Structures 58 (2023) 105488

Contents lists available at ScienceDirect

Structures
journal homepage: www.elsevier.com/locate/structures

A new seismic input method for layered soil and its application in dynamic
response analysis of underground structure
YaFei Zhai a, b, ZhiYu Song a, b, *, Yun Liu a, Tao Gao a, Zhengxiong Bai a, b, ZhongHui Bi c
a
Yellow River Engineering Consulting Co., Ltd., Zhengzhou 450003, China
b
Key Laboratory of Water Management and Water Security for Yellow River Basin of Ministry of Water Resources, Zhengzhou 450003, China
c
College of Water Conservancy and Hydropower Engineering, Hohai University, Nanjing, Jiangsuprovince 210098, China

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

Keywords: With the acceleration of urbanization, the seismic problem of underground engineering is getting more and more
Layered soil attention. In actual projects, foundation soils often show a complex layered distribution. In terms of layered
Infinite element ground motion input, this study propose a new layered ground motion input method based on infinite element
Ground motion input
boundary, which avoids the application of spring and damping elements on the boundary nodes of layered soil,
Nonlinearity
Underground structure
simplifies the ground motion input process and has high accuracy at the same time. To reasonably simulate the
nonlinear seismic response of layered soil, the equivalent linear model of soil is nested to Mohr-Coulomb (MC)
model to establish the viscoelastic-plastic model. The seismic response analysis procedure of the soil-structure
monolithic model considering the nonlinearity of the layered soil is given. On this basis, the soil-structure
nonlinear seismic response analysis is carried out to study the effect of burial depth on the seismic perfor­
mance of underground structures, taking a project as an example. The results show that with the increase of the
burial depth of the structure, the top–bottom displacement difference shows a change law of first increasing and
then decreasing under the action of earthquake; From the internal force response, the most unfavorable burial
depth exists due to the joint influence of the burial depth of the structure and the characteristics of the layered
soil; For the same burial depth condition, the dynamic bending moment caused by the seismic load action is
larger compared to the shear response.

1. Introduction layered foundations [13–18], which uses finite element method to solve
the internal analysis domain and boundary element method at the
Unlike homogeneous foundations, wave propagation in layered soils boundary. To more conveniently solve problems in infinite domains,
requires consideration of reflection and transmission at the interface of scholars have proposed infinite elements on the basis of finite elements,
different soil layers in addition to absorption and elastic restoring forces which are often used together with finite elements to solve complex
of scattered waves by the soil layers [1–4]. In terms of seismic input boundary problems [19–22], and are an effective supplement to finite
methods for layered foundations, there are mainly finite element element in solving unbounded domain problems [23]. It has better co­
methods (FEM) [5–7] and boundary element methods (BEM)[8–12]. ordination with FEM and has advantages over other methods such as
The BEM is a numerical method developed on the basis of classical in­ BEM to solve infinite domain problems.
tegral equations and finite element methods to solve differential equa­ Among the site seismic response analysis methods, the equivalent
tions. It can conveniently handle unbounded domain problems with linearization method is considered as an effective method to solve the
high accuracy, but is generally limited to linear problem solving. The seismic response of nonlinear soil layers. Many experts and scholars
FEM is suitable for complex geometric shapes and boundary conditions, have conducted analytical studies on seismic problems of underground
and is suitable for solving nonlinear and heterogeneous problems. structures based on equivalent linearization of soils [24–26]. However,
Currently, it is widely used in engineering. Combining the advantages of because the model does not reflect the accumulation of soil deformation
FEM and BEM, some scholars have proposed the scaled boundary finite and the effect of considering the stress path on the damping ratio, the
element method (SBFEM) to solve the seismic response problem of calculation error is large when considering large soil strains. Therefore,

* Corresponding author at: Yellow River Engineering Consulting Co., Ltd., Zhengzhou 450003, China.
E-mail address: [email protected] (Z. Song).

