Beruflich Dokumente
Kultur Dokumente
Abstract: Tower–girder transverse pounding was observed on the Yokohama Bay Bridge wind tongue during the 2011 Great East Japan
Earthquake. Damage to the wind tongue due to pounding can potentially lead to damage of other members. However, pounding was not taken
into consideration in the design, and the process of how the damage occurs after the wind tongue damage has not been clarified. Therefore, a
way to model the damage process involving tower–girder transverse pounding is needed. In this study, a frame model of the Yokohama Bay
Bridge was constructed to reproduce the observed damage that occurred as a result of the pounding. A multiscale model was then constructed
by combining the global frame model with a local finite-element model consisting of a shell and solid elements. The model was validated
through comparison with actual measurements. Dynamic analysis of the multiscale model clarified the deformation of the wind tongue during
large earthquakes and the resulting damage to the bridge. The model demonstrated that ductile damage can occur at the base of the wind tongue
during a Level 2 earthquake and that the tower link can then drop off due to its large displacement after the wind tongue damage. DOI:
10.1061/(ASCE)BE.1943-5592.0001336. © 2018 American Society of Civil Engineers.
Transverse direcon
Downloaded from ascelibrary.org by TULANE UNIVERSITY LIBRARY on 12/15/18. Copyright ASCE. For personal use only; all rights reserved.
Cross beam
Wind tongue
Wind shoe
Tower link
Rotate around
longitudinal axis
and transverse axis
T1(x,y) T2(x,y)
To Honmoku To Oguro
200m 460m 200m T1(x,y) T2(x,y)
Yokohama Yokohama
T3(x,y) T4(x,y) Harbor Harbor
S5(x,y,z) S7(x,y,z) T3L(x,y) T3R(x,y) T4L(x,y) T4R(x,y)
S1(x,y,z) S3(x,y,z) S4(y,z) S6(y,z) S9(x,y,z)
K4(x,y,z) K6(x,y,z)
Yokohama
Harbor
S1(x,y,z) S3(x,y,z) S5L(x,y,z) S6L(y,z) S7(x,y,z) S8L(y,z) S9(x,y,z) K3(x,y,z)
S2(y,z) S4(y,z) K5(x,y,z)
P2 P3
S8R(y,z)
S5R(x,y,z) S6R(y,z)
Target Bridge and Observed Response Great East Japan Earthquake, bolts were found to be fractured due to
the large transverse displacement of the tower link.
Yokohama Bay Bridge, which opened in 1989, is a three-span con- Although the latest technologies were used at the time of construc-
tinuous cable-stayed bridge with a center span of 460 m, a side span tion, the design of this bridge, which was constructed before the 1995
of 200 m, and a total length of 860 m. The main girder is a double- Kobe earthquake, did not consider Level 2 earthquakes. After the
deck truss structure, and the upper chord members have a box sec- 1995 Kobe earthquake, a seismic retrofit was carried out based on
tion in which the upper flanges are integrated with the steel deck. Six dynamic analysis using the frame model and static analysis using a
lanes of the Yokohama Highway Gulf line run on the upper layer, detailed shell model. Uplift prevention cables, displacement limiters,
and National Route No. 357 runs on the lower layer; the national and step prevention structures were installed as a result of this retrofit.
route has two lanes under this tentative configuration. The main tow- The bridge has a monitoring system with 85 servo type acceler-
ers are H-shaped rigid frame structures with a height of 172 m and a ometers at 36 locations. Thirty-channel accelerometers were in-
width of 29.25 m. The lower lateral beam of each main tower has a stalled along the girder, 28-channel accelerometers on towers and
convex wind tongue. The transverse displacement is constrained by piers, and 27-channels accelerometers on the ground and foundation
wind shoes that are installed on the main girder crossbeam. The structure. The sampling frequency of measurement was 100 Hz.
girder hangs from the towers by tower links. This type of link con- Fig. 2 provides the sensor layout.
nection plays a seismic-isolation function by moving like a pendu- Table 1 presents the list of earthquake responses observed at
lum in the longitudinal and transverse directions. After the 2011 the Yokohama Bay Bridge during the 2011 Great East Japan
1 base of the piers and at the base of the main towers were used as
10
input signals, and acceleration responses observed at other locations
as output signals.
