Soil Dynamics and Earthquake Engineering 101 (2017) 209–224

Contents lists available at ScienceDirect

Soil Dynamics and Earthquake Engineering

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

Seismic response of clay-pile-raft-superstructure systems subjected to far- MARK

field ground motions
⁎
Lei Zhang, Huabei Liu
School of Civil Engineering and Mechanics, Huazhong University of Science and Technology, Wuhan, Hubei 430074, PR China

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

Keywords: A series of three-dimensional (3D) finite element analyses incorporated with a hyperbolic-hysteretic soil model
Earthquake were performed to investigate the seismic response of pile-raft-superstructure systems constructed on soft clay
Clay stratum, focusing on the seismic pile bending moment and superstructural responses. The seismic pile bending
Pile bending moment moment results suggested that using a lumped mass to represent the superstructure, which has been widely used
Acceleration
in many other studies, could only perform well for a relatively low-rise superstructure; on the other hand, the
Finite element analysis
Superstructure
seismic response of superstructure was found to be significantly affected by the soil-structure interaction, and
both the detrimental and beneficial effects of dynamic soil-structure interaction were observed. Hence, coupled
soil-foundation-superstructure analyses were primarily performed in this study. The influences of peak base
acceleration, pile flexural rigidity and the configuration of superstructure on both the pile bending moment and
superstructural responses were studied. Furthermore, some correlations were derived to relate the maximum pile
bending moment to the influencing factors, which can be used as useful tools for obtaining preliminary and first-
order estimates of the maximum pile bending moment for pile-raft-superstructure systems constructed on soft
clay deposits.

1. Introduction layers of soft clays. The performance of the pile-superstructure system

It is a common practice to assume that the structure is fixed at the
base and to apply the free-field ground motion at the base (e.g. [40])
when investigating the seismic response of a structure. In so doing, the
influence of dynamic soil-structure interaction (SSI) is neglected, which
may induce large prediction errors as the seismically-induced motion at
ground surface is not likely to be the same as that at building founda-
tion level [43,45,46]. The seismic SSI tends to increase the fundamental
resonance period and damping of the system in comparison with the
fixed-base assumption; as a result, the effect of seismic SSI is con-
ventionally considered beneficial and hence neglected as recommended
in many seismic codes such as ATC-3-06 [3] and NEHRP [16]. How-
ever, the detrimental effect of dynamic SSI was also reported in many
published studies (e.g. [41,27]); the effect of seismic SSI can have either
beneficial or detrimental effects on the seismic response of a structure,
depending on the factors of earthquake type, soil type, foundation
configuration and dynamic characteristics of the structure (e.g.
[28,36]). In fact, the seismic SSI has been recognized as being im-
portant and the coupled soil-foundation-superstructure analysis has
been recommended by many researchers [21,28,36,38].
Pile foundations have been widely used for buildings built on thick
against seismic shaking is an important area of study, which involves
complex dynamic soil-pile-superstructure interaction (SPSI) mechan-
isms. During an earthquake, piles are subjected to kinematic and in-
ertial forces respectively imposed by the surrounding soils and the su-
perstructure that they support, which may result in the piles being
subjected to structural distress leading to cracking or the formation of
plastic hinges as observed in many postearthquake investigations (e.g.
[9,17,26]). Besides, a major concern in seismic SSI is the amplification
of ground motion induced by the soft soil layer [8,34,39,47], which
may result in the piles and the superstructure being subjected to am-
plified loading even under small to moderate earthquakes. Further-
more, the performance of pile-superstructure system constructed on soft
clay stratum against earthquake is also an important consideration
under many national design codes, e.g. NEHRP [16], GB50011-2010
[37] and Eurocode 8 [14].
Significant works have been done in the area of seismic soil-pile-
structure interaction over the past decades. Most of the laboratory ex-
periments including centrifuge tests (e.g. [1,10,11,19,22,50]) and 1-g
shaking table tests (e.g. [4,18,23,24,25,30,48]) were focused on the
seismic response of soil-pile-superstructure installed in predominantly
sandy (liquefiable or dry) soil, while the relevant studies involving soft

⁎
Corresponding author.
E-mail addresses: zhangleigeo@hust.edu.cn (L. Zhang), hbliu@hust.edu.cn (H. Liu).

http://dx.doi.org/10.1016/j.soildyn.2017.08.004
Received 21 March 2017; Received in revised form 16 May 2017; Accepted 6 August 2017
0267-7261/ © 2017 Elsevier Ltd. All rights reserved.
L. Zhang, H. Liu Soil Dynamics and Earthquake Engineering 101 (2017) 209–224

clays are still relatively limited. One significant work in this area was superstructure were presented. Given the fact that the relevant studies
carried out by Meymand [35], who conducted a series of large scale 1-g involving soft clays are still relatively few, the findings obtained from
shaking table tests to study the seismic interaction of soft clay-pile-su- the present study can provide a useful reference for practical seismic
perstructure; Hokmabadi et al. [27] also performed a series of 1-g design of pile-raft-superstructure systems constructed on soft clays
shaking table tests to investigate the seismic response of superstructure subjected to far-field ground motions.
supported by a 4×4 pile group installed in a synthetic clay bed.
Banerjee [7] and Banerjee et al. [8] performed a series of centrifuge 2. Numerical modelling procedure
tests to study the dynamic response of pile-raft system embedded in soft
kaolin clay subjected to short-duration far-field earthquakes. In their 2.1. General information
work, the pile spacing along the shaking direction was 10 times the pile
diameter or more, and their test results were largely representative of Fig. 2 shows the configuration of the clay-pile-raft-superstructure
the seismic response of single piles embedded in soft clay. Zhang [51], systems adopted in the present numerical study, which contains a su-
Zhang et al. [52,53] also performed a series of centrifuge tests to in- perstructure with storey number ranging from 0 to 20 and supported by
vestigate the influence of pile group configuration on the seismic re- a 5×5 pile-raft system installed in a soft clay bed, the properties of
sponse of clay-pile-raft systems subjected to both long- and short- which are shown in Table 1. The piles were embedded in a pure clay
duration far-field ground motions. With the exception of the studies by bed, with the toes sitting atop a 0.5 m-tick sand layer. The properties of
Ayothiraman et al. [4] and Hokmabadi et al. [27], most of the afore- the sand were adopted following Banerjee [7], which are listed in
mentioned experimental studies treated the superstructure as either a Table 2. As can be seen, these clay-pile-raft-superstructure systems are
lumped mass or a simplified single degree of freedom oscillator and self-symmetrical with respect to the ground motion orientation, hence
hence the effect of higher modes of the superstructure was not ac- only a half 3D finite model of each system was set up using ABAQUS/
counted for in these studies. On the other hand, in order to fully account Explicit 6.13. Fig. 3 shows the 3D finite element model of the 20-storey
for the nonlinear behaviour of the soil under seismic loadings, numer- building supported by a clay-pile-raft system, which contains 19200
ical simulations such as finite element analysis (e.g. [8,21,33]) and fi- linear hexahedral elements, 5760 linear quadrilateral elements, and
nite difference analysis (e.g. [27,29]) are commonly performed in time 3225 linear beam elements.
domain to investigate the seismic SPSI. In addition, beam-on-dynamic-
Winkler-foundation model or dynamic p-y method is also a popular
2.2. Soil model
approach to account for the dynamic SPSI (e.g. [12,38]), for which the
parameters assigned to the springs and dashpots used for the p-y curve
The behaviour of the soft clay was simulated using a hyperbolic-
are usually back-calculated from the measured pile response.
hysteretic soil model as shown in Fig. 4, which was proposed by
In this study, a total of 90 three-dimensional (3D) finite element
Banerjee [7] and calibrated using laboratory test data from cyclic
analyses were performed using ABAQUS/Explicit to investigate the
triaxial and resonant column tests on kaolin clay. The basic shear stress-
seismic response of different pile-raft-superstructure systems con-
strain relationship for this hyperbolic-hysteresis model is shown in Eq.
structed on soft clay subjected to far-field ground motions. A newly
(1).
developed VUMAT subroutine was incorporated to account for the
hyperbolic-hysteretic soil behaviour that was proposed by Banerjee [7]
⎧q − 1
⎡ 3G max
⎤ Initial loading (backbone) path
on the basis of a series of resonant column and cyclic triaxial tests for ⎪ f 3G max / qf 1 + 3G max εs / qf
⎣ ⎦
soft clay. Two approaches, namely lumped mass and detailed modelling ⎪
⎪ 2 3G max
of the superstructure, which are respectively termed "added mass" and q = − 2qf + 3G / q ⎡ 1 + 3G (ε − ε ) / 2q ⎤ + qr1 Unloading path
⎨ max f
⎣ max r 1 s f⎦
"detailed model", were used to account for the inertial effect imposed by ⎪
⎪ 2qf − 2 3G max
a superstructure during the seismic shaking event. As Fig. 1 shows, the ⎡ ⎤ − qr 2 Reloading path
⎪ 3G max / qf 1 + 3G max (εs − εr 2) / 2qf
⎣ ⎦ (1)
reference ground motion (PBA = 0.06 g) adopted in the present study ⎩
is similar to that used by Banerjee et al. [8], which represents the type where q and εs are the current deviator stress and generalized shear
of shaking that may be experienced in Singapore due to a typical far- strain, respectively;
field earthquake arising from the strike-slip Great Sumatran Fault. In
order to study the effect of different earthquake intensity, the reference qr1 and qr2 are the respective deviator stresses at the reversal points;
ground motion was scaled down to two other different peak ground εr1 and εr2 are the respective generalized shear strains at the reversal
accelerations about 0.01 g and 0.03 g. In addition, some other influ- points;
encing factors such as pile flexural rigidity, mass of the raft and storey Gmax is the small-strain shear modulus;
number of the superstructure were varied in the numerical simulations. qf is the deviator stress at failure.
The computed results of pile bending moment, deflection of the su-
perstructure, inter-storey drift ratio and shear force of the column of For normally consolidated kaolin clay, the small-strain shear

