Chung-Bang Yun Professor, Department of Civil and Environmental Engineering Director, Smart Infrastructure Technology Center KAIST, Korea International Symposium on Stochastic Analysis for Risk Management Tokyo, Japan on Dec 23, 2010
Chung-Bang YunProfessor, Department of Civil and Environmental Engineering
Director, Smart Infrastructure Technology Center
KAIST, Korea
International Symposium on Stochastic Analysis for Risk Management
Tokyo, Japan on Dec 23, 2010
2
1• Introduction
2
• Vibration-based SHM- Long-term SHM systems for large civil infrastructure
- Large-scale wireless sensor network system on a cable stayed bridge
- Kalman filtering technique for damage identification under earthquake
3
• Innovative Nondestructive Evaluation Techniques- Innovative sensing: OFS, Microwaves, Vision-based sensing
- Robotic trenchless rehabilitation of underground pipes
- Multifunctional wireless impedance sensor node
- Wireless power/data transmission for guided wave sensing
4• Concluding Remarks
4
Dedication to Prof. Shinozuka
Congratulation to Prof. Shinozukafor His 80th birthday, and
for His wonderful career as a professor and researcher leading the world technology developments for the mitigation and management of the natural and man-made disasters!!
Whole-hearted appreciationfor His great teaching and guidance to us:7 Korean PhD students & 10 visiting scholars
5
Identification of Linear Dynamic System(ASCE J. Eng. Mech. Div., 1982)
Identification of Nonlinear Dynamic System(J. Struct. Mech., 1980)
Bridge deck model and Wind forces
1 2 3 4
1 2 3 4
( ) ( ) ( ) ( ) ( )
( ) ( ) ( ) ( ) ( )
se
se
F t H h t H t H h t H t
Q t Ah t A t A h t A t
A fixed offshore tower and its model
Estimated parameters
1
0 0
1
0
1
0 0
1
0
( )
( )
( )
( )
a
a D
a
a M
J M M C
D M M C
K M M K
L M M C
ARMAX model &
Maximum likelihood method
0 0 0 {( ) }a D MM C K M C v v C v
Extended Kalman filter for
parameter estimation
Observed and Estimated responses in & ( )h t ( )t
Early Researches on System Identification
by Prof. Shinozuka and Yun (in late 1970)
I. Long-term SHM systems for large civil infrastructures
II. Large-scale wireless sensor network system on a cable stayed bridge
III. Kalman filtering techniques for damage identification under earthquake
7
Explore Sensor-Based Monitoring for
•Real-Time Structural Health Monitoring
Condition-Based Timely Maintenance
•Rapid & Remote Post-Event Damage Assessment
Effective Emergency Response
I. Long-term SHM Systems for Large Civil Infrastructures(M Feng at UCI)
•Soil-Structure Interaction
•Amplitude of Ground Motion
•Modeling of
Traffic Excitation•5-Year
Monitoring
•Vibration Test•Vehicle-Structure
Interaction•8-Year Monitoring
•Bridge Doctor
Software
8
Detection of Fatigue Damage in Steel Structures(M Shinozuka at UCI)
RepeaterReceiver
Non-Gaussian Conditional
Simulation along the Deck
A1A2
A3
Anemometer and
Transmitter
9
Non-Invasive Damage Detection of Underground Pipes(M Shinozuka at UCI)
(a) 2D contour map
(b) Visualized graphic
Based on Non-Invasive Acceleration Measurements of Pipe Motion rather than
Water Pressure in Pipe and Wireless Transmission of Data (NIST)
10
Advanced Diagnostics and Prognostics(M Feng at UCI)
Signal Processing and System ID Methods for Post-Event Damage Assessment
Validation by Seismic Shaking Table Tests
Field validation of state-of-the-art wireless smart sensor technology for a cable-stayed bridge in KoreaConstruction of an international SHM test-bedParticipants :
Korea : H.J. Jung & C.B. Yun (KAIST),
H.K. Kim (SNU), J.J. Lee (Sejong Univ.)
J.W. Seo (Hyundai Institute of Const. Tech.)