0
Fig. 4 gives the frequency change of the first transverse mode
Acceleration[m/ s2]
10
during the main shock. In the initial stage, the boundary condition
-1
of the girder at the wind tongue was considered as free in the trans-
10 verse direction before wind tongue contact of the wind shoe. Once
these parts contacted each other due to the pounding, the girder
-2 Main shock motion was constrained; the frequency then decreased to 0.27 Hz in
10
aftershock1 the last stage because the response became smaller and pounding
aftershock2 䡚㻥 stopped. Similar pounding responses have been observed on other
Retrofit design wave
-3
10 - 1 bridges (Ganev et al. 1998; Smyth et al. 2003). However, the pound-
0 1
10 10 10 ing response of this bridge was not reproduced by numerical simu-
Period[s] lation in the design model that was used for the retrofit design. To
reproduce the frequency change, pounding must be considered.
Fig. 3. Response spectra of the observed responses and the retrofit
design wave.
Modeling of Pounding Force in the Global
Frame Model
Earthquake main shock and aftershocks. Ten earthquakes including
the main shock and nine aftershocks were observed. Fig. 3 gives the In order to reproduce the response that occurred in the 2011 Great
acceleration response spectra of the seismic motions that were East Japan Earthquake, a frame model of the Yokohama Bay Bridge
observed at the engineering foundation surface (G1) and the was constructed based on the model that was used for the retrofit
response spectra of the input ground motion for the retrofit design. design. The constructed frame model is presented in Fig. 5. Truss
As presented in Fig. 3, the amplitude levels of Aftershocks 2–9 members were modeled by beam elements, and cables were mod-
were smaller than that of the main shock and Aftershock 1. Besides, eled by truss elements that resist only tension. The initial axial force
the response spectrum of the main shock was smaller than the input of each member, which contributes to the geometric stiffness matrix,
wave used for the retrofit design. This result indicates that the earth- was introduced based on the gravity force. The lower road floor was
quake responses observed during the 2011 Great East Japan modeled based on the tentative configuration of the two-lane road.
Earthquake did not cause serious damage to the bridge. The corners of the piers and the main towers were modeled as rigid
On the other hand, circular contact scratches were observed after connections. The piers and the horizontal beam of the towers were
the 2011 Great East Japan Earthquake. In addition, periodic spikes modeled by linear beam elements, and the other parts of main towers
were observed during the main shock and aftershocks (Siringoringo were modeled by nonlinear beam elements. The nonlinearity was
et al. 2014). These observations indicate that pounding in the trans- expressed in terms of trilinear M- f relationship. The base of the
verse direction occurred at the wind tongue. As for the longitudinal tower and piers were modeled as SR springs to represent the defor-
direction, pounding was not observed and the isolation system of mation of the base structure. The mass of the step prevention struc-
the link bearing is considered to have functioned appropriately ture and the displacement limiter were taken into account. The uplift
(Siringoringo et al. 2014). To investigate the effect of pounding, the prevention cables were modeled by the truss elements based on the
observed response was divided using a 50-s short-time moving win- design drawings. In order to reproduce the pounding between the
dow with 50% overlap, and a system identification method was wind shoe and the wind tongue, a pounding spring was introduced.
applied to investigate the frequency change in the time domain. For
the system identification method, the System Realization using
Modeling of Pounding Force
Information Matrix (SRIM) was used (Juang 1997; Siringoringo
and Fujino 2006, 2008). Vibration characteristics such as natural The pounding between the wind tongue and the wind shoe was eval-
frequencies and mode shapes were obtained. Although the pound- uated by introducing the nonlinear pounding spring. The pounding
ing occurred, the global responses had large periodic components; was modeled phenomenologically by considering following phenom-
therefore, an equivalent linear system was assumed and the system ena. In the case of small earthquakes, the pounding was expressed by
identification was applied. Acceleration responses observed at the a linear pounding spring; however, in large earthquakes, not only does
Frequency[Hz]
0.32
0.3
0.28
0.26
0 100 200 300 400 500 600
Time[s]
Observed Design model Frame model Multi- scale model
Fig. 4. Comparison of the frequency change between the observed data and analysis results.
Downloaded from ascelibrary.org by TULANE UNIVERSITY LIBRARY on 12/15/18. Copyright ASCE. For personal use only; all rights reserved.
