Beruflich Dokumente
Kultur Dokumente
com/
Vibration and Control
Published by:
http://www.sagepublications.com
Additional services and information for Journal of Vibration and Control can be found at:
Email Alerts: http://jvc.sagepub.com/cgi/alerts
Subscriptions: http://jvc.sagepub.com/subscriptions
Reprints: http://www.sagepub.com/journalsReprints.nav
Permissions: http://www.sagepub.com/journalsPermissions.nav
Citations: http://jvc.sagepub.com/content/early/2014/11/13/1077546314557689.refs.html
(JVC)
[114]
[PREPRINTER stage]
Article
Abstract
This paper presents a two-step approach based on modal strain energy and response sensitivity analysis to identify the
local damages in isotropic plates with moderate thickness. The first step focuses on detection of damage location.
The local damage is simulated by a reduction in the elemental Youngs modulus of the plate. It is determined from the
modal strain energy change ratio approach. A method to weaken the vicinity effect is proposed to reduce the false
alarms in the localization of damage. In the second step, an approach based on response sensitivity-based finite element
model updating is used to further identify the locations and extents of the local damages in time domain. The identified
results are obtained iteratively with Tikhonov regularization using the measured structural dynamic responses. Two
numerical examples are investigated to illustrate the correctness and efficiency of the proposed method. Both single and
multiple damages can be identified successfully and the effect of measurement noise on the identification
results is investigated. Good identified results can be obtained from the short time histories of a few number of
measurement points.
Keywords
Damage identification, plate; modal strain energy change, response sensitivity, model updating, time domain
1. Introduction
As an important type of structural component, plate
appears in a broad range of engineering applications,
including aerospace, automotive, civil and mechanical
engineering. Development of an early damage detection
method for plates is of great importance in maintaining
the integrity and safety of the whole structures.
Techniques based on vibration have been widely
developed for structural damage identication and
health monitoring in the last few decades. By examining
changes in the dynamic properties of a structure,
Doebling et al. (1998) brought about a comprehensive
review of the damage detection methods. Zou et al.
(2000) reviewed the progress on structural condition
monitoring and damage identication for composite
structures. Worden et al. (2008) summarized that a
review of nonlinear dynamics applications to structural
health monitoring.
Generally speaking, damage detection requires a
mathematical model of the structure in conjunction
with experimental modal parameters of the structure.
The identication approaches are mainly based on the
following aspects: the changes in the natural frequencies (Cawley and Adams, 1979; Khiem and Lien, 2004),
mode shapes and their derivatives (Pandey et al., 1991;
Stubbs and Kim, 1996), measured dynamic exibility
(Pandey and Biswas, 1994; Doebling et al., 1996;
Jaishi and Ren, 2006), or frequency response function
(Liu et al., 2009; Huang et al., 2012). Shi et al. (1998)
developed the modal strain energy change ratio-based
damage detection algorithm which examples that the
ratio is the signicant indicator to damage location.
Later, an improvement was made by Shi et al. (2002)
based on the elemental modal strain energy change
before and after the occurrence of damage in a structure. The algorithm includes the analytical stiness and
(JVC)
[114]
[PREPRINTER stage]
mass matrices of the system in the damage quantication. It reduces signicantly the modal truncation error
and the nite-element modeling error from higher analytical modes in the computation, and thus improves
the convergence properties.
In addition to approaches in frequency domain,
there have been large amount of studies on nondestructive methods for structural damage detection
in time domain. Cattarius and Inman (1997) made
use of the time histories of vibration response of the
structure to identify damage in smart structures. Lu
and Law (2007a) proposed a structural damage identication approach based on response sensitivity analysis
in time domain. Research focusing on some other properties of structural vibration provided an impressive
view on damage detection. An and Ou (2014) presented
a signal energy change-based damage localization
approach for beam structures using the accelerations
on measured nodes.