0.2 Fig. 1. Input base motions adopted in this study: (a)

PBA = 0.01 g, 0.03 g and 0.06 g, time history, (b) response spectrum for the base
Spectral acceleration (g)

respectively
Acceleration (g)

motion with peak base acceleration (PBA) of 0.06 g

0.15 (5% damping).

0 0.1

0.05
PBA
0
0 5 10 15 20 25 0 2 4 6 8 10
Time (s)
Period (s)
(a) (b)

210
L. Zhang, H. Liu Soil Dynamics and Earthquake Engineering 101 (2017) 209–224

6m Fig. 2. Elevation and plan views of the soil-pile-raft-superstructure

Storey n systems adopted in the three-dimensional (3D) finite element analyses
3m (n =0, 4, 10, 15 and 20).
Storey (n-1)
0.1-m thick slab
Storey 3

Storey 2
Column
Storey 1

1.2 m Raft

Pile

21.7 m 20 m Pure clay

1m

Elevation View
65 m

Raft 7.5 m 4.5 m

4.5 m Pure clay

Pile
30 m 15 m 25 m

Section: 0.6 m (height)

3m
× 0.3 m (width)
3m
15 m
Plan View
Shaking direction

Table 1 Table 2
Properties of clay adopted in the numerical simulations. Properties of the other materials adopted in the numerical simulations.

Property Values/Range Property Values/Range

Poisson's ratio 0.3 Pile Young's modulus (GPa) 30; 70; 210
Density (kg/m3) 1600 Young's moduli of raft, slab, column and 30
Frictional angle of clay 25° girder (GPa)
Small-strain modulus (kPa) 2060(p0’)0.653 Young's modulus of sand (MPa) 100
Pile flexural rigidity (kNm2) 1472,622; 3436,117;
Note: p0’is the initial mean effective normal stress, with the unit of kPa. 10,308,351
Poisson's ratio of pile, raft, slab, column and 0.2
girder
modulus can be expressed by the following equation [7], modified from
Poisson's ratio of sand 0.49
Viggiani and Atkinson [49]: Density of pile, slab, column and girder (kg/ 2500
m3)
Gmax = 2060(p0′ )0.653 (2) Density of raft for "detailed model" model 2500
(kg/m3)
Density of raft for "added mass" model (kg/ 2500–9960
where p'0 is the initial mean effective normal stress, and the units for
m3)
both Gmax and p'0 are kPa. Density of sand (kg/m3) 1600
Besides, the degradation of the backbone curve under repeated Frictional angle of sand 35°
loading was also taken into account in this soil model. The interested
readers may refer to Banerjee [7] for more detailed information on this
hyperbolic-hysteretic soil model. 2.3. Modelling of pile-raft system
In this study, the hyperbolic-hysteresis model was recoded into a
VUMAT subroutine for the 3D finite element analyses using the explicit The piles and raft were modelled as linear elastic materials, the
time integration scheme, which yields significant savings of the com- properties of which are listed in Table 2. The piles were extended
putational time [51,52]. through the overlying raft, with the same nodes shared at the interface
between the piles and raft. Three different piles with flexural rigidities
ranging from 1,472,622 to 10,308,351 kNm2 were used; according to
Singapore Standard CP4 [44], the bearing capacity of the 5×5 pile-raft

