STRUCTURAL DAMAGE IDENTIFICATION BY USING
ENERGY TRANSFER RATIO
Zhou-Hong Zong1, Tong-Lo Wang2
1
College of Civil Engineering and Architecture, Fuzhou University, Fuzhou City, Fujian Province,
35002, People’s Republic of China. Email: zy0520@fzu.edu.cn
2
Department of Civil and Environmental Engineering, Florida International University,
Miami, Florida, 33199, USA. Email: wang@eng.fiu.edu
ABSTRACT
The energy transfer ratio (ETR) is referred to as the
ratio of energy transfers to total energies in each mode for
the non-proportional damped system. Although previous
studies have established the possibility that the change of
the energy transfer ratio may be used for damage
assessment of highway bridges, it is still not simple tasks to
extract ETR from the measured signals and to establish a
direct relationship between various damage patterns and
changes of the ETR. In this paper, the damage detection of a
model highway bridge are conducted by using ETR index,
the modal parameters, such as frequencies and damping
ratios, as well as the static data, such as the strain and
deflections. A time-domain process of modal parameter
identification method associated with the ETR is presented.
Based on this new process, the modal parameters, including
the frequencies, damping ratios and ETR, can be extracted.
The dynamic testing of a 1/6-scaled steel-concrete model
bridge with different damage scenarios are carried out to
verify the process. It is demonstrated that the ETR is
sensitive to the bearing damage and steel cracking than
other parameters. The comparisons between the testing and
analysis correlate very well.
1 INTRODUCTION
Most damage detection theories and practices are
formulated based on the following assumption: that failure or
deterioration would primarily affect the stiffness and therefore
affect the modal characteristics of the dynamic response of
the structure. If this kind of changes can be detected and
classified, this measure can be further implemented for a
bridge monitoring system to indicate the condition, or
damage, or remaining capacity of the structures. It can also
be used to evaluate the seismic behavior of the structures.
The damage diagnosis and health monitoring of existing
highway bridges are active areas of research in recent years.
However, conventionally defined modal parameters have
been shown by many that they are not sensitive enough to
detect many types of bridge damage. Furthermore, the
modal parameters of conventional modal testing are global
parameters which can not locate the damages. In civil
engineering applications, it is extremely difficult to measure
the input excitation forces acting on a large-scale structure.
The experimental modal identification through ambient
vibration measurements has become a very attractive
technique nowadays. Ambient excitations such as traffic,
wind, earthquake, micro-vibration of the earth, and their
combination are natural or environmental agents. The
ambient vibration has the advantage of being inexpensive
and convenient since no equipment is needed to excite the
structure. The service needs not have to be interrupted to
use this technique. The experimental modal analysis by
ambient vibrations was successfully applied to many bridges.
Modal testing has some weakness as well. One of these is
high level of noise, as compared with the signals.
The energy transfer ratio (ETR), which was proposed by
Prof.Liang and Prof.Lee [1,2], is referred to as the ratio of
energy transfers to total energies in each mode for the
non-proportional damped system. A pilot study at the
University at Buffalo [3,4] has shown that , before and after a
damage occurs, it is expected to have a large change in the
modal energy transfer ratio when the measurement positions
are very close to a specific damage. When the measurement
positions are far away from damage, the change of the
modal energy transfer ratio may become smaller. Extensive
studies [5] reveal that the change of ETR is close to 200
percent if a 5% local change of stiffness occurs. The
changing of stiffness is used to simulate the “damage”, which
can only cause about 2% change in natural frequencies, 6%
change in damping ratios and 7.5% change in mode shapes.
It is possible to detect the change of physical conditions of a
highway bridge, by comparing the degree of parameter
change at different positions. Three bridges were measured
under ambient and impact excitation by Lee and Liang [5].
Since there was no real damage existed in these bridges,
there were no notable changes in the measured and the
baseline signatures. These field tests were intended to
establish good measurements repeatability in ETR
extraction.
ETR is a modal parameter related to energy dissipation and
transfer. Real engineering system belongs to
non-proportionally systems. Therefore, energies are not only
absorbed by damping but also transferred to damping and
 P1Λ1 P1* Λ1* 
ETR exists in real world systems. Estimation of ETR requires
frequencies, damping and mode shapes mathematically. In
[P] =  * 
(7)
addition, ETR will include energy phenomena and local  P1 P1 
properties in physical terms. Since ETR is a local and energy Λ 0
parameter, it cannot only identify energy-related damage but [Λ ] =  1 * 
also locate damage. Also ETR can denote the damage  0 Λ1 
location. It has been theoretically demonstrated that ETR is
more sensitive to structural damage. In this paper, the and [Λ1 ]= diag (λi ) n×n (8)
dynamic and static testing of a 1/6-scaled steel-concrete
model bridge with different damage scenarios are carried out. Where P1 is the complex mode shape
The damage detection of the model highway bridge are 2
conducted by using ETR index, and the modal parameters, matrix, λi = −ξ iω i ± j 1− ξ i ω i is the i th eigen- value
such as frequencies and damping ratios, as well as the static of generally damped system, or is called the complex
data, such as the strain and deflections. Some testing frequency [6], and the superscript * refers to the complex
aspects are also specified. conjugate operation. The subscript i refers to the modal
order. By using the same idea, the modified stiffness matrix

