Soil Dynamics and Earthquake Engineering 82 (2016) 111–122

Contents lists available at ScienceDirect

Soil Dynamics and Earthquake Engineering

journal homepage: www.elsevier.com/locate/soildyn

Shaking table test of a multi-story subway station under pulse-like

ground motions
Zhiyi Chen a,b,n, Wei Chen b, Yueyang Li b, Yong Yuan a,b
a
State Key Laboratory of Disaster Reduction in Civil Engineering, Tongji University, Shanghai 200092, China
b
Department of Geotechnical Engineering, Tongji University, Shanghai 200092, China

art ic l e i nf o a b s t r a c t

Article history: A series of shaking table tests were conducted to investigate the effect of pulse-like ground motion on a
Received 13 April 2015 multi-story subway station. Dynamic response data, including internal forces, column drift, and settle-
Received in revised form ment and deformation of the soil were obtained and analyzed. Results show that the pulse-like ground
16 November 2015
motion increases dynamic responses of the subway station and surrounding soils mainly owing to its
Accepted 7 December 2015
inherent rich low-frequency component and high energy. In terms of the structure, central columns,
especially central columns on a floor with large story height, are vulnerable components of a multi-story
Keywords: subway station. Both the dynamic earth pressure and the deformation mode of the side wall were
Shaking table test analyzed.
Multi-story subway station
& 2015 Elsevier Ltd. All rights reserved.
Pulse-like ground motion
Dynamic responses

1. Introduction damage during a strong earthquake. It is commonly believed that

the centrifuge test is an attractive way for seismic performance
With the rapid development of the economy and society in evaluations due to its ability of reproducing the in-site stress state
China, modern underground transportation, represented by the of soils. Researches have been conducted by using a centrifuge,
subway, is continuously developing towards having a deeper and and good results were obtained [6,7]. In addition, shaking table is
multilevel structural form. Huaihai Road Station on metro line 13 subtlety in loading, control and observation [8]. Hence, shaking
in Shanghai, for example, is a six-story island-platform station table test is also a common way for studying seismic performance
having height of nearly 30 m and diaphragm walls that are 71 m of underground structures [6,9,10]. These studies are of great help
deep [1]. More problems tend to arise for a deeper structural form to understand soil–structure interaction or responses of structures.
that has multiple layers and a larger story height. First, the water In studies of superstructures, it is also found that pulse-like
and earth pressures imposed on side walls of the structure ground motions may induce more severe damage to structures
increase with an increase in depth. Second, owing to the accu- compared with other ground motions, such as far-field ground
motions. Pulse-like ground motion is defined as ground motion
mulation of load transferred from top to bottom, the axial com-
whose PGV/PGA (the ratio of peak ground velocity to peak ground
pression ratio of central columns increases. Third, the structural
acceleration) is greater than 0.2 while ordinary ground motion has
configuration tends to be more complicated so as to provide
a ratio smaller than 0.15 [11]. If the rupture propagates in the
multiple functions, which increases the number of latent vulner-
direction of the recording station, coherently traveling long-period
able points. Finally, a large story height greatly reduces the lateral
waves produce high ground velocities and large displacements in
stiffness of central columns and side walls. Hence, the seismic the fault-normal direction [12], and most of the seismic energy in
performance of such an underground structure is worthy of ground motion is concentrated in the pulse [13]. Many studies
attention. have verified the effects of pulse-like ground motion on the
In recent years, centrifuge and shaking table tests have been superstructure. Bertero et al. [14] showed that pulse-like ground
conducted for subway stations to study the seismic performance motion can induce a dramatically strong response in fixed-base
and failure mechanism of their underground structures [2–5]. buildings. Anderson and Bertero [15], in their study of the non-
Results show that underground structures may suffer severe linear dynamic response of a 10-story steel frame, revealed that
the lower floors of buildings with such structure can suffer great
n
Corresponding author at: Department of Geotechnical Engineering, Tongji
damage if subjected to pulse-like ground motion. Makris and Black
University, Shanghai 200092, China. [16] found that local, distinguishable acceleration pulses result in
E-mail address: zhiyichen@tongji.edu.cn (Z. Chen). unusual demands of structures. Sehhati et al. [17] stated that

http://dx.doi.org/10.1016/j.soildyn.2015.12.002
0267-7261/& 2015 Elsevier Ltd. All rights reserved.
112 Z. Chen et al. / Soil Dynamics and Earthquake Engineering 82 (2016) 111–122

pulse-like ground motions impose a larger ductility demand on a which supported the upper ring plate, to allow the ring plate to
structure compared with ordinary ground motions. Additionally, deform laterally. To minimize the relative slip between the soil and
studies have been conducted on the effect of pulse-like ground the container on the base surface, crushed rock was bonded to the
motions on isolated structures and bridges [18–22]. With regard to base steel plate to roughen the surface. In Ref. [24], Lu et al. con-
underground structures, Chen and Wei [23] studied the effect of ducted three free field-shaking table tests to verify the boundary
pulse-like ground motion on mountain tunnels and concluded that effect of the flexible container, which was the same one used
the velocity pulses are the main factor determining damage to herein. They indicated that the boundary effect can be ignored
tunnel linings. However, from the perspective of the structural when the distance between the structure and the boundary was
form, the subway station has a framed structure. Hence, the sub- more than 600 mm. The distance in this paper's tests was more
way station and tunnel differ in terms of their mechanical and than 1.2 m. Hence, the effect of boundary on dynamic responses of
vibration characteristics. Furthermore, the framed structure con- the structure could be ignored.
figuration does not transmit static loads as effectively as a circular
lining. As a result, the high-energy impulse of pulse-like ground 2.3. Sensors and data acquisition system
motion poses a great threat to the structural members of a framed
structure, doing damage to the undetected vulnerable spots and To study the dynamic response of the model structure and the
even to the whole structure. Additionally, the impulse may dynamic soil–structure interaction, accelerometers, strain gauges,
increase the shear deformation of soil notably and thus enlarge the displacement meters, laser displacement meters and soil pressure
story drift of the station and cause further damage. gauges were used. The strain gauge was an FLA-3-11 produced by
In this paper, shaking table tests of a multi-story subway sta- the Japanese company TML. The gauge backing was made of epoxy
tion under pulse-like ground motions are conducted. On the basis resin with thickness of 0.03 mm, and the length, width, backing
of the elastic response of a subway station under different ground length and backing width of the gauge were 0.3, 1.4, 3 and 2 mm,
motions, the effects of pulse-like ground motion on the internal respectively. The laser displacement meter was a CP08MHT80
force and deformation of structural members are discussed. The produced by the German company Wenglor and had dimensions
dynamic earth pressure and deformation pattern of the side wall of 50 mm
50 mm
20 mm, a measuring range of 50 mm, reso-
are investigated. Moreover, the seismic performance of a deep lution finer than 8 μm, and response time less than 660 μs. The
subway station under different levels of ground motion are soil pressure gauge had an outside diameter of 30 mm, capacity of
evaluated. 200 kPa, and precision of 0.5% of full scale. The data acquisition
system with 128 channels was produced by MTS Company, and
the sampling rate used in the test was 512 Hz.
2. Experimental setup