211
L. Zhang, H. Liu
C
B
Fig. 3. 3D finite element model of 20-storey superstructure supported by a 5 × 5 pile-raft
system (19200 linear hexahedral elements of type C3D8R, 5760 linear quadrilateral
elements of type S4R, and 3225 linear line elements of type B31).
Fig. 4. The shear stress-strain relationship for the hyperbolic-hysteretic soil constitutive
model (q and εs are the current deviator stress and generalized shear strain, respectively;
Gmax is the small-strain shear modulus; qf is the deviator stress at failure; R= Gmax/ qf.).
212
Soil Dynamics and Earthquake Engineering 101 (2017) 209–224
system is more than 31,000 kN based on the undrained shear strength
profile of the clay measured by T-bar testing during the centrifuge test
[53], which can sufficiently support the superstructures considered in
this study. Rayleigh damping coefficients are applied to the pile and raft
materials to yield a damping ratio of 5% to the two characteristic fre-
quencies corresponding to the two peaks in the acceleration response
spectrum plot of the input ground motion shown in Fig. 1(b). Following
the study of Banerjee et al. [8], this study assumed a perfect bonding
between the pile and soil under the seismic shaking events. Hence the
pile and soil elements were modelled as sharing the same nodes at the
interface between the pile circumference and the soil. This method has
also been successfully applied in many previous numerical pile-soil
interaction analyses, such as those reported by Alsaleh and Shahrour
[2], Rovithis et al. [42] and Dai and Roesset [20]. In this study, to
combine the merits of both solid and beam elements, a "hybrid" mod-
elling method (e.g. [8]) in which the relatively flexible beam elements
are embedded in the surrounding solid pile elements to jointly model
the pile; the surrounding solid pile elements share 90% of the flexural
rigidity and mass of the pile, and the left 10% of the flexural rigidity
and mass of the pile is assigned to the embedded beam elements. The
final pile bending moments are 10 times the computed bending mo-
ments obtained from the embedded beam elements. This approach of
modelling piles have been favourably examined by a series of centrifuge
tests on different pile-raft systems embedded in kaolin clay bed [51,52].
2.4. Modelling of superstructure
Two approaches, namely "added mass" and "detailed model", were
used to account for the inertial effect imposed by a superstructure
during the seismic shaking event in the present numerical study. The
term "added mass" originally referred to affixing masses on the top of a
raft to model the weight of a superstructure in centrifuge tests (e.g.
[8,51,52,53]), and then was extended to numerical simulations in
which the superstructure weight was accounted for by increasing the
density of raft (e.g. [8,51,52]). However, this approach may only per-
form well for comparatively stiff buildings where the response due to
higher modes is less significant. On the other hand, to more realistically
represent a superstructure, "detailed model" which implies the detailed
modelling of the superstructure (e.g. [27]) was primarily adopted in the
present study. In this paper, the focus is placed on the seismic response
of pile-raft-superstructure system built on soft clay stratum under small
to moderate levels of seismic shaking, and the superstructures were
assumed to respond within the linear elastic range; thereby the slabs,
columns and girders were modelled as linear elastic materials, the
properties of which are listed in Table 2. General geometrical properties
of the superstructure are provided in Fig. 2, and the section dimensions
of the columns for different superstructures are shown in Table 3. The
dimensions and elastic properties of the structural components were
adopted in accordance with Eurocodes 1 [13] and 2 [15], with the
consideration of both the permeant and imposed loads associated with
the superstructure; they are also consistent with the design practice in
Singapore as reported by Balendra et al. [5], Balendra et al. [6], and
Shao and Tiong [54]. Rayleigh damping coefficients are applied to the
superstructures to yield a damping ratio of 5% to the respective first-
and second-mode frequencies of the superstructure as listed in Table 4,
which were obtained from the modal analyses of the superstructures
under fixed-base condition.
2.5. Lateral boundary condition
In the numerical simulation of soil-structure systems subjected to
seismic base shaking, the lateral boundaries perpendicular to the
shaking direction should be treated in a manner so that the influence of
wave reflection from the lateral boundaries can be minimized and also
the analysis is computationally efficient. In this study, the boundary
condition in a laminar container commonly used in seismic shaking
L. Zhang, H. Liu Soil Dynamics and Earthquake Engineering 101 (2017) 209–224

Table 3
Section dimensions of the columns in the four superstructures.

Storey level Column section: depth (mm) × width (mm)

4 storeys 10 storeys 15 storeys 20 storeys

1 400 × 400 640 × 640 800 × 800 910 × 910

2 400 × 400 640 × 640 800 × 800 910 × 910
3 400 × 400 600 × 600 750 × 750 850 × 850
4 400 × 400 600 × 600 750 × 750 850 × 850
5 – 550 × 550 700 × 700 800 × 800
6 – 550 × 550 700 × 700 800 × 800
7 – 500 × 500 650 × 650 750 × 750
8 – 500 × 500 650 × 650 750 × 750
9 – 450 × 450 600 × 600 700 × 700
10 – 450 × 450 600 × 600 700 × 700
11 – – 550 × 550 650 × 650
12 – – 550 × 550 650 × 650
13 – – 500 × 500 600 × 600
14 – – 500 × 500 600 × 600
15 – – 450 × 450 550 × 550
16 – – – 550 × 550
17 – – – 500 × 500
18 – – – 500 × 500
19 – – – 450 × 450
20 – – – 450 × 450

Table 4
First- and second-mode frequencies of the fixed-base superstructures.

Superstructure First mode frequency (Hz) Second mode frequency (Hz)

4 storeys 2.28 7.02
10 storeys 1.17 3.42
15 storeys 0.80 2.32
20 storeys 0.60 1.70
table tests was replicated in the numerical simulations; the vertically
collocated nodes on the two lateral boundaries perpendicular to the
shaking direction were connected using rigid tie-rods so that the cor-
responding nodes on these two faces are constrained to undergo the
same motion in the direction of shaking. This approach was adopted in
many previous studies (e.g. [8,31,32]) to replicate the free-field ground
motion.
Fig. 5 shows the acceleration response at the top of the lateral
boundary for the pile-foundation model with 20 storeys subjected to the
0.06-g PBA ground motion, along with the corresponding free-field
ground acceleration responses; the locations of Points B and C are in-
dicated in Fig. 3, which represent ground surface nodes respectively
located at and far away from the lateral boundary. The free-field ac-
celeration response was computed at the clay surface from a pure clay
stratum model without embedded pile-raft-superstructure system. As
can be seen from Fig. 5, the free-field acceleration responses computed
from Points B and C are almost the same, which are also very
Fig. 6. 3D finite element model of the centrifuge test sample with 4 × 3 pile-raft system
(7200 linear hexahedral elements of type C3D8R, 328 linear line elements of type B31.).
comparable with that at the top of the lateral boundary for the pile-
foundation model with 20-storey superstructure; hence, a consistent
free-field acceleration response can be ensured with the lateral
boundaries connected using rigid tie-rods as adopted in this study.
Besides, several 3D finite element models with different dimensions
of the lateral boundary were used to investigation the influence of di-
mension of the lateral boundary on the raft acceleration and pile
bending moment responses [51], which indicate that the lateral
boundary with a dimension of 65 m is long enough to minimise the
effect of wave reflection induced by the truncated boundaries in nu-
merical modelling.
3. Numerical validation analysis
The test results from a series of centrifuge tests on both single piles
[8] and pile groups [51,53] embedded in soft clay stratum can be used
to examine the numerical modelling procedure adopted in this study.
Fig. 6 shows the 3D finite element half-model of a 4 × 3 aluminum pile-
raft system installed in soft kaolin bed, which follows the same

Fig. 5. Comparison of the acceleration response at

the top of lateral boundary for the pile-foundation
model with 20 storeys against the free-field ground
surface accelerations (EpIp = 1,472,622 kNm2 for
the pile-foundation model with 20 storeys, PBA =
0.06 g, locations of Points B and C are shown in
Fig. 3.): (a) acceleration time histories, (b) response
spectra.

213
L. Zhang, H. Liu
numerical modelling procedure adopted in this study. Fig. 7 shows the
measured and computed raft acceleration responses for the 4 × 3
aluminum pile-raft model with a structural mass of 597 t, subjected to
the 0.06-g PBA ground motion. As can be seen, the computed and
measured acceleration data compare well, with favourable agreement
observed in the peak values and the overall trend. Furthermore, Fig. 8
shows the generally favourable agreement between the computed and
measured maximum pile bending moment profiles for both the outer
and inner piles, in which the maximum pile bending moment profiles
are obtained by selecting the instantaneous bending moments along the
pile shaft corresponding to the instant when the uppermost strain gauge
reading attains its maximum value. It is noted that, due to the pile
group effect, the inertial forces transferred from the raft to the sup-
porting piles are uneven, which results in the discrepancy in the
bending moment response between the outer and inner piles.
Although not shown here, some more similarly favourable com-
parisons of the ground acceleration and pile bending moment response
between the centrifuge tests and numerical simulations can be found in
other studies [8,51,52]. In the subsequent sections, the same numerical
procedure is used to perform a total of 90 3D finite element analyses,
with the consideration of some varying factors including peak base
acceleration, pile flexural rigidity, and storey mass and number. The
results of the pile bending moment, superstructural deflection, inter-
storey drift ratio and column shear force are presented in the following
sections.
214
Soil Dynamics and Earthquake Engineering 101 (2017) 209–224
Fig. 7. Computed and measured acceleration re-
sponses at raft top for 4×3 pile group (structural
mass, mstr = 597 t; pile flexural rigidity, EpIp =
3,436,117 kNm2), subjected to the 0.06-g PBA
ground motion: (a) acceleration time histories, (b)
acceleration response spectra.
Fig. 8. Computed and measured maximum pile
bending moment profiles for 4×3 pile group (mstr =
597 t, EpIp = 3,436,117 kNm2), subjected to the
0.06-g PBA ground motion: (a) outer pile, (b) inner
pile.
4. Seismic response of pile bending moment
As indicated in Fig. 3, the outer pile termed "reference pile" was
selected to present the results of pile bending moment during the
seismic shaking events. However, as also mentioned in the previous
section, due to the pile group effect, the pile bending moment response
near pile head can be influenced by the pile location; more detailed
discussions on the difference in the pile bending moment response due
to the influence of pile location can be found in [51].
Fig. 9 shows the bending moment time history computed at the pile
head (EpIp = 3,436,117 kNm2) for pure raft model along with the
corresponding input base acceleration time history. As can be seen,
there is a time lag between the peak base acceleration and peak bending
moment, which is attributed to the seismic wave propagation through
the embedded piles and clay stratum. In the following subsections, the
results of the pile bending moment will be presented in the form of
envelop of pile bending moment along the pile shaft and the maximum
bending moment experienced at pile head. In addition, some compar-
isons of the computed maximum shear force distribution along pile
shaft between the “added mass” and “detailed model” models are also
provided in Subsection 4.1.
4.1. Typical computed envelopes of the pile bending moment and shear force
Fig. 10 shows some of the computed envelopes of the pile bending
moment, which were obtained by selecting the maximum bending
moments experienced at each of the points (nodes) along the pile shaft
L. Zhang, H. Liu Soil Dynamics and Earthquake Engineering 101 (2017) 209–224