2 ENERGY TRANSFER RATIO (ETR)

[K ] in Eq. (2) can also be eigen-decomposed and its
eigenvector Q can be obtained. The relationship of the
Consider only the homogeneous form of the equation of
motion:
stiffness matrix [K ] and its eigenvector Q is shown as:

[M ]X + [C ]X + [K ]X = 0 (1) Q T
[K ]Q = diag (ω ) ni
2
(9)

where [M ] 、[C ] 、[K ] are respectively the mass, damping, Where ω ni are the i th natural frequencies in the system.
For a proportionally damped system, Eq. (2) can be
and stiffness matrix. X 、 X 、 X are acceleration, velocity, diagonalized easily into a set of single degree freedom
and displacement, respectively. Without loss of generality, T
equations only by pre-multiplying eigenvector Q . However,
we may rewrite the Eq(1) as follows:
[] [ ]
Y + CY + KY =0 (2)
for a non-proportionally damped system, the coupling pairs in
the damping matrix will exist if we only pre-multiply Q ,
T

where: Y = [M ] X
0.5
; [C ] = [M ]
−0.5
[C ][M ]−0.5
; and Eq. (2) cannot be diagonalized. In order to decouple the

[K ] = [M ] −0.5
[K ][M ]−0.5 ; [C ] , [K ] are the new
non-proportionally damped system and obtain a more
general decoupled form, the generalized Rayleigh quotient
should be applied. The generalized Rayleigh quotient
modified damping and stiffness matrices. Y , Y , Y are the
acceleration, velocity, and displacement vectors in the new
includes the eigenvector Q , the couple mode shape P1 ,
coordinate after transformation. and the modified damping matrix [C ]. It is shown as
[]
Using equation 2, the state-space equation can be
T
expressed as Q C P1
[Z ] = [A][Z ] (3)
Ψ=
Q T P1
(10)

  Applying the generalized Rayleigh quotient, Liang et al [1]

where [Z ] = Y  (4) decoupled and diagonalized Eq. (2) into the following set of
Y  single degree freedom equations.


[A] = − C [ ] − [K ] (5)
ϖ i + 2ϑiω i ϖ i + ω ni 2ϖ i = 0 (11)
Ψ
 I 0  ϑi = ξ i + jζ i = ii (12a)
2ω i
Where [A] is called the continuous state matrix and [Z ] is
ω i = ω ni exp(ζ i )
called the state vector. The continuous state matrix [ A] can
and (12b)

be eigen-decomposed into the following form: where ϖ ,ϖ ,ϖ are the acceleration, velocity, and
[A]= [P ][Λ][P ] −1
(6) displacement in the i th mode after decoupling
Where [P ] is the eigen-vector of the continuous state matrix and j = −1 . ϑi , ξi , and ζi are the complex

