Computers and Geotechnics 55 (2014) 172–186

Contents lists available at ScienceDirect

Computers and Geotechnics

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

Assessment of soil–pile–structure interaction influencing seismic

response of mid-rise buildings sitting on floating pile foundations
Aslan S. Hokmabadi ⇑, Behzad Fatahi, Bijan Samali
Centre for Built Infrastructure Research (CBIR), School of Civil and Environmental Engineering, University of Technology Sydney (UTS), P.O. Box 123, Sydney, Australia

a r t i c l e i n f o a b s t r a c t

Article history: The role of the seismic soil–pile–structure interaction (SSPSI) is usually considered beneficial to the struc-
Received 22 April 2013 tural system under seismic loading since it lengthens the lateral fundamental period and leads to higher
Received in revised form 10 July 2013 damping of the system in comparison with the fixed-base assumption. Lessons learned from recent
Accepted 29 August 2013
earthquakes show that fixed-base assumption could be misleading, and neglecting the influence of SSPSI
Available online 23 September 2013
could lead to unsafe design particularly for structures founded on soft soils. In this study, in order to bet-
ter understand the SSPSI phenomena, a series of shaking table tests have been conducted for three differ-
Keywords:
ent cases, namely: (i) fixed-base structure representing the situation excluding the soil–structure
Soil–pile–structure interaction
Seismic response
interaction; (ii) structure supported by shallow foundation on soft soil; and (iii) structure supported
Shaking table test by floating (frictional) pile foundation in soft soil. A laminar soil container has been designed and con-
FLAC3D structed to simulate the free field soil response by minimising boundary effects during shaking table
Laminar soil container tests. In addition, a fully nonlinear three dimensional numerical model employing FLAC3D has been
Frictional piles adopted to perform time-history analysis on the mentioned three cases. The numerical model adopts hys-
Floating piles teretic damping algorithm representing the variation of the shear modulus and damping ratio of the soil
with the cyclic shear strain capturing the energy absorbing characteristics of the soil. Results are pre-
sented in terms of the structural response parameters most significant for the damage such as foundation
rocking, base shear, floor deformation, and inter-storey drifts. Comparison of the numerical predictions
and the experimental data shows a good agreement confirming the reliability of the numerical model.
Both experimental and numerical results indicate that soil–structure interaction amplifies the lateral
deflections and inter-storey drifts of the structures supported by floating pile foundations in comparison
to the fixed base structures. However, the floating pile foundations contribute to the reduction in the lat-
eral displacements in comparison to the shallow foundation case, due to the reduced rocking
components.
Ó 2013 Elsevier Ltd. All rights reserved.

1. Introduction
The seismic design of buildings has been undergoing a critical
reappraisal in recent years, with change of emphasis from
‘‘strength’’ to ‘‘performance’’. The development of capacity design
principles in the 1970s [1] was an expression of the realisation that
the distribution of strength through a building was more important
than the absolute value of the design base shear which can be iden-
tified as the key point in the performance-based seismic design [2],
where the overall performance of the building is controlled during
the seismic design process.
For determining the seismic response of structures, it is a com-
mon practice to assume the structure is fixed at the base. In fact, if
the ground is stiff enough (e.g. structure founded on solid rock) it is
reasonable to assume that the input motion of the structure due to
a design earthquake is essentially identical to the motion of the
free field, which is defined as the motion experienced at the same
point before the structure is built. However, for structures con-
structed on soft soils, two modifications need to be considered
for determining the seismic response. First, the imposed motion
to the structure differs from the free field motion due to the pres-
ence of the structure. Secondly, additional dynamic deformations
are induced within the structure due to the underneath soft soil.
The process, in which response of the soil influences the motion
of the structure and response of the structure influences the mo-
tion of the soil is referred to as soil–structure interaction [3].
The dynamic equation of the motion for the structure (Fig. 1)
can be written as:

_ þ ½Kfug ¼ ½M1ug þ F v
½Mfug þ ½Cfug ð1Þ
⇑ Corresponding author. Tel.: +61 411603532; fax: +61 295142633.
E-mail addresses: aslan.sadeghihokmabadi@student.uts.edu.au (A.S. Hokmaba- where [M], [C] and [K] are the mass, damping, and stiffness matrices
di), Behzad.fatahi@uts.edu.au (B. Fatahi), Bijan.samali@uts.edu.au (B. Samali). _ and {ü} are the relative nodal
of the structure, respectively. {u}, {u},

