Received: 24 February 2023 | Revised: 19 April 2023 | Accepted: 8 May 2023

DOI: 10.1002/dug2.12054

RESEARCH ARTICLE

Nonlinear anisotropic finite element analysis of liquefiable

tunnel–sand–pile interaction under seismic excitation

Md. Foisal Haque

Department of Civil Engineering, Bangladesh Abstract

University of Engineering and Technology,
Dhaka, Bangladesh Nonlinear time‐history analysis can be used to determine the liquefiable behaviors of
the tunnel–sand–pile interaction (TSPI) model with the consideration of sand
Correspondence anisotropy. This study presents the nonlinear response of the TSPI model with the
Md. Foisal Haque, Department of Civil existence of liquefaction under seismic excitation. The analysis reveals that tunnel and
Engineering, Bangladesh University of pile behave as isotropic elements, while sand shows isotropic, orthotropic, and
Engineering and Technology, Dhaka 1000,
Bangladesh. anisotropic characteristics. Three constitutive models including UBC3D‐PLM (two
Email: [email protected] yield surfaces associated with the hardening rule), NGI‐ADP (yielding with associated
plastic potential function), and a user‐specified constitutive model are adopted to
Funding information
None evaluate the isotropic, orthotropic, and anisotropic behaviors of sand. On this basis,
two finite element‐based codes (ETABS 18.1.1 and Plaxis 3D) are used to evaluate
sand behaviors and responses. Responses of the tunnel, sand, pile, and excess pore
pressure ratio are recorded in the interaction zone by varying the pile diameter, tunnel
diameter, and tunnel–pile clearance. Compared with the orthotropic and isotropic
conditions, lower variations of results are found in the anisotropic condition, except
for the case of generation of excess pore pressure. In addition, the present reanalysis
results are in agreement with previous analytical and case study results, which further
indicates the effectiveness of the finite element‐based numerical codes.

KEYWORDS
anisotropy, finite element, liquefaction, nonlinear, seismic excitation, tunnel–sand–pile interaction

Highlights
• FEA of the tunnel–sand–pile interaction model is carried out for various
material conditions of the sand.
• A simplified analytical formula is proposed to calculate liquefaction effect.
• Square root of the sum of squares responses of the tunnel, pile, and sand under
seismic excitation are predicted.
• The liquefaction phenomenon of sand under seismic excitation is evaluated.

1 | INTRODUCTION Pasternak foundation model can be used to study lateral

deformation of soil, whereas the Winkler model can only be
Sand shows anisotropic behaviors, while concrete is used to examine the vertical movement of soil. However,
considered as an isotropic material. Therefore, the both lateral and vertical deformations will occur in soil
tunnel–sand–pile interaction (TSPI) shows anisotropic/ under seismic loading in numerical and analytical studies.
orthogonally anisotropic and isotropic mechanisms because The dynamic soil–structure interaction is affected by these
of the material behaviors of soil and concrete. It is difficult deformation characteristics (Choudhury et al., 2018;
to clearly describe the complex interaction due to the Lakirouhani & Valioskooyi, 2017; Lu et al., 2018).
limitations of numerical methods. Several analytical formu- Tunneling‐induced vertical and lateral movements of
lae (Franza et al., 2017; Marshall & Haji, 2015; Zhang liquefiable sand are higher compared with the pure solid
et al., 2018) of TSPI have been developed with consideration state of the soil. In addition, liquefaction phenomena of
of isotropic conditions and assumptions based on Winkler different ground conditions, including (a) soil with pile, (b)
and Pasternak foundation models under static loading. The soil with a tunnel, (c) soil with a tunnel and pile, and (d)

This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided
the original work is properly cited.
© 2023 The Authors. Deep Underground Science and Engineering published by John Wiley & Sons Australia, Ltd on behalf of China University of Mining and
Technology.

Deep Undergr Sci Eng. 2023;2:275–285. wileyonlinelibrary.com/journal/duse | 275

 27701328, 2023, 3, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/dug2.12054 by CochraneChina, Wiley Online Library on [15/11/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
276 | HAQUE