2.1. Shaking table 3. Test design

The shaking table test was carried out using the MTS Company 3.1. Scale factor design
shaking table facility at the State Key Laboratory for Disaster
Reduction in Civil Engineering, Tongji University. The table can be The prototype design of the model structure is a modern sub-
input with three-dimensional and six-degree-of-freedom motions. way station with height of 28.3 m. The station was designed ori-
The dimensions of the table are 4 m
4 m. The working frequency ginally to be a six-story island platform station, and then because
ranges from 0.1 to 50 Hz. The shaking table vibrates with two of the need for parking, the first to third floors underground were
maximum horizontal direction accelerations of 1.2 g and 0.8 g, and merged into one layer to function as a stereo garage. The second
a maximum acceleration of 0.7 g vertically. floor is the lobby floor, the third is a floor that houses equipment,
and the fourth is an island platform. The total length of the station
2.2. Model soil container is 155 m, and the width varies from 23.6 to 28.35 m. The prototype
structure was made of reinforcement concrete. Concrete of Grade
To minimize the box effect, a flexible container was used in the C45 was used for central columns and C35 for the rest parts of the
test. The cylindrical soil container was 3000 mm in diameter, see station [25]. Steel rebar of HRB400 was used in central columns
Fig. 1. Its lateral rubber membrane was 5 mm thick, and reinfor- and HRB335 for the other parts [25].
cement bars having a diameter of 4 mm and spacing of 60 mm The scale factors of the model structure are listed in Table 1.
were used to strengthen the outside of the box. The membrane According to similarity theory, three aspects of the simulation of
was fixed with an upper ring plate and a base plate by bolts. A the soil–structure interaction should be considered primarily:
height-adjustable screw rod was installed to adjust the cylinder to geometric similarity, physical similarity and mechanical similarity.
a proper state. A universal joint was set on the top of the columns, On account of the differences in dimensions between a modern
subway station and typical one, the scale factor design should be
based on the size and bearing capacity of the shaking table, size of
the soil container, boundary effect, and convenience of model
manufacturing. The length scale factor is set to 0.02. Fig. 2 presents
the dimensions of the model structure. Then scale factors of dis-
placement and area can be determined.
In the shaking table test, organic glass was chosen as the
material of the model structure owing to its good homogeneity,
high strength and low elastic modulus, providing flexibility to the
design of the scale factor. This material is also suited to accurate
manufacturing. Thus elastic modulus and density scale factors can
be determined according to material tests of the organic glass.
After the scale factor of geometry, elastic modulus and density
Fig. 1. Schematic diagram of the soil container. are decided, scale factors among the physical quantities can be
 Z. Chen et al. / Soil Dynamics and Earthquake Engineering 82 (2016) 111–122 113

Table 1
24
Scale factors of the model structure.

16

13
Type Physical quantity Scale factor

Geometry properties Length 0.020

Linear displacement 0.020 side wall central stereo

192
Angular displacement 1.000 column garage
Area 4.00
10  4
Material properties Elastic modulus 0.106
Equivalent density 1.765

8
Stress 0.106

13
Strain 1.000
lobby

92
floor

568
Poisson's ratio 1.000 2
Loading Force 4.24
10  5

8
Linear load 2.12
10  3

13
device

92
Area load 0.106 floor
Moment 8.48
10  7 4
1.41
10  5

8
Dynamic properties Mass
Stiffness 2.12
10  3 island

17
platform

125
Duration 8.16
10  2
Frequency 12.253
16 18 floor
8

24
Velocity 0.245

30
Acceleration 3.003

146 114 146

deduced using the Buckingham π law [26]: 470
8
>S ¼ SE
> σ
>
>
> 1 1
<St ¼ Sl Sρ 2 =SE 2
1 1 ð1Þ
>Sv ¼ SE 2 =Sρ 2

150
>
>
>
>
:Sa ¼ SE =Sl Sρ central longitudinal
column beam
where Sσ ; SE ; St ; Sl ; Sv ; Sa denote the stress scale factor, elastic
modulus scale factor, time scale factor, geometric scale factor,
velocity scale factor, and acceleration scale factor, respectively.

160 ¡Á 5
According to a numerical simulation of the shaking table test

1100
on a subway station [27], the error introduced when using the
plane strain hypothesis can be ignored when the distance between
the observation plane and the end of the model is equal to the
structure width. The widths of the model structure and the col- A A
umn separation are 0.47 and 0.16 m, respectively, and the long-
itudinal length of the model structure is thus determined to be
1.1 m. A diaphragm wall mainly behaves as a flexural member, as

150
does the central column according to the results of a previous
study [28]. In addition, the model structure was designed on the
basis of the similarity principle of bending stiffness to introduce 169 96 169
the action of steel bars [28]: 18 470 18
3
Ep I p =lp
Sk ¼ Sσ Sl ¼ 3
ð2Þ Fig. 2. Dimensions of the model structure: (a) cross section A–A; (b) top view
Em I m =lm (unit: mm).

where Sk ; Sσ ; Sl are the stiffness scale factor, stress modulus scale

Table 2
factor, and geometric scale factor, respectively, Ep ; I p ; lp are the Scale factors of the model soil.
prototype's elastic modulus, moment of inertia, and geometry,
respectively, and Em ; I m ; lm are the model's elastic modulus, Physical quantity Symbols Scale factor

moment of inertia, and geometry, respectively.