0266-352X/$ - see front matter Ó 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.compgeo.2013.08.011
 A.S. Hokmabadi et al. / Computers and Geotechnics 55 (2014) 172–186
(a)
Fig. 1. Schematic modelling of the multi degree freedom structure considering: (a) structure supported by floating pile foundation employing foundation springs; (b) lateral
deformation and rocking of the structure supported by floating pile foundation; (c) lateral deformation of the fixed-base structure.
displacements, velocities and accelerations of the structure with re-
spect to ground, respectively. {üg} is ground acceleration, and {Fv} is
the force vector corresponding to the viscous boundaries. This vec-
tor is nonzero only when there is a difference between the motion
on the near side of the artificial boundary and the motion in the free
field [4]. The role of the seismic soil–pile–structure interaction
(SSPSI) is usually considered beneficial to the structural system un-
der seismic loading since it elongates the period of the structure and
increases the damping of the structural system, so the consideration
of SSPSI tends to reduce the base shear and in turn structural de-
mand of the superstructure in comparison to the fixed-base condi-
tion. In contrast, as shown in Fig. 1, SSPSI may increase the overall
displacement of the superstructure in comparison to the fixed-base
condition due to translation and rotation of the foundation (e.g.
Guin and Banerjee [5]; Yingcai [6]). The rocking stiffness is devel-
oped due to the resistance of the piles to vertical movement [7],
as shown particularly in Fig. 1b. Ma et al. [8] showed that rocking
may be the most critical mode of vibration for a foundation because
of the very low hysteretic (material) damping, which will lead to
high motion amplitude when the excitation frequencies are near
the resonance state. The increase in the lateral deformation of the
building can change the performance level of the structure and is
especially important for tall, slender structures or for closely spaced
structures that can be subjected to pounding when relative dis-
placements become large [3]. Moreover, increase in the total defor-
mation of the structure and in turn secondary P  D effect
influences the total stability of the structure. The lessons learned
from post seismic observations of the past earthquakes such as
1985 Mexico City, 1994 Northridge, and 1995 Kobe earthquakes
provided sufficient reason to believe that the SSPSI effects should
be investigated with greater rigour and precision (e.g. Mendoza
and Romo [9]; Mizuno et al. [10]).
Pile foundations are usually employed to transmit foundation
loads through soil strata of low bearing capacity to deeper soil or
rock strata having a high bearing capacity and stiffness. End bear-
173
(c)
(b)
ing piles terminate in hard, relatively impenetrable materials such
as rock or very dense sand and gravel, while floating piles obtain a
greater part of their capacity by skin friction or adhesion and are
mostly employed in situations where the bedrock is deep. Determi-
nation of the pile foundation seismic response is a complex process
involving inertial interaction between the structure and the pile
foundation, kinematic interaction between piles and soils, and
the non-linear response of soils to strong earthquake motions [7].
However, simple methods such as Winkler computational model
are often used in engineering practice in which soil–pile interac-
tion is modelled using either linear or non-linear springs. The reli-
ability of these constitutive models has been questioned by many
due to the simplifying assumptions regularly used [11,12]. At first,
the applied earthquake motion in the time history analysis is de-
rived from the free field motion ignoring the presence of super-
structure and pile elements. Secondly, Winkler springs which
have been developed initially to model single pile–soil interaction,
are not directly applicable to simulate pile groups due to the over-
lapping displacement fields of piles affecting the individual pile
stiffness [13]. The limitations of Winkler methods and availability
of advanced computational tools lead the researcher to conduct
fully-nonlinear analysis to study the seismic response of pile foun-
dations. As mentioned by Chu [14], for systems with strong nonlin-
ear behaviour, coupled soil–pile–structure response analysis is
highly desirable which can explicitly express the relationship be-
tween the soil and the structural responses, especially when the
stiffness of the pile foundation significantly affect the overall soil
response.
Although a number of works dealing with the SSPSI effects on
the seismic response of structures are available in the literature,
most of them adopt simplified models (e.g. single degree of free-
dom system for superstructure or linear analysis) [15–19]. The
present research aims to study the effects of SSPSI on the seismic
response of the superstructure by employing the fully nonlinear
method in which main components of the interaction including
174 A.S. Hokmabadi et al. / Computers and Geotechnics 55 (2014) 172–186
subsoil, floating pile foundation, and superstructure, are modelled
simultaneously. For this purpose, a three-dimensional explicit fi-
nite-difference program, FLAC3D [20], has been used to numeri-
cally model and examine the influence of the soil–structure
interaction on the seismic response of a 15-storey moment resiting
building. Two types of foundations, shallow foundations and float-
ing pile foundations, have been considered. The proposed numeri-
cal soil–structure model has been verified and validated against
experimental shaking table results.
2. Shaking table experimental tests
Model tests in geotechnical engineering offers the advantage of
simulating complex systems under controlled conditions providing
the opportunity of better understanding the fundamental mecha-
nisms of these systems. Such tests are often used as calibration
benchmarks for numerical or analytical methods, or to make quan-
titative predictions of the prototype response [21]. In most previ-
ously conducted shaking table tests (e.g. Chau et al. [22];
Ishimura et al. [23]; Jakrapiyanun [24]; Pitilakis et al. [25]; Mey-
mand [26]) the superstructure is simplified as a single degree of
freedom oscillator in which the behaviour of the soil–structure sys-
tem may not be completely conforming to reality and the effect of
higher modes would not be captured. In the current model tests,
unlike the previous efforts, a multi-storey frame for the superstruc-
ture is adopted, representing the dynamic properties of the proto-
type structure such as natural frequency of the first and higher
modes, number of stories, and density. Moreover, an advanced
laminar soil container has been designed to simulate the free field
soil response by minimising boundary effects. Consequently, the
current shaking table tests which simulate two common types of
foundations (shallow foundation and floating pile foundation) on
the soft soil together with the fixed-base model (excluding the
soil–structure interaction) provide unique and valuable data to
investigate the influence of the soil–pile–structure interaction on
the seismic response of buildings.
The experimental model tests have been carried out utilising
the shaking table facilities located at the structures laboratory of
the University of Technology Sydney (UTS). The size of the shaking
table is 3 m  3 m, with maximum payload of 10 tonnes and over-
turning moment of 100 kN m. Furthermore, the shaking table can
apply maximum acceleration of ±2.5 g with testing frequency
range of 0.1–100 Hz.
2.1. Prototype characteristics and scaling factors
A fifteen-storey concrete moment resisting building frame with
the total height of 45 m and width of 12 m consisting of three
spans, representing the conventional types of mid-rise moment
resisting buildings, is selected for this study as shown in Fig. 2.
The spacing between the frames into the page is 4 m. Natural fre-
quency of the prototype building is 0.384 Hz and its total mass is
953 tonnes. The soil medium beneath the structure is a clayey soil
with the shear wave velocity of 200 m/s and density of 1470 kg/m3.
The horizontal distance of the soil lateral boundaries and bedrock
depth has been selected to be 60 m and 30 m, respectively. The
building is resting on a footing which is 1 m thick and 15 m wide.
For the pile foundation case (Fig. 2c), a 4  4 reinforced concrete
pile group with pile diameter and length of 1.2 m and 20 m,
respectively, and equal spacing of four time the diameter (4d) are
considered. The piles are closed-end and have rigid connection
with the pile cap representing typical floating pile foundations.
In order to achieve a reasonable scale model, a dynamic simili-
tude between the model and the prototype should be applied as
described in the literature (e.g. Harris and Sabnis [27]; Langhaar
[28]; Meymand [26]). Dynamic similitude governs a condition
where homologous parts of the model and prototype experience
homologous net forces. The scaling relations for the variables con-
tributing to the primary modes of system response are presented
in Table 1.
Adopting an appropriate geometric scaling factor (k) is one of
the important steps in scale modelling on shaking table. Although
small scale models could save cost, the precision of the results
could be substantially reduced. Considering the mentioned specifi-
cations of UTS shaking table, scaling factor of 1:30 provides the
largest achievable scale model with rational scales, maximum pay-
load, and overturning moment meeting the facility limitations.
Thus, geometric scaling factor (k) of 1:30 is adopted for experimen-
tal shaking table tests on the scale model in this study. According
to Table 1, apart from the geometric scaling which should be im-
posed to all the components, the required scaled natural frequency
for the structural model and the required scaled shear wave veloc-
ity and density of the soil mix should be 2.11 Hz, 36 m/s and
1470 kg/m3, respectively. Moreover, the required scaled natural
frequency of the soil mix inside the soil container needs to be
10 Hz which is used as a benchmark to design the laminar soil
container.
2.2. Model components of shaking table tests
The developed soil–structure model for shaking table tests pos-
sesses four main components including the model structure, the
model pile foundations, the laminar soil container, and the soil
mix together with the imposed shaking events. Details and charac-
teristics of these components are explained below.
2.2.1. Model structure
Employing geometric scaling factor of 1:30, height, length, and
width of the structural model are determined to be 1.50 m, 0.40 m,
and 0.40 m, respectively. In addition, according to the scaling rela-
tionship as shown in Table 1, the required natural frequency of the
structural model is 2.11 Hz. In addition, the density of the model
and prototype should be equal. Thus, the total mass of 106 kg for
the model structure is obtained.
In order to simulate the prototype structure more accurately on
the shaking table, the model structure has been designed employ-
ing SAP2000 [29] software considering the required characteristics
of the model structure. The 3D numerical model consists of fifteen
horizontal steel plates as the floors and four vertical steel plates as
the columns. Steel plate grade 250, according to Australian stan-
dards [30], with the minimum yield stress of 280 MPa and the min-
imum tensile strength of 410 MPa, has been adopted in the design.
The thickness of the steel plates has been determined in the design
process after several cycles of trial and error in order to fit the re-
quired natural frequency and mass of the model structure. The
finalised base plate is a 500  500  10 mm steel plate while the
floors consist of 400  400  5 mm plates and four
500  40  2 mm steel plates are used for the columns. The con-
nections between the columns and floors are provided using stain-
less steel metal screws with 2.5 mm diameter and 15 mm length.
After the numerical modelling and design, the structural model
was constructed in house. The completed structural model is
shown in Fig. 3.
2.2.2. Pile foundation
Similar to the model structure, the model pile should be sub-
jected to the competing scale model criteria. In order to achieve
a successful model pile design, the principal governing factors of
pile response such as slenderness ratio L/d, moment curvature rela-
tionship, flexural stiffness EI, relative soil/pile stiffness, yielding
behaviour/mechanism, and natural frequency of vibration should
 A.S. Hokmabadi et al. / Computers and Geotechnics 55 (2014) 172–186 175

(a)

(b) (c)
Fig. 2. (a) Prototype fixed-base structure; (b) prototype structure supported by shallow foundation; (c) prototype structure supported by floating (frictional) pile foundation.

Table 1 replicated at the model scale. Thus, by considering the geometric

Scaling relations in terms of geometric scaling factor (k). scaling factor (k) of 1:30 in this study, the model piles should have
Mass density 1 Acceleration 1 Length k a diameter of 40 mm with L/d ratio of 16.6.
Force k3 Shear wave velocity k1/2 Stress k The moment-curvature relation criterion represents pile re-
Stiffness k2 Time k1/2 Strain 1 sponse to the lateral loading which is a function of the flexural
Modulus k Frequency k-1/2 EI k5
rigidity and yielding behaviour. Since in the present study piles
are intended to respond in the elastic range (this assumption is
confirmed numerically), this criterion is achieved by scaling the
flexural rigidity (EI) of the piles according to Table 1 (k5, k = 1/30)
in addition to ensuring that the yielding point of the model pile
is equal to or greater than the scaled prototype. Furthermore, by
scaling the stiffness of the soil and pile consistently, the relative
soil/pile stiffness parameter will be satisfied inevitably. Therefore,
the soil–pile interaction should then be accurately reproduced in
the model.
Previous researchers (e.g. Bao et al. [31]; Chau et al. [22]; Tao
et al. [32]) have used different types of materials like aluminium
tubes, steel bars, and reinforced concrete to build a model pile.
Considering the selected scaling factor in this study (k = 1/30)
and in turn the required stiffness and yielding stress for the model
piles, a commercial Polyethylene high pressure pipe with Standard
Dimension Ratio (SDR) of 7.4 according to the Australian Standard
[33], is the selected candidate which falls in the range of acceptable
criteria with 5% deviation from the target value for EI. Moreover,
Polyethylene pipes can tolerate large deformation prior to the
yielding point without any brittle failure. Characteristics of the
model pile used in this study are summarised in Table 2.

2.2.3. Soil mix

A synthetic clay mixture was designed to provide soil medium
for the shaking table testing. Previous researchers (Meymand
[26]; Turan et al. [34]; and Moss et al., [35]) reported that a
reconstituted soil would not be able to satisfy the competing scale

Fig. 3. The completed model structure for shaking table tests. Table 2
Characteristics of the model pile built from polyethylene pressure pipe.