7650 DOF
1279 nodes
2253elements
z
y x
where F = restoring force of the pounding spring; x = displacement where Fþi and F–i = force of ith yield point; ki = ith stiffness; and
of the pounding spring; xþ 0 and x–0 = gap of the wind tongue in the xþi = displacement of ith yield point. Because a hexa-linear model
positive and negative directions, respectively; and k1 = initial stiff- is employed, i = 1, 2,…5. Fþi, F–i, xþi, x–i are initially set as
ness of the wind tongue.
Fþi ¼ Fi (4)
xþ0 ¼ ug
Fi ¼ Fi (5)
x0 ug (2)
xþi ¼ xi (6)
where ug = gap of the wind tongue. After the material goes into the
plastic range
xi ¼ xi (7)
(
Fþi þ kiþ1 ðx xþi Þ ðxi < x < xiþ1 Þ
F¼ (3) where Fi and xi = yielding point defined from the skeleton curve pre-
Fi þ kiþ1 ðx xi Þ ðxðiþ1Þ < x < xi Þ sented in Fig. 7(a).
(a) (b)
(
In addition, the unloading stiffness was the same as the initial xi ðxmin > xi Þ
stiffness k1, and the yielding points were updated by following xi ¼ (18)
xunloading ðxmin xi Þ
equations.
Consider the unloading process after the force takes positive
peak value, Funloading at xunloading, xþ0, and x–0 were updated as xþi ¼ xi þ 2xi (19)
follows:
Funloading where xmin = minimum displacement; and Fmin = minimum force.
xþ0 ¼ xunloading (8) The Interactive Simulator and Analyzer for Structures (ISAS)—
k1
which was developed by the authors (Takeda 2017) using
x0 ¼ xþ0 2ug (9) MATLAB® (version R2013b)—was used for the analysis of the
frame model.
The updated displacements of ith yield point, xþi and x–i, were
further updated as Reproduction of the Pounding Response by the Frame
( Model
xþi ðxmax < xþi Þ
xþi ¼ (10) To check the validity of the model, an eigenvalue analysis was con-
xunloading ðxmax xþi Þ
ducted; obtained natural frequencies and mode shapes were com-
pared with the vibration characteristics identified from measure-
xi ¼ xþi 2xi (11) ment data. Table 2 gives the comparison of the natural frequencies
identified from each earthquake. From this table, the natural fre-
where xmax = maximum displacement. The updated force of ith quencies of the sway mode, the first transverse mode, and the sec-
yield point, Fþi, and F–i, were updated as ond vertical mode obtained from eigenvalue analysis of the frame
( model were in the variation range of the identified frequencies. In
Fþi ðFmax < Fi Þ
Fþi ¼ (12) order to investigate the accuracy of the mode shape, the Modal
Funloading ðFmax Fi Þ Assurance Criterion (MAC) was calculated using the following
equation (Allemang and Brown 1982):
Fi ¼ Fþi 2Fi (13) j f Tobserve f model j
MAC ¼ (20)
f Tobserve f observe ð f model f model Þ
where Fmax = maximum force. Consider then the unloading process
after the force takes a negative peak value, Funloading.
where f observe and f model = modal vector identified from the
Funloading observed data and the modal vector identified from the eigenvalue
x0 ¼ xunloading (14)
k1 analysis, respectively.
Table 3 gives the MAC of each identified mode. From this table,
it can be seen that the MAC of the first mode is around 0.9 and that
xþ0 ¼ x0 þ 2ug (15)
the mode shape of constructed frame model is in good agreement
with the mode shape identified from the observed data.
( Then, dynamic analysis using the main shock ground motion
Fi ðFmin > Fi Þ
Fi ¼ (16) was conducted to check the reproducibility of the observed
Funloading ðFmin Fi Þ responses. Triaxial accelerations as observed at K2, K4, and K6
(Fig. 2) were used as input waves. The input wave at Pier P4 was
Fþi ¼ Fi þ 2Fi (17) not directly available because the accelerometer at K8 was not
Damping
Main shock 0.142 0.323 0.313 0.493 0.03
Aftershock 1 — 0.323 0.335 0.488
Aftershock 2 — 0.261 0.342 0.532 0.02
Aftershock 3 — 0.266 0.342 —
Aftershock 4 — 0.283 0.343 0.533
0.01
Aftershock 5 — 0.265 0.349 0.419
Aftershock 6 — 0.288 0.347 —
Downloaded from ascelibrary.org by TULANE UNIVERSITY LIBRARY on 12/15/18. Copyright ASCE. For personal use only; all rights reserved.
