Download as pdf or txt
Download as pdf or txt
You are on page 1of 17

Seismic Response Analyses of the Yokohama Bay Cable-Stayed

Bridge in the 2011 Great East Japan Earthquake


Dionysius M. Siringoringo1; Yozo Fujino, M.ASCE2; and Kenji Namikawa3

Abstract: Yokohama Bay Bridge, with a total span length of 860 m, is the second longest span cable-stayed bridge in East Japan, and one of the
most densely instrumented bridges in Japan. On March 11, 2011, the Great East Japan (Tohoku) Earthquake hit northeastern Japan with mag-
Downloaded from ascelibrary.org by Tokyo Univ Seisan Gijutsu on 02/17/14. Copyright ASCE. For personal use only; all rights reserved.

nitude of MW 9.0, notably the largest earthquake in Japan’s modern history. Intensity 51 ðPGA 1:4e2:5 m=s2 Þ of the maximum scale 7 of Japan
Meteorogical Agency’s seismic intensity was recorded on the bridge site. This paper describes seismic response analyses of the bridge, system
identification, performance evaluation of link-bearing connections (a seismic isolation system), and postearthquake field observation. Response
analyses show that transverse vibration dominated the response of girder and tower, with a maximum girder displacement of 62 cm. The large
transverse vibrations induced pounding between tower and girder on the tower–girder connections as shown by the periodic impulses on
acceleration records. Meanwhile, longitudinal accelerations indicate that the link-bearing connections functioned properly during the
earthquake. System identification reveals nonlinearity of the response as evidenced by variations in natural frequencies and mode shapes
during large excitation. Despite these conditions, the bridge did not suffer any structural damages, since the ground motions experienced during
the earthquake were less than the design and seismic retrofit ground motions. DOI: 10.1061/(ASCE)BE.1943-5592.0000508. © 2014
American Society of Civil Engineers.
Author keywords: Seismic response; Cable-stayed bridge; 2011 Great East Japan (Tohoku) earthquake; MIMO system identification;
Instrumented bridge.

Introduction the most powerful earthquake in Japan’s modern history. The


epicenter was located offshore at 38.103°N, 142.860°E with a focal
Many long-span cable-supported bridges in seismically active re- depth of 24 km, about 398 km away from the Yokohama Bay Bridge.
gions are now instrumented with monitoring systems. Some notable The earthquake fault rupture consisted of an area about 440 km long
examples are the Golden Gate Bridge (Çelebi 2012), the Vincent north to south and up to 220 km wide off the northeast coast of Japan
Thomas Suspension Bridge (Smyth et al. 2003), the Cape Girardeau (Ide et al. 2011). This means that the closest distance from the bridge
Bridge (Celebi 2006) in the United States and the Higashi-Kobe site to the rupture fault area was about 180 km. Japan Meteorological
Bridge (Ganev et al. 1998), Katsushika Harp Bridge (Siringoringo and Agency (JMA) seismic intensity 51 ðPGA 1:40e2:50 m=s2 Þ of the
Fujino 2007), Yokohama Bay Bridge, Rainbow Bridge, and Tsurumi maximum scale of 7 was recorded on the bridge site; that is
Fairway Bridge in Tokyo metropolitan area, Japan (Siringoringo and equivalent to scale VII in modified Mercalli intensity (MMI).
Fujino 2008). Analyses of seismic responses of these bridges during This study describes seismic response analyses of the Yokohama
moderate and large earthquakes, such as during the 1987 Whittier Bay Bridge under the 2011 Great East Japan earthquake. The earth-
Narrows and the 1994 Northridge earthquake for Vincent Thomas quake foreshock, main shock, and aftershocks have generated
Suspension Bridge and the 1995 Hyogo-Nambu earthquake for a comprehensive dataset of long-span cable-stayed bridge responses
Higashi-Kobe Bridge, have improved our understanding of seismic subjected to multiple-support excitations. The high-quality seismic
performances of long-span bridges, especially the nonlinearity of responses recorded on the bridge are significant, considering that
bridge response (Smyth et al. 2003; Ganev et al. 1998), soil–structure seismic intensity of the main shock is one of the largest ever recorded
interaction (Ganev et al. 1998), and performance of structural con- on a completed long-span cable-stayed bridge in Japan. Moreover, the
nections (Siringoringo and Fujino 2008). number of vibration sensors on the bridge is larger than on any other
On March 11, 2011, at 14:46 JST, the Great East Japan (Tohoku) instrumented cable-stayed bridge in Japan. The paper describes
Earthquake hit northeastern Japan with magnitude MW 9.0, notably (1) temporal and spectral analysis of seismic responses; (2) system
identification and observation of changes in modal parameters with
1
respect to earthquake amplitude; (3) investigation on response non-
Research Assistant Professor, Dept. Civil Engineering, Univ. of Tokyo,
linearity; (4) performance evaluation of link-bearing connection (a
7-3-1 Hongo, Bunkyo-ku, Tokyo 113-8656, Japan (corresponding author).
E-mail: dion@bridge.t.u-tokyo.ac.jp
seismic isolation device); and (5) postearthquake field investigation.
2
Professor, Dept. Civil Engineering, Univ. of Tokyo, 7-3-1 Hongo,
Bunkyo-ku, Tokyo 113-8656, Japan. E-mail: fujino@sogo.t.u-tokyo.ac.jp Description of Yokohama Bay Bridge
3
Manager, Design Eng. Div., Metropolitan Expressway Co. Ltd., 1-4-1
Kasumigaseki, Chiyoda-ku, Tokyo 100-8930, Japan. E-mail: k.namikawa610@
Yokohama Bay Bridge links Tokyo to Yokohama Harbor area as
shutoko.jp
Note. This manuscript was submitted on September 20, 2012; approved on a part of Tokyo–Yokohama Bay shore expressway. The bridge is
May 9, 2013; published online on May 11, 2013. Discussion period open until a continuous, double-deck, three-span cable-stayed with a total span
June 13, 2014; separate discussions must be submitted for individual papers. of 860 m (200=460=200 m) (Fig. 1). The girder consists of a steel
This paper is part of the Journal of Bridge Engineering, © ASCE, ISSN truss-box with double deck: the upper deck for six-lane Yokohama
1084-0702/A4014006(17)/$25.00. Expressway Bay shore route and the lower deck for the two-lane

© ASCE A4014006-1 J. Bridge Eng.

J. Bridge Eng.
Downloaded from ascelibrary.org by Tokyo Univ Seisan Gijutsu on 02/17/14. Copyright ASCE. For personal use only; all rights reserved.

