Soil Dynamics and Earthquake Engineering 97 (2017) 184–204

Contents lists available at ScienceDirect

Soil Dynamics and Earthquake Engineering

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

Centrifuge modeling of seismic foundation-soil-foundation interaction on MARK

liquefiable sand
⁎
Y. Jafariana, , B. Mehrzadb, C.J. Leec, A.H. Haddadb
a
Geotechnical Engineering Research Center, International Institute of Earthquake Engineering and Seismology, Tehran, Iran
b
Department of Civil Engineering, Semnan University, Semnan, Iran
c
Department of Civil Engineering, National Central University, Taiwan

A R T I C L E I N F O A BS T RAC T

Keywords: Liquefaction-induced settlement and bearing capacity failure have been reported as leading causes of damages
Shallow foundation in shallow foundations during earthquakes. Previous studies of this problem have mainly focused on the
Liquefaction performance of isolated shallow foundations. In urban areas, however, foundations are generally located in close
Settlement proximity. In this study, three series of centrifuge tests were conducted to investigate the effect of foundation-
Foundation-soil-foundation interaction
soil-foundation interaction (FSFI) on the seismic and post-seismic settlement of shallow foundations on
Centrifuge modeling
saturated sand. Two rigid foundations with different surcharge loads (as heavy and light foundations) were
placed with different spacing. Multiple shaking events were applied to achieve different extents of soil
liquefaction. The results indicate that significant part of foundation settlement occurred before soil reconsolida-
tion. Furthermore, the time period after shaking, wherein excess pore water pressure sustains, plays an
important role in the total settlement of foundations. The acceleration responses experienced by the foundations
were significantly larger than those observed in the free-field. The heavy foundation fluctuated more strongly
than the light one. Moreover, adjacency considerably affected the seismic response of foundations whereas
stronger acceleration response on the ground level was observed for the closer cases. The Clear asymmetric
settlement was observed for the adjacent foundations. It is demonstrated that settlement of foundations not
only is dependent on foundations' proximity but also is a function of shaking intensity. Influence of foundations'
spacing on the generation-dissipation mechanism of excess pore water pressure (EPWP) and liquefaction extent
was described by the time-dependent contours plotted by interpolation of the recorded data.

1. Introduction Luzon Philippines earthquake, wherein many buildings experienced

Numerous catastrophic failures were reported in the past earth-
quakes in which soil liquefaction caused major damages to the super-
structures resting on shallow foundations. Earthquake-induced pore
pressure buildup and associated shear strength reduction of liquefied
sands may result in bearing capacity degradation and seismic settle-
ment accumulation of shallow foundations. Most of the damaged
structures in the past earthquakes were located close to each other in
urban areas. Considerable damages have been reported for such
foundations which suffered improper design of shallow footing on
liquefiable soils. Accordingly, numerous studies have been carried out
to assess these complicated mechanisms (e.g., [1–4]).
Approximately 340 reinforced concrete buildings were damaged by
excessive settlement and tilting as a result of liquefaction, during the
Niigata 1964 earthquake (e.g., [1,5,6]). Remarkable extents of subsoil
liquefaction were also observed in the city of Dagupan during the 1990
excessive settlement (e.g., [7–9]). Significant settlements were ob-
served in corner buildings, in building without adjacent structures in
one or both sides, in buildings surrounded by lightweight structures,
and in those part of the area where there was greater separation
between adjacent buildings [10]. Tokimatsu et al. [11] reported that
structures movements were related to foundation dimensions, confin-
ing pressure, and the shear stress imposed by the buildings and their
adjacent structures. During the Adapazari, Turkey, 1999 earthquake,
many structures were damaged by the liquefaction of shallow and
relatively thin layers of saturated sand [12].
In addition to the field studies, numerous experimental efforts
including 1g shaking table and centrifuge experiments were conducted
for a better understanding of this problem. Yoshimi and Tokimatsu [1]
studied the factors such as pore pressure development and structure’s
width, height and contact pressure which influenced the settlement of
the structure. Liu and Dobry [2] focused on soil densification and soil

⁎
Correspondence to: International Institute of Earthquake Engineering and Seismology (IIEES), P.O. Box: 19395-3913, Tehran, Iran.
E-mail addresses: yjafarianm@iiees.ac.ir (Y. Jafarian), beh_mehrzads@semnan.ac.ir (B. Mehrzad), cjleeciv@cc.ncu.edu.tw (C.J. Lee), ahaddad@semnan.ac.ir (A.H. Haddad).