saturated soil with a tunnel of the sea bed, have been The main objective of this research is to evaluate the
described by several authors based on various research nonlinear response of tunnel and sand of the TSPI model
methods such as numerical and analytical analyses, the in liquefiable conditions under seismic excitation based
centrifuge test, and the shaking table test (Finn & on sand anisotropy. The results are confirmed on the
Fujita, 2002; Liyanapathirana & Poulos, 2002; basis of the isotropic and orthotropic properties of sand
Liyanathirana & Poulos, 2002; Orang et al., 2021; Taylor because the results of sand under several uncertain
& Madabhushi, 2020; Wang et al., 2022; Yue et al., 2020; conditions are identical to those obtained through field
Zhao et al., 2021). Normally, liquefaction occurs due to investigations.
seismic shaking, which impacts surface and underground
structures. Analytical formulae for tunnels only under
seismic excitation without consideration of liquefaction have 2 | MECHANISM OF THE TSPI
been derived by several authors (Hashash et al., 2001; John F IN IT E E L E ME N T M O D E L
& Zahrah, 1985). However, the analytical formulae of
liquefaction have not yet been formulated to describe the 2.1 | Geometric and material properties
behaviors of TSPI. The probability of liquefaction of
cohesive soil is lower compared with that of soil with lower The 3D finite element model of the TSPI system shows
cohesion. Therefore, the response of a tunnel containing the isotropic and anisotropic properties of sand. The
group piles on both sides in cohesive soil was excellent under orthotropic behavior of sand is represented by the
seismic shaking (Meguid & Mattar, 2009). Vertical seismic proposed 2D equivalent system, while nonlinear behav-
excitation exerts less impact on an underground tunnel iors of sand exist in every type of analysis procedure.
station than horizontal transverse earthquake motion Therefore, the geometric properties of 3D and idealized
(Zhang et al., 2019). Moreover, the influence of the tunnel 2D models of the TSPI system are shown in Figure 1.
on the pile is greater (Abdullah & Taha, 2013). However, the This model consists of a single circular tunnel with four
liquefaction mechanism of anisotropic sand is very complex, equal‐diameter piles with a pile cap. The equivalent
and currently, no formulae/numerical methods are available system is composed of a tunnel and a single pile that is
that properly describe this mechanism, although the rectangular in shape. In this system, the tunnel and pile
University of British Colombia 3D‐Plaxis Liquefaction are inside the sand. The area of the equivalent tunnel is
Model (UBC 3D‐PLM) material model of sand can describe equal to the circumferential area of the circular tunnel,
this phenomenon under isotropic conditions by considering and the equivalent area of the pile is equal to the entire
the two‐yield surface mechanism of nonlinearity longitudinal surface area of the pile of the main system,
(Plaxis, 2020). In addition, the NGI‐ADP material model including the pile cap. Meanwhile, the length of the pile
shows anisotropic behaviors of clay in the nonlinear range in the equivalent system is the same as that in the main
without consideration of liquefaction (Plaxis, 2020). The system. Seismic shaking of the recorded Kobe earth-
equivalent composite plate element of the TSPI model is quake (January 17, 1995) with a maximum 8.2 m/s2
analyzed by ETABS, 2017 (18.1.1) in lieu of the 3D full acceleration and 48 s duration is applied at the bottom of
model such as Plaxis. This plate can demonstrate ortho- the TSPI models along the transverse direction of the
tropic/orthogonally anisotropic characteristics. The excess tunnel, as shown in Figure 1c. The 500 kN static loading
pore pressure caused by liquefaction can be generated is applied at the top of the pile cap. The liquefaction
based on the function of stress and strain of sand effect of sand in the 2D model is determined using the
(Kinney, 2018). Time‐dependent properties of an isotropic modified analytical formula. For the finite element
material (concrete) are selected according to the standard analysis, constant parameters of the TSPI model are
during the equivalent plane strain analysis (CEB‐FIP, 1998). shown in Table 1.
It is difficult to examine the liquefaction effect without an Meanwhile, variable parameters are the tunnel outer
appropriate material model for finite element analysis. diameters (DT) of 8, 12, and 16 m, pile diameters (dP) of
The modified formula of excess pore pressure ratio is used 400, 500, and 600 mm, and tunnel–pile clear distances
during the plane strain analysis of the TSPI model to (dT–P) of 1, 2, 3, 4, 5, 6, 7, 8, 9, and 10 m. In addition,
address this phenomenon. The shaking table test has been other parameters of the 3D and equivalent 2D models of
conducted to evaluate the interaction of a soil–double‐box the TSPI system are obtained from general algebraic
utility tunnel (with and without joint) under seismic formulations: (a) length (L), width (B), and height (H) of
excitation and dynamic loadings (Han et al., 2022; Zhang the 3D model, (b) width (BT) and height (HT) of the
et al., 2020). The maximum bending moments of the equivalent solid tunnel, (c) diameter of the equivalent pile
utility tunnel are determined at each corner for jointed and (DP), (d) length (LPC) and width (BPC) of the pile cap,
jointless conditions. In addition, reliability analysis is and (e) TSPI zone width of the equivalent system. In all
necessary to predict the precise seismic impact on the soil cases, the TSPI zone width of the equivalent system is
as well as the structure. Several scholars (Zhang et al., higher than that of the 3D model, and every element of
2021, 2022) studied reliability and probabilistic stability to the TSPI system is considered a solid plate element in the
evaluate the behavior of sloping ground under seismic equivalent system. Moreover, this plate element is
excitations and observed satisfactory responses to the slope. divided by the four‐noded rectangular element with a
The seismic landslide response of the sand has been recorded size of 100 mm × 100 mm. In the 3D finite element model
by the shaking table test (Chen et al., 2019), and the of the TSPI system, sand, tunnel, and pile are considered
maximum particle displacement was obtained at a 30° to be 10‐noded tetrahedral elements that follow the four‐
sloping angle of sand. point Gauss integration rule, a six‐noded triangular
 27701328, 2023, 3, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/dug2.12054 by CochraneChina, Wiley Online Library on [15/11/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
DEEP UNDERGROUND SCIENCE AND ENGINEERING | 277

F I G U R E 1 Geometric properties and loading arrangements of the tunnel–sand–pile interaction (TSPI) system with 3D and 2D equivalent
configurations. (a) Equivalent 2D form of the TSPI model, (b) 3D configuration of the TSPI model, (c) input seimic excitation, (d) 3D TSPI
modelling by Plaxis, and (e) equivalent 2D TSPI modelling by ETABS.

