An Artificially-damaged Real Steel Truss Bridge and Its Numerical Modeling for Vibration-based Damage Detection

Abstract

In this study, a field damage experiment was carried out on an existing steel truss bridge to investigate feasibility of vibration-based damage detection (VBDD) of bridges. Two tension members were cut sequentially to serve as artificial damage. For each damage scenario, the dynamic
characteristics, specifically the dominant frequencies and mode shapes, of the bridge were identified from the dynamic responses excited by a passing experiment vehicle. On the other hand, finite-element (FE) models were constructed with commercial FE-analysis software, and then their eigen-frequencies
and corresponding mode shapes were compared with field-experiment results. It could be concluded that, firstly, in the field experiment, the modal frequencies and mode shapes of the bridge were identified with high precision and accuracy. Secondly, the eigen-frequencies and corresponding mode
shapes calculated with the FE models matched with the experiment results very well, indicating that those FE models could reliably model the real bridges and serve as alternatives for VBDD studies. However it did not apply to the case that the initial conditions between the FE model and real
bridge are inconsistent. Thirdly, the change in the identified modal frequencies and mode shapes was observed: modal frequencies decreased as damage causing high stress redistribution was applied, especially obvious as damage was applied asymmetrically; mode shapes were distorted as damage
was applied asymmetrically.

Full text

An Artificially
-
damaged Real Steel Truss Bridge and Its
Numerical Modeling for Vibration-based Damage Detection
Journal:
IABSE Nara 2015
Manuscript ID:
Nara-0432-2015.R1
Theme:
New technological advances on sustainability
Date Submitted by the Author:
15-Dec-2014
Complete List of Authors:
Kim, Chul-Woo; Kyoto University, Dept. of Civil and Earth Resources
Engineering
Chang, Kai-Chun; Kyoto University, Dept. of Civil and Earth Resources
Engineering
Material and Equipment:
Steel
Type of Structure:
Bridges
Other Aspects:
Inspection and Maintenance, Dynamic effects / vibrations

An Artificially-damaged Real Steel Truss Bridge and Its Numerical Modelling
for Vibration-based Damage Detection
Chul
-
Woo KIM
Professor
Kyoto University
Kyoto, Japan
kim.chulwoo.5u@kyoto-u.ac.jp
Chul-Woo Kim, born 1965,
received his Dr. Eng. degree from
Kobe University of Japan. His
main area of research is related to
bridge dynamics, structural health
monitoring and structural
reliability.
K.C. Chang
Researcher
Kyoto University
Kyoto, Japan
chang.kaichun.4z@kyoto-u.ac.jp
K.C. Chang, born 1980, received
his Ph.D. degree from National
Taiwan University. His research
interests are bridge dynamics,
vibration-based damage detection,
and structural health monitoring.
Summary
In this study, a field damage experiment was carried out on an existing steel truss bridge to
investigate feasibility of vibration-based damage detection (VBDD) of bridges. Two tension
members were cut sequentially to serve as artificial damage. For each damage scenario, the
dynamic characteristics, specifically the dominant frequencies and mode shapes, of the bridge were
identified from the dynamic responses excited by a passing experiment vehicle. On the other hand,
finite-element (FE) models were constructed with commercial FE-analysis software, and then their
eigen-frequencies and corresponding mode shapes were compared with field-experiment results. It
could be concluded that, firstly, in the field experiment, the modal frequencies and mode shapes of
the bridge were identified with high precision and accuracy. Secondly, the eigen-frequencies and
corresponding mode shapes calculated with the FE models matched with the experiment results
very well, indicating that those FE models could reliably model the real bridges and serve as
alternatives for VBDD studies. However it did not apply to the case that the initial conditions
between the FE model and real bridge are inconsistent. Thirdly, the change in the identified modal
frequencies and mode shapes was observed: modal frequencies decreased as damage causing high
stress redistribution was applied, especially obvious as damage was applied asymmetrically; mode
shapes were distorted as damage was applied asymmetrically.
Keywords: field experiment; modal analysis; FE model; vibration-based damage detection.
1. Introduction
In facing the ageing issues of bridge infrastructures, bridge damage detection has become an
important research and practical topic, especially for those developed countries that numerous
bridges were constructed in highly economic growth periods. Taking Japan for example, it is
statistically estimated that, up to 2012, those bridges having served for over 50 years occupy 9% of
the total bridges; in 10 years, those may reach 28%; and in 20 years, those may exceed 53% [1]. To
monitor the health conditions of those ageing bridges as well as of certain important newly-
constructed bridges, needs for inspection tasks that aim to detect potential bridge damage keep
increasing and pressing. Moreover, a huge amount of emergent inspection tasks may be required
once a natural disaster like earthquake or tsunami strikes.
One popular damage detection method utilizes structural vibrations, based on the intuitive
knowledge that damage may change the bridge’s mechanical properties (e.g. stiffness, mass or
energy dissipation mechanism), and hence change its dynamic responses in time domain as well as
dynamic characteristics (e.g. modal parameters or their relatives) in either modal or frequency
Page 1 of 10
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60