bending moment response refers to that associated with the super-

structure with detailed modelling unless otherwise stated.
The normalized maximum pile bending moment can be employed to
more conveniently investigate the influence of storey number, which is
defined as shown below.
Mmax
MN1 =
Mmax , 0 (3)
where Mmax and Mmax, 0 are the maximum pile bending moments for
pile foundation supporting an n-storey (n = 0, 4, 10, 15 and 20) su-
perstructure and a pure raft, respectively.
As Fig. 13 shows, from a total of 45 analyses associated with the
detailed superstructures, the normalized maximum pile bending mo-
ment and the storey number can be approximately represented by the
following expression.
Mmax
Fig. 9. Bending moment time history computed at the pile head (EpIp = 3,436,117 kNm2) = e 0.032n
for the numerical model with a pure raft versus the input base acceleration time history. Mmax , 0 (4)
where n is the storey number of a superstructure.
during the respective seismic shaking events. As can be seen, the piles Despite the scatter shown in Fig. 13, Eq. (4) can be employed to
experience the largest bending moments at pile head while the smallest estimate the influence of the storey number on the maximum pile
bending moments at pile tip, which is due to the fact that the piles are bending moment response with a maximum prediction error less than
effectively fixed at the heads by the overlying raft and nearly pin- 25%, which is still acceptable from a practical engineering standpoint.
connected at the tips with the underneath soils. Besides, the use of
"added mass" to model the superstructure tends to underestimate the 4.3. Influence of peak base acceleration
pile bending moment response compared with the approach using
"detailed model"; as Fig. 10(b) and (d) show, the underestimation ratios As Fig. 14 shows, the maximum pile bending moment can be sig-
of the maximum pile head bending moment involving "added mass" nificantly influenced by the peak base acceleration following a clear
approach are about 15% and 51% respectively for the 10- and 20-stroey increasing trend. Similarly, the maximum pile bending moments are
buildings, albeit Fig. 10(a) also suggests that these two approaches lead normalized to study the effect of PBA, following the expression shown
to very similar pile bending moment response for the superstructure below.
with 4 storeys. Similarly, as shown in Fig. 11 of the computed envelopes
Mmax
of the shear force along pile shaft, while these two approaches lead to MN 2 = 0.032n
quite comparable maximum shear force distributions for the 4-storey Mmax , 0.01 × e (5)
building, the use of "added mass" to model the superstructure generally where Mmax, 0.01 is the maximum pile bending moment corresponding
tends to underestimate the shear force response especially at the pile to the PBA of 0.01 g.
head. In addition, as shown in Fig. 11, the maximum shear force is not As Fig. 15 shows, the normalized maximum pile bending moment
necessarily experienced at the pile head, but can also be experienced at can be well correlated with the PBA following a linear line, which is
a depth between 5 and 8 m 5–8 times pile diameter). This trait is dif- expressed as:
ferent from that of the pile bending moment response plot shown in
Mmax
Fig. 10 and indicates the importance of considering the kinematic effect = 77.641 × PBA
Mmax , 0.032n
arising from the surrounding soils. 0.01 × e (6)
As can be inferred from Fig. 15, Eq. (6) can be used to estimate the
4.2. Influence of storey number effect of PBA on the maximum pile bending moment response, with a
prediction error ranging from −23% to 20%.
Fig. 12 shows the influence of storey number of superstructure on
the maximum bending moment computed at pile head, in which storey 4.4. Influence of pile flexural rigidity
number "0" denotes a pure raft without overlying superstructure. As can
be seen, the maximum bending moment computed at pile head can be Fig. 16 plots the maximum pile bending moment against the pile
significantly influenced by the storey number and peak base accelera- flexural rigidity for models with different superstructures subjected to
tion with an overall increasing trend, regardless of approaches used to the 0.06-g PBA base motion, which suggests an overall increasing trend,
model the superstructure. However, it is also noted that the increasing regardless of the storey number. Using the same data shown in Fig. 16,
trend becomes much less evident when comparing the maximum pile Fig. 17 plots the maximum pile curvature against the stiffness-to-mass
bending moments between the 15- and 20-storey superstructures ratio, which can be reasonably fitted using a power law shown below.
modelled using “detailed model”. Besides, similar to that observed in
−0.706
Fig. 10, using "added mass" to model a superstructure tends to under- Mmax ΣEp Ip ⎤
estimate the maximum pile bending moment response in comparison = 0.0012 ⎡
⎢ m l3 ⎥
Ep Ip ⎣ str p ⎦ (7)
with the approach using "detailed model". However, these two methods
lead to very similar pile bending moment response for the super- where EpIp is the pile flexural rigidity, with unit kPa;
structure with storey number of 4. This finding suggests that the use of
"added mass” to account for the inertial effect of superstructure may be mstr is the mass of the raft and the superstructure, with unit tonne;
only valid for relatively stiff or low-rise superstructures for which the lp is the pile length, with unit m;
effect of higher modes of the superstructure is comparatively less sig- Mmax/(EpIp) is the maximum pile curvature, with unit m−1;
nificant; to more realistically account for the inertial effect imposed by ∑EpIp/(mstrlp3) is the stiffness-to-mass ratio, with unit s−2.
a high-rise superstructure, a detailed model of the superstructure
should be used in the numerical simulations. Hereafter, seismic pile The maximum curvature of pile decreases sharply for small stiffness-

