Professional Documents
Culture Documents
Lei Su, Hua-Ping Wan*, Shaghayegh Abtahi, Yong Li, and Xian-Zhang Ling*
Corresponding author
Dr. Lei Su, Assistant Professor
School of Civil Engineering, Qingdao University of Technology, Qingdao, China
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by TU MUENCHEN on 05/27/19
E-mail: sulei@qut.edu.cn.
-1-
Page 2 of 81
Abstract
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.
system behind a quay wall in liquefiable soil and laterally spreading ground through
both large-scale shaking table test and parallel finite element (FE) simulation. A
system using the parallel modeling technique with high computational efficiency. This
FE model of the SPSI system is validated by the shaking table test results. The
validated FE model is firstly used to further explore the dynamic behavior of the SPSI
system with details on the global responses of the SPSI system and the local
responses. Secondly, the validated FE model is used for global sensitivity analysis
(GSA) to fully assess the effects of uncertain parameters on the interested dynamic
responses of the SPSI system. The experimental and numerical investigations show
clay crust at the landside and the internal forces in piles behind the quay wall. GSA
results show that the relative importance of system parameters depends on the
dynamic responses of interest, while the interaction effects among system parameters
Introduction
Single pile or pile group is one of the most common foundation solutions for
-2-
Page 3 of 81
and loss of function owing to liquefaction-induced lateral spreading, when they are
saturated sand strata. For example, both lateral deformation of the surrounding soil
and inertial loads transmitted from superstructures during an earthquake can cause
extensive damage to the pile foundation, as observed in past earthquake events, such
as the 1989 Loma Prieta earthquake, the 1995 Kobe earthquake, and the 1999 Chi-Chi
earthquake (Boulanger et al. 1995; Sugimura et al. 2004; Chu et al. 2006). These
affects the motion of the PSS, and in the meanwhile the response of PSS affects the
motion of the surrounding soil (Naggar and Bentley 2000; Maheshwari et al. 2004).
The significant role of SPSI in altering the dynamic characteristics of the PSS and the
effective seismic motions imposed at the foundation has been confirmed by a large
volume of research work (Makris et al. 1996; Stewart et al. 1999; Nikolaou et al.
2001; Su et al. 2019a, b). Traditionally, the interaction between soil, pile, and
structure has in fact been decoupled in analysis and partially considered with bias in
-3-
Page 4 of 81
design. Specifically, the inertial force resulting from the superstructure is applied as
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.
an external load to the soil-pile system, which is modeled as a beam on an elastic base
(Yao et al. 2004). Zhang and Tang (2009) pointed out that the SPSI plays either a
These research findings necessitate fundamental and practical studies of SPSI, aiming
at attaining a more reliable and economic design procedure for the PSSs with higher
confidence.
Earlier researchers (Veletsos and Meek 1974, Boulanger et al. 1999) confirmed
that SPSI can be essential in seismic analysis of PSSs, particularly in liquefiable soil,
where SPSI is a more complex process. This is because it also involves the
superstructure and pile foundation, and kinematic interaction between soil and piles.
are frequently triggered in waterfront area and reclaimed land due to seismic
excitation in the past two decades. Thus, in recent years, there has been an increasing
Laboratory test (e.g., shaking table and centrifuge tests) and numerical simulation
are effective tools for disclosing the SPSI mechanism of PSS in liquefaction-induced
-4-
Page 5 of 81
lateral spreading ground. Laboratory test allows for simulating the complicated
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.
structural systems under the controlled experimental conditions. For example, Yao et
structure in liquefiable ground using a shaking table test. Cubrinovski et al. (2006)
soil using shake table test. Dobry et al. (2010) studied liquefaction-induced lateral
Motamed and Towhata (2010a, 2010b) performed shake table test to investigate the
of pile group behind quay wall, and their seismic performance. Their findings showed
that the position of individual piles has considerable effect on the magnitude of lateral
forces on pile, and the displacement and velocity of soil most significantly affect the
distribution of the lateral forces on pile. Haeri et al. (2012) also conducted a
spreading. They found that the seismic performance of a pile group in an infinite mild
slope far from a free face is different from those located behind a quay wall. Motamed
et al. (2013) performed an E-Defense large-scale shake table test to explore the
large ground deformation. Durante et al. (2015) explored the dynamic response and
characteristics of single pile and pile group subjected to various input motions by
shaking table test. Apart from the shaking table test, the dynamic centrifuge test is
-5-
Page 6 of 81
also widely employed for the investigation of SPSI effect on the pile foundation and
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.
PSSs with a few examples as follows. Abdoun et al. (2003) conducted eight centrifuge
models on single pile and pile group subjected to earthquake-induced liquefaction and
lateral spreading. It was found that the maximum bending moment occurred at the
interface between liquefied and non-liquefied layers. Imamura et al. (2004) carried out
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by TU MUENCHEN on 05/27/19
centrifuge model test on pile group against seismically induced lateral spreading and
found that the total lateral force on pile was significantly influenced by the number of
pile and pile spacing. Brandenberg et al. (2005) conducted a total of eight dynamic
model tests on a 9-m radius centrifuge to study the behavior of single pile and pile
group in liquefiable and lateral spreading ground. González et al. (2009) conducted
investigate the effect of soil permeability on the response of single pile and pile group
deformation of quay wall and pile group through the centrifuge model test and
tests to study dynamic response of single pile as well as pile group embedded in
poorly graded sand and found that the pile position and excitation frequency are the
two main factors that affect dynamic response of the piles. Taghavi et al. (2017)
seismic behavior of pile group in soft clay consisting of over-consolidated clay layer
-6-
Page 7 of 81
method to analyze the nonlinear behavior of soil and pile subjected to lateral
spreading using a typical three-layer model. Using the beam on nonlinear Winkler
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by TU MUENCHEN on 05/27/19
foundation method, Brandenberg et al. (2007) evaluated the static response of pile
three-dimensional (3D) effective stress analysis to predict the results of shaking table
test and soil pressure on pile foundation with lateral spreading. Their results revealed
that the peak bending moment of pile foundation was significantly affected by the
non-liquefied crust layer. Ashour and Ardalan (2011) presented a new analysis
procedure for the assessment of the lateral response of the isolated pile subjected to
lateral spreading. Chang et al. (2013) and Varun et al. (2013) developed nonlinear
dynamic finite element (FE) models to study the behavior of the SPSI system in
liquefied and laterally spreading ground and validated these models against test results
shake-table experiment, and fully explored the effect of soil permeability on the
liquefaction-induced lateral response. The analysis showed that the soil permeability
-7-
Page 8 of 81
SPSI system to assess the effects of the target model parameters on the dynamic
responses of interest (Maheshwari and Sarkar 2011; Su et al. 2016; He et al. 2017; Li
et al. 2017; Su et al. 2018). In practice, the model parameters (e.g., soil and pile
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by TU MUENCHEN on 05/27/19
properties) have a random nature to some extent (Nour et al. 2003; Andrade and Borja
2006; Ching et al. 2016; Khosravikia et al. 2017; Mirzaie et al. 2017). The traditional
SA technique measures the local effects of the inputs on model outputs by perturbing
each single input slightly in turn while all other inputs are kept constant at their
nominal value, which refers to as local SA (LSA). As a result, LSA is not suitable
since it fails to explore the effects of entire space of input variation. In contrast to
LSA, global SA (GSA), as its name indicates, is able to evaluate the impacts of the
whole variations of the model parameters over their entire domain on model outputs.
