Soil Dynamics and Earthquake Engineering 104 (2018) 210–227

Contents lists available at ScienceDirect

Soil Dynamics and Earthquake Engineering

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

An investigation about seismic behavior of piled raft foundation for oil T

storage tanks using centrifuge modelling
⁎
Seyed Mohammad Sadegh Sahraeiana,b, , Jiro Takemurab, Sakae Sekib
a
Department of Civil and Environmental Engineering, Shiraz University, Shiraz, Iran
b
Department of Civil and Environmental Engineering, Tokyo Institute of Technology, Tokyo, Japan

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

Keywords: Some level of settlement is allowed in the design of oil tank if the uneven settlement can be controlled in an
Oil storage tank allowable value. Considering a critical condition of piled raft foundation (PRF), that is, secure contact of raft
Liquefaction base to the ground surface, and the expected function of piles to impose additional resistance against the local
Piled raft foundation settlement, PRF is considered as one of the rational foundation systems for the oil tanks. However, PRF has a
Centrifuge modelling
complex interaction with soil under horizontal seismic loading, especially if the tank rests on a liquefiable soil,
which may cause an extreme change of the soil stiffness underneath the tank. In this study, a series of centrifuge
model tests was performed to investigate the mechanical behavior of oil tank supported by piled raft foundation
on liquefiable saturated sand and non-liquefiable dry sand. In the tests, two types of foundation were modelled; a
slab foundation, and a piled raft foundation. Using the observed results, such as accelerations of the tank and
ground, dynamic and permanent displacement of the foundation, and excess pore water pressures of the ground,
advantages and limitations of piled raft foundation for application to oil tanks on sandy soil are discussed.

1. Introduction effective pile spacing [8,9]. Mechanical behavior of this foundation

Majority of existing oil storage tanks in Japan were constructed
before the early 1970's when soil liquefaction was first considered in
the design of tank foundations. Since the 1964 Niigata earthquake, the
1978 Miyagiken-oki earthquake [1] and the 1995 Hyogoken-Nambu
Earthquake, it has become an urgent matter for geotechnical engineers
to assess the seismic stability of existing oil storage tanks and imple-
ment proper countermeasures against soil liquefaction.
Piled raft foundations (PRFs) have received considerable attention
in the recent years, especially since Burland et al. [2] introduced the
settlement reducer concept of PRF. The raft in this foundation system
has adequate bearing capacity; therefore, the main objective of in-
troducing the pile elements is to control or minimize the settlement,
especially uneven settlement, rather than to carry the major portion of
the vertical loads. Therefore, a major concern in the design of PRF is
how to design the piles optimally to control the settlement [3–5]. Some
researchers utilized finite element modelling (FEM) to study the effect
of raft and pile dimension on the behavior of this foundation system
[6,7]. Also, PRFs have been used for building design and some case
studies of buildings have been reported. Field measurements were
employed in these cases to estimate several parameters such as settle-
ment, uneven settlement, load sharing between piles and the raft, and
system under various loading conditions has been also studied by
physical modelling. Static lateral loading tests were conducted in 1g
condition to evaluate the application of pile groups and PRF, and dis-
cuss the optimized parameters, e.g. raft size, number of piles, piles
spacing [10,11]. Furthermore, 1g experimental and analytical studies
were performed for static lateral loading conditions to investigate the
effects of pile head connection conditions between the raft and piles
[12,13]. Similar researches were also made for dynamic loading con-
ditions to investigate the performance of piled raft foundation [14,15].
In addition, some studies were accomplished about the performance of
piled raft foundations, which experienced real earthquake loadings
[9,16,17].
Centrifuge modelling is a prevalent approach for various studies in
fields of geotechnical engineering, including soil liquefaction [18,19]
and soil-structure-interaction problems [20,21]. To study the mechan-
ical behavior of piled raft foundation as a complex soil-structure-in-
teraction problem, centrifuge model tests have also been conducted
under not only static loadings but also dynamic loadings. Horikoshi
et al. [22] and Sawada and Takemura [23] used centrifuge modelling to
compare the behavior of PRF with pile group and raft foundations
under static horizontal loadings. On the other hand, Horikoshi et al.
[14] and Nakai et al. [24] conducted dynamic centrifuge model tests to

⁎
Corresponding author at: Department of Civil and Environmental Engineering, Shiraz University, Shiraz, Iran.
E-mail addresses: sadegh.sahraeian@shirazu.ac.ir, sadegh.sahraeian@gmail.com (S.M.S. Sahraeian), jtakemur@cv.titech.ac.jp (J. Takemura), seki@cv.titech.ac.jp (S. Seki).