T A B L E 1 Constant parameters of the tunnel–sand–pile interaction element that follows the three‐point Gauss integration
(TSPI) model in this research. rule, and a two‐noded line element that follows the three‐
Constant parameters Values point Gauss integration rule, respectively. The maximum
and minimum sizes of elements for the 3D analysis are
Right side cover, C (m) 3
set to 100 and 50 mm, respectively. Elastic moduli of the
Backside cover, E (m) 3 tunnel and pile are considered to be 70 000 and
Front side cover, F (m) 3 22 000 MPa, respectively. In addition, unit weight and
Poisson's ratio of the tunnel and pile materials are set to
Thickness of the tunnel, tT (mm) 400 24 kN/m3 and 0.22, respectively. Some parameters of the
Thickness of the pile cap, tP (mm) 500 isotropic tunnel and piles are chosen for the 2D
equivalent analysis, which are as follows: (a) thermal
Length of the pile, LP (m) 20
expansion of the tunnel and pile of 0.0000085, (b)
Cover of the pile cap, Z1 (m) 2 crushing strength of the tunnel and pile of 40 and
Left cover of the tunnel, X1 (m) 3 21 MPa, respectively, (c) negligible tension capacity of
concrete for the tunnel and pile, (d) acceptance levels of
Cover of the tunnel invert, Z3 (m) 8
the tunnel and pile in compressions such as immediate
Cover of the tunnel crown, Z4 (m) 8 occupancy (IO), with 0.003, life safety (LS), with 0.006,
Clear spacing of pile‐to‐pile 1.5 times of pile diameter and collapse prevention (CP), with 0.015, respectively, (e)
hysteresis type of tunnel and pile of concrete, (f) friction
Number of piles 4 nos. and dilatational angles of the Drucker–Prager parame-
Side cover of the pile cap (mm) 150 ters of 0°, (g) damping of the tunnel and pile of 5%, (h)
design period of the tunnel and pile of 150 years, and (i)
 27701328, 2023, 3, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/dug2.12054 by CochraneChina, Wiley Online Library on [15/11/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
278 | HAQUE

creep and shrinkage of the tunnel and pile depending on three sides. Hinges are applied on one side to confirm the
the standard (CEB‐FIP, 1998). Similarly, some parame- base of the TSPI model and the other two sides are
ters are chosen for the 2D equivalent analysis of assumed to be roller support to ensure uniformity of the
orthotropic sand, which are as follows: (a) unit weight adjacent flexible layers. The static load is applied at the
of 16 kN/m3, (b) modulus of elasticities in various top of the equivalent pile and the seismic load is applied
directions of 10.0, 9.0, and 9.5 MPa, respectively, (c) at the bottom of this system. The equivalent pile is the
thermal expansions of 0.0000045, 0.0000035, and combination of the actual number of piles with the cap.
0.0000040, respectively, (d) Poisson's ratios of 0.450, This is shown in Figure 1a. The area of the equivalent
0.480, and 0.465, respectively, (e) no tension acceptance tunnel is equal to the circumferential area of the circular
criteria of sand, (f) acceptance levels of sand in tunnel.
compression for immediate occupancy, with 0.008, life
safety, with 0.015, and collapse prevention, with 0.025,
respectively, (g) kinematic hysteresis model, (h) 3 | C O N S T I T U T I V E MO D E L S
Drucker–Prager friction and dilatational angles of 30° OF SAND
and 0°, and (i) radiation damping of the sand of 30%.
Materials of tunnel, pile, and sand for the 3D analysis are The constitutive model of material describes stress–strain
taken to be the same as those for the 2D analysis. To behaviors of the structures under external loadings in the
perform the plane strain analysis, the thickness of both numerical analysis. However, due to the anisotropic
sand and tunnel is considered to be 1 m. Also, the characteristics of sand, it is difficult to develop a material
thicknesses of the pile are set as 330, 400, and 474 mm. model to study the liquefaction effect. This research
These above‐mentioned geometric, material, and loading explores the liquefiable anisotropic phenomenon based
properties are selected for parametric studies in this on user‐specified parameters in ETABS and the NGI‐
research. Implicit finite element analysis is performed ADP model in Plaxis.
until convergence is achieved. The Newmark method is
considered for the dynamic numerical analysis in the
undrained condition of the TSPI model. When saturated 3.1 | Isotropic UBC3D‐PLM (liquefaction
properties of sand are considered, the Newmark method model)
is more effective. In addition, boundary conditions of the
TSPI model are as follows: (a) fixed at the bottom, (b) The UBC3D‐PLM model was developed for cohesionless
free at the top, and (c) roller support at both sides to isotropic soil under undrained conditions subjected to
ensure flexibility of the adjacent medium of all sides of seismic excitation. This model describes the effective
this model. stress under the elastoplastic condition to simulate the
The reason for choosing the finite element model is liquefaction behavior of sand under seismic loading. This
that the model can predict the variations of the model includes the Mohr–Coulomb yield condition in a
interactive response of the tunnel and pile by considering 3D principal stress space for primary loading and a yield
the isotropic, orthotropic, and anisotropic characteristics surface with a simplified kinematic hardening rule for
of sand. In addition, values of various dimensional secondary loading. Also, a modified nonassociated
parameters of the TSPI model are taken from the plastic potential function based on Drucker–Prager's
standard format to reflect the tunnel size and pile criterion was used for the primary yield surface with the
diameter in the equivalent system. assumption of stress–strain coaxially in the deviatoric
plane for a stress path starting from the isotropic line.
Two yield surfaces of the UBC 3D‐PLM model ensured
2.2 | Modeling process smooth transition to the liquefied state of soil to allow
distinction between primary and secondary loadings. In
2.2.1 | 3D modeling in Plaxis addition, this model incorporated a densification law
through a secondary yield surface with a kinematic
Three steps are followed in Plaxis modeling. First, the hardening rule, which improved the precision of the
volume of sand is created by the volume element. evolution of the excess pore pressure, while this surface
Second, sand is removed from the inner portion of the generated lower plastic deformations compared to the
tunnel. Third, pile along with the cap is inserted in the primary yield surface. Therefore, this model was used to
sand. Two loading steps are applied after completing the predict the liquefiable interactive response of the TSPI
modeling process. First, static loading is applied on top model in an isotropic condition.
of the pile cap. Then, seismic excitation is applied at the
bottom of the model along the transverse direction of the
tunnel. Figure 1d shows the 3D model of Plaxis. 3.2 | User‐specified orthotropic equivalent
plate model (orthogonally anisotropic)
2.2.2 | 2D equivalent modeling in ETABS Sand is considered an orthotropic/orthogonally aniso-
tropic material in the equivalent plane strain analysis; it
2D equivalent modeling is performed using ETABS behaves like a plate element to follow the kinematic
18.1.1 in Figure 1e. Sand, pile, and tunnel are treated as hardening rule during the nonlinear analysis. Shear
equivalent plate elements and boundaries are applied on modulus is added with the stiffness matrix of sand, and
 27701328, 2023, 3, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/dug2.12054 by CochraneChina, Wiley Online Library on [15/11/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
DEEP UNDERGROUND SCIENCE AND ENGINEERING | 279

