Soil Dynamics and Earthquake Engineering 138 (2020) 106333

Contents lists available at ScienceDirect

Soil Dynamics and Earthquake Engineering

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

Experimental study and numerical simulation on dynamic soil-structure

interaction under earthquake excitations
Shutong Liu a, Peizhen Li a, Wenyang Zhang b, Zheng Lu a, *
a
State Key Laboratory of Disaster Reduction in Civil Engineering, Tongji University, Shanghai, China
b
Civil & Environmental Engineering Department, University of California, Los Angeles, USA

A R T I C L E I N F O A B S T R A C T

Keywords: Soil-structure interaction (SSI) phenomena subjected to seismic loadings are investigated in this study. Firstly,
Soil-structure interaction shaking table tests on dynamic soil-structure interaction subjected to seismic loadings are carried out, and SSI
Shaking table test effects are analyzed through a comparative study between the soil-structure system and the rigid foundation
Seismic response
condition. Then, a corresponding numerical model on the SSI tests is presented. In the finite element model, a
Parametric study
bounding surface plasticity model and an equivalent linear method are separately used for simulating the
nonlinear behavior of soil. It turns out that the calculated results by the bounding surface plasticity model agree
much better with the experimental results. Finally, the effects of different soil types and structural properties on
soil-structure interaction are discussed by numerical parametric study. The results show that SSI effects
noticeably mitigate the dynamic response of structure, and the closer the frequency of superstructure to the soil
frequency, the more obvious the effect of soil-structure interaction. With increased shear wave velocity of soil,
the impact of soil-structure interaction weakens. It is important for studying the impact of SSI on structural
responses and practical engineering design purposes.

1. Introduction site descriptions for earthquake databases. Ansal et al. [8] estimated the
site amplification from microtremor spectral ratios and proved the
The soil-structure interaction (SSI) is a complicated phenomenon for applicability of microtremor measurements to access the site condition.
the structure with a pile foundation coupled to the soil medium, which is Fatahi et al. [9] carried out comparative analysis for seismic design of
also an essential issue in the field of earthquake engineering. The soil- high buildings on soft soil considering soil-structure interaction or site
structure interaction generally performs two primary effects-kinematic effect, and the results showed only including the local site effect
interaction and inertial interaction [1,2]. Over the past 40 years, excluding SSI isn’t enough to ensure its safety.
concern over soil-structure interaction has received considerable Some techniques are used as a ground-improvement method to
attention. Nevertheless, different methodologies may lead to different improve soil quality for engineering problems. Güllü [10] attempted to
analysis results of the SSI problem, which usually puzzle engineers. It is use the superplasticiser for stabilisation of a fine-grained soil, and the
still a controversial issue whether the seismic soil-structure interaction test results showed that the superplasticisers improved engineering
plays a beneficial or detrimental role [3,4]. characteristics for stabilisation of clay for pre- and post-freezing and
Site effect is a significant topic in the field of SSI. Different from soil- thawing conditions. Besides, Güllü et al. [11] investigated the rheolog­
structure interaction, the site effects only modify the earthquake re­ ical capability of geopolymer grout, which is made by some alkaline
cords. Güllü et al. [5] plotted seismic hazard maps for Gaziantep city activator solutions and some precursor of aluminosilicate sources,
considering the site conditions by the GIS technique, and compared providing alternatives for feasible grouting compared to native cement.
ground-motion prediction equations using the weighting factors to On the other hand, Güllü [12,13] proposed gene expression program­
further conducted a seismic hazard study [6]. Güllü [7] innovatively ming to predict the rheological characteristics of grout with bottom ash
applied artificial neural network and genetic expression programming to and the grout mixtures with various stabilizers for jet grouting columns,
predict average shear wave velocity, which is the widely defining achieving estimations with excellent precision. Soil after ground
method to reflect the site effects. This work is beneficial to improve the improvement-structure interaction may be an interesting and

* Corresponding author. State Key Laboratory of Disaster Reduction in Civil Engineering, Tongji University, Shanghai, 200092, China.
E-mail address: [email protected] (Z. Lu).

https://doi.org/10.1016/j.soildyn.2020.106333
Received 15 November 2019; Received in revised form 26 June 2020; Accepted 18 July 2020
Available online 3 August 2020
0267-7261/© 2020 Elsevier Ltd. All rights reserved.
S. Liu et al. Soil Dynamics and Earthquake Engineering 138 (2020) 106333

Fig. 1. (a) Photo of SF model; (b) the arrangement of measuring points.