domain [2]. Such methods are referred to as the vibration-based bridge damage detection (VBDD)
methods. Among many damage-sensitive features, modal parameters are one of the features that
received researchers’ interest the earliest, as early as 1970s [3], because they are the physical
quantities dependent to the structural stiffness.
For investigating the effectiveness of VBDD techniques, field experiments on real bridges are
important and of high reference value because they are conducted in an environment that is most
similar to those within which the VBDD systems will be operated. Such environments are generally
not as well-controlled as those in numerical simulations and laboratories. However, most existing
studies examine these VBDD techniques by means of numerical simulations and laboratory
experiments, while still relatively few studies report their practical validity for real bridges, which
are likely to be subject to budget limitations and service conditions that prevent relevant authorities
from granting permission to apply damage to the bridge.
In viewing the limited number of practical cases, we carried out a field experiment on a real simply-
supported steel Warren-truss bridge in Japan with four artificial damage scenarios applied
sequentially. In addition, a finite-element (FE) model were constructed with the commercial FE-
analysis software ABAQUS
®
. The objective of this paper is firstly to summarize the modal
parameters identified from the field experiment and secondly to investigate whether the numerical
model can appropriately serve as an alternative tools for testing VBDD techniques. To this end, the
field experiment is briefly described in the next section, followed by the summary of identified
modal frequencies and mode shapes. Then the FE model is described, and their eigen-frequencies
and corresponding mode shapes are compared with field-experiment results. Finally several
concluding remarks are highlighted regarding to the discrepancy between experimental and
numerical results and to the change in the modal parameters from either the field experiments or
numerical analyses due to the artificial damage.
2. Field Experiment
The experiment bridge was a simply-
supported through-type steel Warren truss
bridge, as shown in Fig. 1. It is 59.2 m in
span length, 8.2 m in maximum height, and
3.6 m in width, designed for a single lane.
The bridge was constructed in 1959 and
planned to remove after a new replacement
bridge was open to traffic in 2012. Before
its removal, the bridge was allowed an
experiment with artificial damage. During
the experiment, the bridge had been closed
to traffic, and therefore no other vehicles
besides the experiment vehicle were
allowed.
The experiment vehicle was a two-axle
recreation vehicle, Serena model produced by Nissan Motor Company Ltd. The total weight was
about 21 kN. The fundamental frequencies of the vehicle were roughly identified in an independent
drop test. They were identified as 1.7~1.8 Hz for the front sprung bouncing mode and 2.1~2.3 Hz
for the rear sprung bouncing mode.
Five scenarios were considered in this study, as sketched in Fig. 2 and briefly summarized in Table
1. Initially, the INT scenario denoted the intact bridge with no artificial damage. Note that the term
‘intact’ is equivalent to ‘undamaged’ herein; it is used simply for expressing a state relative to the
8@7400= 59200 mm
P1 P2
A1 A2 A3 A4 A5
A6 A7 A8
DMG1
DMG2 DMG3
Passing direction
Ai: Accelerometer No. i (Vert.)
DMGi: damage scenario i
Pi: Pier No.i
Fig. 1 Experiment bridge with sensor layout
Page 2 of 10
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60

