897119897)

(ii) Compute possible combinations of 1198971198972 119889119889 and 119907119907119889119889 with constraints

(0 lt 119907119907119889119889 lt 119907119907119894119894 119889119889 gt 0) (Eq 5 6 and 7)

- 11 -

1198971198972 = 1198971198971119905119905 minus 1199051199051 minus 119905119905119889119889

1199051199051 (6)

119889119889 = 119897119897 minus 1198971198971 minus 1198971198972 (7)

Since a PZT for pulse-echo method acts as both an actuator and a sensor the obtained sig-

nal contains both the excitation and the response signal Thus the reflected wave from the

entrance of damage can be buried into the tail of excitation signal and cause error on arrival

time of reflected wave from damage Even in this case damage can be localized and quan-

tized using the constraints When first wave packet reflected from the entrance of damage is

buried into the excitation signal the next wave packet reflected from the exit of damage can

be utilized alternatively If the algorithm regards the second reflected wave as the first re-

flected wave there are no combinations of estimated variable meeting the constraints Then

the algorithm repeats the process by regarding the wave as the second reflected wave the

constraints can select proper combination of the estimated variables

The estimated values have range bounded by the constraints When additional information

of 119907119907119894119894 is utilized the damage can be characterized more accurately In this study three meth-

ods are proposed and verified as information level of 119907119907119894119894

Method 1 Solely based on the current data measurement (Reference-free)

Method 2 Limit the possible range of 119907119907119894119894 (Prior knowledge of the system)

Method 3 Measure 119907119907119894119894 (Additional data need to be collected)

Method 1 described in this section can be utilized without any information of 119907119907119894119894 Meth-

od 2 produces more accurate estimation results using prior knowledge of the system such as

- 12 -

119907119907119894119894 variation under temperature changes Method 3 gives unique estimated value using meas-

ured 119907119907119894119894 from other path in the same time In section 4 and 5 damage is characterized using

these three methods and the results are compared each another

36 Matching pursuit method considering the dispersion effect of Lamb

wave

Matching pursuit method has been developed for signal compression approximation and

characterization of obtained signal and function [21-23] The key concept of the matching

pursuit method is that any signal (function) can be decomposed into a linear expansion of

signals selected from a pool of basis signals (functions) called lsquodictionaryrsquo

Let 119891119891 be a function to be analyzed then 119891119891 can be approximated a set of functions select-

ed from the dictionary 119863119863 in order to best match This is done by successive approximations

with orthogonal projections Let1198921198921205741205740 isin 119863119863 the function 119891119891 can be decomposed into

119891119891 = lang1198911198911198921198921205741205740rang1198921198921205741205740 + 119877119877119891119891 (8)

where 119877119877119891119891 is the residual function after approximating 119891119891 using 1198921198921205741205740 1198921198921205741205740 is orthogonal to

119877119877119891119891 hence

1198911198912 = lang1198911198911198921198921205741205740rang2

+ 1198771198771198911198912 (9)

To minimize 119877119877119891119891 1198921198921205741205740 is chosen to maximize lang1198911198911198921198921205741205740rang The matching pursuit method is

an iterative algorithm that decomposes the residue 119877119877119891119891 by projecting it on a function of 119863119863

that matches 119877119877119891119891 almost at best as it was done for 119891119891 The residue 119877119877119899119899119891119891 is decomposed into

119877119877119899119899119891119891 = lang119877119877119899119899119891119891119892119892120574120574119899119899rang119892119892120574120574119899119899 + 119877119877119899119899+1119891119891 (10)

- 13 -

the decomposition is carried up to the order 119898119898 119891119891 can be expressed as

119891119891 = lang119877119877119899119899119891119891119892119892120574120574119899119899rang119892119892120574120574119899119899

119898119898minus1

119899119899=0

+ 119877119877119898119898119891119891 (11)

The original function 119891119891 is decomposed into a sum of dictionary elements that are chosen to

best match its residues

In this study the matching pursuit method is utilized to accurately decompose of the over-

lapped Gaussian pulses and estimate the parameters of the pulse which used for damage

characterization The 119894119894119905119905ℎ Gaussian pulse in the dictionary for approximating the obtained

signal is expressed as

119892119892120574120574119894119894 = 119891119891(119860119860119894119894 119906119906119894119894 119904119904119894119894 119889119889119894119894120601120601119894119894) (12)

where 119860119860119894119894119906119906119894119894119904119904119894119894119889119889119894119894 and 120601120601119894119894 represent amplitude arrival time scale (pulse duration) excita-

tion(central) frequency and phase of the Gaussian pulse respectively The parameters of the

pulse are estimated using a nonlinear regression based optimization method (Gauss-Newton

method) based on an initial guess of the parameters Detail procedure for estimating the pa-

rameters of the pulse is expressed in Hong et al [23]

Among the estimated parameters of the pulses pulse duration (scale) represents the disper-

sion effect If a wave-packet is spread out the estimated pulse duration is increased Using

this feature the matching pursuit method is modified to capture waves which have longer

pulse duration than transmitted A0 Lamb wave mode If the estimate pulse duration of a wave

is less than the transmitted wave the modified matching pursuit method regards it as error

- 14 -

CHAPTER 4 NUMERICAL SIMULATION

41 Composite modeling method

Composite materials can be modeled from nano to macro level depends on its engineering

purpose In general four methods are utilized for modeling of composite materials First

composite is modeled as a homogenous material expressed in terms of effective material

properties [24] The effective material properties are estimated using the modulus and the

volume fraction of each constituent theoretically (Rule of mixture) and measured experimen-

tally Second a model is established that each laminate which have own material properties

are totally bonded together In this model the stacking sequence and direction of each lami-

nates are considered additionally [25-27] Third method for composite modeling in micro

level called micromechanics [28-30] This model considers the interaction between fibers and

matrix in laminate Last multiscale modeling which the scope of modeling is extended to

molecular or quantum level of constituents [31-33] This model starts from molecular me-

chanics or molecular dynamics then the scale of model leads to the nano micro and macro

level by incorporating small scale models As the scale of models is smaller it provides more

logical way than other large scale of models while the computational cost is more expensive

The scope of this study is focused on the dynamic behavior of composite structure the beam

or plate model was utilized for the numerical simulation

- 15 -

42 Damage modeling method

Generally three damage models have been developed and utilized in the real applications

First the stiffness reduction model called as the equivalent element model was introduced

[17 34-35] This model is simply realized by replacing the region around damage area by a

region with adjusted material properties (Youngrsquos modulus shear modulus density etc)

globally Next the duplicated node model is done by disconnecting the element connectivity

assuming there is no interaction among the delaminated surface This model can be utilized

when the location and the extension of damage is important issue [34-36] Last the contact

mechanics model using the principles of contact mechanics the two interfacial forces (con-

tact resistance and sliding resistance) in the delaminated surface [37-38] In this model the

delaminated surfaces are modeled as nonlinear functions of the interlaminar normal strain

and the interlaminar shear strain This model is computationally expensive while provides

more logical way than other models There is however no universally accepted model for

general impact-induced damage in composites In the case of impact-induced damage it is

impossible to predict or estimate the location and number of delaminated layers Thus from

the uncertainty of impact-induced damage the equivalent element model can be considered

as the effective damage model

- 16 -

43 Numerical simulation description

TABLE I Material property of simulation model and test specimen

Tensile Strength 657 MPa Tensile Modulus 59 GPa Shear Strength 804 MPa Shear Modulus 24 Gpa Specific Gravity 16

PZT A

PZT C

PZT BDamage

3

250 125 125

500 Unit mm

Fig 6 Description of the numerical simulation model

The feasibility of the proposed damage characterization technique is verified by 2-D FEM

simulation using MSC PATRAN and NASTRAN Fig 6 presents the dimension PZTs and

damage location of the simulation model In the model 05mm x 05mm square shell element

was used for modeling of the A0 mode whose flexural wave characteristic The element size

was determined by considering the wavelength of the propagation Lamb wave modes The

material properties of the simulation model and damage cases are presented in Table I and

Table II respectively

PZT A and C act as both actuator and sensor for collecting pulse-echo signal They were

installed on the plate which two of them are both-side collocated for the A0 Lamb wave

- 17 -

mode decomposition [39-40] PZT B was only used as a sensor for collecting pitch-catch sig-

nal For simulating the excitation and sensing of PZTs the pin-force model was adopted that

an input force apples to the end nodes of excitation PZT and averaging element strains corre-

sponding sensing area A Gaussian pulse (tone-burst) 50 kHz central frequency was excited

as an input signal

TABLE II Damage cases for numerical simulation

119949119949120783120783 (mm) 119941119941 (mm) G value reduction ()

Case 1 650 200 90 darr

Case 2 675 150 90 darr

Case 3 675 150 93 darr

Case 4 900 150 90 darr

- 18 -

44 Numerical simulation results

015 02 025 03 035

-2

0

2

Time (ms)

Am

plitu

de

IntactDamage

Reflected wave from damage

(a) Pitch-catch signal

015 02 025 03 035

-2

0

2

Time (ms)

Am

plitu

de

IntactDamage

Reflected waves from damage

(b) Pulse-echo signal Fig 7 Comparison of A0 Lamb wave mode signals from numerical simulation between intact and damage cases Time delay attenuation and multiple reflections due to damage simulated successfully using the damage modeling method described in the previous section

Fig 7 shows that comparison of A0 Lamb wave mode signals between intact and one of

damage cases in the simulation The features of A0 Lamb wave mode passing through dam-

age area described in section 2 are successfully simulated Thus the equivalent element

method is proper for modeling damage in composites In the damage cases the additional

waves reflected from damage is observed in both pitch-catch and pulse echo signal

Table III presents the result of the proposed damage characterization algorithm using the

numerical simulation cases shown in Table II In lsquoMethod 2rsquo 119907119907119894119894 variation is assumed to plusmn

5 of the normal value (1190~1315ms) The normal value of 119907119907119894119894 for lsquoMethod 3rsquo is meas-

ured as 1253 ms In each case the damage is characterized using the proposed technique

Note that as the information of 119907119907119894119894 is specified the estimated value is closer to the exact

value

- 19 -

TABLE III Damage characterization for the numerical simulation models

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

Case 1

Exact 650 200

Method 1 248 50~70 79~414 160~834

Method 2 248 603~667 135~241 271~486

Method 3 248 635 188 379

Case 2

Exact 675 150

Method 1 193 60~75 01~251 37~651

Method 2 193 618~682 114~222 295~575

Method 3 193 65 168 435

Case 3

Exact 675 150

Method 1 233 60~75 66~302 140~649

Method 2 233 668~738 85~196 182~419

Method 3 233 703 140 301

Case 4

Exact 900 150

Method 1 204 75~100 48~349 118~856

Method 2 204 860~951 107~216 264~530

Method 3 204 906 162 397

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1253 ms 1190~1315 ms Measured 119907119907119894119894 1253 ms

- 20 -

CHAPTER 5 EXPERIMENTAL VERIFICATION

51 Overview

In this section experimental study is investigated so that the proposed damage characteri-

zation technique Overall experimental setup and test specimen are shown in Fig 8 and 9

The experimental setup consists of test specimens data acquisition (DAQ) system and self-

sensing circuit Self-sensing circuit was installed for pulse-echo measurement using PZTs

[40-41]

PZT A(PZT C) PZT B

500

500

140

75

250

Teflon Inserted Damage (d =25)

Unit mm

Thickness = 3

PZT A(PZT C) PZT B

500

500

150

90

250

Impact Location

Unit mm

Thickness = 3

(a) Teflon inserted damage specimen

(b) Impact-induced damage specimen

Fig 8 Test specimens for the experimental verification In the Teflon inserted damage Teflon tapes with 25mm diameter were inserted at the center of the specimen The impact-induced damage was introduced to the specimen using the impact test equipment

- 21 -

511 Specimen

Two carbon composite plates with a dimension of 500times500times3mm3 were manufactured by

Nexcomes Inc by stacking 12 layers of CF3327 3K prepreg (woven fabrics type) manufac-

tured by Hankuk Fiber Inc The material properties of the specimen are presented in Table I

One of the specimens Teflon tapes with 25mm diameter was inserted in every 3 layers at the

center of the specimen to mimic the damage The other specimen is utilized for the character-

ization of real impact-induced damage

512 Data acquisition (DAQ) system and PZT

The data acquisition (DAQ) system consists of an arbitrary waveform generator (AWG) a

high speed signal digitizer (DIG) and two multiplexers (MUX) Using the 16-bits AWG

tone-burst input signal with a plusmn10 peak-to-peak voltage were generated The response signal

was measured by the 14-bit DIG For the signal-to-noise ratio reduction both pitch-catch and

pulse-echo signals were measured 150 times and averaged in the time domain 3 identical

circular PZTs of 10mm diameter manufactured by APC International Ltd were installed on

the specimen A Gaussian pulse (tone-burst) 50 kHz central frequency was excited as an in-

put signal The excitation frequency was chosen in order to avoid overlapping of the S0 and

A0 Lamb wave modes and generating of the higher modes (S1 A1hellip)

- 22 -

PZT Actuator

PZT Sensor

AWG

DIG

DAQ System Specimen

SensingActuatingMUX

1

23

4

1 Sending signal to MUX

4 Sending signal to DIG

Self-SensingCircuit

18 Unit mm

10

(a)

(b)

Fig 9 Overall experimental setup (a) Data acquisition (DAQ) system and (b) Installed circu-lar PZT transducer on the specimen with its dimension

513 Impact test

The impact-induced damage was introduced to the specimen using the impact test equip-

ment presented in Fig 10 A 5kg mass with 10mm tip was dropped from 015m height onto

the same location of the specimen The signals were obtained at each number of impacts

Since the reflected wave is not detected at the first impact case the data obtained from the

second and third impact were utilized for the analysis

10

Unit mm (a)

(b)

Fig 10 (a) Impact test equipment for introducing impact-induced damage and (b) Tip of the impact tester with 10mm diameter

- 23 -

514 Temperature effect test

In reality structures are subjected to changing environmental conditions such as tempera-

ture that affect measured signal and cause false-alarm Thus the robustness of the proposed

damage characterization technique under temperature variation in investigated (Fig 11) The

specimens were placed inside a temperature chamber and a thermocouple was installed on

the specimen to measure temperature of surface of the specimen The signal was obtained

under three temperature cases (0 20 and 50degC) and the humidity was kept at 30

Temperature Chamber

Test Specimen

Fig 11 Test specimen in temperature chamber

- 24 -

52 Teflon inserted damage

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Reflected wave from damage

(a)

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Transmitted WaveReflected Wave

Detected reflected wave

(b)

Fig 12 A0 Lamb wave mode obtained using pitch-catch method from the Teflon inserted damage specimen (a) The reflected wave from damage is shown (b) The Teflon inserted damage is detected by capturing the reflected wave

Fig 12 shows the obtained pitch-catch signal from the Teflon inserted damage specimen

and damage diagnosis result The Teflon inserted damage is detected successfully by captur-

ing the reflected wave Table IV presents the result of the Teflon inserted damage characteri-

zation 119907119907119894119894 is measured as 1213ms and the variation is assumed to plusmn 5 of the normal val-

ue (1152~1273ms) The results indicate that the proposed damage characterization approach

works well in real application

TABLE IV Damage characterization for the Teflon inserted damage specimen

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

Exact 635 25

Method 1 123 60~75 73~338 298~1220

Method 2 123 619~662 229~310 934~1188

Method 3 123 631 284 1161

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1213 ms

- 25 -

53 Impact-induced damage

11 115 12 125 13-4

-2

0

2

4

Time (ms)

Vol

tage

(mV

)

IntactImpact2Impact3

Reflected wave from damage

(a) Pitch-catch signal

11 115 12 125 13 135

-05

0

05

1

Time (ms)

Volta

ge (m

V)

IntactImpact2Impact3

Boundary reflection

Reflected wave from damage

(b) Pulse-echo signal

Fig 13 The signal obtained from the test specimen subjected to impact The impact was in-troduced at the same location of the specimen and the signal was obtained after each impacts As the number of impact is increased the transmitted A0 Lamb wave mode is delayed and attenuated

The obtained signals from impact-induced damage specimen are presented in Fig 13 It is

observed that as the number of impact is increased the transmitted A0 Lamb wave mode is

more delayed because the wave velocity in the damage area is reduced Fig 14 shows the

false-alarm test result whether the proposed damage detection algorithm works even in pris-

tine condition of the test specimen It is shown that no false alarm is occurred because no ad-

ditional reflected wave is captured

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Transmitted Wave

(a) Transmitted wave

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Reflected Wave

No reflected wave

(b) Reflected wave

Fig 14 False alarm test for the proposed damage detection algorithm in pristine condition of the specimen Reflected wave from damage is not captured false alarm is not occurred

- 26 -

TABLE V Damage characterization for the impact-induced damage specimen

of Impacts m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

2

(1470 J)

Method 1 080 65~90 75~246 470~1157

Method 2 080 801~885 114~245 711~1134

Method 3 080 843 179 1122

3

(2205 J)

Method 1 148 50~100 06~380 21~1134

Method 2 148 758~838 249~368 843~1125

Method 3 148 798 308 1044

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1213 ms

Table V presents the characterization results of the impact-induced damage case Here the

measured 119907119907119894119894 and the assumed variation are same as the artificial damage case It is observed

that as the number of impact is increased 119889119889 is increased while 119907119907119889119889 is reduced Thus it is

validated that the proposed damage characterization approach estimates the increment of the

damage well

54 Temperature effect test

541 Teflon inserted damage

11 115 12 125 13-10

-5

0

5

10

Time (ms)

Vol

tage

(mV

)

0 oC20 oC50 oC

(a) Pitch-catch signal

11 115 12 125 13 135

-1

0

1

2

3

Time(ms)

Volta

ge (m

V)

0 oC20 oC50 oC

(b) Pulse-echo signal

Fig 15 Signals obtained from the Teflon inserted damage specimen under temperature varia-tion Time delay and amplitude variation are occurred

Fig 15 shows the obtained signals from the Teflon inserted damage specimen subjected to

- 27 -

temperature variation Time delay and amplitude variations are occurred there is no addition-

al damage is introduced The analyzed results of the obtained signals are reported in Table VI

The measured 119907119907119894119894 is 1247 ms at 0 degC 1213 ms at 20 degC and 1168 ms at 50 degC respectively

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms (1152~1273 ms)

The estimated results are also changed under temperature variations the fluctuation of esti-

mated range and exact value is small

TABLE VI Temperature effect test for the Teflon inserted damage specimen

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

0 degC

Exact 635 25

Method 1 113 60~70 79~268 352~1291

Method 2 113 560~619 233~344 1034~1172

Method 3 113 606 257 1141

20 degC

Exact 635 25

Method 1 123 60~75 73~338 298~1220

Method 2 123 619~662 229~310 934~1188

Method 3 123 631 284 1161

50 degC

Exact 635 25

Method 1 137 60~75 72~328 263~1228

Method 2 137 622~687 183~299 666~1086

Method 3 137 642 264 960

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1247 ms at 0 degC 1213 ms at 20 degC and 1168 ms at 50 degC

- 28 -

542 Impact-induced damage

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Vol

tage

(mV

)

Transmitted WaveReflected Wave

No reflected wave

(a) 0 degC

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Transmitted WaveReflected Wave

No reflected wave

(b) 50 degC

Fig 16 False alarm test under temperature variations It is shown that proposed damage char-acterization algorithm is robust under temperature variations

The false alarm test in pristine condition of the impact-induced damage specimen under

temperature variation was investigated (Fig 16) It is validated that the proposed damage di-

agnosis algorithm is robust under temperature variations Fig 17 shows the obtained signals

from the impact-induced damage (impacted 3 times) specimen under temperature variation

The analyzed results of the obtained signals are reported in Table VII The measured 119907119907119894119894 in

each temperature and the assumed variation are same as the artificial damage case

11 115 12 125 13-5

0

5

Time (ms)

Volta

ge (m

V)

0 oC20 oC50 oC

(a) Pitch-catch signal

11 115 12 125 13 135

-1

0

1

2

3

Time(ms)

Vol

tage

(mV

)

0 oC20 oC50 oC

(b) Pulse-echo signal

Fig 17 Signals obtained from impact-induced damage specimen (Impact 3) under tempera-ture variation

- 29 -

TABLE VII Temperature effect test for the impact induced damage specimen (Impact 3 case)

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

0 degC

Method 1 135 75~90 72~310 267~1148

Method 2 135 705~779 264~381 978~1126

Method 3 135 763 289 1071

20 degC

Method 1 148 50~100 06~380 21~1134

Method 2 148 758~838 249~368 843~1125

Method 3 148 798 308 1044

50 degC

Method 1 153 75~100 53~415 174~1227

Method 2 153 804~889 214~336 702~1103

Method 3 153 815 320 1050

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1247 ms at 0 degC 1213 ms at 20 degC and 1168 ms at 50 degC

- 30 -

CHAPTER 6 CONCLUDING REMARKS

61 Executive summary

In this study the reference-free damage detection localization and quantification tech-

nique in a multi layered composite plate is proposed Additional reflected A0 Lamb wave

mode from the damage due to multiple reflections in damage area is used as a feature for

characterizing the damage First damage is characterized in terms of a damage index (m-

value) defined as the ratio of the physical length of the damage to the wavelength of the A0

Lamb wave mode Then the range of possible damage location length and severity are esti-

mated with constraints This procedure can be accomplished without prior baseline data To

realize the proposed technique the matching pursuit technique is modified for reduction of

the error by considering the dispersive characteristic of Lamb wave The matching pursuit

method estimates the arrival time of each wave packets accurately even two wave packets are

overlapped The numerical simulations and experimental tests including Teflon tape inserted

and impact-induced damage are investigated for the feasibility of the proposed technique

Moreover the effectiveness of the proposed technique under changing temperature is exper-

imentally verified

- 31 -

62 Suggestion for further study

The limitations of this study and the possible improvement and further studies are listed as

follow

1 Measurement of impact-induced damage length using conventional NDT technique

Additional experimental verification for impact-induced damage characterization is

needed by measuring the real length of damage using conventional NDT technique

such as ultrasonic scan or infrared camera

2 Limitation of both side collocated PZT installation In this study excitation PZTs are

installed both side and collocated for decomposing A0 Lamb wave mode However it

can produce decomposition error due to imperfection of PZT location and their installa-

tion Moreover in many cases assess to both surface of structures is not allowed such

as aircraft applications To improve this limitation a reference-free damage characteri-

zation using single side PZTs will be considered

3 Complex structure application In this study damage is characterized in simple plate-

like structure When the monitored structure becomes complex the response signals al-

so becomes complex due to its geometry and boundary Thus the investigations for

complex structure application such as stiffener spar and curvature are needed

- 32 -

REFERENCES

[1] Cawley P and Alleyne D (1996) ldquoThe use of Lamb waves for the long range inspection of large structuresrdquo Ultrasonics 34 pp287-290

[2] Ghosh T Kundu T and Karpur P (1998) ldquoEfficient use of Lamb waves for detecting defects in large platesrdquo Ultrasonics 36 pp791-801

[3] Guirguitui V (2003) ldquoLamb wave generation with piezoelectric wafer active sensors for structural health monitoringrdquo SPIE Conference on Smart Structures and Materials San Diego CA 5391 pp111-422

[4] Ragvahan A and Cesnik CE S (2004) ldquoModeling of piezoelectric-based Lamb wave generation and sensing for structural health monitoringrdquo SPIE Conference on Smart Struc-tures and Materials San Diego CA 5391 pp419-430

[5] Rose JL Pilarski A and Dirit JJ (1993) ldquoAn approach to guided wave mode selection for inspection of laminated platerdquo Journal of Reinforced Plastics and Composites 12(5) pp536-544

[6] Guo N and Cawley P (1993) ldquoThe interaction of Lamb waves with delaminations in composite laminatesrdquo Journal of Acoustical Society of America 94(4) pp 2240-2246

[7] Valdez SHD and Soutis C (2002) ldquoReal-time nondestructive evaluation of fiber com-posite laminates using low-frequency Lamb wavesrdquo Journal of Acoustical Society of Amer-ica 111(5) pp 2026-2033

[8] Yang M and Qiao P (2005) ldquoModeling and experimental detection of damage in virous materials using the pulse-echo method and piezoelectric sensoractuatorsrdquo Smart Materials and Structures 14 pp1083-1100

[9] Lestari W and Qiao P (2005) ldquoApplication of wave propagation analysis for damage identification in composite laminated beamsrdquo Journal of Composite Materials 39(22) pp1967-1984

[10] Kessler S S Spearing S M and Soutis C (2002) ldquoDamage detection in composite ma-terials using Lamb wave methodsrdquo Smart Materials and Structures 11 pp269-278

[11] Sohn H Park G Wail J R Limback N P and Farrar C R (2004) ldquoWavelet-based active sensing for delamination detection in composite structuresrdquo Smart Materials and Structures 13 pp 153-160

[12] Petculescu G Krishnaswamy S and Achenback J D (2008) ldquoGroup delay measure-ments using modally selective Lamb wave transducers for detection and sizing of delamina-tions in compositesrdquo Smart Materials and Structures 17

[13] Sohn H Park H W Law K H and Farrar C R (2007) ldquoDamage detection in compo-

- 33 -

site plates by using an enhanced time reversal methodrdquo Journal of Aerospace Engineering (ASCE) pp 141-151

[14] Sohn H (2007) ldquoEffects of environmental and operational variability on structural health monitoringrdquo Philosophical Transactions of the Royal Society Ardquo 365 pp539-560

[15] Rytter A (1993) ldquoVibration based inspection of civil engineering structuresrdquo PhD The-sis Department of Building Technology and Structural Engineering University of Aalborg Aalborg Denmark

[16] Hayashi T and Kawashima K (2002) ldquoMultiple reflections of Lamb waves at a delami-nationrdquo Ultrasonics 40 pp193-197

[17] Wang C H and Rose L R F (2003) ldquoWave reflection and transmission in beams con-taining delamination and inhomogeneityrdquo Journal of Sound and Vibration 264 pp 851-872

[18] Ramadas C Balasubramaniam K Joshi M and Krishnamurthy C V (2009) ldquoInterac-tion of the primary anti-symmetric Lamb mode (Ao) with symmetric delaminations numer-ical and experimental studiesrdquo Smart Materials and Structures 18

[19] Ramadas C Balasubramaniam K Joshi M and Krishnamurthy C V (2010) ldquoInterac-tion of guided Lamb waves with an asymmetrically located delamination in a laminated composite platerdquo Smart Materials and Structures 19

[20] Wilcox P Lowe M and Cawely P (2001) ldquoThe effect of dispersion on long-range in-spection using ultrasonic guided wavesrdquo NDTampE International 34 pp1-9

[21] Mallat S G and Zhang Z (1993) ldquoMatching pursuits with time-frequency dictionariesrdquo IEEE Tansactions on Signal Processing 41(12) pp3397-3415

[22] Zhang G Zhang S and Wang Y (2000) ldquoApplication of adaptive time-frequency de-composition in ultrasonic NDE of highly scattering materialsrdquo Ultrasonics 38 pp 961-964

[23] Hong J C Sun K H and Kim Y Y (2005) ldquoThe matching pursuit approach based on the modulated Gaussian pulse for efficient guided-wave damage inspectionrdquo Smart Mate-rials and Structures 14 pp548-560

[24] Herakovich C T (1997) Mechanics of Fibrous Composites Wiley 480 pages

[25] Reddy J N (1985) ldquoA review of the literature on finite-element modeling of laminated composite platesrdquo The Shock and Vibration Digest 17(4) pp 3-8

[26] Reddy J N (2003) Mechanics of Laminated Composite Plates and Shells Theory and Analysis CRC Press 856 pages

[27] Palacz M Krawczuk M and Ostachowicz W (2005) ldquoThe spectral finite element mod-el for analysis of flexure-shear coupled wave propagation Part 1 laminated multilayer

- 34 -

composite beamrdquo Composite Structures 68 pp37-44

[28] Zhang Y C and Harding J (1990) ldquoA numerical micromechanics analysis of the me-chanical properties of a plain weave compositerdquo Computer amp Structures 36(5) pp 839-844

[29] Nemat-Nasser S and Hori M (1999) Micromechanics Overall Properties of Heteroge-neous Materials Elsevier 810 pages

[30] Scida D Aboura Z Benzeggagh M L and Bocherens E (1999) ldquoA micromechanics model for 3D elasticity and failure of woven-fibre composite materialrdquo Composites Science and Technology 59 pp 505-517

[31] Thomas Y and Wu X H (1997) ldquoA multiscale finite element method for elliptic prob-lems in composite materials and porous mediardquo Journal of Computational Physics 134 pp 169-189

[32] Lee K H Moorthy S and Ghosh S (1999) ldquoMultiple scale computational model for damage in composite materialsrdquo Computer Methods in Applied Mechanics and Engineer-ing 172 pp 175-201

[33] Namilae S and Chandra N (2005) ldquoMultiscale model to study the effect of interfaces in carbon nanotube-based compositesrdquo Journal of Engineering Materials and Technology (ASME) 127 pp222-232

[34] Adams D E (2007) Health monitoring of Structural Materials and Components Wiley 460 pages

[35] Gopalakrishnan S (2009) ldquoModeling Aspects in Finite Elementrdquo Encyclopedia of Struc-tural Health Monitoring 2 pp791-809

[36] Palacz M Krawczuk M and Ostachowicz W (2005) ldquoThe spectral finite element mod-el for analysis of flexure-shear coupled wave propagation Part 2 Deminated multilayer composite beamrdquo Composite Structures 68 pp45-51

[37] Baik J M and Thopson R B (1984) ldquoUltrasonic scattering from imperfect interfaces A quasi-static modelrdquo Journal of Nondestructive Evaluation 4(3) pp177-196

[38] Luo H and Hanagud S (2000) ldquoDynamics of delamincated beamsrdquo International Jour-nal of Solids and Structures 37 pp1501-1519

[39] Kim S B and Sohn H (2007) ldquoInstantaneous reference-free crack detection based on polarization characteristics of piezoelectric materialsrdquo Smart Materials and Structures 16 pp2375-2387

[40] An Y K and Sohn H (2010) ldquoInstantaneous crack detection under varying temperature and static loading conditionsrdquo The Special Issue of Structural Control and Health Monitor-ing 17(7) pp 730-741

- 35 -

[41] Lee S J and Sohn H (2006) ldquoActive self-sensing scheme development for structural health monitoringrdquo Smart Materials and Structures 15(6) pp 1734-1746

- 36 -

요 약 문

복합재료 손상의 무기저 진단 및 정량화

복합재료는 재료의 경량성 내구성 고강도의 특성으로 인하여 항공기에서부터

교량에까지 널리 사용되고 있다 하지만 외부의 충격으로 인한 복합재료 내부의

손상은 재료의 강도를 감소시켜 구조물의 안정성을 저하시키는 요인이 된다 하지만

복합재료의 손상은 재료 내부에서 발생하여 육안으로는 식별하기도 힘들며 표면에서

보이는 것만으로 내부의 손상 정도를 판단하기는 불가능하다는 단점을 가지고 있다

본 연구에서는 손상 내부에서 반사로 인해 발생하는 파를 이용하여 기저자료를

사용하지 않고 특정 시점에서 얻은 자료만으로 복합재료의 손상을 진단하고 정량화

할 수 있는 기법을 제안한다 제안된 기법의 적용성을 검증하기 위하여 유한요소법을

이용한 수치해석을 수행하였고 Teflon tape 이 삽입된 시편과 실제 충격하중으로

인한 손상을 입은 시편을 이용한 실험을 진행하였다 수치해석 및 실험 결과 제안된

기법이 복합재료의 손상을 성공적으로 진단 및 정량화할 수 있음을 확인하였다 또한

다양한 외부 온도 환경에서 제안된 기법의 환경영향 실험을 수행하였고 다양한

환경에서 손상의 진단 및 정량화가 성공적으로 이루어짐을 검증하였다

핵심어 무기저 손상진단 및 정량화 압전센서 복합재료 유한요소법 Matching pursuit

method

- 37 -

감 사 의 글

기대와 두려움을 안고 시작한 석사과정이 어느덧 2 년 이라는 세월 속에서 추억으로 자리잡아

가고 있습니다 학부생 기간을 포함한 지난 3 년 동안 SHM 이라는 새로운 분야를 접하고 연구를

수행 할 수 있도록 도와주신 고마운 분들이 있어 그 감사를 표하고자 합니다

우선 저의 많은 부족한 점을 일깨워 주시고 연구에 대한 뜻을 더욱 확고히 세울 수 있게 해

주신 지도교수님이신 손 훈 교수님께 깊은 감사의 뜻을 전합니다 짧은 기간이었지만 교수님의

도움으로 많은 것을 배우고 느끼고 깨달을 수 있었습니다 또한 본 연구를 수행할 수 있도록 지

원해주신 국토해양부 (U-City 석 박사과정 지원사업)에 깊은 감사 드립니다

때로 힘들고 지칠 때 항상 저를 믿고 응원해 주시는 가족이 있었기에 지금의 글을 쓸 수 있었

습니다 또한 지금 연구실에서 함께 생활하고 있는 윤규형 민구형 철민이형 현석이형 진열이형

호민이형 병진이 선혜누나와 성대에 계신 창길이형 미국에 계신 의재형과 Debaditiya 에게도 감

사의 말씀을 전합니다 그리고 모두 나열할 수는 없지만 카이스트 생활의 활력소가 되어준 카이

스트 건설 및 환경공학과 선후배들 직원 분들과 SOS 멤버들에게도 감사를 표합니다

앞서 거론한 많은 이들의 도움이 없었다면 저 또한 지금 이 자리에 있을 수 없었을 것입니다

많이 부족하기에 더 노력하고 배우고자 했던 초심을 다시 한번 생각하며 더 큰 결실을 맺기 위해

앞으로 나아가겠습니다 도움을 주신 모든 분들께 이 감사의 글을 바칩니다

2011년 6월 카이스트에서 임형진 올림

- 38 -

CURRICULUM VITAE

Hyung Jin Lim

Research Assistant

The Civil and Environmental Engineering Department

Korea Advanced Institute of Science and Technology (KAIST)

291 Daehak-Ro Guseong-Dong Yuseong-Gu

Daejeon Republic of Korea 305-701

Phone (82)+42-350-3665 Fax (82)+42-350-3610

Email limhj87gmailcom

RESEARCH INTERESTS

Damage Detection Localization and Quantification on Composite and Metallic Structures Pattern Recognition Finite Element Modeling (FEM)

EDUCATION

2009 ndash 2011 MS Civil and Environmental Eng Korea Advanced Institute of Science and Technology (KAIST) Korea Rep

2006 ndash 2009 BS Civil and Environmental Eng Korea Advanced Institute of Science and Technology (KAIST) Korea Rep

JOURNAL PUBLICATIONS The corresponding author is underlined 1 Hyung Jin Lim Hoon Sohn ldquoReference-free damage detection localization and quantification

using in compositerdquo in preparation for Journal of the Acoustical Society of America 2011

2 Chul Min Yeum Hoon Sohn Jeong Bum Ihn Hyung Jin Lim ldquoInstantaneous delamination detection in a composite using a dual piezoelectric transducer networkrdquo Submitted to Compo-sites Science and Technology 2011

3 Yun-Kyu An Hyung Jin Lim Min Koo Kim Jin Yeol Yang Hoon Sohn Chang Guen Lee ldquoInvestigation of applicability of local reference-free damage detection techniques to in-situ

- 39 -

bridgesrdquo Submitted to ASCE Journal of Structural Engineering 2011

4 Hyung Jin Lim Min Koo Kim Hoon Sohn Chan Yik Park ldquoImpedance based damage detec-tion under varying temperature and loading conditionsrdquo Submitted to NDTampE International 2011

5 Hyung Jin Lim Hoon Sohn ldquoQuantitative estimation of transmitted and reflected lamb waves at discontinuityrdquo Journal of the Korean Society for Nondestructive Testing Vol30 No 4 pp1-8 2010

CONFERENCE PROCEEDINGS

The corresponding author is underlined 1 Hyung Jin Lim and Hoon Sohn ldquoReference-free damage detection localization and quantifi-

cation in compositesrdquo ASCE Engineering Mechanics Institute Conference Boston June 2-4 2011

2 Hyung Jin Lim and Hoon Sohn ldquoReference-free delamination quantification within multi-layer composite materialsrdquo KSNT Spring Conference May 26-27 Gyeong-Ju Korea 2011

3 Min Koo Kim Hyun Jin Lim Hoon Sohn Chan Yik Park ldquoImpedance-based bolt loosening detection under varying temperature and loadingrdquo Asian Pacific Workshop on Structural Health Monitoring Tokyo Japan November 30-December 2 2010

4 Hyung Jin Lim Chul Min Yeum and Hoon Sohn ldquoModeling of Impact-induced Delamination in a Multilayer Composite Platerdquo The 23rd KKCNN Symposium on Civil Engineering Taipei (Taiwan) November 13-15 2010

5 Yun-Kyu An Hyung Jin Lim and Hoon Sohn ldquoTemperature independent Delamination Detec-tion using data normalizationrdquo The 3rd International Conference on Multi-Functional Materi-als and Structures Chonbuk National Univ Jeonju Korea September 14-18 2010

6 Hoon Sohn Yun-Kyu An Chang Gil Lee Min Koo Kim Hyung Jin Lim ldquoApplication of online NDT techniques to decommissioned bridge testingrdquo the 5th European Workshop on Structural Health Monitoring Sorrento Italy June 29-July 02 2010

7 Min Koo Kim Hyung Jin Lim Hoon Sohn Chan Yik Park ldquoImpedance based damage diag-nosis of a complex composite aircraft wing under changing loading and temperature condi-tionsrdquo the 5th European Workshop on Structural Health Monitoring Sorrento Italy June 29-July 02 2010

8 Yun-Kyu An Min Koo Kim Hyung Jin Lim and Hoon Sohn ldquoValidations of reference-free crack detection techniques through a decommissioned bridge testrdquo 2010 COSEIK Annual Con-ference Jeju Korea April 8-9 2010

- 40 -

PARTICIPANT PROJECTS

1 Development of Smart Steel Members for Online Steel Structure Monitoring POSCO Corporation (Funded 41000000 KRW (41000 USD) for 010511 to 123111)

2 Development of Enabling Technologies for Structural Health Monitoring Korean Agency for Defense Development (Funded 240000000 KRW (240000 USD) out of total 3200000000 KRW (3200000 USD) for 060108-053111)

3 Development of On-board SHM Technologies for Composite Air Vehicles The Boeing Company (Funded 400000 USD (400000000 KRW) for 080108 to 073111)

EDUCATIONAL EXPERIENCES

Fall 2009

Teaching Assistant CE207 ldquoElementary Structural Engineering and Laboratoryrdquo KAIST 3 unit undergraduate course

EXPERIMENTAL EXPERIENTCES

2010 Monitoring of Real Aircraft Composite Wing Segment Korea Agency for Defense De-velopment (ADD) Daejeon Korea Real time monitoring of composite air vehicles fund-ed by Korea Agency for Defense Development (ADD)

2009-2011

Field Monitoring of Real Bridge Structures Gimpo (Ramp G bridge) amp Icheon (Sam-seung bridge) Korea US-Korea collaborative research for bridge monitoring testbeds

HONORS amp AWARDS

Scholarships for outstanding students granted KAIST (2007-2009) The three prominent students in the civil and environmental engineering department of KAIST were selected as recipients This scholarship provided 2000$ per year for three years

- 41 -

Page 17: 복합재료 손상의 무기저 진단 및 정량화 - KAISTssslab.kaist.ac.kr/article/pdf/Thesis_2.pdf · 1 Declaration of Ethical Conduct in Research: I, as a graduate student

22 Multiple reflections in the damage area

Standing wave patterns due to multiple reflections of A0 Lamb wave mode in the damage

area were observed through numerical simulations and ultrasonic wave-field imaging using a

scanning laser vibrometer [16] The formation of standing waves in damage is illustrated

through Fig 2 When the incident waves enter the damage area a significant portion of the

waves are reflected back from the exit Then the reflected waves travel through the damage

and undergo reflection again at the original entrance As a consequence of multiple reflec-

tions at the entrance and the exit a considerable amount of ultrasonic energy is trapped inside

the damage area [17-19]

Multiple Reflections

Incident Wave Transmitted Wave

Damage

Fig 2 Multiple reflections in the damage area The multiple reflections of A0 Lamb waves in the damage area are utilize as a feature for characterization of the damage in this study

- 6 -

CHAPTER 3 THEORETICAL DEVELOPMENT

31 Overview

Damage

l = l1 + d + l2

d = vdtd l1 = vit1 l2 = vit2

t = t1 + td + t2

Excitation and Sensing PZT Sensing PZT

l

Transmitted Wave

Reflected Waves

Incident Wave

Fig 3 Parameters and notations of the proposed damage characterization technique Damage is detected by identifying the additional reflected wave from the damage due to multiple re-flections (dashed lines) Localization and quantification of the damage is carried out by measuring the arrival time of transmitted and reflected waves from the damage

In this section the reference-free damage characterization algorithm is proposed Damage

diagnosis and quantification are achieved by identifying the additional reflected waves due to

multiple reflections in the damage area (dashed lines in Fig 3) Moreover damage is local-

ized using additional information from pulse-echo method implement For the localization

and the quantification of damage the variables such as the path length and A0 Lamb wave

mode velocity in damage area are defined (Fig 3) For example 119897119897 is expressed as

119897119897 = 1198971198971 + 119889119889 + 1198971198972 (1)

where 1198971198971 119889119889 and 1198971198972 are the distance from the excitation PZTs to the entrance of damage the

damage length and the distance from the exit of damage to the sensing PZT respectively

- 7 -

Moreover 119907119907119894119894 and 119907119907119889119889 are also defined as the A0 Lamb wave mode velocity in pristine area

and damage area respectively Note that the whole procedure of the damage characterization

is accomplished based solely on the signals obtained from the current state of the structure

32 Damage index (m-value)

The proposed reference-free damage characterization technique utilizes the waves reflect-

ed additionally from the damage area as a feature The arrival time difference of the transmit-

ted and reflected A0 Lamb wave modes reflected from the exit and the entrance of the delam-

ination area one times each (∆119905119905) contains the information of the physical length of the dam-

age (119889119889) and A0 Lamb wave mode velocity in the damage area (119907119907119889119889) The wave velocity in the

damage area which is related to the shear modulus (G) represents the severity of the damage

∆119905119905 = 2119905119905119889119889 =2119889119889119907119907119889119889

(2)

However based on the measured data available the 119889119889 and 119907119907119889119889 (or G) cannot be uniquely

determined Fig 4 shows two FEM models which have different physical length and wave

velocity (G) of the damage Detail explanation of delamination modeling technique will be

covered in section 4 The similar signal in the time domain and ∆119905119905 are obtained from the

two different damage cases

- 8 -

Damage 1

Damage 2

20

Unit mm15

90 G value reduction

943 G value reduction

m-value = 248

m-value = 248

015 02 025 03 035

-2

-1

0

1

2

Time (ms)

Ampl

itude

Damage 1Damage 2

Fig 4 Two different FEM models which have different 119889119889 and 119907119907119889119889 (or G) This represents that the damage cannot be characterized uniquely based on the measured data available

Thus in this study the damage index (m-value) which can be uniquely defined based on

the available data is introduced The m-value is defined as the ratio of the physical length of

the damage and wavelength of the A0 Lamb wave mode in the damage area (120582120582119889119889)

119898119898 =119889119889120582120582119889119889

=119889119889119889119889119907119907119889119889

(3)

In many cases the transmitted and reflected waves are overlapped and it is hard to decom-

pose and calculate the arrival times of them Matching pursuit method is an alternative way

to accurately decompose of the overlapped waves and estimate the parameters of each wave

such as arrival time (Fig 5) Detail explanation of the matching pursuit method is covered in

section 36

015 02 025 03-2

-1

0

1

2

Time (ms)

Am

plitu

de

OriginalApproximated

(a)

015 02 025 03-2

-1

0

1

2

Time (ms)

Ampl

itude

Wave 1Wave 2

(b)

Fig 5 (a) Example of the overlapped signal with comparing approximated signal using the matching pursuit method (b) Decomposed signal using the matching pursuit method The matching pursuit method provides a conventional way to decompose and identify each wave

- 9 -

33 Damage detection algorithm (Level 1)

The damage diagnosis is accomplished by extracting the additional reflected A0 Lamb

wave modes from damage due to multiple reflections If A0 Lamb wave modes are decom-

posed perfectly it is trivial that one transmitted A0 Lamb wave mode is always captured in

the response signal If damage exists other waves in the response signal behind the transmit-

ted A0 Lamb wave mode are presented Since this approach does not rely on previously col-

lected data the diagnosis is independent from environmental variation such as temperature

changes

It is reported that the dispersion of Lamb waves cause wave-packets to spread out in space

and time as they propagate through a structure [20] Thus the reflected waves from the dam-

age have longer pulse duration than the transmitted wave due to their longer propagating dis-

tance In this study the matching pursuit method is modified to capture more spread wave

than the transmitted A0 Lamb wave mode Detail explanation of the modified matching pur-

suit method is described in latter section

Let 1199041199041 and 1199041199042 as the pulse duration (scale) of the transmitted and additional reflected A0

Lamb wave mode respectively

ldquoIf 1199041199042 gt 1199041199041 damage exists

Otherwise there is no damagerdquo

- 10 -

34 Damage quantification algorithm (Level 3)

As mentioned in the previous section damage is quantized by estimating 119889119889 and 119907119907119889119889 using

pitch-catch method The procedure to estimate 119889119889 and 119907119907119889119889 is

(i) Measure 119897119897 119905119905 and 119905119905119889119889

(ii) Assume the range of 119889119889 (0 lt 119889119889 lt 119897119897)

(iii) Calculate 119907119907119894119894 and 119907119907119889119889 with a constraints (0 lt 119907119907119889119889 lt 119907119907119894119894) (Eq 4 and 5)

(iv) Estimate m-value and find a feasible range of 119889119889 and 119907119907119889119889

119907119907119894119894 =119897119897 minus 119889119889119905119905 minus 119905119905119889119889

(4)

119907119907119889119889 =119889119889119905119905119889119889

(5)

35 Damage localization algorithm (Level 2amp3)

For damage localization 1198971198971 is estimated by measuring the arrival time from the excitation

PZT to the entrance of delamination (1199051199051) with additional pulse-echo method implementation

The procedure is similar with the previous damage quantification algorithm

(i) Measure 119897119897 119905119905 119905119905119889119889 and 1199051199051

(ii) Assume the range of 1198971198971 (0 lt 1198971198971 lt 119897119897)

(ii) Compute possible combinations of 1198971198972 119889119889 and 119907119907119889119889 with constraints

(0 lt 119907119907119889119889 lt 119907119907119894119894 119889119889 gt 0) (Eq 5 6 and 7)

- 11 -

1198971198972 = 1198971198971119905119905 minus 1199051199051 minus 119905119905119889119889

1199051199051 (6)

119889119889 = 119897119897 minus 1198971198971 minus 1198971198972 (7)

Since a PZT for pulse-echo method acts as both an actuator and a sensor the obtained sig-

nal contains both the excitation and the response signal Thus the reflected wave from the

entrance of damage can be buried into the tail of excitation signal and cause error on arrival

time of reflected wave from damage Even in this case damage can be localized and quan-

tized using the constraints When first wave packet reflected from the entrance of damage is

buried into the excitation signal the next wave packet reflected from the exit of damage can

be utilized alternatively If the algorithm regards the second reflected wave as the first re-

flected wave there are no combinations of estimated variable meeting the constraints Then

the algorithm repeats the process by regarding the wave as the second reflected wave the

constraints can select proper combination of the estimated variables

The estimated values have range bounded by the constraints When additional information

of 119907119907119894119894 is utilized the damage can be characterized more accurately In this study three meth-

ods are proposed and verified as information level of 119907119907119894119894

Method 1 Solely based on the current data measurement (Reference-free)

Method 2 Limit the possible range of 119907119907119894119894 (Prior knowledge of the system)

Method 3 Measure 119907119907119894119894 (Additional data need to be collected)

Method 1 described in this section can be utilized without any information of 119907119907119894119894 Meth-

od 2 produces more accurate estimation results using prior knowledge of the system such as

- 12 -

119907119907119894119894 variation under temperature changes Method 3 gives unique estimated value using meas-

ured 119907119907119894119894 from other path in the same time In section 4 and 5 damage is characterized using

these three methods and the results are compared each another

36 Matching pursuit method considering the dispersion effect of Lamb

wave

Matching pursuit method has been developed for signal compression approximation and

characterization of obtained signal and function [21-23] The key concept of the matching

pursuit method is that any signal (function) can be decomposed into a linear expansion of

signals selected from a pool of basis signals (functions) called lsquodictionaryrsquo

Let 119891119891 be a function to be analyzed then 119891119891 can be approximated a set of functions select-

ed from the dictionary 119863119863 in order to best match This is done by successive approximations

with orthogonal projections Let1198921198921205741205740 isin 119863119863 the function 119891119891 can be decomposed into

119891119891 = lang1198911198911198921198921205741205740rang1198921198921205741205740 + 119877119877119891119891 (8)

where 119877119877119891119891 is the residual function after approximating 119891119891 using 1198921198921205741205740 1198921198921205741205740 is orthogonal to

119877119877119891119891 hence

1198911198912 = lang1198911198911198921198921205741205740rang2

+ 1198771198771198911198912 (9)

To minimize 119877119877119891119891 1198921198921205741205740 is chosen to maximize lang1198911198911198921198921205741205740rang The matching pursuit method is

an iterative algorithm that decomposes the residue 119877119877119891119891 by projecting it on a function of 119863119863

that matches 119877119877119891119891 almost at best as it was done for 119891119891 The residue 119877119877119899119899119891119891 is decomposed into

119877119877119899119899119891119891 = lang119877119877119899119899119891119891119892119892120574120574119899119899rang119892119892120574120574119899119899 + 119877119877119899119899+1119891119891 (10)

- 13 -

the decomposition is carried up to the order 119898119898 119891119891 can be expressed as

119891119891 = lang119877119877119899119899119891119891119892119892120574120574119899119899rang119892119892120574120574119899119899

119898119898minus1

119899119899=0

+ 119877119877119898119898119891119891 (11)

The original function 119891119891 is decomposed into a sum of dictionary elements that are chosen to

best match its residues

In this study the matching pursuit method is utilized to accurately decompose of the over-

lapped Gaussian pulses and estimate the parameters of the pulse which used for damage

characterization The 119894119894119905119905ℎ Gaussian pulse in the dictionary for approximating the obtained

signal is expressed as

119892119892120574120574119894119894 = 119891119891(119860119860119894119894 119906119906119894119894 119904119904119894119894 119889119889119894119894120601120601119894119894) (12)

where 119860119860119894119894119906119906119894119894119904119904119894119894119889119889119894119894 and 120601120601119894119894 represent amplitude arrival time scale (pulse duration) excita-

tion(central) frequency and phase of the Gaussian pulse respectively The parameters of the

pulse are estimated using a nonlinear regression based optimization method (Gauss-Newton

method) based on an initial guess of the parameters Detail procedure for estimating the pa-

rameters of the pulse is expressed in Hong et al [23]

Among the estimated parameters of the pulses pulse duration (scale) represents the disper-

sion effect If a wave-packet is spread out the estimated pulse duration is increased Using

this feature the matching pursuit method is modified to capture waves which have longer

pulse duration than transmitted A0 Lamb wave mode If the estimate pulse duration of a wave

is less than the transmitted wave the modified matching pursuit method regards it as error

- 14 -

CHAPTER 4 NUMERICAL SIMULATION

41 Composite modeling method

Composite materials can be modeled from nano to macro level depends on its engineering

purpose In general four methods are utilized for modeling of composite materials First

composite is modeled as a homogenous material expressed in terms of effective material

properties [24] The effective material properties are estimated using the modulus and the

volume fraction of each constituent theoretically (Rule of mixture) and measured experimen-

tally Second a model is established that each laminate which have own material properties

are totally bonded together In this model the stacking sequence and direction of each lami-

nates are considered additionally [25-27] Third method for composite modeling in micro

level called micromechanics [28-30] This model considers the interaction between fibers and

matrix in laminate Last multiscale modeling which the scope of modeling is extended to

molecular or quantum level of constituents [31-33] This model starts from molecular me-

chanics or molecular dynamics then the scale of model leads to the nano micro and macro

level by incorporating small scale models As the scale of models is smaller it provides more

logical way than other large scale of models while the computational cost is more expensive

The scope of this study is focused on the dynamic behavior of composite structure the beam

or plate model was utilized for the numerical simulation

- 15 -

42 Damage modeling method

Generally three damage models have been developed and utilized in the real applications

First the stiffness reduction model called as the equivalent element model was introduced

[17 34-35] This model is simply realized by replacing the region around damage area by a

region with adjusted material properties (Youngrsquos modulus shear modulus density etc)

globally Next the duplicated node model is done by disconnecting the element connectivity

assuming there is no interaction among the delaminated surface This model can be utilized

when the location and the extension of damage is important issue [34-36] Last the contact

mechanics model using the principles of contact mechanics the two interfacial forces (con-

tact resistance and sliding resistance) in the delaminated surface [37-38] In this model the

delaminated surfaces are modeled as nonlinear functions of the interlaminar normal strain

and the interlaminar shear strain This model is computationally expensive while provides

more logical way than other models There is however no universally accepted model for

general impact-induced damage in composites In the case of impact-induced damage it is

impossible to predict or estimate the location and number of delaminated layers Thus from

the uncertainty of impact-induced damage the equivalent element model can be considered

as the effective damage model

- 16 -

43 Numerical simulation description

TABLE I Material property of simulation model and test specimen

Tensile Strength 657 MPa Tensile Modulus 59 GPa Shear Strength 804 MPa Shear Modulus 24 Gpa Specific Gravity 16

PZT A

PZT C

PZT BDamage

3

250 125 125

500 Unit mm

Fig 6 Description of the numerical simulation model

The feasibility of the proposed damage characterization technique is verified by 2-D FEM

simulation using MSC PATRAN and NASTRAN Fig 6 presents the dimension PZTs and

damage location of the simulation model In the model 05mm x 05mm square shell element

was used for modeling of the A0 mode whose flexural wave characteristic The element size

was determined by considering the wavelength of the propagation Lamb wave modes The

material properties of the simulation model and damage cases are presented in Table I and

Table II respectively

PZT A and C act as both actuator and sensor for collecting pulse-echo signal They were

installed on the plate which two of them are both-side collocated for the A0 Lamb wave

- 17 -

mode decomposition [39-40] PZT B was only used as a sensor for collecting pitch-catch sig-

nal For simulating the excitation and sensing of PZTs the pin-force model was adopted that

an input force apples to the end nodes of excitation PZT and averaging element strains corre-

sponding sensing area A Gaussian pulse (tone-burst) 50 kHz central frequency was excited

as an input signal

TABLE II Damage cases for numerical simulation

119949119949120783120783 (mm) 119941119941 (mm) G value reduction ()

Case 1 650 200 90 darr

Case 2 675 150 90 darr

Case 3 675 150 93 darr

Case 4 900 150 90 darr

- 18 -

44 Numerical simulation results

015 02 025 03 035

-2

0

2

Time (ms)

Am

plitu

de

IntactDamage

Reflected wave from damage

(a) Pitch-catch signal

015 02 025 03 035

-2

0

2

Time (ms)

Am

plitu

de

IntactDamage

Reflected waves from damage

(b) Pulse-echo signal Fig 7 Comparison of A0 Lamb wave mode signals from numerical simulation between intact and damage cases Time delay attenuation and multiple reflections due to damage simulated successfully using the damage modeling method described in the previous section

Fig 7 shows that comparison of A0 Lamb wave mode signals between intact and one of

damage cases in the simulation The features of A0 Lamb wave mode passing through dam-

age area described in section 2 are successfully simulated Thus the equivalent element

method is proper for modeling damage in composites In the damage cases the additional

waves reflected from damage is observed in both pitch-catch and pulse echo signal

Table III presents the result of the proposed damage characterization algorithm using the

numerical simulation cases shown in Table II In lsquoMethod 2rsquo 119907119907119894119894 variation is assumed to plusmn

5 of the normal value (1190~1315ms) The normal value of 119907119907119894119894 for lsquoMethod 3rsquo is meas-

ured as 1253 ms In each case the damage is characterized using the proposed technique

Note that as the information of 119907119907119894119894 is specified the estimated value is closer to the exact

value

- 19 -

TABLE III Damage characterization for the numerical simulation models

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

Case 1

Exact 650 200

Method 1 248 50~70 79~414 160~834

Method 2 248 603~667 135~241 271~486

Method 3 248 635 188 379

Case 2

Exact 675 150

Method 1 193 60~75 01~251 37~651

Method 2 193 618~682 114~222 295~575

Method 3 193 65 168 435

Case 3

Exact 675 150

Method 1 233 60~75 66~302 140~649

Method 2 233 668~738 85~196 182~419

Method 3 233 703 140 301

Case 4

Exact 900 150

Method 1 204 75~100 48~349 118~856

Method 2 204 860~951 107~216 264~530

Method 3 204 906 162 397

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1253 ms 1190~1315 ms Measured 119907119907119894119894 1253 ms

- 20 -

CHAPTER 5 EXPERIMENTAL VERIFICATION

51 Overview

In this section experimental study is investigated so that the proposed damage characteri-

zation technique Overall experimental setup and test specimen are shown in Fig 8 and 9

The experimental setup consists of test specimens data acquisition (DAQ) system and self-

sensing circuit Self-sensing circuit was installed for pulse-echo measurement using PZTs

[40-41]

PZT A(PZT C) PZT B

500

500

140

75

250

Teflon Inserted Damage (d =25)

Unit mm

Thickness = 3

PZT A(PZT C) PZT B

500

500

150

90

250

Impact Location

Unit mm

Thickness = 3

(a) Teflon inserted damage specimen

(b) Impact-induced damage specimen

Fig 8 Test specimens for the experimental verification In the Teflon inserted damage Teflon tapes with 25mm diameter were inserted at the center of the specimen The impact-induced damage was introduced to the specimen using the impact test equipment

- 21 -

511 Specimen

Two carbon composite plates with a dimension of 500times500times3mm3 were manufactured by

Nexcomes Inc by stacking 12 layers of CF3327 3K prepreg (woven fabrics type) manufac-

tured by Hankuk Fiber Inc The material properties of the specimen are presented in Table I

One of the specimens Teflon tapes with 25mm diameter was inserted in every 3 layers at the

center of the specimen to mimic the damage The other specimen is utilized for the character-

ization of real impact-induced damage

512 Data acquisition (DAQ) system and PZT

The data acquisition (DAQ) system consists of an arbitrary waveform generator (AWG) a

high speed signal digitizer (DIG) and two multiplexers (MUX) Using the 16-bits AWG

tone-burst input signal with a plusmn10 peak-to-peak voltage were generated The response signal

was measured by the 14-bit DIG For the signal-to-noise ratio reduction both pitch-catch and

pulse-echo signals were measured 150 times and averaged in the time domain 3 identical

circular PZTs of 10mm diameter manufactured by APC International Ltd were installed on

the specimen A Gaussian pulse (tone-burst) 50 kHz central frequency was excited as an in-

put signal The excitation frequency was chosen in order to avoid overlapping of the S0 and

A0 Lamb wave modes and generating of the higher modes (S1 A1hellip)

- 22 -

PZT Actuator

PZT Sensor

AWG

DIG

DAQ System Specimen

SensingActuatingMUX

1

23

4

1 Sending signal to MUX

4 Sending signal to DIG

Self-SensingCircuit

18 Unit mm

10

(a)

(b)

Fig 9 Overall experimental setup (a) Data acquisition (DAQ) system and (b) Installed circu-lar PZT transducer on the specimen with its dimension

513 Impact test

The impact-induced damage was introduced to the specimen using the impact test equip-

ment presented in Fig 10 A 5kg mass with 10mm tip was dropped from 015m height onto

the same location of the specimen The signals were obtained at each number of impacts

Since the reflected wave is not detected at the first impact case the data obtained from the

second and third impact were utilized for the analysis

10

Unit mm (a)

(b)

Fig 10 (a) Impact test equipment for introducing impact-induced damage and (b) Tip of the impact tester with 10mm diameter

- 23 -

514 Temperature effect test

In reality structures are subjected to changing environmental conditions such as tempera-

ture that affect measured signal and cause false-alarm Thus the robustness of the proposed

damage characterization technique under temperature variation in investigated (Fig 11) The

specimens were placed inside a temperature chamber and a thermocouple was installed on

the specimen to measure temperature of surface of the specimen The signal was obtained

under three temperature cases (0 20 and 50degC) and the humidity was kept at 30

Temperature Chamber

Test Specimen

Fig 11 Test specimen in temperature chamber

- 24 -

52 Teflon inserted damage

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Reflected wave from damage

(a)

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Transmitted WaveReflected Wave

Detected reflected wave

(b)

Fig 12 A0 Lamb wave mode obtained using pitch-catch method from the Teflon inserted damage specimen (a) The reflected wave from damage is shown (b) The Teflon inserted damage is detected by capturing the reflected wave

Fig 12 shows the obtained pitch-catch signal from the Teflon inserted damage specimen

and damage diagnosis result The Teflon inserted damage is detected successfully by captur-

ing the reflected wave Table IV presents the result of the Teflon inserted damage characteri-

zation 119907119907119894119894 is measured as 1213ms and the variation is assumed to plusmn 5 of the normal val-

ue (1152~1273ms) The results indicate that the proposed damage characterization approach

works well in real application

TABLE IV Damage characterization for the Teflon inserted damage specimen

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

Exact 635 25

Method 1 123 60~75 73~338 298~1220

Method 2 123 619~662 229~310 934~1188

Method 3 123 631 284 1161

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1213 ms

- 25 -

53 Impact-induced damage

11 115 12 125 13-4

-2

0

2

4

Time (ms)

Vol

tage

(mV

)

IntactImpact2Impact3

Reflected wave from damage

(a) Pitch-catch signal

11 115 12 125 13 135

-05

0

05

1

Time (ms)

Volta

ge (m

V)

IntactImpact2Impact3

Boundary reflection

Reflected wave from damage

(b) Pulse-echo signal

Fig 13 The signal obtained from the test specimen subjected to impact The impact was in-troduced at the same location of the specimen and the signal was obtained after each impacts As the number of impact is increased the transmitted A0 Lamb wave mode is delayed and attenuated

The obtained signals from impact-induced damage specimen are presented in Fig 13 It is

observed that as the number of impact is increased the transmitted A0 Lamb wave mode is

more delayed because the wave velocity in the damage area is reduced Fig 14 shows the

false-alarm test result whether the proposed damage detection algorithm works even in pris-

tine condition of the test specimen It is shown that no false alarm is occurred because no ad-

ditional reflected wave is captured

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Transmitted Wave

(a) Transmitted wave

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Reflected Wave

No reflected wave

(b) Reflected wave

Fig 14 False alarm test for the proposed damage detection algorithm in pristine condition of the specimen Reflected wave from damage is not captured false alarm is not occurred

- 26 -

TABLE V Damage characterization for the impact-induced damage specimen

of Impacts m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

2

(1470 J)

Method 1 080 65~90 75~246 470~1157

Method 2 080 801~885 114~245 711~1134

Method 3 080 843 179 1122

3

(2205 J)

Method 1 148 50~100 06~380 21~1134

Method 2 148 758~838 249~368 843~1125

Method 3 148 798 308 1044

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1213 ms

Table V presents the characterization results of the impact-induced damage case Here the

measured 119907119907119894119894 and the assumed variation are same as the artificial damage case It is observed

that as the number of impact is increased 119889119889 is increased while 119907119907119889119889 is reduced Thus it is

validated that the proposed damage characterization approach estimates the increment of the

damage well

54 Temperature effect test

541 Teflon inserted damage

11 115 12 125 13-10

-5

0

5

10

Time (ms)

Vol

tage

(mV

)

0 oC20 oC50 oC

(a) Pitch-catch signal

11 115 12 125 13 135

-1

0

1

2

3

Time(ms)

Volta

ge (m

V)

0 oC20 oC50 oC

(b) Pulse-echo signal

Fig 15 Signals obtained from the Teflon inserted damage specimen under temperature varia-tion Time delay and amplitude variation are occurred

Fig 15 shows the obtained signals from the Teflon inserted damage specimen subjected to

- 27 -

temperature variation Time delay and amplitude variations are occurred there is no addition-

al damage is introduced The analyzed results of the obtained signals are reported in Table VI

The measured 119907119907119894119894 is 1247 ms at 0 degC 1213 ms at 20 degC and 1168 ms at 50 degC respectively

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms (1152~1273 ms)

The estimated results are also changed under temperature variations the fluctuation of esti-

mated range and exact value is small

TABLE VI Temperature effect test for the Teflon inserted damage specimen

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

0 degC

Exact 635 25

Method 1 113 60~70 79~268 352~1291

Method 2 113 560~619 233~344 1034~1172

Method 3 113 606 257 1141

20 degC

Exact 635 25

Method 1 123 60~75 73~338 298~1220

Method 2 123 619~662 229~310 934~1188

Method 3 123 631 284 1161

50 degC

Exact 635 25

Method 1 137 60~75 72~328 263~1228

Method 2 137 622~687 183~299 666~1086

Method 3 137 642 264 960

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1247 ms at 0 degC 1213 ms at 20 degC and 1168 ms at 50 degC

- 28 -

542 Impact-induced damage

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Vol

tage

(mV

)

Transmitted WaveReflected Wave

No reflected wave

(a) 0 degC

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Transmitted WaveReflected Wave

No reflected wave

(b) 50 degC

Fig 16 False alarm test under temperature variations It is shown that proposed damage char-acterization algorithm is robust under temperature variations

The false alarm test in pristine condition of the impact-induced damage specimen under

temperature variation was investigated (Fig 16) It is validated that the proposed damage di-

agnosis algorithm is robust under temperature variations Fig 17 shows the obtained signals

from the impact-induced damage (impacted 3 times) specimen under temperature variation

The analyzed results of the obtained signals are reported in Table VII The measured 119907119907119894119894 in

each temperature and the assumed variation are same as the artificial damage case

11 115 12 125 13-5

0

5

Time (ms)

Volta

ge (m

V)

0 oC20 oC50 oC

(a) Pitch-catch signal

11 115 12 125 13 135

-1

0

1

2

3

Time(ms)

Vol

tage

(mV

)

0 oC20 oC50 oC

(b) Pulse-echo signal

Fig 17 Signals obtained from impact-induced damage specimen (Impact 3) under tempera-ture variation

- 29 -

TABLE VII Temperature effect test for the impact induced damage specimen (Impact 3 case)

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

0 degC

Method 1 135 75~90 72~310 267~1148

Method 2 135 705~779 264~381 978~1126

Method 3 135 763 289 1071

20 degC

Method 1 148 50~100 06~380 21~1134

Method 2 148 758~838 249~368 843~1125

Method 3 148 798 308 1044

50 degC

Method 1 153 75~100 53~415 174~1227

Method 2 153 804~889 214~336 702~1103

Method 3 153 815 320 1050

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1247 ms at 0 degC 1213 ms at 20 degC and 1168 ms at 50 degC

- 30 -

CHAPTER 6 CONCLUDING REMARKS

61 Executive summary

In this study the reference-free damage detection localization and quantification tech-

nique in a multi layered composite plate is proposed Additional reflected A0 Lamb wave

mode from the damage due to multiple reflections in damage area is used as a feature for

characterizing the damage First damage is characterized in terms of a damage index (m-

value) defined as the ratio of the physical length of the damage to the wavelength of the A0

Lamb wave mode Then the range of possible damage location length and severity are esti-

mated with constraints This procedure can be accomplished without prior baseline data To

realize the proposed technique the matching pursuit technique is modified for reduction of

the error by considering the dispersive characteristic of Lamb wave The matching pursuit

method estimates the arrival time of each wave packets accurately even two wave packets are

overlapped The numerical simulations and experimental tests including Teflon tape inserted

and impact-induced damage are investigated for the feasibility of the proposed technique

Moreover the effectiveness of the proposed technique under changing temperature is exper-

imentally verified

- 31 -

62 Suggestion for further study

The limitations of this study and the possible improvement and further studies are listed as

follow

1 Measurement of impact-induced damage length using conventional NDT technique

Additional experimental verification for impact-induced damage characterization is

needed by measuring the real length of damage using conventional NDT technique

such as ultrasonic scan or infrared camera

2 Limitation of both side collocated PZT installation In this study excitation PZTs are

installed both side and collocated for decomposing A0 Lamb wave mode However it

can produce decomposition error due to imperfection of PZT location and their installa-

tion Moreover in many cases assess to both surface of structures is not allowed such

as aircraft applications To improve this limitation a reference-free damage characteri-

zation using single side PZTs will be considered

3 Complex structure application In this study damage is characterized in simple plate-

like structure When the monitored structure becomes complex the response signals al-

so becomes complex due to its geometry and boundary Thus the investigations for

complex structure application such as stiffener spar and curvature are needed

- 32 -

REFERENCES

[1] Cawley P and Alleyne D (1996) ldquoThe use of Lamb waves for the long range inspection of large structuresrdquo Ultrasonics 34 pp287-290

[2] Ghosh T Kundu T and Karpur P (1998) ldquoEfficient use of Lamb waves for detecting defects in large platesrdquo Ultrasonics 36 pp791-801

[3] Guirguitui V (2003) ldquoLamb wave generation with piezoelectric wafer active sensors for structural health monitoringrdquo SPIE Conference on Smart Structures and Materials San Diego CA 5391 pp111-422

[4] Ragvahan A and Cesnik CE S (2004) ldquoModeling of piezoelectric-based Lamb wave generation and sensing for structural health monitoringrdquo SPIE Conference on Smart Struc-tures and Materials San Diego CA 5391 pp419-430

[5] Rose JL Pilarski A and Dirit JJ (1993) ldquoAn approach to guided wave mode selection for inspection of laminated platerdquo Journal of Reinforced Plastics and Composites 12(5) pp536-544

[6] Guo N and Cawley P (1993) ldquoThe interaction of Lamb waves with delaminations in composite laminatesrdquo Journal of Acoustical Society of America 94(4) pp 2240-2246

[7] Valdez SHD and Soutis C (2002) ldquoReal-time nondestructive evaluation of fiber com-posite laminates using low-frequency Lamb wavesrdquo Journal of Acoustical Society of Amer-ica 111(5) pp 2026-2033

[8] Yang M and Qiao P (2005) ldquoModeling and experimental detection of damage in virous materials using the pulse-echo method and piezoelectric sensoractuatorsrdquo Smart Materials and Structures 14 pp1083-1100

[9] Lestari W and Qiao P (2005) ldquoApplication of wave propagation analysis for damage identification in composite laminated beamsrdquo Journal of Composite Materials 39(22) pp1967-1984

[10] Kessler S S Spearing S M and Soutis C (2002) ldquoDamage detection in composite ma-terials using Lamb wave methodsrdquo Smart Materials and Structures 11 pp269-278

[11] Sohn H Park G Wail J R Limback N P and Farrar C R (2004) ldquoWavelet-based active sensing for delamination detection in composite structuresrdquo Smart Materials and Structures 13 pp 153-160

[12] Petculescu G Krishnaswamy S and Achenback J D (2008) ldquoGroup delay measure-ments using modally selective Lamb wave transducers for detection and sizing of delamina-tions in compositesrdquo Smart Materials and Structures 17

[13] Sohn H Park H W Law K H and Farrar C R (2007) ldquoDamage detection in compo-

- 33 -

site plates by using an enhanced time reversal methodrdquo Journal of Aerospace Engineering (ASCE) pp 141-151

[14] Sohn H (2007) ldquoEffects of environmental and operational variability on structural health monitoringrdquo Philosophical Transactions of the Royal Society Ardquo 365 pp539-560

[15] Rytter A (1993) ldquoVibration based inspection of civil engineering structuresrdquo PhD The-sis Department of Building Technology and Structural Engineering University of Aalborg Aalborg Denmark

[16] Hayashi T and Kawashima K (2002) ldquoMultiple reflections of Lamb waves at a delami-nationrdquo Ultrasonics 40 pp193-197

[17] Wang C H and Rose L R F (2003) ldquoWave reflection and transmission in beams con-taining delamination and inhomogeneityrdquo Journal of Sound and Vibration 264 pp 851-872

[18] Ramadas C Balasubramaniam K Joshi M and Krishnamurthy C V (2009) ldquoInterac-tion of the primary anti-symmetric Lamb mode (Ao) with symmetric delaminations numer-ical and experimental studiesrdquo Smart Materials and Structures 18

[19] Ramadas C Balasubramaniam K Joshi M and Krishnamurthy C V (2010) ldquoInterac-tion of guided Lamb waves with an asymmetrically located delamination in a laminated composite platerdquo Smart Materials and Structures 19

[20] Wilcox P Lowe M and Cawely P (2001) ldquoThe effect of dispersion on long-range in-spection using ultrasonic guided wavesrdquo NDTampE International 34 pp1-9

[21] Mallat S G and Zhang Z (1993) ldquoMatching pursuits with time-frequency dictionariesrdquo IEEE Tansactions on Signal Processing 41(12) pp3397-3415

[22] Zhang G Zhang S and Wang Y (2000) ldquoApplication of adaptive time-frequency de-composition in ultrasonic NDE of highly scattering materialsrdquo Ultrasonics 38 pp 961-964

[23] Hong J C Sun K H and Kim Y Y (2005) ldquoThe matching pursuit approach based on the modulated Gaussian pulse for efficient guided-wave damage inspectionrdquo Smart Mate-rials and Structures 14 pp548-560

[24] Herakovich C T (1997) Mechanics of Fibrous Composites Wiley 480 pages

[25] Reddy J N (1985) ldquoA review of the literature on finite-element modeling of laminated composite platesrdquo The Shock and Vibration Digest 17(4) pp 3-8

[26] Reddy J N (2003) Mechanics of Laminated Composite Plates and Shells Theory and Analysis CRC Press 856 pages

[27] Palacz M Krawczuk M and Ostachowicz W (2005) ldquoThe spectral finite element mod-el for analysis of flexure-shear coupled wave propagation Part 1 laminated multilayer

- 34 -

composite beamrdquo Composite Structures 68 pp37-44

[28] Zhang Y C and Harding J (1990) ldquoA numerical micromechanics analysis of the me-chanical properties of a plain weave compositerdquo Computer amp Structures 36(5) pp 839-844

[29] Nemat-Nasser S and Hori M (1999) Micromechanics Overall Properties of Heteroge-neous Materials Elsevier 810 pages

[30] Scida D Aboura Z Benzeggagh M L and Bocherens E (1999) ldquoA micromechanics model for 3D elasticity and failure of woven-fibre composite materialrdquo Composites Science and Technology 59 pp 505-517

[31] Thomas Y and Wu X H (1997) ldquoA multiscale finite element method for elliptic prob-lems in composite materials and porous mediardquo Journal of Computational Physics 134 pp 169-189

[32] Lee K H Moorthy S and Ghosh S (1999) ldquoMultiple scale computational model for damage in composite materialsrdquo Computer Methods in Applied Mechanics and Engineer-ing 172 pp 175-201

[33] Namilae S and Chandra N (2005) ldquoMultiscale model to study the effect of interfaces in carbon nanotube-based compositesrdquo Journal of Engineering Materials and Technology (ASME) 127 pp222-232

[34] Adams D E (2007) Health monitoring of Structural Materials and Components Wiley 460 pages

[35] Gopalakrishnan S (2009) ldquoModeling Aspects in Finite Elementrdquo Encyclopedia of Struc-tural Health Monitoring 2 pp791-809

[36] Palacz M Krawczuk M and Ostachowicz W (2005) ldquoThe spectral finite element mod-el for analysis of flexure-shear coupled wave propagation Part 2 Deminated multilayer composite beamrdquo Composite Structures 68 pp45-51

[37] Baik J M and Thopson R B (1984) ldquoUltrasonic scattering from imperfect interfaces A quasi-static modelrdquo Journal of Nondestructive Evaluation 4(3) pp177-196

[38] Luo H and Hanagud S (2000) ldquoDynamics of delamincated beamsrdquo International Jour-nal of Solids and Structures 37 pp1501-1519

[39] Kim S B and Sohn H (2007) ldquoInstantaneous reference-free crack detection based on polarization characteristics of piezoelectric materialsrdquo Smart Materials and Structures 16 pp2375-2387

[40] An Y K and Sohn H (2010) ldquoInstantaneous crack detection under varying temperature and static loading conditionsrdquo The Special Issue of Structural Control and Health Monitor-ing 17(7) pp 730-741

- 35 -

[41] Lee S J and Sohn H (2006) ldquoActive self-sensing scheme development for structural health monitoringrdquo Smart Materials and Structures 15(6) pp 1734-1746

- 36 -

요 약 문

복합재료 손상의 무기저 진단 및 정량화

복합재료는 재료의 경량성 내구성 고강도의 특성으로 인하여 항공기에서부터

교량에까지 널리 사용되고 있다 하지만 외부의 충격으로 인한 복합재료 내부의

손상은 재료의 강도를 감소시켜 구조물의 안정성을 저하시키는 요인이 된다 하지만

복합재료의 손상은 재료 내부에서 발생하여 육안으로는 식별하기도 힘들며 표면에서

보이는 것만으로 내부의 손상 정도를 판단하기는 불가능하다는 단점을 가지고 있다

본 연구에서는 손상 내부에서 반사로 인해 발생하는 파를 이용하여 기저자료를

사용하지 않고 특정 시점에서 얻은 자료만으로 복합재료의 손상을 진단하고 정량화

할 수 있는 기법을 제안한다 제안된 기법의 적용성을 검증하기 위하여 유한요소법을

이용한 수치해석을 수행하였고 Teflon tape 이 삽입된 시편과 실제 충격하중으로

인한 손상을 입은 시편을 이용한 실험을 진행하였다 수치해석 및 실험 결과 제안된

기법이 복합재료의 손상을 성공적으로 진단 및 정량화할 수 있음을 확인하였다 또한

다양한 외부 온도 환경에서 제안된 기법의 환경영향 실험을 수행하였고 다양한

환경에서 손상의 진단 및 정량화가 성공적으로 이루어짐을 검증하였다

핵심어 무기저 손상진단 및 정량화 압전센서 복합재료 유한요소법 Matching pursuit

method

- 37 -

감 사 의 글

기대와 두려움을 안고 시작한 석사과정이 어느덧 2 년 이라는 세월 속에서 추억으로 자리잡아

가고 있습니다 학부생 기간을 포함한 지난 3 년 동안 SHM 이라는 새로운 분야를 접하고 연구를

수행 할 수 있도록 도와주신 고마운 분들이 있어 그 감사를 표하고자 합니다

우선 저의 많은 부족한 점을 일깨워 주시고 연구에 대한 뜻을 더욱 확고히 세울 수 있게 해

주신 지도교수님이신 손 훈 교수님께 깊은 감사의 뜻을 전합니다 짧은 기간이었지만 교수님의

도움으로 많은 것을 배우고 느끼고 깨달을 수 있었습니다 또한 본 연구를 수행할 수 있도록 지

원해주신 국토해양부 (U-City 석 박사과정 지원사업)에 깊은 감사 드립니다

때로 힘들고 지칠 때 항상 저를 믿고 응원해 주시는 가족이 있었기에 지금의 글을 쓸 수 있었

습니다 또한 지금 연구실에서 함께 생활하고 있는 윤규형 민구형 철민이형 현석이형 진열이형

호민이형 병진이 선혜누나와 성대에 계신 창길이형 미국에 계신 의재형과 Debaditiya 에게도 감

사의 말씀을 전합니다 그리고 모두 나열할 수는 없지만 카이스트 생활의 활력소가 되어준 카이

스트 건설 및 환경공학과 선후배들 직원 분들과 SOS 멤버들에게도 감사를 표합니다

앞서 거론한 많은 이들의 도움이 없었다면 저 또한 지금 이 자리에 있을 수 없었을 것입니다

많이 부족하기에 더 노력하고 배우고자 했던 초심을 다시 한번 생각하며 더 큰 결실을 맺기 위해

앞으로 나아가겠습니다 도움을 주신 모든 분들께 이 감사의 글을 바칩니다

2011년 6월 카이스트에서 임형진 올림

- 38 -

CURRICULUM VITAE

Hyung Jin Lim

Research Assistant

The Civil and Environmental Engineering Department

Korea Advanced Institute of Science and Technology (KAIST)

291 Daehak-Ro Guseong-Dong Yuseong-Gu

Daejeon Republic of Korea 305-701

Phone (82)+42-350-3665 Fax (82)+42-350-3610

Email limhj87gmailcom

RESEARCH INTERESTS

Damage Detection Localization and Quantification on Composite and Metallic Structures Pattern Recognition Finite Element Modeling (FEM)

EDUCATION

2009 ndash 2011 MS Civil and Environmental Eng Korea Advanced Institute of Science and Technology (KAIST) Korea Rep

2006 ndash 2009 BS Civil and Environmental Eng Korea Advanced Institute of Science and Technology (KAIST) Korea Rep

JOURNAL PUBLICATIONS The corresponding author is underlined 1 Hyung Jin Lim Hoon Sohn ldquoReference-free damage detection localization and quantification

using in compositerdquo in preparation for Journal of the Acoustical Society of America 2011

2 Chul Min Yeum Hoon Sohn Jeong Bum Ihn Hyung Jin Lim ldquoInstantaneous delamination detection in a composite using a dual piezoelectric transducer networkrdquo Submitted to Compo-sites Science and Technology 2011

3 Yun-Kyu An Hyung Jin Lim Min Koo Kim Jin Yeol Yang Hoon Sohn Chang Guen Lee ldquoInvestigation of applicability of local reference-free damage detection techniques to in-situ

- 39 -

bridgesrdquo Submitted to ASCE Journal of Structural Engineering 2011

4 Hyung Jin Lim Min Koo Kim Hoon Sohn Chan Yik Park ldquoImpedance based damage detec-tion under varying temperature and loading conditionsrdquo Submitted to NDTampE International 2011

5 Hyung Jin Lim Hoon Sohn ldquoQuantitative estimation of transmitted and reflected lamb waves at discontinuityrdquo Journal of the Korean Society for Nondestructive Testing Vol30 No 4 pp1-8 2010

CONFERENCE PROCEEDINGS

The corresponding author is underlined 1 Hyung Jin Lim and Hoon Sohn ldquoReference-free damage detection localization and quantifi-

cation in compositesrdquo ASCE Engineering Mechanics Institute Conference Boston June 2-4 2011

2 Hyung Jin Lim and Hoon Sohn ldquoReference-free delamination quantification within multi-layer composite materialsrdquo KSNT Spring Conference May 26-27 Gyeong-Ju Korea 2011

3 Min Koo Kim Hyun Jin Lim Hoon Sohn Chan Yik Park ldquoImpedance-based bolt loosening detection under varying temperature and loadingrdquo Asian Pacific Workshop on Structural Health Monitoring Tokyo Japan November 30-December 2 2010

4 Hyung Jin Lim Chul Min Yeum and Hoon Sohn ldquoModeling of Impact-induced Delamination in a Multilayer Composite Platerdquo The 23rd KKCNN Symposium on Civil Engineering Taipei (Taiwan) November 13-15 2010

5 Yun-Kyu An Hyung Jin Lim and Hoon Sohn ldquoTemperature independent Delamination Detec-tion using data normalizationrdquo The 3rd International Conference on Multi-Functional Materi-als and Structures Chonbuk National Univ Jeonju Korea September 14-18 2010

6 Hoon Sohn Yun-Kyu An Chang Gil Lee Min Koo Kim Hyung Jin Lim ldquoApplication of online NDT techniques to decommissioned bridge testingrdquo the 5th European Workshop on Structural Health Monitoring Sorrento Italy June 29-July 02 2010

7 Min Koo Kim Hyung Jin Lim Hoon Sohn Chan Yik Park ldquoImpedance based damage diag-nosis of a complex composite aircraft wing under changing loading and temperature condi-tionsrdquo the 5th European Workshop on Structural Health Monitoring Sorrento Italy June 29-July 02 2010

8 Yun-Kyu An Min Koo Kim Hyung Jin Lim and Hoon Sohn ldquoValidations of reference-free crack detection techniques through a decommissioned bridge testrdquo 2010 COSEIK Annual Con-ference Jeju Korea April 8-9 2010

- 40 -

PARTICIPANT PROJECTS

1 Development of Smart Steel Members for Online Steel Structure Monitoring POSCO Corporation (Funded 41000000 KRW (41000 USD) for 010511 to 123111)

2 Development of Enabling Technologies for Structural Health Monitoring Korean Agency for Defense Development (Funded 240000000 KRW (240000 USD) out of total 3200000000 KRW (3200000 USD) for 060108-053111)

3 Development of On-board SHM Technologies for Composite Air Vehicles The Boeing Company (Funded 400000 USD (400000000 KRW) for 080108 to 073111)

EDUCATIONAL EXPERIENCES

Fall 2009

Teaching Assistant CE207 ldquoElementary Structural Engineering and Laboratoryrdquo KAIST 3 unit undergraduate course

EXPERIMENTAL EXPERIENTCES

2010 Monitoring of Real Aircraft Composite Wing Segment Korea Agency for Defense De-velopment (ADD) Daejeon Korea Real time monitoring of composite air vehicles fund-ed by Korea Agency for Defense Development (ADD)

2009-2011

Field Monitoring of Real Bridge Structures Gimpo (Ramp G bridge) amp Icheon (Sam-seung bridge) Korea US-Korea collaborative research for bridge monitoring testbeds

HONORS amp AWARDS

Scholarships for outstanding students granted KAIST (2007-2009) The three prominent students in the civil and environmental engineering department of KAIST were selected as recipients This scholarship provided 2000$ per year for three years

- 41 -

Page 18: 복합재료 손상의 무기저 진단 및 정량화 - KAISTssslab.kaist.ac.kr/article/pdf/Thesis_2.pdf · 1 Declaration of Ethical Conduct in Research: I, as a graduate student

CHAPTER 3 THEORETICAL DEVELOPMENT

31 Overview

Damage

l = l1 + d + l2

d = vdtd l1 = vit1 l2 = vit2

t = t1 + td + t2

Excitation and Sensing PZT Sensing PZT

l

Transmitted Wave

Reflected Waves

Incident Wave

Fig 3 Parameters and notations of the proposed damage characterization technique Damage is detected by identifying the additional reflected wave from the damage due to multiple re-flections (dashed lines) Localization and quantification of the damage is carried out by measuring the arrival time of transmitted and reflected waves from the damage

In this section the reference-free damage characterization algorithm is proposed Damage

diagnosis and quantification are achieved by identifying the additional reflected waves due to

multiple reflections in the damage area (dashed lines in Fig 3) Moreover damage is local-

ized using additional information from pulse-echo method implement For the localization

and the quantification of damage the variables such as the path length and A0 Lamb wave

mode velocity in damage area are defined (Fig 3) For example 119897119897 is expressed as

119897119897 = 1198971198971 + 119889119889 + 1198971198972 (1)

where 1198971198971 119889119889 and 1198971198972 are the distance from the excitation PZTs to the entrance of damage the

damage length and the distance from the exit of damage to the sensing PZT respectively

- 7 -

Moreover 119907119907119894119894 and 119907119907119889119889 are also defined as the A0 Lamb wave mode velocity in pristine area

and damage area respectively Note that the whole procedure of the damage characterization

is accomplished based solely on the signals obtained from the current state of the structure

32 Damage index (m-value)

The proposed reference-free damage characterization technique utilizes the waves reflect-

ed additionally from the damage area as a feature The arrival time difference of the transmit-

ted and reflected A0 Lamb wave modes reflected from the exit and the entrance of the delam-

ination area one times each (∆119905119905) contains the information of the physical length of the dam-

age (119889119889) and A0 Lamb wave mode velocity in the damage area (119907119907119889119889) The wave velocity in the

damage area which is related to the shear modulus (G) represents the severity of the damage

∆119905119905 = 2119905119905119889119889 =2119889119889119907119907119889119889

(2)

However based on the measured data available the 119889119889 and 119907119907119889119889 (or G) cannot be uniquely

determined Fig 4 shows two FEM models which have different physical length and wave

velocity (G) of the damage Detail explanation of delamination modeling technique will be

covered in section 4 The similar signal in the time domain and ∆119905119905 are obtained from the

two different damage cases

- 8 -

Damage 1

Damage 2

20

Unit mm15

90 G value reduction

943 G value reduction

m-value = 248

m-value = 248

015 02 025 03 035

-2

-1

0

1

2

Time (ms)

Ampl

itude

Damage 1Damage 2

Fig 4 Two different FEM models which have different 119889119889 and 119907119907119889119889 (or G) This represents that the damage cannot be characterized uniquely based on the measured data available

Thus in this study the damage index (m-value) which can be uniquely defined based on

the available data is introduced The m-value is defined as the ratio of the physical length of

the damage and wavelength of the A0 Lamb wave mode in the damage area (120582120582119889119889)

119898119898 =119889119889120582120582119889119889

=119889119889119889119889119907119907119889119889

(3)

In many cases the transmitted and reflected waves are overlapped and it is hard to decom-

pose and calculate the arrival times of them Matching pursuit method is an alternative way

to accurately decompose of the overlapped waves and estimate the parameters of each wave

such as arrival time (Fig 5) Detail explanation of the matching pursuit method is covered in

section 36

015 02 025 03-2

-1

0

1

2

Time (ms)

Am

plitu

de

OriginalApproximated

(a)

015 02 025 03-2

-1

0

1

2

Time (ms)

Ampl

itude

Wave 1Wave 2

(b)

Fig 5 (a) Example of the overlapped signal with comparing approximated signal using the matching pursuit method (b) Decomposed signal using the matching pursuit method The matching pursuit method provides a conventional way to decompose and identify each wave

- 9 -

33 Damage detection algorithm (Level 1)

The damage diagnosis is accomplished by extracting the additional reflected A0 Lamb

wave modes from damage due to multiple reflections If A0 Lamb wave modes are decom-

posed perfectly it is trivial that one transmitted A0 Lamb wave mode is always captured in

the response signal If damage exists other waves in the response signal behind the transmit-

ted A0 Lamb wave mode are presented Since this approach does not rely on previously col-

lected data the diagnosis is independent from environmental variation such as temperature

changes

It is reported that the dispersion of Lamb waves cause wave-packets to spread out in space

and time as they propagate through a structure [20] Thus the reflected waves from the dam-

age have longer pulse duration than the transmitted wave due to their longer propagating dis-

tance In this study the matching pursuit method is modified to capture more spread wave

than the transmitted A0 Lamb wave mode Detail explanation of the modified matching pur-

suit method is described in latter section

Let 1199041199041 and 1199041199042 as the pulse duration (scale) of the transmitted and additional reflected A0

Lamb wave mode respectively

ldquoIf 1199041199042 gt 1199041199041 damage exists

Otherwise there is no damagerdquo

- 10 -

34 Damage quantification algorithm (Level 3)

As mentioned in the previous section damage is quantized by estimating 119889119889 and 119907119907119889119889 using

pitch-catch method The procedure to estimate 119889119889 and 119907119907119889119889 is

(i) Measure 119897119897 119905119905 and 119905119905119889119889

(ii) Assume the range of 119889119889 (0 lt 119889119889 lt 119897119897)

(iii) Calculate 119907119907119894119894 and 119907119907119889119889 with a constraints (0 lt 119907119907119889119889 lt 119907119907119894119894) (Eq 4 and 5)

(iv) Estimate m-value and find a feasible range of 119889119889 and 119907119907119889119889

119907119907119894119894 =119897119897 minus 119889119889119905119905 minus 119905119905119889119889

(4)

119907119907119889119889 =119889119889119905119905119889119889

(5)

35 Damage localization algorithm (Level 2amp3)

For damage localization 1198971198971 is estimated by measuring the arrival time from the excitation

PZT to the entrance of delamination (1199051199051) with additional pulse-echo method implementation

The procedure is similar with the previous damage quantification algorithm

(i) Measure 119897119897 119905119905 119905119905119889119889 and 1199051199051

(ii) Assume the range of 1198971198971 (0 lt 1198971198971 lt 119897119897)

(ii) Compute possible combinations of 1198971198972 119889119889 and 119907119907119889119889 with constraints

(0 lt 119907119907119889119889 lt 119907119907119894119894 119889119889 gt 0) (Eq 5 6 and 7)

- 11 -

1198971198972 = 1198971198971119905119905 minus 1199051199051 minus 119905119905119889119889

1199051199051 (6)

119889119889 = 119897119897 minus 1198971198971 minus 1198971198972 (7)

Since a PZT for pulse-echo method acts as both an actuator and a sensor the obtained sig-

nal contains both the excitation and the response signal Thus the reflected wave from the

entrance of damage can be buried into the tail of excitation signal and cause error on arrival

time of reflected wave from damage Even in this case damage can be localized and quan-

tized using the constraints When first wave packet reflected from the entrance of damage is

buried into the excitation signal the next wave packet reflected from the exit of damage can

be utilized alternatively If the algorithm regards the second reflected wave as the first re-

flected wave there are no combinations of estimated variable meeting the constraints Then

the algorithm repeats the process by regarding the wave as the second reflected wave the

constraints can select proper combination of the estimated variables

The estimated values have range bounded by the constraints When additional information

of 119907119907119894119894 is utilized the damage can be characterized more accurately In this study three meth-

ods are proposed and verified as information level of 119907119907119894119894

Method 1 Solely based on the current data measurement (Reference-free)

Method 2 Limit the possible range of 119907119907119894119894 (Prior knowledge of the system)

Method 3 Measure 119907119907119894119894 (Additional data need to be collected)

Method 1 described in this section can be utilized without any information of 119907119907119894119894 Meth-

od 2 produces more accurate estimation results using prior knowledge of the system such as

- 12 -

119907119907119894119894 variation under temperature changes Method 3 gives unique estimated value using meas-

ured 119907119907119894119894 from other path in the same time In section 4 and 5 damage is characterized using

these three methods and the results are compared each another

36 Matching pursuit method considering the dispersion effect of Lamb

wave

Matching pursuit method has been developed for signal compression approximation and

characterization of obtained signal and function [21-23] The key concept of the matching

pursuit method is that any signal (function) can be decomposed into a linear expansion of

signals selected from a pool of basis signals (functions) called lsquodictionaryrsquo

Let 119891119891 be a function to be analyzed then 119891119891 can be approximated a set of functions select-

ed from the dictionary 119863119863 in order to best match This is done by successive approximations

with orthogonal projections Let1198921198921205741205740 isin 119863119863 the function 119891119891 can be decomposed into

119891119891 = lang1198911198911198921198921205741205740rang1198921198921205741205740 + 119877119877119891119891 (8)

where 119877119877119891119891 is the residual function after approximating 119891119891 using 1198921198921205741205740 1198921198921205741205740 is orthogonal to

119877119877119891119891 hence

1198911198912 = lang1198911198911198921198921205741205740rang2

+ 1198771198771198911198912 (9)

To minimize 119877119877119891119891 1198921198921205741205740 is chosen to maximize lang1198911198911198921198921205741205740rang The matching pursuit method is

an iterative algorithm that decomposes the residue 119877119877119891119891 by projecting it on a function of 119863119863

that matches 119877119877119891119891 almost at best as it was done for 119891119891 The residue 119877119877119899119899119891119891 is decomposed into

119877119877119899119899119891119891 = lang119877119877119899119899119891119891119892119892120574120574119899119899rang119892119892120574120574119899119899 + 119877119877119899119899+1119891119891 (10)

- 13 -

the decomposition is carried up to the order 119898119898 119891119891 can be expressed as

119891119891 = lang119877119877119899119899119891119891119892119892120574120574119899119899rang119892119892120574120574119899119899

119898119898minus1

119899119899=0

+ 119877119877119898119898119891119891 (11)

The original function 119891119891 is decomposed into a sum of dictionary elements that are chosen to

best match its residues

In this study the matching pursuit method is utilized to accurately decompose of the over-

lapped Gaussian pulses and estimate the parameters of the pulse which used for damage

characterization The 119894119894119905119905ℎ Gaussian pulse in the dictionary for approximating the obtained

signal is expressed as

119892119892120574120574119894119894 = 119891119891(119860119860119894119894 119906119906119894119894 119904119904119894119894 119889119889119894119894120601120601119894119894) (12)

where 119860119860119894119894119906119906119894119894119904119904119894119894119889119889119894119894 and 120601120601119894119894 represent amplitude arrival time scale (pulse duration) excita-

tion(central) frequency and phase of the Gaussian pulse respectively The parameters of the

pulse are estimated using a nonlinear regression based optimization method (Gauss-Newton

method) based on an initial guess of the parameters Detail procedure for estimating the pa-

rameters of the pulse is expressed in Hong et al [23]

Among the estimated parameters of the pulses pulse duration (scale) represents the disper-

sion effect If a wave-packet is spread out the estimated pulse duration is increased Using

this feature the matching pursuit method is modified to capture waves which have longer

pulse duration than transmitted A0 Lamb wave mode If the estimate pulse duration of a wave

is less than the transmitted wave the modified matching pursuit method regards it as error

- 14 -

CHAPTER 4 NUMERICAL SIMULATION

41 Composite modeling method

Composite materials can be modeled from nano to macro level depends on its engineering

purpose In general four methods are utilized for modeling of composite materials First

composite is modeled as a homogenous material expressed in terms of effective material

properties [24] The effective material properties are estimated using the modulus and the

volume fraction of each constituent theoretically (Rule of mixture) and measured experimen-

tally Second a model is established that each laminate which have own material properties

are totally bonded together In this model the stacking sequence and direction of each lami-

nates are considered additionally [25-27] Third method for composite modeling in micro

level called micromechanics [28-30] This model considers the interaction between fibers and

matrix in laminate Last multiscale modeling which the scope of modeling is extended to

molecular or quantum level of constituents [31-33] This model starts from molecular me-

chanics or molecular dynamics then the scale of model leads to the nano micro and macro

level by incorporating small scale models As the scale of models is smaller it provides more

logical way than other large scale of models while the computational cost is more expensive

The scope of this study is focused on the dynamic behavior of composite structure the beam

or plate model was utilized for the numerical simulation

- 15 -

42 Damage modeling method

Generally three damage models have been developed and utilized in the real applications

First the stiffness reduction model called as the equivalent element model was introduced

[17 34-35] This model is simply realized by replacing the region around damage area by a

region with adjusted material properties (Youngrsquos modulus shear modulus density etc)

globally Next the duplicated node model is done by disconnecting the element connectivity

assuming there is no interaction among the delaminated surface This model can be utilized

when the location and the extension of damage is important issue [34-36] Last the contact

mechanics model using the principles of contact mechanics the two interfacial forces (con-

tact resistance and sliding resistance) in the delaminated surface [37-38] In this model the

delaminated surfaces are modeled as nonlinear functions of the interlaminar normal strain

and the interlaminar shear strain This model is computationally expensive while provides

more logical way than other models There is however no universally accepted model for

general impact-induced damage in composites In the case of impact-induced damage it is

impossible to predict or estimate the location and number of delaminated layers Thus from

the uncertainty of impact-induced damage the equivalent element model can be considered

as the effective damage model

- 16 -

43 Numerical simulation description

TABLE I Material property of simulation model and test specimen

Tensile Strength 657 MPa Tensile Modulus 59 GPa Shear Strength 804 MPa Shear Modulus 24 Gpa Specific Gravity 16

PZT A

PZT C

PZT BDamage

3

250 125 125

500 Unit mm

Fig 6 Description of the numerical simulation model

The feasibility of the proposed damage characterization technique is verified by 2-D FEM

simulation using MSC PATRAN and NASTRAN Fig 6 presents the dimension PZTs and

damage location of the simulation model In the model 05mm x 05mm square shell element

was used for modeling of the A0 mode whose flexural wave characteristic The element size

was determined by considering the wavelength of the propagation Lamb wave modes The

material properties of the simulation model and damage cases are presented in Table I and

Table II respectively

PZT A and C act as both actuator and sensor for collecting pulse-echo signal They were

installed on the plate which two of them are both-side collocated for the A0 Lamb wave

- 17 -

mode decomposition [39-40] PZT B was only used as a sensor for collecting pitch-catch sig-

nal For simulating the excitation and sensing of PZTs the pin-force model was adopted that

an input force apples to the end nodes of excitation PZT and averaging element strains corre-

sponding sensing area A Gaussian pulse (tone-burst) 50 kHz central frequency was excited

as an input signal

TABLE II Damage cases for numerical simulation

119949119949120783120783 (mm) 119941119941 (mm) G value reduction ()

Case 1 650 200 90 darr

Case 2 675 150 90 darr

Case 3 675 150 93 darr

Case 4 900 150 90 darr

- 18 -

44 Numerical simulation results

015 02 025 03 035

-2

0

2

Time (ms)

Am

plitu

de

IntactDamage

Reflected wave from damage

(a) Pitch-catch signal

015 02 025 03 035

-2

0

2

Time (ms)

Am

plitu

de

IntactDamage

Reflected waves from damage

(b) Pulse-echo signal Fig 7 Comparison of A0 Lamb wave mode signals from numerical simulation between intact and damage cases Time delay attenuation and multiple reflections due to damage simulated successfully using the damage modeling method described in the previous section

Fig 7 shows that comparison of A0 Lamb wave mode signals between intact and one of

damage cases in the simulation The features of A0 Lamb wave mode passing through dam-

age area described in section 2 are successfully simulated Thus the equivalent element

method is proper for modeling damage in composites In the damage cases the additional

waves reflected from damage is observed in both pitch-catch and pulse echo signal

Table III presents the result of the proposed damage characterization algorithm using the

numerical simulation cases shown in Table II In lsquoMethod 2rsquo 119907119907119894119894 variation is assumed to plusmn

5 of the normal value (1190~1315ms) The normal value of 119907119907119894119894 for lsquoMethod 3rsquo is meas-

ured as 1253 ms In each case the damage is characterized using the proposed technique

Note that as the information of 119907119907119894119894 is specified the estimated value is closer to the exact

value

- 19 -

TABLE III Damage characterization for the numerical simulation models

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

Case 1

Exact 650 200

Method 1 248 50~70 79~414 160~834

Method 2 248 603~667 135~241 271~486

Method 3 248 635 188 379

Case 2

Exact 675 150

Method 1 193 60~75 01~251 37~651

Method 2 193 618~682 114~222 295~575

Method 3 193 65 168 435

Case 3

Exact 675 150

Method 1 233 60~75 66~302 140~649

Method 2 233 668~738 85~196 182~419

Method 3 233 703 140 301

Case 4

Exact 900 150

Method 1 204 75~100 48~349 118~856

Method 2 204 860~951 107~216 264~530

Method 3 204 906 162 397

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1253 ms 1190~1315 ms Measured 119907119907119894119894 1253 ms

- 20 -

CHAPTER 5 EXPERIMENTAL VERIFICATION

51 Overview

In this section experimental study is investigated so that the proposed damage characteri-

zation technique Overall experimental setup and test specimen are shown in Fig 8 and 9

The experimental setup consists of test specimens data acquisition (DAQ) system and self-

sensing circuit Self-sensing circuit was installed for pulse-echo measurement using PZTs

[40-41]

PZT A(PZT C) PZT B

500

500

140

75

250

Teflon Inserted Damage (d =25)

Unit mm

Thickness = 3

PZT A(PZT C) PZT B

500

500

150

90

250

Impact Location

Unit mm

Thickness = 3

(a) Teflon inserted damage specimen

(b) Impact-induced damage specimen

Fig 8 Test specimens for the experimental verification In the Teflon inserted damage Teflon tapes with 25mm diameter were inserted at the center of the specimen The impact-induced damage was introduced to the specimen using the impact test equipment

- 21 -

511 Specimen

Two carbon composite plates with a dimension of 500times500times3mm3 were manufactured by

Nexcomes Inc by stacking 12 layers of CF3327 3K prepreg (woven fabrics type) manufac-

tured by Hankuk Fiber Inc The material properties of the specimen are presented in Table I

One of the specimens Teflon tapes with 25mm diameter was inserted in every 3 layers at the

center of the specimen to mimic the damage The other specimen is utilized for the character-

ization of real impact-induced damage

512 Data acquisition (DAQ) system and PZT

The data acquisition (DAQ) system consists of an arbitrary waveform generator (AWG) a

high speed signal digitizer (DIG) and two multiplexers (MUX) Using the 16-bits AWG

tone-burst input signal with a plusmn10 peak-to-peak voltage were generated The response signal

was measured by the 14-bit DIG For the signal-to-noise ratio reduction both pitch-catch and

pulse-echo signals were measured 150 times and averaged in the time domain 3 identical

circular PZTs of 10mm diameter manufactured by APC International Ltd were installed on

the specimen A Gaussian pulse (tone-burst) 50 kHz central frequency was excited as an in-

put signal The excitation frequency was chosen in order to avoid overlapping of the S0 and

A0 Lamb wave modes and generating of the higher modes (S1 A1hellip)

- 22 -

PZT Actuator

PZT Sensor

AWG

DIG

DAQ System Specimen

SensingActuatingMUX

1

23

4

1 Sending signal to MUX

4 Sending signal to DIG

Self-SensingCircuit

18 Unit mm

10

(a)

(b)

Fig 9 Overall experimental setup (a) Data acquisition (DAQ) system and (b) Installed circu-lar PZT transducer on the specimen with its dimension

513 Impact test

The impact-induced damage was introduced to the specimen using the impact test equip-

ment presented in Fig 10 A 5kg mass with 10mm tip was dropped from 015m height onto

the same location of the specimen The signals were obtained at each number of impacts

Since the reflected wave is not detected at the first impact case the data obtained from the

second and third impact were utilized for the analysis

10

Unit mm (a)

(b)

Fig 10 (a) Impact test equipment for introducing impact-induced damage and (b) Tip of the impact tester with 10mm diameter

- 23 -

514 Temperature effect test

In reality structures are subjected to changing environmental conditions such as tempera-

ture that affect measured signal and cause false-alarm Thus the robustness of the proposed

damage characterization technique under temperature variation in investigated (Fig 11) The

specimens were placed inside a temperature chamber and a thermocouple was installed on

the specimen to measure temperature of surface of the specimen The signal was obtained

under three temperature cases (0 20 and 50degC) and the humidity was kept at 30

Temperature Chamber

Test Specimen

Fig 11 Test specimen in temperature chamber

- 24 -

52 Teflon inserted damage

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Reflected wave from damage

(a)

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Transmitted WaveReflected Wave

Detected reflected wave

(b)

Fig 12 A0 Lamb wave mode obtained using pitch-catch method from the Teflon inserted damage specimen (a) The reflected wave from damage is shown (b) The Teflon inserted damage is detected by capturing the reflected wave

Fig 12 shows the obtained pitch-catch signal from the Teflon inserted damage specimen

and damage diagnosis result The Teflon inserted damage is detected successfully by captur-

ing the reflected wave Table IV presents the result of the Teflon inserted damage characteri-

zation 119907119907119894119894 is measured as 1213ms and the variation is assumed to plusmn 5 of the normal val-

ue (1152~1273ms) The results indicate that the proposed damage characterization approach

works well in real application

TABLE IV Damage characterization for the Teflon inserted damage specimen

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

Exact 635 25

Method 1 123 60~75 73~338 298~1220

Method 2 123 619~662 229~310 934~1188

Method 3 123 631 284 1161

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1213 ms

- 25 -

53 Impact-induced damage

11 115 12 125 13-4

-2

0

2

4

Time (ms)

Vol

tage

(mV

)

IntactImpact2Impact3

Reflected wave from damage

(a) Pitch-catch signal

11 115 12 125 13 135

-05

0

05

1

Time (ms)

Volta

ge (m

V)

IntactImpact2Impact3

Boundary reflection

Reflected wave from damage

(b) Pulse-echo signal

Fig 13 The signal obtained from the test specimen subjected to impact The impact was in-troduced at the same location of the specimen and the signal was obtained after each impacts As the number of impact is increased the transmitted A0 Lamb wave mode is delayed and attenuated

The obtained signals from impact-induced damage specimen are presented in Fig 13 It is

observed that as the number of impact is increased the transmitted A0 Lamb wave mode is

more delayed because the wave velocity in the damage area is reduced Fig 14 shows the

false-alarm test result whether the proposed damage detection algorithm works even in pris-

tine condition of the test specimen It is shown that no false alarm is occurred because no ad-

ditional reflected wave is captured

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Transmitted Wave

(a) Transmitted wave

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Reflected Wave

No reflected wave

(b) Reflected wave

Fig 14 False alarm test for the proposed damage detection algorithm in pristine condition of the specimen Reflected wave from damage is not captured false alarm is not occurred

- 26 -

TABLE V Damage characterization for the impact-induced damage specimen

of Impacts m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

2

(1470 J)

Method 1 080 65~90 75~246 470~1157

Method 2 080 801~885 114~245 711~1134

Method 3 080 843 179 1122

3

(2205 J)

Method 1 148 50~100 06~380 21~1134

Method 2 148 758~838 249~368 843~1125

Method 3 148 798 308 1044

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1213 ms

Table V presents the characterization results of the impact-induced damage case Here the

measured 119907119907119894119894 and the assumed variation are same as the artificial damage case It is observed

that as the number of impact is increased 119889119889 is increased while 119907119907119889119889 is reduced Thus it is

validated that the proposed damage characterization approach estimates the increment of the

damage well

54 Temperature effect test

541 Teflon inserted damage

11 115 12 125 13-10

-5

0

5

10

Time (ms)

Vol

tage

(mV

)

0 oC20 oC50 oC

(a) Pitch-catch signal

11 115 12 125 13 135

-1

0

1

2

3

Time(ms)

Volta

ge (m

V)

0 oC20 oC50 oC

(b) Pulse-echo signal

Fig 15 Signals obtained from the Teflon inserted damage specimen under temperature varia-tion Time delay and amplitude variation are occurred

Fig 15 shows the obtained signals from the Teflon inserted damage specimen subjected to

- 27 -

temperature variation Time delay and amplitude variations are occurred there is no addition-

al damage is introduced The analyzed results of the obtained signals are reported in Table VI

The measured 119907119907119894119894 is 1247 ms at 0 degC 1213 ms at 20 degC and 1168 ms at 50 degC respectively

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms (1152~1273 ms)

The estimated results are also changed under temperature variations the fluctuation of esti-

mated range and exact value is small

TABLE VI Temperature effect test for the Teflon inserted damage specimen

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

0 degC

Exact 635 25

Method 1 113 60~70 79~268 352~1291

Method 2 113 560~619 233~344 1034~1172

Method 3 113 606 257 1141

20 degC

Exact 635 25

Method 1 123 60~75 73~338 298~1220

Method 2 123 619~662 229~310 934~1188

Method 3 123 631 284 1161

50 degC

Exact 635 25

Method 1 137 60~75 72~328 263~1228

Method 2 137 622~687 183~299 666~1086

Method 3 137 642 264 960

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1247 ms at 0 degC 1213 ms at 20 degC and 1168 ms at 50 degC

- 28 -

542 Impact-induced damage

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Vol

tage

(mV

)

Transmitted WaveReflected Wave

No reflected wave

(a) 0 degC

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Transmitted WaveReflected Wave

No reflected wave

(b) 50 degC

Fig 16 False alarm test under temperature variations It is shown that proposed damage char-acterization algorithm is robust under temperature variations

The false alarm test in pristine condition of the impact-induced damage specimen under

temperature variation was investigated (Fig 16) It is validated that the proposed damage di-

agnosis algorithm is robust under temperature variations Fig 17 shows the obtained signals

from the impact-induced damage (impacted 3 times) specimen under temperature variation

The analyzed results of the obtained signals are reported in Table VII The measured 119907119907119894119894 in

each temperature and the assumed variation are same as the artificial damage case

11 115 12 125 13-5

0

5

Time (ms)

Volta

ge (m

V)

0 oC20 oC50 oC

(a) Pitch-catch signal

11 115 12 125 13 135

-1

0

1

2

3

Time(ms)

Vol

tage

(mV

)

0 oC20 oC50 oC

(b) Pulse-echo signal

Fig 17 Signals obtained from impact-induced damage specimen (Impact 3) under tempera-ture variation

- 29 -

TABLE VII Temperature effect test for the impact induced damage specimen (Impact 3 case)

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

0 degC

Method 1 135 75~90 72~310 267~1148

Method 2 135 705~779 264~381 978~1126

Method 3 135 763 289 1071

20 degC

Method 1 148 50~100 06~380 21~1134

Method 2 148 758~838 249~368 843~1125

Method 3 148 798 308 1044

50 degC

Method 1 153 75~100 53~415 174~1227

Method 2 153 804~889 214~336 702~1103

Method 3 153 815 320 1050

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1247 ms at 0 degC 1213 ms at 20 degC and 1168 ms at 50 degC

- 30 -

CHAPTER 6 CONCLUDING REMARKS

61 Executive summary

In this study the reference-free damage detection localization and quantification tech-

nique in a multi layered composite plate is proposed Additional reflected A0 Lamb wave

mode from the damage due to multiple reflections in damage area is used as a feature for

characterizing the damage First damage is characterized in terms of a damage index (m-

value) defined as the ratio of the physical length of the damage to the wavelength of the A0

Lamb wave mode Then the range of possible damage location length and severity are esti-

mated with constraints This procedure can be accomplished without prior baseline data To

realize the proposed technique the matching pursuit technique is modified for reduction of

the error by considering the dispersive characteristic of Lamb wave The matching pursuit

method estimates the arrival time of each wave packets accurately even two wave packets are

overlapped The numerical simulations and experimental tests including Teflon tape inserted

and impact-induced damage are investigated for the feasibility of the proposed technique

Moreover the effectiveness of the proposed technique under changing temperature is exper-

imentally verified

- 31 -

62 Suggestion for further study

The limitations of this study and the possible improvement and further studies are listed as

follow

1 Measurement of impact-induced damage length using conventional NDT technique

Additional experimental verification for impact-induced damage characterization is

needed by measuring the real length of damage using conventional NDT technique

such as ultrasonic scan or infrared camera

2 Limitation of both side collocated PZT installation In this study excitation PZTs are

installed both side and collocated for decomposing A0 Lamb wave mode However it

can produce decomposition error due to imperfection of PZT location and their installa-

tion Moreover in many cases assess to both surface of structures is not allowed such

as aircraft applications To improve this limitation a reference-free damage characteri-

zation using single side PZTs will be considered

3 Complex structure application In this study damage is characterized in simple plate-

like structure When the monitored structure becomes complex the response signals al-

so becomes complex due to its geometry and boundary Thus the investigations for

complex structure application such as stiffener spar and curvature are needed

- 32 -

REFERENCES

[1] Cawley P and Alleyne D (1996) ldquoThe use of Lamb waves for the long range inspection of large structuresrdquo Ultrasonics 34 pp287-290

[2] Ghosh T Kundu T and Karpur P (1998) ldquoEfficient use of Lamb waves for detecting defects in large platesrdquo Ultrasonics 36 pp791-801

[3] Guirguitui V (2003) ldquoLamb wave generation with piezoelectric wafer active sensors for structural health monitoringrdquo SPIE Conference on Smart Structures and Materials San Diego CA 5391 pp111-422

[4] Ragvahan A and Cesnik CE S (2004) ldquoModeling of piezoelectric-based Lamb wave generation and sensing for structural health monitoringrdquo SPIE Conference on Smart Struc-tures and Materials San Diego CA 5391 pp419-430

[5] Rose JL Pilarski A and Dirit JJ (1993) ldquoAn approach to guided wave mode selection for inspection of laminated platerdquo Journal of Reinforced Plastics and Composites 12(5) pp536-544

[6] Guo N and Cawley P (1993) ldquoThe interaction of Lamb waves with delaminations in composite laminatesrdquo Journal of Acoustical Society of America 94(4) pp 2240-2246

[7] Valdez SHD and Soutis C (2002) ldquoReal-time nondestructive evaluation of fiber com-posite laminates using low-frequency Lamb wavesrdquo Journal of Acoustical Society of Amer-ica 111(5) pp 2026-2033

[8] Yang M and Qiao P (2005) ldquoModeling and experimental detection of damage in virous materials using the pulse-echo method and piezoelectric sensoractuatorsrdquo Smart Materials and Structures 14 pp1083-1100

[9] Lestari W and Qiao P (2005) ldquoApplication of wave propagation analysis for damage identification in composite laminated beamsrdquo Journal of Composite Materials 39(22) pp1967-1984

[10] Kessler S S Spearing S M and Soutis C (2002) ldquoDamage detection in composite ma-terials using Lamb wave methodsrdquo Smart Materials and Structures 11 pp269-278

[11] Sohn H Park G Wail J R Limback N P and Farrar C R (2004) ldquoWavelet-based active sensing for delamination detection in composite structuresrdquo Smart Materials and Structures 13 pp 153-160

[12] Petculescu G Krishnaswamy S and Achenback J D (2008) ldquoGroup delay measure-ments using modally selective Lamb wave transducers for detection and sizing of delamina-tions in compositesrdquo Smart Materials and Structures 17

[13] Sohn H Park H W Law K H and Farrar C R (2007) ldquoDamage detection in compo-

- 33 -

site plates by using an enhanced time reversal methodrdquo Journal of Aerospace Engineering (ASCE) pp 141-151

[14] Sohn H (2007) ldquoEffects of environmental and operational variability on structural health monitoringrdquo Philosophical Transactions of the Royal Society Ardquo 365 pp539-560

[15] Rytter A (1993) ldquoVibration based inspection of civil engineering structuresrdquo PhD The-sis Department of Building Technology and Structural Engineering University of Aalborg Aalborg Denmark

[16] Hayashi T and Kawashima K (2002) ldquoMultiple reflections of Lamb waves at a delami-nationrdquo Ultrasonics 40 pp193-197

[17] Wang C H and Rose L R F (2003) ldquoWave reflection and transmission in beams con-taining delamination and inhomogeneityrdquo Journal of Sound and Vibration 264 pp 851-872

[18] Ramadas C Balasubramaniam K Joshi M and Krishnamurthy C V (2009) ldquoInterac-tion of the primary anti-symmetric Lamb mode (Ao) with symmetric delaminations numer-ical and experimental studiesrdquo Smart Materials and Structures 18

[19] Ramadas C Balasubramaniam K Joshi M and Krishnamurthy C V (2010) ldquoInterac-tion of guided Lamb waves with an asymmetrically located delamination in a laminated composite platerdquo Smart Materials and Structures 19

[20] Wilcox P Lowe M and Cawely P (2001) ldquoThe effect of dispersion on long-range in-spection using ultrasonic guided wavesrdquo NDTampE International 34 pp1-9

[21] Mallat S G and Zhang Z (1993) ldquoMatching pursuits with time-frequency dictionariesrdquo IEEE Tansactions on Signal Processing 41(12) pp3397-3415

[22] Zhang G Zhang S and Wang Y (2000) ldquoApplication of adaptive time-frequency de-composition in ultrasonic NDE of highly scattering materialsrdquo Ultrasonics 38 pp 961-964

[23] Hong J C Sun K H and Kim Y Y (2005) ldquoThe matching pursuit approach based on the modulated Gaussian pulse for efficient guided-wave damage inspectionrdquo Smart Mate-rials and Structures 14 pp548-560

[24] Herakovich C T (1997) Mechanics of Fibrous Composites Wiley 480 pages

[25] Reddy J N (1985) ldquoA review of the literature on finite-element modeling of laminated composite platesrdquo The Shock and Vibration Digest 17(4) pp 3-8

[26] Reddy J N (2003) Mechanics of Laminated Composite Plates and Shells Theory and Analysis CRC Press 856 pages

[27] Palacz M Krawczuk M and Ostachowicz W (2005) ldquoThe spectral finite element mod-el for analysis of flexure-shear coupled wave propagation Part 1 laminated multilayer

- 34 -

composite beamrdquo Composite Structures 68 pp37-44

[28] Zhang Y C and Harding J (1990) ldquoA numerical micromechanics analysis of the me-chanical properties of a plain weave compositerdquo Computer amp Structures 36(5) pp 839-844

[29] Nemat-Nasser S and Hori M (1999) Micromechanics Overall Properties of Heteroge-neous Materials Elsevier 810 pages

[30] Scida D Aboura Z Benzeggagh M L and Bocherens E (1999) ldquoA micromechanics model for 3D elasticity and failure of woven-fibre composite materialrdquo Composites Science and Technology 59 pp 505-517

[31] Thomas Y and Wu X H (1997) ldquoA multiscale finite element method for elliptic prob-lems in composite materials and porous mediardquo Journal of Computational Physics 134 pp 169-189

[32] Lee K H Moorthy S and Ghosh S (1999) ldquoMultiple scale computational model for damage in composite materialsrdquo Computer Methods in Applied Mechanics and Engineer-ing 172 pp 175-201

[33] Namilae S and Chandra N (2005) ldquoMultiscale model to study the effect of interfaces in carbon nanotube-based compositesrdquo Journal of Engineering Materials and Technology (ASME) 127 pp222-232

[34] Adams D E (2007) Health monitoring of Structural Materials and Components Wiley 460 pages

[35] Gopalakrishnan S (2009) ldquoModeling Aspects in Finite Elementrdquo Encyclopedia of Struc-tural Health Monitoring 2 pp791-809