uncertain model parameters, especially when the model is nonlinear, the parameter
uncertainties are significant, or the interaction among model parameters are notable
(Saltelli et al. 2008). Therefore, GSA is proposed for reliable assessment of the
influences of system properties over their whole range on the dynamic responses of
In this study, both shaking table test and numerical simulation are used to
-8-
Page 9 of 81
column pier supported on a 2×2 pile group embedded in a two-layer soil stratum. A
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.
large-scale shaking table test on the target SPSI system, is conducted subjected to
efficient simulation of the SPSI system. The developed parallel FE model is calibrated
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by TU MUENCHEN on 05/27/19
using the shaking table test results. Then, the validated parallel FE model is used to
fully understand influences of the system parameters on the dynamic responses of the
SPSI system through GSA. However, the powerful GSA comes with the high
computational cost issue because a large number of model runs are required to ensure
A shaking table test of a SPSI system behind a quay wall (see Fig. 1) has been
specifications of the shaking table are summarized in Table 1. The SPSI system under
investigation is composed of a two-layer soil stratum, a 2×2 pile group, a column pier,
and a superstructure. The soil profile consists of a non-liquefiable clay crust overlying
a saturated sand layer with a relative density of around 45%. The mass density of the
clay and sand is 1700 kg/m3 and 1900 kg/m3, respectively. The clay deposit is
-9-
Page 10 of 81
reconstituted using Harbin clay, and the sand layer is a poorly-graded medium Harbin
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.
sand classified according to the grain size distribution depicted in Fig. 2. To ensure
the uniformity and full saturation of the sand layer, the sand stratum is constructed
using the water sedimentation method (Ishihara 1993). The water table is located at
the interface between clay and sand layers, shown in Fig. 1(b), which indicates that
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by TU MUENCHEN on 05/27/19
the both sides of quay wall have the same elevation of water table.
Fig. 3 shows the schematic details of the SPSI system fabricated through
fine-aggregate concrete with the compressive strength of 12.7 MPa and mass density
of 2400 kg/m3 and steel bars (i.e., the galvanized iron wires). In this physical model,
the design of pile foundation is based on the damaged Shengli Bridge in 1976
Tangshan Earthquake. This bridge has the circular cross-section, reinforced concrete
pile foundation, and the diameter and length of prototype pile are 1.0 m and 18 m,
respectively. For the model experiment, the scaling ratio of length is set to 10.
According to the scaling law, the diameter and length of model pile is 0.1 m and 1.8
width 2.2m × height 1.6 m) and the scaling design, the piles in the SPSI system are
1350 mm long and of a circular cross-section with a diameter of 100 mm, and are
spaced with the center-to-center distance of 0.26 m. The pile cap has a block
dimension of 460 mm (length) × 460 mm (width) × 150 mm (height), and the column
pier is 450 mm in height and of a circular cross-section with a diameter of 150 mm.
The column pier and piles employ 32 and 20 longitudinal reinforcing steel bars with a
- 10 -
Page 11 of 81
diameter of 5 mm, respectively. The longitudinal reinforcing steels in piles and the
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.
column pier are embedded in the pile cap and are evenly distributed around a circle
with a concrete cover of 5 mm. The longitudinal reinforcing steel bars of the column
pier and piles are confined by 2-mm-diameter spiral stirrups. The pile cap is
reinforced with 8-mm-diameter crossing steel bars, with concrete cover thickness of
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by TU MUENCHEN on 05/27/19
10 mm at the bottom and top sides, and 20 mm at the lateral sides. More details can be
found from the configuration of the pile-cap-pier system shown in Fig. 3. The
superstructure mass is 240-kg lumped at the pier top, and the column pier and the pile
group are connected by the embedded low pile cap, which can provide a
base so that it can rotate freely about its base within the limited size of the soil
container, which allows for triggering the lateral spreading of liquefiable soil during
dynamic excitation. Since the quay wall may rotate towards the waterside during the
shaking phase, the lateral spreading would occur during dynamic excitation. Fig. 4
displays the construction process of the pile-cap-pier system, and the completed
shaking table test setup for this SPSI system is shown in Fig. 5.
OpenSees (Mckenna et al. 2000). Owing to symmetry in both geometry and loading
pattern of the SPSI system studied, one-half domain is explicitly modeled with the
- 11 -
Page 12 of 81
accordance with the dimensions of SPSI system tested on the shaking table.
Specifically, the numerical model employed in this study is validated directly using
transformed to the actual scale). Thus, it is widely accepted that the reduced-scale
tests can be directly used to verify the numerical methodologies, as oftentimes found
in the literature (e.g., Uzuoka et al. 2008; He et al. 2017; Mohajeri et al. 2004). In the
FE modeling, the soil-pile interaction is modeled by a series of rigid link elements and
presented in Fig. 7. The rigid link elements are used to characterize the effect of pile
diameter, and specifically, the length of the rigid link elements is equal to the pile
radius. Two types of zero length elements (i.e., zeroLength and zeroLengthSection in
interface. The zeroLength elements aim to axially connect the rigid link elements to
the corresponding soil nodes. Along the soil-pile interface, the zeroLengthSection
elements provide the skin-friction yield shear force to simulate the interface slip. Such
yield shear force depends on the length and depth of pile elements as well as soil
properties (i.e., friction angle and cohesion). The yield shear force of soil-pile
interface is limited by Fskin friction = (cA + σ tan ) l h / N, where l is the pile perimeter;
h is the center to center contributing height (according to the adjacent soil element
- 12 -
Page 13 of 81
heights), is the soil-pile friction angle, cA is the soil-pile adhesion, σ is the lateral
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.
effective stress, and N is the number of zeroLengthSection elements along the pile
plastic material. Herein, the friction angle and adhesion of interface is assumed to the
soil properties. Note that the end nodes of rigid link elements near soil elements have
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by TU MUENCHEN on 05/27/19
the same coordinates (i.e., soil nodes) as the nodes of zero length elements. The
inertia of the pile is simulated by defining a density parameter associated with the
pile. More details regarding the soil-pile interface modeling is referred to Su et al.