215
L. Zhang, H. Liu
to-mass ratios up to about 5 and remains relatively unchanged for large
stiffness-to-mass ratios greater than about 20, which is consistent with
the findings observed from the centrifuge tests performed on different
pile-raft systems installed in soft kaolin clay bed [53].
5. Seismic response of the superstructure
Fig. 18 shows the typical comparisons of time history and response
spectrum between raft acceleration and the corresponding input base
acceleration. As can be seen, the raft accelerations are generally dif-
ferent from the free-field accelerations plotted in Fig. 5, and they are
significantly amplified due to the presence of soft clay deposit, which is
216
Soil Dynamics and Earthquake Engineering 101 (2017) 209–224
Fig. 10. Computed envelopes of pile bending mo-
ment for numerical models using "added mass" and
"detailed model" to simulate the inertial effect of
superstructure (EpIp = 1,472,622 kNm2, PBA =
0.06 g): (a) 4 storeys, (b) 10 storeys, (c) 15 storeys,
(d) 20 storeys.
consistent with the findings of many previous studies performed on soft
clays [8,34,39,47]; the response spectra of the raft accelerations un-
derlying different superstructures may have different magnitudes, but
they have almost the same two characteristic frequencies of about
0.68 Hz and 1.1 Hz and mainly depend on the input ground motion as
well as the dynamic characteristics of clay deposit. As a result, the su-
perstructures supported by pile-raft systems installed in soft clay de-
posit likely also experience amplified deformation and load during the
seismic shaking events.
In the following subsections, some results of the maximum lateral
deflection, maximum inter-storey drift ratio and maximum column
shear force of the superstructure are to be presented, which were
L. Zhang, H. Liu
calculated based on the numerical models with detailed modelling of
the superstructure. The maximum lateral deflection of the super-
structure and maximum inter-storey drift ratio are respectively defined
by the following two expressions.
Defi = max Ui (t ) − U0 (t )
Dri i = max Ui (t ) − Ui − 1 (t ) / h
wherein Defi, Drii are the maximum lateral deflection and maximum
inter-storey drift ratio at the ith level, respectively;
i is the storey level, ranging from 1 to 20;
t is time, ranging from 0 to 25 s;
Ui(t) and U0(t) are the lateral displacements of the ith floor and raft
(building base) at time t, respectively;
Soil Dynamics and Earthquake Engineering 101 (2017) 209–224
Fig. 11. Computed envelopes of pile shear force for
numerical models using "added mass" and "detailed
model" to simulate the inertial effect of super-
structure (EpIp = 1,472,622 kNm2, PBA = 0.06 g):
(a) 4 storeys, (b) 10 storeys, (c) 15 storeys, (d) 20
storeys.
h is the height of the storey.
Similar to the pile bending moment response, the columns at the
same storey level usually do not evenly share the inertial force arising
from the above storeys; the inner columns C2 and C5 shown in Fig. 3
(8)
tend to experience comparatively larger shear force responses than the
(9) outer columns C1, C3, C4 and C6 shown in the same figure. In this
study, the averaged column shear force of columns C1 to C6 is pre-
sented rather than that of any individual column, which can better
reflect the overall trend of shear force demand for the superstructures.
5.1. Influence of soil-structure interaction
To investigate the influence of soil-structure interaction (SSI) on the
seismic response of superstructure, two numerical models namely
217
L. Zhang, H. Liu
Fig. 12. Influence of storey number of superstructure on the maximum bending moments
computed at the pile head for analyses involving both "added mass" and "detailed model"
(EpIp = 1,472,622 kNm2).
Fig. 13. Normalized maximum bending moment at the pile head against the storey
number of superstructure for analyses involving "detailed model" (totally 45 analyses).
Fig. 14. Influence of PBA on the maximum bending moment computed at the pile head
(EpIp = 1,472,622 kNm2).
"fixed-base" and "pile-foundation" models which respectively represent
the superstructure model with fixed base and full model of the clay-pile-
raft-superstructure system were used. Time domain analyses were
performed for the pile-foundation models, with input base acceleration
time history shown in Fig. 1; both time domain and modal response
spectrum analyses (recommended in most seismic design codes for
building such as Eurocode 8) were performed for the fixed-base models,
with the free-filed ground acceleration time history and response
spectrum (shown in Fig. 5) applied at the tips of the first-level columns
of the superstructure. Fig. 19 shows the distributions of maximum lat-
eral deflection, maximum inter-storey drift ratio and maximum column
shear force for the 4- and 20-storey superstructures computed from
218
Soil Dynamics and Earthquake Engineering 101 (2017) 209–224
Fig. 15. Normalized maximum pile bending moment at the pile head versus PBA (totally
45 analyses).
Fig. 16. Influence of pile flexural rigidity on the maximum bending moment at pile head
(PBA = 0.06 g).
Fig. 17. Maximum pile curvature plotted against stiffness-to-mass ratio of the pile-raft-
superstructure system, for different pile types (PBA = 0.06 g).
both the fixed-base and pile-foundation models. As can be seen, for
fixed-base models, the time domain and modal response spectrum
analyses generally leads to more comparable results of the super-
structure response, which significantly differ from those computed from
the pile-foundation models. For the 4-storey superstructure, the SSI is
detrimental as the time domain analysis performed on the fixed-base
model tend to underestimate the response of superstructural deflection
and column shear force, with the maximum underestimation ratios
respectively being about 51% and 31%. On the other hand, without
consideration of SSI, the fixed-base model leads to too conservative
results for the 20-storey superstructure, the time domain analysis per-
formed on which overpredicts the maximum superstructural deflection
and maximum column shear force by up to 65% and 126%, respec-
tively. Besides, as Fig. 19(a)–(d) show, the fixed-base models also tend
to underestimate the maximum superstructural deflection and max-
imum inter-storey drift ratio at the first storey, regardless of the storey
number, which is likely due to that base rotation of the superstructure is
fully constrained for the fixed-base models. The difference in the effect
L. Zhang, H. Liu
of SSI between the 4- and 20-storey superstructures can be partly ex-
plained by comparing the first-mode frequencies of these two super-
structures (respectively 2.28 Hz and 0.60 Hz as listed in Table 4) with
the dominant frequency of the raft/free-field motion of about
0.66–0.68 Hz as shown in Figs. 5 and 18; the SSI tends to reduce the
fundamental frequency of the superstructure, which results in the
219
Soil Dynamics and Earthquake Engineering 101 (2017) 209–224
Fig. 18. Typical comparison of the raft accelerations
with the input base acceleration (EpIp =
1,472,622 kNm2, PBA = 0.06 g): (a) acceleration
time histories, (b) acceleration response spectra.
Fig. 19. Seismic response of superstructure with and
without consideration of soil-structure interaction
(SSI) (EpIp = 1,472,622 kNm2, PBA = 0.06 g, TD:
time domain analysis, MRS: modal response spec-
trum analysis): (a) maximum lateral deflection
(MLD) for 4 storeys, (b) MLD for 20 storeys, (c)
maximum inter-storey drift ratio (MISDR) for 4
storeys, (d) MISDR for 20 storeys, (e) maximum
column shear force (MCSF) for 4 storeys, (f) MCSF
for 20 storeys.
fundamental frequencies of the 4- and 20-storey superstructures being
respectively close to and departing from the dominant frequency of the
raft motion. In general, decreasing difference between the first-mode
frequency of a superstructure and the dominant frequency of the base
motion leads to more intense seismic response, and vice versa.
The foregoing discussions and comparisons suggest that SSI has a
L. Zhang, H. Liu
significant influence on the seismic response of superstructure, re-
gardless of the detrimental or beneficial effects. Hence, SSI should be
taken into consideration when studying the seismic response of a su-
perstructure, which was also proposed by many other researchers (e.g.
[21,26,28,36,38]). Hereafter, the seismic response of a superstructure
refers to that with the consideration of SSI unless otherwise stated.
5.2. Influence of peak base acceleration
As Fig. 20 shows, the distribution profiles of the maximum lateral
deflection, maximum inter-storey drift ratio and maximum column
shear force for the 4- and 15-storey superstructures increase sig-
nificantly with the increasing peak base acceleration. Besides, Fig. 20(f)
suggests that the maximum shear force of the 15-storey superstructure
may not necessarily occur at the first-level columns as that for the 4-
storey superstructure as shown in Fig. 20(e), but can also be experi-
enced by the second-level columns, depending on the levels of PBA of
the motions applied.
The column shear forces mainly arise from the inertial forces ex-
erted by the floors during the seismic shaking events, the averaged
value of which can be expressed by the following equation:
220
Soil Dynamics and Earthquake Engineering 101 (2017) 209–224
Fig. 20. Influence of PBA on the seismic response of
superstructure (EpIp = 1,472,622 kNm2): (a) MLD
for 4 storeys, (b) MLD for 15 storeys, (c) MISDR for 4
storeys, (d) MISDR for 15 storeys, (e) MCSF for 4
storeys, (f) MCSF for 15 storeys.
1 ⎛
n
⎞
Fi (t ) = abs ∑ [m i × Ai (t )]
N ⎜ ⎟
⎝ j=i ⎠ (10)
where i is the storey level;
t is time, ranging from 0 s to 25 s in this study;
Fi(t) is the averaged shear force for columns below the ith storey at
time t;
N is the total number of columns below the ith storey ( N = 9 in this
study);
n is the total number of floors (n = 4, 10, 15 and 20 in this study);
mi is the mass of the ith floor;
Ai(t) is acceleration of the ith floor at time t.
Fig. 21 shows the acceleration distribution profiles of the 15-storey
superstructure when the first-storey columns experience the maximum
averaged shear force for the superstructure subjected to different PBA
ground motions. As can be seen, the first floor experiences the accel-
eration towards the opposite direction in comparison with the upper
floors associated with the PBA of 0.01 g and 0.03 g, while all the 15
floors experience the accelerations in the same direction for the PBA of
0.06 g. Hence, by use of Eq. (10), the discrepancy in the maximum
L. Zhang, H. Liu
Fig. 21. Computed accelerations at different floors and raft for the 15-storey super-
structure subjected to different base motions (EpIp = 1,472,622 kNm2, time ≈ 6.5 s).
column shear force distribution as shown in Fig. 20(f) among analyses
involving different PBA ground motions can be reasonably explained.
5.3. Influence of pile flexural rigidity
Fig. 22(a) and (b) show the maximum superstructural deflection
221
Soil Dynamics and Earthquake Engineering 101 (2017) 209–224
profiles of the 4- and 20-storey superstructures supported by pile
foundations with varying pile flexural rigidities, which suggest a con-
sistent decreasing trend of the maximum superstructural deflection
with the pile flexural rigidity for storey levels up to four; however, as
Fig. 22(b) shows, for the storey levels ranging from 5 to 20, the max-
imum lateral deflection of the 20-storey superstructure seems to be
relatively insensitive to the variation in pile flexural rigidity. Similar
findings are also found in the corresponding maximum inter-storey drift
ratio profiles shown in Fig. 22(c) and (d). Besides, as Fig. 22(d) shows,
for storey levels ranging from 5 to 15 of the 20-storey superstructure,
the maximum inter-storey drift ratio tends to increase with the in-
creasing pile flexural rigidity. Fig. 22(e) and (f) show the plots of the
maximum column shear force against pile flexural rigidity, which
suggest the opposite trends of the maximum column shear force re-
sponse with the increasing pile flexural rigidity between the 4- and 20-
storey superstructures. The discrepancy in the seismic response of su-
perstructure against the pile flexural rigidity is largely dependent on its
fundamental frequency and the dominant frequency of the building
base motion. As can be seen from Table 4, the first-mode frequencies of
the 4- and 20-storey superstructures under fixed-base condition are
2.28 Hz and 0.6 Hz, respectively. The actual fundamental frequencies of
the 4- and 20-storey superstructures overlying pile foundation likely are
respectively larger and smaller than the dominant frequency of the
Fig. 22. Influence of pile flexural rigidity on the
seismic response of superstructure (PBA = 0.06 g):
(a) MLD for 4 storeys, (b) MLD for 20 storeys, (c)
MISDR for 4 storeys, (d) MISDR for 20 storeys, (e)
MCSF for 4 storeys, (f) MCSF for 20 storeys.
L. Zhang, H. Liu Soil Dynamics and Earthquake Engineering 101 (2017) 209–224