[36] Palacz M Krawczuk M and Ostachowicz W (2005) ldquoThe spectral finite element mod-el for analysis of flexure-shear coupled wave propagation Part 2 Deminated multilayer composite beamrdquo Composite Structures 68 pp45-51

[37] Baik J M and Thopson R B (1984) ldquoUltrasonic scattering from imperfect interfaces A quasi-static modelrdquo Journal of Nondestructive Evaluation 4(3) pp177-196

[38] Luo H and Hanagud S (2000) ldquoDynamics of delamincated beamsrdquo International Jour-nal of Solids and Structures 37 pp1501-1519

[39] Kim S B and Sohn H (2007) ldquoInstantaneous reference-free crack detection based on polarization characteristics of piezoelectric materialsrdquo Smart Materials and Structures 16 pp2375-2387

[40] An Y K and Sohn H (2010) ldquoInstantaneous crack detection under varying temperature and static loading conditionsrdquo The Special Issue of Structural Control and Health Monitor-ing 17(7) pp 730-741

- 35 -

[41] Lee S J and Sohn H (2006) ldquoActive self-sensing scheme development for structural health monitoringrdquo Smart Materials and Structures 15(6) pp 1734-1746

- 36 -

요 약 문

복합재료 손상의 무기저 진단 및 정량화

복합재료는 재료의 경량성 내구성 고강도의 특성으로 인하여 항공기에서부터

교량에까지 널리 사용되고 있다 하지만 외부의 충격으로 인한 복합재료 내부의

손상은 재료의 강도를 감소시켜 구조물의 안정성을 저하시키는 요인이 된다 하지만

복합재료의 손상은 재료 내부에서 발생하여 육안으로는 식별하기도 힘들며 표면에서

보이는 것만으로 내부의 손상 정도를 판단하기는 불가능하다는 단점을 가지고 있다

본 연구에서는 손상 내부에서 반사로 인해 발생하는 파를 이용하여 기저자료를

사용하지 않고 특정 시점에서 얻은 자료만으로 복합재료의 손상을 진단하고 정량화

할 수 있는 기법을 제안한다 제안된 기법의 적용성을 검증하기 위하여 유한요소법을

이용한 수치해석을 수행하였고 Teflon tape 이 삽입된 시편과 실제 충격하중으로

인한 손상을 입은 시편을 이용한 실험을 진행하였다 수치해석 및 실험 결과 제안된

기법이 복합재료의 손상을 성공적으로 진단 및 정량화할 수 있음을 확인하였다 또한

다양한 외부 온도 환경에서 제안된 기법의 환경영향 실험을 수행하였고 다양한

환경에서 손상의 진단 및 정량화가 성공적으로 이루어짐을 검증하였다

핵심어 무기저 손상진단 및 정량화 압전센서 복합재료 유한요소법 Matching pursuit

method

- 37 -

감 사 의 글

기대와 두려움을 안고 시작한 석사과정이 어느덧 2 년 이라는 세월 속에서 추억으로 자리잡아

가고 있습니다 학부생 기간을 포함한 지난 3 년 동안 SHM 이라는 새로운 분야를 접하고 연구를

수행 할 수 있도록 도와주신 고마운 분들이 있어 그 감사를 표하고자 합니다

우선 저의 많은 부족한 점을 일깨워 주시고 연구에 대한 뜻을 더욱 확고히 세울 수 있게 해

주신 지도교수님이신 손 훈 교수님께 깊은 감사의 뜻을 전합니다 짧은 기간이었지만 교수님의

도움으로 많은 것을 배우고 느끼고 깨달을 수 있었습니다 또한 본 연구를 수행할 수 있도록 지

원해주신 국토해양부 (U-City 석 박사과정 지원사업)에 깊은 감사 드립니다

때로 힘들고 지칠 때 항상 저를 믿고 응원해 주시는 가족이 있었기에 지금의 글을 쓸 수 있었

습니다 또한 지금 연구실에서 함께 생활하고 있는 윤규형 민구형 철민이형 현석이형 진열이형

호민이형 병진이 선혜누나와 성대에 계신 창길이형 미국에 계신 의재형과 Debaditiya 에게도 감

사의 말씀을 전합니다 그리고 모두 나열할 수는 없지만 카이스트 생활의 활력소가 되어준 카이

스트 건설 및 환경공학과 선후배들 직원 분들과 SOS 멤버들에게도 감사를 표합니다

앞서 거론한 많은 이들의 도움이 없었다면 저 또한 지금 이 자리에 있을 수 없었을 것입니다

많이 부족하기에 더 노력하고 배우고자 했던 초심을 다시 한번 생각하며 더 큰 결실을 맺기 위해

앞으로 나아가겠습니다 도움을 주신 모든 분들께 이 감사의 글을 바칩니다

2011년 6월 카이스트에서 임형진 올림

- 38 -

CURRICULUM VITAE

Hyung Jin Lim

Research Assistant

The Civil and Environmental Engineering Department

Korea Advanced Institute of Science and Technology (KAIST)

291 Daehak-Ro Guseong-Dong Yuseong-Gu

Daejeon Republic of Korea 305-701

Phone (82)+42-350-3665 Fax (82)+42-350-3610

Email limhj87gmailcom

RESEARCH INTERESTS

Damage Detection Localization and Quantification on Composite and Metallic Structures Pattern Recognition Finite Element Modeling (FEM)

EDUCATION

2009 ndash 2011 MS Civil and Environmental Eng Korea Advanced Institute of Science and Technology (KAIST) Korea Rep

2006 ndash 2009 BS Civil and Environmental Eng Korea Advanced Institute of Science and Technology (KAIST) Korea Rep

JOURNAL PUBLICATIONS The corresponding author is underlined 1 Hyung Jin Lim Hoon Sohn ldquoReference-free damage detection localization and quantification

using in compositerdquo in preparation for Journal of the Acoustical Society of America 2011

2 Chul Min Yeum Hoon Sohn Jeong Bum Ihn Hyung Jin Lim ldquoInstantaneous delamination detection in a composite using a dual piezoelectric transducer networkrdquo Submitted to Compo-sites Science and Technology 2011

3 Yun-Kyu An Hyung Jin Lim Min Koo Kim Jin Yeol Yang Hoon Sohn Chang Guen Lee ldquoInvestigation of applicability of local reference-free damage detection techniques to in-situ

- 39 -

bridgesrdquo Submitted to ASCE Journal of Structural Engineering 2011

4 Hyung Jin Lim Min Koo Kim Hoon Sohn Chan Yik Park ldquoImpedance based damage detec-tion under varying temperature and loading conditionsrdquo Submitted to NDTampE International 2011

5 Hyung Jin Lim Hoon Sohn ldquoQuantitative estimation of transmitted and reflected lamb waves at discontinuityrdquo Journal of the Korean Society for Nondestructive Testing Vol30 No 4 pp1-8 2010

CONFERENCE PROCEEDINGS

The corresponding author is underlined 1 Hyung Jin Lim and Hoon Sohn ldquoReference-free damage detection localization and quantifi-

cation in compositesrdquo ASCE Engineering Mechanics Institute Conference Boston June 2-4 2011

2 Hyung Jin Lim and Hoon Sohn ldquoReference-free delamination quantification within multi-layer composite materialsrdquo KSNT Spring Conference May 26-27 Gyeong-Ju Korea 2011

3 Min Koo Kim Hyun Jin Lim Hoon Sohn Chan Yik Park ldquoImpedance-based bolt loosening detection under varying temperature and loadingrdquo Asian Pacific Workshop on Structural Health Monitoring Tokyo Japan November 30-December 2 2010

4 Hyung Jin Lim Chul Min Yeum and Hoon Sohn ldquoModeling of Impact-induced Delamination in a Multilayer Composite Platerdquo The 23rd KKCNN Symposium on Civil Engineering Taipei (Taiwan) November 13-15 2010

5 Yun-Kyu An Hyung Jin Lim and Hoon Sohn ldquoTemperature independent Delamination Detec-tion using data normalizationrdquo The 3rd International Conference on Multi-Functional Materi-als and Structures Chonbuk National Univ Jeonju Korea September 14-18 2010

6 Hoon Sohn Yun-Kyu An Chang Gil Lee Min Koo Kim Hyung Jin Lim ldquoApplication of online NDT techniques to decommissioned bridge testingrdquo the 5th European Workshop on Structural Health Monitoring Sorrento Italy June 29-July 02 2010

7 Min Koo Kim Hyung Jin Lim Hoon Sohn Chan Yik Park ldquoImpedance based damage diag-nosis of a complex composite aircraft wing under changing loading and temperature condi-tionsrdquo the 5th European Workshop on Structural Health Monitoring Sorrento Italy June 29-July 02 2010

8 Yun-Kyu An Min Koo Kim Hyung Jin Lim and Hoon Sohn ldquoValidations of reference-free crack detection techniques through a decommissioned bridge testrdquo 2010 COSEIK Annual Con-ference Jeju Korea April 8-9 2010

- 40 -

PARTICIPANT PROJECTS

1 Development of Smart Steel Members for Online Steel Structure Monitoring POSCO Corporation (Funded 41000000 KRW (41000 USD) for 010511 to 123111)

2 Development of Enabling Technologies for Structural Health Monitoring Korean Agency for Defense Development (Funded 240000000 KRW (240000 USD) out of total 3200000000 KRW (3200000 USD) for 060108-053111)

3 Development of On-board SHM Technologies for Composite Air Vehicles The Boeing Company (Funded 400000 USD (400000000 KRW) for 080108 to 073111)

EDUCATIONAL EXPERIENCES

Fall 2009

Teaching Assistant CE207 ldquoElementary Structural Engineering and Laboratoryrdquo KAIST 3 unit undergraduate course

EXPERIMENTAL EXPERIENTCES

2010 Monitoring of Real Aircraft Composite Wing Segment Korea Agency for Defense De-velopment (ADD) Daejeon Korea Real time monitoring of composite air vehicles fund-ed by Korea Agency for Defense Development (ADD)

2009-2011

Field Monitoring of Real Bridge Structures Gimpo (Ramp G bridge) amp Icheon (Sam-seung bridge) Korea US-Korea collaborative research for bridge monitoring testbeds

HONORS amp AWARDS

Scholarships for outstanding students granted KAIST (2007-2009) The three prominent students in the civil and environmental engineering department of KAIST were selected as recipients This scholarship provided 2000$ per year for three years

- 41 -

Page 19: 복합재료 손상의 무기저 진단 및 정량화 - KAISTssslab.kaist.ac.kr/article/pdf/Thesis_2.pdf · 1 Declaration of Ethical Conduct in Research: I, as a graduate student

Moreover 119907119907119894119894 and 119907119907119889119889 are also defined as the A0 Lamb wave mode velocity in pristine area

and damage area respectively Note that the whole procedure of the damage characterization

is accomplished based solely on the signals obtained from the current state of the structure

32 Damage index (m-value)

The proposed reference-free damage characterization technique utilizes the waves reflect-

ed additionally from the damage area as a feature The arrival time difference of the transmit-

ted and reflected A0 Lamb wave modes reflected from the exit and the entrance of the delam-

ination area one times each (∆119905119905) contains the information of the physical length of the dam-

age (119889119889) and A0 Lamb wave mode velocity in the damage area (119907119907119889119889) The wave velocity in the

damage area which is related to the shear modulus (G) represents the severity of the damage

∆119905119905 = 2119905119905119889119889 =2119889119889119907119907119889119889

(2)

However based on the measured data available the 119889119889 and 119907119907119889119889 (or G) cannot be uniquely

determined Fig 4 shows two FEM models which have different physical length and wave

velocity (G) of the damage Detail explanation of delamination modeling technique will be

covered in section 4 The similar signal in the time domain and ∆119905119905 are obtained from the

two different damage cases

- 8 -

Damage 1

Damage 2

20

Unit mm15

90 G value reduction

943 G value reduction

m-value = 248

m-value = 248

015 02 025 03 035

-2

-1

0

1

2

Time (ms)

Ampl

itude

Damage 1Damage 2

Fig 4 Two different FEM models which have different 119889119889 and 119907119907119889119889 (or G) This represents that the damage cannot be characterized uniquely based on the measured data available

Thus in this study the damage index (m-value) which can be uniquely defined based on

the available data is introduced The m-value is defined as the ratio of the physical length of

the damage and wavelength of the A0 Lamb wave mode in the damage area (120582120582119889119889)

119898119898 =119889119889120582120582119889119889

=119889119889119889119889119907119907119889119889

(3)

In many cases the transmitted and reflected waves are overlapped and it is hard to decom-

pose and calculate the arrival times of them Matching pursuit method is an alternative way

to accurately decompose of the overlapped waves and estimate the parameters of each wave

such as arrival time (Fig 5) Detail explanation of the matching pursuit method is covered in

section 36

015 02 025 03-2

-1

0

1

2

Time (ms)

Am

plitu

de

OriginalApproximated

(a)

015 02 025 03-2

-1

0

1

2

Time (ms)

Ampl

itude

Wave 1Wave 2

(b)

Fig 5 (a) Example of the overlapped signal with comparing approximated signal using the matching pursuit method (b) Decomposed signal using the matching pursuit method The matching pursuit method provides a conventional way to decompose and identify each wave

- 9 -

33 Damage detection algorithm (Level 1)

The damage diagnosis is accomplished by extracting the additional reflected A0 Lamb

wave modes from damage due to multiple reflections If A0 Lamb wave modes are decom-

posed perfectly it is trivial that one transmitted A0 Lamb wave mode is always captured in

the response signal If damage exists other waves in the response signal behind the transmit-

ted A0 Lamb wave mode are presented Since this approach does not rely on previously col-

lected data the diagnosis is independent from environmental variation such as temperature

changes

It is reported that the dispersion of Lamb waves cause wave-packets to spread out in space

and time as they propagate through a structure [20] Thus the reflected waves from the dam-

age have longer pulse duration than the transmitted wave due to their longer propagating dis-

tance In this study the matching pursuit method is modified to capture more spread wave

than the transmitted A0 Lamb wave mode Detail explanation of the modified matching pur-

suit method is described in latter section

Let 1199041199041 and 1199041199042 as the pulse duration (scale) of the transmitted and additional reflected A0

Lamb wave mode respectively

ldquoIf 1199041199042 gt 1199041199041 damage exists

Otherwise there is no damagerdquo

- 10 -

34 Damage quantification algorithm (Level 3)

As mentioned in the previous section damage is quantized by estimating 119889119889 and 119907119907119889119889 using

pitch-catch method The procedure to estimate 119889119889 and 119907119907119889119889 is

(i) Measure 119897119897 119905119905 and 119905119905119889119889

(ii) Assume the range of 119889119889 (0 lt 119889119889 lt 119897119897)

(iii) Calculate 119907119907119894119894 and 119907119907119889119889 with a constraints (0 lt 119907119907119889119889 lt 119907119907119894119894) (Eq 4 and 5)

(iv) Estimate m-value and find a feasible range of 119889119889 and 119907119907119889119889

119907119907119894119894 =119897119897 minus 119889119889119905119905 minus 119905119905119889119889

(4)

119907119907119889119889 =119889119889119905119905119889119889

(5)

35 Damage localization algorithm (Level 2amp3)

For damage localization 1198971198971 is estimated by measuring the arrival time from the excitation

PZT to the entrance of delamination (1199051199051) with additional pulse-echo method implementation

The procedure is similar with the previous damage quantification algorithm

(i) Measure 119897119897 119905119905 119905119905119889119889 and 1199051199051

(ii) Assume the range of 1198971198971 (0 lt 1198971198971 lt 119897119897)

(ii) Compute possible combinations of 1198971198972 119889119889 and 119907119907119889119889 with constraints

(0 lt 119907119907119889119889 lt 119907119907119894119894 119889119889 gt 0) (Eq 5 6 and 7)

- 11 -

1198971198972 = 1198971198971119905119905 minus 1199051199051 minus 119905119905119889119889

1199051199051 (6)

119889119889 = 119897119897 minus 1198971198971 minus 1198971198972 (7)

Since a PZT for pulse-echo method acts as both an actuator and a sensor the obtained sig-

nal contains both the excitation and the response signal Thus the reflected wave from the

entrance of damage can be buried into the tail of excitation signal and cause error on arrival

time of reflected wave from damage Even in this case damage can be localized and quan-

tized using the constraints When first wave packet reflected from the entrance of damage is

buried into the excitation signal the next wave packet reflected from the exit of damage can

be utilized alternatively If the algorithm regards the second reflected wave as the first re-

flected wave there are no combinations of estimated variable meeting the constraints Then

the algorithm repeats the process by regarding the wave as the second reflected wave the

constraints can select proper combination of the estimated variables

The estimated values have range bounded by the constraints When additional information

of 119907119907119894119894 is utilized the damage can be characterized more accurately In this study three meth-

ods are proposed and verified as information level of 119907119907119894119894

Method 1 Solely based on the current data measurement (Reference-free)

Method 2 Limit the possible range of 119907119907119894119894 (Prior knowledge of the system)

Method 3 Measure 119907119907119894119894 (Additional data need to be collected)

Method 1 described in this section can be utilized without any information of 119907119907119894119894 Meth-

od 2 produces more accurate estimation results using prior knowledge of the system such as

- 12 -

119907119907119894119894 variation under temperature changes Method 3 gives unique estimated value using meas-

ured 119907119907119894119894 from other path in the same time In section 4 and 5 damage is characterized using

these three methods and the results are compared each another

36 Matching pursuit method considering the dispersion effect of Lamb

wave

Matching pursuit method has been developed for signal compression approximation and

characterization of obtained signal and function [21-23] The key concept of the matching

pursuit method is that any signal (function) can be decomposed into a linear expansion of

signals selected from a pool of basis signals (functions) called lsquodictionaryrsquo

Let 119891119891 be a function to be analyzed then 119891119891 can be approximated a set of functions select-

ed from the dictionary 119863119863 in order to best match This is done by successive approximations

with orthogonal projections Let1198921198921205741205740 isin 119863119863 the function 119891119891 can be decomposed into

119891119891 = lang1198911198911198921198921205741205740rang1198921198921205741205740 + 119877119877119891119891 (8)

where 119877119877119891119891 is the residual function after approximating 119891119891 using 1198921198921205741205740 1198921198921205741205740 is orthogonal to

119877119877119891119891 hence

1198911198912 = lang1198911198911198921198921205741205740rang2

+ 1198771198771198911198912 (9)

To minimize 119877119877119891119891 1198921198921205741205740 is chosen to maximize lang1198911198911198921198921205741205740rang The matching pursuit method is

an iterative algorithm that decomposes the residue 119877119877119891119891 by projecting it on a function of 119863119863

that matches 119877119877119891119891 almost at best as it was done for 119891119891 The residue 119877119877119899119899119891119891 is decomposed into

119877119877119899119899119891119891 = lang119877119877119899119899119891119891119892119892120574120574119899119899rang119892119892120574120574119899119899 + 119877119877119899119899+1119891119891 (10)

- 13 -

the decomposition is carried up to the order 119898119898 119891119891 can be expressed as

119891119891 = lang119877119877119899119899119891119891119892119892120574120574119899119899rang119892119892120574120574119899119899

119898119898minus1

119899119899=0

+ 119877119877119898119898119891119891 (11)

The original function 119891119891 is decomposed into a sum of dictionary elements that are chosen to

best match its residues

In this study the matching pursuit method is utilized to accurately decompose of the over-

lapped Gaussian pulses and estimate the parameters of the pulse which used for damage

characterization The 119894119894119905119905ℎ Gaussian pulse in the dictionary for approximating the obtained

signal is expressed as

119892119892120574120574119894119894 = 119891119891(119860119860119894119894 119906119906119894119894 119904119904119894119894 119889119889119894119894120601120601119894119894) (12)

where 119860119860119894119894119906119906119894119894119904119904119894119894119889119889119894119894 and 120601120601119894119894 represent amplitude arrival time scale (pulse duration) excita-

tion(central) frequency and phase of the Gaussian pulse respectively The parameters of the

pulse are estimated using a nonlinear regression based optimization method (Gauss-Newton

method) based on an initial guess of the parameters Detail procedure for estimating the pa-

rameters of the pulse is expressed in Hong et al [23]

Among the estimated parameters of the pulses pulse duration (scale) represents the disper-

sion effect If a wave-packet is spread out the estimated pulse duration is increased Using

this feature the matching pursuit method is modified to capture waves which have longer

pulse duration than transmitted A0 Lamb wave mode If the estimate pulse duration of a wave

is less than the transmitted wave the modified matching pursuit method regards it as error

- 14 -

CHAPTER 4 NUMERICAL SIMULATION

41 Composite modeling method

Composite materials can be modeled from nano to macro level depends on its engineering

purpose In general four methods are utilized for modeling of composite materials First

composite is modeled as a homogenous material expressed in terms of effective material

properties [24] The effective material properties are estimated using the modulus and the

volume fraction of each constituent theoretically (Rule of mixture) and measured experimen-

tally Second a model is established that each laminate which have own material properties

are totally bonded together In this model the stacking sequence and direction of each lami-

nates are considered additionally [25-27] Third method for composite modeling in micro

level called micromechanics [28-30] This model considers the interaction between fibers and

matrix in laminate Last multiscale modeling which the scope of modeling is extended to

molecular or quantum level of constituents [31-33] This model starts from molecular me-

chanics or molecular dynamics then the scale of model leads to the nano micro and macro

level by incorporating small scale models As the scale of models is smaller it provides more

logical way than other large scale of models while the computational cost is more expensive

The scope of this study is focused on the dynamic behavior of composite structure the beam

or plate model was utilized for the numerical simulation

- 15 -

42 Damage modeling method

Generally three damage models have been developed and utilized in the real applications

First the stiffness reduction model called as the equivalent element model was introduced

[17 34-35] This model is simply realized by replacing the region around damage area by a

region with adjusted material properties (Youngrsquos modulus shear modulus density etc)

globally Next the duplicated node model is done by disconnecting the element connectivity

assuming there is no interaction among the delaminated surface This model can be utilized

when the location and the extension of damage is important issue [34-36] Last the contact

mechanics model using the principles of contact mechanics the two interfacial forces (con-

tact resistance and sliding resistance) in the delaminated surface [37-38] In this model the

delaminated surfaces are modeled as nonlinear functions of the interlaminar normal strain

and the interlaminar shear strain This model is computationally expensive while provides

more logical way than other models There is however no universally accepted model for

general impact-induced damage in composites In the case of impact-induced damage it is

impossible to predict or estimate the location and number of delaminated layers Thus from

the uncertainty of impact-induced damage the equivalent element model can be considered

as the effective damage model

- 16 -

43 Numerical simulation description

TABLE I Material property of simulation model and test specimen

Tensile Strength 657 MPa Tensile Modulus 59 GPa Shear Strength 804 MPa Shear Modulus 24 Gpa Specific Gravity 16

PZT A

PZT C

PZT BDamage

3

250 125 125

500 Unit mm

Fig 6 Description of the numerical simulation model

The feasibility of the proposed damage characterization technique is verified by 2-D FEM

simulation using MSC PATRAN and NASTRAN Fig 6 presents the dimension PZTs and

damage location of the simulation model In the model 05mm x 05mm square shell element

was used for modeling of the A0 mode whose flexural wave characteristic The element size

was determined by considering the wavelength of the propagation Lamb wave modes The

material properties of the simulation model and damage cases are presented in Table I and

Table II respectively

PZT A and C act as both actuator and sensor for collecting pulse-echo signal They were

installed on the plate which two of them are both-side collocated for the A0 Lamb wave

- 17 -

mode decomposition [39-40] PZT B was only used as a sensor for collecting pitch-catch sig-

nal For simulating the excitation and sensing of PZTs the pin-force model was adopted that

an input force apples to the end nodes of excitation PZT and averaging element strains corre-

sponding sensing area A Gaussian pulse (tone-burst) 50 kHz central frequency was excited

as an input signal

TABLE II Damage cases for numerical simulation

119949119949120783120783 (mm) 119941119941 (mm) G value reduction ()

Case 1 650 200 90 darr

Case 2 675 150 90 darr

Case 3 675 150 93 darr

Case 4 900 150 90 darr

- 18 -

44 Numerical simulation results

015 02 025 03 035

-2

0

2

Time (ms)

Am

plitu

de

IntactDamage

Reflected wave from damage

(a) Pitch-catch signal

015 02 025 03 035

-2

0

2

Time (ms)

Am

plitu

de

IntactDamage

Reflected waves from damage

(b) Pulse-echo signal Fig 7 Comparison of A0 Lamb wave mode signals from numerical simulation between intact and damage cases Time delay attenuation and multiple reflections due to damage simulated successfully using the damage modeling method described in the previous section

Fig 7 shows that comparison of A0 Lamb wave mode signals between intact and one of

damage cases in the simulation The features of A0 Lamb wave mode passing through dam-

age area described in section 2 are successfully simulated Thus the equivalent element

method is proper for modeling damage in composites In the damage cases the additional

waves reflected from damage is observed in both pitch-catch and pulse echo signal

Table III presents the result of the proposed damage characterization algorithm using the

numerical simulation cases shown in Table II In lsquoMethod 2rsquo 119907119907119894119894 variation is assumed to plusmn

5 of the normal value (1190~1315ms) The normal value of 119907119907119894119894 for lsquoMethod 3rsquo is meas-

ured as 1253 ms In each case the damage is characterized using the proposed technique

Note that as the information of 119907119907119894119894 is specified the estimated value is closer to the exact

value

- 19 -

TABLE III Damage characterization for the numerical simulation models

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

Case 1

Exact 650 200

Method 1 248 50~70 79~414 160~834

Method 2 248 603~667 135~241 271~486

Method 3 248 635 188 379

Case 2

Exact 675 150

Method 1 193 60~75 01~251 37~651

Method 2 193 618~682 114~222 295~575

Method 3 193 65 168 435

Case 3

Exact 675 150

Method 1 233 60~75 66~302 140~649

Method 2 233 668~738 85~196 182~419

Method 3 233 703 140 301

Case 4

Exact 900 150

Method 1 204 75~100 48~349 118~856

Method 2 204 860~951 107~216 264~530

Method 3 204 906 162 397

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1253 ms 1190~1315 ms Measured 119907119907119894119894 1253 ms

- 20 -

CHAPTER 5 EXPERIMENTAL VERIFICATION

51 Overview

In this section experimental study is investigated so that the proposed damage characteri-

zation technique Overall experimental setup and test specimen are shown in Fig 8 and 9

The experimental setup consists of test specimens data acquisition (DAQ) system and self-

sensing circuit Self-sensing circuit was installed for pulse-echo measurement using PZTs

[40-41]

PZT A(PZT C) PZT B

500

500

140

75

250

Teflon Inserted Damage (d =25)

Unit mm

Thickness = 3

PZT A(PZT C) PZT B

500

500

150

90

250

Impact Location

Unit mm

Thickness = 3

(a) Teflon inserted damage specimen

(b) Impact-induced damage specimen

Fig 8 Test specimens for the experimental verification In the Teflon inserted damage Teflon tapes with 25mm diameter were inserted at the center of the specimen The impact-induced damage was introduced to the specimen using the impact test equipment

- 21 -

511 Specimen

Two carbon composite plates with a dimension of 500times500times3mm3 were manufactured by

Nexcomes Inc by stacking 12 layers of CF3327 3K prepreg (woven fabrics type) manufac-

tured by Hankuk Fiber Inc The material properties of the specimen are presented in Table I

One of the specimens Teflon tapes with 25mm diameter was inserted in every 3 layers at the

center of the specimen to mimic the damage The other specimen is utilized for the character-

ization of real impact-induced damage

512 Data acquisition (DAQ) system and PZT

The data acquisition (DAQ) system consists of an arbitrary waveform generator (AWG) a

high speed signal digitizer (DIG) and two multiplexers (MUX) Using the 16-bits AWG

tone-burst input signal with a plusmn10 peak-to-peak voltage were generated The response signal

was measured by the 14-bit DIG For the signal-to-noise ratio reduction both pitch-catch and

pulse-echo signals were measured 150 times and averaged in the time domain 3 identical

circular PZTs of 10mm diameter manufactured by APC International Ltd were installed on

the specimen A Gaussian pulse (tone-burst) 50 kHz central frequency was excited as an in-

put signal The excitation frequency was chosen in order to avoid overlapping of the S0 and

A0 Lamb wave modes and generating of the higher modes (S1 A1hellip)

- 22 -

PZT Actuator

PZT Sensor

AWG

DIG

DAQ System Specimen

SensingActuatingMUX

1

23

4

1 Sending signal to MUX

4 Sending signal to DIG

Self-SensingCircuit

18 Unit mm

10

(a)

(b)

Fig 9 Overall experimental setup (a) Data acquisition (DAQ) system and (b) Installed circu-lar PZT transducer on the specimen with its dimension

513 Impact test

The impact-induced damage was introduced to the specimen using the impact test equip-

ment presented in Fig 10 A 5kg mass with 10mm tip was dropped from 015m height onto

the same location of the specimen The signals were obtained at each number of impacts

Since the reflected wave is not detected at the first impact case the data obtained from the

second and third impact were utilized for the analysis

10

Unit mm (a)

(b)

Fig 10 (a) Impact test equipment for introducing impact-induced damage and (b) Tip of the impact tester with 10mm diameter

- 23 -

514 Temperature effect test

In reality structures are subjected to changing environmental conditions such as tempera-

ture that affect measured signal and cause false-alarm Thus the robustness of the proposed

damage characterization technique under temperature variation in investigated (Fig 11) The

specimens were placed inside a temperature chamber and a thermocouple was installed on

the specimen to measure temperature of surface of the specimen The signal was obtained

under three temperature cases (0 20 and 50degC) and the humidity was kept at 30

Temperature Chamber

Test Specimen

Fig 11 Test specimen in temperature chamber

- 24 -

52 Teflon inserted damage

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Reflected wave from damage

(a)

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Transmitted WaveReflected Wave

Detected reflected wave

(b)

Fig 12 A0 Lamb wave mode obtained using pitch-catch method from the Teflon inserted damage specimen (a) The reflected wave from damage is shown (b) The Teflon inserted damage is detected by capturing the reflected wave

Fig 12 shows the obtained pitch-catch signal from the Teflon inserted damage specimen

and damage diagnosis result The Teflon inserted damage is detected successfully by captur-

ing the reflected wave Table IV presents the result of the Teflon inserted damage characteri-

zation 119907119907119894119894 is measured as 1213ms and the variation is assumed to plusmn 5 of the normal val-

ue (1152~1273ms) The results indicate that the proposed damage characterization approach

works well in real application

TABLE IV Damage characterization for the Teflon inserted damage specimen

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

Exact 635 25

Method 1 123 60~75 73~338 298~1220

Method 2 123 619~662 229~310 934~1188

Method 3 123 631 284 1161

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1213 ms

- 25 -

53 Impact-induced damage

11 115 12 125 13-4

-2

0

2

4

Time (ms)

Vol

tage

(mV

)

IntactImpact2Impact3

Reflected wave from damage

(a) Pitch-catch signal

11 115 12 125 13 135

-05

0

05

1

Time (ms)

Volta

ge (m

V)

IntactImpact2Impact3

Boundary reflection

Reflected wave from damage

(b) Pulse-echo signal

Fig 13 The signal obtained from the test specimen subjected to impact The impact was in-troduced at the same location of the specimen and the signal was obtained after each impacts As the number of impact is increased the transmitted A0 Lamb wave mode is delayed and attenuated

The obtained signals from impact-induced damage specimen are presented in Fig 13 It is

observed that as the number of impact is increased the transmitted A0 Lamb wave mode is

more delayed because the wave velocity in the damage area is reduced Fig 14 shows the

false-alarm test result whether the proposed damage detection algorithm works even in pris-

tine condition of the test specimen It is shown that no false alarm is occurred because no ad-

ditional reflected wave is captured

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Transmitted Wave

(a) Transmitted wave

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Reflected Wave

No reflected wave

(b) Reflected wave

Fig 14 False alarm test for the proposed damage detection algorithm in pristine condition of the specimen Reflected wave from damage is not captured false alarm is not occurred

- 26 -

TABLE V Damage characterization for the impact-induced damage specimen

of Impacts m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

2

(1470 J)

Method 1 080 65~90 75~246 470~1157

Method 2 080 801~885 114~245 711~1134

Method 3 080 843 179 1122

3

(2205 J)

Method 1 148 50~100 06~380 21~1134

Method 2 148 758~838 249~368 843~1125

Method 3 148 798 308 1044

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1213 ms

Table V presents the characterization results of the impact-induced damage case Here the

measured 119907119907119894119894 and the assumed variation are same as the artificial damage case It is observed

that as the number of impact is increased 119889119889 is increased while 119907119907119889119889 is reduced Thus it is

validated that the proposed damage characterization approach estimates the increment of the

damage well

54 Temperature effect test

541 Teflon inserted damage

11 115 12 125 13-10

-5

0

5

10

Time (ms)

Vol

tage

(mV

)

0 oC20 oC50 oC

(a) Pitch-catch signal

11 115 12 125 13 135

-1

0

1

2

3

Time(ms)

Volta

ge (m

V)

0 oC20 oC50 oC

(b) Pulse-echo signal

Fig 15 Signals obtained from the Teflon inserted damage specimen under temperature varia-tion Time delay and amplitude variation are occurred

Fig 15 shows the obtained signals from the Teflon inserted damage specimen subjected to

- 27 -

temperature variation Time delay and amplitude variations are occurred there is no addition-

al damage is introduced The analyzed results of the obtained signals are reported in Table VI

The measured 119907119907119894119894 is 1247 ms at 0 degC 1213 ms at 20 degC and 1168 ms at 50 degC respectively

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms (1152~1273 ms)

The estimated results are also changed under temperature variations the fluctuation of esti-

mated range and exact value is small

TABLE VI Temperature effect test for the Teflon inserted damage specimen

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

0 degC

Exact 635 25

Method 1 113 60~70 79~268 352~1291

Method 2 113 560~619 233~344 1034~1172

Method 3 113 606 257 1141

20 degC

Exact 635 25

Method 1 123 60~75 73~338 298~1220

Method 2 123 619~662 229~310 934~1188

Method 3 123 631 284 1161

50 degC

Exact 635 25

Method 1 137 60~75 72~328 263~1228

Method 2 137 622~687 183~299 666~1086

Method 3 137 642 264 960

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1247 ms at 0 degC 1213 ms at 20 degC and 1168 ms at 50 degC

- 28 -

542 Impact-induced damage

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Vol

tage

(mV

)

Transmitted WaveReflected Wave

No reflected wave

(a) 0 degC

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Transmitted WaveReflected Wave

No reflected wave

(b) 50 degC

Fig 16 False alarm test under temperature variations It is shown that proposed damage char-acterization algorithm is robust under temperature variations

The false alarm test in pristine condition of the impact-induced damage specimen under

temperature variation was investigated (Fig 16) It is validated that the proposed damage di-

agnosis algorithm is robust under temperature variations Fig 17 shows the obtained signals

from the impact-induced damage (impacted 3 times) specimen under temperature variation

The analyzed results of the obtained signals are reported in Table VII The measured 119907119907119894119894 in

each temperature and the assumed variation are same as the artificial damage case

11 115 12 125 13-5

0

5

Time (ms)

Volta

ge (m

V)

0 oC20 oC50 oC

(a) Pitch-catch signal

11 115 12 125 13 135

-1

0

1

2

3

Time(ms)

Vol

tage

(mV

)

0 oC20 oC50 oC

(b) Pulse-echo signal

Fig 17 Signals obtained from impact-induced damage specimen (Impact 3) under tempera-ture variation

- 29 -

TABLE VII Temperature effect test for the impact induced damage specimen (Impact 3 case)

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

0 degC

Method 1 135 75~90 72~310 267~1148

Method 2 135 705~779 264~381 978~1126

Method 3 135 763 289 1071

20 degC

Method 1 148 50~100 06~380 21~1134

Method 2 148 758~838 249~368 843~1125

Method 3 148 798 308 1044

50 degC

Method 1 153 75~100 53~415 174~1227

Method 2 153 804~889 214~336 702~1103

Method 3 153 815 320 1050

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1247 ms at 0 degC 1213 ms at 20 degC and 1168 ms at 50 degC

- 30 -

CHAPTER 6 CONCLUDING REMARKS

61 Executive summary

In this study the reference-free damage detection localization and quantification tech-

nique in a multi layered composite plate is proposed Additional reflected A0 Lamb wave

mode from the damage due to multiple reflections in damage area is used as a feature for

characterizing the damage First damage is characterized in terms of a damage index (m-

value) defined as the ratio of the physical length of the damage to the wavelength of the A0

Lamb wave mode Then the range of possible damage location length and severity are esti-

mated with constraints This procedure can be accomplished without prior baseline data To

realize the proposed technique the matching pursuit technique is modified for reduction of

the error by considering the dispersive characteristic of Lamb wave The matching pursuit

method estimates the arrival time of each wave packets accurately even two wave packets are

overlapped The numerical simulations and experimental tests including Teflon tape inserted

and impact-induced damage are investigated for the feasibility of the proposed technique

Moreover the effectiveness of the proposed technique under changing temperature is exper-

imentally verified

- 31 -

62 Suggestion for further study

The limitations of this study and the possible improvement and further studies are listed as

follow

1 Measurement of impact-induced damage length using conventional NDT technique

Additional experimental verification for impact-induced damage characterization is

needed by measuring the real length of damage using conventional NDT technique

such as ultrasonic scan or infrared camera

2 Limitation of both side collocated PZT installation In this study excitation PZTs are

installed both side and collocated for decomposing A0 Lamb wave mode However it

can produce decomposition error due to imperfection of PZT location and their installa-

tion Moreover in many cases assess to both surface of structures is not allowed such

as aircraft applications To improve this limitation a reference-free damage characteri-

zation using single side PZTs will be considered

3 Complex structure application In this study damage is characterized in simple plate-

like structure When the monitored structure becomes complex the response signals al-

so becomes complex due to its geometry and boundary Thus the investigations for

complex structure application such as stiffener spar and curvature are needed

- 32 -

REFERENCES

[1] Cawley P and Alleyne D (1996) ldquoThe use of Lamb waves for the long range inspection of large structuresrdquo Ultrasonics 34 pp287-290

[2] Ghosh T Kundu T and Karpur P (1998) ldquoEfficient use of Lamb waves for detecting defects in large platesrdquo Ultrasonics 36 pp791-801

[3] Guirguitui V (2003) ldquoLamb wave generation with piezoelectric wafer active sensors for structural health monitoringrdquo SPIE Conference on Smart Structures and Materials San Diego CA 5391 pp111-422

[4] Ragvahan A and Cesnik CE S (2004) ldquoModeling of piezoelectric-based Lamb wave generation and sensing for structural health monitoringrdquo SPIE Conference on Smart Struc-tures and Materials San Diego CA 5391 pp419-430

[5] Rose JL Pilarski A and Dirit JJ (1993) ldquoAn approach to guided wave mode selection for inspection of laminated platerdquo Journal of Reinforced Plastics and Composites 12(5) pp536-544

[6] Guo N and Cawley P (1993) ldquoThe interaction of Lamb waves with delaminations in composite laminatesrdquo Journal of Acoustical Society of America 94(4) pp 2240-2246

[7] Valdez SHD and Soutis C (2002) ldquoReal-time nondestructive evaluation of fiber com-posite laminates using low-frequency Lamb wavesrdquo Journal of Acoustical Society of Amer-ica 111(5) pp 2026-2033

[8] Yang M and Qiao P (2005) ldquoModeling and experimental detection of damage in virous materials using the pulse-echo method and piezoelectric sensoractuatorsrdquo Smart Materials and Structures 14 pp1083-1100

[9] Lestari W and Qiao P (2005) ldquoApplication of wave propagation analysis for damage identification in composite laminated beamsrdquo Journal of Composite Materials 39(22) pp1967-1984

[10] Kessler S S Spearing S M and Soutis C (2002) ldquoDamage detection in composite ma-terials using Lamb wave methodsrdquo Smart Materials and Structures 11 pp269-278

[11] Sohn H Park G Wail J R Limback N P and Farrar C R (2004) ldquoWavelet-based active sensing for delamination detection in composite structuresrdquo Smart Materials and Structures 13 pp 153-160

[12] Petculescu G Krishnaswamy S and Achenback J D (2008) ldquoGroup delay measure-ments using modally selective Lamb wave transducers for detection and sizing of delamina-tions in compositesrdquo Smart Materials and Structures 17

[13] Sohn H Park H W Law K H and Farrar C R (2007) ldquoDamage detection in compo-

- 33 -

site plates by using an enhanced time reversal methodrdquo Journal of Aerospace Engineering (ASCE) pp 141-151

[14] Sohn H (2007) ldquoEffects of environmental and operational variability on structural health monitoringrdquo Philosophical Transactions of the Royal Society Ardquo 365 pp539-560

[15] Rytter A (1993) ldquoVibration based inspection of civil engineering structuresrdquo PhD The-sis Department of Building Technology and Structural Engineering University of Aalborg Aalborg Denmark

[16] Hayashi T and Kawashima K (2002) ldquoMultiple reflections of Lamb waves at a delami-nationrdquo Ultrasonics 40 pp193-197

[17] Wang C H and Rose L R F (2003) ldquoWave reflection and transmission in beams con-taining delamination and inhomogeneityrdquo Journal of Sound and Vibration 264 pp 851-872

[18] Ramadas C Balasubramaniam K Joshi M and Krishnamurthy C V (2009) ldquoInterac-tion of the primary anti-symmetric Lamb mode (Ao) with symmetric delaminations numer-ical and experimental studiesrdquo Smart Materials and Structures 18

[19] Ramadas C Balasubramaniam K Joshi M and Krishnamurthy C V (2010) ldquoInterac-tion of guided Lamb waves with an asymmetrically located delamination in a laminated composite platerdquo Smart Materials and Structures 19

[20] Wilcox P Lowe M and Cawely P (2001) ldquoThe effect of dispersion on long-range in-spection using ultrasonic guided wavesrdquo NDTampE International 34 pp1-9

[21] Mallat S G and Zhang Z (1993) ldquoMatching pursuits with time-frequency dictionariesrdquo IEEE Tansactions on Signal Processing 41(12) pp3397-3415

[22] Zhang G Zhang S and Wang Y (2000) ldquoApplication of adaptive time-frequency de-composition in ultrasonic NDE of highly scattering materialsrdquo Ultrasonics 38 pp 961-964

[23] Hong J C Sun K H and Kim Y Y (2005) ldquoThe matching pursuit approach based on the modulated Gaussian pulse for efficient guided-wave damage inspectionrdquo Smart Mate-rials and Structures 14 pp548-560

[24] Herakovich C T (1997) Mechanics of Fibrous Composites Wiley 480 pages

[25] Reddy J N (1985) ldquoA review of the literature on finite-element modeling of laminated composite platesrdquo The Shock and Vibration Digest 17(4) pp 3-8

[26] Reddy J N (2003) Mechanics of Laminated Composite Plates and Shells Theory and Analysis CRC Press 856 pages

[27] Palacz M Krawczuk M and Ostachowicz W (2005) ldquoThe spectral finite element mod-el for analysis of flexure-shear coupled wave propagation Part 1 laminated multilayer

- 34 -

composite beamrdquo Composite Structures 68 pp37-44

[28] Zhang Y C and Harding J (1990) ldquoA numerical micromechanics analysis of the me-chanical properties of a plain weave compositerdquo Computer amp Structures 36(5) pp 839-844

[29] Nemat-Nasser S and Hori M (1999) Micromechanics Overall Properties of Heteroge-neous Materials Elsevier 810 pages

[30] Scida D Aboura Z Benzeggagh M L and Bocherens E (1999) ldquoA micromechanics model for 3D elasticity and failure of woven-fibre composite materialrdquo Composites Science and Technology 59 pp 505-517

[31] Thomas Y and Wu X H (1997) ldquoA multiscale finite element method for elliptic prob-lems in composite materials and porous mediardquo Journal of Computational Physics 134 pp 169-189

[32] Lee K H Moorthy S and Ghosh S (1999) ldquoMultiple scale computational model for damage in composite materialsrdquo Computer Methods in Applied Mechanics and Engineer-ing 172 pp 175-201

[33] Namilae S and Chandra N (2005) ldquoMultiscale model to study the effect of interfaces in carbon nanotube-based compositesrdquo Journal of Engineering Materials and Technology (ASME) 127 pp222-232

[34] Adams D E (2007) Health monitoring of Structural Materials and Components Wiley 460 pages

[35] Gopalakrishnan S (2009) ldquoModeling Aspects in Finite Elementrdquo Encyclopedia of Struc-tural Health Monitoring 2 pp791-809

[36] Palacz M Krawczuk M and Ostachowicz W (2005) ldquoThe spectral finite element mod-el for analysis of flexure-shear coupled wave propagation Part 2 Deminated multilayer composite beamrdquo Composite Structures 68 pp45-51

[37] Baik J M and Thopson R B (1984) ldquoUltrasonic scattering from imperfect interfaces A quasi-static modelrdquo Journal of Nondestructive Evaluation 4(3) pp177-196

[38] Luo H and Hanagud S (2000) ldquoDynamics of delamincated beamsrdquo International Jour-nal of Solids and Structures 37 pp1501-1519

[39] Kim S B and Sohn H (2007) ldquoInstantaneous reference-free crack detection based on polarization characteristics of piezoelectric materialsrdquo Smart Materials and Structures 16 pp2375-2387

[40] An Y K and Sohn H (2010) ldquoInstantaneous crack detection under varying temperature and static loading conditionsrdquo The Special Issue of Structural Control and Health Monitor-ing 17(7) pp 730-741

- 35 -

[41] Lee S J and Sohn H (2006) ldquoActive self-sensing scheme development for structural health monitoringrdquo Smart Materials and Structures 15(6) pp 1734-1746

- 36 -

요 약 문

복합재료 손상의 무기저 진단 및 정량화

복합재료는 재료의 경량성 내구성 고강도의 특성으로 인하여 항공기에서부터

교량에까지 널리 사용되고 있다 하지만 외부의 충격으로 인한 복합재료 내부의

손상은 재료의 강도를 감소시켜 구조물의 안정성을 저하시키는 요인이 된다 하지만

복합재료의 손상은 재료 내부에서 발생하여 육안으로는 식별하기도 힘들며 표면에서

보이는 것만으로 내부의 손상 정도를 판단하기는 불가능하다는 단점을 가지고 있다

본 연구에서는 손상 내부에서 반사로 인해 발생하는 파를 이용하여 기저자료를

사용하지 않고 특정 시점에서 얻은 자료만으로 복합재료의 손상을 진단하고 정량화

할 수 있는 기법을 제안한다 제안된 기법의 적용성을 검증하기 위하여 유한요소법을

이용한 수치해석을 수행하였고 Teflon tape 이 삽입된 시편과 실제 충격하중으로

인한 손상을 입은 시편을 이용한 실험을 진행하였다 수치해석 및 실험 결과 제안된

기법이 복합재료의 손상을 성공적으로 진단 및 정량화할 수 있음을 확인하였다 또한

다양한 외부 온도 환경에서 제안된 기법의 환경영향 실험을 수행하였고 다양한

환경에서 손상의 진단 및 정량화가 성공적으로 이루어짐을 검증하였다

핵심어 무기저 손상진단 및 정량화 압전센서 복합재료 유한요소법 Matching pursuit

method

- 37 -

감 사 의 글

기대와 두려움을 안고 시작한 석사과정이 어느덧 2 년 이라는 세월 속에서 추억으로 자리잡아

가고 있습니다 학부생 기간을 포함한 지난 3 년 동안 SHM 이라는 새로운 분야를 접하고 연구를

수행 할 수 있도록 도와주신 고마운 분들이 있어 그 감사를 표하고자 합니다

우선 저의 많은 부족한 점을 일깨워 주시고 연구에 대한 뜻을 더욱 확고히 세울 수 있게 해

주신 지도교수님이신 손 훈 교수님께 깊은 감사의 뜻을 전합니다 짧은 기간이었지만 교수님의

도움으로 많은 것을 배우고 느끼고 깨달을 수 있었습니다 또한 본 연구를 수행할 수 있도록 지

원해주신 국토해양부 (U-City 석 박사과정 지원사업)에 깊은 감사 드립니다

때로 힘들고 지칠 때 항상 저를 믿고 응원해 주시는 가족이 있었기에 지금의 글을 쓸 수 있었

습니다 또한 지금 연구실에서 함께 생활하고 있는 윤규형 민구형 철민이형 현석이형 진열이형

호민이형 병진이 선혜누나와 성대에 계신 창길이형 미국에 계신 의재형과 Debaditiya 에게도 감

사의 말씀을 전합니다 그리고 모두 나열할 수는 없지만 카이스트 생활의 활력소가 되어준 카이

스트 건설 및 환경공학과 선후배들 직원 분들과 SOS 멤버들에게도 감사를 표합니다

앞서 거론한 많은 이들의 도움이 없었다면 저 또한 지금 이 자리에 있을 수 없었을 것입니다

많이 부족하기에 더 노력하고 배우고자 했던 초심을 다시 한번 생각하며 더 큰 결실을 맺기 위해

앞으로 나아가겠습니다 도움을 주신 모든 분들께 이 감사의 글을 바칩니다

2011년 6월 카이스트에서 임형진 올림

- 38 -

CURRICULUM VITAE

Hyung Jin Lim

Research Assistant

The Civil and Environmental Engineering Department

Korea Advanced Institute of Science and Technology (KAIST)

291 Daehak-Ro Guseong-Dong Yuseong-Gu

Daejeon Republic of Korea 305-701

Phone (82)+42-350-3665 Fax (82)+42-350-3610

Email limhj87gmailcom

RESEARCH INTERESTS

Damage Detection Localization and Quantification on Composite and Metallic Structures Pattern Recognition Finite Element Modeling (FEM)

EDUCATION

2009 ndash 2011 MS Civil and Environmental Eng Korea Advanced Institute of Science and Technology (KAIST) Korea Rep

2006 ndash 2009 BS Civil and Environmental Eng Korea Advanced Institute of Science and Technology (KAIST) Korea Rep

JOURNAL PUBLICATIONS The corresponding author is underlined 1 Hyung Jin Lim Hoon Sohn ldquoReference-free damage detection localization and quantification

using in compositerdquo in preparation for Journal of the Acoustical Society of America 2011

2 Chul Min Yeum Hoon Sohn Jeong Bum Ihn Hyung Jin Lim ldquoInstantaneous delamination detection in a composite using a dual piezoelectric transducer networkrdquo Submitted to Compo-sites Science and Technology 2011

3 Yun-Kyu An Hyung Jin Lim Min Koo Kim Jin Yeol Yang Hoon Sohn Chang Guen Lee ldquoInvestigation of applicability of local reference-free damage detection techniques to in-situ

- 39 -

bridgesrdquo Submitted to ASCE Journal of Structural Engineering 2011

4 Hyung Jin Lim Min Koo Kim Hoon Sohn Chan Yik Park ldquoImpedance based damage detec-tion under varying temperature and loading conditionsrdquo Submitted to NDTampE International 2011

5 Hyung Jin Lim Hoon Sohn ldquoQuantitative estimation of transmitted and reflected lamb waves at discontinuityrdquo Journal of the Korean Society for Nondestructive Testing Vol30 No 4 pp1-8 2010

CONFERENCE PROCEEDINGS

The corresponding author is underlined 1 Hyung Jin Lim and Hoon Sohn ldquoReference-free damage detection localization and quantifi-

cation in compositesrdquo ASCE Engineering Mechanics Institute Conference Boston June 2-4 2011

2 Hyung Jin Lim and Hoon Sohn ldquoReference-free delamination quantification within multi-layer composite materialsrdquo KSNT Spring Conference May 26-27 Gyeong-Ju Korea 2011

3 Min Koo Kim Hyun Jin Lim Hoon Sohn Chan Yik Park ldquoImpedance-based bolt loosening detection under varying temperature and loadingrdquo Asian Pacific Workshop on Structural Health Monitoring Tokyo Japan November 30-December 2 2010

4 Hyung Jin Lim Chul Min Yeum and Hoon Sohn ldquoModeling of Impact-induced Delamination in a Multilayer Composite Platerdquo The 23rd KKCNN Symposium on Civil Engineering Taipei (Taiwan) November 13-15 2010

5 Yun-Kyu An Hyung Jin Lim and Hoon Sohn ldquoTemperature independent Delamination Detec-tion using data normalizationrdquo The 3rd International Conference on Multi-Functional Materi-als and Structures Chonbuk National Univ Jeonju Korea September 14-18 2010

6 Hoon Sohn Yun-Kyu An Chang Gil Lee Min Koo Kim Hyung Jin Lim ldquoApplication of online NDT techniques to decommissioned bridge testingrdquo the 5th European Workshop on Structural Health Monitoring Sorrento Italy June 29-July 02 2010

7 Min Koo Kim Hyung Jin Lim Hoon Sohn Chan Yik Park ldquoImpedance based damage diag-nosis of a complex composite aircraft wing under changing loading and temperature condi-tionsrdquo the 5th European Workshop on Structural Health Monitoring Sorrento Italy June 29-July 02 2010

8 Yun-Kyu An Min Koo Kim Hyung Jin Lim and Hoon Sohn ldquoValidations of reference-free crack detection techniques through a decommissioned bridge testrdquo 2010 COSEIK Annual Con-ference Jeju Korea April 8-9 2010

- 40 -

PARTICIPANT PROJECTS

1 Development of Smart Steel Members for Online Steel Structure Monitoring POSCO Corporation (Funded 41000000 KRW (41000 USD) for 010511 to 123111)

2 Development of Enabling Technologies for Structural Health Monitoring Korean Agency for Defense Development (Funded 240000000 KRW (240000 USD) out of total 3200000000 KRW (3200000 USD) for 060108-053111)

3 Development of On-board SHM Technologies for Composite Air Vehicles The Boeing Company (Funded 400000 USD (400000000 KRW) for 080108 to 073111)

EDUCATIONAL EXPERIENCES

Fall 2009

Teaching Assistant CE207 ldquoElementary Structural Engineering and Laboratoryrdquo KAIST 3 unit undergraduate course

EXPERIMENTAL EXPERIENTCES

2010 Monitoring of Real Aircraft Composite Wing Segment Korea Agency for Defense De-velopment (ADD) Daejeon Korea Real time monitoring of composite air vehicles fund-ed by Korea Agency for Defense Development (ADD)

2009-2011

Field Monitoring of Real Bridge Structures Gimpo (Ramp G bridge) amp Icheon (Sam-seung bridge) Korea US-Korea collaborative research for bridge monitoring testbeds

HONORS amp AWARDS

Scholarships for outstanding students granted KAIST (2007-2009) The three prominent students in the civil and environmental engineering department of KAIST were selected as recipients This scholarship provided 2000$ per year for three years

- 41 -

Page 20: 복합재료 손상의 무기저 진단 및 정량화 - KAISTssslab.kaist.ac.kr/article/pdf/Thesis_2.pdf · 1 Declaration of Ethical Conduct in Research: I, as a graduate student

Damage 1

Damage 2

20

Unit mm15

90 G value reduction

943 G value reduction

m-value = 248

m-value = 248

015 02 025 03 035

-2

-1

0

1

2

Time (ms)

Ampl

itude

Damage 1Damage 2

Fig 4 Two different FEM models which have different 119889119889 and 119907119907119889119889 (or G) This represents that the damage cannot be characterized uniquely based on the measured data available

Thus in this study the damage index (m-value) which can be uniquely defined based on

the available data is introduced The m-value is defined as the ratio of the physical length of

the damage and wavelength of the A0 Lamb wave mode in the damage area (120582120582119889119889)

119898119898 =119889119889120582120582119889119889

=119889119889119889119889119907119907119889119889

(3)

In many cases the transmitted and reflected waves are overlapped and it is hard to decom-

pose and calculate the arrival times of them Matching pursuit method is an alternative way

to accurately decompose of the overlapped waves and estimate the parameters of each wave

such as arrival time (Fig 5) Detail explanation of the matching pursuit method is covered in

section 36

015 02 025 03-2

-1

0

1

2

Time (ms)

Am

plitu

de

OriginalApproximated

(a)

015 02 025 03-2

-1

0

1

2

Time (ms)

Ampl

itude

Wave 1Wave 2

(b)

Fig 5 (a) Example of the overlapped signal with comparing approximated signal using the matching pursuit method (b) Decomposed signal using the matching pursuit method The matching pursuit method provides a conventional way to decompose and identify each wave

- 9 -

33 Damage detection algorithm (Level 1)

The damage diagnosis is accomplished by extracting the additional reflected A0 Lamb

wave modes from damage due to multiple reflections If A0 Lamb wave modes are decom-

posed perfectly it is trivial that one transmitted A0 Lamb wave mode is always captured in

the response signal If damage exists other waves in the response signal behind the transmit-

ted A0 Lamb wave mode are presented Since this approach does not rely on previously col-

lected data the diagnosis is independent from environmental variation such as temperature

changes

It is reported that the dispersion of Lamb waves cause wave-packets to spread out in space

and time as they propagate through a structure [20] Thus the reflected waves from the dam-

age have longer pulse duration than the transmitted wave due to their longer propagating dis-

tance In this study the matching pursuit method is modified to capture more spread wave

than the transmitted A0 Lamb wave mode Detail explanation of the modified matching pur-

suit method is described in latter section

Let 1199041199041 and 1199041199042 as the pulse duration (scale) of the transmitted and additional reflected A0

Lamb wave mode respectively

ldquoIf 1199041199042 gt 1199041199041 damage exists

Otherwise there is no damagerdquo

- 10 -

34 Damage quantification algorithm (Level 3)

As mentioned in the previous section damage is quantized by estimating 119889119889 and 119907119907119889119889 using

pitch-catch method The procedure to estimate 119889119889 and 119907119907119889119889 is

(i) Measure 119897119897 119905119905 and 119905119905119889119889

(ii) Assume the range of 119889119889 (0 lt 119889119889 lt 119897119897)

(iii) Calculate 119907119907119894119894 and 119907119907119889119889 with a constraints (0 lt 119907119907119889119889 lt 119907119907119894119894) (Eq 4 and 5)

(iv) Estimate m-value and find a feasible range of 119889119889 and 119907119907119889119889

119907119907119894119894 =119897119897 minus 119889119889119905119905 minus 119905119905119889119889

(4)

119907119907119889119889 =119889119889119905119905119889119889

(5)

35 Damage localization algorithm (Level 2amp3)

For damage localization 1198971198971 is estimated by measuring the arrival time from the excitation

PZT to the entrance of delamination (1199051199051) with additional pulse-echo method implementation

The procedure is similar with the previous damage quantification algorithm

(i) Measure 119897119897 119905119905 119905119905119889119889 and 1199051199051

(ii) Assume the range of 1198971198971 (0 lt 1198971198971 lt 119897119897)

(ii) Compute possible combinations of 1198971198972 119889119889 and 119907119907119889119889 with constraints

(0 lt 119907119907119889119889 lt 119907119907119894119894 119889119889 gt 0) (Eq 5 6 and 7)

- 11 -

1198971198972 = 1198971198971119905119905 minus 1199051199051 minus 119905119905119889119889

1199051199051 (6)

119889119889 = 119897119897 minus 1198971198971 minus 1198971198972 (7)

Since a PZT for pulse-echo method acts as both an actuator and a sensor the obtained sig-

nal contains both the excitation and the response signal Thus the reflected wave from the

entrance of damage can be buried into the tail of excitation signal and cause error on arrival

time of reflected wave from damage Even in this case damage can be localized and quan-

tized using the constraints When first wave packet reflected from the entrance of damage is

buried into the excitation signal the next wave packet reflected from the exit of damage can

be utilized alternatively If the algorithm regards the second reflected wave as the first re-

flected wave there are no combinations of estimated variable meeting the constraints Then

the algorithm repeats the process by regarding the wave as the second reflected wave the

constraints can select proper combination of the estimated variables

The estimated values have range bounded by the constraints When additional information

of 119907119907119894119894 is utilized the damage can be characterized more accurately In this study three meth-

ods are proposed and verified as information level of 119907119907119894119894

Method 1 Solely based on the current data measurement (Reference-free)

Method 2 Limit the possible range of 119907119907119894119894 (Prior knowledge of the system)

Method 3 Measure 119907119907119894119894 (Additional data need to be collected)

Method 1 described in this section can be utilized without any information of 119907119907119894119894 Meth-

od 2 produces more accurate estimation results using prior knowledge of the system such as

- 12 -

119907119907119894119894 variation under temperature changes Method 3 gives unique estimated value using meas-

ured 119907119907119894119894 from other path in the same time In section 4 and 5 damage is characterized using

these three methods and the results are compared each another

36 Matching pursuit method considering the dispersion effect of Lamb

wave

Matching pursuit method has been developed for signal compression approximation and

characterization of obtained signal and function [21-23] The key concept of the matching

pursuit method is that any signal (function) can be decomposed into a linear expansion of

signals selected from a pool of basis signals (functions) called lsquodictionaryrsquo

Let 119891119891 be a function to be analyzed then 119891119891 can be approximated a set of functions select-

ed from the dictionary 119863119863 in order to best match This is done by successive approximations

with orthogonal projections Let1198921198921205741205740 isin 119863119863 the function 119891119891 can be decomposed into

119891119891 = lang1198911198911198921198921205741205740rang1198921198921205741205740 + 119877119877119891119891 (8)

where 119877119877119891119891 is the residual function after approximating 119891119891 using 1198921198921205741205740 1198921198921205741205740 is orthogonal to

119877119877119891119891 hence

1198911198912 = lang1198911198911198921198921205741205740rang2

+ 1198771198771198911198912 (9)

To minimize 119877119877119891119891 1198921198921205741205740 is chosen to maximize lang1198911198911198921198921205741205740rang The matching pursuit method is

an iterative algorithm that decomposes the residue 119877119877119891119891 by projecting it on a function of 119863119863

that matches 119877119877119891119891 almost at best as it was done for 119891119891 The residue 119877119877119899119899119891119891 is decomposed into

119877119877119899119899119891119891 = lang119877119877119899119899119891119891119892119892120574120574119899119899rang119892119892120574120574119899119899 + 119877119877119899119899+1119891119891 (10)

- 13 -

the decomposition is carried up to the order 119898119898 119891119891 can be expressed as

119891119891 = lang119877119877119899119899119891119891119892119892120574120574119899119899rang119892119892120574120574119899119899

119898119898minus1

119899119899=0

+ 119877119877119898119898119891119891 (11)

The original function 119891119891 is decomposed into a sum of dictionary elements that are chosen to

best match its residues

In this study the matching pursuit method is utilized to accurately decompose of the over-

lapped Gaussian pulses and estimate the parameters of the pulse which used for damage

characterization The 119894119894119905119905ℎ Gaussian pulse in the dictionary for approximating the obtained

signal is expressed as

119892119892120574120574119894119894 = 119891119891(119860119860119894119894 119906119906119894119894 119904119904119894119894 119889119889119894119894120601120601119894119894) (12)

where 119860119860119894119894119906119906119894119894119904119904119894119894119889119889119894119894 and 120601120601119894119894 represent amplitude arrival time scale (pulse duration) excita-

tion(central) frequency and phase of the Gaussian pulse respectively The parameters of the

pulse are estimated using a nonlinear regression based optimization method (Gauss-Newton

method) based on an initial guess of the parameters Detail procedure for estimating the pa-

rameters of the pulse is expressed in Hong et al [23]

Among the estimated parameters of the pulses pulse duration (scale) represents the disper-

sion effect If a wave-packet is spread out the estimated pulse duration is increased Using

this feature the matching pursuit method is modified to capture waves which have longer

pulse duration than transmitted A0 Lamb wave mode If the estimate pulse duration of a wave

is less than the transmitted wave the modified matching pursuit method regards it as error

- 14 -

CHAPTER 4 NUMERICAL SIMULATION

41 Composite modeling method

Composite materials can be modeled from nano to macro level depends on its engineering

purpose In general four methods are utilized for modeling of composite materials First

composite is modeled as a homogenous material expressed in terms of effective material

properties [24] The effective material properties are estimated using the modulus and the

volume fraction of each constituent theoretically (Rule of mixture) and measured experimen-

tally Second a model is established that each laminate which have own material properties

are totally bonded together In this model the stacking sequence and direction of each lami-

nates are considered additionally [25-27] Third method for composite modeling in micro

level called micromechanics [28-30] This model considers the interaction between fibers and

matrix in laminate Last multiscale modeling which the scope of modeling is extended to

molecular or quantum level of constituents [31-33] This model starts from molecular me-

chanics or molecular dynamics then the scale of model leads to the nano micro and macro

level by incorporating small scale models As the scale of models is smaller it provides more

logical way than other large scale of models while the computational cost is more expensive

The scope of this study is focused on the dynamic behavior of composite structure the beam

or plate model was utilized for the numerical simulation

- 15 -

42 Damage modeling method

Generally three damage models have been developed and utilized in the real applications

First the stiffness reduction model called as the equivalent element model was introduced

[17 34-35] This model is simply realized by replacing the region around damage area by a

region with adjusted material properties (Youngrsquos modulus shear modulus density etc)

globally Next the duplicated node model is done by disconnecting the element connectivity

assuming there is no interaction among the delaminated surface This model can be utilized

when the location and the extension of damage is important issue [34-36] Last the contact

mechanics model using the principles of contact mechanics the two interfacial forces (con-

tact resistance and sliding resistance) in the delaminated surface [37-38] In this model the

delaminated surfaces are modeled as nonlinear functions of the interlaminar normal strain

and the interlaminar shear strain This model is computationally expensive while provides

more logical way than other models There is however no universally accepted model for

general impact-induced damage in composites In the case of impact-induced damage it is

impossible to predict or estimate the location and number of delaminated layers Thus from

the uncertainty of impact-induced damage the equivalent element model can be considered

as the effective damage model

- 16 -

43 Numerical simulation description

TABLE I Material property of simulation model and test specimen

Tensile Strength 657 MPa Tensile Modulus 59 GPa Shear Strength 804 MPa Shear Modulus 24 Gpa Specific Gravity 16

PZT A

PZT C

PZT BDamage

3

250 125 125

500 Unit mm

Fig 6 Description of the numerical simulation model

The feasibility of the proposed damage characterization technique is verified by 2-D FEM

simulation using MSC PATRAN and NASTRAN Fig 6 presents the dimension PZTs and

damage location of the simulation model In the model 05mm x 05mm square shell element

was used for modeling of the A0 mode whose flexural wave characteristic The element size

was determined by considering the wavelength of the propagation Lamb wave modes The

material properties of the simulation model and damage cases are presented in Table I and

Table II respectively

PZT A and C act as both actuator and sensor for collecting pulse-echo signal They were

installed on the plate which two of them are both-side collocated for the A0 Lamb wave

- 17 -

mode decomposition [39-40] PZT B was only used as a sensor for collecting pitch-catch sig-

nal For simulating the excitation and sensing of PZTs the pin-force model was adopted that

an input force apples to the end nodes of excitation PZT and averaging element strains corre-

sponding sensing area A Gaussian pulse (tone-burst) 50 kHz central frequency was excited

as an input signal

TABLE II Damage cases for numerical simulation

119949119949120783120783 (mm) 119941119941 (mm) G value reduction ()

Case 1 650 200 90 darr

Case 2 675 150 90 darr

Case 3 675 150 93 darr

Case 4 900 150 90 darr

- 18 -

44 Numerical simulation results

015 02 025 03 035

-2

0

2

Time (ms)

Am

plitu

de

IntactDamage

Reflected wave from damage

(a) Pitch-catch signal

015 02 025 03 035

-2

0

2

Time (ms)

Am

plitu

de

IntactDamage

Reflected waves from damage

(b) Pulse-echo signal Fig 7 Comparison of A0 Lamb wave mode signals from numerical simulation between intact and damage cases Time delay attenuation and multiple reflections due to damage simulated successfully using the damage modeling method described in the previous section

Fig 7 shows that comparison of A0 Lamb wave mode signals between intact and one of

damage cases in the simulation The features of A0 Lamb wave mode passing through dam-

age area described in section 2 are successfully simulated Thus the equivalent element

method is proper for modeling damage in composites In the damage cases the additional

waves reflected from damage is observed in both pitch-catch and pulse echo signal

Table III presents the result of the proposed damage characterization algorithm using the

numerical simulation cases shown in Table II In lsquoMethod 2rsquo 119907119907119894119894 variation is assumed to plusmn

5 of the normal value (1190~1315ms) The normal value of 119907119907119894119894 for lsquoMethod 3rsquo is meas-

ured as 1253 ms In each case the damage is characterized using the proposed technique

Note that as the information of 119907119907119894119894 is specified the estimated value is closer to the exact

value

- 19 -

TABLE III Damage characterization for the numerical simulation models

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

Case 1

Exact 650 200

Method 1 248 50~70 79~414 160~834

Method 2 248 603~667 135~241 271~486

Method 3 248 635 188 379

Case 2

Exact 675 150

Method 1 193 60~75 01~251 37~651

Method 2 193 618~682 114~222 295~575

Method 3 193 65 168 435

Case 3

Exact 675 150

Method 1 233 60~75 66~302 140~649

Method 2 233 668~738 85~196 182~419

Method 3 233 703 140 301

Case 4

Exact 900 150

Method 1 204 75~100 48~349 118~856

Method 2 204 860~951 107~216 264~530

Method 3 204 906 162 397

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1253 ms 1190~1315 ms Measured 119907119907119894119894 1253 ms

- 20 -

CHAPTER 5 EXPERIMENTAL VERIFICATION

51 Overview

In this section experimental study is investigated so that the proposed damage characteri-

zation technique Overall experimental setup and test specimen are shown in Fig 8 and 9

The experimental setup consists of test specimens data acquisition (DAQ) system and self-

sensing circuit Self-sensing circuit was installed for pulse-echo measurement using PZTs

[40-41]

PZT A(PZT C) PZT B

500

500

140

75

250

Teflon Inserted Damage (d =25)

Unit mm

Thickness = 3

PZT A(PZT C) PZT B

500

500

150

90

250

Impact Location

Unit mm

Thickness = 3

(a) Teflon inserted damage specimen

(b) Impact-induced damage specimen

Fig 8 Test specimens for the experimental verification In the Teflon inserted damage Teflon tapes with 25mm diameter were inserted at the center of the specimen The impact-induced damage was introduced to the specimen using the impact test equipment

- 21 -

511 Specimen

Two carbon composite plates with a dimension of 500times500times3mm3 were manufactured by

Nexcomes Inc by stacking 12 layers of CF3327 3K prepreg (woven fabrics type) manufac-

tured by Hankuk Fiber Inc The material properties of the specimen are presented in Table I

One of the specimens Teflon tapes with 25mm diameter was inserted in every 3 layers at the

center of the specimen to mimic the damage The other specimen is utilized for the character-

ization of real impact-induced damage

512 Data acquisition (DAQ) system and PZT

The data acquisition (DAQ) system consists of an arbitrary waveform generator (AWG) a

high speed signal digitizer (DIG) and two multiplexers (MUX) Using the 16-bits AWG

tone-burst input signal with a plusmn10 peak-to-peak voltage were generated The response signal

was measured by the 14-bit DIG For the signal-to-noise ratio reduction both pitch-catch and

pulse-echo signals were measured 150 times and averaged in the time domain 3 identical

circular PZTs of 10mm diameter manufactured by APC International Ltd were installed on

the specimen A Gaussian pulse (tone-burst) 50 kHz central frequency was excited as an in-

put signal The excitation frequency was chosen in order to avoid overlapping of the S0 and

A0 Lamb wave modes and generating of the higher modes (S1 A1hellip)

- 22 -

PZT Actuator

PZT Sensor

AWG

DIG

DAQ System Specimen

SensingActuatingMUX

1

23

4

1 Sending signal to MUX

4 Sending signal to DIG

Self-SensingCircuit

18 Unit mm

10

(a)

(b)

Fig 9 Overall experimental setup (a) Data acquisition (DAQ) system and (b) Installed circu-lar PZT transducer on the specimen with its dimension

513 Impact test

The impact-induced damage was introduced to the specimen using the impact test equip-

ment presented in Fig 10 A 5kg mass with 10mm tip was dropped from 015m height onto

the same location of the specimen The signals were obtained at each number of impacts

Since the reflected wave is not detected at the first impact case the data obtained from the

second and third impact were utilized for the analysis

10

Unit mm (a)

(b)

Fig 10 (a) Impact test equipment for introducing impact-induced damage and (b) Tip of the impact tester with 10mm diameter

- 23 -

514 Temperature effect test

In reality structures are subjected to changing environmental conditions such as tempera-

ture that affect measured signal and cause false-alarm Thus the robustness of the proposed

damage characterization technique under temperature variation in investigated (Fig 11) The

specimens were placed inside a temperature chamber and a thermocouple was installed on

the specimen to measure temperature of surface of the specimen The signal was obtained

under three temperature cases (0 20 and 50degC) and the humidity was kept at 30

Temperature Chamber

Test Specimen

Fig 11 Test specimen in temperature chamber

- 24 -

52 Teflon inserted damage

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Reflected wave from damage

(a)

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Transmitted WaveReflected Wave

Detected reflected wave

(b)

Fig 12 A0 Lamb wave mode obtained using pitch-catch method from the Teflon inserted damage specimen (a) The reflected wave from damage is shown (b) The Teflon inserted damage is detected by capturing the reflected wave

Fig 12 shows the obtained pitch-catch signal from the Teflon inserted damage specimen

and damage diagnosis result The Teflon inserted damage is detected successfully by captur-

ing the reflected wave Table IV presents the result of the Teflon inserted damage characteri-

zation 119907119907119894119894 is measured as 1213ms and the variation is assumed to plusmn 5 of the normal val-

ue (1152~1273ms) The results indicate that the proposed damage characterization approach

works well in real application

TABLE IV Damage characterization for the Teflon inserted damage specimen

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

Exact 635 25

Method 1 123 60~75 73~338 298~1220

Method 2 123 619~662 229~310 934~1188

Method 3 123 631 284 1161

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1213 ms

- 25 -

53 Impact-induced damage

11 115 12 125 13-4

-2

0

2

4

Time (ms)

Vol

tage

(mV

)

IntactImpact2Impact3

Reflected wave from damage

(a) Pitch-catch signal

11 115 12 125 13 135

-05

0

05

1

Time (ms)

Volta

ge (m

V)

IntactImpact2Impact3

Boundary reflection

Reflected wave from damage

(b) Pulse-echo signal

Fig 13 The signal obtained from the test specimen subjected to impact The impact was in-troduced at the same location of the specimen and the signal was obtained after each impacts As the number of impact is increased the transmitted A0 Lamb wave mode is delayed and attenuated

The obtained signals from impact-induced damage specimen are presented in Fig 13 It is

observed that as the number of impact is increased the transmitted A0 Lamb wave mode is

more delayed because the wave velocity in the damage area is reduced Fig 14 shows the

false-alarm test result whether the proposed damage detection algorithm works even in pris-

tine condition of the test specimen It is shown that no false alarm is occurred because no ad-

ditional reflected wave is captured

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Transmitted Wave

(a) Transmitted wave

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Reflected Wave

No reflected wave

(b) Reflected wave

Fig 14 False alarm test for the proposed damage detection algorithm in pristine condition of the specimen Reflected wave from damage is not captured false alarm is not occurred

- 26 -

TABLE V Damage characterization for the impact-induced damage specimen

of Impacts m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

2

(1470 J)

Method 1 080 65~90 75~246 470~1157

Method 2 080 801~885 114~245 711~1134

Method 3 080 843 179 1122

3

(2205 J)

Method 1 148 50~100 06~380 21~1134

Method 2 148 758~838 249~368 843~1125

Method 3 148 798 308 1044

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1213 ms

Table V presents the characterization results of the impact-induced damage case Here the

measured 119907119907119894119894 and the assumed variation are same as the artificial damage case It is observed

that as the number of impact is increased 119889119889 is increased while 119907119907119889119889 is reduced Thus it is

validated that the proposed damage characterization approach estimates the increment of the

damage well

54 Temperature effect test

541 Teflon inserted damage

11 115 12 125 13-10

-5

0

5

10

Time (ms)

Vol

tage

(mV

)

0 oC20 oC50 oC

(a) Pitch-catch signal

11 115 12 125 13 135

-1

0

1

2

3

Time(ms)

Volta

ge (m

V)

0 oC20 oC50 oC

(b) Pulse-echo signal

Fig 15 Signals obtained from the Teflon inserted damage specimen under temperature varia-tion Time delay and amplitude variation are occurred

Fig 15 shows the obtained signals from the Teflon inserted damage specimen subjected to

- 27 -

temperature variation Time delay and amplitude variations are occurred there is no addition-

al damage is introduced The analyzed results of the obtained signals are reported in Table VI

The measured 119907119907119894119894 is 1247 ms at 0 degC 1213 ms at 20 degC and 1168 ms at 50 degC respectively

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms (1152~1273 ms)

The estimated results are also changed under temperature variations the fluctuation of esti-

mated range and exact value is small

TABLE VI Temperature effect test for the Teflon inserted damage specimen

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

0 degC

Exact 635 25

Method 1 113 60~70 79~268 352~1291

Method 2 113 560~619 233~344 1034~1172

Method 3 113 606 257 1141

20 degC

Exact 635 25

Method 1 123 60~75 73~338 298~1220

Method 2 123 619~662 229~310 934~1188

Method 3 123 631 284 1161

50 degC

Exact 635 25

Method 1 137 60~75 72~328 263~1228

Method 2 137 622~687 183~299 666~1086

Method 3 137 642 264 960

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1247 ms at 0 degC 1213 ms at 20 degC and 1168 ms at 50 degC

- 28 -

542 Impact-induced damage

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Vol

tage

(mV

)

Transmitted WaveReflected Wave

No reflected wave

(a) 0 degC

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Transmitted WaveReflected Wave

No reflected wave

(b) 50 degC

Fig 16 False alarm test under temperature variations It is shown that proposed damage char-acterization algorithm is robust under temperature variations

The false alarm test in pristine condition of the impact-induced damage specimen under

temperature variation was investigated (Fig 16) It is validated that the proposed damage di-

agnosis algorithm is robust under temperature variations Fig 17 shows the obtained signals

from the impact-induced damage (impacted 3 times) specimen under temperature variation

The analyzed results of the obtained signals are reported in Table VII The measured 119907119907119894119894 in

each temperature and the assumed variation are same as the artificial damage case

11 115 12 125 13-5

0

5

Time (ms)

Volta

ge (m

V)

0 oC20 oC50 oC

(a) Pitch-catch signal

11 115 12 125 13 135

-1

0

1

2

3

Time(ms)

Vol

tage

(mV

)

0 oC20 oC50 oC

(b) Pulse-echo signal

Fig 17 Signals obtained from impact-induced damage specimen (Impact 3) under tempera-ture variation

- 29 -

TABLE VII Temperature effect test for the impact induced damage specimen (Impact 3 case)

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

0 degC

Method 1 135 75~90 72~310 267~1148

Method 2 135 705~779 264~381 978~1126

Method 3 135 763 289 1071

20 degC

Method 1 148 50~100 06~380 21~1134

Method 2 148 758~838 249~368 843~1125

Method 3 148 798 308 1044

50 degC

Method 1 153 75~100 53~415 174~1227

Method 2 153 804~889 214~336 702~1103

Method 3 153 815 320 1050

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1247 ms at 0 degC 1213 ms at 20 degC and 1168 ms at 50 degC

- 30 -

CHAPTER 6 CONCLUDING REMARKS

61 Executive summary

In this study the reference-free damage detection localization and quantification tech-

nique in a multi layered composite plate is proposed Additional reflected A0 Lamb wave

mode from the damage due to multiple reflections in damage area is used as a feature for

characterizing the damage First damage is characterized in terms of a damage index (m-

value) defined as the ratio of the physical length of the damage to the wavelength of the A0

Lamb wave mode Then the range of possible damage location length and severity are esti-

mated with constraints This procedure can be accomplished without prior baseline data To

realize the proposed technique the matching pursuit technique is modified for reduction of

the error by considering the dispersive characteristic of Lamb wave The matching pursuit

method estimates the arrival time of each wave packets accurately even two wave packets are

overlapped The numerical simulations and experimental tests including Teflon tape inserted

and impact-induced damage are investigated for the feasibility of the proposed technique

Moreover the effectiveness of the proposed technique under changing temperature is exper-

imentally verified

- 31 -

62 Suggestion for further study

The limitations of this study and the possible improvement and further studies are listed as

follow

1 Measurement of impact-induced damage length using conventional NDT technique

Additional experimental verification for impact-induced damage characterization is

needed by measuring the real length of damage using conventional NDT technique

such as ultrasonic scan or infrared camera

2 Limitation of both side collocated PZT installation In this study excitation PZTs are

installed both side and collocated for decomposing A0 Lamb wave mode However it

can produce decomposition error due to imperfection of PZT location and their installa-

tion Moreover in many cases assess to both surface of structures is not allowed such

as aircraft applications To improve this limitation a reference-free damage characteri-

zation using single side PZTs will be considered

3 Complex structure application In this study damage is characterized in simple plate-

like structure When the monitored structure becomes complex the response signals al-

so becomes complex due to its geometry and boundary Thus the investigations for

complex structure application such as stiffener spar and curvature are needed

- 32 -

REFERENCES

[1] Cawley P and Alleyne D (1996) ldquoThe use of Lamb waves for the long range inspection of large structuresrdquo Ultrasonics 34 pp287-290

[2] Ghosh T Kundu T and Karpur P (1998) ldquoEfficient use of Lamb waves for detecting defects in large platesrdquo Ultrasonics 36 pp791-801

[3] Guirguitui V (2003) ldquoLamb wave generation with piezoelectric wafer active sensors for structural health monitoringrdquo SPIE Conference on Smart Structures and Materials San Diego CA 5391 pp111-422

[4] Ragvahan A and Cesnik CE S (2004) ldquoModeling of piezoelectric-based Lamb wave generation and sensing for structural health monitoringrdquo SPIE Conference on Smart Struc-tures and Materials San Diego CA 5391 pp419-430

[5] Rose JL Pilarski A and Dirit JJ (1993) ldquoAn approach to guided wave mode selection for inspection of laminated platerdquo Journal of Reinforced Plastics and Composites 12(5) pp536-544

[6] Guo N and Cawley P (1993) ldquoThe interaction of Lamb waves with delaminations in composite laminatesrdquo Journal of Acoustical Society of America 94(4) pp 2240-2246

[7] Valdez SHD and Soutis C (2002) ldquoReal-time nondestructive evaluation of fiber com-posite laminates using low-frequency Lamb wavesrdquo Journal of Acoustical Society of Amer-ica 111(5) pp 2026-2033

[8] Yang M and Qiao P (2005) ldquoModeling and experimental detection of damage in virous materials using the pulse-echo method and piezoelectric sensoractuatorsrdquo Smart Materials and Structures 14 pp1083-1100

[9] Lestari W and Qiao P (2005) ldquoApplication of wave propagation analysis for damage identification in composite laminated beamsrdquo Journal of Composite Materials 39(22) pp1967-1984

[10] Kessler S S Spearing S M and Soutis C (2002) ldquoDamage detection in composite ma-terials using Lamb wave methodsrdquo Smart Materials and Structures 11 pp269-278

[11] Sohn H Park G Wail J R Limback N P and Farrar C R (2004) ldquoWavelet-based active sensing for delamination detection in composite structuresrdquo Smart Materials and Structures 13 pp 153-160

[12] Petculescu G Krishnaswamy S and Achenback J D (2008) ldquoGroup delay measure-ments using modally selective Lamb wave transducers for detection and sizing of delamina-tions in compositesrdquo Smart Materials and Structures 17

[13] Sohn H Park H W Law K H and Farrar C R (2007) ldquoDamage detection in compo-

- 33 -

site plates by using an enhanced time reversal methodrdquo Journal of Aerospace Engineering (ASCE) pp 141-151

[14] Sohn H (2007) ldquoEffects of environmental and operational variability on structural health monitoringrdquo Philosophical Transactions of the Royal Society Ardquo 365 pp539-560

[15] Rytter A (1993) ldquoVibration based inspection of civil engineering structuresrdquo PhD The-sis Department of Building Technology and Structural Engineering University of Aalborg Aalborg Denmark

[16] Hayashi T and Kawashima K (2002) ldquoMultiple reflections of Lamb waves at a delami-nationrdquo Ultrasonics 40 pp193-197

[17] Wang C H and Rose L R F (2003) ldquoWave reflection and transmission in beams con-taining delamination and inhomogeneityrdquo Journal of Sound and Vibration 264 pp 851-872

[18] Ramadas C Balasubramaniam K Joshi M and Krishnamurthy C V (2009) ldquoInterac-tion of the primary anti-symmetric Lamb mode (Ao) with symmetric delaminations numer-ical and experimental studiesrdquo Smart Materials and Structures 18

[19] Ramadas C Balasubramaniam K Joshi M and Krishnamurthy C V (2010) ldquoInterac-tion of guided Lamb waves with an asymmetrically located delamination in a laminated composite platerdquo Smart Materials and Structures 19

[20] Wilcox P Lowe M and Cawely P (2001) ldquoThe effect of dispersion on long-range in-spection using ultrasonic guided wavesrdquo NDTampE International 34 pp1-9

[21] Mallat S G and Zhang Z (1993) ldquoMatching pursuits with time-frequency dictionariesrdquo IEEE Tansactions on Signal Processing 41(12) pp3397-3415

[22] Zhang G Zhang S and Wang Y (2000) ldquoApplication of adaptive time-frequency de-composition in ultrasonic NDE of highly scattering materialsrdquo Ultrasonics 38 pp 961-964

[23] Hong J C Sun K H and Kim Y Y (2005) ldquoThe matching pursuit approach based on the modulated Gaussian pulse for efficient guided-wave damage inspectionrdquo Smart Mate-rials and Structures 14 pp548-560

[24] Herakovich C T (1997) Mechanics of Fibrous Composites Wiley 480 pages

[25] Reddy J N (1985) ldquoA review of the literature on finite-element modeling of laminated composite platesrdquo The Shock and Vibration Digest 17(4) pp 3-8

[26] Reddy J N (2003) Mechanics of Laminated Composite Plates and Shells Theory and Analysis CRC Press 856 pages

[27] Palacz M Krawczuk M and Ostachowicz W (2005) ldquoThe spectral finite element mod-el for analysis of flexure-shear coupled wave propagation Part 1 laminated multilayer

- 34 -

composite beamrdquo Composite Structures 68 pp37-44

[28] Zhang Y C and Harding J (1990) ldquoA numerical micromechanics analysis of the me-chanical properties of a plain weave compositerdquo Computer amp Structures 36(5) pp 839-844

[29] Nemat-Nasser S and Hori M (1999) Micromechanics Overall Properties of Heteroge-neous Materials Elsevier 810 pages

[30] Scida D Aboura Z Benzeggagh M L and Bocherens E (1999) ldquoA micromechanics model for 3D elasticity and failure of woven-fibre composite materialrdquo Composites Science and Technology 59 pp 505-517

[31] Thomas Y and Wu X H (1997) ldquoA multiscale finite element method for elliptic prob-lems in composite materials and porous mediardquo Journal of Computational Physics 134 pp 169-189

[32] Lee K H Moorthy S and Ghosh S (1999) ldquoMultiple scale computational model for damage in composite materialsrdquo Computer Methods in Applied Mechanics and Engineer-ing 172 pp 175-201

[33] Namilae S and Chandra N (2005) ldquoMultiscale model to study the effect of interfaces in carbon nanotube-based compositesrdquo Journal of Engineering Materials and Technology (ASME) 127 pp222-232

[34] Adams D E (2007) Health monitoring of Structural Materials and Components Wiley 460 pages

[35] Gopalakrishnan S (2009) ldquoModeling Aspects in Finite Elementrdquo Encyclopedia of Struc-tural Health Monitoring 2 pp791-809

[36] Palacz M Krawczuk M and Ostachowicz W (2005) ldquoThe spectral finite element mod-el for analysis of flexure-shear coupled wave propagation Part 2 Deminated multilayer composite beamrdquo Composite Structures 68 pp45-51

[37] Baik J M and Thopson R B (1984) ldquoUltrasonic scattering from imperfect interfaces A quasi-static modelrdquo Journal of Nondestructive Evaluation 4(3) pp177-196

[38] Luo H and Hanagud S (2000) ldquoDynamics of delamincated beamsrdquo International Jour-nal of Solids and Structures 37 pp1501-1519

[39] Kim S B and Sohn H (2007) ldquoInstantaneous reference-free crack detection based on polarization characteristics of piezoelectric materialsrdquo Smart Materials and Structures 16 pp2375-2387

[40] An Y K and Sohn H (2010) ldquoInstantaneous crack detection under varying temperature and static loading conditionsrdquo The Special Issue of Structural Control and Health Monitor-ing 17(7) pp 730-741

- 35 -

[41] Lee S J and Sohn H (2006) ldquoActive self-sensing scheme development for structural health monitoringrdquo Smart Materials and Structures 15(6) pp 1734-1746

- 36 -

요 약 문

복합재료 손상의 무기저 진단 및 정량화

복합재료는 재료의 경량성 내구성 고강도의 특성으로 인하여 항공기에서부터

교량에까지 널리 사용되고 있다 하지만 외부의 충격으로 인한 복합재료 내부의

손상은 재료의 강도를 감소시켜 구조물의 안정성을 저하시키는 요인이 된다 하지만

복합재료의 손상은 재료 내부에서 발생하여 육안으로는 식별하기도 힘들며 표면에서

보이는 것만으로 내부의 손상 정도를 판단하기는 불가능하다는 단점을 가지고 있다

본 연구에서는 손상 내부에서 반사로 인해 발생하는 파를 이용하여 기저자료를

사용하지 않고 특정 시점에서 얻은 자료만으로 복합재료의 손상을 진단하고 정량화

할 수 있는 기법을 제안한다 제안된 기법의 적용성을 검증하기 위하여 유한요소법을

이용한 수치해석을 수행하였고 Teflon tape 이 삽입된 시편과 실제 충격하중으로

인한 손상을 입은 시편을 이용한 실험을 진행하였다 수치해석 및 실험 결과 제안된

기법이 복합재료의 손상을 성공적으로 진단 및 정량화할 수 있음을 확인하였다 또한

다양한 외부 온도 환경에서 제안된 기법의 환경영향 실험을 수행하였고 다양한

환경에서 손상의 진단 및 정량화가 성공적으로 이루어짐을 검증하였다

핵심어 무기저 손상진단 및 정량화 압전센서 복합재료 유한요소법 Matching pursuit

method

- 37 -

감 사 의 글

기대와 두려움을 안고 시작한 석사과정이 어느덧 2 년 이라는 세월 속에서 추억으로 자리잡아

가고 있습니다 학부생 기간을 포함한 지난 3 년 동안 SHM 이라는 새로운 분야를 접하고 연구를

수행 할 수 있도록 도와주신 고마운 분들이 있어 그 감사를 표하고자 합니다

우선 저의 많은 부족한 점을 일깨워 주시고 연구에 대한 뜻을 더욱 확고히 세울 수 있게 해

주신 지도교수님이신 손 훈 교수님께 깊은 감사의 뜻을 전합니다 짧은 기간이었지만 교수님의

도움으로 많은 것을 배우고 느끼고 깨달을 수 있었습니다 또한 본 연구를 수행할 수 있도록 지

원해주신 국토해양부 (U-City 석 박사과정 지원사업)에 깊은 감사 드립니다

때로 힘들고 지칠 때 항상 저를 믿고 응원해 주시는 가족이 있었기에 지금의 글을 쓸 수 있었

습니다 또한 지금 연구실에서 함께 생활하고 있는 윤규형 민구형 철민이형 현석이형 진열이형

호민이형 병진이 선혜누나와 성대에 계신 창길이형 미국에 계신 의재형과 Debaditiya 에게도 감

사의 말씀을 전합니다 그리고 모두 나열할 수는 없지만 카이스트 생활의 활력소가 되어준 카이

스트 건설 및 환경공학과 선후배들 직원 분들과 SOS 멤버들에게도 감사를 표합니다

앞서 거론한 많은 이들의 도움이 없었다면 저 또한 지금 이 자리에 있을 수 없었을 것입니다

많이 부족하기에 더 노력하고 배우고자 했던 초심을 다시 한번 생각하며 더 큰 결실을 맺기 위해

앞으로 나아가겠습니다 도움을 주신 모든 분들께 이 감사의 글을 바칩니다

2011년 6월 카이스트에서 임형진 올림

- 38 -

CURRICULUM VITAE

Hyung Jin Lim

Research Assistant

The Civil and Environmental Engineering Department

Korea Advanced Institute of Science and Technology (KAIST)

291 Daehak-Ro Guseong-Dong Yuseong-Gu

Daejeon Republic of Korea 305-701

Phone (82)+42-350-3665 Fax (82)+42-350-3610

Email limhj87gmailcom

RESEARCH INTERESTS

Damage Detection Localization and Quantification on Composite and Metallic Structures Pattern Recognition Finite Element Modeling (FEM)

EDUCATION

2009 ndash 2011 MS Civil and Environmental Eng Korea Advanced Institute of Science and Technology (KAIST) Korea Rep

2006 ndash 2009 BS Civil and Environmental Eng Korea Advanced Institute of Science and Technology (KAIST) Korea Rep

JOURNAL PUBLICATIONS The corresponding author is underlined 1 Hyung Jin Lim Hoon Sohn ldquoReference-free damage detection localization and quantification

using in compositerdquo in preparation for Journal of the Acoustical Society of America 2011

2 Chul Min Yeum Hoon Sohn Jeong Bum Ihn Hyung Jin Lim ldquoInstantaneous delamination detection in a composite using a dual piezoelectric transducer networkrdquo Submitted to Compo-sites Science and Technology 2011

3 Yun-Kyu An Hyung Jin Lim Min Koo Kim Jin Yeol Yang Hoon Sohn Chang Guen Lee ldquoInvestigation of applicability of local reference-free damage detection techniques to in-situ

- 39 -

bridgesrdquo Submitted to ASCE Journal of Structural Engineering 2011

4 Hyung Jin Lim Min Koo Kim Hoon Sohn Chan Yik Park ldquoImpedance based damage detec-tion under varying temperature and loading conditionsrdquo Submitted to NDTampE International 2011

5 Hyung Jin Lim Hoon Sohn ldquoQuantitative estimation of transmitted and reflected lamb waves at discontinuityrdquo Journal of the Korean Society for Nondestructive Testing Vol30 No 4 pp1-8 2010

CONFERENCE PROCEEDINGS

The corresponding author is underlined 1 Hyung Jin Lim and Hoon Sohn ldquoReference-free damage detection localization and quantifi-

cation in compositesrdquo ASCE Engineering Mechanics Institute Conference Boston June 2-4 2011

2 Hyung Jin Lim and Hoon Sohn ldquoReference-free delamination quantification within multi-layer composite materialsrdquo KSNT Spring Conference May 26-27 Gyeong-Ju Korea 2011

3 Min Koo Kim Hyun Jin Lim Hoon Sohn Chan Yik Park ldquoImpedance-based bolt loosening detection under varying temperature and loadingrdquo Asian Pacific Workshop on Structural Health Monitoring Tokyo Japan November 30-December 2 2010

4 Hyung Jin Lim Chul Min Yeum and Hoon Sohn ldquoModeling of Impact-induced Delamination in a Multilayer Composite Platerdquo The 23rd KKCNN Symposium on Civil Engineering Taipei (Taiwan) November 13-15 2010

5 Yun-Kyu An Hyung Jin Lim and Hoon Sohn ldquoTemperature independent Delamination Detec-tion using data normalizationrdquo The 3rd International Conference on Multi-Functional Materi-als and Structures Chonbuk National Univ Jeonju Korea September 14-18 2010

6 Hoon Sohn Yun-Kyu An Chang Gil Lee Min Koo Kim Hyung Jin Lim ldquoApplication of online NDT techniques to decommissioned bridge testingrdquo the 5th European Workshop on Structural Health Monitoring Sorrento Italy June 29-July 02 2010

7 Min Koo Kim Hyung Jin Lim Hoon Sohn Chan Yik Park ldquoImpedance based damage diag-nosis of a complex composite aircraft wing under changing loading and temperature condi-tionsrdquo the 5th European Workshop on Structural Health Monitoring Sorrento Italy June 29-July 02 2010

8 Yun-Kyu An Min Koo Kim Hyung Jin Lim and Hoon Sohn ldquoValidations of reference-free crack detection techniques through a decommissioned bridge testrdquo 2010 COSEIK Annual Con-ference Jeju Korea April 8-9 2010

- 40 -

PARTICIPANT PROJECTS

1 Development of Smart Steel Members for Online Steel Structure Monitoring POSCO Corporation (Funded 41000000 KRW (41000 USD) for 010511 to 123111)

2 Development of Enabling Technologies for Structural Health Monitoring Korean Agency for Defense Development (Funded 240000000 KRW (240000 USD) out of total 3200000000 KRW (3200000 USD) for 060108-053111)

3 Development of On-board SHM Technologies for Composite Air Vehicles The Boeing Company (Funded 400000 USD (400000000 KRW) for 080108 to 073111)

EDUCATIONAL EXPERIENCES

Fall 2009

Teaching Assistant CE207 ldquoElementary Structural Engineering and Laboratoryrdquo KAIST 3 unit undergraduate course

EXPERIMENTAL EXPERIENTCES

2010 Monitoring of Real Aircraft Composite Wing Segment Korea Agency for Defense De-velopment (ADD) Daejeon Korea Real time monitoring of composite air vehicles fund-ed by Korea Agency for Defense Development (ADD)

2009-2011

Field Monitoring of Real Bridge Structures Gimpo (Ramp G bridge) amp Icheon (Sam-seung bridge) Korea US-Korea collaborative research for bridge monitoring testbeds

HONORS amp AWARDS

Scholarships for outstanding students granted KAIST (2007-2009) The three prominent students in the civil and environmental engineering department of KAIST were selected as recipients This scholarship provided 2000$ per year for three years

- 41 -

Page 21: 복합재료 손상의 무기저 진단 및 정량화 - KAISTssslab.kaist.ac.kr/article/pdf/Thesis_2.pdf · 1 Declaration of Ethical Conduct in Research: I, as a graduate student

33 Damage detection algorithm (Level 1)

The damage diagnosis is accomplished by extracting the additional reflected A0 Lamb

wave modes from damage due to multiple reflections If A0 Lamb wave modes are decom-

posed perfectly it is trivial that one transmitted A0 Lamb wave mode is always captured in

the response signal If damage exists other waves in the response signal behind the transmit-

ted A0 Lamb wave mode are presented Since this approach does not rely on previously col-

lected data the diagnosis is independent from environmental variation such as temperature

changes

It is reported that the dispersion of Lamb waves cause wave-packets to spread out in space

and time as they propagate through a structure [20] Thus the reflected waves from the dam-

age have longer pulse duration than the transmitted wave due to their longer propagating dis-

tance In this study the matching pursuit method is modified to capture more spread wave

than the transmitted A0 Lamb wave mode Detail explanation of the modified matching pur-

suit method is described in latter section

Let 1199041199041 and 1199041199042 as the pulse duration (scale) of the transmitted and additional reflected A0

Lamb wave mode respectively

ldquoIf 1199041199042 gt 1199041199041 damage exists

Otherwise there is no damagerdquo

- 10 -

34 Damage quantification algorithm (Level 3)

As mentioned in the previous section damage is quantized by estimating 119889119889 and 119907119907119889119889 using

pitch-catch method The procedure to estimate 119889119889 and 119907119907119889119889 is

(i) Measure 119897119897 119905119905 and 119905119905119889119889

(ii) Assume the range of 119889119889 (0 lt 119889119889 lt 119897119897)

(iii) Calculate 119907119907119894119894 and 119907119907119889119889 with a constraints (0 lt 119907119907119889119889 lt 119907119907119894119894) (Eq 4 and 5)

(iv) Estimate m-value and find a feasible range of 119889119889 and 119907119907119889119889

119907119907119894119894 =119897119897 minus 119889119889119905119905 minus 119905119905119889119889

(4)

119907119907119889119889 =119889119889119905119905119889119889

(5)

35 Damage localization algorithm (Level 2amp3)

For damage localization 1198971198971 is estimated by measuring the arrival time from the excitation

PZT to the entrance of delamination (1199051199051) with additional pulse-echo method implementation

The procedure is similar with the previous damage quantification algorithm

(i) Measure 119897119897 119905119905 119905119905119889119889 and 1199051199051

(ii) Assume the range of 1198971198971 (0 lt 1198971198971 lt 119897119897)

(ii) Compute possible combinations of 1198971198972 119889119889 and 119907119907119889119889 with constraints

(0 lt 119907119907119889119889 lt 119907119907119894119894 119889119889 gt 0) (Eq 5 6 and 7)

- 11 -

1198971198972 = 1198971198971119905119905 minus 1199051199051 minus 119905119905119889119889

1199051199051 (6)

119889119889 = 119897119897 minus 1198971198971 minus 1198971198972 (7)

Since a PZT for pulse-echo method acts as both an actuator and a sensor the obtained sig-

nal contains both the excitation and the response signal Thus the reflected wave from the

entrance of damage can be buried into the tail of excitation signal and cause error on arrival

time of reflected wave from damage Even in this case damage can be localized and quan-

tized using the constraints When first wave packet reflected from the entrance of damage is

buried into the excitation signal the next wave packet reflected from the exit of damage can

be utilized alternatively If the algorithm regards the second reflected wave as the first re-

flected wave there are no combinations of estimated variable meeting the constraints Then

the algorithm repeats the process by regarding the wave as the second reflected wave the

constraints can select proper combination of the estimated variables

The estimated values have range bounded by the constraints When additional information

of 119907119907119894119894 is utilized the damage can be characterized more accurately In this study three meth-

ods are proposed and verified as information level of 119907119907119894119894

Method 1 Solely based on the current data measurement (Reference-free)

Method 2 Limit the possible range of 119907119907119894119894 (Prior knowledge of the system)

Method 3 Measure 119907119907119894119894 (Additional data need to be collected)

Method 1 described in this section can be utilized without any information of 119907119907119894119894 Meth-

od 2 produces more accurate estimation results using prior knowledge of the system such as

- 12 -

119907119907119894119894 variation under temperature changes Method 3 gives unique estimated value using meas-

ured 119907119907119894119894 from other path in the same time In section 4 and 5 damage is characterized using

these three methods and the results are compared each another

36 Matching pursuit method considering the dispersion effect of Lamb

wave

Matching pursuit method has been developed for signal compression approximation and

characterization of obtained signal and function [21-23] The key concept of the matching

pursuit method is that any signal (function) can be decomposed into a linear expansion of

signals selected from a pool of basis signals (functions) called lsquodictionaryrsquo

Let 119891119891 be a function to be analyzed then 119891119891 can be approximated a set of functions select-

ed from the dictionary 119863119863 in order to best match This is done by successive approximations

with orthogonal projections Let1198921198921205741205740 isin 119863119863 the function 119891119891 can be decomposed into

119891119891 = lang1198911198911198921198921205741205740rang1198921198921205741205740 + 119877119877119891119891 (8)

where 119877119877119891119891 is the residual function after approximating 119891119891 using 1198921198921205741205740 1198921198921205741205740 is orthogonal to

119877119877119891119891 hence

1198911198912 = lang1198911198911198921198921205741205740rang2

+ 1198771198771198911198912 (9)

To minimize 119877119877119891119891 1198921198921205741205740 is chosen to maximize lang1198911198911198921198921205741205740rang The matching pursuit method is

an iterative algorithm that decomposes the residue 119877119877119891119891 by projecting it on a function of 119863119863

that matches 119877119877119891119891 almost at best as it was done for 119891119891 The residue 119877119877119899119899119891119891 is decomposed into

119877119877119899119899119891119891 = lang119877119877119899119899119891119891119892119892120574120574119899119899rang119892119892120574120574119899119899 + 119877119877119899119899+1119891119891 (10)

- 13 -

the decomposition is carried up to the order 119898119898 119891119891 can be expressed as

119891119891 = lang119877119877119899119899119891119891119892119892120574120574119899119899rang119892119892120574120574119899119899

119898119898minus1

119899119899=0

+ 119877119877119898119898119891119891 (11)

The original function 119891119891 is decomposed into a sum of dictionary elements that are chosen to

best match its residues

In this study the matching pursuit method is utilized to accurately decompose of the over-

lapped Gaussian pulses and estimate the parameters of the pulse which used for damage

characterization The 119894119894119905119905ℎ Gaussian pulse in the dictionary for approximating the obtained

signal is expressed as

119892119892120574120574119894119894 = 119891119891(119860119860119894119894 119906119906119894119894 119904119904119894119894 119889119889119894119894120601120601119894119894) (12)

where 119860119860119894119894119906119906119894119894119904119904119894119894119889119889119894119894 and 120601120601119894119894 represent amplitude arrival time scale (pulse duration) excita-

tion(central) frequency and phase of the Gaussian pulse respectively The parameters of the

pulse are estimated using a nonlinear regression based optimization method (Gauss-Newton

method) based on an initial guess of the parameters Detail procedure for estimating the pa-

rameters of the pulse is expressed in Hong et al [23]

Among the estimated parameters of the pulses pulse duration (scale) represents the disper-

sion effect If a wave-packet is spread out the estimated pulse duration is increased Using

this feature the matching pursuit method is modified to capture waves which have longer

pulse duration than transmitted A0 Lamb wave mode If the estimate pulse duration of a wave

is less than the transmitted wave the modified matching pursuit method regards it as error

- 14 -

CHAPTER 4 NUMERICAL SIMULATION

41 Composite modeling method

Composite materials can be modeled from nano to macro level depends on its engineering

purpose In general four methods are utilized for modeling of composite materials First

composite is modeled as a homogenous material expressed in terms of effective material

properties [24] The effective material properties are estimated using the modulus and the

volume fraction of each constituent theoretically (Rule of mixture) and measured experimen-

tally Second a model is established that each laminate which have own material properties

are totally bonded together In this model the stacking sequence and direction of each lami-

nates are considered additionally [25-27] Third method for composite modeling in micro

level called micromechanics [28-30] This model considers the interaction between fibers and

matrix in laminate Last multiscale modeling which the scope of modeling is extended to

molecular or quantum level of constituents [31-33] This model starts from molecular me-

chanics or molecular dynamics then the scale of model leads to the nano micro and macro

level by incorporating small scale models As the scale of models is smaller it provides more

logical way than other large scale of models while the computational cost is more expensive

The scope of this study is focused on the dynamic behavior of composite structure the beam

or plate model was utilized for the numerical simulation

- 15 -

42 Damage modeling method

Generally three damage models have been developed and utilized in the real applications

First the stiffness reduction model called as the equivalent element model was introduced

[17 34-35] This model is simply realized by replacing the region around damage area by a

region with adjusted material properties (Youngrsquos modulus shear modulus density etc)

globally Next the duplicated node model is done by disconnecting the element connectivity

assuming there is no interaction among the delaminated surface This model can be utilized

when the location and the extension of damage is important issue [34-36] Last the contact

mechanics model using the principles of contact mechanics the two interfacial forces (con-

tact resistance and sliding resistance) in the delaminated surface [37-38] In this model the

delaminated surfaces are modeled as nonlinear functions of the interlaminar normal strain

and the interlaminar shear strain This model is computationally expensive while provides

more logical way than other models There is however no universally accepted model for

general impact-induced damage in composites In the case of impact-induced damage it is

impossible to predict or estimate the location and number of delaminated layers Thus from

the uncertainty of impact-induced damage the equivalent element model can be considered

as the effective damage model

- 16 -

43 Numerical simulation description

TABLE I Material property of simulation model and test specimen

Tensile Strength 657 MPa Tensile Modulus 59 GPa Shear Strength 804 MPa Shear Modulus 24 Gpa Specific Gravity 16

PZT A

PZT C

PZT BDamage

3

250 125 125

500 Unit mm

Fig 6 Description of the numerical simulation model

The feasibility of the proposed damage characterization technique is verified by 2-D FEM

simulation using MSC PATRAN and NASTRAN Fig 6 presents the dimension PZTs and

damage location of the simulation model In the model 05mm x 05mm square shell element

was used for modeling of the A0 mode whose flexural wave characteristic The element size

was determined by considering the wavelength of the propagation Lamb wave modes The

material properties of the simulation model and damage cases are presented in Table I and

Table II respectively

PZT A and C act as both actuator and sensor for collecting pulse-echo signal They were

installed on the plate which two of them are both-side collocated for the A0 Lamb wave

- 17 -

mode decomposition [39-40] PZT B was only used as a sensor for collecting pitch-catch sig-

nal For simulating the excitation and sensing of PZTs the pin-force model was adopted that

an input force apples to the end nodes of excitation PZT and averaging element strains corre-

sponding sensing area A Gaussian pulse (tone-burst) 50 kHz central frequency was excited

as an input signal

TABLE II Damage cases for numerical simulation

119949119949120783120783 (mm) 119941119941 (mm) G value reduction ()

Case 1 650 200 90 darr

Case 2 675 150 90 darr

Case 3 675 150 93 darr

Case 4 900 150 90 darr

- 18 -

44 Numerical simulation results

015 02 025 03 035

-2

0

2

Time (ms)

Am

plitu

de

IntactDamage

Reflected wave from damage

(a) Pitch-catch signal

015 02 025 03 035

-2

0

2

Time (ms)

Am

plitu

de

IntactDamage

Reflected waves from damage

(b) Pulse-echo signal Fig 7 Comparison of A0 Lamb wave mode signals from numerical simulation between intact and damage cases Time delay attenuation and multiple reflections due to damage simulated successfully using the damage modeling method described in the previous section

Fig 7 shows that comparison of A0 Lamb wave mode signals between intact and one of

damage cases in the simulation The features of A0 Lamb wave mode passing through dam-

age area described in section 2 are successfully simulated Thus the equivalent element

method is proper for modeling damage in composites In the damage cases the additional

waves reflected from damage is observed in both pitch-catch and pulse echo signal

Table III presents the result of the proposed damage characterization algorithm using the

numerical simulation cases shown in Table II In lsquoMethod 2rsquo 119907119907119894119894 variation is assumed to plusmn

5 of the normal value (1190~1315ms) The normal value of 119907119907119894119894 for lsquoMethod 3rsquo is meas-

ured as 1253 ms In each case the damage is characterized using the proposed technique

Note that as the information of 119907119907119894119894 is specified the estimated value is closer to the exact

value

- 19 -

TABLE III Damage characterization for the numerical simulation models

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

Case 1

Exact 650 200

Method 1 248 50~70 79~414 160~834

Method 2 248 603~667 135~241 271~486

Method 3 248 635 188 379

Case 2

Exact 675 150

Method 1 193 60~75 01~251 37~651

Method 2 193 618~682 114~222 295~575

Method 3 193 65 168 435

Case 3

Exact 675 150

Method 1 233 60~75 66~302 140~649

Method 2 233 668~738 85~196 182~419

Method 3 233 703 140 301

Case 4

Exact 900 150

Method 1 204 75~100 48~349 118~856

Method 2 204 860~951 107~216 264~530

Method 3 204 906 162 397

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1253 ms 1190~1315 ms Measured 119907119907119894119894 1253 ms

- 20 -

CHAPTER 5 EXPERIMENTAL VERIFICATION

51 Overview

In this section experimental study is investigated so that the proposed damage characteri-

zation technique Overall experimental setup and test specimen are shown in Fig 8 and 9

The experimental setup consists of test specimens data acquisition (DAQ) system and self-

sensing circuit Self-sensing circuit was installed for pulse-echo measurement using PZTs

[40-41]

PZT A(PZT C) PZT B

500

500

140

75

250

Teflon Inserted Damage (d =25)

Unit mm

Thickness = 3

PZT A(PZT C) PZT B

500

500

150

90

250

Impact Location

Unit mm

Thickness = 3

(a) Teflon inserted damage specimen

(b) Impact-induced damage specimen

Fig 8 Test specimens for the experimental verification In the Teflon inserted damage Teflon tapes with 25mm diameter were inserted at the center of the specimen The impact-induced damage was introduced to the specimen using the impact test equipment

- 21 -

511 Specimen

Two carbon composite plates with a dimension of 500times500times3mm3 were manufactured by

Nexcomes Inc by stacking 12 layers of CF3327 3K prepreg (woven fabrics type) manufac-

tured by Hankuk Fiber Inc The material properties of the specimen are presented in Table I

One of the specimens Teflon tapes with 25mm diameter was inserted in every 3 layers at the

center of the specimen to mimic the damage The other specimen is utilized for the character-

ization of real impact-induced damage

512 Data acquisition (DAQ) system and PZT

The data acquisition (DAQ) system consists of an arbitrary waveform generator (AWG) a

high speed signal digitizer (DIG) and two multiplexers (MUX) Using the 16-bits AWG

tone-burst input signal with a plusmn10 peak-to-peak voltage were generated The response signal

was measured by the 14-bit DIG For the signal-to-noise ratio reduction both pitch-catch and

pulse-echo signals were measured 150 times and averaged in the time domain 3 identical

circular PZTs of 10mm diameter manufactured by APC International Ltd were installed on

the specimen A Gaussian pulse (tone-burst) 50 kHz central frequency was excited as an in-

put signal The excitation frequency was chosen in order to avoid overlapping of the S0 and

A0 Lamb wave modes and generating of the higher modes (S1 A1hellip)

- 22 -

PZT Actuator

PZT Sensor

AWG

DIG

DAQ System Specimen

SensingActuatingMUX

1

23

4

1 Sending signal to MUX

4 Sending signal to DIG

Self-SensingCircuit

18 Unit mm

10

(a)

(b)

Fig 9 Overall experimental setup (a) Data acquisition (DAQ) system and (b) Installed circu-lar PZT transducer on the specimen with its dimension

513 Impact test

The impact-induced damage was introduced to the specimen using the impact test equip-

ment presented in Fig 10 A 5kg mass with 10mm tip was dropped from 015m height onto

the same location of the specimen The signals were obtained at each number of impacts

Since the reflected wave is not detected at the first impact case the data obtained from the

second and third impact were utilized for the analysis

10

Unit mm (a)

(b)

Fig 10 (a) Impact test equipment for introducing impact-induced damage and (b) Tip of the impact tester with 10mm diameter

- 23 -

514 Temperature effect test

In reality structures are subjected to changing environmental conditions such as tempera-

ture that affect measured signal and cause false-alarm Thus the robustness of the proposed

damage characterization technique under temperature variation in investigated (Fig 11) The

specimens were placed inside a temperature chamber and a thermocouple was installed on

the specimen to measure temperature of surface of the specimen The signal was obtained

under three temperature cases (0 20 and 50degC) and the humidity was kept at 30

Temperature Chamber

Test Specimen

Fig 11 Test specimen in temperature chamber

- 24 -

52 Teflon inserted damage

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Reflected wave from damage

(a)

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Transmitted WaveReflected Wave

Detected reflected wave

(b)

Fig 12 A0 Lamb wave mode obtained using pitch-catch method from the Teflon inserted damage specimen (a) The reflected wave from damage is shown (b) The Teflon inserted damage is detected by capturing the reflected wave

Fig 12 shows the obtained pitch-catch signal from the Teflon inserted damage specimen

and damage diagnosis result The Teflon inserted damage is detected successfully by captur-

ing the reflected wave Table IV presents the result of the Teflon inserted damage characteri-

zation 119907119907119894119894 is measured as 1213ms and the variation is assumed to plusmn 5 of the normal val-

ue (1152~1273ms) The results indicate that the proposed damage characterization approach

works well in real application

TABLE IV Damage characterization for the Teflon inserted damage specimen

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

Exact 635 25

Method 1 123 60~75 73~338 298~1220

Method 2 123 619~662 229~310 934~1188

Method 3 123 631 284 1161

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1213 ms

- 25 -

53 Impact-induced damage

11 115 12 125 13-4

-2

0

2

4

Time (ms)

Vol

tage

(mV

)

IntactImpact2Impact3

Reflected wave from damage

(a) Pitch-catch signal

11 115 12 125 13 135

-05

0

05

1

Time (ms)

Volta

ge (m

V)

IntactImpact2Impact3

Boundary reflection

Reflected wave from damage

(b) Pulse-echo signal

Fig 13 The signal obtained from the test specimen subjected to impact The impact was in-troduced at the same location of the specimen and the signal was obtained after each impacts As the number of impact is increased the transmitted A0 Lamb wave mode is delayed and attenuated

The obtained signals from impact-induced damage specimen are presented in Fig 13 It is

observed that as the number of impact is increased the transmitted A0 Lamb wave mode is

more delayed because the wave velocity in the damage area is reduced Fig 14 shows the

false-alarm test result whether the proposed damage detection algorithm works even in pris-

tine condition of the test specimen It is shown that no false alarm is occurred because no ad-

ditional reflected wave is captured

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Transmitted Wave

(a) Transmitted wave

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Reflected Wave

No reflected wave

(b) Reflected wave

Fig 14 False alarm test for the proposed damage detection algorithm in pristine condition of the specimen Reflected wave from damage is not captured false alarm is not occurred

- 26 -

TABLE V Damage characterization for the impact-induced damage specimen

of Impacts m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

2

(1470 J)

Method 1 080 65~90 75~246 470~1157

Method 2 080 801~885 114~245 711~1134

Method 3 080 843 179 1122

3

(2205 J)

Method 1 148 50~100 06~380 21~1134

Method 2 148 758~838 249~368 843~1125

Method 3 148 798 308 1044

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1213 ms

Table V presents the characterization results of the impact-induced damage case Here the

measured 119907119907119894119894 and the assumed variation are same as the artificial damage case It is observed

that as the number of impact is increased 119889119889 is increased while 119907119907119889119889 is reduced Thus it is

validated that the proposed damage characterization approach estimates the increment of the

damage well

54 Temperature effect test

541 Teflon inserted damage

11 115 12 125 13-10

-5

0

5

10

Time (ms)

Vol

tage

(mV

)

0 oC20 oC50 oC

(a) Pitch-catch signal

11 115 12 125 13 135

-1

0

1

2

3

Time(ms)

Volta

ge (m

V)

0 oC20 oC50 oC

(b) Pulse-echo signal

Fig 15 Signals obtained from the Teflon inserted damage specimen under temperature varia-tion Time delay and amplitude variation are occurred

Fig 15 shows the obtained signals from the Teflon inserted damage specimen subjected to

- 27 -

temperature variation Time delay and amplitude variations are occurred there is no addition-

al damage is introduced The analyzed results of the obtained signals are reported in Table VI

The measured 119907119907119894119894 is 1247 ms at 0 degC 1213 ms at 20 degC and 1168 ms at 50 degC respectively

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms (1152~1273 ms)

The estimated results are also changed under temperature variations the fluctuation of esti-

mated range and exact value is small

TABLE VI Temperature effect test for the Teflon inserted damage specimen

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

0 degC

Exact 635 25

Method 1 113 60~70 79~268 352~1291

Method 2 113 560~619 233~344 1034~1172

Method 3 113 606 257 1141

20 degC

Exact 635 25

Method 1 123 60~75 73~338 298~1220

Method 2 123 619~662 229~310 934~1188

Method 3 123 631 284 1161

50 degC

Exact 635 25

Method 1 137 60~75 72~328 263~1228

Method 2 137 622~687 183~299 666~1086

Method 3 137 642 264 960

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1247 ms at 0 degC 1213 ms at 20 degC and 1168 ms at 50 degC

- 28 -

542 Impact-induced damage

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Vol

tage

(mV

)

Transmitted WaveReflected Wave

No reflected wave

(a) 0 degC

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Transmitted WaveReflected Wave

No reflected wave

(b) 50 degC

Fig 16 False alarm test under temperature variations It is shown that proposed damage char-acterization algorithm is robust under temperature variations

The false alarm test in pristine condition of the impact-induced damage specimen under

temperature variation was investigated (Fig 16) It is validated that the proposed damage di-

agnosis algorithm is robust under temperature variations Fig 17 shows the obtained signals

from the impact-induced damage (impacted 3 times) specimen under temperature variation

The analyzed results of the obtained signals are reported in Table VII The measured 119907119907119894119894 in

each temperature and the assumed variation are same as the artificial damage case

11 115 12 125 13-5

0

5

Time (ms)

Volta

ge (m

V)

0 oC20 oC50 oC

(a) Pitch-catch signal

11 115 12 125 13 135

-1

0

1

2

3

Time(ms)

Vol

tage

(mV

)

0 oC20 oC50 oC

(b) Pulse-echo signal

Fig 17 Signals obtained from impact-induced damage specimen (Impact 3) under tempera-ture variation

- 29 -

TABLE VII Temperature effect test for the impact induced damage specimen (Impact 3 case)

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

0 degC

Method 1 135 75~90 72~310 267~1148

Method 2 135 705~779 264~381 978~1126

Method 3 135 763 289 1071

20 degC

Method 1 148 50~100 06~380 21~1134

Method 2 148 758~838 249~368 843~1125

Method 3 148 798 308 1044

50 degC

Method 1 153 75~100 53~415 174~1227

Method 2 153 804~889 214~336 702~1103

Method 3 153 815 320 1050

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1247 ms at 0 degC 1213 ms at 20 degC and 1168 ms at 50 degC

