Вы находитесь на странице: 1из 12

Effect of Damage on Modal Parameters Using Full-Scale Test Data

Babak Moaveni1, Xianfei He1, Joel P. Conte1, and Eric Udd2


1
Department of Structural Engineering, University of California, San Diego
2
Blue Road Research, Gresham, Oregon

ABSTRACT

Structural damage may be defined as a weakening of the structure that negatively affects its planned performance
or as any deviation in the structure’s original geometric or material properties that may cause undesirable structural
response to static and/or dynamic loads. The fact that damage induces changes in the dynamic characteristics of a
structure has been widely exploited for damage assessment of structures. This study leveraged a full-scale sub-
component experimental research program conducted in the Charles Lee Powell Structural Research Laboratories
at the University of California, San Diego (UCSD). This full-scale experiment allowed to generate, as a payload
project, an excellent set of vibration response data of a full-scale composite beam at different damage levels. An
Eigensystem Realization Algorithm (ERA) was applied to identify the modal parameters (natural frequencies, mode
shapes, damping ratios, and macro-strain mode shapes) of the composite beam based on its impulse responses
recorded by accelerometers and long-gage fiber Bragg grating (FBG) strain sensors in its undamaged and dam-
aged states. These identified modal parameters are presented and compared at different levels of damage, so as
to investigate their sensitivity to actual structural damage. The results obtained in this study can be used as bench-
mark for developing or validating damage detection algorithms which represent a crucial component of vibration-
based structural health monitoring systems for civil structures.

Introduction

In recent years, structural health monitoring is receiving increasing attention in the civil engineering research com-
munity as a potential tool to detect damage at the earliest possible stage and evaluate the remaining useful life of
structures. Civil structures may be damaged due to natural and man-made hazards such as earthquakes, hurri-
canes, and explosions. Also under service load conditions, they undergo progressive damage in the form of ageing
and deterioration due to environmental conditions. The basic idea of vibration-based, non-destructive damage
identification is to measure the changes in dynamic characteristics of a structure during its lifetime and use them as
a basis for identifying structural damage. Damage identification consists of detecting the occurrence of damage,
localizing the damage area(s), and estimating the extent of damage. Standard damage identification procedures
involve conducting repeated vibration surveys on the structure during its lifetime.

Experimental modal analysis has been explored as a technology for identifying mechanical characteristics as well
as condition assessment and damage identification of structures. If it is properly designed and executed, the
dynamic properties (natural frequencies, damping ratios and mode shapes) of a structure can be measured within
the frequency band of interest and with a reasonably fine spatial resolution. Experimental modal analysis has been
widely used for system and damage identification of bridges [1-8]. Extensive literature reviews were provided by
Doebling et al. [9-10] and Sohn et al. [11] on damage identification, based on changes in modal characteristics.
Sensitivity of measured modal properties to potential damage has been studied in reference [12]. It should be indi-
cated that the success of condition assessment or damage detection based on experimental modal analysis
depends strongly on the accuracy and completeness of the identified structural dynamic properties. Any damage
identification algorithm or method can only be as good as the estimated modal properties of the system consid-
ered. If the uncertainty of the identified modal properties (due to estimation errors or changes in environmental/
operational conditions) is of the same order as the changes in these properties due to damage, there is little value
in attempting to use experimental modal analysis for condition assessment or damage detection.

The study presented in this paper leveraged a full-scale sub-component experimental research program conducted
in the Charles Lee Powell Structural Research Laboratories at the University of California, San Diego (UCSD). This
full-scale experiment allowed the authors to generate, as a payload project, an excellent set of vibration response
data of a full-scale composite beam at different damage levels. The objective of this experimental program was to
validate the design of a concrete-filled composite beam component of the planned I-5/Gilman Advanced Technol-
ogy Bridge [13]. For this purpose, quasi-static uni-directional cyclic load tests (i.e., load-unload cycles) of increas-
ing amplitude were applied to the beam, gradually introducing damage in the beam. After each of several
sequences of quasi-static loading-unloading cycles, a set of dynamic tests was performed in order to investigate
the changes in dynamic characteristics (extracted from the vibration response data) as a function of structural dam-
age. Two different sources of dynamic excitation were used: (1) a computer-controlled electro-dynamic shaker, and
(2) an impact hammer. The vibration data obtained from the impact tests revealed to be the most informative to
identify the beam modal parameters with increasing level of damage. This small-strain vibration response data was
obtained at several damage levels using a set of four long-gage (1m) fiber Bragg grating (FBG) strain sensors
complemented by a set of eight single channel piezoelectric accelerometers.