http://dx.doi.org/10.1016/j.soildyn.2017.03.019
Received 7 November 2015; Received in revised form 11 March 2017; Accepted 13 March 2017
Available online 21 March 2017
0267-7261/ © 2017 Elsevier Ltd. All rights reserved.
Y. Jafarian et al.
permeability on pore pressure buildup and foundation settlement using
several centrifuge tests and reported a significant amount of negative
excess pore pressure generated beneath the shallow foundations.
Partial drainage was shown to occur simultaneously with excess pore
pressure buildup, as fast pore water pressure redistribution took place
in a three-dimensional pattern in response to transient hydraulic
gradients. Coelho et al. [13,14] reported the issue of post-earthquake
pore pressure migration as a key factor in bridge foundation settlement
during earthquake excitation. Adalier et al. [15] discussed the mechan-
isms involved in shallow foundation settlement and the influence of
stone column as liquefaction countermeasure. Dashti et al. [3,16]
centrifuge studies confirm the combined role of the shear strains
imparted by the superstructure and the post-liquefaction volumetric
strains as dominant mechanisms of total settlement. Coelho [17] and
Marques et al. [18] mentioned that the initial static shear stress
imparted by rigid foundation influences excess pore pressure history.
Mehrzad et al. [19] investigated the effect of soil permeability and
contact pressure on foundation response through the centrifuge and
numerical models and reported that settlement of foundations in-
creased with the increase of soil permeability. Entire soil profile was
liquefied during their centrifuge tests. In contrast to Dashti et al. [3]
observation, Marques et al. [18] and Mehrzad et al. [19] found that
foundations settlement continues even after shaking is stopped. The
mechanism and the amount of post-seismic settlement may be related
to the properties of soil, structure, and input motion.
All of the reviewed studies deal with liquefaction-induced settle-
ment of isolated foundations, without adjacent structures. However, in
urban areas, structures are located in close proximity which may affect
the seismic response of foundations. There are a few studies which
address the influence of adjacent structures on settlement mechanism
of shallow foundations. Mason et al. [20] examined seismic soil–
foundation–structure interaction of framed structures through centri-
fuge experiments. They found that structure-soil-structure interaction
(SSSI) can be beneficial or detrimental, depending on the earthquake
motion and the structural system. Tsukamoto et al. [4] conducted two
series of seismic 1g shaking table tests on rigid circular foundations to
examine the effects of shaking duration and the group effects of
foundations. Settlements during the shaking time remained almost
the same, regardless of the foundations spacing. The post-shaking
settlements, however, were different for different spacing. Foundations
settled often less when located close to each other. This phenomenon is
referred to as site-city effects [21]. Hayden et al. [22] conducted
centrifuge experiments to observe the performance of adjacent struc-
tures affected by liquefaction. They found that adjacent structures
tended to tilt away from one another and settled less than isolated
structures. Although, they mentioned that the proximity of foundations
would have complex effects, including the tilting of each foundation;
they did not discuss factors which influence its mechanisms.
This paper focuses on foundation-soil-foundation interaction
(FSFI) in the liquefiable sand. Three centrifuge experiments were
designed to examine the effect of foundation proximity on seismic
response and liquefaction-induced settlement of foundations. Two rigid
foundations with two different static contact pressures (representative
of light and heavy foundations) were located in three different spacing
from each other. Experimental setups for each test will be discussed in
detail. This includes soil and foundation properties, model preparation
method, scaling laws, centrifuge boundary effects, and instrumentation
arrangement. Testing procedure and the results will be fully explained
for each test series. Settlement of the both foundations (in different
spacing) and the free-field, excess pore water pressure (EPWP), and
recorded accelerations will be reported precisely during shaking and
post-shaking periods.
2. Centrifuge shaking table tests
The purpose of geotechnical centrifuge modeling is to simulate soil
185
Soil Dynamics and Earthquake Engineering 97 (2017) 184–204
systems in smaller scales. Centrifuge modeling data provides a basis for
calibration of design and computational modeling procedures [23].
Details of the centrifuge modeling in this study are described in the
following sections.
2.1. Testing equipment
The experiments were performed in the 100g-ton centrifuge at the
National Central University (NCU), Taiwan. The NCU Centrifuge has a
nominal radius of 3 m and has an in-flight 1-D servo-hydraulically
controlled shaker integrated into a swing basket to impart base
dynamic excitation in a high gravity field. The shaker has a maximum
nominal shaking force of 53.4 kN with a maximum table displacement
of 76.4 mm and operates up to the acceleration of 80g. The nominal
operating frequency range of shaking is 0–250 Hz. The table-payload
mounting area is 1000×546×500 mm (L×W×H).
A laminar container, with inner dimensions of 711×356×353 mm
(Length×Width×Height) and constructed of 38 light-weight aluminum
alloy frames, was employed for these experiments. Each frame is
8.9 mm in height, separated from adjacent rings by roller bearings.
Roller bearings were specially designed to permit translation in the
longitudinal direction with minimal frictional resistance. A relative
displacement of up to 2.5 mm between adjacent frames was possible,
for a total overall shear strain of up to 15%. Each frame had a high
lateral stiffness to maintain overall conditions of zero lateral strain and
a constant horizontal cross-section during shaking at 80g centrifugal
acceleration. The laminar container was designed for dry or saturated
soil models and permits the development of stresses and strains
associated with 1-D shear wave propagation. A flexible 0.3 mm thick
latex membrane bag was used to retain the soil and the pore fluid
within the laminar container.
2.2. Similitude laws
The main principle in centrifuge modeling is that a 1/N scale model
subjected to a gravitational acceleration of Ng (where g is acceleration of
gravity) could experience the same stress as the prototype. If the same soil
with the same density, ρ, is used in the model and in the prototype, then for
a centrifuge model subjected to an inertial acceleration field N times earth’s
gravity, the vertical stress at depth hm (the subscript m denotes the
centrifuge model) is identical to that of the corresponding prototype at
depth hp (the subscript p denotes the prototype), where hp=Nhm and the
scale factor (model: prototype) for linear dimensions is 1: N. This
relationship is the scaling law of centrifuge modeling; i.e. stress and
pressure similarities are achieved at homologous points. The key scaling
relationships for dynamic events are shown in Table 1. In fact, the scaling
factors are relationships between a prototype subjected to base shaking (the
amplitude of the base acceleration, ap, and the frequency, fp) in earth’s
gravity (1g) and the corresponding 1/N centrifuge model tested at an
acceleration of Ng. The centrifuge model is subjected to a scaled base
shaking with the acceleration amplitude of am=Nap and the frequency of
fm=Nfp. The scale factors retain the stress and pressure similarities of the
linear dimensions and base acceleration of the centrifuge model and the
prototype are 1: N and 1: N-1, respectively.
Table 1
Scaling relationships for dynamic centrifuge modeling.
Parameter Prototype Model (Ng)
Stress and pressure 1 1
Displacement 1 1/N
Velocity 1 1
Acceleration 1 N
Frequency 1 N
Time (dynamic) 1 1/N
Time (consolidation) 1 1/N2
Y. Jafarian et al.
In liquefaction problems involving the dissipation of excess pore
water pressure, time scaling is derived from the governing equation of
the consolidation problem. The rate of dissipation takes place N2 times
faster in the model than in the full scale. Thus, there is a conflict
between the scaling of inertial effects in shaking timespan and the
scaling factor associated with the dissipation of excess pore water
pressure. In centrifuge modeling, use of viscous pore fluids, which is N
times more viscous than the pure water, is a well-established method
for satisfying the scaling laws for the movement of pore fluids through
the soil during dynamic events. As the experiments were conducted in
80g centrifugal gravity, a solution of Methocel cellulose powder mixed
in water was used as the pore fluid with a viscosity of approximately 80
± 2 times greater than that of water.
2.3. Test setup
The objective of the centrifuge experiments was to examine FSFI
(foundation-soil-foundation interaction) effects on foundation re-
sponse by changing the foundations’ spacing. Moreover, the experi-
ment program was designed to observe liquefaction-induced settle-
ments of both light and heavy foundations in limited and complete
liquefaction conditions. Three tests were conducted to attain these
purposes. The model rigid foundations were rectangular steel plates
with a flat bottom. Foundations were designed and fabricated with the
same dimensions and different static contact pressures (or weights).
Each rigid foundation was 6.4 m (8 cm in model scale) in width and
9.6 m (12 cm in model scale) in length. Heavy foundation (Foundation
H) represents a nine-story building with a static contact pressure of
88.5 kPa and the light one (Foundation L) represents a three-story
building with a static contact pressure of 31.5 kPa. Rigid foundations
with low heights were employed to minimize the probable settlements
due to inertial soil-structure interaction and the resultant ratcheting.
Therefore, the complex mechanism involved in foundation response
due to SSI effect is almost canceled out in these experiments for
simplification. The soil and model foundations were instrumented with
accelerometers, displacement transducers (LVDT), and pore pressure
transducers (PPT), as shown in Fig. 1. Model scale dimensions are
presented in centimeter and the prototype scale dimensions are in
meter in the parenthesis. All of the instruments were calibrated before
each test.
In Test-1, Foundation H and Foundation L were located by a center
to center distance of 3.875B, where B is foundations' width. This
distance was chosen to maximize the spacing between the foundations
and the spacing between each foundation and the laminar box
boundaries. Schematic view of experimental model Test-1 is demon-
strated in Fig. 1(a), which includes important dimensions of the model
and locations of the instruments. Vertical arrays of accelerometers and
PPTs were placed up to a depth of 12 m under the centerline of the
foundations and across a vertical line in the middle. PPTs were also
installed to full depth of soil model in the zone between the foundations
to observe the extent of liquefaction progress during the test. With
sufficient distance to the foundations, the intermediate vertical soil
column between the foundations has been regarded as free-field in the
previous studies (e.g. [3,24]). This hypothesis, however, needs to be
verified by analysis of the experimental data because the seismic
response of the foundations may affect the intermediate region.
On the other extreme, foundations H and L were placed close to
each other in Test-2 (there was 10 cm spacing between two foundations
in prototype scale). It is an interesting question what distance from
shallow foundation can be assumed as free-field. To answer this
question, instruments (accelerometers and PPTs) were placed at
different distances from the foundation’s edge in this test. Lee et al.
[25] investigated the boundary effects of the currently used laminar
container on the seismic response of sand bed at various depths and
distances from the end walls. They concluded that the changes in the
maximum amplitudes, maximum Fourier amplitudes, and the main
186
Soil Dynamics and Earthquake Engineering 97 (2017) 184–204
frequencies of the accelerations measured at a distance of approxi-
mately 1/20th of the model container length from the end walls were
less than 7%. There was no obvious discrepancy in the time histories of
the excess pore water pressure ratio at this zone. Accordingly, PPTs
were placed in order that satisfy this criterion and as far as possible
from foundation’s edge to capture the results in free-field condition.
Fig. 1(b) represents experimental setup of Test-2 in detail; PPTs were
placed in different distances of B, 2B and 3B from the edge of
Foundation H. In the other side, one PPT and one accelerometer were
placed in a distance of B from the edge of Foundation L. In addition to
the PPTs, arrays of some accelerometers were placed at different
distances, as shown in Fig. 1(b).
In an intermediate case, Foundation H and Foundation L were
placed with a center-to-center spacing of 2B in Test-3. This test was
designed to investigate how an intermediate foundations spacing can
affect their seismic response and settlement. Details of setup and
instrumentations of Test-3 are shown in Fig. 1(c).
2.4. Soil properties and model preparation
A kind of fine crushed quartz sand known as the No. 306 sand was
used to prepare the sand beds within the laminar container. The grains
size distribution of the sand is shown in Fig. 2. It is classified as poorly
graded sand, SP, according to the Unified Soil Classification System
(USCS). Its specific gravity (Gs), effective grain size (D10), and mean
grain size (D50), are 2.65, 0.147 mm, and 0.193 mm, respectively, as
summarized in Table 2. The maximum and minimum dry unit weights,
which were determined according to the JGS-0161 method, are equal
to 16.3 kN/m3 and 13.5 kN/m3, respectively.
The quartz sand was pluviated along a regular path into the
container to prepare a fairly uniform sand bed with a relative density
of 55%. The sand was dropped at a constant flow rate using a sand
rainer. The air pluviation process was interrupted as needed to embed
the accelerometers and the pore water pressure transducers (PPTs) at
the specified points. After completing the sand bed preparation
(300 mm height in the model scale, 24 m depth in the prototype scale),
the package was mounted on the shaker. An acrylic plate was used to
tightly cover the container for saturation of the sand model. Air was
simultaneously and continuously vacuumed from both the inner and
outer chambers of the container. Simultaneously, de-aired fluid (water-
Methocel cellulose solution with a viscosity 80 ± 2 times greater than
the water viscosity) was carefully dripped into the container to saturate
the sand bed until the water level reached 2 mm above the sand
surface. This process required about 36 h. The acrylic plate was then
removed and the LVDTs were attached to a mounting unit in order to
measure settlements of the ground surface and foundations. The
saturated sand model was tested at the centrifugal acceleration of
80g. A sample photograph of the model is presented in Fig. 3; this is
the model Test-1 before spinning.
2.5. Input motions and testing conditions
After saturation, the instruments were connected to the data
acquisition system, and the centrifuge was accelerated up to 80g with
steps of 10g. The model was maintained at each g-level for 5 min to
ensure that the sand bed was fully consolidated at each overburden
stress. The test conditions listed in Table 3 were scheduled in order to
study the major factors contributing to the seismic response of adjacent
foundations (i.e., FSFI effects) which are subjected to the dynamic base
excitation of sand deposit.
Each model was shaken one-dimensionally and horizontally along
its longer axis using an electro-hydraulic shaker. Before the tests, the
model sand bed was subjected to pre-shaking driven by the shaker to
measure the profile of the shear wave velocity (Vs) and natural
frequency of sand bed through the vertical accelerometer array.
Shear wave velocity was estimated through the time delay measured
Y. Jafarian et al.
Fig. 1. Centrifuge test setup with soil instrumentation for (a) Test-1, (b) Test-2 and (c) Test-3. Model scale in cm (prototype scale in m).
by each adjacent accelerometer. During pre-shaking event, shaker fired
one cycle of small amplitude sinusoidal wave (around 0.01g in
amplitude and 2 Hz in frequency, both in the prototype scale). A 1-D
shear beam approach estimated the corresponding maximum strain
levels on the first cycle to be 0.0008–0.002% from the bottom to the top
of the sand bed. Average shear wave velocity for the model profile was
191 m/s, hence, with a profile depth of 24 m, natural frequency can be
estimated by:
Vs
f=
4H
where f is model’s natural frequency, H is the depth of model sand and
Soil Dynamics and Earthquake Engineering 97 (2017) 184–204
Vs is average shear wave velocity of the sand profile. Therefore, the
model profile has a natural frequency of 1.81 Hz. The models were
shaken multi times in succession with 15-cycle sinusoidal base input
motions of constant amplitudes and 2 Hz frequency (in prototype
scale). An equivalent uniform earthquake motion is considered to
represent the corresponding irregular earthquake motion using reason-
able amplitude, equivalent cycle number, and proper predominant
frequency. A single equivalent uniform earthquake motion might be
representative of a great number of irregular earthquake motions. In
this way, centrifuge modeling test results are presented in simpler
(1)
patterns and interpretation of seismic behavior is more straightfor-
ward. Ample time was allowed between each stage of shaking in order
187
Y. Jafarian et al. Soil Dynamics and Earthquake Engineering 97 (2017) 184–204