[ ]
and Λ is the eigen-value matrix of the continuous state
damping ratio, dissipation damping ratio, and energy transfer

matrix. They can be interpreted into the following formulas by ratio (ETR) in the i th modal, respectively. ω i and ω ni are
the complex model shape and the complex frequency: th
the i natural frequencies of the non-proportionally and the
corresponding proportionally damped system., respectively.
According to Eqs. (1)-(11), the modal parameters, including concrete slab added mass
conventional parameters (natural frequencies and
dissipation damping ratio) and energy transfer ratios (ETR),
are defined theoretically.
steel girder

It can be shown that ζ i is the ratio of energy transferred hinge diaphragm roller

during one cycle to 4π times of the energy of the mode

before the cycle. ζ i can be measured by several methods 4-inch 10-feet 4-inch

either locally or globally. Local ETR can detect damage in a

small region and global ETR indicates a change of the bridge Figure1 1:6 Scaled Model Bridge
physical conditions. Now, suppose a damage breaks out in a
structure. It often does not significantly affect the value of
mass, damping, and stiffness. Therefore, both the concrete slab added mass
undamaged structure and the damaged structure can be
seen as variants from their corresponding proportionally
damped system. Denote the i th natural frequency of the bolt

undamaged and the damaged systems by ωi0 and ω ij (j

demonstrates different damage extent or damage condition),
we can write as follows.
ω i 0 = ω ni exp(ζ i 0 ) (13a) stiffener diaphragm

ω ij = ω ni exp(ζ ij ) (13b)
Figure 2 Connection of the Model Bridge.
Now suppose the damage creates a change between ωi0
and ω ij by the factor 0.1%, that is to say:
exp(ζ ij ) − exp(ζ i 0 )
ω ij − ω io = = 0.1%
exp(ζ i 0 )
ζ ij − ζ i 0
If ζ i 0 = 0.001 , then: = 100%
ζ i0
It can be seen that the result will be 1000 times more
sensitive than the change of natural frequencies due to the
damage.

3 TEST PROGRAM [7]

Description of Test Model Bridge Figure3 Artificial Damage

A real single slab-on-girder bridge was selected to be the
prototype bridge. In this experimental program, the length
top view of model bridge
scale factor is 1:6. The whole modal bridge is schematically
support support
shown in Fig. 1. The details of diaphragm and the connection
of girders are shown in Fig. 2. girder 4(S4)
⑧ ④
⑦ ③
Damage Simulation for the Model Bridge girder3 damage girder 3(S3)
⑩ ⑥ ②
bearing damage girder 2(S2)
Two types of artificial damages were introduced in the model ⑨ ⑤
girder1 damage
①
bridge. The first one was to simulate the loss of a bearing girder 1(S1)
after a severe earthquake or the settlement of bents. A loss 2 3
of a bearing was simulated by taking a central roller (at girder
2) off, while the other three rollers still remain in their original
positions. The second type of artificial damage is the Figure 4 Deformation Measure Points and
out-of-plane buckling and cracking of the steel girders. They Damage Positions.
are simulated to cut one exterior girder and one inner girder
to 1/2 height of the web from the lower flange by using
 Table 1 Damage Conditions analytical model to capture the dynamic behavior of the
model bridge. The shell elements with membrane and plate
Conditions Content bending behavior were utilized to model the flange and web
Intact No damage of I-beam, the channel diaphragms and girder stiffeners. The
Damage I Bearing damage block elements were used to model the added iron block
Damage II Only girder 1 damage because iron blocks were connected to the concrete slab
with cement mortar. The mass per unit volume for shell and
block element is used respectively. The steel girder bearings
electric saw near the one-third span (girder 1) and two-third were simulated using the “constraint command” in the
span (girder 3). At the beginning, the cracks were recovered ANSYS 5.5 program packages [8]. The material parameters,
by the cover plates and bolts. Then the damage are such as the moduli of elasticity of concrete and steel, were
simulated by loosening the bolts. The detail of this artificial determined according to the material testing results. The
damage is demonstrated in Fig.3 and Fig.4. The combination place where the girder is cut is modeled in detail in Fig. 6.
of damage conditions is shown in Table 1. The first mode shape for the undamaged is shown in Fig7.