An Eigensystem Realization Algorithm (ERA) [14] was applied to identify the modal parameters (natural frequen-
cies, mode shapes, damping ratios, macro-strain mode shapes) of the composite beam based on its impulse
responses (free vibration responses) recorded by accelerometers and FBG strain sensors in its undamaged and
damaged states. These identified modal parameters are presented and compared at different levels of damage, so
as to investigate their sensitivity to actual structural damage. These results can be used as benchmark for develop-
ing or validating damage identification algorithms which represent a crucial component of vibration-based struc-
tural health monitoring systems for civil structures.

Composite Beam Experiment

The designed I-5/Gilman Advanced Technology Bridge is a 137m (450ft) long cable-stayed bridge supported by a
59m (193ft) high A-frame pylon, utilizing fiber reinforced polymer composite materials [13]. The bridge system is a
dual plane, asymmetric cable-stayed design as shown in Figure 1. Before the I-5/Gilman Advanced Technology
Bridge can be constructed, a validation test program to evaluate the performance of the bridge was performed. The
prototype test program, which was conducted at the Charles Lee Powell Structural Research Laboratories at
UCSD, evaluates the manufactured fiber reinforced polymer (FRP) components at the material level, through cou-
pon testing, and at the element level on full-scale component, connection and system tests. The test considered
here was conducted on a full-scale longitudinal girder of the bridge (test L2).

Fig. 1 Elevation of I-5/Gilman Advanced Technology Bridge [13]


The longitudinal girders for the I-5/Gilman Advanced Technology Bridge consist of prefabricated carbon/epoxy
shells filled with concrete. These girder shells are 0.91m (3ft) in diameter, 9.75m (32ft) long, and are spliced
together with mild steel reinforcement such that ductile connection and system behavior can be achieved. In the
FRP shell, two rows of 51mm (2 in) diameter holes are drilled along the top edge of the girder and shear stirrups
are embedded in the concrete core to provide interfacial shear resistance between the girder and the deck. The
longitudinal girder test program was composed of two phases. The first test phase consisted of two stages of test-
ing, which were performed on a single longitudinal girder shell specimen. The first stage (test L1a) tested a hollow
FRP shell where only the ends of the girder were grouted with concrete to support the specimen. Once the initial
stiffness test was completed, the girder was fully grouted with concrete and a failure test of the steel-free, concrete-
filled girder was conducted (test L1b). For the second phase of the longitudinal girder test program (test L2), which
is considered in this study, a second girder specimen was cut into two equal halves, spliced together with mild steel
reinforcement, and filled with concrete. A schematic of the test setup is provided in Figure 2.

9.75m (32')
8.53m (28')

0.61m (2') Gap 76 mm (3") 1.22 m (4')


0.254 m (10")

0.91m (3')

Spiral Ø10" #4 1.22 m (4')


Pipe 0.349 m (13.75")
3.66 m (12')

Strong Floor
Roller Fixture Pin Fixture

2.74m (9') 3.05 m (10') 2.74m (9')

Fig. 2 Schematic test setup for L2 girder test: Side elevation [15]

