USING WAVELETS ON THE RESPONSE OF A BEAM TO A CALIBRATED VEHICLE FOR DAMAGE DETECTION
David Hester
Arturo Gonzalez
Nantes, 2nd July 2009
Overview of presentation
• Rational behind research / Introduction to technique
• Computer models used• Description of Continuous Wavelet
Transform (CWT)• Performance of different wavelets• Approach for detecting small damage
01
Rational behind study
• Research in Structural Health Monitoring (SHM) increasing, typically requires Non Destructive Testing
• Bridges are a particularly interesting set of structures, in service for long period, traffic loads are steadily increasing
• Financially beneficial if service life of existing bridges can be maximised
• Reliable early damage detection technique significant step toward achieving this
02
Introduction to technique
• Ultimately would like to be able to detect damage in a bridge by monitoring it’s dynamic response
• Fundamental principal: Damage causes change in mechanical properties of structure
• Potential to use Wavelets to detect damage in a beam by performing a wavelet transform on the deflection signal of the beam
03Damage Deflection gauge
0 1 2 3 4 5 6 7 8 9 10-3.5
-3
-2.5
-2
-1.5
-1
-0.5
0
0.5x 10
-3
Time
Def
lect
ion
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
-2.5
-2
-1.5
-1
-0.5
0
0.5
1
1.5
2
2.5
x 10-4
Wav
elet
Coe
ffici
ent
Normalised Position of Load
Wavelet
Transform
04
Modelling of the Structural Response to a Moving Load
•Discretized model of a simply supported beam
[Kg], [Mg]
1
2 3
4
L
05
Crack Modelling
•Crack = Loss of Stiffness
Sinha’s method
•Sinha approximates the
exponential curve of
Christides and Barr with
a straight line.
•lc= 1.5d
06
Load ModellingVehicle modelled as a
constant moving Load
[Mg]{d2
y/dt2
}+[Kg]{y}={F}
Introduction to wavelets
• Wavelet transform was developed to extract Time-Frequency information from a signal
• A wavelet is a waveform of limited duration
07
Mexican hatDb 5
Gauss 2 Morlet
Figures taken from MATLAB
Outline of Wavelet Transform1. Wavelet compared to a section at start of
the original signal
2. Calculate wavelet coefficient ‘C’, which represents how closely correlated the wavelet is with this section of the signal.
3. Shift the wavelet to the right and repeat steps 1 & 2
4. Scale (stretch) the wavelet and repeat steps 1 through 3
5. Repeat steps 1-4 for all scales. Result of the WT are many wavelet coefficients ‘C’
08Figures taken from MATLAB
Using wavelets to detect damage
• Wavelets can detect local discontinuities in a signal
• Discontinuity in deflection-time response of a bridge as load passes over cracked section
• Basic principal is to use wavelet to detect the discontinuity in the signal and thereby locate damage
09
Mid-span deflection response of a beam cracked at the 1/3 point subject to moving P-Load (1st beam freq=0.9Hz)
0 0.2 0.4 0.6 0.8 1-4
-3.5
-3
-2.5
-2
-1.5
-1
-0.5
0
0.5x 10
-3
NORMALISED POSITION OF LOAD ON BRIDGE ( x(t)/L )
DE
FLE
CT
ION
AT
MID
-SP
AN
10
CrackDeflection sensor
WT applied to the midspan deflection signal of a beam subject to P-Load. Beam has a crack at 1/3rd point,
Increase in wavelet coefficients at 0.33L
→There is a localised discontinuity in deflection signal at 0.33L
