Earthquake Geotechnical Engineering for Protection and Development of

Environment and Constructions – Silvestri & Moraci (Eds)

© 2019 Associazione Geotecnica Italiana, Rome, Italy, ISBN 978-0-367-14328-2

Dynamic centrifuge tests for soil-foundation-structure interaction

problems: Disconnected Piled Raft (DPR)

D.S. Kim, K.W. Ko & S. Manandhar

Department of Civil and Environmental Engineering, KAIST, Daejeon, Korea

H.J. Park
School of Urban and Environmental Engineering, UNIST, Ulsan, Korea

J.G. Ha
Korea Atomic Energy Research Institute, Daejeon, Korea

ABSTRACT: The evaluation of seismic behavior of structure considering soil-foundation-

structure interaction (SFSI) has emerged as an important area of research for effective seismic
design. In order to investigate SFSI problems, it could be advantageous to perform the centri-
fuge tests owing to not only simulating the soil-foundation-structure system easily, but also
observing the large-strain, non-linear behavior at in-situ stress conditions. In this study, the
effectiveness of the centrifuge tests for the SFSI problems is validated and its application to
the SFSI problems is studied by focusing on seismic response of the structure on the discon-
nected piled raft (DPR) and the dynamic bending moment of the DPR.

1 INTRODUCTION

Considering soil-foundation-structure interaction (SFSI) in seismic design enables structures

to be constructed efficiently and precisely. In particular, the seismic design method for the
shallow foundation, which allows the non-linear soil-foundation interface behaviors during
strong earthquake, has been introduced as an effective seismic design, commonly referred as
“rocking foundation” (Gazetas, 2015; Kim et al., 2015a). Various methods to reduce the seis-
mic response of a structure using a “rocking foundation” or “rocking isolation” have been
studied (Gajan et al., 2005; Mergos & Kawashima, 2005; Anastasopoulos et al., 2010; Gaze-
tas, 2015; Ko et al., 2018d). In these techniques, the foundation is designed to exhibit rocking
motion, which allows yielding behavior during an earthquake. Thus, the rocking effect reduces
the seismic load on the structure during a strong earthquake (Gazetas & Apostolou, 2004;
Kim et al., 2015b; Ko et al., 2018b).
The challenge for practical application of the rocking foundation concept is estimating and
minimizing the permanent deformation of the foundation after a strong earthquake. In this
regard, the disconnected piled raft (DPR) has been viewed as an effective solution to reduce
permanent settlement (Allmond & Kutter, 2014; Antonellis & Panagiotou, 2014). In DPR, the
piles, which are disconnected from the raft, support the vertical load not as a structural com-
ponent but as a reinforcement for the subsoil (Wong et al., 2000). During an earthquake,
DPR allow the foundation yielding behavior; thus, foundation yielding can be used to reduce
structural loads during a strong earthquake. Most previous studies on the dynamic perform-
ance of the DPR have focused on structural seismic load using rocking behavior of founda-
tion; the geotechnical problems including piles need to be considered for its practical
application to construction.
The SFSI problem is not easily evaluated since the numerous parameters relevant to the
soil-foundation-structure system are simultaneously correlated. In this sense, physical

105
modeling has advantages of simulating the soil-foundation-structure system easily and observ-
ing the non-linear behavior of a system under a large strain range. In particular, dynamic cen-
trifuge testing is appropriate for assessing SFSI problems. Field stress conditions can be
simulated easily, and the time, cost and effort required for the soil, foundation, and structure
modeling are reduced. Dynamic centrifuge testing for SFSI problems has been conducted in
various studies (Morris, 1981; Chen et al., 2013; Martakis et al., 2017).
The objective of this study is to assess the seismic behavior of DPR through centrifuge tests.
The validity of the dynamic centrifuge test for SFSI problems is discussed through the case
study of a nuclear power plant containment structure. The evaluation of the DPR is divided
into structural and foundation levels. In the structural level of the DPR, the seismic load of
the structure on the DPR and the permanent settlement of the foundation during a strong
earthquake are investigated by comparing with the results of the shallow foundation. In the
foundation level of the DPR, the dynamic bending moment of the DPR is evaluated through
dynamic centrifuge tests; four different models of DPR and a Connected Piled Raft (CPR)
model with aluminum and steel piles are tested.

2 CENTRIFUGE EQUIPMENT AND VALIDATION FOR SFSI

2.1 KAIST geo-centrifuge facility

The basic idea of physical modeling using a centrifuge is to accelerate a reduced-scale geotech-
nical structure to an appropriate high g-level to simulate the prototype-scale stress field in the
model structure. Centrifuge modeling with shaking table equipment provides an excellent
opportunity to observe the SFSI in a scaled model.
The beam-type centrifuge facility in KAIST was used to perform all the tests reported
herein. This facility has a platform radius of 5 m and a maximum capacity of 240 g-tons (Kim
et al., 2013a). A scaling law was applied to the results to express them on the prototype scale
(Madabhushi, 2014). A self-balanced electro-hydraulic earthquake simulator mounted on the
centrifuge can generate sinusoidal and real earthquake motions up to 0.5 g in the prototype
scale. The allowable frequency range is from 20 Hz to 300 Hz on the model scale, and the
utilization of the self-balanced shaking table was verified (Kim et al., 2013b). Using North-
ridge earthquake, sinusoidal, and sweep signal, the verification tests of the earthquake simula-
tor were performed. The targeted input signals were well matched with the measured signal
from the earthquake simulator in both time and frequency domains.

