Modelling of defects and failure in composites - Stephen Hallett.pdf ·  · 2016-02-23Modelling of defects and failure in composites ... –Shape of resin pocket at ply termination

Modelling of defects and failure in composites

Prof. Stephen Hallett

www.bris.ac.uk/composites

Virtual Testing

Coupons

Generic features

Specific features

Sub-components

Components

Test Analysis

Current Future

Test Analysis

Test backed up by Analysis Analysis backed up by Test

Defects and Features

• Composites’ structural properties are influenced

by ply scale features in the geometry

• The as-manufactured component varies from the

design intent

• Knowledge of the effect of defects is important

for :

– Certification

– Design process to inform stress allowables

– NDT and quality control for sentencing of defects

– Test work to design representative specimens

• Large range of possible defects means testing of

all cases is not possible – development of

hi-fidelity modelling tools

As-Manufactured Virtual Composites

• Process

modelling

• Formation

of defects

• Influence of

process

parameters

• Advanced

failure

prediction

• High fidelity

simulations

• Effect of defects

• Complex loading

• As

manufactured geometry

• Mesh creation

• Acceptance limits

• Inspection limits

6 mm

Effect of wrinkles

Ply drops and embedded delaminations

Textile composites

Process modelling

FE models from NDTdata

• Out of plane waviness

– Critical defect in terms of reduction in

compressive strength

• Gaps and overlaps

– Caused by automated tape laying process

• Delamination defects

– E.g. small inclusions or imperfections during layup

– Simplified specimen used, combined with ply drops

– Low velocity impact damage

• Voids and Porosity

Defect Types

Wrinkle Experiments

• Coupons with wrinkle angle θ (8° and 12° for tension, 10° and 12°for compression)

• Produced by varying the number of inserted 90° strips, together with pristine specimen for comparison.

0

50

100

150

200

250

300

350

400

0 0.5 1 1.5 2

Ten

sile

str

ess

(M

Pa)

Crosshead displacement (mm)

* Mukhopadhyay S., Jones M.I. & Hallett S.R. Composites 2013, Azores

Wrinkle Modelling

• Modelled with ply-by-ply hi-fidelity analysis in Abaqus/explicit

• Cohesive interface elements for delamination

• In-plane damage models for matrix cracking and fibre failure

6 mm

+45°

90°

-45°

0°

Compressive stress at final failure

* Mukhopadhyay S., Jones M.I. & Hallett S.R. Composites 2013, Azores

A

A’

B

B’

Hi-fidelity Gaps & Overlaps Modelling

preprocessor1 preprocessor2 preprocessor3 preprocessor4

A

A’

B

B’

A

A’

B

B’

preprocessor5

AA’

Improved Ply waviness by Spline curve interpolating in X, Y direction-1st derivative of waviness is continuous at every node

• Internal geometry of plies and Gaps & Overlaps features is highly complex

• Requires custom tools for generating models with Gaps & Overlaps

* Li X. et al, ICMAC, 2015, Bristol

Gaps & Overlaps Models

• Cross-section images

A

A’

A

A’

A

A’

* Li X. et al, ICMAC, 2015, Bristol

Compression Testing of Gaps and Overlaps

* Li X. et al, ICMAC, 2015, Bristol

Delaminations and Ply-drops

• Ply-drops are an unavoidable feature in variable thickness laminates

• Simplified specimen developed to examine failure

• Hi-fidelity modelling with cohesive elementsused for assessment

• Local geometry has a strong effect on strength

– Ply waviness from layup

– Shape of resin pocket at ply termination

– Manufacturing defects leading to small delaminations at the end of ply terminations

* Kawashita L et al, ICCM18, 2011, Jeju Island

Ply-Level Meshing Algorithms

• As-manufactured geometries and internal features captured initially by an image-based meshing technique

• Virtual testing can then be used to augment the building block approach –here considering a wide range of delamination defect locations

* Kawashita L et al, ICCM18, 2011, Jeju Island

Virtual Testing of Defects• Example: 4mm long delaminations (NDT detection threshold) introduced

near ply terminations

• Automatic mesh generation, job submission & post-processing in a Linux cluster; hundreds of runs completed overnight

Runtime for each

slice model:

20-40 min*

*previous generation

8-core HPC node

01a

defects

resin

pockets

01b 02a 02b

03a 03a 04a 04a

05a 05a 06a 06a

07a 07a 08a 08a

* Kawashita L et al, ICCM18, 2011, Jeju Island

Virtual Testing of Defects: Static Strength

• Maximum strength knockdown, critical defect locations and respective delamination mode-mixities identified

• Methodology validated against experiments for various thickness and taper angles

worst

defect

location

pristine

worst defect

location

02a

Symmetric

specimens

* Kawashita L et al, ICCM18, 2011, Jeju Island

Laminated Composite Plate

Supporting Window

Indenter/ Impactor

m n

b a

d

X Y

Z

90o

0o

-45o

45o

0

1

2

3

4

5

6

0 1 2 3 4 5

Load (

kN

)

Time (ms)

FS-6J

FS-4J

* Hallett S.R. and Sun X., ICILLS,2014, Cape Town

Impact Damage

• Low velocity impact is well known for introducing delamination damage (BVID)

• This dramatically reduces the Compression After Impact (CAI) strength

• Hi-fidelity models are able to accurately predict impact damage in the form of delamination, matrix cracks and fibrefailure

• Impact induced damage can be input into models either from simulation or test data

• In-plane loading of damaged panel caused growth of critical delaminations and ultimate failure

• Models can be used to predict failure and also study mechanisms in detail as well as effect of layup etc.

CAI Strength

Voids

• Modelling of voids is not as well advanced as other defect types

• More challenging because the exact mechanism of failure from voids is not well understood

• Most work considers the knock-down due to global void volume fraction, but not local void morphology

• Current work is focussed on the detailed understanding of failure from voids to embed into future models

Process Modelling

• As well as failure behaviour, for design it is also important to understand the origin of defects

• Modelling can be used to predict the formation of defects, especially fibre waviness

• Compaction during processsing is a major driver of the excess length needed for creation of wrinkles

• FE models recently developed that can capture this behaviour

Future Challenges

• How to model a full component when a feature model takes >500,000 elements?– Homogenised models

– Shell elements

• Bridging the length scales– Micro-meso

– Meso-macro

– Multi-scale models

• Modelling the as manufactured condition– Statistical variance

• New materials and manufacturing processes– 3D woven textiles

– Fibre placement

• Computational resource– Very large numbers of CPU

– Used efficiently

• Advanced Numerical Methods– XFEM