Outer diameter (mm) 40 Young’s modulus (MPa) 1.16E+3

be addressed [26]. By adopting the geometric similitude, the Wall thickness (mm) 5.5 Density (kg/m3) 955
overall pile slenderness and relative contact surface area would Cross-sectional area 5.78E+2 Poisson’s ratio 0.4
(mm2)
be preserved in the model. This also guarantees that pile group
Moment of inertia (mm4) 8.33E+4 Flexural yield stress (MPa) 32
relative spacing and consequent group interaction would be
176 A.S. Hokmabadi et al. / Computers and Geotechnics 55 (2014) 172–186
modelling criterion of shear wave velocity with enough bearing
capacity for the foundation in shaking table tests while synthetic
clay mix can provide adequate undrained shear strength to mobi-
lise the required bearing capacity underneath the structural model
meeting the scale modelling criterion of the shear wave velocity. It
should be noted that, without providing enough bearing capacity
for the structural model foundation, the underneath soil may expe-
rience failure or excessive settlements while testing process is
being undertaken. In order to find out the most appropriate mix
for the test program, three different mixes (A, B, and C) were pro-
duced and examined in the UTS soils laboratory. The proportion of
different mix components for the three mixes are summarised in
Table 3. Mix A, which is the closest mix to what proposed by Mey-
mand [26], has higher percentages of kaolinite and bentinite, lower
percentage of class F fly ash and lime, and the same percentage of
water content comparing to Mix B. Mix B and C have the same dry
component percentages, but the water content was increased by
20% in Mix C in comparison to Mix B in order to achieve better mix-
ibility and workability for the mix.
Each proposed mix was prepared three times to control repeat-
ability of the test and each time three cylindrical test specimens of
size D = 50 mm and h = 100 mm were taken for the bender element
test which was performed to measure the shear wave velocity of
the soil over the curing age. The elapsed time from specimen prep-
aration to testing is termed ‘‘curing age’’. To carry out bender ele-
ment tests, the soil specimens were placed between bender
elements as shown in Fig. 4a; and the shear wave velocity of each
soil specimen was obtained by measuring the time required for the
wave to travel between two bender elements using PC running
GDS bender element control software. The adopted system has a
data acquisition speed of 2 MSamples/s, 16 bit resolution of data
acquisition and the connection to the control box through USB link.
In this study, the propagated shear wave type has been sine waves
with amplitude of 10 V and a period of 1 s. Fig. 4b shows the sche-
matic graphical signal processing to measure the shear wave travel
time at the bender element test.
The extracted average shear wave velocities versus curing ages
for the three different mixes over the period of two weeks have
been illustrated in Fig. 5. As shown in Fig. 5, the examined soil
mixes gain stiffness, and consequently shear wave velocity in-
creases with the curing age as expected (e.g. Wartman, [36]; Rie-
mer, et al. [37]; Moss et al. [35]). However, only Mix C produces
the required shear wave velocity of 36 m/s for the soil model on
the second day of its curing age while the other two mixes are un-
able to produce such a low shear wave velocity as required. In
addition, in order to ensure that the undrained shear strength of
the proposed soil mix is adequate to satisfy the required founda-
tion bearing capacity underneath the structural model, Unconfined
Compression Tests were performed on three soil specimens in
accordance with AS5101.4-2008 [38]. Eventually, desired soil mix
consisting of 60% Q38 kaolinite clay, 20% Active-bond 23 Bentonite,
20% class F fly ash and lime, and 120% water (% of the dry mix) has
been adopted for the shaking table tests in this study. Table 4 sum-
marised the soil mix properties at the second day of its curing age.
Accordingly, soil density on the second day was determined to be
Table 3
Proportion of different components for the examined mixtures.
Mix Components Mix A (%) Mix B (%)
Q38 kaolinite clay 67.5 60
ActiveBond 23 bentonite 22.5 20
Class F fly ash + Lime 10 20
Watera 100 100
a
% of the dry mix.
1450 kg/m3 being almost equal to the prototype soil density
(1470 kg/m3) as required. Therefore, the designed soil mix
possesses the required dynamic similitude characteristics. It
should be noted that the prototype soil acquires the properties of
cemented soil that can be found in nature or treated soil.
2.2.4. Shaking events
The input ground motions in this study are represented by a set
of real earthquakes defined at the outcropping bedrock. Each test
model was subjected to two near field shaking events including
Kobe, 1995, Northridge, 1994, two far field earthquakes including
El Centro, 1940, and Hachinohe, 1968, and Sine Sweep test. It is
well known that the intensity of shaking decreases as the distance
increases from the seismic fault where the earthquake shaking is
generated [39]. In addition, high frequency components lose en-
ergy more quickly than low frequency components while travelling
through the ground. As a result, near field earthquakes generate
higher ground peak acceleration and frequency component in com-
parison with the far filed earthquakes. The characteristics of the
mentioned earthquakes suggested by the International Association
for Structural Control and Monitoring for benchmark seismic stud-
ies [40] are summarised in Table 5. Referring to Table 2, although
the model earthquake magnitude remains the same as the proto-
type, time intervals of the original records should be reduced by
the factor of 5.48 (k1/2, k = 1/30) which means that the scaled earth-
quakes contain higher frequencies and shorter durations. The
scaled acceleration records of the four adopted earthquakes
together with the relevant frequency content obtained from Fast
Fourier Transform are illustrated in Fig. 6a–d. In addition, exponen-
tial sine sweep wave with amplitude of 0.05 g, exponential
increase rate of 0.5 Hz, and frequency range of 1–50 Hz has been
applied to the test models in order to identify the dynamic charac-
teristics of the systems. Fig. 6e displays a schematic view of the
applied exponential sine sweep waves.
2.2.5. Laminar soil container
A soil container is required to hold the soil in place during shak-
ing table tests and provide confinement. The ideal soil container
should simulate the free field soil response by minimising bound-
ary effects. Since the seismic behaviour of the soil container affects
the interaction between the soil and the structure, the performance
of the soil container is of key importance for conducting seismic
soil–structure interaction model tests successfully [25]. Well-de-
signed laminar soil container as figured out by many researchers
(e.g. Chau et al. [22]; Taylor, [42]) has advantage among other
types in which lateral motion of the entire depth of the laminar soil
container follows a sinusoidal shape representing authentic condi-
tions of the free field ground motion. Therefore, in order to perform
rigorous and reliable experimental shaking table tests, a laminar
soil container has been employed in this study.
Considering the adopted geometric scaling factor (1:30) and
allowing a further 10 mm on each side for construction purposes
similar to Prasad et al. [43], the final length, width, and depth of
the laminar soil container are selected to be 2.10 m, 1.30 m, and
1.10 m, respectively. Same as the model structure, the laminar soil
container is initially designed employing a 3D numerical model.
The key parameter in the design of the soil container is the natural
frequency of the container itself which should be close to the nat-
ural frequency of the soil deposit inside the container (approxi-
mately 10 Hz for this study) in order to minimise the interaction
Mix C (%) between the soil and the container during the shaking table test.
60 The employed laminar soil container consists of a rectangular
20 laminar box made up of aluminium rectangular hollow section
20
frames separated by rubber layers. The aluminium frames provide
120
lateral confinement of the soil, while the rubber layers allow the
container to deform in a shear beam manner. The soil container
 A.S. Hokmabadi et al. / Computers and Geotechnics 55 (2014) 172–186 177

(a) (b)
Fig. 4. (a) Bender element test setup; (b) schematic graphical signal processing to measure the shear wave travel time between the sender and receiver bender elements.

soil mix ensuring negligible relative slip between the soil and the
Bender Element Test bottom surface of the container and to justify the fixed-base
(Amplitude= 10 V, period= 1 sec.)
assumption in the computer model. Fig. 7 shows the laminar soil
Shear Wave Velocity (m/s)

container adopted in this study.

2.3. Shaking table tests program

Mix A (67.5% kaolinite, 22.5% Bentonite, 10% fly ash and lime, and 100% water)
Mix B (60% kaolinite, 20% Bentonite, 20% fly ash and lime, and 100% water)
Mix C (60% kaolinite, 20% Bentonite, 20% fly ash and lime, and 120% water)
Curing Age (Days)
Fig. 5. Average shear wave velocity for three mixes obtained from Bender Element
Test.
Table 4
Properties of the soil mix on the second day of curing.
Soil properties
Mass density (kg/m3)
Shear wave velocity (m/s)
Maximum shear modulus, Gmax (kPa)
Undrained shear strength, Su (kPa)
Plasticity Index, PI (%)
was fixed and secured on the shaking table using eight M38 bolts
passing through the provided holes. Then, the internal surface of
the soil container was covered and sealed with two layers of black
plastic sheeting. According to Gohl and Finn [44] and Valsangkar
et al. [45], 25 mm thick absorbing layers of Polystyrene foam
sheets have been installed at the end walls of the soil container
to simulate viscous boundaries in the free field condition and min-
imise the reflection of the outward propagating waves back into
the model. In addition, a layer of well graded gravel was glued to
the bottom of the soil container to create a rough interface be-
tween the soil and the base during the test. This layer provides fric-
tion between the timber base plate (as a bedrock) and the in-situ
The shaking table tests have been carried out under three con-
ditions: fixed-base condition, with shallow foundation, and with
floating pile foundation. In the first case, a fixed base model (con-
structed structure directly fixed on top of the shaking table) has
been tested in order to ensure the structural model possesses the
target natural frequency and determine the damping ratio of the
structural model. To achieve the above, constructed structural
model was fixed and secured on the shaking table as shown in
Fig. 1. Displacement transducers (levels 3, 5, 7, 11, 13, and 15)
and accelerometers (at levels 3, 5, 7, 9, 11, 13, and 15) were in-
stalled on the structure in order to monitor the dynamic response
of the structure and to primarily measure the structural lateral dis-
placements. The recorded accelerations can be used to check the
Value consistency and accuracy of the obtained displacements through
1450 a double integration in time domain. In addition, by recording
36 the accelerometers which are installed on two edges of the top
1776 floor, any possible torsion of the structure during the seismic exci-
3.1 tations could be monitored.
42
Initially, Sine Sweep test was performed on the structural mod-
el to determine the natural frequency of the model. Sine Sweep test
involves a logarithmic frequency sweep holding a specified accel-
eration constant at the base of the structure. For the current Sine
Sweep test, by increasing the frequency of the shaking table from
0.1 Hz to 50 Hz, the first resonance between the shaking table
and the structural model frequencies showed the fundamental nat-
ural frequency of the model. The test was repeated three times to
ensure the determined natural frequency is adequately accurate.
The resulting natural frequency of the constructed structural mod-
el obtained from Sine Sweep test results was 2.19 Hz which is in a
very good agreement with the desired natural frequency of the
structural model (2.11 Hz). Therefore, the constructed structural
model, with the natural frequency of 2.19 Hz and the total mass

Table 5
Utilised earthquake base motions.