2.2 Validation for centrifuge modeling for SFSI study

The reliability of centrifuge testing system for simulating SFSI as well as free-field motion
should be validated in advance. Therefore, an equivalent shear beam (ESB) model box, which
can simulate boundary wall movement that is similar to the soil deposit using bearings and
rubber located between the wall layers, was used for the dynamic centrifuge tests. The dimen-
sions of the ESB payload platform were 670 mm × 670 mm × 650 mm (length, width, and
height, respectively).
Lee et al. (2013) performed the dynamic centrifuge tests to evaluate the free-field motion
and dynamic performance of the ESB model container. As shown in Figure 1a, five acceler-
ometers, from A1 to A5, were placed in three arrays, one on the inside of the end wall of the
container (ESB), another in the middle of the soil deposit (Center), and the third 12 cm away
from the end wall (Side). Figure 1b compares typical acceleration time-histories and response
spectrum, measured at the three accelerometer arrays for an event subjected to maximum
shaking acceleration of 0.235 g in prototype scale. The acceleration time traces recorded at the
base of the model container (A0) were used as the input base-shaking motions of the soil pro-
file. The measured accelerations at different locations match reasonably well and the ESB
model container functions properly. In addition, the response spectrum of the free-field soil
response is well matched with the ESB box response; that is, the end walls of the model

106
Figure 1. Performance of an ESB model box for dynamic centrifuge tests: (a) soil models and instru-
mentation layouts for the test; (b) response spectrum; (c) acceleration time histories for input shaking
acceleration (bedrock PGA: 0.235 g in prototype scale) (Lee et al., 2013).

container act as a shear beam having stiffness equivalent to that of the adjacent soil layers.
Moreover, it shows that the free-field motions measured at soil model in the ESB box match
well with the motions estimated by 1-D site response analysis using equivalent-linear program
SHAKE91 (Idriss & Sun, 1992) and thus the ESB model container can provide a reliable lat-
eral boundary for dynamic site response studies (Lee et al. 2013).
Ha et al. (2014) simulated the Hualien large-scale seismic test (LSST) to assess the effective-
ness of centrifuge modeling for SFSI study. The Hualien LSST program was an international
project to observe real SFSI for a 1/4 scale nuclear containment structure and began recording
earthquake data from 1990 with well-investigated soil information. If the centrifuge modeling
can properly simulate the SFSI during an earthquake, centrifuge modeling can be a powerful
tool to understand the SFSI mechanism and can be utilized to simulate the seismic behavior
at a particular site. Figure 2 compares the typical acceleration time histories and frequency
response measured at the foundation and the free-field surface during the Chi-Chi earthquake
and those recorded at the corresponding locations in the centrifuge test. The amplification
characteristics and time-histories are well matched at the soil layer and foundation of the
structure. This shows the potential of dynamic centrifuge test to simulate the Hualien LSST
and validate the reliability of the centrifuge test method for SFSI research.

3 SEISMIC PERFORMANCE OF STRUCTURE ON THE DISCONNECTED PILED

RAFT

When the foundation is allowed to rock and slide like a shallow foundation on soft soil, the
structural responses are limited (Kim et al., 2015b). So, allowing rocking motion for founda-
tion has been proposed by several researchers as an effective method of seismic isolation
(Gajan et al., 2005, Gazetas, 2015) and has been applied in the design of a modern bridge
(Rion Antirion Bridge). However, this concept has the potential for permanent settlement and
failure in the foundation. On the other hand, pile foundations are widely used with structures
to control settlement (Eslami et al., 2012). Most pile foundations have a rigid connection

107
Figure 2. Centrifuge simulation of Haulien LSST (Ha et al., 2014).

between the piles and the raft. This coupling provides a way for transmission of high horizon-
tal shear forces and mobilized moments to the structure when an earthquake occurs (Fiora-
vante & Giretti, 2010).
To solve the challenges of large settlements of shallow foundations and large seismic load of
structure founded on the CPR, the DPR that releases the connection between the piles and
the raft presents an innovative alternative, which has already been used in projects. The foun-
dation system of the Rion-Antirion Bridge pier in Greece is a famous application of DPR. It
is designed to reduce seismic loading and settlement by inserting an interposed gravel layer
between the piles and the raft (Garnier & Pecker, 1999; Pecker, 2006).
The objective of this section is to investigate the behavior of DPR under various conditions
through centrifuge model tests. The influence factors for the tests are the pile head conditions,
the rigidity of the interposed layer and the level of base shaking. A series of centrifuge model
tests were conducted on dry sand with a model consisting of a foundation and a single degree-
of-freedom structure. The seismic performances of the DPR are compared to other foundation
configurations in terms of the seismic load of the superstructure and the permanent settlement
of the ground.