Shear modulus SG 0.020
The synthetic model soil was a mixture of sand and sawdust. Length SL 0.020
According to trial tests, adding sawdust to sand can reduce both Density Sρ 0.333
the density and dynamic shear modulus, which complies with Acceleration Sa 3.003

similitude requirements. Employing the Buckingham π law, the

scale factors of geometry, density, shear modulus, and inertial 3.2. Material and designation of the model
acceleration were selected as essential parameters and adjusted to
satisfy [26]. 3.2.1. Model structure
SG =ðSl Sρ Þ ¼ Sa ; ð3Þ To measure the elastic modulus of the organic glass, a material
test was carried out as shown in Fig. 3. The elastic moduli of three
where SG ; Sl ; Sρ ; Sa denote the shear modulus ratio, geometry ratio, specimens were 3.60, 3.21, and 3.19 GPa, respectively. The aver-
density ratio, and inertial acceleration ratio, respectively. The scale age value was 3.33 MPa, and the corresponding scale factor of the
factors of soil are presented in Table 2. elastic modulus was 0.106. To match the performances of the
114 Z. Chen et al. / Soil Dynamics and Earthquake Engineering 82 (2016) 111–122

1.0 0.2
Prototype soil
Test specimen Test machine Model soil
0.8

Gd /Gdmax
0.1
0.6

0.4
0.0
1E-4 1E-3 0.01
d

Fig. 5. Dynamic properties of the model soil.

acceleration scale factor is 3.00 and the frequency ratio is 12.253,

the density scale factor should be 1.765. In this case, the original
model structure mass was 88 kg but should be 324 kg according
to the scale factor, and 236 kg of mass thus needed to be added.
Owing to the arrangement of sensors, the actual mass added was
Fig. 3. Material test of organic glass.
208 kg. Fig. 4 shows the layout of added mass.

3.2.2. Model soil

Model structure The prototype soil was typical soil deposits found in Shanghai.
To satisfy the scale factor equation and taking the maximum
dynamic shear modulus, Gd =Gd max  γ d curve, and λ  γ d curve into
consideration, the most appropriate mass ratio of sawdust to sand
was 1:2.5, where Gd ; Gd max ; λ; γ d denote the dynamic shear mod-
ulus, maximum dynamic shear modulus, damping ratio, and
dynamic shear strain respectively. The density of the mixture was
700 kg/m3, the density scale factor was 0.39, the confining pres-
sure ratio was 0.02, and the modulus obtained in the test was
1.81 MPa (target modulus: 1.72 MPa, error: 5.2%). The Gd =Gd max 
γ d curve and λ  γ d curve obtained in a dynamic tri-axial test are
presented in Fig. 5.
During the experiment, the soil was placed into the soil con-
tainer layer by layer. Each layer was compacted to have thickness
of approximately 20 cm. Before the main shaking table test, a trial
shaking table test was conducted to analyze the degree of com-
paction and settlement, and to ensure the soil density in the main
test remained accurate and consistent.

Added mass 3.3. Layouts of sensors

Fig. 6 presents the soil container and observation plane. The

layouts of 20 strain gauges, 18 accelerometers, 10 displacement
meters, eight laser displacement meters, and 10 soil pressure
gauges in the model soil and model structure are shown in
Figs. 7 and 8. Strain gauges were used to investigate the dynamic
response of the structure. Accelerometers A3, A6, A9, and A12
were set to investigate the propagation law of seismic waves in
soil, accelerometers A2, A5, A8, and A11 to investigate the effect of
structure on the propagation law, accelerometers A10, A11, and
A12 to validate the boundary effect, and accelerometers A15–A18
to record the dynamic acceleration response of each story. Dis-
placement meters were placed to measure the settlement of soil
and lateral displacement of the soil container. Laser displacement
meters were fixed on the upper surface of each floor and circular
Fig. 4. (a) Model structure; and (b) layout of added mass. tubes made of organic glass were fixed on the lower surface of
each floor. Hence, the laser displacement meters provided data on
the relative displacement of the central column and side wall. Soil
shaking table and to keep the acceleration scale factor and fre- pressure gauges P1–P8 were arranged to explore the distribution
quency scale factor within reasonable ranges, mass was added to of the dynamic earth pressure and P9 and P10 were used to check
meet the requirement of the density scale factor. When the the dynamic earth pressure.
 Z. Chen et al. / Soil Dynamics and Earthquake Engineering 82 (2016) 111–122 115

soil box

model structure
central column
observation plane

direction of vibration

Fig. 6. (a) Soil container; and (b) location of the observation plane.