Fig. 1. Yokohama Bay Bridge sensor codes and locations

national route. The bridge has two H-shaped towers 172 m tall and a large earthquake such as the 1923 Great Kanto Earthquake, the
29.25 m wide, welded as a monolithic section. weak ground condition, and the high center of gravity of the bridge,
The bridge was opened in September 1989, and in 2005, a seis- a special seismic design measure was adopted. The measure includes
mic retrofit program was conducted for safety assurance of a Level 2 the use of LBCs, namely 10-m-long end-links at the end-piers (P1
earthquake according to Japan’s bridge seismic code. The retrofit and P4) and 2-m-long tower-links at the towers (P2 and P3), to ac-
program considered two types of maximum credible earthquakes: commodate girder longitudinal motion. The LBC system maintains
magnitude 8 far-field or moderately far-field large earthquakes long longitudinal fundamental period and allows the girder to be
taking place in the subduction zone of the Pacific plate with a near- suspended from towers and piers [Maeda et al. 1991; Metropolitan
field inland earthquake occurring beneath the site or close to the site. Expressway Public Corporation (MEPC) 1991]. As a result, the ef-
Based on the seismological model (Wald and Somerville 1995), the fect of inertia force of superstructure on substructure during an
ground motions were simulated assuming seven scenarios with fault earthquake can be minimized. By lengthening the natural period, the
geometry that resembles the 1923 Great Kanto Earthquake (Fujino structure experiences smaller acceleration but larger displacement.
et al. 2005). Simulations identified potential damages for both types Shorter LBCs at the tower function to restrict excessive longitudinal
of ground motion and concluded that significant damage would displacement that could occur during a large earthquake.
occur on the towers and bearings under such excitations. Further- The LBCs are made of steel and consist of a solid main body in
more, the far-field ground motion would create more damage and the middle part and circular bearings at both ends. Inside the bear-
induce 1.5-m longitudinal displacement of the girder. Accordingly, a ings, steel rods are placed through an inner ring and tightened by
fail-safe design concept was introduced, and the following strategies bolts to two circular steel plates located on both sides of the main
for major structural elements were implemented: body (Fig. 2). The LBC is designed as a tension link. The pin
1. Providing adequate seating on the approach span to avoid connections at both ends accommodate in-plane rotations that allow
unseating during a large earthquake; girder longitudinal movement as in a pendulum system.
2. Constructing additional cables connecting girder and end- The original design model describes LBCs as longitudinal hinge
piers in case a large transverse excitation damages the two connections, and they are expected to function as such in any level of
link-bearing connections (LBCs) at the end-piers; earthquake. This implies that relative displacement between pier and
3. Adding stiffeners inside the towers and piers to increase the girder will have to be sufficiently large to ensure that high-frequency
ultimate strength and ductility in case a large longitudinal components of pier and tower vibration are not transferred to the
excitation damages the edge of steel frame; girder. When such a condition is satisfied, inertia force from girder
4. Installing cable inside the lateral upper beam near the top of on the piers and towers could be minimized, and the amount of
tower to prevent the beam from falling in case of a large moment force on the substructures can be significantly reduced. This
deformation; and will create an isolation effect commonly found in base-isolated
5. Providing additional seats on the lateral beam under the girder to bridges. The fundamental longitudinal mode that captures such link
prevent impact force from girder in case a large transverse ex- behavior is noted as the longitudinal slip mode. The finite-element
citation damages the tower links and stay cables near the towers. model generates such a mode with period of 7.7 s (Maeda et al.
1991; MEPC 1991).
Characteristics of Tower–Girder and In transverse direction, the girder is suspended over the piers
End-Pier–Girder Connections and towers. Girder transverse movements are restricted by wind
shoes located on the pier–girder and tower–girder connections. A
Earthquake resistance was one of the main concerns in the bridge small gap exists between wind shoes and girder to accommodate
design. Considering seismicity of the area, possible recurrence of small relative motion. Part of girder connections that face wind

© ASCE A4014006-2 J. Bridge Eng.

J. Bridge Eng.
Downloaded from ascelibrary.org by Tokyo Univ Seisan Gijutsu on 02/17/14. Copyright ASCE. For personal use only; all rights reserved.

Fig. 2. (a) Characteristics of pier–girder and tower–girder connections and location of sensors on the connections; (b) detail figure of end-link; (c) detail
figure of tower-link (image by Dionysius M. Siringoringo)

shoes are covered with side bearings with Teflon polytetra- locations (Fig. 1). Along the girder, sensors are installed at 12
fluoroethylene (PTFE) surface, whereas the surfaces of wind shoes locations with spacing of 115 m. Two types of accelerometers
are made of stainless steel. Both side bearings and wind shoes are used: servo-type accelerometers SA-355CT (triaxial) and
are designed to resist maximum transverse load of 27,654 and SA-255CT (biaxial) produced by manufacturer Tokyo Sokushin.
48,346 kN for pier–girder and tower–girder connections, respec- These accelerometers have maximum amplitude capability 20 m=s2 ,
tively (MEPC 1991). sensitivity 2 m=s2 =V, and linearity 0.03% of the full scale. All
sensors have frequency bandwidth of 0.05–35 Hz and operate at
sampling frequency 100 Hz. The sensors are connected through
Description of Seismic Monitoring System a wired network system, and responses are recorded every time
accelerations exceed the preset trigger level. The monitoring system
The bridge has a permanent monitoring system consisting of 85 is operated and maintained by Tokyo Metropolitan Expressway
vibration sensors measuring acceleration and displacement on 36 Public Corporation.

© ASCE A4014006-3 J. Bridge Eng.

J. Bridge Eng.
Seismic Records Description Response spectra of free-field ground motions G1 located on the
hard-rock layer about 100 m below the surface are plotted in Fig. 4.
This study describes analyses of seismic responses recorded from Also plotted are the original design response spectrum for the hard-
March 9, 2011, foreshock; March 11, 2011, main shock at 14:47 rock layer and two seismic retrofit ground motions that produce the
JST; and nine aftershocks with JMA seismic intensity equal to or largest acceleration spectra for period longer than 3 s. As shown in
larger than 3 that occurred on March 11 and a few days after (Table the figure, acceleration spectra of the main shock are smaller than
1). Seismic responses from the main shock last for about 10 min, the both the original design and the retrofit. Considering that the first 10
longest among the seismic data recorded. The first part of the paper natural periods of Yokohama Bay Bridge are longer than 1 s, their
will focus on the response analyses and system identification of the corresponding spectra amplitudes are thus smaller than those of
main shock, and the later part of the paper will cover the results from the design and retrofit spectra, indicating that ground motions of
analysis of the foreshock and aftershocks. the present earthquake are not potentially damaging to the bridge.
Furthermore, response spectra calculated at two different soil levels
located on the pier and tower foundations (Fig. 5) clearly indicate
Characteristics of Recorded Ground Motions
Downloaded from ascelibrary.org by Tokyo Univ Seisan Gijutsu on 02/17/14. Copyright ASCE. For personal use only; all rights reserved.

significant spatial variation of ground motions in a short period range


that is less than 3 s. However, the variations become less significant
Fig. 3 shows the main shock accelerations recorded on three levels
for periods longer than 3 s, the range that is pertinent to the bridge
of tower P3 foundation: hard-soil layer (246:42 m), midfoundation
fundamental modes.
level (218:05 m), and base of the tower. The figures show that
amplitude of accelerations in longitudinal and transverse directions
are in the same order, and the peak amplitudes of accelerations Characteristics of Recorded Girder Responses
located on the silt layer are almost twice that of the hard-soil layer,
indicating amplification of ground motion. Similar conditions are The girder responses are characterized by large transverse vibration.
also observed on the responses of tower P2 foundation (not shown on The maximum recorded accelerations in the middle of the center
the figure). Generally, the ground motions are dominated by low- span are 299, 194, and 51 cm=s2 in transverse, vertical, and longi-
frequency peaks between 0.2 and 1.5 Hz. tudinal directions, respectively (Table 2). Large girder transverse

Table 1. List of Earthquake Records Used in Analysis.