US : B.F. Spencer, Jr. & G. Agha (UIUC)
Japan : T. Nagayama & Y. Fujino (U. of Tokyo)
US-Korea-Japan Collaborative Project (2009-2011)
II. Bridge SHM Test-bed Using Large-scale Wireless Sensor Network(CB Yun, HJ Jung at KAIST; BF Spencer at UIUC; & T Nagayama at UT)
Daejeon (KAIST)
Seoul
Jindo Island
Jindo Bridges
Jeju Island
12
External Input
Connector
SHM-A Sensor Board
(bottom)
Basic
Connect
or
Programmable
Filter w/ ADC
Quickfilter
QF4A512
SHM-A Sensor Board
(top)
Accelerometer
ST Microelectronics
LIS344ALH
OP AMP
TI OPA 4344
Humidity & Temp.
Sensor
SHT11
Light Sensor
TAOS 2561
Basic
Connector
SHM-A Board (rev. 4.0)
External Antenna
(Antenova 2.4GHz)
iMote2
Battery Board
(IBB2400CA)
w/ 3 AAA
Batteries
Assembled iMote2
Multi-scale SHM-A Board
Hierarchical Data/Information Processing
Data acquisition
Outcome forwarding
processing
Energy Harvester
Wireless Smart Sensor: Hardware & Software
Data & Power Management
Effective Power
ManagementSnoozeAlarm
Data Inundation
MitigationThresholdSentry
Continuous /Autonomous
OperationAutoMonitor
13
Cable : 8
Deck : 22
Pylon : 3
Total : 33
Cable : 7
Deck : 26
Pylon : 3
Wind: 1
Total : 37
70 wireless sensor nodes
(207 Acceleration Ch. & 3 Wind Ch.)
Nodes on pylon top(powered by solar cell)Deck Nodes Nodes on pylons Nodes on cables
Nodes on cables(powered by solar cell) Wind-Sentry
Amemometerinterfaced with
Wind-Sentry
Various Types of Sensor Nodes
Sensor Deployments
14
Output-only Modal Analysis Combined Mode Shapes
Click Mouse!! (Left: Skip, Right: Change Phase of Mode Shape)
Click Mouse!! (Left: Skip, Right: Change Phase of Mode Shape)
Click Mouse!! (Left: Skip, Right: Change Phase of Mode Shape)
Click Mouse!! (Left: Skip, Right: Change Phase of Mode Shape)
Click Mouse!! (Left: Skip, Right: Change Phase of Mode Shape)
Click Mouse!! (Left: Skip, Right: Change Phase of Mode Shape)
Click Mouse!! (Left: Skip, Right: Change Phase of Mode Shape)
Click Mouse!! (Left: Skip, Right: Change Phase of Mode Shape)
Click Mouse!! (Left: Skip, Right: Change Phase of Mode Shape)
Click Mouse!! (Left: Skip, Right: Change Phase of Mode Shape)
DV1
DV2
DV3
DV6
DV9
Output-only Modal Identification
0 0.5 1 1.5 2 2.5 30
20
40
60
80
100
Frequency (Hz)
Syste
m O
rder
0
20
40
60
80
The 1
st
Sin
gula
r V
alu
e
DL1 DV1 DV2 NC1 DV3 DV4 DV5NC2
DV6 DT1DV7
DV8 PB1NC3
DV9
Click Mouse!! (Left: Skip, Right: Change Phase of Mode Shape)
DV1
SSI: 0.4380Hz
FDD: 0.4492Hz
FEM: 0.4293Hz
Click Mouse!! (Left: Skip, Right: Change Phase of Mode Shape)
DV2
SSI: 0.6439Hz
FDD: 0.6445Hz
FEM: 0.6375Hz
Click Mouse!! (Left: Skip, Right: Change Phase of Mode Shape)
DT1
SSI: 1.8410Hz
FDD: Not found
FEM: 1.9592Hz
DV3
SSI: 1.0364Hz
FDD: 1.0352Hz
FEM: 0.9958Hz