100 Table 3
Summary of centrifuge test conditions.
Percent finer by weight (%)

80 Test No. Shaking Base acc./ Pore Dr (%)/eb Foundation

event frequency fluida spacing (center-
/number of to-center) (×B)c
60 cycles (g/Hz/
cycle)

Test-1 Pre- 0.01/2/1 MCE 55/0.7421 3.875

40 shaking 0.04/2/15 MCE 55/0.7421
E1 0.07/2/15 MCE 57/0.7352
E2 0.17/2/15 MCE 58/0.7319
20 E3

Test-2 Pre- 0.01/2/1 MCE 55/0.7421 1

0 shaking 0.01/2/15 MCE 55/0.7421
0.01 0.1 1 10 E1 0.02/2/15 MCE 55/0.7418
E2 0.03/2/15 MCE 57/0.7354
Grain size (mm)
E3 0.06/2/15 MCE 59/0.7301
Fig. 2. Grains size distribution of No. 306 quartz sand. E4 0.18/2/15 MCE 61/0.7238
E5

Test-3 Pre- 0.01/2/1 MCE 55/0.7421 2

shaking 0.01/2/15 MCE 55/0.7421
Table 2
E1 0.02/2/15 MCE 55/0.7421
Soil properties of No. 306 fine quartz sand.
E2 0.03/2/15 MCE 57/0.7354
E3 0.07/2/15 MCE 59/0.7287
Gs 2.65
E4 0.18/2/15 MCE 63/0.7164
Dr (%) 55
E5
D50 (mm) 0.193
D10 (mm) 0.147 a
MCE=a water-Methocel cellulose solution with a viscosity of 80 times the viscosity of
γmax 16.3
water.
γmin 13.5 b
Relative density of the tested sand bed before each shaking event.
Permeability (cm/s) 7.47×10-3 c
Foundations center-to-center distance.
Void ratio, e 0.742

3. Test results and discussion

3.1. Free-field response

The liquefaction and post-liquefaction behaviors of sand deposits

Fig. 3. Photograph of Test-1 model prepared for the spinning stage. Laminar box,
LVDTs, foundations plate, and colored sand are observed in the figure. (For interpreta-
tion of the references to color in this figure, the reader is referred to the web version of
this article.)
are basically governed by the generation and dissipation of excess pore
water pressure. The presence of two foundation models in the contain-
er may result in 3D stress non-homogeneities and lateral water flow
especially in Test-1 and Test-3 that may influence the free-field
response. Other researchers (e.g., [3,17,24]) have noted the stress
non-homogeneities around the superstructure during their centrifuge
experiments. Nevertheless, they reported the data recorded at the
distance of approximately 2B from foundation center as the free-field
response. The excess pore water pressure results, presented herein as
the free-field response, are the results of Test-2 at the vertical array of
instruments far from the foundations (see Fig. 1(b)). It seems that the
adjacent foundations have a minor effect on the vertical array of PPTs
near the container boundary (i.e. P1, P2, P3 and, P4). In the other
words, the FSI (foundation-soil interaction) and boundary effects seem
to be relatively negligible in this zone. Excess pore water pressure
results are shown in Fig. 4(a) at various depths of the free-field for
different shaking events. These figures are drawn in semi-logarithmic
plots in order to clearly demonstrate generation and dissipation of pore
water pressure during the shaking and the post-shaking timespans; the
latter time is considerably more. The shaded section shown in Fig. 4(a)
indicates the duration of the base shaking. Triggering of liquefaction is
determined by excess pore water pressure ratio. This parameter is
defined as the measured excess pore water pressure, Δu, divided by the
corresponding effective overburden pressure, σ'v0:
∆u
ru =
σ′v0 (2)

to monitor full EPWP dissipation. Time histories of acceleration, pore The “complete liquefaction” term is referred to the condition in
water pressure at various elevations, and surface settlement were which ru is equal to one and ru values less than one denote “limited
recorded simultaneously. The measurements in the following sections liquefaction”.
are presented in prototype units unless otherwise specifically noted. Full depth of free-field sand profile was liquefied in the strongest

188
Y. Jafarian et al. Soil Dynamics and Earthquake Engineering 97 (2017) 184–204

Fig. 4. (a) Free-field EPWP time histories, hollow circles indicate the time at which solidification commences at a specific depth; (b) settlement time histories for Test-2.