Earthquake Country Year PGA (g) Mw (R) Duration (s) Type Hypocentral distancea (km)
Northridge USA 1994 0.843 6.7 30.0 Near field 9.2
Kobe Japan 1995 0.833 6.8 56.0 Near field 7.4
El Centro USA 1940 0.349 6.9 56.5 Far field 15.69
Hachinohe Japan 1968 0.229 7.5 36.0 Far field 14.1
a
Obtained from PEER [41].
178 A.S. Hokmabadi et al. / Computers and Geotechnics 55 (2014) 172–186

Amplitude (g/Hz×10-4 )
1.0
Acceleration (g) 6.0
0.8
0.6
Scaled Northridge Earthquake 5.0
0.4 Scaling factor = 1/5.48
0.2 4.0
0.0 3.0
-0.2
-0.4 2.0
-0.6 1.0
-0.8
-1.0 0.0
0 1 2 3 4 5 6 0 10 20 30 40 50
Time (sec) Frequency (Hz)

Amplitude (g/Hz×10-4)
1.0
Acceleration (g)

8.0
0.8
0.6
Scaled Kobe Earthquake 7.0
0.4 Scaling factor = 1/5.48 6.0
0.2 5.0
0.0 4.0
-0.2 3.0
-0.4
2.0
-0.6
-0.8 1.0
-1.0 0.0
0 2 4 6 8 10 0 10 20 30 40 50
Time (sec) Frequency (Hz)

Amplitude (g/Hz×10-4)
0.6
0.4
0.5
Acceleration (g)

0.3 Scaled El Centro Earthquake

0.2 Scaling factor = 1/5.48 0.4
0.1
0.3
0.0
-0.1 0.2
-0.2
0.1
-0.3
-0.4 0.0
0 2 4 6 8 10 12 0 10 20 30 40 50
Time (sec) Frequency (Hz)
Amplitude (g/Hz×10 -4)

0.4
Acceleration (g)

2.0
0.3 Scaled Hachinohe Earthquake
0.2 Scaling factor = 1/5.48 1.5
0.1
0.0 1.0
-0.1
-0.2 0.5
-0.3
-0.4 0.0
0 2 4 6 8 10 0 10 20 30 40 50
Time (sec) Frequency (Hz)

Fig. 6. Adopted shaking events in this study: (a) scaled Northridge earthquake; (b) scaled Kobe earthquake; (c) scaled El Centro earthquake; (d) scaled Hachinohe earthquake;
(e) exponential sine sweep wave.