Click Mouse!! (Left: Skip, Right: Change Phase of Mode Shape)
Jindo-side
Subnetwork
15
Jindo Haenam70m 70m344m
f 7x139f 7x109f 7x73f 7x151
JC4(C-JE1)
JC6(C-JE2)
JC9(C-JE4)
JC13(C-JE7)
JC15(C-JE8)
HC15(C-HE7)
HC13(C-HE6)
HC9(C-HE4)
HC6(C-HE2)
HC4(C-HE1)
4 parallel cables
4 parallel cables
15
Using high mode frequencies
Cable
Tension force (tonf)
Routine inspection Installed WSSNsin 2009 (Avg.)In 2007 In 2008
HC4 262.7 268.4 274.0 (2.04)*
HC6 304.6 304.1 294.7 (-3.19)*
HC9 86.9 88.5 89.3 (0.90)*
HC13 164.0 165.1 170.2 (3.00)*
HC15 219.9 220.0 224.9 (2.18)*
JC4 245.1 250.7 254.0 (1.30)*
JC6 282.0 277.5 274.5 (-1.09)*
JC9 85.5 86.6 88.5 (2.15)*
JC13 148.3 150.7 154.3 (2.33)*
JC15 214.1 216.5 216.8 (0.14)*
*Differences (%) compared with tensions in 2008
0 5 10 15 20
50
100
150
200
250
300
350
400
450
500
Measurements
Tensio
n F
orc
e (
Tonf)
Estimated Tension from Haenam-side Cables
HC4
HC6
HC9
HC13
HC15
0 5 10 15 20 25
50
100
150
200
250
300
350
400
450
500
Measurements
Tensio
n F
orc
e (
Tonf)
Estimated Tension from Jindo-side Cables
JC4
JC6
JC9
JC13
JC15
2 2 22
2 44 4
n
eff eff
f T EI na bn
n mL mL
24 effT mL aLinear regression
to find a and b
Cable Tension Force Estimation
Vibration-based Tension Force Estimation
Estimated Tension Forces
16
*
1
m
j jj
a=
= b Sa s;踐 踐D D鉗 ・ 鉗鉗顏 顏
r pS
r p
HPC based on Cosine Similarity FE Model Updating Using GA
jq
js
is
j ja s
i i
a si
V*b
jV
iq
Two Single Vectors
*b
js
is
iV j
V
g
is
g
iq
g
iV
i jq q=
Grouped Vector with
Same Cosine Similarity
2*
j j jV a= -b s * /T T
j j j ja = s b s s
Least Square
Estimation:
( )2
*
* * * * 2
* *1 (1 cos )
( )( )
T
jT T
j jT T
j j
V q
踐 = - = - 顏
s bb b b b
s s b b
Always smaller
than
0 5 10 15 20 25 30 35 40 45 50
0.090.1
0.2
0.3
0.4
0.5 0.6
Generation
Norm
aliz
ed O
bje
ctive F
unction V
alu
e
Objective Function
No Group:27P
Conv. Group:11P
HPC1:18P
HPC2:25P
Average Adjustment Rate
No Group: 50.5%, HPC1(18P)=38.6%
fG1 fG5 fG10 fG15 fG20 fG25PJ PH
0.5
1
1.5
2
Parameters
Pupdate
d /
Pin
itia
l
No Group:27P
HPC1:18P
FE Model Updating with Hierarchical Parameter Clustering
No Grouping: 27P
10 16 1 25 5 21 9 17 6 20 13 23 3 11 15 7 19 4 22 8 12 14 18 2 24 27 260
0.01
0.02
0.03
0.04
0.05
0.06
0.07
0.08
Parameters
Cosin
e D
ista
nce
Hierarchical Clustering (Threshold Distance = 0.0038053)
Hierarchical Parameter
Clustering (HPC1:18P)
17
- Modified Takeda Model: Non-analytic Form – Identification of , , &
Stiffness Degradation : Pinching : Strength Deterioration :
yM
Elasticcurvature
Inelasticcurvature
60m 60m
20m
① ② ③ ④ ⑤
⑥ y
xz
Ground acceleration (pga=0.4g)
M ( , )x xM Mf ( , )x xM Mf
M
fff1( )EI
2( )EI
"
3( )EI
'
3( )EI
*( , )p pM Mf
0.00 2.00 4.00 6.00Time(sec)
-5.0E+0
0.0E+0
5.0E+0
Exc
it ati o
n(m
/ sec
2)
gu pL
yfyf f