(2017). The quay wall is modeled using the parallel linear-elastic beam-column
elements to ensure no direct interaction between the sand strata on both sides.
adopted to simulate the energy dissipation and enhance the numerical system stability
(Lu 2006). The Rayleigh damping with a relatively low level of stiffness proportional
with Von Mises multi-yield surface and kinematic hardening. This model is available
dynamic behaviors of sand and clay, respectively. It is powerful for characterizing the
- 13 -
Page 14 of 81
under dynamic loading (Parra 1996). The PDMY soil model is used to characterize
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.
the nonlinear behavior of the sand (Yang 2000; Elgamal et al. 2003, 2009; Yang et al.
2003; Khosravifar et al. 2018). To be specific, the yield function of the PDMY model
follows the classical plasticity convention. It is assumed that the material elasticity is
linear and isotropic and the material plasticity is nonlinearity and anisotropy. The
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by TU MUENCHEN on 05/27/19
yield function forms a conical surface in the stress space with its apex on the
hydrostatic axis. A number of similar yield surface with a common apex and different
sizes form the hardening zone, and the outermost surface is the envelop of peak shear
strength. The flow rule of the PDMY model defines the direction of plastic strain
increments using the normality rule. The soil contractive/dilative behavior is handled
by a non-associative flow rule. The main parameters to define the PDMY model
include shear modulus, friction angle and calibration constants governing the pore
pressure buildup rate, dilative tendency and the level of liquefaction-induced cyclic
shear strain (He 2005). In contrast, the PIMY soil model is used to capture the shear
typical parameters for clay (i.e., shear modulus, cohesion, and density) are obtained
by laboratory test. Subsequently, the modulus and damping curves of clay with shear
strain are automatically fitted in the PIMY clay constitutive model based on the test
results. Such curves will be further adjusted to match the results of shake-table
experiment. It is noted that the main damping emanates from the soil nonlinear shear
- 14 -
Page 15 of 81
performed based on the following steps. First, the reasonable range of calibration
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.
the reasonable range are designed. Then, each set of parameter values are entered into
numerical model to obtain the corresponding dynamic response. Finally, the set of
parameter values with the smallest discrepancy between the computed and
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by TU MUENCHEN on 05/27/19
Table 2 summarizes the soil properties used in the constitutive model of soil.
The piles and the pier are modeled through nonlinear beam-column elements with
fiber sections while the pile cap is modeled by the solid elements. The pile elements
insert into the solid elements of pile cap, and as such they are connected through the
plastic hinges is avoided for pile design for design level earthquakes. However, the
pile foundation is often damaged during extreme earthquake events, which means that
the pile may enter into the plastic region. As such, in this modeling, the nonlinear
characteristics of pile foundation are explored from an analysis point of view. Pier
elements and pile elements are directly connected using rigid link elements. These
connection details are shown in Fig. 6(c). The mass of piles and pier is lumped into
the nodes of corresponding elements. For the fiber sections of piles and the pier, the
(i.e., Steel02 material in OpenSees), and the core and cover concrete fibers are
- 15 -
Page 16 of 81
neglected. Table 3 summarizes the properties of concrete and steel used in the fiber
section and the hysterical behavior of the uniaxial steel and concrete fibers are
illustrated in Fig. 8. Under the cyclic loading (Fig. 8a), the cyclic stress-strain
behavior of steel is demonstrated in Fig. 8(b), and the cyclic stress-strain behavior of
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by TU MUENCHEN on 05/27/19
core and cover concrete are shown in Figs. 8(c) and (d), respectively.
In line with the geometrical configuration of piles and the column pier as well as
Fig. 9. The moment-curvature behavior of the pile and pier sections under different
axial load ratios is illustrated in Fig. 10, which exhibits an obvious nonlinear feature
capacity and initial modulus of the pier is significantly larger than those of the pile
under the same curvature. This observation can be illustrated by the different
reinforcement arrangements and geometries of the pier and pile. Overall, the moment
capacities of the pile and pier both increase with an increment of the curvature.
Specifically, the moment capacities increase quickly before the curvature of 0.02, but
after that, the moment capacities grow slowly and gradually approach a constant level.
simulations (Lu et al. 2004, 2006, 2011). To reduce the computational time, the
- 16 -
Page 17 of 81
computation, the parallel computation for analysis of this target SPSI system
maintains several distinct features including (1) specifying the number of processors
for the subsequent analysis in advance, (2) partitioning the soil mesh into subdomains
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by TU MUENCHEN on 05/27/19
that are mapped to processors, (3) adding the structural mesh to each processor, and
(4) post-processing the results obtained from all processors. It is noteworthy that the
boundary nodes across different processors should be simultaneously defined for the
processors involved. To reduce the communication time among the processors, the
number of the boundary nodes is minimized by partitioning the FE mesh along the
divided into three sub-domains in total as shown in Fig. 6(d), in which the different
color meshes are mapped to the different processors. To ensure the computational
accuracy and efficiency, Chiaramonte et al. (2013) suggested that the height of the
brick elements should be less than one-quarter of the shear wave length of the softest
material (i.e., helement wave 4 ). In this simulation, the heights of the brick elements
associated with sand and clay layers are set to 0.1 m and 0.075 m, respectively. Such
element height setting satisfies the above-mentioned rule since the shear wave length
of the softest one is around 0.96 m. The mesh is properly refined near the zone of pile
Boundary conditions
- 17 -
Page 18 of 81
Different boundary conditions are applied at the two phases of the analyses (i.e.,
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.
static and dynamic) for appropriate simulation purposes as elaborated later in next
section on analysis procedures. During the phase of static analysis, the two end-planes
(i.e., perpendicular to the symmetry plane) of the FE model are constrained with
(DOFs) are constrained at each node along the base of the FE model. During the
phase of dynamic analysis, the shear beam boundary condition is applied to both sides
in the horizontal direction to ensure the same horizontal movement for the nodes at
the same height throughout the shaking process, and the x-directional DOFs of the
nodes at the base are released since the horizontal (x-directional) dynamic excitation
frequency of 2.0 Hz, and a duration of 10 sec. The pore pressure DOFs of clay deposit
above the water table are fixed, and the lateral boundary of model is impervious.
To realistically simulate the soil behavior before and after shaking, the staged
analyses are employed for the parallel FE modeling of the SPSI system to guarantee
appropriate initial condition before excitation. This is needed because the soil is
initially in the stressed state due to self-weight but has zero displacement. Therefore, a
gravity analysis for self-weight modeling is carried out first, followed by the initial
state analysis to maintain the soil stress states while resetting the soil displacement to
be zero. The obtained soil stress states are then used as the initial condition for the
- 18 -
Page 19 of 81
the first stage, the soil domain is divided into three subdomains, each being mapped
into one processor. The self-gravity is applied to each soil subdomain, and the nodal
force and pore pressure on the ground surface are applied at the waterside for gravity
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by TU MUENCHEN on 05/27/19
analysis. Higher permeability (i.e., 1.0 m/s) is assigned to the sand stratum. In the
second stage, all elements keep the same properties as those in the first stage and the
initial state analysis is employed to ensure zero displacement but non-zero stress for
the soil. In the third stage, the pile elements and the soil-pile interface elements (i.e.,
zero length elements) are added into the corresponding processor, and then the linear
elastic analysis is performed. In the fourth stage, only the soil properties are switched
from elastic to plastic whereas the other properties remain the same as those in the
previous stage. Finally, in the last stage, the actual permeability is updated for the soil
(see Table 2) while all other material properties are the same in the fourth stage.