Table 3. MAC of each identified mode 2–9. Fig. 8 provides the comparison of the observed frequency-
damping relationship with Rayleigh damping. Rayleigh damping pa-
First First Second
rameters of the foundations were set to large values to reduce the high
Earthquake Sway transverse vertical vertical
frequency component that corresponds to the base mode. The whole
Main shock 0.809 0.975 0.815 0.960 time history response was divided using the short moving time win-
(0.870) (0.968) (0.815) (0.960) dow of 50 s; frequency changes were identified by applying the
Aftershock 1 — 0.974 0.763 0.916 SRIM on each response. The results were reasonably consistent with
(0.966) (0.763) (0.915) the estimates reported by Siringoringo et al. (2014). The overlap ratio
Aftershock 2 — 0.998 0.968 0.527 of the time window was 50%.
(0.998) (0.968) (0.528) Fig. 4 provides the comparison of the frequency change in the
Aftershock 3 — 0.996 0.745 0.706 time domain during the main shock. The design model, which does
(0.995) (0.745) (0.705) not take into consideration the pounding phenomena, did not repro-
Aftershock 4 — 0.990 0.987 0.807 duce the observed frequency change. On the other hand, the result
(0.989) (0.987) (0.807) of the constructed frame model was in good agreement with the
Aftershock 5 — 0.989 0.943 0.879 observed frequency change. The constructed frame model did
(0.992) (0.943) (0.879) reproduce the observed pounding responses.
Aftershock 6 — 0.994 0.966 — The reproducibility of the time history response was also investi-
(0.991) (0.966) — gated by comparing the observed response and the analysis
Aftershock 7 — 0.996 0.948 — response. Figs. 9(a and b) give the transverse acceleration time his-
(0.996) (0.948) — tory and Fig. 9 (c) gives the Fourier spectra of the transverse accel-
Aftershock 8 — 0.966 0.725 0.739 eration at the center of the span (S5). Analysis results roughly repro-
(0.963) (0.725) (0.739) duced the observed response and maximum value and the time of
Aftershock 9 — 0.969 0.891 0.878 the maximum value. Moreover, the first mode amplitude of the
(0.970) (0.891) (0.878) Fourier spectrum was in good agreement with the observed results.
Average (Aftershock 2–9) 809 0.987 0.896 0.756 Fig. 9(d) gives the transverse acceleration at the wind tongue
(0.870) (0.987) (0.897) (0.756) (T5), and Fig. 9(e) presents a zoom view of Fig. 9(d) from 145 to
Note: First value is the MAC of the frame model; value in parentheses is 155 s. The timing of the peak values (i.e., the timing of the pound-
the MAC of the multiscale model. ings) and the maximum peak value were reproduced well. The dis-
crepancy in the timing of the maximum peak value was considered
to be due to high frequency components. As a result, it can be con-
working. Horizontal acceleration and vertical acceleration is cluded that the constructed frame model can reproduce the observed
affected by the surrounding soil and stiffness of the basement, pounding response. In addition, Fig. 9(f) provides the comparison
respectively. Therefore, the horizontal acceleration at K6—the of the acceleration Root Mean Square (RMS) between the observed
functioning measurement point just beneath the pile cap closest to data and analysis results at the center of the girder. The design model
K8—was used as a substitute for the horizontal acceleration at underestimated the acceleration RMS during the large earthquakes
K8. The vertical acceleration at K2 was used as a substitute for with pounding. On the other hand, the estimation accuracy of the
the vertical acceleration at K8 because Piers P1 and P4 were simi- constructed model that can take into consideration the pounding is
lar and the vertical ground motions around the piers were consid- higher than that of the design model. From these results, by consider-
ered similar. ing the pounding, it can be seen that the constructed frame model can
Elemental Rayleigh damping was applied for the dynamic analy- reproduce not only the small earthquake response with the same ac-
sis. Rayleigh damping parameters of the elements excluding the foun- curacy as the design model but also the pounding response with a
dations were defined by the least-squares method based on the higher accuracy than the design model. To estimate the response
frequency-damping relationship that was identified from Aftershocks under large earthquakes, the pounding must be considered.