http://dx.doi.org/10.1016/j.soildyn.2017.10.010
Received 2 March 2016; Received in revised form 2 August 2017; Accepted 16 October 2017
Available online 05 November 2017
0267-7261/ © 2017 Elsevier Ltd. All rights reserved.
S.M.S. Sahraeian et al.
study the dynamic behavior of PRFs and pile group foundations, in-
cluding the effect of pile head connection. Despite numerous studies on
piled raft foundations, optimal and rational design methods of piled raft
foundations have not been extended to the civil engineering infra-
structures. This is partly due to the complex soil-structure interaction
between raft, ground and piles during an earthquake. In particular, if
the piled raft rests on a liquefiable ground, the soil-foundation inter-
action becomes more complex. Because of this complexity and possible
large settlement, the practical implementation of piled raft foundation
is further hindered.
Another concern in the seismic design of piled raft foundation is to
secure the contact of the raft with the subsoil. Without the contact, the
contribution of raft cannot be assured against the horizontal load. To
ensure the secure contact, the foundation settlement should be greater
than the ground settlement. In the design of oil tank foundations, un-
even settlement is a greater concern than maximum settlement. For
example, an allowable uneven settlement is 1/300 of tank diameter
[25], implying that some level of tank foundation settlement is per-
mitted as long as uneven settlement is maintained below the allowable
value. Therefore, piled raft foundation is considered one of the rational
foundation systems for the oil storage tanks. Some studies have been
conducted about oil tank foundations. For example, performances of
pile foundation of storage tanks were investigated in some case studies
[26,27]. Sento et al. [28] reported case studies about oil tanks on li-
quefiable sandy soil using compaction method as the countermeasure. A
few researchers have considered piled raft foundations for the oil tanks.
A case study of oil storage tank with piled raft foundation was done by
Liew et al. [29]. Chaudhary [30] utilized FEM to study the behavior of
piled raft foundation for a huge storage tank. As few researches on PRF
of oil tanks on the liquefiable sand, Imamura et al. [31] and Takemura
et al. [32] investigated about the dynamic response of oil tank sup-
ported by PRF using centrifuge modelling. From these researches, dy-
namic and permanent behavior of foundations were well observed.
However, the observations were made in the shaking direction only, not
in the different directions.
In this study, dynamic centrifuge model tests were performed to
investigate the mechanical behavior of oil tank supported by piled raft
foundation on liquefiable saturated sand and non-liquefiable dry sand.
In the tests, two types of foundations were modelled for oil storage
tanks, namely, slab foundation (SF) and piled raft foundation (PRF).
From the observed behavior, such as excess pore water pressures and
accelerations of the ground, and accelerations, rotation and settlement
of the tank, typical dynamic behavior and permanent displacements of
the tank with PRF were studied and compared with those of the slab
foundation not only in the shaking direction but also in the transverse
direction. From these investigations and comparisons, the advantages
and limitations of piled raft foundation for the application to oil tanks
on sandy soil are discussed.
2. Dynamic centrifuge tests
2.1. Equipment, model foundations and test cases
The centrifuge tests were conducted using Tokyo Tech Mark III
centrifuge and a shaking table [33], under 50 g centrifugal acceleration.
For modelling of the ground, a laminar box consisted of 15 laminas and
rubber membrane bag with inner dimensions 600 mm in length,
250 mm in width and 438 mm in depth was used as shown in Fig. 1.
Because the main objective in the current research was to model
ground without liquefaction and with complete liquefaction, a simple
uniform sandy ground with a moderate relative density was modelled
beneath the tank. To this end, five model tests were performed
(Table 1). In Case 1 and Case 2, a slab foundation (SF) and a piled raft
foundation (PRF) were placed on dry sand, respectively. The SF and
PRF were also modelled in Cases 3a and 3b and Case 4 for saturated
sand. Case 3b was conducted in almost same conditions as Case 3a. The
211
Soil Dynamics and Earthquake Engineering 104 (2018) 210–227
sensors were placed in two different sections; one section at the center
line of the model in the shaking direction; and the other, in the trans-
verse direction. Model dimensions and instrumentation details are
shown in Fig. 1(a), (b) and (c).
2.2. Tank, pile, raft and ground modelling
Characteristics of the tank, pile and raft model used for the tests are
presented in both the model and prototype scales in Table 2 (for more
details about scaling factors in geotechnical centrifuge modelling refer
to Garnier et al. [34]). The tank model (Fig. 2(a)) is made of an acrylic
cylinder with 140 mm outer diameter, 160 mm height and 3 mm
thickness. These dimensions were selected to model a small size tank
considering the capacity of the model box. It was glued to the slab/raft
model made of an aluminum disk with 150 mm diameter and 10 mm
thickness (Fig. 2(a) and (c)). The raft model has 12 conical shape
concave holes which are put onto the pile heads (Fig. 2(d) and Fig. 3).
Silica sand No.8 (Table 3), which was used for the model ground, was
glued to the bottom surface of the model raft to create a rough surface
condition. Water was used as a liquid in the tank with a height of
140 mm. The total weight of the water, tank and raft (2.9 kg), created
1.42 kN of weight and 81 kPa of the average raft base pressure under
50 g centrifugal acceleration.
The piled raft foundation has 12 identical piles, made of an alu-
minum tube with outer diameter of 6 mm, a thickness of 0.5 mm, and
length of 100 mm as shown in Fig. 3. The rough piles shaft surface was
also made by gluing silica sand No.8. These piles were arranged sym-
metrically as shown in Fig. 2(d). Utilizing this number and configura-
tion of piles, the spacing/diameter ratio of piles (s/d) for most of the
piles is 5.4. Friction angle of sand (φ’) with relatively medium condition
(Dr = 65%) is about 40° [35]. The calculated vertical bearing capacities
of the raft, assuming the full mobilization and partial mobilization (tan
φ* = 2/3tan φ’ [36], φ* = 30°) of the friction angle of the sand, range
from 29 to 147 MN and 18 to 92 MN for dry and saturated sand, re-
spectively, in prototype scale. The vertical bearing capacity of one pile
for these friction angles ranges from 0.32 to 1.7 MN and 0.19–1.1 MN
for dry and saturated sand, respectively, in prototype scale. The total
load of tank, including the tank and raft, is about 3.6 MN, which is
much smaller than the bearing capacity of the raft alone, but almost
larger than the total bearing capacities of the 12 piles. From these
calculations, the expected function of piles as a settlement reducer
which is the major objective of piles in PRF, can be confirmed. The pile
heads were not rigidly fixed onto the raft, but simply capped by the
concave hole, which allows free rotation like pinned connection
(Fig. 3). In this way, the piles were subjected mostly to large axial and
lateral forces and a small bending moment at the connection point to
the raft. This condition is close to the actual situation of normal piled-
foundation of oil tank [37]. In the model pile design, flexural rigidity
and axial stiffness of concrete piles were targeted, but not failure of
piles. As confirmed in Table 2, the axial load causing yield of the pile
material is larger than the most of expected pile bearing capacity,
shown above, and also much higher than the total load divided by pile
number, i.e., 0.3 MN (= 3.6/12). The raft made by aluminum can be
considered as a rigid plate which corresponds to a small diameter tank
foundation. These conditions of structure components were made to
focus on the effects of soil failure rather than the structural failures.
In order to measure the pile axial load and shaft friction, the piles
used in Case 2 were instrumented by axial strain gages at the head and
tip as shown in Fig. 3. However, in Case 4, to prevent the non-uni-
formity of the ground made by sand pouring due to the wires connected
to the piles, non-instrumented piles were substituted while 5 external
(non-built-in) earth pressure cells (E.P.s) were glued on the raft base to
measure the raft contact pressure (Fig. 2(d)). The raft model with non-
built-in E.P.s was also used in Case 3a. To improve the reliability of
earth pressure measurement by eliminating the stress concentration on
the attached E.P.s, a new raft model with 5 built-in E.P.s covered by
S.M.S. Sahraeian et al. Soil Dynamics and Earthquake Engineering 104 (2018) 210–227

Fig. 1. Model setup, instrumentation and laminar box used for the tests.

thin silicon rubber were employed in Case 3b (Fig. 2(b)). Table 2

Characteristics of tank, raft and pile used in model and prototype in 50 g.

2.3. Model preparation, and test procedure Model Prototype

Fine silica sand (No.8) was used for the liquefiable sand layer and Tank material acrylic cylinder steel
outer diameter 140 mm 7.0 m
coarse silica sand (No.3) for the bottom drainage layer. Detailed
thickness 3 mm
properties of silica sands No.3 and No.8 are presented in Table 3. Dry height 160 mm 8.0 m
silica sand No.3 was first placed at the bottom of the laminar box and weight (liquid & raft) 1.42 kN 3.6 MN
then the fine silica sand layer was made by air pluviation. In the model, tank average pressure 81 kPa 81 kPa
Raft material aluminum RC
de-aired water was used as pore fluid of the sand. In order to ensure the
diameter 150 mm 7.5 m
ground saturation, CO2 gas firstly was passed from the bottom of la- thickness 10 mm 0.5 m
minar box through the dry sands slowly to replace the air with CO2 gas base surface rough rough
in the pores. The box was then placed in a vacuum tank. Having va- Pile material aluminum RC
cuum conditions less than − 95 kPa, de-aired water was introduced outer diameter 6 mm (0.5 mm) 0.3 m
thickness 0.5 mm 25 mm
from bottom of the box [38]. The coarse grain size silica sand No.3 was
length 10 mm 5m
utilized as the drainage layer at the model bottom to supply water axial rigidity: EA 596 kN 1.49 GN
evenly into the model ground during the saturation process. The sand yielding axial load 0.60 kN 1.50 MN
layer with a relative density of 65% was aimed as target density, but in bending rigidity: EI 0.0023 kNm2 14.2 MNm2
shaft surface rough rough
some cases, the final relative density had a few deviations from the
target value (Table 1). The piles were fixed in the center of the mod-
elling box by an aluminum guide during pouring the sand (Fig. 4). viscous fluid and the relatively high prototype permeability of soil in
During the sand preparation, the accelerometers and pore water pres- the model might have some effect at pile vicinity. However, the suction
sure transducers were placed at the prescribed locations as shown in around the pile is considered to be small because of less relative dis-
Fig. 1. placement between pile and soil in the model of this study, because of
As water was used for pore fluid in the model, the prototype per- less lateral movement of soil and floating pile in the liquefied soil.
meability of silica sand No.8 was about 1.0 × 10− 3 m/s in 50 g cen- Having made the model ground and placing the model tank on the
trifugal acceleration (Table 3). Although this value is relatively high, it ground, the tank was vertically loaded by an electrical jack to have a
is low enough to accumulate excess pore water pressure and produce secure contact between the raft base and the ground surface. The load-
soil liquefaction in the early stage of the shaking. The permeability of displacement curves measured in the preloading for all cases except for
soil affects the soil pile interaction, especially excess pore water mi- Case 3a (the data were not recorded) are presented in Fig. 5. As shown
gration in the vicinity of the pile. Gonzalez et al. [39] reported the in this figure, the maximum preload for different cases was chosen
effects of pore fluid (water and viscus fluid) on the displacement of piles based on the sand stiffness and foundation type. In the cases with the
subjected to lateral spreading by liquefaction. The difference of the pile piled raft foundation (PRF), it was assumed that the raft load proportion
displacement was attributed to extent of dissipation of suction which (RLP) will be about 50% and the preload about twice that of slab
was generated by the dilatancy due to local shearing around the pile. foundation (SF) was applied. Also, in saturated cases the preload was
The larger suction can be sustained for the less permeable case with

Table 1
Test cases.