3.1 Test program

The beam-type centrifuge facility in KAIST was used to perform all the tests reported here in
50 g centrifugal acceleration field. The centrifuge test model consisted of a single-degree-of-
freedom (SDOF) structure model, a foundation, and subsoil deposits in an ESB container.
For manufacturing the SDOF model, two thin steel plates and a lumped mass were used.
The natural period of the structure is about 0.6 sec in the prototype scale. Four types of foun-
dation were selected in these tests; shallow foundation, CPR, and two DPR with different
materials in the interposed layer between the piles and the mat. Except for the CPR in which
the pile and the mat were connected using bolts, three foundation types can easily move hori-
zontally and vertically. The size of the mat (shallow foundation) is 9 cm, 9 cm, and 3 cm for
width, length, height in the model scale, which corresponds to 4.5 m, 4.5 m, and 1.5 m in the
prototype scale. The prototype of the pile foundation is prestressed high strength concrete
(PHC) 500 pile with outer diameter of 500 mm. The model pile is 10 mm in outer diameter
and has 1 mm thickness. A total of nine piles for each foundation type were used in uniform
square arrangement of 3 by 3. The mat and pile foundation is made of aluminum.
In the centrifuge tests for the disconnected pile condition, the settlement of foundation is an
important factor for evaluating the effect of SFSI. The material in the interposed layer between
the piles and the mat can affect the amount of settlement. So, the material was changed from
loose sand to small gravels. The parameters in the centrifuge tests are shown in Table. 1.

108
 Dry silica sand, a fine and uniform sand artificially produced by a hammer crusher, was
used in the experiments. The soil model was made using sand pluviation method at 45% rela-
tive density. The bender elements were installed in the testing box during model preparation
to obtain the shear wave velocity profiles of the soil layer. The initial average shear wave vel-
ocities were about 200 m/s in 50 g-level. The site period of the prototype soil was estimated as
0.53 s. Accelerometers and LVDTs were installed for measuring the seismic behaviors of soil,
foundation and structure, and the schematic diagram of the centrifuge tests are depicted in
Figure 3.
Several real earthquake motions were utilized for the base input motion such as Northridge,
Hachinohe, and Ofunato earthquakes, and the response spectrum of input motions was
described in Ha et al. (2018). At the beginning of the test, strong earthquake motions larger
than 0.3 g intensity were applied for observing the initial settlement during the earthquake.
Then, earthquake intensity in terms of peak ground acceleration increased from about 0.03g
to about 0.3g in stages. These variations in input motion intensity could show the effect of the
soil nonlinearity on soil-foundation-structure interaction.

3.2 Seismic behaviors of SDOF structure

It is important to evaluate the seismic behavior of structure on the foundation with and with-
out considering SFSI. To evaluate the effects of SFSI on seismic behavior of the foundation-
structure system, the fixed base motion (without considering SFSI) and the flexible base
motion (with considering SFSI) needs to be compared with each other. The fixed base motion,
for which the base boundary condition is fixed, is calculated from the equation of motion for
the SDOF structure. In this section, the measured free-field surface motion was utilized to cal-
culate the fixed base motion with known structural properties, such as mass, stiffness and
damping. Meanwhile, when the foundation and structure are lying on the loose sand, the
SFSI effects become larger and the boundary condition of the structure is more flexible. So,
the measured motion on the top mass of the structure includes the SFSI effects. This is the
flexible base motion. The detailed process of calculating the net lateral displacement of the
structure is described in Kim et al. (2014).

Table 1. Parameters for the dynamic centrifuge test (Foundation, Structure)

Test No. Test 1 Test 2

Foundation Shallow foundation (SF- Disconnected pile Connected Disconnected pile

Type AL) foundation (F) DPF- pile foundation (F) DPF-G
S (Sand medium) foundation (Gravel medium)
(F) CPF
Structure Single-degree-of-freedom structure (Tn = 0.6 sec in prototype scale)

Figure 3. Schematic diagram for the centrifuge tests in prototype scale (unit: m).

109
 Figure 4 shows the displacement time history of the fixed and the flexible base motion
during the initial strong earthquake; Northridge earthquake with peak acceleration of about
0.35 g. The gray line is the fixed base motion and the black line is the flexible base motion. In
Figure 4c, the peak values of the fixed base motion and flexible base motion for the case of
CPR are similar. However, for the other three cases, the flexible base motions are smaller than
the fixed base motion, and the amplitudes of peak values are near 0.08 m. These results indi-
cate that the foundation movement can reduce the seismic load on the structure resting on the
flexible foundation. This is because these three foundation types permit foundation move-
ment, especially the foundation rocking phenomena. This is a similar trend with previous
research about the rocking foundation and the DPR (Kim et al., 2015b, Allmond et al., 2014.)
After initial strong earthquake test, stage tests with increasing the intensity of the bedrock
motion were performed. To observe the variation trend in the fixed and the flexible base
motion, the peak value of pseudo-acceleration during Northridge earthquake tests are plotted
as shown in Figure 5. The pseudo-accelerations were determined from the net lateral displace-
ment by multiplying the structural stiffness and dividing by the lumped mass. The hollow
markers are the results of the stage tests and the solid markers are those of the initial strong
earthquake test.
In the connected foundation case, both the fixed and the flexible base acceleration were
increased with an increase in input motion intensity, so the markers are on 1 to 1 line, which is
denoted by a dotted line. On the other hand, the other three cases show different trends for
different input earthquake intensities. The fixed and flexible base accelerations are similar for
the weak input motion, but the flexible accelerations are quite smaller than the fixed base
motion for the strong input motion and they converge to about 0.7g. The results from the
initial strong earthquake follow the general trend. When the foundation rocks, the overturn-
ing moment from the superstructure cannot exceed the ultimate moment capacity of soil-foun-
dation system, so that the flexible base acceleration converged to this limitation. It means that
the seismic load of structure dramatically reduced during strong earthquakes when foundation
rocking is permitted.