0.1
Acceleration[m/ s2]
Acceleration[m/ s ]
Fourier amplitude
2
2 2
0 0 0.05
-2 -2
Downloaded from ascelibrary.org by TULANE UNIVERSITY LIBRARY on 12/15/18. Copyright ASCE. For personal use only; all rights reserved.
0
0 200 400 600 130 140 150 160 170 0 0.5 1 1.5 2
Time[s] Time[s] Frequency[Hz]
4 4
Acceleration[m/ s ]
Design model
Acceleration[m/ s ]
2
2
2
Analysis RMS[m/ s ]
2 2 0.6 Constructed model
0 0
0.4
-2 -2
-4 -4 0.2
0 200 400 600 145 150 155
Time[s] Time[s]
0
0 0.2 0.4 0.6
Observed MAX=2.28 Analysis MAX=3.01 2
Observed RMS[m/ s ]
(d) (e) (f)
Fig. 9. Comparison of the transverse acceleration response and the observed data and analysis: (a) acceleration at S5; (b) zoom view of a; (c) Fourier
spectra at S5; (d) acceleration at T5; (e) zoom view of d; and (f) RMS of each earthquake at the center of the girder.
MPC beam
Gap
5mm
Contact
surface
Crossbeam MPC beam
Wind tongue
Wind shoe
Hanging member
MPC beam
Crossbeam
MPC beam
by multipoint constraint multipoint constraint (MPC) equations, responses of the multiscale model. The first 500-s part of the main
which are widely used in constructing multiscale models (Shim shock was used as the input wave in order to compare the increasing
et al. 2002; Li et al. 2007). and decreasing of the frequency due to pounding. Multipoint input
The locations of the multipoint constraints were the boundaries was applied as in the dynamic analysis of the frame model.
of the beam elements and shell elements at the towers, link brackets, In order to examine whether the frequency change during the
and crossbeams. To consider the pounding, the contact surface is main shock is reproduced, the natural frequencies were identified
defined as between the wind tongue and wind shoe. The constructed with SRIM using the moving window to the acceleration response
multiscale model is presented in Fig. 10. FE analysis software, of multiscale model. The frequency change is presented in Fig. 4
ABAQUS (version 6.14) was used for the multiscale model analysis. with the green line. The increasing and decreasing of the frequency
In order to clarify the validity of the multiscale model, an eigen- due to pounding is reproduced similar to as seen for the frame
value analysis was conducted, and the obtained natural frequencies model analysis.
were compared with the measured data. The comparison of the nat- Fig. 11(a) provides the transverse acceleration response at the
ural frequencies between the eigenvalue analysis results and all center of the girder. The multiscale model reproduces the observed
observed natural frequencies is presented in Table 2. response in the same manner as the frame model. Moreover, the
4
Acceleration[m/ s ]
2
Fourier amplitude
0.1
2
0
0.05
-2
-4 0
100 110 120 130 140 150 0 0.2 0.4 0.6 0.8 1
Time[s] Frequency[Hz]
Observed Multi- scale model Frame model Observed Multi- scale model Frame model
(a) (b)
0.04
Acceleration[m/ s2]
Fourier amplitude
2 0.03
0 0.02
0.01
-2
0
100 110 120 130 140 150 0 0.2 0.4 0.6 0.8 1
Time[s] Frequency[Hz]
Observed Multi- scale model Frame model Observed Multi- scale model Frame model
(c) (d)
0.02
Acceleration[m/ s ]
2
Fourier amplitude
5 0.015
0 0.01
-5 0.005
0
100 110 120 130 140 150 0 0.2 0.4 0.6 0.8 1
Time[s] Frequency[Hz]
Observed Multi- scale model Frame model Observed Multi- scale model Frame model
(e) (f)
Fig. 11. Comparison of the transverse acceleration response at S5 during the main shock: (a) time history at S5; (b) Fourier spectra at S5;
(c) time history at S3; Fourier spectra at S3; (e) time history at T5; and (f) Fourier spectra at T5.