500 500 500 500 500 500 Table 3

D6 D1 D2 D3 D4 D5 A4 Input ground motions.
A1 A2 A3
400

400

D7 A5 A6 Information Ground motion

A: Accelerometer
400

400

D8 Hector Mine ChiChi El Centro

A8 A9
A7 D: Displacement meter
400
400

PGA (g) 0.1 0.1 0.1

D9 A10 A11 A12
500 500 PGV (m/s) 0.766 0.496 0.115
400

PGV/PGA (s) 0.766 0.496 0.115

D10 A13 A14
Record sequence number 1803 1542 172

Fig. 7. Layouts of accelerometers embedded in soil and displacement meters

(unit: mm). 0.1 g and 0.2 g, the amplification factors were greater closer to the
ground surface. Additionally, amplification factors at any depth
3.4. Loading method and test cases under pulse-like ground motion were slightly larger than those
under ordinary ground motion. However, in the case of 0.6 g, the
To investigate the effect of pulse-like ground motion on the amplification factors were obviously lower. Huang et al. [33], in
subway station, three records were selected from the Pacific their study of the dynamic response of soil deposits in Shanghai,
Earthquake Engineering Research Center database [29] and the found that soil had damping characteristics when the peak
information is shown in Table 3. Specifically, El Centro ground acceleration of the input wave was larger than 0.3 g. In this paper,
motion is ordinary ground motion while Hector Mine and ChiChi methods proposed by Zeghal et al. [34] and Brennan et al. [35]
ground motions are pulse-like ground motions. Definitions of were used to calculate the shear modulus and damping ratios of
pulse-like ground motion are consistent with those given by the the soil. It is found that with the increase of acceleration, shear
papers [30,31]. To define the pulse-like characteristics of these modulus decreases significantly, resulting in the natural frequency
motions more directly, the ratio of peak ground velocity to peak getting far away from the domain frequency of the motion. This
ground acceleration (PGV/PGA) was selected as the index. might cause a reduction in amplification factor of soil acceleration.
According to studies[11,32], pulse-like ground motion is defined as The peak acceleration of the model structure is shown in Fig. 12.
ground motion whose PGV/PGA is greater than 0.2 while ordinary Generally, the model structure embedded in soil underwent the
ground motion has a ratio smaller than 0.15. The pulse-like effect magnified acceleration effect. The accelerations of different stories
of Hector Mine ground motion is greater than that of ChiChi were magnified to various degrees. The peak accelerations recorded
ground motion. Fig. 9 presents the acceleration time histories and by A16 and A17 were much higher than those recorded by other
Fourier spectra of the three records. accelerometers. The effect of different types of ground motion on the
For the purpose of investigating the dynamic earth pressure peak accelerations of the structure is not clear.
under different intensities and types of ground motions, the three
ground motions were scaled to four levels (0.1, 0.2, 0.6, and 1.0 g). 4.2. Lateral displacement and settlement of soil
White noise was used to check the changes in the model. Table 4
gives the test cases. Fig. 13 shows the maximum lateral displacements of the soil
container under different ground motions with the peak accel-
eration of 0.1 g. To better understand the lateral displacement of
4. Test verification the soil container under ground motions, typical time histories of
displacement meter D6 are presented in Fig. 13(b). Fig. 13(a) shows
4.1. Amplification of the acceleration of soil shear-type deformation of the soil container, which satisfies the
lateral deformations of soils as semi-infinite half space media
According to tests results, peak values of accelerometers A7, A8 subjected to the ground motion of a real earthquake [36]. In
and A9 under El Centro ground motion (peak acceleration was addition, it could be find from Fig. 13(a) and (b) that the effect of
0.1 g) were 0.085, 0.081 and 0.082, respectively (see Fig. 10), which the pulse-like ground motion on the overall soil deformation was
indicated that influences of boundary effect could be ignored. much greater than that of ordinary ground motion.
Fig. 11 presents the peak acceleration amplification factors for Taking the record of the Hector Mine ground motion as an
different depths of soil with different peak accelerations. Because example, Fig. 14(a) shows the instantaneous soil deformation under
there was no signal for A12, it is replaced by the signal for A11 the Hector Mine ground motion when the peak acceleration was
temporarily in Fig. 11. It is seen that, when peak acceleration were 0.1 g. In Fig. 14, T symbol stands for the duration of earthquakes. It is
116 Z. Chen et al. / Soil Dynamics and Earthquake Engineering 82 (2016) 111–122

A: Accelerometer S: Strain gauge P: Soil pressure gauge LD: Laser displacement meter

P1
S1 S2 S3 S4

76
P2

77
S8
LD1 LD2 P3
S5 S6 S7 (A15)

77
S9 S10 S11 S12
P4
A16 LD3 LD4

77
S13 S14 S15 S16 P9 P5

77
A17 LD5 LD6
P6

231
S18 S19 S20

77
S17
P7
A18 LD7 LD8

77
P10 P8
Fig. 8. Layouts of sensors on the cross section of the model structure: (a) accelerometers and strain gauges; and (b) soil pressure meters and laser displacement meters.

10
1.0 Hector Mine Hector Mine
8
Fourier amplitude(g)
Acceleration(g)

0.5
6
0.0
4
-0.5
2
-1.0
0
0 30 60 90 120 0 5 10 15 20 25 30
Time(s) Frequency(Hz)

4
1.0 ChiChi ChiChi
Fourier amplitude(g)

3
Acceleration(g)

0.5

0.0 2

-0.5
1

-1.0
0
0 20 40 60 80 0 5 10 15 20 25 30
Time(s) Frequency(Hz)

0.6
1.0 El Centro El Centro
Fourier amplitude(g)
Acceleration(g)

0.5
0.4

0.0

0.2
-0.5

-1.0
0.0
0 10 20 30 40 0 5 10 15 20 25 30
Time(s) Frequency(Hz)
Fig. 9. Acceleration–time histories and Fourier spectra of (a) Hector mine, (b) ChiChi, and (c) El Centro ground motions.
 Z. Chen et al. / Soil Dynamics and Earthquake Engineering 82 (2016) 111–122 117

seen that the container had a shear-type deformation mode In the test, the obvious settlement of soil implies a significant
throughout the duration of the earthquake. Fig. 14(b) shows the change in the soil density, which may affect the test results. Fig. 15
maximum soil deformation under the Hector Mine ground motion shows the maximum settlements of soil for the three records with
with different peak accelerations. It can be concluded that the
deformations were large close to the ground, especially when the
A1
0.6
earthquake was strong. Moreover, owing to the large story height of

Height of the structure (m)

El Centro
the first floor underground, this kind of deformation can do more ChiChi
harm to the subway station. In the actual project, the density of soil Hector Mine
0.4
near the surface is lower than that at greater depth, and the defor- A15

mation of soil near the surface might thus be even larger.

A16

0.2
Table 4 A17
Test program.

Test sequence Test no. Ground motion Peak acceleration (g) A18
0.0
1 WN-1 White Noise 0.07
1.0 1.5 2.0 2.5
2 El-x0.1 El Centro 0.10
3 Chi-x0.1 ChiChi 0.10 Amplification
4 HM-x0.1 Hector Mine 0.10
Fig. 12. Peak acceleration amplification factors of the model structure.
5 WN-2 White Noise 0.07
6 El-x0.2 El Centro 0.20
7 Chi-x0.2 ChiChi 0.20
8 HM-x0.2 Hector Mine 0.20

Height of the soil container(m)

9 WN-3 White Noise 0.07
10 El-x0.6 g El Centro 0.60 1.6 El Centro D6
11 Chi-x0.6 g Hector Mine 0.60 ChiChi
12 HM-x0.6 ChiChi 0.60 1.2 Hector D7
13 WN-4 White Noise 0.07 Mine
14 El-x1.0 g El Centro 1.00 0.8 D8
15 Chi-x1.0 g ChiChi 1.00
16 HM-x1.0 Hector Mine 1.00 0.4
D9
0.0
0 2 4 6 8 D10
0.10 Lateral displacement(mm)

8
0.05 4 El Centro 0.1g
Acceleration (g)

0
Displacement (mm)

-4
-8
0.00 8
4 ChiChi 0.1g
0
-4
A7 -8
-0.05 8
A8 Hector Mine 0.1g
4
A9 0
-4
-0.10 -8
0 1 2 3 4 0 5 10 15 20
Time (s) Time (s)
Fig. 10. Time histories of accelerometers A7, A8 and A9 under El Centro ground Fig. 13. Lateral displacements of the model soil container under different ground
motion with peak acceleration of 0.1 g. motions and (b) time history of lateral displacement of D6.