3.3 Moment-rotation angle responses of foundation

Figure 6 shows the hysteresis loop of the foundation rotation angle and overturning moment.
When foundation rocking is allowed, the hysteresis loop shows nonlinear behavior with
reduced rocking stiffness and increased energy dissipation characteristics. Furthermore, there
is no overturning moment greater than about 3000 kNm, which is the ultimate moment

Figure 4. Displacement time history of the fixed and flexible motion during initial strong earthquake
(Northridge, Bedrock PGA = 0.35g)

110
Figure 5. Comparison of maximum pseudo-accelerations between fixed base motion and flexible base
motion

Figure 6. Comparison of moment-rotation curves between different foundation types

capacity. However, in the case of CPR, the overturning moment increases to 7000 kNm and a
linear behavior is achieved despite strong earthquake input motion.

3.4 Variations in rocking stiffness and damping with foundation rotation

The variations in rocking stiffness and damping ratio of the soil-foundation system are
important parameters for determining the rocking response in the displacement-based design
(Paolucci et al., 2013). For each moment-rotation hysteresis loop from all the centrifuge tests,
a representative rocking stiffness was calculated from the ratio of the average of the maximum
and minimum moments to the average of the maximum and the minimum rotation angles.
The hysteresis damping ratio was calculated by dividing the area of the hysteresis curve by the
area of elastic energy (Kokkali et al., 2015). Figure 7a shows the variations in the rocking stiff-
ness with the rotation angle. For the shallow foundation and DPR, the fitted curves for the
largest safety factor introduced in Paolucci et al. (2013) were also compared. The initial rock-
ing stiffness at a small rotation angle has a significant deviation depending on the connected
condition between the pile heads and the raft. The largest initial rocking stiffness appeared for
the floating connected pile foundation ((F)CPF). In the case of the shallow foundation and
the DPR with the interposed layer ((F)DPF-S, (F)DPF-G, and SF), the initial rocking stiff-
ness is small, about 3,000,000 kN-m/rad, and shows a relatively smooth decreasing tendency,

111
Figure 7. Rocking stiffness and damping ratio versus rotation angle of each foundation system from
the all stage test data; (a) Rocking stiffness and fitted curves depending on the factor of safety in Paolucci
et al. (2013) (b) Damping Ratio (Ha et al., 2018)

similar to the fitted curve. The relationship between the hysteresis damping ratio and the foun-
dation rotation angle is shown in Figure 7b along with the fitted curve introduced by Paolucci
et al. (2013). The damping ratio is very difficult to obtain from the experiment using actual
recorded seismic motion due to the scattered data. For the foundations on which rocking is
allowed, a similar tendency with Paolucci et al. (2013) was observed in which the damping
ratio increased with increasing rotation angle. In the case of the CPR, the damping ratio was
limited to 5 % and the rotation angle was less than 0.001 rad; only small rotation occurred
during the strong earthquake.

3.5 Foundation settlement during strong earthquake

An important point to evaluate the performance of a DPR is to check its stability against ver-
tical settlement during earthquakes. In Figure 8, the time history of settlement of various
foundation types is described. As shown in Figure 8, the trend of measured vertical settlements
during strong earthquakes shows highest settlement in the case of the shallow foundation, fol-
lowed by the free-field surface, then the DPR with interposed layers and lastly the CPR. In

Figure 8. Settlements of various foundation types for the initial strong earthquake in prototype scale
(Northridge, Bedrock PGA = 0.35g)

112
the case of floating pile groups, the settlement is due to the subsidence of the ground during
the earthquake. In the case of disconnected piles with interposed layers, the settlement was
less than the free-field settlement because of the interposed layer. The effect of the stiffness of
the interposed layer on the DPR can be clearly seen from the comparison with the settlement
of the free-field surface. The settlement of the DPR with a stiff gravel layer was smaller than
the settlement with the loose sand layer having low stiffness. This demonstrates that utilizing
an intervening layer such as a stiff gravel layer in the DPR design, like in the Rion-Antirion
Bridge (Pecker, 2006), is more advantageous with respect to vertical settlement performance.