The quasi-static cyclic loading was applied to the girder using four 1335 KN (300 kips) displacement-controlled
hydraulic jacks in a four-point bending test (Figure 2). The increasing level of load during the quasi-static cyclic
loading progressively introduces damage in the beam through inelastic (irreversible) deformations. The total load
versus the girder mid-span displacement is shown in Figure 3. After certain load cycles (2, 5, 7, 8, 12), the loading
fixtures were removed to perform a sequence of dynamic tests to be used in system identification and damage
detection studies. This sequence of dynamic tests was performed twice before starting the quasi-static loading
cycles and the corresponding undamaged states of the beam are referred to as damage states S0 and S1. Dam-
age states S2 to S6 refer to the state of the beam after loading cycles 2, 5, 7, 8 and 12, respectively, shown in Fig-
ure 3. The repeated sequence of dynamic tests consisted of a set of forced vibration tests using a 50 lbs force
electro-dynamic shaker followed by a set of impact (free vibration) tests using an impact hammer with integrated
load cell recording the applied force. The forced vibration tests using the shaker consisted of a set of sixteen
(Gaussian) white noise excitations followed by three (linear) sine sweeps in the frequency ranges [12-22 Hz], [38-
48 Hz], and [93-103 Hz]. The free vibration tests using the impact hammer consisted of three vertical impact tests
at each of four locations along the top edge of the girder as shown in Figure 4. Therefore, a total of 12 vertical
impact tests was performed on the beam for each of 7 different damage states S0-S6. The modal identification
results presented herein are based only on the data generated from these vertical impact tests.
4000

Second hole failure S6


3500 S5
First hole failure
3000

S4
2500
S3

Total Load (kN)


2000

1500
S2
1000

500

0
0 50 100 150 200 250 300

Deflection (mm)

Fig. 3 Total load vs. displacement at mid-span [15]

FBG#4

FBG#1 FBG#2 FBG#3

Fig. 4 Locations of accelerometers, FBG sensors, vertical hammer impacts and electro-dynamic shaker

The girder was instrumented with strain gages, linear potentiometers, and inclinometers for the whole duration of
the quasi-static cyclic test. In addition, the girder was instrumented with four long-gage FBG strain sensors and
eight accelerometers for the purpose of the payload dynamic tests. The FBG strain sensors were surface mounted
utilizing brackets with a pair of sensors located on the top and bottom of the beam at mid-span, and the remaining
two sensors located along the bottom edge of the beam at approximately 1/3 the length on either side of mid-span
as shown in Figure 4. The sensors are pre-strained in their long-gage packages in order to measure compression
as well as tension. In addition to the FBG strain sensors, eight accelerometers were attached to the structure mea-
suring the vertical acceleration. Seven of these accelerometers were approximately equally spaced along the bot-
tom edge of the girder and one was mounted on the moving mass of the electro-dynamic shaker so as to measure
the dynamic force applied to the beam (see Figure 4). For every dynamic test performed at each of the damage
states S0 through S6, the vibration response of the composite girder was measured by accelerometers a2 through
a8, while the four FBG strain sensors measured the vibration response for damage states S1 through S5.
Identification of Modal Parameters

An Eigensystem Realization Algorithm (ERA) followed by a Least Square Optimization [16] was employed for iden-
tifying the modal parameters (natural frequencies, mode shapes, damping ratios, and macro-strain mode shapes)
of the composite beam in its undamaged and damaged states. The identified modal parameters using ERA are
based on the impact test (free vibration) data recorded by the accelerometers and FBG strain sensors. A total of 12
vertical impact tests was performed on the beam for each of the 7 damage states S0-S6, with damage states S0
and S1 representing the beam in its undamaged condition. Two sets of identification results are presented herein
for the accelerometer and the FBG sensor data: (1) from applying ERA to one impact test at a time, and (2) from
applying ERA to 12 impact tests in one identification.

Figure 5 shows the identified natural frequencies of the first 5 modes using ERA based on accelerometer data for
each of 12 × 7 = 84 impact tests (each circle corresponds to an identified frequency from one impact test). In
each ERA realization, a Hankel matrix of size ( 7 × 250 ) × 250 was constructed based on the impulse response
data sampled at 512Hz. Then, after performing a singular value decomposition, a system of order n = 16 was real-
ized, from which a maximum of 8 physical modes of vibration could be extracted. The following observations can
be made from Figure 5: (1) at each damage state, the identified modal frequencies from all 12 different impact tests
are generally in close agreement. If an identified modal frequency is not consistent with the others, it could be due
to the fact that the corresponding vibration mode is not significantly excited (resulting in a low signal-to-noise ratio).
For example, the 4th, 5th and 6th values of the identified second mode frequency exhibit some discrepancy for dam-
age states S0, S1 and S6, due to the fact that the 4th, 5th and 6th impact tests are applied at the second vertical
hammer impact location (see Figure 4) which is close to a node of the second vibration mode. (2) As expected, the
identified natural frequencies for damage state S0 and S1 are almost identical, since both sets of results corre-
spond to the undamaged state of the beam. (3) With increasing level of damage in the beam, the identified modal
frequencies decrease (with an exception for the 5th mode at damage state S2), consistent with the stiffness degra-
dation due to damage. The statistics of these identified modal frequencies (each based on 12 sample data points)
are reported in Table 1.