of 104 kg, possesses the required characteristics to meet the dy-
namic similitude criteria. The estimated value of the structural
damping ratio of the constructed structural model was determined
to be equal to 1.1%, obtained from the free vibration lateral dis-
placement records of the structural model using the Taylor series
expansion [46].
After ensuring the adequacy of the structural model character-
istics, shaking table tests were performed by applying scaled earth-
quake acceleration records of Northridge, 1994 (Fig. 6a), Kobe,
1995 (Fig. 6b), El Centro, 1940 (Fig. 6c), and Hachinohe, 1968
(Fig. 6d) to the fixed base structural model and the results in terms
of maximum lateral deflections are presented in Section 4.
The second case of the shaking table tests was to study the ef-
fects of the soil–structure interaction under the shallow foundation
case. After securing the laminar soil container on the shaking table,
2 cubic meters of the designed soil mix (Mix C: 60% Q38 kaolinite
clay, 20% Active-bond 23 Bentonite, 20% class F fly ash and lime,
and water, 120% of the dry mix) was produced and placed into
the laminar soil container. As explained in Section 2.2.3, the de-
sired soil mix acquires the required stiffness and consequently
the shear wave velocity after two days of curing. As a result, the
time frame for the testing process was very tight and time sensi-
tive. Therefore, soil mixing and placement needed to be carried
out in one day in order to produce a homogenous soil mix, and
after two days of curing, the final tests had to be performed.
During the soil mixing process, ten cylindrical soil samples of
D = 50 mm and h = 100 mm were taken from the soil mix for qual-
ity control of the mix. The entire mixing process and filling the
laminar soil container were completed in one day. Then, the soil
mix inside the container was left to be cured for two days while
 A.S. Hokmabadi et al. / Computers and Geotechnics 55 (2014) 172–186
Fig. 7. Final setup of the shaking table tests for the structure with floating
(frictional) pile foundation.
the surface of the soil container was covered and sealed. On the
second day, the structural model was lifted up and placed on the
designated location, without observing any excessive settlement
or failure underneath the base plate as predicted. In addition to
the instrumentation used on the structure, vertical displacement
transducers were placed on the level of base plate of the structure
(simulating the foundation) to determine the vertical displace-
ments of the structure during the testing process as shown in
Fig. 8a. Similar shaking events including Sine Sweep test and four
scaled earthquake records (Fig. 6) have been applied to the system.
The natural frequency of the soil–structure model from the per-
formed Sine Sweep test was measured to be 1.60 Hz.
The last case of the shaking table tests was to consider the float-
ing pile foundation and investigate the influence of soil–pile–struc-
ture interaction on the seismic response of the superstructure by
comparing this case with the previously mentioned fixed-base
and shallow foundation cases. Since the properties of the designed
soil mix is time dependent, this stage should be carried out at the
(a)
Fig. 8. Shaking table tests setup and connections for: (a) model shallow foundation; (b) model pile foundation.
179
same age as for the shallow foundation case in order to make the
results comparable, without any variation of the dynamic soil
properties.
As mentioned in Section 2.2.2, commercial Polyethylene pres-
sure pipe was employed to build the model piles. The length of
the model floating piles is 66 cm, leaving 34 cm distance between
the piles toe and the base. Wooden tips were fitted to the model
piles to provide a closed end condition. The model piles were dri-
ven into the soil through a 150 mm tall wooden template to ensure
location and verticality. Moreover, employing template during the
installation process helps to achieve full connection between the
piles and the surrounding soil without generating any gap due to
the installation process. The template was constructed with special
cut outs to accommodate a few millimetres of extra room for piles
with external strain gages aiming to prevent any possible damage
to the strain gages during installation.
After installation of the model piles the template was removed
and the steel plate (simulating the foundation) with prefabricated
holes was fitted over the group. Sixteen M12 bolts were used to
provide fixed connection between the piles head and the steel
plate as shown in Fig. 8b. The required nuts were fixed to the pile
top with strong glue and steel rings before the test and the strength
and capability of this connection technique was examined success-
fully. Then, the model structure was suspended from the overhead
crane and connected to the steel plate from the pre-located con-
nections (Fig. 9b) similar to the fixed-base and shallow foundation
cases.
Consequently, all the components of the system including the
container, soil, piles, and superstructure were installed. The same
arrangement of displacement transducers and accelerometers has
been used on the structure and the steel plate (simulating the foun-
dation). In addition, twelve strain gauges were installed on the piles
and four 3D accelerometers were embedded inside the soil body.
Since the influence of the soil–structure interaction on the response
of the superstructure is the main objective of this research, just the
data obtained from the instrumentation on the structure itself, not
including the soil and piles sensors, are reported in this paper. Sim-
ilar shaking events including Sine Sweep test and four scaled earth-
quake records have been applied to the floating pile foundation
system. The natural frequency of the soil–pile–structure model
from the performed Sine Sweep test was measured to be 1.8 Hz.
The results of the conducted shaking table tests under the influence
of four scaled earthquake acceleration records in terms of the max-
imum lateral deflections of various stories of the structure are pre-
sented and discussed in Section 4. The final setup of the tests
including the displacement transducers and accelerometers at dif-
ferent levels of the structural model for the floating pile foundation
system on the shaking table are shown in Fig. 7.
(b)
180 A.S. Hokmabadi et al. / Computers and Geotechnics 55 (2014) 172–186
(a)
(b)
(c)
Fig. 9. Numerical grid and model components in FLAC3D for: (a) fixed-base
structure; (b) structure supported by shallow foundation; (c) structure supported
by floating (frictional) pile foundation.
3. Development of 3d numerical model
In order to conduct a fully coupled analysis of the entire soil–
pile–structure system, a three dimensional numerical soil–struc-
ture model has been developed which treats the behaviours of
the soil and the structure with equal rigour. Adopting direct method
of analysis, the numerical model can perform fully nonlinear time
history dynamic analysis to simulate realistic dynamic behaviour
of soil and structure under seismic excitations. According to Chu
[14], time domain analysis in necessary to compute the nonlinear
dynamic responses of soil–pile–structure systems as the frequency
domain analysis can deal only with linear responses without con-
sidering any nonlinearities. In this study, three-dimensional expli-
cit finite difference based program FLAC3D, Fast Lagrangian
Analysis of Continua, version 4.0 [20] has been employed following
the other researchers experience (e.g. Comodromos and Papado-
poulou [47]; Rayhani and El Naggar, [48]). This program can simu-
late behaviour of different types of structures and materials by
elements which can be adjusted to fit the geometry of the model.
Each element behaves according to a prescribed constitutive model
in response to the applied forces or boundary restraints. The pro-
gram offers a wide range of capabilities to solve complex problems
in mechanics such as inelastic analysis including plastic moment
and simulation of hinges for structural systems.
Three cases including: fixed-base condition, the structure sup-
ported by the shallow foundation, and the structure supported
by the floating pile foundation have been modelled separately
and the results are compared. The dimensions of the numerical
models were chosen to be similar to the experimental tests. The
reason for choosing the soil deposit thickness of 30 m for the pro-
totype is that most amplification occurs within the first 30 m of the
soil profile, which is in agreement with most modern seismic codes
calculating local site effects based on the properties of the top 30 m
of the soil profile [48].
Experience gained from the parametric studies helped to final-
ise the adopted mesh size and the maximum unbalanced force at
the grid points to optimise the accuracy and the computation
speed, simultaneously. For the floating pile foundation model, the
generated mesh comprised 10,868 zones and 16,356 grid points.
Fast computation facilities at University of Technology Sydney
were employed to conduct the time-history analysis, and the com-
putation took approximately 20 h for a single analysis. The numer-
ical grid and model components in FLAC3D are shown in Fig. 9.
Adjusting the boundary conditions for the static analysis, in
which the system is under the gravity loads only, the bottom face
of the model is fixed in all directions, while the side boundaries are
fixed in the horizontal directions. During the dynamic time-history
analysis, in order to avoid reflection of outward propagating waves
back into the model, quiet (viscous) boundaries comprising inde-
pendent dashpots in the normal and shear directions are placed
at the lateral boundaries of the soil medium. The lateral boundaries
of the main grid are coupled to the free-field grids by viscous dash-
pots of quiet boundaries at the sides of the model, as shown in
Fig. 2, to simulate the free field motion which would exist in the
absence of the structure and pile foundation. Rigid boundary con-
ditions is adopted to simulate the bedrock in the seismic soil–
structure interaction analysis as suggested by other researchers
(e.g. Dutta and Roy [49]; Spyrakos et al. [50]), and the earthquake
input motions are applied at the bedrock assuming horizontally
polarised shear waves propagating vertically. Lu et al. [51] empha-
sised on the influence of the gravity load on contact state of the
soil–structure interface mentioning that significant error in the
analysis may occur if gravity is not taken into account in the dy-
namic analysis.
Solid elements are used to model the soil deposits, and Mohr-
Coulomb failure criterion is adopted. In addition, the built-in tan-
gent modulus function developed by Hardin and Drnevich [52] is
adopted to implement hysteretic damping of the soil representing
the variation of the shear modulus reduction factor and damping
ratio with cyclic shear strain of the soil. This model is defined as
follows:
Ms ¼ 1=ð1 þ ðc=cref ÞÞ ð2Þ
 A.S. Hokmabadi et al. / Computers and Geotechnics 55 (2014) 172–186
where Ms is the secant modulus (G/Gmax), c is the cyclic shear strain,
and cref is Hardin/Drnevich constant. In this study, cref = 0.234 which
gives the best numerical fit to the backbone curves suggested by
Sun et al. [53] for the fine grained soils (cohesive soils) are adopted,
and the comparison has been presented in Fig. 10.
There are two main analytical procedures for dynamic analysis
of soil–structure systems under seismic loads, equivalent-linear
and fully nonlinear method. The equivalent-linear method (e.g.
Seed and Idriss [54]) cannot capture directly any nonlinearity ef-
fects during the solution process and uses linear properties for
each element that remain constant throughout the history of shak-
ing, and are estimated from the mean level of dynamic motion.
Therefore, the above mentioned strain-dependent modulus and
damping functions for the soil are only taken into account in an
average sense, in order to approximate some effects of nonlinear-
ity. As a result, during the low amplitude shakings in the excitation
history, soil elements will be modelled overdamped and too soft,
and during the strong shaking soil elements will be modelled
underdamped and too stiff. In contrast, employing fully nonlinear
method, nonlinearity in the stress-strain law is followed directly
by each element and the dependence of damping and apparent
modulus on strain level is automatically modelled. Byrne et al.
[55] and Beaty [56] provided some overviews of the above men-
tioned methods and concluded that the most appropriate method
for a dynamic analysis of soil–structure system is a fully nonlinear
method. In addition, Lu et al. [57] studies illustrated the potential
for further reliance on the computer simulation in the assessment
of the nonlinear seismic ground response using the nonlinear dy-
namic analysis. Consequently, fully nonlinear method for dynamic
analysis of soil–structure systems has been employed in this study.
The common soil tests such as bender element and density test
were conducted on the obtained samples during the mixing pro-
cess on the second day of curing age as described in Section 2.3.
The results are in good conformity with the initial laboratory test
results and adopted in the numerical model as summarised in
Table 4.
The pile elements and superstructure are modelled with solid
elements considering elastic-perfectly plastic behaviour with
yielding criteria for the elements to control the possibly of inelastic
behaviour in both superstructure and piles. The formulation
adopted in this study to simulate the inelastic behaviour of pile
and structural elements assumes that the material behaves linear
elastically until reaching the defined yield stress and after reaching
this yield stress, the material deforms without inducing additional
resistance. The yield stress for the material used in the model
structure is 280 MPa. As a calibration, a FLAC3D analysis was first
conducted on a cantilever pile while the pile was fixed at one end
into the ground without the surrounding soil and different lateral
Damping Ratio
Backbone curve for
cohesive soils
G / G max
Sun et al. (1988) (Utilised test: Resonant column test)
----Adopted in this study for γref =0.234
Cyclic Shear Strain (%)
(a)
Fig. 10. Adopted fitting curve for fine grained soil in this study (after Sun et al. [57]): (a) Relations between G/Gmax versus cyclic shear strain; (b) relations between damping
ratio versus cyclic shear strain.
181
loads were applied on the free end of the cantilever pile. The
recorded deflection from the FLAC3D model showed less than 2%
difference from the existing analytical predictions, confirming the
accuracy of the model.
Because of the different characteristics of the soil and the super-
structure/piles, sliding and separation may occur at the soil–struc-
ture interfaces [58]. Two sets of interface elements are modelled in
this study. For the shallow foundation case, the interface elements
are placed between the foundation and the soil surface (Fig. 11a).
However, for the pile foundation case, the interface elements were
attached to the outer perimeter and bottom of the piles as shown
in Fig. 11b. It should be noted that in the pile foundation case there
is no interface or attachment between the foundation and the sur-
face soil as some gap in the shaking table tests was considered to
avoid any pile-raft behaviour. Therefore, there is no direct stress
transfer between the foundation slab and the subsoil in the pile
foundation case. According to Fig. 11c, the normal and shear forces
that describe the interface response in the elastic range are deter-
mined at calculation time (t + Dt) using the following relations
[20]:
F ðtþDtÞn ¼ kn un A þ rn A ð3Þ
F ðtþDtÞsi ¼ F ðtÞsi þ ks Duðtþð1=2ÞDtÞsi A þ rsi A ð4Þ
ðtþDtÞn ðtþDtÞsi
where F and F are the normal and shear force vector at
time (t + Dt), respectively. un is the absolute normal penetration
of the interface node into the target face. Dusi is the incremental rel-
ative shear displacement vector. rn and rsi are the additional nor-
mal and shear stresses added due to interface stress initialisation,
respectively. kn and ks are the normal and shear stiffnesses, respec-
tively, and A is the representative area associated with the interface
node. The lateral and axial stiffnesses of the interface elements are
set to ten times the equivalent stiffness of the neighbouring zone,
based on the recommended relationship by Rayhani and El Naggar
[48] and Itasca Consulting Group [20] for the isotropic soil medium,
as follows:
 