meaningful research direction in the future. performance-based plastic design method to analyze the SSI effects of
Recently it has made some progress in the aspect of theoretical steel moment-resisting frames. Fatahi et al. [33] and Tabatabaiefar et al.
studies, numerical analysis and experimental studies on soil-structure [34] employed finite difference method by FLAC 2D software to inves­
interaction. Some theoretical models and analytical methods have tigate the effects of Plasticity Index on the seismic performance of
been developed to investigate the complex problem. Carbonari et al. mid-rise building frames resting on soft soil, and the soil characteristics
[14] proposed a lumped parameter model for the time domain SSI effects on seismic response of moment resisting building frames. Also,
analysis of the frequency-dependent dynamic stiffness of pile group Far et al. [35] made a comprehensive critical review on modelling
foundations. Lu et al. [15] presented an improved replacement oscillator techniques and computation methods for dynamic SSI analysis.
approach transforming a flexible-base single-degree-of-freedom (SDOF) With the goal of developing theories or validating analytical solu­
structure into an equivalent fixed-base SDOF oscillator. Jin et al. [16] tions, experimental studies of SSI systems have been carried out. They
derived a closed-form analytical solution of a SDOF oscillator on can be divided into two categories: full-scale field test and reduced-scale
semi-circular flexible composite foundation in half-space. Arias et al. model test. Full-scale field testing, such as Millikan Library forced vi­
[17] introduced an SSI model accounting for the response of all vibra­ bration tests [36], in situ forced vibration and pushover tests of school
tion modes. Besides, boundary reaction method [18] and near-field buildings in Taiwan [37], can be more accurate to study the actual
method [19] are presented for nonlinear analysis of structures behavior, but be restricted due to high expense and long timescales. The
including soil-structure interaction. alternative is reduced-scale model test. A 1/4-scale model of a nuclear
Numerical analyses have been conducted to access the SSI effects on power containment building constructed at Hualien [38] was adopted
the structural dynamic responses under earthquake excitations. Taba­ for investigation of the SSI effects during strong earthquake events.
tabaiefar et al. [20] reviewed research studies on seismic behaviour of Hushmand et al. [39] performed centrifuge tests to study the seismic
braced and unbraced building structures considering SSI, and analyzed response of stiff-unyielding buried reservoir structures. The shake table
the dynamic behaviour of unbraced steel frames resting on soft ground tests are widely used to measure and evaluate the actual behaviors of
[21]. Cruz et al. [22] and Medina et al. [23] evaluated dynamic prop­ structure under seismic excitations. A shaking table test was purposely
erties for structures in soil-structure system subjected to earthquakes. designed by Pitilakis et al. [40] to prove the ability of the numerical code
Abell et al. [24] compared the differences in response to 3-D, 3�1-D, and MISS3D to simulate the SSI phenomenon. Durante et al. [41] assessed
1-D excitations in the seismic soil-structure interaction system. Factors the modal dynamic response of the soil–pile–structure system in terms of
such as topographic effect, near-fault and far-fault earthquakes, reso­ period elongation and system damping ratio through the shaking table.
nance effect, regional subsidence effects, which impact the SSI effects, Shaking table SSI studies have also been reported by George et al. [42],
are studied by Alitalesh et al. [25] and Mayoral et al. [26]. Güllü et al. Zhang et al. [43] and Shen et al. [44]. In these shaking table tests, the
[27]discussed the soil–structure resonance effects on a structure superstructures of most models used for studying SSI effects have mainly
including SSI, indicating that the increased soil layer thickness leads to been designed with too small scale, single degree of freedom or special
prominent resonance effects on the structure. Güllü et al. [28] per­ building, which is hardly representative and cannot truly reflect the
formed SSI analysis of a historical stone masonry mosque to compare the seismic responses of majority buildings.
effects of near and far-fault earthquakes on its seismic responses, and Besides, the aforementioned past studies, always separately studied
also carried out full 3D nonlinear time history analysis for a historical the SSI effects in limited side-analytical method, experimental study or
masonry stone arch bridge considering SSI effects [29]. These efforts are numerical analyses. Most of previous numerical studies are lack of
helpful for the protection of historical structures under seismic waves. experimental validation. Without experimental verification, either
Furthermore, some scholars made efforts on comprehensive para­ analytical method or numerical analyses may be not convincing, and
metric study for the dynamic responses of structure considering soil- cannot be adopted into practical engineering. In turn, validated nu­
structure interaction, to provide references for seismic design codes merical analyses of entire soil–foundation–structure systems subjected
and provisions, especially for performance-based seismic design. Both to the real earthquake motions provide an alternative to shaking table
Bararnia et al. [30] and Hassani et al. [31] estimated the inelastic tests for extending research on this topic. Therefore, it is necessary to
displacement ratios of soil-structure systems, which is a critical role in carry out a large-scale shaking table test on SSI effect combined with the
performance-based seismic design. Ganjavi et al. [32] used a corresponding numerical simulation.

2
S. Liu et al. Soil Dynamics and Earthquake Engineering 138 (2020) 106333

Fig. 2. Gd =G0 eγd and λeγd curves of the soil.

Fig. 3. The acceleration time-history and corresponding Fourier spectrum (a) SJ; (b) IW; (c) CC.

In the present study, shaking table tests featuring soft soil and a six- 2. Material and method
story steel frame structure were specifically designed to consider soil-
structure interaction, and conducted at Tongji University, China. A nu­ 2.1. Test
merical model on the SSI tests is also validated in this paper. Further­
more, with the verified model, extensive parametric studies with The test model in this study was designed according to scaling laws.
different nonlinear soil types and structural properties are considered. The geometric and time similarity coefficients were 1/4. Models with
The main objectives are to comprehensively evaluate SSI effect on the two types of foundation were designed: the superstructure was con­
dynamic response of structure and to identify how some parameters structed on the rigid foundation (RF model), and an SSI test model
affect soil-structure interaction. consisting of soil, 3 by 3 pile group foundation, and a superstructure (SF
model). Fig. 1(a) shows the configuration of SF model.
Following the Meymand’s practice [45] and considering the facility

3
S. Liu et al. Soil Dynamics and Earthquake Engineering 138 (2020) 106333

Fig. 4. (a) The first natural frequency of superstructure; (b) the first natural frequency of soil; (c) the first mode damping ratio of superstructure.

Fig. 5. Acceleration time histories and corresponding Fourier spectra of points at various soil depths under CC5.

bearing capacity of the shaking table, Lu et al. [46] designed the later­ dimensions of 10 � 80 � 600 mm. The slabs consisted of steel plates
ally flexible but radially stiff cylindrical container in this study, which (Q345) with plane dimensions of 0.7 � 0.7 m and a thickness of 50 mm.
can reduce the undesirable boundary effects, and the boundary di­ A mixture with a mass content of 1:2.5 of sawdust and sand [50] was
mensions were taken as 3000 � 1500 mm. Chen et al. [49] set the ratio adopted as the model soil. Fig. 2 provides the Gd =G0 eγd and λeγd curves
D/d of the soil plane diameter D to the structure plane dimension of the test soil through cyclic triaxial tests, where Gd , G0 , λ and γ d are the
d greater than 5 in SSI study, and the error caused by lateral boundary dynamic shear modulus, initial shear modulus, damping ratio, and shear
effects was very small and tended to be stable. The ratio D/d in this study strain, respectively. The effective confining pressure applied to the
was very close to 5. A lot of shaking table tests on SSI study were con­ specimens in the cyclic triaxial tests is 20 kPa, and the measured values
ducted using this container [44,46–48]. The superstructure was a G0 of the two specimens are 8.80 MPa and 8.96 MPa respectively.
six-story steel frame structure, with the total mass and height being Displacement meters (label D) and accelerometers (label A for su­
1500 kg and 5.48 m respectively. The frame columns consisted of perstructure, label S for soil) were adopted to measure the seismic re­
high-strength steel plates (Q690) with width � length � height sponses of SSI system, as shown in Fig. 1(b).