the liquefaction phenomenon is examined using the

proposed simplified/modified analytical formula. The
excess pore water pressure ratio is the function of the
modulus of elasticity (E), stress, and strain of the plate
according to the proposed formula. The simplified form
of the excess pore water pressure ratio (RU) (Equation 1)
is established by considering the Poisson's ratio and
internal friction angle of 0.450 and 30°, respectively. F I G U R E 2 Various locations for the evaluation of tunnel, sand,
Poisson's ratio varies because of the orthotropic and pile responses within the interaction zone.
characteristics.

RU = 0.6(Eε P /σP ) − 1. (1) displacements, sand displacements, pile lateral displace-

ment, and excess pore water pressure ratio decrease
For this plate element, effective stress can be gradually with an increase in the tunnel–pile clearance
expressed as stress (σP) of the equivalent plate of sand because of the increase in the sand volume within the
at any stage of the nonlinear analysis. Also, shear strain interaction zone. The anisotropic characteristic of sand
of sand is equal to the major axis strain (εP) of the plate. shows minimum displacements of the tunnel, pile, and
The kinematic hysteresis model is related to the sand because of the inverse proportionality relation
kinematic hardening behavior, which is commonly between the material stiffness matrix and displacement.
observed in the metals. This model is used in this In addition, the excess pore pressure ratio shows the
research for the equivalent plate (metal) of the sand. maximum response due to the direct proportionality
Meanwhile, this model is also appropriate for ductile relation.
material as it is capable of dissipating a significant
amount of energy.
4.1 | Lateral response of the tunnel in the
interaction zone
3.3 | Anisotropic NGI‐ADP model
Lateral displacement of the tunnel is recorded at a
The NGI‐ADP model is developed for undrained loading specific point along the center line of the interaction zone
of clay, which can describe the soil–structure interaction of the right face on the tunnel body. The square root of
behavior excluding liquefaction. As a combination of the the sum of squares (SRSS) value of the displacement
Tresca approximation and the modified Von Mises under the seismic excitation is determined because SRSS
plastic potential function, this model is also used in the represents the resultant displacement among all displace-
present research for the evaluation of the liquefaction ments at a point during the entire duration of seismic
behaviors of sand. The NGI‐ADP model is created for shaking. Lateral displacement of the tunnel increases
the 1D‐anisotropy condition. Parameters for the devel- gradually with increasing tunnel diameter (Figure 3),
opment of the yield and plastic potential functions are as because a larger diameter of the tunnel reduces the ratio
follows: (a) shear strength, (b) vertical stress, (c) of model width/height and the diameter of the tunnel.
horizontal stress, and (d) shear strain. The stress path‐ For this reason, lateral disturbance of the tunnel under
dependent hardening parameter shows varied plastic seismic excitation is higher than that of the lower‐
shear strain during compression and extension. Also, the diameter tunnel. The displacement increases with
hardening of the yield curves is characterized by slightly increasing pile diameter because a larger‐diameter pile
distorted elliptical shapes, while the shape is dependent reduces the width of the interaction zone, which
on the interpolation function. increased the effect of seismic shaking on the tunnel
body. Compared with the isotropic and anisotropic
characteristics of sand, the orthotropic property of sand
4 | NONLINEAR R ESPONSE OF reduces displacements and lateral spreading. A larger
T H E TS P I MO D E L I N T H E diameter of pile shows the maximum variations of
INTERACTION ZONE results. Sometimes, a smaller‐diameter pile shows higher
variations of results because of the isotropic, orthotropic,
The nonlinear response of the TSPI model in the and anisotropic characteristics of sand. However, as
interaction zone indicates the response of the tunnel, shown in Figure 3a, tunnel lateral displacement due to
sand, and pile within this zone, which is shown in the orthotropic condition is increased by 25% from
Figure 2. Sand shows isotropic (I), orthotropic (O), and anisotropy for pile and tunnel diameters of 600 mm and
anisotropic (A) behaviors within this zone. Tunnel and 8 m, respectively. Variations of results in the orthotropic
pile behave as isotropic materials. The response is and isotropic properties of sand are less than 5% at the
denoted by the lateral and vertical movements of the location of 1 m tunnel–pile clearance for 400 mm
tunnel, pile, and sand. Lateral and vertical movements of diameter of the pile and 12 m diameter of the tunnel in
the tunnel and sand are shown at locations F1 and F2, Figure 3b. Similarly, the maximum variation of lateral
respectively, in Figure 2. Similarly, pile lateral displace- spreading of the tunnel is observed to be approximately
ment and excess pore water pressure ratio are expressed 37% for the isotropic and anisotropic properties of sand
at locations F3 and F2, respectively, in Figure 2. Tunnel at the tunnel–pile clearance of 1 m for the pile diameter
 27701328, 2023, 3, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/dug2.12054 by CochraneChina, Wiley Online Library on [15/11/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
280 | HAQUE