Earthquake Source coordinate Epicenter distance to JMA seismic Data Peak input acceleration
Name (March 2011 JST) MW (depth in km) the bridge (km) intensity length (s) at footing level (cm=s2 )
FS 9, 11:45 7.3 38.33N, 143.28E (8) 480 2 120 4.11
MS 11, 14:47 9.0 38.10N 142.86E (24) 398 51 600 83.32
AS1 11, 15:16 7.7 36.11 N, 141.26 E (43) 192 4 400 35.27
AS2 11, 15:27 7.5 37.84N, 144.89E (34) 567 3 200 6.06
AS3 11, 16:29 6.5 39.06N, 142.28E (36) 603 2 60 3.84
AS4 11, 17:20 6.1 37.42N, 141.32E (27) 295 3 150 7.16
AS5 12, 04:00 6.7 36.98N, 138.6E (8) 205 3 120 3.11
AS6 13, 10:27 6.4 35.8N, 141.9E (10) 233 3 120 3.82
AS7 14, 10:03 6.2 36.46N, 141.12E (10) 204 4 60 5.31
AS8 15, 22:32 6.4 35.31N, 138.71E (14) 64 4 120 16.65
AS9 16, 12:52 6.1 35.8N, 140.9E (10) 148 3 120 5.62
Note: FS 5 foreshock; MS 5 main shock; AS 5 aftershock.

Fig. 3. Ground motions recorded on foundation of tower P3 during the main shock in longitudinal direction: (a) hard-soil layer; (b) silt layer; and
(c) footing level, and transverse direction (d) hard-soil layer; (e) silt layer; and (f) footing level

© ASCE A4014006-4 J. Bridge Eng.

J. Bridge Eng.
vibration is mainly due to the first transverse mode at 0.28–0.32 Hz girder accelerations during the main shock, however, indicate that
(Fig. 6). Meanwhile, the girder vertical vibration is dominated by the predominant girder frequencies are closely spaced around
five vertical bending modes between 0.32 and 1.2 Hz. It should be 0.32 Hz for both vertical and transverse modes. Time-frequency
mentioned that three accelerometers on the Oguro side, S8R(Y,Z), analysis by wavelet transform is performed to evaluate possible
S8L(Y,Z), and S9(X,Y,Z), failed, which means only 25 of the total changes in natural frequency during the entire main shock response
32 channels of accelerometer on the girder worked properly during as illustrated in Fig. 7. The figure shows that in transverse direction,
the main shock. the frequency peak appears around 0.28 Hz at the initial response
Previous investigations during smaller earthquakes (Siringor- and increases up to 0.32 Hz during the largest excitation at
ingo and Fujino 2006, 2008) identify the first vertical and transverse t 5 100e200 s. The peak then decreases to 0.28 Hz near the end of
frequency at 0.34 and 0.28 Hz, respectively. Frequency spectra of response. For vertical acceleration, the frequency peak appears around
0.34 Hz at the initial response and decreases slightly to 0.32 Hz during
the largest excitation, before increasing up to 0.34 Hz near the end of
the response. The changes in transverse and vertical frequencies
Downloaded from ascelibrary.org by Tokyo Univ Seisan Gijutsu on 02/17/14. Copyright ASCE. For personal use only; all rights reserved.

indicate the time-invariant nature of the responses that could be caused


by nonlinearity of structure or conditions of input excitations. To have
a better understanding of the characteristics of the structure and the
change in modal parameters throughout the excitation, a system
identification technique is used as discussed in later sections.

Characteristics of Recorded Tower Responses

The local tower mode at 0.42 Hz dominated the tower in-plane


motions (Fig. 8) in addition to the bridge global transverse modes
at 0.28 and 0.58 Hz. Maximum in-plane accelerations larger than
600 cm=s2 were recorded on the top of both towers. The tower out-
of-plane vibration was mainly caused by bridge global modes: the
girder vertical and torsional modes, in which the largest component
Fig. 4. Comparison of acceleration response spectra for design, seismic
was due to the second vertical bending modes at 0.48 to 0.52 Hz.
retrofit, and recorded main shock (G1)
For the out-of-plane accelerations, the maximum responses were

Fig. 5. Comparison of acceleration response spectra from ground motions recorded on foundations of piers and towers during the main shock for silt
layer: (a) longitudinal direction; (b) transverse direction, and hard-soil layer: (c) longitudinal direction; (d) transverse direction

© ASCE A4014006-5 J. Bridge Eng.

J. Bridge Eng.
Table 2. List of Recorded Accelerations and Displacements on Several Points on the Bridge during Main Shock March 11, 2011, 14:47
Max acceleration [RMS of
Sensor code and location Direction acceleration] (cm=s2 ) Max displacement (cm)
S5 R Longitudinal 51.14 [5.2] 19.60
(girder, main span midpoint) Transverse 299.17 [39.4] 61.80
Vertical 194.25 [19.6] 19.50
S4 Transverse 250.78 [32.9] 47.15
(girder, quarter of main span) Vertical 163.58 [18.9] 11.32
S2 Transverse 335.77 [34.4] 20.50
(girder, midpoint of side span) Vertical 165.80 [19.6] 6.94
T1 Out-of-plane 252.99 [26.6] 25.00
Top of tower In-plane 635.94 [78.6] 54.60
Downloaded from ascelibrary.org by Tokyo Univ Seisan Gijutsu on 02/17/14. Copyright ASCE. For personal use only; all rights reserved.

T2 Out-of-plane 418.67 [30.1] 24.28


Top of tower In-plane 656.87 [70.3] 48.36
T3 Longitudinal 124.37 [15.6] 23.40
Middle of tower Transverse 344.41 [47.2] 43.40
T4 Longitudinal 138.80 [16.2] 22.50
Middle of tower Transverse 411.73 [41.1] 42.70
T8 (bottom of tower, footing slab) Longitudinal 71.38 [7.4] 17.50
Transverse 67.27 [4.5] 18.00

only one tenth of the pier out-of-plane maximum acceleration.