189
Y. Jafarian et al.
shaking event (i.e. Test-2 E5 as listed in Table 3 and shown in
Fig. 4(a)). The upper layers were liquefied at a very early stage and
then the liquefaction front propagated rapidly from a depth of 2.4 m
(where the shallowest PPTs were located) downward to the full depth of
soil profile (as shown in Fig. 4(a), E5 Event). Less than 4 s was required
for the whole model to be liquefied. The entire sand profile then
remained liquefied for the rest of shaking period (i.e., up to 7.5 s) and
the superficial sands remained in liquefaction state for about 200 s
after the end of shaking. EPWP began to decrease from the bottom of
the container and the solidification process was consequently propa-
gated upward during which the grains sink and the shear strength is
restored.
The theory of sinking sand grains to simulate dissipation of EPWP
was proposed by Florin and Ivanov [26]. This theory was modified
through the consolidation theory by Scott [27] and Kim et al. [28]. The
boundary between the upper liquefied region and the lower solidified
region is termed as the solidification front [26,27]. The solidification
front was determined as the point in each time history at which EPWP
in the liquefied deposit began to decrease. These points are shown as
hollow circles in Fig. 4(a) and indicate the time at which solidification
commences at a specific depth. It can be seen that the solidification
front moves upward from a depth of 24 m with a velocity of
approximately 89.4 mm/s as the dissipation of EPWP progresses.
EPWP was fully dissipated in approximately 1000 s at the end of
reconsolidation phase, as shown in Fig. 4(a), E5 Event.
The soil was liquefied up to the level of 7.2 m in event E4, which is
considered as a relatively strong event. Although the lower layers were
not completely liquefied in this event, they experienced limited
liquefaction with relatively high excess pore water pressure ratios of
about 0.7. Once the base excitation ceased, EPWP began to decrease in
7.2 m and deeper layers, but soil remained liquefied at depth of 2.4 m
for about 50 s after the end of shaking. EPWP was fully dissipated in
approximately 500 s in event E4. The EPWP time histories were
relatively the same for the weaker events. Near the ground surface at
depth 2.4 m, sand remained liquefied in less time under events E3 and
E2, compared to the two strongest events. The durations of liquefaction
state were about 20 and 10 s for the events E3 and E2, respectively.
During these events, limited liquefaction was observed in deeper layers
with relatively low excess pore water pressure ratios.
The acceleration time histories recorded in the free-field for E5, E4
and E3 in Test-2 are shown in Fig. 5(a–c). It is apparent that
liquefaction caused severe deterioration of soil stiffness in various
events, especially in the strongest one (E5). The arrows shown in the
plots indicate the instant at which the soil is liquefied (based on the
EPWP records at the corresponding depth). Softening of underlying
soil resulted in the significant decay of accelerations at different levels
when a few cycles of input shaking passed. Eventually, motions fully
decayed due to liquefaction of underlying sand. The rate and extent of
soil softening and timing of liquefaction were greatly influenced by
shaking intensity. Liquefaction occurred in upper layers firstly and then
propagated downwards, eventually affecting the entire sand stratum.
On the other hand, ground acceleration not only did not decay during
shaking in the weakest event (E1) but also amplified with the
amplification factor of about 2. Moreover, high values of EPWP were
not developed during this event. Hence, in the weak shaking events,
degradation of soil stiffness was not considerable for preventing
transmission of shear waves to the upper layers.
Fig. 4(b) shows semi-logarithmic settlement time histories of the
free-field ground surface (measured by LVDT1 in Test-2) for five
shaking events (E1–E5). Although foundations may affect LVDT1,
their interaction with this instrument is minor in Test-2 compared with
Test-1 and Test-3; hence, their results are reported herein as a free-
field response. Free-field settlement commenced immediately after the
onset of liquefaction and continued until excitation ceased. The amount
of the settlement observed during the shaking decreased with the
increase of shaking intensity. Soil behavior during the shaking time of
190
Soil Dynamics and Earthquake Engineering 97 (2017) 184–204
the stronger event is more likely to be undrained in which most of the
settlement occurs during reconsolidation phase. Ground surface heave
was observed during shaking in the strongest shaking (E5). Hausler
[24] also reported ground heave in the so-called free-field ground
located between two foundations. The observed ground heave can be
related to lateral movements of soil because of partial bearing capacity
failure, but this is not the case for the free-field settlement measured by
LVDT1 (in Test-2) which was placed relatively far from the founda-
tions. In the free-field, the settlement was negligible during the time
period in which the sand was liquefied and EPWP was sustained.
Thereafter, EPWP started to dissipate and the solidification front
moved from the bottom of the container towards the ground surface.
The upward flow of fluid caused seepage-induced liquefaction in the
sand; thus, there was a delay in EPWP dissipation in the upper layers.
In fact, the bottom layers reconsolidated (or solidified) and settled first
and the upper layers settled after a while. Total settlement of the
ground surface was dependent to shaking intensity whereas more
settlement occurred by the stronger shaking events. The main part of
the total settlement in the free-field occurred when EPWP dissipation
and the consequent reconsolidation mechanisms were in progress. This
part of the settlement was more considerable for the stronger shaking
events during which more excess pore pressures were generated. In the
two strongest events (i.e., E5 and E4), in which large amounts of sand
were liquefied, about 70% of total settlement occurred during the
reconsolidation phase.
It should be noticed that sand’s relative density changed in each
stage of loading. Ground surface and foundations experienced perma-
nent settlements in loading stages. Therefore, the relative density of
sand in each stage was higher than that in the previous stage. The
subsequent increase of relative density per loading stage may change
soil behavior and liquefaction-induced settlements. Furthermore, it will
be discussed in the subsequent sections of the paper that shallow
foundations cause non-homogeneities in the sand fabric due to the
differential settlement between the foundation and the free-field
ground. These features must be considered in data analysis and future
numerical simulations. However, change in relative density was
negligible in weak shaking events and it is also reasonable to assume
the sand model as homogeneous media in the moderate shaking events.
3.2. Test-1 (response of foundations)
Test-1 involves two foundations with the center to center spacing of
3.875B. Three sinusoidal input motions were imparted to the model
with the peak base shaking amplitude of 0.04 g, 0.07 g, and 0.17 g as
events E1, E2, and E3, respectively. The time histories of EPWP
observed under Foundation L and Foundation H along with their
settlement time histories in the Test-1 are shown in Fig. 6(a–c).
Unfortunately, PPT10 which was located under the light foundation
at the depth of 7.2 m malfunctioned in Test-1; and thus, its data were
not plotted in this figure.
Fig. 6(a and b) indicate that similar EPWP behavior is observed
under both foundations and in the free-field, at the depth of 12 m;
indicating that foundations’ impact was negligible at this level and soil
behavior can be assumed as free-field for the deeper levels. Like the
free-field response, EPWP time histories can be divided into three
phases: EPWP buildup during shaking (Phase I), the timespan in which
EPWP is sustained (Phase II), and the timespan in which EPWP
dissipates (Phase III). Following the shaking time, there is a time
period in which the underlying sand remains in the liquefied state. The
stronger the input motion, the longer the sustained EPWP time period.
Tsukamoto et al. [4] also noted that the time period after shaking at
which high excess pore pressure is maintained tends to increase with
the duration of shaking.
No liquefaction occurred beneath the foundations during shaking
events E1, E2, and E3. Although soil beneath the light foundation
wasn’t liquefied during shaking, it experienced a generation of large
Y. Jafarian et al.
Fig. 5. Variations of acceleration time histories for events (a) E3, (b) E4 and (c) E5 in Test-2.
positive EPWP during different input motions. Average value of ru
under Foundation L was ranging from 0.5 to 0.7 for the three events at
the depth of 2.4 m. On the other hand, large negative EPWP was
generated under foundation H (at 2.4 m depth) during shaking. EPWP
behavior was affected by the heavy footing up to the depth of 7.2 m
(approximately equal to the foundation width). Average value of ru in
the depths of 2.4 m and 7.2 m under the heavy foundation centerline
were −0.1 and 0.3, respectively. Negative excess pore pressure within
the soil beneath the foundation has also been reported in the recent
centrifuge and 1g shaking table studies (e.g. [1,2,15]). Compared with
the free-field condition, the higher confining pressure and the devia-
toric stress applied by foundation surcharge are two beneficial compo-
nents which prevent pore water pressure buildup during shaking [29].
Furthermore, negative excess pore water pressure can be explained by
the dilative response of soil under the foundation. The soil under and
around the foundation tend to move laterally away from the footing
midline due to large shear stress imposed in these regions. This large
deformation (or shear strain) dictates the soil to dilate and the EPWP
would consequently reduce. Away (vertically and horizontally) from
these zones, significant positive EPWP was produced in the free-field
with ru values of about 1. Immediately after excitation ceased, vertical
and horizontal hydraulic gradient was developed and surrounding
water flowed towards the foundations. After a relatively long time,
large positive EPWP was generated under the foundations due to
seepage. Once the water pressure equalized in each level and horizontal
hydraulic gradients vanished, the soil began to reconsolidate starting
from the container base towards the soil surface.
Fig. 7(a and b) shows acceleration time histories for event E3
recorded by the vertical array of accelerometers placed under
Foundation L and Foundation H. The dashed line curves in Fig. 7
represent input base motion. The amplitude of acceleration measured
at the depth of 7.2 m decreased significantly after the third cycle, but
wave amplification was observed in the upper layers, especially beneath
the heavy footing. As discussed before, the layers deeper than 7.2 m
seems to behave like the free-field. The pore pressure builds up at these
191
Soil Dynamics and Earthquake Engineering 97 (2017) 184–204
levels and soil softening causes acceleration de-amplification and
doesn’t permit shear waves to propagate towards the upper layers.
Since negative pore pressure was generated under foundations due to
the dilative behavior of soil, these zones don’t soften during shaking
and have larger shear modulus compared with the deeper layers. Thus,
the weakened vertical waves amplified when passing through the upper
stiffer layers (beneath the foundations). The heavy foundation experi-
enced larger amplitudes of acceleration than the light one. Acceleration
time histories under the light footing were more similar to those
measured in the free-field, compared with the heavy one. In the other
words, the light surcharge has a minor impact on the underlying soil
behavior. Major accelerations were completely vanished under the light
foundation, the same as the free-field response. In contrast to the large
events, de-amplification was not observed in the weaker events. The
extent of soil softening was minor in the weaker events (E1 and E2) and
the shear waves could propagate easily to the upper layers with
amplification. The heavy and the light footings oscillated in the same
phase during events E1 and E2 and amplification occurred from
container base towards the ground surface.
Settlement time histories of light and heavy foundations in Test-1
are shown in Fig. 6(c). Foundations settlement commenced immedi-
ately after the first cycle of shaking and continued with time.
Foundations continued to settle, even after excitation ceased, until
the end of reconsolidation in the upper layers. The settlements
measured during and after the shaking timespan increased with the
increase of shaking intensity. The rate of settlement accumulation
decreased after shaking ceased. Negligible settlement of foundation
occurred during soil reconsolidation. Such behavior was observed and
reported in the recent centrifuge experiments (e.g, [4]). Dashti et al. [3]
discussed in detail what mechanisms of foundations settlement can be
dominant during and after shaking. Partial drainage and inertial force
imposed by foundation surcharge seem to be two dominant mechan-
isms of foundation settlement during shaking.
Results of the first experiment (i.e., Test-1 as shown in Fig. 6(a and
b)) indicate that pore water pressure under the foundations increases
Y. Jafarian et al. Soil Dynamics and Earthquake Engineering 97 (2017) 184–204

Fig. 6. EPWP time histories (a) under Foundation L, (b) under Foundation H, and (c) settlement time histories measured during events E1, E2, and E3 for Test-1.

Fig. 7. Acceleration time histories under (a) Foundation L (b) Foundation H for event E3 of Test-1 (The dashed-line curves represent input base motion).