Amplification
0.5 1.0 1.5 2.0 2.5
A3 0.0

A6
-0.4
Soil depth(m)

A9
-0.8
A11 A12 0.1g El Centro
(No signal) -1.2 0.2g El Centro
0.6g El Centro
1.0g El Centro
-1.6

Fig. 11. Peak acceleration amplification factors of the model soil.

118 Z. Chen et al. / Soil Dynamics and Earthquake Engineering 82 (2016) 111–122

ground motions. Hector Mine ground motion carries much larger

1.6 0T 2/8 T energy than El Centro, which will be explained in detail later.
1/8 T
1.2
0.8
Height of soil container(m)

0.4 5. Test result and analysis

0.0
1.6 5.1. Deformation mode of the model structure
3/8 T 4/8 T 5/8 T
1.2
0.8 The presented results of displacements, accelerations, and soil
0.4 pressures mean dynamic results. Internal forces presented include
0.0 both static and dynamic forces. To comply with the real project, all
1.6
6/8 T 7/8 T T results obtained in the test were converted from model to proto-
1.2
type according to the scale factors. As shown in Fig. 16, the model
0.8
structure had a racking deformation mode. In Fig. 16, T symbol
0.4
stands for the duration of earthquakes. Fig. 16(a) presents the
0.0
-3 -2 -1 0 1 2 -3
3 -2 -1 0 1 2 -3
3 -2 -1 0 1 2 3
instantaneous deformation of the side wall in steps of one-fourth
of the period of the El Centro record when the peak acceleration
Lateral displacement of model box(mm)
was 0.033 g. Fig. 16(b) shows the maximum deformation of the
side wall throughout the entire period under different records
with different peak accelerations.
Height of the soil container(m)

Both the instantaneous and maximum deformations of the

1.6 side wall turned out to have a nice racking form, especially
when the peak acceleration was small. When subjected to shear
1.2
distortions during an earthquake, a rectangular box structure
will undergo transverse racking deformation [37]. Although the
0.8 0.1g multi-story subway station has relatively large height, it still has
0.2g a nice racking deformation mode. This is mainly because of the
0.4 0.6g large stiffness imparted by its thick side wall. Chen et al. [38]
1.0g came to the same conclusion on the basis of numerical results.
0.0
The deformations increased with the peak acceleration. By
0 30 60 90 120
comparing the deformations under the El Centro ground motion
Lateral displacement(mm)
with those under the ChiChi and Hector Mine ground motions, it
Fig. 14. Lateral displacements of the model soil container under the Hector Mine is found that the deformations of the side wall under pulse-like
ground motion (a) in steps of one-eighth of the duration of the ground motion; ground motions were obviously larger than those under ordin-
(b) with different peak accelerations.
ary ground motions. Comparison between the deformations for
the ChiChi record with those for the Hector Mine record shows
D1 D2 D3 D4 D5 that the pulse-like effect may remarkably increase the seismic
response of the understructure.

5.2. Dynamic response characteristics of the internal force of the

0 central column

Fig. 17 shows the relationship between the maximum moment

Settlement(mm)

-2
of the top-story central column and the peak acceleration of input
-4 motions.
The maximum moment of the central column on the top story
increased with the peak acceleration. The maximum moments
-6
under pulse-like ground motions were obviously larger than those
El Centro 1.0g under ordinary ground motions. The maximum moment increased
-8 ChiChi 1.0g with the magnitude of the pulse-like effect. In addition, the pulse-
Hector Mine 1.0g like effect is remarkable especially in a destructive earthquake. In
-10
0 1 2 3 Fig. 17, when the peak acceleration was 0.033 g, the maximum
Distance from the edge (m) moment of the first-story central column induced by the Hector
Mine (ChiChi) ground motion was 26.44% (20.20%) larger than that
Fig. 15. Settlement of the model soil under three ground motions when peak
induced by the El Centro ground motion. When the peak accel-
accelerations are 1.0 g.
eration was 0.333 g, the difference reached 80.64% (29.43%).
The above results are mainly explained as follows. First, although
the maximum peak acceleration. It is concluded that the changes
seismic waves have the same peak acceleration, pulse-like ground
in density can be ignored. Under the El Centro and ChiChi ground
motion carries more energy than the ordinary motion. Various
motions, the maximum settlement was less than 1.0 mm. In the
parameters are used to measure earthquake energy. The Arias
case of the Hector Mine record, the maximum settlement was intensity [39] is an important measure of the strength of ground
7.8 mm, which is small compared with the height of the soil, 1.6 m. motion, as it simultaneously reflects multiple characteristics of the
In particular, the effect of strong pulse-like ground motion (in the motion [40]. The Arias intensity describes the cumulative energy
Hector Mine record) on the settlement of soil was much stronger per unit weight absorbed by an infinite set of single-degree-of-
than the effects of ground motion in the El Centro and ChiChi freedom oscillators [41]. Moustafa and Takewaki [42] used the Arias
records. This difference in settlement was caused by the energy of intensity to describe the characteristics of pulse-like motions.
 Z. Chen et al. / Soil Dynamics and Earthquake Engineering 82 (2016) 111–122 119

Height of the structure (m)

Height of the structure (m)
25 0.25T 25
0.50T
20 20
0.75T
15 15
10 0.033g El Centro
10 0.067g El Centro
0.200g El Centro
5 0.333g El Centro
5
0.333g ChiChi
0 0.333g Hector Mine
0
-7.5 -5.0 -2.5 0.0 -25 0 25 50 75 100 125 150
Lateral displacement (mm) Lateral displacement (mm)
Fig. 16. Lateral displacements of the side wall (a) in steps of one-fourth of the ground motion duration; and (b) under three ground motions.