Test Setup
The testing of the model bridge includes both static tests and
dynamic tests. First of all, the testing was conducted on the
non-damaged model bridge, then the damaged model bridge.
After the necessary coefficients were determined from the
pre-testing, the normal testing would be performed.
According to the combinations of damage, the static testing
was generated first with MTS Structural Testing System,
then followed by forced vibration testing and ambient
vibration testing. The forced vibration was conducted by an
energy hammer. The ambient vibration was conducted by a
scaled cart at different speed. Hanning window function was
used during signal processing to limit leakage. The sampling
rate was 3000Hz, and the cut-off frequency was 200Hz.
Moving window average technique was used to enhance the
signal-to-noise ratio. The power density functions or Figure 6 The FE Model of Crack in the Girder
frequency response spectrum was calculated after 57
averages with 87.5% overlap. The overlap allows for a more
optimal use of the time history. The normal testing setup was
shown in Fig.5.

Figure 7 The First Mode Shape for the Undamaged

5 TESTING RESULTS AND DAMAGE

IDENTIFICATION

5.1 DAMAGE IDENTIFICATION FROM STATIC DATA

Figure 5 Set-up of the Model Bridge
The changes of deformation before and after damage
may indicate presence of damage.
4 FINITE ELEMENT MODELLING
BEARING DAMAGE
The program ANSYS 5.5, a finite element analysis software The changes of deflection and strain under different loading
package, was used to create a detailed three-dimensional
via positions due to bearing damage are shown in Fig.8. The The frequencies, damping ratios, and ETR index identified
bearing damage at girder 2 caused more deflection change by using the Polyreference method [11] in the time domain
at near measured position than at other positions in Fig. 8(a). under two damage conditions are compared with those at
But similar condition doesn’t occur in Fig. 8(b). This indicates intact condition, and damage will be diagnosed from the
that deflection is more sensitive than strain to bearing changes of these parameters.
removal. We may conclude that bearing damage occurred at
the roller of girder 2, which is near the largest deflection Table 2 Comparison of Test and Calculation Results
position.
Natural Frequency (Hz)
Mode
GIRDER DAMAGE Ambient FE
No. Impact Test
Test Modelling
Due to girder damages, the changes of deflection and strain NO.1 33.723 33.722 35.382
under different loading via positions are shown in Fig.9. The
No.2 50.100 52.664 69.617
damage of girder 1 caused larger deflection changes at most
No.3 89.795 93.250 90.538
of the measured positions in Fig. 9(a), not only nears the
No.4 103.160 105.724 100.357
damage position. But similar condition doesn’t occur in Fig. 9
(b). The strain change near damage position in girder 1 is No.5 134.082 140.516 138.838
greater than that at any other positions. It indicates that No.6 169.437 169.243 153.142
strain is more sensitive than deflection to girder cracking. It No.7 185.492 188.713 176.870
is not possible to know exactly where the damage occurred.
We can only obtain that the damage occurred around
measured positions where the largest strain occurred. BEARING DAMAGE

The changes of natural frequencies and damping ratios for

10
the first three models before and after bearing failure are
50
summarized in Table 3. It can be inferred from the results
that the bearing failure affects the second bending mode
40 5

30
(mode 3) most. The first torsion mode (mode 2) also has
0

20

significant change after bearing is removed. All damping

-5
1
10
1

2 0 3
S4
-10
ratios increase after central bearing is damaged. Therefore,
3

S1
S2
S3
S4
S1
S2
S3
the natural frequency and damping ratio are indicators for the
damaged and /or condition changes. This conclusion is the
same as that from the SDOFI method.
(a) Deflection Changes (%) (b) Strain Changes (%)
Figure 8 ETR can be extract through any 3 position measurements.
The main objective of this study is to locate the damage
through the changes of ETR before and after damage is
introduced. To achieve this goal, the ETR is calculated
transversely by using the response data from those
80 10