Nonlinear dynamic time history analysis is performed with the dynamic excitation
responses with the measured responses during the shaking in order to confirm the
Before validating the FE model for the dynamic analysis phase, the results from the
- 19 -
Page 20 of 81
static analysis phase are investigated in order to ensure that initial state of the parallel
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.
FE model is authentic before the dynamic analysis. The results of the initial state
analysis are shown in Fig. 11. Fig. 11(a) shows the contour of vertical effective stress,
which is generally uniform for the same depth, while it is not strictly uniform around
the pile cap due to the self-gravity of the superstructure. Since a high permeability
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by TU MUENCHEN on 05/27/19
(i.e., 1.0 m/s) is assigned to soil in order to quickly achieve the initial condition of the
parallel FE model, the hydrostatic pressure is not influenced by the self-gravity of the
superstructure and thus is strictly uniform, as displayed in Fig. 11(b). The pore
pressure of the clay crust at the landside above the water table is zero, while the pore
pressure of the ground surface at the waterside is non-zero due to the influence of
0.2-m-deep water. It is clear that the hydrostatic pressure depends on the water table
only. To summarize, observations reveal that the initial state of the established
In the second phase of dynamic analysis, the measured responses include the
acceleration and excess pore pressure (EPP) time histories of the soil deposit as well
as the displacement histories from the piles and the pier column. The comparison of
the experiment with the computed EPPs at different depths is demonstrated in Fig. 12,
which reveals that the EPP accumulation is fast and the sand stratum liquefies during
the first few cycles of shaking. The liquefaction-level EPPs are maintained until the
end of shaking. Overall, the computed EPPs correlate well with the experimental
- 20 -
Page 21 of 81
lag behind the experimental ones before liquefaction, and fluctuate less than the
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.
experimental ones after liquefaction. The fluctuations in the experimental results are
not well captured by the numerical model. Such phenomenon may be explained by the
fact that the experimental pore pressure is significantly influenced by the shear
movement of soil container, especially near the soil container wall. Likewise, the
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by TU MUENCHEN on 05/27/19
accelerations, as illustrated in Fig. 13. The acceleration at the sand deposit close to the
ground surface attenuates quickly with time once the sand deposit liquefies, and the
acceleration amplitude of upper clay layer also keeps a low but relatively constant
level after initial liquefaction. This is mainly because the energy of base excitation
fails to transfer to the upper clay layer due to the liquefaction of underlying sand
layer. Fig. 14 demonstrates comparison between the experimental and the computed
displacements of the pile cap and the pier top. It is observed that the computed
displacement correlates well with the experimental counterpart. Based on the classical
beam theory, the experimental bending moment on pile can be obtained from the
computed bending moment on Piles 1 and 2. It can be seen that the maximum bending
moment occurs at the pile top (i.e., bottom of pile cap). Overall, the experimental and
between the experimental and computed bending moments may be due to the contact
problem of the strain gauges. In summary, the reasonably good agreement between
- 21 -
Page 22 of 81
the computed and experimental responses indicates that the parallel FE model has
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.
been successfully validated. This validated parallel FE model can be used to fully
explore the dynamic behavior of the SPSI system in the following section. It should
be noted that some local response is not well captured in the established numerical
location P (Fig. 6b) at various depths near quay wall. Obviously, the lateral soil
displacement increases cyclic by cyclic with the application of base excitation. The
maximum soil displacement occurs at the ground surface. The computed pore
pressure of location P1 (shown in Fig. 6b) near pile group is depicted in Fig. 17. In
general, the pore pressures near pile group are similar to the free field ones shown in
Fig. 12. However, the pore pressures near pile group have more fluctuation compared
to free field ones, especially at depths of 0.45 m and 1.35 m. This may be mainly
because the pore pressure near pile group is influenced by the structure vibration.
Overall, the dilation spikes of experimental and computed pore pressure in this study
are not obvious, compared to Chaloulos et al. (2013). Based on the Gonzalez et al.
(2009), the soil pressure can be obtained by back-calculating of the computed bending
moments. Fig. 18 provides the back-computed soil pressure on Piles 1 and 2. From
- 22 -
Page 23 of 81
Fig. 18, the soil pressure presents the sinusoidal trend. Such sinusoidal response
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.
pattern may be related to the overlying clay layer configuration, although the sand
stratum near pile group has liquefied (Fig. 17). This is because such clay layer can
effectively prevent the dissipation of pore pressure. The soil pressures on Pile 2 are
greater than those on Pile 1. The soil pressures on the tip and top of pile are
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by TU MUENCHEN on 05/27/19
significantly larger than that on the other depths. Furthermore, responses of the SPSI
instants are considered: before shaking, at the initial liquefaction (i.e., about 2.6 sec
indicated from Fig. 12), and at the end of shaking. Fig. 19 demonstrates the deformed
shape of the SPSI system at the initial liquefaction and the end of shaking. From the
Fig. 19(a), several observations can be made: (1) the clay crust at the landside moves
significantly towards the quay wall due to the lateral flow of the underlying
liquefiable sand; (2) the lateral displacement of the soil deposit increases with the
decrease of distance from quay wall, with the maximum (i.e., 65 mm) occurring near
the quay wall; and (3) the lateral deformation of soil near the pile cap is relatively
small, which indicates that the pile-cap system has a confinement effect on soil
deposit behind the pile. Fig. 19(b) details the deformed shape of the SPSI system at
the end of shaking. By comparing Figs. 19(a) and (b), it is found that (1) the
deformation patterns of the SPSI system associated with these two instants are
similar; (2) the lateral deformation at the end of shaking is larger than that at the
initial liquefaction, which is caused by further declination of the quay wall shown
- 23 -
Page 24 of 81
from the right panel of Fig. 19; and (3) the maximum soil lateral displacement at the
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.
end of shaking is approximately twice the displacement at the initial liquefaction (i.e.,
130 mm).
Fig. 20 shows the profiles of pile responses (i.e., axial force, shear force, and
bending moment) before shaking, at the initial liquefaction, and at the end of shaking.