Test code Foundation Ground Details

Case 1 Slab Dry sand (Dr = 66%) Slab w/o E.P.s

Case 2 Piled Raft Dry sand (Dr = 68%) 12 instrumented piles & raft w/o E.P.s
Case 3a Slab Saturated sand (Dr = 65%) Slab with 5 non-built-in E.P.s
Case 3b Slab Saturated sand (Dr = 68%) Slab with 5 built-in E.P.s
Case 4 Piled Raft Saturated sand (Dr = 69%) 12 non-instrumented piles & raft with 5 non-built-in E.P.s

212
S.M.S. Sahraeian et al. Soil Dynamics and Earthquake Engineering 104 (2018) 210–227

Fig. 2. Tank and raft model.

Fig. 3. Pile model.

Table 3
Properties of silica sands.

No.8 No.3

Specific gravity Gs 2.65 2.56

Mean grain size D50 (mm) 0.1 1.47
Effective grain size D10 (mm) 0.041 1.21
Uniformity coefficient Uc 2.93 1.26
Maximum void ratio emax 1.333 0.971
Minimum void ratio emin 0.703 0.702
Permeability coefficient(prototype for k(m/s) 2.0 × 10− 5 4.6 × 10− 3
50 g) (1.0 × 10− (2.3 × 10− 1)
3)

half of the dry cases due to the looseness of the saturated sand in
comparison to the dry sand. In this way, the maximum preloads on the
dry sand were 490 N and 980 N for SF and PRF, respectively; while, in
the saturated cases, about half of these loads were exerted on the
foundations. The measured earth pressures in Case 3b with the built-in
cells and Case 4 with the non-built-in cells during the preloading pro-
cess are presented in Fig. 6, together with the average raft contact
pressure exerted by the jack. In Case 4, the pressure could not be re-
corded by E.P.5 (the raft center). In Case 4, all the earth-pressure cells
Fig. 4. Guide and piles during sand pouring.
recorded larger values than the calculated average pressure (25 kPa),
which was calculated neglecting the load supported by the piles. These
larger stresses can be attributed to the stress concentration on the non-
built-in earth-pressure cells. However, the trends of variation in the
measured contact pressures were well comparable to that of the average
pressure, meaning that even those non-built-in cells could provide
qualitative useful data during the test. In order to eliminate this

213
S.M.S. Sahraeian et al.
Fig. 5. Exerted loads in preloading process.
undesirable stress concentration, the built-in earth-pressure cells were
implemented at the raft base in Case 3b. Although the measured contact
pressures by the built-in cells showed a large difference, the average
value of the measured pressures was much closer to the average exerted
pressure. The inevitable uneven ground surface condition could be a
reason for the large variation of the measured pressures.
After this preloading process, the jack was detached and the whole
setup was mounted on the shaking table on the swing platform of the
centrifuge. Filling the tank with water, the centrifugal acceleration was
increased up to 50 g. For the dry sand cases, white noise vibration was
applied to the model, but not for the saturated sand cases to avoid the
change of structure of sand. After confirming the steadiness of all sensor
outputs, the shaking tests were conducted. The target input wave of the
main shock used in the tests is the EW component of the acceleration
recorded at Kurikoma, Kurihara city in 2008 Iwate-Miyagi Nairiku
earthquake [40]. After the first shake, the second shake with about
fifteen percent higher amplitude was applied to the model. The com-
parison of target acceleration and its Fourier spectrum with those of
input motions in the tests are presented in the prototype scale in Fig. 7.
Due to the limited performance of the shaker, high frequency compo-
nent of the targeted motion could not be made. Furthermore, there
were some differences in the input acceleration, which can be clearly
seen in the variation of Arias intensity of the input accelerations in
Fig. 8. Arias intensity (Ia) firstly proposed by Arias (1970) [41] is a
measure of intensity of shaking defined as:
π ∞
Ia =
2g
∫0 [a (t )]2 dt
(1)
where a(t) is shaking acceleration and t is time. In the dry sand cases,
almost the same input motion could be exerted for the two test cases in
each shake, but in the saturated sand cases, the different motions were
inputted between the test cases depending on the first and second
shakes. In Shake 1, the input motions in Case 3b and Case 4 are nearly
similar, especially until 7 s; but in Case 3a, it was significantly larger
214
Soil Dynamics and Earthquake Engineering 104 (2018) 210–227
than the others. In Shake 2, Case 3a and Case 3b had almost the same
level of input motion, but they were remarkably larger than Case 4.
From the time variation of Arias intensity, it can be seen that the ma-
jority of major input acceleration had been exerted until 10–15 s and
thereafter, the input acceleration amplitudes were so small and the
differences between Case 3a, 3b and 4 were negligible.
In the shaking tests, the ground and tank accelerations, the hor-
izontal and vertical displacements of the tank, and the excess pore
water pressures in the ground were measured as shown in Fig. 1. In the
saturated cases, these instrumentations were made not only in the
longitudinal section of the shaking direction, but also in the transverse
section (Fig. 1(c)). In the following discussions, the results of the model
tests are given in the prototype scale unless otherwise stated.
3. Results and discussions
3.1. Tank on dry sand
3.1.1. Ground response
Ground accelerations observed beneath the tank in Shake 1 for the
SF and PRF models on dry sand are shown with the input accelerations
in Fig. 9. Fourier amplitudes of the acceleration beneath the tank (A6)
are compared to those of the input in Fig. 10. The ground accelerations
were amplified towards the shallow depth in both cases, and they were
amplified more in low periods (high frequencies) in Case 1 with SF. This
behavior could be related to higher confinement pressure of the soil
underneath the tank in SF than PRF (Case 2). As shown in subsequent
section, in Case 2 the majority of tank load was supported by piles and
small portion of the tank load was transferred to the soil beneath the
tank; therefore, the stiffness of the soil under the tank with large con-
fining pressure in Case 1 was larger than that in Case 2, resulting in the
large predominant frequency of the subsoil.
3.1.2. Tank response, settlement and rotation
Accelerations at the tank top and bottom in the shaking direction
(A8, A9) during Shake 1 and their Fourier amplitude are shown in
Fig. 11. The tank top acceleration was larger than that of the bottom in
Case 1 with SF, implying a significant rocking motion. While in Case 2
with PRF, the difference between the top and bottom accelerations and
Fourier amplitudes are not as much as Case 1 which show the efficiency
of the piled raft foundation in reducing rocking motion of the tank. The
spectra of accelerations at the tank bottom and just beneath the tank are
almost the same in Case 1. While in Case 2, the acceleration of tank
bottom is larger than that beneath the tank, implying that the shaking
motion was transmitted through the piles and the tank vibrated rather
independently from the ground. The raft contact condition to the
ground caused the differences in the acceleration predominant period
of the tank between the two foundations. The predominant period was
smaller in Case 1 than Case 2 and the tank acceleration was largely
Fig. 6. Raft base pressures during preloading in Cases 3b & 4.
S.M.S. Sahraeian et al.
Fig. 7. Input accelerations and their Fourier spectra.
amplified in this small period range (0.3–0.5 s). This tendency of tank
response was also observed in the white noise vibration as shown in
Fig. 12. But the predominant period of the white noise vibration was
shorter than that in the shaking test, suggesting the reduction of soil
stiffness by the large input motion in the latter than the former.
Tank settlements measured by three laser displacement transducers
(L1, L2 and L3, Fig. 1) are compared for the slab and piled raft foun-
dations in Fig. 13. In Fig. 13(a) the settlements measured at the op-
posite top edges of the tank in the shaking direction (L1, L2) are shown,
while in Fig. 13(b) the tank edge settlements in the transverse direction
(L3) are shown with the tank center settlement, that is, the average of
L1 and L2. The tank center settlement and L2 are not plotted in the
figures for Case 2 as the settlement could not be measured by L2 in this
case due to the dislocation of the laser from the target plate. However, it
is confirmed from the figures that the settlements of PRF are much
smaller than those of SF, which is a good evidence of settlement reducer
function of the piles. In the beginning of shaking until 2.5 s, although
the input accelerations were not so large, both cases showed rapid
settlements, which could be attributed to the densification of sand.
After the initial settlement, the settlement rate once decreased in both
cases, but in Case 1 the rate increased again from around 6–8 s and then
the tank showed very small settlement increment until the end of
shaking. This large increase of settlement is not clearly seen for Case 2
in the figure. The difference in the trend of settlement from 6 to 8 s
Fig. 8. Arias intensity of input motions.
215
Soil Dynamics and Earthquake Engineering 104 (2018) 210–227
could be caused by the rocking motion enhanced during this period
(Fig. 11) and the load transfer by the piles to deep stiff layer. The LDTs
data show dynamic variation in both cases, but the amplitude of var-
iations and its duration is partly larger in the case of slab foundation
which is another evidence of large rocking motion in this case. Al-
though the tank accelerations in the transverse direction were not
measured in the dry sand cases, it can be confirmed from the small
amplitude of L3 that the rocking should be smaller in the transverse
direction than the shaking direction.
In Fig. 14 the tank rotations in the two dry sand cases are shown.
The rotation of the tank in the shaking direction (L1 - L2 direction) and
L1 - L3 direction are presented together with the maximum rotation for
Case 1 while in Case 2 due to the absence of L2 data the rotation is
presented only in L1-L3 direction. Considerable difference between the
two foundations can be also seen in the rotational behavior. Both the
rotation amplitude in dynamic response and the residual rotation of the
slab foundation were larger than those of the piled raft foundation,
which verify a better performance of PRF in reducing the tank rotation
during dynamic loading. The rotation of SF ceased at t = 8 s and re-
mained constant, despite the further shaking at the later part of the
shaking. The same trend can be confirmed in PRF, but the time of be-
coming the constant rotation angle was earlier than SF. It should be
noted that the rotation in the shaking direction was much smaller than
that in L1-L3 direction. The rotational behavior will be discussed more
S.M.S. Sahraeian et al. Soil Dynamics and Earthquake Engineering 104 (2018) 210–227