18
16 mode 1
14
S0 S1 S2 S3 S4 S5 S6
60
mode 2
40
20
Frequency [Hz]

S0 S1 S2 S3 S4 S5 S6
100 mode 3
80
S0 S1 S2 S3 S4 S5 S6
180
mode 4
160
140
S0 S1 S2 S3 S4 S5 S6
300
mode 5
250
200
S0 S1 S2 S3 S4 S5 S6
Damage State
Fig. 5 Identified natural frequencies of the first 5 modes applying ERA to 84 different impact
tests from accelerometers (with ERA applied one test at a time)
Table 1: Mean / coefficient of variation [%] of the identified frequencies [Hz] from ERA for damage states S0 to S6

S0 S1 S2 S3 S4 S5 S6
Mode 1 17.35/0.07 17.34/0.10 16.10/0.08 15.37/0.23 15.17/0.06 15.02/0.12 14.09/0.18
Mode 2 42.97/3.53 42.78/3.49 40.93/0.76 39.47/2.24 38.93/0.52 38.58/0.92 35.68/17.17
Mode 3 97.58/0.12 97.49/0.13 89.26/0.11 86.82/0.15 84.90/0.22 84.05/0.17 79.91/0.18
Mode 4 167.67/0.03 167.72/0.04 158.87/0.06 153.64/0.26 143.23/0.23 143.21/0.07 142.81/0.12
Mode 5 246.42/0.26 246.22/0.25 256.92/0.66 244.85/0.08 232.39/0.08 228.35/0.20 219.80/0.09

Another system realization is performed based on the same free vibration acceleration data, but considering all
twelve impact tests (at each damage state) in a single identification. The basic idea behind this identification strat-
egy is to use the information from all impact tests to identify each modal frequency. Thus, if a single test does not
contain significant information about a given modal frequency (for example due to low modal participation), the lat-
ter can still be identified well through other impact tests that are more informative about this mode. To do so, ERA
is applied in its multiple input, multiple output (MIMO) formulation, but instead of forming the Hankel matrix based
on free vibration data from a truly multiple input test, the block Hankel matrix is formed by including the responses
from (r) single input impact tests as shown in the equation below.

1 2 r 1 r 1 r
g (1) g (1) … g (1) g (2) … g (2) … g (N) … g (N)
1 2 r 1 r 1 r
H = g (2) g (2) … g (2) g (3) … g (3) … g (N + 1) … g (N + 1) (1)
… … … … … … … … … … …
g ( N ) g ( N ) … g ( N ) g ( N + 1 ) … g ( N + 1 ) … g ( 2N – 1 ) … g ( 2N – 1 )
1 2 r 1 r 1 r

i
where g ( k ) denotes the impulse response vector (at time t = k ⋅ ∆T ) from the ith impact test. In the present
case, the block Hankel matrix was built including the data from all (r = 12) impact tests at each damage state. A
model of order n = 16 (to extract a maximum of 8 physical modes of vibration) was fit to the data, and then the
modal parameters were extracted from the realized state-space model. The identified modal parameters (natural
frequencies and damping ratios) are reported in Table 2, while Figure 6(a) shows the mode shapes (real part of
complex modal vectors) corresponding to damage state S1 (undamaged state of the beam). In the latter plot, the
circles correspond to the identified mode shape at the sensor locations and the dashed lines represent cubic fits
through these circles. In order to validate and better understand the modal parameters identified from the experi-
mental data, a three-dimensional finite element model of the composite girder was created in SAP2000 [17]. The
mode shapes and natural frequencies obtained from this finite element model are presented in Figure 6(b). It can
be seen that the analytical and identified (from acceleration data) natural frequencies and mode shapes are in
good agreement for the undamaged state of the beam. It is worth noting that, due to flexibility of the beam sup-
ports, the mode shapes at both supports are not zero.