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by TU MUENCHEN on 05/27/19
As seen from Fig. 20, the observations can be made: (1) before shaking, the shear
force and bending moment of piles are exactly zero and the axial forces are constant
(i.e., negative as compressive), since the piles are mainly subjected to the self-gravity
of the SPSI system only at this instant; (2) at the initial liquefaction and at the end of
shaking, the axial force, shear force, and bending moment of Piles 1 and 2 exhibit a
similar profile; (3) the Pile 1 is subject to the compressive axial force while the Pile 2
is subjected to the tensile axial force, which is caused by the overturning moment
induced by liquefaction-induced lateral spreading; and (4) the bending moment of the
Pile 1 is larger than that of the Pile 2 owing to the compressive axial force in Pile 1
and the tensile axial force in Pile 2. This is consistent with the observation made from
Fig. 10 that the larger compressive axial force will give rise to larger bending moment
capacity. The maximum bending moment occurs on the top cross-section of the pile,
and it first achieves its maximum value, which depends on the axial force, shown in
Fig. 10(a).
Fig. 21 demonstrates the profiles of responses (i.e., axial force, shear force, and
bending moment) of the pier column at the different instants. As seen from Fig. 21,
- 24 -
Page 25 of 81
several observations are emphasized: (1) the axial forces of the pier column for these
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.
three scenarios are identical because there is no acceleration along the vertical
direction during shaking; (2) the shear force and bending moment of the pier column
is zero before shaking since there is no external action on the pier at this time; (3) the
axial force, shear force, and bending moment of the pier column decrease with the
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by TU MUENCHEN on 05/27/19
height of the column; and (4) the shear force and bending moment at the end of
shaking are larger than those at the initial liquefaction because of the increased lateral
Moment-curvature responses
and the pier. Overall, the bending moment and curvature responses present a similar
periodicity. As seen from Figs. 22(a) and (b), the following observations can be made:
(1) from the depth of 0.75 m to 1.35 m, the nonlinear behavior of the
moment-curvature response on piles, especially for the Pile 1, becomes more obvious,
since the permanent bending moment and curvature become larger; and (2) compared
to Pile 2, the nonlinearity and energy dissipation of Pile 1 is more significant. Fig.
22(c) indicates that the bending moment and curvature of the section at the pile top
are obviously greater than those on pile shaft. The bending moment and curvature of
the section at the top of Pile 1 are larger than those for Pile 2 due to the compressive
axial force in Pile 1 and the tensile axial force in Pile 2. The bending moment and
- 25 -
Page 26 of 81
curvature of the section at the bottom of the pier column, is shown in Fig. 22(d). It is
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.
The pile nonlinearity is further investigated here through the strain responses of
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by TU MUENCHEN on 05/27/19
the outermost concrete and steel fibers (Fig. 9b) in the section of pile. As shown in
Fig. 23(a), below the depth of 0.75 m, the fiber strains of Pile 1 rise with the increase
of depth. The fiber strains of both core and cover concrete of Pile 1 are smaller than
their corresponding strains at the compressive strength listed in Table 3. However, the
fiber strain of reinforcing steel of Pile 1 is higher than the yield strain at yield strength
shown in Fig. 8(b), which implies that the reinforcing steel has yielded, particularly at
depths of 1.05 m and 1.35 m. Such behaviors are consistent with the
The larger fiber strain of reinforcing steel of Pile 2 at the depth of 1.35 m accounts for
the high nonlinearity of its moment-curvature response at this depth (Fig. 22b).
Likewise, the fiber strains of core and cover concrete from the section at the top
of Pile 1 are substantially larger than counterparts of Pile 2, shown in Fig. 23(c).
Obviously, all fiber strains from the section at the top of a pile are greater than those
response at the top of the pile (Fig. 22c). The comparison between Figs. 23(c) and 8(b)
shows that the fiber strain of steel bars from the section at the top of the pile exceeds
- 26 -
Page 27 of 81
the yield strain, which means that the top of the pile enters into the nonlinear stage
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.
during shaking. The fiber strains of the cover concrete at the top of Piles 1 and 2 reach
the strain at crushing strength (i.e., cu 0.004 listed in Table 3), whereas the fiber
strains of core concrete at the top of Piles 1 and 2 are below the strain at the
nonlinearity at the top of Piles 1 and 2 is mainly caused by the yield of reinforcing
steel and cover concrete. From the Fig. 23(d), all fiber strains on pier bottom are
below their corresponding yield strains (Table 3), which reveals that the pier bottom
For the shake table test model considered in this study, it is assumed that the
here. In contrast, the model parameter uncertainty (e.g., the variability in soil and steel
properties) may have significant impact on model outputs and thus studied through
(variance) of the model output. The idea behind GSA is that the total variance of the
model output can be decomposed into a collection of partial variances attributed to the
main effects of individual inputs as well as their interaction effects (Wan et al. 2017a).
- 27 -
Page 28 of 81
V V
1i d
i
1i j d
Vi , j
1i j k d
Vi , j ,k L V1,2,,d (1)
where
V V ( y)
Vi V ( E( y | xi ))
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by TU MUENCHEN on 05/27/19
Vi , j V ( E( y | x i , j )) Vi V j (2)
Vi , j ,k V ( E( y | x i , j ,k )) Vi , j Vi ,k V j ,k Vi V j Vk
L
set of xi , x j and xk ; and E() and V () denote the expectation and variance
operators, respectively.
contribution to the variance with respect to the input x u , is defined as the ratio of the
Vu
Su . (3)
V
S
1i d
i
1i j d
Si , j
1i j k d
Si , j ,k L S1,2,,d 1 . (4)
partial variance due to the interaction effect of xi and x j ; and the higher order
sensitivity index Si , j ,, p quantifies the joint influences of larger sets of inputs.
For a particular input xi , the first-order sensitivity index Si and the total
- 28 -
Page 29 of 81
Vi
Si (5)
V
V
STi 1 i (6)
V
where V i V ( E( y | x i )) , in which x i indicates the set of all inputs excluding xi .
The difference between the first-order sensitivity index Si and the total sensitivity
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by TU MUENCHEN on 05/27/19
index STi is that apart from the main effect, the latter also takes into account the
joint effects. Therefore, the larger the discrepancy between them, the more
Although GSA is a powerful and robust means for assessing the importance of
engineering system as the one studied in this work, since a huge number of model
overcome the problem of the high computational cost, polynomial chaos expansion
1 1 2
M ( ) 0 0 1 ( x ) 2 ( x , x ) 3 ( x , x , x ) L (7)
1 1 1 2 1 2 1 2 3 1 2 3
1 1 1 1 2 1 1 1 2 1 3 1
or in a compact form
- 29 -
Page 30 of 81
where the multi-dimensional indices 1 2 K d are d -tuples; are the PCE
coefficients; and (x ) are the basis functions belonging to the Askey scheme of
where mn represents the Kronecker delta that is one if m n and zero otherwise;
(x ) i 1 ( xi )
d
i
(11)
where the subscript i refers to the i -th degree of the -th univariate polynomial
In practice, the PCE representation of the model response is truncated such that
the total degree does not exceed the finite degree p , expressed as
regression method, which is very effective for solving the high-dimensional problems.