192
Y. Jafarian et al.
even after the shaking phase by seepage process. Hence, sand is
softened in the second phase due to pore pressure buildup and
foundations settle under their own weight in this phase. Shear
deformations due to partial bearing capacity failure seem to be
dominant mechanism in this phase.
Dashti et al. [3] 's centrifuge studies showed that about 98 ± 1% of
total foundations settlement occurs during shaking; however, in the
current centrifuge experiments, a significant part of foundation settle-
ment occurred after shaking and prior to the reconsolidation phase. It
should be noted that Dashti et al. [3] employed structures with more
realistic fundamental frequencies and inertial forces on a thin non-
liquefiable sand layer which may affect soil-structure-interaction and
total settlement, subsequently. However, fundamental period of super-
structure barely affects the post-seismic settlement. In general, that
part of settlement associated with the post-shaking mechanisms is a
matter of utmost consideration, especially when the liquefied layer is
thick. Depends on the shaking intensity, the thickness of the liquefied
layer, and foundation weight, settlements in the second phase of the
current test are ranged between 20% and 35% of the total settlement.
Tsukamoto et al. [4] mentioned a distinct difference between the
settlement mechanisms during the shaking and the time period in
which fully developed EPWP are sustained. The amount of hydraulic
gradient, the extent of liquefaction (thickness of the liquefied layer),
soil permeability, and foundation weight may control foundations
settlement during this phase. Approximately no settlement was ob-
served after EPWP started to be dissipated, during the reconsolidation
phase. The foundation and free-field settlements in each phase are
presented in Table 4. It should be noted that relative densities changed
after each stage of the test. The approximate values of relative densities
for each test are presented in Table 3. As can be seen in Table 4, most of
the settlement of foundations was observed in the first two phases.
Approximately, 60, 30, and 10 percentages of the total foundation
settlements occurred during shaking (Phase I), Phase II, and Phase III,
respectively. Most of the free-field ground settlement occurred during
Phase I and Phase III in each event. More than 90% of free-field
settlement in the strongest event, by which the whole model was
liquefied, occurred in reconsolidation phase (i.e., Phase III). Therefore,
there is a clear difference between settlement mechanism of free-field
and foundations. Total settlement of Foundation H, Foundation L, and
free-field in the strongest event was 25.8, 15.8, and 5.6 cm, respec-
tively.
3.3. Test-2 (foundation response)
Two foundations were located close to each other in Test-2 to study
the interaction of adjacent shallow foundations rested on liquefiable
soil (see Fig. 1(b)). This centrifuge model was excited by five sinusoidal
input motions with peak base shaking amplitude of 0.01g, 0.02g,
0.03g, 0.06g, and 0.18g as events E1, E2, E3, E4, and E5, respectively.
The time histories of EPWP observed under Foundation L and
Table 4
Foundation and free-field settlements of Test-1 (prototype).
Event Max acc. (g) ru Foundation H cm (%)
Phase I Phase II Phase III
E3 0.18 1 13.2 (51) 8.7 (34) 3.9 (15)
E2 0.07 1 13.8 (57) 9.0 (37) 1.5
(6)
E1 0.04 0.85 15.1 (73) 4.6 (22) 1.1
(5)
Phase I: during shaking.
Phase II: after shaking and before reconsolidation.
Phase III: after reconsolidation.
193
Soil Dynamics and Earthquake Engineering 97 (2017) 184–204
Foundation H along with their settlement time histories are shown in
Fig. 8(a–c).
In Test-2, pore pressure generation and dissipation were the same
as those of Test-1 and can be divided into three phases: (I) during
shaking, (II) time period during which EPWP sustains and (III)
reconsolidation. EPWP under the both (L and H) foundations were
smaller than that observed in the free-field, but mechanisms of EPWP
generation during shaking were similar under Foundation L and
Foundation H; specifically it was almost identical to the largest shaking
event (E5). This is in contrast to the results of Test-1 in which
mechanism of pore pressure buildup was significantly different under
the widely spaced foundations (see results of Test-1 in Fig. 6(a and b)).
After shaking ceased, EPWP increased under the heavy footing in
different shaking events because of vertical and horizontal hydraulic
gradients and the resultant water flow towards the foundations. If the
water inflow is larger than the water outflow, EPWP will increase and
vice versa. EPWP was equalized under the heavy and light foundations
by about 10 s after shaking ceased and then reconsolidation phase
started with upward seepage.
The free-field response is often used as a benchmark for the near-
field response. Free-field earthquake motion is defined as a motion
recorded at least at a distance that is an order of magnitude larger than
the characteristic dimension of the structure-of-interest [30]. Such
criteria cannot be satisfied by the available dimensions of the NCU
centrifuge container. It is an interesting question how far from the
structure can be fairly assumed as the free-field response in centrifuge
studies. Mason et al. [20] centrifuge studies on seismic soil–founda-
tion–structure interaction in dry sand revealed that acceleration
records at a distance equal to foundation width from the foundation
edge can be adequately judged as free-field motion. Other centrifuge
studies on the performance of shallow foundation on liquefied sand
assumed a distance of about two times of foundation width from the
foundation center as free-field results (e.g. [3,24]). Fig. 9(a and b)
illustrate the EPWP time histories and acceleration response spectra
from PPTs and accelerometers located at three different points (i.e., P1,
P7, and P11) at the same elevation (i.e., 2.4 m) in Test-2. As shown in
Fig. 9, the three free-field records are very similar for event E4. The
same negligible difference was observed for the rest events, not
presented in Fig. 9. It is found that these records at the distances
larger than B (i.e., foundation width) relative to the foundation's edge
were not affected by dynamic response of the foundation. Hence, all
three records seem to be adequate, if not perfect, representative of the
free-field response.
Acceleration time histories under Foundation L and Foundation H
for the strongest shaking event in Test-2 are shown in Fig. 10(a and b).
It is noting from the free-field test results (in Test-1) that acceleration
decayed (or even vanished) after one cycle in the free-field which is the
sign of liquefaction in the underlying layers. Acceleration amplitudes
under the foundations decayed after five cycles, but not completely
vanished because complete liquefaction did not occur under the both
Free-field cm (%) Foundation L cm (%)
Phase I Phase II Phase III Phase I Phase II Phase III
0.2 0 5.4 (96) 7.2 (46) 5.2 (33) 3.4 (21)
(04) (0)
1.9 (34) 0.2 3.5 (62) 132 (65) 5.2 (25) 2.0 (10)
(4)
8.2 (69) 0.9 2.8 (23) 13.1 (80) 2.9 (18) 0.3
(8) (2)
Y. Jafarian et al. Soil Dynamics and Earthquake Engineering 97 (2017) 184–204

Fig. 8. EPWP time histories under (a) light foundation, (b) heavy foundation, and (c) settlement time histories of Test-2.

194
Y. Jafarian et al.
Fig. 9. (a) EPWP time histories (b) Acceleration response spectra (5% damped) of horizontal ground motions recorded at a depth of 2.4 m in different locations (i.e. light side (LS),
heavy side (HS), and free-field (FF)) during event E4 for Test-2.
(light and heavy) foundations. Acceleration time histories at levels of
2.4 m and 7.2 m under the heavy and light footings are mostly the
same, as seen in Fig. 10(a and b). It means that the materials under
these two footings behaved like a stiff block. Moreover, the adjacent
light and heavy foundations oscillated in the same phase and the same
amplitude like a mono-foundation.
Settlement time histories of the adjacent light and heavy
foundations in Test-2 are plotted in Fig. 8(c). Settlement of both
foundations commenced immediately after the first cycle of input
motion and accumulated with time, resembling Test-1. The rate of
settlement accumulation was relatively high in the first phase (i.e.,
2.67 cm/s during the shaking). The settlement rate decreased to
0.09 cm/s once shaking ceased. Settlement accumulation continued
during the second phase wherein EPWP remained constant and
sand was completely liquefied. Accumulation of foundations set-
tlement approximately stopped once reconsolidation started in the
third phase. Settlement magnitude of both foundations in the first
phase was about the same for different shaking events; they settled
about 12 cm in this phase. Settlement magnitude in the second
phase increased with the increase of shaking intensity. It seems
that extent of liquefaction affected the magnitude of foundation
settlement in the second phase. In the other words, the deeper the
Fig. 10. Acceleration time histories under (a) Foundation L, (b) Foundation H for event E5 of Test-2. (The dashed-line curves represent input base motion).
195
Soil Dynamics and Earthquake Engineering 97 (2017) 184–204
liquefied depth was, the larger the foundation settlement was
observed. Amounts of foundations settlement per different phases
in Test-2 together with the free-field settlement are presented in
Table 5. Most of the foundation settlement happened in the first
and the second phases wherein shear deformations seem to be the
dominant mechanism. On the other hand, most of the free-field
settlement occurred during the first and the third phases by
volumetric deformations as the dominant mechanism. It is notable
that both heavy and light foundations settled approximately with
the same magnitude in each phase. The data presented in Table 5
provide valuable insights into the mechanism of the liquefaction-
induced settlement of shallow foundations and free-field ground.
Foundation behavior could be assumed independent of each other
in Test-1, but there is a clear interaction between the adjacent (light
and heavy) foundations in Test-2. There was a distinct rotation in
Foundation H during Test-2. Schematic view of the initial and final
locations of the heavy and light foundations is shown in Fig. 11 for
comparison. The heavy foundation settled more in the free-field side
than in the light foundation side and experienced a clockwise rotation
of about 4 degrees. Instantaneous rotation of the foundations was
impossible to be measured because the number of the available LVDTs
for each test was limited. The foundations were placed horizontally
Y. Jafarian et al. Soil Dynamics and Earthquake Engineering 97 (2017) 184–204

Table 5
Foundation and free-field settlements of Test-2 (prototype).