Table 2: Identified natural frequencies and damping ratios based on a single identification including all impact test acceleration
data for each damage state S0 to S6

Mode 1 Mode 2 Mode 3 Mode 4 Mode 5


Frequency Damping Frequency Damping Frequency Damping Frequency Damping Frequency Damping
[Hz] Ratio [%] [Hz] Ratio [%] [Hz] Ratio [%] [Hz] Ratio [%] [Hz] Ratio [%]
S0 17.35 1.51 43.48 4.25 97.57 1.69 167.70 0.84 246.63 0.42
S1 17.34 1.49 43.24 4.57 97.50 1.68 167.74 0.83 246.16 0.49
Table 2: Identified natural frequencies and damping ratios based on a single identification including all impact test acceleration
data for each damage state S0 to S6

S2 16.09 1.65 40.72 2.68 89.24 1.99 158.90 0.92 254.51 1.14
S3 15.35 1.71 39.09 2.04 86.80 1.71 153.32 1.84 244.72 0.62
S4 15.17 1.25 38.91 2.21 84.82 2.21 149.29 2.62 232.25 0.76
S5 15.00 1.28 38.98 3.19 84.12 1.88 143.20 0.93 227.94 0.88
S6 14.08 1.43 33.04 2.09 79.93 1.72 142.84 1.42 219.71 0.92

Mode 1 at f = 17.34Hz f1 = 17.1Hz


1.5
1
0.5
0 x
Mode 2 at f = 43.24Hz f2 = 51.6Hz
1
0
−1
Mode 3 at f = 97.50Hz f3 = 97.2Hz
4
0
−4
Mode 4 at f = 167.74Hz
1 f4 = 158.9Hz
0
−1

Mode 5 at f = 246.15Hz
1 f5 = 244Hz
0
−1
0 50 100 150 200 250 300
x [in]
(b)
(a)
Fig. 6 Natural frequencies and mode shapes of the composite girder, (a) identified from the experimental
data and (b) computed from the finite element model

Figure 7 plots the identified complex modal vectors in a polar plot format. This representation indicates whether a
vibration mode is identified as classically damped or not. If the components (each representing an observed DOF)
of a complex modal vector are collinear (i.e., in phase or out of phase), then this mode is classically (or proportion-
ally) damped. The more the components of a modal vector are scattered in the complex plane, the more this mode
is non-classically damped. From Figure 7, it can be observed that the first four identified modes are classically
damped, while the fifth identified mode is non-classically damped. It is worth mentioning that in the presence of low
signal-to-noise ratio and/or due to identification errors, a truly classically-damped mode could be identified as non-
classically damped, which could be the case here for the fifth mode.

f1 = 17.34Hz f2 = 43.24Hz f3 = 97.50Hz f4 = 167.74Hz f5 = 246.15Hz


90 90 90 90 90
0.5 1 1 1 1

180 0

270 270 270 270 270


Fig. 7 Complex modal vectors in complex plane (polar plot representation) for damage state S1
The mode shapes identified (from a single state-space realization incorporating all twelve impact tests) at increas-
ing level of damage (S0, S2 and S4) are shown in Figure 8 for the first four modes. It can be observed that the
mode shapes do not change dramatically due to the damage. However, for some modes these changes are suffi-
cient to detect damage.

The impulse response data transduced from the FBG strain sensors were also used to identify the modal parame-
ters of the composite beam in this study. Figure 9 plots the natural frequencies identified using ERA based on
impulse response data measured from the four FBG sensors for damage states S1 to S5. As can be observed from
this figure, not all modal frequencies can be identified from each impact test. For example, at damage state S4, the
second and fourth modal frequencies are not identified and also the third mode can only be identified for the first
three and last three of the 12 impact tests (impacts at location one and four in Figure 4). Comparing the identifica-
tion results obtained from accelerometer data and FBG sensor data, it can be concluded that the second mode fre-
quency is better identified from accelerometer data than from FBG sensor data, and the fifth mode frequency
cannot be identified using the FBG sensor data. The natural frequencies and macro-strain modal vectors identified
using all twelve impact test data in a single realization are shown in Figure 10 for damage state S1. The fact that
modes 2 to 4 are identified as non-classically damped could be due to low signal-to-noise ratio at these modal fre-
quencies. These identified natural frequencies are in good agreement with the ones obtained from accelerometer
data (see Figure 7). Figure 11 presents the identified macro-strain mode shapes (normalized to a unit length) for
damage states S1, S2, S3 and S5. It should be noted that the second and fourth macro-strain mode shapes could
not be identified at damage state S3. From this figure, it can be seen that the change in macro-strain mode shape
at FBG sensor #2 (see Figure 4) due to damage is significant beyond damage state S2.