artificial damaged state, instead of an absolutely intact state. Firstly, two damage scenarios were
applied consecutively. The first damage (DMG1) denoted a half cut (see Fig. 2) applied by an
Oxyacetylene cutting torch in the vertical member at the midspan (see Fig. 1); the second damage
(DMG2) denoted a full cut applied in the same member. Those applied cuts were designed to
imitate real damage patterns that had been inspected previously, probably caused by corrosion or
overloading. After DMG2, the full cut was recovered, which was denoted as the RCV scenario. A
full cut applied in the vertical member located at the 5/8th-span is denoted as DMG3.
Fig. 2 Experiment bridge with sensor layout
Table 1. Damage scenario
Scenario Description
INT Intact bridge
DMG1 Half cut in vertical member @midspan
DMG2
Full cut in vertical member
@
midspan
RCV
Recovery of the cut member
DMG3
Full cut in vertical member
@
5/8
-
span
Table 2. Identified modal frequencies for the first five modes.
Mode No.
1 2 3 4 5
INT
2.975±0.09
%
6.865±0.17
%
9.629
±
0.31
%
10.547±0.14
%
13.428±0.12
%
DMG1 2.976±0.05% 6.887±0.04% 9.685±0.07% 10.594±0.14% 13.494±0.11%
DMG2 2.885±0.04% 6.876±0.03% 9.663±0.02% 10.568±0.04% 13.461±0.06%
RCV 2.967±0.04% 6.841±0.04% 9.567±0.05% 10.452±0.09% 13.393±0.24%
DMG3 2.922±0.06% 6.457±0.14% 8.651±0.15% 10.040±0.06% 13.397±0.54%
Note. Frequencies are expressed in mean (Hz) ± coefficient of variation.
As shown in Fig. 1, eight uni-axial accelerometers were installed vertically on the deck of the
experiment bridge, five at the damage side (A1-A5) and three at the opposite side (A6-A8). Limited
to the number of accelerometers available, the A6-A8 side serves to offer a clue to the judgment of
torsion modes, rather than to construct complete mode shapes. Two optical sensors were installed at
two ends of the bridge, respectively, to detect the entrance and exit of the vehicle. The sampling
frequency was set as 200 Hz for all sensors.
A moving vehicle experiment was performed on the experiment bridge. During the experiment, the
bridge was excited by the moving experimental vehicle, and the bridge vertical acceleration
responses, both during and after the vehicle passage, were measured by the accelerometers. Ten
runs were conducted for all scenarios except DMG1 scenario, for which twelve runs were
conducted. Although the vehicle was planned to move at a constant speed, it is difficult, even nearly
impossible, to maintain the constant speed manually. Subjected to this difficulty, the travelling
speed was calculated in an average way by dividing the known bridge span length by the travelling
time length measured by the optical sensors. The average speed such calculated ranged from 36 to
41 km/hr.
In this study, modal parameters were identified with a multivariate autoregressive (AR) time-series
method [4][4] aided by stabilization diagrams [5] for automatic selection of meaningful structural
modes. Bridge free vibration responses, i.e. the responses after the vehicle left the bridge, were
taken into modal-parameter identification to avoid any potential effect of passing vehicle on the
identified results, despite that the effect has been reported to be very little. The identified modal
frequencies for the first five modes are summarized in Table 2.
Page 3 of 10
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60

(b)
(c)
(
d)
Fig. 3 Views of FE model of the experiment bridge
:
(a) isotropic, (b) elevation, (c) top plane, (d) bottom plane.
(a)
BC: bottom chord
BLB: bottom lateral
bracing
D: diagonal
FB: floor beam
ST: stringer
TC: top chord
TLB: top lateral bracing
TS: top strut;
V: vertical
3. Finite Element Models
A three-dimensional finite element (FE) model (see Fig. 3) was developed with the FE analysis
software ABAQUS
®
. In the model, the truss members were modelled by beam elements (see Table
3 for details) and assigned with SS400 steel material properties (density 7900 kg/m
3
and elasticity
modulus 200 GPa); the deck was modelled by shell elements and assigned with concrete material
properties (density 2400 kg/m
3
and elasticity 21 GPa). In total, there were 3590 elements, including
2950 beam elements and 640 shell elements, and 3850 nodes. All the geometrical properties (cross-
sectional dimension, length, etc.), joint and boundary conditions were set up according to original
design drawings.
INT, DMG2 and DMG3 scenarios were considered in numerical analysis. The bridge model
developed above was regarded as the INT scenario, even though the actual properties of the real
bridge that has served for decades may not perfectly consistent with those revealed on the original
design drawings. The DMG2 scenario, a full cut in the vertical member at mid-span in the field
experiment, was simulated in FE model by removing the whole corresponding member. By doing
so, the break of force transmission pathways could be easily simulated without losing generalities.
The mass loss of the removed member was about 0.15% of the whole bridge mass, being small
enough to be disregarded herein. Similarly, the DMG3 scenario, a full cut in the vertical member at
5/8th-span in the field experiment, was simulated in FE model by removing the whole
corresponding member.
Page 4 of 10
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60