- 30 -

CHAPTER 6 CONCLUDING REMARKS

61 Executive summary

In this study the reference-free damage detection localization and quantification tech-

nique in a multi layered composite plate is proposed Additional reflected A0 Lamb wave

mode from the damage due to multiple reflections in damage area is used as a feature for

characterizing the damage First damage is characterized in terms of a damage index (m-

value) defined as the ratio of the physical length of the damage to the wavelength of the A0

Lamb wave mode Then the range of possible damage location length and severity are esti-

mated with constraints This procedure can be accomplished without prior baseline data To

realize the proposed technique the matching pursuit technique is modified for reduction of

the error by considering the dispersive characteristic of Lamb wave The matching pursuit

method estimates the arrival time of each wave packets accurately even two wave packets are

overlapped The numerical simulations and experimental tests including Teflon tape inserted

and impact-induced damage are investigated for the feasibility of the proposed technique

Moreover the effectiveness of the proposed technique under changing temperature is exper-

imentally verified

- 31 -

62 Suggestion for further study

The limitations of this study and the possible improvement and further studies are listed as

follow

1 Measurement of impact-induced damage length using conventional NDT technique

Additional experimental verification for impact-induced damage characterization is

needed by measuring the real length of damage using conventional NDT technique

such as ultrasonic scan or infrared camera

2 Limitation of both side collocated PZT installation In this study excitation PZTs are

installed both side and collocated for decomposing A0 Lamb wave mode However it

can produce decomposition error due to imperfection of PZT location and their installa-

tion Moreover in many cases assess to both surface of structures is not allowed such

as aircraft applications To improve this limitation a reference-free damage characteri-

zation using single side PZTs will be considered

3 Complex structure application In this study damage is characterized in simple plate-

like structure When the monitored structure becomes complex the response signals al-

so becomes complex due to its geometry and boundary Thus the investigations for

complex structure application such as stiffener spar and curvature are needed

- 32 -

REFERENCES

[1] Cawley P and Alleyne D (1996) ldquoThe use of Lamb waves for the long range inspection of large structuresrdquo Ultrasonics 34 pp287-290

[2] Ghosh T Kundu T and Karpur P (1998) ldquoEfficient use of Lamb waves for detecting defects in large platesrdquo Ultrasonics 36 pp791-801

[3] Guirguitui V (2003) ldquoLamb wave generation with piezoelectric wafer active sensors for structural health monitoringrdquo SPIE Conference on Smart Structures and Materials San Diego CA 5391 pp111-422

[4] Ragvahan A and Cesnik CE S (2004) ldquoModeling of piezoelectric-based Lamb wave generation and sensing for structural health monitoringrdquo SPIE Conference on Smart Struc-tures and Materials San Diego CA 5391 pp419-430

[5] Rose JL Pilarski A and Dirit JJ (1993) ldquoAn approach to guided wave mode selection for inspection of laminated platerdquo Journal of Reinforced Plastics and Composites 12(5) pp536-544

[6] Guo N and Cawley P (1993) ldquoThe interaction of Lamb waves with delaminations in composite laminatesrdquo Journal of Acoustical Society of America 94(4) pp 2240-2246

[7] Valdez SHD and Soutis C (2002) ldquoReal-time nondestructive evaluation of fiber com-posite laminates using low-frequency Lamb wavesrdquo Journal of Acoustical Society of Amer-ica 111(5) pp 2026-2033

[8] Yang M and Qiao P (2005) ldquoModeling and experimental detection of damage in virous materials using the pulse-echo method and piezoelectric sensoractuatorsrdquo Smart Materials and Structures 14 pp1083-1100

[9] Lestari W and Qiao P (2005) ldquoApplication of wave propagation analysis for damage identification in composite laminated beamsrdquo Journal of Composite Materials 39(22) pp1967-1984

[10] Kessler S S Spearing S M and Soutis C (2002) ldquoDamage detection in composite ma-terials using Lamb wave methodsrdquo Smart Materials and Structures 11 pp269-278

[11] Sohn H Park G Wail J R Limback N P and Farrar C R (2004) ldquoWavelet-based active sensing for delamination detection in composite structuresrdquo Smart Materials and Structures 13 pp 153-160

[12] Petculescu G Krishnaswamy S and Achenback J D (2008) ldquoGroup delay measure-ments using modally selective Lamb wave transducers for detection and sizing of delamina-tions in compositesrdquo Smart Materials and Structures 17

[13] Sohn H Park H W Law K H and Farrar C R (2007) ldquoDamage detection in compo-

- 33 -

site plates by using an enhanced time reversal methodrdquo Journal of Aerospace Engineering (ASCE) pp 141-151

[14] Sohn H (2007) ldquoEffects of environmental and operational variability on structural health monitoringrdquo Philosophical Transactions of the Royal Society Ardquo 365 pp539-560

[15] Rytter A (1993) ldquoVibration based inspection of civil engineering structuresrdquo PhD The-sis Department of Building Technology and Structural Engineering University of Aalborg Aalborg Denmark

[16] Hayashi T and Kawashima K (2002) ldquoMultiple reflections of Lamb waves at a delami-nationrdquo Ultrasonics 40 pp193-197

[17] Wang C H and Rose L R F (2003) ldquoWave reflection and transmission in beams con-taining delamination and inhomogeneityrdquo Journal of Sound and Vibration 264 pp 851-872

[18] Ramadas C Balasubramaniam K Joshi M and Krishnamurthy C V (2009) ldquoInterac-tion of the primary anti-symmetric Lamb mode (Ao) with symmetric delaminations numer-ical and experimental studiesrdquo Smart Materials and Structures 18

[19] Ramadas C Balasubramaniam K Joshi M and Krishnamurthy C V (2010) ldquoInterac-tion of guided Lamb waves with an asymmetrically located delamination in a laminated composite platerdquo Smart Materials and Structures 19

[20] Wilcox P Lowe M and Cawely P (2001) ldquoThe effect of dispersion on long-range in-spection using ultrasonic guided wavesrdquo NDTampE International 34 pp1-9

[21] Mallat S G and Zhang Z (1993) ldquoMatching pursuits with time-frequency dictionariesrdquo IEEE Tansactions on Signal Processing 41(12) pp3397-3415

[22] Zhang G Zhang S and Wang Y (2000) ldquoApplication of adaptive time-frequency de-composition in ultrasonic NDE of highly scattering materialsrdquo Ultrasonics 38 pp 961-964

[23] Hong J C Sun K H and Kim Y Y (2005) ldquoThe matching pursuit approach based on the modulated Gaussian pulse for efficient guided-wave damage inspectionrdquo Smart Mate-rials and Structures 14 pp548-560

[24] Herakovich C T (1997) Mechanics of Fibrous Composites Wiley 480 pages

[25] Reddy J N (1985) ldquoA review of the literature on finite-element modeling of laminated composite platesrdquo The Shock and Vibration Digest 17(4) pp 3-8

[26] Reddy J N (2003) Mechanics of Laminated Composite Plates and Shells Theory and Analysis CRC Press 856 pages

[27] Palacz M Krawczuk M and Ostachowicz W (2005) ldquoThe spectral finite element mod-el for analysis of flexure-shear coupled wave propagation Part 1 laminated multilayer

- 34 -

composite beamrdquo Composite Structures 68 pp37-44

[28] Zhang Y C and Harding J (1990) ldquoA numerical micromechanics analysis of the me-chanical properties of a plain weave compositerdquo Computer amp Structures 36(5) pp 839-844

[29] Nemat-Nasser S and Hori M (1999) Micromechanics Overall Properties of Heteroge-neous Materials Elsevier 810 pages

[30] Scida D Aboura Z Benzeggagh M L and Bocherens E (1999) ldquoA micromechanics model for 3D elasticity and failure of woven-fibre composite materialrdquo Composites Science and Technology 59 pp 505-517

[31] Thomas Y and Wu X H (1997) ldquoA multiscale finite element method for elliptic prob-lems in composite materials and porous mediardquo Journal of Computational Physics 134 pp 169-189

[32] Lee K H Moorthy S and Ghosh S (1999) ldquoMultiple scale computational model for damage in composite materialsrdquo Computer Methods in Applied Mechanics and Engineer-ing 172 pp 175-201

[33] Namilae S and Chandra N (2005) ldquoMultiscale model to study the effect of interfaces in carbon nanotube-based compositesrdquo Journal of Engineering Materials and Technology (ASME) 127 pp222-232

[34] Adams D E (2007) Health monitoring of Structural Materials and Components Wiley 460 pages

[35] Gopalakrishnan S (2009) ldquoModeling Aspects in Finite Elementrdquo Encyclopedia of Struc-tural Health Monitoring 2 pp791-809

[36] Palacz M Krawczuk M and Ostachowicz W (2005) ldquoThe spectral finite element mod-el for analysis of flexure-shear coupled wave propagation Part 2 Deminated multilayer composite beamrdquo Composite Structures 68 pp45-51

[37] Baik J M and Thopson R B (1984) ldquoUltrasonic scattering from imperfect interfaces A quasi-static modelrdquo Journal of Nondestructive Evaluation 4(3) pp177-196

[38] Luo H and Hanagud S (2000) ldquoDynamics of delamincated beamsrdquo International Jour-nal of Solids and Structures 37 pp1501-1519

[39] Kim S B and Sohn H (2007) ldquoInstantaneous reference-free crack detection based on polarization characteristics of piezoelectric materialsrdquo Smart Materials and Structures 16 pp2375-2387

[40] An Y K and Sohn H (2010) ldquoInstantaneous crack detection under varying temperature and static loading conditionsrdquo The Special Issue of Structural Control and Health Monitor-ing 17(7) pp 730-741

- 35 -

[41] Lee S J and Sohn H (2006) ldquoActive self-sensing scheme development for structural health monitoringrdquo Smart Materials and Structures 15(6) pp 1734-1746

- 36 -

요 약 문

복합재료 손상의 무기저 진단 및 정량화

복합재료는 재료의 경량성 내구성 고강도의 특성으로 인하여 항공기에서부터

교량에까지 널리 사용되고 있다 하지만 외부의 충격으로 인한 복합재료 내부의

손상은 재료의 강도를 감소시켜 구조물의 안정성을 저하시키는 요인이 된다 하지만

복합재료의 손상은 재료 내부에서 발생하여 육안으로는 식별하기도 힘들며 표면에서

보이는 것만으로 내부의 손상 정도를 판단하기는 불가능하다는 단점을 가지고 있다

본 연구에서는 손상 내부에서 반사로 인해 발생하는 파를 이용하여 기저자료를

사용하지 않고 특정 시점에서 얻은 자료만으로 복합재료의 손상을 진단하고 정량화

할 수 있는 기법을 제안한다 제안된 기법의 적용성을 검증하기 위하여 유한요소법을

이용한 수치해석을 수행하였고 Teflon tape 이 삽입된 시편과 실제 충격하중으로

인한 손상을 입은 시편을 이용한 실험을 진행하였다 수치해석 및 실험 결과 제안된

기법이 복합재료의 손상을 성공적으로 진단 및 정량화할 수 있음을 확인하였다 또한

다양한 외부 온도 환경에서 제안된 기법의 환경영향 실험을 수행하였고 다양한

환경에서 손상의 진단 및 정량화가 성공적으로 이루어짐을 검증하였다

핵심어 무기저 손상진단 및 정량화 압전센서 복합재료 유한요소법 Matching pursuit

method

- 37 -

감 사 의 글

기대와 두려움을 안고 시작한 석사과정이 어느덧 2 년 이라는 세월 속에서 추억으로 자리잡아

가고 있습니다 학부생 기간을 포함한 지난 3 년 동안 SHM 이라는 새로운 분야를 접하고 연구를

수행 할 수 있도록 도와주신 고마운 분들이 있어 그 감사를 표하고자 합니다

우선 저의 많은 부족한 점을 일깨워 주시고 연구에 대한 뜻을 더욱 확고히 세울 수 있게 해

주신 지도교수님이신 손 훈 교수님께 깊은 감사의 뜻을 전합니다 짧은 기간이었지만 교수님의

도움으로 많은 것을 배우고 느끼고 깨달을 수 있었습니다 또한 본 연구를 수행할 수 있도록 지

원해주신 국토해양부 (U-City 석 박사과정 지원사업)에 깊은 감사 드립니다

때로 힘들고 지칠 때 항상 저를 믿고 응원해 주시는 가족이 있었기에 지금의 글을 쓸 수 있었

습니다 또한 지금 연구실에서 함께 생활하고 있는 윤규형 민구형 철민이형 현석이형 진열이형

호민이형 병진이 선혜누나와 성대에 계신 창길이형 미국에 계신 의재형과 Debaditiya 에게도 감

사의 말씀을 전합니다 그리고 모두 나열할 수는 없지만 카이스트 생활의 활력소가 되어준 카이

스트 건설 및 환경공학과 선후배들 직원 분들과 SOS 멤버들에게도 감사를 표합니다

앞서 거론한 많은 이들의 도움이 없었다면 저 또한 지금 이 자리에 있을 수 없었을 것입니다

많이 부족하기에 더 노력하고 배우고자 했던 초심을 다시 한번 생각하며 더 큰 결실을 맺기 위해

앞으로 나아가겠습니다 도움을 주신 모든 분들께 이 감사의 글을 바칩니다

2011년 6월 카이스트에서 임형진 올림

- 38 -

CURRICULUM VITAE

Hyung Jin Lim

Research Assistant

The Civil and Environmental Engineering Department

Korea Advanced Institute of Science and Technology (KAIST)

291 Daehak-Ro Guseong-Dong Yuseong-Gu

Daejeon Republic of Korea 305-701

Phone (82)+42-350-3665 Fax (82)+42-350-3610

Email limhj87gmailcom

RESEARCH INTERESTS

Damage Detection Localization and Quantification on Composite and Metallic Structures Pattern Recognition Finite Element Modeling (FEM)

EDUCATION

2009 ndash 2011 MS Civil and Environmental Eng Korea Advanced Institute of Science and Technology (KAIST) Korea Rep

2006 ndash 2009 BS Civil and Environmental Eng Korea Advanced Institute of Science and Technology (KAIST) Korea Rep

JOURNAL PUBLICATIONS The corresponding author is underlined 1 Hyung Jin Lim Hoon Sohn ldquoReference-free damage detection localization and quantification

using in compositerdquo in preparation for Journal of the Acoustical Society of America 2011

2 Chul Min Yeum Hoon Sohn Jeong Bum Ihn Hyung Jin Lim ldquoInstantaneous delamination detection in a composite using a dual piezoelectric transducer networkrdquo Submitted to Compo-sites Science and Technology 2011

3 Yun-Kyu An Hyung Jin Lim Min Koo Kim Jin Yeol Yang Hoon Sohn Chang Guen Lee ldquoInvestigation of applicability of local reference-free damage detection techniques to in-situ

- 39 -

bridgesrdquo Submitted to ASCE Journal of Structural Engineering 2011

4 Hyung Jin Lim Min Koo Kim Hoon Sohn Chan Yik Park ldquoImpedance based damage detec-tion under varying temperature and loading conditionsrdquo Submitted to NDTampE International 2011

5 Hyung Jin Lim Hoon Sohn ldquoQuantitative estimation of transmitted and reflected lamb waves at discontinuityrdquo Journal of the Korean Society for Nondestructive Testing Vol30 No 4 pp1-8 2010

CONFERENCE PROCEEDINGS

The corresponding author is underlined 1 Hyung Jin Lim and Hoon Sohn ldquoReference-free damage detection localization and quantifi-

cation in compositesrdquo ASCE Engineering Mechanics Institute Conference Boston June 2-4 2011

2 Hyung Jin Lim and Hoon Sohn ldquoReference-free delamination quantification within multi-layer composite materialsrdquo KSNT Spring Conference May 26-27 Gyeong-Ju Korea 2011

3 Min Koo Kim Hyun Jin Lim Hoon Sohn Chan Yik Park ldquoImpedance-based bolt loosening detection under varying temperature and loadingrdquo Asian Pacific Workshop on Structural Health Monitoring Tokyo Japan November 30-December 2 2010

4 Hyung Jin Lim Chul Min Yeum and Hoon Sohn ldquoModeling of Impact-induced Delamination in a Multilayer Composite Platerdquo The 23rd KKCNN Symposium on Civil Engineering Taipei (Taiwan) November 13-15 2010

5 Yun-Kyu An Hyung Jin Lim and Hoon Sohn ldquoTemperature independent Delamination Detec-tion using data normalizationrdquo The 3rd International Conference on Multi-Functional Materi-als and Structures Chonbuk National Univ Jeonju Korea September 14-18 2010

6 Hoon Sohn Yun-Kyu An Chang Gil Lee Min Koo Kim Hyung Jin Lim ldquoApplication of online NDT techniques to decommissioned bridge testingrdquo the 5th European Workshop on Structural Health Monitoring Sorrento Italy June 29-July 02 2010

7 Min Koo Kim Hyung Jin Lim Hoon Sohn Chan Yik Park ldquoImpedance based damage diag-nosis of a complex composite aircraft wing under changing loading and temperature condi-tionsrdquo the 5th European Workshop on Structural Health Monitoring Sorrento Italy June 29-July 02 2010

8 Yun-Kyu An Min Koo Kim Hyung Jin Lim and Hoon Sohn ldquoValidations of reference-free crack detection techniques through a decommissioned bridge testrdquo 2010 COSEIK Annual Con-ference Jeju Korea April 8-9 2010

- 40 -

PARTICIPANT PROJECTS

1 Development of Smart Steel Members for Online Steel Structure Monitoring POSCO Corporation (Funded 41000000 KRW (41000 USD) for 010511 to 123111)

2 Development of Enabling Technologies for Structural Health Monitoring Korean Agency for Defense Development (Funded 240000000 KRW (240000 USD) out of total 3200000000 KRW (3200000 USD) for 060108-053111)

3 Development of On-board SHM Technologies for Composite Air Vehicles The Boeing Company (Funded 400000 USD (400000000 KRW) for 080108 to 073111)

EDUCATIONAL EXPERIENCES

Fall 2009

Teaching Assistant CE207 ldquoElementary Structural Engineering and Laboratoryrdquo KAIST 3 unit undergraduate course

EXPERIMENTAL EXPERIENTCES

2010 Monitoring of Real Aircraft Composite Wing Segment Korea Agency for Defense De-velopment (ADD) Daejeon Korea Real time monitoring of composite air vehicles fund-ed by Korea Agency for Defense Development (ADD)

2009-2011

Field Monitoring of Real Bridge Structures Gimpo (Ramp G bridge) amp Icheon (Sam-seung bridge) Korea US-Korea collaborative research for bridge monitoring testbeds

HONORS amp AWARDS

Scholarships for outstanding students granted KAIST (2007-2009) The three prominent students in the civil and environmental engineering department of KAIST were selected as recipients This scholarship provided 2000$ per year for three years

- 41 -

Page 22: 복합재료 손상의 무기저 진단 및 정량화 - KAISTssslab.kaist.ac.kr/article/pdf/Thesis_2.pdf · 1 Declaration of Ethical Conduct in Research: I, as a graduate student

34 Damage quantification algorithm (Level 3)

As mentioned in the previous section damage is quantized by estimating 119889119889 and 119907119907119889119889 using

pitch-catch method The procedure to estimate 119889119889 and 119907119907119889119889 is

(i) Measure 119897119897 119905119905 and 119905119905119889119889

(ii) Assume the range of 119889119889 (0 lt 119889119889 lt 119897119897)

(iii) Calculate 119907119907119894119894 and 119907119907119889119889 with a constraints (0 lt 119907119907119889119889 lt 119907119907119894119894) (Eq 4 and 5)

(iv) Estimate m-value and find a feasible range of 119889119889 and 119907119907119889119889

119907119907119894119894 =119897119897 minus 119889119889119905119905 minus 119905119905119889119889

(4)

119907119907119889119889 =119889119889119905119905119889119889

(5)

35 Damage localization algorithm (Level 2amp3)

For damage localization 1198971198971 is estimated by measuring the arrival time from the excitation

PZT to the entrance of delamination (1199051199051) with additional pulse-echo method implementation

The procedure is similar with the previous damage quantification algorithm

(i) Measure 119897119897 119905119905 119905119905119889119889 and 1199051199051

(ii) Assume the range of 1198971198971 (0 lt 1198971198971 lt 119897119897)

(ii) Compute possible combinations of 1198971198972 119889119889 and 119907119907119889119889 with constraints

(0 lt 119907119907119889119889 lt 119907119907119894119894 119889119889 gt 0) (Eq 5 6 and 7)

- 11 -

1198971198972 = 1198971198971119905119905 minus 1199051199051 minus 119905119905119889119889

1199051199051 (6)

119889119889 = 119897119897 minus 1198971198971 minus 1198971198972 (7)

Since a PZT for pulse-echo method acts as both an actuator and a sensor the obtained sig-

nal contains both the excitation and the response signal Thus the reflected wave from the

entrance of damage can be buried into the tail of excitation signal and cause error on arrival

time of reflected wave from damage Even in this case damage can be localized and quan-

tized using the constraints When first wave packet reflected from the entrance of damage is

buried into the excitation signal the next wave packet reflected from the exit of damage can

be utilized alternatively If the algorithm regards the second reflected wave as the first re-

flected wave there are no combinations of estimated variable meeting the constraints Then

the algorithm repeats the process by regarding the wave as the second reflected wave the

constraints can select proper combination of the estimated variables

The estimated values have range bounded by the constraints When additional information

of 119907119907119894119894 is utilized the damage can be characterized more accurately In this study three meth-

ods are proposed and verified as information level of 119907119907119894119894

Method 1 Solely based on the current data measurement (Reference-free)

Method 2 Limit the possible range of 119907119907119894119894 (Prior knowledge of the system)

Method 3 Measure 119907119907119894119894 (Additional data need to be collected)

Method 1 described in this section can be utilized without any information of 119907119907119894119894 Meth-

od 2 produces more accurate estimation results using prior knowledge of the system such as

- 12 -

119907119907119894119894 variation under temperature changes Method 3 gives unique estimated value using meas-

ured 119907119907119894119894 from other path in the same time In section 4 and 5 damage is characterized using

these three methods and the results are compared each another

36 Matching pursuit method considering the dispersion effect of Lamb

wave

Matching pursuit method has been developed for signal compression approximation and

characterization of obtained signal and function [21-23] The key concept of the matching

pursuit method is that any signal (function) can be decomposed into a linear expansion of

signals selected from a pool of basis signals (functions) called lsquodictionaryrsquo

Let 119891119891 be a function to be analyzed then 119891119891 can be approximated a set of functions select-

ed from the dictionary 119863119863 in order to best match This is done by successive approximations

with orthogonal projections Let1198921198921205741205740 isin 119863119863 the function 119891119891 can be decomposed into

119891119891 = lang1198911198911198921198921205741205740rang1198921198921205741205740 + 119877119877119891119891 (8)

where 119877119877119891119891 is the residual function after approximating 119891119891 using 1198921198921205741205740 1198921198921205741205740 is orthogonal to

119877119877119891119891 hence

1198911198912 = lang1198911198911198921198921205741205740rang2

+ 1198771198771198911198912 (9)

To minimize 119877119877119891119891 1198921198921205741205740 is chosen to maximize lang1198911198911198921198921205741205740rang The matching pursuit method is

an iterative algorithm that decomposes the residue 119877119877119891119891 by projecting it on a function of 119863119863

that matches 119877119877119891119891 almost at best as it was done for 119891119891 The residue 119877119877119899119899119891119891 is decomposed into

119877119877119899119899119891119891 = lang119877119877119899119899119891119891119892119892120574120574119899119899rang119892119892120574120574119899119899 + 119877119877119899119899+1119891119891 (10)

- 13 -

the decomposition is carried up to the order 119898119898 119891119891 can be expressed as

119891119891 = lang119877119877119899119899119891119891119892119892120574120574119899119899rang119892119892120574120574119899119899

119898119898minus1

119899119899=0

+ 119877119877119898119898119891119891 (11)

The original function 119891119891 is decomposed into a sum of dictionary elements that are chosen to

best match its residues

In this study the matching pursuit method is utilized to accurately decompose of the over-

lapped Gaussian pulses and estimate the parameters of the pulse which used for damage

characterization The 119894119894119905119905ℎ Gaussian pulse in the dictionary for approximating the obtained

signal is expressed as

119892119892120574120574119894119894 = 119891119891(119860119860119894119894 119906119906119894119894 119904119904119894119894 119889119889119894119894120601120601119894119894) (12)

where 119860119860119894119894119906119906119894119894119904119904119894119894119889119889119894119894 and 120601120601119894119894 represent amplitude arrival time scale (pulse duration) excita-

tion(central) frequency and phase of the Gaussian pulse respectively The parameters of the

pulse are estimated using a nonlinear regression based optimization method (Gauss-Newton

method) based on an initial guess of the parameters Detail procedure for estimating the pa-

rameters of the pulse is expressed in Hong et al [23]

Among the estimated parameters of the pulses pulse duration (scale) represents the disper-

sion effect If a wave-packet is spread out the estimated pulse duration is increased Using

this feature the matching pursuit method is modified to capture waves which have longer

pulse duration than transmitted A0 Lamb wave mode If the estimate pulse duration of a wave

is less than the transmitted wave the modified matching pursuit method regards it as error

- 14 -

CHAPTER 4 NUMERICAL SIMULATION

41 Composite modeling method

Composite materials can be modeled from nano to macro level depends on its engineering

purpose In general four methods are utilized for modeling of composite materials First

composite is modeled as a homogenous material expressed in terms of effective material

properties [24] The effective material properties are estimated using the modulus and the

volume fraction of each constituent theoretically (Rule of mixture) and measured experimen-

tally Second a model is established that each laminate which have own material properties

are totally bonded together In this model the stacking sequence and direction of each lami-

nates are considered additionally [25-27] Third method for composite modeling in micro

level called micromechanics [28-30] This model considers the interaction between fibers and

matrix in laminate Last multiscale modeling which the scope of modeling is extended to

molecular or quantum level of constituents [31-33] This model starts from molecular me-

chanics or molecular dynamics then the scale of model leads to the nano micro and macro

level by incorporating small scale models As the scale of models is smaller it provides more

logical way than other large scale of models while the computational cost is more expensive

The scope of this study is focused on the dynamic behavior of composite structure the beam

or plate model was utilized for the numerical simulation

- 15 -

42 Damage modeling method

Generally three damage models have been developed and utilized in the real applications

First the stiffness reduction model called as the equivalent element model was introduced

[17 34-35] This model is simply realized by replacing the region around damage area by a

region with adjusted material properties (Youngrsquos modulus shear modulus density etc)

globally Next the duplicated node model is done by disconnecting the element connectivity

assuming there is no interaction among the delaminated surface This model can be utilized

when the location and the extension of damage is important issue [34-36] Last the contact

mechanics model using the principles of contact mechanics the two interfacial forces (con-

tact resistance and sliding resistance) in the delaminated surface [37-38] In this model the

delaminated surfaces are modeled as nonlinear functions of the interlaminar normal strain

and the interlaminar shear strain This model is computationally expensive while provides

more logical way than other models There is however no universally accepted model for

general impact-induced damage in composites In the case of impact-induced damage it is

impossible to predict or estimate the location and number of delaminated layers Thus from

the uncertainty of impact-induced damage the equivalent element model can be considered

as the effective damage model

- 16 -

43 Numerical simulation description

TABLE I Material property of simulation model and test specimen

Tensile Strength 657 MPa Tensile Modulus 59 GPa Shear Strength 804 MPa Shear Modulus 24 Gpa Specific Gravity 16

PZT A

PZT C

PZT BDamage

3

250 125 125

500 Unit mm

Fig 6 Description of the numerical simulation model

The feasibility of the proposed damage characterization technique is verified by 2-D FEM

simulation using MSC PATRAN and NASTRAN Fig 6 presents the dimension PZTs and

damage location of the simulation model In the model 05mm x 05mm square shell element

was used for modeling of the A0 mode whose flexural wave characteristic The element size

was determined by considering the wavelength of the propagation Lamb wave modes The

material properties of the simulation model and damage cases are presented in Table I and

Table II respectively

PZT A and C act as both actuator and sensor for collecting pulse-echo signal They were

installed on the plate which two of them are both-side collocated for the A0 Lamb wave

- 17 -

mode decomposition [39-40] PZT B was only used as a sensor for collecting pitch-catch sig-

nal For simulating the excitation and sensing of PZTs the pin-force model was adopted that

an input force apples to the end nodes of excitation PZT and averaging element strains corre-

sponding sensing area A Gaussian pulse (tone-burst) 50 kHz central frequency was excited

as an input signal

TABLE II Damage cases for numerical simulation

119949119949120783120783 (mm) 119941119941 (mm) G value reduction ()

Case 1 650 200 90 darr

Case 2 675 150 90 darr

Case 3 675 150 93 darr

Case 4 900 150 90 darr

- 18 -

44 Numerical simulation results

015 02 025 03 035

-2

0

2

Time (ms)

Am

plitu

de

IntactDamage

Reflected wave from damage

(a) Pitch-catch signal

015 02 025 03 035

-2

0

2

Time (ms)

Am

plitu

de

IntactDamage

Reflected waves from damage

(b) Pulse-echo signal Fig 7 Comparison of A0 Lamb wave mode signals from numerical simulation between intact and damage cases Time delay attenuation and multiple reflections due to damage simulated successfully using the damage modeling method described in the previous section

Fig 7 shows that comparison of A0 Lamb wave mode signals between intact and one of

damage cases in the simulation The features of A0 Lamb wave mode passing through dam-

age area described in section 2 are successfully simulated Thus the equivalent element

method is proper for modeling damage in composites In the damage cases the additional

waves reflected from damage is observed in both pitch-catch and pulse echo signal

Table III presents the result of the proposed damage characterization algorithm using the

numerical simulation cases shown in Table II In lsquoMethod 2rsquo 119907119907119894119894 variation is assumed to plusmn

5 of the normal value (1190~1315ms) The normal value of 119907119907119894119894 for lsquoMethod 3rsquo is meas-

ured as 1253 ms In each case the damage is characterized using the proposed technique

Note that as the information of 119907119907119894119894 is specified the estimated value is closer to the exact

value

- 19 -

TABLE III Damage characterization for the numerical simulation models

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

Case 1

Exact 650 200

Method 1 248 50~70 79~414 160~834

Method 2 248 603~667 135~241 271~486

Method 3 248 635 188 379

Case 2

Exact 675 150

Method 1 193 60~75 01~251 37~651

Method 2 193 618~682 114~222 295~575

Method 3 193 65 168 435

Case 3

Exact 675 150

Method 1 233 60~75 66~302 140~649

Method 2 233 668~738 85~196 182~419

Method 3 233 703 140 301

Case 4

Exact 900 150

Method 1 204 75~100 48~349 118~856

Method 2 204 860~951 107~216 264~530

Method 3 204 906 162 397

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1253 ms 1190~1315 ms Measured 119907119907119894119894 1253 ms

- 20 -

CHAPTER 5 EXPERIMENTAL VERIFICATION

51 Overview

In this section experimental study is investigated so that the proposed damage characteri-

zation technique Overall experimental setup and test specimen are shown in Fig 8 and 9

The experimental setup consists of test specimens data acquisition (DAQ) system and self-

sensing circuit Self-sensing circuit was installed for pulse-echo measurement using PZTs

[40-41]

PZT A(PZT C) PZT B

500

500

140

75

250

Teflon Inserted Damage (d =25)

Unit mm

Thickness = 3

PZT A(PZT C) PZT B

500

500

150

90

250

Impact Location

Unit mm

Thickness = 3

(a) Teflon inserted damage specimen

(b) Impact-induced damage specimen

Fig 8 Test specimens for the experimental verification In the Teflon inserted damage Teflon tapes with 25mm diameter were inserted at the center of the specimen The impact-induced damage was introduced to the specimen using the impact test equipment

- 21 -

511 Specimen

Two carbon composite plates with a dimension of 500times500times3mm3 were manufactured by

Nexcomes Inc by stacking 12 layers of CF3327 3K prepreg (woven fabrics type) manufac-

tured by Hankuk Fiber Inc The material properties of the specimen are presented in Table I

One of the specimens Teflon tapes with 25mm diameter was inserted in every 3 layers at the

center of the specimen to mimic the damage The other specimen is utilized for the character-

ization of real impact-induced damage

512 Data acquisition (DAQ) system and PZT

The data acquisition (DAQ) system consists of an arbitrary waveform generator (AWG) a

high speed signal digitizer (DIG) and two multiplexers (MUX) Using the 16-bits AWG

tone-burst input signal with a plusmn10 peak-to-peak voltage were generated The response signal

was measured by the 14-bit DIG For the signal-to-noise ratio reduction both pitch-catch and

pulse-echo signals were measured 150 times and averaged in the time domain 3 identical

circular PZTs of 10mm diameter manufactured by APC International Ltd were installed on

the specimen A Gaussian pulse (tone-burst) 50 kHz central frequency was excited as an in-

put signal The excitation frequency was chosen in order to avoid overlapping of the S0 and

A0 Lamb wave modes and generating of the higher modes (S1 A1hellip)

- 22 -

PZT Actuator

PZT Sensor

AWG

DIG

DAQ System Specimen

SensingActuatingMUX

1

23

4

1 Sending signal to MUX

4 Sending signal to DIG

Self-SensingCircuit

18 Unit mm

10

(a)

(b)

Fig 9 Overall experimental setup (a) Data acquisition (DAQ) system and (b) Installed circu-lar PZT transducer on the specimen with its dimension

513 Impact test

The impact-induced damage was introduced to the specimen using the impact test equip-

ment presented in Fig 10 A 5kg mass with 10mm tip was dropped from 015m height onto

the same location of the specimen The signals were obtained at each number of impacts

Since the reflected wave is not detected at the first impact case the data obtained from the

second and third impact were utilized for the analysis

10

Unit mm (a)

(b)

Fig 10 (a) Impact test equipment for introducing impact-induced damage and (b) Tip of the impact tester with 10mm diameter

- 23 -

514 Temperature effect test

In reality structures are subjected to changing environmental conditions such as tempera-

ture that affect measured signal and cause false-alarm Thus the robustness of the proposed

damage characterization technique under temperature variation in investigated (Fig 11) The

specimens were placed inside a temperature chamber and a thermocouple was installed on

the specimen to measure temperature of surface of the specimen The signal was obtained

under three temperature cases (0 20 and 50degC) and the humidity was kept at 30

Temperature Chamber

Test Specimen

Fig 11 Test specimen in temperature chamber

- 24 -

52 Teflon inserted damage

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Reflected wave from damage

(a)

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Transmitted WaveReflected Wave

Detected reflected wave

(b)

Fig 12 A0 Lamb wave mode obtained using pitch-catch method from the Teflon inserted damage specimen (a) The reflected wave from damage is shown (b) The Teflon inserted damage is detected by capturing the reflected wave

Fig 12 shows the obtained pitch-catch signal from the Teflon inserted damage specimen

and damage diagnosis result The Teflon inserted damage is detected successfully by captur-

ing the reflected wave Table IV presents the result of the Teflon inserted damage characteri-

zation 119907119907119894119894 is measured as 1213ms and the variation is assumed to plusmn 5 of the normal val-

ue (1152~1273ms) The results indicate that the proposed damage characterization approach

works well in real application

TABLE IV Damage characterization for the Teflon inserted damage specimen

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

Exact 635 25

Method 1 123 60~75 73~338 298~1220

Method 2 123 619~662 229~310 934~1188

Method 3 123 631 284 1161

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1213 ms

- 25 -

53 Impact-induced damage

11 115 12 125 13-4

-2

0

2

4

Time (ms)

Vol

tage

(mV

)

IntactImpact2Impact3

Reflected wave from damage

(a) Pitch-catch signal

11 115 12 125 13 135

-05

0

05

1

Time (ms)

Volta

ge (m

V)

IntactImpact2Impact3

Boundary reflection

Reflected wave from damage

(b) Pulse-echo signal

Fig 13 The signal obtained from the test specimen subjected to impact The impact was in-troduced at the same location of the specimen and the signal was obtained after each impacts As the number of impact is increased the transmitted A0 Lamb wave mode is delayed and attenuated

The obtained signals from impact-induced damage specimen are presented in Fig 13 It is

observed that as the number of impact is increased the transmitted A0 Lamb wave mode is

more delayed because the wave velocity in the damage area is reduced Fig 14 shows the

false-alarm test result whether the proposed damage detection algorithm works even in pris-

tine condition of the test specimen It is shown that no false alarm is occurred because no ad-

ditional reflected wave is captured

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Transmitted Wave

(a) Transmitted wave

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Reflected Wave

No reflected wave

(b) Reflected wave

Fig 14 False alarm test for the proposed damage detection algorithm in pristine condition of the specimen Reflected wave from damage is not captured false alarm is not occurred

- 26 -

TABLE V Damage characterization for the impact-induced damage specimen

of Impacts m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

2

(1470 J)

Method 1 080 65~90 75~246 470~1157

Method 2 080 801~885 114~245 711~1134

Method 3 080 843 179 1122

3

(2205 J)

Method 1 148 50~100 06~380 21~1134

Method 2 148 758~838 249~368 843~1125

Method 3 148 798 308 1044

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1213 ms

Table V presents the characterization results of the impact-induced damage case Here the

measured 119907119907119894119894 and the assumed variation are same as the artificial damage case It is observed

that as the number of impact is increased 119889119889 is increased while 119907119907119889119889 is reduced Thus it is

validated that the proposed damage characterization approach estimates the increment of the

damage well

54 Temperature effect test

541 Teflon inserted damage

11 115 12 125 13-10

-5

0

5

10

Time (ms)

Vol

tage

(mV

)

0 oC20 oC50 oC

(a) Pitch-catch signal

11 115 12 125 13 135

-1

0

1

2

3

Time(ms)

Volta

ge (m

V)

0 oC20 oC50 oC

(b) Pulse-echo signal

Fig 15 Signals obtained from the Teflon inserted damage specimen under temperature varia-tion Time delay and amplitude variation are occurred

Fig 15 shows the obtained signals from the Teflon inserted damage specimen subjected to

- 27 -

temperature variation Time delay and amplitude variations are occurred there is no addition-

al damage is introduced The analyzed results of the obtained signals are reported in Table VI

The measured 119907119907119894119894 is 1247 ms at 0 degC 1213 ms at 20 degC and 1168 ms at 50 degC respectively

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms (1152~1273 ms)

The estimated results are also changed under temperature variations the fluctuation of esti-

mated range and exact value is small

TABLE VI Temperature effect test for the Teflon inserted damage specimen

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

0 degC

Exact 635 25

Method 1 113 60~70 79~268 352~1291

Method 2 113 560~619 233~344 1034~1172

Method 3 113 606 257 1141

20 degC

Exact 635 25

Method 1 123 60~75 73~338 298~1220

Method 2 123 619~662 229~310 934~1188

Method 3 123 631 284 1161

50 degC

Exact 635 25

Method 1 137 60~75 72~328 263~1228

Method 2 137 622~687 183~299 666~1086

Method 3 137 642 264 960

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1247 ms at 0 degC 1213 ms at 20 degC and 1168 ms at 50 degC

- 28 -

542 Impact-induced damage

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Vol

tage

(mV

)

Transmitted WaveReflected Wave

No reflected wave

(a) 0 degC

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Transmitted WaveReflected Wave

No reflected wave

(b) 50 degC

Fig 16 False alarm test under temperature variations It is shown that proposed damage char-acterization algorithm is robust under temperature variations

The false alarm test in pristine condition of the impact-induced damage specimen under

temperature variation was investigated (Fig 16) It is validated that the proposed damage di-

agnosis algorithm is robust under temperature variations Fig 17 shows the obtained signals

from the impact-induced damage (impacted 3 times) specimen under temperature variation

The analyzed results of the obtained signals are reported in Table VII The measured 119907119907119894119894 in

each temperature and the assumed variation are same as the artificial damage case

11 115 12 125 13-5

0

5

Time (ms)

Volta

ge (m

V)

0 oC20 oC50 oC

(a) Pitch-catch signal

11 115 12 125 13 135

-1

0

1

2

3

Time(ms)

Vol

tage

(mV

)

0 oC20 oC50 oC

(b) Pulse-echo signal

Fig 17 Signals obtained from impact-induced damage specimen (Impact 3) under tempera-ture variation

- 29 -

TABLE VII Temperature effect test for the impact induced damage specimen (Impact 3 case)

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

0 degC

Method 1 135 75~90 72~310 267~1148

Method 2 135 705~779 264~381 978~1126

Method 3 135 763 289 1071

20 degC

Method 1 148 50~100 06~380 21~1134

Method 2 148 758~838 249~368 843~1125

Method 3 148 798 308 1044

50 degC

Method 1 153 75~100 53~415 174~1227

Method 2 153 804~889 214~336 702~1103

Method 3 153 815 320 1050

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1247 ms at 0 degC 1213 ms at 20 degC and 1168 ms at 50 degC

- 30 -

CHAPTER 6 CONCLUDING REMARKS

61 Executive summary

In this study the reference-free damage detection localization and quantification tech-

nique in a multi layered composite plate is proposed Additional reflected A0 Lamb wave

mode from the damage due to multiple reflections in damage area is used as a feature for

characterizing the damage First damage is characterized in terms of a damage index (m-

value) defined as the ratio of the physical length of the damage to the wavelength of the A0

Lamb wave mode Then the range of possible damage location length and severity are esti-

mated with constraints This procedure can be accomplished without prior baseline data To

realize the proposed technique the matching pursuit technique is modified for reduction of

the error by considering the dispersive characteristic of Lamb wave The matching pursuit

method estimates the arrival time of each wave packets accurately even two wave packets are

overlapped The numerical simulations and experimental tests including Teflon tape inserted

and impact-induced damage are investigated for the feasibility of the proposed technique

Moreover the effectiveness of the proposed technique under changing temperature is exper-

imentally verified

- 31 -

62 Suggestion for further study

The limitations of this study and the possible improvement and further studies are listed as

follow

1 Measurement of impact-induced damage length using conventional NDT technique

Additional experimental verification for impact-induced damage characterization is

needed by measuring the real length of damage using conventional NDT technique

such as ultrasonic scan or infrared camera

2 Limitation of both side collocated PZT installation In this study excitation PZTs are

installed both side and collocated for decomposing A0 Lamb wave mode However it

can produce decomposition error due to imperfection of PZT location and their installa-

tion Moreover in many cases assess to both surface of structures is not allowed such

as aircraft applications To improve this limitation a reference-free damage characteri-

zation using single side PZTs will be considered

3 Complex structure application In this study damage is characterized in simple plate-

like structure When the monitored structure becomes complex the response signals al-

so becomes complex due to its geometry and boundary Thus the investigations for

complex structure application such as stiffener spar and curvature are needed

- 32 -

REFERENCES

[1] Cawley P and Alleyne D (1996) ldquoThe use of Lamb waves for the long range inspection of large structuresrdquo Ultrasonics 34 pp287-290

[2] Ghosh T Kundu T and Karpur P (1998) ldquoEfficient use of Lamb waves for detecting defects in large platesrdquo Ultrasonics 36 pp791-801

[3] Guirguitui V (2003) ldquoLamb wave generation with piezoelectric wafer active sensors for structural health monitoringrdquo SPIE Conference on Smart Structures and Materials San Diego CA 5391 pp111-422

[4] Ragvahan A and Cesnik CE S (2004) ldquoModeling of piezoelectric-based Lamb wave generation and sensing for structural health monitoringrdquo SPIE Conference on Smart Struc-tures and Materials San Diego CA 5391 pp419-430

[5] Rose JL Pilarski A and Dirit JJ (1993) ldquoAn approach to guided wave mode selection for inspection of laminated platerdquo Journal of Reinforced Plastics and Composites 12(5) pp536-544

[6] Guo N and Cawley P (1993) ldquoThe interaction of Lamb waves with delaminations in composite laminatesrdquo Journal of Acoustical Society of America 94(4) pp 2240-2246

[7] Valdez SHD and Soutis C (2002) ldquoReal-time nondestructive evaluation of fiber com-posite laminates using low-frequency Lamb wavesrdquo Journal of Acoustical Society of Amer-ica 111(5) pp 2026-2033

[8] Yang M and Qiao P (2005) ldquoModeling and experimental detection of damage in virous materials using the pulse-echo method and piezoelectric sensoractuatorsrdquo Smart Materials and Structures 14 pp1083-1100

[9] Lestari W and Qiao P (2005) ldquoApplication of wave propagation analysis for damage identification in composite laminated beamsrdquo Journal of Composite Materials 39(22) pp1967-1984

[10] Kessler S S Spearing S M and Soutis C (2002) ldquoDamage detection in composite ma-terials using Lamb wave methodsrdquo Smart Materials and Structures 11 pp269-278

[11] Sohn H Park G Wail J R Limback N P and Farrar C R (2004) ldquoWavelet-based active sensing for delamination detection in composite structuresrdquo Smart Materials and Structures 13 pp 153-160

[12] Petculescu G Krishnaswamy S and Achenback J D (2008) ldquoGroup delay measure-ments using modally selective Lamb wave transducers for detection and sizing of delamina-tions in compositesrdquo Smart Materials and Structures 17

[13] Sohn H Park H W Law K H and Farrar C R (2007) ldquoDamage detection in compo-

- 33 -

site plates by using an enhanced time reversal methodrdquo Journal of Aerospace Engineering (ASCE) pp 141-151

[14] Sohn H (2007) ldquoEffects of environmental and operational variability on structural health monitoringrdquo Philosophical Transactions of the Royal Society Ardquo 365 pp539-560

[15] Rytter A (1993) ldquoVibration based inspection of civil engineering structuresrdquo PhD The-sis Department of Building Technology and Structural Engineering University of Aalborg Aalborg Denmark

[16] Hayashi T and Kawashima K (2002) ldquoMultiple reflections of Lamb waves at a delami-nationrdquo Ultrasonics 40 pp193-197

[17] Wang C H and Rose L R F (2003) ldquoWave reflection and transmission in beams con-taining delamination and inhomogeneityrdquo Journal of Sound and Vibration 264 pp 851-872

[18] Ramadas C Balasubramaniam K Joshi M and Krishnamurthy C V (2009) ldquoInterac-tion of the primary anti-symmetric Lamb mode (Ao) with symmetric delaminations numer-ical and experimental studiesrdquo Smart Materials and Structures 18

[19] Ramadas C Balasubramaniam K Joshi M and Krishnamurthy C V (2010) ldquoInterac-tion of guided Lamb waves with an asymmetrically located delamination in a laminated composite platerdquo Smart Materials and Structures 19

[20] Wilcox P Lowe M and Cawely P (2001) ldquoThe effect of dispersion on long-range in-spection using ultrasonic guided wavesrdquo NDTampE International 34 pp1-9

[21] Mallat S G and Zhang Z (1993) ldquoMatching pursuits with time-frequency dictionariesrdquo IEEE Tansactions on Signal Processing 41(12) pp3397-3415

[22] Zhang G Zhang S and Wang Y (2000) ldquoApplication of adaptive time-frequency de-composition in ultrasonic NDE of highly scattering materialsrdquo Ultrasonics 38 pp 961-964

[23] Hong J C Sun K H and Kim Y Y (2005) ldquoThe matching pursuit approach based on the modulated Gaussian pulse for efficient guided-wave damage inspectionrdquo Smart Mate-rials and Structures 14 pp548-560

[24] Herakovich C T (1997) Mechanics of Fibrous Composites Wiley 480 pages

[25] Reddy J N (1985) ldquoA review of the literature on finite-element modeling of laminated composite platesrdquo The Shock and Vibration Digest 17(4) pp 3-8

[26] Reddy J N (2003) Mechanics of Laminated Composite Plates and Shells Theory and Analysis CRC Press 856 pages

[27] Palacz M Krawczuk M and Ostachowicz W (2005) ldquoThe spectral finite element mod-el for analysis of flexure-shear coupled wave propagation Part 1 laminated multilayer

- 34 -

composite beamrdquo Composite Structures 68 pp37-44

[28] Zhang Y C and Harding J (1990) ldquoA numerical micromechanics analysis of the me-chanical properties of a plain weave compositerdquo Computer amp Structures 36(5) pp 839-844

[29] Nemat-Nasser S and Hori M (1999) Micromechanics Overall Properties of Heteroge-neous Materials Elsevier 810 pages

[30] Scida D Aboura Z Benzeggagh M L and Bocherens E (1999) ldquoA micromechanics model for 3D elasticity and failure of woven-fibre composite materialrdquo Composites Science and Technology 59 pp 505-517

[31] Thomas Y and Wu X H (1997) ldquoA multiscale finite element method for elliptic prob-lems in composite materials and porous mediardquo Journal of Computational Physics 134 pp 169-189

[32] Lee K H Moorthy S and Ghosh S (1999) ldquoMultiple scale computational model for damage in composite materialsrdquo Computer Methods in Applied Mechanics and Engineer-ing 172 pp 175-201

[33] Namilae S and Chandra N (2005) ldquoMultiscale model to study the effect of interfaces in carbon nanotube-based compositesrdquo Journal of Engineering Materials and Technology (ASME) 127 pp222-232

[34] Adams D E (2007) Health monitoring of Structural Materials and Components Wiley 460 pages

[35] Gopalakrishnan S (2009) ldquoModeling Aspects in Finite Elementrdquo Encyclopedia of Struc-tural Health Monitoring 2 pp791-809

[36] Palacz M Krawczuk M and Ostachowicz W (2005) ldquoThe spectral finite element mod-el for analysis of flexure-shear coupled wave propagation Part 2 Deminated multilayer composite beamrdquo Composite Structures 68 pp45-51

[37] Baik J M and Thopson R B (1984) ldquoUltrasonic scattering from imperfect interfaces A quasi-static modelrdquo Journal of Nondestructive Evaluation 4(3) pp177-196

[38] Luo H and Hanagud S (2000) ldquoDynamics of delamincated beamsrdquo International Jour-nal of Solids and Structures 37 pp1501-1519

[39] Kim S B and Sohn H (2007) ldquoInstantaneous reference-free crack detection based on polarization characteristics of piezoelectric materialsrdquo Smart Materials and Structures 16 pp2375-2387

[40] An Y K and Sohn H (2010) ldquoInstantaneous crack detection under varying temperature and static loading conditionsrdquo The Special Issue of Structural Control and Health Monitor-ing 17(7) pp 730-741

- 35 -

[41] Lee S J and Sohn H (2006) ldquoActive self-sensing scheme development for structural health monitoringrdquo Smart Materials and Structures 15(6) pp 1734-1746

- 36 -

요 약 문

복합재료 손상의 무기저 진단 및 정량화

복합재료는 재료의 경량성 내구성 고강도의 특성으로 인하여 항공기에서부터

교량에까지 널리 사용되고 있다 하지만 외부의 충격으로 인한 복합재료 내부의

손상은 재료의 강도를 감소시켜 구조물의 안정성을 저하시키는 요인이 된다 하지만

복합재료의 손상은 재료 내부에서 발생하여 육안으로는 식별하기도 힘들며 표면에서

보이는 것만으로 내부의 손상 정도를 판단하기는 불가능하다는 단점을 가지고 있다

본 연구에서는 손상 내부에서 반사로 인해 발생하는 파를 이용하여 기저자료를

사용하지 않고 특정 시점에서 얻은 자료만으로 복합재료의 손상을 진단하고 정량화

할 수 있는 기법을 제안한다 제안된 기법의 적용성을 검증하기 위하여 유한요소법을

이용한 수치해석을 수행하였고 Teflon tape 이 삽입된 시편과 실제 충격하중으로

인한 손상을 입은 시편을 이용한 실험을 진행하였다 수치해석 및 실험 결과 제안된

기법이 복합재료의 손상을 성공적으로 진단 및 정량화할 수 있음을 확인하였다 또한

다양한 외부 온도 환경에서 제안된 기법의 환경영향 실험을 수행하였고 다양한

환경에서 손상의 진단 및 정량화가 성공적으로 이루어짐을 검증하였다

핵심어 무기저 손상진단 및 정량화 압전센서 복합재료 유한요소법 Matching pursuit

method

- 37 -

감 사 의 글

기대와 두려움을 안고 시작한 석사과정이 어느덧 2 년 이라는 세월 속에서 추억으로 자리잡아

가고 있습니다 학부생 기간을 포함한 지난 3 년 동안 SHM 이라는 새로운 분야를 접하고 연구를

수행 할 수 있도록 도와주신 고마운 분들이 있어 그 감사를 표하고자 합니다

우선 저의 많은 부족한 점을 일깨워 주시고 연구에 대한 뜻을 더욱 확고히 세울 수 있게 해

주신 지도교수님이신 손 훈 교수님께 깊은 감사의 뜻을 전합니다 짧은 기간이었지만 교수님의

도움으로 많은 것을 배우고 느끼고 깨달을 수 있었습니다 또한 본 연구를 수행할 수 있도록 지

원해주신 국토해양부 (U-City 석 박사과정 지원사업)에 깊은 감사 드립니다

때로 힘들고 지칠 때 항상 저를 믿고 응원해 주시는 가족이 있었기에 지금의 글을 쓸 수 있었

습니다 또한 지금 연구실에서 함께 생활하고 있는 윤규형 민구형 철민이형 현석이형 진열이형

호민이형 병진이 선혜누나와 성대에 계신 창길이형 미국에 계신 의재형과 Debaditiya 에게도 감

사의 말씀을 전합니다 그리고 모두 나열할 수는 없지만 카이스트 생활의 활력소가 되어준 카이

스트 건설 및 환경공학과 선후배들 직원 분들과 SOS 멤버들에게도 감사를 표합니다

앞서 거론한 많은 이들의 도움이 없었다면 저 또한 지금 이 자리에 있을 수 없었을 것입니다

많이 부족하기에 더 노력하고 배우고자 했던 초심을 다시 한번 생각하며 더 큰 결실을 맺기 위해

앞으로 나아가겠습니다 도움을 주신 모든 분들께 이 감사의 글을 바칩니다

2011년 6월 카이스트에서 임형진 올림

- 38 -

CURRICULUM VITAE

Hyung Jin Lim

Research Assistant

The Civil and Environmental Engineering Department

Korea Advanced Institute of Science and Technology (KAIST)

291 Daehak-Ro Guseong-Dong Yuseong-Gu

Daejeon Republic of Korea 305-701

Phone (82)+42-350-3665 Fax (82)+42-350-3610

Email limhj87gmailcom

RESEARCH INTERESTS

Damage Detection Localization and Quantification on Composite and Metallic Structures Pattern Recognition Finite Element Modeling (FEM)

EDUCATION

2009 ndash 2011 MS Civil and Environmental Eng Korea Advanced Institute of Science and Technology (KAIST) Korea Rep

2006 ndash 2009 BS Civil and Environmental Eng Korea Advanced Institute of Science and Technology (KAIST) Korea Rep

JOURNAL PUBLICATIONS The corresponding author is underlined 1 Hyung Jin Lim Hoon Sohn ldquoReference-free damage detection localization and quantification

using in compositerdquo in preparation for Journal of the Acoustical Society of America 2011

2 Chul Min Yeum Hoon Sohn Jeong Bum Ihn Hyung Jin Lim ldquoInstantaneous delamination detection in a composite using a dual piezoelectric transducer networkrdquo Submitted to Compo-sites Science and Technology 2011

3 Yun-Kyu An Hyung Jin Lim Min Koo Kim Jin Yeol Yang Hoon Sohn Chang Guen Lee ldquoInvestigation of applicability of local reference-free damage detection techniques to in-situ

- 39 -

bridgesrdquo Submitted to ASCE Journal of Structural Engineering 2011

4 Hyung Jin Lim Min Koo Kim Hoon Sohn Chan Yik Park ldquoImpedance based damage detec-tion under varying temperature and loading conditionsrdquo Submitted to NDTampE International 2011

5 Hyung Jin Lim Hoon Sohn ldquoQuantitative estimation of transmitted and reflected lamb waves at discontinuityrdquo Journal of the Korean Society for Nondestructive Testing Vol30 No 4 pp1-8 2010

CONFERENCE PROCEEDINGS

The corresponding author is underlined 1 Hyung Jin Lim and Hoon Sohn ldquoReference-free damage detection localization and quantifi-

cation in compositesrdquo ASCE Engineering Mechanics Institute Conference Boston June 2-4 2011

2 Hyung Jin Lim and Hoon Sohn ldquoReference-free delamination quantification within multi-layer composite materialsrdquo KSNT Spring Conference May 26-27 Gyeong-Ju Korea 2011

3 Min Koo Kim Hyun Jin Lim Hoon Sohn Chan Yik Park ldquoImpedance-based bolt loosening detection under varying temperature and loadingrdquo Asian Pacific Workshop on Structural Health Monitoring Tokyo Japan November 30-December 2 2010

4 Hyung Jin Lim Chul Min Yeum and Hoon Sohn ldquoModeling of Impact-induced Delamination in a Multilayer Composite Platerdquo The 23rd KKCNN Symposium on Civil Engineering Taipei (Taiwan) November 13-15 2010

5 Yun-Kyu An Hyung Jin Lim and Hoon Sohn ldquoTemperature independent Delamination Detec-tion using data normalizationrdquo The 3rd International Conference on Multi-Functional Materi-als and Structures Chonbuk National Univ Jeonju Korea September 14-18 2010

6 Hoon Sohn Yun-Kyu An Chang Gil Lee Min Koo Kim Hyung Jin Lim ldquoApplication of online NDT techniques to decommissioned bridge testingrdquo the 5th European Workshop on Structural Health Monitoring Sorrento Italy June 29-July 02 2010

7 Min Koo Kim Hyung Jin Lim Hoon Sohn Chan Yik Park ldquoImpedance based damage diag-nosis of a complex composite aircraft wing under changing loading and temperature condi-tionsrdquo the 5th European Workshop on Structural Health Monitoring Sorrento Italy June 29-July 02 2010

8 Yun-Kyu An Min Koo Kim Hyung Jin Lim and Hoon Sohn ldquoValidations of reference-free crack detection techniques through a decommissioned bridge testrdquo 2010 COSEIK Annual Con-ference Jeju Korea April 8-9 2010

- 40 -

PARTICIPANT PROJECTS

1 Development of Smart Steel Members for Online Steel Structure Monitoring POSCO Corporation (Funded 41000000 KRW (41000 USD) for 010511 to 123111)

2 Development of Enabling Technologies for Structural Health Monitoring Korean Agency for Defense Development (Funded 240000000 KRW (240000 USD) out of total 3200000000 KRW (3200000 USD) for 060108-053111)

3 Development of On-board SHM Technologies for Composite Air Vehicles The Boeing Company (Funded 400000 USD (400000000 KRW) for 080108 to 073111)

EDUCATIONAL EXPERIENCES

Fall 2009

Teaching Assistant CE207 ldquoElementary Structural Engineering and Laboratoryrdquo KAIST 3 unit undergraduate course

EXPERIMENTAL EXPERIENTCES

2010 Monitoring of Real Aircraft Composite Wing Segment Korea Agency for Defense De-velopment (ADD) Daejeon Korea Real time monitoring of composite air vehicles fund-ed by Korea Agency for Defense Development (ADD)

2009-2011

Field Monitoring of Real Bridge Structures Gimpo (Ramp G bridge) amp Icheon (Sam-seung bridge) Korea US-Korea collaborative research for bridge monitoring testbeds

HONORS amp AWARDS

Scholarships for outstanding students granted KAIST (2007-2009) The three prominent students in the civil and environmental engineering department of KAIST were selected as recipients This scholarship provided 2000$ per year for three years

- 41 -

Page 23: 복합재료 손상의 무기저 진단 및 정량화 - KAISTssslab.kaist.ac.kr/article/pdf/Thesis_2.pdf · 1 Declaration of Ethical Conduct in Research: I, as a graduate student

1198971198972 = 1198971198971119905119905 minus 1199051199051 minus 119905119905119889119889

1199051199051 (6)

119889119889 = 119897119897 minus 1198971198971 minus 1198971198972 (7)

Since a PZT for pulse-echo method acts as both an actuator and a sensor the obtained sig-

nal contains both the excitation and the response signal Thus the reflected wave from the

entrance of damage can be buried into the tail of excitation signal and cause error on arrival

time of reflected wave from damage Even in this case damage can be localized and quan-

tized using the constraints When first wave packet reflected from the entrance of damage is

buried into the excitation signal the next wave packet reflected from the exit of damage can

be utilized alternatively If the algorithm regards the second reflected wave as the first re-

flected wave there are no combinations of estimated variable meeting the constraints Then

the algorithm repeats the process by regarding the wave as the second reflected wave the

constraints can select proper combination of the estimated variables

The estimated values have range bounded by the constraints When additional information

of 119907119907119894119894 is utilized the damage can be characterized more accurately In this study three meth-

ods are proposed and verified as information level of 119907119907119894119894

Method 1 Solely based on the current data measurement (Reference-free)

Method 2 Limit the possible range of 119907119907119894119894 (Prior knowledge of the system)

Method 3 Measure 119907119907119894119894 (Additional data need to be collected)

Method 1 described in this section can be utilized without any information of 119907119907119894119894 Meth-

od 2 produces more accurate estimation results using prior knowledge of the system such as

- 12 -

119907119907119894119894 variation under temperature changes Method 3 gives unique estimated value using meas-

ured 119907119907119894119894 from other path in the same time In section 4 and 5 damage is characterized using

these three methods and the results are compared each another

36 Matching pursuit method considering the dispersion effect of Lamb

wave

Matching pursuit method has been developed for signal compression approximation and

characterization of obtained signal and function [21-23] The key concept of the matching

pursuit method is that any signal (function) can be decomposed into a linear expansion of

signals selected from a pool of basis signals (functions) called lsquodictionaryrsquo

Let 119891119891 be a function to be analyzed then 119891119891 can be approximated a set of functions select-

ed from the dictionary 119863119863 in order to best match This is done by successive approximations

with orthogonal projections Let1198921198921205741205740 isin 119863119863 the function 119891119891 can be decomposed into

119891119891 = lang1198911198911198921198921205741205740rang1198921198921205741205740 + 119877119877119891119891 (8)

where 119877119877119891119891 is the residual function after approximating 119891119891 using 1198921198921205741205740 1198921198921205741205740 is orthogonal to

119877119877119891119891 hence

1198911198912 = lang1198911198911198921198921205741205740rang2

+ 1198771198771198911198912 (9)

To minimize 119877119877119891119891 1198921198921205741205740 is chosen to maximize lang1198911198911198921198921205741205740rang The matching pursuit method is

an iterative algorithm that decomposes the residue 119877119877119891119891 by projecting it on a function of 119863119863

that matches 119877119877119891119891 almost at best as it was done for 119891119891 The residue 119877119877119899119899119891119891 is decomposed into

119877119877119899119899119891119891 = lang119877119877119899119899119891119891119892119892120574120574119899119899rang119892119892120574120574119899119899 + 119877119877119899119899+1119891119891 (10)

- 13 -

the decomposition is carried up to the order 119898119898 119891119891 can be expressed as

119891119891 = lang119877119877119899119899119891119891119892119892120574120574119899119899rang119892119892120574120574119899119899

119898119898minus1

119899119899=0

+ 119877119877119898119898119891119891 (11)

The original function 119891119891 is decomposed into a sum of dictionary elements that are chosen to

best match its residues

In this study the matching pursuit method is utilized to accurately decompose of the over-

lapped Gaussian pulses and estimate the parameters of the pulse which used for damage

characterization The 119894119894119905119905ℎ Gaussian pulse in the dictionary for approximating the obtained

signal is expressed as

119892119892120574120574119894119894 = 119891119891(119860119860119894119894 119906119906119894119894 119904119904119894119894 119889119889119894119894120601120601119894119894) (12)

where 119860119860119894119894119906119906119894119894119904119904119894119894119889119889119894119894 and 120601120601119894119894 represent amplitude arrival time scale (pulse duration) excita-

tion(central) frequency and phase of the Gaussian pulse respectively The parameters of the

pulse are estimated using a nonlinear regression based optimization method (Gauss-Newton

method) based on an initial guess of the parameters Detail procedure for estimating the pa-

rameters of the pulse is expressed in Hong et al [23]

Among the estimated parameters of the pulses pulse duration (scale) represents the disper-

sion effect If a wave-packet is spread out the estimated pulse duration is increased Using

this feature the matching pursuit method is modified to capture waves which have longer

pulse duration than transmitted A0 Lamb wave mode If the estimate pulse duration of a wave

is less than the transmitted wave the modified matching pursuit method regards it as error

- 14 -

CHAPTER 4 NUMERICAL SIMULATION

41 Composite modeling method

Composite materials can be modeled from nano to macro level depends on its engineering

purpose In general four methods are utilized for modeling of composite materials First

composite is modeled as a homogenous material expressed in terms of effective material

properties [24] The effective material properties are estimated using the modulus and the

volume fraction of each constituent theoretically (Rule of mixture) and measured experimen-

tally Second a model is established that each laminate which have own material properties

are totally bonded together In this model the stacking sequence and direction of each lami-

nates are considered additionally [25-27] Third method for composite modeling in micro

level called micromechanics [28-30] This model considers the interaction between fibers and

matrix in laminate Last multiscale modeling which the scope of modeling is extended to

molecular or quantum level of constituents [31-33] This model starts from molecular me-

chanics or molecular dynamics then the scale of model leads to the nano micro and macro

level by incorporating small scale models As the scale of models is smaller it provides more

logical way than other large scale of models while the computational cost is more expensive

The scope of this study is focused on the dynamic behavior of composite structure the beam

or plate model was utilized for the numerical simulation

- 15 -

42 Damage modeling method

Generally three damage models have been developed and utilized in the real applications

First the stiffness reduction model called as the equivalent element model was introduced

[17 34-35] This model is simply realized by replacing the region around damage area by a

region with adjusted material properties (Youngrsquos modulus shear modulus density etc)

globally Next the duplicated node model is done by disconnecting the element connectivity

assuming there is no interaction among the delaminated surface This model can be utilized

when the location and the extension of damage is important issue [34-36] Last the contact

mechanics model using the principles of contact mechanics the two interfacial forces (con-

tact resistance and sliding resistance) in the delaminated surface [37-38] In this model the

delaminated surfaces are modeled as nonlinear functions of the interlaminar normal strain

and the interlaminar shear strain This model is computationally expensive while provides

more logical way than other models There is however no universally accepted model for

general impact-induced damage in composites In the case of impact-induced damage it is

impossible to predict or estimate the location and number of delaminated layers Thus from

the uncertainty of impact-induced damage the equivalent element model can be considered

as the effective damage model

- 16 -

43 Numerical simulation description

TABLE I Material property of simulation model and test specimen

Tensile Strength 657 MPa Tensile Modulus 59 GPa Shear Strength 804 MPa Shear Modulus 24 Gpa Specific Gravity 16

PZT A

PZT C

PZT BDamage

3

250 125 125

500 Unit mm

Fig 6 Description of the numerical simulation model

The feasibility of the proposed damage characterization technique is verified by 2-D FEM

simulation using MSC PATRAN and NASTRAN Fig 6 presents the dimension PZTs and

damage location of the simulation model In the model 05mm x 05mm square shell element

was used for modeling of the A0 mode whose flexural wave characteristic The element size

was determined by considering the wavelength of the propagation Lamb wave modes The

material properties of the simulation model and damage cases are presented in Table I and

Table II respectively

PZT A and C act as both actuator and sensor for collecting pulse-echo signal They were

installed on the plate which two of them are both-side collocated for the A0 Lamb wave

- 17 -

mode decomposition [39-40] PZT B was only used as a sensor for collecting pitch-catch sig-

nal For simulating the excitation and sensing of PZTs the pin-force model was adopted that

an input force apples to the end nodes of excitation PZT and averaging element strains corre-

sponding sensing area A Gaussian pulse (tone-burst) 50 kHz central frequency was excited

as an input signal

TABLE II Damage cases for numerical simulation

119949119949120783120783 (mm) 119941119941 (mm) G value reduction ()

Case 1 650 200 90 darr

Case 2 675 150 90 darr

Case 3 675 150 93 darr

Case 4 900 150 90 darr

- 18 -

44 Numerical simulation results

015 02 025 03 035

-2

0

2

Time (ms)

Am

plitu

de

IntactDamage

Reflected wave from damage

(a) Pitch-catch signal

015 02 025 03 035

-2

0

2

Time (ms)

Am

plitu

de

IntactDamage

Reflected waves from damage

(b) Pulse-echo signal Fig 7 Comparison of A0 Lamb wave mode signals from numerical simulation between intact and damage cases Time delay attenuation and multiple reflections due to damage simulated successfully using the damage modeling method described in the previous section

Fig 7 shows that comparison of A0 Lamb wave mode signals between intact and one of

damage cases in the simulation The features of A0 Lamb wave mode passing through dam-

age area described in section 2 are successfully simulated Thus the equivalent element

method is proper for modeling damage in composites In the damage cases the additional

waves reflected from damage is observed in both pitch-catch and pulse echo signal

Table III presents the result of the proposed damage characterization algorithm using the

numerical simulation cases shown in Table II In lsquoMethod 2rsquo 119907119907119894119894 variation is assumed to plusmn

5 of the normal value (1190~1315ms) The normal value of 119907119907119894119894 for lsquoMethod 3rsquo is meas-

ured as 1253 ms In each case the damage is characterized using the proposed technique

Note that as the information of 119907119907119894119894 is specified the estimated value is closer to the exact

value

- 19 -

TABLE III Damage characterization for the numerical simulation models

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

Case 1

Exact 650 200

Method 1 248 50~70 79~414 160~834

Method 2 248 603~667 135~241 271~486

Method 3 248 635 188 379

Case 2

Exact 675 150

Method 1 193 60~75 01~251 37~651

Method 2 193 618~682 114~222 295~575

Method 3 193 65 168 435

Case 3

Exact 675 150

Method 1 233 60~75 66~302 140~649

Method 2 233 668~738 85~196 182~419

Method 3 233 703 140 301

Case 4

Exact 900 150

Method 1 204 75~100 48~349 118~856

Method 2 204 860~951 107~216 264~530

Method 3 204 906 162 397

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1253 ms 1190~1315 ms Measured 119907119907119894119894 1253 ms

- 20 -

CHAPTER 5 EXPERIMENTAL VERIFICATION

51 Overview

In this section experimental study is investigated so that the proposed damage characteri-

zation technique Overall experimental setup and test specimen are shown in Fig 8 and 9

The experimental setup consists of test specimens data acquisition (DAQ) system and self-

sensing circuit Self-sensing circuit was installed for pulse-echo measurement using PZTs

[40-41]

PZT A(PZT C) PZT B

500

500

140

75

250

Teflon Inserted Damage (d =25)

Unit mm

Thickness = 3

PZT A(PZT C) PZT B

500

500

150

90

250

Impact Location

Unit mm

Thickness = 3

(a) Teflon inserted damage specimen

(b) Impact-induced damage specimen

Fig 8 Test specimens for the experimental verification In the Teflon inserted damage Teflon tapes with 25mm diameter were inserted at the center of the specimen The impact-induced damage was introduced to the specimen using the impact test equipment

- 21 -

511 Specimen

Two carbon composite plates with a dimension of 500times500times3mm3 were manufactured by

Nexcomes Inc by stacking 12 layers of CF3327 3K prepreg (woven fabrics type) manufac-

tured by Hankuk Fiber Inc The material properties of the specimen are presented in Table I

One of the specimens Teflon tapes with 25mm diameter was inserted in every 3 layers at the

center of the specimen to mimic the damage The other specimen is utilized for the character-

ization of real impact-induced damage

512 Data acquisition (DAQ) system and PZT

The data acquisition (DAQ) system consists of an arbitrary waveform generator (AWG) a

high speed signal digitizer (DIG) and two multiplexers (MUX) Using the 16-bits AWG

tone-burst input signal with a plusmn10 peak-to-peak voltage were generated The response signal

was measured by the 14-bit DIG For the signal-to-noise ratio reduction both pitch-catch and

pulse-echo signals were measured 150 times and averaged in the time domain 3 identical

circular PZTs of 10mm diameter manufactured by APC International Ltd were installed on

the specimen A Gaussian pulse (tone-burst) 50 kHz central frequency was excited as an in-

put signal The excitation frequency was chosen in order to avoid overlapping of the S0 and

A0 Lamb wave modes and generating of the higher modes (S1 A1hellip)

- 22 -

PZT Actuator

PZT Sensor

AWG

DIG

DAQ System Specimen

SensingActuatingMUX

1

23

4

1 Sending signal to MUX

4 Sending signal to DIG

Self-SensingCircuit

18 Unit mm

10

(a)

(b)

Fig 9 Overall experimental setup (a) Data acquisition (DAQ) system and (b) Installed circu-lar PZT transducer on the specimen with its dimension

513 Impact test

The impact-induced damage was introduced to the specimen using the impact test equip-

ment presented in Fig 10 A 5kg mass with 10mm tip was dropped from 015m height onto

the same location of the specimen The signals were obtained at each number of impacts

Since the reflected wave is not detected at the first impact case the data obtained from the

second and third impact were utilized for the analysis

10

Unit mm (a)

(b)

Fig 10 (a) Impact test equipment for introducing impact-induced damage and (b) Tip of the impact tester with 10mm diameter

- 23 -

514 Temperature effect test

In reality structures are subjected to changing environmental conditions such as tempera-

ture that affect measured signal and cause false-alarm Thus the robustness of the proposed

damage characterization technique under temperature variation in investigated (Fig 11) The

specimens were placed inside a temperature chamber and a thermocouple was installed on

the specimen to measure temperature of surface of the specimen The signal was obtained

under three temperature cases (0 20 and 50degC) and the humidity was kept at 30

Temperature Chamber

Test Specimen

Fig 11 Test specimen in temperature chamber

- 24 -

52 Teflon inserted damage

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Reflected wave from damage

(a)

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Transmitted WaveReflected Wave

Detected reflected wave

(b)

Fig 12 A0 Lamb wave mode obtained using pitch-catch method from the Teflon inserted damage specimen (a) The reflected wave from damage is shown (b) The Teflon inserted damage is detected by capturing the reflected wave

Fig 12 shows the obtained pitch-catch signal from the Teflon inserted damage specimen

and damage diagnosis result The Teflon inserted damage is detected successfully by captur-

ing the reflected wave Table IV presents the result of the Teflon inserted damage characteri-

zation 119907119907119894119894 is measured as 1213ms and the variation is assumed to plusmn 5 of the normal val-

ue (1152~1273ms) The results indicate that the proposed damage characterization approach

works well in real application

TABLE IV Damage characterization for the Teflon inserted damage specimen

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

Exact 635 25

Method 1 123 60~75 73~338 298~1220

Method 2 123 619~662 229~310 934~1188

Method 3 123 631 284 1161

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1213 ms

- 25 -

53 Impact-induced damage

11 115 12 125 13-4

-2

0

2

4

Time (ms)

Vol

tage

(mV

)

IntactImpact2Impact3

Reflected wave from damage

(a) Pitch-catch signal

11 115 12 125 13 135

-05

0

05

1

Time (ms)

Volta

ge (m

V)

IntactImpact2Impact3

Boundary reflection

Reflected wave from damage

(b) Pulse-echo signal

Fig 13 The signal obtained from the test specimen subjected to impact The impact was in-troduced at the same location of the specimen and the signal was obtained after each impacts As the number of impact is increased the transmitted A0 Lamb wave mode is delayed and attenuated

The obtained signals from impact-induced damage specimen are presented in Fig 13 It is

observed that as the number of impact is increased the transmitted A0 Lamb wave mode is

more delayed because the wave velocity in the damage area is reduced Fig 14 shows the

false-alarm test result whether the proposed damage detection algorithm works even in pris-

tine condition of the test specimen It is shown that no false alarm is occurred because no ad-

ditional reflected wave is captured

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Transmitted Wave

(a) Transmitted wave

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Reflected Wave

No reflected wave

(b) Reflected wave

Fig 14 False alarm test for the proposed damage detection algorithm in pristine condition of the specimen Reflected wave from damage is not captured false alarm is not occurred

- 26 -

TABLE V Damage characterization for the impact-induced damage specimen

of Impacts m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

2

(1470 J)

Method 1 080 65~90 75~246 470~1157

Method 2 080 801~885 114~245 711~1134

Method 3 080 843 179 1122

3

(2205 J)

Method 1 148 50~100 06~380 21~1134

Method 2 148 758~838 249~368 843~1125

Method 3 148 798 308 1044

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1213 ms

Table V presents the characterization results of the impact-induced damage case Here the

measured 119907119907119894119894 and the assumed variation are same as the artificial damage case It is observed

that as the number of impact is increased 119889119889 is increased while 119907119907119889119889 is reduced Thus it is

validated that the proposed damage characterization approach estimates the increment of the

damage well

54 Temperature effect test

541 Teflon inserted damage

11 115 12 125 13-10

-5

0

5

10

Time (ms)

Vol

tage

(mV

)

0 oC20 oC50 oC

(a) Pitch-catch signal

11 115 12 125 13 135

-1

0

1

2

3

Time(ms)

Volta

ge (m

V)

0 oC20 oC50 oC

(b) Pulse-echo signal

Fig 15 Signals obtained from the Teflon inserted damage specimen under temperature varia-tion Time delay and amplitude variation are occurred

Fig 15 shows the obtained signals from the Teflon inserted damage specimen subjected to

- 27 -

temperature variation Time delay and amplitude variations are occurred there is no addition-

al damage is introduced The analyzed results of the obtained signals are reported in Table VI

The measured 119907119907119894119894 is 1247 ms at 0 degC 1213 ms at 20 degC and 1168 ms at 50 degC respectively

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms (1152~1273 ms)

The estimated results are also changed under temperature variations the fluctuation of esti-

mated range and exact value is small

TABLE VI Temperature effect test for the Teflon inserted damage specimen

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

0 degC

Exact 635 25

Method 1 113 60~70 79~268 352~1291

Method 2 113 560~619 233~344 1034~1172

Method 3 113 606 257 1141

20 degC

Exact 635 25

Method 1 123 60~75 73~338 298~1220

Method 2 123 619~662 229~310 934~1188

Method 3 123 631 284 1161

50 degC

Exact 635 25

Method 1 137 60~75 72~328 263~1228

Method 2 137 622~687 183~299 666~1086

Method 3 137 642 264 960

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1247 ms at 0 degC 1213 ms at 20 degC and 1168 ms at 50 degC

- 28 -

542 Impact-induced damage

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Vol

tage

(mV

)

Transmitted WaveReflected Wave

No reflected wave

(a) 0 degC

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Transmitted WaveReflected Wave

No reflected wave

(b) 50 degC

Fig 16 False alarm test under temperature variations It is shown that proposed damage char-acterization algorithm is robust under temperature variations

The false alarm test in pristine condition of the impact-induced damage specimen under

temperature variation was investigated (Fig 16) It is validated that the proposed damage di-

agnosis algorithm is robust under temperature variations Fig 17 shows the obtained signals

from the impact-induced damage (impacted 3 times) specimen under temperature variation

The analyzed results of the obtained signals are reported in Table VII The measured 119907119907119894119894 in

each temperature and the assumed variation are same as the artificial damage case

11 115 12 125 13-5

0

5

Time (ms)

Volta

ge (m

V)

0 oC20 oC50 oC

(a) Pitch-catch signal

11 115 12 125 13 135

-1

0

1

2

3

Time(ms)

Vol

tage

(mV

)

0 oC20 oC50 oC

(b) Pulse-echo signal

Fig 17 Signals obtained from impact-induced damage specimen (Impact 3) under tempera-ture variation

- 29 -

TABLE VII Temperature effect test for the impact induced damage specimen (Impact 3 case)

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

0 degC

Method 1 135 75~90 72~310 267~1148

Method 2 135 705~779 264~381 978~1126

Method 3 135 763 289 1071

20 degC

Method 1 148 50~100 06~380 21~1134

Method 2 148 758~838 249~368 843~1125

Method 3 148 798 308 1044

50 degC

Method 1 153 75~100 53~415 174~1227

Method 2 153 804~889 214~336 702~1103

Method 3 153 815 320 1050

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1247 ms at 0 degC 1213 ms at 20 degC and 1168 ms at 50 degC

- 30 -

CHAPTER 6 CONCLUDING REMARKS

61 Executive summary

In this study the reference-free damage detection localization and quantification tech-

nique in a multi layered composite plate is proposed Additional reflected A0 Lamb wave

mode from the damage due to multiple reflections in damage area is used as a feature for

characterizing the damage First damage is characterized in terms of a damage index (m-

value) defined as the ratio of the physical length of the damage to the wavelength of the A0

Lamb wave mode Then the range of possible damage location length and severity are esti-

mated with constraints This procedure can be accomplished without prior baseline data To

realize the proposed technique the matching pursuit technique is modified for reduction of

the error by considering the dispersive characteristic of Lamb wave The matching pursuit

method estimates the arrival time of each wave packets accurately even two wave packets are

overlapped The numerical simulations and experimental tests including Teflon tape inserted

and impact-induced damage are investigated for the feasibility of the proposed technique

Moreover the effectiveness of the proposed technique under changing temperature is exper-

imentally verified

- 31 -

62 Suggestion for further study

The limitations of this study and the possible improvement and further studies are listed as

follow

1 Measurement of impact-induced damage length using conventional NDT technique

Additional experimental verification for impact-induced damage characterization is

needed by measuring the real length of damage using conventional NDT technique

such as ultrasonic scan or infrared camera

2 Limitation of both side collocated PZT installation In this study excitation PZTs are

installed both side and collocated for decomposing A0 Lamb wave mode However it

can produce decomposition error due to imperfection of PZT location and their installa-

tion Moreover in many cases assess to both surface of structures is not allowed such

as aircraft applications To improve this limitation a reference-free damage characteri-

zation using single side PZTs will be considered

3 Complex structure application In this study damage is characterized in simple plate-

like structure When the monitored structure becomes complex the response signals al-

so becomes complex due to its geometry and boundary Thus the investigations for

complex structure application such as stiffener spar and curvature are needed

- 32 -

REFERENCES

[1] Cawley P and Alleyne D (1996) ldquoThe use of Lamb waves for the long range inspection of large structuresrdquo Ultrasonics 34 pp287-290

[2] Ghosh T Kundu T and Karpur P (1998) ldquoEfficient use of Lamb waves for detecting defects in large platesrdquo Ultrasonics 36 pp791-801

[3] Guirguitui V (2003) ldquoLamb wave generation with piezoelectric wafer active sensors for structural health monitoringrdquo SPIE Conference on Smart Structures and Materials San Diego CA 5391 pp111-422

[4] Ragvahan A and Cesnik CE S (2004) ldquoModeling of piezoelectric-based Lamb wave generation and sensing for structural health monitoringrdquo SPIE Conference on Smart Struc-tures and Materials San Diego CA 5391 pp419-430

[5] Rose JL Pilarski A and Dirit JJ (1993) ldquoAn approach to guided wave mode selection for inspection of laminated platerdquo Journal of Reinforced Plastics and Composites 12(5) pp536-544

[6] Guo N and Cawley P (1993) ldquoThe interaction of Lamb waves with delaminations in composite laminatesrdquo Journal of Acoustical Society of America 94(4) pp 2240-2246

[7] Valdez SHD and Soutis C (2002) ldquoReal-time nondestructive evaluation of fiber com-posite laminates using low-frequency Lamb wavesrdquo Journal of Acoustical Society of Amer-ica 111(5) pp 2026-2033

[8] Yang M and Qiao P (2005) ldquoModeling and experimental detection of damage in virous materials using the pulse-echo method and piezoelectric sensoractuatorsrdquo Smart Materials and Structures 14 pp1083-1100

[9] Lestari W and Qiao P (2005) ldquoApplication of wave propagation analysis for damage identification in composite laminated beamsrdquo Journal of Composite Materials 39(22) pp1967-1984

[10] Kessler S S Spearing S M and Soutis C (2002) ldquoDamage detection in composite ma-terials using Lamb wave methodsrdquo Smart Materials and Structures 11 pp269-278

[11] Sohn H Park G Wail J R Limback N P and Farrar C R (2004) ldquoWavelet-based active sensing for delamination detection in composite structuresrdquo Smart Materials and Structures 13 pp 153-160

[12] Petculescu G Krishnaswamy S and Achenback J D (2008) ldquoGroup delay measure-ments using modally selective Lamb wave transducers for detection and sizing of delamina-tions in compositesrdquo Smart Materials and Structures 17

[13] Sohn H Park H W Law K H and Farrar C R (2007) ldquoDamage detection in compo-

- 33 -

site plates by using an enhanced time reversal methodrdquo Journal of Aerospace Engineering (ASCE) pp 141-151

[14] Sohn H (2007) ldquoEffects of environmental and operational variability on structural health monitoringrdquo Philosophical Transactions of the Royal Society Ardquo 365 pp539-560

[15] Rytter A (1993) ldquoVibration based inspection of civil engineering structuresrdquo PhD The-sis Department of Building Technology and Structural Engineering University of Aalborg Aalborg Denmark

[16] Hayashi T and Kawashima K (2002) ldquoMultiple reflections of Lamb waves at a delami-nationrdquo Ultrasonics 40 pp193-197

[17] Wang C H and Rose L R F (2003) ldquoWave reflection and transmission in beams con-taining delamination and inhomogeneityrdquo Journal of Sound and Vibration 264 pp 851-872

[18] Ramadas C Balasubramaniam K Joshi M and Krishnamurthy C V (2009) ldquoInterac-tion of the primary anti-symmetric Lamb mode (Ao) with symmetric delaminations numer-ical and experimental studiesrdquo Smart Materials and Structures 18

[19] Ramadas C Balasubramaniam K Joshi M and Krishnamurthy C V (2010) ldquoInterac-tion of guided Lamb waves with an asymmetrically located delamination in a laminated composite platerdquo Smart Materials and Structures 19

[20] Wilcox P Lowe M and Cawely P (2001) ldquoThe effect of dispersion on long-range in-spection using ultrasonic guided wavesrdquo NDTampE International 34 pp1-9

[21] Mallat S G and Zhang Z (1993) ldquoMatching pursuits with time-frequency dictionariesrdquo IEEE Tansactions on Signal Processing 41(12) pp3397-3415

[22] Zhang G Zhang S and Wang Y (2000) ldquoApplication of adaptive time-frequency de-composition in ultrasonic NDE of highly scattering materialsrdquo Ultrasonics 38 pp 961-964

[23] Hong J C Sun K H and Kim Y Y (2005) ldquoThe matching pursuit approach based on the modulated Gaussian pulse for efficient guided-wave damage inspectionrdquo Smart Mate-rials and Structures 14 pp548-560

[24] Herakovich C T (1997) Mechanics of Fibrous Composites Wiley 480 pages

[25] Reddy J N (1985) ldquoA review of the literature on finite-element modeling of laminated composite platesrdquo The Shock and Vibration Digest 17(4) pp 3-8

[26] Reddy J N (2003) Mechanics of Laminated Composite Plates and Shells Theory and Analysis CRC Press 856 pages

[27] Palacz M Krawczuk M and Ostachowicz W (2005) ldquoThe spectral finite element mod-el for analysis of flexure-shear coupled wave propagation Part 1 laminated multilayer

- 34 -

composite beamrdquo Composite Structures 68 pp37-44

[28] Zhang Y C and Harding J (1990) ldquoA numerical micromechanics analysis of the me-chanical properties of a plain weave compositerdquo Computer amp Structures 36(5) pp 839-844

[29] Nemat-Nasser S and Hori M (1999) Micromechanics Overall Properties of Heteroge-neous Materials Elsevier 810 pages

[30] Scida D Aboura Z Benzeggagh M L and Bocherens E (1999) ldquoA micromechanics model for 3D elasticity and failure of woven-fibre composite materialrdquo Composites Science and Technology 59 pp 505-517

[31] Thomas Y and Wu X H (1997) ldquoA multiscale finite element method for elliptic prob-lems in composite materials and porous mediardquo Journal of Computational Physics 134 pp 169-189

[32] Lee K H Moorthy S and Ghosh S (1999) ldquoMultiple scale computational model for damage in composite materialsrdquo Computer Methods in Applied Mechanics and Engineer-ing 172 pp 175-201

[33] Namilae S and Chandra N (2005) ldquoMultiscale model to study the effect of interfaces in carbon nanotube-based compositesrdquo Journal of Engineering Materials and Technology (ASME) 127 pp222-232

[34] Adams D E (2007) Health monitoring of Structural Materials and Components Wiley 460 pages

[35] Gopalakrishnan S (2009) ldquoModeling Aspects in Finite Elementrdquo Encyclopedia of Struc-tural Health Monitoring 2 pp791-809

[36] Palacz M Krawczuk M and Ostachowicz W (2005) ldquoThe spectral finite element mod-el for analysis of flexure-shear coupled wave propagation Part 2 Deminated multilayer composite beamrdquo Composite Structures 68 pp45-51

[37] Baik J M and Thopson R B (1984) ldquoUltrasonic scattering from imperfect interfaces A quasi-static modelrdquo Journal of Nondestructive Evaluation 4(3) pp177-196

[38] Luo H and Hanagud S (2000) ldquoDynamics of delamincated beamsrdquo International Jour-nal of Solids and Structures 37 pp1501-1519

[39] Kim S B and Sohn H (2007) ldquoInstantaneous reference-free crack detection based on polarization characteristics of piezoelectric materialsrdquo Smart Materials and Structures 16 pp2375-2387

[40] An Y K and Sohn H (2010) ldquoInstantaneous crack detection under varying temperature and static loading conditionsrdquo The Special Issue of Structural Control and Health Monitor-ing 17(7) pp 730-741

- 35 -

[41] Lee S J and Sohn H (2006) ldquoActive self-sensing scheme development for structural health monitoringrdquo Smart Materials and Structures 15(6) pp 1734-1746

- 36 -

요 약 문

복합재료 손상의 무기저 진단 및 정량화

복합재료는 재료의 경량성 내구성 고강도의 특성으로 인하여 항공기에서부터

교량에까지 널리 사용되고 있다 하지만 외부의 충격으로 인한 복합재료 내부의

손상은 재료의 강도를 감소시켜 구조물의 안정성을 저하시키는 요인이 된다 하지만

복합재료의 손상은 재료 내부에서 발생하여 육안으로는 식별하기도 힘들며 표면에서

보이는 것만으로 내부의 손상 정도를 판단하기는 불가능하다는 단점을 가지고 있다

본 연구에서는 손상 내부에서 반사로 인해 발생하는 파를 이용하여 기저자료를

사용하지 않고 특정 시점에서 얻은 자료만으로 복합재료의 손상을 진단하고 정량화

할 수 있는 기법을 제안한다 제안된 기법의 적용성을 검증하기 위하여 유한요소법을

이용한 수치해석을 수행하였고 Teflon tape 이 삽입된 시편과 실제 충격하중으로

인한 손상을 입은 시편을 이용한 실험을 진행하였다 수치해석 및 실험 결과 제안된

기법이 복합재료의 손상을 성공적으로 진단 및 정량화할 수 있음을 확인하였다 또한

다양한 외부 온도 환경에서 제안된 기법의 환경영향 실험을 수행하였고 다양한

환경에서 손상의 진단 및 정량화가 성공적으로 이루어짐을 검증하였다

핵심어 무기저 손상진단 및 정량화 압전센서 복합재료 유한요소법 Matching pursuit

method

- 37 -

감 사 의 글

기대와 두려움을 안고 시작한 석사과정이 어느덧 2 년 이라는 세월 속에서 추억으로 자리잡아

가고 있습니다 학부생 기간을 포함한 지난 3 년 동안 SHM 이라는 새로운 분야를 접하고 연구를

수행 할 수 있도록 도와주신 고마운 분들이 있어 그 감사를 표하고자 합니다

우선 저의 많은 부족한 점을 일깨워 주시고 연구에 대한 뜻을 더욱 확고히 세울 수 있게 해

주신 지도교수님이신 손 훈 교수님께 깊은 감사의 뜻을 전합니다 짧은 기간이었지만 교수님의

도움으로 많은 것을 배우고 느끼고 깨달을 수 있었습니다 또한 본 연구를 수행할 수 있도록 지

원해주신 국토해양부 (U-City 석 박사과정 지원사업)에 깊은 감사 드립니다

때로 힘들고 지칠 때 항상 저를 믿고 응원해 주시는 가족이 있었기에 지금의 글을 쓸 수 있었

습니다 또한 지금 연구실에서 함께 생활하고 있는 윤규형 민구형 철민이형 현석이형 진열이형

호민이형 병진이 선혜누나와 성대에 계신 창길이형 미국에 계신 의재형과 Debaditiya 에게도 감

사의 말씀을 전합니다 그리고 모두 나열할 수는 없지만 카이스트 생활의 활력소가 되어준 카이

스트 건설 및 환경공학과 선후배들 직원 분들과 SOS 멤버들에게도 감사를 표합니다

앞서 거론한 많은 이들의 도움이 없었다면 저 또한 지금 이 자리에 있을 수 없었을 것입니다

많이 부족하기에 더 노력하고 배우고자 했던 초심을 다시 한번 생각하며 더 큰 결실을 맺기 위해

앞으로 나아가겠습니다 도움을 주신 모든 분들께 이 감사의 글을 바칩니다

2011년 6월 카이스트에서 임형진 올림

- 38 -

CURRICULUM VITAE

Hyung Jin Lim

Research Assistant

The Civil and Environmental Engineering Department

Korea Advanced Institute of Science and Technology (KAIST)

291 Daehak-Ro Guseong-Dong Yuseong-Gu

Daejeon Republic of Korea 305-701

Phone (82)+42-350-3665 Fax (82)+42-350-3610

Email limhj87gmailcom

RESEARCH INTERESTS

Damage Detection Localization and Quantification on Composite and Metallic Structures Pattern Recognition Finite Element Modeling (FEM)

EDUCATION

2009 ndash 2011 MS Civil and Environmental Eng Korea Advanced Institute of Science and Technology (KAIST) Korea Rep

2006 ndash 2009 BS Civil and Environmental Eng Korea Advanced Institute of Science and Technology (KAIST) Korea Rep

JOURNAL PUBLICATIONS The corresponding author is underlined 1 Hyung Jin Lim Hoon Sohn ldquoReference-free damage detection localization and quantification

using in compositerdquo in preparation for Journal of the Acoustical Society of America 2011

2 Chul Min Yeum Hoon Sohn Jeong Bum Ihn Hyung Jin Lim ldquoInstantaneous delamination detection in a composite using a dual piezoelectric transducer networkrdquo Submitted to Compo-sites Science and Technology 2011

3 Yun-Kyu An Hyung Jin Lim Min Koo Kim Jin Yeol Yang Hoon Sohn Chang Guen Lee ldquoInvestigation of applicability of local reference-free damage detection techniques to in-situ

- 39 -

bridgesrdquo Submitted to ASCE Journal of Structural Engineering 2011

4 Hyung Jin Lim Min Koo Kim Hoon Sohn Chan Yik Park ldquoImpedance based damage detec-tion under varying temperature and loading conditionsrdquo Submitted to NDTampE International 2011

5 Hyung Jin Lim Hoon Sohn ldquoQuantitative estimation of transmitted and reflected lamb waves at discontinuityrdquo Journal of the Korean Society for Nondestructive Testing Vol30 No 4 pp1-8 2010

CONFERENCE PROCEEDINGS

The corresponding author is underlined 1 Hyung Jin Lim and Hoon Sohn ldquoReference-free damage detection localization and quantifi-

cation in compositesrdquo ASCE Engineering Mechanics Institute Conference Boston June 2-4 2011

2 Hyung Jin Lim and Hoon Sohn ldquoReference-free delamination quantification within multi-layer composite materialsrdquo KSNT Spring Conference May 26-27 Gyeong-Ju Korea 2011

3 Min Koo Kim Hyun Jin Lim Hoon Sohn Chan Yik Park ldquoImpedance-based bolt loosening detection under varying temperature and loadingrdquo Asian Pacific Workshop on Structural Health Monitoring Tokyo Japan November 30-December 2 2010

4 Hyung Jin Lim Chul Min Yeum and Hoon Sohn ldquoModeling of Impact-induced Delamination in a Multilayer Composite Platerdquo The 23rd KKCNN Symposium on Civil Engineering Taipei (Taiwan) November 13-15 2010

5 Yun-Kyu An Hyung Jin Lim and Hoon Sohn ldquoTemperature independent Delamination Detec-tion using data normalizationrdquo The 3rd International Conference on Multi-Functional Materi-als and Structures Chonbuk National Univ Jeonju Korea September 14-18 2010

6 Hoon Sohn Yun-Kyu An Chang Gil Lee Min Koo Kim Hyung Jin Lim ldquoApplication of online NDT techniques to decommissioned bridge testingrdquo the 5th European Workshop on Structural Health Monitoring Sorrento Italy June 29-July 02 2010

7 Min Koo Kim Hyung Jin Lim Hoon Sohn Chan Yik Park ldquoImpedance based damage diag-nosis of a complex composite aircraft wing under changing loading and temperature condi-tionsrdquo the 5th European Workshop on Structural Health Monitoring Sorrento Italy June 29-July 02 2010

8 Yun-Kyu An Min Koo Kim Hyung Jin Lim and Hoon Sohn ldquoValidations of reference-free crack detection techniques through a decommissioned bridge testrdquo 2010 COSEIK Annual Con-ference Jeju Korea April 8-9 2010

- 40 -

PARTICIPANT PROJECTS

1 Development of Smart Steel Members for Online Steel Structure Monitoring POSCO Corporation (Funded 41000000 KRW (41000 USD) for 010511 to 123111)

2 Development of Enabling Technologies for Structural Health Monitoring Korean Agency for Defense Development (Funded 240000000 KRW (240000 USD) out of total 3200000000 KRW (3200000 USD) for 060108-053111)

3 Development of On-board SHM Technologies for Composite Air Vehicles The Boeing Company (Funded 400000 USD (400000000 KRW) for 080108 to 073111)

EDUCATIONAL EXPERIENCES

Fall 2009

Teaching Assistant CE207 ldquoElementary Structural Engineering and Laboratoryrdquo KAIST 3 unit undergraduate course

EXPERIMENTAL EXPERIENTCES

2010 Monitoring of Real Aircraft Composite Wing Segment Korea Agency for Defense De-velopment (ADD) Daejeon Korea Real time monitoring of composite air vehicles fund-ed by Korea Agency for Defense Development (ADD)

2009-2011

Field Monitoring of Real Bridge Structures Gimpo (Ramp G bridge) amp Icheon (Sam-seung bridge) Korea US-Korea collaborative research for bridge monitoring testbeds

HONORS amp AWARDS

Scholarships for outstanding students granted KAIST (2007-2009) The three prominent students in the civil and environmental engineering department of KAIST were selected as recipients This scholarship provided 2000$ per year for three years

- 41 -

Page 24: 복합재료 손상의 무기저 진단 및 정량화 - KAISTssslab.kaist.ac.kr/article/pdf/Thesis_2.pdf · 1 Declaration of Ethical Conduct in Research: I, as a graduate student

119907119907119894119894 variation under temperature changes Method 3 gives unique estimated value using meas-

ured 119907119907119894119894 from other path in the same time In section 4 and 5 damage is characterized using

these three methods and the results are compared each another

36 Matching pursuit method considering the dispersion effect of Lamb

wave

Matching pursuit method has been developed for signal compression approximation and

characterization of obtained signal and function [21-23] The key concept of the matching

pursuit method is that any signal (function) can be decomposed into a linear expansion of

signals selected from a pool of basis signals (functions) called lsquodictionaryrsquo

Let 119891119891 be a function to be analyzed then 119891119891 can be approximated a set of functions select-

ed from the dictionary 119863119863 in order to best match This is done by successive approximations

with orthogonal projections Let1198921198921205741205740 isin 119863119863 the function 119891119891 can be decomposed into

119891119891 = lang1198911198911198921198921205741205740rang1198921198921205741205740 + 119877119877119891119891 (8)

where 119877119877119891119891 is the residual function after approximating 119891119891 using 1198921198921205741205740 1198921198921205741205740 is orthogonal to

119877119877119891119891 hence

1198911198912 = lang1198911198911198921198921205741205740rang2

+ 1198771198771198911198912 (9)

To minimize 119877119877119891119891 1198921198921205741205740 is chosen to maximize lang1198911198911198921198921205741205740rang The matching pursuit method is

an iterative algorithm that decomposes the residue 119877119877119891119891 by projecting it on a function of 119863119863

that matches 119877119877119891119891 almost at best as it was done for 119891119891 The residue 119877119877119899119899119891119891 is decomposed into

119877119877119899119899119891119891 = lang119877119877119899119899119891119891119892119892120574120574119899119899rang119892119892120574120574119899119899 + 119877119877119899119899+1119891119891 (10)

- 13 -

the decomposition is carried up to the order 119898119898 119891119891 can be expressed as

119891119891 = lang119877119877119899119899119891119891119892119892120574120574119899119899rang119892119892120574120574119899119899

119898119898minus1

119899119899=0

+ 119877119877119898119898119891119891 (11)

The original function 119891119891 is decomposed into a sum of dictionary elements that are chosen to

best match its residues

In this study the matching pursuit method is utilized to accurately decompose of the over-

lapped Gaussian pulses and estimate the parameters of the pulse which used for damage

characterization The 119894119894119905119905ℎ Gaussian pulse in the dictionary for approximating the obtained

signal is expressed as

119892119892120574120574119894119894 = 119891119891(119860119860119894119894 119906119906119894119894 119904119904119894119894 119889119889119894119894120601120601119894119894) (12)

where 119860119860119894119894119906119906119894119894119904119904119894119894119889119889119894119894 and 120601120601119894119894 represent amplitude arrival time scale (pulse duration) excita-

tion(central) frequency and phase of the Gaussian pulse respectively The parameters of the

pulse are estimated using a nonlinear regression based optimization method (Gauss-Newton

method) based on an initial guess of the parameters Detail procedure for estimating the pa-

rameters of the pulse is expressed in Hong et al [23]

Among the estimated parameters of the pulses pulse duration (scale) represents the disper-

sion effect If a wave-packet is spread out the estimated pulse duration is increased Using

this feature the matching pursuit method is modified to capture waves which have longer

pulse duration than transmitted A0 Lamb wave mode If the estimate pulse duration of a wave

is less than the transmitted wave the modified matching pursuit method regards it as error

- 14 -

CHAPTER 4 NUMERICAL SIMULATION

41 Composite modeling method

Composite materials can be modeled from nano to macro level depends on its engineering

purpose In general four methods are utilized for modeling of composite materials First

composite is modeled as a homogenous material expressed in terms of effective material

properties [24] The effective material properties are estimated using the modulus and the

volume fraction of each constituent theoretically (Rule of mixture) and measured experimen-

tally Second a model is established that each laminate which have own material properties

are totally bonded together In this model the stacking sequence and direction of each lami-

nates are considered additionally [25-27] Third method for composite modeling in micro

level called micromechanics [28-30] This model considers the interaction between fibers and

matrix in laminate Last multiscale modeling which the scope of modeling is extended to

molecular or quantum level of constituents [31-33] This model starts from molecular me-

chanics or molecular dynamics then the scale of model leads to the nano micro and macro

level by incorporating small scale models As the scale of models is smaller it provides more

logical way than other large scale of models while the computational cost is more expensive

The scope of this study is focused on the dynamic behavior of composite structure the beam

or plate model was utilized for the numerical simulation

- 15 -

42 Damage modeling method

Generally three damage models have been developed and utilized in the real applications

First the stiffness reduction model called as the equivalent element model was introduced

[17 34-35] This model is simply realized by replacing the region around damage area by a

region with adjusted material properties (Youngrsquos modulus shear modulus density etc)

globally Next the duplicated node model is done by disconnecting the element connectivity

assuming there is no interaction among the delaminated surface This model can be utilized

when the location and the extension of damage is important issue [34-36] Last the contact

mechanics model using the principles of contact mechanics the two interfacial forces (con-

tact resistance and sliding resistance) in the delaminated surface [37-38] In this model the

delaminated surfaces are modeled as nonlinear functions of the interlaminar normal strain

and the interlaminar shear strain This model is computationally expensive while provides

more logical way than other models There is however no universally accepted model for

general impact-induced damage in composites In the case of impact-induced damage it is

impossible to predict or estimate the location and number of delaminated layers Thus from

the uncertainty of impact-induced damage the equivalent element model can be considered

as the effective damage model

- 16 -

43 Numerical simulation description

TABLE I Material property of simulation model and test specimen

Tensile Strength 657 MPa Tensile Modulus 59 GPa Shear Strength 804 MPa Shear Modulus 24 Gpa Specific Gravity 16

PZT A

PZT C

PZT BDamage

3

250 125 125

500 Unit mm

Fig 6 Description of the numerical simulation model

The feasibility of the proposed damage characterization technique is verified by 2-D FEM

simulation using MSC PATRAN and NASTRAN Fig 6 presents the dimension PZTs and

damage location of the simulation model In the model 05mm x 05mm square shell element

was used for modeling of the A0 mode whose flexural wave characteristic The element size

was determined by considering the wavelength of the propagation Lamb wave modes The

material properties of the simulation model and damage cases are presented in Table I and

Table II respectively

PZT A and C act as both actuator and sensor for collecting pulse-echo signal They were

installed on the plate which two of them are both-side collocated for the A0 Lamb wave

- 17 -

mode decomposition [39-40] PZT B was only used as a sensor for collecting pitch-catch sig-

nal For simulating the excitation and sensing of PZTs the pin-force model was adopted that

an input force apples to the end nodes of excitation PZT and averaging element strains corre-

sponding sensing area A Gaussian pulse (tone-burst) 50 kHz central frequency was excited

as an input signal

TABLE II Damage cases for numerical simulation

119949119949120783120783 (mm) 119941119941 (mm) G value reduction ()

Case 1 650 200 90 darr

Case 2 675 150 90 darr

Case 3 675 150 93 darr

Case 4 900 150 90 darr

- 18 -

44 Numerical simulation results

015 02 025 03 035

-2

0

2

Time (ms)

Am

plitu

de

IntactDamage

Reflected wave from damage

(a) Pitch-catch signal

015 02 025 03 035

-2

0

2

Time (ms)

Am

plitu

de

IntactDamage

Reflected waves from damage

(b) Pulse-echo signal Fig 7 Comparison of A0 Lamb wave mode signals from numerical simulation between intact and damage cases Time delay attenuation and multiple reflections due to damage simulated successfully using the damage modeling method described in the previous section

Fig 7 shows that comparison of A0 Lamb wave mode signals between intact and one of

damage cases in the simulation The features of A0 Lamb wave mode passing through dam-

age area described in section 2 are successfully simulated Thus the equivalent element

method is proper for modeling damage in composites In the damage cases the additional

waves reflected from damage is observed in both pitch-catch and pulse echo signal

Table III presents the result of the proposed damage characterization algorithm using the

numerical simulation cases shown in Table II In lsquoMethod 2rsquo 119907119907119894119894 variation is assumed to plusmn

5 of the normal value (1190~1315ms) The normal value of 119907119907119894119894 for lsquoMethod 3rsquo is meas-

ured as 1253 ms In each case the damage is characterized using the proposed technique

Note that as the information of 119907119907119894119894 is specified the estimated value is closer to the exact

value

- 19 -

TABLE III Damage characterization for the numerical simulation models

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

Case 1

Exact 650 200

Method 1 248 50~70 79~414 160~834

Method 2 248 603~667 135~241 271~486

Method 3 248 635 188 379

Case 2

Exact 675 150

Method 1 193 60~75 01~251 37~651

Method 2 193 618~682 114~222 295~575

Method 3 193 65 168 435

Case 3

Exact 675 150

Method 1 233 60~75 66~302 140~649

Method 2 233 668~738 85~196 182~419

Method 3 233 703 140 301

Case 4

Exact 900 150

Method 1 204 75~100 48~349 118~856

Method 2 204 860~951 107~216 264~530

Method 3 204 906 162 397

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1253 ms 1190~1315 ms Measured 119907119907119894119894 1253 ms

- 20 -

CHAPTER 5 EXPERIMENTAL VERIFICATION

51 Overview

In this section experimental study is investigated so that the proposed damage characteri-

zation technique Overall experimental setup and test specimen are shown in Fig 8 and 9

The experimental setup consists of test specimens data acquisition (DAQ) system and self-

sensing circuit Self-sensing circuit was installed for pulse-echo measurement using PZTs

[40-41]

PZT A(PZT C) PZT B

500

500

140

75

250

Teflon Inserted Damage (d =25)

Unit mm

Thickness = 3

PZT A(PZT C) PZT B

500

500

150

90

250

Impact Location

Unit mm

Thickness = 3

(a) Teflon inserted damage specimen

(b) Impact-induced damage specimen

Fig 8 Test specimens for the experimental verification In the Teflon inserted damage Teflon tapes with 25mm diameter were inserted at the center of the specimen The impact-induced damage was introduced to the specimen using the impact test equipment

- 21 -

511 Specimen

Two carbon composite plates with a dimension of 500times500times3mm3 were manufactured by

Nexcomes Inc by stacking 12 layers of CF3327 3K prepreg (woven fabrics type) manufac-

tured by Hankuk Fiber Inc The material properties of the specimen are presented in Table I

One of the specimens Teflon tapes with 25mm diameter was inserted in every 3 layers at the

center of the specimen to mimic the damage The other specimen is utilized for the character-

ization of real impact-induced damage

512 Data acquisition (DAQ) system and PZT

The data acquisition (DAQ) system consists of an arbitrary waveform generator (AWG) a

high speed signal digitizer (DIG) and two multiplexers (MUX) Using the 16-bits AWG

tone-burst input signal with a plusmn10 peak-to-peak voltage were generated The response signal

was measured by the 14-bit DIG For the signal-to-noise ratio reduction both pitch-catch and

pulse-echo signals were measured 150 times and averaged in the time domain 3 identical

circular PZTs of 10mm diameter manufactured by APC International Ltd were installed on

the specimen A Gaussian pulse (tone-burst) 50 kHz central frequency was excited as an in-

put signal The excitation frequency was chosen in order to avoid overlapping of the S0 and

A0 Lamb wave modes and generating of the higher modes (S1 A1hellip)

- 22 -

PZT Actuator

PZT Sensor

AWG

DIG

DAQ System Specimen

SensingActuatingMUX

1

23

4

1 Sending signal to MUX

4 Sending signal to DIG

Self-SensingCircuit

18 Unit mm

10

(a)

(b)

Fig 9 Overall experimental setup (a) Data acquisition (DAQ) system and (b) Installed circu-lar PZT transducer on the specimen with its dimension

513 Impact test

The impact-induced damage was introduced to the specimen using the impact test equip-

ment presented in Fig 10 A 5kg mass with 10mm tip was dropped from 015m height onto

the same location of the specimen The signals were obtained at each number of impacts

Since the reflected wave is not detected at the first impact case the data obtained from the

second and third impact were utilized for the analysis

10

Unit mm (a)

(b)

Fig 10 (a) Impact test equipment for introducing impact-induced damage and (b) Tip of the impact tester with 10mm diameter

- 23 -

514 Temperature effect test

In reality structures are subjected to changing environmental conditions such as tempera-

ture that affect measured signal and cause false-alarm Thus the robustness of the proposed

damage characterization technique under temperature variation in investigated (Fig 11) The

specimens were placed inside a temperature chamber and a thermocouple was installed on

the specimen to measure temperature of surface of the specimen The signal was obtained

under three temperature cases (0 20 and 50degC) and the humidity was kept at 30

Temperature Chamber

Test Specimen

Fig 11 Test specimen in temperature chamber

- 24 -

52 Teflon inserted damage

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Reflected wave from damage

(a)

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Transmitted WaveReflected Wave

Detected reflected wave

(b)

Fig 12 A0 Lamb wave mode obtained using pitch-catch method from the Teflon inserted damage specimen (a) The reflected wave from damage is shown (b) The Teflon inserted damage is detected by capturing the reflected wave

Fig 12 shows the obtained pitch-catch signal from the Teflon inserted damage specimen

and damage diagnosis result The Teflon inserted damage is detected successfully by captur-

ing the reflected wave Table IV presents the result of the Teflon inserted damage characteri-

zation 119907119907119894119894 is measured as 1213ms and the variation is assumed to plusmn 5 of the normal val-

ue (1152~1273ms) The results indicate that the proposed damage characterization approach

works well in real application

TABLE IV Damage characterization for the Teflon inserted damage specimen

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

Exact 635 25

Method 1 123 60~75 73~338 298~1220

Method 2 123 619~662 229~310 934~1188

Method 3 123 631 284 1161

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1213 ms

- 25 -

53 Impact-induced damage

11 115 12 125 13-4

-2

0

2

4

Time (ms)

Vol

tage

(mV

)

IntactImpact2Impact3

Reflected wave from damage

(a) Pitch-catch signal

11 115 12 125 13 135

-05

0

05

1

Time (ms)

Volta

ge (m

V)

IntactImpact2Impact3

Boundary reflection

Reflected wave from damage

(b) Pulse-echo signal

Fig 13 The signal obtained from the test specimen subjected to impact The impact was in-troduced at the same location of the specimen and the signal was obtained after each impacts As the number of impact is increased the transmitted A0 Lamb wave mode is delayed and attenuated

The obtained signals from impact-induced damage specimen are presented in Fig 13 It is

observed that as the number of impact is increased the transmitted A0 Lamb wave mode is

more delayed because the wave velocity in the damage area is reduced Fig 14 shows the

false-alarm test result whether the proposed damage detection algorithm works even in pris-

tine condition of the test specimen It is shown that no false alarm is occurred because no ad-

ditional reflected wave is captured

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Transmitted Wave

(a) Transmitted wave

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Reflected Wave

No reflected wave

(b) Reflected wave

Fig 14 False alarm test for the proposed damage detection algorithm in pristine condition of the specimen Reflected wave from damage is not captured false alarm is not occurred

- 26 -

TABLE V Damage characterization for the impact-induced damage specimen

of Impacts m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

2

(1470 J)

Method 1 080 65~90 75~246 470~1157

Method 2 080 801~885 114~245 711~1134

Method 3 080 843 179 1122

3

(2205 J)

Method 1 148 50~100 06~380 21~1134

Method 2 148 758~838 249~368 843~1125

Method 3 148 798 308 1044

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1213 ms

Table V presents the characterization results of the impact-induced damage case Here the

measured 119907119907119894119894 and the assumed variation are same as the artificial damage case It is observed

that as the number of impact is increased 119889119889 is increased while 119907119907119889119889 is reduced Thus it is

validated that the proposed damage characterization approach estimates the increment of the

damage well

54 Temperature effect test

541 Teflon inserted damage

11 115 12 125 13-10

-5

0

5

10

Time (ms)

Vol

tage

(mV

)

0 oC20 oC50 oC

(a) Pitch-catch signal

11 115 12 125 13 135

-1

0

1

2

3

Time(ms)

Volta

ge (m

V)

0 oC20 oC50 oC

(b) Pulse-echo signal

Fig 15 Signals obtained from the Teflon inserted damage specimen under temperature varia-tion Time delay and amplitude variation are occurred

Fig 15 shows the obtained signals from the Teflon inserted damage specimen subjected to

- 27 -

temperature variation Time delay and amplitude variations are occurred there is no addition-

al damage is introduced The analyzed results of the obtained signals are reported in Table VI

The measured 119907119907119894119894 is 1247 ms at 0 degC 1213 ms at 20 degC and 1168 ms at 50 degC respectively

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms (1152~1273 ms)

The estimated results are also changed under temperature variations the fluctuation of esti-

mated range and exact value is small

TABLE VI Temperature effect test for the Teflon inserted damage specimen

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

0 degC

Exact 635 25

Method 1 113 60~70 79~268 352~1291

Method 2 113 560~619 233~344 1034~1172

Method 3 113 606 257 1141

20 degC

Exact 635 25

Method 1 123 60~75 73~338 298~1220

Method 2 123 619~662 229~310 934~1188

Method 3 123 631 284 1161

50 degC

Exact 635 25

Method 1 137 60~75 72~328 263~1228

Method 2 137 622~687 183~299 666~1086

Method 3 137 642 264 960

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1247 ms at 0 degC 1213 ms at 20 degC and 1168 ms at 50 degC

- 28 -

542 Impact-induced damage

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Vol

tage

(mV

)

Transmitted WaveReflected Wave

No reflected wave

(a) 0 degC

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Transmitted WaveReflected Wave

No reflected wave

(b) 50 degC

Fig 16 False alarm test under temperature variations It is shown that proposed damage char-acterization algorithm is robust under temperature variations

The false alarm test in pristine condition of the impact-induced damage specimen under

temperature variation was investigated (Fig 16) It is validated that the proposed damage di-

agnosis algorithm is robust under temperature variations Fig 17 shows the obtained signals

from the impact-induced damage (impacted 3 times) specimen under temperature variation

The analyzed results of the obtained signals are reported in Table VII The measured 119907119907119894119894 in

each temperature and the assumed variation are same as the artificial damage case

11 115 12 125 13-5

0

5

Time (ms)

Volta

ge (m

V)

0 oC20 oC50 oC

(a) Pitch-catch signal

11 115 12 125 13 135

-1

0

1

2

3

Time(ms)

Vol

tage

(mV

)

0 oC20 oC50 oC

(b) Pulse-echo signal

Fig 17 Signals obtained from impact-induced damage specimen (Impact 3) under tempera-ture variation

- 29 -

TABLE VII Temperature effect test for the impact induced damage specimen (Impact 3 case)

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

0 degC

Method 1 135 75~90 72~310 267~1148

Method 2 135 705~779 264~381 978~1126

Method 3 135 763 289 1071

20 degC

Method 1 148 50~100 06~380 21~1134

Method 2 148 758~838 249~368 843~1125

Method 3 148 798 308 1044

50 degC

Method 1 153 75~100 53~415 174~1227

Method 2 153 804~889 214~336 702~1103

Method 3 153 815 320 1050

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1247 ms at 0 degC 1213 ms at 20 degC and 1168 ms at 50 degC

- 30 -

CHAPTER 6 CONCLUDING REMARKS

61 Executive summary

In this study the reference-free damage detection localization and quantification tech-

nique in a multi layered composite plate is proposed Additional reflected A0 Lamb wave

mode from the damage due to multiple reflections in damage area is used as a feature for

characterizing the damage First damage is characterized in terms of a damage index (m-

value) defined as the ratio of the physical length of the damage to the wavelength of the A0

Lamb wave mode Then the range of possible damage location length and severity are esti-

mated with constraints This procedure can be accomplished without prior baseline data To

realize the proposed technique the matching pursuit technique is modified for reduction of

the error by considering the dispersive characteristic of Lamb wave The matching pursuit

method estimates the arrival time of each wave packets accurately even two wave packets are

overlapped The numerical simulations and experimental tests including Teflon tape inserted

and impact-induced damage are investigated for the feasibility of the proposed technique

Moreover the effectiveness of the proposed technique under changing temperature is exper-

imentally verified

- 31 -

62 Suggestion for further study

The limitations of this study and the possible improvement and further studies are listed as

follow

1 Measurement of impact-induced damage length using conventional NDT technique

Additional experimental verification for impact-induced damage characterization is

needed by measuring the real length of damage using conventional NDT technique

such as ultrasonic scan or infrared camera

2 Limitation of both side collocated PZT installation In this study excitation PZTs are

installed both side and collocated for decomposing A0 Lamb wave mode However it

can produce decomposition error due to imperfection of PZT location and their installa-

tion Moreover in many cases assess to both surface of structures is not allowed such

as aircraft applications To improve this limitation a reference-free damage characteri-

zation using single side PZTs will be considered

3 Complex structure application In this study damage is characterized in simple plate-

like structure When the monitored structure becomes complex the response signals al-

so becomes complex due to its geometry and boundary Thus the investigations for

complex structure application such as stiffener spar and curvature are needed

- 32 -

REFERENCES

[1] Cawley P and Alleyne D (1996) ldquoThe use of Lamb waves for the long range inspection of large structuresrdquo Ultrasonics 34 pp287-290

[2] Ghosh T Kundu T and Karpur P (1998) ldquoEfficient use of Lamb waves for detecting defects in large platesrdquo Ultrasonics 36 pp791-801

[3] Guirguitui V (2003) ldquoLamb wave generation with piezoelectric wafer active sensors for structural health monitoringrdquo SPIE Conference on Smart Structures and Materials San Diego CA 5391 pp111-422

[4] Ragvahan A and Cesnik CE S (2004) ldquoModeling of piezoelectric-based Lamb wave generation and sensing for structural health monitoringrdquo SPIE Conference on Smart Struc-tures and Materials San Diego CA 5391 pp419-430

[5] Rose JL Pilarski A and Dirit JJ (1993) ldquoAn approach to guided wave mode selection for inspection of laminated platerdquo Journal of Reinforced Plastics and Composites 12(5) pp536-544

[6] Guo N and Cawley P (1993) ldquoThe interaction of Lamb waves with delaminations in composite laminatesrdquo Journal of Acoustical Society of America 94(4) pp 2240-2246

[7] Valdez SHD and Soutis C (2002) ldquoReal-time nondestructive evaluation of fiber com-posite laminates using low-frequency Lamb wavesrdquo Journal of Acoustical Society of Amer-ica 111(5) pp 2026-2033