Most of the techniques for damage identication
mentioned above are related to one-dimensional (1D)
structures, for example, beams and trusses, etc. While
the research on two-dimensional (2D) plate-like structures is relatively fewer in the literature concerning
damage identication. Cawley and Adams (1979) developed a damage detection algorithm for plate structures
from frequency shifts. Cornwell et al. (1999) extended
the modal energy method to detect damage in plate
structures. Li et al. (2002) presented a strain mode technique for damage identication in plate-like structures.
Yam et al. (2002) conducted a sensitivity analysis on
static and dynamic response parameters for damage
identication in plate-like structures. Basing on changes
in uniform load surface, Wu and Law (2005) presented
damage identication in plate structures. Yoon et al.
(2005) extended the gapped-smoothing method to
identify damages in 2D plate-like structures, which
was originally developed for damage detection in 1D
structures. Bayissa and Haritos (2007) presented a new
damage-sensitive parameter which was based on bending moment response power spectral density for
damage identication in 2D plate-like structures.
Coppotelli et al. (2007) identied structural damage
using a sensitivity approach with output-only data.
Qiao et al. (2008) developed a new combined static/
dynamic technique for improved damage detection of
laminated composite plates. The promise of the technique is that the abnormality of dynamic response due
to damage may become more pronounced and easier to
be detected under the sustaining static load. Fan and
Qiao (2009) presented a 2D continuous wavelet transform-based damage detection algorithm using
Dergauss 2 d wavelet for plate-type structures.
Kazemi et al. (2010) proposed a two-stage procedure
to localize various faults and their corresponding
(JVC)
[114]
[PREPRINTER stage]
Fu et al.
where K and M are the system stiness and mass matrices of the plate, li and (i the ith eigenvalue and eigenvector, respectively.
The modal strain energy of intact and damaged
structure of jth element in ith mode can be dened as
MSEij (Ti Kj (i ,
3a; 3b
MSEdij MSEij
MSEij
m
1X
MSECRij
m i1
Shi et al. (1998) have proved that the following properties do exist in MSECR which can be shown in
Figure 1. Suppose there is only one damage in the
(JVC)
[114]
[PREPRINTER stage]
where @d=@E
i , @d=@Ei , @d=@Ei are the acceleration, velocity and displacement sensitivities concerning the
Youngs modulus of the ith element. As the global stiness matrix K is the function of the damage parameter
Ei , we can obtain the partial derivative @K=@Ei directly.
The fourth and fth terms in equation (8) can be moved
@d=@E
@K _ @K
d
d
@Ei
@Ei
@d
@d_
@d
@K _ @K
C
K
2
d0
d
@Ei
@Ei
@Ei
@Ei
@Ei
min ga
l X
nt
T
1X
^ ij Rij
^ ij Rij W R
R
2 j1 i1
10
11
[114]
[PREPRINTER stage]
(JVC)
Fu et al.
12
@d1 ti
6 @E1
6 @d2 ti
6 @E
6 1
Stti
6 .
6 .
6
6 .
6 @di ti
6 @E1
6
6 .
6 ..
4
@dl ti
@E1
@d1 ti
@E2
@d2 ti
@E2
..
.
@di ti
@E2
..
.
@dl ti
@E2
@d1 ti
@Ei
@d2 ti
@Ei
..
.
@di ti
@Ei
..
.
@dl ti
@Ei
Ek1 Ek "Ek
@d1 ti
@EN 7
@d2 ti 7
7
@EN 7
.. 7
7
. 7
7
@di ti 7
@EN 7
7
.. 7
. 7
5
14
@dl ti
@EN
15
16
17
13
18
3. Numerical simulations
3.1. A cantilevered plate
Several studies related to damage identication for a
steel cantilevered plate are done in this numerical example. The dimensions of the plate under study are
1000 mm 1000 mm 60 mm as shown in Figure 3.