4
S. Liu et al. Soil Dynamics and Earthquake Engineering 138 (2020) 106333

Fig. 6. The magnification coefficients of peak acceleration for soil. (a) SJ; (b) IW; (c) CC.

Fig. 7. The acceleration time-history and corresponding 5%-damped spectral acceleration at the top floor (a) IW4; (b) CC6; (c) SJ8.

According to Code for seismic design of buildings (GB50011-2010) in acceleration increased gradually from 0.05 to 0.4 g (g is the acceleration
China [51], the artificial Shanghai bedrock wave SHW2, 2008 Iwate due to gravity). The complete loading sequences included eight excita­
seismic wave and 1999 Chichi seismic wave were utilized as the input tion levels that were labeled from SJ1 to SJ8, IW1 to IW8 and CC1 to
excitations for SSI system, labeled SJ, IW and CC hereafter. It should be CC8. Before and after each level of acceleration was input, white noise
noted that to demonstrate the SSI effects reasonably, the ground motions (WN) with small amplitude was used as input to see the dynamic
of S13 under the SSI case, were used as the input excitations for corre­ characteristics, labeled from WN1 to WN9.
sponding rigid base cases. The soil surface S13 point is 0.9 m distant
from the container center, and represents the free-field ground response.
Fig. 3 shows the acceleration time history and corresponding Fourier 2.2. Numerical method
spectrum. Each type of seismic wave acted in one direction, and the peak
A three-dimensional model of soil-structure interaction shaking table

5
S. Liu et al. Soil Dynamics and Earthquake Engineering 138 (2020) 106333

Fig. 8. The peak floor acceleration of superstructure (a) SJ; (b) IW; (c) CC; (d)Average.

test is built using ABAQUS software. In the numerical model, three degrees of freedom between the two elements is solved by the constraint
dimensional solid elements are used for the group pile foundation and equation.
soil, whereas columns and slabs of the superstructure are meshed using The simulation of soil’s nonlinear behavior is very critical in SSI
three-dimensional beam elements and shell elements, respectively. problem. The impact of effective confined pressure is considered for the
Surface-to-surface contact with finite sliding is used for soil-structure initial shear modulus of each soil layer [55].
interface, wherein its normal and tangential behavior adopts hard con­
tact and penalty friction formulation respectively, and the friction co­ 2.2.1. Equivalent linear method
efficient is set to 0.33 [52]. The nodes at the same height along the The nonlinear phase of soil under earthquake wave is considered by
cylinder, are the same displacement in the direction of seismic wave the equivalent linearization method using the procedure SHAKE91 [56],
input, to represent the physical conditions imposed by the flexible which is an equivalent linear site response analysis program and widely
container. Similar way was also used by other researchers in numerical used in seismic response analysis of soil [24,57,58]. The effective dy­
simulation for test [47,48,52–54]. The equation function of ABAQUS namic shear strain in calculation is taken as 0.65 times the maximum
software is adopted to achieve it. In addition, the column elements-B31 dynamic shear strain. Rayleigh damping is used in ABAQUS. Rayleigh
has six degrees of freedom, while the cap elements-C3D8R has only three damping could be shown in Equation (1):
translational degrees of freedom. The problem of the uncoordinated

6
S. Liu et al. Soil Dynamics and Earthquake Engineering 138 (2020) 106333

½C� ¼ a0 ½M� þ a1 ½K� (1)

where a0 and a1 are the mass and stiffness damping coefficient,

respectively, ½M� is the mass matrix and ½K� is the stiffness matrix. By
means of orthogonality and taking the same damping ratio of each order,
i.e. ξm ¼ ξn ¼ ξ, it can be obtained:
� � � �
a0 2ξ ωm ωn
¼ (2)
a1 ωm þ ωn 1

where ωm is the fundamental circular frequency of the SSI system, and

ωn is the fundamental circular frequency of the earthquake excitation.
With the help of SHAKE91, the damping ratio of soil layer is obtained
using the curves from the above tests (Fig. 2).

2.2.2. Bounding surface plasticity model

A nonlinear soil model is also used for the nonlinear analyses, which
is a multiaxial plastic with viscous damping soil model with a vanishing
elastic region and coupled fully nonlinear behavior proposed by Borja
et al. [59]. It was also calibrated and validated using centrifuge test data
by Zhang et al. [52]. According to Borja et al. [60], the soil parameters
are calibrated by equation (3):
Z 2τ 0 " �pffiffi !m # 1
G 3 τ0 τ þ R 2
¼1 h þ H0 dτ (3)
Gmax 2γ0 0 τ

where G ¼ ​ τ0 =γ0 is the secant shear stiffness, and R is the radius of the
bounding surface. Parameters h, m, and H0 are related to the intensity of
the hardening. When the shear strain is large, the secant stiffness is going
to be zero, which causes H0 ¼ 0. As shown in Zeghal et al. [61], the
radius of the bounding surface could be calculated in equation (4):
pffiffi pffiffi
R ¼ 2 τmax � 2 Gγmax ¼5% (4)

Based on the test result of soil, the parameters for the bounding
surface plasticity model are chosen as ​ R � 0:0235Gmax , h ¼ 0:11Gmax ,
Fig. 9. The components of displacement.
m ¼ 0:93, ξ0 ¼ 0.014.

Fig. 10. The time history of the displacement component at the top of superstructure (CC8). (a) Ug ; (b) Hθ; (c) Ue ; (d) U.

Fig. 11. The displacement time-history at the top floor (a) IW4; (b) SJ8.

7
S. Liu et al. Soil Dynamics and Earthquake Engineering 138 (2020) 106333

Fig. 12. The peak value of elastic floor displacement (a) SJ; (b) IW; (c) CC; (d)Average.