[8] Yang M and Qiao P (2005) ldquoModeling and experimental detection of damage in virous materials using the pulse-echo method and piezoelectric sensoractuatorsrdquo Smart Materials and Structures 14 pp1083-1100

[9] Lestari W and Qiao P (2005) ldquoApplication of wave propagation analysis for damage identification in composite laminated beamsrdquo Journal of Composite Materials 39(22) pp1967-1984

[10] Kessler S S Spearing S M and Soutis C (2002) ldquoDamage detection in composite ma-terials using Lamb wave methodsrdquo Smart Materials and Structures 11 pp269-278

[11] Sohn H Park G Wail J R Limback N P and Farrar C R (2004) ldquoWavelet-based active sensing for delamination detection in composite structuresrdquo Smart Materials and Structures 13 pp 153-160

[12] Petculescu G Krishnaswamy S and Achenback J D (2008) ldquoGroup delay measure-ments using modally selective Lamb wave transducers for detection and sizing of delamina-tions in compositesrdquo Smart Materials and Structures 17

[13] Sohn H Park H W Law K H and Farrar C R (2007) ldquoDamage detection in compo-

- 33 -

site plates by using an enhanced time reversal methodrdquo Journal of Aerospace Engineering (ASCE) pp 141-151

[14] Sohn H (2007) ldquoEffects of environmental and operational variability on structural health monitoringrdquo Philosophical Transactions of the Royal Society Ardquo 365 pp539-560

[15] Rytter A (1993) ldquoVibration based inspection of civil engineering structuresrdquo PhD The-sis Department of Building Technology and Structural Engineering University of Aalborg Aalborg Denmark

[16] Hayashi T and Kawashima K (2002) ldquoMultiple reflections of Lamb waves at a delami-nationrdquo Ultrasonics 40 pp193-197

[17] Wang C H and Rose L R F (2003) ldquoWave reflection and transmission in beams con-taining delamination and inhomogeneityrdquo Journal of Sound and Vibration 264 pp 851-872

[18] Ramadas C Balasubramaniam K Joshi M and Krishnamurthy C V (2009) ldquoInterac-tion of the primary anti-symmetric Lamb mode (Ao) with symmetric delaminations numer-ical and experimental studiesrdquo Smart Materials and Structures 18

[19] Ramadas C Balasubramaniam K Joshi M and Krishnamurthy C V (2010) ldquoInterac-tion of guided Lamb waves with an asymmetrically located delamination in a laminated composite platerdquo Smart Materials and Structures 19

[20] Wilcox P Lowe M and Cawely P (2001) ldquoThe effect of dispersion on long-range in-spection using ultrasonic guided wavesrdquo NDTampE International 34 pp1-9

[21] Mallat S G and Zhang Z (1993) ldquoMatching pursuits with time-frequency dictionariesrdquo IEEE Tansactions on Signal Processing 41(12) pp3397-3415

[22] Zhang G Zhang S and Wang Y (2000) ldquoApplication of adaptive time-frequency de-composition in ultrasonic NDE of highly scattering materialsrdquo Ultrasonics 38 pp 961-964

[23] Hong J C Sun K H and Kim Y Y (2005) ldquoThe matching pursuit approach based on the modulated Gaussian pulse for efficient guided-wave damage inspectionrdquo Smart Mate-rials and Structures 14 pp548-560

[24] Herakovich C T (1997) Mechanics of Fibrous Composites Wiley 480 pages

[25] Reddy J N (1985) ldquoA review of the literature on finite-element modeling of laminated composite platesrdquo The Shock and Vibration Digest 17(4) pp 3-8

[26] Reddy J N (2003) Mechanics of Laminated Composite Plates and Shells Theory and Analysis CRC Press 856 pages

[27] Palacz M Krawczuk M and Ostachowicz W (2005) ldquoThe spectral finite element mod-el for analysis of flexure-shear coupled wave propagation Part 1 laminated multilayer

- 34 -

composite beamrdquo Composite Structures 68 pp37-44

[28] Zhang Y C and Harding J (1990) ldquoA numerical micromechanics analysis of the me-chanical properties of a plain weave compositerdquo Computer amp Structures 36(5) pp 839-844

[29] Nemat-Nasser S and Hori M (1999) Micromechanics Overall Properties of Heteroge-neous Materials Elsevier 810 pages

[30] Scida D Aboura Z Benzeggagh M L and Bocherens E (1999) ldquoA micromechanics model for 3D elasticity and failure of woven-fibre composite materialrdquo Composites Science and Technology 59 pp 505-517

[31] Thomas Y and Wu X H (1997) ldquoA multiscale finite element method for elliptic prob-lems in composite materials and porous mediardquo Journal of Computational Physics 134 pp 169-189

[32] Lee K H Moorthy S and Ghosh S (1999) ldquoMultiple scale computational model for damage in composite materialsrdquo Computer Methods in Applied Mechanics and Engineer-ing 172 pp 175-201

[33] Namilae S and Chandra N (2005) ldquoMultiscale model to study the effect of interfaces in carbon nanotube-based compositesrdquo Journal of Engineering Materials and Technology (ASME) 127 pp222-232

[34] Adams D E (2007) Health monitoring of Structural Materials and Components Wiley 460 pages

[35] Gopalakrishnan S (2009) ldquoModeling Aspects in Finite Elementrdquo Encyclopedia of Struc-tural Health Monitoring 2 pp791-809

[36] Palacz M Krawczuk M and Ostachowicz W (2005) ldquoThe spectral finite element mod-el for analysis of flexure-shear coupled wave propagation Part 2 Deminated multilayer composite beamrdquo Composite Structures 68 pp45-51

[37] Baik J M and Thopson R B (1984) ldquoUltrasonic scattering from imperfect interfaces A quasi-static modelrdquo Journal of Nondestructive Evaluation 4(3) pp177-196

[38] Luo H and Hanagud S (2000) ldquoDynamics of delamincated beamsrdquo International Jour-nal of Solids and Structures 37 pp1501-1519

[39] Kim S B and Sohn H (2007) ldquoInstantaneous reference-free crack detection based on polarization characteristics of piezoelectric materialsrdquo Smart Materials and Structures 16 pp2375-2387

[40] An Y K and Sohn H (2010) ldquoInstantaneous crack detection under varying temperature and static loading conditionsrdquo The Special Issue of Structural Control and Health Monitor-ing 17(7) pp 730-741

- 35 -

[41] Lee S J and Sohn H (2006) ldquoActive self-sensing scheme development for structural health monitoringrdquo Smart Materials and Structures 15(6) pp 1734-1746

- 36 -

요 약 문

복합재료 손상의 무기저 진단 및 정량화

복합재료는 재료의 경량성 내구성 고강도의 특성으로 인하여 항공기에서부터

교량에까지 널리 사용되고 있다 하지만 외부의 충격으로 인한 복합재료 내부의

손상은 재료의 강도를 감소시켜 구조물의 안정성을 저하시키는 요인이 된다 하지만

복합재료의 손상은 재료 내부에서 발생하여 육안으로는 식별하기도 힘들며 표면에서

보이는 것만으로 내부의 손상 정도를 판단하기는 불가능하다는 단점을 가지고 있다

본 연구에서는 손상 내부에서 반사로 인해 발생하는 파를 이용하여 기저자료를

사용하지 않고 특정 시점에서 얻은 자료만으로 복합재료의 손상을 진단하고 정량화

할 수 있는 기법을 제안한다 제안된 기법의 적용성을 검증하기 위하여 유한요소법을

이용한 수치해석을 수행하였고 Teflon tape 이 삽입된 시편과 실제 충격하중으로

인한 손상을 입은 시편을 이용한 실험을 진행하였다 수치해석 및 실험 결과 제안된

기법이 복합재료의 손상을 성공적으로 진단 및 정량화할 수 있음을 확인하였다 또한

다양한 외부 온도 환경에서 제안된 기법의 환경영향 실험을 수행하였고 다양한

환경에서 손상의 진단 및 정량화가 성공적으로 이루어짐을 검증하였다

핵심어 무기저 손상진단 및 정량화 압전센서 복합재료 유한요소법 Matching pursuit

method

- 37 -

감 사 의 글

기대와 두려움을 안고 시작한 석사과정이 어느덧 2 년 이라는 세월 속에서 추억으로 자리잡아

가고 있습니다 학부생 기간을 포함한 지난 3 년 동안 SHM 이라는 새로운 분야를 접하고 연구를

수행 할 수 있도록 도와주신 고마운 분들이 있어 그 감사를 표하고자 합니다

우선 저의 많은 부족한 점을 일깨워 주시고 연구에 대한 뜻을 더욱 확고히 세울 수 있게 해

주신 지도교수님이신 손 훈 교수님께 깊은 감사의 뜻을 전합니다 짧은 기간이었지만 교수님의

도움으로 많은 것을 배우고 느끼고 깨달을 수 있었습니다 또한 본 연구를 수행할 수 있도록 지

원해주신 국토해양부 (U-City 석 박사과정 지원사업)에 깊은 감사 드립니다

때로 힘들고 지칠 때 항상 저를 믿고 응원해 주시는 가족이 있었기에 지금의 글을 쓸 수 있었

습니다 또한 지금 연구실에서 함께 생활하고 있는 윤규형 민구형 철민이형 현석이형 진열이형

호민이형 병진이 선혜누나와 성대에 계신 창길이형 미국에 계신 의재형과 Debaditiya 에게도 감

사의 말씀을 전합니다 그리고 모두 나열할 수는 없지만 카이스트 생활의 활력소가 되어준 카이

스트 건설 및 환경공학과 선후배들 직원 분들과 SOS 멤버들에게도 감사를 표합니다

앞서 거론한 많은 이들의 도움이 없었다면 저 또한 지금 이 자리에 있을 수 없었을 것입니다

많이 부족하기에 더 노력하고 배우고자 했던 초심을 다시 한번 생각하며 더 큰 결실을 맺기 위해

앞으로 나아가겠습니다 도움을 주신 모든 분들께 이 감사의 글을 바칩니다

2011년 6월 카이스트에서 임형진 올림

- 38 -

CURRICULUM VITAE

Hyung Jin Lim

Research Assistant

The Civil and Environmental Engineering Department

Korea Advanced Institute of Science and Technology (KAIST)

291 Daehak-Ro Guseong-Dong Yuseong-Gu

Daejeon Republic of Korea 305-701

Phone (82)+42-350-3665 Fax (82)+42-350-3610

Email limhj87gmailcom

RESEARCH INTERESTS

Damage Detection Localization and Quantification on Composite and Metallic Structures Pattern Recognition Finite Element Modeling (FEM)

EDUCATION

2009 ndash 2011 MS Civil and Environmental Eng Korea Advanced Institute of Science and Technology (KAIST) Korea Rep

2006 ndash 2009 BS Civil and Environmental Eng Korea Advanced Institute of Science and Technology (KAIST) Korea Rep

JOURNAL PUBLICATIONS The corresponding author is underlined 1 Hyung Jin Lim Hoon Sohn ldquoReference-free damage detection localization and quantification

using in compositerdquo in preparation for Journal of the Acoustical Society of America 2011

2 Chul Min Yeum Hoon Sohn Jeong Bum Ihn Hyung Jin Lim ldquoInstantaneous delamination detection in a composite using a dual piezoelectric transducer networkrdquo Submitted to Compo-sites Science and Technology 2011

3 Yun-Kyu An Hyung Jin Lim Min Koo Kim Jin Yeol Yang Hoon Sohn Chang Guen Lee ldquoInvestigation of applicability of local reference-free damage detection techniques to in-situ

- 39 -

bridgesrdquo Submitted to ASCE Journal of Structural Engineering 2011

4 Hyung Jin Lim Min Koo Kim Hoon Sohn Chan Yik Park ldquoImpedance based damage detec-tion under varying temperature and loading conditionsrdquo Submitted to NDTampE International 2011

5 Hyung Jin Lim Hoon Sohn ldquoQuantitative estimation of transmitted and reflected lamb waves at discontinuityrdquo Journal of the Korean Society for Nondestructive Testing Vol30 No 4 pp1-8 2010

CONFERENCE PROCEEDINGS

The corresponding author is underlined 1 Hyung Jin Lim and Hoon Sohn ldquoReference-free damage detection localization and quantifi-

cation in compositesrdquo ASCE Engineering Mechanics Institute Conference Boston June 2-4 2011

2 Hyung Jin Lim and Hoon Sohn ldquoReference-free delamination quantification within multi-layer composite materialsrdquo KSNT Spring Conference May 26-27 Gyeong-Ju Korea 2011

3 Min Koo Kim Hyun Jin Lim Hoon Sohn Chan Yik Park ldquoImpedance-based bolt loosening detection under varying temperature and loadingrdquo Asian Pacific Workshop on Structural Health Monitoring Tokyo Japan November 30-December 2 2010

4 Hyung Jin Lim Chul Min Yeum and Hoon Sohn ldquoModeling of Impact-induced Delamination in a Multilayer Composite Platerdquo The 23rd KKCNN Symposium on Civil Engineering Taipei (Taiwan) November 13-15 2010

5 Yun-Kyu An Hyung Jin Lim and Hoon Sohn ldquoTemperature independent Delamination Detec-tion using data normalizationrdquo The 3rd International Conference on Multi-Functional Materi-als and Structures Chonbuk National Univ Jeonju Korea September 14-18 2010

6 Hoon Sohn Yun-Kyu An Chang Gil Lee Min Koo Kim Hyung Jin Lim ldquoApplication of online NDT techniques to decommissioned bridge testingrdquo the 5th European Workshop on Structural Health Monitoring Sorrento Italy June 29-July 02 2010

7 Min Koo Kim Hyung Jin Lim Hoon Sohn Chan Yik Park ldquoImpedance based damage diag-nosis of a complex composite aircraft wing under changing loading and temperature condi-tionsrdquo the 5th European Workshop on Structural Health Monitoring Sorrento Italy June 29-July 02 2010

8 Yun-Kyu An Min Koo Kim Hyung Jin Lim and Hoon Sohn ldquoValidations of reference-free crack detection techniques through a decommissioned bridge testrdquo 2010 COSEIK Annual Con-ference Jeju Korea April 8-9 2010

- 40 -

PARTICIPANT PROJECTS

1 Development of Smart Steel Members for Online Steel Structure Monitoring POSCO Corporation (Funded 41000000 KRW (41000 USD) for 010511 to 123111)

2 Development of Enabling Technologies for Structural Health Monitoring Korean Agency for Defense Development (Funded 240000000 KRW (240000 USD) out of total 3200000000 KRW (3200000 USD) for 060108-053111)

3 Development of On-board SHM Technologies for Composite Air Vehicles The Boeing Company (Funded 400000 USD (400000000 KRW) for 080108 to 073111)

EDUCATIONAL EXPERIENCES

Fall 2009

Teaching Assistant CE207 ldquoElementary Structural Engineering and Laboratoryrdquo KAIST 3 unit undergraduate course

EXPERIMENTAL EXPERIENTCES

2010 Monitoring of Real Aircraft Composite Wing Segment Korea Agency for Defense De-velopment (ADD) Daejeon Korea Real time monitoring of composite air vehicles fund-ed by Korea Agency for Defense Development (ADD)

2009-2011

Field Monitoring of Real Bridge Structures Gimpo (Ramp G bridge) amp Icheon (Sam-seung bridge) Korea US-Korea collaborative research for bridge monitoring testbeds

HONORS amp AWARDS

Scholarships for outstanding students granted KAIST (2007-2009) The three prominent students in the civil and environmental engineering department of KAIST were selected as recipients This scholarship provided 2000$ per year for three years

- 41 -

Page 25: 복합재료 손상의 무기저 진단 및 정량화 - KAISTssslab.kaist.ac.kr/article/pdf/Thesis_2.pdf · 1 Declaration of Ethical Conduct in Research: I, as a graduate student

the decomposition is carried up to the order 119898119898 119891119891 can be expressed as

119891119891 = lang119877119877119899119899119891119891119892119892120574120574119899119899rang119892119892120574120574119899119899

119898119898minus1

119899119899=0

+ 119877119877119898119898119891119891 (11)

The original function 119891119891 is decomposed into a sum of dictionary elements that are chosen to

best match its residues

In this study the matching pursuit method is utilized to accurately decompose of the over-

lapped Gaussian pulses and estimate the parameters of the pulse which used for damage

characterization The 119894119894119905119905ℎ Gaussian pulse in the dictionary for approximating the obtained

signal is expressed as

119892119892120574120574119894119894 = 119891119891(119860119860119894119894 119906119906119894119894 119904119904119894119894 119889119889119894119894120601120601119894119894) (12)

where 119860119860119894119894119906119906119894119894119904119904119894119894119889119889119894119894 and 120601120601119894119894 represent amplitude arrival time scale (pulse duration) excita-

tion(central) frequency and phase of the Gaussian pulse respectively The parameters of the

pulse are estimated using a nonlinear regression based optimization method (Gauss-Newton

method) based on an initial guess of the parameters Detail procedure for estimating the pa-

rameters of the pulse is expressed in Hong et al [23]

Among the estimated parameters of the pulses pulse duration (scale) represents the disper-

sion effect If a wave-packet is spread out the estimated pulse duration is increased Using

this feature the matching pursuit method is modified to capture waves which have longer

pulse duration than transmitted A0 Lamb wave mode If the estimate pulse duration of a wave

is less than the transmitted wave the modified matching pursuit method regards it as error

- 14 -

CHAPTER 4 NUMERICAL SIMULATION

41 Composite modeling method

Composite materials can be modeled from nano to macro level depends on its engineering

purpose In general four methods are utilized for modeling of composite materials First

composite is modeled as a homogenous material expressed in terms of effective material

properties [24] The effective material properties are estimated using the modulus and the

volume fraction of each constituent theoretically (Rule of mixture) and measured experimen-

tally Second a model is established that each laminate which have own material properties

are totally bonded together In this model the stacking sequence and direction of each lami-

nates are considered additionally [25-27] Third method for composite modeling in micro

level called micromechanics [28-30] This model considers the interaction between fibers and

matrix in laminate Last multiscale modeling which the scope of modeling is extended to

molecular or quantum level of constituents [31-33] This model starts from molecular me-

chanics or molecular dynamics then the scale of model leads to the nano micro and macro

level by incorporating small scale models As the scale of models is smaller it provides more

logical way than other large scale of models while the computational cost is more expensive

The scope of this study is focused on the dynamic behavior of composite structure the beam

or plate model was utilized for the numerical simulation

- 15 -

42 Damage modeling method

Generally three damage models have been developed and utilized in the real applications

First the stiffness reduction model called as the equivalent element model was introduced

[17 34-35] This model is simply realized by replacing the region around damage area by a

region with adjusted material properties (Youngrsquos modulus shear modulus density etc)

globally Next the duplicated node model is done by disconnecting the element connectivity

assuming there is no interaction among the delaminated surface This model can be utilized

when the location and the extension of damage is important issue [34-36] Last the contact

mechanics model using the principles of contact mechanics the two interfacial forces (con-

tact resistance and sliding resistance) in the delaminated surface [37-38] In this model the

delaminated surfaces are modeled as nonlinear functions of the interlaminar normal strain

and the interlaminar shear strain This model is computationally expensive while provides

more logical way than other models There is however no universally accepted model for

general impact-induced damage in composites In the case of impact-induced damage it is

impossible to predict or estimate the location and number of delaminated layers Thus from

the uncertainty of impact-induced damage the equivalent element model can be considered

as the effective damage model

- 16 -

43 Numerical simulation description

TABLE I Material property of simulation model and test specimen

Tensile Strength 657 MPa Tensile Modulus 59 GPa Shear Strength 804 MPa Shear Modulus 24 Gpa Specific Gravity 16

PZT A

PZT C

PZT BDamage

3

250 125 125

500 Unit mm

Fig 6 Description of the numerical simulation model

The feasibility of the proposed damage characterization technique is verified by 2-D FEM

simulation using MSC PATRAN and NASTRAN Fig 6 presents the dimension PZTs and

damage location of the simulation model In the model 05mm x 05mm square shell element

was used for modeling of the A0 mode whose flexural wave characteristic The element size

was determined by considering the wavelength of the propagation Lamb wave modes The

material properties of the simulation model and damage cases are presented in Table I and

Table II respectively

PZT A and C act as both actuator and sensor for collecting pulse-echo signal They were

installed on the plate which two of them are both-side collocated for the A0 Lamb wave

- 17 -

mode decomposition [39-40] PZT B was only used as a sensor for collecting pitch-catch sig-

nal For simulating the excitation and sensing of PZTs the pin-force model was adopted that

an input force apples to the end nodes of excitation PZT and averaging element strains corre-

sponding sensing area A Gaussian pulse (tone-burst) 50 kHz central frequency was excited

as an input signal

TABLE II Damage cases for numerical simulation

119949119949120783120783 (mm) 119941119941 (mm) G value reduction ()

Case 1 650 200 90 darr

Case 2 675 150 90 darr

Case 3 675 150 93 darr

Case 4 900 150 90 darr

- 18 -

44 Numerical simulation results

015 02 025 03 035

-2

0

2

Time (ms)

Am

plitu

de

IntactDamage

Reflected wave from damage

(a) Pitch-catch signal

015 02 025 03 035

-2

0

2

Time (ms)

Am

plitu

de

IntactDamage

Reflected waves from damage

(b) Pulse-echo signal Fig 7 Comparison of A0 Lamb wave mode signals from numerical simulation between intact and damage cases Time delay attenuation and multiple reflections due to damage simulated successfully using the damage modeling method described in the previous section

Fig 7 shows that comparison of A0 Lamb wave mode signals between intact and one of

damage cases in the simulation The features of A0 Lamb wave mode passing through dam-

age area described in section 2 are successfully simulated Thus the equivalent element

method is proper for modeling damage in composites In the damage cases the additional

waves reflected from damage is observed in both pitch-catch and pulse echo signal

Table III presents the result of the proposed damage characterization algorithm using the

numerical simulation cases shown in Table II In lsquoMethod 2rsquo 119907119907119894119894 variation is assumed to plusmn

5 of the normal value (1190~1315ms) The normal value of 119907119907119894119894 for lsquoMethod 3rsquo is meas-

ured as 1253 ms In each case the damage is characterized using the proposed technique

Note that as the information of 119907119907119894119894 is specified the estimated value is closer to the exact

value

- 19 -

TABLE III Damage characterization for the numerical simulation models

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

Case 1

Exact 650 200

Method 1 248 50~70 79~414 160~834

Method 2 248 603~667 135~241 271~486

Method 3 248 635 188 379

Case 2

Exact 675 150

Method 1 193 60~75 01~251 37~651

Method 2 193 618~682 114~222 295~575

Method 3 193 65 168 435

Case 3

Exact 675 150

Method 1 233 60~75 66~302 140~649

Method 2 233 668~738 85~196 182~419

Method 3 233 703 140 301

Case 4

Exact 900 150

Method 1 204 75~100 48~349 118~856

Method 2 204 860~951 107~216 264~530

Method 3 204 906 162 397

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1253 ms 1190~1315 ms Measured 119907119907119894119894 1253 ms

- 20 -

CHAPTER 5 EXPERIMENTAL VERIFICATION

51 Overview

In this section experimental study is investigated so that the proposed damage characteri-

zation technique Overall experimental setup and test specimen are shown in Fig 8 and 9

The experimental setup consists of test specimens data acquisition (DAQ) system and self-

sensing circuit Self-sensing circuit was installed for pulse-echo measurement using PZTs

[40-41]

PZT A(PZT C) PZT B

500

500

140

75

250

Teflon Inserted Damage (d =25)

Unit mm

Thickness = 3

PZT A(PZT C) PZT B

500

500

150

90

250

Impact Location

Unit mm

Thickness = 3

(a) Teflon inserted damage specimen

(b) Impact-induced damage specimen

Fig 8 Test specimens for the experimental verification In the Teflon inserted damage Teflon tapes with 25mm diameter were inserted at the center of the specimen The impact-induced damage was introduced to the specimen using the impact test equipment

- 21 -

511 Specimen

Two carbon composite plates with a dimension of 500times500times3mm3 were manufactured by

Nexcomes Inc by stacking 12 layers of CF3327 3K prepreg (woven fabrics type) manufac-

tured by Hankuk Fiber Inc The material properties of the specimen are presented in Table I

One of the specimens Teflon tapes with 25mm diameter was inserted in every 3 layers at the

center of the specimen to mimic the damage The other specimen is utilized for the character-

ization of real impact-induced damage

512 Data acquisition (DAQ) system and PZT

The data acquisition (DAQ) system consists of an arbitrary waveform generator (AWG) a

high speed signal digitizer (DIG) and two multiplexers (MUX) Using the 16-bits AWG

tone-burst input signal with a plusmn10 peak-to-peak voltage were generated The response signal

was measured by the 14-bit DIG For the signal-to-noise ratio reduction both pitch-catch and

pulse-echo signals were measured 150 times and averaged in the time domain 3 identical

circular PZTs of 10mm diameter manufactured by APC International Ltd were installed on

the specimen A Gaussian pulse (tone-burst) 50 kHz central frequency was excited as an in-

put signal The excitation frequency was chosen in order to avoid overlapping of the S0 and

A0 Lamb wave modes and generating of the higher modes (S1 A1hellip)

- 22 -

PZT Actuator

PZT Sensor

AWG

DIG

DAQ System Specimen

SensingActuatingMUX

1

23

4

1 Sending signal to MUX

4 Sending signal to DIG

Self-SensingCircuit

18 Unit mm

10

(a)

(b)

Fig 9 Overall experimental setup (a) Data acquisition (DAQ) system and (b) Installed circu-lar PZT transducer on the specimen with its dimension

513 Impact test

The impact-induced damage was introduced to the specimen using the impact test equip-

ment presented in Fig 10 A 5kg mass with 10mm tip was dropped from 015m height onto

the same location of the specimen The signals were obtained at each number of impacts

Since the reflected wave is not detected at the first impact case the data obtained from the

second and third impact were utilized for the analysis

10

Unit mm (a)

(b)

Fig 10 (a) Impact test equipment for introducing impact-induced damage and (b) Tip of the impact tester with 10mm diameter

- 23 -

514 Temperature effect test

In reality structures are subjected to changing environmental conditions such as tempera-

ture that affect measured signal and cause false-alarm Thus the robustness of the proposed

damage characterization technique under temperature variation in investigated (Fig 11) The

specimens were placed inside a temperature chamber and a thermocouple was installed on

the specimen to measure temperature of surface of the specimen The signal was obtained

under three temperature cases (0 20 and 50degC) and the humidity was kept at 30

Temperature Chamber

Test Specimen

Fig 11 Test specimen in temperature chamber

- 24 -

52 Teflon inserted damage

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Reflected wave from damage

(a)

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Transmitted WaveReflected Wave

Detected reflected wave

(b)

Fig 12 A0 Lamb wave mode obtained using pitch-catch method from the Teflon inserted damage specimen (a) The reflected wave from damage is shown (b) The Teflon inserted damage is detected by capturing the reflected wave

Fig 12 shows the obtained pitch-catch signal from the Teflon inserted damage specimen

and damage diagnosis result The Teflon inserted damage is detected successfully by captur-

ing the reflected wave Table IV presents the result of the Teflon inserted damage characteri-

zation 119907119907119894119894 is measured as 1213ms and the variation is assumed to plusmn 5 of the normal val-

ue (1152~1273ms) The results indicate that the proposed damage characterization approach

works well in real application

TABLE IV Damage characterization for the Teflon inserted damage specimen

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

Exact 635 25

Method 1 123 60~75 73~338 298~1220

Method 2 123 619~662 229~310 934~1188

Method 3 123 631 284 1161

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1213 ms

- 25 -

53 Impact-induced damage

11 115 12 125 13-4

-2

0

2

4

Time (ms)

Vol

tage

(mV

)

IntactImpact2Impact3

Reflected wave from damage

(a) Pitch-catch signal

11 115 12 125 13 135

-05

0

05

1

Time (ms)

Volta

ge (m

V)

IntactImpact2Impact3

Boundary reflection

Reflected wave from damage

(b) Pulse-echo signal

Fig 13 The signal obtained from the test specimen subjected to impact The impact was in-troduced at the same location of the specimen and the signal was obtained after each impacts As the number of impact is increased the transmitted A0 Lamb wave mode is delayed and attenuated

The obtained signals from impact-induced damage specimen are presented in Fig 13 It is

observed that as the number of impact is increased the transmitted A0 Lamb wave mode is

more delayed because the wave velocity in the damage area is reduced Fig 14 shows the

false-alarm test result whether the proposed damage detection algorithm works even in pris-

tine condition of the test specimen It is shown that no false alarm is occurred because no ad-

ditional reflected wave is captured

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Transmitted Wave

(a) Transmitted wave

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Reflected Wave

No reflected wave

(b) Reflected wave

Fig 14 False alarm test for the proposed damage detection algorithm in pristine condition of the specimen Reflected wave from damage is not captured false alarm is not occurred

- 26 -

TABLE V Damage characterization for the impact-induced damage specimen

of Impacts m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

2

(1470 J)

Method 1 080 65~90 75~246 470~1157

Method 2 080 801~885 114~245 711~1134

Method 3 080 843 179 1122

3

(2205 J)

Method 1 148 50~100 06~380 21~1134

Method 2 148 758~838 249~368 843~1125

Method 3 148 798 308 1044

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1213 ms

Table V presents the characterization results of the impact-induced damage case Here the

measured 119907119907119894119894 and the assumed variation are same as the artificial damage case It is observed

that as the number of impact is increased 119889119889 is increased while 119907119907119889119889 is reduced Thus it is

validated that the proposed damage characterization approach estimates the increment of the

damage well

54 Temperature effect test

541 Teflon inserted damage

11 115 12 125 13-10

-5

0

5

10

Time (ms)

Vol

tage

(mV

)

0 oC20 oC50 oC

(a) Pitch-catch signal

11 115 12 125 13 135

-1

0

1

2

3

Time(ms)

Volta

ge (m

V)

0 oC20 oC50 oC

(b) Pulse-echo signal

Fig 15 Signals obtained from the Teflon inserted damage specimen under temperature varia-tion Time delay and amplitude variation are occurred

Fig 15 shows the obtained signals from the Teflon inserted damage specimen subjected to

- 27 -

temperature variation Time delay and amplitude variations are occurred there is no addition-

al damage is introduced The analyzed results of the obtained signals are reported in Table VI

The measured 119907119907119894119894 is 1247 ms at 0 degC 1213 ms at 20 degC and 1168 ms at 50 degC respectively

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms (1152~1273 ms)

The estimated results are also changed under temperature variations the fluctuation of esti-

mated range and exact value is small

TABLE VI Temperature effect test for the Teflon inserted damage specimen

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

0 degC

Exact 635 25

Method 1 113 60~70 79~268 352~1291

Method 2 113 560~619 233~344 1034~1172

Method 3 113 606 257 1141

20 degC

Exact 635 25

Method 1 123 60~75 73~338 298~1220

Method 2 123 619~662 229~310 934~1188

Method 3 123 631 284 1161

50 degC

Exact 635 25

Method 1 137 60~75 72~328 263~1228

Method 2 137 622~687 183~299 666~1086

Method 3 137 642 264 960

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1247 ms at 0 degC 1213 ms at 20 degC and 1168 ms at 50 degC

- 28 -

542 Impact-induced damage

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Vol

tage

(mV

)

Transmitted WaveReflected Wave

No reflected wave

(a) 0 degC

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Transmitted WaveReflected Wave

No reflected wave

(b) 50 degC

Fig 16 False alarm test under temperature variations It is shown that proposed damage char-acterization algorithm is robust under temperature variations

The false alarm test in pristine condition of the impact-induced damage specimen under

temperature variation was investigated (Fig 16) It is validated that the proposed damage di-

agnosis algorithm is robust under temperature variations Fig 17 shows the obtained signals

from the impact-induced damage (impacted 3 times) specimen under temperature variation

The analyzed results of the obtained signals are reported in Table VII The measured 119907119907119894119894 in

each temperature and the assumed variation are same as the artificial damage case

11 115 12 125 13-5

0

5

Time (ms)

Volta

ge (m

V)

0 oC20 oC50 oC

(a) Pitch-catch signal

11 115 12 125 13 135

-1

0

1

2

3

Time(ms)

Vol

tage

(mV

)

0 oC20 oC50 oC

(b) Pulse-echo signal

Fig 17 Signals obtained from impact-induced damage specimen (Impact 3) under tempera-ture variation

- 29 -

TABLE VII Temperature effect test for the impact induced damage specimen (Impact 3 case)

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

0 degC

Method 1 135 75~90 72~310 267~1148

Method 2 135 705~779 264~381 978~1126

Method 3 135 763 289 1071

20 degC

Method 1 148 50~100 06~380 21~1134

Method 2 148 758~838 249~368 843~1125

Method 3 148 798 308 1044

50 degC

Method 1 153 75~100 53~415 174~1227

Method 2 153 804~889 214~336 702~1103

Method 3 153 815 320 1050

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1247 ms at 0 degC 1213 ms at 20 degC and 1168 ms at 50 degC

- 30 -

CHAPTER 6 CONCLUDING REMARKS

61 Executive summary

In this study the reference-free damage detection localization and quantification tech-

nique in a multi layered composite plate is proposed Additional reflected A0 Lamb wave

mode from the damage due to multiple reflections in damage area is used as a feature for

characterizing the damage First damage is characterized in terms of a damage index (m-

value) defined as the ratio of the physical length of the damage to the wavelength of the A0

Lamb wave mode Then the range of possible damage location length and severity are esti-

mated with constraints This procedure can be accomplished without prior baseline data To

realize the proposed technique the matching pursuit technique is modified for reduction of

the error by considering the dispersive characteristic of Lamb wave The matching pursuit

method estimates the arrival time of each wave packets accurately even two wave packets are

overlapped The numerical simulations and experimental tests including Teflon tape inserted

and impact-induced damage are investigated for the feasibility of the proposed technique

Moreover the effectiveness of the proposed technique under changing temperature is exper-

imentally verified

- 31 -

62 Suggestion for further study

The limitations of this study and the possible improvement and further studies are listed as

follow

1 Measurement of impact-induced damage length using conventional NDT technique

Additional experimental verification for impact-induced damage characterization is

needed by measuring the real length of damage using conventional NDT technique

such as ultrasonic scan or infrared camera

2 Limitation of both side collocated PZT installation In this study excitation PZTs are

installed both side and collocated for decomposing A0 Lamb wave mode However it

can produce decomposition error due to imperfection of PZT location and their installa-

tion Moreover in many cases assess to both surface of structures is not allowed such

as aircraft applications To improve this limitation a reference-free damage characteri-

zation using single side PZTs will be considered

3 Complex structure application In this study damage is characterized in simple plate-

like structure When the monitored structure becomes complex the response signals al-

so becomes complex due to its geometry and boundary Thus the investigations for

complex structure application such as stiffener spar and curvature are needed

- 32 -

REFERENCES

[1] Cawley P and Alleyne D (1996) ldquoThe use of Lamb waves for the long range inspection of large structuresrdquo Ultrasonics 34 pp287-290

[2] Ghosh T Kundu T and Karpur P (1998) ldquoEfficient use of Lamb waves for detecting defects in large platesrdquo Ultrasonics 36 pp791-801

[3] Guirguitui V (2003) ldquoLamb wave generation with piezoelectric wafer active sensors for structural health monitoringrdquo SPIE Conference on Smart Structures and Materials San Diego CA 5391 pp111-422

[4] Ragvahan A and Cesnik CE S (2004) ldquoModeling of piezoelectric-based Lamb wave generation and sensing for structural health monitoringrdquo SPIE Conference on Smart Struc-tures and Materials San Diego CA 5391 pp419-430

[5] Rose JL Pilarski A and Dirit JJ (1993) ldquoAn approach to guided wave mode selection for inspection of laminated platerdquo Journal of Reinforced Plastics and Composites 12(5) pp536-544

[6] Guo N and Cawley P (1993) ldquoThe interaction of Lamb waves with delaminations in composite laminatesrdquo Journal of Acoustical Society of America 94(4) pp 2240-2246

[7] Valdez SHD and Soutis C (2002) ldquoReal-time nondestructive evaluation of fiber com-posite laminates using low-frequency Lamb wavesrdquo Journal of Acoustical Society of Amer-ica 111(5) pp 2026-2033

[8] Yang M and Qiao P (2005) ldquoModeling and experimental detection of damage in virous materials using the pulse-echo method and piezoelectric sensoractuatorsrdquo Smart Materials and Structures 14 pp1083-1100

[9] Lestari W and Qiao P (2005) ldquoApplication of wave propagation analysis for damage identification in composite laminated beamsrdquo Journal of Composite Materials 39(22) pp1967-1984

[10] Kessler S S Spearing S M and Soutis C (2002) ldquoDamage detection in composite ma-terials using Lamb wave methodsrdquo Smart Materials and Structures 11 pp269-278

[11] Sohn H Park G Wail J R Limback N P and Farrar C R (2004) ldquoWavelet-based active sensing for delamination detection in composite structuresrdquo Smart Materials and Structures 13 pp 153-160

[12] Petculescu G Krishnaswamy S and Achenback J D (2008) ldquoGroup delay measure-ments using modally selective Lamb wave transducers for detection and sizing of delamina-tions in compositesrdquo Smart Materials and Structures 17

[13] Sohn H Park H W Law K H and Farrar C R (2007) ldquoDamage detection in compo-

- 33 -

site plates by using an enhanced time reversal methodrdquo Journal of Aerospace Engineering (ASCE) pp 141-151

[14] Sohn H (2007) ldquoEffects of environmental and operational variability on structural health monitoringrdquo Philosophical Transactions of the Royal Society Ardquo 365 pp539-560

[15] Rytter A (1993) ldquoVibration based inspection of civil engineering structuresrdquo PhD The-sis Department of Building Technology and Structural Engineering University of Aalborg Aalborg Denmark

[16] Hayashi T and Kawashima K (2002) ldquoMultiple reflections of Lamb waves at a delami-nationrdquo Ultrasonics 40 pp193-197

[17] Wang C H and Rose L R F (2003) ldquoWave reflection and transmission in beams con-taining delamination and inhomogeneityrdquo Journal of Sound and Vibration 264 pp 851-872

[18] Ramadas C Balasubramaniam K Joshi M and Krishnamurthy C V (2009) ldquoInterac-tion of the primary anti-symmetric Lamb mode (Ao) with symmetric delaminations numer-ical and experimental studiesrdquo Smart Materials and Structures 18

[19] Ramadas C Balasubramaniam K Joshi M and Krishnamurthy C V (2010) ldquoInterac-tion of guided Lamb waves with an asymmetrically located delamination in a laminated composite platerdquo Smart Materials and Structures 19

[20] Wilcox P Lowe M and Cawely P (2001) ldquoThe effect of dispersion on long-range in-spection using ultrasonic guided wavesrdquo NDTampE International 34 pp1-9

[21] Mallat S G and Zhang Z (1993) ldquoMatching pursuits with time-frequency dictionariesrdquo IEEE Tansactions on Signal Processing 41(12) pp3397-3415

[22] Zhang G Zhang S and Wang Y (2000) ldquoApplication of adaptive time-frequency de-composition in ultrasonic NDE of highly scattering materialsrdquo Ultrasonics 38 pp 961-964

[23] Hong J C Sun K H and Kim Y Y (2005) ldquoThe matching pursuit approach based on the modulated Gaussian pulse for efficient guided-wave damage inspectionrdquo Smart Mate-rials and Structures 14 pp548-560

[24] Herakovich C T (1997) Mechanics of Fibrous Composites Wiley 480 pages

[25] Reddy J N (1985) ldquoA review of the literature on finite-element modeling of laminated composite platesrdquo The Shock and Vibration Digest 17(4) pp 3-8

[26] Reddy J N (2003) Mechanics of Laminated Composite Plates and Shells Theory and Analysis CRC Press 856 pages

[27] Palacz M Krawczuk M and Ostachowicz W (2005) ldquoThe spectral finite element mod-el for analysis of flexure-shear coupled wave propagation Part 1 laminated multilayer

- 34 -

composite beamrdquo Composite Structures 68 pp37-44

[28] Zhang Y C and Harding J (1990) ldquoA numerical micromechanics analysis of the me-chanical properties of a plain weave compositerdquo Computer amp Structures 36(5) pp 839-844

[29] Nemat-Nasser S and Hori M (1999) Micromechanics Overall Properties of Heteroge-neous Materials Elsevier 810 pages

[30] Scida D Aboura Z Benzeggagh M L and Bocherens E (1999) ldquoA micromechanics model for 3D elasticity and failure of woven-fibre composite materialrdquo Composites Science and Technology 59 pp 505-517

[31] Thomas Y and Wu X H (1997) ldquoA multiscale finite element method for elliptic prob-lems in composite materials and porous mediardquo Journal of Computational Physics 134 pp 169-189

[32] Lee K H Moorthy S and Ghosh S (1999) ldquoMultiple scale computational model for damage in composite materialsrdquo Computer Methods in Applied Mechanics and Engineer-ing 172 pp 175-201

[33] Namilae S and Chandra N (2005) ldquoMultiscale model to study the effect of interfaces in carbon nanotube-based compositesrdquo Journal of Engineering Materials and Technology (ASME) 127 pp222-232

[34] Adams D E (2007) Health monitoring of Structural Materials and Components Wiley 460 pages

[35] Gopalakrishnan S (2009) ldquoModeling Aspects in Finite Elementrdquo Encyclopedia of Struc-tural Health Monitoring 2 pp791-809

[36] Palacz M Krawczuk M and Ostachowicz W (2005) ldquoThe spectral finite element mod-el for analysis of flexure-shear coupled wave propagation Part 2 Deminated multilayer composite beamrdquo Composite Structures 68 pp45-51

[37] Baik J M and Thopson R B (1984) ldquoUltrasonic scattering from imperfect interfaces A quasi-static modelrdquo Journal of Nondestructive Evaluation 4(3) pp177-196

[38] Luo H and Hanagud S (2000) ldquoDynamics of delamincated beamsrdquo International Jour-nal of Solids and Structures 37 pp1501-1519

[39] Kim S B and Sohn H (2007) ldquoInstantaneous reference-free crack detection based on polarization characteristics of piezoelectric materialsrdquo Smart Materials and Structures 16 pp2375-2387

[40] An Y K and Sohn H (2010) ldquoInstantaneous crack detection under varying temperature and static loading conditionsrdquo The Special Issue of Structural Control and Health Monitor-ing 17(7) pp 730-741

- 35 -

[41] Lee S J and Sohn H (2006) ldquoActive self-sensing scheme development for structural health monitoringrdquo Smart Materials and Structures 15(6) pp 1734-1746

- 36 -

요 약 문

복합재료 손상의 무기저 진단 및 정량화

복합재료는 재료의 경량성 내구성 고강도의 특성으로 인하여 항공기에서부터

교량에까지 널리 사용되고 있다 하지만 외부의 충격으로 인한 복합재료 내부의

손상은 재료의 강도를 감소시켜 구조물의 안정성을 저하시키는 요인이 된다 하지만

복합재료의 손상은 재료 내부에서 발생하여 육안으로는 식별하기도 힘들며 표면에서

보이는 것만으로 내부의 손상 정도를 판단하기는 불가능하다는 단점을 가지고 있다

본 연구에서는 손상 내부에서 반사로 인해 발생하는 파를 이용하여 기저자료를

사용하지 않고 특정 시점에서 얻은 자료만으로 복합재료의 손상을 진단하고 정량화

할 수 있는 기법을 제안한다 제안된 기법의 적용성을 검증하기 위하여 유한요소법을

이용한 수치해석을 수행하였고 Teflon tape 이 삽입된 시편과 실제 충격하중으로

인한 손상을 입은 시편을 이용한 실험을 진행하였다 수치해석 및 실험 결과 제안된

기법이 복합재료의 손상을 성공적으로 진단 및 정량화할 수 있음을 확인하였다 또한

다양한 외부 온도 환경에서 제안된 기법의 환경영향 실험을 수행하였고 다양한

환경에서 손상의 진단 및 정량화가 성공적으로 이루어짐을 검증하였다

핵심어 무기저 손상진단 및 정량화 압전센서 복합재료 유한요소법 Matching pursuit

method

- 37 -

감 사 의 글

기대와 두려움을 안고 시작한 석사과정이 어느덧 2 년 이라는 세월 속에서 추억으로 자리잡아

가고 있습니다 학부생 기간을 포함한 지난 3 년 동안 SHM 이라는 새로운 분야를 접하고 연구를

수행 할 수 있도록 도와주신 고마운 분들이 있어 그 감사를 표하고자 합니다

우선 저의 많은 부족한 점을 일깨워 주시고 연구에 대한 뜻을 더욱 확고히 세울 수 있게 해

주신 지도교수님이신 손 훈 교수님께 깊은 감사의 뜻을 전합니다 짧은 기간이었지만 교수님의

도움으로 많은 것을 배우고 느끼고 깨달을 수 있었습니다 또한 본 연구를 수행할 수 있도록 지

원해주신 국토해양부 (U-City 석 박사과정 지원사업)에 깊은 감사 드립니다

때로 힘들고 지칠 때 항상 저를 믿고 응원해 주시는 가족이 있었기에 지금의 글을 쓸 수 있었

습니다 또한 지금 연구실에서 함께 생활하고 있는 윤규형 민구형 철민이형 현석이형 진열이형

호민이형 병진이 선혜누나와 성대에 계신 창길이형 미국에 계신 의재형과 Debaditiya 에게도 감

사의 말씀을 전합니다 그리고 모두 나열할 수는 없지만 카이스트 생활의 활력소가 되어준 카이

스트 건설 및 환경공학과 선후배들 직원 분들과 SOS 멤버들에게도 감사를 표합니다

앞서 거론한 많은 이들의 도움이 없었다면 저 또한 지금 이 자리에 있을 수 없었을 것입니다

많이 부족하기에 더 노력하고 배우고자 했던 초심을 다시 한번 생각하며 더 큰 결실을 맺기 위해

앞으로 나아가겠습니다 도움을 주신 모든 분들께 이 감사의 글을 바칩니다

2011년 6월 카이스트에서 임형진 올림

- 38 -

CURRICULUM VITAE

Hyung Jin Lim

Research Assistant

The Civil and Environmental Engineering Department

Korea Advanced Institute of Science and Technology (KAIST)

291 Daehak-Ro Guseong-Dong Yuseong-Gu

Daejeon Republic of Korea 305-701

Phone (82)+42-350-3665 Fax (82)+42-350-3610

Email limhj87gmailcom

RESEARCH INTERESTS

Damage Detection Localization and Quantification on Composite and Metallic Structures Pattern Recognition Finite Element Modeling (FEM)

EDUCATION

2009 ndash 2011 MS Civil and Environmental Eng Korea Advanced Institute of Science and Technology (KAIST) Korea Rep

2006 ndash 2009 BS Civil and Environmental Eng Korea Advanced Institute of Science and Technology (KAIST) Korea Rep

JOURNAL PUBLICATIONS The corresponding author is underlined 1 Hyung Jin Lim Hoon Sohn ldquoReference-free damage detection localization and quantification

using in compositerdquo in preparation for Journal of the Acoustical Society of America 2011

2 Chul Min Yeum Hoon Sohn Jeong Bum Ihn Hyung Jin Lim ldquoInstantaneous delamination detection in a composite using a dual piezoelectric transducer networkrdquo Submitted to Compo-sites Science and Technology 2011

3 Yun-Kyu An Hyung Jin Lim Min Koo Kim Jin Yeol Yang Hoon Sohn Chang Guen Lee ldquoInvestigation of applicability of local reference-free damage detection techniques to in-situ

- 39 -

bridgesrdquo Submitted to ASCE Journal of Structural Engineering 2011

4 Hyung Jin Lim Min Koo Kim Hoon Sohn Chan Yik Park ldquoImpedance based damage detec-tion under varying temperature and loading conditionsrdquo Submitted to NDTampE International 2011

5 Hyung Jin Lim Hoon Sohn ldquoQuantitative estimation of transmitted and reflected lamb waves at discontinuityrdquo Journal of the Korean Society for Nondestructive Testing Vol30 No 4 pp1-8 2010

CONFERENCE PROCEEDINGS

The corresponding author is underlined 1 Hyung Jin Lim and Hoon Sohn ldquoReference-free damage detection localization and quantifi-

cation in compositesrdquo ASCE Engineering Mechanics Institute Conference Boston June 2-4 2011

2 Hyung Jin Lim and Hoon Sohn ldquoReference-free delamination quantification within multi-layer composite materialsrdquo KSNT Spring Conference May 26-27 Gyeong-Ju Korea 2011

3 Min Koo Kim Hyun Jin Lim Hoon Sohn Chan Yik Park ldquoImpedance-based bolt loosening detection under varying temperature and loadingrdquo Asian Pacific Workshop on Structural Health Monitoring Tokyo Japan November 30-December 2 2010

4 Hyung Jin Lim Chul Min Yeum and Hoon Sohn ldquoModeling of Impact-induced Delamination in a Multilayer Composite Platerdquo The 23rd KKCNN Symposium on Civil Engineering Taipei (Taiwan) November 13-15 2010

5 Yun-Kyu An Hyung Jin Lim and Hoon Sohn ldquoTemperature independent Delamination Detec-tion using data normalizationrdquo The 3rd International Conference on Multi-Functional Materi-als and Structures Chonbuk National Univ Jeonju Korea September 14-18 2010

6 Hoon Sohn Yun-Kyu An Chang Gil Lee Min Koo Kim Hyung Jin Lim ldquoApplication of online NDT techniques to decommissioned bridge testingrdquo the 5th European Workshop on Structural Health Monitoring Sorrento Italy June 29-July 02 2010

7 Min Koo Kim Hyung Jin Lim Hoon Sohn Chan Yik Park ldquoImpedance based damage diag-nosis of a complex composite aircraft wing under changing loading and temperature condi-tionsrdquo the 5th European Workshop on Structural Health Monitoring Sorrento Italy June 29-July 02 2010

8 Yun-Kyu An Min Koo Kim Hyung Jin Lim and Hoon Sohn ldquoValidations of reference-free crack detection techniques through a decommissioned bridge testrdquo 2010 COSEIK Annual Con-ference Jeju Korea April 8-9 2010

- 40 -

PARTICIPANT PROJECTS

1 Development of Smart Steel Members for Online Steel Structure Monitoring POSCO Corporation (Funded 41000000 KRW (41000 USD) for 010511 to 123111)

2 Development of Enabling Technologies for Structural Health Monitoring Korean Agency for Defense Development (Funded 240000000 KRW (240000 USD) out of total 3200000000 KRW (3200000 USD) for 060108-053111)

3 Development of On-board SHM Technologies for Composite Air Vehicles The Boeing Company (Funded 400000 USD (400000000 KRW) for 080108 to 073111)

EDUCATIONAL EXPERIENCES

Fall 2009

Teaching Assistant CE207 ldquoElementary Structural Engineering and Laboratoryrdquo KAIST 3 unit undergraduate course

EXPERIMENTAL EXPERIENTCES

2010 Monitoring of Real Aircraft Composite Wing Segment Korea Agency for Defense De-velopment (ADD) Daejeon Korea Real time monitoring of composite air vehicles fund-ed by Korea Agency for Defense Development (ADD)

2009-2011

Field Monitoring of Real Bridge Structures Gimpo (Ramp G bridge) amp Icheon (Sam-seung bridge) Korea US-Korea collaborative research for bridge monitoring testbeds

HONORS amp AWARDS

Scholarships for outstanding students granted KAIST (2007-2009) The three prominent students in the civil and environmental engineering department of KAIST were selected as recipients This scholarship provided 2000$ per year for three years

- 41 -

Page 26: 복합재료 손상의 무기저 진단 및 정량화 - KAISTssslab.kaist.ac.kr/article/pdf/Thesis_2.pdf · 1 Declaration of Ethical Conduct in Research: I, as a graduate student

CHAPTER 4 NUMERICAL SIMULATION

41 Composite modeling method

Composite materials can be modeled from nano to macro level depends on its engineering

purpose In general four methods are utilized for modeling of composite materials First

composite is modeled as a homogenous material expressed in terms of effective material

properties [24] The effective material properties are estimated using the modulus and the

volume fraction of each constituent theoretically (Rule of mixture) and measured experimen-

tally Second a model is established that each laminate which have own material properties

are totally bonded together In this model the stacking sequence and direction of each lami-

nates are considered additionally [25-27] Third method for composite modeling in micro

level called micromechanics [28-30] This model considers the interaction between fibers and

matrix in laminate Last multiscale modeling which the scope of modeling is extended to

molecular or quantum level of constituents [31-33] This model starts from molecular me-

chanics or molecular dynamics then the scale of model leads to the nano micro and macro

level by incorporating small scale models As the scale of models is smaller it provides more

logical way than other large scale of models while the computational cost is more expensive

The scope of this study is focused on the dynamic behavior of composite structure the beam

or plate model was utilized for the numerical simulation

- 15 -

42 Damage modeling method

Generally three damage models have been developed and utilized in the real applications

First the stiffness reduction model called as the equivalent element model was introduced

[17 34-35] This model is simply realized by replacing the region around damage area by a

region with adjusted material properties (Youngrsquos modulus shear modulus density etc)

globally Next the duplicated node model is done by disconnecting the element connectivity

assuming there is no interaction among the delaminated surface This model can be utilized

when the location and the extension of damage is important issue [34-36] Last the contact

mechanics model using the principles of contact mechanics the two interfacial forces (con-

tact resistance and sliding resistance) in the delaminated surface [37-38] In this model the

delaminated surfaces are modeled as nonlinear functions of the interlaminar normal strain

and the interlaminar shear strain This model is computationally expensive while provides

more logical way than other models There is however no universally accepted model for

general impact-induced damage in composites In the case of impact-induced damage it is

impossible to predict or estimate the location and number of delaminated layers Thus from

the uncertainty of impact-induced damage the equivalent element model can be considered

as the effective damage model

- 16 -

43 Numerical simulation description

TABLE I Material property of simulation model and test specimen

Tensile Strength 657 MPa Tensile Modulus 59 GPa Shear Strength 804 MPa Shear Modulus 24 Gpa Specific Gravity 16

PZT A

PZT C

PZT BDamage

3

250 125 125

500 Unit mm

Fig 6 Description of the numerical simulation model

The feasibility of the proposed damage characterization technique is verified by 2-D FEM

simulation using MSC PATRAN and NASTRAN Fig 6 presents the dimension PZTs and

damage location of the simulation model In the model 05mm x 05mm square shell element

was used for modeling of the A0 mode whose flexural wave characteristic The element size

was determined by considering the wavelength of the propagation Lamb wave modes The

material properties of the simulation model and damage cases are presented in Table I and

Table II respectively

PZT A and C act as both actuator and sensor for collecting pulse-echo signal They were

installed on the plate which two of them are both-side collocated for the A0 Lamb wave

- 17 -

mode decomposition [39-40] PZT B was only used as a sensor for collecting pitch-catch sig-

nal For simulating the excitation and sensing of PZTs the pin-force model was adopted that

an input force apples to the end nodes of excitation PZT and averaging element strains corre-

sponding sensing area A Gaussian pulse (tone-burst) 50 kHz central frequency was excited

as an input signal

TABLE II Damage cases for numerical simulation

119949119949120783120783 (mm) 119941119941 (mm) G value reduction ()

Case 1 650 200 90 darr

Case 2 675 150 90 darr

Case 3 675 150 93 darr

Case 4 900 150 90 darr

- 18 -

44 Numerical simulation results

015 02 025 03 035

-2

0

2

Time (ms)

Am

plitu

de

IntactDamage

Reflected wave from damage

(a) Pitch-catch signal

015 02 025 03 035

-2

0

2

Time (ms)

Am

plitu

de

IntactDamage

Reflected waves from damage

(b) Pulse-echo signal Fig 7 Comparison of A0 Lamb wave mode signals from numerical simulation between intact and damage cases Time delay attenuation and multiple reflections due to damage simulated successfully using the damage modeling method described in the previous section

Fig 7 shows that comparison of A0 Lamb wave mode signals between intact and one of

damage cases in the simulation The features of A0 Lamb wave mode passing through dam-

age area described in section 2 are successfully simulated Thus the equivalent element

method is proper for modeling damage in composites In the damage cases the additional

waves reflected from damage is observed in both pitch-catch and pulse echo signal

Table III presents the result of the proposed damage characterization algorithm using the

numerical simulation cases shown in Table II In lsquoMethod 2rsquo 119907119907119894119894 variation is assumed to plusmn

5 of the normal value (1190~1315ms) The normal value of 119907119907119894119894 for lsquoMethod 3rsquo is meas-

ured as 1253 ms In each case the damage is characterized using the proposed technique

Note that as the information of 119907119907119894119894 is specified the estimated value is closer to the exact

value

- 19 -

TABLE III Damage characterization for the numerical simulation models

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

Case 1

Exact 650 200

Method 1 248 50~70 79~414 160~834

Method 2 248 603~667 135~241 271~486

Method 3 248 635 188 379

Case 2

Exact 675 150

Method 1 193 60~75 01~251 37~651

Method 2 193 618~682 114~222 295~575

Method 3 193 65 168 435

Case 3

Exact 675 150

Method 1 233 60~75 66~302 140~649

Method 2 233 668~738 85~196 182~419

Method 3 233 703 140 301

Case 4

Exact 900 150

Method 1 204 75~100 48~349 118~856

Method 2 204 860~951 107~216 264~530

Method 3 204 906 162 397

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1253 ms 1190~1315 ms Measured 119907119907119894119894 1253 ms

- 20 -

CHAPTER 5 EXPERIMENTAL VERIFICATION

51 Overview

In this section experimental study is investigated so that the proposed damage characteri-

zation technique Overall experimental setup and test specimen are shown in Fig 8 and 9

The experimental setup consists of test specimens data acquisition (DAQ) system and self-

sensing circuit Self-sensing circuit was installed for pulse-echo measurement using PZTs

[40-41]

PZT A(PZT C) PZT B

500

500

140

75

250

Teflon Inserted Damage (d =25)

Unit mm

Thickness = 3

PZT A(PZT C) PZT B

500

500

150

90

250

Impact Location

Unit mm

Thickness = 3

(a) Teflon inserted damage specimen

(b) Impact-induced damage specimen

Fig 8 Test specimens for the experimental verification In the Teflon inserted damage Teflon tapes with 25mm diameter were inserted at the center of the specimen The impact-induced damage was introduced to the specimen using the impact test equipment

- 21 -

511 Specimen

Two carbon composite plates with a dimension of 500times500times3mm3 were manufactured by

Nexcomes Inc by stacking 12 layers of CF3327 3K prepreg (woven fabrics type) manufac-

tured by Hankuk Fiber Inc The material properties of the specimen are presented in Table I

One of the specimens Teflon tapes with 25mm diameter was inserted in every 3 layers at the

center of the specimen to mimic the damage The other specimen is utilized for the character-

ization of real impact-induced damage

512 Data acquisition (DAQ) system and PZT

The data acquisition (DAQ) system consists of an arbitrary waveform generator (AWG) a

high speed signal digitizer (DIG) and two multiplexers (MUX) Using the 16-bits AWG

tone-burst input signal with a plusmn10 peak-to-peak voltage were generated The response signal

was measured by the 14-bit DIG For the signal-to-noise ratio reduction both pitch-catch and

pulse-echo signals were measured 150 times and averaged in the time domain 3 identical

circular PZTs of 10mm diameter manufactured by APC International Ltd were installed on

the specimen A Gaussian pulse (tone-burst) 50 kHz central frequency was excited as an in-

put signal The excitation frequency was chosen in order to avoid overlapping of the S0 and

A0 Lamb wave modes and generating of the higher modes (S1 A1hellip)

- 22 -

PZT Actuator

PZT Sensor

AWG

DIG

DAQ System Specimen

SensingActuatingMUX

1

23

4

1 Sending signal to MUX

4 Sending signal to DIG

Self-SensingCircuit

18 Unit mm

10

(a)

(b)

Fig 9 Overall experimental setup (a) Data acquisition (DAQ) system and (b) Installed circu-lar PZT transducer on the specimen with its dimension

513 Impact test

The impact-induced damage was introduced to the specimen using the impact test equip-

ment presented in Fig 10 A 5kg mass with 10mm tip was dropped from 015m height onto

the same location of the specimen The signals were obtained at each number of impacts

Since the reflected wave is not detected at the first impact case the data obtained from the

second and third impact were utilized for the analysis

10

Unit mm (a)

(b)

Fig 10 (a) Impact test equipment for introducing impact-induced damage and (b) Tip of the impact tester with 10mm diameter

- 23 -

514 Temperature effect test

In reality structures are subjected to changing environmental conditions such as tempera-

ture that affect measured signal and cause false-alarm Thus the robustness of the proposed

damage characterization technique under temperature variation in investigated (Fig 11) The

specimens were placed inside a temperature chamber and a thermocouple was installed on

the specimen to measure temperature of surface of the specimen The signal was obtained

under three temperature cases (0 20 and 50degC) and the humidity was kept at 30

Temperature Chamber

Test Specimen

Fig 11 Test specimen in temperature chamber

- 24 -

52 Teflon inserted damage

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Reflected wave from damage

(a)

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Transmitted WaveReflected Wave

Detected reflected wave

(b)

Fig 12 A0 Lamb wave mode obtained using pitch-catch method from the Teflon inserted damage specimen (a) The reflected wave from damage is shown (b) The Teflon inserted damage is detected by capturing the reflected wave

Fig 12 shows the obtained pitch-catch signal from the Teflon inserted damage specimen

and damage diagnosis result The Teflon inserted damage is detected successfully by captur-

ing the reflected wave Table IV presents the result of the Teflon inserted damage characteri-

zation 119907119907119894119894 is measured as 1213ms and the variation is assumed to plusmn 5 of the normal val-

ue (1152~1273ms) The results indicate that the proposed damage characterization approach

works well in real application

TABLE IV Damage characterization for the Teflon inserted damage specimen

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

Exact 635 25

Method 1 123 60~75 73~338 298~1220

Method 2 123 619~662 229~310 934~1188

Method 3 123 631 284 1161

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1213 ms

- 25 -

53 Impact-induced damage

11 115 12 125 13-4

-2

0

2

4

Time (ms)

Vol

tage

(mV

)

IntactImpact2Impact3

Reflected wave from damage

(a) Pitch-catch signal

11 115 12 125 13 135

-05

0

05

1

Time (ms)

Volta

ge (m

V)

IntactImpact2Impact3

Boundary reflection

Reflected wave from damage

(b) Pulse-echo signal

Fig 13 The signal obtained from the test specimen subjected to impact The impact was in-troduced at the same location of the specimen and the signal was obtained after each impacts As the number of impact is increased the transmitted A0 Lamb wave mode is delayed and attenuated

The obtained signals from impact-induced damage specimen are presented in Fig 13 It is

observed that as the number of impact is increased the transmitted A0 Lamb wave mode is

more delayed because the wave velocity in the damage area is reduced Fig 14 shows the

false-alarm test result whether the proposed damage detection algorithm works even in pris-

tine condition of the test specimen It is shown that no false alarm is occurred because no ad-

ditional reflected wave is captured

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Transmitted Wave

(a) Transmitted wave

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Reflected Wave

No reflected wave

(b) Reflected wave

Fig 14 False alarm test for the proposed damage detection algorithm in pristine condition of the specimen Reflected wave from damage is not captured false alarm is not occurred

- 26 -

TABLE V Damage characterization for the impact-induced damage specimen

of Impacts m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

2

(1470 J)

Method 1 080 65~90 75~246 470~1157

Method 2 080 801~885 114~245 711~1134

Method 3 080 843 179 1122

3

(2205 J)

Method 1 148 50~100 06~380 21~1134

Method 2 148 758~838 249~368 843~1125

Method 3 148 798 308 1044

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1213 ms

Table V presents the characterization results of the impact-induced damage case Here the

measured 119907119907119894119894 and the assumed variation are same as the artificial damage case It is observed

that as the number of impact is increased 119889119889 is increased while 119907119907119889119889 is reduced Thus it is

validated that the proposed damage characterization approach estimates the increment of the

damage well

54 Temperature effect test

541 Teflon inserted damage

11 115 12 125 13-10

-5

0

5

10

Time (ms)

Vol

tage

(mV

)

0 oC20 oC50 oC

(a) Pitch-catch signal

11 115 12 125 13 135

-1

0

1

2

3

Time(ms)

Volta

ge (m

V)

0 oC20 oC50 oC

(b) Pulse-echo signal

Fig 15 Signals obtained from the Teflon inserted damage specimen under temperature varia-tion Time delay and amplitude variation are occurred

Fig 15 shows the obtained signals from the Teflon inserted damage specimen subjected to

- 27 -

temperature variation Time delay and amplitude variations are occurred there is no addition-

al damage is introduced The analyzed results of the obtained signals are reported in Table VI

The measured 119907119907119894119894 is 1247 ms at 0 degC 1213 ms at 20 degC and 1168 ms at 50 degC respectively

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms (1152~1273 ms)

The estimated results are also changed under temperature variations the fluctuation of esti-

mated range and exact value is small

TABLE VI Temperature effect test for the Teflon inserted damage specimen

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

0 degC

Exact 635 25

Method 1 113 60~70 79~268 352~1291

Method 2 113 560~619 233~344 1034~1172

Method 3 113 606 257 1141

20 degC

Exact 635 25

Method 1 123 60~75 73~338 298~1220

Method 2 123 619~662 229~310 934~1188

Method 3 123 631 284 1161

50 degC

Exact 635 25

Method 1 137 60~75 72~328 263~1228

Method 2 137 622~687 183~299 666~1086

Method 3 137 642 264 960

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1247 ms at 0 degC 1213 ms at 20 degC and 1168 ms at 50 degC

- 28 -

542 Impact-induced damage

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Vol

tage

(mV

)

Transmitted WaveReflected Wave

No reflected wave

(a) 0 degC

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Transmitted WaveReflected Wave

No reflected wave

(b) 50 degC

Fig 16 False alarm test under temperature variations It is shown that proposed damage char-acterization algorithm is robust under temperature variations

The false alarm test in pristine condition of the impact-induced damage specimen under

temperature variation was investigated (Fig 16) It is validated that the proposed damage di-

agnosis algorithm is robust under temperature variations Fig 17 shows the obtained signals

from the impact-induced damage (impacted 3 times) specimen under temperature variation

The analyzed results of the obtained signals are reported in Table VII The measured 119907119907119894119894 in

each temperature and the assumed variation are same as the artificial damage case

11 115 12 125 13-5

0

5

Time (ms)

Volta

ge (m

V)

0 oC20 oC50 oC

(a) Pitch-catch signal

11 115 12 125 13 135

-1

0

1

2

3

Time(ms)

Vol

tage

(mV

)

0 oC20 oC50 oC

(b) Pulse-echo signal

Fig 17 Signals obtained from impact-induced damage specimen (Impact 3) under tempera-ture variation

- 29 -

TABLE VII Temperature effect test for the impact induced damage specimen (Impact 3 case)

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

0 degC

Method 1 135 75~90 72~310 267~1148

Method 2 135 705~779 264~381 978~1126

Method 3 135 763 289 1071

20 degC

Method 1 148 50~100 06~380 21~1134

Method 2 148 758~838 249~368 843~1125

Method 3 148 798 308 1044

50 degC

Method 1 153 75~100 53~415 174~1227

Method 2 153 804~889 214~336 702~1103

Method 3 153 815 320 1050

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1247 ms at 0 degC 1213 ms at 20 degC and 1168 ms at 50 degC

- 30 -

CHAPTER 6 CONCLUDING REMARKS

61 Executive summary

In this study the reference-free damage detection localization and quantification tech-

nique in a multi layered composite plate is proposed Additional reflected A0 Lamb wave

mode from the damage due to multiple reflections in damage area is used as a feature for

characterizing the damage First damage is characterized in terms of a damage index (m-

value) defined as the ratio of the physical length of the damage to the wavelength of the A0

Lamb wave mode Then the range of possible damage location length and severity are esti-

mated with constraints This procedure can be accomplished without prior baseline data To

realize the proposed technique the matching pursuit technique is modified for reduction of

the error by considering the dispersive characteristic of Lamb wave The matching pursuit

method estimates the arrival time of each wave packets accurately even two wave packets are

overlapped The numerical simulations and experimental tests including Teflon tape inserted

and impact-induced damage are investigated for the feasibility of the proposed technique

Moreover the effectiveness of the proposed technique under changing temperature is exper-

imentally verified

- 31 -

62 Suggestion for further study

The limitations of this study and the possible improvement and further studies are listed as

follow

1 Measurement of impact-induced damage length using conventional NDT technique

Additional experimental verification for impact-induced damage characterization is

needed by measuring the real length of damage using conventional NDT technique

such as ultrasonic scan or infrared camera

2 Limitation of both side collocated PZT installation In this study excitation PZTs are

installed both side and collocated for decomposing A0 Lamb wave mode However it

can produce decomposition error due to imperfection of PZT location and their installa-

tion Moreover in many cases assess to both surface of structures is not allowed such

as aircraft applications To improve this limitation a reference-free damage characteri-

zation using single side PZTs will be considered

3 Complex structure application In this study damage is characterized in simple plate-

like structure When the monitored structure becomes complex the response signals al-

so becomes complex due to its geometry and boundary Thus the investigations for

complex structure application such as stiffener spar and curvature are needed

- 32 -

REFERENCES

[1] Cawley P and Alleyne D (1996) ldquoThe use of Lamb waves for the long range inspection of large structuresrdquo Ultrasonics 34 pp287-290

[2] Ghosh T Kundu T and Karpur P (1998) ldquoEfficient use of Lamb waves for detecting defects in large platesrdquo Ultrasonics 36 pp791-801

[3] Guirguitui V (2003) ldquoLamb wave generation with piezoelectric wafer active sensors for structural health monitoringrdquo SPIE Conference on Smart Structures and Materials San Diego CA 5391 pp111-422

[4] Ragvahan A and Cesnik CE S (2004) ldquoModeling of piezoelectric-based Lamb wave generation and sensing for structural health monitoringrdquo SPIE Conference on Smart Struc-tures and Materials San Diego CA 5391 pp419-430

[5] Rose JL Pilarski A and Dirit JJ (1993) ldquoAn approach to guided wave mode selection for inspection of laminated platerdquo Journal of Reinforced Plastics and Composites 12(5) pp536-544

[6] Guo N and Cawley P (1993) ldquoThe interaction of Lamb waves with delaminations in composite laminatesrdquo Journal of Acoustical Society of America 94(4) pp 2240-2246

[7] Valdez SHD and Soutis C (2002) ldquoReal-time nondestructive evaluation of fiber com-posite laminates using low-frequency Lamb wavesrdquo Journal of Acoustical Society of Amer-ica 111(5) pp 2026-2033

[8] Yang M and Qiao P (2005) ldquoModeling and experimental detection of damage in virous materials using the pulse-echo method and piezoelectric sensoractuatorsrdquo Smart Materials and Structures 14 pp1083-1100

[9] Lestari W and Qiao P (2005) ldquoApplication of wave propagation analysis for damage identification in composite laminated beamsrdquo Journal of Composite Materials 39(22) pp1967-1984

[10] Kessler S S Spearing S M and Soutis C (2002) ldquoDamage detection in composite ma-terials using Lamb wave methodsrdquo Smart Materials and Structures 11 pp269-278

[11] Sohn H Park G Wail J R Limback N P and Farrar C R (2004) ldquoWavelet-based active sensing for delamination detection in composite structuresrdquo Smart Materials and Structures 13 pp 153-160

[12] Petculescu G Krishnaswamy S and Achenback J D (2008) ldquoGroup delay measure-ments using modally selective Lamb wave transducers for detection and sizing of delamina-tions in compositesrdquo Smart Materials and Structures 17

[13] Sohn H Park H W Law K H and Farrar C R (2007) ldquoDamage detection in compo-

- 33 -

site plates by using an enhanced time reversal methodrdquo Journal of Aerospace Engineering (ASCE) pp 141-151

[14] Sohn H (2007) ldquoEffects of environmental and operational variability on structural health monitoringrdquo Philosophical Transactions of the Royal Society Ardquo 365 pp539-560

[15] Rytter A (1993) ldquoVibration based inspection of civil engineering structuresrdquo PhD The-sis Department of Building Technology and Structural Engineering University of Aalborg Aalborg Denmark

[16] Hayashi T and Kawashima K (2002) ldquoMultiple reflections of Lamb waves at a delami-nationrdquo Ultrasonics 40 pp193-197

[17] Wang C H and Rose L R F (2003) ldquoWave reflection and transmission in beams con-taining delamination and inhomogeneityrdquo Journal of Sound and Vibration 264 pp 851-872

[18] Ramadas C Balasubramaniam K Joshi M and Krishnamurthy C V (2009) ldquoInterac-tion of the primary anti-symmetric Lamb mode (Ao) with symmetric delaminations numer-ical and experimental studiesrdquo Smart Materials and Structures 18

[19] Ramadas C Balasubramaniam K Joshi M and Krishnamurthy C V (2010) ldquoInterac-tion of guided Lamb waves with an asymmetrically located delamination in a laminated composite platerdquo Smart Materials and Structures 19

[20] Wilcox P Lowe M and Cawely P (2001) ldquoThe effect of dispersion on long-range in-spection using ultrasonic guided wavesrdquo NDTampE International 34 pp1-9

[21] Mallat S G and Zhang Z (1993) ldquoMatching pursuits with time-frequency dictionariesrdquo IEEE Tansactions on Signal Processing 41(12) pp3397-3415

[22] Zhang G Zhang S and Wang Y (2000) ldquoApplication of adaptive time-frequency de-composition in ultrasonic NDE of highly scattering materialsrdquo Ultrasonics 38 pp 961-964

[23] Hong J C Sun K H and Kim Y Y (2005) ldquoThe matching pursuit approach based on the modulated Gaussian pulse for efficient guided-wave damage inspectionrdquo Smart Mate-rials and Structures 14 pp548-560

[24] Herakovich C T (1997) Mechanics of Fibrous Composites Wiley 480 pages

[25] Reddy J N (1985) ldquoA review of the literature on finite-element modeling of laminated composite platesrdquo The Shock and Vibration Digest 17(4) pp 3-8

[26] Reddy J N (2003) Mechanics of Laminated Composite Plates and Shells Theory and Analysis CRC Press 856 pages

[27] Palacz M Krawczuk M and Ostachowicz W (2005) ldquoThe spectral finite element mod-el for analysis of flexure-shear coupled wave propagation Part 1 laminated multilayer

- 34 -

composite beamrdquo Composite Structures 68 pp37-44

[28] Zhang Y C and Harding J (1990) ldquoA numerical micromechanics analysis of the me-chanical properties of a plain weave compositerdquo Computer amp Structures 36(5) pp 839-844

[29] Nemat-Nasser S and Hori M (1999) Micromechanics Overall Properties of Heteroge-neous Materials Elsevier 810 pages

[30] Scida D Aboura Z Benzeggagh M L and Bocherens E (1999) ldquoA micromechanics model for 3D elasticity and failure of woven-fibre composite materialrdquo Composites Science and Technology 59 pp 505-517

[31] Thomas Y and Wu X H (1997) ldquoA multiscale finite element method for elliptic prob-lems in composite materials and porous mediardquo Journal of Computational Physics 134 pp 169-189

[32] Lee K H Moorthy S and Ghosh S (1999) ldquoMultiple scale computational model for damage in composite materialsrdquo Computer Methods in Applied Mechanics and Engineer-ing 172 pp 175-201

[33] Namilae S and Chandra N (2005) ldquoMultiscale model to study the effect of interfaces in carbon nanotube-based compositesrdquo Journal of Engineering Materials and Technology (ASME) 127 pp222-232

[34] Adams D E (2007) Health monitoring of Structural Materials and Components Wiley 460 pages

[35] Gopalakrishnan S (2009) ldquoModeling Aspects in Finite Elementrdquo Encyclopedia of Struc-tural Health Monitoring 2 pp791-809

[36] Palacz M Krawczuk M and Ostachowicz W (2005) ldquoThe spectral finite element mod-el for analysis of flexure-shear coupled wave propagation Part 2 Deminated multilayer composite beamrdquo Composite Structures 68 pp45-51

[37] Baik J M and Thopson R B (1984) ldquoUltrasonic scattering from imperfect interfaces A quasi-static modelrdquo Journal of Nondestructive Evaluation 4(3) pp177-196

[38] Luo H and Hanagud S (2000) ldquoDynamics of delamincated beamsrdquo International Jour-nal of Solids and Structures 37 pp1501-1519

[39] Kim S B and Sohn H (2007) ldquoInstantaneous reference-free crack detection based on polarization characteristics of piezoelectric materialsrdquo Smart Materials and Structures 16 pp2375-2387

[40] An Y K and Sohn H (2010) ldquoInstantaneous crack detection under varying temperature and static loading conditionsrdquo The Special Issue of Structural Control and Health Monitor-ing 17(7) pp 730-741

- 35 -

[41] Lee S J and Sohn H (2006) ldquoActive self-sensing scheme development for structural health monitoringrdquo Smart Materials and Structures 15(6) pp 1734-1746

- 36 -

요 약 문

복합재료 손상의 무기저 진단 및 정량화

복합재료는 재료의 경량성 내구성 고강도의 특성으로 인하여 항공기에서부터

교량에까지 널리 사용되고 있다 하지만 외부의 충격으로 인한 복합재료 내부의

손상은 재료의 강도를 감소시켜 구조물의 안정성을 저하시키는 요인이 된다 하지만

복합재료의 손상은 재료 내부에서 발생하여 육안으로는 식별하기도 힘들며 표면에서

보이는 것만으로 내부의 손상 정도를 판단하기는 불가능하다는 단점을 가지고 있다

본 연구에서는 손상 내부에서 반사로 인해 발생하는 파를 이용하여 기저자료를

사용하지 않고 특정 시점에서 얻은 자료만으로 복합재료의 손상을 진단하고 정량화

할 수 있는 기법을 제안한다 제안된 기법의 적용성을 검증하기 위하여 유한요소법을

이용한 수치해석을 수행하였고 Teflon tape 이 삽입된 시편과 실제 충격하중으로

인한 손상을 입은 시편을 이용한 실험을 진행하였다 수치해석 및 실험 결과 제안된

기법이 복합재료의 손상을 성공적으로 진단 및 정량화할 수 있음을 확인하였다 또한

다양한 외부 온도 환경에서 제안된 기법의 환경영향 실험을 수행하였고 다양한

환경에서 손상의 진단 및 정량화가 성공적으로 이루어짐을 검증하였다

핵심어 무기저 손상진단 및 정량화 압전센서 복합재료 유한요소법 Matching pursuit

method

- 37 -

감 사 의 글

기대와 두려움을 안고 시작한 석사과정이 어느덧 2 년 이라는 세월 속에서 추억으로 자리잡아

가고 있습니다 학부생 기간을 포함한 지난 3 년 동안 SHM 이라는 새로운 분야를 접하고 연구를

수행 할 수 있도록 도와주신 고마운 분들이 있어 그 감사를 표하고자 합니다

우선 저의 많은 부족한 점을 일깨워 주시고 연구에 대한 뜻을 더욱 확고히 세울 수 있게 해

주신 지도교수님이신 손 훈 교수님께 깊은 감사의 뜻을 전합니다 짧은 기간이었지만 교수님의

도움으로 많은 것을 배우고 느끼고 깨달을 수 있었습니다 또한 본 연구를 수행할 수 있도록 지

원해주신 국토해양부 (U-City 석 박사과정 지원사업)에 깊은 감사 드립니다

때로 힘들고 지칠 때 항상 저를 믿고 응원해 주시는 가족이 있었기에 지금의 글을 쓸 수 있었

습니다 또한 지금 연구실에서 함께 생활하고 있는 윤규형 민구형 철민이형 현석이형 진열이형

호민이형 병진이 선혜누나와 성대에 계신 창길이형 미국에 계신 의재형과 Debaditiya 에게도 감

사의 말씀을 전합니다 그리고 모두 나열할 수는 없지만 카이스트 생활의 활력소가 되어준 카이

스트 건설 및 환경공학과 선후배들 직원 분들과 SOS 멤버들에게도 감사를 표합니다

앞서 거론한 많은 이들의 도움이 없었다면 저 또한 지금 이 자리에 있을 수 없었을 것입니다

많이 부족하기에 더 노력하고 배우고자 했던 초심을 다시 한번 생각하며 더 큰 결실을 맺기 위해

앞으로 나아가겠습니다 도움을 주신 모든 분들께 이 감사의 글을 바칩니다

2011년 6월 카이스트에서 임형진 올림

- 38 -

CURRICULUM VITAE

Hyung Jin Lim

Research Assistant

The Civil and Environmental Engineering Department

Korea Advanced Institute of Science and Technology (KAIST)

291 Daehak-Ro Guseong-Dong Yuseong-Gu

Daejeon Republic of Korea 305-701

Phone (82)+42-350-3665 Fax (82)+42-350-3610

Email limhj87gmailcom

RESEARCH INTERESTS

Damage Detection Localization and Quantification on Composite and Metallic Structures Pattern Recognition Finite Element Modeling (FEM)

EDUCATION

2009 ndash 2011 MS Civil and Environmental Eng Korea Advanced Institute of Science and Technology (KAIST) Korea Rep

2006 ndash 2009 BS Civil and Environmental Eng Korea Advanced Institute of Science and Technology (KAIST) Korea Rep

JOURNAL PUBLICATIONS The corresponding author is underlined 1 Hyung Jin Lim Hoon Sohn ldquoReference-free damage detection localization and quantification

using in compositerdquo in preparation for Journal of the Acoustical Society of America 2011

2 Chul Min Yeum Hoon Sohn Jeong Bum Ihn Hyung Jin Lim ldquoInstantaneous delamination detection in a composite using a dual piezoelectric transducer networkrdquo Submitted to Compo-sites Science and Technology 2011

3 Yun-Kyu An Hyung Jin Lim Min Koo Kim Jin Yeol Yang Hoon Sohn Chang Guen Lee ldquoInvestigation of applicability of local reference-free damage detection techniques to in-situ

- 39 -

bridgesrdquo Submitted to ASCE Journal of Structural Engineering 2011

4 Hyung Jin Lim Min Koo Kim Hoon Sohn Chan Yik Park ldquoImpedance based damage detec-tion under varying temperature and loading conditionsrdquo Submitted to NDTampE International 2011

5 Hyung Jin Lim Hoon Sohn ldquoQuantitative estimation of transmitted and reflected lamb waves at discontinuityrdquo Journal of the Korean Society for Nondestructive Testing Vol30 No 4 pp1-8 2010

CONFERENCE PROCEEDINGS

The corresponding author is underlined 1 Hyung Jin Lim and Hoon Sohn ldquoReference-free damage detection localization and quantifi-

cation in compositesrdquo ASCE Engineering Mechanics Institute Conference Boston June 2-4 2011

2 Hyung Jin Lim and Hoon Sohn ldquoReference-free delamination quantification within multi-layer composite materialsrdquo KSNT Spring Conference May 26-27 Gyeong-Ju Korea 2011

3 Min Koo Kim Hyun Jin Lim Hoon Sohn Chan Yik Park ldquoImpedance-based bolt loosening detection under varying temperature and loadingrdquo Asian Pacific Workshop on Structural Health Monitoring Tokyo Japan November 30-December 2 2010

4 Hyung Jin Lim Chul Min Yeum and Hoon Sohn ldquoModeling of Impact-induced Delamination in a Multilayer Composite Platerdquo The 23rd KKCNN Symposium on Civil Engineering Taipei (Taiwan) November 13-15 2010

5 Yun-Kyu An Hyung Jin Lim and Hoon Sohn ldquoTemperature independent Delamination Detec-tion using data normalizationrdquo The 3rd International Conference on Multi-Functional Materi-als and Structures Chonbuk National Univ Jeonju Korea September 14-18 2010

6 Hoon Sohn Yun-Kyu An Chang Gil Lee Min Koo Kim Hyung Jin Lim ldquoApplication of online NDT techniques to decommissioned bridge testingrdquo the 5th European Workshop on Structural Health Monitoring Sorrento Italy June 29-July 02 2010

7 Min Koo Kim Hyung Jin Lim Hoon Sohn Chan Yik Park ldquoImpedance based damage diag-nosis of a complex composite aircraft wing under changing loading and temperature condi-tionsrdquo the 5th European Workshop on Structural Health Monitoring Sorrento Italy June 29-July 02 2010

8 Yun-Kyu An Min Koo Kim Hyung Jin Lim and Hoon Sohn ldquoValidations of reference-free crack detection techniques through a decommissioned bridge testrdquo 2010 COSEIK Annual Con-ference Jeju Korea April 8-9 2010

- 40 -

PARTICIPANT PROJECTS

1 Development of Smart Steel Members for Online Steel Structure Monitoring POSCO Corporation (Funded 41000000 KRW (41000 USD) for 010511 to 123111)

2 Development of Enabling Technologies for Structural Health Monitoring Korean Agency for Defense Development (Funded 240000000 KRW (240000 USD) out of total 3200000000 KRW (3200000 USD) for 060108-053111)

3 Development of On-board SHM Technologies for Composite Air Vehicles The Boeing Company (Funded 400000 USD (400000000 KRW) for 080108 to 073111)

EDUCATIONAL EXPERIENCES

Fall 2009

Teaching Assistant CE207 ldquoElementary Structural Engineering and Laboratoryrdquo KAIST 3 unit undergraduate course

EXPERIMENTAL EXPERIENTCES

2010 Monitoring of Real Aircraft Composite Wing Segment Korea Agency for Defense De-velopment (ADD) Daejeon Korea Real time monitoring of composite air vehicles fund-ed by Korea Agency for Defense Development (ADD)

2009-2011

Field Monitoring of Real Bridge Structures Gimpo (Ramp G bridge) amp Icheon (Sam-seung bridge) Korea US-Korea collaborative research for bridge monitoring testbeds

HONORS amp AWARDS

Scholarships for outstanding students granted KAIST (2007-2009) The three prominent students in the civil and environmental engineering department of KAIST were selected as recipients This scholarship provided 2000$ per year for three years

- 41 -

Page 27: 복합재료 손상의 무기저 진단 및 정량화 - KAISTssslab.kaist.ac.kr/article/pdf/Thesis_2.pdf · 1 Declaration of Ethical Conduct in Research: I, as a graduate student

42 Damage modeling method

Generally three damage models have been developed and utilized in the real applications

First the stiffness reduction model called as the equivalent element model was introduced

[17 34-35] This model is simply realized by replacing the region around damage area by a

region with adjusted material properties (Youngrsquos modulus shear modulus density etc)

globally Next the duplicated node model is done by disconnecting the element connectivity

assuming there is no interaction among the delaminated surface This model can be utilized

when the location and the extension of damage is important issue [34-36] Last the contact

mechanics model using the principles of contact mechanics the two interfacial forces (con-

tact resistance and sliding resistance) in the delaminated surface [37-38] In this model the

delaminated surfaces are modeled as nonlinear functions of the interlaminar normal strain

and the interlaminar shear strain This model is computationally expensive while provides

more logical way than other models There is however no universally accepted model for

general impact-induced damage in composites In the case of impact-induced damage it is

impossible to predict or estimate the location and number of delaminated layers Thus from

the uncertainty of impact-induced damage the equivalent element model can be considered

as the effective damage model

- 16 -

43 Numerical simulation description

TABLE I Material property of simulation model and test specimen

Tensile Strength 657 MPa Tensile Modulus 59 GPa Shear Strength 804 MPa Shear Modulus 24 Gpa Specific Gravity 16

PZT A

PZT C

PZT BDamage

3

250 125 125

500 Unit mm

Fig 6 Description of the numerical simulation model

The feasibility of the proposed damage characterization technique is verified by 2-D FEM

simulation using MSC PATRAN and NASTRAN Fig 6 presents the dimension PZTs and

damage location of the simulation model In the model 05mm x 05mm square shell element

was used for modeling of the A0 mode whose flexural wave characteristic The element size

was determined by considering the wavelength of the propagation Lamb wave modes The

material properties of the simulation model and damage cases are presented in Table I and

Table II respectively

PZT A and C act as both actuator and sensor for collecting pulse-echo signal They were

installed on the plate which two of them are both-side collocated for the A0 Lamb wave

- 17 -

mode decomposition [39-40] PZT B was only used as a sensor for collecting pitch-catch sig-

nal For simulating the excitation and sensing of PZTs the pin-force model was adopted that

an input force apples to the end nodes of excitation PZT and averaging element strains corre-

sponding sensing area A Gaussian pulse (tone-burst) 50 kHz central frequency was excited

as an input signal

TABLE II Damage cases for numerical simulation

119949119949120783120783 (mm) 119941119941 (mm) G value reduction ()

Case 1 650 200 90 darr

Case 2 675 150 90 darr

Case 3 675 150 93 darr

Case 4 900 150 90 darr

- 18 -

44 Numerical simulation results

015 02 025 03 035

-2

0

2

Time (ms)

Am

plitu

de

IntactDamage

Reflected wave from damage

(a) Pitch-catch signal

015 02 025 03 035

-2

0

2

Time (ms)

Am

plitu

de

IntactDamage

Reflected waves from damage

(b) Pulse-echo signal Fig 7 Comparison of A0 Lamb wave mode signals from numerical simulation between intact and damage cases Time delay attenuation and multiple reflections due to damage simulated successfully using the damage modeling method described in the previous section

Fig 7 shows that comparison of A0 Lamb wave mode signals between intact and one of

damage cases in the simulation The features of A0 Lamb wave mode passing through dam-

age area described in section 2 are successfully simulated Thus the equivalent element

method is proper for modeling damage in composites In the damage cases the additional

waves reflected from damage is observed in both pitch-catch and pulse echo signal

Table III presents the result of the proposed damage characterization algorithm using the

numerical simulation cases shown in Table II In lsquoMethod 2rsquo 119907119907119894119894 variation is assumed to plusmn

5 of the normal value (1190~1315ms) The normal value of 119907119907119894119894 for lsquoMethod 3rsquo is meas-

ured as 1253 ms In each case the damage is characterized using the proposed technique

Note that as the information of 119907119907119894119894 is specified the estimated value is closer to the exact

value

- 19 -

TABLE III Damage characterization for the numerical simulation models

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

Case 1

Exact 650 200

Method 1 248 50~70 79~414 160~834

Method 2 248 603~667 135~241 271~486

Method 3 248 635 188 379

Case 2

Exact 675 150

Method 1 193 60~75 01~251 37~651

Method 2 193 618~682 114~222 295~575

Method 3 193 65 168 435

Case 3

Exact 675 150

Method 1 233 60~75 66~302 140~649

Method 2 233 668~738 85~196 182~419

Method 3 233 703 140 301

Case 4

Exact 900 150

Method 1 204 75~100 48~349 118~856

Method 2 204 860~951 107~216 264~530

Method 3 204 906 162 397

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1253 ms 1190~1315 ms Measured 119907119907119894119894 1253 ms

- 20 -

CHAPTER 5 EXPERIMENTAL VERIFICATION

51 Overview

In this section experimental study is investigated so that the proposed damage characteri-

zation technique Overall experimental setup and test specimen are shown in Fig 8 and 9

The experimental setup consists of test specimens data acquisition (DAQ) system and self-

sensing circuit Self-sensing circuit was installed for pulse-echo measurement using PZTs

[40-41]

PZT A(PZT C) PZT B

500

500

140

75

250

Teflon Inserted Damage (d =25)

Unit mm

Thickness = 3

PZT A(PZT C) PZT B

500

500

150

90

250

Impact Location

Unit mm

Thickness = 3

(a) Teflon inserted damage specimen

(b) Impact-induced damage specimen

Fig 8 Test specimens for the experimental verification In the Teflon inserted damage Teflon tapes with 25mm diameter were inserted at the center of the specimen The impact-induced damage was introduced to the specimen using the impact test equipment

- 21 -

511 Specimen

Two carbon composite plates with a dimension of 500times500times3mm3 were manufactured by

Nexcomes Inc by stacking 12 layers of CF3327 3K prepreg (woven fabrics type) manufac-

tured by Hankuk Fiber Inc The material properties of the specimen are presented in Table I

One of the specimens Teflon tapes with 25mm diameter was inserted in every 3 layers at the

center of the specimen to mimic the damage The other specimen is utilized for the character-

ization of real impact-induced damage

512 Data acquisition (DAQ) system and PZT

The data acquisition (DAQ) system consists of an arbitrary waveform generator (AWG) a

high speed signal digitizer (DIG) and two multiplexers (MUX) Using the 16-bits AWG

tone-burst input signal with a plusmn10 peak-to-peak voltage were generated The response signal

was measured by the 14-bit DIG For the signal-to-noise ratio reduction both pitch-catch and

pulse-echo signals were measured 150 times and averaged in the time domain 3 identical

circular PZTs of 10mm diameter manufactured by APC International Ltd were installed on

the specimen A Gaussian pulse (tone-burst) 50 kHz central frequency was excited as an in-

put signal The excitation frequency was chosen in order to avoid overlapping of the S0 and

A0 Lamb wave modes and generating of the higher modes (S1 A1hellip)

- 22 -

PZT Actuator

PZT Sensor

AWG

DIG

DAQ System Specimen

SensingActuatingMUX

1

23

4

1 Sending signal to MUX

4 Sending signal to DIG

Self-SensingCircuit

18 Unit mm

10

(a)

(b)

Fig 9 Overall experimental setup (a) Data acquisition (DAQ) system and (b) Installed circu-lar PZT transducer on the specimen with its dimension

513 Impact test

The impact-induced damage was introduced to the specimen using the impact test equip-

ment presented in Fig 10 A 5kg mass with 10mm tip was dropped from 015m height onto

the same location of the specimen The signals were obtained at each number of impacts

Since the reflected wave is not detected at the first impact case the data obtained from the

second and third impact were utilized for the analysis

10

Unit mm (a)

(b)

Fig 10 (a) Impact test equipment for introducing impact-induced damage and (b) Tip of the impact tester with 10mm diameter

- 23 -

514 Temperature effect test

In reality structures are subjected to changing environmental conditions such as tempera-

ture that affect measured signal and cause false-alarm Thus the robustness of the proposed

damage characterization technique under temperature variation in investigated (Fig 11) The

specimens were placed inside a temperature chamber and a thermocouple was installed on

the specimen to measure temperature of surface of the specimen The signal was obtained

under three temperature cases (0 20 and 50degC) and the humidity was kept at 30

Temperature Chamber

Test Specimen

Fig 11 Test specimen in temperature chamber

- 24 -

52 Teflon inserted damage

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Reflected wave from damage

(a)

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Transmitted WaveReflected Wave

Detected reflected wave

(b)

Fig 12 A0 Lamb wave mode obtained using pitch-catch method from the Teflon inserted damage specimen (a) The reflected wave from damage is shown (b) The Teflon inserted damage is detected by capturing the reflected wave

Fig 12 shows the obtained pitch-catch signal from the Teflon inserted damage specimen

and damage diagnosis result The Teflon inserted damage is detected successfully by captur-

ing the reflected wave Table IV presents the result of the Teflon inserted damage characteri-

zation 119907119907119894119894 is measured as 1213ms and the variation is assumed to plusmn 5 of the normal val-

ue (1152~1273ms) The results indicate that the proposed damage characterization approach

works well in real application

TABLE IV Damage characterization for the Teflon inserted damage specimen

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

Exact 635 25

Method 1 123 60~75 73~338 298~1220

Method 2 123 619~662 229~310 934~1188

Method 3 123 631 284 1161

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1213 ms

- 25 -

53 Impact-induced damage

11 115 12 125 13-4

-2

0

2

4

Time (ms)

Vol

tage

(mV

)

IntactImpact2Impact3

Reflected wave from damage

(a) Pitch-catch signal

11 115 12 125 13 135

-05

0

05

1

Time (ms)

Volta

ge (m

V)

IntactImpact2Impact3

Boundary reflection

Reflected wave from damage

(b) Pulse-echo signal

Fig 13 The signal obtained from the test specimen subjected to impact The impact was in-troduced at the same location of the specimen and the signal was obtained after each impacts As the number of impact is increased the transmitted A0 Lamb wave mode is delayed and attenuated

The obtained signals from impact-induced damage specimen are presented in Fig 13 It is

observed that as the number of impact is increased the transmitted A0 Lamb wave mode is

more delayed because the wave velocity in the damage area is reduced Fig 14 shows the

false-alarm test result whether the proposed damage detection algorithm works even in pris-

tine condition of the test specimen It is shown that no false alarm is occurred because no ad-

ditional reflected wave is captured

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Transmitted Wave

(a) Transmitted wave

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Reflected Wave

No reflected wave

(b) Reflected wave

Fig 14 False alarm test for the proposed damage detection algorithm in pristine condition of the specimen Reflected wave from damage is not captured false alarm is not occurred

- 26 -

TABLE V Damage characterization for the impact-induced damage specimen

of Impacts m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

2

(1470 J)

Method 1 080 65~90 75~246 470~1157

Method 2 080 801~885 114~245 711~1134

Method 3 080 843 179 1122

3

(2205 J)

Method 1 148 50~100 06~380 21~1134

Method 2 148 758~838 249~368 843~1125

Method 3 148 798 308 1044

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1213 ms

Table V presents the characterization results of the impact-induced damage case Here the

measured 119907119907119894119894 and the assumed variation are same as the artificial damage case It is observed

that as the number of impact is increased 119889119889 is increased while 119907119907119889119889 is reduced Thus it is

validated that the proposed damage characterization approach estimates the increment of the

damage well

54 Temperature effect test

541 Teflon inserted damage

11 115 12 125 13-10

-5

0

5

10

Time (ms)

Vol

tage

(mV

)

0 oC20 oC50 oC

(a) Pitch-catch signal

11 115 12 125 13 135

-1

0

1

2

3

Time(ms)

Volta

ge (m

V)

0 oC20 oC50 oC

(b) Pulse-echo signal

Fig 15 Signals obtained from the Teflon inserted damage specimen under temperature varia-tion Time delay and amplitude variation are occurred

Fig 15 shows the obtained signals from the Teflon inserted damage specimen subjected to

- 27 -

temperature variation Time delay and amplitude variations are occurred there is no addition-

al damage is introduced The analyzed results of the obtained signals are reported in Table VI

The measured 119907119907119894119894 is 1247 ms at 0 degC 1213 ms at 20 degC and 1168 ms at 50 degC respectively

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms (1152~1273 ms)

The estimated results are also changed under temperature variations the fluctuation of esti-

mated range and exact value is small

TABLE VI Temperature effect test for the Teflon inserted damage specimen

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

0 degC

Exact 635 25

Method 1 113 60~70 79~268 352~1291

Method 2 113 560~619 233~344 1034~1172

Method 3 113 606 257 1141

20 degC

Exact 635 25

Method 1 123 60~75 73~338 298~1220

Method 2 123 619~662 229~310 934~1188

Method 3 123 631 284 1161

50 degC

Exact 635 25

Method 1 137 60~75 72~328 263~1228

Method 2 137 622~687 183~299 666~1086

Method 3 137 642 264 960

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1247 ms at 0 degC 1213 ms at 20 degC and 1168 ms at 50 degC

- 28 -

542 Impact-induced damage

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Vol

tage

(mV

)

Transmitted WaveReflected Wave

No reflected wave

(a) 0 degC

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Transmitted WaveReflected Wave

No reflected wave

(b) 50 degC

Fig 16 False alarm test under temperature variations It is shown that proposed damage char-acterization algorithm is robust under temperature variations

The false alarm test in pristine condition of the impact-induced damage specimen under

temperature variation was investigated (Fig 16) It is validated that the proposed damage di-

agnosis algorithm is robust under temperature variations Fig 17 shows the obtained signals

from the impact-induced damage (impacted 3 times) specimen under temperature variation

The analyzed results of the obtained signals are reported in Table VII The measured 119907119907119894119894 in

each temperature and the assumed variation are same as the artificial damage case

11 115 12 125 13-5

0

5

Time (ms)

Volta

ge (m

V)

0 oC20 oC50 oC

(a) Pitch-catch signal

11 115 12 125 13 135

-1

0

1

2

3

Time(ms)

Vol

tage

(mV

)

0 oC20 oC50 oC

(b) Pulse-echo signal

Fig 17 Signals obtained from impact-induced damage specimen (Impact 3) under tempera-ture variation

- 29 -

TABLE VII Temperature effect test for the impact induced damage specimen (Impact 3 case)

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

0 degC

Method 1 135 75~90 72~310 267~1148

Method 2 135 705~779 264~381 978~1126

Method 3 135 763 289 1071

20 degC

Method 1 148 50~100 06~380 21~1134

Method 2 148 758~838 249~368 843~1125

Method 3 148 798 308 1044

50 degC

Method 1 153 75~100 53~415 174~1227

Method 2 153 804~889 214~336 702~1103

Method 3 153 815 320 1050

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1247 ms at 0 degC 1213 ms at 20 degC and 1168 ms at 50 degC

- 30 -

CHAPTER 6 CONCLUDING REMARKS

61 Executive summary

In this study the reference-free damage detection localization and quantification tech-

nique in a multi layered composite plate is proposed Additional reflected A0 Lamb wave

mode from the damage due to multiple reflections in damage area is used as a feature for

characterizing the damage First damage is characterized in terms of a damage index (m-

value) defined as the ratio of the physical length of the damage to the wavelength of the A0

Lamb wave mode Then the range of possible damage location length and severity are esti-

mated with constraints This procedure can be accomplished without prior baseline data To

realize the proposed technique the matching pursuit technique is modified for reduction of

the error by considering the dispersive characteristic of Lamb wave The matching pursuit

method estimates the arrival time of each wave packets accurately even two wave packets are

overlapped The numerical simulations and experimental tests including Teflon tape inserted

and impact-induced damage are investigated for the feasibility of the proposed technique

Moreover the effectiveness of the proposed technique under changing temperature is exper-

imentally verified

- 31 -

62 Suggestion for further study

The limitations of this study and the possible improvement and further studies are listed as

follow

1 Measurement of impact-induced damage length using conventional NDT technique

Additional experimental verification for impact-induced damage characterization is

needed by measuring the real length of damage using conventional NDT technique

such as ultrasonic scan or infrared camera

2 Limitation of both side collocated PZT installation In this study excitation PZTs are

installed both side and collocated for decomposing A0 Lamb wave mode However it

can produce decomposition error due to imperfection of PZT location and their installa-

tion Moreover in many cases assess to both surface of structures is not allowed such

as aircraft applications To improve this limitation a reference-free damage characteri-

zation using single side PZTs will be considered

3 Complex structure application In this study damage is characterized in simple plate-

like structure When the monitored structure becomes complex the response signals al-

so becomes complex due to its geometry and boundary Thus the investigations for

complex structure application such as stiffener spar and curvature are needed

- 32 -

REFERENCES

[1] Cawley P and Alleyne D (1996) ldquoThe use of Lamb waves for the long range inspection of large structuresrdquo Ultrasonics 34 pp287-290

[2] Ghosh T Kundu T and Karpur P (1998) ldquoEfficient use of Lamb waves for detecting defects in large platesrdquo Ultrasonics 36 pp791-801

[3] Guirguitui V (2003) ldquoLamb wave generation with piezoelectric wafer active sensors for structural health monitoringrdquo SPIE Conference on Smart Structures and Materials San Diego CA 5391 pp111-422

[4] Ragvahan A and Cesnik CE S (2004) ldquoModeling of piezoelectric-based Lamb wave generation and sensing for structural health monitoringrdquo SPIE Conference on Smart Struc-tures and Materials San Diego CA 5391 pp419-430

[5] Rose JL Pilarski A and Dirit JJ (1993) ldquoAn approach to guided wave mode selection for inspection of laminated platerdquo Journal of Reinforced Plastics and Composites 12(5) pp536-544

[6] Guo N and Cawley P (1993) ldquoThe interaction of Lamb waves with delaminations in composite laminatesrdquo Journal of Acoustical Society of America 94(4) pp 2240-2246

[7] Valdez SHD and Soutis C (2002) ldquoReal-time nondestructive evaluation of fiber com-posite laminates using low-frequency Lamb wavesrdquo Journal of Acoustical Society of Amer-ica 111(5) pp 2026-2033

[8] Yang M and Qiao P (2005) ldquoModeling and experimental detection of damage in virous materials using the pulse-echo method and piezoelectric sensoractuatorsrdquo Smart Materials and Structures 14 pp1083-1100

[9] Lestari W and Qiao P (2005) ldquoApplication of wave propagation analysis for damage identification in composite laminated beamsrdquo Journal of Composite Materials 39(22) pp1967-1984

[10] Kessler S S Spearing S M and Soutis C (2002) ldquoDamage detection in composite ma-terials using Lamb wave methodsrdquo Smart Materials and Structures 11 pp269-278

[11] Sohn H Park G Wail J R Limback N P and Farrar C R (2004) ldquoWavelet-based active sensing for delamination detection in composite structuresrdquo Smart Materials and Structures 13 pp 153-160

[12] Petculescu G Krishnaswamy S and Achenback J D (2008) ldquoGroup delay measure-ments using modally selective Lamb wave transducers for detection and sizing of delamina-tions in compositesrdquo Smart Materials and Structures 17

[13] Sohn H Park H W Law K H and Farrar C R (2007) ldquoDamage detection in compo-

- 33 -

site plates by using an enhanced time reversal methodrdquo Journal of Aerospace Engineering (ASCE) pp 141-151

[14] Sohn H (2007) ldquoEffects of environmental and operational variability on structural health monitoringrdquo Philosophical Transactions of the Royal Society Ardquo 365 pp539-560

[15] Rytter A (1993) ldquoVibration based inspection of civil engineering structuresrdquo PhD The-sis Department of Building Technology and Structural Engineering University of Aalborg Aalborg Denmark

[16] Hayashi T and Kawashima K (2002) ldquoMultiple reflections of Lamb waves at a delami-nationrdquo Ultrasonics 40 pp193-197

[17] Wang C H and Rose L R F (2003) ldquoWave reflection and transmission in beams con-taining delamination and inhomogeneityrdquo Journal of Sound and Vibration 264 pp 851-872

[18] Ramadas C Balasubramaniam K Joshi M and Krishnamurthy C V (2009) ldquoInterac-tion of the primary anti-symmetric Lamb mode (Ao) with symmetric delaminations numer-ical and experimental studiesrdquo Smart Materials and Structures 18

[19] Ramadas C Balasubramaniam K Joshi M and Krishnamurthy C V (2010) ldquoInterac-tion of guided Lamb waves with an asymmetrically located delamination in a laminated composite platerdquo Smart Materials and Structures 19

[20] Wilcox P Lowe M and Cawely P (2001) ldquoThe effect of dispersion on long-range in-spection using ultrasonic guided wavesrdquo NDTampE International 34 pp1-9

[21] Mallat S G and Zhang Z (1993) ldquoMatching pursuits with time-frequency dictionariesrdquo IEEE Tansactions on Signal Processing 41(12) pp3397-3415

[22] Zhang G Zhang S and Wang Y (2000) ldquoApplication of adaptive time-frequency de-composition in ultrasonic NDE of highly scattering materialsrdquo Ultrasonics 38 pp 961-964

[23] Hong J C Sun K H and Kim Y Y (2005) ldquoThe matching pursuit approach based on the modulated Gaussian pulse for efficient guided-wave damage inspectionrdquo Smart Mate-rials and Structures 14 pp548-560

[24] Herakovich C T (1997) Mechanics of Fibrous Composites Wiley 480 pages

[25] Reddy J N (1985) ldquoA review of the literature on finite-element modeling of laminated composite platesrdquo The Shock and Vibration Digest 17(4) pp 3-8

[26] Reddy J N (2003) Mechanics of Laminated Composite Plates and Shells Theory and Analysis CRC Press 856 pages

[27] Palacz M Krawczuk M and Ostachowicz W (2005) ldquoThe spectral finite element mod-el for analysis of flexure-shear coupled wave propagation Part 1 laminated multilayer

- 34 -

composite beamrdquo Composite Structures 68 pp37-44

[28] Zhang Y C and Harding J (1990) ldquoA numerical micromechanics analysis of the me-chanical properties of a plain weave compositerdquo Computer amp Structures 36(5) pp 839-844

[29] Nemat-Nasser S and Hori M (1999) Micromechanics Overall Properties of Heteroge-neous Materials Elsevier 810 pages

[30] Scida D Aboura Z Benzeggagh M L and Bocherens E (1999) ldquoA micromechanics model for 3D elasticity and failure of woven-fibre composite materialrdquo Composites Science and Technology 59 pp 505-517

[31] Thomas Y and Wu X H (1997) ldquoA multiscale finite element method for elliptic prob-lems in composite materials and porous mediardquo Journal of Computational Physics 134 pp 169-189

[32] Lee K H Moorthy S and Ghosh S (1999) ldquoMultiple scale computational model for damage in composite materialsrdquo Computer Methods in Applied Mechanics and Engineer-ing 172 pp 175-201

[33] Namilae S and Chandra N (2005) ldquoMultiscale model to study the effect of interfaces in carbon nanotube-based compositesrdquo Journal of Engineering Materials and Technology (ASME) 127 pp222-232

[34] Adams D E (2007) Health monitoring of Structural Materials and Components Wiley 460 pages

[35] Gopalakrishnan S (2009) ldquoModeling Aspects in Finite Elementrdquo Encyclopedia of Struc-tural Health Monitoring 2 pp791-809

[36] Palacz M Krawczuk M and Ostachowicz W (2005) ldquoThe spectral finite element mod-el for analysis of flexure-shear coupled wave propagation Part 2 Deminated multilayer composite beamrdquo Composite Structures 68 pp45-51

[37] Baik J M and Thopson R B (1984) ldquoUltrasonic scattering from imperfect interfaces A quasi-static modelrdquo Journal of Nondestructive Evaluation 4(3) pp177-196

[38] Luo H and Hanagud S (2000) ldquoDynamics of delamincated beamsrdquo International Jour-nal of Solids and Structures 37 pp1501-1519

[39] Kim S B and Sohn H (2007) ldquoInstantaneous reference-free crack detection based on polarization characteristics of piezoelectric materialsrdquo Smart Materials and Structures 16 pp2375-2387

[40] An Y K and Sohn H (2010) ldquoInstantaneous crack detection under varying temperature and static loading conditionsrdquo The Special Issue of Structural Control and Health Monitor-ing 17(7) pp 730-741

- 35 -

[41] Lee S J and Sohn H (2006) ldquoActive self-sensing scheme development for structural health monitoringrdquo Smart Materials and Structures 15(6) pp 1734-1746

- 36 -

요 약 문

복합재료 손상의 무기저 진단 및 정량화

복합재료는 재료의 경량성 내구성 고강도의 특성으로 인하여 항공기에서부터

교량에까지 널리 사용되고 있다 하지만 외부의 충격으로 인한 복합재료 내부의

손상은 재료의 강도를 감소시켜 구조물의 안정성을 저하시키는 요인이 된다 하지만

복합재료의 손상은 재료 내부에서 발생하여 육안으로는 식별하기도 힘들며 표면에서

보이는 것만으로 내부의 손상 정도를 판단하기는 불가능하다는 단점을 가지고 있다

본 연구에서는 손상 내부에서 반사로 인해 발생하는 파를 이용하여 기저자료를

사용하지 않고 특정 시점에서 얻은 자료만으로 복합재료의 손상을 진단하고 정량화

할 수 있는 기법을 제안한다 제안된 기법의 적용성을 검증하기 위하여 유한요소법을

이용한 수치해석을 수행하였고 Teflon tape 이 삽입된 시편과 실제 충격하중으로

인한 손상을 입은 시편을 이용한 실험을 진행하였다 수치해석 및 실험 결과 제안된

기법이 복합재료의 손상을 성공적으로 진단 및 정량화할 수 있음을 확인하였다 또한

다양한 외부 온도 환경에서 제안된 기법의 환경영향 실험을 수행하였고 다양한

환경에서 손상의 진단 및 정량화가 성공적으로 이루어짐을 검증하였다

핵심어 무기저 손상진단 및 정량화 압전센서 복합재료 유한요소법 Matching pursuit

method

- 37 -

감 사 의 글

기대와 두려움을 안고 시작한 석사과정이 어느덧 2 년 이라는 세월 속에서 추억으로 자리잡아

가고 있습니다 학부생 기간을 포함한 지난 3 년 동안 SHM 이라는 새로운 분야를 접하고 연구를

수행 할 수 있도록 도와주신 고마운 분들이 있어 그 감사를 표하고자 합니다

우선 저의 많은 부족한 점을 일깨워 주시고 연구에 대한 뜻을 더욱 확고히 세울 수 있게 해

주신 지도교수님이신 손 훈 교수님께 깊은 감사의 뜻을 전합니다 짧은 기간이었지만 교수님의

도움으로 많은 것을 배우고 느끼고 깨달을 수 있었습니다 또한 본 연구를 수행할 수 있도록 지

원해주신 국토해양부 (U-City 석 박사과정 지원사업)에 깊은 감사 드립니다

때로 힘들고 지칠 때 항상 저를 믿고 응원해 주시는 가족이 있었기에 지금의 글을 쓸 수 있었

습니다 또한 지금 연구실에서 함께 생활하고 있는 윤규형 민구형 철민이형 현석이형 진열이형

호민이형 병진이 선혜누나와 성대에 계신 창길이형 미국에 계신 의재형과 Debaditiya 에게도 감

사의 말씀을 전합니다 그리고 모두 나열할 수는 없지만 카이스트 생활의 활력소가 되어준 카이

스트 건설 및 환경공학과 선후배들 직원 분들과 SOS 멤버들에게도 감사를 표합니다

앞서 거론한 많은 이들의 도움이 없었다면 저 또한 지금 이 자리에 있을 수 없었을 것입니다

많이 부족하기에 더 노력하고 배우고자 했던 초심을 다시 한번 생각하며 더 큰 결실을 맺기 위해

앞으로 나아가겠습니다 도움을 주신 모든 분들께 이 감사의 글을 바칩니다

2011년 6월 카이스트에서 임형진 올림

- 38 -

CURRICULUM VITAE

Hyung Jin Lim

Research Assistant

The Civil and Environmental Engineering Department

Korea Advanced Institute of Science and Technology (KAIST)

291 Daehak-Ro Guseong-Dong Yuseong-Gu

Daejeon Republic of Korea 305-701

Phone (82)+42-350-3665 Fax (82)+42-350-3610

Email limhj87gmailcom

RESEARCH INTERESTS

Damage Detection Localization and Quantification on Composite and Metallic Structures Pattern Recognition Finite Element Modeling (FEM)

EDUCATION

2009 ndash 2011 MS Civil and Environmental Eng Korea Advanced Institute of Science and Technology (KAIST) Korea Rep

2006 ndash 2009 BS Civil and Environmental Eng Korea Advanced Institute of Science and Technology (KAIST) Korea Rep

JOURNAL PUBLICATIONS The corresponding author is underlined 1 Hyung Jin Lim Hoon Sohn ldquoReference-free damage detection localization and quantification

using in compositerdquo in preparation for Journal of the Acoustical Society of America 2011

2 Chul Min Yeum Hoon Sohn Jeong Bum Ihn Hyung Jin Lim ldquoInstantaneous delamination detection in a composite using a dual piezoelectric transducer networkrdquo Submitted to Compo-sites Science and Technology 2011

3 Yun-Kyu An Hyung Jin Lim Min Koo Kim Jin Yeol Yang Hoon Sohn Chang Guen Lee ldquoInvestigation of applicability of local reference-free damage detection techniques to in-situ

- 39 -

bridgesrdquo Submitted to ASCE Journal of Structural Engineering 2011

4 Hyung Jin Lim Min Koo Kim Hoon Sohn Chan Yik Park ldquoImpedance based damage detec-tion under varying temperature and loading conditionsrdquo Submitted to NDTampE International 2011

5 Hyung Jin Lim Hoon Sohn ldquoQuantitative estimation of transmitted and reflected lamb waves at discontinuityrdquo Journal of the Korean Society for Nondestructive Testing Vol30 No 4 pp1-8 2010

CONFERENCE PROCEEDINGS

The corresponding author is underlined 1 Hyung Jin Lim and Hoon Sohn ldquoReference-free damage detection localization and quantifi-

cation in compositesrdquo ASCE Engineering Mechanics Institute Conference Boston June 2-4 2011

2 Hyung Jin Lim and Hoon Sohn ldquoReference-free delamination quantification within multi-layer composite materialsrdquo KSNT Spring Conference May 26-27 Gyeong-Ju Korea 2011

3 Min Koo Kim Hyun Jin Lim Hoon Sohn Chan Yik Park ldquoImpedance-based bolt loosening detection under varying temperature and loadingrdquo Asian Pacific Workshop on Structural Health Monitoring Tokyo Japan November 30-December 2 2010

4 Hyung Jin Lim Chul Min Yeum and Hoon Sohn ldquoModeling of Impact-induced Delamination in a Multilayer Composite Platerdquo The 23rd KKCNN Symposium on Civil Engineering Taipei (Taiwan) November 13-15 2010

5 Yun-Kyu An Hyung Jin Lim and Hoon Sohn ldquoTemperature independent Delamination Detec-tion using data normalizationrdquo The 3rd International Conference on Multi-Functional Materi-als and Structures Chonbuk National Univ Jeonju Korea September 14-18 2010

6 Hoon Sohn Yun-Kyu An Chang Gil Lee Min Koo Kim Hyung Jin Lim ldquoApplication of online NDT techniques to decommissioned bridge testingrdquo the 5th European Workshop on Structural Health Monitoring Sorrento Italy June 29-July 02 2010

7 Min Koo Kim Hyung Jin Lim Hoon Sohn Chan Yik Park ldquoImpedance based damage diag-nosis of a complex composite aircraft wing under changing loading and temperature condi-tionsrdquo the 5th European Workshop on Structural Health Monitoring Sorrento Italy June 29-July 02 2010

8 Yun-Kyu An Min Koo Kim Hyung Jin Lim and Hoon Sohn ldquoValidations of reference-free crack detection techniques through a decommissioned bridge testrdquo 2010 COSEIK Annual Con-ference Jeju Korea April 8-9 2010

- 40 -

PARTICIPANT PROJECTS

1 Development of Smart Steel Members for Online Steel Structure Monitoring POSCO Corporation (Funded 41000000 KRW (41000 USD) for 010511 to 123111)

2 Development of Enabling Technologies for Structural Health Monitoring Korean Agency for Defense Development (Funded 240000000 KRW (240000 USD) out of total 3200000000 KRW (3200000 USD) for 060108-053111)

3 Development of On-board SHM Technologies for Composite Air Vehicles The Boeing Company (Funded 400000 USD (400000000 KRW) for 080108 to 073111)

EDUCATIONAL EXPERIENCES

Fall 2009

Teaching Assistant CE207 ldquoElementary Structural Engineering and Laboratoryrdquo KAIST 3 unit undergraduate course

EXPERIMENTAL EXPERIENTCES

2010 Monitoring of Real Aircraft Composite Wing Segment Korea Agency for Defense De-velopment (ADD) Daejeon Korea Real time monitoring of composite air vehicles fund-ed by Korea Agency for Defense Development (ADD)

2009-2011

Field Monitoring of Real Bridge Structures Gimpo (Ramp G bridge) amp Icheon (Sam-seung bridge) Korea US-Korea collaborative research for bridge monitoring testbeds

HONORS amp AWARDS

Scholarships for outstanding students granted KAIST (2007-2009) The three prominent students in the civil and environmental engineering department of KAIST were selected as recipients This scholarship provided 2000$ per year for three years

- 41 -

Page 28: 복합재료 손상의 무기저 진단 및 정량화 - KAISTssslab.kaist.ac.kr/article/pdf/Thesis_2.pdf · 1 Declaration of Ethical Conduct in Research: I, as a graduate student

43 Numerical simulation description

TABLE I Material property of simulation model and test specimen

Tensile Strength 657 MPa Tensile Modulus 59 GPa Shear Strength 804 MPa Shear Modulus 24 Gpa Specific Gravity 16

PZT A

PZT C

PZT BDamage

3

250 125 125

500 Unit mm

Fig 6 Description of the numerical simulation model

The feasibility of the proposed damage characterization technique is verified by 2-D FEM

simulation using MSC PATRAN and NASTRAN Fig 6 presents the dimension PZTs and

damage location of the simulation model In the model 05mm x 05mm square shell element

was used for modeling of the A0 mode whose flexural wave characteristic The element size

was determined by considering the wavelength of the propagation Lamb wave modes The

material properties of the simulation model and damage cases are presented in Table I and

Table II respectively

PZT A and C act as both actuator and sensor for collecting pulse-echo signal They were

installed on the plate which two of them are both-side collocated for the A0 Lamb wave

- 17 -

mode decomposition [39-40] PZT B was only used as a sensor for collecting pitch-catch sig-

nal For simulating the excitation and sensing of PZTs the pin-force model was adopted that

an input force apples to the end nodes of excitation PZT and averaging element strains corre-

sponding sensing area A Gaussian pulse (tone-burst) 50 kHz central frequency was excited

as an input signal

TABLE II Damage cases for numerical simulation

119949119949120783120783 (mm) 119941119941 (mm) G value reduction ()