S0
mode 1 S0

0.5 S2 17 mode 1
0.4 S2 16
0.3 S4
15
0.2 S4
S1 S2 S3 S4 S5
mode 2
0.5 42 mode 2
38
0
34
−0.5 S1 S2 S3 S4 S5
Frequency [Hz]

mode3
95 mode 3
0.5
90
0 85

−0.5 S1 S2 S3 S4 S5

mode 4 165
mode 4
0.5 155
0 145
S1 S2 S3 S4 S5
−0.5
Damage State
0 50 100 150 200 250 300
x [in] Fig. 9 Identified natural frequencies of the first 4 modes
Fig. 8 Normalized mode shapes of the beam for using ERA applied to 60 different impact tests (12 for
damage states S0, S2 and S4 each damage state) recorded from FBG sensors
f1 = 17.35Hz f2 = 39.55Hz f3 = 97.32Hz f4 = 167.67Hz
90 90 90 90

180 0

270 270 270 270


Fig. 10 Complex modal macro-strain vectors in complex plane (polar plot representation) for damage state S1

Sensitivity Analysis of Identified Modal Parameters to Damage

In this section, the sensitivity of identified modal parameters such as natural frequencies, mode shapes, and modal
assurance criterion (MAC) values (between corresponding mode shapes at undamaged and damaged states) are
investigated with increasing level of damage in the beam. Based on the force-deformation curve in Figure 3
obtained from the quasi-static tests, the global tangent stiffness of the beam structure is determined at several
points along the envelope curve of the hysteresis loops (i.e., damage levels). The normalized change of stiffness
∆K Si and normalized change of frequency are defined respectively as

k k
K S1 – K Si k f S1 – f Si
∆K Si = -----------------------
-, ∆f Si = ---------------------
- (2)
K S1 k
f S1

where the subscript Si refers to the damage state (e.g., S1, S2, ..., S6), KSi represents the global stiffness at dam-
age state Si, and fSik denotes the identified natural frequency of the kth mode at damage state Si. The normalized
change of stiffness and normalized change of natural frequencies are plotted in Figures 12(a) and 12(b), respec-
tively, versus the damage level (S0-S6). It is observed that with increasing damage, the identified natural frequen-
cies decrease monotonically due to loss of stiffness in the beam, with an exception for the 5th mode at damage
state S2.
Normalized ∆K [%]

1 100
(a)
S1

0
50
−1
1 0
S0 S1 S2 S3 S4 S5 S6
20
S2

0 (b)

−1
Normalized ∆f [%]

1 FBG#1 10
FBG#2
S3

FBG#3
0
FBG#4 Mode1
Mode2
−1 0
Mode3
1 Mode4
Mode5
S5

0
−10
−1 S0 S1 S2 S3 S4 S5 S6
Mode 1 Mode 2 Mode 3 Mode 4
Fig. 11 Macro-strain mode shapes (normalized to a Fig. 12 (a) Normalized change of stiffness and (b)
unit length) for damage states S1, S2, S3 and S5 normalized change of frequency
MAC values are calculated between the identified mode shapes at damaged states the beam (S2-S6) and the cor-
responding mode shape at the undamaged state (S1). MAC values are bounded between 0 and 1 and measure
the degree of correlation between corresponding mode shapes in the undamaged and damaged states (MAC
value of 1 for unchanged mode shapes). Figure 13(a) displays the MAC values for the first four vibration modes at
different damage states. Figure 13(b) shows the change of MAC values due to increasing level of damage,
k
∆MAC = 1 – M AC Si , for the first four modes. From Figure 13(a), it can be seen that the MAC value of the first
mode is the most insensitive to damage, while the MAC value of the second mode is the most sensitive to damage.
It is also observed (from Figures 13(a) and (b)) that the MAC values obtained at damage states S4 and S6 are out-
liers of the trend of the MAC value versus damage state/level.