3. Results remarkable caused by the soil getting softer.

3.1. Test results 3.1.2. Acceleration response

As shown in Fig. 1 (b), S7–S13 soil accelerometers are arranged at the
This study mainly concentrates on the impact of soil-structure same plane position and different depths in SF model to measure the
interaction on the dynamic responses of structure. The dynamic char­ free-field response. It should be noted that the free-field responses
acteristics under white noise, the acceleration and displacement re­ represent the site effects in the test and don’t include SSI effects. Fig. 5
sponses of superstructure subjected to seismic loadings are discussed to shows the acceleration time histories and corresponding Fourier spectra
evaluate SSI effects. of S7–S13 under CC5. As the decrease of the depth of the soil, the peak
value of the ground motion increases obviously. Moreover, the ground
3.1.1. Dynamic characteristic motion is obviously amplified at low-frequency components, especially
It is well recognized that compared with rigid foundation, the dy­ at soil fundamental frequency, while the amplification at medium-
namic characteristics of the SSI system would change due to the rotation frequency and high-frequency components is relatively limited.
and translation of foundation [36]. Fig. 4 (a) and (c) present the The peak acceleration of the free field response is given compared to
fundamental frequency and damping ratio of RF and SF model, and that of the container bottom by magnification coefficient of peak ac­
Fig. 4 (b) shows the fundamental frequency of soil. The first natural celeration in Fig. 6. The peak values of soil acceleration increase first,
frequency of superstructure on rigid base is 1.56 Hz, while that of SSI then decreases, and then increases from the bottom to the top of the soil.
system is about 1.47 Hz, decreasing about 6%. The reason why signifi­ When the seismic wave propagation gets close to the soil surface, the
cant reduction extent of superstructure’s natural frequency doesn’t peak acceleration appears evident amplification. The magnification
appear is that the superstructure in the test is relatively flexible. In Fig. 4 coefficient of soil acceleration at each point is greater than 1, which
(b), with the increase of seismic peak value, the soil frequency decreases demonstrates the amplification effect of soil on seismic peak value. In
obviously, and the soil frequency also decreases to a certain extent under addition, the stiffness of soft clay degrades with increasing excitation,
white noise, showing the degradation of soil stiffness. It is observed that and soil nonlinearity increases, leading to the smaller acceleration
the damping ratio of SF model is significantly larger than that of RF magnification coefficient.
model, demonstrating that SSI effect increases the damping ratio of 1st Absolute acceleration could reflect the floor inertia force under
mode. As the amplitudes of input seismic wave increases, the increase of earthquake excitations to some extent. This section analyzes the abso­
the damping ratio by soil-structure interaction becomes more lute acceleration of each floor, i.e. the measured results of

8
S. Liu et al. Soil Dynamics and Earthquake Engineering 138 (2020) 106333

Fig. 13. The peak value of storey drift angle (a) SJ; (b) IW; (c) CC; (d) Average.

Fig. 14. Comparison between the peak ground acceleration versus depth in the free-field column. (a) SJ4; (b) IW4; (c) CC4.

accelerometers arranged at each floor. Fig. 7 is the time history of the more obvious at the late seismic input.
absolute acceleration and the corresponding 5%-damped spectral ac­ As can be seen in Fig. 8, the larger magnitude of the excitation, the
celeration of the top floor under IW4, CC6 and SJ8 excitations. Fig. 8 larger difference of the peak floor acceleration between RF and SF
compares the peak floor acceleration at each floor of RF and SF models. models, and the mitigation effect of SSI on acceleration response per­
It can be seen that the absolute acceleration time history and its forms better. Besides, SSI has considerable influence on the distribution
corresponding 5%-damped spectral acceleration of SF model are much law of peak floor acceleration. Under SJ and CC waves, the maximum
smaller than that of RF model, which indicates that SSI effectively acceleration of superstructure appears at the first floor, regardless of SSI
mitigates the seismic acceleration response of floor. Additionally, as a system or rigid foundation, but in the case of rigid foundation, another
result of the dissipated energy by soil, the spectral acceleration at soil maximum point appears at the fifth floor, which does not appear in SSI
frequency (around 10 Hz) of RF model is rather larger than that of SF system. From the average absolute acceleration, the mitigation effect
model, and the mitigation effect of SSI on acceleration response becomes under IW wave is not as good as that under SJ and CC waves, as IW wave

9
S. Liu et al. Soil Dynamics and Earthquake Engineering 138 (2020) 106333

Fig. 15. Point S13 acceleration time history and corresponding spectral acceleration of computation and experimental results under (a) SJ4; (b) IW4; (c) CC4.

consists more high-frequency component. The average absolute accel­ slightly smaller than that of SF model, which also demonstrates SSI
eration of RF model is 1.5–2 times that of SF model. extends the movement period of superstructure.
Fig. 12 shows the peak value comparison of elastic displacement
3.1.3. Displacement response between RF and SF model, and that of storey drift angle is plotted in
The components of displacement at the top of the superstructure in Fig. 13, which is one of the widely recognized seismic performance
SSI system are shown in Equation (5) and Fig. 9. parameters and can be effectively used to assess the level of structural
damage under earthquake actions [32]. From Fig. 10, it can be seen that
U ¼ Ug þ Hθ þ Ue (5)
higher acceleration peak values yield larger peak elastic displacement of
RF model, and in most cases, that of SF model also conforms to this law.
where U is the total displacement of superstructure. Hθ is the rocking
However, the peak elastic displacement of SF model in middle earth­
component due to foundation rotation. Ug is the translational compo­
quake case is greater than that in large earthquake case. This is due to
nent, and Ue is the elastic deformation component of superstructure.
the phase difference between the rotation displacement and the total
The time histories of the displacement components at the top of su­
displacement, and they may be out of phase which results in the increase
perstructure under CC8 excitation are shown in Fig. 10. The three
of the peak elastic displacement [48].
components may be in phase or out of phase, and do not fully syn­
In Figs. 12 and 13, the elastic displacement and storey drift angle at
chronize with each other. It is evident that structural elastic displace­
each floor in SF model, both the peak and average value, are obviously
ment accounts for the largest proportion of the total displacement, and
smaller than that in RF model, which implies that SSI effects remarkably
the rotation displacement also accounts for a large proportion, which is
decrease the floor elastic displacement and storey drift angle, and the
consistent with the observation of forced vibration tests on Millikan
difference between SF model and RF model becomes larger as acceler­
Library [36]. Therefore, comparing the elastic displacement between
ation peak values increase. But the SSI effects on the elastic displace­
the two models is more reasonable.
ment and storey drift angle at each floor are quite different under
Fig. 11 shows the time history of elastic displacement at the top floor
different seismic waves, as a result of the difference of earthquake
in IW4 and SJ8 cases. Considering the soil-structure interaction, there is
spectrum characteristics, and it is most obvious under CC wave. In
a significant decrease of the elastic displacement at the top floor, espe­
addition, the peak elastic displacement from lower to upper floor in SF
cially at the late seismic input. The displacement period of RF model is
model is almost a straight line, which implies that its distribution of