https://doi.org/10.1016/j.istruc.2023.105488
Received 15 May 2023; Received in revised form 28 September 2023; Accepted 30 October 2023
Available online 11 November 2023
2352-0124/© 2023 Institution of Structural Engineers. Published by Elsevier Ltd. All rights reserved.
Y. Zhai et al. Structures 58 (2023) 105488

some scholars proposed to combine the equivalent linear viscoelastic 2.1. Infinite element boundary theory
model with the MC model to establish a viscoelastic-plastic model to
solve the dynamic response problem of underground structures [27]. The concept of infinite element boundary was first proposed by
Tsinidis et al. [28,29] considered the soil as a viscoelastic-plastic model Ungless [34]. In the infinite element theory of dynamic analysis, the
and compared the numerical calculations with the results of dynamic effect of wave reflection on the near-field region can be neglected by
tests to analyze, which showed that the liner stiffness, soil-tunnel con­ introducing damping on the boundary to absorb the radiated energy of
tact characteristics and soil plasticity have significant effects on the the wave. Taking a compressional wave (P-wave) as an example [35], its
structural dynamic response; Argyroudis et al. [30] used the equation of motion can be expressed as:
viscoelastic-plastic model to simulate the nonlinear behavior of soils
under seismic action, and proposed a method for the numerical con­ ∂2 ε λ + 2G 2
= ∇ ε = v2p ∇2 ε (1)
struction of seismic susceptibility curves for shallow buried underpass ∂t2 ρ
tunnels considering soil-structure interaction (SSI) and liner aging ef­
Where, ε is the volumetric strain; vp is the compressional wave velocity;
fects, with engineering applications; Huang et al. [1] revealed the in­
λ, G are the elastic constants; ρ is the propagation medium density.
fluence law of regional site conditions on the analysis of tunnel seismic
Considering a compressional wave propagating only in the x-direc­
susceptibility based on the viscoelastic-plastic model. In most studies,
tion, its displacement at the truncation boundary can be expressed as:
the vibration method is used as the ground motion input, that is, both
{
sides use equal displacement boundary to input seismic waves from the u = f1 (x − vp t)
ux = ux1 + ux2 , uy = uz = 0 x1 (2)
bottom of the model, and the influence of boundary effect is attenuated ux2 = f2 (x + vp t)
by increasing the range of the model foundation. For layered soils, the
average shear modulus and damping ratio of each soil layer are obtained Where, ux, uy, uz indicate different directions and displacement, ux1 in­
by solving the one-dimensional (1D) free field reaction analysis. The dicates the displacement caused by the incident wave, ux2 indicates the
vibration input method is only applicable to rigid bedrock although the displacement caused by the reflected wave.
calculation steps are simple [31]. The seismic response of soil under From the physical equations of elastic mechanics:
seismic action is closely related to the depth position of the soil layer,
σ x = 2Gεx + λε (3)
and when the soil layer is thick, the use of the average shear modulus
method also causes certain calculation errors. Where, εx represents the normal strain in the x-direction, there is εx =
When solving the dynamic response of underground structures, both
ε = u′x1 + u′x2 .
the model ground motion input method and the simulation method of
Introducing distributed damping at the truncated boundary of the
soil nonlinearity will directly affect the calculation accuracy and the
foundation, the damping force is obtained as:
reliability of the results. On the basis of existing research results and
combined with infinite element method, the author proposes an infinite σ damp = − dp u̇x (4)
element substructure wave input method for the seismic input problem
of layered foundations, and provides an analysis process for solving the Where, u̇x indicates velocity, there is u̇x = − vp (u′x1 − u′x2 ).
seismic response of layered foundations. When solving nonlinear soil At the truncated boundary, the soil node is subjected to the damping
problems, an equivalent linearized viscoelastic model nested in the force σdamp provided by the outer infinite element and the normal stress
Mohr Coulomb model is used to establish a viscoelastic-plastic model. By σx transmitted by the internal soil. From the equilibrium of forces, it can
solving one-dimensional free field response analysis, the equivalent be obtained that σdamp = σx at the truncated boundary, as follows:
shear modulus and equivalent damping ratio of soil at different depths ( ) ( )
are obtained. Combined with Abaqus secondary development technol­ λ + 2G − dp vp u′x1 + λ + 2G + dp vp u′x2 = 0 (5)
ogy, corresponding material properties are assigned to soil elements at To ensure that the reflected wave is completely absorbed at the
different depths in layered soil. Taking a certain underground station
boundary (ux2 = 0,u′x2 = 0), make:
structure as the research object, the dynamic response analysis of the soil
structure model was conducted using the layered seismic input method λ + 2G
dp = = ρv p (6)
proposed in this study. The research results revealed the influence of vp
burial depth on the overall seismic performance of the underground
In the shear wave incident condition, the same is obtained, ds = ρvs.
structure, providing a reference basis for the seismic design of under­
The infinite element boundary is able to absorb well the reflected wave
ground structure engineering.
effects on the near-domain foundation at the truncated boundary due to
the consideration of the damping constants ds and dp.
2. Infinite element-substructure based ground motion input
method
2.2. Infinite element-substructure ground motion input method
In the seismic analysis of underground structures, finite near-domain
foundations need to be intercepted from infinite domain foundations for In the seismic response analysis of structures, especially under­
simulation, for which truncated foundation boundaries need to be ground structures, the fluctuation input method is a common method for
treated to eliminate the effects of scattered waves on near-domain ground motion input, which meets the requirements in terms of accu­
foundations [1,30,32,33]. The common method is to set a viscoelastic racy but is cumbersome to implement. For this reason, some scholars
artificial boundary at the truncated foundation boundary to simulate the propose the substructure input method, which has a clear concept and
absorption and elastic restoring force of the scattered waves on the simplifies the operation process of the original fluctuation input method.
distant foundation, but for the complex layered distribution foundation, In this study, based on the infinite element boundary, the infinite
the spring damping unit at the truncated foundation boundary is rela­ element-substructure ground motion input method is proposed in
tively complicated to impose, and the spring stiffness and damping co­ combination with the substructure input method in the literature [11].
efficient at the medium partition are difficult to determine accurately. The specific operation process is as follows:
Therefore, this study propose a new seismic input method for layered
ground base. Compared with the viscoelastic artificial boundary, the (1) Determine the finite domain range of the foundation model ac­
new method avoids the application of spring and damping units to the cording to the spatial dimensions of the subsurface structure,
boundary nodes of layered soils and simplifies the operation process. establish the overall soil-structure finite element analysis model,