4 EVALUATION OF DYNAMIC BENDING MOMENT OF

DISCONNECTED PILED RAFT

In conventional design, a group of piles is used to ensure the bearing capacity of the founda-
tion, where each pile carries a small amount of the applied load. However, if the pile founda-
tion does not require a safety margin, the piles can be designed to reduce the settlement of the
raft, where each pile carries a large amount of the load. To reduce the number of piles com-
pared with the conventional design, Randolph & Clancy (1993) studied the proportion of the
load carried by the piles acting to reduce the settlement and the raft. Because the vertical load
is directly applied to the top of the pile and a large amount of the load is carried by each pile,
the pile material has to endure an axial load.
In contrast to conventional design, DPR has been introduced as an effective design for pile
foundation; piles disconnected from the raft support the vertical load not as a structural com-
ponent but as reinforcement of the subsoil. The load transfer platform (LTP) is a granular soil
layer between the raft and piles that prevents the vertical load from being applied directly to
the piles. This makes the vertical load transfer mechanism of the DPR slightly different from
that of the connected piled raft (CPR) foundation. In the case of the DPR, the difference
between the settlements of the piles and soil near the surface causes negative skin friction on
the upper part of the piles. This changes the distribution of the axial load on the piles (Fiora-
vante & Giretti, 2010; Tradigo et al., 2015). However, most of studies on DPR has been
focused on the pile behavior under the static condition.
On the other hand, Azizkandi et al. (2018) compared the dynamic performance of DPR and
CPR using 1 g shaking table tests and numerical analysis, but did not focus on the dynamic
behavior of piles. DPR piles generally would not be subjected to large bending moment
during earthquakes; however, experimental evaluation of the dynamic bending moment of
piles in DPR is required to enhance the reliability of DPR for practical construction.
In this study, centrifuge test models were designed using the virtual prototype model of a
bridge pier. With respect to the single-degree-of-freedom (SDOF) structure, four group pile
models (i.e., DPR with aluminum, CPR with aluminum, DPR with steel, and CPR with steel)
were subjected to centrifuge tests. To observe variations in pile bending moment with associ-
ated conditions, DPR and CPR (two cases) with the same pile material were tested in one
model box. In DPR, the LTP layer was inserted between the piles and the raft. The effects of
pile location and material (aluminum and steel) on the dynamic behavior of DPR and CPR
were evaluated using dynamic centrifuge tests (Ko et al., 2018c).

4.1 Test program

4.1.1 Ground modeling

Dry silica sand with relatively angular particles was used in this study. This is an artificial
sand that is produced with a hammer crusher and is classified as SP (poorly graded clean
sand) according to the Unified Soil Classification System (USCS). The sand had a mean grain
size (D50) of 0.22 mm, coefficient of uniformity (Cu) of 1.96, and coefficient of curvature (Cc)
of 1.11. The target relative density (Dr) of the soil was 50% to represent the loose soil condi-
tion. Gravel with a particle diameter of 2–3 mm was used as the material of the LTP for the

113
DPR. A sand raining system was used to pour dry silica sand into the equivalent shear beam
(ESB) box for the dynamic centrifuge tests.

4.1.2 Pile modeling

Aluminum and steel piles were used to simulate the dynamic behavior of the pretensioned
spun high strength concrete (PHC) and steel pipe. These two types of piles were selected as
representative piles that are generally used in practical construction. In the PHC pile model,
aluminum was used instead of concrete, as concrete is not easy to fabricate in small scale. The
bending moment of piles during earthquakes depends on the flexural rigidity (EI; E is the
Young’s modulus of the pile, and I is the second moment of area) of the pile. Although the
dimensions of the virtual pile prototype were given (i.e., the diameter and the length of the
virtual pile prototype are 1 m and 20 m, respectively), the thickness of the pile was controlled
to match the flexural rigidity of the prototype pile model. The properties of the pile model are
described in Table 2.
The piles and the raft in CPR were designed to be fixed using bolts. Figure 9a shows the
locations of the strain gauges attached to the piles. Seven pairs of strain gauges were attached
at each depth to measure the bending moment of the pile foundation. For the dynamic test,
the strain gauges were densely attached to the upper part of the pile rather than at the end of
the pile for observing the bending moment close to the raft. The dynamic bending moment of
the pile at each depth was calculated from the strain variation as follows:

EI
Mp ¼ εp ð1Þ
y

Table 2. Properties of pile model

Aluminum pile Steel pile
Testing model Prototype Model at 50 g Prototype Model at 50 g

Material PHC Aluminum Steel Stainless steel

Diameter 1m 20 mm 1m 20 mm
Thickness - 1 mm - 2 mm
Length 20 m 400 mm 20 m 400 mm
Flexural rigidity (NM2) 1.35×104 1.91×102 5.89×109 9.09×102

Figure 9. Plan and cross section of foundation model [prototype scale (model scale)]: (a) front and side
view of pile model; (b) plan view of raft model (Ko et al., 2018c).

114
 where Mp is the bending moment of the pile, εp is the strain variation of the pile, and y is
the pile radius.

4.1.3 Foundation modeling

The raft was had dimensions of 7.5 m × 7.5 m × 2 m (length × width × height). The weight of
the raft was 299 tons. Each foundation had nine piles in a three × three arrangement
(Figure 9b). Each pile was located with an equivalent center-to-center distance of 2.5 m. For
the CPR, the piles and the foundation model were fixed with bolts. For the DPR, the LTP
was inserted between the piles and the foundation model to disconnect them. Piles located at
the edge with respect to the shaking direction were instrumented with strain gauges.

4.1.4 Dynamic centrifuge tests

Figure 10 shows a sectional-view schematic of the dynamic centrifuge test. The centrifugal
acceleration was 50 g. The soil depth was 25 m on the prototype scale. The end condition of
the piles was floating, and the distances between the end of the pile and the bedrock for the
DPR and CPR were 4 m and 5 m, respectively. For the DPR, the LTP was modeled with the
dimensions of 7.5 m × 7.5 m × 1 m (length × width × depth) in prototype scale.
The DPR and CPR foundation models were tested in the same ESB box. An accelerometer
was attached to the bottom of the ESB box to measure the input earthquake motion, and
another accelerometer was installed on the soil surface to measure the free field motion. An
accelerometer was installed on the LTP at the same position as the accelerometer on the soil
surface to measure the dynamic response. Strain gauges were attached to four piles in total,
and two pile groups were each arranged in the CPR and the DPR configuration, respectively.
Each pile was placed at the middle or edge of the foundation with respect to the shaking direc-
tion. The centrifuge test models were subjected to real earthquake motions; Ofunato (short-
period dominated) and Hachinohe (long-period dominated) earthquakes. The intensities of
the input motions were increased in stages from small to large for both the Ofunato and
Hachinohe earthquake signals.