Moreover, spectra of acceleration of the girder show that vibration
energy is significantly smaller compared with that of the pier. That
girder longitudinal acceleration is dominated by low-frequency mo-
tion, significant high-frequency reduction on girder longitudinal ac-
celeration, and large difference of vibration amplitude between pier
and girder indicate a functioning isolation effect of LBCs. In transverse
direction, the existing small gap between girder and pier at wind shoes
was exceeded during the main shock, resulting in a transfer of vi-
bration energy from pier or tower to girder. Such a condition is
demonstrated by the same frequency peaks at 0.57 Hz on the trans-
verse accelerations of the pier-cap and girder [Fig. 9(f and h)].
Similar to the pier–girder longitudinal connection, Fig. 10 shows
that high-frequency peaks around 1.0 and 3.0 Hz dominate tower
acceleration at the tower–girder connection, whereas girder accel-
eration is dominated by frequency peaks less than 1.0 Hz. Moreover,
amplitude of high-frequency components on the girder longitudinal
acceleration are significantly reduced, indicating that the LBC has
been effective as an isolation of longitudinal vibration. In transverse
direction, the figure reveals that both tower and girder accelerations
have very similar frequency characteristics. The tower transverse
accelerations are characterized by many periodic spikes resembling
Fig. 6. Fourier spectra amplitude of acceleration recorded in middle of impulses, especially during the largest excitation between 100 and
center span in (a) transverse and (b) vertical direction 300 s (Fig. 11). Periodic impulses indicate occurrence of transverse
pounding between tower and girder, and the impulses also appear on
the girder vertical accelerations (i.e., sensors S3-Z and S7-Z located
252 and 418 cm=s2 for P2 and P3, respectively. The different peak above the wind shoes). In addition to the main shock, the same pe-
amplitude of tower P3 was due to instantaneous spike on the re- riodic impulses appear on the tower–girder accelerations only during
sponse, but in general, vibration amplitudes of tower P2 and P3 were the first aftershock (AS1). Fig. 11(b and d) highlight the occurrences
not so different, with the RMS of 26 and 30 cm=s2 for P2 and P3, of impulse during the large excitation. In the figures, the time of
respectively. impulse occurrence is defined as the point when instantaneous sharp
peak with large amplitude appears on either positive or negative value
Characteristics of Recorded Responses on of the accelerations. By observing the time interval between two
Pier–Girder and Tower–Girder Connections successive impulses, one can estimate the structural mode that triggers
the impulse. From the figure, it is evident that the girder first transverse
Characteristics of accelerations recorded on the pier-caps and the mode triggers the pounding, since the average time interval between
wind shoes describe performance of LBCs during the main shock. two consecutive impulses is about 3.2 s (0.31–0.32 Hz).
Fig. 9 shows that dominant frequency peaks between 1.5 and 2 Hz
characterize the pier longitudinal (out-of-plane B3-X) acceleration, Displacements of Girder and Tower
whereas frequency peaks between 0.1 and 0.15 Hz dominate the
girder longitudinal acceleration (S1-X). The girder longitudinal Table 2 summarizes the maximum displacements of girder, tower,
acceleration is significantly smaller, with the maximum acceleration and piers. Displacements were computed from accelerations by

© ASCE A4014006-6 J. Bridge Eng.

J. Bridge Eng.
Downloaded from ascelibrary.org by Tokyo Univ Seisan Gijutsu on 02/17/14. Copyright ASCE. For personal use only; all rights reserved.

Fig. 7. Girder midspan responses during the main shock: (a) transverse acceleration; (b) ridge of Morlet wavelet scalogram of transverse acceleration;
(c) vertical acceleration; (d) ridge of Morlet wavelet scalogram of vertical acceleration; (e) ridge of transverse acceleration scalogram for the main shock
beginning and coda; (f) ridge of transverse acceleration scalogram during the largest excitation; (g) ridge of vertical acceleration scalogram for the main
shock beginning and coda; (h) ridge of vertical acceleration scalogram during the largest excitation

© ASCE A4014006-7 J. Bridge Eng.

J. Bridge Eng.
Downloaded from ascelibrary.org by Tokyo Univ Seisan Gijutsu on 02/17/14. Copyright ASCE. For personal use only; all rights reserved.

Fig. 8. Accelerations and Fourier amplitude spectra recorded on top of the tower during main shock

Fig. 9. Characteristics of longitudinal and transverse accelerations at the pier–girder connection (P1)

double integration and removing frequency components less than largest excitation (t 5 150e200 s), suggesting that the girder was
0.1 Hz. For the girder, maximum displacements in the middle of completely decoupled from pier movement at the LBC for that
center span were 62 and 20 cm in transverse and vertical directions, period of time.
respectively. The girder transverse displacement was mainly in- Tower in-plane displacement was significantly large during the
duced by the first girder transverse mode at 0.28–0.32 Hz, whereas main shock, with the peak displacement of 55 cm recorded on the top
the vertical displacement was mainly due to the first vertical bending of tower P2, or about twice the peak out-of-plane displacement. The
mode at 0.32–0.34 Hz. large tower transverse displacement was mainly due to the local
Girder longitudinal displacement is one major concern, since an tower in-phase mode at 0.42 Hz. Meanwhile, the out-of-plane tower
excessive longitudinal displacement could initiate pounding on the displacement was dominated by girder vertical mode that coupled
approach spans and result in damage to the pier–girder connections. with tower out-of-plane mode such as the girder first vertical mode at
The results show, however, that the maximum longitudinal dis- 0.32 Hz.
placement during the main shock was about 20 cm, which is still far
below the allowable peak displacement of 1.5 m anticipated in the
seismic retrofit program. Relative pier–girder and tower–girder System Identification
longitudinal displacements were within 5–10 and 1–5 cm, re-
spectively (Fig. 12). The figure also shows that pier-cap displace- Multiple-input multiple-output (MIMO) system identification was
ment (B3-X) lags behind girder (S1-X) displacement during the used to evaluate fundamental vibration modes pertinent to the bridge

© ASCE A4014006-8 J. Bridge Eng.

J. Bridge Eng.
Downloaded from ascelibrary.org by Tokyo Univ Seisan Gijutsu on 02/17/14. Copyright ASCE. For personal use only; all rights reserved.

Fig. 10. Characteristics of longitudinal and transverse accelerations at the tower–girder connection (P2)

Fig. 11. Close-up look at the tower transverse (in-plane) acceleration at the tower–girder connections shows periodic impulse response for (a) main
shock and (c) aftershock 1; time interval between consecutive impulses during (b) main shock and (d) aftershock 1

seismic responses. The identification is based on the system re- composed by correlation functions of input and output data. To
alization using information matrix (SRIM) (Juang 1997) algorithm determine ½Op , one can start by obtaining the matrices of input-
that uses correlations between input and output data to estimate output correlation data
modal parameters of a dynamical system through a realization pro-
cess. The identification procedure starts by estimating the observ-    T
^ xx Op
Rhh ¼ Op R (1)
ability matrix ½Op  from the so-called information matrix that is

© ASCE A4014006-9 J. Bridge Eng.