III. Extended Kalman Filter for Damage Identification of Bridge Pier
(KJ Lee at Daelim Co., CB Yun at KAIST)
18
Observation Equations for Acceleration Response Measurement
State Vector and Parameter Vector
θ( ) { ( ) ( ) ( ) ( )}yk M k k k k
1 1 1[ ( ) ( ) ( ) ( ) ( ) (k) θ( )]T
k l l lq k q k q k q k q k q k
1 ( , ; , )k k k k kG f k w
Y (Χ ; )k k kh k v
Extended Kalman Filter Formulation
Two Step Approach
Sequential Prediction Error Method for Parameters, θ(k)
Extended Kalman Filter for State, X(k)1Yk 1Yk
1/ 1 / 1 1 1ˆˆ [ ( )]k k k k k k k kX X K Y Y
1 1 1 1 1 1/ 1ˆ ˆ[ ( , )]k k k k k k k k kB Y Y X
1ˆk
1Xk
19
Parameter Identification by Extended Kalman Filter
Estimated nonlinear parameters
Nonlinear Parameters My(KNm) α β γ
Exact Values 1200.0 0.01000 1.0000 0.2000
Initial guesses 600.0 0.00500 0.5000 0.1000
w/ 2-modes 1156.8 0.0096 0.4956 0.1674
w/ 5-mode3 1174.2 0.0096 0.4894 0.1789
w/ 12-modes 1167.8 0.0096 0.5509 0.1789
Parameter estimation Estimated M-φ
5-modes 12-modes
-4.0E-4 0.0E+0 4.0E-4Curvature
-2.0E+6
0.0E+0
2.0E+6
Mom
ent (K
N*m
)
Exact
w/ SMEKF
-4.0E-4 0.0E+0 4.0E-4Curvature
-2.0E+6
0.0E+0
2.0E+6
Mom
ent (K
N*m
)
Exact
w/ SMEKF
Estimated M-φ relationship by SMEKF Estimated responses
2.00 3.00 4.00 5.00 6.00Time (sec)
-5.0E+0
0.0E+0
5.0E+0
Acce
lerati
on (m
/sec^2
)
Exact
w/ SMEKF
2.00 3.00 4.00 5.00 6.00Time (sec)
-5.0E+0
0.0E+0
5.0E+0
Accel
eration
(m/se
c^2)
Exact
w/ SMEKF
at Node 3
at Node 6
I. Innovative sensing: OFS, Microwaves, Vision-based sensing
II. Robotic trenchless rehabilitation of underground pipes
III. Multifunctional wireless impedance sensor node
IV. Wireless power/data transmission for guided wave sensing
I. Innovative Sensors and NDE Technologies(M Feng at UCI)
Microwave Imaging System
Distributed Fiber Optic Strain SensorMoiré Fringe–based Fiber Optic Accelerometers
C
r
ac
kA
C
r
B
X [mm]
0 50 100 150 200
Y [m
m]
0
50
100
150
200
250
300
0
20
40
60
80
100
120
140
160
Strain []
Non-Contact Vision-Based Disp. Sensors
• Spin carbon fibers on internal wall of pipes 11 times faster than
manual application
• Reduce cost by 15 times
• Eliminate environmental impact
Robotics Trenchless Rehabilitation of Underground Pipes(M Feng at UCI)
2 1
31
( )( ) [ (1 )]
( ) ( )
s
sAZ ZZ i
ZC
• Electromechanical Impedance
PZT
Impedance Analyzer PC
GPIB Interface
Card & Cable
- +
PZT PZT
MFC
Thermo-
coupler
3.05 3.1 3.15
x 104
-2
-1
0
1
2
3
4
5
6
7
8
Frequency [Hz]
Re
al (
Z(
) )
Test #1 at 22.6oC
Test #338 at 10.3oC
Max. CC=0.9858
CC=0.9900
3.05 3.1 3.15
x 104
85
90
95
100
105
110
115
120
125
130
135
140
Frequency [Hz]
Re
al (
Z(
) )
Test #1 at 22.6oC
Test #338 at 10.3oCCC=-0.0986
0 200 400 600 800 1000 1200 1400 1600
0.65
0.7
0.75
0.8
0.85
0.9
0.95
1
1.05
Test No.