Fig. 23. Influence of storey number of superstructure

on the seismic response of superstructure (EpIp =
1,472,622 kN m2, PBA = 0.06 g): (a) MLD, (b)
MISDR, (c) MCSF.

building base motion of about 0.68 Hz as shown in Fig. 18. As the pile
flexural rigidity increases, the fundamental frequency increases; as a
result, the fundamental frequencies of the 4- and 20-storey super-
structures supported by pile foundation approach and depart from the
dominant frequency of the building base motion, respectively.
From the above discussions, the pile flexural rigidity can effectively
influence the seismic response of a superstructure. However, its influ-
ence on the seismic response of a superstructure is less significant as
compared to that on the pile bending moment response.
with a fundamental frequency of 0.8 Hz under the fixed-base condition.
It is largely due to SSI effect which can effectively reduce the funda-
mental frequency of the superstructure; as a result, the 15-storey su-
perstructure likely has an actual fundamental frequency comparatively
closer to the dominant frequency of the building base motion.
Hence, the influence of storey number on the seismic response of
superstructure is also frequency-dependent, which depends on the dy-
namic characteristics of the input ground motion as well as the soil-
foundation-superstructure system.

5.4. Influence of storey number

6. Discussions

As Fig. 23 shows, the maximum lateral deflection, maximum inter-

Section 4 presents the computed results of pile bending moment for
storey drift ratio and maximum column shear force of the super-
different pile-raft-superstructure systems constructed on soft clay sub-
structure can be significantly influenced by the storey number of su-
jected to far-field ground motions. The pile bending moment response
perstructure. As Fig. 23(a) shows, at the same storey level, the max-
was found to be significantly influenced by peak base acceleration, pile
imum lateral superstructural deflection tends to increase as the storey
flexural rigidity, and the inertial effect imposed by a superstructure,
number increases up to 15, and then becomes much smaller when the
with a general increasing trend. The approach using "added mass” to
storey number changes from 15 to 20. As Fig. 23(b) shows, the 15-
account for the inertial effect imposed by a superstructure may be only
storey superstructure experiences the largest maximum inter-storey
valid for low-rise superstructures; to more realistically account for the
drift ratio for storey levels below 12; for storey levels above 12, the 20-
inertial effect imposed by a relatively high-rise superstructure, detailed
storey superstructure experiences the largest maximum inter-storey
modelling of the superstructure should be adopted. Furthermore, Eqs.
drift ratio. Similar findings are also observed in Fig. 23(c) of the max-
(4), (6) and (7) were derived to favourably respectively represent the
imum column shear force profile, except that the transition is at the
effects of the three factors on the maximum pile bending moment re-
10th floor. Similarly, the discrepancy in the seismic response of the
sponse experienced at the pile head. It should be noted that, however,
superstructure among these four superstructures can be partly ex-
upon the derivation of Eqs. (4), (6) and (7), variations in some other
plained by comparing the fundamental frequencies of the super-
factors such as pile length and clay thickness were not taken into
structures with the dominant frequency of the base motion of the su-
consideration.
perstructure. As Table 4 shows, under the fixed-base condition, the
In the previous study [51], based on a series of centrifuge tests and
fundamental frequencies of the 4-, 10- and 15-storey superstructure
numerical parametric studies, a semi-empirical equation was derived to
change from 2.28 to 0.8 Hz, which gradually become closer to the
predict the maximum pile bending moment at the pile head for pile
dominant frequency (0.68 Hz as shown in Fig. 18) of the building base
foundations installed in soft clay deposits, which can be expressed as:
motion. Furthermore, the 20-storey superstructure under fixed-base
condition has a fundamental frequency of 0.6 Hz which is the nearest to
the dominant frequency of the building base motion; however, it gen-
erally has less intense seismic responses than 15-storey superstructure