J. Bridge Eng.
After the transformation, one can estimate the natural frequency (vi )
and modal damping ratio (ji )
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
vi ¼ Reðli Þ2 þ Imðli Þ2 , ji ¼ 2Reðli Þ=vi (8)

The mode shapes matrix in coordinate system is obtained by trans-


forming the eigenvectors in z-domain into coordinate domain using
the output transformation matrix R

^
F ¼ RF (9)

Detailed information on application of the system identification al-


Downloaded from ascelibrary.org by Tokyo Univ Seisan Gijutsu on 02/17/14. Copyright ASCE. For personal use only; all rights reserved.

gorithm to Yokohama Bay Bridge, including numerical simulations


Fig. 12. (a) Longitudinal displacement of girder-end and pier cap; to verify its performance under various conditions, is given else-
(b) longitudinal displacement of tower P2 at girder connection where (Siringoringo and Fujino 2008).
To implement the identification, one needs to select a set of input-
output data. In this study, responses from triaxial accelerometers
^ xx are defined as located at the bottom of the end-pier (P1: sensor K2) and the towers
where the quantities Rhh and R
(P2: sensor K4; P3: sensor K6) were selected as inputs. These re-
sponses were used instead of the free-field responses (G1) to mini-
Rhh ¼ Ryy 2 Ryz R21
zz Ryz
T
(2a) mize the effect of soil–structure interaction and realize the MIMO
^ xx ¼
R Rxx 2 Rxz R21 system. Unfortunately, sensors at the bottom of P4 failed during the
zz Rxz
T
(2b)
main shock, so they were not included in the analysis. As the outputs,
As shown in Eq. (2), the quantity Rhh is determined from the input 41 channels that include girder (23 channels), pier P1 (two channels),
(z) autocorrelation matrix Rzz , the input-output correlation matrix tower P2 (eight channels), and tower P3 (eight channels) were se-
Ryz , and the output ( y) autocorrelation matrix Ryy , and it exists only lected. Because the input and output data are derived from multiaxial
if the input autocorrelation matrix Rzz is a nonsingular matrix. To accelerations, the identification yields three-dimensional mode shapes.
obtain the solution for matrix ½Op , further factorization of Eq. (1) is It should be mentioned that the system identification is based on
required by using the singular value decomposition as the assumption that modal parameters remain constant during a
specific time window in which the input-output dataset is analyzed.
   T However, considering that the bridge may enter a nonlinear region
^ xx Op ½:, 1: ð p 2 1Þm
Rhh ½:, 1: ð p 2 1Þm ¼ Op R during large excitation, this assumption may not be satisfied for the
whole response. Therefore, piecewise linear analysis was conducted
¼ H2N S22N VT2N (3)
on shorter moving time windows during which the modal parame-
ters were assumed to be constant. For this reason, the total time
Following the identity in Eq. (3), the observability matrix can be history responses were divided into several time windows.
obtained as Two important factors determine accuracy of modal parameter
estimates: model selection and p value selection. Model selection is
    performed using a systematic procedure of eliminating the fictitious
Op ¼ H2N and ^ xx Op T ½:, 1: ð p 2 1Þm ¼ S2 VT (4)
R 2N 2N modes as explained in (Siringoringo and Fujino 2008). The p value
represents selection of block matrix data included in analysis [Eq. (5)],
Furthermore, given the observability matrix, one can estimate sys- whose actual value could be larger than the theoretical value ow-
tem matrix ½A as ing to measurement noise. The appropriate p value is determined
  from a convergent point in a typical stabilization diagram that
½A ¼ ½Opp ½1: ð p 2 1Þm,:  Op ðm þ 1: pm,: Þ (5) plots the results of p value versus identification error. The
minimum number of columns in input and output block matrix
where p 5 pseudo inverse matrix. The integer p should be selected data were evaluated in a numerical study on the sensitivity of
such that matrix ½Op ðm 1 1: pm,:Þ of dimension ð p 2 1Þm 3 2N has modal parameter estimates with respect to the length of input-
rank larger than or equal to 2N, hence p $ 2N=m 1 1. output in block matrices, and it was found that setting the number
Modal parameters of the structural system are estimated by of columns to 5,000 (50 s) gives reasonably consistent estimates.
solving the eigenvalues problem of matrix A To illustrate the previously mentioned procedure, Fig. 13 shows
the stabilization diagrams that describe consistency of identification
½AF ~F
^ ¼L ^ (6) results of the first transverse and vertical modes with respect to p
value. Records with various lengths were evaluated by varying the
where matrices L ~ and F^ 5 eigenvalues and eigenvectors of matrix number of rows (i.e., p value) in input and output block matrices
[A], respectively. The eigenvalues and eigenvectors can be real or while setting number of columns to 5,000 (50 s). The figure shows
complex; in case of the latter, they appear as complex conjugate that the identified values approach steady results as the p value
pairs. The eigenvalues l ~i are expressed in z- domain and can be increases, and they converge to constant values when p 5 80. As
related to the modal characteristics of the dynamics system using the a result, p 5 80 was selected as the representative p value for the
following transformation: analysis. Histograms show that the frequency estimates are con-
fined in a narrow band of frequencies, indicating small variation
  of identified values and high certainty. The mean values for iden-
~i =Dt
li ¼ ln l (7) tified frequencies are 0.323 and 0.345 Hz for the first and second

© ASCE A4014006-10 J. Bridge Eng.

J. Bridge Eng.
Downloaded from ascelibrary.org by Tokyo Univ Seisan Gijutsu on 02/17/14. Copyright ASCE. For personal use only; all rights reserved.

Fig. 13. (a) Stabilization diagram of identified natural frequencies with respect to p value; (b) stabilization diagram of identified damping ratios with
respect to p value; (c) distribution of identified results of natural frequencies; (d) distribution of identified results of damping ratios