10
S. Liu et al. Soil Dynamics and Earthquake Engineering 138 (2020) 106333

Fig. 16. Point A1 acceleration time history and corresponding spectral acceleration of computation and experimental results under (a) SJ4; (b) IW4; (c) CC4.

inter-story drift is uniform. So it is obvious in Fig. 13 that the peak storey around soil frequencies too large and estimates the seismic response at
drift angle from lower to upper floor in SF model is almost constant, other input frequencies too small. Hence, the modeling method is
while that in RF model reaches maximum at the second floor then it will reasonable, and the nonlinear model is more capable of conducting the
decrease as the floor rises. numerical analysis of the SSI system.

3.2. Comparison of the numerical and experimental results 3.3. Calculation results

Fig. 14 shows the comparison between the peak ground acceleration Based on the modeling method described in the previous section, in
versus depth in the free-field column S7–S13. Figs. 15–17 compare the which the soil adopts bounding surface plasticity model, the influence of
acceleration time history and the corresponding 5%-damped spectral some important parameters, including nonlinear soil type and structural
acceleration comparison between simulation results and experimental property, are presented.
results under moderate earthquake, in which Point S13 represents the
free-field ground response, Point A1 is at superstructure base and Point 3.3.1. Effects of structural property
A7 is at the superstructure top. The numerical results from equivalent Steel frame structures with different frequencies are obtained by
linear method (EL) and nonlinear model with viscous damping (NLV) changing the section size of frame columns, and the effects of SSI on steel
are compared with the test data. frame structures with different frequencies are compared. The SJ wave,
From the figures, it is found that acceleration response predicted by IW wave and CC wave in shaking table tests are used as the input seismic
equivalent linear method and nonlinear model generally agree with wave, and the peak values are adjusted to 0.2 g. The geometries and
experimental data, but the bounding surface plasticity model is superior materials in shaking table test remain unchanged. The cross-sections of
to the equivalent linear method in simulating the seismic responses of frame columns are taken as 10 � 80 mm, 20 � 100 mm and 40 � 120
the SSI system, both in terms of peak acceleration and its 5%-damped mm, respectively. The corresponding frequencies of steel frame on rigid
spectral acceleration. It can be seen from the spectral acceleration that base are 1.56 Hz, 4.38 Hz and 8.47 Hz, respectively.
bounding surface plasticity model performs more stably, and can accu­ Considering soil-structure interaction, the frequencies of soil-
rately predict the responses over a broad range of input frequencies structure interaction system with frame column sections of 10 � 80
However, the equivalent linear model predicts the seismic responses mm, 20 � 100 mm and 40 � 120 mm are 1.49 Hz, 3.25 Hz and 3.94 Hz,

11
S. Liu et al. Soil Dynamics and Earthquake Engineering 138 (2020) 106333

Fig. 17. Point A7 acceleration time history and corresponding spectral acceleration of computation and experimental results under (a) SJ4; (b) IW4; (c) CC4.

respectively. With higher frequency of steel frame structure, the fre­

quency of steel frame structure becomes closer to the soil frequency and
the frequency of soil-structure interaction system decreases more
compared to rigid foundation, and the effects of soil-structure interac­
tion are intensified.
The pile locations are indicated by number and are provided in
Fig. 18. The contact pressure amplitude of the No. 3, 5, 2 and 6 piles
along the height of each pile under different excitations are plotted in
Fig. 19.
It can be seen from Fig. 19 that contact pressure amplitude of side
piles is much larger than that of middle piles, and that of side piles 6 is
the largest. The contact pressure amplitudes at the pile-soil interface of
different piles have differences at both ends of the pile while displaying
little difference at the middle of the pile, especially for side piles. The
amplitude of contact pressure between pile and soil basically exhibits a
“K" type distribution.
It is a growing trend of the contact pressure between pile and soil
with a higher frequency of steel frame structure in soil-structure inter­
action system, especially under SJ wave and CC wave, which also im­
plies the enhancement of interaction between pile and soil.

3.3.2. Effects of soil type

From the formula vs ¼ 4Hf, where H is the soil height, f is the
fundamental frequency of soil and vs is the shear wave velocity, the
Fig. 18. Locations of piles by number.
shear wave velocity of the soil used in shaking table test is about 60 m/s.

12
S. Liu et al. Soil Dynamics and Earthquake Engineering 138 (2020) 106333

Fig. 19. The contact pressure amplitude of different piles with different structural properties. (a) SJ; (b) IW; (c) CC.

Fig. 20. The Fourier response spectra of Point A7 under different soil types. (a) SJ4; (b) IW4; (c) CC4.

According to the soil types in the Chinese code [51], the shear wave test belongs to soft soil.
velocities of soil less than 150 m/s2, between 150 m/s2 and 250 m/s2, In order to compare the influence of different soil types on soil-
between 250 m/s2 and 500 m/s2 are defined as soft soil, medium soft soil structure interaction, this section chooses medium-soft soil with vs ¼
and medium dense soil, respectively. Thus, the soil used in shaking table 200 m/s and medium-dense soil with vs ¼ 400 m/s, as well as the soil in

13
S. Liu et al. Soil Dynamics and Earthquake Engineering 138 (2020) 106333

Fig. 21. Peak acceleration magnification coefficients under different soil types. (a) SJ; (b) IW; (c) CC.