→There is damage at 0.33L
WAVELET COEFFICIENTS AT DIFFERENT SCALES AND POSITIONS
NORMALISED POSITION OF LOAD ON BEAM ( x(t)/L )
SC
ALE
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1 14 27 40 53 66 79 92105118131144157170183196209222235248
11
≈ 0.9HzScale=27
Wavelet
Transform
0 1 2 3 4 5 6 7 8 9 10-3.5
-3
-2.5
-2
-1.5
-1
-0.5
0
0.5x 10
-3
Deflection Signal
Performance of different wavelets in detecting damageDamage
LevelWavelet Possible to detect Damage
(Yes/No)Damage
LevelWavelet Possible to detect Damage
(Yes/No)Pure
signalNoisy signal Pure signal Noisy
signala/h=0.1 db2 No No a/h=0.4 db2 Yes No
sym 2 No No sym 2 Yes Nocorfil 1 No No corfil 1 Yes Nogauss2 No No gauss2 Yes Yesmex hat No No mex hat Yes Yes
a/h=0.2 db2 Yes No a/h=0.5 db2 Yes Yessym 2 Yes No sym 2 Yes Yes
corfil 1 Yes No corfil 1 Yes Yesgauss2 Yes No gauss2 Yes Yesmex hat Yes No mex hat Yes Yes
a/h=0.3 db2 Yes NoDamage Level (a/h) = height of the crack / Depth of the beam
sym 2 Yes Nocorfil 1 Yes Nogauss2 Yes Yesmex hat Yes Yes
Gauss 2 Mex hat
12
Coefficient line plot
13
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1-2
-1.5
-1
-0.5
0
0.5
1
1.5
2
2.5x 10
-3
NORMALISED POSITION OF LOAD ON BEAM ( x(t)/L )
WA
VE
LET
CO
EF
FIC
IEN
T
IDENTIFYING DAMAGE LOCATION FOR DIFFERENT LEVELS OF DAMAGE
Undamaged
delta=0.2delta=0.4
delta=0.6Delta=
Crack height / Beam depth
Delta=0.6
Delta=0.4
Delta=0.2
Damage at 1/3 point
WAVELET COEFFICIENTS AT DIFFERENT SCALES AND POSITIONS
NORMALISED POSITION OF LOAD ON BEAM ( x(t)/L )
SC
ALE
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1 14 27 40 53 66 79 92105118131144157170183196209222235248
14
Improvement by using multiple measurements•
Use of one single measurement, delta=0.2
•Use of multiple measurements
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1-2
-1.5
-1
-0.5
0
0.5
1x 10
-3 IDENTIFYING DAMAGE USING MIDSPAN DEFLECTION SIGNAL
NORMALISED POSITION OF LOAD ON BEAM ( x(t)/L )
WA
VE
LET
CO
FF
ICIE
NT
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
-1
-0.5
0
0.5
1x 10
-3
NORMALISED POSITION OF LOAD ON BEAM ( x(t)/L )
WA
VE
LE
T C
OE
FF
ICIE
NT
IDENTIFING DAMAGE USING A NUMBER OF MEASURING POINTS
1/5L
2/5L
3/5L
4/5L
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
-1
-0.5
0
0.5
1x 10
-3 AVERAGE COEFFICIENTS SEVERAL MEASURING POINTS
NORMALISED POSITION OF LOAD ON BEAM ( x(t)/L )
WA
VE
LET
CO
EF
FIC
IEN
T
Min not at damage location
Min at damage location
Average
Crack
Deflection Sensors
Conclusions
• Possible to use a moving load as a form of non destructive testing to detect damage
• Wavelet transform applied to Deflection-Time response can identify and locate damage
• In presence of noise detects large cracks relatively easily
• Multiple measuring locations give better results when detecting small cracks
Wavelet
Transform0 1 2 3 4 5 6 7 8 9 10
-3.5
-3
-2.5
-2
-1.5
-1
-0.5
0
0.5x 10
-3
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
-2.5
-2
-1.5
-1
-0.5
0
0.5
1
1.5
2
2.5
x 10-4
Locate DamageCrack
Deflection GaugeDeflection Signal
15
Acknowledgements
• This investigation has been carried out as part of work program 7 of the ASSET project, Sustainable Surface Transport
ASSET(Advanced Safety and Driver Support in Efficient Road Transport),
16