Table 3. Element profiles for truss members.
Component
Top chords Verticals
Member TC1, TC8 TC2, TC3, TC6,
TC7
TC4, TC5 V1, V2, V3, V4, V5,
V6, V7, V8
Dimension
Box 318×310×19×10 Box 298×309×9×9 Box 298×312×9×12 I 278×170×8×8
Component
Bottom lateral bracings Stringers
Member BLB1, BLB8 BLB2, BLB7 BLB3, BLB4, BLB5,
BLB6 ST1, ST2
Dimension
T 90×200×8×10 T 88×180×8×8 T 78×160×8×8 I 721×170×8×10
Component
Bottom chords Diagonals
Member BC1, BC2,
BC7, BC8
BC3, BC4,
BC5, BC6 D2, D7 D3, D4, D5, D6
Dimension
Box 296×198×8×8 Box 298×202×9×12
I 278×180×8×8 Box 278×200×8×8
Component
Floor beams Top lateral bracings Top struts
Member FB1, FB2, FB3, FB4,
FB5, FB6, FB7 FB end TLB2, TLB3, TLB4,
TLB5, TLB6, TLB7
TS1, TS2, TS3, TS4,
TS5, TS6, TS7
Dimension
I 738
×
220
×
9
×
8
I 736
×
160
×
8
×
8
T
103
×
150
×
8
×
8
I 216
×
150
×
8
×
8
Note. Box: box beam, I: I beam, T: T beam, dimension: height × width × web thickness × flange thickness.
Table 4. Comparison between numerical (FE) and experimental (EX) modal frequencies and mode
shapes (INT scenario).
1st mode
2nd mode
FE
f
1,
n
= 2.978 Hz
f
2,n
= 6.776 Hz
EX
f
1,
e
= 2.975 Hz
f
2,
e
= 6.872 Hz
∆ 0.1 % 1.4 %
3rd mode 4th mode
FE
f
3,n
= 9.831 Hz
f
4,
n
= 10.957 Hz
EX
f
3,
e
= 9.608 Hz
f
4,
e
= 10.559 Hz
∆
2.3 %
3.6 %
Note: f
i,n
: numerical modal frequency of the i-th mode, f
i,e
: experimental modal frequency of the i-th
mode, ∆: discrepancy; gray line: undeformed shape; green line: deformed shape.
4. Experimental vs. Numerical Results
By eigenvalue analysis, the modal frequencies and corresponding mode shapes of the numerical
bridge model were calculated. Comparing the numerical and experimental modal frequencies and
Page 5 of 10
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60