222
L. Zhang, H. Liu
E I l p −0.595 ⎛ (∑ EP IP ) ρsoil Hsoil
2
Mmax , 0 = 6.68 × 10−3 ⎛ P P ⎞ ⎛ ⎞ ⎜ ⎟
⎜ 3
⎝ r r
⎠⎝ ⎠ ⎝ G m
soil str p l
−0.3
mstr ⎞
× PBA1.092 ⎛⎜ ⎟
⎝ ρsoil l p Araft ⎠
ρpile ⎞0.29 ⎛ Hsoil ⎞0.893
× M−0.032 ⎛⎜ ⎟ ⎜ ⎟
⎝ ρsoil ⎠ ⎝ l p ⎠
wherein Mmax, 0 is the predicted maximum pile bending moment for
pile foundation supporting a pure raft, same as that defined in Eqs. (3)
and (4);
Araft is the area of raft footprint;
Hsoil is the thickness of soil;
M is the slope of critical state line, which can be expressed as M =
6sinϕ/(3-sinϕ) (ϕ is the effective internal friction angle of soil);
ρpile and ρsoil are the densities of the pile and soil, respectively;
Gsoil is the average maximum soil shear modulus which can be re-
presented by the following equation:
Hsoil
∫0 Gmax dz
Gsoil =
Hsoil
where Gmax is the depth-dependent maximum or small-strain shear
modulus of the soil, which can be obtained from Eq. (2).
Apart from the three factors as those considered in this study, five
more factors namely pile length, soil stiffness, soil frictional angle, soil
thickness and pile density were considered in Eq. (11).
The seismic maximum pile bending moment for pile-raft-super-
structure systems constructed on soft clay can be approximately esti-
mated following the procedure proposed herein. Firstly, a maximum
pile bending moment for pile foundation supporting a pure raft can be
estimated using Eq. (11); secondly, the maximum pile bending moment
for the same pile foundation supporting a superstructure can be ap-
proximately estimated using Eq. (4); thirdly, Eqs. (6) and (7) can be
employed to make further adjustments to the predicted maximum pile
bending moment to account for more complex situations. The proce-
dure proposed herein could provide a quick and first-order prediction of
the seismic maximum pile bending moment prior to performing more
detailed and rigorous analyses, which likely can serve as a useful tool
for practical design of pile foundation against seismic shaking.
On the other hand, as Section 5 presents, the seismic response of a
superstructure is not so straightforward; similar correlations as those
derived for the seismic pile bending moment response were not ob-
tained at this time. The seismic response of a superstructure is highly
frequency-dependent, and the effect of seismic SSI is significant and can
be either beneficial or detrimental to the seismic response of a super-
structure; hence, SSI should be taken into consideration when studying
the seismic response of a structure supported by pile foundation. Fur-
thermore, the superstructures considered in this study are rather
simple, and the applicability of the findings drawn in this study may
need to be further examined when more complex superstructures with
significantly different dynamic characteristics are encountered.
7. Conclusions
As an extension to a previous study [51] in which the superstructure
was modelled by added mass and increasing the raft density respec-
tively in the centrifuge tests and numerical parametric studies, a series
of 3D finite element analyses incorporated with a hyperbolic-hysteretic
soil model were performed to investigate the seismic response of pile-
raft-superstructure systems constructed on soft clay stratum. Three
factors, namely peak base acceleration, configuration of superstructure,
and flexural rigidity of pile, were accounted for. It is observed that all
the three factors have significant influences on the pile bending mo-
ment response; the superstructural response can also be significantly
Soil Dynamics and Earthquake Engineering 101 (2017) 209–224
−0.509 influenced by the former two factors, while affected to a lesser extent by
⎞
⎟ the pile flexural rigidity. Besides, the coupled soil-foundation-super-
⎠ structure analysis with a detailed modelling of the superstructure is
highly recommended as the SSI and the configuration of a super-
structure can significantly influence the seismic responses of both the
foundation and superstructure.
This study provides a useful reference for the practical seismic de-
(11) sign of pile-raft-superstructure systems constructed on soft clay de-
posits. The findings drawn from the present study may be only valid for
pile-raft-superstructure systems subjected to small to moderate far-field
ground motions, where the pile-raft-superstructure systems can still be
modelled by linear elastic materials. Future studies can be performed to
further examine the seismic response of the pile-raft-superstructure
systems when strong ground motions, more sophisticated super-
structure and plastic deformation of the pile-raft-superstructure systems
are taken into consideration.
Acknowledgements
The first author is grateful to Associate Professor Siang Huat Goh,
Professor Fook Hou Lee and Assistant Professor Subhadeep Banerjee for
their invaluable guidance and help in centrifuge and numerical mod-
(12)
elling of soil-structure systems subjected to seismic base shakings. The
centrifuge tests were carried out with the support and assistance of the
technical staff at the Centre for Soft Ground Engineering, NUS.
References
[1] Abdoun T, Dobry R, O'Rourke TD, Goh SH. Pile response to lateral spreads: cen-
trifuge modeling. ASCE J Geotech Geoenviron Eng 2003;129(10):869–78.
[2] Alsaleh H, Shahrour I. Influence of plasticity on the seismic soil-micropiles-structure
interaction. Soil Dyn Earthq Eng 2009;29(3):574–8.
[3] ATC. Amended tentative provisions for the development of seismic regulations for
buildings. ATC Publication ATC 3-06, NBS Special Publication 510, NSF Publication
78-78, Washington, DC; 1978.
[4] Ayothiraman R, Matsagar VA, Kanaujia VK. Influence of superstructure flexibility
on seismic response pile foundation in sand. In: Proceedings of the 15th world
conference on earthquake engineering, 24–28 September 2012, Lisbon; 2012.
[5] Balendra T, Lam NTK, Perry MJ, Lumantarna E, Wilson JL. Simplified displacement
demand prediction of tall asymmetric buildings subjected to long-distance earth-
quakes. Eng Struct 2005;27(3):335–48.
[6] Balendra T, Suyanthi S, Tan KH, Ahmed A. Seismic capacity of typical high-rise
buildings in Singapore. Struct Des Tall Spec Build 2013;22(18):1404–21.
[7] Banerjee S. Centrifuge and numerical modeling of soft clay-pile-raft foundations
subjected to seismic shaking Ph.D. Thesis Singapore: National University of
Singapore; 2009.
[8] Banerjee S, Goh SH, Lee FH. Earthquake-induced bending moment in fixed-head
piles in soft clay. Géotechnique 2014;64(6):431–46.
[9] Bhattacharya S, Adhikari S, Alexander NA. A simplified method for unified buckling
and free vibration analysis of pile-supported structures in seismically liquefiable
soils. Soil Dyn Earthq Eng 2009;29(8):1220–35.
[10] Boulanger RW, Curras CJ, Kutter BL, Wilson DW, Abghari A. Seismic soil-pile-
structure interaction experiments and analyses. ASCE J Geotech Geoenviron Eng
1999;125(9):750–9.
[11] Brandenberg SJ, Boulanger RW, Kutter BL, Chang D. Behavior of pile foundations in