Max
. CC
w/ E
FS
0 200 400 600 800 1000 1200 1400 16005
10
15
20
25
30
35
Tem
pera
ture
(o C)
thr1=0.942
thr2=0.884
thr3=0.8072mm cut
4mm cut
8mm cut
0 200 400 600 800 1000 1200 1400 1600
-0.2
0
0.2
0.4
0.6
0.8
1
1.2
Test No.
Max
. CC
w/o
EFS
0 200 400 600 800 1000 1200 1400 1600
10
15
20
25
30
35
Tem
pera
ture
(o C)
Intact 2mm cut 4mm cut 8mm cut
Temperature
Compensation
30 35 40
0
10
20
30
Frequency [kHz]
Re
al (
Z(
) )
Intact 2mm 4mm 8mm
II. Multifunctional Wireless Impedance Sensor Node(J Min, CB Yun at KAIST, S Park at SKKU)
125 130 1356
7
8
9
10
11
12
Real part (V/I
)
Frequency (kHz)
BaselineDamage
Baselines Damage
Effective Frequency Shift (EFS)
Temperature Compensation Technique
1
1CC { ( ( ) )( ( ) )}/max max
N
i i i X Y
i
x x y yN
Low-Cost Wireless Impedance Sensor Node
Wireless Impedance Sensor Node (WISN)
* Impedance Analyzer (HP4294A) : $41,000
Impedance Sensor Node (KAIST, UTO)
Control Room
Internet
Base station
Wireless
PZT Sensors
Pattern Recognitionfor Damage Diagnosis
Temperature Compensation
Sensor Self-Diagnosis
Power Supply &Battery Power Monitoring
Multi-Functional
WISNMulti-Channel
Sensing
($300)
($300)
Intelligent Damage Diagnosis Using Neural Network
• It is important to select a proper frequency range sensitive to the expected damage type.
• The sensitive frequency range varies to the type of the damage.
• Damage type and severity are to be determined.
Correlation Coefficients at Multiple Frequency Ranges
Loose Bolts Cracks1
0.95
0.90
0.80
0.75
0.70
Loose Bolt #1
Loose Bolts #1&2
Cut #1
Cut #2
1
0.95
0.90
0.80
0.75
0.70
Unknown
Health Status
of Structures
Damage Classification &
Quantification
Damage
Features (CC)
for Sub-freq. Ranges
Classify
DamagesTraining
NN
Data Base
FR1 FR2 FR3 FR4 FR5 FR6 FR7 FR8 FR9
Full Frequency Range(10-100 kHz)
CC = 0.9911
(2 Cuts)
Impedance Signals
26
Multiple Damage Diagnosis on Bridge
0 50 100 150 200 250 3000.3
0.4
0.5
0.6
0.7
0.8
0.9
1
290 Training Patterns
(5 Cases)
Loose Bolt and Crack Detection on a Bridge (Korea Expressway Corp.)
0 10 20 30 40 50 60 700.4
0.5
0.6
0.7
0.8
0.9
1
No. of Samples
Te
mp
era
ture
Co
mp
en
sa
ted
C.C
.