of 600 mm and 16 m diameter of the tunnel in Figure 3c. condition of the sand for the tunnel diameter of 12 m in
Possible variations of results of various characteristics of Figure 4b and 16 m in Figure 4c. This difference may be
sand, diameters of the tunnel, diameters of the pile, and due to the shear failure of sand in this location due to the
the tunnel–pile clearance are shown in Figure 3. Differ- exclusion of the shear modulus from the formulation of
ences in the results can be obviously observed in these the material stiffness matrix in an isotropic condition.
graphs. The tunnel diameter of 8 m in Figure 4a yields minimum
variations in results. The maximum displacement occurs
at the location of the tunnel–pile clearance of 1 m for the
4.2 | Vertical displacement of the tunnel pile diameter of 600 mm in the isotropic condition of
sand with tunnel diameter of 16 mm in Figure 4c. In
Vertical displacement of the tunnel is smaller than the addition, the minimum displacement occurs at the
lateral displacement. SRSS vertical displacements in tunnel–pile clearance of 10 m for the pile diameter of
various conditions are shown in Figure 4. Differences 400 mm in the anisotropic condition of the sand with
in the results for all cases at the location of tunnel–pile tunnel diameter of 8 m in Figure 4a.
clearance of 10 m are similar to those of other diameters,
except for the diameters of 500 and 600 mm in an
isotropic condition. This is because the disturbance of 4.3 | Lateral spreading of sand in the
sand under seismic excitation is not reduced significantly interaction zone
in this condition for piles of these diameters. This
displacement variation is less for the pile diameter of Lateral spreading of sand in the middle of the interaction
400 mm in the anisotropic condition of sand with a zone is higher compared with that in other locations
tunnel diameter of 8 m. The difference in displacements because resonance may be induced under seismic
for tunnel diameters of 8, 12, and 16 mm is found to be excitation. These variations are shown in Figure 5. As
approximately 40% in the anisotropic condition of sand can be seen, displacements decrease gradually with an
for the pile diameter of 400 mm. The location of the increase in the tunnel–pile clearance. Percentages of
tunnel–pile clearance of 9 m shows the maximum incremental lateral displacements from orthotropic to
displacement difference of approximately 12 mm between anisotropic are found to be 29% of 400 mm diameter pile,
the pile diameters of 400 and 500 mm in the isotropic 32% of 500 mm diameter pile, and 35% of 600 mm

F I G U R E 3 Lateral tunnel displacements for different tunnel diameters of (a) 8 m, (b) 12 m, and (c) 16 m with variations of pile diameters and
material characteristics of sand under seismic excitation.

F I G U R E 4 Tunnel vertical movements for different tunnel diameters of (a) 8 m, (b) 12 m, and (c) 16 m with variations of pile diameters and
material properties of sand under seismic excitation.
 27701328, 2023, 3, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/dug2.12054 by CochraneChina, Wiley Online Library on [15/11/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
DEEP UNDERGROUND SCIENCE AND ENGINEERING | 281

diameter pile at the tunnel–pile clearance of 1 m for the than that of the anisotropic condition. The SRSS vertical
tunnel diameter of 8 m in Figure 5a. Maximum lateral movement of sand is greater for the pile diameter of
spreading of sand occurs at 1 m tunnel–pile clearance for 600 mm at the tunnel–pile clearance of 1 m, which
the pile diameter of 600 mm in an isotropic condition of indicates the poor condition of the sand. Degradation
sand. Magnitudes of these displacements for tunnel of the residual stress from the hysteresis loop may occur,
diameters of 8, 12, and 16 m are observed to be 50, 88, which is generated during seismic loading. In addition,
and 110 mm, respectively, as shown in Figure 5. Dis- lower tunnel–pile clearance may lead to a higher
placement of this location in the isotropic condition is probability of occurrence of liquefaction. The probability
larger compared with the orthotropic and anisotropic decreases gradually with an increase in the width of the
properties of the sand. Higher lateral displacement of interaction zone. The vertical displacements among
sand represents losses of shear strength and the isotropic, orthotropic, and anisotropic characteristics of
probability of occurrence of liquefaction to ensure the sand show little variations at the tunnel–pile clearance
development of excess pore water pressure within the of 10 m in almost all cases, except for the pile diameter of
interaction zone. 600 mm in isotropic and orthotropic conditions of sand.

4.4 | Vertical movement of sand 4.5 | Lateral displacement of pile in the

interaction zone
The difference of the vertical displacement between two
different diameter piles increases with an increase in the The smaller‐diameter pile shows higher lateral spreading
tunnel diameter. Similarly, it decreases with an increase in all cases. The maximum lateral spreading of the pile is
in the tunnel–pile clearance in all cases. Meanwhile, at the lowest location of tunnel–pile clearance for all
maximum vertical displacements in the anisotropic cases. The SRSS displacements for tunnel diameters of 8,
condition of sand for tunnel diameters of 8, 12, and 12, and 16 m in Figure 7a–c are 85, 110, and 123 mm,
16 m in Figure 6a–c are predicted to be 30, 52, and respectively. The lateral displacement of tunnel–pile
70 mm, respectively, for the pile diameter of 600 mm at clearance of 4 and 5 m shows similar variations for the
the tunnel–pile clearance of 1 m. Displacements for isotropic, orthotropic, and anisotropic conditions of
isotropic and orthotropic conditions of sand are higher sand with pile diameters of 600, 500, and 300 mm,

F I G U R E 5 Lateral spreading of sand for different tunnel diameters of (a) 8 m, (b) 12 m, and (c) 16 m with variations of pile diameters and
material properties of sand under seismic excitation.