K þ 43 G
ks ¼ kn ¼ 10 ð5Þ
Dzmin
where K and G are bulk and shear modulus of neighbouring zone,
respectively, and Dzmin is the smallest width of an adjoining zone
in the normal direction. This is a simplifying assumption that has
been used to ensure that the interface stiffness has minimal influ-
ence on system compliance by avoiding the intrusion of adjacent
zones (a numerical effect) and preventing excessive computation
time [48]. In addition, shear strength of the interfaces was defined
by Mohr–Coulomb failure criterion and the tensile strength of the
Sun et al. (1988) (Utilised test: Resonant column test)
----Adopted in this study for γref =0.234
Backbone curve for
cohesive soils
Cyclic Shear Strain (%)
(b)
182 A.S. Hokmabadi et al. / Computers and Geotechnics 55 (2014) 172–186
Ts
ks S Ss
D
kn
S = slider
Ts= tensile strength
Ss= shear strength
D = dilation (assumed zero) A
Ks= shear stiffness
Kn= normal stiffness
(c)
Fig. 11. Interface elements adopted in this study: (a) interfaces between the shallow foundation and the soil; (b) interfaces at the outer perimeter and tip of the floating piles
and surrounding soil; (c) components of the interface constitutive model.
interfaces are set to zero in order to allow gapping between the
piles and the supporting soil in the pile foundation case and uplift
in the shallow foundation case.
Finally, fully nonlinear time-history analysis is conducted under
the influence of the mentioned shaking events (Fig. 6), and results
in terms of the maximum inelastic lateral deflections determined
for three mentioned cases are presented and discussed in the next
section.
4. Results and discussion
Results of the conducted shaking table tests and the 3D numer-
ical predictions for the maximum lateral displacements of the
fixed-base, shallow foundations, and floating pile foundations are
presented in Figs. 12 and 13, respectively. To determine the lateral
deflections, the movement of the shaking table has been subtracted
from the storey movements. Therefore, all the records are relative
to the base movements. It should be noted that the presented data
are based on the lateral deformation of each storey when the max-
imum deflection at the top level occurred. This approach gives more
reasonable pattern of the structural deformation in comparison
with the approach that maximum absolute storey deformation irre-
spective of occurrence time are recorded [59]. Fig. 14 illustrates a
sample of time-history deformation records used to obtain the lat-
eral deformations reported in Figs. 12 and 13.
Comparing the results of the conducted shaking table tests
(Fig. 12) and the 3D numerical predictions (Fig. 13) for the maxi-
mum lateral displacements of the fixed-base, shallow foundations,
and floating pile foundations, it is observed that the trend and the
values of the 3D numerical predictions are in a good agreement
and consistent with the experimental shaking table test results.
Therefore, the developed 3D numerical model can replicate the
behaviour of the soil–pile–structure system with acceptable accu-
racy and is a rational and appropriate tool for further studies of the
soil–pile–structure interaction effects. The observed disparity be-
tween FLAC3D predictions and experimental measurements in
the lower levels of the shallow foundation and floating pile cases
can be due to the nature of the numerical method, adopting elas-
tic-perfectly plastic Mohr–Coulomb model for the soil, assuming
(b)
(a)
ideal rigid connection between the foundation and the pile caps,
and unavoidable experimental uncertainties. Moreover, as an
example, Fig. 15 presents the time-history acceleration records at
the top of the 15-storey model structure for the fixed-base, shallow
foundations, and floating pile foundations under the influence of
1940 El Centro earthquake. Comparison of the measurements
and the predictions indicates that the horizontal acceleration –
time curves obtained from the 3D numerical analysis and the lab-
oratory experiments are in a reasonable agreement.
In order to draw a general conclusion to be used by practicing
engineers, the average values of the 3D numerical predictions
and experimental values of deformations for each case were deter-
mined and compared in Fig. 16. In comparison to the fixed base
structure, the maximum lateral deflection of the structure sup-
ported by floating pile foundations increases by 34%, and 27%
based on the experimental measurements and the 3D numerical
predictions, respectively. Moreover, the maximum lateral deflec-
tion of the structure supported by the shallow foundation is in-
creased by 55% based on the experimental values and 59% based
on the 3D numerical predictions in comparison to the results ob-
tained from the fixed base structure. The natural frequency of
the system reduces due to the soil–structure interaction (2.19 Hz
and 2.11 Hz for the fixed-base condition, 1.8 Hz and 1.88 Hz for
the floating pile foundation, and 1.60 Hz and 1.64 Hz for the shal-
low foundation case based on the experimental results and numer-
ical predictions, respectively). Therefore, such decreases in the
natural frequency (increases in the natural period) considerably al-
ter the response of the building frames under the seismic excita-
tion. This is due to the fact that the natural period of the system
lies in the long period region of the response spectrum curve,
and the displacement response tends to increase. The pile founda-
tions reduce the lateral displacements in comparison to the shal-
low foundation case since the presence of stiff pile elements in
the soft soil increases the equivalent stiffness of the ground and
influences the dynamic properties of the whole system such as
the natural frequency and damping.
Rocking component plays an important role in the lateral defor-
mation of the superstructure. According to Kramer [3], relative
lateral structural displacements under the influence of the soil–
structure interaction consist of rocking and distortion components.
 A.S. Hokmabadi et al. / Computers and Geotechnics 55 (2014) 172–186 183

15 15
14 Scaled Northridge Earthquake (1994) 14
13
12
Near field Eearthquake,
Scaled factor = 1/ 30
13
12
(b) Scaled Kobe Earthquake (1995)
Near field Earthquake

Storey Number
Mw = 6.7 (R), PGA = 0.843 (g) Scaled factor = 1/ 30
11 11
Storey Number

Mw = 6.8 (R), PGA = 0.833 (g)

10 10
9 (a) 9
8 8
7 7
6 6
5 5 Fixed base Experimental Results
4 Fixed base Experimental Results 4
Shallow foundation Experimental Results
3 Shallow foundation Experimental Results 3
2 2 Floating piles Experimental Results
Floating piles Experimental Results
1 1
0 0
0 10 20 30 40 0 10 20 30 40
Maximum Lateral Deflection (mm) Maximum Lateral Deflection (mm)

15 15
14 14
13 (c) Scaled El Centro Earthquake (1940)
Far field Earthquake 13 (d) Scaled Hachinohe Earthquake (1968)
Far field Earthquake
12 Scaled factor = 1/ 30 12 Scaled factor = 1/ 30
11 Mw = 6.9 (R), PGA = 0.349 (g)

Storey Number
Storey Number

11 Mw = 7.2 (R), PGA = 0.229 (g)

10 10
9 9
8 8
7 7
6 6
5 Fixed base Experimental Results 5 Fixed base Experimental Results
4 4
3 Shallow foundation Experimental Results 3 Shallow foundation Experimental Results
2 2
Floating piles Experimental Results Floating piles Experimental Results
1 1
0 0
0 10 20 30 40 0 10 20 30 40
Maximum Lateral Deflection (mm) Maximum Lateral Deflection (mm)

Fig. 12. Recorded maximum lateral deflection of the structure from the shaking table tests for the fixed base, shallow foundation, and end-bearing pile foundation cases
under the influence of: (a) Northridge earthquake; (b) Kobe earthquake; (c) El Centro earthquake; (d) Hachinohe earthquake.

15
14 15
Scaled Northridge Earthquake (1994)
14
13
12
Near field Eearthquake,
Scaled factor = 1/ 30 13 (b) Scaled Kobe Earthquake (1995)
Near field Earthquake
Mw = 6.7 (R), PGA = 0.843 (g) 12
11 Scaled factor = 1/ 30
Storey Number

11
Storey Number

Mw = 6.8 (R), PGA = 0.833 (g)

10
10
9 (a) 9
8
8
7 7
6 6
5 Fixed base 3D Numerical Results 5 Fixed base 3D Numerical Results
4 4
3 Shallow foundation 3D Numerical Results
3 Shallow foundation 3D Numerical Results
2 2
Floating piles 3D Numerical Results Floating pile 3D Numerical Results
1 1
0 0
0 10 20 30 40 0 10 20 30 40
Maximum Lateral Deflection (mm) Maximum Lateral Deflection (mm)

15 15
14 14
13 (c) Scaled El Centro Earthquake (1940)
Far field Earthquake 13 (d) Scaled Hachinohe Earthquake (1968)
Far field Earthquake
12 Scaled factor = 1/ 30 12 Scaled factor = 1/ 30
11 Mw = 6.9 (R), PGA = 0.349 (g)
Storey Number

11 Mw = 7.2 (R), PGA = 0.229 (g)

Storey Number

10 10
9 9
8 8
7 7
6 6
5 Fixed base 3D Numerical Results 5 Fixed base 3D Numerical Results
4 4
Shallow foundation 3D Numerical Results Shallow foundation 3D Numerical Results
3 3
2 Floating piles 3D numerical Results 2 Floating piles 3D Numerical Results
1 1
0 0
0 10 20 30 40 0 10 20 30 40
Maximum Lateral Deflection (mm) Maximum Lateral Deflection (mm)

Fig. 13. 3D numerical predictions of the maximum lateral deformation under the influence of: (a) Northridge earthquake; (b) Kobe earthquake; (c) El Centro earthquake; (d)
Hachinohe earthquake.

The maximum vertical displacement and the rocking angles of the measured to be 13.63 mm due to distortion component, while
foundation in the instant of the maximum deformation at the top the maximum lateral deflection at the top of the structure sup-
of the structure are summarised in Table 6. Accordingly, for the ported by floating pile foundation was 20.40 mm with 7.62 mm
floating pile foundation cases, approximately 27% of the maximum of that value being due to rocking component and 12.78 mm took
lateral deflections were due to the rocking component, while 73% place due to distortion component. In the floating pile foundation
took place due to the distortion component. These values for the cases, rocking occurs due to the axial deformation of the pile
shallow foundation cases are 37% and 63%, respectively. For exam- elements together with the deformation of the surrounding and
ple, under the influence of 1940 El Centro earthquake, the maxi- beneath soil elements. The area replacement ratio of the pile group
mum lateral deflection at the top of the fixed base model was is 8% in this study and as a result piles attract significant axial
184 A.S. Hokmabadi et al. / Computers and Geotechnics 55 (2014) 172–186

15
14
13
12
11

Storey Number
10
9
8
7
6 Fixed base Numerical Results
5 Fixed base Exp. Results
4 Shallow foundation Numerical Results
3 Shallow foundation Exp. Results
2 Floating piles Numerical Results
1 Floating piles Exp. Results
0
0 10 20 30
Maximum Lateral Deflection (mm)

Fig. 16. Average values of maximum lateral displacements base on shaking table
experimental measurements versus 3D numerical predictions.
Fig. 14. Sample experimental time-history displacement results for the fixed base
model under the influence of El Centro earthquake.