70 0

accelerometers along the cross-section of the model bridge

60 -10

50 -20

40
30
-30
-40 (see Fig.10). Table 4 demonstrates the change of ETRs
1
20
10
1
-50
-60 between the first three modes after bearing damage is
introduced (also shown in Fig.11). It was found that, although
0 -70
2 -10 2 -80

the ETRs at all positions have certain changes after the

3 S4 3
S4
S3 S3
S2 S2
S1 S1

bearing damage is introduced, only the ETR of mode 3 at

position 1 has a remarkable change. Referring to the artificial
(a) Deflection Changes (%) (b) Strain Changes (%) damage definition, it is found that position 1 is the place
Figure 9 where the central roller was removed. The following
conclusions may be obtained from above results:
5.2 DYNAMIC PARAMETER EXTRACTION IN THE
FREQUENCY DOMAIN
girder＃4
24 23 22 21 20 19
The modified single degree of freedom identification (SDOFI)
girder＃3
method in the frequency domain was used to extract the 18 17 16 15 14 13
dynamic parameters [9,10]. Only the vertical vibration mode girder＃2 12 11 10 9 8 7
shapes were measured and analyzed. The comparison of
the frequencies of the model bridge with uniformly girder＃1 6 5 4 3 2 1
25
distribution mass without under two different excitations is ＃1 ＃2 ＃3 ＃4 ＃5 ＃6
summarized in Table 2. Reasonably good agreements have Positions
been achieved.

5.3 DAMAGE IDENTIFICATION USING ETR IN THE TIME Figure 10 Accelerometer Placement and
DOMIAN Positions for ETR extraction along the transverse
direction

1. ETR is very sensitive to the bearing removal. The
biggest change of ETR before and after bearing
removal is around 4500%, and the smallest change of
ETR is above 18%. It can be concluded that the change
of ETR can be used to indicate the damage and /or
condition changes of bridge structures.
2. The location where the biggest change of ETR occurs
can indicate the location of damage. In our case, the
roller near position 1 was removed and the biggest
change of ETR in Table 4 is also around position 1. This
agreement is first observed and is very promising for
ETR to be used to locate damage and condition
changes.
3. Through the current ETR extraction strategy and
interpretation methodology, it is not possible to know
exactly which roller was removed. There are totally 4
rollers at position 1. From the test results shown in
Table 4, we could informed that the damage occurred
around position 1, but we don’t know which roller at
position 1 is damaged.
4. Natural frequencies and damping ratios are global
vibrational signatures and can well indicate the
happening of damage and/or condition change of
bridge structures. But they can not be used to locate
the position of damage and/ or condition change. They
should be used together with ETR for the purpose of
damage and/or condition changes. They should be
used together with ETR for the purpose of bridge
damage detection.

Table 3 Changes of Modal Parameters before and after Bearing Damage

Mode Natural frequency Damping Ratio

Number Intact Damaged Changes Intact Damaged Changes
Mode 1 34.63 32.65 5.72% 10.202% 6.287% 38.37%
Mode 2 53.25 57.74 8.43% 5.989% 10.422% 74.02%
Mode 3 79.86 74.82 6.31% 3.136% 7.008% 123.47%

Table 4 Changes of ETR Index before and after Bearing Damage

Mode Change of ETR index

Number Position 1 Position 2 Position 3 Position 4 Position 5 Position 6
Mode 1 592.5% 75.13% 100% 96.91% 124.2% 87.48%
Mode 2 81.73% 1478% 938.3% 27.62% 31.59% 46.01%
Mode 3 4524% 18.78% 128.1% 55.14% 10.48% 32.72%

Figure 11 ETR changes by Bearing Damage Figure 12 ETR changes by Girder Cracking

Table 5 Changes of Modal Parameters before and after Girder 1 Damage

Mode Natural frequency Damping Ratio

Number Intact Damaged Changes Intact Damaged Changes
Mode 1 34.63 32.05 7.45% 10.202% 10.335% 1.30%
Mode 2 53.25 54.73 2.78% 5.989% 10.017% 67.26%
Mode 3 79.86 82.69 3.54% 3.136% 3.729% 18.91%
 Table 6 Changes of ETR Index before and after Girder 1 Damage

Mode Change of ETR index

Number Position 1 Position 2 Position 3 Position 4 Position 5 Position 6
Mode 1 100.0% 100.1% 62.93% 1614.7% 288.9% 99.08%
Mode 2 86.92% 214.3% 100.3% 820.2% 99.79% 99.63%
Mode 3 23.35% 34.03% 32.66% 68.21% 63.89% 42.24%