F I G U R E 6 Vertical displacements of sand for different tunnel diameters of (a) 8 m, (b) 12 m, and (c) 16 m with variations of pile diameters and
material properties of sand under the seismic excitation.
 27701328, 2023, 3, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/dug2.12054 by CochraneChina, Wiley Online Library on [15/11/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
282 | HAQUE

F I G U R E 7 Lateral displacements of the pile for different tunnel diameters of (a) 8 m, (b) 12 m, and (c) 16 m with variations of pile diameters and
material properties of sand under seismic excitation.

F I G U R E 8 Excess pore pressure generation for different tunnel diameters of (a) 8 m, (b) 12 m, and (c) 16 m with variations of pile diameters and
material properties of sand under seismic excitation.

respectively. This is found for piles of all diameters. This of the interaction zone may rapidly absorb seismic
variation indicates the poorer shearing properties of the excitation so as to reduce the development of excess pore
material stiffness matrix of sand during the successive pressure within the sand particle. As the tunnel–pile
iteration of the nonlinear analysis procedure. In addition, clearance of 10 m decreases (Figure 8b), sand exists as a
the pile lateral displacement is less than 2 mm at the fluid in the tunnel and pile of 12 m and 400 mm diameters,
tunnel–pile clearance of 10 m for 600 mm pile diameter in respectively. Liquefaction occurs in all cases for the 8 m
all cases, which ensures lower interaction effect on the tunnel diameter in Figure 8a, except for the 400 mm pile
pile. In some locations, lateral displacements are very diameter in the isotropic condition of sand for 8, 9, and
close because of the rapid degradation of the hysteresis 10 m tunnel–pile clearance. In addition, sand exists as a
loop during the nonlinear analysis, which indicates the fluid for all diameters of the pile and the tunnel–pile
growth of residual stress within the sand particle. Lower clearance of the TSPI model with 16 m tunnel diameter in
lateral spreading of the pile in various conditions Figure 8c, which indicates that the TSPI model is
indicates the reduction of the interaction of the TSPI unstable. The liquefaction phenomenon of the interaction
model with an increase in the tunnel–pile clearance. zone grows rapidly in other locations because of the
combined effect of tunnel and pile within this zone. It is
very difficult to ensure that no liquefaction occurs in other
4.6 | Variation of excess pore pressure in the zones. However, the fully fluid state of the interaction
interaction zone zone may indicate the marked decrease in the shear
strength of the TSPI model.
The excess pore pressure ratio is estimated to predict the
existence of liquefaction. The anisotropic property of sand
shows the maximum excess pore pressure ratio. The excess 5 | VERIFICATION OF
pore pressure ratio increases with an increase in the width NUMERICAL CODES W ITH
of the interaction zone, pile diameter, and tunnel PREVIOUS STUDIES
diameter. In almost all cases, the excess pore pressure
ratio exceeds the unit value, which leads to liquefaction. Two different finite element‐based numerical codes
The effect of liquefaction can be optimized with the width (ETABS and Plaxis 3D) were used to perform the
increment of the interaction zone because the higher width analysis of the TSPI model. These codes have been
 27701328, 2023, 3, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/dug2.12054 by CochraneChina, Wiley Online Library on [15/11/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
DEEP UNDERGROUND SCIENCE AND ENGINEERING | 283

verified by previously published work, which indicates previous studies may indicate the high level of
the effectiveness of the analysis procedure in this study. accuracy of the present numerical analysis. In addi-
tion, the difference in the results between previous
studies and the present study for orthotropic and
5.1 | Verification with the case study isotropic conditions of sand is higher than the
of experimental work anisotropy due to the absence of shear modulus from
the material stiffness matrix.
Fifty case study reports were reviewed based on the
physical model, the full‐scale test, and the centrifugal
test, while the response of the interaction zone such as 5.2 | Verification with the previously solved
settlement and relative distance was recorded (Dias & example by the GEPAN 2D
Bezuijen, 2015). The present study re‐analyzed previ-
ous work by using two different codes to confirm the The previous studies addressed a general problem of
level of accuracy. The loading and other data were 2 × 2 existing piles in a group, which was constructed in
determined based on the case studies. A common the adjacent tunnel. No external load was included in the
model was re‐considered for the evaluation of the analysis by the GEPAN computer program (Loganathan
tunnel–soil–pile interaction, as shown in Figure 9a. All et al., 2001). However, the present study conducts this
variable dimensions in Figure 9a were obtained from analysis again using ETABS and Plaxis 3D. Figure 10a
the 50 case study reports found from a previous paper shows the geometry of the model. Variations of results
(Dias & Bezuijen, 2015). The dimensions are as between the present study and previous studies are
follows: (a) pile length (Zp), (b) tunnel diameter (Dt), shown in Figure 10b. Material properties are as follows:
(c) pile diameter (Dp), (d) distance from the center line (a) modulus of elasticity of the pile: 30 GPa, (b) Poisson's
of the tunnel to the top surface of the ground (Zt), (e) ratio of the pile: 0.25, (c) undrained cohesion of soil:
lateral distance from the tunnel center line to the pile 60 kPa, (d) modulus of elasticity of soil: 24 MPa, and (e)
center line (Ld), and (f) concentrated load on the pile Poisson's ratio of soil: 0.50. Similarly, other parameters
(WL). The 2D analysis of the common model is are as follows: (a) group pile assumed to be fixed rotation
performed by considering the same approximation of at the pile cap, (b) infinite thickness of the soil layer, and
the present TSPI model, and the width of the common (c) consideration of only the elastic behavior of the pile
model is considered to be 1 m to conduct the 3D and soil. Similar parameters for isotropic, orthotropic,
numerical analysis. The results obtained are slightly and anisotropic conditions of the soil are used for
different from the previous study in Figure 9b. The performing the finite element analysis as in the present
interaction effect is mitigated with the tunnel–pile TSPI model. Lateral displacements of the pile cap are
center‐to‐center distance of 1 m based on the previous recorded with variations of the relative depth of the
study because of the smaller settlement ratio of the pile tunnel corresponding to the center‐to‐center distance of
and soil. However, this study obtains identical results the group piles and tunnel in Figure 10b. According to
at approximately 2 m tunnel–pile center‐to‐center Figure 10b, displacements decrease gradually with an
distance. The settlement ratio decreases exponentially, increase in the magnitude of the relative depth of the
starting from the 2 m tunnel‐pile center‐to‐center tunnel for all material conditions of soil and previous
distance in Figure 9b. The settlement ratio for studies. The minimum difference in the results between
anisotropic, orthotropic, and isotropic conditions of the orthotropic condition of soil and the previous study
soil increases by 14.3%, 50%, and 60%, respectively, is found to be approximately 3% along the whole relative
compared with the previous study. The smaller depth of the tunnel. The minimum difference in the
difference in anisotropy between the present and results between the isotropy of soil and the previous