Table 6
Maximum vertical displacement and rocking angle of the base plate.
1.5
Fixed base Scaled Maximum vertical displacement (rocking angle) of the base
Accelaration (g)

1.0
earthquake plate
0.5 acceleration
Fixed Shallow foundation Floating pile foundation
0.0 record
base
-0.5 Experimental Numerical Experimental Numerical

-1.0 Northridge 0 2.54 mm 2.68 mm 1.9 mm 1.98 mm

-1.5
(a) Numerical Results Exp. Results (0.58°) (0.61°) (0.43°) (0.45°)
0.0 0.5 1.0 1.5 2.0 2.5 Kobe 0 1.32 mm 1.45 mm 0.43 mm 0.52 mm
(0.30°) (0.33°) (0.1°) (0.12°)
Time (sec)
El Centro 0 1.98 mm 2.06 mm 1.27 mm 1.34 mm
3.0 (0.45°) (0.47°) (0.29°) (0.31°)
Shallow foundation Hachinohe 0 1.47 mm 1.58 mm 0.93 mm 1.02 mm
Accelaration (g)

2.0
(0.33°) (0.36°) (0.21°) (0.23°)
1.0
0.0
-1.0
base structure excluding soil–structure interaction. Therefore,
-2.0
(b) Numerical Results Exp. Results
presence of floating pile foundation increases the base shear and
-3.0 in turn demands of the superstructure, in comparison with the case
0.0 0.5 1.0 1.5 2.0 2.5
Time (sec)
supported by shallow foundation.
Since the adopted model is a Multi Degree of Freedom (MDOF)
3.0 structure, inter-storey drifts can be estimated and employed to
Floating piles
investigate the performance levels of the building structures under
Accelaration (g)

2.0
1.0 the influence of dynamic soil–structure interaction. The corre-
0.0 sponding inter-storey drifts of the average values of 3D numerical
-1.0 model (Fig. 18) have been calculated using the following equation
based on the Australian standard [60]:
-2.0
(c) Numerical Results Exp. Results
-3.0
0.0 0.5 1.0 1.5 2.0 2.5 Drift ¼ ðdiþ1  di Þ=h ð6Þ
Time (sec)
where di+1 is deflection at (i + 1) level, di is deflection at (i) level, and
Fig. 15. Time-history acceleration records at top of the 15-storey model structure
h is the storey height. In the performance-based seismic design, the
under the influence of El Centro earthquake for: (a) fixed-base structure; (b)
structure supported by shallow foundation; (c) structure supported by floating seismic performance (performance level) is described by consider-
(frictional) pile foundation. ing the maximum allowable damage state (damage performance)
for an identified seismic hazard (hazard level). Performance levels
describe the state of structures after being subjected to a certain
forces. However, clearly the rocking of the structure in the shallow hazard level, and based on FEMA273/274 [61] are classified as: fully
foundation case, without pile elements, is much more than the case operational, operational, life safe, near collapse, or collapse. Overall
with pile foundations resulting in further amplification of the lat- lateral deflection, ductility demand, and inter-storey drifts are the
eral deflection. most commonly used damage parameters. The above mentioned
Fig. 17 compares the 3D numerical predictions of the structural five qualitative levels are related to the corresponding quantitative
demand in terms of the base shear for the fixed-base structure and maximum inter-storey drifts (as a damage parameter) of: <0.2%,
the structure supported by two types of foundations, shallow foun- <0.5%, <1.5%, <2.5%, and >2.5%, respectively [61]. In addition, most
dations and floating pile foundations. In general, the ratio of the of the force-based design codes employ an additional check in
base shear for cases including soil–structure interaction to that of terms of limiting inter-storey drifts to ensure that particular defor-
fixed-base is less than one, demonstrating the effect of soil–struc- mation-based criteria are met. For example, ASCE [62] defines
ture interaction in reducing the base shear of the structure. The allowable storey drift for structures considering type and risk cate-
base shear of the structure supported by the floating pile founda- gory of the structure. Australian Earthquake Code [60] indicates
tion and shallow foundation is on average 78% and 70% of the fixed 1.5% as the maximum allowable storey drift. According to Fig. 18,
 A.S. Hokmabadi et al. / Computers and Geotechnics 55 (2014) 172–186 185

700 modulus and damping ratio of the soil with the cyclic shear strain.
Fixed base Another advantage of the current numerical modelling technique
600
is performing the SSPSI analysis in a fully-coupled manner in
Shallow foundation which main components of the interaction including subsoil, pile
500
foundation, and superstructure are modelled simultaneously with-
Base Shear (N)

Floating pile foundation

400 out resorting to independent calculations of site or superstructure
response, or application of the pile group interaction factors.
300

200 5. Conclusions
100
The pile foundations are usually considered to be rigid enough
0 to guarantee the restraint against rocking motions and conven-
Northridge Kobe El Centro Hachinohe
tional fixed-base models are used to predict the seismic response
Earthquake
of these systems although the behaviour of system can be signifi-
Fig. 17. Base shear of the model structure obtained from 3D numerical analysis for: cantly affected by seismic soil–pile–structure interaction. A series
fixed-base structure; structure supported by shallow foundation; and structure of shaking table experimental tests have been conducted and a
supported by floating (frictional) pile foundation. laminar soil container has been designed to simulate the free field
soil response by minimising the boundary effects, and the super-
structure has been simulated as a multi-storey frame representing
15
14 most of the dynamic properties of the prototype structure.
13 Employing FLAC3D, a fully nonlinear three dimensional numerical
12
11 model has been adopted to perform time-history analyses imple-
Storey Number