Fig. 9. Ground accelerations beneath the tank in Shake 1 (Dry sand).

in later section.
3.1.3. Piles behavior
Using the instrumented piles in Case 2, the head axial load, tip re-
sistance and shaft friction carried by each pile were measured.
Fig. 15(a) shows the axial forces time histories of piles 1, 5 and 9
(Fig. 2(d)). The pile at the inner part of the raft (No. 9) carried larger
bearing load in the static condition. On the other hand, the pile at outer
edge in the shaking direction (No. 1) showed larger amplitude during
the shaking, implying larger contribution of the pile resistance against
the rotational moment from the tank as illustrated in Fig. 15(c). Total
loads carried by all piles are presented in Fig. 15(b). In the figure, the
measured pile head resistance, that is, the tank load carried by the piles,
the pile tip, and shaft resistance components are shown together with
the tank load. It should be noted that due to the interference of the
moment strain to the axial strain measurement near the pile top, the
measured total pile load was overestimated, which could be seen in the
static and dynamic components. Namely, the measured pile loads were

Fig. 10. Ground responses in Shake 1 (Dry sand).

216
S.M.S. Sahraeian et al. Soil Dynamics and Earthquake Engineering 104 (2018) 210–227

Fig. 11. Tank responses in Shake 1 (Dry sand).

Fig. 12. Ground and tank responses during White noise (Dry sand).

larger than the tank load in static conditions and during the shaking,
though the shaking motion was applied in the horizontal direction, not
vertical direction. However, the pile axial load bearing behavior can be
discussed by the measured data and it can be said from the figure that
the majority of the tank load was carried by piles in the static conditions
for this particular PRF.
About 70% of pile resistance was mobilized by the tips and 30% by
the shaft in the static condition. In order to verify the effects of piles and
raft on the rotational behavior of the tank, the time-histories of the tank
rotational moment, piles resistant moment and raft resistant moment

Fig. 13. Tank settlements in Shake 1 (Dry sand).

217
S.M.S. Sahraeian et al. Soil Dynamics and Earthquake Engineering 104 (2018) 210–227

Fig. 14. Tank rotations in Shake 1 (Dry sand).

are shown in Fig. 15(d) and the tank rotational moment and the piles
resistant moment are compared in Fig. 15(e). The tank rotational mo-
ment and piles resistant moment are calculated from the tank inertia
force and the piles axial forces respectively around the center of the raft
(Fig. 15(c)). The difference between these two moments is the raft re-
sistant moment. Although the error in the measurement of pile axial
forces could cause the uncertainty in the estimation of the moment
resistance, the mobilization trend of the resistance can be confirmed in
Fig. 15(d) and (e). Both of the tip and shaft resistances almost evenly
contributed in preventing the tank rotation (Fig. 15(e)). The piles re-
sistant moment had the main role in bearing the tank rotational mo-
ment. Despite very small raft contact pressure in the static condition, it
is confirmed that the raft also resisted the moment load.
Considering the contribution of piles against static and dynamic
loads in the dry sand conditions, more rational design of PRF could be
made by reducing the number of piles. However, for reducing the pile
number, the other critical conditions rather than the foundation set-
tlement, such as failure of the piles and punching of the raft should be
assessed for the design of PRF for oil tanks on dry sand, in which the
relatively large axial pile load could be generated with the static and
dynamic component.
3.2. Tank on saturated sand
3.2.1. Ground response
Fig. 16 presents the ground responses at shallow depth (Z =

Fig. 15. Piles behavior in Case 2 (PRF, Dry sand). (a) Variation of piles loads, (b) Total axial force carried by piles, (c) Diagram of piles resistant moment (PRM) and tank rotational
moment (TRM), (d) Tank rotational moment, piles and raft resistant moment during the shaking, (e) Piles resistant moment vs. tank rotational moment.