- 30 -
Page 31 of 81
Let x i i1 be the experimental design, which is usually achieved by the space-filling
N
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.
sequence). Then the original model solver is executed at each sample to collect the
coefficients involves solving the minimization problem of the l 2 -norm of the residual
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by TU MUENCHEN on 05/27/19
2
N
ˆ arg min yi (xi ) . (13)
i 1 A p ,d
Denoting
Due to the orthogonal nature of the basis functions of PCE, the variance-based
Sˆi
kL i
k2 k , k
(16)
kL
k2 k , k
SˆTi
kL Ti
k2 k , k
(17)
kL
k2 k , k
L 0 k A p ,d
.
- 31 -
Page 32 of 81
For this SPSI system, the maximum lateral soil displacement, the maximum pier
displacement, the maximum pile bending moment, and the maximum pile curvature
are investigated for GSA. We choose these four types of dynamic responses given the
fact that the lateral soil displacement is the primary factor causing the damage and
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by TU MUENCHEN on 05/27/19
failure to the piles, and the maximum displacement, the maximum bending moment,
and the maximum curvature are three essential responses that can be used to evaluate
associated with the SPSI system are considered as random variables, listed in Table 4.
Note that inclusion of material parameters alone is because (1) it is well recognized
that the material parameters consisting of soil, steel, and concrete properties are
corresponding to the given shake table model under the exact base input, and thus, the
geometrical configuration variability and the ground input variability are not included.
table model because this study focuses on investigating the dynamic response of
that although the parametric study employs the numerical model of reduced-scale
shake-table experiment, it can still be used to learn about the system partially and
- 32 -
Page 33 of 81
PCE-based anaytical GSA procedure consists of three main steps: preparing the
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.
training data, constructing PCE, and performing GSA, which are detailed in Fig. 24.
(1) Sobol sequence-based experimental design (Wan and Ren 2015), maintaining an
of samples is set to 100 based on the well-known “10d” rule, where d is number of
input (Konakli and Sudret 2016, Pan and Daniel 2017). The generated Sobol
points falling in the unit hypercube are needed to be transformed into the real
PCE coefficients can be estimated from the training data by the least square
estimator. Before the constructed PCE model is used as the surrogate of the
(3) Once the PCE is built, the variance-based sensitivity indices Si and STi can be
sensitivity indices allow for global assessment of the relative importance of model
- 33 -
Page 34 of 81
Following the implentation procedures, the GSA results of the SPSI system are
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.
shown in Fig. 25, from which the following observations can be made:
(1) For the maximum pier displacement (see Fig. 25a) and the maximum pile bending
moment (see Fig. 25c), the parameter f y is the dominant contributor indicating
for the maximum pier displacement and the maximum pile bending moment is
equal to 0.909 and 0.921, respectively, which means that it accounts for 90.9%
parameters are less influential. It is also observed that interaction effects among
parameters are not obvious since there is almost no differences between the
(2) For the maximum lateral soil displacement (see Fig. 25b), the parameters s and
f y is the most and the second most influential, respectively, while the rest are
pile bending moment, the mutual effects among parameters on the maximum
lateral soil displacement are relatively noticeable since the total sensitivity index
(3) For the maximum pile curvature (see Fig. 25d), the relative importance of the
parameters based on the sensitivity measures are ranked, from high to low, as
Es , f y , s , f ccr , f cco and the others. Like the maximum lateral soil
- 34 -
Page 35 of 81
pronounced.
The presented PCE-based GSA results provide insights into the significance of
system. GSA exposed the intrinsic relationship between the system properties and the
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by TU MUENCHEN on 05/27/19
on the GSA results. Specifically, substantial attention should be paid to the dominant
parameters since they mainly account for the variation of the dynamic responses,
Concluding remarks
soil-pile-structure interaction (SPSI) system, and SPSI has been found to have
comprehensive investigation into the dynamic behavior of the SPSI system in terms of
large-scale shaking table test, parallel finite element (FE) modeling, and global
FE analysis of the SPSI system. In general, the FE-predicted responses exhibit a good
- 35 -
Page 36 of 81
agreement with the experimental counterparts, which confirms that the developed 3D
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.
parallel FE model is validated for simulating the SPSI system subjected to the
further investigated using the validated 3D parallel FE model. Based on the validated
dynamic responses of the SPSI system. The key findings from this study are:
The clay crust at the landside moves significantly towards the quay wall because
of the lateral flow of the underlying liquefiable sand during shaking. The pile-cap
system has a confinement effect on soil deposit behind the pile, which can be
confirmed by the relatively small lateral deformation of soil near the pile cap.
The bending moment developed in the Pile 1 closer to waterfront is larger than
that of the Pile 2, since the liquefaction-induced lateral spreading leads to the
compressive axial force in Pile 1 and the tensile axial force in Pile 2. The shear
force and bending moment at the end of shaking are larger than those at the initial
especially for the Pile 1, becomes more notable with increased depth because the
permanent bending moment and curvature increase. The bending moment and
curvature on pile top are considerably greater than those on pile shaft, and its
- 36 -
Page 37 of 81
Certain parameters play a dominant role in affecting dynamic responses, and the
influences of the same system parameter on the different dynamic responses can
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by TU MUENCHEN on 05/27/19
target dynamic responses are not evident since their first-order and total sensitivity
It is worth mentioning that the presented numerical model in this study has a
further development. However, the provided analysis results can contribute to the
Acknowledgements
This work was financially supported by the National Natural Science Foundation
References
Abdoun, T., Dobry, R., O'rourke, T.D., and Goh, S. 2003. Pile response to lateral
- 37 -
Page 38 of 81
Andrade, J.E., and Borja, R.I. 2006. Quantifying sensitivity of local site response
Ashour, M., and Ardalan, H. 2011. Piles in fully liquefied soils with lateral spread.
Biondini, F., Bontempi, F., Frangopol, D.M., and Malerba, P.G. 2004. Reliability of
Boulanger R.W., Curras C.J., Kutter B.L., Wilson D.W., and Abghari, A. 1999.
Boulanger R.W., Idriss I.M., and Mejia L.H. 1995. Investigation and evaluation of
Brandenberg, S.J., Boulanger, R.W., Kutter, B.L., and Chang, D. 2005. Behavior of
Brandenberg, S.J., Boulanger, R.W., Kutter, B.L., and Chang, D. 2007. Static
- 38 -
Page 39 of 81
133(9): 1055-1066.
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.
Cardoso, J.B., de Almeida, J.R., Dias, J.M., and Coelho, P.G. 2008. Structural
reliability analysis using Monte Carlo simulation and neural networks. Advances
Chaloulos, Y.K., Bouckovalas, G.D., and Karamitros, D.K. 2013. Pile response in
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by TU MUENCHEN on 05/27/19
Chang, D., Boulanger, R.W., Brandenberg, S.J., and Kutter, B.L. 2013. FEM analysis
Chiaramonte, M.M., Arduino, P., Lehman, D.E., and Roeder, C.W. 2013. Seismic
Ching, J., Lin, G.H., Chen, J.R., and Phoon, K.K. 2016. Transformation models for
effective friction angle and relative density calibrated based on generic database
Chu, D.B., Stewart, J.P., Youd, T.L., and Chu, B.L. 2006. Liquefaction-induced
132(12): 1549-1565.