Fig. 22. The contact pressure amplitude of different piles under different soil types. (a) SJ; (b) IW; (c) CC.

shaking table test. The shear modulus of soil is adjusted according to the consideration. It is evident that (a) for soil, the magnification coefficient
pffiffiffiffiffiffiffiffi
formula vs ¼ G=ρ , where G is the shear modulus of soil, and ρ is the of peak acceleration is greater than 1, and the peak acceleration of
soil density. The superstructure of the calculation model is a steel frame seismic wave is amplified by the site effects of soil, which is consistent
with 40 � 120 mm column section and 8.47 Hz fundamental frequency. with the research conducted by Seed et al. [62]; (b) due to remarkable
In the case of soft soil, medium soft soil and medium dense soil, the soil-structure interaction, the peak floor acceleration of the floor in soft
frequencies of soil-structure interaction systems are 3.94 Hz, 7.08 Hz soil is the smallest; and (c) its variation law is obviously different from
and 7.85 Hz, respectively. Fig. 20 displays the Fourier spectra of accel­ that in the other two soil types, where the magnification coefficient of
eration response at the top floor under different soil types, which in­ floor acceleration in soft soil decreases first and then increases, while in
dicates that the smaller the soil stiffness, the more the SSI frequencies medium-dense soil and medium-soft soil, the magnification factor of
shift to the low. Compared with the superstructure frequency on rigid floor acceleration increases first, then decreases and then increases
foundation, as shear wave velocity of soil increases, the declining scale again.
of the SSI system frequency decreases, and the effect of soil-structure As shown in Fig. 22, the pile-soil contact pressure of the same pile in
interaction weakens. medium-soft soil or medium-dense soil is much higher than that in soft
Fig. 21 displays the acceleration magnification coefficients along soil. As the soil stiffness increases, the restraint effect of soil on pile
with the height of the SSI system with different soil types under different becomes more intensive, leading to the larger contact pressure between
seismic excitations. The peak accelerations of each story and the free- pile and soil. However, it is obvious that for the soil-structure system
field are given compared to that of the container bottom in the figure. with medium-soft soil and medium-dense soil, results can remarkably
The free-field responses represent the site effects, which doesn’t include vary from one ground motion to another.
SSI effects, while the superstructure responses take SSI effects into As mentioned before, in the SSI system, the displacement of super­
structure includes three parts: the rocking component, the translational

14
S. Liu et al. Soil Dynamics and Earthquake Engineering 138 (2020) 106333

Fig. 23. Peak displacement of different components under different soil types (a) total; (b) the rocking; (c) elastic.

component, and the elastic deformation component. The amplitude of in estimating the seismic responses of SSI system over a broad range of
the total, the rocking, and the elastic displacement are illustrated in input frequencies.
Fig. 23. It can be seen from Fig. 23: the total displacement of the When the frequency of superstructure is close to the soil frequency,
structure in the soft soil is much larger than that in the other two soil the effect of soil-structure interaction is very considerable, and the soil-
types, which should be taken seriously for practical engineering on soft structure interaction significantly changes the dynamic characteristics
soil; harder soil yields less rotational displacement of superstructure; of the structure. In practical engineering, especially on soft soil, the
while the minimum elastic displacement of superstructure is observed in design depends on the dynamic characteristics of the structures, like
the soft soil. Decreasing the soil stiffness is accompanied by reducing the TMDs, should take the SSI effects into consideration. As the super­
restraint effect of soil on pile foundation and increasing the proportion structure frequency in the soil-structure interaction system increases,
of rotational displacement. It should be remarked that although SSI the contact pressure between pile and soil increases gradually, and the
decreases the elastic displacement of from the test results. soil-structure interaction becomes more significant.
The soil amplifies the peak acceleration of input motion from the
4. Discussion shaking table under moderate earthquake for the soil types explored in
this paper. Higher shear wave velocity of soil, compared with the su­
From the shaking table test results, for rigid foundation, the funda­ perstructure frequency on rigid foundation, causes a smaller decline
mental frequency of SSI systems slightly decreases, but the damping scale of the SSI system frequency, and the effect of soil-structure inter­
ratio significantly increases. The results suggested that SSI has a positive action weakens. Compared to those in medium-soft soil and medium-
mitigation effect on the dynamic response of superstructure, both in dense soil, the peak floor acceleration, the peak elastic displacement
terms of acceleration and displacement response. Therefore, the rigid and the pile-soil contact pressure in soft soil is the smallest, but the peak
foundation assumption is conservative in practical design. SSI also total displacement and rotation displacement is the largest. The restraint
changes the distribution law of maximum floor acceleration, floor effect of the soil on the pile foundation is intensified by increasing the
displacement and storey drift ratio. soil stiffness, and the relative movement between pile and soil becomes
Compared with data generated by shaking table tests, a three- more difficult.
dimensional model with bounding surface plasticity model for soil is
validated, which has better performance to the equivalent linear method