218
S.M.S. Sahraeian et al.
Fig. 16. Ground accelerations in Shake 1 (saturated sand).
1.25 m) in the saturated cases (Case 3a, 3b and 4) during Shake 1. The
accelerations recorded beneath the tank (A5: X = 0 m), and beside the
tank (A6: X = 6.75 m, A7: X = 10 m) along the center line of the tank
in the shaking direction (Y = 0 m), and at the transverse section (A11:
X = 0 and Y = 4.5 m) are compared with the input motion (A1).
Fourier amplitudes of the accelerations measured by A5, A7 and A11
are compared with that of input motion in Fig. 17. Attenuation of input
motion was observed in the ground, especially for the short period
component, which is a clear evidence of soil stiffness reduction by li-
quefaction. The attenuation was more significant beside the tank than
beneath the tank, which can be attributed to the confinement effect of
the tank to the soil underneath. This prevents significant reduction of
stiffness and increase of damping ratio as happened in the free ground
beside the tank. However, even though the natural period of the ground
219
Soil Dynamics and Earthquake Engineering 104 (2018) 210–227
becomes long due to the liquefaction, which could cause the amplifi-
cation in the long period range as seen beneath the tank in Case 3b and
Case 4. In Case 3a, because of relatively large input motion, the at-
tenuation was more than the other two cases. The acceleration response
of A11 located at 0.75 m out from the raft edge in the transverse section
was similar to that of A5 beneath the tank center, implying that the tank
load could affect the dynamic behavior at the location of A11. In the
results of Cases 3b and 4, which had almost similar input motions, a
significant difference cannot be confirmed between SF and PRF in the
saturated sand while it was observed in the dry sand.
3.2.2. Excess pore water pressure
The excess pore water pressures (EPWP) in the saturated sand cases
(Cases 3a, 3b and 4) at different locations of the ground observed in
S.M.S. Sahraeian et al.
Fig. 17. Ground responses in Shake 1 (saturated sand).
Shake 1 are presented in Fig. 18. The figure shows the EPWP in dif-
ferent depths in four vertical arrays, three in the longitudinal section at
the tank center (P7, P4, P3 and P2: X = 0 m), near (P8 and P5: X =
6.75 m) and far (P9 and P6: X = 10 m) outside of the raft, and one in
the transverse section (P12, P11: X = 0, Y = 4.5 m). In Case 3b, P2 and
P3 could not be recorded. In the figure, the initial overburden stress
(σv0′ = zγ ′, γ ′: effective unit weight of sand) and a nominal vertical stress
(σv′ * ) are presented. σv′ * is the sum of σv0′ and the stress by the tank
pressure, which is calculated by the elastic solution assuming the uni-
formly distributed raft pressure on an elastic half-space and RLP =
100% [42]. Although this assumption (RLP = 100%) is not completely
correct, according to later discussions (Section 3.2.3), during the li-
quefaction period, most of the tank load is carried by the raft; therefore,
this assumption is not so unrealistic during the shakings. The σv′ * value
is a criterion for discussion on the behaviors in different cases and σv′ * –
EPWP is a representative value for the remaining effective stress in the
ground during the liquefaction. In the early stage of shaking, the EPWPs
increased rapidly and then this rapid rise ceased and the water pressure
either became almost constant or increased gradually during the
shaking. Then, the EPWPs started the dissipation. At the locations,
outside of the raft (P8, P5, P9 and P6) where the tank load did not affect
σv′ * remarkably, the EPWP built nearly up to the σv′ * value, meaning
almost zero effective stress. While at the location beneath and near the
tank (P7, P4, P3, P2, P12 and P11), where the tank load affected σv′ *,
the EPWPs were smaller than the σv′ * values, reconfirming the con-
finement effect of the tank on the soil underneath. The larger the dif-
ference between σv′ * and σv0 ′ is (more confinement effect from the tank),
the larger the remaining effective stresses, that is, σv′ * – EPWP. Although
the effective stress in the area beneath the tank didn’t reach zero value
220
Soil Dynamics and Earthquake Engineering 104 (2018) 210–227
as compared to the surrounding area, the effective stress is affected
considerably due to the large EPWP beneath the tank. The pore water
pressure behaviors of Cases 3b and 4 were almost the same, except P7
and P4 where EPWP in the rapid increase was larger in Case 4 than Case
3b while Case 3a showed different behavior especially in the late start
of the dissipation. Furthermore, the residual EPWPs observed at the
shallow depth after dissipation was larger in Case 3a than the others.
These residual EPWPs were due to the tank settlement at the location
beneath the tank and the settlement of PPT due to the relatively large
unit weight at the location out of the raft. The possible reason of the
difference between Case 3a and the other cases is the differences in the
liquefaction level due to the larger input motion and smaller relative
density in Case 3a than the others (See Figs. 7 and 8 and Table 1). The
relatively large EPWP in the early stage of shaking beneath the tank in
PRF (Case 4) than SF (Case 3b) could be attributed to the increase of
raft pressure due to the reduction of pile bearing load.
As above mentioned, the typical variation of EPWPs can be divided
into three parts. The first part is until the end of rapid increase, which is
called “build-up period” and (t1), is the end of this period. The second
part is from (t1) to the time when the pore pressure starts the dissipation
(t2), which is named “liquefaction period”. The third is the dissipation
stage from (t2) until the end of this period (t3). Determining the end of
dissipation (t3) is not straightforward due to the gradual decrease of
EPWP and the residual EPWP. The end of dissipation was determined at
the time when the decrease of EPWP from the maximum value became
99% of that at the end of the measurement. These three times (t1, t2 and
t3) are highlighted in the EPWP graphs of Fig. 18 and are used for the
discussion on the later parts.
S.M.S. Sahraeian et al. Soil Dynamics and Earthquake Engineering 104 (2018) 210–227

Fig. 18. Excess pore water pressures of the ground in Shake 1(saturated sand).

3.2.3. Raft base contact pressures
Fig. 19 shows the variations of raft base contact pressures measured
by five pressure cells during Shake 1. In Case 4, the cell at the tank
center (EP5) and in Case 3a, EP3 could not measure the pressure. The
measured base pressures in Cases 3a and 4 with the non-built-in cells
were more uniform than those measured in Case 3b with the built-in
cells, which was partly because of the effect of initial surface undulation
on the measured pressures which was more significant for the latter cell
type than the former. However, comparing the variations of Case 3a
and 3b, two points can be confirmed. The first is that the built-in cells

Fig. 19. Raft base pressures of Cases 3a, 3b and 4 in Shake 1(saturated sand).