From the changes of the modal frequencies and MAC values, it is possible to determine the presence of damage,
but it is very difficult (if not impossible) to determine the location and extent of damage, since both of these indica-
tors are global in nature. Figure 14 shows the change of the normalized (to a unit length) mode shapes (identified
using a single realization based on all twelve impact tests) with respect to the undamaged state S1 at each dam-
age state of the beam. On the left side of each subplot, the identified undamaged mode shape is plotted (out of
scale). It can be seen that the mode shapes are sensitive to damage, but it is still not possible to localize and quan-
tify damage by simply looking at these plots. It is necessary to employ more robust and advanced damage identifi-
cation techniques, such as modal strain energy and finite element model updating to identify the damage in the
structure.
g p
1 Mode 1 0.05

S0 0
0.95 S1
S2 −0.05
MAC

S3 0.2
S4
Mode 2

0.9
S5 0
S6
−0.2
0.85
Mode1 Mode2 Mode3 Mode4 0.1
(a)
Mode 3

7 0
6
Mode1
5 −0.1
∆MAC [%]

Mode2
4 0.2
Mode3
3
Mode 4

2 Mode4
0
1
0 −0.2
S0 S1S2 S3 S4 S5 S6 Shape S2 S3 S4 S5 S6
(b)
Fig. 13 (a) MAC values and (b) their changes at Fig. 14 Change of mode shapes due to damage
different damage states

Concluding Remarks

The study presented in this paper leveraged a full-scale experiment conducted in the Charles Lee Powell Structural
Research Laboratories at the University of California, San Diego (UCSD), to validate the design of a concrete-filled
composite beam component of the planned I-5/Gilman Advanced Technology Bridge. During the test, several
sequences of quasi-static unidirectional cyclic loads of increasing amplitude were applied to the beam, gradually
introducing damage in the beam through inelastic (irreversible) deformations. After each of several sequences of
quasi-static loading-unloading cycles, a set of dynamic tests was performed in order to investigate the changes in
dynamic characteristics of the beam as a function of structural damage. The repeated sequence of dynamic tests
included a set of impact tests (performed using an impact hammer) generating free vibration data.

Based on free vibration data measured using accelerometers and long-gage fiber Bragg grating (FBG) strain sen-
sors from different impact tests, an Eigensystem Realization Algorithm (ERA) followed by a Least Square Optimi-
zation was employed for identification of modal parameters (natural frequencies, mode shapes, damping ratios,
and macro-strain mode shapes) of the composite beam in its undamaged and various damaged states. The identi-
fied results from different tests using different types of data (acceleration or macro-strain) show good agreement,
therefore validating the system identification approach used in this research as well as the accuracy of the experi-
mental modal identification results obtained. In addition, a three-dimensional analytical model of the composite
beam (in its undamaged state) was developed in SAP2000, a structural analysis software. The modal parameters
computed from this finite element model also served to further validate and better understand the experimental
modal identification results for the beam in its undamaged state. Finally, the sensitivity of the identified modal
parameters such as natural frequencies, mode shapes, and modal assurance criterion (MAC) values, to increasing
damage is investigated. The study presented herein provides insight into the effects of actual localized damage on
the identified modal parameters of a full-scale structural component, which are useful for ongoing research on
vibration-based damage identification methods for civil structures.

Acknowledgements

This work was partially funded by the National Science Foundation, Grant No. DMI-0131967 under a Blue Road
Research STTR Project on which UCSD was the principal subcontractor. The authors would like to thank the
National Science Foundation for their support of this important research. The authors are also grateful to Professor
Vistasp Karbhari, UCSD, and to Dr. Charles Sikorsky at Caltrans for allowing them to perform, as a payload project
to their full-scale sub-component experiment, the dynamic tests used in this research.