2500 affected by the flexibility of the structure will also change the
The maximum moment(kN×m)

El Centro elastic internal forces of central columns [43]. In additional, Piti-

2000 ChiChi lakis and Tsinidis [44] pointed out that the displacement-based
Hector Mine methods were closer to the physics of the problem and presented
1500 several advantages including the proper evaluation of the inelastic
response of the structural components. Consequently, using
1000 deformations to describe the columns performance will be more
rational, and will also show the kinematic nature.
In addition, with respect to the effects of motions on the axial
500
force, there are few differences between pulse-like motions and
ordinary motions. As shown in Fig. 18, the maximum axial forces of
0
0.00 0.07 0.13 0.20 0.27 0.33 the top-story central column under El Centro, ChiChi, and Hector
Peak acceleration(g) Mine ground motions were 4330, 4160, and 4390 kN, respectively.
Moreover, the conclusion is transferable to bottom-story columns.
Fig. 17. Maximum moment of the top-story central column. It is found that under two extreme conditions (0.033-g El Centro
ground motion and 0.333-g Hector Mine ground motion), there
Calculations show that when the peak acceleration is 0.1 g, the Arias was only a 401-kN (1.2%) difference in the axial compression ratio.
intensity of the El Centro, ChiChi, and Hector Mine ground motions In other words, the horizontal inputted pulse-like motion did not
are 0.146, 0.773, and 2.113, respectively. There is clearly a great induce vertical vibration of the central column.
difference in energy among the three motions. Second, the pulse-
like ground motions are rich in low-frequency components if 5.3. Dynamic response characteristics of the displacement of the
compared with the ordinary ground motion, as seen for the Fourier central column
spectrum of the three ground motions in Fig. 9. The soil deposits
filter out a significant portion of the high-frequency content and Column drifts of each story under the ChiChi ground motion
can amplify low-frequency signals. As a result, when seismic waves with different peak accelerations are shown in Fig. 19. It can be
propagate from bedrock to the underground structure, the energy concluded from the gradient in the figure that (1) the drift
attenuation of pulse-like motion is less than that of ordinary increased with the peak acceleration and (2) except in the 0.066-g
motion. Therefore, the input energy of the underground structure is case, the column drifts of the top story and the bottom story were
dramatically different. A large proportion of the energy of ordinary larger than those of the middle stories. For instance, when the
motion is dissipated by soil. peak acceleration was 0.333 g, the column drift for the top story
To evaluate the safety of the subway station under different was 37.68% higher than that for the second story. The lateral
intensities and types of ground motions, the top-story central stiffness values of columns on the top story and second story were
column was selected as an observed object for two reasons. First, 4.71
106 and 6.07
107 N/m, respectively. An increase in height
the large story height of the top story means that the lateral clearly reduced the lateral stiffness, therefore enlarging the
stiffness is much lower for the top story than for the other stories. column drift.
Second, although the axial force of the bottom-story column is the As stated in the previous section, from the perspective of
largest, the size of the top-story column is relatively small and internal forces, central columns seemed safe under the ground
thus the bearing capacity is relatively low. Axial force–moment motions of the considered earthquakes. But it did not take
time histories with peak acceleration of 0.333 g converted from account of stress concentration effect and so on. From the view of
model to prototype and axial force–moment bearing capacity performance-based design, deformation is a more suitable index
curves calculated for the prototype are presented in Fig. 18. With than force with which to evaluate the seismic performance of
the same peak acceleration, the Hector Mine motion, followed by underground structures. In investigating the relationship
ChiChi motion and then El Centro motion, had the greatest effect between the pulse-like effect and column drift, the top-story
on the moment. The prototype structure is made of reinforced central column was selected for its large story height. Fig. 20
concrete (an elasto-plastic material), while the model structure is shows the relation between the column drift and peak accel-
made of organic glass (an elastic material). However, from the eration. The variation in column drift was basically the same as
point of dynamic internal forces, the stress concentration of the that of the moment: (1) the column drift on the top story
prototype structure may affect distributions of internal forces thus increased with the peak acceleration; (2) the column drift under
leading to higher forces in some sections compared to the elastic pulse-like ground motion was obviously larger than that under
responses. Moreover, nonlinearities of soils and interface which ordinary ground motion and increased with an increase in the
120 Z. Chen et al. / Soil Dynamics and Earthquake Engineering 82 (2016) 111–122

3 3 3
Axial force(×104kN) Bearing capacity Bearing capacity Bearing capacity

Axial force(×104kN)

Axial force(×104kN)
El Centro ChiChi Hector Mine
2 2 2

1 1 1

0 0 0
0 1000 2000 3000 0 1000 2000 3000 0 1000 2000 3000
Moment(kN*m) Moment(kN*m) Moment(kN*m)
Fig. 18. Bearing capacity curves and axial force–moment time histories of the top-story central column with peak acceleration of 0.333 g.

0.012
25
The height of structure(m)

El Centro
0.010 ChiChi
20
0.008 Hector Mine

Column drift
15
0.006
10 0.033g 0.004
0.066g
5 0.002
0.200g
0 0.333g
0.000
0 30 60 90 120 150 0.000 0.067 0.133 0.200 0.267 0.333
Displacement of the columns(mm) Peak Acceleration(g)
Fig. 19. Lateral displacements of columns in each story under ChiChi ground Fig. 20. Column drifts of the top-story central column.
motion.

magnitude of the pulse-like effect; and (3) such increment was Table 5
Column drifts of the top-story central column.
greater given the peak acceleration was higher; i.e., the earth-
quake imposed was stronger. Peak acceleration of Peak acceleration of Column drift of the top-story
As stated in the previous section, from the perspective of model structure prototype structure central column
internal forces, central columns were safe under the ground El Centro ChiChi Hector
Mine
motions of the considered earthquakes. This does not entirely
agree with the perspective of deformation. Table 5 presents the 0.1 g 0.033 g 1/2278 1/677 1/677
top-story column drift under the three motions with different 0.2 g 0.066 g 1/797 1/482 1/404
peak accelerations. When the peak acceleration of prototype was 0.6 g 0.200 g 1/268 1/193 1/138
1.0 g 0.333 g 1/206 1/153 1/103
0.2 g (major earthquake), column drifts were already beyond 1/
200. Furthermore, when the peak acceleration was 0.333 g, the
column drift reached 1/103 under Hector Mine ground motion. It
is important to note that, because the model structure is elastic,
the column drifts measured are smaller than those in a real 30
Height of the structure(m)

P1
situation. Hence, the column drift rather than the internal force 25
P2
deserves more attention and improvement.
P3 20
5.4. Dynamic earth pressure distribution P4
15
P5
Fig. 21 shows the maximum dynamic earth pressure distribu- 10
P6
tion on the model structure. The dynamic earth pressure was El Centro 0.2g
calculated by subtracting the static earth pressure from the total P7
5 ChiChi 0.2g
earth pressure. It is seen that the dynamic earth pressure under Hector Mine 0.2g
P8 0
pulse-like motion was higher than that under ordinary motion. 0 1 2 3 4 5 6 7 8
However, the maximum dynamic earth pressure distribution was Peak dynamic earth pressure(kPa)
not clear and it is different from the dynamic earth pressure acting
on a single-double story or double-story underground structure Fig. 21. Distribution of the peak dynamic soil pressure on the side wall.