Case 1 650 200 90 darr

Case 2 675 150 90 darr

Case 3 675 150 93 darr

Case 4 900 150 90 darr

- 18 -

44 Numerical simulation results

015 02 025 03 035

-2

0

2

Time (ms)

Am

plitu

de

IntactDamage

Reflected wave from damage

(a) Pitch-catch signal

015 02 025 03 035

-2

0

2

Time (ms)

Am

plitu

de

IntactDamage

Reflected waves from damage

(b) Pulse-echo signal Fig 7 Comparison of A0 Lamb wave mode signals from numerical simulation between intact and damage cases Time delay attenuation and multiple reflections due to damage simulated successfully using the damage modeling method described in the previous section

Fig 7 shows that comparison of A0 Lamb wave mode signals between intact and one of

damage cases in the simulation The features of A0 Lamb wave mode passing through dam-

age area described in section 2 are successfully simulated Thus the equivalent element

method is proper for modeling damage in composites In the damage cases the additional

waves reflected from damage is observed in both pitch-catch and pulse echo signal

Table III presents the result of the proposed damage characterization algorithm using the

numerical simulation cases shown in Table II In lsquoMethod 2rsquo 119907119907119894119894 variation is assumed to plusmn

5 of the normal value (1190~1315ms) The normal value of 119907119907119894119894 for lsquoMethod 3rsquo is meas-

ured as 1253 ms In each case the damage is characterized using the proposed technique

Note that as the information of 119907119907119894119894 is specified the estimated value is closer to the exact

value

- 19 -

TABLE III Damage characterization for the numerical simulation models

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

Case 1

Exact 650 200

Method 1 248 50~70 79~414 160~834

Method 2 248 603~667 135~241 271~486

Method 3 248 635 188 379

Case 2

Exact 675 150

Method 1 193 60~75 01~251 37~651

Method 2 193 618~682 114~222 295~575

Method 3 193 65 168 435

Case 3

Exact 675 150

Method 1 233 60~75 66~302 140~649

Method 2 233 668~738 85~196 182~419

Method 3 233 703 140 301

Case 4

Exact 900 150

Method 1 204 75~100 48~349 118~856

Method 2 204 860~951 107~216 264~530

Method 3 204 906 162 397

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1253 ms 1190~1315 ms Measured 119907119907119894119894 1253 ms

- 20 -

CHAPTER 5 EXPERIMENTAL VERIFICATION

51 Overview

In this section experimental study is investigated so that the proposed damage characteri-

zation technique Overall experimental setup and test specimen are shown in Fig 8 and 9

The experimental setup consists of test specimens data acquisition (DAQ) system and self-

sensing circuit Self-sensing circuit was installed for pulse-echo measurement using PZTs

[40-41]

PZT A(PZT C) PZT B

500

500

140

75

250

Teflon Inserted Damage (d =25)

Unit mm

Thickness = 3

PZT A(PZT C) PZT B

500

500

150

90

250

Impact Location

Unit mm

Thickness = 3

(a) Teflon inserted damage specimen

(b) Impact-induced damage specimen

Fig 8 Test specimens for the experimental verification In the Teflon inserted damage Teflon tapes with 25mm diameter were inserted at the center of the specimen The impact-induced damage was introduced to the specimen using the impact test equipment

- 21 -

511 Specimen

Two carbon composite plates with a dimension of 500times500times3mm3 were manufactured by

Nexcomes Inc by stacking 12 layers of CF3327 3K prepreg (woven fabrics type) manufac-

tured by Hankuk Fiber Inc The material properties of the specimen are presented in Table I

One of the specimens Teflon tapes with 25mm diameter was inserted in every 3 layers at the

center of the specimen to mimic the damage The other specimen is utilized for the character-

ization of real impact-induced damage

512 Data acquisition (DAQ) system and PZT

The data acquisition (DAQ) system consists of an arbitrary waveform generator (AWG) a

high speed signal digitizer (DIG) and two multiplexers (MUX) Using the 16-bits AWG

tone-burst input signal with a plusmn10 peak-to-peak voltage were generated The response signal

was measured by the 14-bit DIG For the signal-to-noise ratio reduction both pitch-catch and

pulse-echo signals were measured 150 times and averaged in the time domain 3 identical

circular PZTs of 10mm diameter manufactured by APC International Ltd were installed on

the specimen A Gaussian pulse (tone-burst) 50 kHz central frequency was excited as an in-

put signal The excitation frequency was chosen in order to avoid overlapping of the S0 and

A0 Lamb wave modes and generating of the higher modes (S1 A1hellip)

- 22 -

PZT Actuator

PZT Sensor

AWG

DIG

DAQ System Specimen

SensingActuatingMUX

1

23

4

1 Sending signal to MUX

4 Sending signal to DIG

Self-SensingCircuit

18 Unit mm

10

(a)

(b)

Fig 9 Overall experimental setup (a) Data acquisition (DAQ) system and (b) Installed circu-lar PZT transducer on the specimen with its dimension

513 Impact test

The impact-induced damage was introduced to the specimen using the impact test equip-

ment presented in Fig 10 A 5kg mass with 10mm tip was dropped from 015m height onto

the same location of the specimen The signals were obtained at each number of impacts

Since the reflected wave is not detected at the first impact case the data obtained from the

second and third impact were utilized for the analysis

10

Unit mm (a)

(b)

Fig 10 (a) Impact test equipment for introducing impact-induced damage and (b) Tip of the impact tester with 10mm diameter

- 23 -

514 Temperature effect test

In reality structures are subjected to changing environmental conditions such as tempera-

ture that affect measured signal and cause false-alarm Thus the robustness of the proposed

damage characterization technique under temperature variation in investigated (Fig 11) The

specimens were placed inside a temperature chamber and a thermocouple was installed on

the specimen to measure temperature of surface of the specimen The signal was obtained

under three temperature cases (0 20 and 50degC) and the humidity was kept at 30

Temperature Chamber

Test Specimen

Fig 11 Test specimen in temperature chamber

- 24 -

52 Teflon inserted damage

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Reflected wave from damage

(a)

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Transmitted WaveReflected Wave

Detected reflected wave

(b)

Fig 12 A0 Lamb wave mode obtained using pitch-catch method from the Teflon inserted damage specimen (a) The reflected wave from damage is shown (b) The Teflon inserted damage is detected by capturing the reflected wave

Fig 12 shows the obtained pitch-catch signal from the Teflon inserted damage specimen

and damage diagnosis result The Teflon inserted damage is detected successfully by captur-

ing the reflected wave Table IV presents the result of the Teflon inserted damage characteri-

zation 119907119907119894119894 is measured as 1213ms and the variation is assumed to plusmn 5 of the normal val-

ue (1152~1273ms) The results indicate that the proposed damage characterization approach

works well in real application

TABLE IV Damage characterization for the Teflon inserted damage specimen

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

Exact 635 25

Method 1 123 60~75 73~338 298~1220

Method 2 123 619~662 229~310 934~1188

Method 3 123 631 284 1161

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1213 ms

- 25 -

53 Impact-induced damage

11 115 12 125 13-4

-2

0

2

4

Time (ms)

Vol

tage

(mV

)

IntactImpact2Impact3

Reflected wave from damage

(a) Pitch-catch signal

11 115 12 125 13 135

-05

0

05

1

Time (ms)

Volta

ge (m

V)

IntactImpact2Impact3

Boundary reflection

Reflected wave from damage

(b) Pulse-echo signal

Fig 13 The signal obtained from the test specimen subjected to impact The impact was in-troduced at the same location of the specimen and the signal was obtained after each impacts As the number of impact is increased the transmitted A0 Lamb wave mode is delayed and attenuated

The obtained signals from impact-induced damage specimen are presented in Fig 13 It is

observed that as the number of impact is increased the transmitted A0 Lamb wave mode is

more delayed because the wave velocity in the damage area is reduced Fig 14 shows the

false-alarm test result whether the proposed damage detection algorithm works even in pris-

tine condition of the test specimen It is shown that no false alarm is occurred because no ad-

ditional reflected wave is captured

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Transmitted Wave

(a) Transmitted wave

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Reflected Wave

No reflected wave

(b) Reflected wave

Fig 14 False alarm test for the proposed damage detection algorithm in pristine condition of the specimen Reflected wave from damage is not captured false alarm is not occurred

- 26 -

TABLE V Damage characterization for the impact-induced damage specimen

of Impacts m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

2

(1470 J)

Method 1 080 65~90 75~246 470~1157

Method 2 080 801~885 114~245 711~1134

Method 3 080 843 179 1122

3

(2205 J)

Method 1 148 50~100 06~380 21~1134

Method 2 148 758~838 249~368 843~1125

Method 3 148 798 308 1044

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1213 ms

Table V presents the characterization results of the impact-induced damage case Here the

measured 119907119907119894119894 and the assumed variation are same as the artificial damage case It is observed

that as the number of impact is increased 119889119889 is increased while 119907119907119889119889 is reduced Thus it is

validated that the proposed damage characterization approach estimates the increment of the

damage well

54 Temperature effect test

541 Teflon inserted damage

11 115 12 125 13-10

-5

0

5

10

Time (ms)

Vol

tage

(mV

)

0 oC20 oC50 oC

(a) Pitch-catch signal

11 115 12 125 13 135

-1

0

1

2

3

Time(ms)

Volta

ge (m

V)

0 oC20 oC50 oC

(b) Pulse-echo signal

Fig 15 Signals obtained from the Teflon inserted damage specimen under temperature varia-tion Time delay and amplitude variation are occurred

Fig 15 shows the obtained signals from the Teflon inserted damage specimen subjected to

- 27 -

temperature variation Time delay and amplitude variations are occurred there is no addition-

al damage is introduced The analyzed results of the obtained signals are reported in Table VI

The measured 119907119907119894119894 is 1247 ms at 0 degC 1213 ms at 20 degC and 1168 ms at 50 degC respectively

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms (1152~1273 ms)

The estimated results are also changed under temperature variations the fluctuation of esti-

mated range and exact value is small

TABLE VI Temperature effect test for the Teflon inserted damage specimen

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

0 degC

Exact 635 25

Method 1 113 60~70 79~268 352~1291

Method 2 113 560~619 233~344 1034~1172

Method 3 113 606 257 1141

20 degC

Exact 635 25

Method 1 123 60~75 73~338 298~1220

Method 2 123 619~662 229~310 934~1188

Method 3 123 631 284 1161

50 degC

Exact 635 25

Method 1 137 60~75 72~328 263~1228

Method 2 137 622~687 183~299 666~1086

Method 3 137 642 264 960

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1247 ms at 0 degC 1213 ms at 20 degC and 1168 ms at 50 degC

- 28 -

542 Impact-induced damage

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Vol

tage

(mV

)

Transmitted WaveReflected Wave

No reflected wave

(a) 0 degC

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Transmitted WaveReflected Wave

No reflected wave

(b) 50 degC

Fig 16 False alarm test under temperature variations It is shown that proposed damage char-acterization algorithm is robust under temperature variations

The false alarm test in pristine condition of the impact-induced damage specimen under

temperature variation was investigated (Fig 16) It is validated that the proposed damage di-

agnosis algorithm is robust under temperature variations Fig 17 shows the obtained signals

from the impact-induced damage (impacted 3 times) specimen under temperature variation

The analyzed results of the obtained signals are reported in Table VII The measured 119907119907119894119894 in

each temperature and the assumed variation are same as the artificial damage case

11 115 12 125 13-5

0

5

Time (ms)

Volta

ge (m

V)

0 oC20 oC50 oC

(a) Pitch-catch signal

11 115 12 125 13 135

-1

0

1

2

3

Time(ms)

Vol

tage

(mV

)

0 oC20 oC50 oC

(b) Pulse-echo signal

Fig 17 Signals obtained from impact-induced damage specimen (Impact 3) under tempera-ture variation

- 29 -

TABLE VII Temperature effect test for the impact induced damage specimen (Impact 3 case)

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

0 degC

Method 1 135 75~90 72~310 267~1148

Method 2 135 705~779 264~381 978~1126

Method 3 135 763 289 1071

20 degC

Method 1 148 50~100 06~380 21~1134

Method 2 148 758~838 249~368 843~1125

Method 3 148 798 308 1044

50 degC

Method 1 153 75~100 53~415 174~1227

Method 2 153 804~889 214~336 702~1103

Method 3 153 815 320 1050

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1247 ms at 0 degC 1213 ms at 20 degC and 1168 ms at 50 degC

- 30 -

CHAPTER 6 CONCLUDING REMARKS

61 Executive summary

In this study the reference-free damage detection localization and quantification tech-

nique in a multi layered composite plate is proposed Additional reflected A0 Lamb wave

mode from the damage due to multiple reflections in damage area is used as a feature for

characterizing the damage First damage is characterized in terms of a damage index (m-

value) defined as the ratio of the physical length of the damage to the wavelength of the A0

Lamb wave mode Then the range of possible damage location length and severity are esti-

mated with constraints This procedure can be accomplished without prior baseline data To

realize the proposed technique the matching pursuit technique is modified for reduction of

the error by considering the dispersive characteristic of Lamb wave The matching pursuit

method estimates the arrival time of each wave packets accurately even two wave packets are

overlapped The numerical simulations and experimental tests including Teflon tape inserted

and impact-induced damage are investigated for the feasibility of the proposed technique

Moreover the effectiveness of the proposed technique under changing temperature is exper-

imentally verified

- 31 -

62 Suggestion for further study

The limitations of this study and the possible improvement and further studies are listed as

follow

1 Measurement of impact-induced damage length using conventional NDT technique

Additional experimental verification for impact-induced damage characterization is

needed by measuring the real length of damage using conventional NDT technique

such as ultrasonic scan or infrared camera

2 Limitation of both side collocated PZT installation In this study excitation PZTs are

installed both side and collocated for decomposing A0 Lamb wave mode However it

can produce decomposition error due to imperfection of PZT location and their installa-

tion Moreover in many cases assess to both surface of structures is not allowed such

as aircraft applications To improve this limitation a reference-free damage characteri-

zation using single side PZTs will be considered

3 Complex structure application In this study damage is characterized in simple plate-

like structure When the monitored structure becomes complex the response signals al-

so becomes complex due to its geometry and boundary Thus the investigations for

complex structure application such as stiffener spar and curvature are needed

- 32 -

REFERENCES

[1] Cawley P and Alleyne D (1996) ldquoThe use of Lamb waves for the long range inspection of large structuresrdquo Ultrasonics 34 pp287-290

[2] Ghosh T Kundu T and Karpur P (1998) ldquoEfficient use of Lamb waves for detecting defects in large platesrdquo Ultrasonics 36 pp791-801

[3] Guirguitui V (2003) ldquoLamb wave generation with piezoelectric wafer active sensors for structural health monitoringrdquo SPIE Conference on Smart Structures and Materials San Diego CA 5391 pp111-422

[4] Ragvahan A and Cesnik CE S (2004) ldquoModeling of piezoelectric-based Lamb wave generation and sensing for structural health monitoringrdquo SPIE Conference on Smart Struc-tures and Materials San Diego CA 5391 pp419-430

[5] Rose JL Pilarski A and Dirit JJ (1993) ldquoAn approach to guided wave mode selection for inspection of laminated platerdquo Journal of Reinforced Plastics and Composites 12(5) pp536-544

[6] Guo N and Cawley P (1993) ldquoThe interaction of Lamb waves with delaminations in composite laminatesrdquo Journal of Acoustical Society of America 94(4) pp 2240-2246

[7] Valdez SHD and Soutis C (2002) ldquoReal-time nondestructive evaluation of fiber com-posite laminates using low-frequency Lamb wavesrdquo Journal of Acoustical Society of Amer-ica 111(5) pp 2026-2033

[8] Yang M and Qiao P (2005) ldquoModeling and experimental detection of damage in virous materials using the pulse-echo method and piezoelectric sensoractuatorsrdquo Smart Materials and Structures 14 pp1083-1100

[9] Lestari W and Qiao P (2005) ldquoApplication of wave propagation analysis for damage identification in composite laminated beamsrdquo Journal of Composite Materials 39(22) pp1967-1984

[10] Kessler S S Spearing S M and Soutis C (2002) ldquoDamage detection in composite ma-terials using Lamb wave methodsrdquo Smart Materials and Structures 11 pp269-278

[11] Sohn H Park G Wail J R Limback N P and Farrar C R (2004) ldquoWavelet-based active sensing for delamination detection in composite structuresrdquo Smart Materials and Structures 13 pp 153-160

[12] Petculescu G Krishnaswamy S and Achenback J D (2008) ldquoGroup delay measure-ments using modally selective Lamb wave transducers for detection and sizing of delamina-tions in compositesrdquo Smart Materials and Structures 17

[13] Sohn H Park H W Law K H and Farrar C R (2007) ldquoDamage detection in compo-

- 33 -

site plates by using an enhanced time reversal methodrdquo Journal of Aerospace Engineering (ASCE) pp 141-151

[14] Sohn H (2007) ldquoEffects of environmental and operational variability on structural health monitoringrdquo Philosophical Transactions of the Royal Society Ardquo 365 pp539-560

[15] Rytter A (1993) ldquoVibration based inspection of civil engineering structuresrdquo PhD The-sis Department of Building Technology and Structural Engineering University of Aalborg Aalborg Denmark

[16] Hayashi T and Kawashima K (2002) ldquoMultiple reflections of Lamb waves at a delami-nationrdquo Ultrasonics 40 pp193-197

[17] Wang C H and Rose L R F (2003) ldquoWave reflection and transmission in beams con-taining delamination and inhomogeneityrdquo Journal of Sound and Vibration 264 pp 851-872

[18] Ramadas C Balasubramaniam K Joshi M and Krishnamurthy C V (2009) ldquoInterac-tion of the primary anti-symmetric Lamb mode (Ao) with symmetric delaminations numer-ical and experimental studiesrdquo Smart Materials and Structures 18

[19] Ramadas C Balasubramaniam K Joshi M and Krishnamurthy C V (2010) ldquoInterac-tion of guided Lamb waves with an asymmetrically located delamination in a laminated composite platerdquo Smart Materials and Structures 19

[20] Wilcox P Lowe M and Cawely P (2001) ldquoThe effect of dispersion on long-range in-spection using ultrasonic guided wavesrdquo NDTampE International 34 pp1-9

[21] Mallat S G and Zhang Z (1993) ldquoMatching pursuits with time-frequency dictionariesrdquo IEEE Tansactions on Signal Processing 41(12) pp3397-3415

[22] Zhang G Zhang S and Wang Y (2000) ldquoApplication of adaptive time-frequency de-composition in ultrasonic NDE of highly scattering materialsrdquo Ultrasonics 38 pp 961-964

[23] Hong J C Sun K H and Kim Y Y (2005) ldquoThe matching pursuit approach based on the modulated Gaussian pulse for efficient guided-wave damage inspectionrdquo Smart Mate-rials and Structures 14 pp548-560

[24] Herakovich C T (1997) Mechanics of Fibrous Composites Wiley 480 pages

[25] Reddy J N (1985) ldquoA review of the literature on finite-element modeling of laminated composite platesrdquo The Shock and Vibration Digest 17(4) pp 3-8

[26] Reddy J N (2003) Mechanics of Laminated Composite Plates and Shells Theory and Analysis CRC Press 856 pages

[27] Palacz M Krawczuk M and Ostachowicz W (2005) ldquoThe spectral finite element mod-el for analysis of flexure-shear coupled wave propagation Part 1 laminated multilayer

- 34 -

composite beamrdquo Composite Structures 68 pp37-44

[28] Zhang Y C and Harding J (1990) ldquoA numerical micromechanics analysis of the me-chanical properties of a plain weave compositerdquo Computer amp Structures 36(5) pp 839-844

[29] Nemat-Nasser S and Hori M (1999) Micromechanics Overall Properties of Heteroge-neous Materials Elsevier 810 pages

[30] Scida D Aboura Z Benzeggagh M L and Bocherens E (1999) ldquoA micromechanics model for 3D elasticity and failure of woven-fibre composite materialrdquo Composites Science and Technology 59 pp 505-517

[31] Thomas Y and Wu X H (1997) ldquoA multiscale finite element method for elliptic prob-lems in composite materials and porous mediardquo Journal of Computational Physics 134 pp 169-189

[32] Lee K H Moorthy S and Ghosh S (1999) ldquoMultiple scale computational model for damage in composite materialsrdquo Computer Methods in Applied Mechanics and Engineer-ing 172 pp 175-201

[33] Namilae S and Chandra N (2005) ldquoMultiscale model to study the effect of interfaces in carbon nanotube-based compositesrdquo Journal of Engineering Materials and Technology (ASME) 127 pp222-232

[34] Adams D E (2007) Health monitoring of Structural Materials and Components Wiley 460 pages

[35] Gopalakrishnan S (2009) ldquoModeling Aspects in Finite Elementrdquo Encyclopedia of Struc-tural Health Monitoring 2 pp791-809

[36] Palacz M Krawczuk M and Ostachowicz W (2005) ldquoThe spectral finite element mod-el for analysis of flexure-shear coupled wave propagation Part 2 Deminated multilayer composite beamrdquo Composite Structures 68 pp45-51

[37] Baik J M and Thopson R B (1984) ldquoUltrasonic scattering from imperfect interfaces A quasi-static modelrdquo Journal of Nondestructive Evaluation 4(3) pp177-196

[38] Luo H and Hanagud S (2000) ldquoDynamics of delamincated beamsrdquo International Jour-nal of Solids and Structures 37 pp1501-1519

[39] Kim S B and Sohn H (2007) ldquoInstantaneous reference-free crack detection based on polarization characteristics of piezoelectric materialsrdquo Smart Materials and Structures 16 pp2375-2387

[40] An Y K and Sohn H (2010) ldquoInstantaneous crack detection under varying temperature and static loading conditionsrdquo The Special Issue of Structural Control and Health Monitor-ing 17(7) pp 730-741

- 35 -

[41] Lee S J and Sohn H (2006) ldquoActive self-sensing scheme development for structural health monitoringrdquo Smart Materials and Structures 15(6) pp 1734-1746

- 36 -

요 약 문

복합재료 손상의 무기저 진단 및 정량화

복합재료는 재료의 경량성 내구성 고강도의 특성으로 인하여 항공기에서부터

교량에까지 널리 사용되고 있다 하지만 외부의 충격으로 인한 복합재료 내부의

손상은 재료의 강도를 감소시켜 구조물의 안정성을 저하시키는 요인이 된다 하지만

복합재료의 손상은 재료 내부에서 발생하여 육안으로는 식별하기도 힘들며 표면에서

보이는 것만으로 내부의 손상 정도를 판단하기는 불가능하다는 단점을 가지고 있다

본 연구에서는 손상 내부에서 반사로 인해 발생하는 파를 이용하여 기저자료를

사용하지 않고 특정 시점에서 얻은 자료만으로 복합재료의 손상을 진단하고 정량화

할 수 있는 기법을 제안한다 제안된 기법의 적용성을 검증하기 위하여 유한요소법을

이용한 수치해석을 수행하였고 Teflon tape 이 삽입된 시편과 실제 충격하중으로

인한 손상을 입은 시편을 이용한 실험을 진행하였다 수치해석 및 실험 결과 제안된

기법이 복합재료의 손상을 성공적으로 진단 및 정량화할 수 있음을 확인하였다 또한

다양한 외부 온도 환경에서 제안된 기법의 환경영향 실험을 수행하였고 다양한

환경에서 손상의 진단 및 정량화가 성공적으로 이루어짐을 검증하였다

핵심어 무기저 손상진단 및 정량화 압전센서 복합재료 유한요소법 Matching pursuit

method

- 37 -

감 사 의 글

기대와 두려움을 안고 시작한 석사과정이 어느덧 2 년 이라는 세월 속에서 추억으로 자리잡아

가고 있습니다 학부생 기간을 포함한 지난 3 년 동안 SHM 이라는 새로운 분야를 접하고 연구를

수행 할 수 있도록 도와주신 고마운 분들이 있어 그 감사를 표하고자 합니다

우선 저의 많은 부족한 점을 일깨워 주시고 연구에 대한 뜻을 더욱 확고히 세울 수 있게 해

주신 지도교수님이신 손 훈 교수님께 깊은 감사의 뜻을 전합니다 짧은 기간이었지만 교수님의

도움으로 많은 것을 배우고 느끼고 깨달을 수 있었습니다 또한 본 연구를 수행할 수 있도록 지

원해주신 국토해양부 (U-City 석 박사과정 지원사업)에 깊은 감사 드립니다

때로 힘들고 지칠 때 항상 저를 믿고 응원해 주시는 가족이 있었기에 지금의 글을 쓸 수 있었

습니다 또한 지금 연구실에서 함께 생활하고 있는 윤규형 민구형 철민이형 현석이형 진열이형

호민이형 병진이 선혜누나와 성대에 계신 창길이형 미국에 계신 의재형과 Debaditiya 에게도 감

사의 말씀을 전합니다 그리고 모두 나열할 수는 없지만 카이스트 생활의 활력소가 되어준 카이

스트 건설 및 환경공학과 선후배들 직원 분들과 SOS 멤버들에게도 감사를 표합니다

앞서 거론한 많은 이들의 도움이 없었다면 저 또한 지금 이 자리에 있을 수 없었을 것입니다

많이 부족하기에 더 노력하고 배우고자 했던 초심을 다시 한번 생각하며 더 큰 결실을 맺기 위해

앞으로 나아가겠습니다 도움을 주신 모든 분들께 이 감사의 글을 바칩니다

2011년 6월 카이스트에서 임형진 올림

- 38 -

CURRICULUM VITAE

Hyung Jin Lim

Research Assistant

The Civil and Environmental Engineering Department

Korea Advanced Institute of Science and Technology (KAIST)

291 Daehak-Ro Guseong-Dong Yuseong-Gu

Daejeon Republic of Korea 305-701

Phone (82)+42-350-3665 Fax (82)+42-350-3610

Email limhj87gmailcom

RESEARCH INTERESTS

Damage Detection Localization and Quantification on Composite and Metallic Structures Pattern Recognition Finite Element Modeling (FEM)

EDUCATION

2009 ndash 2011 MS Civil and Environmental Eng Korea Advanced Institute of Science and Technology (KAIST) Korea Rep

2006 ndash 2009 BS Civil and Environmental Eng Korea Advanced Institute of Science and Technology (KAIST) Korea Rep

JOURNAL PUBLICATIONS The corresponding author is underlined 1 Hyung Jin Lim Hoon Sohn ldquoReference-free damage detection localization and quantification

using in compositerdquo in preparation for Journal of the Acoustical Society of America 2011

2 Chul Min Yeum Hoon Sohn Jeong Bum Ihn Hyung Jin Lim ldquoInstantaneous delamination detection in a composite using a dual piezoelectric transducer networkrdquo Submitted to Compo-sites Science and Technology 2011

3 Yun-Kyu An Hyung Jin Lim Min Koo Kim Jin Yeol Yang Hoon Sohn Chang Guen Lee ldquoInvestigation of applicability of local reference-free damage detection techniques to in-situ

- 39 -

bridgesrdquo Submitted to ASCE Journal of Structural Engineering 2011

4 Hyung Jin Lim Min Koo Kim Hoon Sohn Chan Yik Park ldquoImpedance based damage detec-tion under varying temperature and loading conditionsrdquo Submitted to NDTampE International 2011

5 Hyung Jin Lim Hoon Sohn ldquoQuantitative estimation of transmitted and reflected lamb waves at discontinuityrdquo Journal of the Korean Society for Nondestructive Testing Vol30 No 4 pp1-8 2010

CONFERENCE PROCEEDINGS

The corresponding author is underlined 1 Hyung Jin Lim and Hoon Sohn ldquoReference-free damage detection localization and quantifi-

cation in compositesrdquo ASCE Engineering Mechanics Institute Conference Boston June 2-4 2011

2 Hyung Jin Lim and Hoon Sohn ldquoReference-free delamination quantification within multi-layer composite materialsrdquo KSNT Spring Conference May 26-27 Gyeong-Ju Korea 2011

3 Min Koo Kim Hyun Jin Lim Hoon Sohn Chan Yik Park ldquoImpedance-based bolt loosening detection under varying temperature and loadingrdquo Asian Pacific Workshop on Structural Health Monitoring Tokyo Japan November 30-December 2 2010

4 Hyung Jin Lim Chul Min Yeum and Hoon Sohn ldquoModeling of Impact-induced Delamination in a Multilayer Composite Platerdquo The 23rd KKCNN Symposium on Civil Engineering Taipei (Taiwan) November 13-15 2010

5 Yun-Kyu An Hyung Jin Lim and Hoon Sohn ldquoTemperature independent Delamination Detec-tion using data normalizationrdquo The 3rd International Conference on Multi-Functional Materi-als and Structures Chonbuk National Univ Jeonju Korea September 14-18 2010

6 Hoon Sohn Yun-Kyu An Chang Gil Lee Min Koo Kim Hyung Jin Lim ldquoApplication of online NDT techniques to decommissioned bridge testingrdquo the 5th European Workshop on Structural Health Monitoring Sorrento Italy June 29-July 02 2010

7 Min Koo Kim Hyung Jin Lim Hoon Sohn Chan Yik Park ldquoImpedance based damage diag-nosis of a complex composite aircraft wing under changing loading and temperature condi-tionsrdquo the 5th European Workshop on Structural Health Monitoring Sorrento Italy June 29-July 02 2010

8 Yun-Kyu An Min Koo Kim Hyung Jin Lim and Hoon Sohn ldquoValidations of reference-free crack detection techniques through a decommissioned bridge testrdquo 2010 COSEIK Annual Con-ference Jeju Korea April 8-9 2010

- 40 -

PARTICIPANT PROJECTS

1 Development of Smart Steel Members for Online Steel Structure Monitoring POSCO Corporation (Funded 41000000 KRW (41000 USD) for 010511 to 123111)

2 Development of Enabling Technologies for Structural Health Monitoring Korean Agency for Defense Development (Funded 240000000 KRW (240000 USD) out of total 3200000000 KRW (3200000 USD) for 060108-053111)

3 Development of On-board SHM Technologies for Composite Air Vehicles The Boeing Company (Funded 400000 USD (400000000 KRW) for 080108 to 073111)

EDUCATIONAL EXPERIENCES

Fall 2009

Teaching Assistant CE207 ldquoElementary Structural Engineering and Laboratoryrdquo KAIST 3 unit undergraduate course

EXPERIMENTAL EXPERIENTCES

2010 Monitoring of Real Aircraft Composite Wing Segment Korea Agency for Defense De-velopment (ADD) Daejeon Korea Real time monitoring of composite air vehicles fund-ed by Korea Agency for Defense Development (ADD)

2009-2011

Field Monitoring of Real Bridge Structures Gimpo (Ramp G bridge) amp Icheon (Sam-seung bridge) Korea US-Korea collaborative research for bridge monitoring testbeds

HONORS amp AWARDS

Scholarships for outstanding students granted KAIST (2007-2009) The three prominent students in the civil and environmental engineering department of KAIST were selected as recipients This scholarship provided 2000$ per year for three years

- 41 -

Page 29: 복합재료 손상의 무기저 진단 및 정량화 - KAISTssslab.kaist.ac.kr/article/pdf/Thesis_2.pdf · 1 Declaration of Ethical Conduct in Research: I, as a graduate student

mode decomposition [39-40] PZT B was only used as a sensor for collecting pitch-catch sig-

nal For simulating the excitation and sensing of PZTs the pin-force model was adopted that

an input force apples to the end nodes of excitation PZT and averaging element strains corre-

sponding sensing area A Gaussian pulse (tone-burst) 50 kHz central frequency was excited

as an input signal

TABLE II Damage cases for numerical simulation

119949119949120783120783 (mm) 119941119941 (mm) G value reduction ()

Case 1 650 200 90 darr

Case 2 675 150 90 darr

Case 3 675 150 93 darr

Case 4 900 150 90 darr

- 18 -

44 Numerical simulation results

015 02 025 03 035

-2

0

2

Time (ms)

Am

plitu

de

IntactDamage

Reflected wave from damage

(a) Pitch-catch signal

015 02 025 03 035

-2

0

2

Time (ms)

Am

plitu

de

IntactDamage

Reflected waves from damage

(b) Pulse-echo signal Fig 7 Comparison of A0 Lamb wave mode signals from numerical simulation between intact and damage cases Time delay attenuation and multiple reflections due to damage simulated successfully using the damage modeling method described in the previous section

Fig 7 shows that comparison of A0 Lamb wave mode signals between intact and one of

damage cases in the simulation The features of A0 Lamb wave mode passing through dam-

age area described in section 2 are successfully simulated Thus the equivalent element

method is proper for modeling damage in composites In the damage cases the additional

waves reflected from damage is observed in both pitch-catch and pulse echo signal

Table III presents the result of the proposed damage characterization algorithm using the

numerical simulation cases shown in Table II In lsquoMethod 2rsquo 119907119907119894119894 variation is assumed to plusmn

5 of the normal value (1190~1315ms) The normal value of 119907119907119894119894 for lsquoMethod 3rsquo is meas-

ured as 1253 ms In each case the damage is characterized using the proposed technique

Note that as the information of 119907119907119894119894 is specified the estimated value is closer to the exact

value

- 19 -

TABLE III Damage characterization for the numerical simulation models

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

Case 1

Exact 650 200

Method 1 248 50~70 79~414 160~834

Method 2 248 603~667 135~241 271~486

Method 3 248 635 188 379

Case 2

Exact 675 150

Method 1 193 60~75 01~251 37~651

Method 2 193 618~682 114~222 295~575

Method 3 193 65 168 435

Case 3

Exact 675 150

Method 1 233 60~75 66~302 140~649

Method 2 233 668~738 85~196 182~419

Method 3 233 703 140 301

Case 4

Exact 900 150

Method 1 204 75~100 48~349 118~856

Method 2 204 860~951 107~216 264~530

Method 3 204 906 162 397

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1253 ms 1190~1315 ms Measured 119907119907119894119894 1253 ms

- 20 -

CHAPTER 5 EXPERIMENTAL VERIFICATION

51 Overview

In this section experimental study is investigated so that the proposed damage characteri-

zation technique Overall experimental setup and test specimen are shown in Fig 8 and 9

The experimental setup consists of test specimens data acquisition (DAQ) system and self-

sensing circuit Self-sensing circuit was installed for pulse-echo measurement using PZTs

[40-41]

PZT A(PZT C) PZT B

500

500

140

75

250

Teflon Inserted Damage (d =25)

Unit mm

Thickness = 3

PZT A(PZT C) PZT B

500

500

150

90

250

Impact Location

Unit mm

Thickness = 3

(a) Teflon inserted damage specimen

(b) Impact-induced damage specimen

Fig 8 Test specimens for the experimental verification In the Teflon inserted damage Teflon tapes with 25mm diameter were inserted at the center of the specimen The impact-induced damage was introduced to the specimen using the impact test equipment

- 21 -

511 Specimen

Two carbon composite plates with a dimension of 500times500times3mm3 were manufactured by

Nexcomes Inc by stacking 12 layers of CF3327 3K prepreg (woven fabrics type) manufac-

tured by Hankuk Fiber Inc The material properties of the specimen are presented in Table I

One of the specimens Teflon tapes with 25mm diameter was inserted in every 3 layers at the

center of the specimen to mimic the damage The other specimen is utilized for the character-

ization of real impact-induced damage

512 Data acquisition (DAQ) system and PZT

The data acquisition (DAQ) system consists of an arbitrary waveform generator (AWG) a

high speed signal digitizer (DIG) and two multiplexers (MUX) Using the 16-bits AWG

tone-burst input signal with a plusmn10 peak-to-peak voltage were generated The response signal

was measured by the 14-bit DIG For the signal-to-noise ratio reduction both pitch-catch and

pulse-echo signals were measured 150 times and averaged in the time domain 3 identical

circular PZTs of 10mm diameter manufactured by APC International Ltd were installed on

the specimen A Gaussian pulse (tone-burst) 50 kHz central frequency was excited as an in-

put signal The excitation frequency was chosen in order to avoid overlapping of the S0 and

A0 Lamb wave modes and generating of the higher modes (S1 A1hellip)

- 22 -

PZT Actuator

PZT Sensor

AWG

DIG

DAQ System Specimen

SensingActuatingMUX

1

23

4

1 Sending signal to MUX

4 Sending signal to DIG

Self-SensingCircuit

18 Unit mm

10

(a)

(b)

Fig 9 Overall experimental setup (a) Data acquisition (DAQ) system and (b) Installed circu-lar PZT transducer on the specimen with its dimension

513 Impact test

The impact-induced damage was introduced to the specimen using the impact test equip-

ment presented in Fig 10 A 5kg mass with 10mm tip was dropped from 015m height onto

the same location of the specimen The signals were obtained at each number of impacts

Since the reflected wave is not detected at the first impact case the data obtained from the

second and third impact were utilized for the analysis

10

Unit mm (a)

(b)

Fig 10 (a) Impact test equipment for introducing impact-induced damage and (b) Tip of the impact tester with 10mm diameter

- 23 -

514 Temperature effect test

In reality structures are subjected to changing environmental conditions such as tempera-

ture that affect measured signal and cause false-alarm Thus the robustness of the proposed

damage characterization technique under temperature variation in investigated (Fig 11) The

specimens were placed inside a temperature chamber and a thermocouple was installed on

the specimen to measure temperature of surface of the specimen The signal was obtained

under three temperature cases (0 20 and 50degC) and the humidity was kept at 30

Temperature Chamber

Test Specimen

Fig 11 Test specimen in temperature chamber

- 24 -

52 Teflon inserted damage

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Reflected wave from damage

(a)

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Transmitted WaveReflected Wave

Detected reflected wave

(b)

Fig 12 A0 Lamb wave mode obtained using pitch-catch method from the Teflon inserted damage specimen (a) The reflected wave from damage is shown (b) The Teflon inserted damage is detected by capturing the reflected wave

Fig 12 shows the obtained pitch-catch signal from the Teflon inserted damage specimen

and damage diagnosis result The Teflon inserted damage is detected successfully by captur-

ing the reflected wave Table IV presents the result of the Teflon inserted damage characteri-

zation 119907119907119894119894 is measured as 1213ms and the variation is assumed to plusmn 5 of the normal val-

ue (1152~1273ms) The results indicate that the proposed damage characterization approach

works well in real application

TABLE IV Damage characterization for the Teflon inserted damage specimen

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

Exact 635 25

Method 1 123 60~75 73~338 298~1220

Method 2 123 619~662 229~310 934~1188

Method 3 123 631 284 1161

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1213 ms

- 25 -

53 Impact-induced damage

11 115 12 125 13-4

-2

0

2

4

Time (ms)

Vol

tage

(mV

)

IntactImpact2Impact3

Reflected wave from damage

(a) Pitch-catch signal

11 115 12 125 13 135

-05

0

05

1

Time (ms)

Volta

ge (m

V)

IntactImpact2Impact3

Boundary reflection

Reflected wave from damage

(b) Pulse-echo signal

Fig 13 The signal obtained from the test specimen subjected to impact The impact was in-troduced at the same location of the specimen and the signal was obtained after each impacts As the number of impact is increased the transmitted A0 Lamb wave mode is delayed and attenuated

The obtained signals from impact-induced damage specimen are presented in Fig 13 It is

observed that as the number of impact is increased the transmitted A0 Lamb wave mode is

more delayed because the wave velocity in the damage area is reduced Fig 14 shows the

false-alarm test result whether the proposed damage detection algorithm works even in pris-

tine condition of the test specimen It is shown that no false alarm is occurred because no ad-

ditional reflected wave is captured

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Transmitted Wave

(a) Transmitted wave

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Reflected Wave

No reflected wave

(b) Reflected wave

Fig 14 False alarm test for the proposed damage detection algorithm in pristine condition of the specimen Reflected wave from damage is not captured false alarm is not occurred

- 26 -

TABLE V Damage characterization for the impact-induced damage specimen

of Impacts m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

2

(1470 J)

Method 1 080 65~90 75~246 470~1157

Method 2 080 801~885 114~245 711~1134

Method 3 080 843 179 1122

3

(2205 J)

Method 1 148 50~100 06~380 21~1134

Method 2 148 758~838 249~368 843~1125

Method 3 148 798 308 1044

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1213 ms

Table V presents the characterization results of the impact-induced damage case Here the

measured 119907119907119894119894 and the assumed variation are same as the artificial damage case It is observed

that as the number of impact is increased 119889119889 is increased while 119907119907119889119889 is reduced Thus it is

validated that the proposed damage characterization approach estimates the increment of the

damage well

54 Temperature effect test

541 Teflon inserted damage

11 115 12 125 13-10

-5

0

5

10

Time (ms)

Vol

tage

(mV

)

0 oC20 oC50 oC

(a) Pitch-catch signal

11 115 12 125 13 135

-1

0

1

2

3

Time(ms)

Volta

ge (m

V)

0 oC20 oC50 oC

(b) Pulse-echo signal

Fig 15 Signals obtained from the Teflon inserted damage specimen under temperature varia-tion Time delay and amplitude variation are occurred

Fig 15 shows the obtained signals from the Teflon inserted damage specimen subjected to

- 27 -

temperature variation Time delay and amplitude variations are occurred there is no addition-

al damage is introduced The analyzed results of the obtained signals are reported in Table VI

The measured 119907119907119894119894 is 1247 ms at 0 degC 1213 ms at 20 degC and 1168 ms at 50 degC respectively

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms (1152~1273 ms)

The estimated results are also changed under temperature variations the fluctuation of esti-

mated range and exact value is small

TABLE VI Temperature effect test for the Teflon inserted damage specimen

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

0 degC

Exact 635 25

Method 1 113 60~70 79~268 352~1291

Method 2 113 560~619 233~344 1034~1172

Method 3 113 606 257 1141

20 degC

Exact 635 25

Method 1 123 60~75 73~338 298~1220

Method 2 123 619~662 229~310 934~1188

Method 3 123 631 284 1161

50 degC

Exact 635 25

Method 1 137 60~75 72~328 263~1228

Method 2 137 622~687 183~299 666~1086

Method 3 137 642 264 960

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1247 ms at 0 degC 1213 ms at 20 degC and 1168 ms at 50 degC

- 28 -

542 Impact-induced damage

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Vol

tage

(mV

)

Transmitted WaveReflected Wave

No reflected wave

(a) 0 degC

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Transmitted WaveReflected Wave

No reflected wave

(b) 50 degC

Fig 16 False alarm test under temperature variations It is shown that proposed damage char-acterization algorithm is robust under temperature variations

The false alarm test in pristine condition of the impact-induced damage specimen under

temperature variation was investigated (Fig 16) It is validated that the proposed damage di-

agnosis algorithm is robust under temperature variations Fig 17 shows the obtained signals

from the impact-induced damage (impacted 3 times) specimen under temperature variation

The analyzed results of the obtained signals are reported in Table VII The measured 119907119907119894119894 in

each temperature and the assumed variation are same as the artificial damage case

11 115 12 125 13-5

0

5

Time (ms)

Volta

ge (m

V)

0 oC20 oC50 oC

(a) Pitch-catch signal

11 115 12 125 13 135

-1

0

1

2

3

Time(ms)

Vol

tage

(mV

)

0 oC20 oC50 oC

(b) Pulse-echo signal

Fig 17 Signals obtained from impact-induced damage specimen (Impact 3) under tempera-ture variation

- 29 -

TABLE VII Temperature effect test for the impact induced damage specimen (Impact 3 case)

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

0 degC

Method 1 135 75~90 72~310 267~1148

Method 2 135 705~779 264~381 978~1126

Method 3 135 763 289 1071

20 degC

Method 1 148 50~100 06~380 21~1134

Method 2 148 758~838 249~368 843~1125

Method 3 148 798 308 1044

50 degC

Method 1 153 75~100 53~415 174~1227

Method 2 153 804~889 214~336 702~1103

Method 3 153 815 320 1050

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1247 ms at 0 degC 1213 ms at 20 degC and 1168 ms at 50 degC

- 30 -

CHAPTER 6 CONCLUDING REMARKS

61 Executive summary

In this study the reference-free damage detection localization and quantification tech-

nique in a multi layered composite plate is proposed Additional reflected A0 Lamb wave

mode from the damage due to multiple reflections in damage area is used as a feature for

characterizing the damage First damage is characterized in terms of a damage index (m-

value) defined as the ratio of the physical length of the damage to the wavelength of the A0

Lamb wave mode Then the range of possible damage location length and severity are esti-

mated with constraints This procedure can be accomplished without prior baseline data To

realize the proposed technique the matching pursuit technique is modified for reduction of

the error by considering the dispersive characteristic of Lamb wave The matching pursuit

method estimates the arrival time of each wave packets accurately even two wave packets are

overlapped The numerical simulations and experimental tests including Teflon tape inserted

and impact-induced damage are investigated for the feasibility of the proposed technique

Moreover the effectiveness of the proposed technique under changing temperature is exper-

imentally verified

- 31 -

62 Suggestion for further study

The limitations of this study and the possible improvement and further studies are listed as

follow

1 Measurement of impact-induced damage length using conventional NDT technique

Additional experimental verification for impact-induced damage characterization is

needed by measuring the real length of damage using conventional NDT technique

such as ultrasonic scan or infrared camera

2 Limitation of both side collocated PZT installation In this study excitation PZTs are

installed both side and collocated for decomposing A0 Lamb wave mode However it

can produce decomposition error due to imperfection of PZT location and their installa-

tion Moreover in many cases assess to both surface of structures is not allowed such

as aircraft applications To improve this limitation a reference-free damage characteri-

zation using single side PZTs will be considered

3 Complex structure application In this study damage is characterized in simple plate-

like structure When the monitored structure becomes complex the response signals al-

so becomes complex due to its geometry and boundary Thus the investigations for

complex structure application such as stiffener spar and curvature are needed

- 32 -

REFERENCES

[1] Cawley P and Alleyne D (1996) ldquoThe use of Lamb waves for the long range inspection of large structuresrdquo Ultrasonics 34 pp287-290

[2] Ghosh T Kundu T and Karpur P (1998) ldquoEfficient use of Lamb waves for detecting defects in large platesrdquo Ultrasonics 36 pp791-801

[3] Guirguitui V (2003) ldquoLamb wave generation with piezoelectric wafer active sensors for structural health monitoringrdquo SPIE Conference on Smart Structures and Materials San Diego CA 5391 pp111-422

[4] Ragvahan A and Cesnik CE S (2004) ldquoModeling of piezoelectric-based Lamb wave generation and sensing for structural health monitoringrdquo SPIE Conference on Smart Struc-tures and Materials San Diego CA 5391 pp419-430

[5] Rose JL Pilarski A and Dirit JJ (1993) ldquoAn approach to guided wave mode selection for inspection of laminated platerdquo Journal of Reinforced Plastics and Composites 12(5) pp536-544

[6] Guo N and Cawley P (1993) ldquoThe interaction of Lamb waves with delaminations in composite laminatesrdquo Journal of Acoustical Society of America 94(4) pp 2240-2246

[7] Valdez SHD and Soutis C (2002) ldquoReal-time nondestructive evaluation of fiber com-posite laminates using low-frequency Lamb wavesrdquo Journal of Acoustical Society of Amer-ica 111(5) pp 2026-2033

[8] Yang M and Qiao P (2005) ldquoModeling and experimental detection of damage in virous materials using the pulse-echo method and piezoelectric sensoractuatorsrdquo Smart Materials and Structures 14 pp1083-1100

[9] Lestari W and Qiao P (2005) ldquoApplication of wave propagation analysis for damage identification in composite laminated beamsrdquo Journal of Composite Materials 39(22) pp1967-1984

[10] Kessler S S Spearing S M and Soutis C (2002) ldquoDamage detection in composite ma-terials using Lamb wave methodsrdquo Smart Materials and Structures 11 pp269-278

[11] Sohn H Park G Wail J R Limback N P and Farrar C R (2004) ldquoWavelet-based active sensing for delamination detection in composite structuresrdquo Smart Materials and Structures 13 pp 153-160

[12] Petculescu G Krishnaswamy S and Achenback J D (2008) ldquoGroup delay measure-ments using modally selective Lamb wave transducers for detection and sizing of delamina-tions in compositesrdquo Smart Materials and Structures 17

[13] Sohn H Park H W Law K H and Farrar C R (2007) ldquoDamage detection in compo-

- 33 -

site plates by using an enhanced time reversal methodrdquo Journal of Aerospace Engineering (ASCE) pp 141-151

[14] Sohn H (2007) ldquoEffects of environmental and operational variability on structural health monitoringrdquo Philosophical Transactions of the Royal Society Ardquo 365 pp539-560

[15] Rytter A (1993) ldquoVibration based inspection of civil engineering structuresrdquo PhD The-sis Department of Building Technology and Structural Engineering University of Aalborg Aalborg Denmark

[16] Hayashi T and Kawashima K (2002) ldquoMultiple reflections of Lamb waves at a delami-nationrdquo Ultrasonics 40 pp193-197

[17] Wang C H and Rose L R F (2003) ldquoWave reflection and transmission in beams con-taining delamination and inhomogeneityrdquo Journal of Sound and Vibration 264 pp 851-872

[18] Ramadas C Balasubramaniam K Joshi M and Krishnamurthy C V (2009) ldquoInterac-tion of the primary anti-symmetric Lamb mode (Ao) with symmetric delaminations numer-ical and experimental studiesrdquo Smart Materials and Structures 18

[19] Ramadas C Balasubramaniam K Joshi M and Krishnamurthy C V (2010) ldquoInterac-tion of guided Lamb waves with an asymmetrically located delamination in a laminated composite platerdquo Smart Materials and Structures 19

[20] Wilcox P Lowe M and Cawely P (2001) ldquoThe effect of dispersion on long-range in-spection using ultrasonic guided wavesrdquo NDTampE International 34 pp1-9

[21] Mallat S G and Zhang Z (1993) ldquoMatching pursuits with time-frequency dictionariesrdquo IEEE Tansactions on Signal Processing 41(12) pp3397-3415

[22] Zhang G Zhang S and Wang Y (2000) ldquoApplication of adaptive time-frequency de-composition in ultrasonic NDE of highly scattering materialsrdquo Ultrasonics 38 pp 961-964

[23] Hong J C Sun K H and Kim Y Y (2005) ldquoThe matching pursuit approach based on the modulated Gaussian pulse for efficient guided-wave damage inspectionrdquo Smart Mate-rials and Structures 14 pp548-560

[24] Herakovich C T (1997) Mechanics of Fibrous Composites Wiley 480 pages

[25] Reddy J N (1985) ldquoA review of the literature on finite-element modeling of laminated composite platesrdquo The Shock and Vibration Digest 17(4) pp 3-8

[26] Reddy J N (2003) Mechanics of Laminated Composite Plates and Shells Theory and Analysis CRC Press 856 pages

[27] Palacz M Krawczuk M and Ostachowicz W (2005) ldquoThe spectral finite element mod-el for analysis of flexure-shear coupled wave propagation Part 1 laminated multilayer

- 34 -

composite beamrdquo Composite Structures 68 pp37-44

[28] Zhang Y C and Harding J (1990) ldquoA numerical micromechanics analysis of the me-chanical properties of a plain weave compositerdquo Computer amp Structures 36(5) pp 839-844

[29] Nemat-Nasser S and Hori M (1999) Micromechanics Overall Properties of Heteroge-neous Materials Elsevier 810 pages

[30] Scida D Aboura Z Benzeggagh M L and Bocherens E (1999) ldquoA micromechanics model for 3D elasticity and failure of woven-fibre composite materialrdquo Composites Science and Technology 59 pp 505-517

[31] Thomas Y and Wu X H (1997) ldquoA multiscale finite element method for elliptic prob-lems in composite materials and porous mediardquo Journal of Computational Physics 134 pp 169-189

[32] Lee K H Moorthy S and Ghosh S (1999) ldquoMultiple scale computational model for damage in composite materialsrdquo Computer Methods in Applied Mechanics and Engineer-ing 172 pp 175-201

[33] Namilae S and Chandra N (2005) ldquoMultiscale model to study the effect of interfaces in carbon nanotube-based compositesrdquo Journal of Engineering Materials and Technology (ASME) 127 pp222-232

[34] Adams D E (2007) Health monitoring of Structural Materials and Components Wiley 460 pages

[35] Gopalakrishnan S (2009) ldquoModeling Aspects in Finite Elementrdquo Encyclopedia of Struc-tural Health Monitoring 2 pp791-809

[36] Palacz M Krawczuk M and Ostachowicz W (2005) ldquoThe spectral finite element mod-el for analysis of flexure-shear coupled wave propagation Part 2 Deminated multilayer composite beamrdquo Composite Structures 68 pp45-51

[37] Baik J M and Thopson R B (1984) ldquoUltrasonic scattering from imperfect interfaces A quasi-static modelrdquo Journal of Nondestructive Evaluation 4(3) pp177-196

[38] Luo H and Hanagud S (2000) ldquoDynamics of delamincated beamsrdquo International Jour-nal of Solids and Structures 37 pp1501-1519

[39] Kim S B and Sohn H (2007) ldquoInstantaneous reference-free crack detection based on polarization characteristics of piezoelectric materialsrdquo Smart Materials and Structures 16 pp2375-2387

[40] An Y K and Sohn H (2010) ldquoInstantaneous crack detection under varying temperature and static loading conditionsrdquo The Special Issue of Structural Control and Health Monitor-ing 17(7) pp 730-741

- 35 -

[41] Lee S J and Sohn H (2006) ldquoActive self-sensing scheme development for structural health monitoringrdquo Smart Materials and Structures 15(6) pp 1734-1746

- 36 -

요 약 문

복합재료 손상의 무기저 진단 및 정량화

복합재료는 재료의 경량성 내구성 고강도의 특성으로 인하여 항공기에서부터

교량에까지 널리 사용되고 있다 하지만 외부의 충격으로 인한 복합재료 내부의

손상은 재료의 강도를 감소시켜 구조물의 안정성을 저하시키는 요인이 된다 하지만

복합재료의 손상은 재료 내부에서 발생하여 육안으로는 식별하기도 힘들며 표면에서

보이는 것만으로 내부의 손상 정도를 판단하기는 불가능하다는 단점을 가지고 있다

본 연구에서는 손상 내부에서 반사로 인해 발생하는 파를 이용하여 기저자료를

사용하지 않고 특정 시점에서 얻은 자료만으로 복합재료의 손상을 진단하고 정량화

할 수 있는 기법을 제안한다 제안된 기법의 적용성을 검증하기 위하여 유한요소법을

이용한 수치해석을 수행하였고 Teflon tape 이 삽입된 시편과 실제 충격하중으로

인한 손상을 입은 시편을 이용한 실험을 진행하였다 수치해석 및 실험 결과 제안된

기법이 복합재료의 손상을 성공적으로 진단 및 정량화할 수 있음을 확인하였다 또한

다양한 외부 온도 환경에서 제안된 기법의 환경영향 실험을 수행하였고 다양한

환경에서 손상의 진단 및 정량화가 성공적으로 이루어짐을 검증하였다

핵심어 무기저 손상진단 및 정량화 압전센서 복합재료 유한요소법 Matching pursuit

method

- 37 -

감 사 의 글

기대와 두려움을 안고 시작한 석사과정이 어느덧 2 년 이라는 세월 속에서 추억으로 자리잡아

가고 있습니다 학부생 기간을 포함한 지난 3 년 동안 SHM 이라는 새로운 분야를 접하고 연구를

수행 할 수 있도록 도와주신 고마운 분들이 있어 그 감사를 표하고자 합니다

우선 저의 많은 부족한 점을 일깨워 주시고 연구에 대한 뜻을 더욱 확고히 세울 수 있게 해

주신 지도교수님이신 손 훈 교수님께 깊은 감사의 뜻을 전합니다 짧은 기간이었지만 교수님의

도움으로 많은 것을 배우고 느끼고 깨달을 수 있었습니다 또한 본 연구를 수행할 수 있도록 지

원해주신 국토해양부 (U-City 석 박사과정 지원사업)에 깊은 감사 드립니다

때로 힘들고 지칠 때 항상 저를 믿고 응원해 주시는 가족이 있었기에 지금의 글을 쓸 수 있었

습니다 또한 지금 연구실에서 함께 생활하고 있는 윤규형 민구형 철민이형 현석이형 진열이형

호민이형 병진이 선혜누나와 성대에 계신 창길이형 미국에 계신 의재형과 Debaditiya 에게도 감

사의 말씀을 전합니다 그리고 모두 나열할 수는 없지만 카이스트 생활의 활력소가 되어준 카이

스트 건설 및 환경공학과 선후배들 직원 분들과 SOS 멤버들에게도 감사를 표합니다

앞서 거론한 많은 이들의 도움이 없었다면 저 또한 지금 이 자리에 있을 수 없었을 것입니다

많이 부족하기에 더 노력하고 배우고자 했던 초심을 다시 한번 생각하며 더 큰 결실을 맺기 위해

앞으로 나아가겠습니다 도움을 주신 모든 분들께 이 감사의 글을 바칩니다

2011년 6월 카이스트에서 임형진 올림

- 38 -

CURRICULUM VITAE

Hyung Jin Lim

Research Assistant

The Civil and Environmental Engineering Department

Korea Advanced Institute of Science and Technology (KAIST)

291 Daehak-Ro Guseong-Dong Yuseong-Gu

Daejeon Republic of Korea 305-701

Phone (82)+42-350-3665 Fax (82)+42-350-3610

Email limhj87gmailcom

RESEARCH INTERESTS

Damage Detection Localization and Quantification on Composite and Metallic Structures Pattern Recognition Finite Element Modeling (FEM)

EDUCATION

2009 ndash 2011 MS Civil and Environmental Eng Korea Advanced Institute of Science and Technology (KAIST) Korea Rep

2006 ndash 2009 BS Civil and Environmental Eng Korea Advanced Institute of Science and Technology (KAIST) Korea Rep

JOURNAL PUBLICATIONS The corresponding author is underlined 1 Hyung Jin Lim Hoon Sohn ldquoReference-free damage detection localization and quantification

using in compositerdquo in preparation for Journal of the Acoustical Society of America 2011

2 Chul Min Yeum Hoon Sohn Jeong Bum Ihn Hyung Jin Lim ldquoInstantaneous delamination detection in a composite using a dual piezoelectric transducer networkrdquo Submitted to Compo-sites Science and Technology 2011

3 Yun-Kyu An Hyung Jin Lim Min Koo Kim Jin Yeol Yang Hoon Sohn Chang Guen Lee ldquoInvestigation of applicability of local reference-free damage detection techniques to in-situ

- 39 -

bridgesrdquo Submitted to ASCE Journal of Structural Engineering 2011

4 Hyung Jin Lim Min Koo Kim Hoon Sohn Chan Yik Park ldquoImpedance based damage detec-tion under varying temperature and loading conditionsrdquo Submitted to NDTampE International 2011

5 Hyung Jin Lim Hoon Sohn ldquoQuantitative estimation of transmitted and reflected lamb waves at discontinuityrdquo Journal of the Korean Society for Nondestructive Testing Vol30 No 4 pp1-8 2010

CONFERENCE PROCEEDINGS

The corresponding author is underlined 1 Hyung Jin Lim and Hoon Sohn ldquoReference-free damage detection localization and quantifi-

cation in compositesrdquo ASCE Engineering Mechanics Institute Conference Boston June 2-4 2011

2 Hyung Jin Lim and Hoon Sohn ldquoReference-free delamination quantification within multi-layer composite materialsrdquo KSNT Spring Conference May 26-27 Gyeong-Ju Korea 2011

3 Min Koo Kim Hyun Jin Lim Hoon Sohn Chan Yik Park ldquoImpedance-based bolt loosening detection under varying temperature and loadingrdquo Asian Pacific Workshop on Structural Health Monitoring Tokyo Japan November 30-December 2 2010

4 Hyung Jin Lim Chul Min Yeum and Hoon Sohn ldquoModeling of Impact-induced Delamination in a Multilayer Composite Platerdquo The 23rd KKCNN Symposium on Civil Engineering Taipei (Taiwan) November 13-15 2010

5 Yun-Kyu An Hyung Jin Lim and Hoon Sohn ldquoTemperature independent Delamination Detec-tion using data normalizationrdquo The 3rd International Conference on Multi-Functional Materi-als and Structures Chonbuk National Univ Jeonju Korea September 14-18 2010

6 Hoon Sohn Yun-Kyu An Chang Gil Lee Min Koo Kim Hyung Jin Lim ldquoApplication of online NDT techniques to decommissioned bridge testingrdquo the 5th European Workshop on Structural Health Monitoring Sorrento Italy June 29-July 02 2010

7 Min Koo Kim Hyung Jin Lim Hoon Sohn Chan Yik Park ldquoImpedance based damage diag-nosis of a complex composite aircraft wing under changing loading and temperature condi-tionsrdquo the 5th European Workshop on Structural Health Monitoring Sorrento Italy June 29-July 02 2010

8 Yun-Kyu An Min Koo Kim Hyung Jin Lim and Hoon Sohn ldquoValidations of reference-free crack detection techniques through a decommissioned bridge testrdquo 2010 COSEIK Annual Con-ference Jeju Korea April 8-9 2010

- 40 -

PARTICIPANT PROJECTS

1 Development of Smart Steel Members for Online Steel Structure Monitoring POSCO Corporation (Funded 41000000 KRW (41000 USD) for 010511 to 123111)

2 Development of Enabling Technologies for Structural Health Monitoring Korean Agency for Defense Development (Funded 240000000 KRW (240000 USD) out of total 3200000000 KRW (3200000 USD) for 060108-053111)

3 Development of On-board SHM Technologies for Composite Air Vehicles The Boeing Company (Funded 400000 USD (400000000 KRW) for 080108 to 073111)

EDUCATIONAL EXPERIENCES

Fall 2009

Teaching Assistant CE207 ldquoElementary Structural Engineering and Laboratoryrdquo KAIST 3 unit undergraduate course

EXPERIMENTAL EXPERIENTCES

2010 Monitoring of Real Aircraft Composite Wing Segment Korea Agency for Defense De-velopment (ADD) Daejeon Korea Real time monitoring of composite air vehicles fund-ed by Korea Agency for Defense Development (ADD)

2009-2011

Field Monitoring of Real Bridge Structures Gimpo (Ramp G bridge) amp Icheon (Sam-seung bridge) Korea US-Korea collaborative research for bridge monitoring testbeds

HONORS amp AWARDS

Scholarships for outstanding students granted KAIST (2007-2009) The three prominent students in the civil and environmental engineering department of KAIST were selected as recipients This scholarship provided 2000$ per year for three years

- 41 -

Page 30: 복합재료 손상의 무기저 진단 및 정량화 - KAISTssslab.kaist.ac.kr/article/pdf/Thesis_2.pdf · 1 Declaration of Ethical Conduct in Research: I, as a graduate student

44 Numerical simulation results

015 02 025 03 035

-2

0

2

Time (ms)

Am

plitu

de

IntactDamage

Reflected wave from damage

(a) Pitch-catch signal

015 02 025 03 035

-2

0

2

Time (ms)

Am

plitu

de

IntactDamage

Reflected waves from damage

(b) Pulse-echo signal Fig 7 Comparison of A0 Lamb wave mode signals from numerical simulation between intact and damage cases Time delay attenuation and multiple reflections due to damage simulated successfully using the damage modeling method described in the previous section

Fig 7 shows that comparison of A0 Lamb wave mode signals between intact and one of

damage cases in the simulation The features of A0 Lamb wave mode passing through dam-

age area described in section 2 are successfully simulated Thus the equivalent element

method is proper for modeling damage in composites In the damage cases the additional

waves reflected from damage is observed in both pitch-catch and pulse echo signal

Table III presents the result of the proposed damage characterization algorithm using the

numerical simulation cases shown in Table II In lsquoMethod 2rsquo 119907119907119894119894 variation is assumed to plusmn

5 of the normal value (1190~1315ms) The normal value of 119907119907119894119894 for lsquoMethod 3rsquo is meas-

ured as 1253 ms In each case the damage is characterized using the proposed technique

Note that as the information of 119907119907119894119894 is specified the estimated value is closer to the exact

value

- 19 -

TABLE III Damage characterization for the numerical simulation models

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

Case 1

Exact 650 200

Method 1 248 50~70 79~414 160~834

Method 2 248 603~667 135~241 271~486

Method 3 248 635 188 379

Case 2

Exact 675 150

Method 1 193 60~75 01~251 37~651

Method 2 193 618~682 114~222 295~575

Method 3 193 65 168 435

Case 3

Exact 675 150

Method 1 233 60~75 66~302 140~649

Method 2 233 668~738 85~196 182~419

Method 3 233 703 140 301

Case 4

Exact 900 150

Method 1 204 75~100 48~349 118~856

Method 2 204 860~951 107~216 264~530

Method 3 204 906 162 397

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1253 ms 1190~1315 ms Measured 119907119907119894119894 1253 ms

- 20 -

CHAPTER 5 EXPERIMENTAL VERIFICATION

51 Overview

In this section experimental study is investigated so that the proposed damage characteri-

zation technique Overall experimental setup and test specimen are shown in Fig 8 and 9

The experimental setup consists of test specimens data acquisition (DAQ) system and self-

sensing circuit Self-sensing circuit was installed for pulse-echo measurement using PZTs

[40-41]

PZT A(PZT C) PZT B

500

500

140

75

250

Teflon Inserted Damage (d =25)

Unit mm

Thickness = 3

PZT A(PZT C) PZT B

500

500

150

90

250

Impact Location

Unit mm

Thickness = 3

(a) Teflon inserted damage specimen

(b) Impact-induced damage specimen

Fig 8 Test specimens for the experimental verification In the Teflon inserted damage Teflon tapes with 25mm diameter were inserted at the center of the specimen The impact-induced damage was introduced to the specimen using the impact test equipment

- 21 -

511 Specimen

Two carbon composite plates with a dimension of 500times500times3mm3 were manufactured by

Nexcomes Inc by stacking 12 layers of CF3327 3K prepreg (woven fabrics type) manufac-

tured by Hankuk Fiber Inc The material properties of the specimen are presented in Table I

One of the specimens Teflon tapes with 25mm diameter was inserted in every 3 layers at the

center of the specimen to mimic the damage The other specimen is utilized for the character-

ization of real impact-induced damage

512 Data acquisition (DAQ) system and PZT

The data acquisition (DAQ) system consists of an arbitrary waveform generator (AWG) a

high speed signal digitizer (DIG) and two multiplexers (MUX) Using the 16-bits AWG

tone-burst input signal with a plusmn10 peak-to-peak voltage were generated The response signal

was measured by the 14-bit DIG For the signal-to-noise ratio reduction both pitch-catch and

pulse-echo signals were measured 150 times and averaged in the time domain 3 identical

circular PZTs of 10mm diameter manufactured by APC International Ltd were installed on

the specimen A Gaussian pulse (tone-burst) 50 kHz central frequency was excited as an in-

put signal The excitation frequency was chosen in order to avoid overlapping of the S0 and

A0 Lamb wave modes and generating of the higher modes (S1 A1hellip)

- 22 -

PZT Actuator

PZT Sensor

AWG

DIG

DAQ System Specimen

SensingActuatingMUX

1

23

4

1 Sending signal to MUX

4 Sending signal to DIG

Self-SensingCircuit

18 Unit mm

10

(a)

(b)

Fig 9 Overall experimental setup (a) Data acquisition (DAQ) system and (b) Installed circu-lar PZT transducer on the specimen with its dimension

513 Impact test

The impact-induced damage was introduced to the specimen using the impact test equip-

ment presented in Fig 10 A 5kg mass with 10mm tip was dropped from 015m height onto

the same location of the specimen The signals were obtained at each number of impacts

Since the reflected wave is not detected at the first impact case the data obtained from the

second and third impact were utilized for the analysis

10

Unit mm (a)

(b)

Fig 10 (a) Impact test equipment for introducing impact-induced damage and (b) Tip of the impact tester with 10mm diameter

- 23 -

514 Temperature effect test

In reality structures are subjected to changing environmental conditions such as tempera-

ture that affect measured signal and cause false-alarm Thus the robustness of the proposed

damage characterization technique under temperature variation in investigated (Fig 11) The

specimens were placed inside a temperature chamber and a thermocouple was installed on

the specimen to measure temperature of surface of the specimen The signal was obtained

under three temperature cases (0 20 and 50degC) and the humidity was kept at 30

Temperature Chamber

Test Specimen

Fig 11 Test specimen in temperature chamber

- 24 -

52 Teflon inserted damage

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Reflected wave from damage

(a)

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Transmitted WaveReflected Wave

Detected reflected wave

(b)

Fig 12 A0 Lamb wave mode obtained using pitch-catch method from the Teflon inserted damage specimen (a) The reflected wave from damage is shown (b) The Teflon inserted damage is detected by capturing the reflected wave

Fig 12 shows the obtained pitch-catch signal from the Teflon inserted damage specimen

and damage diagnosis result The Teflon inserted damage is detected successfully by captur-

ing the reflected wave Table IV presents the result of the Teflon inserted damage characteri-

zation 119907119907119894119894 is measured as 1213ms and the variation is assumed to plusmn 5 of the normal val-

ue (1152~1273ms) The results indicate that the proposed damage characterization approach

works well in real application

TABLE IV Damage characterization for the Teflon inserted damage specimen

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

Exact 635 25

Method 1 123 60~75 73~338 298~1220

Method 2 123 619~662 229~310 934~1188

Method 3 123 631 284 1161

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1213 ms

- 25 -

53 Impact-induced damage

11 115 12 125 13-4

-2

0

2

4

Time (ms)

Vol

tage

(mV

)

IntactImpact2Impact3

Reflected wave from damage

(a) Pitch-catch signal

11 115 12 125 13 135

-05

0

05

1

Time (ms)

Volta

ge (m

V)

IntactImpact2Impact3

Boundary reflection

Reflected wave from damage

(b) Pulse-echo signal

Fig 13 The signal obtained from the test specimen subjected to impact The impact was in-troduced at the same location of the specimen and the signal was obtained after each impacts As the number of impact is increased the transmitted A0 Lamb wave mode is delayed and attenuated

The obtained signals from impact-induced damage specimen are presented in Fig 13 It is

observed that as the number of impact is increased the transmitted A0 Lamb wave mode is

more delayed because the wave velocity in the damage area is reduced Fig 14 shows the

false-alarm test result whether the proposed damage detection algorithm works even in pris-

tine condition of the test specimen It is shown that no false alarm is occurred because no ad-

ditional reflected wave is captured

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Transmitted Wave

(a) Transmitted wave

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Reflected Wave

No reflected wave

(b) Reflected wave

Fig 14 False alarm test for the proposed damage detection algorithm in pristine condition of the specimen Reflected wave from damage is not captured false alarm is not occurred

- 26 -

TABLE V Damage characterization for the impact-induced damage specimen

of Impacts m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

2

(1470 J)

Method 1 080 65~90 75~246 470~1157

Method 2 080 801~885 114~245 711~1134

Method 3 080 843 179 1122

3

(2205 J)

Method 1 148 50~100 06~380 21~1134

Method 2 148 758~838 249~368 843~1125

Method 3 148 798 308 1044

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1213 ms

Table V presents the characterization results of the impact-induced damage case Here the

measured 119907119907119894119894 and the assumed variation are same as the artificial damage case It is observed

that as the number of impact is increased 119889119889 is increased while 119907119907119889119889 is reduced Thus it is

validated that the proposed damage characterization approach estimates the increment of the

damage well

54 Temperature effect test

541 Teflon inserted damage

11 115 12 125 13-10

-5

0

5

10

Time (ms)

Vol

tage

(mV

)

0 oC20 oC50 oC

(a) Pitch-catch signal

11 115 12 125 13 135

-1

0

1

2

3

Time(ms)

Volta

ge (m

V)

0 oC20 oC50 oC

(b) Pulse-echo signal

Fig 15 Signals obtained from the Teflon inserted damage specimen under temperature varia-tion Time delay and amplitude variation are occurred

Fig 15 shows the obtained signals from the Teflon inserted damage specimen subjected to

- 27 -

temperature variation Time delay and amplitude variations are occurred there is no addition-

al damage is introduced The analyzed results of the obtained signals are reported in Table VI

The measured 119907119907119894119894 is 1247 ms at 0 degC 1213 ms at 20 degC and 1168 ms at 50 degC respectively

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms (1152~1273 ms)

The estimated results are also changed under temperature variations the fluctuation of esti-

mated range and exact value is small

TABLE VI Temperature effect test for the Teflon inserted damage specimen

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

0 degC

Exact 635 25

Method 1 113 60~70 79~268 352~1291

Method 2 113 560~619 233~344 1034~1172

Method 3 113 606 257 1141

20 degC

Exact 635 25

Method 1 123 60~75 73~338 298~1220

Method 2 123 619~662 229~310 934~1188

Method 3 123 631 284 1161

50 degC

Exact 635 25

Method 1 137 60~75 72~328 263~1228

Method 2 137 622~687 183~299 666~1086

Method 3 137 642 264 960

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1247 ms at 0 degC 1213 ms at 20 degC and 1168 ms at 50 degC

- 28 -

542 Impact-induced damage

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Vol

tage

(mV

)

Transmitted WaveReflected Wave

No reflected wave

(a) 0 degC

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Transmitted WaveReflected Wave

No reflected wave

(b) 50 degC

Fig 16 False alarm test under temperature variations It is shown that proposed damage char-acterization algorithm is robust under temperature variations

The false alarm test in pristine condition of the impact-induced damage specimen under

temperature variation was investigated (Fig 16) It is validated that the proposed damage di-

agnosis algorithm is robust under temperature variations Fig 17 shows the obtained signals

from the impact-induced damage (impacted 3 times) specimen under temperature variation

The analyzed results of the obtained signals are reported in Table VII The measured 119907119907119894119894 in

each temperature and the assumed variation are same as the artificial damage case

11 115 12 125 13-5

0

5

Time (ms)

Volta

ge (m

V)

0 oC20 oC50 oC

(a) Pitch-catch signal

11 115 12 125 13 135

-1

0

1

2

3

Time(ms)

Vol

tage

(mV

)

0 oC20 oC50 oC

(b) Pulse-echo signal

Fig 17 Signals obtained from impact-induced damage specimen (Impact 3) under tempera-ture variation

- 29 -

TABLE VII Temperature effect test for the impact induced damage specimen (Impact 3 case)

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

0 degC

Method 1 135 75~90 72~310 267~1148

Method 2 135 705~779 264~381 978~1126

Method 3 135 763 289 1071

20 degC

Method 1 148 50~100 06~380 21~1134

Method 2 148 758~838 249~368 843~1125

Method 3 148 798 308 1044

50 degC

Method 1 153 75~100 53~415 174~1227

Method 2 153 804~889 214~336 702~1103

Method 3 153 815 320 1050

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1247 ms at 0 degC 1213 ms at 20 degC and 1168 ms at 50 degC

- 30 -

CHAPTER 6 CONCLUDING REMARKS

61 Executive summary

In this study the reference-free damage detection localization and quantification tech-

nique in a multi layered composite plate is proposed Additional reflected A0 Lamb wave

mode from the damage due to multiple reflections in damage area is used as a feature for

characterizing the damage First damage is characterized in terms of a damage index (m-

value) defined as the ratio of the physical length of the damage to the wavelength of the A0

Lamb wave mode Then the range of possible damage location length and severity are esti-

mated with constraints This procedure can be accomplished without prior baseline data To

realize the proposed technique the matching pursuit technique is modified for reduction of

the error by considering the dispersive characteristic of Lamb wave The matching pursuit

method estimates the arrival time of each wave packets accurately even two wave packets are

overlapped The numerical simulations and experimental tests including Teflon tape inserted

and impact-induced damage are investigated for the feasibility of the proposed technique

Moreover the effectiveness of the proposed technique under changing temperature is exper-

imentally verified

- 31 -

62 Suggestion for further study

The limitations of this study and the possible improvement and further studies are listed as

follow

1 Measurement of impact-induced damage length using conventional NDT technique

Additional experimental verification for impact-induced damage characterization is

needed by measuring the real length of damage using conventional NDT technique

such as ultrasonic scan or infrared camera

2 Limitation of both side collocated PZT installation In this study excitation PZTs are

installed both side and collocated for decomposing A0 Lamb wave mode However it

can produce decomposition error due to imperfection of PZT location and their installa-

tion Moreover in many cases assess to both surface of structures is not allowed such

as aircraft applications To improve this limitation a reference-free damage characteri-

zation using single side PZTs will be considered

3 Complex structure application In this study damage is characterized in simple plate-

like structure When the monitored structure becomes complex the response signals al-

so becomes complex due to its geometry and boundary Thus the investigations for

complex structure application such as stiffener spar and curvature are needed

- 32 -

REFERENCES

[1] Cawley P and Alleyne D (1996) ldquoThe use of Lamb waves for the long range inspection of large structuresrdquo Ultrasonics 34 pp287-290

[2] Ghosh T Kundu T and Karpur P (1998) ldquoEfficient use of Lamb waves for detecting defects in large platesrdquo Ultrasonics 36 pp791-801

[3] Guirguitui V (2003) ldquoLamb wave generation with piezoelectric wafer active sensors for structural health monitoringrdquo SPIE Conference on Smart Structures and Materials San Diego CA 5391 pp111-422

[4] Ragvahan A and Cesnik CE S (2004) ldquoModeling of piezoelectric-based Lamb wave generation and sensing for structural health monitoringrdquo SPIE Conference on Smart Struc-tures and Materials San Diego CA 5391 pp419-430

[5] Rose JL Pilarski A and Dirit JJ (1993) ldquoAn approach to guided wave mode selection for inspection of laminated platerdquo Journal of Reinforced Plastics and Composites 12(5) pp536-544

[6] Guo N and Cawley P (1993) ldquoThe interaction of Lamb waves with delaminations in composite laminatesrdquo Journal of Acoustical Society of America 94(4) pp 2240-2246

[7] Valdez SHD and Soutis C (2002) ldquoReal-time nondestructive evaluation of fiber com-posite laminates using low-frequency Lamb wavesrdquo Journal of Acoustical Society of Amer-ica 111(5) pp 2026-2033

[8] Yang M and Qiao P (2005) ldquoModeling and experimental detection of damage in virous materials using the pulse-echo method and piezoelectric sensoractuatorsrdquo Smart Materials and Structures 14 pp1083-1100

[9] Lestari W and Qiao P (2005) ldquoApplication of wave propagation analysis for damage identification in composite laminated beamsrdquo Journal of Composite Materials 39(22) pp1967-1984

[10] Kessler S S Spearing S M and Soutis C (2002) ldquoDamage detection in composite ma-terials using Lamb wave methodsrdquo Smart Materials and Structures 11 pp269-278

[11] Sohn H Park G Wail J R Limback N P and Farrar C R (2004) ldquoWavelet-based active sensing for delamination detection in composite structuresrdquo Smart Materials and Structures 13 pp 153-160

[12] Petculescu G Krishnaswamy S and Achenback J D (2008) ldquoGroup delay measure-ments using modally selective Lamb wave transducers for detection and sizing of delamina-tions in compositesrdquo Smart Materials and Structures 17

[13] Sohn H Park H W Law K H and Farrar C R (2007) ldquoDamage detection in compo-

- 33 -

site plates by using an enhanced time reversal methodrdquo Journal of Aerospace Engineering (ASCE) pp 141-151

[14] Sohn H (2007) ldquoEffects of environmental and operational variability on structural health monitoringrdquo Philosophical Transactions of the Royal Society Ardquo 365 pp539-560

[15] Rytter A (1993) ldquoVibration based inspection of civil engineering structuresrdquo PhD The-sis Department of Building Technology and Structural Engineering University of Aalborg Aalborg Denmark

[16] Hayashi T and Kawashima K (2002) ldquoMultiple reflections of Lamb waves at a delami-nationrdquo Ultrasonics 40 pp193-197

[17] Wang C H and Rose L R F (2003) ldquoWave reflection and transmission in beams con-taining delamination and inhomogeneityrdquo Journal of Sound and Vibration 264 pp 851-872

[18] Ramadas C Balasubramaniam K Joshi M and Krishnamurthy C V (2009) ldquoInterac-tion of the primary anti-symmetric Lamb mode (Ao) with symmetric delaminations numer-ical and experimental studiesrdquo Smart Materials and Structures 18

[19] Ramadas C Balasubramaniam K Joshi M and Krishnamurthy C V (2010) ldquoInterac-tion of guided Lamb waves with an asymmetrically located delamination in a laminated composite platerdquo Smart Materials and Structures 19

[20] Wilcox P Lowe M and Cawely P (2001) ldquoThe effect of dispersion on long-range in-spection using ultrasonic guided wavesrdquo NDTampE International 34 pp1-9

[21] Mallat S G and Zhang Z (1993) ldquoMatching pursuits with time-frequency dictionariesrdquo IEEE Tansactions on Signal Processing 41(12) pp3397-3415

[22] Zhang G Zhang S and Wang Y (2000) ldquoApplication of adaptive time-frequency de-composition in ultrasonic NDE of highly scattering materialsrdquo Ultrasonics 38 pp 961-964

[23] Hong J C Sun K H and Kim Y Y (2005) ldquoThe matching pursuit approach based on the modulated Gaussian pulse for efficient guided-wave damage inspectionrdquo Smart Mate-rials and Structures 14 pp548-560

[24] Herakovich C T (1997) Mechanics of Fibrous Composites Wiley 480 pages

[25] Reddy J N (1985) ldquoA review of the literature on finite-element modeling of laminated composite platesrdquo The Shock and Vibration Digest 17(4) pp 3-8

[26] Reddy J N (2003) Mechanics of Laminated Composite Plates and Shells Theory and Analysis CRC Press 856 pages

[27] Palacz M Krawczuk M and Ostachowicz W (2005) ldquoThe spectral finite element mod-el for analysis of flexure-shear coupled wave propagation Part 1 laminated multilayer

- 34 -

composite beamrdquo Composite Structures 68 pp37-44

[28] Zhang Y C and Harding J (1990) ldquoA numerical micromechanics analysis of the me-chanical properties of a plain weave compositerdquo Computer amp Structures 36(5) pp 839-844

[29] Nemat-Nasser S and Hori M (1999) Micromechanics Overall Properties of Heteroge-neous Materials Elsevier 810 pages

[30] Scida D Aboura Z Benzeggagh M L and Bocherens E (1999) ldquoA micromechanics model for 3D elasticity and failure of woven-fibre composite materialrdquo Composites Science and Technology 59 pp 505-517

[31] Thomas Y and Wu X H (1997) ldquoA multiscale finite element method for elliptic prob-lems in composite materials and porous mediardquo Journal of Computational Physics 134 pp 169-189

[32] Lee K H Moorthy S and Ghosh S (1999) ldquoMultiple scale computational model for damage in composite materialsrdquo Computer Methods in Applied Mechanics and Engineer-ing 172 pp 175-201

[33] Namilae S and Chandra N (2005) ldquoMultiscale model to study the effect of interfaces in carbon nanotube-based compositesrdquo Journal of Engineering Materials and Technology (ASME) 127 pp222-232

[34] Adams D E (2007) Health monitoring of Structural Materials and Components Wiley 460 pages

[35] Gopalakrishnan S (2009) ldquoModeling Aspects in Finite Elementrdquo Encyclopedia of Struc-tural Health Monitoring 2 pp791-809

[36] Palacz M Krawczuk M and Ostachowicz W (2005) ldquoThe spectral finite element mod-el for analysis of flexure-shear coupled wave propagation Part 2 Deminated multilayer composite beamrdquo Composite Structures 68 pp45-51

[37] Baik J M and Thopson R B (1984) ldquoUltrasonic scattering from imperfect interfaces A quasi-static modelrdquo Journal of Nondestructive Evaluation 4(3) pp177-196

[38] Luo H and Hanagud S (2000) ldquoDynamics of delamincated beamsrdquo International Jour-nal of Solids and Structures 37 pp1501-1519

[39] Kim S B and Sohn H (2007) ldquoInstantaneous reference-free crack detection based on polarization characteristics of piezoelectric materialsrdquo Smart Materials and Structures 16 pp2375-2387

[40] An Y K and Sohn H (2010) ldquoInstantaneous crack detection under varying temperature and static loading conditionsrdquo The Special Issue of Structural Control and Health Monitor-ing 17(7) pp 730-741

- 35 -

[41] Lee S J and Sohn H (2006) ldquoActive self-sensing scheme development for structural health monitoringrdquo Smart Materials and Structures 15(6) pp 1734-1746

- 36 -

요 약 문

복합재료 손상의 무기저 진단 및 정량화

복합재료는 재료의 경량성 내구성 고강도의 특성으로 인하여 항공기에서부터

교량에까지 널리 사용되고 있다 하지만 외부의 충격으로 인한 복합재료 내부의

손상은 재료의 강도를 감소시켜 구조물의 안정성을 저하시키는 요인이 된다 하지만

복합재료의 손상은 재료 내부에서 발생하여 육안으로는 식별하기도 힘들며 표면에서

보이는 것만으로 내부의 손상 정도를 판단하기는 불가능하다는 단점을 가지고 있다

본 연구에서는 손상 내부에서 반사로 인해 발생하는 파를 이용하여 기저자료를

사용하지 않고 특정 시점에서 얻은 자료만으로 복합재료의 손상을 진단하고 정량화

할 수 있는 기법을 제안한다 제안된 기법의 적용성을 검증하기 위하여 유한요소법을

이용한 수치해석을 수행하였고 Teflon tape 이 삽입된 시편과 실제 충격하중으로

인한 손상을 입은 시편을 이용한 실험을 진행하였다 수치해석 및 실험 결과 제안된

기법이 복합재료의 손상을 성공적으로 진단 및 정량화할 수 있음을 확인하였다 또한

다양한 외부 온도 환경에서 제안된 기법의 환경영향 실험을 수행하였고 다양한

환경에서 손상의 진단 및 정량화가 성공적으로 이루어짐을 검증하였다

핵심어 무기저 손상진단 및 정량화 압전센서 복합재료 유한요소법 Matching pursuit

method

- 37 -

감 사 의 글

기대와 두려움을 안고 시작한 석사과정이 어느덧 2 년 이라는 세월 속에서 추억으로 자리잡아

가고 있습니다 학부생 기간을 포함한 지난 3 년 동안 SHM 이라는 새로운 분야를 접하고 연구를

수행 할 수 있도록 도와주신 고마운 분들이 있어 그 감사를 표하고자 합니다

우선 저의 많은 부족한 점을 일깨워 주시고 연구에 대한 뜻을 더욱 확고히 세울 수 있게 해

주신 지도교수님이신 손 훈 교수님께 깊은 감사의 뜻을 전합니다 짧은 기간이었지만 교수님의

도움으로 많은 것을 배우고 느끼고 깨달을 수 있었습니다 또한 본 연구를 수행할 수 있도록 지

원해주신 국토해양부 (U-City 석 박사과정 지원사업)에 깊은 감사 드립니다

때로 힘들고 지칠 때 항상 저를 믿고 응원해 주시는 가족이 있었기에 지금의 글을 쓸 수 있었

습니다 또한 지금 연구실에서 함께 생활하고 있는 윤규형 민구형 철민이형 현석이형 진열이형

호민이형 병진이 선혜누나와 성대에 계신 창길이형 미국에 계신 의재형과 Debaditiya 에게도 감

사의 말씀을 전합니다 그리고 모두 나열할 수는 없지만 카이스트 생활의 활력소가 되어준 카이

스트 건설 및 환경공학과 선후배들 직원 분들과 SOS 멤버들에게도 감사를 표합니다

앞서 거론한 많은 이들의 도움이 없었다면 저 또한 지금 이 자리에 있을 수 없었을 것입니다

많이 부족하기에 더 노력하고 배우고자 했던 초심을 다시 한번 생각하며 더 큰 결실을 맺기 위해

앞으로 나아가겠습니다 도움을 주신 모든 분들께 이 감사의 글을 바칩니다

2011년 6월 카이스트에서 임형진 올림

- 38 -

CURRICULUM VITAE

Hyung Jin Lim

Research Assistant

The Civil and Environmental Engineering Department

Korea Advanced Institute of Science and Technology (KAIST)

291 Daehak-Ro Guseong-Dong Yuseong-Gu

Daejeon Republic of Korea 305-701

Phone (82)+42-350-3665 Fax (82)+42-350-3610

Email limhj87gmailcom

RESEARCH INTERESTS

Damage Detection Localization and Quantification on Composite and Metallic Structures Pattern Recognition Finite Element Modeling (FEM)

EDUCATION

2009 ndash 2011 MS Civil and Environmental Eng Korea Advanced Institute of Science and Technology (KAIST) Korea Rep

2006 ndash 2009 BS Civil and Environmental Eng Korea Advanced Institute of Science and Technology (KAIST) Korea Rep

JOURNAL PUBLICATIONS The corresponding author is underlined 1 Hyung Jin Lim Hoon Sohn ldquoReference-free damage detection localization and quantification

using in compositerdquo in preparation for Journal of the Acoustical Society of America 2011

2 Chul Min Yeum Hoon Sohn Jeong Bum Ihn Hyung Jin Lim ldquoInstantaneous delamination detection in a composite using a dual piezoelectric transducer networkrdquo Submitted to Compo-sites Science and Technology 2011

3 Yun-Kyu An Hyung Jin Lim Min Koo Kim Jin Yeol Yang Hoon Sohn Chang Guen Lee ldquoInvestigation of applicability of local reference-free damage detection techniques to in-situ

- 39 -

bridgesrdquo Submitted to ASCE Journal of Structural Engineering 2011

4 Hyung Jin Lim Min Koo Kim Hoon Sohn Chan Yik Park ldquoImpedance based damage detec-tion under varying temperature and loading conditionsrdquo Submitted to NDTampE International 2011

5 Hyung Jin Lim Hoon Sohn ldquoQuantitative estimation of transmitted and reflected lamb waves at discontinuityrdquo Journal of the Korean Society for Nondestructive Testing Vol30 No 4 pp1-8 2010

CONFERENCE PROCEEDINGS

The corresponding author is underlined 1 Hyung Jin Lim and Hoon Sohn ldquoReference-free damage detection localization and quantifi-

cation in compositesrdquo ASCE Engineering Mechanics Institute Conference Boston June 2-4 2011

2 Hyung Jin Lim and Hoon Sohn ldquoReference-free delamination quantification within multi-layer composite materialsrdquo KSNT Spring Conference May 26-27 Gyeong-Ju Korea 2011

3 Min Koo Kim Hyun Jin Lim Hoon Sohn Chan Yik Park ldquoImpedance-based bolt loosening detection under varying temperature and loadingrdquo Asian Pacific Workshop on Structural Health Monitoring Tokyo Japan November 30-December 2 2010

4 Hyung Jin Lim Chul Min Yeum and Hoon Sohn ldquoModeling of Impact-induced Delamination in a Multilayer Composite Platerdquo The 23rd KKCNN Symposium on Civil Engineering Taipei (Taiwan) November 13-15 2010

5 Yun-Kyu An Hyung Jin Lim and Hoon Sohn ldquoTemperature independent Delamination Detec-tion using data normalizationrdquo The 3rd International Conference on Multi-Functional Materi-als and Structures Chonbuk National Univ Jeonju Korea September 14-18 2010

6 Hoon Sohn Yun-Kyu An Chang Gil Lee Min Koo Kim Hyung Jin Lim ldquoApplication of online NDT techniques to decommissioned bridge testingrdquo the 5th European Workshop on Structural Health Monitoring Sorrento Italy June 29-July 02 2010

7 Min Koo Kim Hyung Jin Lim Hoon Sohn Chan Yik Park ldquoImpedance based damage diag-nosis of a complex composite aircraft wing under changing loading and temperature condi-tionsrdquo the 5th European Workshop on Structural Health Monitoring Sorrento Italy June 29-July 02 2010

8 Yun-Kyu An Min Koo Kim Hyung Jin Lim and Hoon Sohn ldquoValidations of reference-free crack detection techniques through a decommissioned bridge testrdquo 2010 COSEIK Annual Con-ference Jeju Korea April 8-9 2010

- 40 -

PARTICIPANT PROJECTS

1 Development of Smart Steel Members for Online Steel Structure Monitoring POSCO Corporation (Funded 41000000 KRW (41000 USD) for 010511 to 123111)

2 Development of Enabling Technologies for Structural Health Monitoring Korean Agency for Defense Development (Funded 240000000 KRW (240000 USD) out of total 3200000000 KRW (3200000 USD) for 060108-053111)

3 Development of On-board SHM Technologies for Composite Air Vehicles The Boeing Company (Funded 400000 USD (400000000 KRW) for 080108 to 073111)

EDUCATIONAL EXPERIENCES

Fall 2009

Teaching Assistant CE207 ldquoElementary Structural Engineering and Laboratoryrdquo KAIST 3 unit undergraduate course

EXPERIMENTAL EXPERIENTCES

2010 Monitoring of Real Aircraft Composite Wing Segment Korea Agency for Defense De-velopment (ADD) Daejeon Korea Real time monitoring of composite air vehicles fund-ed by Korea Agency for Defense Development (ADD)

2009-2011

Field Monitoring of Real Bridge Structures Gimpo (Ramp G bridge) amp Icheon (Sam-seung bridge) Korea US-Korea collaborative research for bridge monitoring testbeds

HONORS amp AWARDS

Scholarships for outstanding students granted KAIST (2007-2009) The three prominent students in the civil and environmental engineering department of KAIST were selected as recipients This scholarship provided 2000$ per year for three years

- 41 -

Page 31: 복합재료 손상의 무기저 진단 및 정량화 - KAISTssslab.kaist.ac.kr/article/pdf/Thesis_2.pdf · 1 Declaration of Ethical Conduct in Research: I, as a graduate student

TABLE III Damage characterization for the numerical simulation models

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

Case 1

Exact 650 200

Method 1 248 50~70 79~414 160~834

Method 2 248 603~667 135~241 271~486

Method 3 248 635 188 379

Case 2

Exact 675 150

Method 1 193 60~75 01~251 37~651

Method 2 193 618~682 114~222 295~575

Method 3 193 65 168 435

Case 3

Exact 675 150

Method 1 233 60~75 66~302 140~649

Method 2 233 668~738 85~196 182~419

Method 3 233 703 140 301

Case 4

Exact 900 150

Method 1 204 75~100 48~349 118~856

Method 2 204 860~951 107~216 264~530

Method 3 204 906 162 397

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1253 ms 1190~1315 ms Measured 119907119907119894119894 1253 ms

- 20 -

CHAPTER 5 EXPERIMENTAL VERIFICATION

51 Overview

In this section experimental study is investigated so that the proposed damage characteri-

zation technique Overall experimental setup and test specimen are shown in Fig 8 and 9

The experimental setup consists of test specimens data acquisition (DAQ) system and self-

sensing circuit Self-sensing circuit was installed for pulse-echo measurement using PZTs

[40-41]

PZT A(PZT C) PZT B

500

500

140

75

250

Teflon Inserted Damage (d =25)

Unit mm

Thickness = 3

PZT A(PZT C) PZT B

500

500

150

90

250

Impact Location

Unit mm

Thickness = 3

(a) Teflon inserted damage specimen

(b) Impact-induced damage specimen

Fig 8 Test specimens for the experimental verification In the Teflon inserted damage Teflon tapes with 25mm diameter were inserted at the center of the specimen The impact-induced damage was introduced to the specimen using the impact test equipment

- 21 -

511 Specimen

Two carbon composite plates with a dimension of 500times500times3mm3 were manufactured by

Nexcomes Inc by stacking 12 layers of CF3327 3K prepreg (woven fabrics type) manufac-

tured by Hankuk Fiber Inc The material properties of the specimen are presented in Table I

One of the specimens Teflon tapes with 25mm diameter was inserted in every 3 layers at the

center of the specimen to mimic the damage The other specimen is utilized for the character-

ization of real impact-induced damage

512 Data acquisition (DAQ) system and PZT

The data acquisition (DAQ) system consists of an arbitrary waveform generator (AWG) a

high speed signal digitizer (DIG) and two multiplexers (MUX) Using the 16-bits AWG

tone-burst input signal with a plusmn10 peak-to-peak voltage were generated The response signal

was measured by the 14-bit DIG For the signal-to-noise ratio reduction both pitch-catch and

pulse-echo signals were measured 150 times and averaged in the time domain 3 identical

circular PZTs of 10mm diameter manufactured by APC International Ltd were installed on

the specimen A Gaussian pulse (tone-burst) 50 kHz central frequency was excited as an in-

put signal The excitation frequency was chosen in order to avoid overlapping of the S0 and

A0 Lamb wave modes and generating of the higher modes (S1 A1hellip)

- 22 -

PZT Actuator

PZT Sensor

AWG

DIG

DAQ System Specimen

SensingActuatingMUX

1

23

4

1 Sending signal to MUX

4 Sending signal to DIG

Self-SensingCircuit

18 Unit mm

10

(a)

(b)

Fig 9 Overall experimental setup (a) Data acquisition (DAQ) system and (b) Installed circu-lar PZT transducer on the specimen with its dimension

513 Impact test

The impact-induced damage was introduced to the specimen using the impact test equip-

ment presented in Fig 10 A 5kg mass with 10mm tip was dropped from 015m height onto

the same location of the specimen The signals were obtained at each number of impacts

Since the reflected wave is not detected at the first impact case the data obtained from the

second and third impact were utilized for the analysis

10

Unit mm (a)

(b)

Fig 10 (a) Impact test equipment for introducing impact-induced damage and (b) Tip of the impact tester with 10mm diameter

- 23 -

514 Temperature effect test

In reality structures are subjected to changing environmental conditions such as tempera-

ture that affect measured signal and cause false-alarm Thus the robustness of the proposed

damage characterization technique under temperature variation in investigated (Fig 11) The

specimens were placed inside a temperature chamber and a thermocouple was installed on

the specimen to measure temperature of surface of the specimen The signal was obtained

under three temperature cases (0 20 and 50degC) and the humidity was kept at 30

Temperature Chamber

Test Specimen

Fig 11 Test specimen in temperature chamber

- 24 -

52 Teflon inserted damage

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Reflected wave from damage

(a)

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Transmitted WaveReflected Wave

Detected reflected wave

(b)

Fig 12 A0 Lamb wave mode obtained using pitch-catch method from the Teflon inserted damage specimen (a) The reflected wave from damage is shown (b) The Teflon inserted damage is detected by capturing the reflected wave

Fig 12 shows the obtained pitch-catch signal from the Teflon inserted damage specimen

and damage diagnosis result The Teflon inserted damage is detected successfully by captur-

ing the reflected wave Table IV presents the result of the Teflon inserted damage characteri-

zation 119907119907119894119894 is measured as 1213ms and the variation is assumed to plusmn 5 of the normal val-

ue (1152~1273ms) The results indicate that the proposed damage characterization approach

works well in real application

TABLE IV Damage characterization for the Teflon inserted damage specimen

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

Exact 635 25

Method 1 123 60~75 73~338 298~1220

Method 2 123 619~662 229~310 934~1188

Method 3 123 631 284 1161

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1213 ms

- 25 -

53 Impact-induced damage

11 115 12 125 13-4

-2

0

2

4

Time (ms)

Vol

tage

(mV

)

IntactImpact2Impact3

Reflected wave from damage

(a) Pitch-catch signal

11 115 12 125 13 135

-05

0

05

1

Time (ms)

Volta

ge (m

V)

IntactImpact2Impact3

Boundary reflection

Reflected wave from damage

(b) Pulse-echo signal

Fig 13 The signal obtained from the test specimen subjected to impact The impact was in-troduced at the same location of the specimen and the signal was obtained after each impacts As the number of impact is increased the transmitted A0 Lamb wave mode is delayed and attenuated

The obtained signals from impact-induced damage specimen are presented in Fig 13 It is

observed that as the number of impact is increased the transmitted A0 Lamb wave mode is

more delayed because the wave velocity in the damage area is reduced Fig 14 shows the

false-alarm test result whether the proposed damage detection algorithm works even in pris-

tine condition of the test specimen It is shown that no false alarm is occurred because no ad-

ditional reflected wave is captured

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Transmitted Wave

(a) Transmitted wave

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Reflected Wave

No reflected wave

(b) Reflected wave

Fig 14 False alarm test for the proposed damage detection algorithm in pristine condition of the specimen Reflected wave from damage is not captured false alarm is not occurred

- 26 -

TABLE V Damage characterization for the impact-induced damage specimen

of Impacts m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

2

(1470 J)

Method 1 080 65~90 75~246 470~1157

Method 2 080 801~885 114~245 711~1134

Method 3 080 843 179 1122

3

(2205 J)

Method 1 148 50~100 06~380 21~1134

Method 2 148 758~838 249~368 843~1125

Method 3 148 798 308 1044

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1213 ms

Table V presents the characterization results of the impact-induced damage case Here the

measured 119907119907119894119894 and the assumed variation are same as the artificial damage case It is observed

that as the number of impact is increased 119889119889 is increased while 119907119907119889119889 is reduced Thus it is

validated that the proposed damage characterization approach estimates the increment of the

damage well

54 Temperature effect test

541 Teflon inserted damage

11 115 12 125 13-10

-5

0

5

10

Time (ms)

Vol

tage

(mV

)

0 oC20 oC50 oC

(a) Pitch-catch signal

11 115 12 125 13 135

-1

0

1

2

3

Time(ms)

Volta

ge (m

V)

0 oC20 oC50 oC

(b) Pulse-echo signal

Fig 15 Signals obtained from the Teflon inserted damage specimen under temperature varia-tion Time delay and amplitude variation are occurred

Fig 15 shows the obtained signals from the Teflon inserted damage specimen subjected to

- 27 -

temperature variation Time delay and amplitude variations are occurred there is no addition-

al damage is introduced The analyzed results of the obtained signals are reported in Table VI

The measured 119907119907119894119894 is 1247 ms at 0 degC 1213 ms at 20 degC and 1168 ms at 50 degC respectively

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms (1152~1273 ms)

The estimated results are also changed under temperature variations the fluctuation of esti-

mated range and exact value is small

TABLE VI Temperature effect test for the Teflon inserted damage specimen

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

0 degC

Exact 635 25

Method 1 113 60~70 79~268 352~1291

Method 2 113 560~619 233~344 1034~1172

Method 3 113 606 257 1141

20 degC

Exact 635 25

Method 1 123 60~75 73~338 298~1220

Method 2 123 619~662 229~310 934~1188

Method 3 123 631 284 1161

50 degC

Exact 635 25

Method 1 137 60~75 72~328 263~1228

Method 2 137 622~687 183~299 666~1086

Method 3 137 642 264 960

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1247 ms at 0 degC 1213 ms at 20 degC and 1168 ms at 50 degC

- 28 -

542 Impact-induced damage

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Vol

tage

(mV

)

Transmitted WaveReflected Wave

No reflected wave

(a) 0 degC

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Transmitted WaveReflected Wave

No reflected wave

(b) 50 degC

Fig 16 False alarm test under temperature variations It is shown that proposed damage char-acterization algorithm is robust under temperature variations

The false alarm test in pristine condition of the impact-induced damage specimen under

temperature variation was investigated (Fig 16) It is validated that the proposed damage di-

agnosis algorithm is robust under temperature variations Fig 17 shows the obtained signals

from the impact-induced damage (impacted 3 times) specimen under temperature variation

The analyzed results of the obtained signals are reported in Table VII The measured 119907119907119894119894 in

each temperature and the assumed variation are same as the artificial damage case

11 115 12 125 13-5

0

5

Time (ms)

Volta

ge (m

V)

0 oC20 oC50 oC

(a) Pitch-catch signal

11 115 12 125 13 135

-1

0

1

2

3

Time(ms)

Vol

tage

(mV

)

0 oC20 oC50 oC

(b) Pulse-echo signal

Fig 17 Signals obtained from impact-induced damage specimen (Impact 3) under tempera-ture variation

- 29 -

TABLE VII Temperature effect test for the impact induced damage specimen (Impact 3 case)

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

0 degC

Method 1 135 75~90 72~310 267~1148

Method 2 135 705~779 264~381 978~1126

Method 3 135 763 289 1071

20 degC

Method 1 148 50~100 06~380 21~1134

Method 2 148 758~838 249~368 843~1125

Method 3 148 798 308 1044

50 degC

Method 1 153 75~100 53~415 174~1227

Method 2 153 804~889 214~336 702~1103

Method 3 153 815 320 1050

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1247 ms at 0 degC 1213 ms at 20 degC and 1168 ms at 50 degC

- 30 -

CHAPTER 6 CONCLUDING REMARKS

61 Executive summary

In this study the reference-free damage detection localization and quantification tech-

nique in a multi layered composite plate is proposed Additional reflected A0 Lamb wave

mode from the damage due to multiple reflections in damage area is used as a feature for

characterizing the damage First damage is characterized in terms of a damage index (m-

value) defined as the ratio of the physical length of the damage to the wavelength of the A0

Lamb wave mode Then the range of possible damage location length and severity are esti-

mated with constraints This procedure can be accomplished without prior baseline data To

realize the proposed technique the matching pursuit technique is modified for reduction of

the error by considering the dispersive characteristic of Lamb wave The matching pursuit

method estimates the arrival time of each wave packets accurately even two wave packets are

overlapped The numerical simulations and experimental tests including Teflon tape inserted

and impact-induced damage are investigated for the feasibility of the proposed technique

Moreover the effectiveness of the proposed technique under changing temperature is exper-

imentally verified

- 31 -

62 Suggestion for further study

The limitations of this study and the possible improvement and further studies are listed as

follow

1 Measurement of impact-induced damage length using conventional NDT technique

Additional experimental verification for impact-induced damage characterization is

needed by measuring the real length of damage using conventional NDT technique

such as ultrasonic scan or infrared camera

2 Limitation of both side collocated PZT installation In this study excitation PZTs are

installed both side and collocated for decomposing A0 Lamb wave mode However it

can produce decomposition error due to imperfection of PZT location and their installa-

tion Moreover in many cases assess to both surface of structures is not allowed such

as aircraft applications To improve this limitation a reference-free damage characteri-

zation using single side PZTs will be considered

3 Complex structure application In this study damage is characterized in simple plate-

like structure When the monitored structure becomes complex the response signals al-

so becomes complex due to its geometry and boundary Thus the investigations for

complex structure application such as stiffener spar and curvature are needed

- 32 -

REFERENCES

[1] Cawley P and Alleyne D (1996) ldquoThe use of Lamb waves for the long range inspection of large structuresrdquo Ultrasonics 34 pp287-290

[2] Ghosh T Kundu T and Karpur P (1998) ldquoEfficient use of Lamb waves for detecting defects in large platesrdquo Ultrasonics 36 pp791-801

[3] Guirguitui V (2003) ldquoLamb wave generation with piezoelectric wafer active sensors for structural health monitoringrdquo SPIE Conference on Smart Structures and Materials San Diego CA 5391 pp111-422

[4] Ragvahan A and Cesnik CE S (2004) ldquoModeling of piezoelectric-based Lamb wave generation and sensing for structural health monitoringrdquo SPIE Conference on Smart Struc-tures and Materials San Diego CA 5391 pp419-430

[5] Rose JL Pilarski A and Dirit JJ (1993) ldquoAn approach to guided wave mode selection for inspection of laminated platerdquo Journal of Reinforced Plastics and Composites 12(5) pp536-544

[6] Guo N and Cawley P (1993) ldquoThe interaction of Lamb waves with delaminations in composite laminatesrdquo Journal of Acoustical Society of America 94(4) pp 2240-2246

[7] Valdez SHD and Soutis C (2002) ldquoReal-time nondestructive evaluation of fiber com-posite laminates using low-frequency Lamb wavesrdquo Journal of Acoustical Society of Amer-ica 111(5) pp 2026-2033

[8] Yang M and Qiao P (2005) ldquoModeling and experimental detection of damage in virous materials using the pulse-echo method and piezoelectric sensoractuatorsrdquo Smart Materials and Structures 14 pp1083-1100

[9] Lestari W and Qiao P (2005) ldquoApplication of wave propagation analysis for damage identification in composite laminated beamsrdquo Journal of Composite Materials 39(22) pp1967-1984

[10] Kessler S S Spearing S M and Soutis C (2002) ldquoDamage detection in composite ma-terials using Lamb wave methodsrdquo Smart Materials and Structures 11 pp269-278

[11] Sohn H Park G Wail J R Limback N P and Farrar C R (2004) ldquoWavelet-based active sensing for delamination detection in composite structuresrdquo Smart Materials and Structures 13 pp 153-160

[12] Petculescu G Krishnaswamy S and Achenback J D (2008) ldquoGroup delay measure-ments using modally selective Lamb wave transducers for detection and sizing of delamina-tions in compositesrdquo Smart Materials and Structures 17

[13] Sohn H Park H W Law K H and Farrar C R (2007) ldquoDamage detection in compo-

- 33 -

site plates by using an enhanced time reversal methodrdquo Journal of Aerospace Engineering (ASCE) pp 141-151

[14] Sohn H (2007) ldquoEffects of environmental and operational variability on structural health monitoringrdquo Philosophical Transactions of the Royal Society Ardquo 365 pp539-560

[15] Rytter A (1993) ldquoVibration based inspection of civil engineering structuresrdquo PhD The-sis Department of Building Technology and Structural Engineering University of Aalborg Aalborg Denmark

[16] Hayashi T and Kawashima K (2002) ldquoMultiple reflections of Lamb waves at a delami-nationrdquo Ultrasonics 40 pp193-197

[17] Wang C H and Rose L R F (2003) ldquoWave reflection and transmission in beams con-taining delamination and inhomogeneityrdquo Journal of Sound and Vibration 264 pp 851-872

[18] Ramadas C Balasubramaniam K Joshi M and Krishnamurthy C V (2009) ldquoInterac-tion of the primary anti-symmetric Lamb mode (Ao) with symmetric delaminations numer-ical and experimental studiesrdquo Smart Materials and Structures 18

[19] Ramadas C Balasubramaniam K Joshi M and Krishnamurthy C V (2010) ldquoInterac-tion of guided Lamb waves with an asymmetrically located delamination in a laminated composite platerdquo Smart Materials and Structures 19

[20] Wilcox P Lowe M and Cawely P (2001) ldquoThe effect of dispersion on long-range in-spection using ultrasonic guided wavesrdquo NDTampE International 34 pp1-9

[21] Mallat S G and Zhang Z (1993) ldquoMatching pursuits with time-frequency dictionariesrdquo IEEE Tansactions on Signal Processing 41(12) pp3397-3415

[22] Zhang G Zhang S and Wang Y (2000) ldquoApplication of adaptive time-frequency de-composition in ultrasonic NDE of highly scattering materialsrdquo Ultrasonics 38 pp 961-964

[23] Hong J C Sun K H and Kim Y Y (2005) ldquoThe matching pursuit approach based on the modulated Gaussian pulse for efficient guided-wave damage inspectionrdquo Smart Mate-rials and Structures 14 pp548-560

[24] Herakovich C T (1997) Mechanics of Fibrous Composites Wiley 480 pages

[25] Reddy J N (1985) ldquoA review of the literature on finite-element modeling of laminated composite platesrdquo The Shock and Vibration Digest 17(4) pp 3-8

[26] Reddy J N (2003) Mechanics of Laminated Composite Plates and Shells Theory and Analysis CRC Press 856 pages

[27] Palacz M Krawczuk M and Ostachowicz W (2005) ldquoThe spectral finite element mod-el for analysis of flexure-shear coupled wave propagation Part 1 laminated multilayer

- 34 -

composite beamrdquo Composite Structures 68 pp37-44

[28] Zhang Y C and Harding J (1990) ldquoA numerical micromechanics analysis of the me-chanical properties of a plain weave compositerdquo Computer amp Structures 36(5) pp 839-844

[29] Nemat-Nasser S and Hori M (1999) Micromechanics Overall Properties of Heteroge-neous Materials Elsevier 810 pages

[30] Scida D Aboura Z Benzeggagh M L and Bocherens E (1999) ldquoA micromechanics model for 3D elasticity and failure of woven-fibre composite materialrdquo Composites Science and Technology 59 pp 505-517

[31] Thomas Y and Wu X H (1997) ldquoA multiscale finite element method for elliptic prob-lems in composite materials and porous mediardquo Journal of Computational Physics 134 pp 169-189

[32] Lee K H Moorthy S and Ghosh S (1999) ldquoMultiple scale computational model for damage in composite materialsrdquo Computer Methods in Applied Mechanics and Engineer-ing 172 pp 175-201

[33] Namilae S and Chandra N (2005) ldquoMultiscale model to study the effect of interfaces in carbon nanotube-based compositesrdquo Journal of Engineering Materials and Technology (ASME) 127 pp222-232

[34] Adams D E (2007) Health monitoring of Structural Materials and Components Wiley 460 pages

[35] Gopalakrishnan S (2009) ldquoModeling Aspects in Finite Elementrdquo Encyclopedia of Struc-tural Health Monitoring 2 pp791-809

[36] Palacz M Krawczuk M and Ostachowicz W (2005) ldquoThe spectral finite element mod-el for analysis of flexure-shear coupled wave propagation Part 2 Deminated multilayer composite beamrdquo Composite Structures 68 pp45-51

[37] Baik J M and Thopson R B (1984) ldquoUltrasonic scattering from imperfect interfaces A quasi-static modelrdquo Journal of Nondestructive Evaluation 4(3) pp177-196

[38] Luo H and Hanagud S (2000) ldquoDynamics of delamincated beamsrdquo International Jour-nal of Solids and Structures 37 pp1501-1519

[39] Kim S B and Sohn H (2007) ldquoInstantaneous reference-free crack detection based on polarization characteristics of piezoelectric materialsrdquo Smart Materials and Structures 16 pp2375-2387

[40] An Y K and Sohn H (2010) ldquoInstantaneous crack detection under varying temperature and static loading conditionsrdquo The Special Issue of Structural Control and Health Monitor-ing 17(7) pp 730-741

- 35 -

[41] Lee S J and Sohn H (2006) ldquoActive self-sensing scheme development for structural health monitoringrdquo Smart Materials and Structures 15(6) pp 1734-1746

- 36 -

요 약 문

복합재료 손상의 무기저 진단 및 정량화

복합재료는 재료의 경량성 내구성 고강도의 특성으로 인하여 항공기에서부터

교량에까지 널리 사용되고 있다 하지만 외부의 충격으로 인한 복합재료 내부의

손상은 재료의 강도를 감소시켜 구조물의 안정성을 저하시키는 요인이 된다 하지만

복합재료의 손상은 재료 내부에서 발생하여 육안으로는 식별하기도 힘들며 표면에서

보이는 것만으로 내부의 손상 정도를 판단하기는 불가능하다는 단점을 가지고 있다

본 연구에서는 손상 내부에서 반사로 인해 발생하는 파를 이용하여 기저자료를

사용하지 않고 특정 시점에서 얻은 자료만으로 복합재료의 손상을 진단하고 정량화

할 수 있는 기법을 제안한다 제안된 기법의 적용성을 검증하기 위하여 유한요소법을

이용한 수치해석을 수행하였고 Teflon tape 이 삽입된 시편과 실제 충격하중으로

인한 손상을 입은 시편을 이용한 실험을 진행하였다 수치해석 및 실험 결과 제안된

기법이 복합재료의 손상을 성공적으로 진단 및 정량화할 수 있음을 확인하였다 또한

다양한 외부 온도 환경에서 제안된 기법의 환경영향 실험을 수행하였고 다양한

환경에서 손상의 진단 및 정량화가 성공적으로 이루어짐을 검증하였다

핵심어 무기저 손상진단 및 정량화 압전센서 복합재료 유한요소법 Matching pursuit

method

- 37 -

감 사 의 글

기대와 두려움을 안고 시작한 석사과정이 어느덧 2 년 이라는 세월 속에서 추억으로 자리잡아

가고 있습니다 학부생 기간을 포함한 지난 3 년 동안 SHM 이라는 새로운 분야를 접하고 연구를

수행 할 수 있도록 도와주신 고마운 분들이 있어 그 감사를 표하고자 합니다

우선 저의 많은 부족한 점을 일깨워 주시고 연구에 대한 뜻을 더욱 확고히 세울 수 있게 해

주신 지도교수님이신 손 훈 교수님께 깊은 감사의 뜻을 전합니다 짧은 기간이었지만 교수님의

도움으로 많은 것을 배우고 느끼고 깨달을 수 있었습니다 또한 본 연구를 수행할 수 있도록 지

원해주신 국토해양부 (U-City 석 박사과정 지원사업)에 깊은 감사 드립니다

때로 힘들고 지칠 때 항상 저를 믿고 응원해 주시는 가족이 있었기에 지금의 글을 쓸 수 있었

습니다 또한 지금 연구실에서 함께 생활하고 있는 윤규형 민구형 철민이형 현석이형 진열이형

호민이형 병진이 선혜누나와 성대에 계신 창길이형 미국에 계신 의재형과 Debaditiya 에게도 감

사의 말씀을 전합니다 그리고 모두 나열할 수는 없지만 카이스트 생활의 활력소가 되어준 카이

스트 건설 및 환경공학과 선후배들 직원 분들과 SOS 멤버들에게도 감사를 표합니다

앞서 거론한 많은 이들의 도움이 없었다면 저 또한 지금 이 자리에 있을 수 없었을 것입니다

많이 부족하기에 더 노력하고 배우고자 했던 초심을 다시 한번 생각하며 더 큰 결실을 맺기 위해

앞으로 나아가겠습니다 도움을 주신 모든 분들께 이 감사의 글을 바칩니다

2011년 6월 카이스트에서 임형진 올림

- 38 -

CURRICULUM VITAE

Hyung Jin Lim

Research Assistant

The Civil and Environmental Engineering Department

Korea Advanced Institute of Science and Technology (KAIST)

291 Daehak-Ro Guseong-Dong Yuseong-Gu

Daejeon Republic of Korea 305-701

Phone (82)+42-350-3665 Fax (82)+42-350-3610

Email limhj87gmailcom

RESEARCH INTERESTS

Damage Detection Localization and Quantification on Composite and Metallic Structures Pattern Recognition Finite Element Modeling (FEM)

EDUCATION

2009 ndash 2011 MS Civil and Environmental Eng Korea Advanced Institute of Science and Technology (KAIST) Korea Rep

2006 ndash 2009 BS Civil and Environmental Eng Korea Advanced Institute of Science and Technology (KAIST) Korea Rep

JOURNAL PUBLICATIONS The corresponding author is underlined 1 Hyung Jin Lim Hoon Sohn ldquoReference-free damage detection localization and quantification

using in compositerdquo in preparation for Journal of the Acoustical Society of America 2011

2 Chul Min Yeum Hoon Sohn Jeong Bum Ihn Hyung Jin Lim ldquoInstantaneous delamination detection in a composite using a dual piezoelectric transducer networkrdquo Submitted to Compo-sites Science and Technology 2011

3 Yun-Kyu An Hyung Jin Lim Min Koo Kim Jin Yeol Yang Hoon Sohn Chang Guen Lee ldquoInvestigation of applicability of local reference-free damage detection techniques to in-situ

- 39 -

bridgesrdquo Submitted to ASCE Journal of Structural Engineering 2011

4 Hyung Jin Lim Min Koo Kim Hoon Sohn Chan Yik Park ldquoImpedance based damage detec-tion under varying temperature and loading conditionsrdquo Submitted to NDTampE International 2011

5 Hyung Jin Lim Hoon Sohn ldquoQuantitative estimation of transmitted and reflected lamb waves at discontinuityrdquo Journal of the Korean Society for Nondestructive Testing Vol30 No 4 pp1-8 2010

CONFERENCE PROCEEDINGS

The corresponding author is underlined 1 Hyung Jin Lim and Hoon Sohn ldquoReference-free damage detection localization and quantifi-

cation in compositesrdquo ASCE Engineering Mechanics Institute Conference Boston June 2-4 2011

2 Hyung Jin Lim and Hoon Sohn ldquoReference-free delamination quantification within multi-layer composite materialsrdquo KSNT Spring Conference May 26-27 Gyeong-Ju Korea 2011

3 Min Koo Kim Hyun Jin Lim Hoon Sohn Chan Yik Park ldquoImpedance-based bolt loosening detection under varying temperature and loadingrdquo Asian Pacific Workshop on Structural Health Monitoring Tokyo Japan November 30-December 2 2010

4 Hyung Jin Lim Chul Min Yeum and Hoon Sohn ldquoModeling of Impact-induced Delamination in a Multilayer Composite Platerdquo The 23rd KKCNN Symposium on Civil Engineering Taipei (Taiwan) November 13-15 2010

5 Yun-Kyu An Hyung Jin Lim and Hoon Sohn ldquoTemperature independent Delamination Detec-tion using data normalizationrdquo The 3rd International Conference on Multi-Functional Materi-als and Structures Chonbuk National Univ Jeonju Korea September 14-18 2010

6 Hoon Sohn Yun-Kyu An Chang Gil Lee Min Koo Kim Hyung Jin Lim ldquoApplication of online NDT techniques to decommissioned bridge testingrdquo the 5th European Workshop on Structural Health Monitoring Sorrento Italy June 29-July 02 2010

7 Min Koo Kim Hyung Jin Lim Hoon Sohn Chan Yik Park ldquoImpedance based damage diag-nosis of a complex composite aircraft wing under changing loading and temperature condi-tionsrdquo the 5th European Workshop on Structural Health Monitoring Sorrento Italy June 29-July 02 2010