F I G U R E 9 Tunnel–soil–pile interaction model (re‐drawn after, Dias & Bezuijen, 2015) and variations in results. (a) Problem statement and (b)
verification results.
 27701328, 2023, 3, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/dug2.12054 by CochraneChina, Wiley Online Library on [15/11/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
284 | HAQUE

F I G U R E 10 Tunnel–soil–pile interaction model (re‐drawn after, Loganathan et al., 2001) and variations in results between the present study and
previous studies. (a) Statement of the problem and (b) validation results.

study is found to be approximately 14% at a relative 3. According to this study, the order of the displacement
depth of 1.25. In addition, this difference is observed to increment rates based on the sand material model is
be approximately 9% between the anisotropic condition isotropic > orthotropic > anisotropic. Moreover, the
of soil and the previous study at a relative depth of 0.5. order of the excess pore pressure ratio increment rate
is isotropic < orthotropic < anisotropic.
4. A newly developed simplified formula for the evalua-
6 | CONCLUSIONS tion of the excess pore pressure ratio of an equivalent
plate element of sand shows considerable variations in
Tunnel, pile, and sand nonlinear responses are expressed the three‐dimensional analysis, which demonstrates
for isotropic, orthotropic, and anisotropic conditions of the suitability of use of this formula.
sand under seismic excitation. According to parametric
studies, the anisotropic property shows lower variations
for the movements of the tunnel, pile, and sand, although For this research, some parameters are fixed. Future
for this property, higher values are found in the case of studies should consider further variations of these
excess pore pressure ratio for the isotropic and orthotropic parameters, which will supplement the results of the
conditions. Variations of results are recorded based on a current study.
total of 270 finite element models. Ninety models are
developed for anisotropy, and the rest are created for CONF LICT OF INTEREST ST ATE MENT
isotropic and orthotropic conditions of sand. In addition, The author declares no conflict of interest.
the results of the anisotropic property of soil are similar to
those of previous studies, which indicates the appropriate DA TA AV AI LA BI LIT Y S TA TEME NT
use of numerical codes for the soil material model. The Data sharing is not applicable to this article as no new
results of the present study are summarized as follows: data were created or analyzed in this study.

1. Displacements of the tunnel, pile, and sand decrease OR CI D

gradually with an increase of the interaction zone Md. Foisal Haque https://orcid.org/0000-0002-
width for isotropic, orthotropic, and anisotropic 0559-8801
conditions of sand, which indicates lower interaction
effect on displacements. However, the possibility of REFE RENC ES
interaction is not fully mitigated because of the Abdullah MH, Taha MR. A review of the effects of tunneling on
exponential variations of results. adjacent piles. EJGE. 2013;18:2739‐2762.
CEB‐FIP. CEB‐FIP Model Code 1990. 2nd ed. Thomas Telford
2. Excess pore pressure ratio is higher value for the
Services Ltd; 1998.
anisotropic condition of sand than for the isotropic Chen Z, Yang P, Liu H, Zhang W, Wu C. Characteristics analysis of
and orthotropic conditions, which indicates the granular landslide using shaking table model test. Soil Dyn
occurrence of liquefaction under seismic excitation. Earthquake Eng. 2019;126:105761.
It increases with the pile and tunnel diameters and Choudhury D, Patil M, Ranjith PG, Zhao J. Dynamic Tunnel‐Soil
decreases gradually with an increase in the tunnel–pile Interaction in Soft Soils Considering Site‐Specific Seismic Ground
Response. Indian Institute of Science; 2018.
clearance. In addition, excess pore pressure ratio is
Dias TGS, Bezuijen A. Data analysis of pile tunnel interaction.
always greater than the unit value for anisotropy, J Geotech Geoenviron Eng. 2015;141:1‐16.
which indicates the liquefied state of sand for three ETABS. CSI Analysis Reference Manual. Version. 18th ed. Computers
diameters of the tunnel in the present study. & Structures Inc; 2017.
 27701328, 2023, 3, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/dug2.12054 by CochraneChina, Wiley Online Library on [15/11/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
DEEP UNDERGROUND SCIENCE AND ENGINEERING | 285