15
S. Liu et al. Soil Dynamics and Earthquake Engineering 138 (2020) 106333

5. Conclusion [14] Carbonari S, Morici M, Dezi F, Leoni G. A lumped parameter model for time-
domain inertial soil-structure interaction analysis of structures on pile foundations.
Earthq Eng Struct Dynam 2018;47:2147–71.
The present study carried out shaking table tests on dynamic soil- [15] Lu Y, Hajirasouliha I, Marshall AM. An improved replacement oscillator approach
structure interaction subjected to earthquake motions and established for soil-structure interaction analysis considering soft soils. Eng Struct 2018;167:
a numerical simulation method using ABAQUS software. Comparing the 26–38.
[16] Jin LG, Liang JW. The effect of foundation flexibility variation on system response
test and numerical results, the modeling method is validated. Based on of dynamic soil-structure interaction: an analytical solution. Bull Earthq Eng 2018;
the verified model, the influences of soil type and structural property on 16:113–27.
soil-structure interaction are explored. Findings can be summarized as [17] Arias H, Jaramillo JD. Base shear determination using response-spectrum modal
analysis of multi-degree-of-freedom systems with soil-structure interaction. Bull
follows: Earthq Eng 2019;17:3801–14.
SSI effects obviously decrease the structural response under seismic [18] Kim JM, Lee EH, Lee SH. Boundary reaction method for nonlinear analysis of
excitations. A three-dimensional model with bounding surface plasticity soil–structure interaction under earthquake loads. Soil Dynam Earthq Eng 2016;89:
85–90.
model for soil predicts dynamic responses of SSI system more accurately. [19] Ghandil M, Behnamfar F. The near-field method for dynamic analysis of structures
As the frequency of superstructure gets close to the soil frequency or the on soft soils including inelastic soil-structure interaction. Soil Dynam Earthq Eng
soil stiffness decreases, the effects of soil-structure interaction become 2015;75:1–17.
[20] Tabatabaiefar SHR, Clifton T. Significance of considering soil-structure interaction
more considerable. effects on seismic design of unbraced building frames resting on soft soils. Aust
Geomech J 2016;51(1):55–64.
Credit author statement [21] Far H. Dynamic behaviour of unbraced steel frames resting on soft ground. Steel
Construct. 2019;12:135–40.
[22] Cruz C, Miranda E. Evaluation of soil-structure interaction effects on the damping
Shutong Liu: Conceptualization, Methodology, Software, Writing - ratios of buildings subjected to earthquakes. Soil Dynam Earthq Eng 2017;100:
original draft, preparation. Peizhen Li: Data curation, Investigation, 183–95.
Supervision. Wenyang Zhang: Software, Validation. Zheng Lu: Visu­ [23] Medina C, Azn� arez JJ, Padr�
on LA, Maeso O. Effects of soil-structure interaction on
the dynamic properties and seismic response of piled structures. Soil Dynam Earthq
alization, Writing - review & editing, Supervision. Eng 2013;53:160–75.
[24] Abell JA, Orbovi�c N, McCallen DB, Jeremic B. Earthquake soil-structure interaction
Declaration of competing interest of nuclear power plants, differences in response to 3-D, 3 � 1-D, and 1-D
excitations. Earthq Eng Struct Dynam 2018;47(6):1478–95.
[25] Alitalesh M, Shahnazari H, Baziar MH. Parametric study on seismic topography-
The authors declare that they have no known competing financial soil-structure interaction; topographic Effect. Geotech Geol Eng 2018;36:2649–66.
interests or personal relationships that could have appeared to influence [26] Mayoral JM, Casta~ non E, Albarran J. Regional subsidence effects on seismic soil-
structure interaction in soft clay. Soil Dynam Earthq Eng 2017;103:123–40.
the work reported in this paper. [27] Güllü H, Pala M. On the resonance effect by dynamic soil structure interaction a
revelation study. Nat Hazards 2014;72:827–47.
Acknowledgements [28] Güllü H, Karabekmez M. Effect of near-fault and far-fault earthquakes on a
historical masonry mosque through 3D dynamic soil-structure interaction. Eng
Struct 2017;152:465–92.
This study was supported by the National Natural Science Founda­ [29] Güllü H, Jaf HS. Full 3D nonlinear time history analysis of dynamic soil structure
tion of China (Grant No. 51978524), the Ministry of Science and Tech­ interaction for a historical masonry arch bridge. Environ. Earth Sci. 2016;75:1421.
[30] Bararnia M, Hassani N, Ganjavi B, Amiri GG. Estimation of inelastic displacement
nology of China (Grant No. SLDRCE19-B-30). The authors gratefully
ratios for soil-structure systems with embedded foundation considering kinematic
acknowledge financial support from China Scholarship Council. and inertial interaction effects. Eng Struct 2018;159:252–64.
[31] Hassani N, Bararnia M, Amiri GG. Effect of soil-structure interaction on inelastic
References displacement ratios of degrading structures. Soil Dynam Earthq Eng 2018;104:
75–87.
[32] Ganjavi B, Gholamrezatabar A, Hajirasouliha I. Effects of soil-structure interaction
[1] Mirzaie F, Mahsuli M, Ghannad MA. Probabilistic analysis of soil-structure and lateral design load pattern on performance-based plastic design of steel
interaction effects on the seismic performance of structures. Earthq Eng Struct moment resisting frames. Struct Des Tall Special Build 2019;28:e1624.
Dynam 2017;46(4):641–60. [33] Fatahi B, Tabatabaiefar SHR. Effects of soil plasticity on seismic performance of
[2] Bararnia M, Hassani N, Ganjavi B, Amiri GG. Estimation of inelastic displacement mid-rise building frames resting on soft soils. Adv Struct Eng 2014;17(10):
ratios for soil-structure systems with embedded foundation considering kinematic 1387–402.
and inertial interaction effects. Eng Struct 2018;159:252–64. [34] Tabatabaiefar SHR, Fatahi B, Samali B. Finite difference modelling of soil-structure
[3] Renzi S, Madiai C, Vannucchi G. A simplified empirical method for assessing interaction for seismic design of moment resisting building frames. Aust Geomech J
seismic soil-structure interaction effects on ordinary shear-type buildings. Soil 2012;47(3):113–20.
Dynam Earthq Eng 2013;55:100–7. [35] Far H. Advanced computation methods for soil-structure interaction analysis of
[4] Mylonakis G, Gazetas G. Seismic soil-structure interaction: beneficial or structures resting on soft soils. Int J Geotech Eng 2019;13(4):352–9.
detrimental? J Earthq Eng 2000;4(3):277–301. [36] Luco JE, Trifunac MD, Wong HL. Isolation of soil-structure interaction effects by
[5] Güllü H, Ansal AM, Ozbay
€ A. Seismic hazard studies for Gaziantep city in south full-scale forced vibration tests. Earthq Eng Struct Dynam 1988;16:1–21.
anatolia of Turkey. Nat Hazards 2008;44:19–50. [37] Ko YY, Chen CH. Soil-structure interaction effects observed in the in situ forced
[6] Güllü H, Iyisan
_ R. A seismic hazard study through the comparison of ground vibration and pushover tests of school buildings in Taiwan and their modeling
motion prediction equations using the weighting factor of logic tree. J Earthq Eng considering the foundation flexibility. Earthq Eng Struct Dynam 2010;39:945–66.
2016;20(6):861–84. [38] Choi JS, Yun CB, Kim JM. Earthquake response analysis of the Hualien soil-
[7] Güllü H. On the prediction of shear wave velocity at local site of strong ground structure interaction system based on updated soil properties using forced
motion stations an application using artificial intelligence. Bull Earthq Eng 2013; vibration test data. Earthq Eng Struct Dynam 2001;30:1–26.
11:969–97. [39] Hushmand A, Dashti S, Davis C, Hushmand B, Zhang M, Ghayoomi M,
[8] Ansal AM, Iyisan
_ R, Güllü H. Microtremor measurements for the microzonation of McCartney JS, Lee Y, Hu J. Seismic performance of underground reservoir
dinar. Pure Appl Geophys 2001;158:2525–41. structures: insight from centrifuge modeling on the influence of structure stiffness.
[9] Fatahi B, Tabatabaiefar SHR, Samali B. Soil-structure interaction vs Site effect for J Geotech Geoenviron Eng 2016;142(7). 04016020.
seismic design of tall buildings on soft soil. Geomech. Eng. 2014;6(3):293–320. [40] Pitilakis D, Dietz M, Wood DM, Clouteaua D, Modaressia A. Numerical simulation
[10] Güllü H. Unconfined compressive strength and freeze–thaw resistance of fine- of dynamic soil–structure interaction in shaking table testing. Soil Dynam Earthq
grained soil stabilised with bottom ash, lime and superplasticizer. Road Mater Eng 2008;28(6):453–67.
Pavement Des 2015;16(3):608–34. [41] Durante MG, Sarno LD, Mylonakis M, Taylor CA, Simonelli AL. Soil-pile-structure
[11] Güllü H, Cevik A, Al-Ezzi KMA, Gülsan ME. On the rheology of using geopolymer interaction: experimental outcomes from shaking table tests. Earthq Eng Struct
for grouting: a comparative study with cement-based grout included fly ash and Dynam 2016;45:1041–61.
cold bonded fly ash. Construct Build Mater 2019;196:594–610. [42] George G, Stokoe KH. Free vibration of embedded foundations: theory versus
[12] Güllü H. A new prediction method to rheological behavior of grout with bottom experiment. J. Geotech. Eng. 1991;117(9):1382–401.
ash for jet grouting columns. Soils Found 2017;57(3):384–96. [43] Zhang ZY, Wei HY, Qin X. Experimental study on damping characteristics of soil-
[13] Güllü H. A novel approach to prediction of rheological characteristics of jet grout structure interaction system based on shaking table test. Soil Dynam Earthq Eng
cement mixtures via genetic expression programming. Neural Comput Appl 2017; 2017;98:183–90.
28:407–20. [44] Shen C, Qian DL. Dynamic characteristics and seismic response of frame-core tube
structures, considering soil-structure interactions. Struct Des Tall Special Build
2019;28:e1575.