mode shapes of the first five bending modes for INT scenario (see Table 4), it was observed that
they match very well: between them, modal frequencies show less than 4% discrepancy and mode
shapes constructed by measurement points show similar shapes. It is also true for DMG2 scenario,
with the comparison of modal frequencies summarized in Table 5 but that of mode shapes omitted
for simplicity. Such a consistency can mutually verify the reliability of both the experiment results
and FE models.
However, the consistency does not apply to DMG3 scenario. From the comparison between
numerical and experiment modal frequencies (see Table 6), it is observed that the first two modes
and the 5th mode match very well, with discrepancies less than 4%, but the 3rd and 4th mode do
not, with discrepancy up to 12% and 8%.
The probable reasons for the consistency between numerical and experimental results for DMG2
scenario and the inconsistency for DMG3 scenario could be stated as follows. It is supposed that we
have a numerical model that could accurately model the real bridge in intact state as mentiond in the
preceding section. In the field experiment, DMG2 was applied directly on the intact bridge. With
the same initial state (intact state), the numerical model of DMG2 scenario can match with the real
bridge very well and therefore can reliably be taken for further dynamic analyses. For DMG3
scenario, in the field experiment it was applied following DMG2 and a recovery treatment so that
the bridge was not guaranteed to remain its initial intact state; on the other hand, in the numerical
model, DMG3 model was yielded from the INT model. This inconsistency of initial condition might
be the cause that the numerical results did not match with the experimental results very well.
Table 5. Comparison between numerical and experimental modal frequencies (DMG2 scenario).
1st mode 2nd mode 3rd mode 4th mode 5th mode
Numerical (Hz) 2.936 6.772 9.831 10.952 13.425
Experimental (Hz) 2.885 6.876 9.663 10.568 13.461
Discrepancy 1.7 % 1.5 % 1.7 % 3.5 % 0.3 %
Table 6. Comparison between numerical and experimental modal frequencies (DMG3 scenario).
1st mode 2nd mode
3rd mode 4th mode 5th mode
Numerical (Hz) 2.927 6.209 9.831 10.930 13.438
Experimental (Hz) 2.922 6.457 8.651 10.040 13.397
Discrepancy 0.2 % 4.0 % 12.0 % 8.1 % 0.3 %
5. Change in Modal Parameters due to Damage
The changes in modal parameters with respect to each damage scenario can be observed from Table
2 and be summarized in Table 7. Note that the changes are based on INT scenario for DMG1 and
DMG2 scenarios and on RCV scenario for DMG3 scenario; that the changes in modal frequencies
are expressed quantitatively in percentage and those in mode shapes are inspected qualitatively by
appearance.
Several general qualitative observations are made as follows. Firstly, modal parameters are little
affected by DMG1, the slight damage causing low stress redistribution. Secondly, modal
frequencies decrease, signifying a global stiffness loss, as severe damage causing high stress
redistribution is applied, such as those observed for the 1st mode as DMG2 is applied and for the
first four modes as DMG3 is applied. Thirdly, both symmetric and anti-symmetric mode shapes are
distorted as damage is applied asymmetrically (with respect to mid-span), such as DMG3 scenario.
In contrast, the mode shape distortion is not observed as damage is applied symmetrically, such as
DMG2 scenario. Finally, no obvious change is observed in modal parameters of the highest mode,
i.e. the 5th mode, probably due to too low the vibration level excited by the passing vehicle as well
as too sparse the sensor density.
Page 6 of 10
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60

Table 7. Change in mean identified modal frequencies and mode shapes due to damage.
Scenario Frequency Mode shape
1st mode
DMG1 + 0.03% Little variation
DMG2 - 3.03% Conspicuous in damage side
DMG3
-
1.
52
%
Slight d
istortion
2nd mode
DMG1
+
0.
32
%
Little variation
DMG2
+
0.
16
%
Little variation
DMG3
-
5.
61
%
Distortion
3rd mode
DMG1
+
0.
58
%
Little variation
DMG2
+ 0.
35
%
Little variation
DMG3 - 9.57% Distortion
4th mode
DMG1
+0.
45
%
Little variation
DMG2 + 0.20% Little variation
DMG3 - 3.94% Distortion
5th mode
DMG1 +0.49% Little variation
DMG2 + 0.25% Little variation
DMG3 + 0.03% Slight distortion
Table 8. MAC comparing modes identified from the damage and corresponding reference scenarios.
INT
Scenario Mode 1 2 3 4 5
1 1.0000 0.0004 0.2695 0.0001 0.0012
2 0.0001 0.9994 0.0110 0.4468 0.0023
DMG1
3
0.2592
0.0109
0.9976
0.0024
0.4169
4
0.0003
0.4740
0.0001
0.9972
0.0011
5 0.0054 0.0000 0.3686 0.0004 0.9581
1 0.9982 0.0005 0.2672 0.0001 0.0009
2 0.0002 0.9996 0.0104 0.4505 0.0022
DMG2
3
0.2680
0.0120
0.9979
0.0024
0.4221
4
0.0005
0.4540
0.0004
0.9986
0.0012
5 0.0006 0.0002 0.4055 0.0002 0.9609
RCV
1 2 3 4 5
1 0.9956 0.0001 0.2860 0.0009 0.0012
2
0.0004
0.8471
0.0702
0.5092
0.0043
DMG3
3
0.0519
0.0494
0.3167
0.2246
0.0140
4 0.0136 0.0486 0.0308
0.6299
0.0063
5 0.0048 0.0241 0.4337 0.0044 0.8656
One may also be interested in whether the numerical FEM results provide similar trend in frequency
change or not. By comparing the eigen-frequency values for INT and DMG2 scenarios (Tables 4
and 5), it is observed that, as the damage is applied, the 1st frequency decrease and the 5th
frequency increase, which are consistent with experimental results; while the 2nd and 4th
frequencies decrease very slightly and the 3rd frequency remains constant, all of which are
inconsistent with experimental results. Such inconsistency could be insignificant because both the
increase rate in experimental results and the decrease rate in numerical results are very small, nearly
to a degree that those discriminations could be regarded as experiment or calculation errors.
Page 7 of 10
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60