Event Max acc. (g) ru Foundation H cm (%) Free-field cm (%) Foundation L cm (%)

Phase I Phase II Phase III Phase I Phase II Phase III Phase I Phase II Phase III

E5 0.17 1 15.2 14.6 8.6 2.5 1.5 7.8 17.7 11.8 5.9
(40) (38) (22) (21) (13) (66) (50) (33) (17)
E4 0.05 0.9 10.7 3.5 0.4 1.1 0.4 3.2 10.9 4.0 0.6
(73) (24) (3) (23) (9) (68) (70) (26) (4)
E3 0.03 0.7 9.5 1.9 0.2 2.5 0.7 1.7 9.1 1.3 0.4
(82) (16) (2) (51) (14) (35) (84) (12) (4)
E2 0.02 0.4 10.0 1.7 0.3 3.8 0.5 0.7 11.1 1.1 0
(83) (14) (3) (76) (10) (14) (91) (9) (0)
E1 0.01 0.15 0 0.1 0 1.0 0 0 0.5 0 0
(100) (0) (0) (100) (0) (0) (100) (0) (0)

Phase I: during shaking.

Phase II: after shaking and before reconsolidation.
Phase III: after reconsolidation.

before the tests through a precise manner and their final rotation was
measured after the test by changing the LVDT location across the
foundation width.
Hayden et al. [22] reported the same results in centrifuge modeling.
Two different boundary conditions can be assumed for each side of
foundations: restrained boundary in foundation side and free boundary
in the free-field side. The restrained side has more stiffness due to its
more confinement compared with the free side. Moreover, EPWP time
histories indicate that net pore water pressure and excess pore water
pressure ratios are higher in the free side; and thus, this side
experienced more strain softening compared with the restrained side.
The stiffer soil within the restrained side impedes the development of
bearing capacity failure. Under the heavy foundation, sand tended to
move laterally towards the free side and more settlement in this side
was recorded. Thus, the heavy foundation rotated clockwise towards
the free side. These observations indicate important implications for
the seismic response of the adjacent foundations. Results of Test-2
demonstrate that unbalanced restraint conditions of the shallow
foundations may cause excessive differential settlements which may
affect the stability of superstructure. The presence of adjacent founda-
tions and FSFI effects are commonly ignored in design and analysis of
shallow footings. Such effects are more complicated in seismic urban
areas wherein neighbor foundations are located on the saturated sand
with high liquefaction potential.
Fig. 11. Schematic view of (a) the initial and final locations of the heavy and the light
foundations (b) definition of free and restrained boundaries.
3.4. Test-3 (foundation response)
In the third test, two rigid foundations were located with the center-
to-center spacing of 2B (see Fig. 1(c)). It is expected that seismic
behavior of the foundations is affected by interaction. Fifteen cycles of
sinusoidal shaking were imparted to the model base with peak
accelerations of 0.01g, 0.02g, 0.03g, 0.07g, and 0.18g as events E1,
E2, E3, E4, and E5, respectively. Fig. 12 displays the EPWP and vertical
settlements of foundations recorded per different events of Test-3.
Unfortunately, LVDT1 which was located in the free-field point
malfunctioned during this test; and thus, its data wasn’t presented in
this figure. The arrangement of instruments for Test-3 is shown in
Fig. 1(c).
As can be seen in Fig. 12(a and b), extents of negative EPWP
buildup were observed under both foundations in Test-3. However,
away from these zones, significant positive EPWP was recorded. The ru
values at the locations away horizontally from the foundations were
near one in the strong shaking events. Although high values of ru
generated, complete liquefaction was not observed in the space
between two foundations and the depth of 2.4 m during shaking.
There was a very large hydraulic gradient between this zone and the
zone under foundations, especially at shallower levels.
The acceleration time histories under the light and heavy founda-
tions are shown in Fig. 13(a and b). The foundations and their
underlying soil have shown near the same seismic responses per
different shaking events and different levels. Similar to Test-2, it seems
that foundations are located on a very stiff block. Foundations and their
underlying soil vibrated in the same phase and amplitude in each
shaking level. Acceleration amplitude didn’t vanish completely at the
depth of 2.4 m; it means that complete liquefaction didn’t occur in this
zone. It can be attributed to the confinement effect of two neighbor
foundations. Stiffer sands underlying the foundations behave like two
stiff columns that restrain the intermediate sands and prevent EPWP
buildup in this region. Nevertheless, ru values were relatively high
during the large shaking events in this zone.
Except for the strongest event (i.e., E5) in which the heavy
foundation considerably settled more than the light foundation, the
foundations experienced same settlements during the other shaking
events (E1-E4). They settled approximately equal in the first phase, but
the heavy foundation settled more than the light one during the second
phase, especially in the strongest intensity. The difference between the
settlements of the (light and heavy) foundations increased with the
increase of shaking intensity. The amounts of foundation settlement
during different phases in Test-3 are presented in Table 6. Similar to
the previous tests, most parts of the total settlement occurred during
the first and second phases. About one-half of foundation settlement
occurred during the shaking in the first phase and the rest occurred

196
Y. Jafarian et al. Soil Dynamics and Earthquake Engineering 97 (2017) 184–204

Fig. 12. EPWP time histories under (a) Foundation L, (b) Foundation H, and (c) settlement time histories of Test-3.

197
Y. Jafarian et al. Soil Dynamics and Earthquake Engineering 97 (2017) 184–204

Fig. 13. Acceleration time histories under (a) Foundation L, (b) Foundation H for event E5 of Test-3. (The dashed-line curves represent input base motion).

during the second phase in event E5. In contrast to Test-2, no clear
tilting of the foundations was observed during Test-3.
4. Effect of adjacency on acceleration response
individuals, was significantly lower than Test-2 and Test-3 where
foundations produced more interaction with each other. Foundations
response for Test-2 and Test-3 was similar in shape and magnitude.
Generally, it appears that foundations adjacency increases spectral
accelerations of the foundations for most period ranges.

Fig. 14 shows acceleration response spectra (in 5% damping) for

event E5 (i.e. the strongest one) which were calculated from the
acceleration records in the base, foundation, and free-field. As there
was no accelerometer located on the soil surface, the free-field response
is referred to the recorded data at the depth of 2.4 m. All of the tests
have similar acceleration spectra with a peak corresponding to the
period of about 0.5 s which was the initial periods of the system in
these centrifuge tests. The initial system period was measured in the
early stage of loading. Foundations peak acceleration was significantly
lower than the base shaking. The free-field responses were considerably
affected by liquefaction of the underlying sand; and thus, possessed
very small acceleration responses. The heavier foundation experienced
higher acceleration compared with the lighter one. Acceleration
response of foundations for Test-1, wherein foundations reacted as
5. Visualization of excess pore water pressure variations
during and after shaking
The use of animation can reveal important time-varying informa-
tion in complicated physical systems [31]. The animation of the excess
pore water pressure ratio of a tested model could be very helpful for
visualizing the progress of liquefaction and EPWP changes under the
foundations during and after shaking. The field of the excess pore water
pressure ratio was estimated by using linear interpolations among the
measurements of the vertical array of pore water pressure transducers
shown in Fig. 1(a–c).
Figs. 15–17 illustrate contours of the excess pore water pressure
ratio at selected times from the visual animations of the whole time

Table 6
Foundation and free-field settlements of Test-3 (prototype).

Event Max acc. (g) ru Foundation H cm (%) Free-field cm (%) Foundation L cm (%)

Phase I Phase II Phase III Phase I Phase II Phase III Phase I Phase II Phase III

E5 0.17 1 14.7 11.4 1.7 N/A N/A N/A 9.1 6.9 2.5
(53) (41) (6) (49) (37) (14)
E4 0.05 0.9 15.3 5.3 0.6 N/A N/A N/A 14.4 3.6 0.6
(72) (25) (3) (77) (19) (4)
E3 0.03 0.7 10.0 1.4 0.2 N/A N/A N/A 8.8 1.2 0.1
(86) (12) (2) (87) (12) (1)
E2 0.02 0.4 10.5 1.0 0.1 N/A N/A N/A 10.2 0.8 0
(91) (9) (0) (93) (7) (0)
E1 0.01 0.15 0 0 0 N/A N/A N/A 0.8 0 0.1
– – – (89) (0) (11)

Phase I: during shaking.

Phase II: after shaking and before reconsolidation.
Phase III: after reconsolidation.
N/A: not available.