GIRDER DAMAGE
Changes of natural frequencies and damping ratios of the
first three modes are summarized in Table 5. It can be seen
that the natural frequency of mode 1 decreases substantially
while the natural frequencies of the last two modes have just
slight changes. At the same time, however, all damping
ratios increase after girder cracking is introduced.
of vibrational signatures of the first 3 modes after certain
damage was introduced in the model bridge. More
specifically, natural frequency, damping ratio and model
energy transfer ratio (ETR) are selected for the purpose of
sensitivity analysis. Table 7 and 8 summarize the comparison
results for both types of damages and the following
conclusions may be obtained from the above results:

The ETRs of the first three modes after girder crack 1. ETR is much more sensitive to the introduced bridge
simulation are extracted transversely based on the same damage than natural frequencies and damping ratios.
strategy described in the previous section. By using the ETR For the bearing failure case, the biggest changes of
at the intact condition as baseline, the change of ETR before natural frequency, damping ratio and ETR are 8.43%,
and after girder deterioration are calculated and plotted in 123.4%, and 4524%, respectively. At girder cracking
Table 6 and shown in Fig.12. The following conclusions may case, the biggest changes of natural frequencies,
be obtained from the test results: damping ratios and ETR are 7.45%, 67.26% and
1614.7%, respectively.
1. ETR is sensitive to a simulated girder cracking. From 2. The mode at which the ETR has the biggest change is
the result shown in Fig.12, the biggest change of ETR not necessarily the same as the mode where the
is 1614%, and the smallest change is 23%, which is natural frequency and/or damping ratios has the
much bigger than the change of natural frequencies biggest change. For the bearing removal case, mode 2
and damping ratios. For this case, the changes of ETR has the biggest change in natural frequencies while the
can be used to indicate the damage and/or condition biggest ETR change occurs at mode 3.
changes. 3. Comparatively, ETR is more sensitive to the damage
2. The location where the damage was introduced is also case of bearing removal than the case simulated girder
the place where the ETR has the biggest change. In our cracking. This is partly a result of the fact that the
study, the girder crack was introduced around the removing of the roller actually released the boundary
position 4 and 5, and both mode 1 and mode 2 have the condition constraint, which will certainly permit energy
biggest changes of ETR. This shows good agreement transfer more freely and easily.
between the biggest change of ETR and the location of 4. Not all modes and associated ETRs are sensitive to
damage. damage. In the bearing removal case, for example, only
3. The ETRs at both mode 1 and mode 2 are very ETR at mode 3 shows large changes around the
sensitive to girder damage, while the ETR at mode 3 is damage zone, while in the girder cracking case, ETRs
less sensitive. at both mode 1 and 2 are sensitive to damage.
4. The natural frequencies and damping ratios have a
different behavior than those for the bearing failure Table 7 Sensitivity Analysis for Bearing Damage
case. In the bearing failure case, the second and the
third mode has the biggest change in both natural Damping
frequencies and damping ratios, while in girder cracking Mode Frequency ETR
Ratio
case, the first mode has the biggest change in natural Number Changes Change
Changes
frequencies and damping ratios Mode 1 5.72% 38.37% 592.5%
Mode 2 8.43% 74.02% 81.73%
Mode 3 6.31% 123.4% 4524%
5.4 SENSITIVITY ANALYSIS
Table 8 Sensitivity Analysis for Girder 1 Damage
The above bridge damage diagnostic studies have
demonstrated tha ETR is very sensitive to both type of bridge Damping
Mode Frequency ETR
damages which were introduced. At the same time, however, Ratio
Number Changes Change
we don’t know how other vibrational signatures will behavior Changes
under the same type damage. Due to the limited time and Mode 1 7.45% 1.30% 1614.7%
experimental data, only natural frequencies, damping ratios Mode 2 2.78% 67.26% 820.2%
and ETRs of the first 3 modes are compared. The sensitivity Mode 3 3.54% 18.91% 68.21%
analysis was conducted by comparing the changes