frequency, respectively. Histograms of damping estimates, on the dominate girder vibration in three respective directions and are most
other hand, are not confined in a narrow band of frequencies, in- sensitive to the excitation amplitude.
dicating quite large variations of damping estimates. It should be
mentioned, however, that these histograms are plotted considering Characteristics of Identified Longitudinal Mode
all data starting from p 5 40. Damping estimates resulted from
p . 80 have less scatter, and therefore are used as the representative The frequency of the first longitudinal mode varies in a range of
identified values. 0.11–0.18 Hz throughout the entire excitation (Fig. 14). The fre-
By applying the previously mentioned procedure and taking the quency decreases during the largest excitation, and the lowest
stabilization diagram into consideration, the total 600-s main shock frequency is identified in frames 3 and 4. Near the end of the re-
response was divided into 12 frames of input-output dataset, and sponse, where the excitation level becomes smaller, the frequency
modal parameters were identified for each frame. The same pro- increases slightly. The damping ratio also shows variation in that
cedure was also applied to the responses of nine aftershocks listed in larger damping is identified in the beginning and near the end of
Table 1, and the results are discussed in the following sections. the response when the excitation is smaller. Large damping esti-
mates during small excitation are rather unexpected. The source of
damping in longitudinal vibration is mainly friction between wind
Results of System Identification
shoes and the girder at the connections, and their values are expected
System identification generates 14 modes between 0.1 and 2.5 Hz to increase as the excitation level increases. It is noted, however, that
where the girder modal displacement dominates the mode shapes damping ratios identified during small excitation are rather scattered
(Table 3). The lowest frequency is the longitudinal swing mode at in that the p value and damping estimate do not converge to certain
0.11–0.18 Hz, followed by the girder first transverse bending at values as they do during large excitation, so it is suspected that the
frequency range 0.27–0.32 Hz and the first vertical symmetric results are mainly due to numerical error in identification.
bending at frequency range 0.32–0.36 Hz. In vertical direction, there As mentioned previously, the LBC is expected to function as
are eight modes identified consisting of five vertical bending and a longitudinal hinge connection in any levels of earthquake and
three torsional modes. In transverse direction, there are five modes the longitudinal mode shape with such condition is denoted as the
consisting of two girder-dominant modes, and three tower- and pier- slip mode. However, previous observations reveal that in smaller
dominant modes. Their identified natural frequencies were generally earthquakes, higher frequency and smaller relative modal displace-
in good agreement with the results from previous earthquakes and ments between pier and girder were observed (Siringoringo and
the finite-element model (Siringoringo and Fujino 2008; MEPC Fujino 2006). Longitudinal mode shape with such condition is de-
1991). noted as the stick mode. The stick mode suggests that LBCs have
In the following sections, characteristics of the three most dom- not functioned as full hinge connections and as a result force dis-
inant modes, that is, the girder first longitudinal, first transverse, and tribution on the end-piers and tower may be different from what was
first vertical mode, are discussed in more detail. The three modes intended in design. Significant amounts of additional force will be

© ASCE A4014006-11 J. Bridge Eng.

J. Bridge Eng.
Table 3. List of Identified Modal Parameters from Foreshock, Main Shock, and Aftershock 1.
Foreshock Main shock Aftershock 1
Finite-element
Frequency Damping Frequency Damping Frequency Damping model
Mode (Hz) (%) (Hz) (%) (Hz) (%) frequency (Hz) Mode shape characteristics
Longitudinal
1 0.12 9.28 0.11 2.08 0.11 8.30 0.14 Longitudinal swing mode (sliding)
Vertical
1 0.34 4.80 0.33 4.20 0.33 1.41 0.34 First bending (first symmetric), lateral coupling
2 0.52 6.16 0.51 1.21 0.51 0.61 0.49 Second bending (first asymmetric), no lateral
coupling
3 0.80 1.87 0.75 3.49 0.77 1.50 0.77 Third bending (second symmetric), pure no
lateral coupling
Downloaded from ascelibrary.org by Tokyo Univ Seisan Gijutsu on 02/17/14. Copyright ASCE. For personal use only; all rights reserved.

4 0.96 1.48 0.94 0.84 0.94 4.09 1.01 Fourth bending (second asymmetric), lateral
coupling
5 1.12 5.41 1.10 1.34 1.12 1.73 1.21 Fifth bending (third symmetric) no lateral
coupling
Torsion
1 0.84 0.92 0.83 0.59 0.86 4.47 0.86 First torsion of girder
2 1.40 3.78 1.41 2.59 1.40 1.03 1.38 Second torsion of girder
3 2.57 1.91 2.41 1.80 2.53 2.32 2.45 Third torsion of girder
Transverse
1 0.27 2.73 0.32 0.47 0.28 7.67 0.28 Girder lateral first bending (first symmetric)
coupled with girder vertical
2 0.54 5.51 0.53 2.63 0.53 4.61 N/A Tower lateral (P2 and P3 out-of-phase),
coupled with girder vertical
3 0.45 1.56 0.42 1.34 0.46 0.79 0.42 Tower lateral (P2 and P3 in-phase) without
girder vertical and lateral coupling
4 0.58 6.41 0.57 1.16 0.57 0.62 N/A End pier (P1) lateral displacement coupled
with girder
5 1.07 2.31 1.10 1.34 1.07 6.23 1.08 Girder lateral second bending (first
asymmetric) with small vertical coupling
Note: Main shock time-window: 150 to 200 s or Frame 4.

Fig. 14. (a) Longitudinal swing mode identified from Frame 4 of the main shock with the slip-mode characteristic; (b) variation identified frequency
and damping of the first longitudinal mode during the main shock; (c) variation of relative modal displacement between girder and pier/tower for the first
longitudinal mode during the main shock

© ASCE A4014006-12 J. Bridge Eng.

J. Bridge Eng.
redistributed to the pier and tower base if the LBCs remain stuck suggesting the occurrence of a slip mode. Furthermore, the oc-
during a large earthquake. currence of a slip mode indicates an increase in longitudinal flex-
Fig. 14(a) shows an example of longitudinal mode shape iden- ibility and explains the frequency reduction of longitudinal mode
tified from Frame 4 of the main shock. To determine whether during the largest excitation.
a longitudinal mode is classified  as stick or slip
 mode, a relative
modal displacement index g 5 fgirder 2 ftower  is defined. In this
Characteristics of Identified Girder Lateral and
index, f denotes the normalized longitudinal modal displacement
Vertical Modes
of girder, tower, and pier at the LBC locations, i.e., sensor B3-X
and S1-X for P1, sensor T5-X and S3-X for P2, and sensor T6-X Natural frequencies and modal displacements of the first transverse
and S7-X for P3. To calculate the normalized longitudinal modal and the first vertical bending are closely related to the amplitude of
displacement, modal displacements at each LBC position are nor- excitation. Results of system identification show that frequency of
malized to maximum values, and then the normalized values are the first transverse mode increases from 0.27 to 0.32 Hz during
compared with the relative modal displacement between girder and the largest excitation, whereas frequency of the first vertical mode
Downloaded from ascelibrary.org by Tokyo Univ Seisan Gijutsu on 02/17/14. Copyright ASCE. For personal use only; all rights reserved.

pier-cap or between girder and tower at the LBC positions. In case of decreases slightly from 0.35 to 0.33 Hz. The changes in frequencies
slip mode, g will be nearly 1 or larger than 1, whereas for the stick are followed by the change in transverse-vertical coupling of the
mode, g will be closer to 0. Note that the value of g will be larger girder mode shape. During large excitation, strong coupling is ob-
than 1 when pier and girder are out of phase, as in the case of the served in both transverse and vertical modes, in that the vertical
mode identified in Frame 4. Fig. 14(c) shows that index g is nearly mode has significant modal displacement in transverse direction and
equal to or larger than 1 during large excitation, i.e., frames 3 to 6, the transverse mode has significant modal displacement in vertical

Fig. 15. Characteristics of mode shapes with strong transverse-vertical coupling during the largest excitation of main shock (Frame 4, t 5 150e200 s):
(a) girder first transverse mode; (b) girder first vertical mode; characteristics of mode shapes with weak transverse-vertical coupling during the end of
main shock (Frame 11, t 5 500e550 s): (c) girder first transverse mode; (d) girder first vertical mode

© ASCE A4014006-13 J. Bridge Eng.