Loose Bolt #1
Retighten Bolt #1
Cut #1
Cut #1 & Loose Bolt #1
Cut #1 & Loose Bolt #1&2
Cut #1&2 & Loose Bolt #1&2
Single Freq. Range (45-50 kHz) Multiple Freq. Ranges (40-80 kHz)
PZT
(PIC151 Type,
50*50*1 mm)
Thermo-coupler
1 2 3 4 5 6 70
0.5
1
1.5
2
Bolt(Real)
Bolt(Estimated)
Crack(Real)
Crack(Estimated)
Excluded Cases
in NN Training
NN Verification : 7 Test Cases
Damage Type & Severity
CC Values in 8 Frequency Ranges
DamagesOutput 1
TypeOutput 2Severity
No Damage [0 0] [0 0]
Loose Bolt Only [1 0] [N 0]
Crack Only [0 1] [0 ƩW/L]
Multiple Damages(Loose Bolt and Crack)
[1 1] [N ƩW/L]
* N : No. of Loose Bolts; * W/L : Normalized Crack Length
Training
for NN
Autonomous Frequency Range
Selection and Damage Diagnosis
Damage Cases #3 & #6
were excluded
from NN training
27Structural Dynamics Laboratory, Dept. of Civil & Environmental Engineering
SHM System for Han River Railroad Bridge, Seoul, Korea(Demonstration Project)
12 Imote2 Wireless Sensors10 Wireless Impedance Sensors
Antenna
Solar Panel
WISN
PZT
20 FBG Strain Sensors
Solar Panel
Antenna
Sensor Node
16 Accelerometers
28
Wireless power and
data transmission
Wireless node for power and data transmission
PZT transducerLaser Diode
Photodiode
Active Sensor
Signal Generator Laser Diode
Signal Analyzer Photodiode
Power Demand: 40-80mW Power Demand: 600mW
Make the sensor node as “simple” as possible
Microprocessor
RF Transmitter
A/D Converter
Wave Generator
Memory
D/A Converter
Active Sensor
Battery
Microprocessor
RF Transmitter
A/D Converter
Memory
Passive Sensor
Battery
III. Wireless Power/Data Transmission for Guided Wave Sensing
(HJ Park, H Sohn, CB Yun at KAIST)
29
Wireless Power
Transmission
Wireless Data
Transmission
High Power Laser
TransmissionDriver
SignalGenerator
PhotodiodeOscilloscope
PhotodiodeReceiving
Driver
PhotovoltaicPanel
LaserDiode
TransmissionDriver
LaserDiode
ReceivingDriver
PZT
Schematics of Wireless Guided Wave Generation and Sensing System
1 i
p r o
V (ω )Z
( C C ) V (ω )
Impedance measurement using a simple self-sensing circuit.
Oscilloscope
NI-DAQ PXI DC-to-DC converter
Rx driver
Laser diode
Tx driver
Photodiode
PZT transducer
30
1 1.05 1.1 1.15 1.2 1.25 1.3 1.35 1.4 1.45 1.5
x 104
-0.2
0
0.2
0.4
0.6
0.8
1
1.2
Frequency (Hz)
Real o
f N
orm
alized
E/M
Im
ped
an
ce
Impedance analyzer
Laser-based wireless system
Results of Wireless E/M Impedance Sensing
30 – 35 KHz
* Maximum amplitudes have been normalized for comparison.
2 2.05 2.1 2.15 2.2 2.25 2.3 2.35 2.4 2.45 2.5
x 104
-0.2
0
0.2
0.4
0.6
0.8
1
1.2
Frequency (Hz)
Real o
f N
orm
alize
d E
/M Im
ped
an
ce
Impedance analyzer
Laser-based wireless system
20 – 25 KHz
10 – 15 KHz
3 3.05 3.1 3.15 3.2 3.25 3.3 3.35 3.4 3.45 3.5
x 104
-0.4
-0.2
0
0.2
0.4
0.6
0.8
1
1.2
Frequency (Hz)
Real o
f N
orm
alize
d E
/M Im
ped
an
ce
Impedance analyzer
Laser-based wireless system
1.31 1.315 1.32 1.325
x 104
-0.2
0
0.2
0.4
0.6
0.8
1
2.275 2.28 2.285 2.29 2.295 2.3
x 104
-0.2
-0.1
0
0.1
0.2
0.3
0.4
0.5
3.055 3.06 3.065 3.07 3.075 3.08 3.085 3.09
x 104
-0.2
0
0.2
0.4
0.6
0.8
1
Recent research and application activities on the vibration-based
SHM and innovative NDE techniques for civil infrastructures at
KAIST and UCI are introduced:
Vibration-based SHM
Innovative NDE Techniques
- Long-term SHM systems for large civil infrastructures
- Large-scale wireless sensor network system on a cable stayed bridge
- Kalman filtering technique for damage identification under earthquake
- Innovative sensing: OFS, Microwaves, Vision-based sensing
- Robotic trenchless rehabilitation of underground pipes
- Multifunctional wireless impedance sensor node for steel structure
- Wireless power/data transmission for guided wave sensing
Best wishes to Prof. Shinozuka and his family
in Many Years to Come!!
Merry Christmas and Happy New Yearto All the Participants!!