8 Yun-Kyu An Min Koo Kim Hyung Jin Lim and Hoon Sohn ldquoValidations of reference-free crack detection techniques through a decommissioned bridge testrdquo 2010 COSEIK Annual Con-ference Jeju Korea April 8-9 2010

- 40 -

PARTICIPANT PROJECTS

1 Development of Smart Steel Members for Online Steel Structure Monitoring POSCO Corporation (Funded 41000000 KRW (41000 USD) for 010511 to 123111)

2 Development of Enabling Technologies for Structural Health Monitoring Korean Agency for Defense Development (Funded 240000000 KRW (240000 USD) out of total 3200000000 KRW (3200000 USD) for 060108-053111)

3 Development of On-board SHM Technologies for Composite Air Vehicles The Boeing Company (Funded 400000 USD (400000000 KRW) for 080108 to 073111)

EDUCATIONAL EXPERIENCES

Fall 2009

Teaching Assistant CE207 ldquoElementary Structural Engineering and Laboratoryrdquo KAIST 3 unit undergraduate course

EXPERIMENTAL EXPERIENTCES

2010 Monitoring of Real Aircraft Composite Wing Segment Korea Agency for Defense De-velopment (ADD) Daejeon Korea Real time monitoring of composite air vehicles fund-ed by Korea Agency for Defense Development (ADD)

2009-2011

Field Monitoring of Real Bridge Structures Gimpo (Ramp G bridge) amp Icheon (Sam-seung bridge) Korea US-Korea collaborative research for bridge monitoring testbeds

HONORS amp AWARDS

Scholarships for outstanding students granted KAIST (2007-2009) The three prominent students in the civil and environmental engineering department of KAIST were selected as recipients This scholarship provided 2000$ per year for three years

- 41 -

Page 32: 복합재료 손상의 무기저 진단 및 정량화 - KAISTssslab.kaist.ac.kr/article/pdf/Thesis_2.pdf · 1 Declaration of Ethical Conduct in Research: I, as a graduate student

CHAPTER 5 EXPERIMENTAL VERIFICATION

51 Overview

In this section experimental study is investigated so that the proposed damage characteri-

zation technique Overall experimental setup and test specimen are shown in Fig 8 and 9

The experimental setup consists of test specimens data acquisition (DAQ) system and self-

sensing circuit Self-sensing circuit was installed for pulse-echo measurement using PZTs

[40-41]

PZT A(PZT C) PZT B

500

500

140

75

250

Teflon Inserted Damage (d =25)

Unit mm

Thickness = 3

PZT A(PZT C) PZT B

500

500

150

90

250

Impact Location

Unit mm

Thickness = 3

(a) Teflon inserted damage specimen

(b) Impact-induced damage specimen

Fig 8 Test specimens for the experimental verification In the Teflon inserted damage Teflon tapes with 25mm diameter were inserted at the center of the specimen The impact-induced damage was introduced to the specimen using the impact test equipment

- 21 -

511 Specimen

Two carbon composite plates with a dimension of 500times500times3mm3 were manufactured by

Nexcomes Inc by stacking 12 layers of CF3327 3K prepreg (woven fabrics type) manufac-

tured by Hankuk Fiber Inc The material properties of the specimen are presented in Table I

One of the specimens Teflon tapes with 25mm diameter was inserted in every 3 layers at the

center of the specimen to mimic the damage The other specimen is utilized for the character-

ization of real impact-induced damage

512 Data acquisition (DAQ) system and PZT

The data acquisition (DAQ) system consists of an arbitrary waveform generator (AWG) a

high speed signal digitizer (DIG) and two multiplexers (MUX) Using the 16-bits AWG

tone-burst input signal with a plusmn10 peak-to-peak voltage were generated The response signal

was measured by the 14-bit DIG For the signal-to-noise ratio reduction both pitch-catch and

pulse-echo signals were measured 150 times and averaged in the time domain 3 identical

circular PZTs of 10mm diameter manufactured by APC International Ltd were installed on

the specimen A Gaussian pulse (tone-burst) 50 kHz central frequency was excited as an in-

put signal The excitation frequency was chosen in order to avoid overlapping of the S0 and

A0 Lamb wave modes and generating of the higher modes (S1 A1hellip)

- 22 -

PZT Actuator

PZT Sensor

AWG

DIG

DAQ System Specimen

SensingActuatingMUX

1

23

4

1 Sending signal to MUX

4 Sending signal to DIG

Self-SensingCircuit

18 Unit mm

10

(a)

(b)

Fig 9 Overall experimental setup (a) Data acquisition (DAQ) system and (b) Installed circu-lar PZT transducer on the specimen with its dimension

513 Impact test

The impact-induced damage was introduced to the specimen using the impact test equip-

ment presented in Fig 10 A 5kg mass with 10mm tip was dropped from 015m height onto

the same location of the specimen The signals were obtained at each number of impacts

Since the reflected wave is not detected at the first impact case the data obtained from the

second and third impact were utilized for the analysis

10

Unit mm (a)

(b)

Fig 10 (a) Impact test equipment for introducing impact-induced damage and (b) Tip of the impact tester with 10mm diameter

- 23 -

514 Temperature effect test

In reality structures are subjected to changing environmental conditions such as tempera-

ture that affect measured signal and cause false-alarm Thus the robustness of the proposed

damage characterization technique under temperature variation in investigated (Fig 11) The

specimens were placed inside a temperature chamber and a thermocouple was installed on

the specimen to measure temperature of surface of the specimen The signal was obtained

under three temperature cases (0 20 and 50degC) and the humidity was kept at 30

Temperature Chamber

Test Specimen

Fig 11 Test specimen in temperature chamber

- 24 -

52 Teflon inserted damage

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Reflected wave from damage

(a)

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Transmitted WaveReflected Wave

Detected reflected wave

(b)

Fig 12 A0 Lamb wave mode obtained using pitch-catch method from the Teflon inserted damage specimen (a) The reflected wave from damage is shown (b) The Teflon inserted damage is detected by capturing the reflected wave

Fig 12 shows the obtained pitch-catch signal from the Teflon inserted damage specimen

and damage diagnosis result The Teflon inserted damage is detected successfully by captur-

ing the reflected wave Table IV presents the result of the Teflon inserted damage characteri-

zation 119907119907119894119894 is measured as 1213ms and the variation is assumed to plusmn 5 of the normal val-

ue (1152~1273ms) The results indicate that the proposed damage characterization approach

works well in real application

TABLE IV Damage characterization for the Teflon inserted damage specimen

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

Exact 635 25

Method 1 123 60~75 73~338 298~1220

Method 2 123 619~662 229~310 934~1188

Method 3 123 631 284 1161

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1213 ms

- 25 -

53 Impact-induced damage

11 115 12 125 13-4

-2

0

2

4

Time (ms)

Vol

tage

(mV

)

IntactImpact2Impact3

Reflected wave from damage

(a) Pitch-catch signal

11 115 12 125 13 135

-05

0

05

1

Time (ms)

Volta

ge (m

V)

IntactImpact2Impact3

Boundary reflection

Reflected wave from damage

(b) Pulse-echo signal

Fig 13 The signal obtained from the test specimen subjected to impact The impact was in-troduced at the same location of the specimen and the signal was obtained after each impacts As the number of impact is increased the transmitted A0 Lamb wave mode is delayed and attenuated

The obtained signals from impact-induced damage specimen are presented in Fig 13 It is

observed that as the number of impact is increased the transmitted A0 Lamb wave mode is

more delayed because the wave velocity in the damage area is reduced Fig 14 shows the

false-alarm test result whether the proposed damage detection algorithm works even in pris-

tine condition of the test specimen It is shown that no false alarm is occurred because no ad-

ditional reflected wave is captured

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Transmitted Wave

(a) Transmitted wave

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Reflected Wave

No reflected wave

(b) Reflected wave

Fig 14 False alarm test for the proposed damage detection algorithm in pristine condition of the specimen Reflected wave from damage is not captured false alarm is not occurred

- 26 -

TABLE V Damage characterization for the impact-induced damage specimen

of Impacts m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

2

(1470 J)

Method 1 080 65~90 75~246 470~1157

Method 2 080 801~885 114~245 711~1134

Method 3 080 843 179 1122

3

(2205 J)

Method 1 148 50~100 06~380 21~1134

Method 2 148 758~838 249~368 843~1125

Method 3 148 798 308 1044

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1213 ms

Table V presents the characterization results of the impact-induced damage case Here the

measured 119907119907119894119894 and the assumed variation are same as the artificial damage case It is observed

that as the number of impact is increased 119889119889 is increased while 119907119907119889119889 is reduced Thus it is

validated that the proposed damage characterization approach estimates the increment of the

damage well

54 Temperature effect test

541 Teflon inserted damage

11 115 12 125 13-10

-5

0

5

10

Time (ms)

Vol

tage

(mV

)

0 oC20 oC50 oC

(a) Pitch-catch signal

11 115 12 125 13 135

-1

0

1

2

3

Time(ms)

Volta

ge (m

V)

0 oC20 oC50 oC

(b) Pulse-echo signal

Fig 15 Signals obtained from the Teflon inserted damage specimen under temperature varia-tion Time delay and amplitude variation are occurred

Fig 15 shows the obtained signals from the Teflon inserted damage specimen subjected to

- 27 -

temperature variation Time delay and amplitude variations are occurred there is no addition-

al damage is introduced The analyzed results of the obtained signals are reported in Table VI

The measured 119907119907119894119894 is 1247 ms at 0 degC 1213 ms at 20 degC and 1168 ms at 50 degC respectively

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms (1152~1273 ms)

The estimated results are also changed under temperature variations the fluctuation of esti-

mated range and exact value is small

TABLE VI Temperature effect test for the Teflon inserted damage specimen

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

0 degC

Exact 635 25

Method 1 113 60~70 79~268 352~1291

Method 2 113 560~619 233~344 1034~1172

Method 3 113 606 257 1141

20 degC

Exact 635 25

Method 1 123 60~75 73~338 298~1220

Method 2 123 619~662 229~310 934~1188

Method 3 123 631 284 1161

50 degC

Exact 635 25

Method 1 137 60~75 72~328 263~1228

Method 2 137 622~687 183~299 666~1086

Method 3 137 642 264 960

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1247 ms at 0 degC 1213 ms at 20 degC and 1168 ms at 50 degC

- 28 -

542 Impact-induced damage

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Vol

tage

(mV

)

Transmitted WaveReflected Wave

No reflected wave

(a) 0 degC

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Transmitted WaveReflected Wave

No reflected wave

(b) 50 degC

Fig 16 False alarm test under temperature variations It is shown that proposed damage char-acterization algorithm is robust under temperature variations

The false alarm test in pristine condition of the impact-induced damage specimen under

temperature variation was investigated (Fig 16) It is validated that the proposed damage di-

agnosis algorithm is robust under temperature variations Fig 17 shows the obtained signals

from the impact-induced damage (impacted 3 times) specimen under temperature variation

The analyzed results of the obtained signals are reported in Table VII The measured 119907119907119894119894 in

each temperature and the assumed variation are same as the artificial damage case

11 115 12 125 13-5

0

5

Time (ms)

Volta

ge (m

V)

0 oC20 oC50 oC

(a) Pitch-catch signal

11 115 12 125 13 135

-1

0

1

2

3

Time(ms)

Vol

tage

(mV

)

0 oC20 oC50 oC

(b) Pulse-echo signal

Fig 17 Signals obtained from impact-induced damage specimen (Impact 3) under tempera-ture variation

- 29 -

TABLE VII Temperature effect test for the impact induced damage specimen (Impact 3 case)

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

0 degC

Method 1 135 75~90 72~310 267~1148

Method 2 135 705~779 264~381 978~1126

Method 3 135 763 289 1071

20 degC

Method 1 148 50~100 06~380 21~1134

Method 2 148 758~838 249~368 843~1125

Method 3 148 798 308 1044

50 degC

Method 1 153 75~100 53~415 174~1227

Method 2 153 804~889 214~336 702~1103

Method 3 153 815 320 1050

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1247 ms at 0 degC 1213 ms at 20 degC and 1168 ms at 50 degC

- 30 -

CHAPTER 6 CONCLUDING REMARKS

61 Executive summary

In this study the reference-free damage detection localization and quantification tech-

nique in a multi layered composite plate is proposed Additional reflected A0 Lamb wave

mode from the damage due to multiple reflections in damage area is used as a feature for

characterizing the damage First damage is characterized in terms of a damage index (m-

value) defined as the ratio of the physical length of the damage to the wavelength of the A0

Lamb wave mode Then the range of possible damage location length and severity are esti-

mated with constraints This procedure can be accomplished without prior baseline data To

realize the proposed technique the matching pursuit technique is modified for reduction of

the error by considering the dispersive characteristic of Lamb wave The matching pursuit

method estimates the arrival time of each wave packets accurately even two wave packets are

overlapped The numerical simulations and experimental tests including Teflon tape inserted

and impact-induced damage are investigated for the feasibility of the proposed technique

Moreover the effectiveness of the proposed technique under changing temperature is exper-

imentally verified

- 31 -

62 Suggestion for further study

The limitations of this study and the possible improvement and further studies are listed as

follow

1 Measurement of impact-induced damage length using conventional NDT technique

Additional experimental verification for impact-induced damage characterization is

needed by measuring the real length of damage using conventional NDT technique

such as ultrasonic scan or infrared camera

2 Limitation of both side collocated PZT installation In this study excitation PZTs are

installed both side and collocated for decomposing A0 Lamb wave mode However it

can produce decomposition error due to imperfection of PZT location and their installa-

tion Moreover in many cases assess to both surface of structures is not allowed such

as aircraft applications To improve this limitation a reference-free damage characteri-

zation using single side PZTs will be considered

3 Complex structure application In this study damage is characterized in simple plate-

like structure When the monitored structure becomes complex the response signals al-

so becomes complex due to its geometry and boundary Thus the investigations for

complex structure application such as stiffener spar and curvature are needed

- 32 -

REFERENCES

[1] Cawley P and Alleyne D (1996) ldquoThe use of Lamb waves for the long range inspection of large structuresrdquo Ultrasonics 34 pp287-290

[2] Ghosh T Kundu T and Karpur P (1998) ldquoEfficient use of Lamb waves for detecting defects in large platesrdquo Ultrasonics 36 pp791-801

[3] Guirguitui V (2003) ldquoLamb wave generation with piezoelectric wafer active sensors for structural health monitoringrdquo SPIE Conference on Smart Structures and Materials San Diego CA 5391 pp111-422

[4] Ragvahan A and Cesnik CE S (2004) ldquoModeling of piezoelectric-based Lamb wave generation and sensing for structural health monitoringrdquo SPIE Conference on Smart Struc-tures and Materials San Diego CA 5391 pp419-430

[5] Rose JL Pilarski A and Dirit JJ (1993) ldquoAn approach to guided wave mode selection for inspection of laminated platerdquo Journal of Reinforced Plastics and Composites 12(5) pp536-544

[6] Guo N and Cawley P (1993) ldquoThe interaction of Lamb waves with delaminations in composite laminatesrdquo Journal of Acoustical Society of America 94(4) pp 2240-2246

[7] Valdez SHD and Soutis C (2002) ldquoReal-time nondestructive evaluation of fiber com-posite laminates using low-frequency Lamb wavesrdquo Journal of Acoustical Society of Amer-ica 111(5) pp 2026-2033

[8] Yang M and Qiao P (2005) ldquoModeling and experimental detection of damage in virous materials using the pulse-echo method and piezoelectric sensoractuatorsrdquo Smart Materials and Structures 14 pp1083-1100

[9] Lestari W and Qiao P (2005) ldquoApplication of wave propagation analysis for damage identification in composite laminated beamsrdquo Journal of Composite Materials 39(22) pp1967-1984

[10] Kessler S S Spearing S M and Soutis C (2002) ldquoDamage detection in composite ma-terials using Lamb wave methodsrdquo Smart Materials and Structures 11 pp269-278

[11] Sohn H Park G Wail J R Limback N P and Farrar C R (2004) ldquoWavelet-based active sensing for delamination detection in composite structuresrdquo Smart Materials and Structures 13 pp 153-160

[12] Petculescu G Krishnaswamy S and Achenback J D (2008) ldquoGroup delay measure-ments using modally selective Lamb wave transducers for detection and sizing of delamina-tions in compositesrdquo Smart Materials and Structures 17

[13] Sohn H Park H W Law K H and Farrar C R (2007) ldquoDamage detection in compo-

- 33 -

site plates by using an enhanced time reversal methodrdquo Journal of Aerospace Engineering (ASCE) pp 141-151

[14] Sohn H (2007) ldquoEffects of environmental and operational variability on structural health monitoringrdquo Philosophical Transactions of the Royal Society Ardquo 365 pp539-560

[15] Rytter A (1993) ldquoVibration based inspection of civil engineering structuresrdquo PhD The-sis Department of Building Technology and Structural Engineering University of Aalborg Aalborg Denmark

[16] Hayashi T and Kawashima K (2002) ldquoMultiple reflections of Lamb waves at a delami-nationrdquo Ultrasonics 40 pp193-197

[17] Wang C H and Rose L R F (2003) ldquoWave reflection and transmission in beams con-taining delamination and inhomogeneityrdquo Journal of Sound and Vibration 264 pp 851-872

[18] Ramadas C Balasubramaniam K Joshi M and Krishnamurthy C V (2009) ldquoInterac-tion of the primary anti-symmetric Lamb mode (Ao) with symmetric delaminations numer-ical and experimental studiesrdquo Smart Materials and Structures 18

[19] Ramadas C Balasubramaniam K Joshi M and Krishnamurthy C V (2010) ldquoInterac-tion of guided Lamb waves with an asymmetrically located delamination in a laminated composite platerdquo Smart Materials and Structures 19

[20] Wilcox P Lowe M and Cawely P (2001) ldquoThe effect of dispersion on long-range in-spection using ultrasonic guided wavesrdquo NDTampE International 34 pp1-9

[21] Mallat S G and Zhang Z (1993) ldquoMatching pursuits with time-frequency dictionariesrdquo IEEE Tansactions on Signal Processing 41(12) pp3397-3415

[22] Zhang G Zhang S and Wang Y (2000) ldquoApplication of adaptive time-frequency de-composition in ultrasonic NDE of highly scattering materialsrdquo Ultrasonics 38 pp 961-964

[23] Hong J C Sun K H and Kim Y Y (2005) ldquoThe matching pursuit approach based on the modulated Gaussian pulse for efficient guided-wave damage inspectionrdquo Smart Mate-rials and Structures 14 pp548-560

[24] Herakovich C T (1997) Mechanics of Fibrous Composites Wiley 480 pages

[25] Reddy J N (1985) ldquoA review of the literature on finite-element modeling of laminated composite platesrdquo The Shock and Vibration Digest 17(4) pp 3-8

[26] Reddy J N (2003) Mechanics of Laminated Composite Plates and Shells Theory and Analysis CRC Press 856 pages

[27] Palacz M Krawczuk M and Ostachowicz W (2005) ldquoThe spectral finite element mod-el for analysis of flexure-shear coupled wave propagation Part 1 laminated multilayer

- 34 -

composite beamrdquo Composite Structures 68 pp37-44

[28] Zhang Y C and Harding J (1990) ldquoA numerical micromechanics analysis of the me-chanical properties of a plain weave compositerdquo Computer amp Structures 36(5) pp 839-844

[29] Nemat-Nasser S and Hori M (1999) Micromechanics Overall Properties of Heteroge-neous Materials Elsevier 810 pages

[30] Scida D Aboura Z Benzeggagh M L and Bocherens E (1999) ldquoA micromechanics model for 3D elasticity and failure of woven-fibre composite materialrdquo Composites Science and Technology 59 pp 505-517

[31] Thomas Y and Wu X H (1997) ldquoA multiscale finite element method for elliptic prob-lems in composite materials and porous mediardquo Journal of Computational Physics 134 pp 169-189

[32] Lee K H Moorthy S and Ghosh S (1999) ldquoMultiple scale computational model for damage in composite materialsrdquo Computer Methods in Applied Mechanics and Engineer-ing 172 pp 175-201

[33] Namilae S and Chandra N (2005) ldquoMultiscale model to study the effect of interfaces in carbon nanotube-based compositesrdquo Journal of Engineering Materials and Technology (ASME) 127 pp222-232

[34] Adams D E (2007) Health monitoring of Structural Materials and Components Wiley 460 pages

[35] Gopalakrishnan S (2009) ldquoModeling Aspects in Finite Elementrdquo Encyclopedia of Struc-tural Health Monitoring 2 pp791-809

[36] Palacz M Krawczuk M and Ostachowicz W (2005) ldquoThe spectral finite element mod-el for analysis of flexure-shear coupled wave propagation Part 2 Deminated multilayer composite beamrdquo Composite Structures 68 pp45-51

[37] Baik J M and Thopson R B (1984) ldquoUltrasonic scattering from imperfect interfaces A quasi-static modelrdquo Journal of Nondestructive Evaluation 4(3) pp177-196

[38] Luo H and Hanagud S (2000) ldquoDynamics of delamincated beamsrdquo International Jour-nal of Solids and Structures 37 pp1501-1519

[39] Kim S B and Sohn H (2007) ldquoInstantaneous reference-free crack detection based on polarization characteristics of piezoelectric materialsrdquo Smart Materials and Structures 16 pp2375-2387

[40] An Y K and Sohn H (2010) ldquoInstantaneous crack detection under varying temperature and static loading conditionsrdquo The Special Issue of Structural Control and Health Monitor-ing 17(7) pp 730-741

- 35 -

[41] Lee S J and Sohn H (2006) ldquoActive self-sensing scheme development for structural health monitoringrdquo Smart Materials and Structures 15(6) pp 1734-1746

- 36 -

요 약 문

복합재료 손상의 무기저 진단 및 정량화

복합재료는 재료의 경량성 내구성 고강도의 특성으로 인하여 항공기에서부터

교량에까지 널리 사용되고 있다 하지만 외부의 충격으로 인한 복합재료 내부의

손상은 재료의 강도를 감소시켜 구조물의 안정성을 저하시키는 요인이 된다 하지만

복합재료의 손상은 재료 내부에서 발생하여 육안으로는 식별하기도 힘들며 표면에서

보이는 것만으로 내부의 손상 정도를 판단하기는 불가능하다는 단점을 가지고 있다

본 연구에서는 손상 내부에서 반사로 인해 발생하는 파를 이용하여 기저자료를

사용하지 않고 특정 시점에서 얻은 자료만으로 복합재료의 손상을 진단하고 정량화

할 수 있는 기법을 제안한다 제안된 기법의 적용성을 검증하기 위하여 유한요소법을

이용한 수치해석을 수행하였고 Teflon tape 이 삽입된 시편과 실제 충격하중으로

인한 손상을 입은 시편을 이용한 실험을 진행하였다 수치해석 및 실험 결과 제안된

기법이 복합재료의 손상을 성공적으로 진단 및 정량화할 수 있음을 확인하였다 또한

다양한 외부 온도 환경에서 제안된 기법의 환경영향 실험을 수행하였고 다양한

환경에서 손상의 진단 및 정량화가 성공적으로 이루어짐을 검증하였다

핵심어 무기저 손상진단 및 정량화 압전센서 복합재료 유한요소법 Matching pursuit

method

- 37 -

감 사 의 글

기대와 두려움을 안고 시작한 석사과정이 어느덧 2 년 이라는 세월 속에서 추억으로 자리잡아

가고 있습니다 학부생 기간을 포함한 지난 3 년 동안 SHM 이라는 새로운 분야를 접하고 연구를

수행 할 수 있도록 도와주신 고마운 분들이 있어 그 감사를 표하고자 합니다

우선 저의 많은 부족한 점을 일깨워 주시고 연구에 대한 뜻을 더욱 확고히 세울 수 있게 해

주신 지도교수님이신 손 훈 교수님께 깊은 감사의 뜻을 전합니다 짧은 기간이었지만 교수님의

도움으로 많은 것을 배우고 느끼고 깨달을 수 있었습니다 또한 본 연구를 수행할 수 있도록 지

원해주신 국토해양부 (U-City 석 박사과정 지원사업)에 깊은 감사 드립니다

때로 힘들고 지칠 때 항상 저를 믿고 응원해 주시는 가족이 있었기에 지금의 글을 쓸 수 있었

습니다 또한 지금 연구실에서 함께 생활하고 있는 윤규형 민구형 철민이형 현석이형 진열이형

호민이형 병진이 선혜누나와 성대에 계신 창길이형 미국에 계신 의재형과 Debaditiya 에게도 감

사의 말씀을 전합니다 그리고 모두 나열할 수는 없지만 카이스트 생활의 활력소가 되어준 카이

스트 건설 및 환경공학과 선후배들 직원 분들과 SOS 멤버들에게도 감사를 표합니다

앞서 거론한 많은 이들의 도움이 없었다면 저 또한 지금 이 자리에 있을 수 없었을 것입니다

많이 부족하기에 더 노력하고 배우고자 했던 초심을 다시 한번 생각하며 더 큰 결실을 맺기 위해

앞으로 나아가겠습니다 도움을 주신 모든 분들께 이 감사의 글을 바칩니다

2011년 6월 카이스트에서 임형진 올림

- 38 -

CURRICULUM VITAE

Hyung Jin Lim

Research Assistant

The Civil and Environmental Engineering Department

Korea Advanced Institute of Science and Technology (KAIST)

291 Daehak-Ro Guseong-Dong Yuseong-Gu

Daejeon Republic of Korea 305-701

Phone (82)+42-350-3665 Fax (82)+42-350-3610

Email limhj87gmailcom

RESEARCH INTERESTS

Damage Detection Localization and Quantification on Composite and Metallic Structures Pattern Recognition Finite Element Modeling (FEM)

EDUCATION

2009 ndash 2011 MS Civil and Environmental Eng Korea Advanced Institute of Science and Technology (KAIST) Korea Rep

2006 ndash 2009 BS Civil and Environmental Eng Korea Advanced Institute of Science and Technology (KAIST) Korea Rep

JOURNAL PUBLICATIONS The corresponding author is underlined 1 Hyung Jin Lim Hoon Sohn ldquoReference-free damage detection localization and quantification

using in compositerdquo in preparation for Journal of the Acoustical Society of America 2011

2 Chul Min Yeum Hoon Sohn Jeong Bum Ihn Hyung Jin Lim ldquoInstantaneous delamination detection in a composite using a dual piezoelectric transducer networkrdquo Submitted to Compo-sites Science and Technology 2011

3 Yun-Kyu An Hyung Jin Lim Min Koo Kim Jin Yeol Yang Hoon Sohn Chang Guen Lee ldquoInvestigation of applicability of local reference-free damage detection techniques to in-situ

- 39 -

bridgesrdquo Submitted to ASCE Journal of Structural Engineering 2011

4 Hyung Jin Lim Min Koo Kim Hoon Sohn Chan Yik Park ldquoImpedance based damage detec-tion under varying temperature and loading conditionsrdquo Submitted to NDTampE International 2011

5 Hyung Jin Lim Hoon Sohn ldquoQuantitative estimation of transmitted and reflected lamb waves at discontinuityrdquo Journal of the Korean Society for Nondestructive Testing Vol30 No 4 pp1-8 2010

CONFERENCE PROCEEDINGS

The corresponding author is underlined 1 Hyung Jin Lim and Hoon Sohn ldquoReference-free damage detection localization and quantifi-

cation in compositesrdquo ASCE Engineering Mechanics Institute Conference Boston June 2-4 2011

2 Hyung Jin Lim and Hoon Sohn ldquoReference-free delamination quantification within multi-layer composite materialsrdquo KSNT Spring Conference May 26-27 Gyeong-Ju Korea 2011

3 Min Koo Kim Hyun Jin Lim Hoon Sohn Chan Yik Park ldquoImpedance-based bolt loosening detection under varying temperature and loadingrdquo Asian Pacific Workshop on Structural Health Monitoring Tokyo Japan November 30-December 2 2010

4 Hyung Jin Lim Chul Min Yeum and Hoon Sohn ldquoModeling of Impact-induced Delamination in a Multilayer Composite Platerdquo The 23rd KKCNN Symposium on Civil Engineering Taipei (Taiwan) November 13-15 2010

5 Yun-Kyu An Hyung Jin Lim and Hoon Sohn ldquoTemperature independent Delamination Detec-tion using data normalizationrdquo The 3rd International Conference on Multi-Functional Materi-als and Structures Chonbuk National Univ Jeonju Korea September 14-18 2010

6 Hoon Sohn Yun-Kyu An Chang Gil Lee Min Koo Kim Hyung Jin Lim ldquoApplication of online NDT techniques to decommissioned bridge testingrdquo the 5th European Workshop on Structural Health Monitoring Sorrento Italy June 29-July 02 2010

7 Min Koo Kim Hyung Jin Lim Hoon Sohn Chan Yik Park ldquoImpedance based damage diag-nosis of a complex composite aircraft wing under changing loading and temperature condi-tionsrdquo the 5th European Workshop on Structural Health Monitoring Sorrento Italy June 29-July 02 2010

8 Yun-Kyu An Min Koo Kim Hyung Jin Lim and Hoon Sohn ldquoValidations of reference-free crack detection techniques through a decommissioned bridge testrdquo 2010 COSEIK Annual Con-ference Jeju Korea April 8-9 2010

- 40 -

PARTICIPANT PROJECTS

1 Development of Smart Steel Members for Online Steel Structure Monitoring POSCO Corporation (Funded 41000000 KRW (41000 USD) for 010511 to 123111)

2 Development of Enabling Technologies for Structural Health Monitoring Korean Agency for Defense Development (Funded 240000000 KRW (240000 USD) out of total 3200000000 KRW (3200000 USD) for 060108-053111)

3 Development of On-board SHM Technologies for Composite Air Vehicles The Boeing Company (Funded 400000 USD (400000000 KRW) for 080108 to 073111)

EDUCATIONAL EXPERIENCES

Fall 2009

Teaching Assistant CE207 ldquoElementary Structural Engineering and Laboratoryrdquo KAIST 3 unit undergraduate course

EXPERIMENTAL EXPERIENTCES

2010 Monitoring of Real Aircraft Composite Wing Segment Korea Agency for Defense De-velopment (ADD) Daejeon Korea Real time monitoring of composite air vehicles fund-ed by Korea Agency for Defense Development (ADD)

2009-2011

Field Monitoring of Real Bridge Structures Gimpo (Ramp G bridge) amp Icheon (Sam-seung bridge) Korea US-Korea collaborative research for bridge monitoring testbeds

HONORS amp AWARDS

Scholarships for outstanding students granted KAIST (2007-2009) The three prominent students in the civil and environmental engineering department of KAIST were selected as recipients This scholarship provided 2000$ per year for three years

- 41 -

Page 33: 복합재료 손상의 무기저 진단 및 정량화 - KAISTssslab.kaist.ac.kr/article/pdf/Thesis_2.pdf · 1 Declaration of Ethical Conduct in Research: I, as a graduate student

511 Specimen

Two carbon composite plates with a dimension of 500times500times3mm3 were manufactured by

Nexcomes Inc by stacking 12 layers of CF3327 3K prepreg (woven fabrics type) manufac-

tured by Hankuk Fiber Inc The material properties of the specimen are presented in Table I

One of the specimens Teflon tapes with 25mm diameter was inserted in every 3 layers at the

center of the specimen to mimic the damage The other specimen is utilized for the character-

ization of real impact-induced damage

512 Data acquisition (DAQ) system and PZT

The data acquisition (DAQ) system consists of an arbitrary waveform generator (AWG) a

high speed signal digitizer (DIG) and two multiplexers (MUX) Using the 16-bits AWG

tone-burst input signal with a plusmn10 peak-to-peak voltage were generated The response signal

was measured by the 14-bit DIG For the signal-to-noise ratio reduction both pitch-catch and

pulse-echo signals were measured 150 times and averaged in the time domain 3 identical

circular PZTs of 10mm diameter manufactured by APC International Ltd were installed on

the specimen A Gaussian pulse (tone-burst) 50 kHz central frequency was excited as an in-

put signal The excitation frequency was chosen in order to avoid overlapping of the S0 and

A0 Lamb wave modes and generating of the higher modes (S1 A1hellip)

- 22 -

PZT Actuator

PZT Sensor

AWG

DIG

DAQ System Specimen

SensingActuatingMUX

1

23

4

1 Sending signal to MUX

4 Sending signal to DIG

Self-SensingCircuit

18 Unit mm

10

(a)

(b)

Fig 9 Overall experimental setup (a) Data acquisition (DAQ) system and (b) Installed circu-lar PZT transducer on the specimen with its dimension

513 Impact test

The impact-induced damage was introduced to the specimen using the impact test equip-

ment presented in Fig 10 A 5kg mass with 10mm tip was dropped from 015m height onto

the same location of the specimen The signals were obtained at each number of impacts

Since the reflected wave is not detected at the first impact case the data obtained from the

second and third impact were utilized for the analysis

10

Unit mm (a)

(b)

Fig 10 (a) Impact test equipment for introducing impact-induced damage and (b) Tip of the impact tester with 10mm diameter

- 23 -

514 Temperature effect test

In reality structures are subjected to changing environmental conditions such as tempera-

ture that affect measured signal and cause false-alarm Thus the robustness of the proposed

damage characterization technique under temperature variation in investigated (Fig 11) The

specimens were placed inside a temperature chamber and a thermocouple was installed on

the specimen to measure temperature of surface of the specimen The signal was obtained

under three temperature cases (0 20 and 50degC) and the humidity was kept at 30

Temperature Chamber

Test Specimen

Fig 11 Test specimen in temperature chamber

- 24 -

52 Teflon inserted damage

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Reflected wave from damage

(a)

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Transmitted WaveReflected Wave

Detected reflected wave

(b)

Fig 12 A0 Lamb wave mode obtained using pitch-catch method from the Teflon inserted damage specimen (a) The reflected wave from damage is shown (b) The Teflon inserted damage is detected by capturing the reflected wave

Fig 12 shows the obtained pitch-catch signal from the Teflon inserted damage specimen

and damage diagnosis result The Teflon inserted damage is detected successfully by captur-

ing the reflected wave Table IV presents the result of the Teflon inserted damage characteri-

zation 119907119907119894119894 is measured as 1213ms and the variation is assumed to plusmn 5 of the normal val-

ue (1152~1273ms) The results indicate that the proposed damage characterization approach

works well in real application

TABLE IV Damage characterization for the Teflon inserted damage specimen

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

Exact 635 25

Method 1 123 60~75 73~338 298~1220

Method 2 123 619~662 229~310 934~1188

Method 3 123 631 284 1161

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1213 ms

- 25 -

53 Impact-induced damage

11 115 12 125 13-4

-2

0

2

4

Time (ms)

Vol

tage

(mV

)

IntactImpact2Impact3

Reflected wave from damage

(a) Pitch-catch signal

11 115 12 125 13 135

-05

0

05

1

Time (ms)

Volta

ge (m

V)

IntactImpact2Impact3

Boundary reflection

Reflected wave from damage

(b) Pulse-echo signal

Fig 13 The signal obtained from the test specimen subjected to impact The impact was in-troduced at the same location of the specimen and the signal was obtained after each impacts As the number of impact is increased the transmitted A0 Lamb wave mode is delayed and attenuated

The obtained signals from impact-induced damage specimen are presented in Fig 13 It is

observed that as the number of impact is increased the transmitted A0 Lamb wave mode is

more delayed because the wave velocity in the damage area is reduced Fig 14 shows the

false-alarm test result whether the proposed damage detection algorithm works even in pris-

tine condition of the test specimen It is shown that no false alarm is occurred because no ad-

ditional reflected wave is captured

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Transmitted Wave

(a) Transmitted wave

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Reflected Wave

No reflected wave

(b) Reflected wave

Fig 14 False alarm test for the proposed damage detection algorithm in pristine condition of the specimen Reflected wave from damage is not captured false alarm is not occurred

- 26 -

TABLE V Damage characterization for the impact-induced damage specimen

of Impacts m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

2

(1470 J)

Method 1 080 65~90 75~246 470~1157

Method 2 080 801~885 114~245 711~1134

Method 3 080 843 179 1122

3

(2205 J)

Method 1 148 50~100 06~380 21~1134

Method 2 148 758~838 249~368 843~1125

Method 3 148 798 308 1044

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1213 ms

Table V presents the characterization results of the impact-induced damage case Here the

measured 119907119907119894119894 and the assumed variation are same as the artificial damage case It is observed

that as the number of impact is increased 119889119889 is increased while 119907119907119889119889 is reduced Thus it is

validated that the proposed damage characterization approach estimates the increment of the

damage well

54 Temperature effect test

541 Teflon inserted damage

11 115 12 125 13-10

-5

0

5

10

Time (ms)

Vol

tage

(mV

)

0 oC20 oC50 oC

(a) Pitch-catch signal

11 115 12 125 13 135

-1

0

1

2

3

Time(ms)

Volta

ge (m

V)

0 oC20 oC50 oC

(b) Pulse-echo signal

Fig 15 Signals obtained from the Teflon inserted damage specimen under temperature varia-tion Time delay and amplitude variation are occurred

Fig 15 shows the obtained signals from the Teflon inserted damage specimen subjected to

- 27 -

temperature variation Time delay and amplitude variations are occurred there is no addition-

al damage is introduced The analyzed results of the obtained signals are reported in Table VI

The measured 119907119907119894119894 is 1247 ms at 0 degC 1213 ms at 20 degC and 1168 ms at 50 degC respectively

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms (1152~1273 ms)

The estimated results are also changed under temperature variations the fluctuation of esti-

mated range and exact value is small

TABLE VI Temperature effect test for the Teflon inserted damage specimen

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

0 degC

Exact 635 25

Method 1 113 60~70 79~268 352~1291

Method 2 113 560~619 233~344 1034~1172

Method 3 113 606 257 1141

20 degC

Exact 635 25

Method 1 123 60~75 73~338 298~1220

Method 2 123 619~662 229~310 934~1188

Method 3 123 631 284 1161

50 degC

Exact 635 25

Method 1 137 60~75 72~328 263~1228

Method 2 137 622~687 183~299 666~1086

Method 3 137 642 264 960

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1247 ms at 0 degC 1213 ms at 20 degC and 1168 ms at 50 degC

- 28 -

542 Impact-induced damage

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Vol

tage

(mV

)

Transmitted WaveReflected Wave

No reflected wave

(a) 0 degC

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Transmitted WaveReflected Wave

No reflected wave

(b) 50 degC

Fig 16 False alarm test under temperature variations It is shown that proposed damage char-acterization algorithm is robust under temperature variations

The false alarm test in pristine condition of the impact-induced damage specimen under

temperature variation was investigated (Fig 16) It is validated that the proposed damage di-

agnosis algorithm is robust under temperature variations Fig 17 shows the obtained signals

from the impact-induced damage (impacted 3 times) specimen under temperature variation

The analyzed results of the obtained signals are reported in Table VII The measured 119907119907119894119894 in

each temperature and the assumed variation are same as the artificial damage case

11 115 12 125 13-5

0

5

Time (ms)

Volta

ge (m

V)

0 oC20 oC50 oC

(a) Pitch-catch signal

11 115 12 125 13 135

-1

0

1

2

3

Time(ms)

Vol

tage

(mV

)

0 oC20 oC50 oC

(b) Pulse-echo signal

Fig 17 Signals obtained from impact-induced damage specimen (Impact 3) under tempera-ture variation

- 29 -

TABLE VII Temperature effect test for the impact induced damage specimen (Impact 3 case)

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

0 degC

Method 1 135 75~90 72~310 267~1148

Method 2 135 705~779 264~381 978~1126

Method 3 135 763 289 1071

20 degC

Method 1 148 50~100 06~380 21~1134

Method 2 148 758~838 249~368 843~1125

Method 3 148 798 308 1044

50 degC

Method 1 153 75~100 53~415 174~1227

Method 2 153 804~889 214~336 702~1103

Method 3 153 815 320 1050

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1247 ms at 0 degC 1213 ms at 20 degC and 1168 ms at 50 degC

- 30 -

CHAPTER 6 CONCLUDING REMARKS

61 Executive summary

In this study the reference-free damage detection localization and quantification tech-

nique in a multi layered composite plate is proposed Additional reflected A0 Lamb wave

mode from the damage due to multiple reflections in damage area is used as a feature for

characterizing the damage First damage is characterized in terms of a damage index (m-

value) defined as the ratio of the physical length of the damage to the wavelength of the A0

Lamb wave mode Then the range of possible damage location length and severity are esti-

mated with constraints This procedure can be accomplished without prior baseline data To

realize the proposed technique the matching pursuit technique is modified for reduction of

the error by considering the dispersive characteristic of Lamb wave The matching pursuit

method estimates the arrival time of each wave packets accurately even two wave packets are

overlapped The numerical simulations and experimental tests including Teflon tape inserted

and impact-induced damage are investigated for the feasibility of the proposed technique

Moreover the effectiveness of the proposed technique under changing temperature is exper-

imentally verified

- 31 -

62 Suggestion for further study

The limitations of this study and the possible improvement and further studies are listed as

follow

1 Measurement of impact-induced damage length using conventional NDT technique

Additional experimental verification for impact-induced damage characterization is

needed by measuring the real length of damage using conventional NDT technique

such as ultrasonic scan or infrared camera

2 Limitation of both side collocated PZT installation In this study excitation PZTs are

installed both side and collocated for decomposing A0 Lamb wave mode However it

can produce decomposition error due to imperfection of PZT location and their installa-

tion Moreover in many cases assess to both surface of structures is not allowed such

as aircraft applications To improve this limitation a reference-free damage characteri-

zation using single side PZTs will be considered

3 Complex structure application In this study damage is characterized in simple plate-

like structure When the monitored structure becomes complex the response signals al-

so becomes complex due to its geometry and boundary Thus the investigations for

complex structure application such as stiffener spar and curvature are needed

- 32 -

REFERENCES

[1] Cawley P and Alleyne D (1996) ldquoThe use of Lamb waves for the long range inspection of large structuresrdquo Ultrasonics 34 pp287-290

[2] Ghosh T Kundu T and Karpur P (1998) ldquoEfficient use of Lamb waves for detecting defects in large platesrdquo Ultrasonics 36 pp791-801

[3] Guirguitui V (2003) ldquoLamb wave generation with piezoelectric wafer active sensors for structural health monitoringrdquo SPIE Conference on Smart Structures and Materials San Diego CA 5391 pp111-422

[4] Ragvahan A and Cesnik CE S (2004) ldquoModeling of piezoelectric-based Lamb wave generation and sensing for structural health monitoringrdquo SPIE Conference on Smart Struc-tures and Materials San Diego CA 5391 pp419-430

[5] Rose JL Pilarski A and Dirit JJ (1993) ldquoAn approach to guided wave mode selection for inspection of laminated platerdquo Journal of Reinforced Plastics and Composites 12(5) pp536-544

[6] Guo N and Cawley P (1993) ldquoThe interaction of Lamb waves with delaminations in composite laminatesrdquo Journal of Acoustical Society of America 94(4) pp 2240-2246

[7] Valdez SHD and Soutis C (2002) ldquoReal-time nondestructive evaluation of fiber com-posite laminates using low-frequency Lamb wavesrdquo Journal of Acoustical Society of Amer-ica 111(5) pp 2026-2033

[8] Yang M and Qiao P (2005) ldquoModeling and experimental detection of damage in virous materials using the pulse-echo method and piezoelectric sensoractuatorsrdquo Smart Materials and Structures 14 pp1083-1100

[9] Lestari W and Qiao P (2005) ldquoApplication of wave propagation analysis for damage identification in composite laminated beamsrdquo Journal of Composite Materials 39(22) pp1967-1984

[10] Kessler S S Spearing S M and Soutis C (2002) ldquoDamage detection in composite ma-terials using Lamb wave methodsrdquo Smart Materials and Structures 11 pp269-278

[11] Sohn H Park G Wail J R Limback N P and Farrar C R (2004) ldquoWavelet-based active sensing for delamination detection in composite structuresrdquo Smart Materials and Structures 13 pp 153-160

[12] Petculescu G Krishnaswamy S and Achenback J D (2008) ldquoGroup delay measure-ments using modally selective Lamb wave transducers for detection and sizing of delamina-tions in compositesrdquo Smart Materials and Structures 17

[13] Sohn H Park H W Law K H and Farrar C R (2007) ldquoDamage detection in compo-

- 33 -

site plates by using an enhanced time reversal methodrdquo Journal of Aerospace Engineering (ASCE) pp 141-151

[14] Sohn H (2007) ldquoEffects of environmental and operational variability on structural health monitoringrdquo Philosophical Transactions of the Royal Society Ardquo 365 pp539-560

[15] Rytter A (1993) ldquoVibration based inspection of civil engineering structuresrdquo PhD The-sis Department of Building Technology and Structural Engineering University of Aalborg Aalborg Denmark

[16] Hayashi T and Kawashima K (2002) ldquoMultiple reflections of Lamb waves at a delami-nationrdquo Ultrasonics 40 pp193-197

[17] Wang C H and Rose L R F (2003) ldquoWave reflection and transmission in beams con-taining delamination and inhomogeneityrdquo Journal of Sound and Vibration 264 pp 851-872

[18] Ramadas C Balasubramaniam K Joshi M and Krishnamurthy C V (2009) ldquoInterac-tion of the primary anti-symmetric Lamb mode (Ao) with symmetric delaminations numer-ical and experimental studiesrdquo Smart Materials and Structures 18

[19] Ramadas C Balasubramaniam K Joshi M and Krishnamurthy C V (2010) ldquoInterac-tion of guided Lamb waves with an asymmetrically located delamination in a laminated composite platerdquo Smart Materials and Structures 19

[20] Wilcox P Lowe M and Cawely P (2001) ldquoThe effect of dispersion on long-range in-spection using ultrasonic guided wavesrdquo NDTampE International 34 pp1-9

[21] Mallat S G and Zhang Z (1993) ldquoMatching pursuits with time-frequency dictionariesrdquo IEEE Tansactions on Signal Processing 41(12) pp3397-3415

[22] Zhang G Zhang S and Wang Y (2000) ldquoApplication of adaptive time-frequency de-composition in ultrasonic NDE of highly scattering materialsrdquo Ultrasonics 38 pp 961-964

[23] Hong J C Sun K H and Kim Y Y (2005) ldquoThe matching pursuit approach based on the modulated Gaussian pulse for efficient guided-wave damage inspectionrdquo Smart Mate-rials and Structures 14 pp548-560

[24] Herakovich C T (1997) Mechanics of Fibrous Composites Wiley 480 pages

[25] Reddy J N (1985) ldquoA review of the literature on finite-element modeling of laminated composite platesrdquo The Shock and Vibration Digest 17(4) pp 3-8

[26] Reddy J N (2003) Mechanics of Laminated Composite Plates and Shells Theory and Analysis CRC Press 856 pages

[27] Palacz M Krawczuk M and Ostachowicz W (2005) ldquoThe spectral finite element mod-el for analysis of flexure-shear coupled wave propagation Part 1 laminated multilayer

- 34 -

composite beamrdquo Composite Structures 68 pp37-44

[28] Zhang Y C and Harding J (1990) ldquoA numerical micromechanics analysis of the me-chanical properties of a plain weave compositerdquo Computer amp Structures 36(5) pp 839-844

[29] Nemat-Nasser S and Hori M (1999) Micromechanics Overall Properties of Heteroge-neous Materials Elsevier 810 pages

[30] Scida D Aboura Z Benzeggagh M L and Bocherens E (1999) ldquoA micromechanics model for 3D elasticity and failure of woven-fibre composite materialrdquo Composites Science and Technology 59 pp 505-517

[31] Thomas Y and Wu X H (1997) ldquoA multiscale finite element method for elliptic prob-lems in composite materials and porous mediardquo Journal of Computational Physics 134 pp 169-189

[32] Lee K H Moorthy S and Ghosh S (1999) ldquoMultiple scale computational model for damage in composite materialsrdquo Computer Methods in Applied Mechanics and Engineer-ing 172 pp 175-201

[33] Namilae S and Chandra N (2005) ldquoMultiscale model to study the effect of interfaces in carbon nanotube-based compositesrdquo Journal of Engineering Materials and Technology (ASME) 127 pp222-232

[34] Adams D E (2007) Health monitoring of Structural Materials and Components Wiley 460 pages

[35] Gopalakrishnan S (2009) ldquoModeling Aspects in Finite Elementrdquo Encyclopedia of Struc-tural Health Monitoring 2 pp791-809

[36] Palacz M Krawczuk M and Ostachowicz W (2005) ldquoThe spectral finite element mod-el for analysis of flexure-shear coupled wave propagation Part 2 Deminated multilayer composite beamrdquo Composite Structures 68 pp45-51

[37] Baik J M and Thopson R B (1984) ldquoUltrasonic scattering from imperfect interfaces A quasi-static modelrdquo Journal of Nondestructive Evaluation 4(3) pp177-196

[38] Luo H and Hanagud S (2000) ldquoDynamics of delamincated beamsrdquo International Jour-nal of Solids and Structures 37 pp1501-1519

[39] Kim S B and Sohn H (2007) ldquoInstantaneous reference-free crack detection based on polarization characteristics of piezoelectric materialsrdquo Smart Materials and Structures 16 pp2375-2387

[40] An Y K and Sohn H (2010) ldquoInstantaneous crack detection under varying temperature and static loading conditionsrdquo The Special Issue of Structural Control and Health Monitor-ing 17(7) pp 730-741

- 35 -

[41] Lee S J and Sohn H (2006) ldquoActive self-sensing scheme development for structural health monitoringrdquo Smart Materials and Structures 15(6) pp 1734-1746

- 36 -

요 약 문

복합재료 손상의 무기저 진단 및 정량화

복합재료는 재료의 경량성 내구성 고강도의 특성으로 인하여 항공기에서부터

교량에까지 널리 사용되고 있다 하지만 외부의 충격으로 인한 복합재료 내부의

손상은 재료의 강도를 감소시켜 구조물의 안정성을 저하시키는 요인이 된다 하지만

복합재료의 손상은 재료 내부에서 발생하여 육안으로는 식별하기도 힘들며 표면에서

보이는 것만으로 내부의 손상 정도를 판단하기는 불가능하다는 단점을 가지고 있다

본 연구에서는 손상 내부에서 반사로 인해 발생하는 파를 이용하여 기저자료를

사용하지 않고 특정 시점에서 얻은 자료만으로 복합재료의 손상을 진단하고 정량화

할 수 있는 기법을 제안한다 제안된 기법의 적용성을 검증하기 위하여 유한요소법을

이용한 수치해석을 수행하였고 Teflon tape 이 삽입된 시편과 실제 충격하중으로

인한 손상을 입은 시편을 이용한 실험을 진행하였다 수치해석 및 실험 결과 제안된

기법이 복합재료의 손상을 성공적으로 진단 및 정량화할 수 있음을 확인하였다 또한

다양한 외부 온도 환경에서 제안된 기법의 환경영향 실험을 수행하였고 다양한

환경에서 손상의 진단 및 정량화가 성공적으로 이루어짐을 검증하였다

핵심어 무기저 손상진단 및 정량화 압전센서 복합재료 유한요소법 Matching pursuit

method

- 37 -

감 사 의 글

기대와 두려움을 안고 시작한 석사과정이 어느덧 2 년 이라는 세월 속에서 추억으로 자리잡아

가고 있습니다 학부생 기간을 포함한 지난 3 년 동안 SHM 이라는 새로운 분야를 접하고 연구를

수행 할 수 있도록 도와주신 고마운 분들이 있어 그 감사를 표하고자 합니다

우선 저의 많은 부족한 점을 일깨워 주시고 연구에 대한 뜻을 더욱 확고히 세울 수 있게 해

주신 지도교수님이신 손 훈 교수님께 깊은 감사의 뜻을 전합니다 짧은 기간이었지만 교수님의

도움으로 많은 것을 배우고 느끼고 깨달을 수 있었습니다 또한 본 연구를 수행할 수 있도록 지

원해주신 국토해양부 (U-City 석 박사과정 지원사업)에 깊은 감사 드립니다

때로 힘들고 지칠 때 항상 저를 믿고 응원해 주시는 가족이 있었기에 지금의 글을 쓸 수 있었

습니다 또한 지금 연구실에서 함께 생활하고 있는 윤규형 민구형 철민이형 현석이형 진열이형

호민이형 병진이 선혜누나와 성대에 계신 창길이형 미국에 계신 의재형과 Debaditiya 에게도 감

사의 말씀을 전합니다 그리고 모두 나열할 수는 없지만 카이스트 생활의 활력소가 되어준 카이

스트 건설 및 환경공학과 선후배들 직원 분들과 SOS 멤버들에게도 감사를 표합니다

앞서 거론한 많은 이들의 도움이 없었다면 저 또한 지금 이 자리에 있을 수 없었을 것입니다

많이 부족하기에 더 노력하고 배우고자 했던 초심을 다시 한번 생각하며 더 큰 결실을 맺기 위해

앞으로 나아가겠습니다 도움을 주신 모든 분들께 이 감사의 글을 바칩니다

2011년 6월 카이스트에서 임형진 올림

- 38 -

CURRICULUM VITAE

Hyung Jin Lim

Research Assistant

The Civil and Environmental Engineering Department

Korea Advanced Institute of Science and Technology (KAIST)

291 Daehak-Ro Guseong-Dong Yuseong-Gu

Daejeon Republic of Korea 305-701

Phone (82)+42-350-3665 Fax (82)+42-350-3610

Email limhj87gmailcom

RESEARCH INTERESTS

Damage Detection Localization and Quantification on Composite and Metallic Structures Pattern Recognition Finite Element Modeling (FEM)

EDUCATION

2009 ndash 2011 MS Civil and Environmental Eng Korea Advanced Institute of Science and Technology (KAIST) Korea Rep

2006 ndash 2009 BS Civil and Environmental Eng Korea Advanced Institute of Science and Technology (KAIST) Korea Rep

JOURNAL PUBLICATIONS The corresponding author is underlined 1 Hyung Jin Lim Hoon Sohn ldquoReference-free damage detection localization and quantification

using in compositerdquo in preparation for Journal of the Acoustical Society of America 2011

2 Chul Min Yeum Hoon Sohn Jeong Bum Ihn Hyung Jin Lim ldquoInstantaneous delamination detection in a composite using a dual piezoelectric transducer networkrdquo Submitted to Compo-sites Science and Technology 2011

3 Yun-Kyu An Hyung Jin Lim Min Koo Kim Jin Yeol Yang Hoon Sohn Chang Guen Lee ldquoInvestigation of applicability of local reference-free damage detection techniques to in-situ

- 39 -

bridgesrdquo Submitted to ASCE Journal of Structural Engineering 2011

4 Hyung Jin Lim Min Koo Kim Hoon Sohn Chan Yik Park ldquoImpedance based damage detec-tion under varying temperature and loading conditionsrdquo Submitted to NDTampE International 2011

5 Hyung Jin Lim Hoon Sohn ldquoQuantitative estimation of transmitted and reflected lamb waves at discontinuityrdquo Journal of the Korean Society for Nondestructive Testing Vol30 No 4 pp1-8 2010

CONFERENCE PROCEEDINGS

The corresponding author is underlined 1 Hyung Jin Lim and Hoon Sohn ldquoReference-free damage detection localization and quantifi-

cation in compositesrdquo ASCE Engineering Mechanics Institute Conference Boston June 2-4 2011

2 Hyung Jin Lim and Hoon Sohn ldquoReference-free delamination quantification within multi-layer composite materialsrdquo KSNT Spring Conference May 26-27 Gyeong-Ju Korea 2011

3 Min Koo Kim Hyun Jin Lim Hoon Sohn Chan Yik Park ldquoImpedance-based bolt loosening detection under varying temperature and loadingrdquo Asian Pacific Workshop on Structural Health Monitoring Tokyo Japan November 30-December 2 2010

4 Hyung Jin Lim Chul Min Yeum and Hoon Sohn ldquoModeling of Impact-induced Delamination in a Multilayer Composite Platerdquo The 23rd KKCNN Symposium on Civil Engineering Taipei (Taiwan) November 13-15 2010

5 Yun-Kyu An Hyung Jin Lim and Hoon Sohn ldquoTemperature independent Delamination Detec-tion using data normalizationrdquo The 3rd International Conference on Multi-Functional Materi-als and Structures Chonbuk National Univ Jeonju Korea September 14-18 2010

6 Hoon Sohn Yun-Kyu An Chang Gil Lee Min Koo Kim Hyung Jin Lim ldquoApplication of online NDT techniques to decommissioned bridge testingrdquo the 5th European Workshop on Structural Health Monitoring Sorrento Italy June 29-July 02 2010

7 Min Koo Kim Hyung Jin Lim Hoon Sohn Chan Yik Park ldquoImpedance based damage diag-nosis of a complex composite aircraft wing under changing loading and temperature condi-tionsrdquo the 5th European Workshop on Structural Health Monitoring Sorrento Italy June 29-July 02 2010

8 Yun-Kyu An Min Koo Kim Hyung Jin Lim and Hoon Sohn ldquoValidations of reference-free crack detection techniques through a decommissioned bridge testrdquo 2010 COSEIK Annual Con-ference Jeju Korea April 8-9 2010

- 40 -

PARTICIPANT PROJECTS

1 Development of Smart Steel Members for Online Steel Structure Monitoring POSCO Corporation (Funded 41000000 KRW (41000 USD) for 010511 to 123111)

2 Development of Enabling Technologies for Structural Health Monitoring Korean Agency for Defense Development (Funded 240000000 KRW (240000 USD) out of total 3200000000 KRW (3200000 USD) for 060108-053111)

3 Development of On-board SHM Technologies for Composite Air Vehicles The Boeing Company (Funded 400000 USD (400000000 KRW) for 080108 to 073111)

EDUCATIONAL EXPERIENCES

Fall 2009

Teaching Assistant CE207 ldquoElementary Structural Engineering and Laboratoryrdquo KAIST 3 unit undergraduate course

EXPERIMENTAL EXPERIENTCES

2010 Monitoring of Real Aircraft Composite Wing Segment Korea Agency for Defense De-velopment (ADD) Daejeon Korea Real time monitoring of composite air vehicles fund-ed by Korea Agency for Defense Development (ADD)

2009-2011

Field Monitoring of Real Bridge Structures Gimpo (Ramp G bridge) amp Icheon (Sam-seung bridge) Korea US-Korea collaborative research for bridge monitoring testbeds

HONORS amp AWARDS

Scholarships for outstanding students granted KAIST (2007-2009) The three prominent students in the civil and environmental engineering department of KAIST were selected as recipients This scholarship provided 2000$ per year for three years

- 41 -

Page 34: 복합재료 손상의 무기저 진단 및 정량화 - KAISTssslab.kaist.ac.kr/article/pdf/Thesis_2.pdf · 1 Declaration of Ethical Conduct in Research: I, as a graduate student

PZT Actuator

PZT Sensor

AWG

DIG

DAQ System Specimen

SensingActuatingMUX

1

23

4

1 Sending signal to MUX

4 Sending signal to DIG

Self-SensingCircuit

18 Unit mm

10

(a)

(b)

Fig 9 Overall experimental setup (a) Data acquisition (DAQ) system and (b) Installed circu-lar PZT transducer on the specimen with its dimension

513 Impact test

The impact-induced damage was introduced to the specimen using the impact test equip-

ment presented in Fig 10 A 5kg mass with 10mm tip was dropped from 015m height onto

the same location of the specimen The signals were obtained at each number of impacts

Since the reflected wave is not detected at the first impact case the data obtained from the

second and third impact were utilized for the analysis

10

Unit mm (a)

(b)

Fig 10 (a) Impact test equipment for introducing impact-induced damage and (b) Tip of the impact tester with 10mm diameter

- 23 -

514 Temperature effect test

In reality structures are subjected to changing environmental conditions such as tempera-

ture that affect measured signal and cause false-alarm Thus the robustness of the proposed

damage characterization technique under temperature variation in investigated (Fig 11) The

specimens were placed inside a temperature chamber and a thermocouple was installed on

the specimen to measure temperature of surface of the specimen The signal was obtained

under three temperature cases (0 20 and 50degC) and the humidity was kept at 30

Temperature Chamber

Test Specimen

Fig 11 Test specimen in temperature chamber

- 24 -

52 Teflon inserted damage

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Reflected wave from damage

(a)

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Transmitted WaveReflected Wave

Detected reflected wave

(b)

Fig 12 A0 Lamb wave mode obtained using pitch-catch method from the Teflon inserted damage specimen (a) The reflected wave from damage is shown (b) The Teflon inserted damage is detected by capturing the reflected wave

Fig 12 shows the obtained pitch-catch signal from the Teflon inserted damage specimen

and damage diagnosis result The Teflon inserted damage is detected successfully by captur-

ing the reflected wave Table IV presents the result of the Teflon inserted damage characteri-

zation 119907119907119894119894 is measured as 1213ms and the variation is assumed to plusmn 5 of the normal val-

ue (1152~1273ms) The results indicate that the proposed damage characterization approach

works well in real application

TABLE IV Damage characterization for the Teflon inserted damage specimen

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

Exact 635 25

Method 1 123 60~75 73~338 298~1220

Method 2 123 619~662 229~310 934~1188

Method 3 123 631 284 1161

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1213 ms

- 25 -

53 Impact-induced damage

11 115 12 125 13-4

-2

0

2

4

Time (ms)

Vol

tage

(mV

)

IntactImpact2Impact3

Reflected wave from damage

(a) Pitch-catch signal

11 115 12 125 13 135

-05

0

05

1

Time (ms)

Volta

ge (m

V)

IntactImpact2Impact3

Boundary reflection

Reflected wave from damage

(b) Pulse-echo signal

Fig 13 The signal obtained from the test specimen subjected to impact The impact was in-troduced at the same location of the specimen and the signal was obtained after each impacts As the number of impact is increased the transmitted A0 Lamb wave mode is delayed and attenuated

The obtained signals from impact-induced damage specimen are presented in Fig 13 It is

observed that as the number of impact is increased the transmitted A0 Lamb wave mode is

more delayed because the wave velocity in the damage area is reduced Fig 14 shows the

false-alarm test result whether the proposed damage detection algorithm works even in pris-

tine condition of the test specimen It is shown that no false alarm is occurred because no ad-

ditional reflected wave is captured

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Transmitted Wave

(a) Transmitted wave

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Reflected Wave

No reflected wave

(b) Reflected wave

Fig 14 False alarm test for the proposed damage detection algorithm in pristine condition of the specimen Reflected wave from damage is not captured false alarm is not occurred

- 26 -

TABLE V Damage characterization for the impact-induced damage specimen

of Impacts m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

2

(1470 J)

Method 1 080 65~90 75~246 470~1157

Method 2 080 801~885 114~245 711~1134

Method 3 080 843 179 1122

3

(2205 J)

Method 1 148 50~100 06~380 21~1134

Method 2 148 758~838 249~368 843~1125

Method 3 148 798 308 1044

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1213 ms

Table V presents the characterization results of the impact-induced damage case Here the

measured 119907119907119894119894 and the assumed variation are same as the artificial damage case It is observed

that as the number of impact is increased 119889119889 is increased while 119907119907119889119889 is reduced Thus it is

validated that the proposed damage characterization approach estimates the increment of the

damage well

54 Temperature effect test

541 Teflon inserted damage

11 115 12 125 13-10

-5

0

5

10

Time (ms)

Vol

tage

(mV

)

0 oC20 oC50 oC

(a) Pitch-catch signal

11 115 12 125 13 135

-1

0

1

2

3

Time(ms)

Volta

ge (m

V)

0 oC20 oC50 oC

(b) Pulse-echo signal

Fig 15 Signals obtained from the Teflon inserted damage specimen under temperature varia-tion Time delay and amplitude variation are occurred

Fig 15 shows the obtained signals from the Teflon inserted damage specimen subjected to

- 27 -

temperature variation Time delay and amplitude variations are occurred there is no addition-

al damage is introduced The analyzed results of the obtained signals are reported in Table VI

The measured 119907119907119894119894 is 1247 ms at 0 degC 1213 ms at 20 degC and 1168 ms at 50 degC respectively

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms (1152~1273 ms)

The estimated results are also changed under temperature variations the fluctuation of esti-

mated range and exact value is small

TABLE VI Temperature effect test for the Teflon inserted damage specimen

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

0 degC

Exact 635 25

Method 1 113 60~70 79~268 352~1291

Method 2 113 560~619 233~344 1034~1172

Method 3 113 606 257 1141

20 degC

Exact 635 25

Method 1 123 60~75 73~338 298~1220

Method 2 123 619~662 229~310 934~1188

Method 3 123 631 284 1161

50 degC

Exact 635 25

Method 1 137 60~75 72~328 263~1228

Method 2 137 622~687 183~299 666~1086

Method 3 137 642 264 960

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1247 ms at 0 degC 1213 ms at 20 degC and 1168 ms at 50 degC

- 28 -

542 Impact-induced damage

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Vol

tage

(mV

)

Transmitted WaveReflected Wave

No reflected wave

(a) 0 degC

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Transmitted WaveReflected Wave

No reflected wave

(b) 50 degC

Fig 16 False alarm test under temperature variations It is shown that proposed damage char-acterization algorithm is robust under temperature variations

The false alarm test in pristine condition of the impact-induced damage specimen under

temperature variation was investigated (Fig 16) It is validated that the proposed damage di-

agnosis algorithm is robust under temperature variations Fig 17 shows the obtained signals

from the impact-induced damage (impacted 3 times) specimen under temperature variation

The analyzed results of the obtained signals are reported in Table VII The measured 119907119907119894119894 in

each temperature and the assumed variation are same as the artificial damage case

11 115 12 125 13-5

0

5

Time (ms)

Volta

ge (m

V)

0 oC20 oC50 oC

(a) Pitch-catch signal

11 115 12 125 13 135

-1

0

1

2

3

Time(ms)

Vol

tage

(mV

)

0 oC20 oC50 oC

(b) Pulse-echo signal

Fig 17 Signals obtained from impact-induced damage specimen (Impact 3) under tempera-ture variation

- 29 -

TABLE VII Temperature effect test for the impact induced damage specimen (Impact 3 case)

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

0 degC

Method 1 135 75~90 72~310 267~1148

Method 2 135 705~779 264~381 978~1126

Method 3 135 763 289 1071

20 degC

Method 1 148 50~100 06~380 21~1134

Method 2 148 758~838 249~368 843~1125

Method 3 148 798 308 1044

50 degC

Method 1 153 75~100 53~415 174~1227

Method 2 153 804~889 214~336 702~1103

Method 3 153 815 320 1050

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1247 ms at 0 degC 1213 ms at 20 degC and 1168 ms at 50 degC

- 30 -

CHAPTER 6 CONCLUDING REMARKS

61 Executive summary

In this study the reference-free damage detection localization and quantification tech-

nique in a multi layered composite plate is proposed Additional reflected A0 Lamb wave

mode from the damage due to multiple reflections in damage area is used as a feature for

characterizing the damage First damage is characterized in terms of a damage index (m-

value) defined as the ratio of the physical length of the damage to the wavelength of the A0

Lamb wave mode Then the range of possible damage location length and severity are esti-

mated with constraints This procedure can be accomplished without prior baseline data To

realize the proposed technique the matching pursuit technique is modified for reduction of

the error by considering the dispersive characteristic of Lamb wave The matching pursuit

method estimates the arrival time of each wave packets accurately even two wave packets are

overlapped The numerical simulations and experimental tests including Teflon tape inserted

and impact-induced damage are investigated for the feasibility of the proposed technique

Moreover the effectiveness of the proposed technique under changing temperature is exper-

imentally verified

- 31 -

62 Suggestion for further study

The limitations of this study and the possible improvement and further studies are listed as

follow

1 Measurement of impact-induced damage length using conventional NDT technique

Additional experimental verification for impact-induced damage characterization is

needed by measuring the real length of damage using conventional NDT technique

such as ultrasonic scan or infrared camera

2 Limitation of both side collocated PZT installation In this study excitation PZTs are

installed both side and collocated for decomposing A0 Lamb wave mode However it

can produce decomposition error due to imperfection of PZT location and their installa-

tion Moreover in many cases assess to both surface of structures is not allowed such

as aircraft applications To improve this limitation a reference-free damage characteri-

zation using single side PZTs will be considered

3 Complex structure application In this study damage is characterized in simple plate-

like structure When the monitored structure becomes complex the response signals al-

so becomes complex due to its geometry and boundary Thus the investigations for

complex structure application such as stiffener spar and curvature are needed

- 32 -

REFERENCES

[1] Cawley P and Alleyne D (1996) ldquoThe use of Lamb waves for the long range inspection of large structuresrdquo Ultrasonics 34 pp287-290

[2] Ghosh T Kundu T and Karpur P (1998) ldquoEfficient use of Lamb waves for detecting defects in large platesrdquo Ultrasonics 36 pp791-801

[3] Guirguitui V (2003) ldquoLamb wave generation with piezoelectric wafer active sensors for structural health monitoringrdquo SPIE Conference on Smart Structures and Materials San Diego CA 5391 pp111-422

[4] Ragvahan A and Cesnik CE S (2004) ldquoModeling of piezoelectric-based Lamb wave generation and sensing for structural health monitoringrdquo SPIE Conference on Smart Struc-tures and Materials San Diego CA 5391 pp419-430

[5] Rose JL Pilarski A and Dirit JJ (1993) ldquoAn approach to guided wave mode selection for inspection of laminated platerdquo Journal of Reinforced Plastics and Composites 12(5) pp536-544

[6] Guo N and Cawley P (1993) ldquoThe interaction of Lamb waves with delaminations in composite laminatesrdquo Journal of Acoustical Society of America 94(4) pp 2240-2246

[7] Valdez SHD and Soutis C (2002) ldquoReal-time nondestructive evaluation of fiber com-posite laminates using low-frequency Lamb wavesrdquo Journal of Acoustical Society of Amer-ica 111(5) pp 2026-2033

[8] Yang M and Qiao P (2005) ldquoModeling and experimental detection of damage in virous materials using the pulse-echo method and piezoelectric sensoractuatorsrdquo Smart Materials and Structures 14 pp1083-1100

[9] Lestari W and Qiao P (2005) ldquoApplication of wave propagation analysis for damage identification in composite laminated beamsrdquo Journal of Composite Materials 39(22) pp1967-1984

[10] Kessler S S Spearing S M and Soutis C (2002) ldquoDamage detection in composite ma-terials using Lamb wave methodsrdquo Smart Materials and Structures 11 pp269-278

[11] Sohn H Park G Wail J R Limback N P and Farrar C R (2004) ldquoWavelet-based active sensing for delamination detection in composite structuresrdquo Smart Materials and Structures 13 pp 153-160

[12] Petculescu G Krishnaswamy S and Achenback J D (2008) ldquoGroup delay measure-ments using modally selective Lamb wave transducers for detection and sizing of delamina-tions in compositesrdquo Smart Materials and Structures 17

[13] Sohn H Park H W Law K H and Farrar C R (2007) ldquoDamage detection in compo-

- 33 -

site plates by using an enhanced time reversal methodrdquo Journal of Aerospace Engineering (ASCE) pp 141-151

[14] Sohn H (2007) ldquoEffects of environmental and operational variability on structural health monitoringrdquo Philosophical Transactions of the Royal Society Ardquo 365 pp539-560

[15] Rytter A (1993) ldquoVibration based inspection of civil engineering structuresrdquo PhD The-sis Department of Building Technology and Structural Engineering University of Aalborg Aalborg Denmark

[16] Hayashi T and Kawashima K (2002) ldquoMultiple reflections of Lamb waves at a delami-nationrdquo Ultrasonics 40 pp193-197

[17] Wang C H and Rose L R F (2003) ldquoWave reflection and transmission in beams con-taining delamination and inhomogeneityrdquo Journal of Sound and Vibration 264 pp 851-872

[18] Ramadas C Balasubramaniam K Joshi M and Krishnamurthy C V (2009) ldquoInterac-tion of the primary anti-symmetric Lamb mode (Ao) with symmetric delaminations numer-ical and experimental studiesrdquo Smart Materials and Structures 18

[19] Ramadas C Balasubramaniam K Joshi M and Krishnamurthy C V (2010) ldquoInterac-tion of guided Lamb waves with an asymmetrically located delamination in a laminated composite platerdquo Smart Materials and Structures 19

[20] Wilcox P Lowe M and Cawely P (2001) ldquoThe effect of dispersion on long-range in-spection using ultrasonic guided wavesrdquo NDTampE International 34 pp1-9

[21] Mallat S G and Zhang Z (1993) ldquoMatching pursuits with time-frequency dictionariesrdquo IEEE Tansactions on Signal Processing 41(12) pp3397-3415

[22] Zhang G Zhang S and Wang Y (2000) ldquoApplication of adaptive time-frequency de-composition in ultrasonic NDE of highly scattering materialsrdquo Ultrasonics 38 pp 961-964

[23] Hong J C Sun K H and Kim Y Y (2005) ldquoThe matching pursuit approach based on the modulated Gaussian pulse for efficient guided-wave damage inspectionrdquo Smart Mate-rials and Structures 14 pp548-560

[24] Herakovich C T (1997) Mechanics of Fibrous Composites Wiley 480 pages

[25] Reddy J N (1985) ldquoA review of the literature on finite-element modeling of laminated composite platesrdquo The Shock and Vibration Digest 17(4) pp 3-8

[26] Reddy J N (2003) Mechanics of Laminated Composite Plates and Shells Theory and Analysis CRC Press 856 pages

[27] Palacz M Krawczuk M and Ostachowicz W (2005) ldquoThe spectral finite element mod-el for analysis of flexure-shear coupled wave propagation Part 1 laminated multilayer

- 34 -

composite beamrdquo Composite Structures 68 pp37-44

[28] Zhang Y C and Harding J (1990) ldquoA numerical micromechanics analysis of the me-chanical properties of a plain weave compositerdquo Computer amp Structures 36(5) pp 839-844

[29] Nemat-Nasser S and Hori M (1999) Micromechanics Overall Properties of Heteroge-neous Materials Elsevier 810 pages

[30] Scida D Aboura Z Benzeggagh M L and Bocherens E (1999) ldquoA micromechanics model for 3D elasticity and failure of woven-fibre composite materialrdquo Composites Science and Technology 59 pp 505-517

[31] Thomas Y and Wu X H (1997) ldquoA multiscale finite element method for elliptic prob-lems in composite materials and porous mediardquo Journal of Computational Physics 134 pp 169-189

[32] Lee K H Moorthy S and Ghosh S (1999) ldquoMultiple scale computational model for damage in composite materialsrdquo Computer Methods in Applied Mechanics and Engineer-ing 172 pp 175-201

[33] Namilae S and Chandra N (2005) ldquoMultiscale model to study the effect of interfaces in carbon nanotube-based compositesrdquo Journal of Engineering Materials and Technology (ASME) 127 pp222-232

[34] Adams D E (2007) Health monitoring of Structural Materials and Components Wiley 460 pages

[35] Gopalakrishnan S (2009) ldquoModeling Aspects in Finite Elementrdquo Encyclopedia of Struc-tural Health Monitoring 2 pp791-809

[36] Palacz M Krawczuk M and Ostachowicz W (2005) ldquoThe spectral finite element mod-el for analysis of flexure-shear coupled wave propagation Part 2 Deminated multilayer composite beamrdquo Composite Structures 68 pp45-51

[37] Baik J M and Thopson R B (1984) ldquoUltrasonic scattering from imperfect interfaces A quasi-static modelrdquo Journal of Nondestructive Evaluation 4(3) pp177-196

[38] Luo H and Hanagud S (2000) ldquoDynamics of delamincated beamsrdquo International Jour-nal of Solids and Structures 37 pp1501-1519

[39] Kim S B and Sohn H (2007) ldquoInstantaneous reference-free crack detection based on polarization characteristics of piezoelectric materialsrdquo Smart Materials and Structures 16 pp2375-2387

[40] An Y K and Sohn H (2010) ldquoInstantaneous crack detection under varying temperature and static loading conditionsrdquo The Special Issue of Structural Control and Health Monitor-ing 17(7) pp 730-741

- 35 -

[41] Lee S J and Sohn H (2006) ldquoActive self-sensing scheme development for structural health monitoringrdquo Smart Materials and Structures 15(6) pp 1734-1746

- 36 -

요 약 문

복합재료 손상의 무기저 진단 및 정량화

복합재료는 재료의 경량성 내구성 고강도의 특성으로 인하여 항공기에서부터

교량에까지 널리 사용되고 있다 하지만 외부의 충격으로 인한 복합재료 내부의

손상은 재료의 강도를 감소시켜 구조물의 안정성을 저하시키는 요인이 된다 하지만

복합재료의 손상은 재료 내부에서 발생하여 육안으로는 식별하기도 힘들며 표면에서

보이는 것만으로 내부의 손상 정도를 판단하기는 불가능하다는 단점을 가지고 있다

본 연구에서는 손상 내부에서 반사로 인해 발생하는 파를 이용하여 기저자료를

사용하지 않고 특정 시점에서 얻은 자료만으로 복합재료의 손상을 진단하고 정량화

할 수 있는 기법을 제안한다 제안된 기법의 적용성을 검증하기 위하여 유한요소법을

이용한 수치해석을 수행하였고 Teflon tape 이 삽입된 시편과 실제 충격하중으로

인한 손상을 입은 시편을 이용한 실험을 진행하였다 수치해석 및 실험 결과 제안된

기법이 복합재료의 손상을 성공적으로 진단 및 정량화할 수 있음을 확인하였다 또한

다양한 외부 온도 환경에서 제안된 기법의 환경영향 실험을 수행하였고 다양한

환경에서 손상의 진단 및 정량화가 성공적으로 이루어짐을 검증하였다

핵심어 무기저 손상진단 및 정량화 압전센서 복합재료 유한요소법 Matching pursuit

method

- 37 -

감 사 의 글

기대와 두려움을 안고 시작한 석사과정이 어느덧 2 년 이라는 세월 속에서 추억으로 자리잡아

가고 있습니다 학부생 기간을 포함한 지난 3 년 동안 SHM 이라는 새로운 분야를 접하고 연구를

수행 할 수 있도록 도와주신 고마운 분들이 있어 그 감사를 표하고자 합니다

우선 저의 많은 부족한 점을 일깨워 주시고 연구에 대한 뜻을 더욱 확고히 세울 수 있게 해

주신 지도교수님이신 손 훈 교수님께 깊은 감사의 뜻을 전합니다 짧은 기간이었지만 교수님의

도움으로 많은 것을 배우고 느끼고 깨달을 수 있었습니다 또한 본 연구를 수행할 수 있도록 지

원해주신 국토해양부 (U-City 석 박사과정 지원사업)에 깊은 감사 드립니다

때로 힘들고 지칠 때 항상 저를 믿고 응원해 주시는 가족이 있었기에 지금의 글을 쓸 수 있었

습니다 또한 지금 연구실에서 함께 생활하고 있는 윤규형 민구형 철민이형 현석이형 진열이형

호민이형 병진이 선혜누나와 성대에 계신 창길이형 미국에 계신 의재형과 Debaditiya 에게도 감

사의 말씀을 전합니다 그리고 모두 나열할 수는 없지만 카이스트 생활의 활력소가 되어준 카이

스트 건설 및 환경공학과 선후배들 직원 분들과 SOS 멤버들에게도 감사를 표합니다

앞서 거론한 많은 이들의 도움이 없었다면 저 또한 지금 이 자리에 있을 수 없었을 것입니다

많이 부족하기에 더 노력하고 배우고자 했던 초심을 다시 한번 생각하며 더 큰 결실을 맺기 위해

앞으로 나아가겠습니다 도움을 주신 모든 분들께 이 감사의 글을 바칩니다

2011년 6월 카이스트에서 임형진 올림

- 38 -

CURRICULUM VITAE

Hyung Jin Lim

Research Assistant

The Civil and Environmental Engineering Department

Korea Advanced Institute of Science and Technology (KAIST)

291 Daehak-Ro Guseong-Dong Yuseong-Gu

Daejeon Republic of Korea 305-701

Phone (82)+42-350-3665 Fax (82)+42-350-3610

Email limhj87gmailcom

RESEARCH INTERESTS

Damage Detection Localization and Quantification on Composite and Metallic Structures Pattern Recognition Finite Element Modeling (FEM)

EDUCATION

2009 ndash 2011 MS Civil and Environmental Eng Korea Advanced Institute of Science and Technology (KAIST) Korea Rep

2006 ndash 2009 BS Civil and Environmental Eng Korea Advanced Institute of Science and Technology (KAIST) Korea Rep

JOURNAL PUBLICATIONS The corresponding author is underlined 1 Hyung Jin Lim Hoon Sohn ldquoReference-free damage detection localization and quantification

using in compositerdquo in preparation for Journal of the Acoustical Society of America 2011

2 Chul Min Yeum Hoon Sohn Jeong Bum Ihn Hyung Jin Lim ldquoInstantaneous delamination detection in a composite using a dual piezoelectric transducer networkrdquo Submitted to Compo-sites Science and Technology 2011

3 Yun-Kyu An Hyung Jin Lim Min Koo Kim Jin Yeol Yang Hoon Sohn Chang Guen Lee ldquoInvestigation of applicability of local reference-free damage detection techniques to in-situ

- 39 -

bridgesrdquo Submitted to ASCE Journal of Structural Engineering 2011

4 Hyung Jin Lim Min Koo Kim Hoon Sohn Chan Yik Park ldquoImpedance based damage detec-tion under varying temperature and loading conditionsrdquo Submitted to NDTampE International 2011

5 Hyung Jin Lim Hoon Sohn ldquoQuantitative estimation of transmitted and reflected lamb waves at discontinuityrdquo Journal of the Korean Society for Nondestructive Testing Vol30 No 4 pp1-8 2010

CONFERENCE PROCEEDINGS

The corresponding author is underlined 1 Hyung Jin Lim and Hoon Sohn ldquoReference-free damage detection localization and quantifi-

cation in compositesrdquo ASCE Engineering Mechanics Institute Conference Boston June 2-4 2011

2 Hyung Jin Lim and Hoon Sohn ldquoReference-free delamination quantification within multi-layer composite materialsrdquo KSNT Spring Conference May 26-27 Gyeong-Ju Korea 2011

3 Min Koo Kim Hyun Jin Lim Hoon Sohn Chan Yik Park ldquoImpedance-based bolt loosening detection under varying temperature and loadingrdquo Asian Pacific Workshop on Structural Health Monitoring Tokyo Japan November 30-December 2 2010

4 Hyung Jin Lim Chul Min Yeum and Hoon Sohn ldquoModeling of Impact-induced Delamination in a Multilayer Composite Platerdquo The 23rd KKCNN Symposium on Civil Engineering Taipei (Taiwan) November 13-15 2010

5 Yun-Kyu An Hyung Jin Lim and Hoon Sohn ldquoTemperature independent Delamination Detec-tion using data normalizationrdquo The 3rd International Conference on Multi-Functional Materi-als and Structures Chonbuk National Univ Jeonju Korea September 14-18 2010

6 Hoon Sohn Yun-Kyu An Chang Gil Lee Min Koo Kim Hyung Jin Lim ldquoApplication of online NDT techniques to decommissioned bridge testingrdquo the 5th European Workshop on Structural Health Monitoring Sorrento Italy June 29-July 02 2010

7 Min Koo Kim Hyung Jin Lim Hoon Sohn Chan Yik Park ldquoImpedance based damage diag-nosis of a complex composite aircraft wing under changing loading and temperature condi-tionsrdquo the 5th European Workshop on Structural Health Monitoring Sorrento Italy June 29-July 02 2010

8 Yun-Kyu An Min Koo Kim Hyung Jin Lim and Hoon Sohn ldquoValidations of reference-free crack detection techniques through a decommissioned bridge testrdquo 2010 COSEIK Annual Con-ference Jeju Korea April 8-9 2010

- 40 -

PARTICIPANT PROJECTS

1 Development of Smart Steel Members for Online Steel Structure Monitoring POSCO Corporation (Funded 41000000 KRW (41000 USD) for 010511 to 123111)

2 Development of Enabling Technologies for Structural Health Monitoring Korean Agency for Defense Development (Funded 240000000 KRW (240000 USD) out of total 3200000000 KRW (3200000 USD) for 060108-053111)

3 Development of On-board SHM Technologies for Composite Air Vehicles The Boeing Company (Funded 400000 USD (400000000 KRW) for 080108 to 073111)

EDUCATIONAL EXPERIENCES

Fall 2009

Teaching Assistant CE207 ldquoElementary Structural Engineering and Laboratoryrdquo KAIST 3 unit undergraduate course

EXPERIMENTAL EXPERIENTCES

2010 Monitoring of Real Aircraft Composite Wing Segment Korea Agency for Defense De-velopment (ADD) Daejeon Korea Real time monitoring of composite air vehicles fund-ed by Korea Agency for Defense Development (ADD)

2009-2011

Field Monitoring of Real Bridge Structures Gimpo (Ramp G bridge) amp Icheon (Sam-seung bridge) Korea US-Korea collaborative research for bridge monitoring testbeds

HONORS amp AWARDS

Scholarships for outstanding students granted KAIST (2007-2009) The three prominent students in the civil and environmental engineering department of KAIST were selected as recipients This scholarship provided 2000$ per year for three years

- 41 -

Page 35: 복합재료 손상의 무기저 진단 및 정량화 - KAISTssslab.kaist.ac.kr/article/pdf/Thesis_2.pdf · 1 Declaration of Ethical Conduct in Research: I, as a graduate student

514 Temperature effect test

In reality structures are subjected to changing environmental conditions such as tempera-

ture that affect measured signal and cause false-alarm Thus the robustness of the proposed

damage characterization technique under temperature variation in investigated (Fig 11) The

specimens were placed inside a temperature chamber and a thermocouple was installed on

the specimen to measure temperature of surface of the specimen The signal was obtained

under three temperature cases (0 20 and 50degC) and the humidity was kept at 30

Temperature Chamber

Test Specimen

Fig 11 Test specimen in temperature chamber

- 24 -

52 Teflon inserted damage

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Reflected wave from damage

(a)

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Transmitted WaveReflected Wave

Detected reflected wave

(b)

Fig 12 A0 Lamb wave mode obtained using pitch-catch method from the Teflon inserted damage specimen (a) The reflected wave from damage is shown (b) The Teflon inserted damage is detected by capturing the reflected wave

Fig 12 shows the obtained pitch-catch signal from the Teflon inserted damage specimen

and damage diagnosis result The Teflon inserted damage is detected successfully by captur-

ing the reflected wave Table IV presents the result of the Teflon inserted damage characteri-

zation 119907119907119894119894 is measured as 1213ms and the variation is assumed to plusmn 5 of the normal val-

ue (1152~1273ms) The results indicate that the proposed damage characterization approach

works well in real application

TABLE IV Damage characterization for the Teflon inserted damage specimen

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

Exact 635 25

Method 1 123 60~75 73~338 298~1220

Method 2 123 619~662 229~310 934~1188

Method 3 123 631 284 1161

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1213 ms

- 25 -

53 Impact-induced damage

11 115 12 125 13-4

-2

0

2

4

Time (ms)

Vol

tage

(mV

)

IntactImpact2Impact3

Reflected wave from damage

(a) Pitch-catch signal

11 115 12 125 13 135

-05

0

05

1

Time (ms)

Volta

ge (m

V)

IntactImpact2Impact3

Boundary reflection

Reflected wave from damage

(b) Pulse-echo signal

Fig 13 The signal obtained from the test specimen subjected to impact The impact was in-troduced at the same location of the specimen and the signal was obtained after each impacts As the number of impact is increased the transmitted A0 Lamb wave mode is delayed and attenuated

The obtained signals from impact-induced damage specimen are presented in Fig 13 It is

observed that as the number of impact is increased the transmitted A0 Lamb wave mode is

more delayed because the wave velocity in the damage area is reduced Fig 14 shows the

false-alarm test result whether the proposed damage detection algorithm works even in pris-

tine condition of the test specimen It is shown that no false alarm is occurred because no ad-

ditional reflected wave is captured

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Transmitted Wave

(a) Transmitted wave

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Reflected Wave

No reflected wave

(b) Reflected wave

Fig 14 False alarm test for the proposed damage detection algorithm in pristine condition of the specimen Reflected wave from damage is not captured false alarm is not occurred

- 26 -

TABLE V Damage characterization for the impact-induced damage specimen

of Impacts m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

2

(1470 J)

Method 1 080 65~90 75~246 470~1157

Method 2 080 801~885 114~245 711~1134

Method 3 080 843 179 1122

3

(2205 J)

Method 1 148 50~100 06~380 21~1134

Method 2 148 758~838 249~368 843~1125

Method 3 148 798 308 1044

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1213 ms

Table V presents the characterization results of the impact-induced damage case Here the

measured 119907119907119894119894 and the assumed variation are same as the artificial damage case It is observed

that as the number of impact is increased 119889119889 is increased while 119907119907119889119889 is reduced Thus it is

validated that the proposed damage characterization approach estimates the increment of the

damage well

54 Temperature effect test

541 Teflon inserted damage

11 115 12 125 13-10

-5

0

5

10

Time (ms)

Vol

tage

(mV

)

0 oC20 oC50 oC

(a) Pitch-catch signal

11 115 12 125 13 135

-1

0

1

2

3

Time(ms)

Volta

ge (m

V)

0 oC20 oC50 oC

(b) Pulse-echo signal

Fig 15 Signals obtained from the Teflon inserted damage specimen under temperature varia-tion Time delay and amplitude variation are occurred

Fig 15 shows the obtained signals from the Teflon inserted damage specimen subjected to

- 27 -

temperature variation Time delay and amplitude variations are occurred there is no addition-

al damage is introduced The analyzed results of the obtained signals are reported in Table VI

The measured 119907119907119894119894 is 1247 ms at 0 degC 1213 ms at 20 degC and 1168 ms at 50 degC respectively

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms (1152~1273 ms)

The estimated results are also changed under temperature variations the fluctuation of esti-

mated range and exact value is small

TABLE VI Temperature effect test for the Teflon inserted damage specimen

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

0 degC

Exact 635 25

Method 1 113 60~70 79~268 352~1291

Method 2 113 560~619 233~344 1034~1172

Method 3 113 606 257 1141

20 degC

Exact 635 25

Method 1 123 60~75 73~338 298~1220

Method 2 123 619~662 229~310 934~1188

Method 3 123 631 284 1161

50 degC

Exact 635 25

Method 1 137 60~75 72~328 263~1228

Method 2 137 622~687 183~299 666~1086

Method 3 137 642 264 960

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1247 ms at 0 degC 1213 ms at 20 degC and 1168 ms at 50 degC

- 28 -

542 Impact-induced damage

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Vol

tage

(mV

)

Transmitted WaveReflected Wave

No reflected wave

(a) 0 degC

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Transmitted WaveReflected Wave

No reflected wave

(b) 50 degC

Fig 16 False alarm test under temperature variations It is shown that proposed damage char-acterization algorithm is robust under temperature variations

The false alarm test in pristine condition of the impact-induced damage specimen under

temperature variation was investigated (Fig 16) It is validated that the proposed damage di-

agnosis algorithm is robust under temperature variations Fig 17 shows the obtained signals

from the impact-induced damage (impacted 3 times) specimen under temperature variation

The analyzed results of the obtained signals are reported in Table VII The measured 119907119907119894119894 in

each temperature and the assumed variation are same as the artificial damage case

11 115 12 125 13-5

0

5

Time (ms)

Volta

ge (m

V)

0 oC20 oC50 oC

(a) Pitch-catch signal

11 115 12 125 13 135

-1

0

1

2

3

Time(ms)

Vol

tage

(mV

)

0 oC20 oC50 oC

(b) Pulse-echo signal

Fig 17 Signals obtained from impact-induced damage specimen (Impact 3) under tempera-ture variation

- 29 -

TABLE VII Temperature effect test for the impact induced damage specimen (Impact 3 case)

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

0 degC

Method 1 135 75~90 72~310 267~1148

Method 2 135 705~779 264~381 978~1126

Method 3 135 763 289 1071

20 degC

Method 1 148 50~100 06~380 21~1134

Method 2 148 758~838 249~368 843~1125

Method 3 148 798 308 1044

50 degC

Method 1 153 75~100 53~415 174~1227

Method 2 153 804~889 214~336 702~1103

Method 3 153 815 320 1050

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1247 ms at 0 degC 1213 ms at 20 degC and 1168 ms at 50 degC

- 30 -

CHAPTER 6 CONCLUDING REMARKS

61 Executive summary

In this study the reference-free damage detection localization and quantification tech-

nique in a multi layered composite plate is proposed Additional reflected A0 Lamb wave

mode from the damage due to multiple reflections in damage area is used as a feature for

characterizing the damage First damage is characterized in terms of a damage index (m-

value) defined as the ratio of the physical length of the damage to the wavelength of the A0

Lamb wave mode Then the range of possible damage location length and severity are esti-

mated with constraints This procedure can be accomplished without prior baseline data To

realize the proposed technique the matching pursuit technique is modified for reduction of

the error by considering the dispersive characteristic of Lamb wave The matching pursuit

method estimates the arrival time of each wave packets accurately even two wave packets are

overlapped The numerical simulations and experimental tests including Teflon tape inserted

and impact-induced damage are investigated for the feasibility of the proposed technique

Moreover the effectiveness of the proposed technique under changing temperature is exper-

imentally verified

- 31 -

62 Suggestion for further study

The limitations of this study and the possible improvement and further studies are listed as

follow

1 Measurement of impact-induced damage length using conventional NDT technique

Additional experimental verification for impact-induced damage characterization is

needed by measuring the real length of damage using conventional NDT technique

such as ultrasonic scan or infrared camera

2 Limitation of both side collocated PZT installation In this study excitation PZTs are

installed both side and collocated for decomposing A0 Lamb wave mode However it

can produce decomposition error due to imperfection of PZT location and their installa-

tion Moreover in many cases assess to both surface of structures is not allowed such

as aircraft applications To improve this limitation a reference-free damage characteri-

zation using single side PZTs will be considered

3 Complex structure application In this study damage is characterized in simple plate-

like structure When the monitored structure becomes complex the response signals al-

so becomes complex due to its geometry and boundary Thus the investigations for

complex structure application such as stiffener spar and curvature are needed

- 32 -

REFERENCES

[1] Cawley P and Alleyne D (1996) ldquoThe use of Lamb waves for the long range inspection of large structuresrdquo Ultrasonics 34 pp287-290

[2] Ghosh T Kundu T and Karpur P (1998) ldquoEfficient use of Lamb waves for detecting defects in large platesrdquo Ultrasonics 36 pp791-801

[3] Guirguitui V (2003) ldquoLamb wave generation with piezoelectric wafer active sensors for structural health monitoringrdquo SPIE Conference on Smart Structures and Materials San Diego CA 5391 pp111-422

[4] Ragvahan A and Cesnik CE S (2004) ldquoModeling of piezoelectric-based Lamb wave generation and sensing for structural health monitoringrdquo SPIE Conference on Smart Struc-tures and Materials San Diego CA 5391 pp419-430

[5] Rose JL Pilarski A and Dirit JJ (1993) ldquoAn approach to guided wave mode selection for inspection of laminated platerdquo Journal of Reinforced Plastics and Composites 12(5) pp536-544

[6] Guo N and Cawley P (1993) ldquoThe interaction of Lamb waves with delaminations in composite laminatesrdquo Journal of Acoustical Society of America 94(4) pp 2240-2246

[7] Valdez SHD and Soutis C (2002) ldquoReal-time nondestructive evaluation of fiber com-posite laminates using low-frequency Lamb wavesrdquo Journal of Acoustical Society of Amer-ica 111(5) pp 2026-2033

[8] Yang M and Qiao P (2005) ldquoModeling and experimental detection of damage in virous materials using the pulse-echo method and piezoelectric sensoractuatorsrdquo Smart Materials and Structures 14 pp1083-1100

[9] Lestari W and Qiao P (2005) ldquoApplication of wave propagation analysis for damage identification in composite laminated beamsrdquo Journal of Composite Materials 39(22) pp1967-1984

[10] Kessler S S Spearing S M and Soutis C (2002) ldquoDamage detection in composite ma-terials using Lamb wave methodsrdquo Smart Materials and Structures 11 pp269-278

[11] Sohn H Park G Wail J R Limback N P and Farrar C R (2004) ldquoWavelet-based active sensing for delamination detection in composite structuresrdquo Smart Materials and Structures 13 pp 153-160

[12] Petculescu G Krishnaswamy S and Achenback J D (2008) ldquoGroup delay measure-ments using modally selective Lamb wave transducers for detection and sizing of delamina-tions in compositesrdquo Smart Materials and Structures 17

[13] Sohn H Park H W Law K H and Farrar C R (2007) ldquoDamage detection in compo-

- 33 -

site plates by using an enhanced time reversal methodrdquo Journal of Aerospace Engineering (ASCE) pp 141-151

[14] Sohn H (2007) ldquoEffects of environmental and operational variability on structural health monitoringrdquo Philosophical Transactions of the Royal Society Ardquo 365 pp539-560

[15] Rytter A (1993) ldquoVibration based inspection of civil engineering structuresrdquo PhD The-sis Department of Building Technology and Structural Engineering University of Aalborg Aalborg Denmark

[16] Hayashi T and Kawashima K (2002) ldquoMultiple reflections of Lamb waves at a delami-nationrdquo Ultrasonics 40 pp193-197

[17] Wang C H and Rose L R F (2003) ldquoWave reflection and transmission in beams con-taining delamination and inhomogeneityrdquo Journal of Sound and Vibration 264 pp 851-872

[18] Ramadas C Balasubramaniam K Joshi M and Krishnamurthy C V (2009) ldquoInterac-tion of the primary anti-symmetric Lamb mode (Ao) with symmetric delaminations numer-ical and experimental studiesrdquo Smart Materials and Structures 18

[19] Ramadas C Balasubramaniam K Joshi M and Krishnamurthy C V (2010) ldquoInterac-tion of guided Lamb waves with an asymmetrically located delamination in a laminated composite platerdquo Smart Materials and Structures 19

[20] Wilcox P Lowe M and Cawely P (2001) ldquoThe effect of dispersion on long-range in-spection using ultrasonic guided wavesrdquo NDTampE International 34 pp1-9

[21] Mallat S G and Zhang Z (1993) ldquoMatching pursuits with time-frequency dictionariesrdquo IEEE Tansactions on Signal Processing 41(12) pp3397-3415

[22] Zhang G Zhang S and Wang Y (2000) ldquoApplication of adaptive time-frequency de-composition in ultrasonic NDE of highly scattering materialsrdquo Ultrasonics 38 pp 961-964

[23] Hong J C Sun K H and Kim Y Y (2005) ldquoThe matching pursuit approach based on the modulated Gaussian pulse for efficient guided-wave damage inspectionrdquo Smart Mate-rials and Structures 14 pp548-560

[24] Herakovich C T (1997) Mechanics of Fibrous Composites Wiley 480 pages

[25] Reddy J N (1985) ldquoA review of the literature on finite-element modeling of laminated composite platesrdquo The Shock and Vibration Digest 17(4) pp 3-8

[26] Reddy J N (2003) Mechanics of Laminated Composite Plates and Shells Theory and Analysis CRC Press 856 pages

[27] Palacz M Krawczuk M and Ostachowicz W (2005) ldquoThe spectral finite element mod-el for analysis of flexure-shear coupled wave propagation Part 1 laminated multilayer

- 34 -

composite beamrdquo Composite Structures 68 pp37-44

[28] Zhang Y C and Harding J (1990) ldquoA numerical micromechanics analysis of the me-chanical properties of a plain weave compositerdquo Computer amp Structures 36(5) pp 839-844

[29] Nemat-Nasser S and Hori M (1999) Micromechanics Overall Properties of Heteroge-neous Materials Elsevier 810 pages

[30] Scida D Aboura Z Benzeggagh M L and Bocherens E (1999) ldquoA micromechanics model for 3D elasticity and failure of woven-fibre composite materialrdquo Composites Science and Technology 59 pp 505-517

[31] Thomas Y and Wu X H (1997) ldquoA multiscale finite element method for elliptic prob-lems in composite materials and porous mediardquo Journal of Computational Physics 134 pp 169-189

[32] Lee K H Moorthy S and Ghosh S (1999) ldquoMultiple scale computational model for damage in composite materialsrdquo Computer Methods in Applied Mechanics and Engineer-ing 172 pp 175-201

[33] Namilae S and Chandra N (2005) ldquoMultiscale model to study the effect of interfaces in carbon nanotube-based compositesrdquo Journal of Engineering Materials and Technology (ASME) 127 pp222-232

[34] Adams D E (2007) Health monitoring of Structural Materials and Components Wiley 460 pages

[35] Gopalakrishnan S (2009) ldquoModeling Aspects in Finite Elementrdquo Encyclopedia of Struc-tural Health Monitoring 2 pp791-809

[36] Palacz M Krawczuk M and Ostachowicz W (2005) ldquoThe spectral finite element mod-el for analysis of flexure-shear coupled wave propagation Part 2 Deminated multilayer composite beamrdquo Composite Structures 68 pp45-51

[37] Baik J M and Thopson R B (1984) ldquoUltrasonic scattering from imperfect interfaces A quasi-static modelrdquo Journal of Nondestructive Evaluation 4(3) pp177-196

[38] Luo H and Hanagud S (2000) ldquoDynamics of delamincated beamsrdquo International Jour-nal of Solids and Structures 37 pp1501-1519

[39] Kim S B and Sohn H (2007) ldquoInstantaneous reference-free crack detection based on polarization characteristics of piezoelectric materialsrdquo Smart Materials and Structures 16 pp2375-2387

[40] An Y K and Sohn H (2010) ldquoInstantaneous crack detection under varying temperature and static loading conditionsrdquo The Special Issue of Structural Control and Health Monitor-ing 17(7) pp 730-741

- 35 -

[41] Lee S J and Sohn H (2006) ldquoActive self-sensing scheme development for structural health monitoringrdquo Smart Materials and Structures 15(6) pp 1734-1746

- 36 -

요 약 문

복합재료 손상의 무기저 진단 및 정량화

복합재료는 재료의 경량성 내구성 고강도의 특성으로 인하여 항공기에서부터

교량에까지 널리 사용되고 있다 하지만 외부의 충격으로 인한 복합재료 내부의

손상은 재료의 강도를 감소시켜 구조물의 안정성을 저하시키는 요인이 된다 하지만

복합재료의 손상은 재료 내부에서 발생하여 육안으로는 식별하기도 힘들며 표면에서

보이는 것만으로 내부의 손상 정도를 판단하기는 불가능하다는 단점을 가지고 있다

본 연구에서는 손상 내부에서 반사로 인해 발생하는 파를 이용하여 기저자료를

사용하지 않고 특정 시점에서 얻은 자료만으로 복합재료의 손상을 진단하고 정량화

할 수 있는 기법을 제안한다 제안된 기법의 적용성을 검증하기 위하여 유한요소법을

이용한 수치해석을 수행하였고 Teflon tape 이 삽입된 시편과 실제 충격하중으로

인한 손상을 입은 시편을 이용한 실험을 진행하였다 수치해석 및 실험 결과 제안된

기법이 복합재료의 손상을 성공적으로 진단 및 정량화할 수 있음을 확인하였다 또한

다양한 외부 온도 환경에서 제안된 기법의 환경영향 실험을 수행하였고 다양한

환경에서 손상의 진단 및 정량화가 성공적으로 이루어짐을 검증하였다

핵심어 무기저 손상진단 및 정량화 압전센서 복합재료 유한요소법 Matching pursuit

method

- 37 -

감 사 의 글

기대와 두려움을 안고 시작한 석사과정이 어느덧 2 년 이라는 세월 속에서 추억으로 자리잡아

가고 있습니다 학부생 기간을 포함한 지난 3 년 동안 SHM 이라는 새로운 분야를 접하고 연구를

수행 할 수 있도록 도와주신 고마운 분들이 있어 그 감사를 표하고자 합니다

우선 저의 많은 부족한 점을 일깨워 주시고 연구에 대한 뜻을 더욱 확고히 세울 수 있게 해

주신 지도교수님이신 손 훈 교수님께 깊은 감사의 뜻을 전합니다 짧은 기간이었지만 교수님의

도움으로 많은 것을 배우고 느끼고 깨달을 수 있었습니다 또한 본 연구를 수행할 수 있도록 지

원해주신 국토해양부 (U-City 석 박사과정 지원사업)에 깊은 감사 드립니다

때로 힘들고 지칠 때 항상 저를 믿고 응원해 주시는 가족이 있었기에 지금의 글을 쓸 수 있었

습니다 또한 지금 연구실에서 함께 생활하고 있는 윤규형 민구형 철민이형 현석이형 진열이형

호민이형 병진이 선혜누나와 성대에 계신 창길이형 미국에 계신 의재형과 Debaditiya 에게도 감

사의 말씀을 전합니다 그리고 모두 나열할 수는 없지만 카이스트 생활의 활력소가 되어준 카이

스트 건설 및 환경공학과 선후배들 직원 분들과 SOS 멤버들에게도 감사를 표합니다

앞서 거론한 많은 이들의 도움이 없었다면 저 또한 지금 이 자리에 있을 수 없었을 것입니다

많이 부족하기에 더 노력하고 배우고자 했던 초심을 다시 한번 생각하며 더 큰 결실을 맺기 위해

앞으로 나아가겠습니다 도움을 주신 모든 분들께 이 감사의 글을 바칩니다

2011년 6월 카이스트에서 임형진 올림

- 38 -

CURRICULUM VITAE

Hyung Jin Lim

Research Assistant

The Civil and Environmental Engineering Department

Korea Advanced Institute of Science and Technology (KAIST)

291 Daehak-Ro Guseong-Dong Yuseong-Gu

Daejeon Republic of Korea 305-701

Phone (82)+42-350-3665 Fax (82)+42-350-3610

Email limhj87gmailcom

RESEARCH INTERESTS

Damage Detection Localization and Quantification on Composite and Metallic Structures Pattern Recognition Finite Element Modeling (FEM)

EDUCATION

2009 ndash 2011 MS Civil and Environmental Eng Korea Advanced Institute of Science and Technology (KAIST) Korea Rep

2006 ndash 2009 BS Civil and Environmental Eng Korea Advanced Institute of Science and Technology (KAIST) Korea Rep

JOURNAL PUBLICATIONS The corresponding author is underlined 1 Hyung Jin Lim Hoon Sohn ldquoReference-free damage detection localization and quantification

using in compositerdquo in preparation for Journal of the Acoustical Society of America 2011

2 Chul Min Yeum Hoon Sohn Jeong Bum Ihn Hyung Jin Lim ldquoInstantaneous delamination detection in a composite using a dual piezoelectric transducer networkrdquo Submitted to Compo-sites Science and Technology 2011

3 Yun-Kyu An Hyung Jin Lim Min Koo Kim Jin Yeol Yang Hoon Sohn Chang Guen Lee ldquoInvestigation of applicability of local reference-free damage detection techniques to in-situ

- 39 -

bridgesrdquo Submitted to ASCE Journal of Structural Engineering 2011

4 Hyung Jin Lim Min Koo Kim Hoon Sohn Chan Yik Park ldquoImpedance based damage detec-tion under varying temperature and loading conditionsrdquo Submitted to NDTampE International 2011

5 Hyung Jin Lim Hoon Sohn ldquoQuantitative estimation of transmitted and reflected lamb waves at discontinuityrdquo Journal of the Korean Society for Nondestructive Testing Vol30 No 4 pp1-8 2010

CONFERENCE PROCEEDINGS

The corresponding author is underlined 1 Hyung Jin Lim and Hoon Sohn ldquoReference-free damage detection localization and quantifi-

cation in compositesrdquo ASCE Engineering Mechanics Institute Conference Boston June 2-4 2011

2 Hyung Jin Lim and Hoon Sohn ldquoReference-free delamination quantification within multi-layer composite materialsrdquo KSNT Spring Conference May 26-27 Gyeong-Ju Korea 2011

3 Min Koo Kim Hyun Jin Lim Hoon Sohn Chan Yik Park ldquoImpedance-based bolt loosening detection under varying temperature and loadingrdquo Asian Pacific Workshop on Structural Health Monitoring Tokyo Japan November 30-December 2 2010

4 Hyung Jin Lim Chul Min Yeum and Hoon Sohn ldquoModeling of Impact-induced Delamination in a Multilayer Composite Platerdquo The 23rd KKCNN Symposium on Civil Engineering Taipei (Taiwan) November 13-15 2010

5 Yun-Kyu An Hyung Jin Lim and Hoon Sohn ldquoTemperature independent Delamination Detec-tion using data normalizationrdquo The 3rd International Conference on Multi-Functional Materi-als and Structures Chonbuk National Univ Jeonju Korea September 14-18 2010

6 Hoon Sohn Yun-Kyu An Chang Gil Lee Min Koo Kim Hyung Jin Lim ldquoApplication of online NDT techniques to decommissioned bridge testingrdquo the 5th European Workshop on Structural Health Monitoring Sorrento Italy June 29-July 02 2010

7 Min Koo Kim Hyung Jin Lim Hoon Sohn Chan Yik Park ldquoImpedance based damage diag-nosis of a complex composite aircraft wing under changing loading and temperature condi-tionsrdquo the 5th European Workshop on Structural Health Monitoring Sorrento Italy June 29-July 02 2010

8 Yun-Kyu An Min Koo Kim Hyung Jin Lim and Hoon Sohn ldquoValidations of reference-free crack detection techniques through a decommissioned bridge testrdquo 2010 COSEIK Annual Con-ference Jeju Korea April 8-9 2010

- 40 -

PARTICIPANT PROJECTS

1 Development of Smart Steel Members for Online Steel Structure Monitoring POSCO Corporation (Funded 41000000 KRW (41000 USD) for 010511 to 123111)

2 Development of Enabling Technologies for Structural Health Monitoring Korean Agency for Defense Development (Funded 240000000 KRW (240000 USD) out of total 3200000000 KRW (3200000 USD) for 060108-053111)

3 Development of On-board SHM Technologies for Composite Air Vehicles The Boeing Company (Funded 400000 USD (400000000 KRW) for 080108 to 073111)

EDUCATIONAL EXPERIENCES

Fall 2009

Teaching Assistant CE207 ldquoElementary Structural Engineering and Laboratoryrdquo KAIST 3 unit undergraduate course

EXPERIMENTAL EXPERIENTCES

2010 Monitoring of Real Aircraft Composite Wing Segment Korea Agency for Defense De-velopment (ADD) Daejeon Korea Real time monitoring of composite air vehicles fund-ed by Korea Agency for Defense Development (ADD)

2009-2011

Field Monitoring of Real Bridge Structures Gimpo (Ramp G bridge) amp Icheon (Sam-seung bridge) Korea US-Korea collaborative research for bridge monitoring testbeds

HONORS amp AWARDS

Scholarships for outstanding students granted KAIST (2007-2009) The three prominent students in the civil and environmental engineering department of KAIST were selected as recipients This scholarship provided 2000$ per year for three years

- 41 -

Page 36: 복합재료 손상의 무기저 진단 및 정량화 - KAISTssslab.kaist.ac.kr/article/pdf/Thesis_2.pdf · 1 Declaration of Ethical Conduct in Research: I, as a graduate student

52 Teflon inserted damage

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Reflected wave from damage

(a)

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Transmitted WaveReflected Wave

Detected reflected wave

(b)

Fig 12 A0 Lamb wave mode obtained using pitch-catch method from the Teflon inserted damage specimen (a) The reflected wave from damage is shown (b) The Teflon inserted damage is detected by capturing the reflected wave

Fig 12 shows the obtained pitch-catch signal from the Teflon inserted damage specimen

and damage diagnosis result The Teflon inserted damage is detected successfully by captur-

ing the reflected wave Table IV presents the result of the Teflon inserted damage characteri-

zation 119907119907119894119894 is measured as 1213ms and the variation is assumed to plusmn 5 of the normal val-

ue (1152~1273ms) The results indicate that the proposed damage characterization approach

works well in real application

TABLE IV Damage characterization for the Teflon inserted damage specimen

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

Exact 635 25

Method 1 123 60~75 73~338 298~1220

Method 2 123 619~662 229~310 934~1188

Method 3 123 631 284 1161

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1213 ms

- 25 -

53 Impact-induced damage

11 115 12 125 13-4

-2

0

2

4

Time (ms)

Vol

tage

(mV

)

IntactImpact2Impact3

Reflected wave from damage

(a) Pitch-catch signal

11 115 12 125 13 135

-05

0

05

1

Time (ms)

Volta

ge (m

V)

IntactImpact2Impact3

Boundary reflection

Reflected wave from damage

(b) Pulse-echo signal

Fig 13 The signal obtained from the test specimen subjected to impact The impact was in-troduced at the same location of the specimen and the signal was obtained after each impacts As the number of impact is increased the transmitted A0 Lamb wave mode is delayed and attenuated

The obtained signals from impact-induced damage specimen are presented in Fig 13 It is

observed that as the number of impact is increased the transmitted A0 Lamb wave mode is

more delayed because the wave velocity in the damage area is reduced Fig 14 shows the

false-alarm test result whether the proposed damage detection algorithm works even in pris-

tine condition of the test specimen It is shown that no false alarm is occurred because no ad-

ditional reflected wave is captured

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Transmitted Wave

(a) Transmitted wave

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Reflected Wave

No reflected wave

(b) Reflected wave

Fig 14 False alarm test for the proposed damage detection algorithm in pristine condition of the specimen Reflected wave from damage is not captured false alarm is not occurred

- 26 -

TABLE V Damage characterization for the impact-induced damage specimen

of Impacts m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

2

(1470 J)

Method 1 080 65~90 75~246 470~1157

Method 2 080 801~885 114~245 711~1134

Method 3 080 843 179 1122

3

(2205 J)

Method 1 148 50~100 06~380 21~1134

Method 2 148 758~838 249~368 843~1125

Method 3 148 798 308 1044

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1213 ms

Table V presents the characterization results of the impact-induced damage case Here the

measured 119907119907119894119894 and the assumed variation are same as the artificial damage case It is observed

that as the number of impact is increased 119889119889 is increased while 119907119907119889119889 is reduced Thus it is

validated that the proposed damage characterization approach estimates the increment of the

damage well

54 Temperature effect test

541 Teflon inserted damage

11 115 12 125 13-10

-5

0

5

10

Time (ms)

Vol

tage

(mV

)

0 oC20 oC50 oC

(a) Pitch-catch signal

11 115 12 125 13 135

-1

0

1

2

3

Time(ms)

Volta

ge (m

V)

0 oC20 oC50 oC

(b) Pulse-echo signal

Fig 15 Signals obtained from the Teflon inserted damage specimen under temperature varia-tion Time delay and amplitude variation are occurred

Fig 15 shows the obtained signals from the Teflon inserted damage specimen subjected to

- 27 -

temperature variation Time delay and amplitude variations are occurred there is no addition-

al damage is introduced The analyzed results of the obtained signals are reported in Table VI

The measured 119907119907119894119894 is 1247 ms at 0 degC 1213 ms at 20 degC and 1168 ms at 50 degC respectively

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms (1152~1273 ms)

The estimated results are also changed under temperature variations the fluctuation of esti-

mated range and exact value is small

TABLE VI Temperature effect test for the Teflon inserted damage specimen

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

0 degC

Exact 635 25

Method 1 113 60~70 79~268 352~1291

Method 2 113 560~619 233~344 1034~1172

Method 3 113 606 257 1141

20 degC

Exact 635 25

Method 1 123 60~75 73~338 298~1220

Method 2 123 619~662 229~310 934~1188

Method 3 123 631 284 1161

50 degC

Exact 635 25

Method 1 137 60~75 72~328 263~1228

Method 2 137 622~687 183~299 666~1086

Method 3 137 642 264 960

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1247 ms at 0 degC 1213 ms at 20 degC and 1168 ms at 50 degC

- 28 -

542 Impact-induced damage

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Vol

tage

(mV

)

Transmitted WaveReflected Wave

No reflected wave

(a) 0 degC

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Transmitted WaveReflected Wave

No reflected wave

(b) 50 degC

Fig 16 False alarm test under temperature variations It is shown that proposed damage char-acterization algorithm is robust under temperature variations

The false alarm test in pristine condition of the impact-induced damage specimen under

temperature variation was investigated (Fig 16) It is validated that the proposed damage di-

agnosis algorithm is robust under temperature variations Fig 17 shows the obtained signals

from the impact-induced damage (impacted 3 times) specimen under temperature variation

The analyzed results of the obtained signals are reported in Table VII The measured 119907119907119894119894 in

each temperature and the assumed variation are same as the artificial damage case

11 115 12 125 13-5

0

5

Time (ms)

Volta

ge (m

V)

0 oC20 oC50 oC

(a) Pitch-catch signal

11 115 12 125 13 135

-1

0

1

2

3

Time(ms)

Vol

tage

(mV

)

0 oC20 oC50 oC

(b) Pulse-echo signal

Fig 17 Signals obtained from impact-induced damage specimen (Impact 3) under tempera-ture variation

- 29 -

TABLE VII Temperature effect test for the impact induced damage specimen (Impact 3 case)

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

0 degC

Method 1 135 75~90 72~310 267~1148

Method 2 135 705~779 264~381 978~1126

Method 3 135 763 289 1071

20 degC

Method 1 148 50~100 06~380 21~1134

Method 2 148 758~838 249~368 843~1125

Method 3 148 798 308 1044

50 degC

Method 1 153 75~100 53~415 174~1227

Method 2 153 804~889 214~336 702~1103

Method 3 153 815 320 1050

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1247 ms at 0 degC 1213 ms at 20 degC and 1168 ms at 50 degC

- 30 -

CHAPTER 6 CONCLUDING REMARKS

61 Executive summary

In this study the reference-free damage detection localization and quantification tech-

nique in a multi layered composite plate is proposed Additional reflected A0 Lamb wave

mode from the damage due to multiple reflections in damage area is used as a feature for

characterizing the damage First damage is characterized in terms of a damage index (m-

value) defined as the ratio of the physical length of the damage to the wavelength of the A0

Lamb wave mode Then the range of possible damage location length and severity are esti-

mated with constraints This procedure can be accomplished without prior baseline data To

realize the proposed technique the matching pursuit technique is modified for reduction of

the error by considering the dispersive characteristic of Lamb wave The matching pursuit

method estimates the arrival time of each wave packets accurately even two wave packets are

overlapped The numerical simulations and experimental tests including Teflon tape inserted

and impact-induced damage are investigated for the feasibility of the proposed technique

Moreover the effectiveness of the proposed technique under changing temperature is exper-

imentally verified

- 31 -

62 Suggestion for further study

The limitations of this study and the possible improvement and further studies are listed as

follow

1 Measurement of impact-induced damage length using conventional NDT technique

Additional experimental verification for impact-induced damage characterization is

needed by measuring the real length of damage using conventional NDT technique

such as ultrasonic scan or infrared camera

2 Limitation of both side collocated PZT installation In this study excitation PZTs are

installed both side and collocated for decomposing A0 Lamb wave mode However it

can produce decomposition error due to imperfection of PZT location and their installa-

tion Moreover in many cases assess to both surface of structures is not allowed such

as aircraft applications To improve this limitation a reference-free damage characteri-

zation using single side PZTs will be considered

3 Complex structure application In this study damage is characterized in simple plate-

like structure When the monitored structure becomes complex the response signals al-

so becomes complex due to its geometry and boundary Thus the investigations for

complex structure application such as stiffener spar and curvature are needed

- 32 -

REFERENCES

[1] Cawley P and Alleyne D (1996) ldquoThe use of Lamb waves for the long range inspection of large structuresrdquo Ultrasonics 34 pp287-290

[2] Ghosh T Kundu T and Karpur P (1998) ldquoEfficient use of Lamb waves for detecting defects in large platesrdquo Ultrasonics 36 pp791-801

[3] Guirguitui V (2003) ldquoLamb wave generation with piezoelectric wafer active sensors for structural health monitoringrdquo SPIE Conference on Smart Structures and Materials San Diego CA 5391 pp111-422

[4] Ragvahan A and Cesnik CE S (2004) ldquoModeling of piezoelectric-based Lamb wave generation and sensing for structural health monitoringrdquo SPIE Conference on Smart Struc-tures and Materials San Diego CA 5391 pp419-430

[5] Rose JL Pilarski A and Dirit JJ (1993) ldquoAn approach to guided wave mode selection for inspection of laminated platerdquo Journal of Reinforced Plastics and Composites 12(5) pp536-544

[6] Guo N and Cawley P (1993) ldquoThe interaction of Lamb waves with delaminations in composite laminatesrdquo Journal of Acoustical Society of America 94(4) pp 2240-2246

[7] Valdez SHD and Soutis C (2002) ldquoReal-time nondestructive evaluation of fiber com-posite laminates using low-frequency Lamb wavesrdquo Journal of Acoustical Society of Amer-ica 111(5) pp 2026-2033

[8] Yang M and Qiao P (2005) ldquoModeling and experimental detection of damage in virous materials using the pulse-echo method and piezoelectric sensoractuatorsrdquo Smart Materials and Structures 14 pp1083-1100

[9] Lestari W and Qiao P (2005) ldquoApplication of wave propagation analysis for damage identification in composite laminated beamsrdquo Journal of Composite Materials 39(22) pp1967-1984