[27,45,46]. Additionally, the distribution mode with an “S” shape is

inconsistent with the pressure induced by racking deformation.
Nonlinear phenomena of soils and the soil–structure interface stiffness of the sensing plate may affect the readings of earth
during shaking, especially for higher intensities, may also affect pressure cells. And the responses of pressure gauges can be quite
the dynamic earth pressure. Additionally, these recorded earth tricky in cases of dry cohesionless soils due to grain size effects
pressures may be biased to some extent by the response of the and so on.
earth pressure cells. Tsinidis et al. [47] also stated that the relative
 Z. Chen et al. / Soil Dynamics and Earthquake Engineering 82 (2016) 111–122 121

6. Conclusions [13] Wen WP, Zhai CH, Li S, Chang ZW, Xie LL. Constant damage inelastic dis-
placement ratios for the near-fault pulse-like ground motions. Eng Struct
2014;59:599–607. http://dx.doi.org/10.1016/j.engstruct.2013.11.011.
Shaking table tests were conducted to study the effect of pulse- [14] Bertero VV, Mahin SA, Herrera RA. Aseismic design implications of near-fault
like ground motions on a multi-story modern subway station. The San Fernando earthquake records. Earthq Eng Struct Dyn 1978;6(1):31–42.
following conclusions are drawn from the results of the study. [15] Anderson JC, Bertero VV. Uncertainties in establishing design earthquakes.
J Struct Eng 1987;113(8):1709–24.
(1) Because of the strength of low-frequency components and [16] Makris N, Black CJ. Evaluation of peak ground velocity as a “good” intensity
high energy, pulse-like ground motion has a dramatically measure for near-source ground motions. J Eng Mech 2004;130(9):1032–44.
greater dynamic effect than ordinary motion on underground http://dx.doi.org/10.1061/(ASCE)0733-9399(2004)130:9(1032).
[17] Sehhati R, Rodriguez-Marek A, ElGawady M, Cofer WF. Effects of near-fault
structures and surrounding soils, in terms of the internal ground motions and equivalent pulses on multi-story structures. Eng Struct
forces and drift of the central column and deformation of 2011;33(3):767–79. http://dx.doi.org/10.1016/j.engstruct.2010.11.032.
the side wall and soil. [18] Makris N, Chang SP. Effect of viscous, viscoplastic and friction damping on the
response of seismic isolated structures. Earthq Eng Struct Dyn 2000;29(1):85–
(2) The multi-story subway station operates in a racking 107.
deformation mode. [19] Jangid R, Kelly JM. Base isolation for near-fault motions. Earthq Eng Struct Dyn
2001;30(5):691–707. http://dx.doi.org/10.1002/eqe.31.
(3) Central columns, especially columns with large height, are the [20] Hall JF, Heaton TH, Halling MW, Wald DJ. Near-source ground motion and its
points vulnerable to ground motion. Reducing the lateral effects on flexible buildings. Earthq Spectra 1995;11(4):569–605.
stiffness generated by large story height increases the column [21] Ismail M, Casas JR. Novel isolation device for protection of cable-stayed
bridges against near-fault earthquakes. J Bridge Eng 2014;19(8). http://dx.
drift notably. The column may suffer larger dynamic responses
doi.org/10.1061/(ASCE)BE.1943-5592.0000509.
under pulse-like ground motion than under ordinary ground [22] Phan V, Saiidi MS, Anderson J, Ghasemi H. Near-fault ground motion effects on
motion. When the peak acceleration of input motion is 1.0 g, reinforced concrete bridge columns. J Struct Eng 2007;133(7):982–9. http:
the top-story maximum moment and drift under strong pulse- //dx.doi.org/10.1061/(ASCE)0733-9445(2007)133:7(982).
[23] Chen ZY, Wei JS. Correlation between ground motion parameters and lining
like ground motion are 81% and 100% higher than those under damage indices for mountain tunnels. Nat Hazards 2013;65(3):1683–702.
ordinary ground motion, respectively. http://dx.doi.org/10.1007/s11069-012-0437-5.
[24] Lu XL, Li PZ, Chen YQ, Chen B. Chen Shaking table model testing on dynamic
soil–structure interaction system. In: Proceedings of the 13th world con-
ference on earthquake engineering; 2004. p. 1–16.
Acknowledgments [25] GB50010. Code for design of concrete structure. Beijing: China Architecture &
Building Press; 2010.
[26] Moncarz PD, Krawinkler H. Theory and application of experimental model
This research was supported by the State Key Laboratory of
analysis in earthquake engineering. Report No. 50. Stanford, CA: Dept. of Civil
Disaster Reduction in Civil Engineering (SLDRCE14-B-11), National Engineering and Environmental Engineering, Stanford University; 1981.
Natural Science Foundation of China (Grant no. 51278524 and [27] Yang LD, Yang C, Ji QQ, Zheng YL. Shaking table test and numerical calculation
on subway station structures in soft soil. J Tongji Univ 2003;31(10):1135–40.
Grant no. 41472246), and Innovation Program of Shanghai Muni- [28] Lin G, Zhu T, Lin B. Similarity technique for dynamic structural model test.
cipal Education Commission (14ZZ034). The experimental pro- J Dalian Univ Technol 2000;40(1):1–8.
gram carried out in this research could not have been executed [29] Pacific Earthquake Engineering Research Center (PEER). PEER strong motion
database. Berkeley: University of California; 2000.
without the able assistance of Lu Wensheng and Zhao Bin of the [30] Baker JW. Quantitative classification of near-fault ground motions using
State Key Laboratory. All support is gratefully acknowledged. wavelet analysis. Bull Seism Soc Am 2007;97(5):1486–501. http://dx.doi.org/
10.1785/0120060255.