The changes of mode shapes can be evaluated quantitatively by a well-known correlation measure,
the modal assurance criteria (MAC). Table 8 gives the MAC values comparing modes identified for
DMG1 and DMG2 scenarios to INT scenario and comparing those for DMG3 scenario to RCV
scenario. The calculated MAC values indicate little change in mode shape due to DMG1 and
DMG2, with MAC values larger than 0.99 for the first four modes and MAC values about 0.96 for
the 5th mode. Significant changes in mode shape due to DMG3 is observed; the most significant
change occurs in the 3rd mode with MAC decreasing to 0.32, the second significant change occurs
in the 4th mode with MAC to 0.63, and the next significant change occurs in the 2nd and 5th modes
with MAC to about 0.85.
6. Discussion and Conclusions
A field damage experiment was conducted on a real simply-supported steel Warren-truss bridge
with four artificial damage scenarios applied and FE models were constructed with ABAQUS® to
model the bridges. Several concluding remarks can be drawn as follows.
Firstly, in the field experiment, the modal frequencies and mode shapes of the bridge were
identified with high precision and accuracy. The precision was indicated by little variations between
different test runs and the accuracy was verified by the FE numerical model.
Secondly, the eigen-frequencies and corresponding mode shapes calculated with the FE models
match with the experiment results very well for INT and DMG2 scenarios, indicating that those FE
models could reliably model the real bridges and serve as an alternative for VBDD studies.
However it is not true for DMG3 scenario, probably due to the inconsistency of initial conditions
between the FE model and real bridge. To develop a more proper model to model the real bridge of
DMG3 scenario, as well as of RCV scenario, could be one of our current challenges. Existing
model updating techniques could be appropriate tools.
Thirdly, the change in the identified modal frequencies and mode shapes was observed. For modal
frequencies, they decreased as damage causing high stress redistribution was applied, signifying a
global stiffness loss. Such a change was especially obvious as damage was applied asymmetrically.
For mode shapes, both symmetric and anti-symmetric ones were distorted as damage was applied
asymmetrically. To test if those parameters are effective damage sensitive features for damage
detection could be another challenge.
7. References
[1] Ministry of Land, Infrastructure, Transport and Tourism (MLIT), 2013 Annual Report on
Road Statistics: Current State of Bridges, MLIT, Japan.
[2] C.R. Farrar, S.W. Doebling and D.A. (2001), Vibration-based structural damage
identification, Philosophical Transactions of the Royal Society A Vol.359, 131-149.
[3] Cawley, P. and Adams, R. D. (1979), The location of defects in structures from
measurements of natural frequencies, The Journal of Strain Analysis for Engineering
Design, 14(2), 49-57.
[4] Kim C.W., Kawatani M. and Hao J. (2012), Modal parameter identification of short span
bridges under a moving vehicle by means of multivariate AR model, Structure and
Infrastructure Engineering 8(5), 459-472.
[5] Van der Auweraer, H. and Peeters, B. (2004), Discriminating physical poles from
mathematical poles in high order systems: use and automation of the stabilization diagram,
In Instrumentation and Measurement Technology Conference, 2004. IMTC 04. Proceedings
of the 21st IEEE: 3, 2193-2198.
Page 8 of 10
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60