198
Y. Jafarian et al.
Fig. 14. Spectral acceleration (5% damping) for Test-1, Test-2, and Test-3.
histories for Test-1, Test-2, and Test-3, respectively, for the strongest
shaking event (i.e. amax~0.18g). Each image from the visual animation,
which contains a color legend bar on the right, indicates certain times
during the experiments. Dark red denotes on liquefaction (ru=1) and
dark blue indicates that ru value is zero. Initial overburden effective
stress (σ'v0) was calculated using the elastic theory, considering uni-
form foundation’s surcharge. It is noteworthy that the estimated initial
effective overburden stress is a function of foundation rigidity that may
slightly affect estimations of the ru values, but the general behavior will
be the same. It is clear that soil profile started to be liquefied from the
upper layer and liquefaction propagated towards the bottom of the
container. The full depth of the free-field ground in all tests under the
largest event was liquefied in 4.5 s. Large negative EPWP was formed
under the foundations and ru was significantly smaller than the
corresponding values in the free-field. Therefore, a large horizontal
199
Soil Dynamics and Earthquake Engineering 97 (2017) 184–204
hydraulic gradient was generated between the free-field and under
foundations in the first cycles of shaking. Once excitation ceased, water
began to flow horizontally from the free-field with high EPWP potential
towards the foundation sands with negative EPWP. Although remained
non-liquefied during and after shaking, ru value beneath both founda-
tions was increased after the end of shaking phase due to upward
seepage and horizontal flow towards the foundations. After EPWP was
equalized at each level by about 10 s, it remained constant for a while
due to upward flow from deeper levels towards the ground surface. The
timing of sustained EPWP is directly related to shaking intensity (or
extent of liquefaction); the stronger the shaking intensity, the longer
the sustained EPWP time. Finally, EPWP began to dissipate and the
soil beneath the foundations reconsolidated together with the free-field
soil. The solidification front moved upward from a depth of 24 m
towards the ground surface with a velocity of approximately 89.4 mm/
s.
Compared with Test-1 and Test-3, horizontal flow was more
complicated in Test-2 wherein the heavy and light foundations were
located close to each other. As shown in Fig. 16, the adjacent
foundations behaved like a single foundation and ru values under the
foundations are more similar, compared to Test-1 and Test-2. Larger
negative EPWPs were generated under the heavy foundation compared
with the light one. Therefore, water flowed from the free-field and the
light foundation towards the heavy foundation, but EPWP increased
during this timespan under the light foundation. This is due to the fact
that horizontal inflow from the free-field towards the light foundation
was more than the horizontal outflow from light foundation towards
the heavy foundation.
6. Liquefaction-induced settlement mechanisms: free-field
versus foundation
Fig. 18 represents typical settlement time histories of free-field and
rigid foundation which have been observed in the three centrifuge
experiments conducted in this study. There is a clear difference
between the settlement mechanism of the free-field and foundation
in different time periods. It was early discussed that settlement
behavior can be divided into three phases. In each phase, settlements
may be categorized as either volumetric- or deviatoric-induced settle-
ments. Settlement of shallow foundations, as seen in Fig. 18(a), begins
immediately once the seismic loading starts and continues to the end of
shaking in Phase I. Partial drainage and inertial force imposed by
foundation weight seem to be two dominant mechanisms of foundation
settlement during shaking. In Phase II, the pore water pressure under
foundation increases due to large hydraulic gradient generated during
shaking, soil softens due to pore pressure buildup, and foundation
settles under its own weight. Shear deformations due to partial bearing
capacity failure are dominant mechanism in Phase II. The amount of
hydraulic gradient, the extent of liquefaction, soil permeability, and
foundation weight may control foundation settlement during this
phase. Negligible foundation settlement is accumulated after EPWP
began to dissipate in reconsolidation phase (Phase III) in which soil
regains its stiffness.
Settlement mechanism in free-field, as shown in Fig. 18(b), is
completely different from that of the foundation. In the free-field, the
ground surface begins to settle once shaking starts in phase I. The
settlement in this phase is attributed to partial drainage and sedimentation
during shaking. No settlement occurred during the second phase in which
EPWP sustains. The free-field soil settles due to EPWP dissipation and
reconsolidation in Phase III.
Empirical estimation of liquefaction-induced volumetric settle-
ments is based on the results of cyclic triaxial tests. These methods
assume fully undrained mechanism during shaking and fully drained
mechanism after shaking. There are some relationships to estimate
liquefaction-induced volumetric strains based on this method, such as
Tokimatsu and Seed [32], Ishihara and Yoshimine [33], and Wu et al.
Y. Jafarian et al. Soil Dynamics and Earthquake Engineering 97 (2017) 184–204

Fig. 15. Distributions of excess pore pressure ratio at selected times for Test-1 in peak acceleration of 0.17g. (For interpretation of the references to color in this figure, the reader is
referred to the web version of this article.)

[34]. In the current experiments, ground surface experienced extents of estimation especially when the sand permeability is high. Most of the
settlement during different shaking events. It is apparent from the settlement in free-field occurred because of EPWP dissipation and
observations that partial drainage occurred during shaking. This reconsolidation in the third phase. This portion was higher for stronger
mechanism should be considered in liquefaction-induced settlement shaking event. Engineers still estimate foundation settlement by the

200
Y. Jafarian et al. Soil Dynamics and Earthquake Engineering 97 (2017) 184–204

Fig. 16. Distributions of excess pore pressure ratio at selected times for Test-2 in peak acceleration of 0.18g. (For interpretation of the references to color in this figure, the reader is
referred to the web version of this article.)

volumetric strain-based relationships which are mostly suited for the comparison between Fig. 18(a and b), it is apparent that most of the
free-field condition. In this way, the volumetric-induced settlement is foundation settlement occurs in Phase I and Phase II where shear
considered as the dominant mechanism of foundations settlement. By deformations are dominant, but most of the free-field settlement occurs

201
Y. Jafarian et al. Soil Dynamics and Earthquake Engineering 97 (2017) 184–204

Fig. 17. Distributions of excess pore pressure ratio at selected times for Test-3 in peak acceleration of 0.18g. (For interpretation of the references to color in this figure, the reader is
referred to the web version of this article.)

in Phase III wherein volumetric settlements are dominant.
Additionally, the amount of foundation settlement is remarkably larger
than free-field settlement. As a result, it seems that such relationships
may result in unconservative estimates of liquefaction-induced founda-
tion settlement.
7. Effect of spacing on foundation settlement
Effect of spacing on the settlement of foundations is evaluated
through the proximity ratio suggested by Tsukamoto et al. [4]. This
parameter is defined as:

202
Y. Jafarian et al.
Fig. 18. Typical time history of liquefaction-induced settlement for (a) rigid foundation
and (b) free-field.
rp = D/(D + B)
where D is spacing between the two foundations and B is foundation
width. The foundations' settlement measured in the three experiments
were plotted against the values of the proximity ratio rp in Fig. 19(a and
b). The amount of settlements during the shaking as well as the total
settlement have been separated in order to clarify the effect of spacing
on the settlement behavior of foundations. It is seen that foundation
Fig. 19. Settlement of foundation against proximity ratio rp for (a) Foundation L and (b) Foundation H.
Soil Dynamics and Earthquake Engineering 97 (2017) 184–204
settlement during shaking is slightly changed by variation of the
spacing between two foundations, but it has a clear impact on the
amount of total settlement. Therefore, part of settlements accumulates
during Phase II, wherein EPWP sustains, and it merits further
consideration. Foundation settlement decreased in smaller proximity
for all seismic events, except for the strongest event. Although
foundation spacing clearly affects the settlement mechanism, no clear
dependency of settlement on the foundation spacing can be found with
the results of the current study. Such dependency might be relevant to
the extent of the liquefied zone around the foundations. There are
different mechanisms contributed to foundation settlement which
makes the problem more complicated but it seems that amount of
settlement in the second phase is crucial and should be considered in
practice.
8. Summary and conclusions
Three series of centrifuge experiments were conducted to evaluate
the liquefaction-induced and post-liquefaction settlement of shallow
foundations. Two rigid rectangular foundations located on the ground
surface were tested under limited and complete liquefaction conditions.
The spacing between the heavy and light foundations has been changed
in three tests to investigate the effect of foundation-soil-foundation
interaction on the seismic performance of both foundations. Each
model was excited with various shaking events to examine the influence
of the intensity of seismic shaking as well as the extent of liquefied
zones.
Clear differences were observed between settlement mechanism of
(3) free-field and shallow foundations during shaking and the time period
after shaking. Permanent settlement accumulates in three phases: (I)
the shaking timespan, (II) the time period at which EPWP is fully
developed and sustains, and (III) reconsolidation phase where EPWP
starts to dissipate. In the free-field ground, most of the settlement
occurred due to volumetric-induced deformations in the first and the
third phases. In contrast, about 90% of foundation settlement occurred
due to shear-induced deformations in the first and the second phases.
203
Y. Jafarian et al.
Although foundation settlement in the first phase was similar in
different shaking events, it has been increased in the second phase
with the increase of shaking intensity and liquefied depth. Foundations
settled significantly more than the free-field, especially when a large
amount of sand profile was liquefied. It seems that estimation of
liquefaction-induced settlements based on volumetric strains in cyclic
element tests is not applicable for foundations settlement. It has been
found in the current study that foundation settlement during Phase II
is considerable; however, it was less considered during the previous
studies. Settlement of this phase may be directly related to soil
permeability. The amount of hydraulic gradient, the extent of liquefac-
tion, soil permeability, and foundation weight may control foundation
settlement.
Visualization of EPWP demonstrates that soil beneath the founda-
tions was never liquefied during shaking. Large negative ru was
observed under the heavy foundation which is attributed to the dilatant
behavior of sand beneath the foundation. Once shaking ceased, EPWP
increased due to large horizontal and vertical gradients generated
during shaking and caused rising ru beneath the foundations. The
entire soil profile was remained with constant EPWP in each level for a
relatively long time after shaking ceased. Rising ru after shaking caused
a large amount of foundation settlement. After a while, soil reconso-
lidation started from deeper layers towards the shallower layers.
Foundations behaved like a mono-foundation when they became
closer. Close foundations oscillated more strongly compared with the
isolated foundation. In this case, foundations and the underlying soil
oscillated in the same phase and amplitude like a rigid system.
Significant tilting was observed in the case where two foundations
were located close to each other. EPWP and soil stiffness difference
between the foundation side and the free-field side produced different
constraint conditions for each footing. The free-field and the founda-
tion sides behaved like free and restrained boundaries, respectively.
Therefore, the foundation settled more in the free-field side as a result.
Alone structures are rarely found in urban areas. The adjacent
foundations impart additional seismic demands on each other. Results
of this study indicate that neglecting the interaction between two
adjacent foundations where liquefaction potential is high may be
unconservative, and in these cases, FSFI effects should be considered
in analysis and design. It is noteworthy that the results of this paper are
limited to the centrifuge tests condition scheduled in the present study
such as soil type, foundation width, uniformity and depth of foundation
sand, and etc. Moreover, the current research ignored the effect of the
superstructure which may result in a more complicated response of
foundations.
Acknowledgements
Results of this paper were parts of a research project contract
between International Institute of Earthquake Engineering and
Seismology (IIEES) of Iran and Construction & Development of
Transportation Infrastructures Company of Iran under project No.
GEO-93-001. This support is gratefully acknowledged.
References
[1] Yoshimi Y, Tokimatsu K. Settlement of buildings on saturated sand during
earthquakes. Soils Found 1977;171:23–38.
[2] Liu L, Dobry R. Seismic response of shallow foundation on liquefiable sand. J
Geotech Geoenviron Eng ASCE 1997;123(6):557–67.
[3] Dashti S, Bray JD, Pestana JM, Riemer M, Wilson D. Mechanisms of seismically
induced settlement of buildings with shallow foundations on liquefiable soil. J
Geotech Geoenviron Eng 2010;136(1):151–64.
[4] Tsukamoto Y, Ishihara K, Sawada S, Fujiwara S. Settlement of rigid circular
foundations during seismic shaking in shaking table tests. Int J Geomech
2012;12(4):462–70.
[5] Seed HB, Idriss IM. Analysis of soil liquefaction: Niigata earthquake. J Soil Mech
Found Div ASCE 1967;93(3):83–108.
204
Soil Dynamics and Earthquake Engineering 97 (2017) 184–204
[6] Nagase H, Ishihara K. Liquefaction-induced compaction and settlement of sand
during earthquakes. Soils Found 1988;28(1):65–76.
[7] Adachi T, Iwai S, Yasui M, Sato Y. Settlement and inclination of reinforced concrete
buildings in Dagupan City due to liquefaction during the 1990 Philippine earth-
quake. In: Proceedings of the 10th world conference on earthquake engineering,
international association for earthquake engineering (IAEE), Madrid (Spain); 1992.
p. 147–52.
[8] Ishihara K, Acacio A, Towhata I. Liquefaction-induced ground damage in Dagupan
in the July 16, 1990 Luzon earthquake. Soils Found 1993;33(1):133–54.
[9] Acacio AA, Kobayashi Y, Towhata I, Bautista RT, Ishihara K. Subsidence of building
foundation resting upon liquefied subsoil case studies and assessment. Soils Found
2001;41(6):111–28.
[10] Tokimatsu K, Midorikawa S, Tamura S, Kuwayama S, Abe A. Preliminary report on
the geotechnical aspects of the Philipine earthquake of July 16, 1990. In:
Proceedings of the 2nd international conference on recent advances in geotechnical
earthquake engineering and soil dynamics, University of Missouri-Rolla. Vol. 1;
1991. p. 357–64.
[11] Tokimatsu K, Kojima J, Kuwayama AA, Midorikawa S. Liquefaction-induced
damage to buildings I 1990 Luzon Earthquake. J Geotech Eng
1994;120(2):290–307.
[12] Yoshida N, Tokimatsu K, Yasuda S, Kokusho T, Okimura T. Geotechnical aspects of
damage in Adapazari city during 1999 Kocaeli, Turkey earthquake. Soils Found
2001;41(4):25–45.
[13] Coelho P, Haigh SK, Madabhushi SP, O'brien T. Centrifuge modeling of the use of
densification as a liquefaction resistance measure for bridge foundations. In:
Proceedings of the 13 the world conference on earthquake engineering; 2004a.
[14] Coelho P, Haigh SK, Madabhushi SP. Centrifuge modeling of liquefaction of
saturated sand under cyclic loading. In: Proceedings of the international conference
on CBS04, Bochum, Germany; 2004b.
[15] Adalier K, Elgamal A, Meneses J, Baez JI. Stone columns as liquefaction counter-
measure in non-plastic silty soils. Soil Dyn Earthq Eng 2003;23(7):571–84.
[16] Dashti S, Bray JD, Pestana JM, Riemer MR, Wilson D. Centrifuge testing to
evaluate and mitigate liquefaction-induced building settlement mechanisms. J
Geotech Geoenviron Eng 2010;136(7):918–29.
[17] Coelho P. Shallow foundations exposed to seismic liquefaction: a centrifuge based
study on the level and mitigation of the effects. SERIES report; Project No.:
227887; 2013.
[18] Marques A, Coelho P, Cilingir U, Haigh SK, Madabhushi G. Earthquake-induced
liquefaction effects on a shallow foundation. In: Proceedings of the WCEE
conference; 2012.
[19] Mehrzad B, Haddad A, Jafarian Y. Centrifuge and numerical models to investigate
liquefaction-induced response of shallow foundations with different contact
pressures. Int J Civ Eng 2016;14:117–31.
[20] Mason HB, Trombetta NW, Chen Z, Bray JD, Hutchinson BL, Kutter BL. Seismic
soil–foundation–structure interaction observed in geotechnical centrifuge experi-
ments. Soil Dyn Earthq Eng 2013;48:162–74.
[21] Taborda R. Three dimensional nonlinear soil and site-city effects in urban regions.
In: civil and environmental engineering. Pittsburgh (PA): Carnegie Mellon
University; 2010.
[22] Hayden C, Zupan J, Allmond J, Kutter B. Centrifue tests of adjacent mat-supported
buildings affected by liquefaction. J Geotech Geoenviron Eng
2015;141(3):04014118.
[23] Finn WDL. State-of-the-art of geotechnical earthquake engineering practice. Soil
Dyn Earthq Eng 2000;20:1–15.
[24] Hausler EA. Influence of ground improvement on settlement and liquefaction: a
study based on field case history evidence and dynamic geotechnical centrifuge
tests. Berkeley, CA: University of California; 2002.
[25] Lee CJ, Wei YC, Kuo YC. Boundary effects of a laminar container in centrifuge
shaking table tests. Soil Dyn Earthq Eng 2012;34:37–51.
[26] Florin VA, Ivanov PL. Liquefaction of saturated sandy soils. In: Proceedings of the
5th international conference soil and mechanics and foundation engineering. Vol.
1; 1961. p. 107–1.
[27] Scott RF. Solidification and consolidation of a liquefied sand column. Soils Found
1985;26(4):23–31.
[28] Kim SR, Hwang JK, Ko HY, Kim MM. Development of dissipation model of excess
pore pressure in liquefied sandy ground. J Geotech Geoenviron Eng
2009;135(4):544–54.
[29] Adalier K, Zimmie TF, Pamuk A. Seismic behavior of rubble-mound moles on sandy
marine deposits. In: Proceedings of the 21st international conference on offshore
mechanics and arctic engineering, Oslo (Norway); 2002.
[30] Trifunac MD. Comparisons between ambient and forced vibration experiments.
Earthq Eng Struct Dyn 1972;1:133–50.
[31] Wang CR, Wei YC, Hung WY, Lee CJ. GUI based computer programs for analyzing
dynamic signals detected from a physical earthquake model. Res J Appl Sci Eng
Technol 2013;6(1):13–9.
[32] Tokimatsu K, Seed HB. Evaluation of settlements in sands due to earthquake
shaking. J Geotech Geoenviron Eng 1987;1138:861–78.
[33] Ishihara K, Yoshimine M. Evaluation of settlements in sand deposits following
liquefaction during earthquakes. Soils Found 1992;321:173–88.
[34] Wu J, Seed RB, Pestana JM. Liquefaction triggering and post liquefaction
deformations of Monterey 0/30 Sand under unidirectional cyclic simple shear
loading. Geoengineering Research Report No.UCB/GE/2003-01. Deptartment of
Civil and Environmental Engineering, Berkeley: Univ. of California, ; 2003.