J. Bridge Eng.
direction as shown by Fig. 15(a and b). During smaller excitation, as
in the beginning and near the end, the transverse-vertical coupling
becomes weaker [Fig. 15(c and d)]. In such a condition, vertical
mode has small participation of transverse modal displacement and
vice versa.
Coupling condition
 of a mode shape is defined by a coupling
index hV,T 5 fV,T =ðjfV j 1 jfT jÞ, where f 5 modal displacement
of girder at the middle of center span (i.e., node S5R), and sub-
scripts V and T 5 vertical and transverse, respectively. In a weakly
coupled mode, h has a large value (almost 1) in one direction and
a small value (almost 0) in the other direction; and in a strongly
coupled mode, both hT and hV are close to 0.5. As illustrated in
Fig. 16(b and c), the transverse-vertical coupling for both modes
Downloaded from ascelibrary.org by Tokyo Univ Seisan Gijutsu on 02/17/14. Copyright ASCE. For personal use only; all rights reserved.

are strong during the first half of the record, especially between
Frame 2 and Frame 5, which is when the largest excitation occurs. At
Fig. 16. (a) Variation of natural frequency of the first transverse and
the second half of the record, coupling becomes weaker before they
first vertical bending for the main shock; (b) variation of coupling index
are fully decoupled toward the end of excitation.
h of the first transverse mode; (c) variation of coupling index h of the first
The changes in natural frequencies and the transverse-vertical
vertical mode
coupling pattern with respect to input amplitude indicate non-
linearity of bridge response. There are several possible sources of
nonlinearity in a cable-stayed bridge seismic response, such as
material nonlinearity, geometric nonlinearity due to large cable
displacement, soil–structure interaction, and behavior of pier–girder
and tower–girder connections. In the present case, displacement
from seismic response was not too large to cause significant geo-
metric or material nonlinearity. In addition, tower responses did not
show significant indication of nonlinearity as expected in a case of
significant soil–structure interaction. Considering these conditions,
behavior of the tower–girder and the pier–girder connections is
thought to be a more possible cause of nonlinearity.
As in the case of longitudinal mode, relative modal displacement
at tower–girder connection in transverse
 direction can be quantified
by index g 5 fgirder 2 ftower , where f 5 normalized transverse
modal displacement of girder and tower. fgirder is selected from
Fig. 17. Variation of g during main shock response for towers P2 and sensor S3-Y and S7-Y, and the values of ftower are selected from
P3
sensor T5-Y and T6-Y for tower P2 and P3, respectively. Index g
becomes larger when relative modal displacement between tower

Fig. 18. Variation of natural frequencies with respect to input acceleration for all analyzed seismic records: (a) first transverse bending mode; (b) first
vertical bending mode; MS = main shock

© ASCE A4014006-14 J. Bridge Eng.

J. Bridge Eng.
and girder is large and vice versa. Fig. 17 shows variation of index g the figure to the change in the first transverse natural frequency
throughout the main shock for towers P2 and P3. Note that g is [Fig. 16(a)], one can see that the change in the first transverse natural
smaller during the larger excitation, i.e., between Frame 2 and Frame frequency is closely related to the behavior of tower–girder trans-
5, and increases gradually toward the end of excitation. Comparing verse connection.
Based on observations of acceleration records and system iden-
tification, one can conclude two different behaviors of pier–girder
and the tower–girder connections. First, in longitudinal direction,
large excitations activate the LBC, causing the girder to slide over
the pier and tower. This condition creates large relative modal dis-
placements on the pier–girder and tower–girder connections. Sec-
ond, the existing lateral gap becomes insufficient owing to large
transverse motion of the girder, and this triggers periodic transverse
pounding between girder and wind shoes. This condition is sup-
Downloaded from ascelibrary.org by Tokyo Univ Seisan Gijutsu on 02/17/14. Copyright ASCE. For personal use only; all rights reserved.

ported by the evidence of smaller tower–girder relative modal


displacement observed on the first transverse mode and the occur-
rence of periodic impulses at the rate of the girder’s first transverse
frequency.

Results from Foreshock and Aftershocks


In addition to the main shock, changes in frequencies during one
seismic event occur only during the first aftershock. Similar to
the main shock, the changes in frequency are also followed by the
change in transverse-vertical coupling pattern. Excitations of other
aftershocks and foreshock were much smaller than the main shock.
Peak accelerations on the footing level for other aftershocks were
mostly less than 10 cm=s2 . In such small excitation, the responses
were likely to remain linear as evidenced by the natural frequencies
that did not change significantly.
Variations of natural frequencies of the first transverse and ver-
tical bending modes with respect to RMS of the excitation amplitude
Fig. 19. Natural frequencies with respect to input amplitude for all for all records are shown in Fig. 18. In this figure, the values of hT
seismic records analyzed; reduction from (a) to (i), respectively in % is and hV between 0.4 and 0.6 are classified as the coupled mode,
14.5, 6.2, 3.1, 4.3, 4.5, 6.9, 1.7, 4, and 6.7 whereas the other values are classified as uncoupled mode. Table 3

Fig. 20. Damping ratios with respect to input amplitude for all seismic records analyzed

© ASCE A4014006-15 J. Bridge Eng.

J. Bridge Eng.
lists the identified natural frequencies and damping for foreshock, Postearthquake Visual Inspection
main shock, and the largest aftershock (AS1). Note that in general
the natural frequencies identified from main shock are smaller The bridge was briefly closed after the earthquake due to an
than the foreshocks and the AS1, especially for the first few lower overturned cargo truck accident on the lower deck. Large girder
modes. The results indicate that natural frequencies of the bridge transverse vibration and high center of gravity of the truck seem-
are generally sensitive to the excitation amplitude, since amplitude ingly caused the truck to become unbalanced and overturn. Other
of the main shock is several times larger than the foreshock and than this, there were no accidents reported, and the bridge did not
aftershock. show any signs of structural damage. Visual inspection was con-
Fig. 19 summarizes the identified natural frequencies for other ducted a few weeks after the earthquake to investigate the condition
modes plotted with respect to RMS of acceleration at the base of of pier–girder and tower–girder connections. Two important find-
P2. The results are obtained from system identification performed ings can be summarized as follows:
separately for each input-output dataset with time window of 50 s. 1. Circular scratch marks were observed on the surface of pier P1
As can be observed in the figures, natural frequencies generally and tower P2 wind shoes, suggesting that the girder experi-
Downloaded from ascelibrary.org by Tokyo Univ Seisan Gijutsu on 02/17/14. Copyright ASCE. For personal use only; all rights reserved.

decrease with the increase of excitation amplitude. Larger re- enced large relative longitudinal movement of about 8–10 cm
ductions are observed in natural frequencies of vertical modes [Fig. 21(b)].
than in torsional modes. The most significant reduction is on the 2. Nuts and several bolts on the lower head of P2 tower link were
second vertical mode (i.e., first asymmetric mode), where the crushed, and signs of scraped paint were seen on the surface of
natural frequency decreases from about 0.52–0.55 Hz for input the bearing’s circular plate [Fig. 21(d)] that could be caused by
RMS less than 3 cm=s2 to 0.47 Hz for input RMS of 24 cm=s2 transverse pounding between tower and girder and by com-
(about 14% reduction). Damping ratio estimates show large var- bination of excessive transverse and vertical motion of girder
iation within 0–6% (Fig. 20), and drawing a clear relationship at the tower–girder connection.
between damping ratios and excitation amplitude from such a large To produce such relative longitudinal displacement, the LBCs
scatter estimates is rather difficult. must have slipped and functioned as a fully hinged connection.