16
S. Liu et al. Soil Dynamics and Earthquake Engineering 138 (2020) 106333

[45] Meymand P, Riemer M, Seed R. Large scale shaking table test of seismic soil-pile [53] Lanzano G, Bilotta E, Russo G, Silvestri F. Experimental and numerical study on
interaction in soft clay. In: 12th world conference on earthquake engineering; circular tunnels under seismic loading. Eur. J. Environ. Civ. Eng. 2015;19(5):
2000. New Zealand. 539–63.
[46] Lu XL, Chen YQ, Chen B, Li PZ. Shaking table model test on dynamic soil-structure [54] Deng YH, Dashti S, Hushmand A, Davis C, Hushmand B. Seismic response of
interaction system. J Asian Architect Build Eng 2002;1(1):55–64. underground reservoir structures in sand: evaluation of Class-C and C1 numerical
[47] Ge Q, Xiong F, Xie LW, Chen J, Yu MJ. Dynamic interaction of soil-structure simulations using centrifuge experiments. Soil Dynam Earthq Eng 2016;85:202–16.
cluster. Soil Dynam Earthq Eng 2019;123:16–30. [55] Seed HB, Wong RT, Idriss IM, Tokimatsu K. Moduli and damping factors for
[48] Li PZ, Liu ST, Lu Z, Yang JP. Numerical analysis of a shaking table test on dynamic dynamic analyses of cohesionless soils. J. Geotech. Eng. 1986;112(11):1016–32.
structure-soil-structure interaction under earthquake excitations. Struct Des Tall [56] Seed HB, Idriss IM, Sun JI. User’s manual for SHAKE91. Davis: University of
Special Build 2017;26:e1382. California; 1992.
[49] Chen B, Lu XL, Li PZ, Chen YQ. Modeling of dynamic soil-structure interaction by [57] Tuladhar R, Yamazaki F, Warnitchai P, Saita J. Seismic microzonation of the
ANASYS program. J. Earthquake Eng. Eng. Vibrat. 2002;22(1):126–31. greater Bangkok area using microtremor observations. Earthq Eng Struct Dynam
[50] Yan X, Yuan JY, Yu HT, Bobet A, Yuan Y. Multi-point shaking table test design for 2010;33:211–25.
long tunnels under non-uniform seismic loading. Tunn Undergr Space Technol [58] Saffarini HS. Ground motion characteristics of the November 1995 Aqaba
2016;59:114–26. earthquake. Eng Struct 2000;22(4):343–51.
[51] CMC (China Ministry of Construction). Code for seismic design of buildings [59] Borja RI, Amies AP. Multiaxial cyclic plasticity model for clays. J. Geotech. Eng.
(GB50011–2010). Beijing, China: China Architecture & Building Press; 2010. In 1994;120(6):1051–70.
Chinese. [60] Borja RI, Lin CH, Sama KM, Masada GM. Modelling non-linear ground response of
[52] Zhang WY, Seylabi EE, Taciroglu E. Validation of a three-dimensional constitutive non-liquefiable soils. Earthq Eng Struct Dynam 2000;29:63–83.
model for nonlinear site response and soil-structure interaction analyses using [61] Zeghal M, Elgamal AW, Tang HT, Stepp JC. Lotung downhole array. II: evaluation
centrifuge test data. Int J Numer Anal Methods GeoMech 2017;41:1828–47. of soil nonlinear properties. J. Geotech. Eng. 1995;121:363–78.
[62] Seed HB, Idriss IM. Influence of soil conditions on ground motion during
earthquakes. J Soil Mech Found Div 1969;95:99–137.

17