10 menting hysteretic damping of the soil to represent the variation

9
8 of the shear modulus and damping ratio of the soil with the cyclic
7 shear strain. Free field boundary conditions have been assigned to
6
5
the numerical model, and appropriate interface elements, capable
Fixed base
4
Shallow foundation
of modelling sliding and separation between the pile and soil ele-
3
2
Floating pile foundation ments, have been considered.
1 Life safe limit (1.5%) By comparing predicted and observed results, it can be con-
0
0 0.5 1 1.5 2 2.5
cluded that the employed numerical model is appropriate for the
Inter-storey Drift (%)
simulation of the soil–pile–structure interaction under strong
ground motions. In addition, based on the shaking table results
Fig. 18. Average 3D numerical inter-storey drifts for: fixed-base structure; struc- and 3D numerical investigations, it is observed that the lateral
ture supported by shallow foundation; structure supported by floating (frictional) deflections of structures siting on the floating pile foundations
pile foundation.
amplified in comparison to the fixed base model (34% based on
the experimental measurements and 27% based on the 3D numer-
ical predictions). This amplification for the structure siting on the
seismic soil–structure interaction tends to increase the inter-storey
shallow foundation is more severe (55% based on the experimental
drifts of the superstructure. The inter-storey drifts of the structure
measurements and 59% based on the 3D numerical predictions).
supported by the floating pile foundation are more than the
Therefore, the floating pile foundations increase the lateral
fixed-base conditions excluding soil–structure interaction. How-
displacements of the superstructure in comparison with the
ever, the structure supported by floating pile foundation experi-
fixed-base assumption, and reduce the lateral displacements in
ences less inter-storey drifts in comparison to the structure
comparison to the shallow foundation case due to the rocking
supported by the shallow foundation. For example, the maximum
components.
recorded inter-storey drift of the fixed base structure is measured
Consequently, seismic soil–pile–structure interaction affects
to be 1.48%, while the corresponding value for the floating pile
the performance level of structures sitting on the soft soil by
foundation and shallow foundation cases are 1.83% and 2.25%,
increasing the inter-storey drifts, which may shift the performance
respectively. In other words, effects of soil–pile–structure interac-
level of the structure from life safe to near collapse or even collapse
tion (pile foundation) and soil–structure interaction (shallow
levels. Therefore, ignoring the real deformability of the soil–struc-
foundation) induces 23% and 52% increase in the recorded inter-sto-
ture system, in both cases with shallow or pile foundations, may
rey drifts, respectively. As a result, the soil–structure interaction
lead to erroneous evaluations of structural displacements and in-
may affect the performance level of the structure and shift the
ter-storey drifts affecting the predicted damage level of structural
performance level of the structure from life safe zone to near
and non-structural elements as well as the lateral load carrying
collapse or even collapse levels.
mechanism of the soil–structure system.
Therefore, although SSPSI reduces the base shear of the struc-
ture leading to the reduction in the structural distortion in compar-
ison with fixed base structure, considering the effect of SSPSI References
increases the overall lateral deformation and consequently inter- [1] Park R. Reinforced concrete structures. New York: John Wiley & Sons; 1975.
storey drifts of the structure mainly due to the rocking component. [2] Priestley MJN. Performance based siesmic design. In: 12th World conference
Moreover, in the seismic response of pile groups, rocking and on earthquake engineering (12WCEE). Auckland, New Zealand: New Zealand
National Society for Earthquake Engineering; 2000.
translation components are coupled and the response of the under-
[3] Kramer SL. Geotechnical earthquake engineering. Prentice Hall; 1996.
neath soils to strong seismic shaking is strongly nonlinear. [4] Wolf JP. Dynamic soil–structure interaction. Englewood Cliffs, New
Practicing engineers can adopt this verified numerical model- Jersey: Prentice-Hall; 1985.
ling procedure in the design considering the effect of SSPSI with [5] Guin J, Banerjee PK. Coupled soil–pile–structure interaction analysis under
seismic excitation. J Struct Eng 1998;124:434–44.
respect to the interface elements, boundary conditions, and hyster- [6] Yingcai H. Seismic response of tall building considering soil–pile–structure
etic damping of the soil representing the variation of the shear interaction quake engineering and engineering. Vibration 2002;1:57–65.
186 A.S. Hokmabadi et al. / Computers and Geotechnics 55 (2014) 172–186
[7] Finn WD. A study of piles during earthquakes: issues of design and analysis.
Bull Earthq Eng 2005;3:141–234.
[8] Ma XH, Cheng YM, Au SK, Cai YQ, Xu CJ. Rocking vibration of a rigid strip
footing on saturated soil. Comput Geotech 2009;36:928–33.
[9] Mendoza M, Romo M. Behavior of building foundations in Mexico city during
the 1985 earthquake: second stage. Lessons learned from the 1985 Mexico
earthquake: Earthquake Eng. Research Inst.; 1989. p. 66–70.
[10] Mizuno H, Iiba M, Hirade T. Pile damage during the 1995 Hyogoken-Nanbu
earthquake in Japan. In: Proc 11th world conf earthquake eng. Acapulco; 1996.
[11] Hokmabadi AS, Fakher A, Fatahi B. Seismic strain wedge model for analysis of
single piles under lateral seismic loading. Aust Geomech 2011;46:31–41.
[12] Ashour M, Norris G, Pilling P. Lateral loading of a pile in layered soil using the
strain wedge model. J Geotech Geoenviron Eng 1998;124:303–15.
[13] Comodromos EM, Papadopoulou MC. Explicit extension of the p–y method to
pile groups in cohesive soils. Comput Geotech 2013;47:28–41.
[14] Chu D. Three-dimentional nonlinear dynamic analysis of soil–pile–structure
interaction. Saint Louis, Missouri: Washington University; 2006.
[15] Han YC, Cathro D. Seismic behaviour of tall building supported on pile
foundations. Seismic analysis and design for soil–pile–structure interaction.
Geotechnical Special Publications, No. 70. ASCE; 1997. p. 36–51.
[16] Carbonari S, Dezi F, Leoni G. Linear soil–structure interaction of coupled wall-
frame structures on pile foundations. Soil Dyn Earthq Eng 2011;31:1296–309.
[17] Veletsos AS. Dynamics of structure-foundation systems. In: Hall WJ, editor.
Structural and geotechnical mechanics. Englewood Cliffs, New
Jersey: Prentice-Hall; 1977. p. 333–61.
[18] Gazetas G. Formulas and charts for impedances of surface and embedded
foundations. J Geotech Eng 1991;117:1363–81.
[19] Anandarajah A, Rashidi H, Arulanandan K. Elasto-plastic finite element
analyses of a soil–structure system under earthquake excitations. Comput
Geotech 1995;17:301–25.
[20] Itasca. FLAC3D version 4.00 Fast Lagrangian analysis of Continua in three
dimensions, User’s manual. Minneapolis, Minnesota, USA: Itasca Consulting
Group, Inc.; 2009.
[21] Rayhani M, El Naggar M, Tabatabaei S. Nonlinear analysis of local site effects
on seismic ground response in the bam earthquake. Geotech Geol Eng
2008;26:91–100.
[22] Chau KT, Shen CY, Guo X. Nonlinear seismic soil–pile–structure interactions:
shaking table tests and FEM analyses. Soil Dyn Earthq Eng 2009;29:300–10.
[23] Ishimura K, Ohtsuki A, Yokoyama K, Koyanagi Y. Sway-rocking model for
simulating nonlinear response of sandy deposit with structure. In: Tenth
world conference on earthquake engineering; 1992. p. 1897–903.
[24] Jakrapiyanun W. Physical modeling of dynamics soil–foundation–structure-
interaction using a laminar container. Ph.D.; 2002.
[25] Pitilakis D, Dietz M, Wood DM, Clouteau D, Modaressi A. Numerical simulation
of dynamic soil–structure interaction in shaking table testing. Soil Dyn Earthq
Eng 2008;28:453–67.
[26] Meymand PJ. Shaking table scale model tests of nonlinear soil–pile–
superstructure in soft clay. PhD thesis in Civil Engineering. University of
California, Berkley; 1998.
[27] Harris HG, Sabnis GM. Structural modeling and experimental
techniques. USA: CRC Press; 1999.
[28] Langhaar H. Dimensional analysis and theory of models. New York: John Wiley
and Sons; 1951.
[29] CSI. SAP2000 v14 analysis reference manual. California: CSI (Computers and
Structures Inc.), Berkley; 2010.
[30] AS/NZS3678. Structural steel – hot-rolled plates, floorplates and slabs. NSW,
Australia: Standards Australia; 2011. p. 34.
[31] Bao Y, Ye G, Ye B, Zhang F. Seismic evaluation of soil–foundation–
superstructure system considering geometry and material nonlinearities of
both soils and structures. Soils Found 2012;52:257–78.
[32] Tao X, Kagawa T, Minowa C, Abe A. Verification of dynamic soil–pile
interaction. Geotechnical Special Publication No. 75. ASCE; 1998. p. 1199–210.
[33] AS/NZS4130. Polyethylene (PE) pipes for pressure applications. NSW,
Australia: Standards Australia; 2009. p. 37.
[34] Turan A. Physical modeling of seismic soil–structure interaction of embedded
structures [NR50400]. Canada: The University of Western Ontario (Canada);
2009.
[35] Moss RE, Crosariol V, Kuo S. Shake table testing to quantify seismic soil
structure interaction of underground structures. In: International conference
on recent advances in geotechnical earthquake engineering and soil dynamics,
San Diego; 2010.
[36] Wartman J. A laboratory study of the effects of fly ash on the geotechnical
properties of soft clay. University of California, Berkeley; 1996.
[37] Riemer M, Gookin W, Bray J, Wartman J. Using reflected waves to measure
small strain dynamic properties. In: 5th Caltrans seismic research workshop,
Sacramento; 1998. p. 16–8.
[38] AS5101.4. Methods for preparation and testing of stabilised materials, Method
4: unconfined compressive strength of compacted materials. NSW, Australia:
Standards Australian; 2008.
[39] Towhata I. Geotechnical earthquake engineering. Berlin Heidelberg: Springer-
Verlag; 2008.
[40] Karamodin AK, Kazemi HH. Semi-active control of structures using neuro-
predictive algorithm for MR dampers. Struct Control Health Monit
2010;17:237–53.
[41] PEER. PEER ground motion database. Pacific Earthquake Engineering Research
Centre, University of California, Berkeley, CA; 2012.
[42] Taylor CA. Large scale shaking tests of geotechnical structures. Earthquake
Engineering Research Centre, University of Bristol; 1997.
[43] Prasad S, Towhata I, Chandradhara G, Nanjundaswamy P. Shaking table tests in
earthquake geotechnical engineering. Curr Sci 2004;87:1398–404.
[44] Gohl WB, Finn WDL. seismic response of single piles in shaking table studies.
In: Fifith Canadian conference earthquake engineering; 1987. p. 435–44.
[45] Valsangkar AJ, Dawe JL, Mita KA. Shake table studies of sesimic response of
single partially suppported piles. In: Sixth Canadian conference earthquake
engineering; 1991. p. 327–34.
[46] Craig RRJ, Kurdila AJ. Fundamentals of structural dynamics. 2nd ed. New
Jersey: John Wiley & Sons Inc.; 2006.
[47] Comodromos EM, Papadopoulou MC. Response evaluation of horizontally
loaded pile groups in clayey soils. Geotechnique 2012;62:329–39.
[48] Rayhani M, El Naggar M. Numerical modeling of seismic response of rigid
foundation on soft soil. Int J Geomech 2008;8:336–46.
[49] Dutta SC, Roy R. A critical review on idealization and modeling for interaction
among soil–foundation–structure system. Comput Struct 2002;80:1579–94.
[50] Spyrakos CC, Maniatakis CA, Koutromanos IA. Soil–structure interaction effects
on base-isolated buildings founded on soil stratum. Eng Struct
2009;31:729–37.
[51] Lu X, Li P, Chen B, Chen Y. Computer simulation of the dynamic layered soil–
pile–structure interaction system. Can Geotech J 2005;42:742–51.
[52] Hardin BO, Drnevich VP. Shear modulus and damping in soils: design
equations and curves. J Soil Mech Found Div 1972;98:667–92.
[53] Sun J, Golesorkhi R, Seed HB. Dynamic moduli and damping ratio for cohesive
soils. University of California Berkeley-Earthquake Engineering Research
Center; 1988.
[54] Seed HB, Idriss I. Influence of soil conditions on ground motion during
earthquakes. J Soil Mech Found Div ASCE 1969;95:99–137.
[55] Byrne PM, Naesgaard E, Seid-Karbasi M. Analysis and design of earth
structures to resist seismic soil liquefaction. In: 59th Canadian geotechnical
conference & 7th joint CGS/IAH-CNC groundwater specialty conference,
Vancouver, Canada; 2006. p. 1–24.
[56] Beaty MH. A synthesized approach for estimating liquefaction-induced
displacements of geotechnical structures. Ph.D. The University of British
Columbia (Canada), Canada; 2001.
[57] Lu J, Elgamal A, Yan L, Law K, Conte J. Large-scale numerical modeling in
geotechnical earthquake engineering. Int J Geomech 2011;11:490–503.
[58] Maheshwari BK, Watanabe H. Nonlinear dynamic behavior of pile
foundations: effects of separation at the soil–pile interface. Soils Found
2006;46:437–48.
[59] Hokmabadi AS, Fatahi B, Samali B. Recording inter-storey drifts of structures in
time-history approach for seismic design of building frames. Aust J Struct Eng
2012;13:175–9.
[60] AS1170.4. Structural design actions—earthquake actions in Australia. NSW,
Australia: Standards Australia; 2007.
[61] BSSC. NEHRP guidelines for the seismic rehabilitation of buildings, 1997
edition, Part 1: provisions and Part 2: commentary. Federal Emergency
Management Agency; 1997.
[62] ASCE7-10. Minimum design loads for buildings and other structures. American
Society of Civil Engineers; 2010.