Link Main
Body
Pin
Downloaded from ascelibrary.org by TULANE UNIVERSITY LIBRARY on 12/15/18. Copyright ASCE. For personal use only; all rights reserved.
(a) (b)
Fig. 12. Structures of the tower link: (a) members of the tower link; and (b) rotation of the tower link and rotation limit.
Fourier amplitudes in Fig. 11(b) reveal that the first mode, which is
the most dominant component, was in agreement with the observed
response. In addition, as presented in Figs. 11(c–f), the responses
were also reproduced at the pounding location. The constructed
multiscale model can, therefore, reproduce the global response with
an accuracy similar to that of the frame model.
Table 4. Maximum transverse displacement and rotation angle of the over analysis was conducted using the detailed tower link model.
tower link
Because the influence of the contact between the bolt and the link
Location Maximum displacement (m) Rotation angle (°) member on the global response of tower link was considered to be
much smaller than the influence of the contact between the wind
P2 port side 0.0563 1.61
tongue and the wind shoe, the maximum transverse displacement of
P2 sea side 0.0563 1.61
the tower link obtained from the global dynamic analysis was used as
P3 port side 0.0503 1.44
the forced displacement of the pushover analysis. The forced displace-
P3 sea side 0.0504 1.44
ment was uniformly applied to the lower pin of the tower link.
Fix
Bolt
L-plate
A
Outer bearing
Link
bracket
Pin (b)
L-plate
Pin Inner bearing
Structure of link bearing
B
Body
Link plate
(a)
(c)
Fig. 14. Detailed model of the tower link: (a) overall view; (b) side view; and (c) cross-section of A in b.
+1.059e+09
+9.704e+08
+8.821e+08
+7.939e+08 L-plate
+7.057e+08 Contact locaon
+6.175e+08
+5.293e+08 Bolt
+4.411e+08 of analysis
+3.529e+08
+2.646e+08
+1.764e+08
+8.821e+07
+0.000e+00
Translaon
Rotaon
(a) (b)
Fig. 15. Contact condition obtained from the pushover analysis: (a) contact condition of the bolt and the L-plate; and (b) contact location between the
bolt and the L-plate.
16
14 I- III- I
I- III- II
Acceleration[m/ s2]
12 I- III- III
10 WAVE3
WAVE6
8 Sagami Trough
II- III- I
6 II- III- II
4 II- III- III
Design spectrum (Type1)
2 Design spectrum (Type2)
0 -1 0 1
10 10 10
Period[s]
Table 6. Comparison of maximum pounding force between the multiscale model and the frame model for each earthquake
1
Downloaded from ascelibrary.org by TULANE UNIVERSITY LIBRARY on 12/15/18. Copyright ASCE. For personal use only; all rights reserved.
Displacement[m]
0.5
- 0.5
-1
0 50 100 150 200
Time[s]
Without damage progress With damage progress
(c)
Fig. 17. Ductile damage of the wind tongue: (a) distribution of the ductile damage variable; (b) distribution of stiffness reduction ratio after damage;
and (c) comparison of displacement of the wind tongue in the damage progress with that in the case of progress without damage.
waves used for the retrofit design, they were based on the epicen-
ter model of the Kanto earthquake that occurred in 1923. The
plate boundary earthquake motion was set so that the seismic
wave propagation after the destruction of asperity would head to-
ward the bridge location. The size of the asperity and its position
that would produce large seismic wave components of the natural
periods of the bridge dominant modes were employed (Komori
et al. 2005a). Fig. 16 provides the acceleration response spectra
of Level 2 input waves.
Based on the regional correction coefficient, Type 1 earth-
quakes, which are plate boundary earthquake, and Type 2 earth-
quakes, which are inland earthquakes, were amplified 1.2 times and
1.0 times, respectively (Japan Road Association 2012). The input
direction of the Sagami Trough was three directions. The input
direction of the other waves was the transverse direction, which was
the most severe direction for the pounding effect.