4.2 Dynamic bending moment of pile: Time history

Figure 11 shows the time history of the dynamic bending moment of aluminum pile at each
depth during the Hachinohe earthquake. To observe the temporal variations in the dynamic
bending moment of the piles in detail, the time domain was added, and the dynamic bending

Figure 10. Schematic diagram and pictures of dynamic centrifuge test: (a) sectional view; (b) plan view
(Ko et al., 2018c).

115
Figure 11. Time history of dynamic bending moment of aluminum pile for each depth during the
Hachinohe earthquake: (a) peak acceleration of bedrock: 0.137 g; (b) peak acceleration of bedrock: 0.36
g (Ko et al., 2018c).

moment was represented in three dimensions. Figure 11a and 11b represent the dynamic bend-
ing moments of the piles during weak and strong earthquakes, respectively. The absolute
values of the dynamic bending moments of the piles vary according to the PGA of the input
signal, but the general trends in the variation of the bending moment are similar regardless of
the PGA.
In CPR, the dynamic bending moment is largest at the top of the pile. Since the pile is con-
nected to the foundation as a fixed condition, a larger bending moment occurs at the top of
the pile. In addition, the end condition of the pile is a fixed-free (top-bottom) condition, so
that the bending moment at the bottom of the pile is smaller than at other depths. On the
other hand, the dynamic bending moment of the pile at the edge is larger than that in the
middle. Since the outside piles take a greater seismic load than the inner piles along the shak-
ing direction, the bending moment of the piles at the edge of the foundation is larger. This is
known as the shadow effect of group piles (Brown et al., 1988; Ruesta & Townsend, 1997).
In general, the dynamic bending moment of DPR is large at the middle of the pile. Since the
end condition of the pile for DPR is a free-free (top-bottom) condition, the bending moment

116
at the middle of the pile is larger than the bending moment at the ends (top and bottom) of
the pile. The peak dynamic bending moment of the pile does not vary depending on the loca-
tion of the pile in the group due to the disconnected condition between the foundation and the
pile group. At the edge pile for DPR, although there is a larger bending moment at the middle
depth of the pile, a large bending moment also occurs at the top of the pile, and this phenom-
enon is clearly observed for the strong earthquake (Figure 11b).
The peak bending moment of DPR and CPR are not very different, except at the edge pile
for CPR (CPR edge > CPR middle ≈ DPR edge ≈ DPR middle). As expected, the dynamic
bending moment of the pile located at the edge of the foundation can be reduced by effectively
disconnecting the pile and the foundation. However, owing to the shadow effect, the dynamic
bending moment of the CPR pile in the middle of the foundation is reduced, and is not very
different from that of the DPR pile. Moreover, if the number of piles increases as in large
group piles, the shadow effect of the CPR would be significant.

4.3 Dynamic bending moment of pile: Depth

Figure 12 shows the depth at which the peak bending moment occurs with PGA at the soil
surface. The depth of the peak bending moment is independent of the PGA at the soil surface.
As previously discussed, the depth of the peak bending moment clearly differs for DPR and
CPR. In CPR, the depth of the peak bending moment mostly occurs at the top of the pile,
above 2.5 m depth, owing to the fixed-free (top-bottom) condition. As a result, the top of the
CPR pile is generally reinforced to prevent failure during earthquakes. In contrast to CPR,
the depth of the peak bending moment of DPR pile depends on the location of the pile in the
group. If the DPR pile is located in the middle of the foundation, the peak bending moment
mostly occurs at 3.5 m depth. Conversely, if the DPR pile is located at the edge of the founda-
tion, the peak bending moment mostly occurs at the top of the pile, above 2 m depth.

4.4 Dynamic bending moment of pile: Pile material

Figure 13 shows the peak bending moment of piles with PGA at the surface. The peak bend-
ing moment of the pile increases as the surface PGA increases. The trends in the dynamic
bending moment described in the time history section are clearly observed here. In DPR, the
peak bending moment of the edge pile is similar to that of the middle pile. In CPR, the peak
bending moment of the edge pile is larger than that of the middle pile due to the shadow
effect. On the other hand, the peak bending moment difference of the middle pile for the CPR
and DPR cases is not large; however, in the case of the edge pile, the peak bending moment
clearly decreases in the DPR.

Figure 12. The depth of peak bending moment of pile (upper row: aluminum pile, lower row: steel pile)
(Ko et al., 2018c).

117
Figure 13. Peak bending moment of pile for DPR and CPR with PGA at surface (upper row: aluminum
pile, lower row: steel pile) (Ko et al., 2018c).

Figure 14. The variation of the peak strain of the pile for the DPR with PGA at surface (Ko et al., 2018c).