[10] Kessler S S Spearing S M and Soutis C (2002) ldquoDamage detection in composite ma-terials using Lamb wave methodsrdquo Smart Materials and Structures 11 pp269-278

[11] Sohn H Park G Wail J R Limback N P and Farrar C R (2004) ldquoWavelet-based active sensing for delamination detection in composite structuresrdquo Smart Materials and Structures 13 pp 153-160

[12] Petculescu G Krishnaswamy S and Achenback J D (2008) ldquoGroup delay measure-ments using modally selective Lamb wave transducers for detection and sizing of delamina-tions in compositesrdquo Smart Materials and Structures 17

[13] Sohn H Park H W Law K H and Farrar C R (2007) ldquoDamage detection in compo-

- 33 -

site plates by using an enhanced time reversal methodrdquo Journal of Aerospace Engineering (ASCE) pp 141-151

[14] Sohn H (2007) ldquoEffects of environmental and operational variability on structural health monitoringrdquo Philosophical Transactions of the Royal Society Ardquo 365 pp539-560

[15] Rytter A (1993) ldquoVibration based inspection of civil engineering structuresrdquo PhD The-sis Department of Building Technology and Structural Engineering University of Aalborg Aalborg Denmark

[16] Hayashi T and Kawashima K (2002) ldquoMultiple reflections of Lamb waves at a delami-nationrdquo Ultrasonics 40 pp193-197

[17] Wang C H and Rose L R F (2003) ldquoWave reflection and transmission in beams con-taining delamination and inhomogeneityrdquo Journal of Sound and Vibration 264 pp 851-872

[18] Ramadas C Balasubramaniam K Joshi M and Krishnamurthy C V (2009) ldquoInterac-tion of the primary anti-symmetric Lamb mode (Ao) with symmetric delaminations numer-ical and experimental studiesrdquo Smart Materials and Structures 18

[19] Ramadas C Balasubramaniam K Joshi M and Krishnamurthy C V (2010) ldquoInterac-tion of guided Lamb waves with an asymmetrically located delamination in a laminated composite platerdquo Smart Materials and Structures 19

[20] Wilcox P Lowe M and Cawely P (2001) ldquoThe effect of dispersion on long-range in-spection using ultrasonic guided wavesrdquo NDTampE International 34 pp1-9

[21] Mallat S G and Zhang Z (1993) ldquoMatching pursuits with time-frequency dictionariesrdquo IEEE Tansactions on Signal Processing 41(12) pp3397-3415

[22] Zhang G Zhang S and Wang Y (2000) ldquoApplication of adaptive time-frequency de-composition in ultrasonic NDE of highly scattering materialsrdquo Ultrasonics 38 pp 961-964

[23] Hong J C Sun K H and Kim Y Y (2005) ldquoThe matching pursuit approach based on the modulated Gaussian pulse for efficient guided-wave damage inspectionrdquo Smart Mate-rials and Structures 14 pp548-560

[24] Herakovich C T (1997) Mechanics of Fibrous Composites Wiley 480 pages

[25] Reddy J N (1985) ldquoA review of the literature on finite-element modeling of laminated composite platesrdquo The Shock and Vibration Digest 17(4) pp 3-8

[26] Reddy J N (2003) Mechanics of Laminated Composite Plates and Shells Theory and Analysis CRC Press 856 pages

[27] Palacz M Krawczuk M and Ostachowicz W (2005) ldquoThe spectral finite element mod-el for analysis of flexure-shear coupled wave propagation Part 1 laminated multilayer

- 34 -

composite beamrdquo Composite Structures 68 pp37-44

[28] Zhang Y C and Harding J (1990) ldquoA numerical micromechanics analysis of the me-chanical properties of a plain weave compositerdquo Computer amp Structures 36(5) pp 839-844

[29] Nemat-Nasser S and Hori M (1999) Micromechanics Overall Properties of Heteroge-neous Materials Elsevier 810 pages

[30] Scida D Aboura Z Benzeggagh M L and Bocherens E (1999) ldquoA micromechanics model for 3D elasticity and failure of woven-fibre composite materialrdquo Composites Science and Technology 59 pp 505-517

[31] Thomas Y and Wu X H (1997) ldquoA multiscale finite element method for elliptic prob-lems in composite materials and porous mediardquo Journal of Computational Physics 134 pp 169-189

[32] Lee K H Moorthy S and Ghosh S (1999) ldquoMultiple scale computational model for damage in composite materialsrdquo Computer Methods in Applied Mechanics and Engineer-ing 172 pp 175-201

[33] Namilae S and Chandra N (2005) ldquoMultiscale model to study the effect of interfaces in carbon nanotube-based compositesrdquo Journal of Engineering Materials and Technology (ASME) 127 pp222-232

[34] Adams D E (2007) Health monitoring of Structural Materials and Components Wiley 460 pages

[35] Gopalakrishnan S (2009) ldquoModeling Aspects in Finite Elementrdquo Encyclopedia of Struc-tural Health Monitoring 2 pp791-809

[36] Palacz M Krawczuk M and Ostachowicz W (2005) ldquoThe spectral finite element mod-el for analysis of flexure-shear coupled wave propagation Part 2 Deminated multilayer composite beamrdquo Composite Structures 68 pp45-51

[37] Baik J M and Thopson R B (1984) ldquoUltrasonic scattering from imperfect interfaces A quasi-static modelrdquo Journal of Nondestructive Evaluation 4(3) pp177-196

[38] Luo H and Hanagud S (2000) ldquoDynamics of delamincated beamsrdquo International Jour-nal of Solids and Structures 37 pp1501-1519

[39] Kim S B and Sohn H (2007) ldquoInstantaneous reference-free crack detection based on polarization characteristics of piezoelectric materialsrdquo Smart Materials and Structures 16 pp2375-2387

[40] An Y K and Sohn H (2010) ldquoInstantaneous crack detection under varying temperature and static loading conditionsrdquo The Special Issue of Structural Control and Health Monitor-ing 17(7) pp 730-741

- 35 -

[41] Lee S J and Sohn H (2006) ldquoActive self-sensing scheme development for structural health monitoringrdquo Smart Materials and Structures 15(6) pp 1734-1746

- 36 -

요 약 문

복합재료 손상의 무기저 진단 및 정량화

복합재료는 재료의 경량성 내구성 고강도의 특성으로 인하여 항공기에서부터

교량에까지 널리 사용되고 있다 하지만 외부의 충격으로 인한 복합재료 내부의

손상은 재료의 강도를 감소시켜 구조물의 안정성을 저하시키는 요인이 된다 하지만

복합재료의 손상은 재료 내부에서 발생하여 육안으로는 식별하기도 힘들며 표면에서

보이는 것만으로 내부의 손상 정도를 판단하기는 불가능하다는 단점을 가지고 있다

본 연구에서는 손상 내부에서 반사로 인해 발생하는 파를 이용하여 기저자료를

사용하지 않고 특정 시점에서 얻은 자료만으로 복합재료의 손상을 진단하고 정량화

할 수 있는 기법을 제안한다 제안된 기법의 적용성을 검증하기 위하여 유한요소법을

이용한 수치해석을 수행하였고 Teflon tape 이 삽입된 시편과 실제 충격하중으로

인한 손상을 입은 시편을 이용한 실험을 진행하였다 수치해석 및 실험 결과 제안된

기법이 복합재료의 손상을 성공적으로 진단 및 정량화할 수 있음을 확인하였다 또한

다양한 외부 온도 환경에서 제안된 기법의 환경영향 실험을 수행하였고 다양한

환경에서 손상의 진단 및 정량화가 성공적으로 이루어짐을 검증하였다

핵심어 무기저 손상진단 및 정량화 압전센서 복합재료 유한요소법 Matching pursuit

method

- 37 -

감 사 의 글

기대와 두려움을 안고 시작한 석사과정이 어느덧 2 년 이라는 세월 속에서 추억으로 자리잡아

가고 있습니다 학부생 기간을 포함한 지난 3 년 동안 SHM 이라는 새로운 분야를 접하고 연구를

수행 할 수 있도록 도와주신 고마운 분들이 있어 그 감사를 표하고자 합니다

우선 저의 많은 부족한 점을 일깨워 주시고 연구에 대한 뜻을 더욱 확고히 세울 수 있게 해

주신 지도교수님이신 손 훈 교수님께 깊은 감사의 뜻을 전합니다 짧은 기간이었지만 교수님의

도움으로 많은 것을 배우고 느끼고 깨달을 수 있었습니다 또한 본 연구를 수행할 수 있도록 지

원해주신 국토해양부 (U-City 석 박사과정 지원사업)에 깊은 감사 드립니다

때로 힘들고 지칠 때 항상 저를 믿고 응원해 주시는 가족이 있었기에 지금의 글을 쓸 수 있었

습니다 또한 지금 연구실에서 함께 생활하고 있는 윤규형 민구형 철민이형 현석이형 진열이형

호민이형 병진이 선혜누나와 성대에 계신 창길이형 미국에 계신 의재형과 Debaditiya 에게도 감

사의 말씀을 전합니다 그리고 모두 나열할 수는 없지만 카이스트 생활의 활력소가 되어준 카이

스트 건설 및 환경공학과 선후배들 직원 분들과 SOS 멤버들에게도 감사를 표합니다

앞서 거론한 많은 이들의 도움이 없었다면 저 또한 지금 이 자리에 있을 수 없었을 것입니다

많이 부족하기에 더 노력하고 배우고자 했던 초심을 다시 한번 생각하며 더 큰 결실을 맺기 위해

앞으로 나아가겠습니다 도움을 주신 모든 분들께 이 감사의 글을 바칩니다

2011년 6월 카이스트에서 임형진 올림

- 38 -

CURRICULUM VITAE

Hyung Jin Lim

Research Assistant

The Civil and Environmental Engineering Department

Korea Advanced Institute of Science and Technology (KAIST)

291 Daehak-Ro Guseong-Dong Yuseong-Gu

Daejeon Republic of Korea 305-701

Phone (82)+42-350-3665 Fax (82)+42-350-3610

Email limhj87gmailcom

RESEARCH INTERESTS

Damage Detection Localization and Quantification on Composite and Metallic Structures Pattern Recognition Finite Element Modeling (FEM)

EDUCATION

2009 ndash 2011 MS Civil and Environmental Eng Korea Advanced Institute of Science and Technology (KAIST) Korea Rep

2006 ndash 2009 BS Civil and Environmental Eng Korea Advanced Institute of Science and Technology (KAIST) Korea Rep

JOURNAL PUBLICATIONS The corresponding author is underlined 1 Hyung Jin Lim Hoon Sohn ldquoReference-free damage detection localization and quantification

using in compositerdquo in preparation for Journal of the Acoustical Society of America 2011

2 Chul Min Yeum Hoon Sohn Jeong Bum Ihn Hyung Jin Lim ldquoInstantaneous delamination detection in a composite using a dual piezoelectric transducer networkrdquo Submitted to Compo-sites Science and Technology 2011

3 Yun-Kyu An Hyung Jin Lim Min Koo Kim Jin Yeol Yang Hoon Sohn Chang Guen Lee ldquoInvestigation of applicability of local reference-free damage detection techniques to in-situ

- 39 -

bridgesrdquo Submitted to ASCE Journal of Structural Engineering 2011

4 Hyung Jin Lim Min Koo Kim Hoon Sohn Chan Yik Park ldquoImpedance based damage detec-tion under varying temperature and loading conditionsrdquo Submitted to NDTampE International 2011

5 Hyung Jin Lim Hoon Sohn ldquoQuantitative estimation of transmitted and reflected lamb waves at discontinuityrdquo Journal of the Korean Society for Nondestructive Testing Vol30 No 4 pp1-8 2010

CONFERENCE PROCEEDINGS

The corresponding author is underlined 1 Hyung Jin Lim and Hoon Sohn ldquoReference-free damage detection localization and quantifi-

cation in compositesrdquo ASCE Engineering Mechanics Institute Conference Boston June 2-4 2011

2 Hyung Jin Lim and Hoon Sohn ldquoReference-free delamination quantification within multi-layer composite materialsrdquo KSNT Spring Conference May 26-27 Gyeong-Ju Korea 2011

3 Min Koo Kim Hyun Jin Lim Hoon Sohn Chan Yik Park ldquoImpedance-based bolt loosening detection under varying temperature and loadingrdquo Asian Pacific Workshop on Structural Health Monitoring Tokyo Japan November 30-December 2 2010

4 Hyung Jin Lim Chul Min Yeum and Hoon Sohn ldquoModeling of Impact-induced Delamination in a Multilayer Composite Platerdquo The 23rd KKCNN Symposium on Civil Engineering Taipei (Taiwan) November 13-15 2010

5 Yun-Kyu An Hyung Jin Lim and Hoon Sohn ldquoTemperature independent Delamination Detec-tion using data normalizationrdquo The 3rd International Conference on Multi-Functional Materi-als and Structures Chonbuk National Univ Jeonju Korea September 14-18 2010

6 Hoon Sohn Yun-Kyu An Chang Gil Lee Min Koo Kim Hyung Jin Lim ldquoApplication of online NDT techniques to decommissioned bridge testingrdquo the 5th European Workshop on Structural Health Monitoring Sorrento Italy June 29-July 02 2010

7 Min Koo Kim Hyung Jin Lim Hoon Sohn Chan Yik Park ldquoImpedance based damage diag-nosis of a complex composite aircraft wing under changing loading and temperature condi-tionsrdquo the 5th European Workshop on Structural Health Monitoring Sorrento Italy June 29-July 02 2010

8 Yun-Kyu An Min Koo Kim Hyung Jin Lim and Hoon Sohn ldquoValidations of reference-free crack detection techniques through a decommissioned bridge testrdquo 2010 COSEIK Annual Con-ference Jeju Korea April 8-9 2010

- 40 -

PARTICIPANT PROJECTS

1 Development of Smart Steel Members for Online Steel Structure Monitoring POSCO Corporation (Funded 41000000 KRW (41000 USD) for 010511 to 123111)

2 Development of Enabling Technologies for Structural Health Monitoring Korean Agency for Defense Development (Funded 240000000 KRW (240000 USD) out of total 3200000000 KRW (3200000 USD) for 060108-053111)

3 Development of On-board SHM Technologies for Composite Air Vehicles The Boeing Company (Funded 400000 USD (400000000 KRW) for 080108 to 073111)

EDUCATIONAL EXPERIENCES

Fall 2009

Teaching Assistant CE207 ldquoElementary Structural Engineering and Laboratoryrdquo KAIST 3 unit undergraduate course

EXPERIMENTAL EXPERIENTCES

2010 Monitoring of Real Aircraft Composite Wing Segment Korea Agency for Defense De-velopment (ADD) Daejeon Korea Real time monitoring of composite air vehicles fund-ed by Korea Agency for Defense Development (ADD)

2009-2011

Field Monitoring of Real Bridge Structures Gimpo (Ramp G bridge) amp Icheon (Sam-seung bridge) Korea US-Korea collaborative research for bridge monitoring testbeds

HONORS amp AWARDS

Scholarships for outstanding students granted KAIST (2007-2009) The three prominent students in the civil and environmental engineering department of KAIST were selected as recipients This scholarship provided 2000$ per year for three years

- 41 -

Page 37: 복합재료 손상의 무기저 진단 및 정량화 - KAISTssslab.kaist.ac.kr/article/pdf/Thesis_2.pdf · 1 Declaration of Ethical Conduct in Research: I, as a graduate student

53 Impact-induced damage

11 115 12 125 13-4

-2

0

2

4

Time (ms)

Vol

tage

(mV

)

IntactImpact2Impact3

Reflected wave from damage

(a) Pitch-catch signal

11 115 12 125 13 135

-05

0

05

1

Time (ms)

Volta

ge (m

V)

IntactImpact2Impact3

Boundary reflection

Reflected wave from damage

(b) Pulse-echo signal

Fig 13 The signal obtained from the test specimen subjected to impact The impact was in-troduced at the same location of the specimen and the signal was obtained after each impacts As the number of impact is increased the transmitted A0 Lamb wave mode is delayed and attenuated

The obtained signals from impact-induced damage specimen are presented in Fig 13 It is

observed that as the number of impact is increased the transmitted A0 Lamb wave mode is

more delayed because the wave velocity in the damage area is reduced Fig 14 shows the

false-alarm test result whether the proposed damage detection algorithm works even in pris-

tine condition of the test specimen It is shown that no false alarm is occurred because no ad-

ditional reflected wave is captured

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Transmitted Wave

(a) Transmitted wave

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Reflected Wave

No reflected wave

(b) Reflected wave

Fig 14 False alarm test for the proposed damage detection algorithm in pristine condition of the specimen Reflected wave from damage is not captured false alarm is not occurred

- 26 -

TABLE V Damage characterization for the impact-induced damage specimen

of Impacts m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

2

(1470 J)

Method 1 080 65~90 75~246 470~1157

Method 2 080 801~885 114~245 711~1134

Method 3 080 843 179 1122

3

(2205 J)

Method 1 148 50~100 06~380 21~1134

Method 2 148 758~838 249~368 843~1125

Method 3 148 798 308 1044

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1213 ms

Table V presents the characterization results of the impact-induced damage case Here the

measured 119907119907119894119894 and the assumed variation are same as the artificial damage case It is observed

that as the number of impact is increased 119889119889 is increased while 119907119907119889119889 is reduced Thus it is

validated that the proposed damage characterization approach estimates the increment of the

damage well

54 Temperature effect test

541 Teflon inserted damage

11 115 12 125 13-10

-5

0

5

10

Time (ms)

Vol

tage

(mV

)

0 oC20 oC50 oC

(a) Pitch-catch signal

11 115 12 125 13 135

-1

0

1

2

3

Time(ms)

Volta

ge (m

V)

0 oC20 oC50 oC

(b) Pulse-echo signal

Fig 15 Signals obtained from the Teflon inserted damage specimen under temperature varia-tion Time delay and amplitude variation are occurred

Fig 15 shows the obtained signals from the Teflon inserted damage specimen subjected to

- 27 -

temperature variation Time delay and amplitude variations are occurred there is no addition-

al damage is introduced The analyzed results of the obtained signals are reported in Table VI

The measured 119907119907119894119894 is 1247 ms at 0 degC 1213 ms at 20 degC and 1168 ms at 50 degC respectively

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms (1152~1273 ms)

The estimated results are also changed under temperature variations the fluctuation of esti-

mated range and exact value is small

TABLE VI Temperature effect test for the Teflon inserted damage specimen

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

0 degC

Exact 635 25

Method 1 113 60~70 79~268 352~1291

Method 2 113 560~619 233~344 1034~1172

Method 3 113 606 257 1141

20 degC

Exact 635 25

Method 1 123 60~75 73~338 298~1220

Method 2 123 619~662 229~310 934~1188

Method 3 123 631 284 1161

50 degC

Exact 635 25

Method 1 137 60~75 72~328 263~1228

Method 2 137 622~687 183~299 666~1086

Method 3 137 642 264 960

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1247 ms at 0 degC 1213 ms at 20 degC and 1168 ms at 50 degC

- 28 -

542 Impact-induced damage

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Vol

tage

(mV

)

Transmitted WaveReflected Wave

No reflected wave

(a) 0 degC

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Transmitted WaveReflected Wave

No reflected wave

(b) 50 degC

Fig 16 False alarm test under temperature variations It is shown that proposed damage char-acterization algorithm is robust under temperature variations

The false alarm test in pristine condition of the impact-induced damage specimen under

temperature variation was investigated (Fig 16) It is validated that the proposed damage di-

agnosis algorithm is robust under temperature variations Fig 17 shows the obtained signals

from the impact-induced damage (impacted 3 times) specimen under temperature variation

The analyzed results of the obtained signals are reported in Table VII The measured 119907119907119894119894 in

each temperature and the assumed variation are same as the artificial damage case

11 115 12 125 13-5

0

5

Time (ms)

Volta

ge (m

V)

0 oC20 oC50 oC

(a) Pitch-catch signal

11 115 12 125 13 135

-1

0

1

2

3

Time(ms)

Vol

tage

(mV

)

0 oC20 oC50 oC

(b) Pulse-echo signal

Fig 17 Signals obtained from impact-induced damage specimen (Impact 3) under tempera-ture variation

- 29 -

TABLE VII Temperature effect test for the impact induced damage specimen (Impact 3 case)

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

0 degC

Method 1 135 75~90 72~310 267~1148

Method 2 135 705~779 264~381 978~1126

Method 3 135 763 289 1071

20 degC

Method 1 148 50~100 06~380 21~1134

Method 2 148 758~838 249~368 843~1125

Method 3 148 798 308 1044

50 degC

Method 1 153 75~100 53~415 174~1227

Method 2 153 804~889 214~336 702~1103

Method 3 153 815 320 1050

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1247 ms at 0 degC 1213 ms at 20 degC and 1168 ms at 50 degC

- 30 -

CHAPTER 6 CONCLUDING REMARKS

61 Executive summary

In this study the reference-free damage detection localization and quantification tech-

nique in a multi layered composite plate is proposed Additional reflected A0 Lamb wave

mode from the damage due to multiple reflections in damage area is used as a feature for

characterizing the damage First damage is characterized in terms of a damage index (m-

value) defined as the ratio of the physical length of the damage to the wavelength of the A0

Lamb wave mode Then the range of possible damage location length and severity are esti-

mated with constraints This procedure can be accomplished without prior baseline data To

realize the proposed technique the matching pursuit technique is modified for reduction of

the error by considering the dispersive characteristic of Lamb wave The matching pursuit

method estimates the arrival time of each wave packets accurately even two wave packets are

overlapped The numerical simulations and experimental tests including Teflon tape inserted

and impact-induced damage are investigated for the feasibility of the proposed technique

Moreover the effectiveness of the proposed technique under changing temperature is exper-

imentally verified

- 31 -

62 Suggestion for further study

The limitations of this study and the possible improvement and further studies are listed as

follow

1 Measurement of impact-induced damage length using conventional NDT technique

Additional experimental verification for impact-induced damage characterization is

needed by measuring the real length of damage using conventional NDT technique

such as ultrasonic scan or infrared camera

2 Limitation of both side collocated PZT installation In this study excitation PZTs are

installed both side and collocated for decomposing A0 Lamb wave mode However it

can produce decomposition error due to imperfection of PZT location and their installa-

tion Moreover in many cases assess to both surface of structures is not allowed such

as aircraft applications To improve this limitation a reference-free damage characteri-

zation using single side PZTs will be considered

3 Complex structure application In this study damage is characterized in simple plate-

like structure When the monitored structure becomes complex the response signals al-

so becomes complex due to its geometry and boundary Thus the investigations for

complex structure application such as stiffener spar and curvature are needed

- 32 -

REFERENCES

[1] Cawley P and Alleyne D (1996) ldquoThe use of Lamb waves for the long range inspection of large structuresrdquo Ultrasonics 34 pp287-290

[2] Ghosh T Kundu T and Karpur P (1998) ldquoEfficient use of Lamb waves for detecting defects in large platesrdquo Ultrasonics 36 pp791-801

[3] Guirguitui V (2003) ldquoLamb wave generation with piezoelectric wafer active sensors for structural health monitoringrdquo SPIE Conference on Smart Structures and Materials San Diego CA 5391 pp111-422

[4] Ragvahan A and Cesnik CE S (2004) ldquoModeling of piezoelectric-based Lamb wave generation and sensing for structural health monitoringrdquo SPIE Conference on Smart Struc-tures and Materials San Diego CA 5391 pp419-430

[5] Rose JL Pilarski A and Dirit JJ (1993) ldquoAn approach to guided wave mode selection for inspection of laminated platerdquo Journal of Reinforced Plastics and Composites 12(5) pp536-544

[6] Guo N and Cawley P (1993) ldquoThe interaction of Lamb waves with delaminations in composite laminatesrdquo Journal of Acoustical Society of America 94(4) pp 2240-2246

[7] Valdez SHD and Soutis C (2002) ldquoReal-time nondestructive evaluation of fiber com-posite laminates using low-frequency Lamb wavesrdquo Journal of Acoustical Society of Amer-ica 111(5) pp 2026-2033

[8] Yang M and Qiao P (2005) ldquoModeling and experimental detection of damage in virous materials using the pulse-echo method and piezoelectric sensoractuatorsrdquo Smart Materials and Structures 14 pp1083-1100

[9] Lestari W and Qiao P (2005) ldquoApplication of wave propagation analysis for damage identification in composite laminated beamsrdquo Journal of Composite Materials 39(22) pp1967-1984

[10] Kessler S S Spearing S M and Soutis C (2002) ldquoDamage detection in composite ma-terials using Lamb wave methodsrdquo Smart Materials and Structures 11 pp269-278

[11] Sohn H Park G Wail J R Limback N P and Farrar C R (2004) ldquoWavelet-based active sensing for delamination detection in composite structuresrdquo Smart Materials and Structures 13 pp 153-160

[12] Petculescu G Krishnaswamy S and Achenback J D (2008) ldquoGroup delay measure-ments using modally selective Lamb wave transducers for detection and sizing of delamina-tions in compositesrdquo Smart Materials and Structures 17

[13] Sohn H Park H W Law K H and Farrar C R (2007) ldquoDamage detection in compo-

- 33 -

site plates by using an enhanced time reversal methodrdquo Journal of Aerospace Engineering (ASCE) pp 141-151

[14] Sohn H (2007) ldquoEffects of environmental and operational variability on structural health monitoringrdquo Philosophical Transactions of the Royal Society Ardquo 365 pp539-560

[15] Rytter A (1993) ldquoVibration based inspection of civil engineering structuresrdquo PhD The-sis Department of Building Technology and Structural Engineering University of Aalborg Aalborg Denmark

[16] Hayashi T and Kawashima K (2002) ldquoMultiple reflections of Lamb waves at a delami-nationrdquo Ultrasonics 40 pp193-197

[17] Wang C H and Rose L R F (2003) ldquoWave reflection and transmission in beams con-taining delamination and inhomogeneityrdquo Journal of Sound and Vibration 264 pp 851-872

[18] Ramadas C Balasubramaniam K Joshi M and Krishnamurthy C V (2009) ldquoInterac-tion of the primary anti-symmetric Lamb mode (Ao) with symmetric delaminations numer-ical and experimental studiesrdquo Smart Materials and Structures 18

[19] Ramadas C Balasubramaniam K Joshi M and Krishnamurthy C V (2010) ldquoInterac-tion of guided Lamb waves with an asymmetrically located delamination in a laminated composite platerdquo Smart Materials and Structures 19

[20] Wilcox P Lowe M and Cawely P (2001) ldquoThe effect of dispersion on long-range in-spection using ultrasonic guided wavesrdquo NDTampE International 34 pp1-9

[21] Mallat S G and Zhang Z (1993) ldquoMatching pursuits with time-frequency dictionariesrdquo IEEE Tansactions on Signal Processing 41(12) pp3397-3415

[22] Zhang G Zhang S and Wang Y (2000) ldquoApplication of adaptive time-frequency de-composition in ultrasonic NDE of highly scattering materialsrdquo Ultrasonics 38 pp 961-964

[23] Hong J C Sun K H and Kim Y Y (2005) ldquoThe matching pursuit approach based on the modulated Gaussian pulse for efficient guided-wave damage inspectionrdquo Smart Mate-rials and Structures 14 pp548-560

[24] Herakovich C T (1997) Mechanics of Fibrous Composites Wiley 480 pages

[25] Reddy J N (1985) ldquoA review of the literature on finite-element modeling of laminated composite platesrdquo The Shock and Vibration Digest 17(4) pp 3-8

[26] Reddy J N (2003) Mechanics of Laminated Composite Plates and Shells Theory and Analysis CRC Press 856 pages

[27] Palacz M Krawczuk M and Ostachowicz W (2005) ldquoThe spectral finite element mod-el for analysis of flexure-shear coupled wave propagation Part 1 laminated multilayer

- 34 -

composite beamrdquo Composite Structures 68 pp37-44

[28] Zhang Y C and Harding J (1990) ldquoA numerical micromechanics analysis of the me-chanical properties of a plain weave compositerdquo Computer amp Structures 36(5) pp 839-844

[29] Nemat-Nasser S and Hori M (1999) Micromechanics Overall Properties of Heteroge-neous Materials Elsevier 810 pages

[30] Scida D Aboura Z Benzeggagh M L and Bocherens E (1999) ldquoA micromechanics model for 3D elasticity and failure of woven-fibre composite materialrdquo Composites Science and Technology 59 pp 505-517

[31] Thomas Y and Wu X H (1997) ldquoA multiscale finite element method for elliptic prob-lems in composite materials and porous mediardquo Journal of Computational Physics 134 pp 169-189

[32] Lee K H Moorthy S and Ghosh S (1999) ldquoMultiple scale computational model for damage in composite materialsrdquo Computer Methods in Applied Mechanics and Engineer-ing 172 pp 175-201

[33] Namilae S and Chandra N (2005) ldquoMultiscale model to study the effect of interfaces in carbon nanotube-based compositesrdquo Journal of Engineering Materials and Technology (ASME) 127 pp222-232

[34] Adams D E (2007) Health monitoring of Structural Materials and Components Wiley 460 pages

[35] Gopalakrishnan S (2009) ldquoModeling Aspects in Finite Elementrdquo Encyclopedia of Struc-tural Health Monitoring 2 pp791-809

[36] Palacz M Krawczuk M and Ostachowicz W (2005) ldquoThe spectral finite element mod-el for analysis of flexure-shear coupled wave propagation Part 2 Deminated multilayer composite beamrdquo Composite Structures 68 pp45-51

[37] Baik J M and Thopson R B (1984) ldquoUltrasonic scattering from imperfect interfaces A quasi-static modelrdquo Journal of Nondestructive Evaluation 4(3) pp177-196

[38] Luo H and Hanagud S (2000) ldquoDynamics of delamincated beamsrdquo International Jour-nal of Solids and Structures 37 pp1501-1519

[39] Kim S B and Sohn H (2007) ldquoInstantaneous reference-free crack detection based on polarization characteristics of piezoelectric materialsrdquo Smart Materials and Structures 16 pp2375-2387

[40] An Y K and Sohn H (2010) ldquoInstantaneous crack detection under varying temperature and static loading conditionsrdquo The Special Issue of Structural Control and Health Monitor-ing 17(7) pp 730-741

- 35 -

[41] Lee S J and Sohn H (2006) ldquoActive self-sensing scheme development for structural health monitoringrdquo Smart Materials and Structures 15(6) pp 1734-1746

- 36 -

요 약 문

복합재료 손상의 무기저 진단 및 정량화

복합재료는 재료의 경량성 내구성 고강도의 특성으로 인하여 항공기에서부터

교량에까지 널리 사용되고 있다 하지만 외부의 충격으로 인한 복합재료 내부의

손상은 재료의 강도를 감소시켜 구조물의 안정성을 저하시키는 요인이 된다 하지만

복합재료의 손상은 재료 내부에서 발생하여 육안으로는 식별하기도 힘들며 표면에서

보이는 것만으로 내부의 손상 정도를 판단하기는 불가능하다는 단점을 가지고 있다

본 연구에서는 손상 내부에서 반사로 인해 발생하는 파를 이용하여 기저자료를

사용하지 않고 특정 시점에서 얻은 자료만으로 복합재료의 손상을 진단하고 정량화

할 수 있는 기법을 제안한다 제안된 기법의 적용성을 검증하기 위하여 유한요소법을

이용한 수치해석을 수행하였고 Teflon tape 이 삽입된 시편과 실제 충격하중으로

인한 손상을 입은 시편을 이용한 실험을 진행하였다 수치해석 및 실험 결과 제안된

기법이 복합재료의 손상을 성공적으로 진단 및 정량화할 수 있음을 확인하였다 또한

다양한 외부 온도 환경에서 제안된 기법의 환경영향 실험을 수행하였고 다양한

환경에서 손상의 진단 및 정량화가 성공적으로 이루어짐을 검증하였다

핵심어 무기저 손상진단 및 정량화 압전센서 복합재료 유한요소법 Matching pursuit

method

- 37 -

감 사 의 글

기대와 두려움을 안고 시작한 석사과정이 어느덧 2 년 이라는 세월 속에서 추억으로 자리잡아

가고 있습니다 학부생 기간을 포함한 지난 3 년 동안 SHM 이라는 새로운 분야를 접하고 연구를

수행 할 수 있도록 도와주신 고마운 분들이 있어 그 감사를 표하고자 합니다

우선 저의 많은 부족한 점을 일깨워 주시고 연구에 대한 뜻을 더욱 확고히 세울 수 있게 해

주신 지도교수님이신 손 훈 교수님께 깊은 감사의 뜻을 전합니다 짧은 기간이었지만 교수님의

도움으로 많은 것을 배우고 느끼고 깨달을 수 있었습니다 또한 본 연구를 수행할 수 있도록 지

원해주신 국토해양부 (U-City 석 박사과정 지원사업)에 깊은 감사 드립니다

때로 힘들고 지칠 때 항상 저를 믿고 응원해 주시는 가족이 있었기에 지금의 글을 쓸 수 있었

습니다 또한 지금 연구실에서 함께 생활하고 있는 윤규형 민구형 철민이형 현석이형 진열이형

호민이형 병진이 선혜누나와 성대에 계신 창길이형 미국에 계신 의재형과 Debaditiya 에게도 감

사의 말씀을 전합니다 그리고 모두 나열할 수는 없지만 카이스트 생활의 활력소가 되어준 카이

스트 건설 및 환경공학과 선후배들 직원 분들과 SOS 멤버들에게도 감사를 표합니다

앞서 거론한 많은 이들의 도움이 없었다면 저 또한 지금 이 자리에 있을 수 없었을 것입니다

많이 부족하기에 더 노력하고 배우고자 했던 초심을 다시 한번 생각하며 더 큰 결실을 맺기 위해

앞으로 나아가겠습니다 도움을 주신 모든 분들께 이 감사의 글을 바칩니다

2011년 6월 카이스트에서 임형진 올림

- 38 -

CURRICULUM VITAE

Hyung Jin Lim

Research Assistant

The Civil and Environmental Engineering Department

Korea Advanced Institute of Science and Technology (KAIST)

291 Daehak-Ro Guseong-Dong Yuseong-Gu

Daejeon Republic of Korea 305-701

Phone (82)+42-350-3665 Fax (82)+42-350-3610

Email limhj87gmailcom

RESEARCH INTERESTS

Damage Detection Localization and Quantification on Composite and Metallic Structures Pattern Recognition Finite Element Modeling (FEM)

EDUCATION

2009 ndash 2011 MS Civil and Environmental Eng Korea Advanced Institute of Science and Technology (KAIST) Korea Rep

2006 ndash 2009 BS Civil and Environmental Eng Korea Advanced Institute of Science and Technology (KAIST) Korea Rep

JOURNAL PUBLICATIONS The corresponding author is underlined 1 Hyung Jin Lim Hoon Sohn ldquoReference-free damage detection localization and quantification

using in compositerdquo in preparation for Journal of the Acoustical Society of America 2011

2 Chul Min Yeum Hoon Sohn Jeong Bum Ihn Hyung Jin Lim ldquoInstantaneous delamination detection in a composite using a dual piezoelectric transducer networkrdquo Submitted to Compo-sites Science and Technology 2011

3 Yun-Kyu An Hyung Jin Lim Min Koo Kim Jin Yeol Yang Hoon Sohn Chang Guen Lee ldquoInvestigation of applicability of local reference-free damage detection techniques to in-situ

- 39 -

bridgesrdquo Submitted to ASCE Journal of Structural Engineering 2011

4 Hyung Jin Lim Min Koo Kim Hoon Sohn Chan Yik Park ldquoImpedance based damage detec-tion under varying temperature and loading conditionsrdquo Submitted to NDTampE International 2011

5 Hyung Jin Lim Hoon Sohn ldquoQuantitative estimation of transmitted and reflected lamb waves at discontinuityrdquo Journal of the Korean Society for Nondestructive Testing Vol30 No 4 pp1-8 2010

CONFERENCE PROCEEDINGS

The corresponding author is underlined 1 Hyung Jin Lim and Hoon Sohn ldquoReference-free damage detection localization and quantifi-

cation in compositesrdquo ASCE Engineering Mechanics Institute Conference Boston June 2-4 2011

2 Hyung Jin Lim and Hoon Sohn ldquoReference-free delamination quantification within multi-layer composite materialsrdquo KSNT Spring Conference May 26-27 Gyeong-Ju Korea 2011

3 Min Koo Kim Hyun Jin Lim Hoon Sohn Chan Yik Park ldquoImpedance-based bolt loosening detection under varying temperature and loadingrdquo Asian Pacific Workshop on Structural Health Monitoring Tokyo Japan November 30-December 2 2010

4 Hyung Jin Lim Chul Min Yeum and Hoon Sohn ldquoModeling of Impact-induced Delamination in a Multilayer Composite Platerdquo The 23rd KKCNN Symposium on Civil Engineering Taipei (Taiwan) November 13-15 2010

5 Yun-Kyu An Hyung Jin Lim and Hoon Sohn ldquoTemperature independent Delamination Detec-tion using data normalizationrdquo The 3rd International Conference on Multi-Functional Materi-als and Structures Chonbuk National Univ Jeonju Korea September 14-18 2010

6 Hoon Sohn Yun-Kyu An Chang Gil Lee Min Koo Kim Hyung Jin Lim ldquoApplication of online NDT techniques to decommissioned bridge testingrdquo the 5th European Workshop on Structural Health Monitoring Sorrento Italy June 29-July 02 2010

7 Min Koo Kim Hyung Jin Lim Hoon Sohn Chan Yik Park ldquoImpedance based damage diag-nosis of a complex composite aircraft wing under changing loading and temperature condi-tionsrdquo the 5th European Workshop on Structural Health Monitoring Sorrento Italy June 29-July 02 2010

8 Yun-Kyu An Min Koo Kim Hyung Jin Lim and Hoon Sohn ldquoValidations of reference-free crack detection techniques through a decommissioned bridge testrdquo 2010 COSEIK Annual Con-ference Jeju Korea April 8-9 2010

- 40 -

PARTICIPANT PROJECTS

1 Development of Smart Steel Members for Online Steel Structure Monitoring POSCO Corporation (Funded 41000000 KRW (41000 USD) for 010511 to 123111)

2 Development of Enabling Technologies for Structural Health Monitoring Korean Agency for Defense Development (Funded 240000000 KRW (240000 USD) out of total 3200000000 KRW (3200000 USD) for 060108-053111)

3 Development of On-board SHM Technologies for Composite Air Vehicles The Boeing Company (Funded 400000 USD (400000000 KRW) for 080108 to 073111)

EDUCATIONAL EXPERIENCES

Fall 2009

Teaching Assistant CE207 ldquoElementary Structural Engineering and Laboratoryrdquo KAIST 3 unit undergraduate course

EXPERIMENTAL EXPERIENTCES

2010 Monitoring of Real Aircraft Composite Wing Segment Korea Agency for Defense De-velopment (ADD) Daejeon Korea Real time monitoring of composite air vehicles fund-ed by Korea Agency for Defense Development (ADD)

2009-2011

Field Monitoring of Real Bridge Structures Gimpo (Ramp G bridge) amp Icheon (Sam-seung bridge) Korea US-Korea collaborative research for bridge monitoring testbeds

HONORS amp AWARDS

Scholarships for outstanding students granted KAIST (2007-2009) The three prominent students in the civil and environmental engineering department of KAIST were selected as recipients This scholarship provided 2000$ per year for three years

- 41 -

Page 38: 복합재료 손상의 무기저 진단 및 정량화 - KAISTssslab.kaist.ac.kr/article/pdf/Thesis_2.pdf · 1 Declaration of Ethical Conduct in Research: I, as a graduate student

TABLE V Damage characterization for the impact-induced damage specimen

of Impacts m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

2

(1470 J)

Method 1 080 65~90 75~246 470~1157

Method 2 080 801~885 114~245 711~1134

Method 3 080 843 179 1122

3

(2205 J)

Method 1 148 50~100 06~380 21~1134

Method 2 148 758~838 249~368 843~1125

Method 3 148 798 308 1044

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1213 ms

Table V presents the characterization results of the impact-induced damage case Here the

measured 119907119907119894119894 and the assumed variation are same as the artificial damage case It is observed

that as the number of impact is increased 119889119889 is increased while 119907119907119889119889 is reduced Thus it is

validated that the proposed damage characterization approach estimates the increment of the

damage well

54 Temperature effect test

541 Teflon inserted damage

11 115 12 125 13-10

-5

0

5

10

Time (ms)

Vol

tage

(mV

)

0 oC20 oC50 oC

(a) Pitch-catch signal

11 115 12 125 13 135

-1

0

1

2

3

Time(ms)

Volta

ge (m

V)

0 oC20 oC50 oC

(b) Pulse-echo signal

Fig 15 Signals obtained from the Teflon inserted damage specimen under temperature varia-tion Time delay and amplitude variation are occurred

Fig 15 shows the obtained signals from the Teflon inserted damage specimen subjected to

- 27 -

temperature variation Time delay and amplitude variations are occurred there is no addition-

al damage is introduced The analyzed results of the obtained signals are reported in Table VI

The measured 119907119907119894119894 is 1247 ms at 0 degC 1213 ms at 20 degC and 1168 ms at 50 degC respectively

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms (1152~1273 ms)

The estimated results are also changed under temperature variations the fluctuation of esti-

mated range and exact value is small

TABLE VI Temperature effect test for the Teflon inserted damage specimen

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

0 degC

Exact 635 25

Method 1 113 60~70 79~268 352~1291

Method 2 113 560~619 233~344 1034~1172

Method 3 113 606 257 1141

20 degC

Exact 635 25

Method 1 123 60~75 73~338 298~1220

Method 2 123 619~662 229~310 934~1188

Method 3 123 631 284 1161

50 degC

Exact 635 25

Method 1 137 60~75 72~328 263~1228

Method 2 137 622~687 183~299 666~1086

Method 3 137 642 264 960

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1247 ms at 0 degC 1213 ms at 20 degC and 1168 ms at 50 degC

- 28 -

542 Impact-induced damage

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Vol

tage

(mV

)

Transmitted WaveReflected Wave

No reflected wave

(a) 0 degC

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Transmitted WaveReflected Wave

No reflected wave

(b) 50 degC

Fig 16 False alarm test under temperature variations It is shown that proposed damage char-acterization algorithm is robust under temperature variations

The false alarm test in pristine condition of the impact-induced damage specimen under

temperature variation was investigated (Fig 16) It is validated that the proposed damage di-

agnosis algorithm is robust under temperature variations Fig 17 shows the obtained signals

from the impact-induced damage (impacted 3 times) specimen under temperature variation

The analyzed results of the obtained signals are reported in Table VII The measured 119907119907119894119894 in

each temperature and the assumed variation are same as the artificial damage case

11 115 12 125 13-5

0

5

Time (ms)

Volta

ge (m

V)

0 oC20 oC50 oC

(a) Pitch-catch signal

11 115 12 125 13 135

-1

0

1

2

3

Time(ms)

Vol

tage

(mV

)

0 oC20 oC50 oC

(b) Pulse-echo signal

Fig 17 Signals obtained from impact-induced damage specimen (Impact 3) under tempera-ture variation

- 29 -

TABLE VII Temperature effect test for the impact induced damage specimen (Impact 3 case)

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

0 degC

Method 1 135 75~90 72~310 267~1148

Method 2 135 705~779 264~381 978~1126

Method 3 135 763 289 1071

20 degC

Method 1 148 50~100 06~380 21~1134

Method 2 148 758~838 249~368 843~1125

Method 3 148 798 308 1044

50 degC

Method 1 153 75~100 53~415 174~1227

Method 2 153 804~889 214~336 702~1103

Method 3 153 815 320 1050

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1247 ms at 0 degC 1213 ms at 20 degC and 1168 ms at 50 degC

- 30 -

CHAPTER 6 CONCLUDING REMARKS

61 Executive summary

In this study the reference-free damage detection localization and quantification tech-

nique in a multi layered composite plate is proposed Additional reflected A0 Lamb wave

mode from the damage due to multiple reflections in damage area is used as a feature for

characterizing the damage First damage is characterized in terms of a damage index (m-

value) defined as the ratio of the physical length of the damage to the wavelength of the A0

Lamb wave mode Then the range of possible damage location length and severity are esti-

mated with constraints This procedure can be accomplished without prior baseline data To

realize the proposed technique the matching pursuit technique is modified for reduction of

the error by considering the dispersive characteristic of Lamb wave The matching pursuit

method estimates the arrival time of each wave packets accurately even two wave packets are

overlapped The numerical simulations and experimental tests including Teflon tape inserted

and impact-induced damage are investigated for the feasibility of the proposed technique

Moreover the effectiveness of the proposed technique under changing temperature is exper-

imentally verified

- 31 -

62 Suggestion for further study

The limitations of this study and the possible improvement and further studies are listed as

follow

1 Measurement of impact-induced damage length using conventional NDT technique

Additional experimental verification for impact-induced damage characterization is

needed by measuring the real length of damage using conventional NDT technique

such as ultrasonic scan or infrared camera

2 Limitation of both side collocated PZT installation In this study excitation PZTs are

installed both side and collocated for decomposing A0 Lamb wave mode However it

can produce decomposition error due to imperfection of PZT location and their installa-

tion Moreover in many cases assess to both surface of structures is not allowed such

as aircraft applications To improve this limitation a reference-free damage characteri-

zation using single side PZTs will be considered

3 Complex structure application In this study damage is characterized in simple plate-

like structure When the monitored structure becomes complex the response signals al-

so becomes complex due to its geometry and boundary Thus the investigations for

complex structure application such as stiffener spar and curvature are needed

- 32 -

REFERENCES

[1] Cawley P and Alleyne D (1996) ldquoThe use of Lamb waves for the long range inspection of large structuresrdquo Ultrasonics 34 pp287-290

[2] Ghosh T Kundu T and Karpur P (1998) ldquoEfficient use of Lamb waves for detecting defects in large platesrdquo Ultrasonics 36 pp791-801

[3] Guirguitui V (2003) ldquoLamb wave generation with piezoelectric wafer active sensors for structural health monitoringrdquo SPIE Conference on Smart Structures and Materials San Diego CA 5391 pp111-422

[4] Ragvahan A and Cesnik CE S (2004) ldquoModeling of piezoelectric-based Lamb wave generation and sensing for structural health monitoringrdquo SPIE Conference on Smart Struc-tures and Materials San Diego CA 5391 pp419-430

[5] Rose JL Pilarski A and Dirit JJ (1993) ldquoAn approach to guided wave mode selection for inspection of laminated platerdquo Journal of Reinforced Plastics and Composites 12(5) pp536-544

[6] Guo N and Cawley P (1993) ldquoThe interaction of Lamb waves with delaminations in composite laminatesrdquo Journal of Acoustical Society of America 94(4) pp 2240-2246

[7] Valdez SHD and Soutis C (2002) ldquoReal-time nondestructive evaluation of fiber com-posite laminates using low-frequency Lamb wavesrdquo Journal of Acoustical Society of Amer-ica 111(5) pp 2026-2033

[8] Yang M and Qiao P (2005) ldquoModeling and experimental detection of damage in virous materials using the pulse-echo method and piezoelectric sensoractuatorsrdquo Smart Materials and Structures 14 pp1083-1100

[9] Lestari W and Qiao P (2005) ldquoApplication of wave propagation analysis for damage identification in composite laminated beamsrdquo Journal of Composite Materials 39(22) pp1967-1984

[10] Kessler S S Spearing S M and Soutis C (2002) ldquoDamage detection in composite ma-terials using Lamb wave methodsrdquo Smart Materials and Structures 11 pp269-278

[11] Sohn H Park G Wail J R Limback N P and Farrar C R (2004) ldquoWavelet-based active sensing for delamination detection in composite structuresrdquo Smart Materials and Structures 13 pp 153-160

[12] Petculescu G Krishnaswamy S and Achenback J D (2008) ldquoGroup delay measure-ments using modally selective Lamb wave transducers for detection and sizing of delamina-tions in compositesrdquo Smart Materials and Structures 17

[13] Sohn H Park H W Law K H and Farrar C R (2007) ldquoDamage detection in compo-

- 33 -

site plates by using an enhanced time reversal methodrdquo Journal of Aerospace Engineering (ASCE) pp 141-151

[14] Sohn H (2007) ldquoEffects of environmental and operational variability on structural health monitoringrdquo Philosophical Transactions of the Royal Society Ardquo 365 pp539-560

[15] Rytter A (1993) ldquoVibration based inspection of civil engineering structuresrdquo PhD The-sis Department of Building Technology and Structural Engineering University of Aalborg Aalborg Denmark

[16] Hayashi T and Kawashima K (2002) ldquoMultiple reflections of Lamb waves at a delami-nationrdquo Ultrasonics 40 pp193-197

[17] Wang C H and Rose L R F (2003) ldquoWave reflection and transmission in beams con-taining delamination and inhomogeneityrdquo Journal of Sound and Vibration 264 pp 851-872

[18] Ramadas C Balasubramaniam K Joshi M and Krishnamurthy C V (2009) ldquoInterac-tion of the primary anti-symmetric Lamb mode (Ao) with symmetric delaminations numer-ical and experimental studiesrdquo Smart Materials and Structures 18

[19] Ramadas C Balasubramaniam K Joshi M and Krishnamurthy C V (2010) ldquoInterac-tion of guided Lamb waves with an asymmetrically located delamination in a laminated composite platerdquo Smart Materials and Structures 19

[20] Wilcox P Lowe M and Cawely P (2001) ldquoThe effect of dispersion on long-range in-spection using ultrasonic guided wavesrdquo NDTampE International 34 pp1-9

[21] Mallat S G and Zhang Z (1993) ldquoMatching pursuits with time-frequency dictionariesrdquo IEEE Tansactions on Signal Processing 41(12) pp3397-3415

[22] Zhang G Zhang S and Wang Y (2000) ldquoApplication of adaptive time-frequency de-composition in ultrasonic NDE of highly scattering materialsrdquo Ultrasonics 38 pp 961-964

[23] Hong J C Sun K H and Kim Y Y (2005) ldquoThe matching pursuit approach based on the modulated Gaussian pulse for efficient guided-wave damage inspectionrdquo Smart Mate-rials and Structures 14 pp548-560

[24] Herakovich C T (1997) Mechanics of Fibrous Composites Wiley 480 pages

[25] Reddy J N (1985) ldquoA review of the literature on finite-element modeling of laminated composite platesrdquo The Shock and Vibration Digest 17(4) pp 3-8

[26] Reddy J N (2003) Mechanics of Laminated Composite Plates and Shells Theory and Analysis CRC Press 856 pages

[27] Palacz M Krawczuk M and Ostachowicz W (2005) ldquoThe spectral finite element mod-el for analysis of flexure-shear coupled wave propagation Part 1 laminated multilayer

- 34 -

composite beamrdquo Composite Structures 68 pp37-44

[28] Zhang Y C and Harding J (1990) ldquoA numerical micromechanics analysis of the me-chanical properties of a plain weave compositerdquo Computer amp Structures 36(5) pp 839-844

[29] Nemat-Nasser S and Hori M (1999) Micromechanics Overall Properties of Heteroge-neous Materials Elsevier 810 pages

[30] Scida D Aboura Z Benzeggagh M L and Bocherens E (1999) ldquoA micromechanics model for 3D elasticity and failure of woven-fibre composite materialrdquo Composites Science and Technology 59 pp 505-517

[31] Thomas Y and Wu X H (1997) ldquoA multiscale finite element method for elliptic prob-lems in composite materials and porous mediardquo Journal of Computational Physics 134 pp 169-189

[32] Lee K H Moorthy S and Ghosh S (1999) ldquoMultiple scale computational model for damage in composite materialsrdquo Computer Methods in Applied Mechanics and Engineer-ing 172 pp 175-201

[33] Namilae S and Chandra N (2005) ldquoMultiscale model to study the effect of interfaces in carbon nanotube-based compositesrdquo Journal of Engineering Materials and Technology (ASME) 127 pp222-232

[34] Adams D E (2007) Health monitoring of Structural Materials and Components Wiley 460 pages

[35] Gopalakrishnan S (2009) ldquoModeling Aspects in Finite Elementrdquo Encyclopedia of Struc-tural Health Monitoring 2 pp791-809

[36] Palacz M Krawczuk M and Ostachowicz W (2005) ldquoThe spectral finite element mod-el for analysis of flexure-shear coupled wave propagation Part 2 Deminated multilayer composite beamrdquo Composite Structures 68 pp45-51

[37] Baik J M and Thopson R B (1984) ldquoUltrasonic scattering from imperfect interfaces A quasi-static modelrdquo Journal of Nondestructive Evaluation 4(3) pp177-196

[38] Luo H and Hanagud S (2000) ldquoDynamics of delamincated beamsrdquo International Jour-nal of Solids and Structures 37 pp1501-1519

[39] Kim S B and Sohn H (2007) ldquoInstantaneous reference-free crack detection based on polarization characteristics of piezoelectric materialsrdquo Smart Materials and Structures 16 pp2375-2387

[40] An Y K and Sohn H (2010) ldquoInstantaneous crack detection under varying temperature and static loading conditionsrdquo The Special Issue of Structural Control and Health Monitor-ing 17(7) pp 730-741

- 35 -

[41] Lee S J and Sohn H (2006) ldquoActive self-sensing scheme development for structural health monitoringrdquo Smart Materials and Structures 15(6) pp 1734-1746

- 36 -

요 약 문

복합재료 손상의 무기저 진단 및 정량화

복합재료는 재료의 경량성 내구성 고강도의 특성으로 인하여 항공기에서부터

교량에까지 널리 사용되고 있다 하지만 외부의 충격으로 인한 복합재료 내부의

손상은 재료의 강도를 감소시켜 구조물의 안정성을 저하시키는 요인이 된다 하지만

복합재료의 손상은 재료 내부에서 발생하여 육안으로는 식별하기도 힘들며 표면에서

보이는 것만으로 내부의 손상 정도를 판단하기는 불가능하다는 단점을 가지고 있다

본 연구에서는 손상 내부에서 반사로 인해 발생하는 파를 이용하여 기저자료를

사용하지 않고 특정 시점에서 얻은 자료만으로 복합재료의 손상을 진단하고 정량화

할 수 있는 기법을 제안한다 제안된 기법의 적용성을 검증하기 위하여 유한요소법을

이용한 수치해석을 수행하였고 Teflon tape 이 삽입된 시편과 실제 충격하중으로

인한 손상을 입은 시편을 이용한 실험을 진행하였다 수치해석 및 실험 결과 제안된

기법이 복합재료의 손상을 성공적으로 진단 및 정량화할 수 있음을 확인하였다 또한

다양한 외부 온도 환경에서 제안된 기법의 환경영향 실험을 수행하였고 다양한

환경에서 손상의 진단 및 정량화가 성공적으로 이루어짐을 검증하였다

핵심어 무기저 손상진단 및 정량화 압전센서 복합재료 유한요소법 Matching pursuit

method

- 37 -

감 사 의 글

기대와 두려움을 안고 시작한 석사과정이 어느덧 2 년 이라는 세월 속에서 추억으로 자리잡아

가고 있습니다 학부생 기간을 포함한 지난 3 년 동안 SHM 이라는 새로운 분야를 접하고 연구를

수행 할 수 있도록 도와주신 고마운 분들이 있어 그 감사를 표하고자 합니다

우선 저의 많은 부족한 점을 일깨워 주시고 연구에 대한 뜻을 더욱 확고히 세울 수 있게 해

주신 지도교수님이신 손 훈 교수님께 깊은 감사의 뜻을 전합니다 짧은 기간이었지만 교수님의

도움으로 많은 것을 배우고 느끼고 깨달을 수 있었습니다 또한 본 연구를 수행할 수 있도록 지

원해주신 국토해양부 (U-City 석 박사과정 지원사업)에 깊은 감사 드립니다

때로 힘들고 지칠 때 항상 저를 믿고 응원해 주시는 가족이 있었기에 지금의 글을 쓸 수 있었

습니다 또한 지금 연구실에서 함께 생활하고 있는 윤규형 민구형 철민이형 현석이형 진열이형

호민이형 병진이 선혜누나와 성대에 계신 창길이형 미국에 계신 의재형과 Debaditiya 에게도 감

사의 말씀을 전합니다 그리고 모두 나열할 수는 없지만 카이스트 생활의 활력소가 되어준 카이

스트 건설 및 환경공학과 선후배들 직원 분들과 SOS 멤버들에게도 감사를 표합니다

앞서 거론한 많은 이들의 도움이 없었다면 저 또한 지금 이 자리에 있을 수 없었을 것입니다

많이 부족하기에 더 노력하고 배우고자 했던 초심을 다시 한번 생각하며 더 큰 결실을 맺기 위해

앞으로 나아가겠습니다 도움을 주신 모든 분들께 이 감사의 글을 바칩니다

2011년 6월 카이스트에서 임형진 올림

- 38 -

CURRICULUM VITAE

Hyung Jin Lim

Research Assistant

The Civil and Environmental Engineering Department

Korea Advanced Institute of Science and Technology (KAIST)

291 Daehak-Ro Guseong-Dong Yuseong-Gu

Daejeon Republic of Korea 305-701

Phone (82)+42-350-3665 Fax (82)+42-350-3610

Email limhj87gmailcom

RESEARCH INTERESTS

Damage Detection Localization and Quantification on Composite and Metallic Structures Pattern Recognition Finite Element Modeling (FEM)

EDUCATION

2009 ndash 2011 MS Civil and Environmental Eng Korea Advanced Institute of Science and Technology (KAIST) Korea Rep

2006 ndash 2009 BS Civil and Environmental Eng Korea Advanced Institute of Science and Technology (KAIST) Korea Rep

JOURNAL PUBLICATIONS The corresponding author is underlined 1 Hyung Jin Lim Hoon Sohn ldquoReference-free damage detection localization and quantification

using in compositerdquo in preparation for Journal of the Acoustical Society of America 2011

2 Chul Min Yeum Hoon Sohn Jeong Bum Ihn Hyung Jin Lim ldquoInstantaneous delamination detection in a composite using a dual piezoelectric transducer networkrdquo Submitted to Compo-sites Science and Technology 2011

3 Yun-Kyu An Hyung Jin Lim Min Koo Kim Jin Yeol Yang Hoon Sohn Chang Guen Lee ldquoInvestigation of applicability of local reference-free damage detection techniques to in-situ

- 39 -

bridgesrdquo Submitted to ASCE Journal of Structural Engineering 2011

4 Hyung Jin Lim Min Koo Kim Hoon Sohn Chan Yik Park ldquoImpedance based damage detec-tion under varying temperature and loading conditionsrdquo Submitted to NDTampE International 2011

5 Hyung Jin Lim Hoon Sohn ldquoQuantitative estimation of transmitted and reflected lamb waves at discontinuityrdquo Journal of the Korean Society for Nondestructive Testing Vol30 No 4 pp1-8 2010

CONFERENCE PROCEEDINGS

The corresponding author is underlined 1 Hyung Jin Lim and Hoon Sohn ldquoReference-free damage detection localization and quantifi-

cation in compositesrdquo ASCE Engineering Mechanics Institute Conference Boston June 2-4 2011

2 Hyung Jin Lim and Hoon Sohn ldquoReference-free delamination quantification within multi-layer composite materialsrdquo KSNT Spring Conference May 26-27 Gyeong-Ju Korea 2011

3 Min Koo Kim Hyun Jin Lim Hoon Sohn Chan Yik Park ldquoImpedance-based bolt loosening detection under varying temperature and loadingrdquo Asian Pacific Workshop on Structural Health Monitoring Tokyo Japan November 30-December 2 2010

4 Hyung Jin Lim Chul Min Yeum and Hoon Sohn ldquoModeling of Impact-induced Delamination in a Multilayer Composite Platerdquo The 23rd KKCNN Symposium on Civil Engineering Taipei (Taiwan) November 13-15 2010

5 Yun-Kyu An Hyung Jin Lim and Hoon Sohn ldquoTemperature independent Delamination Detec-tion using data normalizationrdquo The 3rd International Conference on Multi-Functional Materi-als and Structures Chonbuk National Univ Jeonju Korea September 14-18 2010

6 Hoon Sohn Yun-Kyu An Chang Gil Lee Min Koo Kim Hyung Jin Lim ldquoApplication of online NDT techniques to decommissioned bridge testingrdquo the 5th European Workshop on Structural Health Monitoring Sorrento Italy June 29-July 02 2010

7 Min Koo Kim Hyung Jin Lim Hoon Sohn Chan Yik Park ldquoImpedance based damage diag-nosis of a complex composite aircraft wing under changing loading and temperature condi-tionsrdquo the 5th European Workshop on Structural Health Monitoring Sorrento Italy June 29-July 02 2010

8 Yun-Kyu An Min Koo Kim Hyung Jin Lim and Hoon Sohn ldquoValidations of reference-free crack detection techniques through a decommissioned bridge testrdquo 2010 COSEIK Annual Con-ference Jeju Korea April 8-9 2010

- 40 -

PARTICIPANT PROJECTS

1 Development of Smart Steel Members for Online Steel Structure Monitoring POSCO Corporation (Funded 41000000 KRW (41000 USD) for 010511 to 123111)

2 Development of Enabling Technologies for Structural Health Monitoring Korean Agency for Defense Development (Funded 240000000 KRW (240000 USD) out of total 3200000000 KRW (3200000 USD) for 060108-053111)

3 Development of On-board SHM Technologies for Composite Air Vehicles The Boeing Company (Funded 400000 USD (400000000 KRW) for 080108 to 073111)

EDUCATIONAL EXPERIENCES

Fall 2009

Teaching Assistant CE207 ldquoElementary Structural Engineering and Laboratoryrdquo KAIST 3 unit undergraduate course

EXPERIMENTAL EXPERIENTCES

2010 Monitoring of Real Aircraft Composite Wing Segment Korea Agency for Defense De-velopment (ADD) Daejeon Korea Real time monitoring of composite air vehicles fund-ed by Korea Agency for Defense Development (ADD)

2009-2011

Field Monitoring of Real Bridge Structures Gimpo (Ramp G bridge) amp Icheon (Sam-seung bridge) Korea US-Korea collaborative research for bridge monitoring testbeds

HONORS amp AWARDS

Scholarships for outstanding students granted KAIST (2007-2009) The three prominent students in the civil and environmental engineering department of KAIST were selected as recipients This scholarship provided 2000$ per year for three years

- 41 -

Page 39: 복합재료 손상의 무기저 진단 및 정량화 - KAISTssslab.kaist.ac.kr/article/pdf/Thesis_2.pdf · 1 Declaration of Ethical Conduct in Research: I, as a graduate student

temperature variation Time delay and amplitude variations are occurred there is no addition-

al damage is introduced The analyzed results of the obtained signals are reported in Table VI

The measured 119907119907119894119894 is 1247 ms at 0 degC 1213 ms at 20 degC and 1168 ms at 50 degC respectively

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms (1152~1273 ms)

The estimated results are also changed under temperature variations the fluctuation of esti-

mated range and exact value is small

TABLE VI Temperature effect test for the Teflon inserted damage specimen

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

0 degC

Exact 635 25

Method 1 113 60~70 79~268 352~1291

Method 2 113 560~619 233~344 1034~1172

Method 3 113 606 257 1141

20 degC

Exact 635 25

Method 1 123 60~75 73~338 298~1220

Method 2 123 619~662 229~310 934~1188

Method 3 123 631 284 1161

50 degC

Exact 635 25

Method 1 137 60~75 72~328 263~1228

Method 2 137 622~687 183~299 666~1086

Method 3 137 642 264 960

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1247 ms at 0 degC 1213 ms at 20 degC and 1168 ms at 50 degC

- 28 -

542 Impact-induced damage

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Vol

tage

(mV

)

Transmitted WaveReflected Wave

No reflected wave

(a) 0 degC

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Transmitted WaveReflected Wave

No reflected wave

(b) 50 degC

Fig 16 False alarm test under temperature variations It is shown that proposed damage char-acterization algorithm is robust under temperature variations

The false alarm test in pristine condition of the impact-induced damage specimen under

temperature variation was investigated (Fig 16) It is validated that the proposed damage di-

agnosis algorithm is robust under temperature variations Fig 17 shows the obtained signals

from the impact-induced damage (impacted 3 times) specimen under temperature variation

The analyzed results of the obtained signals are reported in Table VII The measured 119907119907119894119894 in

each temperature and the assumed variation are same as the artificial damage case

11 115 12 125 13-5

0

5

Time (ms)

Volta

ge (m

V)

0 oC20 oC50 oC

(a) Pitch-catch signal

11 115 12 125 13 135

-1

0

1

2

3

Time(ms)

Vol

tage

(mV

)

0 oC20 oC50 oC

(b) Pulse-echo signal

Fig 17 Signals obtained from impact-induced damage specimen (Impact 3) under tempera-ture variation

- 29 -

TABLE VII Temperature effect test for the impact induced damage specimen (Impact 3 case)

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

0 degC

Method 1 135 75~90 72~310 267~1148

Method 2 135 705~779 264~381 978~1126

Method 3 135 763 289 1071

20 degC

Method 1 148 50~100 06~380 21~1134

Method 2 148 758~838 249~368 843~1125

Method 3 148 798 308 1044

50 degC

Method 1 153 75~100 53~415 174~1227

Method 2 153 804~889 214~336 702~1103

Method 3 153 815 320 1050

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1247 ms at 0 degC 1213 ms at 20 degC and 1168 ms at 50 degC

- 30 -

CHAPTER 6 CONCLUDING REMARKS

61 Executive summary

In this study the reference-free damage detection localization and quantification tech-

nique in a multi layered composite plate is proposed Additional reflected A0 Lamb wave

mode from the damage due to multiple reflections in damage area is used as a feature for

characterizing the damage First damage is characterized in terms of a damage index (m-

value) defined as the ratio of the physical length of the damage to the wavelength of the A0

Lamb wave mode Then the range of possible damage location length and severity are esti-

mated with constraints This procedure can be accomplished without prior baseline data To

realize the proposed technique the matching pursuit technique is modified for reduction of

the error by considering the dispersive characteristic of Lamb wave The matching pursuit

method estimates the arrival time of each wave packets accurately even two wave packets are

overlapped The numerical simulations and experimental tests including Teflon tape inserted

and impact-induced damage are investigated for the feasibility of the proposed technique

Moreover the effectiveness of the proposed technique under changing temperature is exper-

imentally verified

- 31 -

62 Suggestion for further study

The limitations of this study and the possible improvement and further studies are listed as

follow

1 Measurement of impact-induced damage length using conventional NDT technique

Additional experimental verification for impact-induced damage characterization is

needed by measuring the real length of damage using conventional NDT technique

such as ultrasonic scan or infrared camera

2 Limitation of both side collocated PZT installation In this study excitation PZTs are

installed both side and collocated for decomposing A0 Lamb wave mode However it

can produce decomposition error due to imperfection of PZT location and their installa-

tion Moreover in many cases assess to both surface of structures is not allowed such

as aircraft applications To improve this limitation a reference-free damage characteri-

zation using single side PZTs will be considered

3 Complex structure application In this study damage is characterized in simple plate-

like structure When the monitored structure becomes complex the response signals al-

so becomes complex due to its geometry and boundary Thus the investigations for

complex structure application such as stiffener spar and curvature are needed

- 32 -

REFERENCES

[1] Cawley P and Alleyne D (1996) ldquoThe use of Lamb waves for the long range inspection of large structuresrdquo Ultrasonics 34 pp287-290

[2] Ghosh T Kundu T and Karpur P (1998) ldquoEfficient use of Lamb waves for detecting defects in large platesrdquo Ultrasonics 36 pp791-801

[3] Guirguitui V (2003) ldquoLamb wave generation with piezoelectric wafer active sensors for structural health monitoringrdquo SPIE Conference on Smart Structures and Materials San Diego CA 5391 pp111-422

[4] Ragvahan A and Cesnik CE S (2004) ldquoModeling of piezoelectric-based Lamb wave generation and sensing for structural health monitoringrdquo SPIE Conference on Smart Struc-tures and Materials San Diego CA 5391 pp419-430

[5] Rose JL Pilarski A and Dirit JJ (1993) ldquoAn approach to guided wave mode selection for inspection of laminated platerdquo Journal of Reinforced Plastics and Composites 12(5) pp536-544

[6] Guo N and Cawley P (1993) ldquoThe interaction of Lamb waves with delaminations in composite laminatesrdquo Journal of Acoustical Society of America 94(4) pp 2240-2246

[7] Valdez SHD and Soutis C (2002) ldquoReal-time nondestructive evaluation of fiber com-posite laminates using low-frequency Lamb wavesrdquo Journal of Acoustical Society of Amer-ica 111(5) pp 2026-2033

[8] Yang M and Qiao P (2005) ldquoModeling and experimental detection of damage in virous materials using the pulse-echo method and piezoelectric sensoractuatorsrdquo Smart Materials and Structures 14 pp1083-1100

[9] Lestari W and Qiao P (2005) ldquoApplication of wave propagation analysis for damage identification in composite laminated beamsrdquo Journal of Composite Materials 39(22) pp1967-1984

[10] Kessler S S Spearing S M and Soutis C (2002) ldquoDamage detection in composite ma-terials using Lamb wave methodsrdquo Smart Materials and Structures 11 pp269-278

[11] Sohn H Park G Wail J R Limback N P and Farrar C R (2004) ldquoWavelet-based active sensing for delamination detection in composite structuresrdquo Smart Materials and Structures 13 pp 153-160

[12] Petculescu G Krishnaswamy S and Achenback J D (2008) ldquoGroup delay measure-ments using modally selective Lamb wave transducers for detection and sizing of delamina-tions in compositesrdquo Smart Materials and Structures 17

[13] Sohn H Park H W Law K H and Farrar C R (2007) ldquoDamage detection in compo-

- 33 -

site plates by using an enhanced time reversal methodrdquo Journal of Aerospace Engineering (ASCE) pp 141-151

[14] Sohn H (2007) ldquoEffects of environmental and operational variability on structural health monitoringrdquo Philosophical Transactions of the Royal Society Ardquo 365 pp539-560

[15] Rytter A (1993) ldquoVibration based inspection of civil engineering structuresrdquo PhD The-sis Department of Building Technology and Structural Engineering University of Aalborg Aalborg Denmark

[16] Hayashi T and Kawashima K (2002) ldquoMultiple reflections of Lamb waves at a delami-nationrdquo Ultrasonics 40 pp193-197

[17] Wang C H and Rose L R F (2003) ldquoWave reflection and transmission in beams con-taining delamination and inhomogeneityrdquo Journal of Sound and Vibration 264 pp 851-872

[18] Ramadas C Balasubramaniam K Joshi M and Krishnamurthy C V (2009) ldquoInterac-tion of the primary anti-symmetric Lamb mode (Ao) with symmetric delaminations numer-ical and experimental studiesrdquo Smart Materials and Structures 18

[19] Ramadas C Balasubramaniam K Joshi M and Krishnamurthy C V (2010) ldquoInterac-tion of guided Lamb waves with an asymmetrically located delamination in a laminated composite platerdquo Smart Materials and Structures 19

[20] Wilcox P Lowe M and Cawely P (2001) ldquoThe effect of dispersion on long-range in-spection using ultrasonic guided wavesrdquo NDTampE International 34 pp1-9

[21] Mallat S G and Zhang Z (1993) ldquoMatching pursuits with time-frequency dictionariesrdquo IEEE Tansactions on Signal Processing 41(12) pp3397-3415

[22] Zhang G Zhang S and Wang Y (2000) ldquoApplication of adaptive time-frequency de-composition in ultrasonic NDE of highly scattering materialsrdquo Ultrasonics 38 pp 961-964

[23] Hong J C Sun K H and Kim Y Y (2005) ldquoThe matching pursuit approach based on the modulated Gaussian pulse for efficient guided-wave damage inspectionrdquo Smart Mate-rials and Structures 14 pp548-560

[24] Herakovich C T (1997) Mechanics of Fibrous Composites Wiley 480 pages

[25] Reddy J N (1985) ldquoA review of the literature on finite-element modeling of laminated composite platesrdquo The Shock and Vibration Digest 17(4) pp 3-8

[26] Reddy J N (2003) Mechanics of Laminated Composite Plates and Shells Theory and Analysis CRC Press 856 pages

[27] Palacz M Krawczuk M and Ostachowicz W (2005) ldquoThe spectral finite element mod-el for analysis of flexure-shear coupled wave propagation Part 1 laminated multilayer

- 34 -

composite beamrdquo Composite Structures 68 pp37-44

[28] Zhang Y C and Harding J (1990) ldquoA numerical micromechanics analysis of the me-chanical properties of a plain weave compositerdquo Computer amp Structures 36(5) pp 839-844

[29] Nemat-Nasser S and Hori M (1999) Micromechanics Overall Properties of Heteroge-neous Materials Elsevier 810 pages

[30] Scida D Aboura Z Benzeggagh M L and Bocherens E (1999) ldquoA micromechanics model for 3D elasticity and failure of woven-fibre composite materialrdquo Composites Science and Technology 59 pp 505-517

[31] Thomas Y and Wu X H (1997) ldquoA multiscale finite element method for elliptic prob-lems in composite materials and porous mediardquo Journal of Computational Physics 134 pp 169-189

[32] Lee K H Moorthy S and Ghosh S (1999) ldquoMultiple scale computational model for damage in composite materialsrdquo Computer Methods in Applied Mechanics and Engineer-ing 172 pp 175-201

[33] Namilae S and Chandra N (2005) ldquoMultiscale model to study the effect of interfaces in carbon nanotube-based compositesrdquo Journal of Engineering Materials and Technology (ASME) 127 pp222-232

[34] Adams D E (2007) Health monitoring of Structural Materials and Components Wiley 460 pages

[35] Gopalakrishnan S (2009) ldquoModeling Aspects in Finite Elementrdquo Encyclopedia of Struc-tural Health Monitoring 2 pp791-809

[36] Palacz M Krawczuk M and Ostachowicz W (2005) ldquoThe spectral finite element mod-el for analysis of flexure-shear coupled wave propagation Part 2 Deminated multilayer composite beamrdquo Composite Structures 68 pp45-51

[37] Baik J M and Thopson R B (1984) ldquoUltrasonic scattering from imperfect interfaces A quasi-static modelrdquo Journal of Nondestructive Evaluation 4(3) pp177-196

[38] Luo H and Hanagud S (2000) ldquoDynamics of delamincated beamsrdquo International Jour-nal of Solids and Structures 37 pp1501-1519

[39] Kim S B and Sohn H (2007) ldquoInstantaneous reference-free crack detection based on polarization characteristics of piezoelectric materialsrdquo Smart Materials and Structures 16 pp2375-2387

[40] An Y K and Sohn H (2010) ldquoInstantaneous crack detection under varying temperature and static loading conditionsrdquo The Special Issue of Structural Control and Health Monitor-ing 17(7) pp 730-741

- 35 -

[41] Lee S J and Sohn H (2006) ldquoActive self-sensing scheme development for structural health monitoringrdquo Smart Materials and Structures 15(6) pp 1734-1746

- 36 -

요 약 문

복합재료 손상의 무기저 진단 및 정량화

복합재료는 재료의 경량성 내구성 고강도의 특성으로 인하여 항공기에서부터

교량에까지 널리 사용되고 있다 하지만 외부의 충격으로 인한 복합재료 내부의

손상은 재료의 강도를 감소시켜 구조물의 안정성을 저하시키는 요인이 된다 하지만

복합재료의 손상은 재료 내부에서 발생하여 육안으로는 식별하기도 힘들며 표면에서

보이는 것만으로 내부의 손상 정도를 판단하기는 불가능하다는 단점을 가지고 있다

본 연구에서는 손상 내부에서 반사로 인해 발생하는 파를 이용하여 기저자료를

사용하지 않고 특정 시점에서 얻은 자료만으로 복합재료의 손상을 진단하고 정량화

할 수 있는 기법을 제안한다 제안된 기법의 적용성을 검증하기 위하여 유한요소법을

이용한 수치해석을 수행하였고 Teflon tape 이 삽입된 시편과 실제 충격하중으로

인한 손상을 입은 시편을 이용한 실험을 진행하였다 수치해석 및 실험 결과 제안된

기법이 복합재료의 손상을 성공적으로 진단 및 정량화할 수 있음을 확인하였다 또한

다양한 외부 온도 환경에서 제안된 기법의 환경영향 실험을 수행하였고 다양한

환경에서 손상의 진단 및 정량화가 성공적으로 이루어짐을 검증하였다

핵심어 무기저 손상진단 및 정량화 압전센서 복합재료 유한요소법 Matching pursuit

method

- 37 -

감 사 의 글

기대와 두려움을 안고 시작한 석사과정이 어느덧 2 년 이라는 세월 속에서 추억으로 자리잡아

가고 있습니다 학부생 기간을 포함한 지난 3 년 동안 SHM 이라는 새로운 분야를 접하고 연구를

수행 할 수 있도록 도와주신 고마운 분들이 있어 그 감사를 표하고자 합니다

우선 저의 많은 부족한 점을 일깨워 주시고 연구에 대한 뜻을 더욱 확고히 세울 수 있게 해

주신 지도교수님이신 손 훈 교수님께 깊은 감사의 뜻을 전합니다 짧은 기간이었지만 교수님의

도움으로 많은 것을 배우고 느끼고 깨달을 수 있었습니다 또한 본 연구를 수행할 수 있도록 지

원해주신 국토해양부 (U-City 석 박사과정 지원사업)에 깊은 감사 드립니다

때로 힘들고 지칠 때 항상 저를 믿고 응원해 주시는 가족이 있었기에 지금의 글을 쓸 수 있었

습니다 또한 지금 연구실에서 함께 생활하고 있는 윤규형 민구형 철민이형 현석이형 진열이형

호민이형 병진이 선혜누나와 성대에 계신 창길이형 미국에 계신 의재형과 Debaditiya 에게도 감

사의 말씀을 전합니다 그리고 모두 나열할 수는 없지만 카이스트 생활의 활력소가 되어준 카이

스트 건설 및 환경공학과 선후배들 직원 분들과 SOS 멤버들에게도 감사를 표합니다

앞서 거론한 많은 이들의 도움이 없었다면 저 또한 지금 이 자리에 있을 수 없었을 것입니다

많이 부족하기에 더 노력하고 배우고자 했던 초심을 다시 한번 생각하며 더 큰 결실을 맺기 위해

앞으로 나아가겠습니다 도움을 주신 모든 분들께 이 감사의 글을 바칩니다

2011년 6월 카이스트에서 임형진 올림

- 38 -

CURRICULUM VITAE

Hyung Jin Lim

Research Assistant

The Civil and Environmental Engineering Department

Korea Advanced Institute of Science and Technology (KAIST)

291 Daehak-Ro Guseong-Dong Yuseong-Gu

Daejeon Republic of Korea 305-701

Phone (82)+42-350-3665 Fax (82)+42-350-3610

Email limhj87gmailcom

RESEARCH INTERESTS

Damage Detection Localization and Quantification on Composite and Metallic Structures Pattern Recognition Finite Element Modeling (FEM)

EDUCATION

2009 ndash 2011 MS Civil and Environmental Eng Korea Advanced Institute of Science and Technology (KAIST) Korea Rep

2006 ndash 2009 BS Civil and Environmental Eng Korea Advanced Institute of Science and Technology (KAIST) Korea Rep

JOURNAL PUBLICATIONS The corresponding author is underlined 1 Hyung Jin Lim Hoon Sohn ldquoReference-free damage detection localization and quantification

using in compositerdquo in preparation for Journal of the Acoustical Society of America 2011

2 Chul Min Yeum Hoon Sohn Jeong Bum Ihn Hyung Jin Lim ldquoInstantaneous delamination detection in a composite using a dual piezoelectric transducer networkrdquo Submitted to Compo-sites Science and Technology 2011

3 Yun-Kyu An Hyung Jin Lim Min Koo Kim Jin Yeol Yang Hoon Sohn Chang Guen Lee ldquoInvestigation of applicability of local reference-free damage detection techniques to in-situ

- 39 -

bridgesrdquo Submitted to ASCE Journal of Structural Engineering 2011

4 Hyung Jin Lim Min Koo Kim Hoon Sohn Chan Yik Park ldquoImpedance based damage detec-tion under varying temperature and loading conditionsrdquo Submitted to NDTampE International 2011

5 Hyung Jin Lim Hoon Sohn ldquoQuantitative estimation of transmitted and reflected lamb waves at discontinuityrdquo Journal of the Korean Society for Nondestructive Testing Vol30 No 4 pp1-8 2010

CONFERENCE PROCEEDINGS

The corresponding author is underlined 1 Hyung Jin Lim and Hoon Sohn ldquoReference-free damage detection localization and quantifi-

cation in compositesrdquo ASCE Engineering Mechanics Institute Conference Boston June 2-4 2011

2 Hyung Jin Lim and Hoon Sohn ldquoReference-free delamination quantification within multi-layer composite materialsrdquo KSNT Spring Conference May 26-27 Gyeong-Ju Korea 2011

3 Min Koo Kim Hyun Jin Lim Hoon Sohn Chan Yik Park ldquoImpedance-based bolt loosening detection under varying temperature and loadingrdquo Asian Pacific Workshop on Structural Health Monitoring Tokyo Japan November 30-December 2 2010

4 Hyung Jin Lim Chul Min Yeum and Hoon Sohn ldquoModeling of Impact-induced Delamination in a Multilayer Composite Platerdquo The 23rd KKCNN Symposium on Civil Engineering Taipei (Taiwan) November 13-15 2010

5 Yun-Kyu An Hyung Jin Lim and Hoon Sohn ldquoTemperature independent Delamination Detec-tion using data normalizationrdquo The 3rd International Conference on Multi-Functional Materi-als and Structures Chonbuk National Univ Jeonju Korea September 14-18 2010

6 Hoon Sohn Yun-Kyu An Chang Gil Lee Min Koo Kim Hyung Jin Lim ldquoApplication of online NDT techniques to decommissioned bridge testingrdquo the 5th European Workshop on Structural Health Monitoring Sorrento Italy June 29-July 02 2010

7 Min Koo Kim Hyung Jin Lim Hoon Sohn Chan Yik Park ldquoImpedance based damage diag-nosis of a complex composite aircraft wing under changing loading and temperature condi-tionsrdquo the 5th European Workshop on Structural Health Monitoring Sorrento Italy June 29-July 02 2010

8 Yun-Kyu An Min Koo Kim Hyung Jin Lim and Hoon Sohn ldquoValidations of reference-free crack detection techniques through a decommissioned bridge testrdquo 2010 COSEIK Annual Con-ference Jeju Korea April 8-9 2010

- 40 -

PARTICIPANT PROJECTS

1 Development of Smart Steel Members for Online Steel Structure Monitoring POSCO Corporation (Funded 41000000 KRW (41000 USD) for 010511 to 123111)

2 Development of Enabling Technologies for Structural Health Monitoring Korean Agency for Defense Development (Funded 240000000 KRW (240000 USD) out of total 3200000000 KRW (3200000 USD) for 060108-053111)

3 Development of On-board SHM Technologies for Composite Air Vehicles The Boeing Company (Funded 400000 USD (400000000 KRW) for 080108 to 073111)

EDUCATIONAL EXPERIENCES

Fall 2009

Teaching Assistant CE207 ldquoElementary Structural Engineering and Laboratoryrdquo KAIST 3 unit undergraduate course

EXPERIMENTAL EXPERIENTCES

2010 Monitoring of Real Aircraft Composite Wing Segment Korea Agency for Defense De-velopment (ADD) Daejeon Korea Real time monitoring of composite air vehicles fund-ed by Korea Agency for Defense Development (ADD)

2009-2011

Field Monitoring of Real Bridge Structures Gimpo (Ramp G bridge) amp Icheon (Sam-seung bridge) Korea US-Korea collaborative research for bridge monitoring testbeds

HONORS amp AWARDS

Scholarships for outstanding students granted KAIST (2007-2009) The three prominent students in the civil and environmental engineering department of KAIST were selected as recipients This scholarship provided 2000$ per year for three years

- 41 -

Page 40: 복합재료 손상의 무기저 진단 및 정량화 - KAISTssslab.kaist.ac.kr/article/pdf/Thesis_2.pdf · 1 Declaration of Ethical Conduct in Research: I, as a graduate student

542 Impact-induced damage

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Vol

tage

(mV

)

Transmitted WaveReflected Wave

No reflected wave

(a) 0 degC

105 11 115 12 125-10

-5

0

5

10

Time (ms)

Volta

ge (m

V)

Transmitted WaveReflected Wave

No reflected wave

(b) 50 degC

Fig 16 False alarm test under temperature variations It is shown that proposed damage char-acterization algorithm is robust under temperature variations

The false alarm test in pristine condition of the impact-induced damage specimen under

temperature variation was investigated (Fig 16) It is validated that the proposed damage di-

agnosis algorithm is robust under temperature variations Fig 17 shows the obtained signals

from the impact-induced damage (impacted 3 times) specimen under temperature variation

The analyzed results of the obtained signals are reported in Table VII The measured 119907119907119894119894 in

each temperature and the assumed variation are same as the artificial damage case

11 115 12 125 13-5

0

5

Time (ms)

Volta

ge (m

V)

0 oC20 oC50 oC

(a) Pitch-catch signal

11 115 12 125 13 135

-1

0

1

2

3

Time(ms)

Vol

tage

(mV

)

0 oC20 oC50 oC

(b) Pulse-echo signal

Fig 17 Signals obtained from impact-induced damage specimen (Impact 3) under tempera-ture variation

- 29 -

TABLE VII Temperature effect test for the impact induced damage specimen (Impact 3 case)

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

0 degC

Method 1 135 75~90 72~310 267~1148

Method 2 135 705~779 264~381 978~1126

Method 3 135 763 289 1071

20 degC

Method 1 148 50~100 06~380 21~1134

Method 2 148 758~838 249~368 843~1125

Method 3 148 798 308 1044

50 degC

Method 1 153 75~100 53~415 174~1227

Method 2 153 804~889 214~336 702~1103

Method 3 153 815 320 1050

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1247 ms at 0 degC 1213 ms at 20 degC and 1168 ms at 50 degC

- 30 -

CHAPTER 6 CONCLUDING REMARKS

61 Executive summary

In this study the reference-free damage detection localization and quantification tech-

nique in a multi layered composite plate is proposed Additional reflected A0 Lamb wave

mode from the damage due to multiple reflections in damage area is used as a feature for

characterizing the damage First damage is characterized in terms of a damage index (m-

value) defined as the ratio of the physical length of the damage to the wavelength of the A0

Lamb wave mode Then the range of possible damage location length and severity are esti-

mated with constraints This procedure can be accomplished without prior baseline data To

realize the proposed technique the matching pursuit technique is modified for reduction of

the error by considering the dispersive characteristic of Lamb wave The matching pursuit

method estimates the arrival time of each wave packets accurately even two wave packets are

overlapped The numerical simulations and experimental tests including Teflon tape inserted

and impact-induced damage are investigated for the feasibility of the proposed technique

Moreover the effectiveness of the proposed technique under changing temperature is exper-

imentally verified

- 31 -

62 Suggestion for further study

The limitations of this study and the possible improvement and further studies are listed as

follow

1 Measurement of impact-induced damage length using conventional NDT technique

Additional experimental verification for impact-induced damage characterization is

needed by measuring the real length of damage using conventional NDT technique

such as ultrasonic scan or infrared camera

2 Limitation of both side collocated PZT installation In this study excitation PZTs are

installed both side and collocated for decomposing A0 Lamb wave mode However it

can produce decomposition error due to imperfection of PZT location and their installa-

tion Moreover in many cases assess to both surface of structures is not allowed such

as aircraft applications To improve this limitation a reference-free damage characteri-

zation using single side PZTs will be considered

3 Complex structure application In this study damage is characterized in simple plate-

like structure When the monitored structure becomes complex the response signals al-

so becomes complex due to its geometry and boundary Thus the investigations for

complex structure application such as stiffener spar and curvature are needed

- 32 -

REFERENCES

[1] Cawley P and Alleyne D (1996) ldquoThe use of Lamb waves for the long range inspection of large structuresrdquo Ultrasonics 34 pp287-290

[2] Ghosh T Kundu T and Karpur P (1998) ldquoEfficient use of Lamb waves for detecting defects in large platesrdquo Ultrasonics 36 pp791-801

[3] Guirguitui V (2003) ldquoLamb wave generation with piezoelectric wafer active sensors for structural health monitoringrdquo SPIE Conference on Smart Structures and Materials San Diego CA 5391 pp111-422

[4] Ragvahan A and Cesnik CE S (2004) ldquoModeling of piezoelectric-based Lamb wave generation and sensing for structural health monitoringrdquo SPIE Conference on Smart Struc-tures and Materials San Diego CA 5391 pp419-430

[5] Rose JL Pilarski A and Dirit JJ (1993) ldquoAn approach to guided wave mode selection for inspection of laminated platerdquo Journal of Reinforced Plastics and Composites 12(5) pp536-544

[6] Guo N and Cawley P (1993) ldquoThe interaction of Lamb waves with delaminations in composite laminatesrdquo Journal of Acoustical Society of America 94(4) pp 2240-2246

[7] Valdez SHD and Soutis C (2002) ldquoReal-time nondestructive evaluation of fiber com-posite laminates using low-frequency Lamb wavesrdquo Journal of Acoustical Society of Amer-ica 111(5) pp 2026-2033

[8] Yang M and Qiao P (2005) ldquoModeling and experimental detection of damage in virous materials using the pulse-echo method and piezoelectric sensoractuatorsrdquo Smart Materials and Structures 14 pp1083-1100

[9] Lestari W and Qiao P (2005) ldquoApplication of wave propagation analysis for damage identification in composite laminated beamsrdquo Journal of Composite Materials 39(22) pp1967-1984

[10] Kessler S S Spearing S M and Soutis C (2002) ldquoDamage detection in composite ma-terials using Lamb wave methodsrdquo Smart Materials and Structures 11 pp269-278

[11] Sohn H Park G Wail J R Limback N P and Farrar C R (2004) ldquoWavelet-based active sensing for delamination detection in composite structuresrdquo Smart Materials and Structures 13 pp 153-160

[12] Petculescu G Krishnaswamy S and Achenback J D (2008) ldquoGroup delay measure-ments using modally selective Lamb wave transducers for detection and sizing of delamina-tions in compositesrdquo Smart Materials and Structures 17

[13] Sohn H Park H W Law K H and Farrar C R (2007) ldquoDamage detection in compo-

- 33 -

site plates by using an enhanced time reversal methodrdquo Journal of Aerospace Engineering (ASCE) pp 141-151

[14] Sohn H (2007) ldquoEffects of environmental and operational variability on structural health monitoringrdquo Philosophical Transactions of the Royal Society Ardquo 365 pp539-560

[15] Rytter A (1993) ldquoVibration based inspection of civil engineering structuresrdquo PhD The-sis Department of Building Technology and Structural Engineering University of Aalborg Aalborg Denmark

[16] Hayashi T and Kawashima K (2002) ldquoMultiple reflections of Lamb waves at a delami-nationrdquo Ultrasonics 40 pp193-197

[17] Wang C H and Rose L R F (2003) ldquoWave reflection and transmission in beams con-taining delamination and inhomogeneityrdquo Journal of Sound and Vibration 264 pp 851-872

[18] Ramadas C Balasubramaniam K Joshi M and Krishnamurthy C V (2009) ldquoInterac-tion of the primary anti-symmetric Lamb mode (Ao) with symmetric delaminations numer-ical and experimental studiesrdquo Smart Materials and Structures 18

[19] Ramadas C Balasubramaniam K Joshi M and Krishnamurthy C V (2010) ldquoInterac-tion of guided Lamb waves with an asymmetrically located delamination in a laminated composite platerdquo Smart Materials and Structures 19

[20] Wilcox P Lowe M and Cawely P (2001) ldquoThe effect of dispersion on long-range in-spection using ultrasonic guided wavesrdquo NDTampE International 34 pp1-9

[21] Mallat S G and Zhang Z (1993) ldquoMatching pursuits with time-frequency dictionariesrdquo IEEE Tansactions on Signal Processing 41(12) pp3397-3415

[22] Zhang G Zhang S and Wang Y (2000) ldquoApplication of adaptive time-frequency de-composition in ultrasonic NDE of highly scattering materialsrdquo Ultrasonics 38 pp 961-964

[23] Hong J C Sun K H and Kim Y Y (2005) ldquoThe matching pursuit approach based on the modulated Gaussian pulse for efficient guided-wave damage inspectionrdquo Smart Mate-rials and Structures 14 pp548-560

[24] Herakovich C T (1997) Mechanics of Fibrous Composites Wiley 480 pages

[25] Reddy J N (1985) ldquoA review of the literature on finite-element modeling of laminated composite platesrdquo The Shock and Vibration Digest 17(4) pp 3-8

[26] Reddy J N (2003) Mechanics of Laminated Composite Plates and Shells Theory and Analysis CRC Press 856 pages

[27] Palacz M Krawczuk M and Ostachowicz W (2005) ldquoThe spectral finite element mod-el for analysis of flexure-shear coupled wave propagation Part 1 laminated multilayer

- 34 -

composite beamrdquo Composite Structures 68 pp37-44

[28] Zhang Y C and Harding J (1990) ldquoA numerical micromechanics analysis of the me-chanical properties of a plain weave compositerdquo Computer amp Structures 36(5) pp 839-844

[29] Nemat-Nasser S and Hori M (1999) Micromechanics Overall Properties of Heteroge-neous Materials Elsevier 810 pages

[30] Scida D Aboura Z Benzeggagh M L and Bocherens E (1999) ldquoA micromechanics model for 3D elasticity and failure of woven-fibre composite materialrdquo Composites Science and Technology 59 pp 505-517

[31] Thomas Y and Wu X H (1997) ldquoA multiscale finite element method for elliptic prob-lems in composite materials and porous mediardquo Journal of Computational Physics 134 pp 169-189

[32] Lee K H Moorthy S and Ghosh S (1999) ldquoMultiple scale computational model for damage in composite materialsrdquo Computer Methods in Applied Mechanics and Engineer-ing 172 pp 175-201

[33] Namilae S and Chandra N (2005) ldquoMultiscale model to study the effect of interfaces in carbon nanotube-based compositesrdquo Journal of Engineering Materials and Technology (ASME) 127 pp222-232

[34] Adams D E (2007) Health monitoring of Structural Materials and Components Wiley 460 pages

[35] Gopalakrishnan S (2009) ldquoModeling Aspects in Finite Elementrdquo Encyclopedia of Struc-tural Health Monitoring 2 pp791-809

[36] Palacz M Krawczuk M and Ostachowicz W (2005) ldquoThe spectral finite element mod-el for analysis of flexure-shear coupled wave propagation Part 2 Deminated multilayer composite beamrdquo Composite Structures 68 pp45-51

[37] Baik J M and Thopson R B (1984) ldquoUltrasonic scattering from imperfect interfaces A quasi-static modelrdquo Journal of Nondestructive Evaluation 4(3) pp177-196

[38] Luo H and Hanagud S (2000) ldquoDynamics of delamincated beamsrdquo International Jour-nal of Solids and Structures 37 pp1501-1519

[39] Kim S B and Sohn H (2007) ldquoInstantaneous reference-free crack detection based on polarization characteristics of piezoelectric materialsrdquo Smart Materials and Structures 16 pp2375-2387

[40] An Y K and Sohn H (2010) ldquoInstantaneous crack detection under varying temperature and static loading conditionsrdquo The Special Issue of Structural Control and Health Monitor-ing 17(7) pp 730-741

- 35 -

[41] Lee S J and Sohn H (2006) ldquoActive self-sensing scheme development for structural health monitoringrdquo Smart Materials and Structures 15(6) pp 1734-1746

- 36 -

요 약 문

복합재료 손상의 무기저 진단 및 정량화

복합재료는 재료의 경량성 내구성 고강도의 특성으로 인하여 항공기에서부터

교량에까지 널리 사용되고 있다 하지만 외부의 충격으로 인한 복합재료 내부의

손상은 재료의 강도를 감소시켜 구조물의 안정성을 저하시키는 요인이 된다 하지만

복합재료의 손상은 재료 내부에서 발생하여 육안으로는 식별하기도 힘들며 표면에서

보이는 것만으로 내부의 손상 정도를 판단하기는 불가능하다는 단점을 가지고 있다

본 연구에서는 손상 내부에서 반사로 인해 발생하는 파를 이용하여 기저자료를

사용하지 않고 특정 시점에서 얻은 자료만으로 복합재료의 손상을 진단하고 정량화

할 수 있는 기법을 제안한다 제안된 기법의 적용성을 검증하기 위하여 유한요소법을

이용한 수치해석을 수행하였고 Teflon tape 이 삽입된 시편과 실제 충격하중으로

인한 손상을 입은 시편을 이용한 실험을 진행하였다 수치해석 및 실험 결과 제안된

기법이 복합재료의 손상을 성공적으로 진단 및 정량화할 수 있음을 확인하였다 또한

다양한 외부 온도 환경에서 제안된 기법의 환경영향 실험을 수행하였고 다양한

환경에서 손상의 진단 및 정량화가 성공적으로 이루어짐을 검증하였다

핵심어 무기저 손상진단 및 정량화 압전센서 복합재료 유한요소법 Matching pursuit

method

- 37 -

감 사 의 글

기대와 두려움을 안고 시작한 석사과정이 어느덧 2 년 이라는 세월 속에서 추억으로 자리잡아

가고 있습니다 학부생 기간을 포함한 지난 3 년 동안 SHM 이라는 새로운 분야를 접하고 연구를

수행 할 수 있도록 도와주신 고마운 분들이 있어 그 감사를 표하고자 합니다

우선 저의 많은 부족한 점을 일깨워 주시고 연구에 대한 뜻을 더욱 확고히 세울 수 있게 해

주신 지도교수님이신 손 훈 교수님께 깊은 감사의 뜻을 전합니다 짧은 기간이었지만 교수님의

도움으로 많은 것을 배우고 느끼고 깨달을 수 있었습니다 또한 본 연구를 수행할 수 있도록 지

원해주신 국토해양부 (U-City 석 박사과정 지원사업)에 깊은 감사 드립니다

때로 힘들고 지칠 때 항상 저를 믿고 응원해 주시는 가족이 있었기에 지금의 글을 쓸 수 있었

습니다 또한 지금 연구실에서 함께 생활하고 있는 윤규형 민구형 철민이형 현석이형 진열이형

호민이형 병진이 선혜누나와 성대에 계신 창길이형 미국에 계신 의재형과 Debaditiya 에게도 감

사의 말씀을 전합니다 그리고 모두 나열할 수는 없지만 카이스트 생활의 활력소가 되어준 카이

스트 건설 및 환경공학과 선후배들 직원 분들과 SOS 멤버들에게도 감사를 표합니다

앞서 거론한 많은 이들의 도움이 없었다면 저 또한 지금 이 자리에 있을 수 없었을 것입니다

많이 부족하기에 더 노력하고 배우고자 했던 초심을 다시 한번 생각하며 더 큰 결실을 맺기 위해

앞으로 나아가겠습니다 도움을 주신 모든 분들께 이 감사의 글을 바칩니다

2011년 6월 카이스트에서 임형진 올림

- 38 -

CURRICULUM VITAE

Hyung Jin Lim

Research Assistant

The Civil and Environmental Engineering Department

Korea Advanced Institute of Science and Technology (KAIST)

291 Daehak-Ro Guseong-Dong Yuseong-Gu

Daejeon Republic of Korea 305-701

Phone (82)+42-350-3665 Fax (82)+42-350-3610

Email limhj87gmailcom

RESEARCH INTERESTS

Damage Detection Localization and Quantification on Composite and Metallic Structures Pattern Recognition Finite Element Modeling (FEM)

EDUCATION

2009 ndash 2011 MS Civil and Environmental Eng Korea Advanced Institute of Science and Technology (KAIST) Korea Rep

2006 ndash 2009 BS Civil and Environmental Eng Korea Advanced Institute of Science and Technology (KAIST) Korea Rep

JOURNAL PUBLICATIONS The corresponding author is underlined 1 Hyung Jin Lim Hoon Sohn ldquoReference-free damage detection localization and quantification

using in compositerdquo in preparation for Journal of the Acoustical Society of America 2011

2 Chul Min Yeum Hoon Sohn Jeong Bum Ihn Hyung Jin Lim ldquoInstantaneous delamination detection in a composite using a dual piezoelectric transducer networkrdquo Submitted to Compo-sites Science and Technology 2011

3 Yun-Kyu An Hyung Jin Lim Min Koo Kim Jin Yeol Yang Hoon Sohn Chang Guen Lee ldquoInvestigation of applicability of local reference-free damage detection techniques to in-situ

- 39 -

bridgesrdquo Submitted to ASCE Journal of Structural Engineering 2011

4 Hyung Jin Lim Min Koo Kim Hoon Sohn Chan Yik Park ldquoImpedance based damage detec-tion under varying temperature and loading conditionsrdquo Submitted to NDTampE International 2011

5 Hyung Jin Lim Hoon Sohn ldquoQuantitative estimation of transmitted and reflected lamb waves at discontinuityrdquo Journal of the Korean Society for Nondestructive Testing Vol30 No 4 pp1-8 2010

CONFERENCE PROCEEDINGS

The corresponding author is underlined 1 Hyung Jin Lim and Hoon Sohn ldquoReference-free damage detection localization and quantifi-

cation in compositesrdquo ASCE Engineering Mechanics Institute Conference Boston June 2-4 2011

2 Hyung Jin Lim and Hoon Sohn ldquoReference-free delamination quantification within multi-layer composite materialsrdquo KSNT Spring Conference May 26-27 Gyeong-Ju Korea 2011

3 Min Koo Kim Hyun Jin Lim Hoon Sohn Chan Yik Park ldquoImpedance-based bolt loosening detection under varying temperature and loadingrdquo Asian Pacific Workshop on Structural Health Monitoring Tokyo Japan November 30-December 2 2010

4 Hyung Jin Lim Chul Min Yeum and Hoon Sohn ldquoModeling of Impact-induced Delamination in a Multilayer Composite Platerdquo The 23rd KKCNN Symposium on Civil Engineering Taipei (Taiwan) November 13-15 2010

5 Yun-Kyu An Hyung Jin Lim and Hoon Sohn ldquoTemperature independent Delamination Detec-tion using data normalizationrdquo The 3rd International Conference on Multi-Functional Materi-als and Structures Chonbuk National Univ Jeonju Korea September 14-18 2010

6 Hoon Sohn Yun-Kyu An Chang Gil Lee Min Koo Kim Hyung Jin Lim ldquoApplication of online NDT techniques to decommissioned bridge testingrdquo the 5th European Workshop on Structural Health Monitoring Sorrento Italy June 29-July 02 2010

7 Min Koo Kim Hyung Jin Lim Hoon Sohn Chan Yik Park ldquoImpedance based damage diag-nosis of a complex composite aircraft wing under changing loading and temperature condi-tionsrdquo the 5th European Workshop on Structural Health Monitoring Sorrento Italy June 29-July 02 2010

8 Yun-Kyu An Min Koo Kim Hyung Jin Lim and Hoon Sohn ldquoValidations of reference-free crack detection techniques through a decommissioned bridge testrdquo 2010 COSEIK Annual Con-ference Jeju Korea April 8-9 2010

- 40 -

PARTICIPANT PROJECTS

1 Development of Smart Steel Members for Online Steel Structure Monitoring POSCO Corporation (Funded 41000000 KRW (41000 USD) for 010511 to 123111)

2 Development of Enabling Technologies for Structural Health Monitoring Korean Agency for Defense Development (Funded 240000000 KRW (240000 USD) out of total 3200000000 KRW (3200000 USD) for 060108-053111)

3 Development of On-board SHM Technologies for Composite Air Vehicles The Boeing Company (Funded 400000 USD (400000000 KRW) for 080108 to 073111)

EDUCATIONAL EXPERIENCES

Fall 2009

Teaching Assistant CE207 ldquoElementary Structural Engineering and Laboratoryrdquo KAIST 3 unit undergraduate course

EXPERIMENTAL EXPERIENTCES

2010 Monitoring of Real Aircraft Composite Wing Segment Korea Agency for Defense De-velopment (ADD) Daejeon Korea Real time monitoring of composite air vehicles fund-ed by Korea Agency for Defense Development (ADD)

2009-2011

Field Monitoring of Real Bridge Structures Gimpo (Ramp G bridge) amp Icheon (Sam-seung bridge) Korea US-Korea collaborative research for bridge monitoring testbeds

HONORS amp AWARDS

Scholarships for outstanding students granted KAIST (2007-2009) The three prominent students in the civil and environmental engineering department of KAIST were selected as recipients This scholarship provided 2000$ per year for three years

- 41 -

Page 41: 복합재료 손상의 무기저 진단 및 정량화 - KAISTssslab.kaist.ac.kr/article/pdf/Thesis_2.pdf · 1 Declaration of Ethical Conduct in Research: I, as a graduate student

TABLE VII Temperature effect test for the impact induced damage specimen (Impact 3 case)

m-value 119949119949120783120783 (mm) 119941119941 (mm) 119959119959119941119941 (ms)

0 degC

Method 1 135 75~90 72~310 267~1148

Method 2 135 705~779 264~381 978~1126

Method 3 135 763 289 1071

20 degC

Method 1 148 50~100 06~380 21~1134

Method 2 148 758~838 249~368 843~1125

Method 3 148 798 308 1044

50 degC

Method 1 153 75~100 53~415 174~1227

Method 2 153 804~889 214~336 702~1103

Method 3 153 815 320 1050

The variation of 119907119907119894119894 is assumed to plusmn 5 of the normal value 1213 ms 1152~1273 ms Measured 119907119907119894119894 1247 ms at 0 degC 1213 ms at 20 degC and 1168 ms at 50 degC

- 30 -

CHAPTER 6 CONCLUDING REMARKS

61 Executive summary

In this study the reference-free damage detection localization and quantification tech-

nique in a multi layered composite plate is proposed Additional reflected A0 Lamb wave

mode from the damage due to multiple reflections in damage area is used as a feature for

characterizing the damage First damage is characterized in terms of a damage index (m-

value) defined as the ratio of the physical length of the damage to the wavelength of the A0

Lamb wave mode Then the range of possible damage location length and severity are esti-

mated with constraints This procedure can be accomplished without prior baseline data To

realize the proposed technique the matching pursuit technique is modified for reduction of

the error by considering the dispersive characteristic of Lamb wave The matching pursuit

method estimates the arrival time of each wave packets accurately even two wave packets are

overlapped The numerical simulations and experimental tests including Teflon tape inserted

and impact-induced damage are investigated for the feasibility of the proposed technique

Moreover the effectiveness of the proposed technique under changing temperature is exper-

imentally verified

- 31 -

62 Suggestion for further study

The limitations of this study and the possible improvement and further studies are listed as

follow

1 Measurement of impact-induced damage length using conventional NDT technique

Additional experimental verification for impact-induced damage characterization is

needed by measuring the real length of damage using conventional NDT technique

such as ultrasonic scan or infrared camera

2 Limitation of both side collocated PZT installation In this study excitation PZTs are

installed both side and collocated for decomposing A0 Lamb wave mode However it

can produce decomposition error due to imperfection of PZT location and their installa-

tion Moreover in many cases assess to both surface of structures is not allowed such

as aircraft applications To improve this limitation a reference-free damage characteri-

zation using single side PZTs will be considered

3 Complex structure application In this study damage is characterized in simple plate-

like structure When the monitored structure becomes complex the response signals al-

so becomes complex due to its geometry and boundary Thus the investigations for

complex structure application such as stiffener spar and curvature are needed

- 32 -

REFERENCES

[1] Cawley P and Alleyne D (1996) ldquoThe use of Lamb waves for the long range inspection of large structuresrdquo Ultrasonics 34 pp287-290

[2] Ghosh T Kundu T and Karpur P (1998) ldquoEfficient use of Lamb waves for detecting defects in large platesrdquo Ultrasonics 36 pp791-801

[3] Guirguitui V (2003) ldquoLamb wave generation with piezoelectric wafer active sensors for structural health monitoringrdquo SPIE Conference on Smart Structures and Materials San Diego CA 5391 pp111-422

[4] Ragvahan A and Cesnik CE S (2004) ldquoModeling of piezoelectric-based Lamb wave generation and sensing for structural health monitoringrdquo SPIE Conference on Smart Struc-tures and Materials San Diego CA 5391 pp419-430

[5] Rose JL Pilarski A and Dirit JJ (1993) ldquoAn approach to guided wave mode selection for inspection of laminated platerdquo Journal of Reinforced Plastics and Composites 12(5) pp536-544

[6] Guo N and Cawley P (1993) ldquoThe interaction of Lamb waves with delaminations in composite laminatesrdquo Journal of Acoustical Society of America 94(4) pp 2240-2246

[7] Valdez SHD and Soutis C (2002) ldquoReal-time nondestructive evaluation of fiber com-posite laminates using low-frequency Lamb wavesrdquo Journal of Acoustical Society of Amer-ica 111(5) pp 2026-2033

[8] Yang M and Qiao P (2005) ldquoModeling and experimental detection of damage in virous materials using the pulse-echo method and piezoelectric sensoractuatorsrdquo Smart Materials and Structures 14 pp1083-1100

[9] Lestari W and Qiao P (2005) ldquoApplication of wave propagation analysis for damage identification in composite laminated beamsrdquo Journal of Composite Materials 39(22) pp1967-1984

[10] Kessler S S Spearing S M and Soutis C (2002) ldquoDamage detection in composite ma-terials using Lamb wave methodsrdquo Smart Materials and Structures 11 pp269-278

[11] Sohn H Park G Wail J R Limback N P and Farrar C R (2004) ldquoWavelet-based active sensing for delamination detection in composite structuresrdquo Smart Materials and Structures 13 pp 153-160

[12] Petculescu G Krishnaswamy S and Achenback J D (2008) ldquoGroup delay measure-ments using modally selective Lamb wave transducers for detection and sizing of delamina-tions in compositesrdquo Smart Materials and Structures 17

[13] Sohn H Park H W Law K H and Farrar C R (2007) ldquoDamage detection in compo-

- 33 -

site plates by using an enhanced time reversal methodrdquo Journal of Aerospace Engineering (ASCE) pp 141-151

[14] Sohn H (2007) ldquoEffects of environmental and operational variability on structural health monitoringrdquo Philosophical Transactions of the Royal Society Ardquo 365 pp539-560

[15] Rytter A (1993) ldquoVibration based inspection of civil engineering structuresrdquo PhD The-sis Department of Building Technology and Structural Engineering University of Aalborg Aalborg Denmark

[16] Hayashi T and Kawashima K (2002) ldquoMultiple reflections of Lamb waves at a delami-nationrdquo Ultrasonics 40 pp193-197

[17] Wang C H and Rose L R F (2003) ldquoWave reflection and transmission in beams con-taining delamination and inhomogeneityrdquo Journal of Sound and Vibration 264 pp 851-872

[18] Ramadas C Balasubramaniam K Joshi M and Krishnamurthy C V (2009) ldquoInterac-tion of the primary anti-symmetric Lamb mode (Ao) with symmetric delaminations numer-ical and experimental studiesrdquo Smart Materials and Structures 18

[19] Ramadas C Balasubramaniam K Joshi M and Krishnamurthy C V (2010) ldquoInterac-tion of guided Lamb waves with an asymmetrically located delamination in a laminated composite platerdquo Smart Materials and Structures 19

[20] Wilcox P Lowe M and Cawely P (2001) ldquoThe effect of dispersion on long-range in-spection using ultrasonic guided wavesrdquo NDTampE International 34 pp1-9

[21] Mallat S G and Zhang Z (1993) ldquoMatching pursuits with time-frequency dictionariesrdquo IEEE Tansactions on Signal Processing 41(12) pp3397-3415

[22] Zhang G Zhang S and Wang Y (2000) ldquoApplication of adaptive time-frequency de-composition in ultrasonic NDE of highly scattering materialsrdquo Ultrasonics 38 pp 961-964

[23] Hong J C Sun K H and Kim Y Y (2005) ldquoThe matching pursuit approach based on the modulated Gaussian pulse for efficient guided-wave damage inspectionrdquo Smart Mate-rials and Structures 14 pp548-560

[24] Herakovich C T (1997) Mechanics of Fibrous Composites Wiley 480 pages

[25] Reddy J N (1985) ldquoA review of the literature on finite-element modeling of laminated composite platesrdquo The Shock and Vibration Digest 17(4) pp 3-8

[26] Reddy J N (2003) Mechanics of Laminated Composite Plates and Shells Theory and Analysis CRC Press 856 pages

[27] Palacz M Krawczuk M and Ostachowicz W (2005) ldquoThe spectral finite element mod-el for analysis of flexure-shear coupled wave propagation Part 1 laminated multilayer

- 34 -

composite beamrdquo Composite Structures 68 pp37-44

[28] Zhang Y C and Harding J (1990) ldquoA numerical micromechanics analysis of the me-chanical properties of a plain weave compositerdquo Computer amp Structures 36(5) pp 839-844

[29] Nemat-Nasser S and Hori M (1999) Micromechanics Overall Properties of Heteroge-neous Materials Elsevier 810 pages

[30] Scida D Aboura Z Benzeggagh M L and Bocherens E (1999) ldquoA micromechanics model for 3D elasticity and failure of woven-fibre composite materialrdquo Composites Science and Technology 59 pp 505-517

[31] Thomas Y and Wu X H (1997) ldquoA multiscale finite element method for elliptic prob-lems in composite materials and porous mediardquo Journal of Computational Physics 134 pp 169-189

[32] Lee K H Moorthy S and Ghosh S (1999) ldquoMultiple scale computational model for damage in composite materialsrdquo Computer Methods in Applied Mechanics and Engineer-ing 172 pp 175-201

[33] Namilae S and Chandra N (2005) ldquoMultiscale model to study the effect of interfaces in carbon nanotube-based compositesrdquo Journal of Engineering Materials and Technology (ASME) 127 pp222-232

[34] Adams D E (2007) Health monitoring of Structural Materials and Components Wiley 460 pages

[35] Gopalakrishnan S (2009) ldquoModeling Aspects in Finite Elementrdquo Encyclopedia of Struc-tural Health Monitoring 2 pp791-809

[36] Palacz M Krawczuk M and Ostachowicz W (2005) ldquoThe spectral finite element mod-el for analysis of flexure-shear coupled wave propagation Part 2 Deminated multilayer composite beamrdquo Composite Structures 68 pp45-51

[37] Baik J M and Thopson R B (1984) ldquoUltrasonic scattering from imperfect interfaces A quasi-static modelrdquo Journal of Nondestructive Evaluation 4(3) pp177-196

[38] Luo H and Hanagud S (2000) ldquoDynamics of delamincated beamsrdquo International Jour-nal of Solids and Structures 37 pp1501-1519

[39] Kim S B and Sohn H (2007) ldquoInstantaneous reference-free crack detection based on polarization characteristics of piezoelectric materialsrdquo Smart Materials and Structures 16 pp2375-2387

[40] An Y K and Sohn H (2010) ldquoInstantaneous crack detection under varying temperature and static loading conditionsrdquo The Special Issue of Structural Control and Health Monitor-ing 17(7) pp 730-741

- 35 -

[41] Lee S J and Sohn H (2006) ldquoActive self-sensing scheme development for structural health monitoringrdquo Smart Materials and Structures 15(6) pp 1734-1746

- 36 -

요 약 문

복합재료 손상의 무기저 진단 및 정량화

복합재료는 재료의 경량성 내구성 고강도의 특성으로 인하여 항공기에서부터

교량에까지 널리 사용되고 있다 하지만 외부의 충격으로 인한 복합재료 내부의

손상은 재료의 강도를 감소시켜 구조물의 안정성을 저하시키는 요인이 된다 하지만

복합재료의 손상은 재료 내부에서 발생하여 육안으로는 식별하기도 힘들며 표면에서

보이는 것만으로 내부의 손상 정도를 판단하기는 불가능하다는 단점을 가지고 있다

본 연구에서는 손상 내부에서 반사로 인해 발생하는 파를 이용하여 기저자료를

사용하지 않고 특정 시점에서 얻은 자료만으로 복합재료의 손상을 진단하고 정량화

할 수 있는 기법을 제안한다 제안된 기법의 적용성을 검증하기 위하여 유한요소법을

이용한 수치해석을 수행하였고 Teflon tape 이 삽입된 시편과 실제 충격하중으로

인한 손상을 입은 시편을 이용한 실험을 진행하였다 수치해석 및 실험 결과 제안된

기법이 복합재료의 손상을 성공적으로 진단 및 정량화할 수 있음을 확인하였다 또한

다양한 외부 온도 환경에서 제안된 기법의 환경영향 실험을 수행하였고 다양한

환경에서 손상의 진단 및 정량화가 성공적으로 이루어짐을 검증하였다

핵심어 무기저 손상진단 및 정량화 압전센서 복합재료 유한요소법 Matching pursuit

method

- 37 -

감 사 의 글

기대와 두려움을 안고 시작한 석사과정이 어느덧 2 년 이라는 세월 속에서 추억으로 자리잡아

가고 있습니다 학부생 기간을 포함한 지난 3 년 동안 SHM 이라는 새로운 분야를 접하고 연구를

수행 할 수 있도록 도와주신 고마운 분들이 있어 그 감사를 표하고자 합니다

우선 저의 많은 부족한 점을 일깨워 주시고 연구에 대한 뜻을 더욱 확고히 세울 수 있게 해

주신 지도교수님이신 손 훈 교수님께 깊은 감사의 뜻을 전합니다 짧은 기간이었지만 교수님의

도움으로 많은 것을 배우고 느끼고 깨달을 수 있었습니다 또한 본 연구를 수행할 수 있도록 지

원해주신 국토해양부 (U-City 석 박사과정 지원사업)에 깊은 감사 드립니다

때로 힘들고 지칠 때 항상 저를 믿고 응원해 주시는 가족이 있었기에 지금의 글을 쓸 수 있었

습니다 또한 지금 연구실에서 함께 생활하고 있는 윤규형 민구형 철민이형 현석이형 진열이형

호민이형 병진이 선혜누나와 성대에 계신 창길이형 미국에 계신 의재형과 Debaditiya 에게도 감

사의 말씀을 전합니다 그리고 모두 나열할 수는 없지만 카이스트 생활의 활력소가 되어준 카이

스트 건설 및 환경공학과 선후배들 직원 분들과 SOS 멤버들에게도 감사를 표합니다

앞서 거론한 많은 이들의 도움이 없었다면 저 또한 지금 이 자리에 있을 수 없었을 것입니다

많이 부족하기에 더 노력하고 배우고자 했던 초심을 다시 한번 생각하며 더 큰 결실을 맺기 위해

앞으로 나아가겠습니다 도움을 주신 모든 분들께 이 감사의 글을 바칩니다

2011년 6월 카이스트에서 임형진 올림

- 38 -

CURRICULUM VITAE

Hyung Jin Lim

Research Assistant

The Civil and Environmental Engineering Department

Korea Advanced Institute of Science and Technology (KAIST)

291 Daehak-Ro Guseong-Dong Yuseong-Gu

Daejeon Republic of Korea 305-701

Phone (82)+42-350-3665 Fax (82)+42-350-3610

Email limhj87gmailcom

RESEARCH INTERESTS

Damage Detection Localization and Quantification on Composite and Metallic Structures Pattern Recognition Finite Element Modeling (FEM)

EDUCATION

2009 ndash 2011 MS Civil and Environmental Eng Korea Advanced Institute of Science and Technology (KAIST) Korea Rep

2006 ndash 2009 BS Civil and Environmental Eng Korea Advanced Institute of Science and Technology (KAIST) Korea Rep

JOURNAL PUBLICATIONS The corresponding author is underlined 1 Hyung Jin Lim Hoon Sohn ldquoReference-free damage detection localization and quantification

using in compositerdquo in preparation for Journal of the Acoustical Society of America 2011

2 Chul Min Yeum Hoon Sohn Jeong Bum Ihn Hyung Jin Lim ldquoInstantaneous delamination detection in a composite using a dual piezoelectric transducer networkrdquo Submitted to Compo-sites Science and Technology 2011

3 Yun-Kyu An Hyung Jin Lim Min Koo Kim Jin Yeol Yang Hoon Sohn Chang Guen Lee ldquoInvestigation of applicability of local reference-free damage detection techniques to in-situ

- 39 -

bridgesrdquo Submitted to ASCE Journal of Structural Engineering 2011

4 Hyung Jin Lim Min Koo Kim Hoon Sohn Chan Yik Park ldquoImpedance based damage detec-tion under varying temperature and loading conditionsrdquo Submitted to NDTampE International 2011

5 Hyung Jin Lim Hoon Sohn ldquoQuantitative estimation of transmitted and reflected lamb waves at discontinuityrdquo Journal of the Korean Society for Nondestructive Testing Vol30 No 4 pp1-8 2010

CONFERENCE PROCEEDINGS

The corresponding author is underlined 1 Hyung Jin Lim and Hoon Sohn ldquoReference-free damage detection localization and quantifi-

cation in compositesrdquo ASCE Engineering Mechanics Institute Conference Boston June 2-4 2011