Although the degrees of freedom of the multiscale model were
much lower than for models consisting only of shell or solid ele-
ments, the dynamic analysis using the multiscale model was time
consuming. To reduce the calculation time, the pounding force dur-
ing the Level 2 earthquake was first estimated using the frame
model. Then, the detailed deformation and damage conditions of Fig. 18. Tower–tower link integrated model: (a) overview; and
the wind tongue were estimated by inputting into the multiscale (b) zoom view of the tower link.
model the earthquakes that could cause the damage to the wind
tongue. In this analysis, the same wave was used as input to each
foundation. In order to estimate the deformation and damage to the wind
Table 6 provides the comparison of the maximum pounding tongue, dynamic analysis was conducted by inputting Level 2 seis-
force estimated by the frame model analysis for each earthquake mic motion into a multiscale model. Level 2–design earthquakes
and the capacity of the wind tongue. The average value of the I-III-I to I-III-III and II-III-I to II-III-III, which can cause the dam-
maximum pounding force in WAVE3, WAVE6, and the Sagami age to the wind tongue according to frame model analysis, were
Trough did not exceed the capacity of the wind tongue. On the used as the input wave. WAVE3, WAVE6, and the Sagami Trough
other hand, the average value of the maximum pounding force of were not used as the input wave of the multiscale model analysis
Level 2 earthquakes as defined in the specification for highway because the average maximum pounding force did not exceed the
bridges exceeded the capacity of the wind tongue. Therefore, it capacity of the wind tongue according to the frame model analysis.
can be said that the damage to the wind tongue can occur in Level In this analysis, the progress of the damage to the wind tongue was
2–design earthquakes. not considered.
fracture was judged by the state variable of ductile fracture crite- of the elements. In this analysis, when the ductile damage variable
rion. The ductile fracture occurred when the ductile damage reached 1, the stiffness of the element was reduced to 1% of the
Fig. 19. Stress distribution of the tower link: (a) overall view of Case 1; (b) cross-section view of Case 1; (c) upper bearing of Case 1; (d) lower bear-
ing of Case 1; (e) overall view of Case 2; (f) cross-section view of Case 2; (g) upper bearing of Case 2; and (h) lower bearing of Case 2.
Remaining distance[m]
0.5 0.3
0.4 0.2
0.3 0.1
0
0.2
- 0.1
0.1 0 0.1 0.2 0.3 0.4 0.5 0.6
0 0.1 0.2 0.3 0.4 0.5 0.6 Displacement[m]
Displacement[m] Static Quasi- static Remaining distance=0[m]
(a) (b)
Downloaded from ascelibrary.org by TULANE UNIVERSITY LIBRARY on 12/15/18. Copyright ASCE. For personal use only; all rights reserved.
Fig. 21. Relationship between the remaining distance and the displacement of the tower link: (a) Case 1; and (b) Case 2.
upper pin of the tower link comes completely out of the link bracket vehicle frontal impact.” J. Constr. Steel Res. 102 (Nov): 190–203.
when the tower link moves away from the main tower, and there is a https://doi.org/10.1016/j.jcsr.2014.07.005.
possibility of the tower link dropping off after the damage to the Bi, K., and H. Hao. 2013. “Numerical simulation of pounding damage
wind tongue during Level 2 earthquakes. to bridge structures under spatially varying ground motions.” Eng.
Struct. 46 (Jan): 62–76. https://doi.org/10.1016/j.engstruct.2012.07
.012.
Conclusion Chen, L., Y. Xiao, G. Xiao, C. Liu, and A. K. Agrawal. 2015. “Test and
numerical simulation of truck collision with anti-ram bollards.” Int.
J. Impact Eng. 75 (Jan): 30–39. https://doi.org/10.1016/j.ijimpeng
A global frame model that took into consideration the pounding
.2014.07.011.
and the hysteretic behavior of the wind tongue was constructed Ganev, T., F. Yamazaki, H. Ishizaki, and M. Kitazawa. 1998. “Response
for this study. A multiscale model was then constructed by com- analysis of the Higashi-Kobe Bridge and surrounding soil in the 1995
bining the global frame model with detailed local shell and solid Hyogoken-Nambu Earthquake.” Earthquake Eng. Struct. Dyn. 27 (6):
models. Numerical simulation using the observed earthquakes 557–576. https://doi.org/10.1002/(SICI)1096-9845(199806)27:6<557::
and Level 2 earthquakes provided the following findings: AID-EQE742>3.0.CO;2-Z.