[31] Shahi SK, Baker JW. An empirically calibrated framework for including the
effects of near-fault directivity in probabilistic seismic hazard analysis. Bull
Seism Soc Am 2011;101(2):742–55. http://dx.doi.org/10.1785/0120100090.
References [32] Bray JD, Rodriguez-Marek A. Characterization of forward-directivity ground
motions in the near-fault region. Soil Dyn Earthq Eng 2004;24(11):815–28.
http://dx.doi.org/10.1016/j.soildyn.2004.05.001.
[1] Wang W, Zhao GG. 71 m! "The Iron Giant" sets new record on "the Shanghai
[33] Huang Y, Ye WM, Chen ZC. Seismic response analysis of the deep saturated
depth". Wen Hui Bao Newspaper; 2013. 09:1. [In Chinese].
soil deposits in Shanghai. Environ Geol 2009;56(6):1163–9. http://dx.doi.org/
[2] Tamari Y, Towhata I. Seismic soil–structure interaction of cross sections of
10.1007/s00254-008-1216-1.
flexible underground structures subjected to soil liquefaction. Soils Found
[34] Zeghal M, Elgamal AW, Tang HT, Stepp JC. Lotung downhole array. II: eva-
2003;43(2):69–87.
luation of soil nonlinear properties. J Geotech Eng 1995;121(4):363–78.
[3] Kutter BL, Chou JC, Travasarou T. Centrifuge testing of the seismic performance
[35] Brennan AJ, Thusyanthan NI, Madabhushi SPG. Evaluation of shear modulus
of a submerged cut-and-cover tunnel in liquefiable soil. In: Proceedings of the
and damping in dynamic centrifuge tests. J Geotech Geoenviron Eng ASCE,
geotechnical earthquake engineering and soil dynamics IV congress (GSP);
2008. 181. 131; 2005. p. 1488–97.
[4] Chen GX, Wang ZH, Zuo X, Du XL, Gao HM. Shaking table test on the seismic [36] Meymand P. Shaking table scale model test of nonlinear soil–pile–super-
failure characteristics of a subway station structure on liquefiable ground. structure interaction in soft clay [PhD dissertation]. Berkeley: University of
Earthq Eng Struct Dyn 2013;42(10):1489–507. http://dx.doi.org/10.1002/ California; 1998.
eqe.2283. [37] Hashash YMA, Hook JJ, Schmidt B, Yao JIC. Seismic design and analysis of
[5] Moss RES, Crosariol VA. Scale model shake table testing of an underground underground structures. Tunn Undergr Space Technol 2001;16(4):247–93.
tunnel cross section in soft clay. Earthq Spectra 2013;29(4):1413–40. http://dx. http://dx.doi.org/10.1016/S0886-7798(01)00051-7.
doi.org/10.1193/070611EQS162M. [38] Chen ZY, Chen W, Zhang W. Zhang Seismic performance evaluation of multi-
[6] Kagawa T, Sato M, Minowa C, Abe A, Tazoh T. Centrifuge simulations of large- story subway structure based on pushover analysis. Advances in Soil Dynamics
scale shaking table tests: case studies. J Geotech Geoenviron Eng 2004;130 and Foundation Engineering (ASCE); 2014. p. 444–54.
(7):663–72. [39] Arias A. Measure of earthquake intensity. Massachusetts Inst. of Tech.,
[7] Kamata H, Mashimo H. Centrifuge model test of tunnel face reinforcement by Cambridge. Univ. of Chile, Santiago de Chile; 1970.
bolting[J]. Tunn Undergr Space Technol 2003;18(2):205–12. [40] Stafford P, Berrill J, Pettinga J. New predictive equations for Arias intensity
[8] Pitilakis D, Dietz M, Wood DM, Clouteau D, Modaressi A. Numerical simulation from crustal earthquakes in New Zealand. J Seism 2009;13(1):31–52. http:
of dynamic soil-structure interaction in shaking table testing. Soil Dyn Earthq //dx.doi.org/10.1007/s10950-008-9114-2.
Eng 2008;28(6):453–67. [41] Travasarou T, Bray JD, Abrahamson NA. Empirical attenuation relationship for
[9] Chen GX, Chen S, Zuo X, Du XL, Qi CZ, Wang ZH. Shaking-table tests and Arias intensity. Earthq Eng Struct Dyn 2003;32(7):1133–55. http://dx.doi.org/
numerical simulations on a subway structure in soft soil. Soil Dyn Earthq Eng 10.1002/eqe.270.
2015;76:13–28. [42] Moustafa A, Takewaki I. Deterministic and probabilistic representation of
[10] Chen J, Shi X, Li J. Shaking table test of utility tunnel under non-uniform near-field pulse-like ground motion. Soil Dyn Earthq Eng 2010;30(5):412–22.
earthquake wave excitation. Soil Dyn Earthq Eng 2010;30(11):1400–16. http://dx.doi.org/10.1016/j.soildyn.2009.12.013.
[11] Loh CH, Wan S, Liao WI. Effects of hysteretic model on seismic demands: [43] Lanzano G, Bilotta E, Russo G, Silvestri F, Madabhushi SPG. Centrifuge mod-
consideration of near-fault ground motions. Struct Des Tall Build 2002;11 eling of seismic loading on tunnels in sand. Geotech Test J 2012;35(6):854–69.
(3):155–69. http://dx.doi.org/10.1002/tal.182. [44] Pitilakis K, Tsinidis G. Performance and seismic design of underground
[12] Somerville P, Graves R. Conditions that give rise to unusually large long period structures. Earthq Geotech Eng Des 2014;28:279–340 Geotechnical Geological
ground motions. Struct Des Tall Build 1993;2(3):211–32. and Earthquake Engineering, Springer.
122 Z. Chen et al. / Soil Dynamics and Earthquake Engineering 82 (2016) 111–122

[45] Zhai E, Davis CA, Yan LP, Hu JP. Numerical simulations of geotechnical cen- [47] Tsinidis G, Pitilakis K, Madabhushi G, Heron C. Dynamic response of flexible
trifuge modeling of seismic earth pressures on an underground restrained square tunnels: centrifuge testing and validation of existing design meth-
structure. International efforts in lifeline earthquake engineering. In: Pro- odologies. Geotechnique 2015;65:401–17.
ceedings of the 6th China–Japan–US trilateral symposium on lifeline earth-
quake engineering; 2014. p. 369–76. DOI: 10.1061/9780784413234.048.
[46] Jiang LZ, Chen J, Li J. Seismic response of underground utility tunnels: shaking
table testing and FEM analysis. Earthq Eng Eng Vib 2010;9(4):555–67.