Fig. 21. (a) Schematic figure of wind shoe and tower link; (b) circular scratch marks on the surface of wind shoe as a result of side bearing movement
indicate girder longitudinal movement; distance between the marks and side bearing stationary position is about 8 cm; (c) location of head of tower link
(P2) bottom head is connected to girder, top head is connected to tower; (d) scratch marks on the surface of the bottom link head caused by combination
of vertical and transverse movement of the girder; arrows point to locations of three broken bolts caused by the movement (images by Dionysius
M. Siringoringo)

© ASCE A4014006-16 J. Bridge Eng.

J. Bridge Eng.
Moreover, because wind shoe and girder is originally separated by Acknowledgments
a small transverse gap, the marks created by girder on the surface of
wind shoes would be possible only if the wind shoes and girder The authors gratefully acknowledge Metropolitan Expressway Pub-
collided at some point in time during the main shock; this explains lic Corporation, Tokyo, Japan, for collaboration in this study and for
the presence of periodic impulses on acceleration records. access to the seismic monitoring system. Valuable discussion with
Dr. Masaaki Yabe of Chodai Consulting Engineering is also highly
appreciated.
Conclusions and Recommendations References
Seismic responses of Yokohama Bay Bridge under the 2011 Great Celebi, M. (2006). “Real-time seismic monitoring of the new Cape Girar-
East Japan (Tohoku) Earthquake were analyzed in this paper. Based deau Bridge and preliminary analyses of recorded data: An overview.”
on response analysis and system identification, the following con- Earthquake Spectra, 22(3), 609–630.
Downloaded from ascelibrary.org by Tokyo Univ Seisan Gijutsu on 02/17/14. Copyright ASCE. For personal use only; all rights reserved.

clusions are drawn: Çelebi, M. (2012). “Golden Gate Bridge response: A study with low-amplitude
1. The ground motions in the 2011 Great East Japan (Tohoku) data from three earthquakes.” Earthquake Spectra, 28(2), 487–510.
Earthquake were still below the design and seismic-retrofit Fujino, Y., Kikkawa, H., Namikawa, K., and Mizoguchi, T. (2005). “Seismic
ground motions. The bridge experienced intense shaking for retrofit design of long-span bridges on metropolitan expressways in
about 10 min but did not suffer from any structural damage. Tokyo.” Proc., 6th Int. Bridge Eng. Conf. Transportation Research
2. The bridge responses during the main shock were dominated Board (CD-ROM), Transportation Research Board, Washington, DC.
Ganev, T., Yamazaki, F., Ishizaki, H., and Kitazawa, M. (1998). “Response
by transverse movement of girder and tower, with the max-
analysis of the Higashi-Kobe Bridge and surrounding soil in the 1995
imum transverse displacements on the top of the tower and in Hyogoken-Nambu earthquake.” Earthquake Eng. Struct. Dynam., 27(6),
middle of the girder 55 and 62 cm, respectively. The earth- 557–576.
quake did not cause girder unseating, since the maximum Ide, S., Baltay, A., and Beroza, G. C. (2011). “Shallow dynamic overshoot
longitudinal displacement was only about 15 cm; that is, less and energetic deep rupture in the 2011 Mw 9.0 Tohoku-Oki earthquake.”
than 1.5 m maximum allowable longitudinal displacement. Science, 332(6036), 1426–1429.
3. Response nonlinearity was observed during the main shock Juang, J. N. (1997). “System realization using information matrix.” J. Guid.
and the first aftershock as evidenced by time variation of Control Dyn., 20(3), 492–500.
natural frequencies and change in transverse-vertical coupling Maeda, K., Otsuka, A., and Takano, H. (1991). “The design and construction of
of the first transverse and vertical mode shapes. Behavior of Yokohama Bay Bridge.” Cable-stayed bridges: Recent developments and
their futures, M. Ito, et al., eds., Elsevier Science Publisher B.V.,
the tower–girder and pier–girder connections is thought to be
Amsterdam, Netherlands, 377–395.
the cause of response nonlinearity. Metropolitan Expressway Public Corporation (MPEC). (1991). The Yoko-
4. The LBCs performed satisfactorily during the main shock. hama Bay Bridge, MPEC, Tokyo (in Japanese).
Observations on relative displacement response, character- Siringoringo, D. M., and Fujino, Y. (2006). “Observed dynamic perfor-
istics of acceleration spectra, and type of longitudinal mode all mance of the Yokohama-Bay Bridge from system identification using
suggest that the LBCs functioned as intended. Physical evi- seismic records.” Struct. Contr. Health Monit., 13(1), 226–244.
dence such as circular scratch marks on the surface of wind Siringoringo, D. M., and Fujino, Y. (2007). “Dynamic characteristics of
shoes also support this suggestion. a curved cable-stayed bridge identified from strong motion records.”
5. Recorded accelerations, results of system identification, and Eng. Struct., 29(8), 2001–2017.
visual inspection all indicate the occurrence of transverse Siringoringo, D. M., and Fujino, Y. (2008). “System identification applied to
long-span cable-supported bridges using seismic records.” Earthquake
pounding between girder and tower and on the tower–girder
Eng. Struct. Dynam., 37(3), 361–386.
connections. Despite its occurrence, transverse pounding did Smyth, A. W., Jin-Song, P., and Masri, S. F. (2003). “System identification
not cause significant damage to the connections. Nevertheless, of the Vincent Thomas suspension bridge using earthquake records.”
effect of transverse pounding between girder and tower on the Earthquake Eng. Struct. Dynam., 32(3), 339–367.
bridge should be investigated in greater detail, and necessary Wald, D. J., and Somerville, P. G. (1995). “Variable-slip rupture model of
countermeasures must be taken in anticipation of future larger the great 1923 Kanto, Japan, earthquake: Geodetic and body-waveform
earthquakes. analysis.” Bull. Seismol. Soc. Am., 85(1), 159–177.

© ASCE A4014006-17 J. Bridge Eng.

J. Bridge Eng.

You might also like