As shown in Figure 13, the bending moment of the steel pile is higher than that of the alu-
minum pile. Since the flexural rigidity of the steel pile is much larger than that of the alumi-
num pile, the latter is expected to be subjected to a larger bending moment. This can be
explained through the peak strain in the pile. Figure 14 represents the variations in the peak
strain in the pile with surface PGA. The peak strain is important in determining the dynamic
bending moment of the pile as given in Eq. 2. As shown in Figure 14, the peak strain increases
as the surface PGA increases. For the DPR, the peak strain does not depend on the pile
material. This implies that the deformation of the pile is predominantly determined by the
dynamic soil behavior. Consequently, the dynamic bending moment of the steel pile, which
has a larger flexural rigidity, is greater than that of the aluminum pile. Of course, this effect
may differ depending on the pile condition (e.g., end bearing pile, embedment in foundation).

5 CONCLUSIONS

The dynamic centrifuge test, which can simulate the field stress condition and real earthquake
motion, is one of the effective tools to study soil-foundation-structure-interaction (SFSI). In
this study, the seismic response of the disconnected piled raft (DPR) was studied through the
dynamic centrifuge tests. The results are summarized as follows:

118
 (1) In this study, the performance of equivalent shear beam (ESB) box was verified using
dry silica sand; the end walls of the ESB box behave in accordance with the dynamic response
of the soil deposit. In addition, a dynamic centrifuge test was performed to simulate the SFSI
behavior of a Hualien large-scale seismic test during the Chi-Chi earthquake using the ESB
box. This study showed the potential of utilizing dynamic centrifuge tests as an experimental
modeling tool for site-specific SFSI analyses.
(2) DPR has been introduced as an effective seismic design approach for pile foundations
because it reduces the seismic load on the structure, the permanent deformation of the founda-
tion, and the dynamic bending moments on the piles, simultaneously. In this study, the seismic
behavior of the DPR was evaluated by focusing on both the structure and soil-foundation
system using dynamic centrifuge tests. The dynamic centrifuge tests were performed with the
single-degree-of-freedom (SDOF) structure on a shallow foundation, a CPR, and a DPR.
From the structural point of view, using the soil-foundation yielding mechanism, especially
during rocking behavior of shallow foundation, reduced the seismic load on the superstructure
during a strong earthquake. The test results showed the seismic load on the structure was
limited by the ultimate moment capacity of the shallow foundation.
(3) The dynamic bending moment on the piles for the DPR was evaluated using centrifuge
tests. The dynamic bending moments on the pile were evaluated for not only the connected
and the disconnected conditions between pile and raft but also the material of the pile. DPR
reduced the dynamic bending moment on the piles, and the peak bending moment occurred at
mid-depth of the piles. Dynamic deformation of the pile for the DPR was predominantly
determined by not the pile material but the soil behavior. Consequently, DPR appears to be
an appropriate method for safe and effective seismic design.

ACKNOWLEDGEMENT

This research was a part of the project titled ‘ Development of performance-based seismic
design technologies for advancement in design codes for port structures‘, funded by the Minis-
try of Oceans and Fisheries, Korea. This research was supported by an R&D project of Hyun-
dai Engineering & Construction.

REFERENCES

Allmond, J.D. & Kutter, B. L. 2014. Design considerations for rocking foundations on unattached piles.
Journal of Geotechnical and Geoenvironmental Engineering 140(10): 4014058.
Anastasopoulos, I., Gazetas, G., Loli, M., Apostolou, M., & Gerolymos, N. 2010. Soil failure can be
used for seismic protection of structures. Bulletin of Earthquake Engineering 8(2): 309–326.
Antonellis, G. & Panagiotou, M. Seismic response of bridges with rocking foundations compared to
fixed-base bridges at a near-fault site. Journal of Bridge Engineering 19(5): 4014007.
Azizkandi, A.S., Baziar, M.H., & Yeznabad, A.F. 2018. 3D dynamic finite element analyses and 1 g
shaking table tests on seismic performance of connected and nonconnected piled raft foundations.
KSCE Journal of Civil Engineering 22(5): 1750–1762.
Brown, D.A., Morrison, C., & Reese, L.C. 1988. Lateral load behavior of pile group in sand. Journal of
Geotechnical Engineering 114(11): 1261–1276.
Chen, Z., Trombetta, N.W., Hutchinson, T.C., Mason, H. B., Bray, J. D., & Kutter, B. L. 2013. Seismic
system identification using centrifuge-based soil-structure interaction test data. Journal of Earthquake
Engineering 17(4): 469–496.
Fioravante, V. & Giretti, D. 2010. Contact versus noncontact piled raft foundations. Canadian Geotech-
nical Journal 47(11): 1271–1287.
Gajan, S., Kutter, B.L., Phalen, J.D., Hutchinson, T.C., & Martin, G. R. 2005. Centrifuge modeling of
load-deformation behavior of rocking shallow foundations. Soil Dynamics and Earthquake Engineering
25(7): 773–783.
Garnier, J. & Pecker, A. 1999. Use of centrifuge tests for the validation of innovative concepts in founda-
tion engineering. Earthquake Geotechnical Engineering. Edited by Seˆco e Pinto. Balkema, Rotterdam,
the Netherlands: 431–439.