Crespo-Minguillón C., and Casas, J.R. 1998. Fatigue reliability analysis of prestressed
- 39 -
Page 40 of 81
Cubrinovski, M., and Ishihara, K. 2004. Simplified method for analysis of piles
119-133.
Cubrinovski, M., Kokusho, T., and Ishihara, K. 2006. Interpretation from large-scale
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by TU MUENCHEN on 05/27/19
shake table tests on piles undergoing lateral spreading in liquefied soils. Soil
Cubrinovski, M., Uzuoka, R., Sugita, H., Tokimatsu, K., Sato, M., Ishihara, K.,
Dobry, R., Thevanayagam, S., Medina, C., Bethapudi, R., Elgamal, A., Bennett, V.,
Abdoun, T., Zeghal, M., El Shamy, U., and Mercado, V.M. 2010. Mechanics of
Durante, M.G., Di Sarno, L., Mylonakis, G., Taylor, C.A., and Simonelli, A.L. 2015.
Elgamal, A., Lu, J., and Forcellini, D. 2009. Mitigation of liquefaction-induced lateral
- 40 -
Page 41 of 81
Elgamal, A., Yang, Z., Parra, E., and Ragheb, A. 2003. Modeling of cyclic mobility in
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.
Elkateb, T., Chalaturnyk, R., and Robertson, P.K. 2003. An overview of soil
González, L., Abdoun, T., and Dobry, R. 2009. Effect of soil permeability on
Haeri, S.M., Kavand, A., Rahmani, I., and Torabi, H. 2012. Response of a group of
He, L., Ramirez, J., Lu, J., Tang, L., Elgamal, A., and Tokimatsu, K. 2017. Lateral
Hussein, M.N., Tobita, T., Iai, S., and Karray, M. 2016. Soil-pile-structure kinematic
Imamura, S., Hagiwara, T., Tsukamoto, Y., and Ishihara, K. 2004. Response of pile
groups against seismically induced lateral flow in centrifuge model tests. Soils
- 41 -
Page 42 of 81
43 (3): 351-451.
Javankhoshdel, S., and Bathurst, R.J. 2014. Simplified probabilistic slope stability
soil variability and their effects on bearing capacity and settlement of shallow
8(4): 352-364.
Khosravifar, A., Elgamal, A., Lu, J., and Li, J. 2018. A 3D model for
Khosravikia, F., Mahsuli, M., and Ghannad, M.A. 2017. Probabilistic evaluation of
Kwon, O.S., and Elnashai, A. 2006. The effect of material and ground motion
- 42 -
Page 43 of 81
Li, Y., Huang, S., Lin, C., Gu, Q., and Qiu, Z. 2017. Response sensitivity analysis for
Lu, J. 2006. Parallel finite element modeling of earthquake ground response and
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by TU MUENCHEN on 05/27/19
Lu, J., Elgamal, A., Yan, L., Law, K.H., and Conte, J.P. 2011. Large-scale numerical
Lu, J., Peng, J., Elgamal, A., Yang, Z., and Law, K.H. 2004. Parallel finite element
Maheshwari, B.K., Truman, K.Z., Naggar, M.H.E., and Gould, P.L. 2004.
118-133.
- 43 -
Page 44 of 81
Mckenna, F., and Fenves, G.L. 2008. Using the OpenSees interpreter on parallel
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.
McKenna, F., Fenves, G., and Scott, M. 2000. Open System for Earthquake
Mirzaie, F., Mahsuli, M., and Ghannad, M.A. 2017. Probabilistic analysis of
Mohajeri, M., Kobayashi, Y., Kawaguchi, K., and Sato. M. 2004. Numerical study on
lateral spreading of liquefied ground behind a sheet pile model in a large scale
Canada.
Motamed, R., and Towhata, I. 2010a. Shaking table model tests on pile groups behind
Motamed, R., and Towhata, I. 2010b. Mitigation measures for pile groups behind
quay walls subjected to lateral flow of liquefied soil: Shake table model tests. Soil
Motamed, R., Towhata, I., Honda, T., Tabata, K., and Abe, A. 2013. Pile group
- 44 -
Page 45 of 81
Naggar, M.H.E., and Bentley, K.J. 2000. Dynamic analysis for laterally loaded piles
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.
Nikolaou, S., Mylonakis, G., Gazetas, G., and Tazoh, T. 2001. Kinematic pile bending
425-440.
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by TU MUENCHEN on 05/27/19
Nour, A., Slimani, A., Laouami, N., and Afra, H. 2003. Finite element model for the
Pan, Q., and Dias, D. 2017. Sliced inverse regression-based sparse polynomial chaos
including cyclic mobility and dilation response in soil systems. PhD Thesis,
Saltelli, A., Ratto, M., Andres, T., Campolongo, F., Cariboni, J., Gatelli, D., Saisana,
M., and Tarantola, S. 2008. Global sensitivity analysis the primer. John Wiley &
Sons.
Stewart, J.P., Fenves, G.L., and Seed, R.B. 1999. Seismic soil-structure interaction in
- 45 -
Page 46 of 81
Su, L., Lu, J., Elgamal, A., and Arulmoli, A.K. 2017. Seismic performance of a
Su, L., Tang, L., Ling, X., Liu, C., and Zhang, X. 2016. Pile response to
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by TU MUENCHEN on 05/27/19
Su, L., Wan, H. P., Bi, K., Li, Y., Lu, J., Ling X. Z., Elgamal, A., and Arulmoli, A. K.
Su, L., Wan, H.P., Dong, Y., Frangopol, D.M., and Ling, X. 2019b. Seismic fragility
Su, L., Wan, H.P., Li, Y., and Ling, X. 2018. Soil-pile-quay wall system with
- 46 -
Page 47 of 81
Taghavi, A., Muraleetharan, K.K., and Miller, G.A. 2017. Nonlinear seismic behavior
Tasiopoulou, P., Gerolymos, N., Tazoh, T., and Gazetas, G. 2012. Pile-group
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by TU MUENCHEN on 05/27/19
Uzuoka, R., Cubrinovski, M., Sugita, H., Sato, M., Tokimatsu, K., Sento, N., Kazama,
M., Zhang, F., Yashima, A., and Oka, F. 2008. Prediction of pile response to
- 47 -
Page 48 of 81
Wan, H. P., Todd, M. D., and Ren, W. X. 2017a. Statistical framework for sensitivity
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.
143(9):04017093.
Wan, H.P., Ren, W.X., and Todd, M.D. 2017b. An efficient metamodeling approach
Yang, Z., Elgamal, A., and Parra, E. 2003. Computational model for cyclic mobility
Yao, S., Kobayashi, K., Yoshida, N., and Matsuo, H. 2004. Interactive behavior of
shake table tests. Soil Dynamics and Earthquake Engineering, 24(5): 397-409.