Finn WDL, Fujita N. Piles in liquefiable soils: seismic analysis and Zhang W, Meng F, Chen F, Liu H. Effects of spatial variability of weak
design issues. Soil Dyn Earthquake Eng. 2002;22:731‐742. layer and seismic randomness on rock slope stability and
Franza A, Marshall AM, Haji T, Abdelatif AO, Carbonari S, reliability analysis. Soil Dyn Earthquake Eng. 2021;146:106735.
Morici M. A simplified elastic analysis of tunnel‐piled structure Zhang W, Wu J, Gu X, Han L, Wang L. Probabilistic stability analysis
interaction. Tunnel Underground Space Technol. 2017;61:104‐121. of embankment slopes considering the spatial variability of soil
Han L, Liu H, Zhang W, et al. Seismic behaviors of utility tunnel‐soil properties and seismic randomness. J Mountain Sci. 2022;19(5):
system: with and without joint connections. Underground Space. 1464‐1474.
2022;7:798‐811. Zhao K, Zhu S, Bai X, et al. Seismic response of immersed tunnel in
Hashash YMA, Hook JJ, Schmidt B, I‐Chiang Yao J. Seismic design liquefiable seabed considering ocean environmental loads. Tunnel
and analysis of underground structures. Tunnel Underground Underground Space Technol Incorporating Trenchless Technol Res.
Space Technol. 2001;16:247‐293. 2021;115:1‐14.
John CMS, Zahrah TF. Aseismic Design of Underground Structures.
National Science Foundation; 1985.
Kinney LS. Pore Pressure Generation and Shear Modulus Degradation AUTHOR BIOGRAPHY
during Laminar Shear Box Testing with Prefabricated Vertical
Drains. Brigham Young University; 2018.
Lakirouhani A, Valioskooyi S. Seismic analysis of rectangular tunnels
Md. Foisal Haque completed his degree
(cut and cover method), soil‐structure interaction. Amirkabir of Bachelor of Science (B.Sc.) in Civil
J Civil Eng. 2017;49(1):45‐48. Engineering from Rajshahi University
Liyanapathirana DS, Poulos HG. A numerical model for dynamic soil of Engineering and Technology
liquefaction analysis. Soil Dyn Earthquake Eng. 2002;22:1007‐1015. (RUET) in 2014. Also, he finished his
Liyanathirana DS, Poulos HG. Numerical simulation of soil liquefac-
tion due to earthquake loading. Soil Dyn Earthquake Eng.
Master of Science (M.Sc.) degree in
2002;22:511‐523. Geotechnical Engineering from Bangladesh Univer-
Loganathan N, Poulos HG, Xu KJ. Ground and pile‐group responses sity of Engineering and Technology (BUET) in
due to tunnelling. Soils Found. 2001;41(1):57‐67. 2019. Now, he is an active student of Doctor of
Lu Q, Chen S, Chang Y, He C. Comparison between numerical and Philosophy (PhD) in Geotechnical Engineering at
analytical analysis of the dynamic behavior of circular tunnels. BUET. His specialization is structural and tunnel
Earth Sci Res J. 2018;22(2):119‐128.
Marshall AM, Haji T. An analytical study of tunnel–pile interaction.
engineering. Professionally, he has been a structural
Tunnel Underground Space Technol. 2015;45:43‐51. engineer in several consulting firms in Bangladesh
Meguid MA, Mattar J. Investigation of tunnel‐soil‐pile interaction in since January 2015. Now, he is a PhD fellow at
cohesive soils. J Geotech Geoenviron Eng. 2009;135(7):973‐979. BUET. Also, he has teaching experience as a
Orang MJ, Boushehri R, Motamed R, Prabhakaran A, Elgamal A. An lecturer and assistant professor at the World
experimental evaluation of helical piles as a liquefaction‐induced University of Bangladesh (WUB) and the Interna-
building settlement mitigation measure. Soil Dyn Earthquake Eng.
2021;151:106994.
tional University of Business Agriculture and
Plaxis. 2020. Material Models Manual. s.l.:Bentley. Technology (IUBAT), respectively. It covers one
Taylor EJ, Madabhushi SPG. Remediation of liquefaction‐induced and a half years of teaching experience. His main
floatation of non‐circular tunnels. Tunnel Underground Space research interest is the interaction of underground‐
Technol. 2020;98:103301. soil‐aboveground structures including the develop-
Wang R, Zhu T, Yu JK, Zhang JM. Influence of vertical ground ment of the small/large scale shake table test model,
motion on the seismic response of underground structures and
underground‐aboveground structure systems in liquefiable
numerical simulation, analytical formulations, and
ground. Tunnel Underground Space Technol Incorporating finite element formulations considering isotropy/
Trenchless Technol Res. 2022;122:1‐12. orthotropic/anisotropic properties of soil.
Yue F, Liu B, Zhu B, et al. Shaking table test and numerical simulation on
seismic performance of prefabricated corrugated steel utility tunnels
on liquefiable ground. Soil Dyn Earthquake Eng. 2020;141:1‐18.
Zhang Z, Bilotta E, Yuan Y, et al. Experimental assessment of the
effect of vertical earthquake motion on underground metro
station. Appl Sci. 2019;9(25):5182.
How to cite this article: Haque MF. Nonlinear
Zhang W, Han L, Feng L, et al. Study on seismic behaviors of a double anisotropic finite element analysis of liquefiable
box utility tunnel with joint connections using shaking table tunnel–sand–pile interaction under seismic
model tests. Soil Dyn Earthquake Eng. 2020;136:106118. excitation. Deep Undergr Sci Eng. 2023;2(3):
Zhang Z, Huang M, Xu C, Jiang Y, Wang W. Simplified solution for 275‐285. doi:10.1002/dug2.12054
tunnel‐soil‐pile interaction in Pasternak's foundation model.
Tunnel Underground Space Technol. 2018;78:146‐158.