221
S.M.S. Sahraeian et al.
could measure the dynamic pressure better than the non-built-in cells
and the second is that the average value of the five pressures measured
by the built-in cells during the shaking was closer to the average tank
pressure acting on the raft (σT = 81 kPa) while the non-built-in cells all
recorded the pressure more than σT . From these facts, it can be said that
the built-in cells could give more precise pressure measurements than
the non-built-in ones. The base pressures in Case 3b showed very in-
teresting behavior that the pressure at the center (EP5) increased, while
the others at the outer part of the raft decreased, implying the stress
concentration at the center portion during the shaking. It could be at-
tributed to the less reduction of stiffness at the tank center than the
perimeter side as discussed in Fig. 18. Due to the failure in measuring
EP5, this stress concentration could not be confirmed in the piled raft
model (Case 4). However, the different behavior in Case 4 can be seen
from that in Case 3b. All pressures at the raft outer side showed a quick
rise at the EPWP generation period (t1) and then started decreasing
around the time t2 of P7. The measured average raft base pressure by
the four outer pressure cells in Shake 1 and Shake 2 are shown with
EPWP at P7 in Fig. 20. Much clearer trends can be confirmed from the
figure, that is, the raft load proportion increased by the reduction of pile
loads due to the liquefaction, but the raft load decreased, in other
words, the pile load was regained gradually by the recovery of effective
stresses of the soil due to the dissipation of EPWPs. The recovery of pile
load was earlier in Shake 2 than Shake 1, which corresponds to the fact
that t2 in Shake 2 was earlier than that in Shake 1 for Case 4 (Fig. 20).
The average raft base pressure at the outer side of the raft after the
shaking was larger for Shake 1 than Shake 2. There are two possible
reasons for that; one is the larger pile load proportion for Shake 2 than
Shake 1, and another one is more stress concentration on the center part
for Shake 2 than Shake 1. It should be noted that although the variation
of the raft base pressures is not shown for Shake 2, the more stress
concentration on the center part was confirmed after Shake 2 than
Shake 1 from the fact that the residual center pressure in Case 3b after
Shake 2 was about 270 kPa larger than that after Shake 1 (220 kPa).
3.2.4. Tank response
3.2.4.1. Acceleration. The accelerations at the top and bottom of the
tank in the shaking direction (A9, A8) are shown with the input
acceleration in Fig. 21. In the figure, the tank response in the transverse
direction is also shown.
The Fourier amplitude spectra of the acceleration both in the
shaking and transverse directions are indicated in Fig. 22. In the
shaking direction during the rapid increase of EPWP in the build-up
period before t1, difference of the input, and the tank bottom and top
accelerations were relatively small but after t1 the bottom and top ac-
celerations showed difference, bigger at the top and smaller at the
bottom than the input, resulting rocking motion of the tank. The
rocking motion in Case 3a with the large input motion was higher than
both other cases, but Case 3b has less rocking motion in comparison
with Case 4 even though the input motion in this case was almost same
until t = 6 s and slightly larger after that time than Case 4 (Fig. 8). In all
the cases, short period components were significantly attenuated but
the long period components amplified especially in Case 3a and Case 4.
As a difference between SF and PRF, it can be seen that the phase
difference between the tank bottom and top accelerations were more
222
Soil Dynamics and Earthquake Engineering 104 (2018) 210–227
Fig. 20. Average raft base pressures in Case 4 (PRF, saturated
sand).
significant in Case 3a and Case 3b than Case 4. In the transverse di-
rection, the accelerations were much smaller than those in the shaking
direction. In particular, the bottom transverse acceleration is almost
negligible and with no clear dominant frequency. But the top transverse
accelerations showed a certain vibration with a clear dominant period
of about 0.3 s, which is about half of the dominant period in the ac-
celeration of the shaking direction, 0.6 s. This clear response of the tank
top in the transverse direction can be attributed to the deflection of tank
top due to the relatively small hoop stiffness at the top part of the tank
wall.
3.2.4.2. Tank settlements. The tank settlements were measured by the
three laser displacement transducers at the locations shown in Fig. 23.
In Shake 2 of Case 3a, as the LDTs could not record the data for certain
time intervals, three sets of the settlement could be measured only until
t = 4 s and the measurements of L1 and L3 could be resumed near the
end of the test.
In Fig. 23 the settlements at the tank center, which is the average of
L1 and L2 are compared for the entire period and the early stage of
shaking at the top and bottom figures with t2 and t1 of P7 and P8 re-
spectively. The settlements increased gradually during the shaking in
contrast to EPWPs behavior, that is, a quick rise in a short time
(Fig. 18). Comparing the results in Shake 1 and Shake 2, the effect of
densification by the first shake can be confirmed. Even though the input
motion in Shake 2 was larger than that of Shake 1, the settlements in
the second shake were much smaller than those of the first, except for
Case 3b of which the second input motion was significantly larger than
the first one, about 2 times in terms of Arias intensity (Fig. 8). The effect
of pre-shaking was more significantly evidenced in the beginning. In
Shake 1, the tank started settling at time of about 1.5 s, which is the
actual onset of the shaking in terms of Arias intensity (Fig. 8), and the
settlement rate increased with time until t = 3 s while in Shake 2, there
were no substantial settlement until t = 2.3 s and no increase in set-
tlement rate was observed in the beginning. The relatively large set-
tlement of Shake 1 in the beginning of shaking could be attributed to
the poor contact of the raft base to the ground surface, a kind of bed-
ding error, which can be removed by the first shake. After this initial
part, the settlement increased almost linearly with time until t = 8 s at
the time when EPWP dissipation started in the deep depth beneath the
tank (P2 as shown in Fig. 18). Although the settlement rate decreased at
this point, further settlement occurred even after the t2P7, the time
when EPWP just beneath the raft (P7) started decreasing, until t2P8, the
time when EPWP at the shallow depth beside the raft (P8) started de-
creasing. After t2P8, the minor settlement, which was mainly caused by
small shaking and the consolidation of sand, took place. The settlement
after t2P8 seems smaller for Case 4 (PRF) than Cases 3a and 3b (SF).
Recovery of pile bearing load, which can be confirmed in Fig. 20 could
be a reason for the smaller settlement in the late stage of the shaking
and after the shaking in the PRF than the SFs.
In the period of liquefaction, from t1 to t2, the settlements of slab
foundations were larger than that of the piled raft foundation. However,
as the larger settlements of SF could be partially due to the larger input
motions (Fig. 8), the effectiveness of PRF as a settlement reducer
against the soil liquefaction could not be confirmed in these tests.
S.M.S. Sahraeian et al.
Fig. 21. Tank response accelerations in Shake 1 (saturated sand).
3.3. Tank maximum rotation (dry and saturated cases)
In the stability assessment of tank foundation, the uneven settle-
ment is a critical concern. For the relatively small diameter tank sup-
ported by a rigid slab or raft, the uneven settlement is equivalent to the
rotation of the foundation. In the previous dynamic model tests of
foundation using the one directional shaking table, e.g., Takemura et al.
[32], the rotation of tank foundation was only measured in the shaking
direction. In this study, with the settlement at three locations (Fig. 23),
the maximum rotation and its direction were measured.
Fig. 24 shows the revolution of the maximum rotation during the
shaking for the entire period and the early stage of shaking in the top
and bottom figures with t2 and t1 of P7 and P8, respectively. In the
223
Soil Dynamics and Earthquake Engineering 104 (2018) 210–227
figure, results of Case 2 are missed, because L2 could not be measured.
Also, for Shake 2 of Case 3a, limited results are shown, as the three
settlements could be obtained until t = 4 s and L1 and L3 could give the
data at the end of the shaking.
In the beginning of shaking, the rotation gradually increased with
time in all the cases. The increase of rotation, in the beginning, is earlier
and much larger for Shake 1 and Shake 2 of Case 1 (SF, Dry sand) and
Shake 2 of Case 3a than the other cases. The large tank accelerations of
dry sand from the beginning (Fig. 11) can be a reason for this behavior.
As discussed in Fig. 14, the rotation of SF on dry sand ceased at t = 8 s
and was kept constant and the same trend was observed in Case 2 (PRF,
Dry sand). The tank accelerations measured in the beginning of Shake 2
are depicted for the saturated cases in Fig. 25. A large amplification was
S.M.S. Sahraeian et al. Soil Dynamics and Earthquake Engineering 104 (2018) 210–227

Fig. 22. Tank response Fourier spectrum in Shake 1 (saturated sand).

observed at the tank top from the very beginning in Case 3a as com-
pared to the two other cases which can be the reason for the large ro-
tation in Shake 2 of Case 3a. On the other hand, the behavior of the tank
on saturated sand is quite complicated and varied in different cases. In
Case 4 (PRF), the rotation increased monotonically in the liquefaction
stage (from t1 to t2) both for Shake 1 and Shake 2. While in Shake 1 of
Cases 3a and 3b, the rotation increased after t1 but decreased in the
liquefaction period. In Shake 2 of Case 3b, the rotation behavior fluc-
tuated more during the shaking. As a result, the tank rotations after the
shaking was larger for Case 4 than Cases 3a and 3b, except for Shake 2
of Case 3a, which showed the large rotation from the beginning. From
these results, the effect of PRF in preventing uneven settlement cannot
be confirmed against the liquefaction in the entire depth of piles.
However, in the beginning of the shaking in the period of EPWP build-
up, the rotations of PRF were smaller than those of SF, showing the
effectiveness of PRF for the condition that the pile vertical bearing re-
sistance could be mobilized.
Fig. 26 shows revolutions of the direction of maximum rotation, θ,
during the shaking. The definition of θ is given in the figure. As there
was some inevitable error in the estimation of direction from the
measured settlement, especially in the beginning of shaking when the
settlement was very small, therefore the first part of the data with large

Fig. 23. Tank center settlements (Saturated sand), top: entire shaking period with t2, bottom: earlier stage of shaking with t1.