Zhang, J., and Tang, Y. 2009. Dimensional analysis of structures with translating and
- 48 -
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by TU MUENCHEN on 05/27/19
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record. Page 49 of 81
- 49 -
Table 4 Uncertainties in modeling parameters for SPSI system.
Table 3 Properties of concrete and steel used in the pier and piles.
Page 50 of 81
Fig. 1 Configuration and sensor deployment of shaking table test model: (a)
Fig. 2 Grain size distribution curve of Harbin sand used in the test.
Fig. 3 Configuration of the pile-supported structure system (unit: mm): (a) pier
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by TU MUENCHEN on 05/27/19
reinforcement; (b) configuration of pile and cap; (c) reinforcement of pile cap;
(d) pile reinforcement; (e) 1-1 cross section; (f) 2-2 cross section.
Fig. 6 Finite element modeling of soil-pile-structure system with domain partition for
parallel computing: (a) elevation view; (b) 3D view; (c) pile group and quay
Fig. 7 Three-dimensional soil-pile interaction modeling: (a) pile and surrounding soil;
Fig. 8 Stress-strain responses of steel and concrete under cyclic loading: (a) loading
Fig. 9 Pile and pier cross sections and fiber sections: (a) pile cross section; (b) pile
fiber section; (c) pier cross section; (d) pier fiber section.
Fig. 10 Moment-curvature behavior of pile and pier sections under different axial
Fig. 11 Model response contours before shaking: (a) vertical effective stress; (b) pore
- 50 -
Page 51 of 81
pressure.
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.
Fig. 14 Displacement time histories of the pile cap and pier top.
Fig. 15 Comparison of experimental and computed bending moment: (a) Pile 1; (b)
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by TU MUENCHEN on 05/27/19
Pile 2
Fig. 16 Computed lateral soil displacement time histories of location P (Fig. 6b) at
Fig. 17 Computed pore pressure of Location P1 (shown in Figure 6b) near pile group
Fig. 19 Deformation of the SPSI system with the contour indicating the horizontal
Fig. 20 Profiles of the axial force, shear force, and bending moment responses of Piles
1 and 2.
Fig. 21 Profiles of the axial force, shear force, and bending moment responses of the
pier column.
Fig. 22 Bending moment-curvature response: (a) Pile 1; (b) Pile 2; (c) top section of
Fig. 23 Strain responses of steel, core and cover concrete for piles and pier: (a) Pile 1;
(b) Pile 2; (c) top section of the pile; (d) bottom section of the pier bottom.
- 51 -
Page 52 of 81
dynamic responses: (a) the maximum pier top displacement; (b) the maximum
lateral soil displacement; (c) the maximum bending moment in piles; (d) the
- 52 -
Page 53 of 81
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by TU MUENCHEN on 05/27/19
Table 3 Properties of concrete and steel used in the pier and piles.
Parameter Cover Core
Concrete compressive strength (MPa) 8.4 12.7
Strain at concrete compressive strength 0.002 0.005
Concrete crushing strength (MPa) 0 10.2
Strain at concrete crushing strength 0.004 0.018
Steel yield strength (MPa) 47
Steel elastic modulus (MPa) 2.06×105
Strain at tensile strength 2.28×10-4
Steel strain-hardening ratio 0.001
use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official versio Page 56 of 81
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by TU MUENCHEN on 05/27/19
Figure 2. Grain size distribution curve of Harbin sand used in the test.
Page 58 of 81
use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official versio
Page 59 of 81
32Φ5
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by TU MUENCHEN on 05/27/19
Φ2@10 1 2
460 Φ8@35
20
100
20
450
Pile 2 Pile 1
100
260
460
150
Φ8@35
(a)
20
100
1 2
20
150
(b) (c)
20Φ5 Φ8@35
20 20 20 20
Φ8@35
1350
Φ2@10
(e) (f)
10
10
(d)
Figure 3. Configuration of the pile-supported structure system (unit: mm): (a) pier reinforcement; (b) configuration of pile and cap; (c)
reinforcement of pile cap; (d) pile reinforcement; (e) 1-1 cross section; (f) 2-2 cross section.
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by TU MUENCHEN on 05/27/19
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.
(a) (b)
Pile
6DOF Node
Soil-pile
interfac
Pile 4DOF Node
e
element zeroLength element
Figure 7. Three-dimensional soil-pile interaction modeling: (a) pile and surrounding soil; (b)
connection of soil-pile interface.
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by TU MUENCHEN on 05/27/19
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.
5mm
(c)
(a)
5mm
Φ3@10mm
Φ3@10mm
150mm 100mm
20Φ5mm
fibers
Outermost
fibers
Outermost
Figure 9. Pile and pier cross sections and fiber sections: (a) pile cross section; (b) pile fiber section;
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by TU MUENCHEN on 05/27/19
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.
(b) pier.
Figure 10. Moment-curvature behavior of pile and pier sections under different axial forces: (a) pile;
Page 66 of 81
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by TU MUENCHEN on 05/27/19
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record. Page 67 of 81
Figure 11. Model response contours before shaking: (a) vertical effective stress; (b) pore pressure.
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by TU MUENCHEN on 05/27/19
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.
Figure 14. Displacement time histories of the pile cap and pier top.
Page 70 of 81
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by TU MUENCHEN on 05/27/19
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record. Page 71 of 81
Figure 15. Comparison of experimental and computed bending moment: (a) Pile 1; (b) Pile 2.
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by TU MUENCHEN on 05/27/19
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.
Figure 17. Computed pore pressure of Location P1 (shown in Figure 6b) near pile group.
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by TU MUENCHEN on 05/27/19
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.
Figure 18. Computed soil pressure on pile: (a) Pile 1; (b) Pile 2.
Page 74 of 81
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by TU MUENCHEN on 05/27/19
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record. Page 75 of 81
Figure 20. Profiles of the axial force, shear force, and bending moment responses of Piles 1 and 2.
Page 76 of 81
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by TU MUENCHEN on 05/27/19
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record. Page 77 of 81
Figure 21. Profiles of the axial force, shear force, and bending moment responses of the pier column.
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by TU MUENCHEN on 05/27/19
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.
top section of the pile; (d) bottom section of the pier bottom.
Figure 23. Strain responses of steel, core and cover concrete for piles and pier: (a) Pile 1; (b) Pile 2; (c)
use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official versio Page 80 of 81
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by TU MUENCHEN on 05/27/19
Step 1: Prepare training data Step 2: Construct PCE Step 3: Perform GSA
Figure 25. Global sensitivity results with respect to various structural properties of dynamic
responses: (a) the maximum pier top displacement; (b) the maximum lateral soil displacement; (c) the
maximum bending moment in piles; (d) the maximum section curvature in piles.