• The vibration characteristics and the dynamic response of the Japan Road Association. 2012. Specifications for highway bridges part 5
constructed frame model and multiscale model based on the seismic design. [In Japanese.] Tokyo: Japan Road Association.
drawings are in good agreement with the measurement results. Japanese Society of Steel Construction (JSSC). 2000. Stainless steel con-
• The frequency change that occurred during the 2011 Great struction materials and building standards law. [In Japanese.] Tokyo:
East Japan Earthquake and that could not be reproduced by Japanese Society of Steel Construction. http://www.jssc.or.jp/ssba
/generalize/generalize.html.
the design model was reproduced after taking the pounding
Juang, J. N. 1997. “System realization using information matrix.” J.
into consideration. Guidance, Control Dyn. 20 (3): 492–500. https://doi.org/10.2514/2
• The mechanism of the bolt fracture of the tower link that
.4068.
occurred during the 2011 Great East Japan Earthquake was Komori, K., H. Kikkawa, N. Odagiri, T. Kinoshita, T. Mizoguchi, Y.
revealed using the detailed tower link model analysis, which Fujino, and M. Yabe. 2005a. “Basic principles and design ground
was based on the dynamic analysis results of the multiscale motions in seismic retrofit design of large cable-supported bridges
model. on the Tokyo Metropolitan Expressway.” [In Japanese.] Doboku
• The dynamic analysis using the multiscale model revealed that Gakkai Ronbunshu 2005 (794): 1–19. https://doi.org/10.2208/jscej
the pounding force exceeded the capacity of the wind tongue .2005.794_1.
and the ductile damage occurred at the base of the wind tongue Komori, K., H. Kikkawa, N. Odagiri, T. Kinoshita, T. Mizoguchi, Y.
in the case of Level 2–design earthquakes. Fujino, and M. Yabe. 2005b. “Study on the seismic retrofit of large
• As a result of the static analysis using the tower–tower link cable-supported bridges on the Tokyo Metropolitan Expressway.” [In
Japanese.] Doboku Gakkai Ronbunshu 2005 (801): 1–20. https://doi.org
integrated model, the drop off of the tower link was deter-
/10.2208/jscej.2005.801_1.
mined to have occurred after the damage to the wind tongue in Li, Z. X., T. Q. Zhou, T. H. T. Chan, and Y. Yu. 2007. “Multi-scale numeri-
the case of the design earthquake I-III-II. cal analysis on dynamic response and local damage in long span
bridges.” J. Eng. Struct. 29 (7): 1507–1524. https://doi.org/10.1016/j
.engstruct.2006.08.004.
Acknowledgments Li, Z. X., T. Q. Zhou, T. H. T. Chan, and Y. Yu. 2009. “Concurrent multi-
scale modeling of civil infrastructures for analysis on structural dete-
The authors wish to thank Mr. Namikawa (Tokyo Metropolitan rioration—Part I: Modeling methodology and strategy.” Finite Elem.
Expressway) and Dr. Yabe (CHODAI). This work was supported Anal. Des. 45 (11): 782–794. https://doi.org/10.1016/j.finel.2009.06
by JSPS KAKENHI Grant JP16J09550. .013.
Shim, K. W., D. J. Monaghan, and C. G. Armstrong. 2002. “Mixed dimen-
sional coupling in finite element stress analysis.” J. Eng. Comput. 18
References (3): 241–252. https://doi.org/10.1007/s003660200021.
Siringoringo, D. M., and Y. Fujino. 2006. “Observed dynamic performance
Allemang, R. J., and D. L. Brown, 1982. “A correlation coefficient for of the Yokohama-Bay Bridge from system identification using seismic
modal vector analysis.” In Proc., 1st Int. Modal Analysis Conf., 110– records.” J. Struct. Control and Health Monit. 13 (1): 226–244. https://
116. Gijón, Spain: International Operational Modal Analysis doi.org/10.1002/stc.135.
Conference. Siringoringo, D. M., and Y. Fujino. 2008. “System identification applied to
Al-Thairy, H., and Y. C. Wang. 2014. “Simplified FE vehicle model for long-span cable-supported bridges using seismic records.” Earthquake
assessing the vulnerability of axially compressed steel columns against Eng. Struct. Dyn. 37 (3): 361–386. https://doi.org/10.1002/eqe.758.