laterally spreading ground during centrifuge tests. ASCE J Geotech Geoenviron Eng
2005;131(11):1378–91.
[12] Brandenberg SJ, Zhao M, Boulanger RW, Wilson DW. p-y plasticity model for
nonlinear dynamic analysis of piles in liquefiable soil. ASCE J Geotech Geoenviron
Eng 2013;139(8):1262–74.
[13] BSI. Eurocode 1: actions on structures – Part 1-1: general actions – densities, self-
weight, imposed loads for buildings. London, UK: BSI; 2002.
[14] BSI. Eurocode 8: design of structures for earthquake resistance – Part 5: founda-
tions, retaining structures and geotechnical aspects. London, UK: BSI; 2004.
[15] BSI. Eurocode 2: design of concrete structures – Part 1-1: general rules and rules for
buildings. London, UK: BSI; 2004.
[16] BSSC. Recommended provisions for the development of seismic regulations for new
buildings and other structures. Washington, DC: Building Seismic Safety Council;
2000.
[17] Castelli F, Maugeri M. Postearthquake analysis of a piled foundation. ASCE J
Geotech Geoenviron Eng 2013;139(10):1822–7.
[18] Chau KT, Shen CY, Guo X. Nonlinear seismic soil-pile-structure interactions:
shaking table tests and FEM analyses. Soil Dyn Earthq Eng 2009;29(2):300–10.
[19] Curras CJ, Boulanger RW, Kutter BL, Wilson DW. Dynamic experiments and ana-
lyses of a pile-group-supported structure. ASCE J Geotech Geoenviron Eng
2001;127(7):585–96.
[20] Dai W, Roesset JM. Horizontal dynamic stiffness of pile groups: approximate
223
L. Zhang, H. Liu
expressions. Soil Dyn Earthq Eng 2010;30(9):844–50.
[21] De Sanctis L, Maiorano RMS, Aversa S. A method for assessing kinematic bending
moments at the pile head. Earthq Eng Struct Dyn 2010;39(10):1133–54.
[22] Dobry R, Abdoun T, O'Rourke TD, Goh SH. Single piles in lateral spreads: field
bending moment evaluation. ASCE J Geotech Geoenviron Eng
2003;129(10):879–89.
[23] Dungca JR, Kuwano J, Takahashi A, Saruwatari T, Izawa J, Suzuki H, Tokimatsu K.
Shaking table tests on the lateral response of a pile buried in liquefied sand. Soil
Dyn Earthq Eng 2006;26(2–4):287–95.
[24] Durante MG, Di Sarno L, Mylonakis G, Taylor CA, Simonelli AL. Soil-pile-structure
interaction: experimental outcomes from shaking table tests. Earthq Eng Struct Dyn
2015;45(7):1041–61.
[25] Gohl WB. Response of pile foundations to simulated earthquake loading: experi-
mental and analytical results Ph.D. Thesis University of British Columbia; 1991.
[26] Hamada M, Towhata I, Yasuda S, Isoyama R. Study on permanent ground dis-
placement induced by seismic liquefaction. Comput Geotech 1987;4(4):97–220.
[27] Hokmabadi AS, Fatahi B, Samali B. Assessment of soil-pile-structure interaction
influencing seismic response of mid-rise buildings sitting on floating pile founda-
tions. Comput Geotech 2014;55:172–86.
[28] Jeremić B, Jie G, Preisig M, Tafazzoli N. Time domain simulation of soil-foundation-
structure interaction in non-uniform soils. Earthq Eng Struct Dyn
2009;38(5):699–718.
[29] Klar A, Baker R, Frydman S. Seismic soil-pile interaction in liquefiable soil. Soil Dyn
Earthq Eng 2004;24(8):551–64.
[30] Lombardi D, Bhattacharya S. Modal analysis of pile-supported structures during
seismic liquefaction. Earthq Eng Struct Dyn 2014;43:119–38.
[31] Lu JC. Parallel finite element modeling of earthquake ground response and lique-
faction Ph.D. Thesis San Diego: University of California; 2006.
[32] Ma K, Banerjee S, Lee FH, Xie HP. Dynamic soil-pile-raft interaction in normally
consolidated soft clay during earthquakes. J Earthq Tsunami 2012;06(03).
1250031_1-1250031_12.
[33] Maheshwari BK, Truman KZ, El Naggar MH, Gould PL. Three-dimensional nonlinear
analysis for seismic soil-pile-structure interaction. Soil Dyn Earthq Eng
2004;24(4):343–56.
[34] Mayoral JM, Alberto Y, Mendoza MJ, Romo MP. Seismic response of an urban
bridge-support system in soft clay. Soil Dyn Earthq Eng 2009;29(5):925–38.
[35] Meymand PJ. Shaking table scale model test of nonlinear soil-pile-superstructure
interaction in soft clay Ph.D. Thesis Berkeley: University of California; 1998.
[36] Mylonakis G, Gazetas G. Seismic soil-structure interaction: beneficial or detri-
mental? J Earthq Eng 2000;4(3):277–301.
[37] NSPRC. GB50011-2010: code for seismic design of buildings. Beijing, China:
National standard of the People's Republic of China; 2010.
224
Soil Dynamics and Earthquake Engineering 101 (2017) 209–224
[38] Nikolaou S, Mylonakis G, Gazetas G, Tazoh T. Kinematic pile bending during
earthquakes: analysis and field measurements. Géotechnique 2001;51(5):425–40.
[39] Pan TC. Site-dependent building response in Singapore to long-distance Sumatra
earthquakes. Earthq Spectra 1997;13(3):475–88.
[40] Pan TC, Megawati K, Goh KS. Response of high-rise buildings in Singapore due to a
potential giant earthquake in the sumatran megathrust. J Earthq Eng
2011;15(Suppl 1):90–106.
[41] Renzi S, Madiai C, Vannucchi G. A simplified empirical method for assessing seismic
soil-structure interaction effects on ordinary shear-type buildings. Soil Dyn Earthq
Eng 2013;55:100–7.
[42] Rovithis EN, Pitilakis KD, Mylonakis GE. Seismic analysis of coupled soil-pile-
structure systems leading to the definition of a pseudo-natural SSI frequency. Soil
Dyn Earthq Eng 2009;29(6):1005–15.
[43] Sarma SK, Srbulov M. A simplified method for prediction of kinematic soil-foun-
dation interaction effects on peak horizontal acceleration of a rigid foundation.
Earthq Eng Struct Dyn 1996;25(8):815–36.
[44] SSC. CP4: code of practice for foundations. Singapore: Singapore Standards Council;
2003.
[45] Stewart JP. Variations between foundation-level and free-field earthquake ground
motions. Earthq Spectra 2000;16(2):511–32.
[46] Tabatabaiefar HR, Fatahi B, Samali B. Seismic behavior of building frames con-
sidering dynamic soil-structure interaction. ASCE Int J Geomech
2013;13(4):409–20.
[47] Tinawi R, Sarrazin M, Filiatrault A. Influence of soft clays on the response spectra
for structures in eastern Canada. Soil Dyn Earthq Eng 1993;12(8):469–77.
[48] Tokimatsu K, Suzuki H, Sato M. Effects of inertial and kinematic interaction on
seismic behavior of pile with embedded foundation. Soil Dyn Earthq Eng
2005;25(7–10):753–62.
[49] Viggiani G, Atkinson JH. Stiffness of fine-grained soil at very small strains.
Géotechnique 1995;45(2):249–65.
[50] Wilson DW, Boulanger RW, Kutter BL. Observed seismic lateral resistance of li-
quefying sand. ASCE J Geotech Geoenviron Eng 2000;126(10):898–906.
[51] Zhang L. Centrifuge and numerical modelling of the seismic response of pile groups
in soft clays Ph.D. Thesis Singapore: National University of Singapore; 2014.
[52] Zhang L, Goh SH, Liu H. Seismic response of pile-raft-clay system subjected to a
long-duration earthquake: centrifuge test and finite element analysis. Soil Dyn
Earthq Eng 2017;92:488–502.
[53] Zhang L, Goh SH, Yi J. A centrifuge study of the seismic response of pile–raft sys-
tems embedded in soft clay. Géotechnique 2017;67(6):479–90.
[54] Shao Z, Tiong RLK. Estimating potential economic loss from damage to RC buildings
in singapore due to far-field seismic risk. J Earthq Eng 2015;19(5):770–83.