224
S.M.S. Sahraeian et al.
fluctuation is eliminated in the figure. The observed behavior in the
direction of maximum rotation can be divided into two groups. In the
first group, the direction suddenly changed from the shaking direction
(θ = 0 or 180) to the different direction and became constant to a
certain direction, which was observed in the both shakes of Case 1, Case
4 and Shake 2 of Case 3a. While in the second group, the direction
gradually changed from the shaking direction until the end of the
shaking, which was observed in Shake 1 of Case 3a and both shakes of
Case 3b. By comparing Fig. 26 with Fig. 24, it can be said that the large
rotation occurred once the direction was almost fixed in the first group.
While in the second group, the rotation did not monotonically increase,
but decreased during liquefaction stage between t1 and t2. Even in
Shake 2 of Case 3b, the rotation, and the direction both showed fluc-
tuation. From these observations, it can be inferred that once the di-
rection of the rotation is fixed to a direction diverted from the shaking
direction, the rotation will be accumulated by shaking. But while the
rotation mainly takes place in the shaking direction in the beginning of
shaking, the monotonic increase of the rotation may not easily occur
and the rotation behavior becomes very complicated as seen in the slab
foundation cases (Fig. 24). However, it should be noted that even for
the slab foundation, once a relatively large tilting to the diverse di-
rection was triggered by a disturbance, such as a large rocking motion
of the tank, a large rotation occurred inevitably by the shaking as
shown in Shake 2 of Case 3a. There could be several reasons why Case 4
(PRF) tilted in the diverse direction in the early stage of shaking, such
as: inevitable difference of bearing resistance of each pile, non-uni-
formity of raft base contact condition to the ground, which could cause
the rotation of the tank to the area with small pile resistances and poor
contact of the raft base to the ground.
Fig. 25. Tank response accelerations in the beginning of Shake 2 (saturated sand).
225
Soil Dynamics and Earthquake Engineering 104 (2018) 210–227
Fig. 24. Tank maximum rotations, (a) entire shaking with t2 (b)
earlier stage of shaking with t1.
The relationships between the tank maximum rotation and the tank
center settlement were presented in Fig. 27. In the top figure, the re-
lationship obtained from the dry ground cases are shown with the re-
sults of the previous study [32] about the relationship between the
rotation of shaking direction and the tank settlement. While in the
bottom of Fig. 27, the same relationship was plotted for the saturated
cases with the indication at the time of EPWP buildup (t1) and lique-
faction stage (t2) obtained from the location of P8 at the shallow depth
beside the tank. Due to the lack of settlement data, the rotations of L1-
L3 direction are used for some cases, as shown in the figures, but in
Shake 2 of Case 2, even this data could not be recorded. In dry cases, the
trend of relation is almost the same for SF and PRF, except the PRF of
the previous study, which shows much smaller uneven settlement than
the others. This small rotation could be considered as an advantage of
PRF. However, the advantage may be overestimated because of the
uncertainty of the rotation in the diverse direction and the non-mono-
tonic increase of the rotation. Nonetheless, the effectiveness of PRF for
preventing the uneven settlement can be confirmed in the dry sand
cases (Case 1 and Case 2). Namely, the tank settlement caused by the
shaking can be reduced by PRF, which results in the smaller rotation as
compared to SF. In the saturated cases, the relationships are very dif-
ferent for the different cases and between the first and second shakings.
As an overall trend of the relations, the following can be pointed out: 1)
majority of the settlement and rotation took place in the liquefaction
stage, 2) the rotations relative to the settlements were larger for the
second shake than the first shake, 3) in the early stage until t1, the
rotations of PRF were smaller than or equal to those of SF. From the
limited results, the effectiveness of PRF in reducing uneven settlement
cannot be confirmed especially for the condition of liquefaction of the
S.M.S. Sahraeian et al.
entire depth of piles. Therefore, the effectiveness of PRF on the ground
with partial liquefaction, such as, the liquefaction in partial depth, not
entire depth, should be examined to clarify the condition for that PRF
can be positively applied as a foundation of oil tanks.
4. Conclusions
In this study, a series of dynamic centrifuge model tests has been
conducted on a slab foundation (SF) and a piled raft foundation (PRF)
of oil storage tank resting on dry and saturated sand. The following
conclusions were derived from the centrifuge model tests.
1. In dry sand ground, the input motions were amplified in both types
of foundation, but the amplification for the PRF was more in longer
period range than the SF. The less confinement effect of the raft
stress of PRF than that of SF made the natural period of the subsoil
longer for the PRF than the SF.
2. In dry sand, not only the raft resistance but also the pile resistance
bears the loads and rotational moment. Due to the contribution of
piles of the pile raft foundation, the PRF can effectively reduce tank
rocking motion, settlement and uneven settlement of the tank.
3. In the saturated sand, due to the significant stiffness reduction by
liquefaction, the input motion was attenuated in the short period
range, especially beside the tank with less confinement pressure
from the tank. While in the long period range, the natural period of
ground was increased and the input motion could be amplified.
4. In the saturated sand, the excess pore water pressure (EPWP) be-
havior of the sand underneath the tank is affected by the variation of
raft contact stress during the shaking. Due to the decrease of ef-
fective stress caused by the liquefaction, pile resistance of PRF sig-
nificantly decreases and the raft bearing load increases, which
causes the rapid and relatively larger EPWP increase underneath the
tank than that for SF. While in the liquefaction period, due to the
recovery of pile resistance from the deeper depth, the raft bearing
load starts decreasing earlier than the decrease of EPWP of the sand
226
Soil Dynamics and Earthquake Engineering 104 (2018) 210–227
Fig. 26. Directions of tank maximum rotation.
Fig. 27. Tank maximum rotations vs settlements.
just underneath the tank, which also affects the EPWP behavior of
the sand.
5. During the EPWP build-up period, as the pile bearing resistance can
still be mobilized, the PRF can have effectiveness in reducing the
tank uneven settlement. But once the liquefaction occurs in the
entire depth of piles, the PRF cannot secure the effectiveness.
Therefore, the effectiveness of PRF in reducing the tank uneven
settlement on the partially liquefied ground or the ground with non-
liquefied soil layers should be studied to clarify the conditions in
which the PRF can be positively applied as a foundation of oil tanks.
6. From the observations of the maximum rotation of the tank, it can
be confirmed that once the direction of the rotation is fixed to an
orientation diverted from the shaking direction, the rotation will be
accumulated by shaking. But while the rotation mainly takes place
in the shaking direction, the monotonic increase in the rotation may
not easily occur and the rotation behavior becomes very compli-
cated. The direction of maximum settlement can be caused by in-
evitable difference of bearing resistance of each pile, non-uniformity
of raft base contact condition to the ground.
7. Due to the pile load caused by the static and dynamic loadings in
PRF on dry sand, special consideration is necessary for assessment of
pile failure and punching failure of the raft, especially for the PRF