2 Hyung Jin Lim and Hoon Sohn ldquoReference-free delamination quantification within multi-layer composite materialsrdquo KSNT Spring Conference May 26-27 Gyeong-Ju Korea 2011

3 Min Koo Kim Hyun Jin Lim Hoon Sohn Chan Yik Park ldquoImpedance-based bolt loosening detection under varying temperature and loadingrdquo Asian Pacific Workshop on Structural Health Monitoring Tokyo Japan November 30-December 2 2010

4 Hyung Jin Lim Chul Min Yeum and Hoon Sohn ldquoModeling of Impact-induced Delamination in a Multilayer Composite Platerdquo The 23rd KKCNN Symposium on Civil Engineering Taipei (Taiwan) November 13-15 2010

5 Yun-Kyu An Hyung Jin Lim and Hoon Sohn ldquoTemperature independent Delamination Detec-tion using data normalizationrdquo The 3rd International Conference on Multi-Functional Materi-als and Structures Chonbuk National Univ Jeonju Korea September 14-18 2010

6 Hoon Sohn Yun-Kyu An Chang Gil Lee Min Koo Kim Hyung Jin Lim ldquoApplication of online NDT techniques to decommissioned bridge testingrdquo the 5th European Workshop on Structural Health Monitoring Sorrento Italy June 29-July 02 2010

7 Min Koo Kim Hyung Jin Lim Hoon Sohn Chan Yik Park ldquoImpedance based damage diag-nosis of a complex composite aircraft wing under changing loading and temperature condi-tionsrdquo the 5th European Workshop on Structural Health Monitoring Sorrento Italy June 29-July 02 2010

8 Yun-Kyu An Min Koo Kim Hyung Jin Lim and Hoon Sohn ldquoValidations of reference-free crack detection techniques through a decommissioned bridge testrdquo 2010 COSEIK Annual Con-ference Jeju Korea April 8-9 2010

- 40 -

PARTICIPANT PROJECTS

1 Development of Smart Steel Members for Online Steel Structure Monitoring POSCO Corporation (Funded 41000000 KRW (41000 USD) for 010511 to 123111)

2 Development of Enabling Technologies for Structural Health Monitoring Korean Agency for Defense Development (Funded 240000000 KRW (240000 USD) out of total 3200000000 KRW (3200000 USD) for 060108-053111)

3 Development of On-board SHM Technologies for Composite Air Vehicles The Boeing Company (Funded 400000 USD (400000000 KRW) for 080108 to 073111)

EDUCATIONAL EXPERIENCES

Fall 2009

Teaching Assistant CE207 ldquoElementary Structural Engineering and Laboratoryrdquo KAIST 3 unit undergraduate course

EXPERIMENTAL EXPERIENTCES

2010 Monitoring of Real Aircraft Composite Wing Segment Korea Agency for Defense De-velopment (ADD) Daejeon Korea Real time monitoring of composite air vehicles fund-ed by Korea Agency for Defense Development (ADD)

2009-2011

Field Monitoring of Real Bridge Structures Gimpo (Ramp G bridge) amp Icheon (Sam-seung bridge) Korea US-Korea collaborative research for bridge monitoring testbeds

HONORS amp AWARDS

Scholarships for outstanding students granted KAIST (2007-2009) The three prominent students in the civil and environmental engineering department of KAIST were selected as recipients This scholarship provided 2000$ per year for three years

- 41 -

Page 42: 복합재료 손상의 무기저 진단 및 정량화 - KAISTssslab.kaist.ac.kr/article/pdf/Thesis_2.pdf · 1 Declaration of Ethical Conduct in Research: I, as a graduate student

CHAPTER 6 CONCLUDING REMARKS

61 Executive summary

In this study the reference-free damage detection localization and quantification tech-

nique in a multi layered composite plate is proposed Additional reflected A0 Lamb wave

mode from the damage due to multiple reflections in damage area is used as a feature for

characterizing the damage First damage is characterized in terms of a damage index (m-

value) defined as the ratio of the physical length of the damage to the wavelength of the A0

Lamb wave mode Then the range of possible damage location length and severity are esti-

mated with constraints This procedure can be accomplished without prior baseline data To

realize the proposed technique the matching pursuit technique is modified for reduction of

the error by considering the dispersive characteristic of Lamb wave The matching pursuit

method estimates the arrival time of each wave packets accurately even two wave packets are

overlapped The numerical simulations and experimental tests including Teflon tape inserted

and impact-induced damage are investigated for the feasibility of the proposed technique

Moreover the effectiveness of the proposed technique under changing temperature is exper-

imentally verified

- 31 -

62 Suggestion for further study

The limitations of this study and the possible improvement and further studies are listed as

follow

1 Measurement of impact-induced damage length using conventional NDT technique

Additional experimental verification for impact-induced damage characterization is

needed by measuring the real length of damage using conventional NDT technique

such as ultrasonic scan or infrared camera

2 Limitation of both side collocated PZT installation In this study excitation PZTs are

installed both side and collocated for decomposing A0 Lamb wave mode However it

can produce decomposition error due to imperfection of PZT location and their installa-

tion Moreover in many cases assess to both surface of structures is not allowed such

as aircraft applications To improve this limitation a reference-free damage characteri-

zation using single side PZTs will be considered

3 Complex structure application In this study damage is characterized in simple plate-

like structure When the monitored structure becomes complex the response signals al-

so becomes complex due to its geometry and boundary Thus the investigations for

complex structure application such as stiffener spar and curvature are needed

- 32 -

REFERENCES

[1] Cawley P and Alleyne D (1996) ldquoThe use of Lamb waves for the long range inspection of large structuresrdquo Ultrasonics 34 pp287-290

[2] Ghosh T Kundu T and Karpur P (1998) ldquoEfficient use of Lamb waves for detecting defects in large platesrdquo Ultrasonics 36 pp791-801

[3] Guirguitui V (2003) ldquoLamb wave generation with piezoelectric wafer active sensors for structural health monitoringrdquo SPIE Conference on Smart Structures and Materials San Diego CA 5391 pp111-422

[4] Ragvahan A and Cesnik CE S (2004) ldquoModeling of piezoelectric-based Lamb wave generation and sensing for structural health monitoringrdquo SPIE Conference on Smart Struc-tures and Materials San Diego CA 5391 pp419-430

[5] Rose JL Pilarski A and Dirit JJ (1993) ldquoAn approach to guided wave mode selection for inspection of laminated platerdquo Journal of Reinforced Plastics and Composites 12(5) pp536-544

[6] Guo N and Cawley P (1993) ldquoThe interaction of Lamb waves with delaminations in composite laminatesrdquo Journal of Acoustical Society of America 94(4) pp 2240-2246

[7] Valdez SHD and Soutis C (2002) ldquoReal-time nondestructive evaluation of fiber com-posite laminates using low-frequency Lamb wavesrdquo Journal of Acoustical Society of Amer-ica 111(5) pp 2026-2033

[8] Yang M and Qiao P (2005) ldquoModeling and experimental detection of damage in virous materials using the pulse-echo method and piezoelectric sensoractuatorsrdquo Smart Materials and Structures 14 pp1083-1100

[9] Lestari W and Qiao P (2005) ldquoApplication of wave propagation analysis for damage identification in composite laminated beamsrdquo Journal of Composite Materials 39(22) pp1967-1984

[10] Kessler S S Spearing S M and Soutis C (2002) ldquoDamage detection in composite ma-terials using Lamb wave methodsrdquo Smart Materials and Structures 11 pp269-278

[11] Sohn H Park G Wail J R Limback N P and Farrar C R (2004) ldquoWavelet-based active sensing for delamination detection in composite structuresrdquo Smart Materials and Structures 13 pp 153-160

[12] Petculescu G Krishnaswamy S and Achenback J D (2008) ldquoGroup delay measure-ments using modally selective Lamb wave transducers for detection and sizing of delamina-tions in compositesrdquo Smart Materials and Structures 17

[13] Sohn H Park H W Law K H and Farrar C R (2007) ldquoDamage detection in compo-

- 33 -

site plates by using an enhanced time reversal methodrdquo Journal of Aerospace Engineering (ASCE) pp 141-151

[14] Sohn H (2007) ldquoEffects of environmental and operational variability on structural health monitoringrdquo Philosophical Transactions of the Royal Society Ardquo 365 pp539-560

[15] Rytter A (1993) ldquoVibration based inspection of civil engineering structuresrdquo PhD The-sis Department of Building Technology and Structural Engineering University of Aalborg Aalborg Denmark

[16] Hayashi T and Kawashima K (2002) ldquoMultiple reflections of Lamb waves at a delami-nationrdquo Ultrasonics 40 pp193-197

[17] Wang C H and Rose L R F (2003) ldquoWave reflection and transmission in beams con-taining delamination and inhomogeneityrdquo Journal of Sound and Vibration 264 pp 851-872

[18] Ramadas C Balasubramaniam K Joshi M and Krishnamurthy C V (2009) ldquoInterac-tion of the primary anti-symmetric Lamb mode (Ao) with symmetric delaminations numer-ical and experimental studiesrdquo Smart Materials and Structures 18

[19] Ramadas C Balasubramaniam K Joshi M and Krishnamurthy C V (2010) ldquoInterac-tion of guided Lamb waves with an asymmetrically located delamination in a laminated composite platerdquo Smart Materials and Structures 19

[20] Wilcox P Lowe M and Cawely P (2001) ldquoThe effect of dispersion on long-range in-spection using ultrasonic guided wavesrdquo NDTampE International 34 pp1-9

[21] Mallat S G and Zhang Z (1993) ldquoMatching pursuits with time-frequency dictionariesrdquo IEEE Tansactions on Signal Processing 41(12) pp3397-3415

[22] Zhang G Zhang S and Wang Y (2000) ldquoApplication of adaptive time-frequency de-composition in ultrasonic NDE of highly scattering materialsrdquo Ultrasonics 38 pp 961-964

[23] Hong J C Sun K H and Kim Y Y (2005) ldquoThe matching pursuit approach based on the modulated Gaussian pulse for efficient guided-wave damage inspectionrdquo Smart Mate-rials and Structures 14 pp548-560

[24] Herakovich C T (1997) Mechanics of Fibrous Composites Wiley 480 pages

[25] Reddy J N (1985) ldquoA review of the literature on finite-element modeling of laminated composite platesrdquo The Shock and Vibration Digest 17(4) pp 3-8

[26] Reddy J N (2003) Mechanics of Laminated Composite Plates and Shells Theory and Analysis CRC Press 856 pages

[27] Palacz M Krawczuk M and Ostachowicz W (2005) ldquoThe spectral finite element mod-el for analysis of flexure-shear coupled wave propagation Part 1 laminated multilayer

- 34 -

composite beamrdquo Composite Structures 68 pp37-44

[28] Zhang Y C and Harding J (1990) ldquoA numerical micromechanics analysis of the me-chanical properties of a plain weave compositerdquo Computer amp Structures 36(5) pp 839-844

[29] Nemat-Nasser S and Hori M (1999) Micromechanics Overall Properties of Heteroge-neous Materials Elsevier 810 pages

[30] Scida D Aboura Z Benzeggagh M L and Bocherens E (1999) ldquoA micromechanics model for 3D elasticity and failure of woven-fibre composite materialrdquo Composites Science and Technology 59 pp 505-517

[31] Thomas Y and Wu X H (1997) ldquoA multiscale finite element method for elliptic prob-lems in composite materials and porous mediardquo Journal of Computational Physics 134 pp 169-189

[32] Lee K H Moorthy S and Ghosh S (1999) ldquoMultiple scale computational model for damage in composite materialsrdquo Computer Methods in Applied Mechanics and Engineer-ing 172 pp 175-201

[33] Namilae S and Chandra N (2005) ldquoMultiscale model to study the effect of interfaces in carbon nanotube-based compositesrdquo Journal of Engineering Materials and Technology (ASME) 127 pp222-232

[34] Adams D E (2007) Health monitoring of Structural Materials and Components Wiley 460 pages

[35] Gopalakrishnan S (2009) ldquoModeling Aspects in Finite Elementrdquo Encyclopedia of Struc-tural Health Monitoring 2 pp791-809

[36] Palacz M Krawczuk M and Ostachowicz W (2005) ldquoThe spectral finite element mod-el for analysis of flexure-shear coupled wave propagation Part 2 Deminated multilayer composite beamrdquo Composite Structures 68 pp45-51

[37] Baik J M and Thopson R B (1984) ldquoUltrasonic scattering from imperfect interfaces A quasi-static modelrdquo Journal of Nondestructive Evaluation 4(3) pp177-196

[38] Luo H and Hanagud S (2000) ldquoDynamics of delamincated beamsrdquo International Jour-nal of Solids and Structures 37 pp1501-1519

[39] Kim S B and Sohn H (2007) ldquoInstantaneous reference-free crack detection based on polarization characteristics of piezoelectric materialsrdquo Smart Materials and Structures 16 pp2375-2387

[40] An Y K and Sohn H (2010) ldquoInstantaneous crack detection under varying temperature and static loading conditionsrdquo The Special Issue of Structural Control and Health Monitor-ing 17(7) pp 730-741

- 35 -

[41] Lee S J and Sohn H (2006) ldquoActive self-sensing scheme development for structural health monitoringrdquo Smart Materials and Structures 15(6) pp 1734-1746

- 36 -

요 약 문

복합재료 손상의 무기저 진단 및 정량화

복합재료는 재료의 경량성 내구성 고강도의 특성으로 인하여 항공기에서부터

교량에까지 널리 사용되고 있다 하지만 외부의 충격으로 인한 복합재료 내부의

손상은 재료의 강도를 감소시켜 구조물의 안정성을 저하시키는 요인이 된다 하지만

복합재료의 손상은 재료 내부에서 발생하여 육안으로는 식별하기도 힘들며 표면에서

보이는 것만으로 내부의 손상 정도를 판단하기는 불가능하다는 단점을 가지고 있다

본 연구에서는 손상 내부에서 반사로 인해 발생하는 파를 이용하여 기저자료를

사용하지 않고 특정 시점에서 얻은 자료만으로 복합재료의 손상을 진단하고 정량화

할 수 있는 기법을 제안한다 제안된 기법의 적용성을 검증하기 위하여 유한요소법을

이용한 수치해석을 수행하였고 Teflon tape 이 삽입된 시편과 실제 충격하중으로

인한 손상을 입은 시편을 이용한 실험을 진행하였다 수치해석 및 실험 결과 제안된

기법이 복합재료의 손상을 성공적으로 진단 및 정량화할 수 있음을 확인하였다 또한

다양한 외부 온도 환경에서 제안된 기법의 환경영향 실험을 수행하였고 다양한

환경에서 손상의 진단 및 정량화가 성공적으로 이루어짐을 검증하였다

핵심어 무기저 손상진단 및 정량화 압전센서 복합재료 유한요소법 Matching pursuit

method

- 37 -

감 사 의 글

기대와 두려움을 안고 시작한 석사과정이 어느덧 2 년 이라는 세월 속에서 추억으로 자리잡아

가고 있습니다 학부생 기간을 포함한 지난 3 년 동안 SHM 이라는 새로운 분야를 접하고 연구를

수행 할 수 있도록 도와주신 고마운 분들이 있어 그 감사를 표하고자 합니다

우선 저의 많은 부족한 점을 일깨워 주시고 연구에 대한 뜻을 더욱 확고히 세울 수 있게 해

주신 지도교수님이신 손 훈 교수님께 깊은 감사의 뜻을 전합니다 짧은 기간이었지만 교수님의

도움으로 많은 것을 배우고 느끼고 깨달을 수 있었습니다 또한 본 연구를 수행할 수 있도록 지

원해주신 국토해양부 (U-City 석 박사과정 지원사업)에 깊은 감사 드립니다

때로 힘들고 지칠 때 항상 저를 믿고 응원해 주시는 가족이 있었기에 지금의 글을 쓸 수 있었

습니다 또한 지금 연구실에서 함께 생활하고 있는 윤규형 민구형 철민이형 현석이형 진열이형

호민이형 병진이 선혜누나와 성대에 계신 창길이형 미국에 계신 의재형과 Debaditiya 에게도 감

사의 말씀을 전합니다 그리고 모두 나열할 수는 없지만 카이스트 생활의 활력소가 되어준 카이

스트 건설 및 환경공학과 선후배들 직원 분들과 SOS 멤버들에게도 감사를 표합니다

앞서 거론한 많은 이들의 도움이 없었다면 저 또한 지금 이 자리에 있을 수 없었을 것입니다

많이 부족하기에 더 노력하고 배우고자 했던 초심을 다시 한번 생각하며 더 큰 결실을 맺기 위해

앞으로 나아가겠습니다 도움을 주신 모든 분들께 이 감사의 글을 바칩니다

2011년 6월 카이스트에서 임형진 올림

- 38 -

CURRICULUM VITAE

Hyung Jin Lim

Research Assistant

The Civil and Environmental Engineering Department

Korea Advanced Institute of Science and Technology (KAIST)

291 Daehak-Ro Guseong-Dong Yuseong-Gu

Daejeon Republic of Korea 305-701

Phone (82)+42-350-3665 Fax (82)+42-350-3610

Email limhj87gmailcom

RESEARCH INTERESTS

Damage Detection Localization and Quantification on Composite and Metallic Structures Pattern Recognition Finite Element Modeling (FEM)

EDUCATION

2009 ndash 2011 MS Civil and Environmental Eng Korea Advanced Institute of Science and Technology (KAIST) Korea Rep

2006 ndash 2009 BS Civil and Environmental Eng Korea Advanced Institute of Science and Technology (KAIST) Korea Rep

JOURNAL PUBLICATIONS The corresponding author is underlined 1 Hyung Jin Lim Hoon Sohn ldquoReference-free damage detection localization and quantification

using in compositerdquo in preparation for Journal of the Acoustical Society of America 2011

2 Chul Min Yeum Hoon Sohn Jeong Bum Ihn Hyung Jin Lim ldquoInstantaneous delamination detection in a composite using a dual piezoelectric transducer networkrdquo Submitted to Compo-sites Science and Technology 2011

3 Yun-Kyu An Hyung Jin Lim Min Koo Kim Jin Yeol Yang Hoon Sohn Chang Guen Lee ldquoInvestigation of applicability of local reference-free damage detection techniques to in-situ

- 39 -

bridgesrdquo Submitted to ASCE Journal of Structural Engineering 2011

4 Hyung Jin Lim Min Koo Kim Hoon Sohn Chan Yik Park ldquoImpedance based damage detec-tion under varying temperature and loading conditionsrdquo Submitted to NDTampE International 2011

5 Hyung Jin Lim Hoon Sohn ldquoQuantitative estimation of transmitted and reflected lamb waves at discontinuityrdquo Journal of the Korean Society for Nondestructive Testing Vol30 No 4 pp1-8 2010

CONFERENCE PROCEEDINGS

The corresponding author is underlined 1 Hyung Jin Lim and Hoon Sohn ldquoReference-free damage detection localization and quantifi-

cation in compositesrdquo ASCE Engineering Mechanics Institute Conference Boston June 2-4 2011

2 Hyung Jin Lim and Hoon Sohn ldquoReference-free delamination quantification within multi-layer composite materialsrdquo KSNT Spring Conference May 26-27 Gyeong-Ju Korea 2011

3 Min Koo Kim Hyun Jin Lim Hoon Sohn Chan Yik Park ldquoImpedance-based bolt loosening detection under varying temperature and loadingrdquo Asian Pacific Workshop on Structural Health Monitoring Tokyo Japan November 30-December 2 2010

4 Hyung Jin Lim Chul Min Yeum and Hoon Sohn ldquoModeling of Impact-induced Delamination in a Multilayer Composite Platerdquo The 23rd KKCNN Symposium on Civil Engineering Taipei (Taiwan) November 13-15 2010

5 Yun-Kyu An Hyung Jin Lim and Hoon Sohn ldquoTemperature independent Delamination Detec-tion using data normalizationrdquo The 3rd International Conference on Multi-Functional Materi-als and Structures Chonbuk National Univ Jeonju Korea September 14-18 2010

6 Hoon Sohn Yun-Kyu An Chang Gil Lee Min Koo Kim Hyung Jin Lim ldquoApplication of online NDT techniques to decommissioned bridge testingrdquo the 5th European Workshop on Structural Health Monitoring Sorrento Italy June 29-July 02 2010

7 Min Koo Kim Hyung Jin Lim Hoon Sohn Chan Yik Park ldquoImpedance based damage diag-nosis of a complex composite aircraft wing under changing loading and temperature condi-tionsrdquo the 5th European Workshop on Structural Health Monitoring Sorrento Italy June 29-July 02 2010

8 Yun-Kyu An Min Koo Kim Hyung Jin Lim and Hoon Sohn ldquoValidations of reference-free crack detection techniques through a decommissioned bridge testrdquo 2010 COSEIK Annual Con-ference Jeju Korea April 8-9 2010

- 40 -

PARTICIPANT PROJECTS

1 Development of Smart Steel Members for Online Steel Structure Monitoring POSCO Corporation (Funded 41000000 KRW (41000 USD) for 010511 to 123111)

2 Development of Enabling Technologies for Structural Health Monitoring Korean Agency for Defense Development (Funded 240000000 KRW (240000 USD) out of total 3200000000 KRW (3200000 USD) for 060108-053111)

3 Development of On-board SHM Technologies for Composite Air Vehicles The Boeing Company (Funded 400000 USD (400000000 KRW) for 080108 to 073111)

EDUCATIONAL EXPERIENCES

Fall 2009

Teaching Assistant CE207 ldquoElementary Structural Engineering and Laboratoryrdquo KAIST 3 unit undergraduate course

EXPERIMENTAL EXPERIENTCES

2010 Monitoring of Real Aircraft Composite Wing Segment Korea Agency for Defense De-velopment (ADD) Daejeon Korea Real time monitoring of composite air vehicles fund-ed by Korea Agency for Defense Development (ADD)

2009-2011

Field Monitoring of Real Bridge Structures Gimpo (Ramp G bridge) amp Icheon (Sam-seung bridge) Korea US-Korea collaborative research for bridge monitoring testbeds

HONORS amp AWARDS

Scholarships for outstanding students granted KAIST (2007-2009) The three prominent students in the civil and environmental engineering department of KAIST were selected as recipients This scholarship provided 2000$ per year for three years

- 41 -

Page 43: 복합재료 손상의 무기저 진단 및 정량화 - KAISTssslab.kaist.ac.kr/article/pdf/Thesis_2.pdf · 1 Declaration of Ethical Conduct in Research: I, as a graduate student

62 Suggestion for further study

The limitations of this study and the possible improvement and further studies are listed as

follow

1 Measurement of impact-induced damage length using conventional NDT technique

Additional experimental verification for impact-induced damage characterization is

needed by measuring the real length of damage using conventional NDT technique

such as ultrasonic scan or infrared camera

2 Limitation of both side collocated PZT installation In this study excitation PZTs are

installed both side and collocated for decomposing A0 Lamb wave mode However it

can produce decomposition error due to imperfection of PZT location and their installa-

tion Moreover in many cases assess to both surface of structures is not allowed such

as aircraft applications To improve this limitation a reference-free damage characteri-

zation using single side PZTs will be considered

3 Complex structure application In this study damage is characterized in simple plate-

like structure When the monitored structure becomes complex the response signals al-

so becomes complex due to its geometry and boundary Thus the investigations for

complex structure application such as stiffener spar and curvature are needed

- 32 -

REFERENCES

[1] Cawley P and Alleyne D (1996) ldquoThe use of Lamb waves for the long range inspection of large structuresrdquo Ultrasonics 34 pp287-290

[2] Ghosh T Kundu T and Karpur P (1998) ldquoEfficient use of Lamb waves for detecting defects in large platesrdquo Ultrasonics 36 pp791-801

[3] Guirguitui V (2003) ldquoLamb wave generation with piezoelectric wafer active sensors for structural health monitoringrdquo SPIE Conference on Smart Structures and Materials San Diego CA 5391 pp111-422

[4] Ragvahan A and Cesnik CE S (2004) ldquoModeling of piezoelectric-based Lamb wave generation and sensing for structural health monitoringrdquo SPIE Conference on Smart Struc-tures and Materials San Diego CA 5391 pp419-430

[5] Rose JL Pilarski A and Dirit JJ (1993) ldquoAn approach to guided wave mode selection for inspection of laminated platerdquo Journal of Reinforced Plastics and Composites 12(5) pp536-544

[6] Guo N and Cawley P (1993) ldquoThe interaction of Lamb waves with delaminations in composite laminatesrdquo Journal of Acoustical Society of America 94(4) pp 2240-2246

[7] Valdez SHD and Soutis C (2002) ldquoReal-time nondestructive evaluation of fiber com-posite laminates using low-frequency Lamb wavesrdquo Journal of Acoustical Society of Amer-ica 111(5) pp 2026-2033

[8] Yang M and Qiao P (2005) ldquoModeling and experimental detection of damage in virous materials using the pulse-echo method and piezoelectric sensoractuatorsrdquo Smart Materials and Structures 14 pp1083-1100

[9] Lestari W and Qiao P (2005) ldquoApplication of wave propagation analysis for damage identification in composite laminated beamsrdquo Journal of Composite Materials 39(22) pp1967-1984

[10] Kessler S S Spearing S M and Soutis C (2002) ldquoDamage detection in composite ma-terials using Lamb wave methodsrdquo Smart Materials and Structures 11 pp269-278

[11] Sohn H Park G Wail J R Limback N P and Farrar C R (2004) ldquoWavelet-based active sensing for delamination detection in composite structuresrdquo Smart Materials and Structures 13 pp 153-160

[12] Petculescu G Krishnaswamy S and Achenback J D (2008) ldquoGroup delay measure-ments using modally selective Lamb wave transducers for detection and sizing of delamina-tions in compositesrdquo Smart Materials and Structures 17

[13] Sohn H Park H W Law K H and Farrar C R (2007) ldquoDamage detection in compo-

- 33 -

site plates by using an enhanced time reversal methodrdquo Journal of Aerospace Engineering (ASCE) pp 141-151

[14] Sohn H (2007) ldquoEffects of environmental and operational variability on structural health monitoringrdquo Philosophical Transactions of the Royal Society Ardquo 365 pp539-560

[15] Rytter A (1993) ldquoVibration based inspection of civil engineering structuresrdquo PhD The-sis Department of Building Technology and Structural Engineering University of Aalborg Aalborg Denmark

[16] Hayashi T and Kawashima K (2002) ldquoMultiple reflections of Lamb waves at a delami-nationrdquo Ultrasonics 40 pp193-197

[17] Wang C H and Rose L R F (2003) ldquoWave reflection and transmission in beams con-taining delamination and inhomogeneityrdquo Journal of Sound and Vibration 264 pp 851-872

[18] Ramadas C Balasubramaniam K Joshi M and Krishnamurthy C V (2009) ldquoInterac-tion of the primary anti-symmetric Lamb mode (Ao) with symmetric delaminations numer-ical and experimental studiesrdquo Smart Materials and Structures 18

[19] Ramadas C Balasubramaniam K Joshi M and Krishnamurthy C V (2010) ldquoInterac-tion of guided Lamb waves with an asymmetrically located delamination in a laminated composite platerdquo Smart Materials and Structures 19

[20] Wilcox P Lowe M and Cawely P (2001) ldquoThe effect of dispersion on long-range in-spection using ultrasonic guided wavesrdquo NDTampE International 34 pp1-9

[21] Mallat S G and Zhang Z (1993) ldquoMatching pursuits with time-frequency dictionariesrdquo IEEE Tansactions on Signal Processing 41(12) pp3397-3415

[22] Zhang G Zhang S and Wang Y (2000) ldquoApplication of adaptive time-frequency de-composition in ultrasonic NDE of highly scattering materialsrdquo Ultrasonics 38 pp 961-964

[23] Hong J C Sun K H and Kim Y Y (2005) ldquoThe matching pursuit approach based on the modulated Gaussian pulse for efficient guided-wave damage inspectionrdquo Smart Mate-rials and Structures 14 pp548-560

[24] Herakovich C T (1997) Mechanics of Fibrous Composites Wiley 480 pages

[25] Reddy J N (1985) ldquoA review of the literature on finite-element modeling of laminated composite platesrdquo The Shock and Vibration Digest 17(4) pp 3-8

[26] Reddy J N (2003) Mechanics of Laminated Composite Plates and Shells Theory and Analysis CRC Press 856 pages

[27] Palacz M Krawczuk M and Ostachowicz W (2005) ldquoThe spectral finite element mod-el for analysis of flexure-shear coupled wave propagation Part 1 laminated multilayer

- 34 -

composite beamrdquo Composite Structures 68 pp37-44

[28] Zhang Y C and Harding J (1990) ldquoA numerical micromechanics analysis of the me-chanical properties of a plain weave compositerdquo Computer amp Structures 36(5) pp 839-844

[29] Nemat-Nasser S and Hori M (1999) Micromechanics Overall Properties of Heteroge-neous Materials Elsevier 810 pages

[30] Scida D Aboura Z Benzeggagh M L and Bocherens E (1999) ldquoA micromechanics model for 3D elasticity and failure of woven-fibre composite materialrdquo Composites Science and Technology 59 pp 505-517

[31] Thomas Y and Wu X H (1997) ldquoA multiscale finite element method for elliptic prob-lems in composite materials and porous mediardquo Journal of Computational Physics 134 pp 169-189

[32] Lee K H Moorthy S and Ghosh S (1999) ldquoMultiple scale computational model for damage in composite materialsrdquo Computer Methods in Applied Mechanics and Engineer-ing 172 pp 175-201

[33] Namilae S and Chandra N (2005) ldquoMultiscale model to study the effect of interfaces in carbon nanotube-based compositesrdquo Journal of Engineering Materials and Technology (ASME) 127 pp222-232

[34] Adams D E (2007) Health monitoring of Structural Materials and Components Wiley 460 pages

[35] Gopalakrishnan S (2009) ldquoModeling Aspects in Finite Elementrdquo Encyclopedia of Struc-tural Health Monitoring 2 pp791-809

[36] Palacz M Krawczuk M and Ostachowicz W (2005) ldquoThe spectral finite element mod-el for analysis of flexure-shear coupled wave propagation Part 2 Deminated multilayer composite beamrdquo Composite Structures 68 pp45-51

[37] Baik J M and Thopson R B (1984) ldquoUltrasonic scattering from imperfect interfaces A quasi-static modelrdquo Journal of Nondestructive Evaluation 4(3) pp177-196

[38] Luo H and Hanagud S (2000) ldquoDynamics of delamincated beamsrdquo International Jour-nal of Solids and Structures 37 pp1501-1519

[39] Kim S B and Sohn H (2007) ldquoInstantaneous reference-free crack detection based on polarization characteristics of piezoelectric materialsrdquo Smart Materials and Structures 16 pp2375-2387

[40] An Y K and Sohn H (2010) ldquoInstantaneous crack detection under varying temperature and static loading conditionsrdquo The Special Issue of Structural Control and Health Monitor-ing 17(7) pp 730-741

- 35 -

[41] Lee S J and Sohn H (2006) ldquoActive self-sensing scheme development for structural health monitoringrdquo Smart Materials and Structures 15(6) pp 1734-1746

- 36 -

요 약 문

복합재료 손상의 무기저 진단 및 정량화

복합재료는 재료의 경량성 내구성 고강도의 특성으로 인하여 항공기에서부터

교량에까지 널리 사용되고 있다 하지만 외부의 충격으로 인한 복합재료 내부의

손상은 재료의 강도를 감소시켜 구조물의 안정성을 저하시키는 요인이 된다 하지만

복합재료의 손상은 재료 내부에서 발생하여 육안으로는 식별하기도 힘들며 표면에서

보이는 것만으로 내부의 손상 정도를 판단하기는 불가능하다는 단점을 가지고 있다

본 연구에서는 손상 내부에서 반사로 인해 발생하는 파를 이용하여 기저자료를

사용하지 않고 특정 시점에서 얻은 자료만으로 복합재료의 손상을 진단하고 정량화

할 수 있는 기법을 제안한다 제안된 기법의 적용성을 검증하기 위하여 유한요소법을

이용한 수치해석을 수행하였고 Teflon tape 이 삽입된 시편과 실제 충격하중으로

인한 손상을 입은 시편을 이용한 실험을 진행하였다 수치해석 및 실험 결과 제안된

기법이 복합재료의 손상을 성공적으로 진단 및 정량화할 수 있음을 확인하였다 또한

다양한 외부 온도 환경에서 제안된 기법의 환경영향 실험을 수행하였고 다양한

환경에서 손상의 진단 및 정량화가 성공적으로 이루어짐을 검증하였다

핵심어 무기저 손상진단 및 정량화 압전센서 복합재료 유한요소법 Matching pursuit

method

- 37 -

감 사 의 글

기대와 두려움을 안고 시작한 석사과정이 어느덧 2 년 이라는 세월 속에서 추억으로 자리잡아

가고 있습니다 학부생 기간을 포함한 지난 3 년 동안 SHM 이라는 새로운 분야를 접하고 연구를

수행 할 수 있도록 도와주신 고마운 분들이 있어 그 감사를 표하고자 합니다

우선 저의 많은 부족한 점을 일깨워 주시고 연구에 대한 뜻을 더욱 확고히 세울 수 있게 해

주신 지도교수님이신 손 훈 교수님께 깊은 감사의 뜻을 전합니다 짧은 기간이었지만 교수님의

도움으로 많은 것을 배우고 느끼고 깨달을 수 있었습니다 또한 본 연구를 수행할 수 있도록 지

원해주신 국토해양부 (U-City 석 박사과정 지원사업)에 깊은 감사 드립니다

때로 힘들고 지칠 때 항상 저를 믿고 응원해 주시는 가족이 있었기에 지금의 글을 쓸 수 있었

습니다 또한 지금 연구실에서 함께 생활하고 있는 윤규형 민구형 철민이형 현석이형 진열이형

호민이형 병진이 선혜누나와 성대에 계신 창길이형 미국에 계신 의재형과 Debaditiya 에게도 감

사의 말씀을 전합니다 그리고 모두 나열할 수는 없지만 카이스트 생활의 활력소가 되어준 카이

스트 건설 및 환경공학과 선후배들 직원 분들과 SOS 멤버들에게도 감사를 표합니다

앞서 거론한 많은 이들의 도움이 없었다면 저 또한 지금 이 자리에 있을 수 없었을 것입니다

많이 부족하기에 더 노력하고 배우고자 했던 초심을 다시 한번 생각하며 더 큰 결실을 맺기 위해

앞으로 나아가겠습니다 도움을 주신 모든 분들께 이 감사의 글을 바칩니다

2011년 6월 카이스트에서 임형진 올림

- 38 -

CURRICULUM VITAE

Hyung Jin Lim

Research Assistant

The Civil and Environmental Engineering Department

Korea Advanced Institute of Science and Technology (KAIST)

291 Daehak-Ro Guseong-Dong Yuseong-Gu

Daejeon Republic of Korea 305-701

Phone (82)+42-350-3665 Fax (82)+42-350-3610

Email limhj87gmailcom

RESEARCH INTERESTS

Damage Detection Localization and Quantification on Composite and Metallic Structures Pattern Recognition Finite Element Modeling (FEM)

EDUCATION

2009 ndash 2011 MS Civil and Environmental Eng Korea Advanced Institute of Science and Technology (KAIST) Korea Rep

2006 ndash 2009 BS Civil and Environmental Eng Korea Advanced Institute of Science and Technology (KAIST) Korea Rep

JOURNAL PUBLICATIONS The corresponding author is underlined 1 Hyung Jin Lim Hoon Sohn ldquoReference-free damage detection localization and quantification

using in compositerdquo in preparation for Journal of the Acoustical Society of America 2011

2 Chul Min Yeum Hoon Sohn Jeong Bum Ihn Hyung Jin Lim ldquoInstantaneous delamination detection in a composite using a dual piezoelectric transducer networkrdquo Submitted to Compo-sites Science and Technology 2011

3 Yun-Kyu An Hyung Jin Lim Min Koo Kim Jin Yeol Yang Hoon Sohn Chang Guen Lee ldquoInvestigation of applicability of local reference-free damage detection techniques to in-situ

- 39 -

bridgesrdquo Submitted to ASCE Journal of Structural Engineering 2011

4 Hyung Jin Lim Min Koo Kim Hoon Sohn Chan Yik Park ldquoImpedance based damage detec-tion under varying temperature and loading conditionsrdquo Submitted to NDTampE International 2011

5 Hyung Jin Lim Hoon Sohn ldquoQuantitative estimation of transmitted and reflected lamb waves at discontinuityrdquo Journal of the Korean Society for Nondestructive Testing Vol30 No 4 pp1-8 2010

CONFERENCE PROCEEDINGS

The corresponding author is underlined 1 Hyung Jin Lim and Hoon Sohn ldquoReference-free damage detection localization and quantifi-

cation in compositesrdquo ASCE Engineering Mechanics Institute Conference Boston June 2-4 2011

2 Hyung Jin Lim and Hoon Sohn ldquoReference-free delamination quantification within multi-layer composite materialsrdquo KSNT Spring Conference May 26-27 Gyeong-Ju Korea 2011

3 Min Koo Kim Hyun Jin Lim Hoon Sohn Chan Yik Park ldquoImpedance-based bolt loosening detection under varying temperature and loadingrdquo Asian Pacific Workshop on Structural Health Monitoring Tokyo Japan November 30-December 2 2010

4 Hyung Jin Lim Chul Min Yeum and Hoon Sohn ldquoModeling of Impact-induced Delamination in a Multilayer Composite Platerdquo The 23rd KKCNN Symposium on Civil Engineering Taipei (Taiwan) November 13-15 2010

5 Yun-Kyu An Hyung Jin Lim and Hoon Sohn ldquoTemperature independent Delamination Detec-tion using data normalizationrdquo The 3rd International Conference on Multi-Functional Materi-als and Structures Chonbuk National Univ Jeonju Korea September 14-18 2010

6 Hoon Sohn Yun-Kyu An Chang Gil Lee Min Koo Kim Hyung Jin Lim ldquoApplication of online NDT techniques to decommissioned bridge testingrdquo the 5th European Workshop on Structural Health Monitoring Sorrento Italy June 29-July 02 2010

7 Min Koo Kim Hyung Jin Lim Hoon Sohn Chan Yik Park ldquoImpedance based damage diag-nosis of a complex composite aircraft wing under changing loading and temperature condi-tionsrdquo the 5th European Workshop on Structural Health Monitoring Sorrento Italy June 29-July 02 2010

8 Yun-Kyu An Min Koo Kim Hyung Jin Lim and Hoon Sohn ldquoValidations of reference-free crack detection techniques through a decommissioned bridge testrdquo 2010 COSEIK Annual Con-ference Jeju Korea April 8-9 2010

- 40 -

PARTICIPANT PROJECTS

1 Development of Smart Steel Members for Online Steel Structure Monitoring POSCO Corporation (Funded 41000000 KRW (41000 USD) for 010511 to 123111)

2 Development of Enabling Technologies for Structural Health Monitoring Korean Agency for Defense Development (Funded 240000000 KRW (240000 USD) out of total 3200000000 KRW (3200000 USD) for 060108-053111)

3 Development of On-board SHM Technologies for Composite Air Vehicles The Boeing Company (Funded 400000 USD (400000000 KRW) for 080108 to 073111)

EDUCATIONAL EXPERIENCES

Fall 2009

Teaching Assistant CE207 ldquoElementary Structural Engineering and Laboratoryrdquo KAIST 3 unit undergraduate course

EXPERIMENTAL EXPERIENTCES

2010 Monitoring of Real Aircraft Composite Wing Segment Korea Agency for Defense De-velopment (ADD) Daejeon Korea Real time monitoring of composite air vehicles fund-ed by Korea Agency for Defense Development (ADD)

2009-2011

Field Monitoring of Real Bridge Structures Gimpo (Ramp G bridge) amp Icheon (Sam-seung bridge) Korea US-Korea collaborative research for bridge monitoring testbeds

HONORS amp AWARDS

Scholarships for outstanding students granted KAIST (2007-2009) The three prominent students in the civil and environmental engineering department of KAIST were selected as recipients This scholarship provided 2000$ per year for three years

- 41 -

Page 44: 복합재료 손상의 무기저 진단 및 정량화 - KAISTssslab.kaist.ac.kr/article/pdf/Thesis_2.pdf · 1 Declaration of Ethical Conduct in Research: I, as a graduate student

REFERENCES

[1] Cawley P and Alleyne D (1996) ldquoThe use of Lamb waves for the long range inspection of large structuresrdquo Ultrasonics 34 pp287-290

[2] Ghosh T Kundu T and Karpur P (1998) ldquoEfficient use of Lamb waves for detecting defects in large platesrdquo Ultrasonics 36 pp791-801

[3] Guirguitui V (2003) ldquoLamb wave generation with piezoelectric wafer active sensors for structural health monitoringrdquo SPIE Conference on Smart Structures and Materials San Diego CA 5391 pp111-422

[4] Ragvahan A and Cesnik CE S (2004) ldquoModeling of piezoelectric-based Lamb wave generation and sensing for structural health monitoringrdquo SPIE Conference on Smart Struc-tures and Materials San Diego CA 5391 pp419-430

[5] Rose JL Pilarski A and Dirit JJ (1993) ldquoAn approach to guided wave mode selection for inspection of laminated platerdquo Journal of Reinforced Plastics and Composites 12(5) pp536-544

[6] Guo N and Cawley P (1993) ldquoThe interaction of Lamb waves with delaminations in composite laminatesrdquo Journal of Acoustical Society of America 94(4) pp 2240-2246

[7] Valdez SHD and Soutis C (2002) ldquoReal-time nondestructive evaluation of fiber com-posite laminates using low-frequency Lamb wavesrdquo Journal of Acoustical Society of Amer-ica 111(5) pp 2026-2033

[8] Yang M and Qiao P (2005) ldquoModeling and experimental detection of damage in virous materials using the pulse-echo method and piezoelectric sensoractuatorsrdquo Smart Materials and Structures 14 pp1083-1100

[9] Lestari W and Qiao P (2005) ldquoApplication of wave propagation analysis for damage identification in composite laminated beamsrdquo Journal of Composite Materials 39(22) pp1967-1984

[10] Kessler S S Spearing S M and Soutis C (2002) ldquoDamage detection in composite ma-terials using Lamb wave methodsrdquo Smart Materials and Structures 11 pp269-278

[11] Sohn H Park G Wail J R Limback N P and Farrar C R (2004) ldquoWavelet-based active sensing for delamination detection in composite structuresrdquo Smart Materials and Structures 13 pp 153-160

[12] Petculescu G Krishnaswamy S and Achenback J D (2008) ldquoGroup delay measure-ments using modally selective Lamb wave transducers for detection and sizing of delamina-tions in compositesrdquo Smart Materials and Structures 17

[13] Sohn H Park H W Law K H and Farrar C R (2007) ldquoDamage detection in compo-

- 33 -

site plates by using an enhanced time reversal methodrdquo Journal of Aerospace Engineering (ASCE) pp 141-151

[14] Sohn H (2007) ldquoEffects of environmental and operational variability on structural health monitoringrdquo Philosophical Transactions of the Royal Society Ardquo 365 pp539-560

[15] Rytter A (1993) ldquoVibration based inspection of civil engineering structuresrdquo PhD The-sis Department of Building Technology and Structural Engineering University of Aalborg Aalborg Denmark

[16] Hayashi T and Kawashima K (2002) ldquoMultiple reflections of Lamb waves at a delami-nationrdquo Ultrasonics 40 pp193-197

[17] Wang C H and Rose L R F (2003) ldquoWave reflection and transmission in beams con-taining delamination and inhomogeneityrdquo Journal of Sound and Vibration 264 pp 851-872

[18] Ramadas C Balasubramaniam K Joshi M and Krishnamurthy C V (2009) ldquoInterac-tion of the primary anti-symmetric Lamb mode (Ao) with symmetric delaminations numer-ical and experimental studiesrdquo Smart Materials and Structures 18

[19] Ramadas C Balasubramaniam K Joshi M and Krishnamurthy C V (2010) ldquoInterac-tion of guided Lamb waves with an asymmetrically located delamination in a laminated composite platerdquo Smart Materials and Structures 19

[20] Wilcox P Lowe M and Cawely P (2001) ldquoThe effect of dispersion on long-range in-spection using ultrasonic guided wavesrdquo NDTampE International 34 pp1-9

[21] Mallat S G and Zhang Z (1993) ldquoMatching pursuits with time-frequency dictionariesrdquo IEEE Tansactions on Signal Processing 41(12) pp3397-3415

[22] Zhang G Zhang S and Wang Y (2000) ldquoApplication of adaptive time-frequency de-composition in ultrasonic NDE of highly scattering materialsrdquo Ultrasonics 38 pp 961-964

[23] Hong J C Sun K H and Kim Y Y (2005) ldquoThe matching pursuit approach based on the modulated Gaussian pulse for efficient guided-wave damage inspectionrdquo Smart Mate-rials and Structures 14 pp548-560

[24] Herakovich C T (1997) Mechanics of Fibrous Composites Wiley 480 pages

[25] Reddy J N (1985) ldquoA review of the literature on finite-element modeling of laminated composite platesrdquo The Shock and Vibration Digest 17(4) pp 3-8

[26] Reddy J N (2003) Mechanics of Laminated Composite Plates and Shells Theory and Analysis CRC Press 856 pages

[27] Palacz M Krawczuk M and Ostachowicz W (2005) ldquoThe spectral finite element mod-el for analysis of flexure-shear coupled wave propagation Part 1 laminated multilayer

- 34 -

composite beamrdquo Composite Structures 68 pp37-44

[28] Zhang Y C and Harding J (1990) ldquoA numerical micromechanics analysis of the me-chanical properties of a plain weave compositerdquo Computer amp Structures 36(5) pp 839-844

[29] Nemat-Nasser S and Hori M (1999) Micromechanics Overall Properties of Heteroge-neous Materials Elsevier 810 pages

[30] Scida D Aboura Z Benzeggagh M L and Bocherens E (1999) ldquoA micromechanics model for 3D elasticity and failure of woven-fibre composite materialrdquo Composites Science and Technology 59 pp 505-517

[31] Thomas Y and Wu X H (1997) ldquoA multiscale finite element method for elliptic prob-lems in composite materials and porous mediardquo Journal of Computational Physics 134 pp 169-189

[32] Lee K H Moorthy S and Ghosh S (1999) ldquoMultiple scale computational model for damage in composite materialsrdquo Computer Methods in Applied Mechanics and Engineer-ing 172 pp 175-201

[33] Namilae S and Chandra N (2005) ldquoMultiscale model to study the effect of interfaces in carbon nanotube-based compositesrdquo Journal of Engineering Materials and Technology (ASME) 127 pp222-232

[34] Adams D E (2007) Health monitoring of Structural Materials and Components Wiley 460 pages

[35] Gopalakrishnan S (2009) ldquoModeling Aspects in Finite Elementrdquo Encyclopedia of Struc-tural Health Monitoring 2 pp791-809

[36] Palacz M Krawczuk M and Ostachowicz W (2005) ldquoThe spectral finite element mod-el for analysis of flexure-shear coupled wave propagation Part 2 Deminated multilayer composite beamrdquo Composite Structures 68 pp45-51

[37] Baik J M and Thopson R B (1984) ldquoUltrasonic scattering from imperfect interfaces A quasi-static modelrdquo Journal of Nondestructive Evaluation 4(3) pp177-196

[38] Luo H and Hanagud S (2000) ldquoDynamics of delamincated beamsrdquo International Jour-nal of Solids and Structures 37 pp1501-1519

[39] Kim S B and Sohn H (2007) ldquoInstantaneous reference-free crack detection based on polarization characteristics of piezoelectric materialsrdquo Smart Materials and Structures 16 pp2375-2387

[40] An Y K and Sohn H (2010) ldquoInstantaneous crack detection under varying temperature and static loading conditionsrdquo The Special Issue of Structural Control and Health Monitor-ing 17(7) pp 730-741

- 35 -

[41] Lee S J and Sohn H (2006) ldquoActive self-sensing scheme development for structural health monitoringrdquo Smart Materials and Structures 15(6) pp 1734-1746

- 36 -

요 약 문

복합재료 손상의 무기저 진단 및 정량화

복합재료는 재료의 경량성 내구성 고강도의 특성으로 인하여 항공기에서부터

교량에까지 널리 사용되고 있다 하지만 외부의 충격으로 인한 복합재료 내부의

손상은 재료의 강도를 감소시켜 구조물의 안정성을 저하시키는 요인이 된다 하지만

복합재료의 손상은 재료 내부에서 발생하여 육안으로는 식별하기도 힘들며 표면에서

보이는 것만으로 내부의 손상 정도를 판단하기는 불가능하다는 단점을 가지고 있다

본 연구에서는 손상 내부에서 반사로 인해 발생하는 파를 이용하여 기저자료를

사용하지 않고 특정 시점에서 얻은 자료만으로 복합재료의 손상을 진단하고 정량화

할 수 있는 기법을 제안한다 제안된 기법의 적용성을 검증하기 위하여 유한요소법을

이용한 수치해석을 수행하였고 Teflon tape 이 삽입된 시편과 실제 충격하중으로

인한 손상을 입은 시편을 이용한 실험을 진행하였다 수치해석 및 실험 결과 제안된

기법이 복합재료의 손상을 성공적으로 진단 및 정량화할 수 있음을 확인하였다 또한

다양한 외부 온도 환경에서 제안된 기법의 환경영향 실험을 수행하였고 다양한

환경에서 손상의 진단 및 정량화가 성공적으로 이루어짐을 검증하였다

핵심어 무기저 손상진단 및 정량화 압전센서 복합재료 유한요소법 Matching pursuit

method

- 37 -

감 사 의 글

기대와 두려움을 안고 시작한 석사과정이 어느덧 2 년 이라는 세월 속에서 추억으로 자리잡아

가고 있습니다 학부생 기간을 포함한 지난 3 년 동안 SHM 이라는 새로운 분야를 접하고 연구를

수행 할 수 있도록 도와주신 고마운 분들이 있어 그 감사를 표하고자 합니다

우선 저의 많은 부족한 점을 일깨워 주시고 연구에 대한 뜻을 더욱 확고히 세울 수 있게 해

주신 지도교수님이신 손 훈 교수님께 깊은 감사의 뜻을 전합니다 짧은 기간이었지만 교수님의

도움으로 많은 것을 배우고 느끼고 깨달을 수 있었습니다 또한 본 연구를 수행할 수 있도록 지

원해주신 국토해양부 (U-City 석 박사과정 지원사업)에 깊은 감사 드립니다

때로 힘들고 지칠 때 항상 저를 믿고 응원해 주시는 가족이 있었기에 지금의 글을 쓸 수 있었

습니다 또한 지금 연구실에서 함께 생활하고 있는 윤규형 민구형 철민이형 현석이형 진열이형

호민이형 병진이 선혜누나와 성대에 계신 창길이형 미국에 계신 의재형과 Debaditiya 에게도 감

사의 말씀을 전합니다 그리고 모두 나열할 수는 없지만 카이스트 생활의 활력소가 되어준 카이

스트 건설 및 환경공학과 선후배들 직원 분들과 SOS 멤버들에게도 감사를 표합니다

앞서 거론한 많은 이들의 도움이 없었다면 저 또한 지금 이 자리에 있을 수 없었을 것입니다

많이 부족하기에 더 노력하고 배우고자 했던 초심을 다시 한번 생각하며 더 큰 결실을 맺기 위해

앞으로 나아가겠습니다 도움을 주신 모든 분들께 이 감사의 글을 바칩니다

2011년 6월 카이스트에서 임형진 올림

- 38 -

CURRICULUM VITAE

Hyung Jin Lim

Research Assistant

The Civil and Environmental Engineering Department

Korea Advanced Institute of Science and Technology (KAIST)

291 Daehak-Ro Guseong-Dong Yuseong-Gu

Daejeon Republic of Korea 305-701

Phone (82)+42-350-3665 Fax (82)+42-350-3610

Email limhj87gmailcom

RESEARCH INTERESTS

Damage Detection Localization and Quantification on Composite and Metallic Structures Pattern Recognition Finite Element Modeling (FEM)

EDUCATION

2009 ndash 2011 MS Civil and Environmental Eng Korea Advanced Institute of Science and Technology (KAIST) Korea Rep

2006 ndash 2009 BS Civil and Environmental Eng Korea Advanced Institute of Science and Technology (KAIST) Korea Rep

JOURNAL PUBLICATIONS The corresponding author is underlined 1 Hyung Jin Lim Hoon Sohn ldquoReference-free damage detection localization and quantification

using in compositerdquo in preparation for Journal of the Acoustical Society of America 2011

2 Chul Min Yeum Hoon Sohn Jeong Bum Ihn Hyung Jin Lim ldquoInstantaneous delamination detection in a composite using a dual piezoelectric transducer networkrdquo Submitted to Compo-sites Science and Technology 2011

3 Yun-Kyu An Hyung Jin Lim Min Koo Kim Jin Yeol Yang Hoon Sohn Chang Guen Lee ldquoInvestigation of applicability of local reference-free damage detection techniques to in-situ

- 39 -

bridgesrdquo Submitted to ASCE Journal of Structural Engineering 2011

4 Hyung Jin Lim Min Koo Kim Hoon Sohn Chan Yik Park ldquoImpedance based damage detec-tion under varying temperature and loading conditionsrdquo Submitted to NDTampE International 2011

5 Hyung Jin Lim Hoon Sohn ldquoQuantitative estimation of transmitted and reflected lamb waves at discontinuityrdquo Journal of the Korean Society for Nondestructive Testing Vol30 No 4 pp1-8 2010

CONFERENCE PROCEEDINGS

The corresponding author is underlined 1 Hyung Jin Lim and Hoon Sohn ldquoReference-free damage detection localization and quantifi-

cation in compositesrdquo ASCE Engineering Mechanics Institute Conference Boston June 2-4 2011

2 Hyung Jin Lim and Hoon Sohn ldquoReference-free delamination quantification within multi-layer composite materialsrdquo KSNT Spring Conference May 26-27 Gyeong-Ju Korea 2011

3 Min Koo Kim Hyun Jin Lim Hoon Sohn Chan Yik Park ldquoImpedance-based bolt loosening detection under varying temperature and loadingrdquo Asian Pacific Workshop on Structural Health Monitoring Tokyo Japan November 30-December 2 2010

4 Hyung Jin Lim Chul Min Yeum and Hoon Sohn ldquoModeling of Impact-induced Delamination in a Multilayer Composite Platerdquo The 23rd KKCNN Symposium on Civil Engineering Taipei (Taiwan) November 13-15 2010

5 Yun-Kyu An Hyung Jin Lim and Hoon Sohn ldquoTemperature independent Delamination Detec-tion using data normalizationrdquo The 3rd International Conference on Multi-Functional Materi-als and Structures Chonbuk National Univ Jeonju Korea September 14-18 2010

6 Hoon Sohn Yun-Kyu An Chang Gil Lee Min Koo Kim Hyung Jin Lim ldquoApplication of online NDT techniques to decommissioned bridge testingrdquo the 5th European Workshop on Structural Health Monitoring Sorrento Italy June 29-July 02 2010

7 Min Koo Kim Hyung Jin Lim Hoon Sohn Chan Yik Park ldquoImpedance based damage diag-nosis of a complex composite aircraft wing under changing loading and temperature condi-tionsrdquo the 5th European Workshop on Structural Health Monitoring Sorrento Italy June 29-July 02 2010

8 Yun-Kyu An Min Koo Kim Hyung Jin Lim and Hoon Sohn ldquoValidations of reference-free crack detection techniques through a decommissioned bridge testrdquo 2010 COSEIK Annual Con-ference Jeju Korea April 8-9 2010

- 40 -

PARTICIPANT PROJECTS

1 Development of Smart Steel Members for Online Steel Structure Monitoring POSCO Corporation (Funded 41000000 KRW (41000 USD) for 010511 to 123111)

2 Development of Enabling Technologies for Structural Health Monitoring Korean Agency for Defense Development (Funded 240000000 KRW (240000 USD) out of total 3200000000 KRW (3200000 USD) for 060108-053111)

3 Development of On-board SHM Technologies for Composite Air Vehicles The Boeing Company (Funded 400000 USD (400000000 KRW) for 080108 to 073111)

EDUCATIONAL EXPERIENCES

Fall 2009

Teaching Assistant CE207 ldquoElementary Structural Engineering and Laboratoryrdquo KAIST 3 unit undergraduate course

EXPERIMENTAL EXPERIENTCES

2010 Monitoring of Real Aircraft Composite Wing Segment Korea Agency for Defense De-velopment (ADD) Daejeon Korea Real time monitoring of composite air vehicles fund-ed by Korea Agency for Defense Development (ADD)

2009-2011

Field Monitoring of Real Bridge Structures Gimpo (Ramp G bridge) amp Icheon (Sam-seung bridge) Korea US-Korea collaborative research for bridge monitoring testbeds

HONORS amp AWARDS

Scholarships for outstanding students granted KAIST (2007-2009) The three prominent students in the civil and environmental engineering department of KAIST were selected as recipients This scholarship provided 2000$ per year for three years

- 41 -

Page 45: 복합재료 손상의 무기저 진단 및 정량화 - KAISTssslab.kaist.ac.kr/article/pdf/Thesis_2.pdf · 1 Declaration of Ethical Conduct in Research: I, as a graduate student

site plates by using an enhanced time reversal methodrdquo Journal of Aerospace Engineering (ASCE) pp 141-151

[14] Sohn H (2007) ldquoEffects of environmental and operational variability on structural health monitoringrdquo Philosophical Transactions of the Royal Society Ardquo 365 pp539-560

[15] Rytter A (1993) ldquoVibration based inspection of civil engineering structuresrdquo PhD The-sis Department of Building Technology and Structural Engineering University of Aalborg Aalborg Denmark

[16] Hayashi T and Kawashima K (2002) ldquoMultiple reflections of Lamb waves at a delami-nationrdquo Ultrasonics 40 pp193-197

[17] Wang C H and Rose L R F (2003) ldquoWave reflection and transmission in beams con-taining delamination and inhomogeneityrdquo Journal of Sound and Vibration 264 pp 851-872

[18] Ramadas C Balasubramaniam K Joshi M and Krishnamurthy C V (2009) ldquoInterac-tion of the primary anti-symmetric Lamb mode (Ao) with symmetric delaminations numer-ical and experimental studiesrdquo Smart Materials and Structures 18

[19] Ramadas C Balasubramaniam K Joshi M and Krishnamurthy C V (2010) ldquoInterac-tion of guided Lamb waves with an asymmetrically located delamination in a laminated composite platerdquo Smart Materials and Structures 19

[20] Wilcox P Lowe M and Cawely P (2001) ldquoThe effect of dispersion on long-range in-spection using ultrasonic guided wavesrdquo NDTampE International 34 pp1-9

[21] Mallat S G and Zhang Z (1993) ldquoMatching pursuits with time-frequency dictionariesrdquo IEEE Tansactions on Signal Processing 41(12) pp3397-3415

[22] Zhang G Zhang S and Wang Y (2000) ldquoApplication of adaptive time-frequency de-composition in ultrasonic NDE of highly scattering materialsrdquo Ultrasonics 38 pp 961-964

[23] Hong J C Sun K H and Kim Y Y (2005) ldquoThe matching pursuit approach based on the modulated Gaussian pulse for efficient guided-wave damage inspectionrdquo Smart Mate-rials and Structures 14 pp548-560

[24] Herakovich C T (1997) Mechanics of Fibrous Composites Wiley 480 pages

[25] Reddy J N (1985) ldquoA review of the literature on finite-element modeling of laminated composite platesrdquo The Shock and Vibration Digest 17(4) pp 3-8

[26] Reddy J N (2003) Mechanics of Laminated Composite Plates and Shells Theory and Analysis CRC Press 856 pages

[27] Palacz M Krawczuk M and Ostachowicz W (2005) ldquoThe spectral finite element mod-el for analysis of flexure-shear coupled wave propagation Part 1 laminated multilayer

- 34 -

composite beamrdquo Composite Structures 68 pp37-44

[28] Zhang Y C and Harding J (1990) ldquoA numerical micromechanics analysis of the me-chanical properties of a plain weave compositerdquo Computer amp Structures 36(5) pp 839-844

[29] Nemat-Nasser S and Hori M (1999) Micromechanics Overall Properties of Heteroge-neous Materials Elsevier 810 pages

[30] Scida D Aboura Z Benzeggagh M L and Bocherens E (1999) ldquoA micromechanics model for 3D elasticity and failure of woven-fibre composite materialrdquo Composites Science and Technology 59 pp 505-517

[31] Thomas Y and Wu X H (1997) ldquoA multiscale finite element method for elliptic prob-lems in composite materials and porous mediardquo Journal of Computational Physics 134 pp 169-189

[32] Lee K H Moorthy S and Ghosh S (1999) ldquoMultiple scale computational model for damage in composite materialsrdquo Computer Methods in Applied Mechanics and Engineer-ing 172 pp 175-201

[33] Namilae S and Chandra N (2005) ldquoMultiscale model to study the effect of interfaces in carbon nanotube-based compositesrdquo Journal of Engineering Materials and Technology (ASME) 127 pp222-232

[34] Adams D E (2007) Health monitoring of Structural Materials and Components Wiley 460 pages

[35] Gopalakrishnan S (2009) ldquoModeling Aspects in Finite Elementrdquo Encyclopedia of Struc-tural Health Monitoring 2 pp791-809

[36] Palacz M Krawczuk M and Ostachowicz W (2005) ldquoThe spectral finite element mod-el for analysis of flexure-shear coupled wave propagation Part 2 Deminated multilayer composite beamrdquo Composite Structures 68 pp45-51

[37] Baik J M and Thopson R B (1984) ldquoUltrasonic scattering from imperfect interfaces A quasi-static modelrdquo Journal of Nondestructive Evaluation 4(3) pp177-196

[38] Luo H and Hanagud S (2000) ldquoDynamics of delamincated beamsrdquo International Jour-nal of Solids and Structures 37 pp1501-1519

[39] Kim S B and Sohn H (2007) ldquoInstantaneous reference-free crack detection based on polarization characteristics of piezoelectric materialsrdquo Smart Materials and Structures 16 pp2375-2387

[40] An Y K and Sohn H (2010) ldquoInstantaneous crack detection under varying temperature and static loading conditionsrdquo The Special Issue of Structural Control and Health Monitor-ing 17(7) pp 730-741

- 35 -

[41] Lee S J and Sohn H (2006) ldquoActive self-sensing scheme development for structural health monitoringrdquo Smart Materials and Structures 15(6) pp 1734-1746

- 36 -

요 약 문

복합재료 손상의 무기저 진단 및 정량화

복합재료는 재료의 경량성 내구성 고강도의 특성으로 인하여 항공기에서부터

교량에까지 널리 사용되고 있다 하지만 외부의 충격으로 인한 복합재료 내부의

손상은 재료의 강도를 감소시켜 구조물의 안정성을 저하시키는 요인이 된다 하지만

복합재료의 손상은 재료 내부에서 발생하여 육안으로는 식별하기도 힘들며 표면에서

보이는 것만으로 내부의 손상 정도를 판단하기는 불가능하다는 단점을 가지고 있다

본 연구에서는 손상 내부에서 반사로 인해 발생하는 파를 이용하여 기저자료를

사용하지 않고 특정 시점에서 얻은 자료만으로 복합재료의 손상을 진단하고 정량화

할 수 있는 기법을 제안한다 제안된 기법의 적용성을 검증하기 위하여 유한요소법을

이용한 수치해석을 수행하였고 Teflon tape 이 삽입된 시편과 실제 충격하중으로

인한 손상을 입은 시편을 이용한 실험을 진행하였다 수치해석 및 실험 결과 제안된

기법이 복합재료의 손상을 성공적으로 진단 및 정량화할 수 있음을 확인하였다 또한

다양한 외부 온도 환경에서 제안된 기법의 환경영향 실험을 수행하였고 다양한

환경에서 손상의 진단 및 정량화가 성공적으로 이루어짐을 검증하였다

핵심어 무기저 손상진단 및 정량화 압전센서 복합재료 유한요소법 Matching pursuit

method

- 37 -

감 사 의 글

기대와 두려움을 안고 시작한 석사과정이 어느덧 2 년 이라는 세월 속에서 추억으로 자리잡아

가고 있습니다 학부생 기간을 포함한 지난 3 년 동안 SHM 이라는 새로운 분야를 접하고 연구를

수행 할 수 있도록 도와주신 고마운 분들이 있어 그 감사를 표하고자 합니다

우선 저의 많은 부족한 점을 일깨워 주시고 연구에 대한 뜻을 더욱 확고히 세울 수 있게 해

주신 지도교수님이신 손 훈 교수님께 깊은 감사의 뜻을 전합니다 짧은 기간이었지만 교수님의

도움으로 많은 것을 배우고 느끼고 깨달을 수 있었습니다 또한 본 연구를 수행할 수 있도록 지

원해주신 국토해양부 (U-City 석 박사과정 지원사업)에 깊은 감사 드립니다

때로 힘들고 지칠 때 항상 저를 믿고 응원해 주시는 가족이 있었기에 지금의 글을 쓸 수 있었

습니다 또한 지금 연구실에서 함께 생활하고 있는 윤규형 민구형 철민이형 현석이형 진열이형

호민이형 병진이 선혜누나와 성대에 계신 창길이형 미국에 계신 의재형과 Debaditiya 에게도 감

사의 말씀을 전합니다 그리고 모두 나열할 수는 없지만 카이스트 생활의 활력소가 되어준 카이

스트 건설 및 환경공학과 선후배들 직원 분들과 SOS 멤버들에게도 감사를 표합니다

앞서 거론한 많은 이들의 도움이 없었다면 저 또한 지금 이 자리에 있을 수 없었을 것입니다

많이 부족하기에 더 노력하고 배우고자 했던 초심을 다시 한번 생각하며 더 큰 결실을 맺기 위해

앞으로 나아가겠습니다 도움을 주신 모든 분들께 이 감사의 글을 바칩니다

2011년 6월 카이스트에서 임형진 올림

- 38 -

CURRICULUM VITAE

Hyung Jin Lim

Research Assistant

The Civil and Environmental Engineering Department

Korea Advanced Institute of Science and Technology (KAIST)

291 Daehak-Ro Guseong-Dong Yuseong-Gu

Daejeon Republic of Korea 305-701

Phone (82)+42-350-3665 Fax (82)+42-350-3610

Email limhj87gmailcom

RESEARCH INTERESTS

Damage Detection Localization and Quantification on Composite and Metallic Structures Pattern Recognition Finite Element Modeling (FEM)

EDUCATION

2009 ndash 2011 MS Civil and Environmental Eng Korea Advanced Institute of Science and Technology (KAIST) Korea Rep

2006 ndash 2009 BS Civil and Environmental Eng Korea Advanced Institute of Science and Technology (KAIST) Korea Rep

JOURNAL PUBLICATIONS The corresponding author is underlined 1 Hyung Jin Lim Hoon Sohn ldquoReference-free damage detection localization and quantification

using in compositerdquo in preparation for Journal of the Acoustical Society of America 2011

2 Chul Min Yeum Hoon Sohn Jeong Bum Ihn Hyung Jin Lim ldquoInstantaneous delamination detection in a composite using a dual piezoelectric transducer networkrdquo Submitted to Compo-sites Science and Technology 2011

3 Yun-Kyu An Hyung Jin Lim Min Koo Kim Jin Yeol Yang Hoon Sohn Chang Guen Lee ldquoInvestigation of applicability of local reference-free damage detection techniques to in-situ

- 39 -

bridgesrdquo Submitted to ASCE Journal of Structural Engineering 2011

4 Hyung Jin Lim Min Koo Kim Hoon Sohn Chan Yik Park ldquoImpedance based damage detec-tion under varying temperature and loading conditionsrdquo Submitted to NDTampE International 2011

5 Hyung Jin Lim Hoon Sohn ldquoQuantitative estimation of transmitted and reflected lamb waves at discontinuityrdquo Journal of the Korean Society for Nondestructive Testing Vol30 No 4 pp1-8 2010

CONFERENCE PROCEEDINGS

The corresponding author is underlined 1 Hyung Jin Lim and Hoon Sohn ldquoReference-free damage detection localization and quantifi-

cation in compositesrdquo ASCE Engineering Mechanics Institute Conference Boston June 2-4 2011

2 Hyung Jin Lim and Hoon Sohn ldquoReference-free delamination quantification within multi-layer composite materialsrdquo KSNT Spring Conference May 26-27 Gyeong-Ju Korea 2011

3 Min Koo Kim Hyun Jin Lim Hoon Sohn Chan Yik Park ldquoImpedance-based bolt loosening detection under varying temperature and loadingrdquo Asian Pacific Workshop on Structural Health Monitoring Tokyo Japan November 30-December 2 2010

4 Hyung Jin Lim Chul Min Yeum and Hoon Sohn ldquoModeling of Impact-induced Delamination in a Multilayer Composite Platerdquo The 23rd KKCNN Symposium on Civil Engineering Taipei (Taiwan) November 13-15 2010

5 Yun-Kyu An Hyung Jin Lim and Hoon Sohn ldquoTemperature independent Delamination Detec-tion using data normalizationrdquo The 3rd International Conference on Multi-Functional Materi-als and Structures Chonbuk National Univ Jeonju Korea September 14-18 2010

6 Hoon Sohn Yun-Kyu An Chang Gil Lee Min Koo Kim Hyung Jin Lim ldquoApplication of online NDT techniques to decommissioned bridge testingrdquo the 5th European Workshop on Structural Health Monitoring Sorrento Italy June 29-July 02 2010

7 Min Koo Kim Hyung Jin Lim Hoon Sohn Chan Yik Park ldquoImpedance based damage diag-nosis of a complex composite aircraft wing under changing loading and temperature condi-tionsrdquo the 5th European Workshop on Structural Health Monitoring Sorrento Italy June 29-July 02 2010

8 Yun-Kyu An Min Koo Kim Hyung Jin Lim and Hoon Sohn ldquoValidations of reference-free crack detection techniques through a decommissioned bridge testrdquo 2010 COSEIK Annual Con-ference Jeju Korea April 8-9 2010

- 40 -

PARTICIPANT PROJECTS

1 Development of Smart Steel Members for Online Steel Structure Monitoring POSCO Corporation (Funded 41000000 KRW (41000 USD) for 010511 to 123111)

2 Development of Enabling Technologies for Structural Health Monitoring Korean Agency for Defense Development (Funded 240000000 KRW (240000 USD) out of total 3200000000 KRW (3200000 USD) for 060108-053111)

3 Development of On-board SHM Technologies for Composite Air Vehicles The Boeing Company (Funded 400000 USD (400000000 KRW) for 080108 to 073111)

EDUCATIONAL EXPERIENCES

Fall 2009

Teaching Assistant CE207 ldquoElementary Structural Engineering and Laboratoryrdquo KAIST 3 unit undergraduate course

EXPERIMENTAL EXPERIENTCES

2010 Monitoring of Real Aircraft Composite Wing Segment Korea Agency for Defense De-velopment (ADD) Daejeon Korea Real time monitoring of composite air vehicles fund-ed by Korea Agency for Defense Development (ADD)

2009-2011

Field Monitoring of Real Bridge Structures Gimpo (Ramp G bridge) amp Icheon (Sam-seung bridge) Korea US-Korea collaborative research for bridge monitoring testbeds

HONORS amp AWARDS

Scholarships for outstanding students granted KAIST (2007-2009) The three prominent students in the civil and environmental engineering department of KAIST were selected as recipients This scholarship provided 2000$ per year for three years

- 41 -

Page 46: 복합재료 손상의 무기저 진단 및 정량화 - KAISTssslab.kaist.ac.kr/article/pdf/Thesis_2.pdf · 1 Declaration of Ethical Conduct in Research: I, as a graduate student

composite beamrdquo Composite Structures 68 pp37-44

[28] Zhang Y C and Harding J (1990) ldquoA numerical micromechanics analysis of the me-chanical properties of a plain weave compositerdquo Computer amp Structures 36(5) pp 839-844

[29] Nemat-Nasser S and Hori M (1999) Micromechanics Overall Properties of Heteroge-neous Materials Elsevier 810 pages

[30] Scida D Aboura Z Benzeggagh M L and Bocherens E (1999) ldquoA micromechanics model for 3D elasticity and failure of woven-fibre composite materialrdquo Composites Science and Technology 59 pp 505-517

[31] Thomas Y and Wu X H (1997) ldquoA multiscale finite element method for elliptic prob-lems in composite materials and porous mediardquo Journal of Computational Physics 134 pp 169-189

[32] Lee K H Moorthy S and Ghosh S (1999) ldquoMultiple scale computational model for damage in composite materialsrdquo Computer Methods in Applied Mechanics and Engineer-ing 172 pp 175-201

[33] Namilae S and Chandra N (2005) ldquoMultiscale model to study the effect of interfaces in carbon nanotube-based compositesrdquo Journal of Engineering Materials and Technology (ASME) 127 pp222-232

[34] Adams D E (2007) Health monitoring of Structural Materials and Components Wiley 460 pages

[35] Gopalakrishnan S (2009) ldquoModeling Aspects in Finite Elementrdquo Encyclopedia of Struc-tural Health Monitoring 2 pp791-809

[36] Palacz M Krawczuk M and Ostachowicz W (2005) ldquoThe spectral finite element mod-el for analysis of flexure-shear coupled wave propagation Part 2 Deminated multilayer composite beamrdquo Composite Structures 68 pp45-51

[37] Baik J M and Thopson R B (1984) ldquoUltrasonic scattering from imperfect interfaces A quasi-static modelrdquo Journal of Nondestructive Evaluation 4(3) pp177-196

[38] Luo H and Hanagud S (2000) ldquoDynamics of delamincated beamsrdquo International Jour-nal of Solids and Structures 37 pp1501-1519

[39] Kim S B and Sohn H (2007) ldquoInstantaneous reference-free crack detection based on polarization characteristics of piezoelectric materialsrdquo Smart Materials and Structures 16 pp2375-2387

[40] An Y K and Sohn H (2010) ldquoInstantaneous crack detection under varying temperature and static loading conditionsrdquo The Special Issue of Structural Control and Health Monitor-ing 17(7) pp 730-741

- 35 -

[41] Lee S J and Sohn H (2006) ldquoActive self-sensing scheme development for structural health monitoringrdquo Smart Materials and Structures 15(6) pp 1734-1746

- 36 -

요 약 문

복합재료 손상의 무기저 진단 및 정량화

복합재료는 재료의 경량성 내구성 고강도의 특성으로 인하여 항공기에서부터

교량에까지 널리 사용되고 있다 하지만 외부의 충격으로 인한 복합재료 내부의

손상은 재료의 강도를 감소시켜 구조물의 안정성을 저하시키는 요인이 된다 하지만

복합재료의 손상은 재료 내부에서 발생하여 육안으로는 식별하기도 힘들며 표면에서

보이는 것만으로 내부의 손상 정도를 판단하기는 불가능하다는 단점을 가지고 있다

본 연구에서는 손상 내부에서 반사로 인해 발생하는 파를 이용하여 기저자료를

사용하지 않고 특정 시점에서 얻은 자료만으로 복합재료의 손상을 진단하고 정량화

할 수 있는 기법을 제안한다 제안된 기법의 적용성을 검증하기 위하여 유한요소법을

이용한 수치해석을 수행하였고 Teflon tape 이 삽입된 시편과 실제 충격하중으로

인한 손상을 입은 시편을 이용한 실험을 진행하였다 수치해석 및 실험 결과 제안된

기법이 복합재료의 손상을 성공적으로 진단 및 정량화할 수 있음을 확인하였다 또한

다양한 외부 온도 환경에서 제안된 기법의 환경영향 실험을 수행하였고 다양한

환경에서 손상의 진단 및 정량화가 성공적으로 이루어짐을 검증하였다

핵심어 무기저 손상진단 및 정량화 압전센서 복합재료 유한요소법 Matching pursuit

method

- 37 -

감 사 의 글

기대와 두려움을 안고 시작한 석사과정이 어느덧 2 년 이라는 세월 속에서 추억으로 자리잡아

가고 있습니다 학부생 기간을 포함한 지난 3 년 동안 SHM 이라는 새로운 분야를 접하고 연구를

수행 할 수 있도록 도와주신 고마운 분들이 있어 그 감사를 표하고자 합니다

우선 저의 많은 부족한 점을 일깨워 주시고 연구에 대한 뜻을 더욱 확고히 세울 수 있게 해

주신 지도교수님이신 손 훈 교수님께 깊은 감사의 뜻을 전합니다 짧은 기간이었지만 교수님의

도움으로 많은 것을 배우고 느끼고 깨달을 수 있었습니다 또한 본 연구를 수행할 수 있도록 지

원해주신 국토해양부 (U-City 석 박사과정 지원사업)에 깊은 감사 드립니다

때로 힘들고 지칠 때 항상 저를 믿고 응원해 주시는 가족이 있었기에 지금의 글을 쓸 수 있었

습니다 또한 지금 연구실에서 함께 생활하고 있는 윤규형 민구형 철민이형 현석이형 진열이형

호민이형 병진이 선혜누나와 성대에 계신 창길이형 미국에 계신 의재형과 Debaditiya 에게도 감

사의 말씀을 전합니다 그리고 모두 나열할 수는 없지만 카이스트 생활의 활력소가 되어준 카이

스트 건설 및 환경공학과 선후배들 직원 분들과 SOS 멤버들에게도 감사를 표합니다

앞서 거론한 많은 이들의 도움이 없었다면 저 또한 지금 이 자리에 있을 수 없었을 것입니다

많이 부족하기에 더 노력하고 배우고자 했던 초심을 다시 한번 생각하며 더 큰 결실을 맺기 위해

앞으로 나아가겠습니다 도움을 주신 모든 분들께 이 감사의 글을 바칩니다

2011년 6월 카이스트에서 임형진 올림

- 38 -

CURRICULUM VITAE

Hyung Jin Lim

Research Assistant

The Civil and Environmental Engineering Department

Korea Advanced Institute of Science and Technology (KAIST)

291 Daehak-Ro Guseong-Dong Yuseong-Gu

Daejeon Republic of Korea 305-701

Phone (82)+42-350-3665 Fax (82)+42-350-3610

Email limhj87gmailcom

RESEARCH INTERESTS

Damage Detection Localization and Quantification on Composite and Metallic Structures Pattern Recognition Finite Element Modeling (FEM)

EDUCATION

2009 ndash 2011 MS Civil and Environmental Eng Korea Advanced Institute of Science and Technology (KAIST) Korea Rep

2006 ndash 2009 BS Civil and Environmental Eng Korea Advanced Institute of Science and Technology (KAIST) Korea Rep

JOURNAL PUBLICATIONS The corresponding author is underlined 1 Hyung Jin Lim Hoon Sohn ldquoReference-free damage detection localization and quantification

using in compositerdquo in preparation for Journal of the Acoustical Society of America 2011

2 Chul Min Yeum Hoon Sohn Jeong Bum Ihn Hyung Jin Lim ldquoInstantaneous delamination detection in a composite using a dual piezoelectric transducer networkrdquo Submitted to Compo-sites Science and Technology 2011

3 Yun-Kyu An Hyung Jin Lim Min Koo Kim Jin Yeol Yang Hoon Sohn Chang Guen Lee ldquoInvestigation of applicability of local reference-free damage detection techniques to in-situ

- 39 -

bridgesrdquo Submitted to ASCE Journal of Structural Engineering 2011

4 Hyung Jin Lim Min Koo Kim Hoon Sohn Chan Yik Park ldquoImpedance based damage detec-tion under varying temperature and loading conditionsrdquo Submitted to NDTampE International 2011

5 Hyung Jin Lim Hoon Sohn ldquoQuantitative estimation of transmitted and reflected lamb waves at discontinuityrdquo Journal of the Korean Society for Nondestructive Testing Vol30 No 4 pp1-8 2010

CONFERENCE PROCEEDINGS

The corresponding author is underlined 1 Hyung Jin Lim and Hoon Sohn ldquoReference-free damage detection localization and quantifi-

cation in compositesrdquo ASCE Engineering Mechanics Institute Conference Boston June 2-4 2011

2 Hyung Jin Lim and Hoon Sohn ldquoReference-free delamination quantification within multi-layer composite materialsrdquo KSNT Spring Conference May 26-27 Gyeong-Ju Korea 2011

3 Min Koo Kim Hyun Jin Lim Hoon Sohn Chan Yik Park ldquoImpedance-based bolt loosening detection under varying temperature and loadingrdquo Asian Pacific Workshop on Structural Health Monitoring Tokyo Japan November 30-December 2 2010

4 Hyung Jin Lim Chul Min Yeum and Hoon Sohn ldquoModeling of Impact-induced Delamination in a Multilayer Composite Platerdquo The 23rd KKCNN Symposium on Civil Engineering Taipei (Taiwan) November 13-15 2010

5 Yun-Kyu An Hyung Jin Lim and Hoon Sohn ldquoTemperature independent Delamination Detec-tion using data normalizationrdquo The 3rd International Conference on Multi-Functional Materi-als and Structures Chonbuk National Univ Jeonju Korea September 14-18 2010

6 Hoon Sohn Yun-Kyu An Chang Gil Lee Min Koo Kim Hyung Jin Lim ldquoApplication of online NDT techniques to decommissioned bridge testingrdquo the 5th European Workshop on Structural Health Monitoring Sorrento Italy June 29-July 02 2010

7 Min Koo Kim Hyung Jin Lim Hoon Sohn Chan Yik Park ldquoImpedance based damage diag-nosis of a complex composite aircraft wing under changing loading and temperature condi-tionsrdquo the 5th European Workshop on Structural Health Monitoring Sorrento Italy June 29-July 02 2010

8 Yun-Kyu An Min Koo Kim Hyung Jin Lim and Hoon Sohn ldquoValidations of reference-free crack detection techniques through a decommissioned bridge testrdquo 2010 COSEIK Annual Con-ference Jeju Korea April 8-9 2010

- 40 -

PARTICIPANT PROJECTS

1 Development of Smart Steel Members for Online Steel Structure Monitoring POSCO Corporation (Funded 41000000 KRW (41000 USD) for 010511 to 123111)

2 Development of Enabling Technologies for Structural Health Monitoring Korean Agency for Defense Development (Funded 240000000 KRW (240000 USD) out of total 3200000000 KRW (3200000 USD) for 060108-053111)

3 Development of On-board SHM Technologies for Composite Air Vehicles The Boeing Company (Funded 400000 USD (400000000 KRW) for 080108 to 073111)

EDUCATIONAL EXPERIENCES

Fall 2009

Teaching Assistant CE207 ldquoElementary Structural Engineering and Laboratoryrdquo KAIST 3 unit undergraduate course

EXPERIMENTAL EXPERIENTCES

2010 Monitoring of Real Aircraft Composite Wing Segment Korea Agency for Defense De-velopment (ADD) Daejeon Korea Real time monitoring of composite air vehicles fund-ed by Korea Agency for Defense Development (ADD)

2009-2011

Field Monitoring of Real Bridge Structures Gimpo (Ramp G bridge) amp Icheon (Sam-seung bridge) Korea US-Korea collaborative research for bridge monitoring testbeds

HONORS amp AWARDS

Scholarships for outstanding students granted KAIST (2007-2009) The three prominent students in the civil and environmental engineering department of KAIST were selected as recipients This scholarship provided 2000$ per year for three years

- 41 -

Page 47: 복합재료 손상의 무기저 진단 및 정량화 - KAISTssslab.kaist.ac.kr/article/pdf/Thesis_2.pdf · 1 Declaration of Ethical Conduct in Research: I, as a graduate student

[41] Lee S J and Sohn H (2006) ldquoActive self-sensing scheme development for structural health monitoringrdquo Smart Materials and Structures 15(6) pp 1734-1746

- 36 -

요 약 문

복합재료 손상의 무기저 진단 및 정량화

복합재료는 재료의 경량성 내구성 고강도의 특성으로 인하여 항공기에서부터

교량에까지 널리 사용되고 있다 하지만 외부의 충격으로 인한 복합재료 내부의

손상은 재료의 강도를 감소시켜 구조물의 안정성을 저하시키는 요인이 된다 하지만

복합재료의 손상은 재료 내부에서 발생하여 육안으로는 식별하기도 힘들며 표면에서

보이는 것만으로 내부의 손상 정도를 판단하기는 불가능하다는 단점을 가지고 있다

본 연구에서는 손상 내부에서 반사로 인해 발생하는 파를 이용하여 기저자료를

사용하지 않고 특정 시점에서 얻은 자료만으로 복합재료의 손상을 진단하고 정량화

할 수 있는 기법을 제안한다 제안된 기법의 적용성을 검증하기 위하여 유한요소법을

이용한 수치해석을 수행하였고 Teflon tape 이 삽입된 시편과 실제 충격하중으로

인한 손상을 입은 시편을 이용한 실험을 진행하였다 수치해석 및 실험 결과 제안된

기법이 복합재료의 손상을 성공적으로 진단 및 정량화할 수 있음을 확인하였다 또한

다양한 외부 온도 환경에서 제안된 기법의 환경영향 실험을 수행하였고 다양한

환경에서 손상의 진단 및 정량화가 성공적으로 이루어짐을 검증하였다

핵심어 무기저 손상진단 및 정량화 압전센서 복합재료 유한요소법 Matching pursuit

method

- 37 -

감 사 의 글

기대와 두려움을 안고 시작한 석사과정이 어느덧 2 년 이라는 세월 속에서 추억으로 자리잡아

가고 있습니다 학부생 기간을 포함한 지난 3 년 동안 SHM 이라는 새로운 분야를 접하고 연구를

수행 할 수 있도록 도와주신 고마운 분들이 있어 그 감사를 표하고자 합니다

우선 저의 많은 부족한 점을 일깨워 주시고 연구에 대한 뜻을 더욱 확고히 세울 수 있게 해

주신 지도교수님이신 손 훈 교수님께 깊은 감사의 뜻을 전합니다 짧은 기간이었지만 교수님의

도움으로 많은 것을 배우고 느끼고 깨달을 수 있었습니다 또한 본 연구를 수행할 수 있도록 지

원해주신 국토해양부 (U-City 석 박사과정 지원사업)에 깊은 감사 드립니다

때로 힘들고 지칠 때 항상 저를 믿고 응원해 주시는 가족이 있었기에 지금의 글을 쓸 수 있었

습니다 또한 지금 연구실에서 함께 생활하고 있는 윤규형 민구형 철민이형 현석이형 진열이형

호민이형 병진이 선혜누나와 성대에 계신 창길이형 미국에 계신 의재형과 Debaditiya 에게도 감

사의 말씀을 전합니다 그리고 모두 나열할 수는 없지만 카이스트 생활의 활력소가 되어준 카이

스트 건설 및 환경공학과 선후배들 직원 분들과 SOS 멤버들에게도 감사를 표합니다

앞서 거론한 많은 이들의 도움이 없었다면 저 또한 지금 이 자리에 있을 수 없었을 것입니다

많이 부족하기에 더 노력하고 배우고자 했던 초심을 다시 한번 생각하며 더 큰 결실을 맺기 위해

앞으로 나아가겠습니다 도움을 주신 모든 분들께 이 감사의 글을 바칩니다

2011년 6월 카이스트에서 임형진 올림

- 38 -

CURRICULUM VITAE

Hyung Jin Lim

Research Assistant

The Civil and Environmental Engineering Department

Korea Advanced Institute of Science and Technology (KAIST)

291 Daehak-Ro Guseong-Dong Yuseong-Gu

Daejeon Republic of Korea 305-701

Phone (82)+42-350-3665 Fax (82)+42-350-3610

Email limhj87gmailcom

RESEARCH INTERESTS

Damage Detection Localization and Quantification on Composite and Metallic Structures Pattern Recognition Finite Element Modeling (FEM)

EDUCATION

2009 ndash 2011 MS Civil and Environmental Eng Korea Advanced Institute of Science and Technology (KAIST) Korea Rep

2006 ndash 2009 BS Civil and Environmental Eng Korea Advanced Institute of Science and Technology (KAIST) Korea Rep

JOURNAL PUBLICATIONS The corresponding author is underlined 1 Hyung Jin Lim Hoon Sohn ldquoReference-free damage detection localization and quantification

using in compositerdquo in preparation for Journal of the Acoustical Society of America 2011

2 Chul Min Yeum Hoon Sohn Jeong Bum Ihn Hyung Jin Lim ldquoInstantaneous delamination detection in a composite using a dual piezoelectric transducer networkrdquo Submitted to Compo-sites Science and Technology 2011

3 Yun-Kyu An Hyung Jin Lim Min Koo Kim Jin Yeol Yang Hoon Sohn Chang Guen Lee ldquoInvestigation of applicability of local reference-free damage detection techniques to in-situ

- 39 -

bridgesrdquo Submitted to ASCE Journal of Structural Engineering 2011

4 Hyung Jin Lim Min Koo Kim Hoon Sohn Chan Yik Park ldquoImpedance based damage detec-tion under varying temperature and loading conditionsrdquo Submitted to NDTampE International 2011

5 Hyung Jin Lim Hoon Sohn ldquoQuantitative estimation of transmitted and reflected lamb waves at discontinuityrdquo Journal of the Korean Society for Nondestructive Testing Vol30 No 4 pp1-8 2010

CONFERENCE PROCEEDINGS

The corresponding author is underlined 1 Hyung Jin Lim and Hoon Sohn ldquoReference-free damage detection localization and quantifi-

cation in compositesrdquo ASCE Engineering Mechanics Institute Conference Boston June 2-4 2011

2 Hyung Jin Lim and Hoon Sohn ldquoReference-free delamination quantification within multi-layer composite materialsrdquo KSNT Spring Conference May 26-27 Gyeong-Ju Korea 2011

3 Min Koo Kim Hyun Jin Lim Hoon Sohn Chan Yik Park ldquoImpedance-based bolt loosening detection under varying temperature and loadingrdquo Asian Pacific Workshop on Structural Health Monitoring Tokyo Japan November 30-December 2 2010

4 Hyung Jin Lim Chul Min Yeum and Hoon Sohn ldquoModeling of Impact-induced Delamination in a Multilayer Composite Platerdquo The 23rd KKCNN Symposium on Civil Engineering Taipei (Taiwan) November 13-15 2010

5 Yun-Kyu An Hyung Jin Lim and Hoon Sohn ldquoTemperature independent Delamination Detec-tion using data normalizationrdquo The 3rd International Conference on Multi-Functional Materi-als and Structures Chonbuk National Univ Jeonju Korea September 14-18 2010

6 Hoon Sohn Yun-Kyu An Chang Gil Lee Min Koo Kim Hyung Jin Lim ldquoApplication of online NDT techniques to decommissioned bridge testingrdquo the 5th European Workshop on Structural Health Monitoring Sorrento Italy June 29-July 02 2010

7 Min Koo Kim Hyung Jin Lim Hoon Sohn Chan Yik Park ldquoImpedance based damage diag-nosis of a complex composite aircraft wing under changing loading and temperature condi-tionsrdquo the 5th European Workshop on Structural Health Monitoring Sorrento Italy June 29-July 02 2010

8 Yun-Kyu An Min Koo Kim Hyung Jin Lim and Hoon Sohn ldquoValidations of reference-free crack detection techniques through a decommissioned bridge testrdquo 2010 COSEIK Annual Con-ference Jeju Korea April 8-9 2010

- 40 -

PARTICIPANT PROJECTS

1 Development of Smart Steel Members for Online Steel Structure Monitoring POSCO Corporation (Funded 41000000 KRW (41000 USD) for 010511 to 123111)

2 Development of Enabling Technologies for Structural Health Monitoring Korean Agency for Defense Development (Funded 240000000 KRW (240000 USD) out of total 3200000000 KRW (3200000 USD) for 060108-053111)

3 Development of On-board SHM Technologies for Composite Air Vehicles The Boeing Company (Funded 400000 USD (400000000 KRW) for 080108 to 073111)

EDUCATIONAL EXPERIENCES

Fall 2009

Teaching Assistant CE207 ldquoElementary Structural Engineering and Laboratoryrdquo KAIST 3 unit undergraduate course

EXPERIMENTAL EXPERIENTCES

2010 Monitoring of Real Aircraft Composite Wing Segment Korea Agency for Defense De-velopment (ADD) Daejeon Korea Real time monitoring of composite air vehicles fund-ed by Korea Agency for Defense Development (ADD)

2009-2011

Field Monitoring of Real Bridge Structures Gimpo (Ramp G bridge) amp Icheon (Sam-seung bridge) Korea US-Korea collaborative research for bridge monitoring testbeds

HONORS amp AWARDS

Scholarships for outstanding students granted KAIST (2007-2009) The three prominent students in the civil and environmental engineering department of KAIST were selected as recipients This scholarship provided 2000$ per year for three years

- 41 -

Page 48: 복합재료 손상의 무기저 진단 및 정량화 - KAISTssslab.kaist.ac.kr/article/pdf/Thesis_2.pdf · 1 Declaration of Ethical Conduct in Research: I, as a graduate student

요 약 문

복합재료 손상의 무기저 진단 및 정량화

복합재료는 재료의 경량성 내구성 고강도의 특성으로 인하여 항공기에서부터

교량에까지 널리 사용되고 있다 하지만 외부의 충격으로 인한 복합재료 내부의

손상은 재료의 강도를 감소시켜 구조물의 안정성을 저하시키는 요인이 된다 하지만

복합재료의 손상은 재료 내부에서 발생하여 육안으로는 식별하기도 힘들며 표면에서

보이는 것만으로 내부의 손상 정도를 판단하기는 불가능하다는 단점을 가지고 있다

본 연구에서는 손상 내부에서 반사로 인해 발생하는 파를 이용하여 기저자료를

사용하지 않고 특정 시점에서 얻은 자료만으로 복합재료의 손상을 진단하고 정량화

할 수 있는 기법을 제안한다 제안된 기법의 적용성을 검증하기 위하여 유한요소법을

이용한 수치해석을 수행하였고 Teflon tape 이 삽입된 시편과 실제 충격하중으로

인한 손상을 입은 시편을 이용한 실험을 진행하였다 수치해석 및 실험 결과 제안된

기법이 복합재료의 손상을 성공적으로 진단 및 정량화할 수 있음을 확인하였다 또한

다양한 외부 온도 환경에서 제안된 기법의 환경영향 실험을 수행하였고 다양한

환경에서 손상의 진단 및 정량화가 성공적으로 이루어짐을 검증하였다

핵심어 무기저 손상진단 및 정량화 압전센서 복합재료 유한요소법 Matching pursuit

method

- 37 -

감 사 의 글

기대와 두려움을 안고 시작한 석사과정이 어느덧 2 년 이라는 세월 속에서 추억으로 자리잡아

가고 있습니다 학부생 기간을 포함한 지난 3 년 동안 SHM 이라는 새로운 분야를 접하고 연구를

수행 할 수 있도록 도와주신 고마운 분들이 있어 그 감사를 표하고자 합니다

우선 저의 많은 부족한 점을 일깨워 주시고 연구에 대한 뜻을 더욱 확고히 세울 수 있게 해

주신 지도교수님이신 손 훈 교수님께 깊은 감사의 뜻을 전합니다 짧은 기간이었지만 교수님의

도움으로 많은 것을 배우고 느끼고 깨달을 수 있었습니다 또한 본 연구를 수행할 수 있도록 지

원해주신 국토해양부 (U-City 석 박사과정 지원사업)에 깊은 감사 드립니다

때로 힘들고 지칠 때 항상 저를 믿고 응원해 주시는 가족이 있었기에 지금의 글을 쓸 수 있었

습니다 또한 지금 연구실에서 함께 생활하고 있는 윤규형 민구형 철민이형 현석이형 진열이형

호민이형 병진이 선혜누나와 성대에 계신 창길이형 미국에 계신 의재형과 Debaditiya 에게도 감

사의 말씀을 전합니다 그리고 모두 나열할 수는 없지만 카이스트 생활의 활력소가 되어준 카이

스트 건설 및 환경공학과 선후배들 직원 분들과 SOS 멤버들에게도 감사를 표합니다

앞서 거론한 많은 이들의 도움이 없었다면 저 또한 지금 이 자리에 있을 수 없었을 것입니다

많이 부족하기에 더 노력하고 배우고자 했던 초심을 다시 한번 생각하며 더 큰 결실을 맺기 위해

앞으로 나아가겠습니다 도움을 주신 모든 분들께 이 감사의 글을 바칩니다

2011년 6월 카이스트에서 임형진 올림

- 38 -

CURRICULUM VITAE

Hyung Jin Lim

Research Assistant

The Civil and Environmental Engineering Department

Korea Advanced Institute of Science and Technology (KAIST)

291 Daehak-Ro Guseong-Dong Yuseong-Gu

Daejeon Republic of Korea 305-701

Phone (82)+42-350-3665 Fax (82)+42-350-3610

Email limhj87gmailcom

RESEARCH INTERESTS

Damage Detection Localization and Quantification on Composite and Metallic Structures Pattern Recognition Finite Element Modeling (FEM)

EDUCATION

2009 ndash 2011 MS Civil and Environmental Eng Korea Advanced Institute of Science and Technology (KAIST) Korea Rep

2006 ndash 2009 BS Civil and Environmental Eng Korea Advanced Institute of Science and Technology (KAIST) Korea Rep

JOURNAL PUBLICATIONS The corresponding author is underlined 1 Hyung Jin Lim Hoon Sohn ldquoReference-free damage detection localization and quantification

using in compositerdquo in preparation for Journal of the Acoustical Society of America 2011

2 Chul Min Yeum Hoon Sohn Jeong Bum Ihn Hyung Jin Lim ldquoInstantaneous delamination detection in a composite using a dual piezoelectric transducer networkrdquo Submitted to Compo-sites Science and Technology 2011

3 Yun-Kyu An Hyung Jin Lim Min Koo Kim Jin Yeol Yang Hoon Sohn Chang Guen Lee ldquoInvestigation of applicability of local reference-free damage detection techniques to in-situ

- 39 -

bridgesrdquo Submitted to ASCE Journal of Structural Engineering 2011

4 Hyung Jin Lim Min Koo Kim Hoon Sohn Chan Yik Park ldquoImpedance based damage detec-tion under varying temperature and loading conditionsrdquo Submitted to NDTampE International 2011

5 Hyung Jin Lim Hoon Sohn ldquoQuantitative estimation of transmitted and reflected lamb waves at discontinuityrdquo Journal of the Korean Society for Nondestructive Testing Vol30 No 4 pp1-8 2010

CONFERENCE PROCEEDINGS

The corresponding author is underlined 1 Hyung Jin Lim and Hoon Sohn ldquoReference-free damage detection localization and quantifi-

cation in compositesrdquo ASCE Engineering Mechanics Institute Conference Boston June 2-4 2011

2 Hyung Jin Lim and Hoon Sohn ldquoReference-free delamination quantification within multi-layer composite materialsrdquo KSNT Spring Conference May 26-27 Gyeong-Ju Korea 2011

3 Min Koo Kim Hyun Jin Lim Hoon Sohn Chan Yik Park ldquoImpedance-based bolt loosening detection under varying temperature and loadingrdquo Asian Pacific Workshop on Structural Health Monitoring Tokyo Japan November 30-December 2 2010

4 Hyung Jin Lim Chul Min Yeum and Hoon Sohn ldquoModeling of Impact-induced Delamination in a Multilayer Composite Platerdquo The 23rd KKCNN Symposium on Civil Engineering Taipei (Taiwan) November 13-15 2010

5 Yun-Kyu An Hyung Jin Lim and Hoon Sohn ldquoTemperature independent Delamination Detec-tion using data normalizationrdquo The 3rd International Conference on Multi-Functional Materi-als and Structures Chonbuk National Univ Jeonju Korea September 14-18 2010

6 Hoon Sohn Yun-Kyu An Chang Gil Lee Min Koo Kim Hyung Jin Lim ldquoApplication of online NDT techniques to decommissioned bridge testingrdquo the 5th European Workshop on Structural Health Monitoring Sorrento Italy June 29-July 02 2010

7 Min Koo Kim Hyung Jin Lim Hoon Sohn Chan Yik Park ldquoImpedance based damage diag-nosis of a complex composite aircraft wing under changing loading and temperature condi-tionsrdquo the 5th European Workshop on Structural Health Monitoring Sorrento Italy June 29-July 02 2010

8 Yun-Kyu An Min Koo Kim Hyung Jin Lim and Hoon Sohn ldquoValidations of reference-free crack detection techniques through a decommissioned bridge testrdquo 2010 COSEIK Annual Con-ference Jeju Korea April 8-9 2010

- 40 -

PARTICIPANT PROJECTS

1 Development of Smart Steel Members for Online Steel Structure Monitoring POSCO Corporation (Funded 41000000 KRW (41000 USD) for 010511 to 123111)

2 Development of Enabling Technologies for Structural Health Monitoring Korean Agency for Defense Development (Funded 240000000 KRW (240000 USD) out of total 3200000000 KRW (3200000 USD) for 060108-053111)

3 Development of On-board SHM Technologies for Composite Air Vehicles The Boeing Company (Funded 400000 USD (400000000 KRW) for 080108 to 073111)

EDUCATIONAL EXPERIENCES

Fall 2009

Teaching Assistant CE207 ldquoElementary Structural Engineering and Laboratoryrdquo KAIST 3 unit undergraduate course

EXPERIMENTAL EXPERIENTCES

2010 Monitoring of Real Aircraft Composite Wing Segment Korea Agency for Defense De-velopment (ADD) Daejeon Korea Real time monitoring of composite air vehicles fund-ed by Korea Agency for Defense Development (ADD)

2009-2011

Field Monitoring of Real Bridge Structures Gimpo (Ramp G bridge) amp Icheon (Sam-seung bridge) Korea US-Korea collaborative research for bridge monitoring testbeds

HONORS amp AWARDS

Scholarships for outstanding students granted KAIST (2007-2009) The three prominent students in the civil and environmental engineering department of KAIST were selected as recipients This scholarship provided 2000$ per year for three years

- 41 -

Page 49: 복합재료 손상의 무기저 진단 및 정량화 - KAISTssslab.kaist.ac.kr/article/pdf/Thesis_2.pdf · 1 Declaration of Ethical Conduct in Research: I, as a graduate student

감 사 의 글

기대와 두려움을 안고 시작한 석사과정이 어느덧 2 년 이라는 세월 속에서 추억으로 자리잡아

가고 있습니다 학부생 기간을 포함한 지난 3 년 동안 SHM 이라는 새로운 분야를 접하고 연구를

수행 할 수 있도록 도와주신 고마운 분들이 있어 그 감사를 표하고자 합니다

우선 저의 많은 부족한 점을 일깨워 주시고 연구에 대한 뜻을 더욱 확고히 세울 수 있게 해

주신 지도교수님이신 손 훈 교수님께 깊은 감사의 뜻을 전합니다 짧은 기간이었지만 교수님의

도움으로 많은 것을 배우고 느끼고 깨달을 수 있었습니다 또한 본 연구를 수행할 수 있도록 지

원해주신 국토해양부 (U-City 석 박사과정 지원사업)에 깊은 감사 드립니다

때로 힘들고 지칠 때 항상 저를 믿고 응원해 주시는 가족이 있었기에 지금의 글을 쓸 수 있었

습니다 또한 지금 연구실에서 함께 생활하고 있는 윤규형 민구형 철민이형 현석이형 진열이형

호민이형 병진이 선혜누나와 성대에 계신 창길이형 미국에 계신 의재형과 Debaditiya 에게도 감

사의 말씀을 전합니다 그리고 모두 나열할 수는 없지만 카이스트 생활의 활력소가 되어준 카이

스트 건설 및 환경공학과 선후배들 직원 분들과 SOS 멤버들에게도 감사를 표합니다

앞서 거론한 많은 이들의 도움이 없었다면 저 또한 지금 이 자리에 있을 수 없었을 것입니다

많이 부족하기에 더 노력하고 배우고자 했던 초심을 다시 한번 생각하며 더 큰 결실을 맺기 위해

앞으로 나아가겠습니다 도움을 주신 모든 분들께 이 감사의 글을 바칩니다

2011년 6월 카이스트에서 임형진 올림

- 38 -

CURRICULUM VITAE

Hyung Jin Lim

Research Assistant

The Civil and Environmental Engineering Department

Korea Advanced Institute of Science and Technology (KAIST)

291 Daehak-Ro Guseong-Dong Yuseong-Gu

Daejeon Republic of Korea 305-701

Phone (82)+42-350-3665 Fax (82)+42-350-3610

Email limhj87gmailcom

RESEARCH INTERESTS

Damage Detection Localization and Quantification on Composite and Metallic Structures Pattern Recognition Finite Element Modeling (FEM)

EDUCATION

2009 ndash 2011 MS Civil and Environmental Eng Korea Advanced Institute of Science and Technology (KAIST) Korea Rep

2006 ndash 2009 BS Civil and Environmental Eng Korea Advanced Institute of Science and Technology (KAIST) Korea Rep

JOURNAL PUBLICATIONS The corresponding author is underlined 1 Hyung Jin Lim Hoon Sohn ldquoReference-free damage detection localization and quantification

using in compositerdquo in preparation for Journal of the Acoustical Society of America 2011

2 Chul Min Yeum Hoon Sohn Jeong Bum Ihn Hyung Jin Lim ldquoInstantaneous delamination detection in a composite using a dual piezoelectric transducer networkrdquo Submitted to Compo-sites Science and Technology 2011

3 Yun-Kyu An Hyung Jin Lim Min Koo Kim Jin Yeol Yang Hoon Sohn Chang Guen Lee ldquoInvestigation of applicability of local reference-free damage detection techniques to in-situ

- 39 -

bridgesrdquo Submitted to ASCE Journal of Structural Engineering 2011

4 Hyung Jin Lim Min Koo Kim Hoon Sohn Chan Yik Park ldquoImpedance based damage detec-tion under varying temperature and loading conditionsrdquo Submitted to NDTampE International 2011

5 Hyung Jin Lim Hoon Sohn ldquoQuantitative estimation of transmitted and reflected lamb waves at discontinuityrdquo Journal of the Korean Society for Nondestructive Testing Vol30 No 4 pp1-8 2010

CONFERENCE PROCEEDINGS

The corresponding author is underlined 1 Hyung Jin Lim and Hoon Sohn ldquoReference-free damage detection localization and quantifi-

cation in compositesrdquo ASCE Engineering Mechanics Institute Conference Boston June 2-4 2011

2 Hyung Jin Lim and Hoon Sohn ldquoReference-free delamination quantification within multi-layer composite materialsrdquo KSNT Spring Conference May 26-27 Gyeong-Ju Korea 2011

3 Min Koo Kim Hyun Jin Lim Hoon Sohn Chan Yik Park ldquoImpedance-based bolt loosening detection under varying temperature and loadingrdquo Asian Pacific Workshop on Structural Health Monitoring Tokyo Japan November 30-December 2 2010

4 Hyung Jin Lim Chul Min Yeum and Hoon Sohn ldquoModeling of Impact-induced Delamination in a Multilayer Composite Platerdquo The 23rd KKCNN Symposium on Civil Engineering Taipei (Taiwan) November 13-15 2010

5 Yun-Kyu An Hyung Jin Lim and Hoon Sohn ldquoTemperature independent Delamination Detec-tion using data normalizationrdquo The 3rd International Conference on Multi-Functional Materi-als and Structures Chonbuk National Univ Jeonju Korea September 14-18 2010

6 Hoon Sohn Yun-Kyu An Chang Gil Lee Min Koo Kim Hyung Jin Lim ldquoApplication of online NDT techniques to decommissioned bridge testingrdquo the 5th European Workshop on Structural Health Monitoring Sorrento Italy June 29-July 02 2010

7 Min Koo Kim Hyung Jin Lim Hoon Sohn Chan Yik Park ldquoImpedance based damage diag-nosis of a complex composite aircraft wing under changing loading and temperature condi-tionsrdquo the 5th European Workshop on Structural Health Monitoring Sorrento Italy June 29-July 02 2010

8 Yun-Kyu An Min Koo Kim Hyung Jin Lim and Hoon Sohn ldquoValidations of reference-free crack detection techniques through a decommissioned bridge testrdquo 2010 COSEIK Annual Con-ference Jeju Korea April 8-9 2010

- 40 -

PARTICIPANT PROJECTS

1 Development of Smart Steel Members for Online Steel Structure Monitoring POSCO Corporation (Funded 41000000 KRW (41000 USD) for 010511 to 123111)

2 Development of Enabling Technologies for Structural Health Monitoring Korean Agency for Defense Development (Funded 240000000 KRW (240000 USD) out of total 3200000000 KRW (3200000 USD) for 060108-053111)

3 Development of On-board SHM Technologies for Composite Air Vehicles The Boeing Company (Funded 400000 USD (400000000 KRW) for 080108 to 073111)

EDUCATIONAL EXPERIENCES

Fall 2009

Teaching Assistant CE207 ldquoElementary Structural Engineering and Laboratoryrdquo KAIST 3 unit undergraduate course

EXPERIMENTAL EXPERIENTCES

2010 Monitoring of Real Aircraft Composite Wing Segment Korea Agency for Defense De-velopment (ADD) Daejeon Korea Real time monitoring of composite air vehicles fund-ed by Korea Agency for Defense Development (ADD)

2009-2011

Field Monitoring of Real Bridge Structures Gimpo (Ramp G bridge) amp Icheon (Sam-seung bridge) Korea US-Korea collaborative research for bridge monitoring testbeds

HONORS amp AWARDS

Scholarships for outstanding students granted KAIST (2007-2009) The three prominent students in the civil and environmental engineering department of KAIST were selected as recipients This scholarship provided 2000$ per year for three years

- 41 -

Page 50: 복합재료 손상의 무기저 진단 및 정량화 - KAISTssslab.kaist.ac.kr/article/pdf/Thesis_2.pdf · 1 Declaration of Ethical Conduct in Research: I, as a graduate student

CURRICULUM VITAE

Hyung Jin Lim

Research Assistant

The Civil and Environmental Engineering Department

Korea Advanced Institute of Science and Technology (KAIST)

291 Daehak-Ro Guseong-Dong Yuseong-Gu

Daejeon Republic of Korea 305-701

Phone (82)+42-350-3665 Fax (82)+42-350-3610

Email limhj87gmailcom

RESEARCH INTERESTS

Damage Detection Localization and Quantification on Composite and Metallic Structures Pattern Recognition Finite Element Modeling (FEM)

EDUCATION

2009 ndash 2011 MS Civil and Environmental Eng Korea Advanced Institute of Science and Technology (KAIST) Korea Rep

2006 ndash 2009 BS Civil and Environmental Eng Korea Advanced Institute of Science and Technology (KAIST) Korea Rep

JOURNAL PUBLICATIONS The corresponding author is underlined 1 Hyung Jin Lim Hoon Sohn ldquoReference-free damage detection localization and quantification

using in compositerdquo in preparation for Journal of the Acoustical Society of America 2011

2 Chul Min Yeum Hoon Sohn Jeong Bum Ihn Hyung Jin Lim ldquoInstantaneous delamination detection in a composite using a dual piezoelectric transducer networkrdquo Submitted to Compo-sites Science and Technology 2011

3 Yun-Kyu An Hyung Jin Lim Min Koo Kim Jin Yeol Yang Hoon Sohn Chang Guen Lee ldquoInvestigation of applicability of local reference-free damage detection techniques to in-situ

- 39 -

bridgesrdquo Submitted to ASCE Journal of Structural Engineering 2011

4 Hyung Jin Lim Min Koo Kim Hoon Sohn Chan Yik Park ldquoImpedance based damage detec-tion under varying temperature and loading conditionsrdquo Submitted to NDTampE International 2011

5 Hyung Jin Lim Hoon Sohn ldquoQuantitative estimation of transmitted and reflected lamb waves at discontinuityrdquo Journal of the Korean Society for Nondestructive Testing Vol30 No 4 pp1-8 2010

CONFERENCE PROCEEDINGS

The corresponding author is underlined 1 Hyung Jin Lim and Hoon Sohn ldquoReference-free damage detection localization and quantifi-

cation in compositesrdquo ASCE Engineering Mechanics Institute Conference Boston June 2-4 2011

2 Hyung Jin Lim and Hoon Sohn ldquoReference-free delamination quantification within multi-layer composite materialsrdquo KSNT Spring Conference May 26-27 Gyeong-Ju Korea 2011

3 Min Koo Kim Hyun Jin Lim Hoon Sohn Chan Yik Park ldquoImpedance-based bolt loosening detection under varying temperature and loadingrdquo Asian Pacific Workshop on Structural Health Monitoring Tokyo Japan November 30-December 2 2010

4 Hyung Jin Lim Chul Min Yeum and Hoon Sohn ldquoModeling of Impact-induced Delamination in a Multilayer Composite Platerdquo The 23rd KKCNN Symposium on Civil Engineering Taipei (Taiwan) November 13-15 2010

5 Yun-Kyu An Hyung Jin Lim and Hoon Sohn ldquoTemperature independent Delamination Detec-tion using data normalizationrdquo The 3rd International Conference on Multi-Functional Materi-als and Structures Chonbuk National Univ Jeonju Korea September 14-18 2010

6 Hoon Sohn Yun-Kyu An Chang Gil Lee Min Koo Kim Hyung Jin Lim ldquoApplication of online NDT techniques to decommissioned bridge testingrdquo the 5th European Workshop on Structural Health Monitoring Sorrento Italy June 29-July 02 2010

7 Min Koo Kim Hyung Jin Lim Hoon Sohn Chan Yik Park ldquoImpedance based damage diag-nosis of a complex composite aircraft wing under changing loading and temperature condi-tionsrdquo the 5th European Workshop on Structural Health Monitoring Sorrento Italy June 29-July 02 2010

8 Yun-Kyu An Min Koo Kim Hyung Jin Lim and Hoon Sohn ldquoValidations of reference-free crack detection techniques through a decommissioned bridge testrdquo 2010 COSEIK Annual Con-ference Jeju Korea April 8-9 2010

- 40 -

PARTICIPANT PROJECTS

1 Development of Smart Steel Members for Online Steel Structure Monitoring POSCO Corporation (Funded 41000000 KRW (41000 USD) for 010511 to 123111)

2 Development of Enabling Technologies for Structural Health Monitoring Korean Agency for Defense Development (Funded 240000000 KRW (240000 USD) out of total 3200000000 KRW (3200000 USD) for 060108-053111)

3 Development of On-board SHM Technologies for Composite Air Vehicles The Boeing Company (Funded 400000 USD (400000000 KRW) for 080108 to 073111)

EDUCATIONAL EXPERIENCES

Fall 2009

Teaching Assistant CE207 ldquoElementary Structural Engineering and Laboratoryrdquo KAIST 3 unit undergraduate course

EXPERIMENTAL EXPERIENTCES

2010 Monitoring of Real Aircraft Composite Wing Segment Korea Agency for Defense De-velopment (ADD) Daejeon Korea Real time monitoring of composite air vehicles fund-ed by Korea Agency for Defense Development (ADD)

2009-2011

Field Monitoring of Real Bridge Structures Gimpo (Ramp G bridge) amp Icheon (Sam-seung bridge) Korea US-Korea collaborative research for bridge monitoring testbeds

HONORS amp AWARDS

Scholarships for outstanding students granted KAIST (2007-2009) The three prominent students in the civil and environmental engineering department of KAIST were selected as recipients This scholarship provided 2000$ per year for three years

- 41 -

Page 51: 복합재료 손상의 무기저 진단 및 정량화 - KAISTssslab.kaist.ac.kr/article/pdf/Thesis_2.pdf · 1 Declaration of Ethical Conduct in Research: I, as a graduate student

bridgesrdquo Submitted to ASCE Journal of Structural Engineering 2011

4 Hyung Jin Lim Min Koo Kim Hoon Sohn Chan Yik Park ldquoImpedance based damage detec-tion under varying temperature and loading conditionsrdquo Submitted to NDTampE International 2011

5 Hyung Jin Lim Hoon Sohn ldquoQuantitative estimation of transmitted and reflected lamb waves at discontinuityrdquo Journal of the Korean Society for Nondestructive Testing Vol30 No 4 pp1-8 2010

CONFERENCE PROCEEDINGS

The corresponding author is underlined 1 Hyung Jin Lim and Hoon Sohn ldquoReference-free damage detection localization and quantifi-

cation in compositesrdquo ASCE Engineering Mechanics Institute Conference Boston June 2-4 2011

2 Hyung Jin Lim and Hoon Sohn ldquoReference-free delamination quantification within multi-layer composite materialsrdquo KSNT Spring Conference May 26-27 Gyeong-Ju Korea 2011

3 Min Koo Kim Hyun Jin Lim Hoon Sohn Chan Yik Park ldquoImpedance-based bolt loosening detection under varying temperature and loadingrdquo Asian Pacific Workshop on Structural Health Monitoring Tokyo Japan November 30-December 2 2010

4 Hyung Jin Lim Chul Min Yeum and Hoon Sohn ldquoModeling of Impact-induced Delamination in a Multilayer Composite Platerdquo The 23rd KKCNN Symposium on Civil Engineering Taipei (Taiwan) November 13-15 2010

5 Yun-Kyu An Hyung Jin Lim and Hoon Sohn ldquoTemperature independent Delamination Detec-tion using data normalizationrdquo The 3rd International Conference on Multi-Functional Materi-als and Structures Chonbuk National Univ Jeonju Korea September 14-18 2010

6 Hoon Sohn Yun-Kyu An Chang Gil Lee Min Koo Kim Hyung Jin Lim ldquoApplication of online NDT techniques to decommissioned bridge testingrdquo the 5th European Workshop on Structural Health Monitoring Sorrento Italy June 29-July 02 2010

7 Min Koo Kim Hyung Jin Lim Hoon Sohn Chan Yik Park ldquoImpedance based damage diag-nosis of a complex composite aircraft wing under changing loading and temperature condi-tionsrdquo the 5th European Workshop on Structural Health Monitoring Sorrento Italy June 29-July 02 2010

8 Yun-Kyu An Min Koo Kim Hyung Jin Lim and Hoon Sohn ldquoValidations of reference-free crack detection techniques through a decommissioned bridge testrdquo 2010 COSEIK Annual Con-ference Jeju Korea April 8-9 2010

- 40 -

PARTICIPANT PROJECTS

1 Development of Smart Steel Members for Online Steel Structure Monitoring POSCO Corporation (Funded 41000000 KRW (41000 USD) for 010511 to 123111)

2 Development of Enabling Technologies for Structural Health Monitoring Korean Agency for Defense Development (Funded 240000000 KRW (240000 USD) out of total 3200000000 KRW (3200000 USD) for 060108-053111)

3 Development of On-board SHM Technologies for Composite Air Vehicles The Boeing Company (Funded 400000 USD (400000000 KRW) for 080108 to 073111)

EDUCATIONAL EXPERIENCES

Fall 2009

Teaching Assistant CE207 ldquoElementary Structural Engineering and Laboratoryrdquo KAIST 3 unit undergraduate course

EXPERIMENTAL EXPERIENTCES

2010 Monitoring of Real Aircraft Composite Wing Segment Korea Agency for Defense De-velopment (ADD) Daejeon Korea Real time monitoring of composite air vehicles fund-ed by Korea Agency for Defense Development (ADD)

2009-2011

Field Monitoring of Real Bridge Structures Gimpo (Ramp G bridge) amp Icheon (Sam-seung bridge) Korea US-Korea collaborative research for bridge monitoring testbeds

HONORS amp AWARDS

Scholarships for outstanding students granted KAIST (2007-2009) The three prominent students in the civil and environmental engineering department of KAIST were selected as recipients This scholarship provided 2000$ per year for three years

- 41 -

Page 52: 복합재료 손상의 무기저 진단 및 정량화 - KAISTssslab.kaist.ac.kr/article/pdf/Thesis_2.pdf · 1 Declaration of Ethical Conduct in Research: I, as a graduate student

PARTICIPANT PROJECTS

1 Development of Smart Steel Members for Online Steel Structure Monitoring POSCO Corporation (Funded 41000000 KRW (41000 USD) for 010511 to 123111)

2 Development of Enabling Technologies for Structural Health Monitoring Korean Agency for Defense Development (Funded 240000000 KRW (240000 USD) out of total 3200000000 KRW (3200000 USD) for 060108-053111)

3 Development of On-board SHM Technologies for Composite Air Vehicles The Boeing Company (Funded 400000 USD (400000000 KRW) for 080108 to 073111)

EDUCATIONAL EXPERIENCES

Fall 2009

Teaching Assistant CE207 ldquoElementary Structural Engineering and Laboratoryrdquo KAIST 3 unit undergraduate course

EXPERIMENTAL EXPERIENTCES

2010 Monitoring of Real Aircraft Composite Wing Segment Korea Agency for Defense De-velopment (ADD) Daejeon Korea Real time monitoring of composite air vehicles fund-ed by Korea Agency for Defense Development (ADD)

2009-2011

Field Monitoring of Real Bridge Structures Gimpo (Ramp G bridge) amp Icheon (Sam-seung bridge) Korea US-Korea collaborative research for bridge monitoring testbeds

HONORS amp AWARDS

Scholarships for outstanding students granted KAIST (2007-2009) The three prominent students in the civil and environmental engineering department of KAIST were selected as recipients This scholarship provided 2000$ per year for three years

- 41 -