References

[1] Abdel-Ghaffar, A. M., and Housner, G. W. “Ambient vibration tests of suspension bridge.” Journal of the Engi-
neering Mechanics Division, ASCE, Vol. 104, (5), 983-999 (1978).
[2] Farrar, C. R., and James III, G. H. “System identification from ambient vibration measurements on a bridge.”
Journal of Sound and Vibration, 205 (1): 1-18 (1997).
[3] Lus, H., Betti, R., and Longman, R.W. “Identification of Linear Structural Systems Using Earthquake Induced
Vibration Data.” Earthquake Engineering and Structural Dynamics, Vol. 28, pp1449-1467 (1999).
[4] Peeters, J. M. B., and De Roeck, D. “Damage identification on the Z24-Bridge using vibration monitoring.” Pro-
ceedings of the European COST F3 Conference on System Identification and Structural Health Monitoring, Madrid,
Spain, pp. 233–242 (2000).
[5] Cunha Á., Caetano, E., Calçada, R., De Roeck, G., and Peeters, B. “Dynamic measurements on bridges:
design, rehabilitation and monitoring.” Proceeding of the Institution of Civil Engineers, Bridge Engineering I56,
BE3: 135-148 (2003).
[6] Smyth, A. W., Pei, J. S., and Masri, S. F. “System identification of the Vincent Thomas suspension bridge using
earthquake records.” Earthquake Engineering and Structural Dynamics, 32: 339-367 (2003).
[7] Caicedo, J. M., Dyke, S. J., and Johnson, E. A. “Natural excitation technique and eigensystem realization algo-
rithm for phase I of the IASC-ASCE benchmark problem: simulated data.” Journal of Engineering Mechanics,
ASCE, Vol. 130, No. 1: 49-60 (2004).
[8] Catbas, F. N., Brown, D. L., and Aktan, A. E. “Parameter Estimation for Multiple-Input Multiple-Output Modal
Analysis of Large Structures.” Journal of Engineering Mechanics, ASCE, Vol. 130, No. 8: 121-130 (2004).
[9] Doebling, S. W., Farrar, C. R., Prime, M. B., and Shevitz, D. W. “Damage identification in structures and
mechanical systems based on changes in their vibration characteristics: a detailed literature survey.” Los Alamos
National Laboratory Rep. No. LA-13070-MS, Los Alamos, N.M. (1996).
[10] Doebling, S. W., Farrar, C. R., and Prime, M. B. “A Summary Review of Vibration-Based Damage
Identification Methods.” The Shock and Vibration Digest, Vol. 30, No. 2, pp. 99-105 (1998).
[11] Sohn, H., Farrar, C. R., Hemez, F. M., Shunk, D. D., Stinemates, D. W., and Nadler, B. R. “A review of
structural health monitoring literature: 1996-2001.” Los Alamos National Laboratory Report LA-13976-MS
(2003).
[12] Zhu H. P., and Xu, Y. L. “Damage detection of mono-coupled periodic structures based on sensitivity
analysis of modal parameters.” Journal of Sound and Vibration 285: 365–390 (2005).
[13] Seible, F., Hegemier, G. A., Karbhari, V. M., Davol, A., Burgueño, R., Wernli, M., and Zhao, L. “The I-5/
Gilman Advanced Composite Cable Stayed Bridge Study.” University of California, San Diego, SSRP-96/
05, (1996).
[14] Juang J. N. and Pappa R. S. “An eigensystem realization algorithm for model parameter identification
and model reduction.” Journal of Guidance, Control, and Dynamics, 8(5), 620-627 (1985).
[15] Brestel, D., Van Den Einde, Y., Karbhari, V. M., and Seible, F. "Characterization of Concrete Filled
Structural Formwork." Proceeding of the 48th International SAMPE Symposium, Long Beach, CA, Book 2,
pp. 2115-2128, May 11-15, (2003).
[16] De Callafon, R., Moaveni, B., and Conte, J. P. “General Realization Algorithm for Structural Modal
Identification.” In Preparation.
[17] Computers and Structures, Inc. “SAP 2000 Linear and Nonlinear, Static and Dynamic Analysis and
Design of Three-Dimensional Structures: Getting Started (Version 9).” Berkeley, California, USA, (2004).

Вам также может понравиться