119
Gazetas, G. 2015. 4th Ishihara Lecture: soil–foundation–structure systems beyond conventional seismic
failure thresholds. Soil Dynamics and Earthquake Engineering 68: 23–39.
Gazetas, G. & Apostolou, M. 2004. Nonlinear soil – structure interaction: foundation uplifting and soil
yielding. Proceedings 3rd UJNR Workshop on Soil-Structure Interaction, Menlo Park, California, USA:
1–16.
Ha, J.G., Lee, S.H., Kim, D.S., & Choo, Y.W. 2014. Simulation of soil-foundation-structure interaction
of hualien large-scale seismic test using dynamic centrifuge test. Soil Dynamics and Earthquake Engin-
eering 61–62:176–187.
Ha, J.G., Ko, K.W., Jo, S.B., Park, H.J., & Kim, D.S. 2018. Investigation of seismic performances of
unconnected pile foundations using dynamic centrifuge tests. Bulletin of Earthquake Engineering doi:
10.1007/s10518-018-00530-y.
Idriss, I.M., & Sun, J.I. 1992. User‘s manual for SHAKE9I. Ctr. for Geotech. Modeling, Dept. of Civ.
and Envir. Engrg., Univ. of California, Davis, Calif.
Kim, D.S., Kim, N.R., Choo, Y.W., & Cho, G.C. 2013a. A newly developed state-of-the-art geotechnical
centrifuge in Korea. KSCE Journal of Civil Engineering 17(1): 77–84.
Kim, D.S., Lee, S.H., Choo, Y.W., & Perdriat, J. 2013b. Self-balanced earthquake simulator on centri-
fuge and dynamic performance verification. KSCE Journal of Civil Engineering 17(4): 651–661.
Kim, D.S., Kim, D.K., & Ha, J.G. 2015a. Evaluation of seismic loads on structures considering soil-
foundation-structure interaction via centrifuge. Japanese Geotechnical Society Special Publication 2
(1):41–53, 15th Asian Regional Conference on Soil Mechanics and Geotechnical Engineering, ARC 2015:
New Innovations and Sustainability.
Kim, D.K., Lee, S.H., Kim, D.S., Choo, Y.W., & Park, H. G. 2015b. Rocking effect of a mat foundation
on the earthquake response of structures. Journal of Geotechnical and Geoenvironmental Engineering
141(1): 04014085
Ko, K.W., Ha, J.G., Park, H.J., and Kim, D.S. 2018a. Centrifuge modeling of improved design for rock-
ing foundation using short piles. Journal of Geotechnical and Geoenvironmental Engineering (in-press).
Ko, K.W., Ha, J.G., Park, H.J., and Kim, D.S. 2018b. Comparison between cyclic and dynamic rocking
behavior for embedded shallow foundation using centrifuge tests. Bulletin of Earthquake Engineering
16(11): 5171–5193.
Ko, K.W., Park, H.J., Ha, J.G., Jin, S., Song, Y.H., Song, M.J., & Kim, D.S. 2018c. Evaluation of
dynamic bending moment of disconnected piled raft via centrifuge tests. Canadian Geotechnical Jour-
nal (in-press).
Ko, K.W., Ha, J.G., Park, H.J., and Kim, D.S. 2018d. Soil-rounding effect on embedded rocking foun-
dation via horizontal slow cyclic tests. Journal of Geotechnical and Geoenvironmental Engineering 144
(3): 04018004.
Lee, S.H., Choo, Y.W., & Kim, D.S. 2013. Performance of an equivalent shear beam (ESB) model con-
tainer for dynamic geotechnical centrifuge tests. Soil Dynamics and Earthquake Engineering 44:
102–114.
Madabhushi, G. 2014. Centrifuge modelling for civil engineers. Taylor & Francis Group, London.
Martakis, P., Taeseri, D., Chatzi, E., & Laue, J. 2017. A centrifuge-based experimental verification of
soil-structure interaction effects. Soil Dynamics and Earthquake Engineering 103: 1–14.
Mergos, P.E. & Kawashima, K. 2005. Rocking isolation of a typical bridge pier on spread foundation.
Journal of Earthquake Engineering 9(sup2): 395–414.
Morris, D. V. 1981. Dynamic soil-structure interaction modelled experimentally on a geotechnical centri-
fuge. Canadian Geotechnical Journal 18(1): 40–51.
Paolucci, R., Figini, R., & Petrini, L. 2013. Introducing Dynamic Nonlinear Soil-Foundation-Structure
Interaction Effects in Displacement-Based Seismic Design. Earthquake Spectra 29(2): 475–496.
Pecker, A. 2006. Enhanced seismic design of shallow foundations: Example of the Rion Antirion bridge.
In Proceedings of the 4th Athenian Lecture on Geotechnical Engineering, Hellenic Society of Soil Mech-
anics and Geotechnical Engineering, Athens: 1–23.
Randolph, M.F. & Clancy, P. 1993. Efficient design of piled rafts. Proceedings 2nd International Geotech-
nical Seminar on Deep Foundations on Bored Auger Piles, Ghent: 119–130.
Ruesta, P.F. & Townsend, F.C. 1997. Evaluation of laterally loaded pile group at roosevelt bridge. Jour-
nal of Geotechnical and Geoenvironmental Engineering 123(12): 1153–1161.
Schofield, A.N. 1980. Cambridge geotechnical centrifuge operations. Geotechnique 25:229–267.
Tradigo, F., Pisanò, F., di Prisco, C., & Mussi, A. 2015. Non-linear soil–structure interaction in discon-
nected piled raft foundations. Computers and Geotechnics 63: 121–134.
Wong, I. H., Chang, M. F., & Cao, X.D. 2000. Raft foundations with disconnected settlement reducing
piles. Design application of raft foundations and ground slabs. Thomas Telford, London: 469–486.

120