The physical material properties of the plate are:
Youngs modulus E 210 GPa, mass density
7:8 103 kg=m3 and Poissons ratio 0:3. The
nite element model of the plate was established by
employing the MATLAB software package. The plate
was discretized into 100 four-node Reissner-Mindlin
plate elements. The rst six natural frequencies of the
plate are 10.0, 325.92, 783.03, 2004.61, 2530.02 and
2851.78 Hz, respectively. An impulsive force acts at
the 121st node of the plate in the global z direction with
Ft 103 t 0:02N
0:02s 5 t 0:04s
Ft 103 0:06 tN
0:04s 5 t 0:06s
(JVC)
[114]
[PREPRINTER stage]
Figure 3. A cantilever steel plate ((1), (2), . . . , (121) denote node number of FEM; 1,2, . . . ,100 denote element number).
Interval
The selected
value of !
Figure of the
weakening result
1
2
3
4
00.30
0.300.60
0.600.90
0.901.00
0.15
0.45
0.75
0.95
Figure
Figure
Figure
Figure
4
5
6
7
(JVC)
[114]
[PREPRINTER stage]
Fu et al.
treatment. One can nd the 46th element has been identied as a damaged one from the improved MSECR. In
the second step response sensitivity-based approach is
used to identify the extent of the damaged element. As
the damage location has been determined in the rst
step, we only need the elemental Youngs modulus of
the damaged element. As shown in Table 2, three acceleration measurements adjacent to the 46th element are
used in the identication. The identied result converged after 3 iterations and is shown in Figure 10.
The optimal regularization parameter lopt is calculated
to be 9:5753 1017 . A comparison is made for the
computational time needed when using the proposed
method and the original response sensitivity method
(JVC)
[114]
[PREPRINTER stage]
Proposed method
Original response
sensitivity method
1
6
13 minutes 53 seconds
58 minutes 16 seconds
Measurement points
Cases
Cases
Cases
Cases
2
3
46
78
(JVC)
[114]
[PREPRINTER stage]
Fu et al.
(JVC)
[114]
[PREPRINTER stage]
10
19
20
where ij and ij are the mode shape components of the
jth mode at ith degree of freedom with noise and without noise, respectively; d^ is the vectors of measured
structural acceleration response; Ep is the noise level;
Noise is standard normal distribution
vector with zero
mean and unit standard deviation; j max is the largest
absolute value of component in the jth mode shape;
var is the variance of the time history.
The last case is re-examined and 2% noise levels are
included to the calculated modeshapes and acceleration
responses to simulate the measured modeshapes and
accelerations. Figures 18 and 19 show the damage
localization results with noise from the MSECR
method and the improved MSECR method. When
the measurement noise is included, there will be some
false alarms in the damage localization.
It should be pointed out that from Figure 19, one
can nd that elements 20 and 28 are also suspicious
damaged, thus in the model updating procedure, the
stiness parameters of these two elements are also
included. Figure 20 shows the identied results with
noisy measurements. The identied results converged
after ve iterations with a max identied error 1.81%
in the 34th element with the optimal regularization parameter lopt equal to 3:5760 1016 . This shows the
robustness of the proposed method.
Ft 105 t 0:02N
Ft 105 0:06 tN
0:02s 5 t 0:04s
:
0:04s 5 t 0:06s
Ft 105 t 1:02N
5
Ft 10 1:06 tN
1:02s 5 t 1:04s
1:04s 5 t 1:06s
(JVC)
[114]
[PREPRINTER stage]
Fu et al.
11
Figure 17. Identification results of multiple local damages for each iteration (noise free).
(JVC)
[114]
[PREPRINTER stage]
12
in the model updating procedure, the stiness parameters of these four elements are also included.
The nal identied results for damage quantication
are obtained after six iterations with a max identied
Figure 21. Sketch of a two-span plate ((1), (2), . . . , (231) denote node number of the FEM; 1,2, . . . ,200 denote element number)
(Dimensions not scaled).
(JVC)
[114]
[PREPRINTER stage]
Fu et al.
13
4. Conclusions
Funding
error 1.32% in the 124st element with the optimal regularization parameter lopt is found to be 9:5344 1012 .
Figure 27 shows even with 2% noise level, all the local
damages have been identied with good accuracy.
References
An Y and Ou J (2014) A signal energy change-based damage
localization approach for beam structures. Measurement
48: 208219.
(JVC)
[114]
[PREPRINTER stage]
14