6 CONCLUSION
Physically speaking, the change of local stiffness and
damping properties will cause the redistribution of stiffness
matrix [K] and damping matrix [C]. Since ETR comes directly
from the complex damping theory and is directly related to
the distribution of damping matrix, the ETR will change
according to the type of damage.
From above, we obtained the following remarks:
1. The deflection is sensitive to the bearing damage, but
not to girder cracking damage. The strain is sensitive to
girder cracking damage, but not to the bearing damage.
Deflection is a global static signature, while strain is a
local static signature. Changes of deflection or strain
may indicate the happening of damage and location of
damage. But these indications strongly rely on the
measured positions. If the measured points are very
close to the damage positions, the indications are clear.
Finding the accurate location may still need to be
developed.
2. ETR is much more sensitive to the introduced bridge
damage than natural frequencies and damping ratios.
The location where the biggest change of ETR occurs
can indicate the location of damage. But it is not
possible to know exactly where damage occurs.
3. Natural frequencies and damping ratios are global
vibrational signatures and can well indicate the
happening of damage and/or condition change of
bridge structures. But they can not be used to locate
the position of damage and/ or condition change. They
should be used together with ETR for the purpose of
damage and/or condition changes. They should be
used together with ETR for the purpose of bridge
damage detection.
4. Not all modes and associated ETRs are sensitive to
damage. In the bearing removal case, for example, only
ETR at mode 3 shows large changes around the
damage zone, while in the girder cracking case, ETRs
at both mode 1 and 2 are sensitive to damage.
Nashville, Tennesse, FL, USA. 226-232.
[3]. Kong, F. (1996), “The application of energy transfer
ratio in the bridge condition assessment,” Ph.D
dissertation submitted to the faculty of the Graduate
School of State University of New York at Buffalo,.New
York.
[4]. Huang, T.J. (1997), “Damage probes in structural and
mechanisms Utilizing energy-based modal parameter
identification”, Ph.D dissertation submitted to the faculty
of the Graduate School of State university of New York
at Buffalo, New York.
[5]. Lee, G.C., and Liang, Z. (1999), “Development of a
bridge monitoring system”, Proceedings of the 2
nd
international Workshop on Structural Health Monitoring,
Standford University, Standford, CA, Sept. 1999,
pp349-358.
[6]. Liang, Z., Tong, M., and Lee, G.C.(1992), “Complex
modes in damped linear dynamic system.”, The
International Journal of Analytical and Experimental
Modal Analysis. Vol7, No.1,1992, pp1-20.
[7]. Wang, T.L., and Zong, Z.H (2002), “Improvement of
evaluation method for existing highway bridges” (Final
report No. FL/DOT/RMC/6672-818), Florida
Department of Transportation, Tallahassee, FL, 32399,
USA.
[8]. Edited by Peter Kohnke,et al (1998), ANYSY theory
Reference ——Release 5.5. ANSYS Inc.
[9]. Ewins, D.J. (1995). Modal Testing: Theory and Practice.
Research Studies Press Ltd., England.
[10]. McConnel, K.G.(1995), “Vibration Testing---Theory and
Practice”, JohnWilly & Sons, Inc. NewYork, NY, 1995.

Finally, the authors wish to express their sincere appreciation
to the Florida Department of Transportation (FDOT) for
funding this research. Special thanks are also extended to Mr.
Marcus Ansley, Dr. Dongzhou Huang, and Dr. Z. Liang for
their valuable advice, suggestions, and comments during the
course of this study.
[11]. Deblauwe, F., Allemang, R.J., “The Polyreference
th
time-domain technique”, Proceedings of the 10
International Seminar on Modl Analysis, Part IV,
Katholike Universiteit Leuven, Nelgium, 1985.

REFERENCES

[1]. Liang, Z., and Lee, G.C.(1991), “Damping of

structures,Part I”, National Center for Earthquake
Engineering Research, NCEER Report 91-0004, State
University of New York at Buffalo, Buffalo, NY, 1991.

[2]. Liang, Z., Tong, M., and Lee, G.C. (1995), “Modal
th
energy measurement of a long steel bridge.”, 13 IMAC,