An Artificially-damaged Real Steel Truss Bridge and Its Numerical Modelling
for Vibration-based Damage Detection
Chul
-
Woo KIM
Professor
Kyoto University
Kyoto, Japan
kim.chulwoo.5u@kyoto-u.ac.jp
Chul-Woo Kim, born 1965,
received his Dr. Eng. degree from
Kobe University of Japan. His
main area of research is related to
bridge dynamics, structural health
monitoring and structural
reliability.
K.C. Chang
Researcher
Kyoto University
Kyoto, Japan
chang.kaichun.4z@kyoto-u.ac.jp
K.C. Chang, born 1980, received
his Ph.D. degree from National
Taiwan University. His research
interests are bridge dynamics,
vibration-based damage
detection, and structural health
monitoring.
In this study, a field damage experiment was conducted on a real simply-supported steel Warren-
truss bridge with four artificial damage scenarios applied. The elevation and plan views of the
experiment bridge and the layout of sensors are shown in Fig. 1. The damage scenarios are
summarized in Table 1. For each damage scenario, the dynamic characteristics, specifically the
dominant frequencies and mode shapes, of the bridge were identified from the dynamic responses
excited by a passing experiment vehicle. On the other hand, finite-element (FE) models (see Fig. 2)
were constructed with commercial FE-analysis software ABAQUS
®
, and then their eigen-
frequencies and corresponding mode shapes were compared with field-experiment results. Several
concluding remarks were drawn as follows.
8@7400= 59200 mm
P1 P2
A1 A2 A3 A4 A5
A6 A7 A8
DMG1
DMG2 DMG3
Passing direction
Ai: Accelerometer No. i (Vert.)
DMGi: damage scenario i
Pi: Pier No.i
Fig. 1 Experiment bridge with sensor layout.
Fig. 2 FE model of the experiment bridge.
Firstly, in the field experiment, the modal
frequencies and mode shapes of the bridge
were identified with high precision and
accuracy. The precision was indicated by little
variations between different test runs and the
accuracy was verified by the FE numerical
model.
Table 1. Damage scenario.
Scenario Description
INT Intact bridge
DMG1 Half cut in vertical member @midspan
DMG2 Full cut in vertical member @midspan
RCV Recovery of the cut member (DMG2)
DMG3
Full cut in vertical member
@
5/8
-
span
Page 9 of 10
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60

Table 2. Comparison between numerical (FE) and experimental (EX) modal frequencies and mode
shapes (INT scenario).
1st mode
2nd mode
FE
f
1,
n
= 2.978 Hz
f
2,n
= 6.776 Hz
EX
f
1,
e
= 2.975 Hz
f
2,
e
= 6.872 Hz
∆ 0.1 % 1.4 %
Table 3. Comparison between numerical and experimental modal frequencies (DMG3 scenario).
1st mode 2nd mode
3rd mode 4th mode 5th mode
Numerical (Hz) 2.927 6.209 9.831 10.930 13.438
Experimental (Hz) 2.922 6.457 8.651 10.040 13.397
Discrepancy 0.2 % 4.0 % 12.0 % 8.1 % 0.3 %
Secondly, the eigen-frequencies and
corresponding mode shapes calculated
with the FE models match with the
experiment results very well for INT (e.g.
the first two modes as shown in Table 2)
and DMG2 scenarios, indicating that
those FE models could serve as an
alternative for vibration-based damage
detection studies. However it is not true
for DMG3 scenario (see Table 3 for
example), probably due to the
inconsistency of initial conditions
between the FE model and real bridge. To
develop a more proper model to model
the real bridge of DMG3 scenario, as well
as of RCV scenario, could be one of our
current challenges. Existing model
updating techniques could be appropriate
tools.
Thirdly, changes in the identified modal
frequencies and mode shapes were
observed. For modal frequencies, they
decreased as damage causing high stress
redistribution was applied, signifying a global stiffness loss. Such a change was especially obvious
as damage was applied asymmetrically. For mode shapes, both symmetric and anti-symmetric ones
were distorted as damage was applied asymmetrically. To test if those parameters are effective
damage sensitive features for damage detection could be another challenge.
Table 4. Change in
identified modal frequencies and
mode shapes due to damage.
Scenario Frequency Mode shape
1st mode
DMG1 + 0.03% Little variation
DMG2 - 3.03% Conspicuous in damage side
DMG3
-
1.
52
%
Slight d
istortion
2nd mode
DMG1 + 0.32% Little variation
DMG2
+
0.
16
%
Little variation
DMG3
-
5.
61
%
Distortion
3rd mode
DMG1
+
0.
58
%
Little variation
DMG2
+ 0.
35
%
Little variation
DMG3
-
9.57
%
D
istortio
n
4th mode
DMG1
+0.
45
%
Little variation
DMG2 + 0.20% Little variation
DMG3 - 3.94% Distortion
5th mode
DMG1 +0.49% Little variation
DMG2 + 0.25% Little variation
DMG3 + 0.03% Slight distortion
Page 10 of 10
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60