with small number of piles. While for the case of soil liquefaction,
the critical conditions are settlement and uneven settlement. For the
actual implementation of the concept of PRF in the design of tank
foundation, further study is necessary covering the various factors,
especially level of liquefaction such as liquefaction potential (PL
value) and duration of shaking.
References
[1] Ishihara K, Kawase Y, Nakajima M. Liquefaction characteristics of sand deposits at
an oil tank site during the 1978 Miyagiken-oki earthquake. Soils Found
1980;20(2):97–111.
[2] J. B. Burland, B. B. Broms, and V. F. B. DeMello, Behaviour of foundations and
structures, In: Proceedings of the 9th ICSMFE, 1977, vol. 2, pp. 496–546.
S.M.S. Sahraeian et al.
[3] Poulos HG. Piled raft foundations: design and applications. Géotechnique
2001;51(2):95–113.
[4] Poulos HG, Small JC, Chow H. Piled raft foundations for tall building. Geotech Eng J
SEAGS AGSSEA 2011;42(2):78–84.
[5] Horikoshi K, Randolph MF. A contribution to optimum design of piled rafts.
Géotechnique 1998;48(3):301–17.
[6] Ziaie-Moayed R, Kamalzare M, Safavian M. Evaluation of piled raft foundations
behavior with different dimensions of piles. J Appl Sci 2010.
[7] A. Alnuiam, H. El Naggar, and M. H. El Naggar, Performance of Piled-Raft System
under Axial Load, In: Proceedings of the 18th Inter Con Soil Mech. Geotech. Eng.,
2013, p. 2663–66.
[8] Yamashita K, Yamada T, Hamada J. Investigation of settlement and load sharing on
piled rafts by monitoring full-scale structures. Soils Found 2011;51(3):513–32.
[9] K. Yamashita, Field measurements on piled raft foundations in Japan, In:
Proceedings of the 9th Inter. Conference on Testing and Design Methods for Deep
Foundations (IS-Kanazawa 2012), Japan, Kanazawa, 2012, vol. 1, pp. 79–94.
[10] K. Pastsakorn, Y. Hashizume, T. Matsumoto, Lateral load tests on model pile groups
and piled raft foundations in sand, In: Proceedings of the International Conference
on Physical Modelling in Geotechnique Geotech, 2002, 709–714.
[11] J. Hamada, T. Tsuchiya, T. Tanikawa, and K. Yamashita, Lateral loading model tests
on piled rafts and their evaluation with simplified theoretical equations, In:
Proceedings of the 9th Inter. Conference on Testing and Design Methods for Deep
Foundations (IS-Kanazawa 2012), Japan, Kanazawa, 2012, vol. 1, p. 467–76.
[12] Matsumoto T, Fukumura K, Pastsakorn K, Horikoshi K, Oki A. Experimental and
analytical study on behaviour of model piled rafts in sand subjected to horizontal
and moment loading. Inter J Phys Model Geotech 2004;4(3):1–19.
[13] Matsumoto T, Nemoto H, Mikami H, Yaegashi K, Arai T, Kitiyodom P. Load tests of
piled raft models with different pile head connection conditions and their analyses.
Soils Found 2010;50(1):63–81.
[14] Horikoshi K, Matsumoto T, Hashizume Y, Watanabe T. Performance of piled raft
foundations subjected to dynamic loading. Int J Phys Model Geotech
2003;3(2):51–62.
[15] Matsumoto T, Fukumura K, Horikoshi K, Oki A. Shaking table tests on model piled
rafts in sand considering influence of superstructures. Inter J Phys Model Geotech
2004;4(3):21–38.
[16] Yamashita K, Hamada J, Onimaru S, Higashino M. Seismic behavior of piled raft
with ground improvement supporting a base-isolated building on soft ground in
Tokyo. Soils Found 2012;52(5):1000–15.
[17] Yamashita K, Hashiba T, Ito H, Tanikawa T. Performance of piled Raft Foundation
subjected to Strong seismic motion. Geotech Eng J SEAGS AGSSEA 2014;45(2).
[18] R. Dobry and L. Liu, Centrifuge modelling of soil liquefaction, In: Proceedings of the
Tenth World Conference on Earthquake Engineering, Madrid, Spain, 1992, p.
6801–9.
[19] Dashti S, Bray JD, Pestana JM, Riemer M, Wilson D. Centrifuge testing to evaluate
and mitigate liquefaction-induced building settlement mechanisms. J Geotech
Geoenviron Eng 2010;136(7):918–29.
[20] Dashti S, Hashash YMA, Gillis K, Musgrove M, Walker M. Development of dynamic
centrifuge models of underground structures near tall buildings. Soil Dyn Earthq
Eng . 2016;86:89–105.
[21] Olarte J, Paramasivam B, Dashti S, Liel A, Zannin J. Centrifuge modeling of miti-
gation-soil-foundation-structure interaction on liquefiable ground. Soil Dyn Earthq
Eng . 2017;97:304–23.
[22] Horikoshi K, Matsumoto T, Hashizume Y, Watanabe T, Fukuyama H. Performance
227
Soil Dynamics and Earthquake Engineering 104 (2018) 210–227
of piled raft foundations subjected to static horizontal loads. Int J Phys Model
Geotech 2003;3(2):37–50.
[23] Sawada K, Takemura J. Centrifuge model tests on piled raft foundation in sand
subjected to lateral and moment loads. Soils Found 2014;54(2):126–40.
[24] S. Nakai, H. Kato, R. Ishida, H. Mano, and M. Nagata, Load Bearing Mechanism of
Piled Raft Foundation during Earthquake, In: Proceedings of the Third UJNR
Workshop Soil-Struct. Interact., 2004.
[25] Notification of technical specifications regarding to the regulations on hazardous
materials. FDMA (Fire Disaster Management Agency, Japan), 1974.
[26] M. Cubrinovski and K. Ishihara, Analysis of the performance of an oil-tank pile
foundation in liquefied deposits, XV ICSMGE TC4 Satell. In: Proceedings of the
Conference Lessons Learn. Recent Strong Earthq., 2001.
[27] B. H. Fellenius, Tech, and M. Ochoa, Large liquid storage tanks on piled foundation,
In: Proceedings of the Inter Conference Found. Soft Ground Eng.-Chall. Mekong
Delta, 2013, p. 3–17.
[28] N. Sento, S. Yasuda, N. Yoshida, and K. Harada, Case studies for oil tank on li-
quefiable sandy ground subjected to extremely large earthquakes and counter-
measure effects by compaction, In: Proceedings of the 13th World Conference
Earthq. Eng., 2004.
[29] S. Liew, S. Gue, and Y. Tan, Design and instrumentation results of a reinforcement
concrete piled raft supporting 2500 t oil storage tank on very soft alluvium deposits,
In: Proceedings of the 9th Inter Conference Piling Deep Found., 2002.
[30] Chaudhary MTA. FEM modelling of a large piled raft for settlement control in weak
rock. Eng Struct 2007;29(11):2901–7.
[31] S. Imamura, T. Yagi, and J. Takemura, Dynamic stability of oil tank supported by
piled-raft foundation on liquefiable sand, In: Proceedings of the Physical Modelling
in Geotechnics, Zurich, 2010, vol. 2, p. 1409–14.
[32] J. Takemura, M. Yamada, and S. Seki, Dynamic response and settlement behavior of
piled raft foundation of oil storage tank, In: Proceedings of the Physical Modelling
in Geotechnique, Perth, Australia, 2014, vol. 1, p. 613–9.
[33] J. Takemura, M. Yamada, and S. Sakae, Dynamic response and settlement behavior
of piled raft foundation of oil storage tank, In: Proceedings of the Joint 9th Inter.
Conference of CUEE/4th Asia Conference on Earthquake Eng., Tokyo, 2012, p.
533–44.
[34] Garnier J, et al. Catalogue of scaling laws and similitude questions in geotechnical
centrifuge modelling. Int J Phys Model Geotech . 2007;7(3):01–23.
[35] Yamada M. A study on dynamic response and seismic settlement behavior of piled
raft foundation for oil storage tank (M.Sc. Dissertation). Tokyo, Japan: Tokyo
Institute of Technology; 2012.
[36] Terzaghi K. Theoretical soil mechanics. New York, USA: John Wiley & Sons, Inc;
1943.
[37] Ishimatsu S, Yagi T, Yoshimi T, Takemura J. Filed observation of pile behavior
during the liquid level variation in an oil tank. Kisoko 2009;37(10):76–9.
[38] T. Kimura, J. Takemura, A. Hiro-oka, M. Okamura, and T. Matsuda,
Countermeasures against liquefaction of sand deposits with structures, In:
Proceedings of the IS-TOKYO’ 95, Tokyo, Japan, 1995, vol. 3, pp. 1203–1224.
[39] Gonzalez L, Abdoun T, Dobry R. Effect of soil permeability on centrifuge modeling
of pile response to lateral spreading. J Geotech Geoenviron Eng .
2009;135(1):62–73.
[40] JMA, Japan Meteorological Agency, 2008 〈http://www.seisvol.kishou.go.jp/eq/
kyoshin/jishin/080614_iwate-miyagi/index.html〉.
[41] Kramer SL. Geotechnical earthquake engineering. Pearson Education; 1996.
[42] B. M. Das, Fundamentals of Geotechnical Engineering, 3rd edition. Chris Carson.