2
Y. Zhai et al. Structures 58 (2023) 105488

dissect the mesh of the model and establish the infinite element
boundary.
(2) According to the finite element foundation model in step (1), a 1D
free field analysis model containing the same soil depth and
properties is established (Fig. 1), where the frequency domain
DeepSoil [36] or Shake91 [37] can be used for soil seismic
response analysis to obtain the acceleration, velocity, and
displacement responses at different soil depth locations.
(3) The finite element analysis model in step (1) is processed, keeping
the outer infinite element unit and one layer of soil unit (Fig. 2),
and the seismic response (acceleration, velocity or displacement)
of the soil layer obtained in step (2) is input into the processed
soil unit model. The equivalent nodal force loads of the outer soil
nodes are calculated.

Input the external equivalent nodal force loads obtained in step 3 to

the overall soil-structure finite element model established in step (1).
For static + dynamic load conditions, it is also necessary to add the
reaction load at the truncated boundary under static load, that is, the
seismic response of the structure under the layered foundation can be
analyzed. Fig. 2. Infinite element-substructure input model.
The infinite element boundary belongs to the viscous artificial
boundary, and the equivalent node load obtained by step (3) consists of
F C (t) = Cu̇(x, y, z, t) (7)
two parts: the damping force generated by the infinite element dynamic
boundary and the free field force [38]. Free Field Force
Damping force
F τ (t) = τ(x, y, z, t) (8)
Equivalent load force

F(t) = F C (t) + Fτ (t) = Cu̇(x, y, z, t) + τ(x, y, z, t) (9)

Where, u is the sum of the input wave displacement and the reflected
wave displacement; u̇ is the velocity, which is obtained by the derivative
of the free field displacement u with respect to time t; τ is the free field
stress, which can be calculated from the free field displacement u ac­
cording to the medium elastic instanton relationship.

3. Seismic response analysis method considering soil

nonlinearity

The operational procedure of the infinite element-substructure

ground shaking input method for elastic half-space is given above. In
fact, under stronger seismic loads, soils tend to exhibit certain modulus
attenuation as well as damping increase effects, and currently, consid­
ering the equivalent linearization of soils is the main method to solve the
seismic response analysis of nonlinear soils [39,40]. In applying the
infinite element-substructure method described above for nonlinear
seismic response analysis of soils, to better simulate the Nonlinear
behavior of the soil under seismic loading, this study adopts the nesting
of the equivalent linear model into the MC model to establish a
viscoelastic-plastic model to solve the dynamic response problem of the
underground structure. This reflects both the modulus decay and the
damping increase effect of the soil under dynamic action, on the other
hand, it also reflects to a certain extent the plastic damage of the soil
under earthquake load.

3.1. Soil equivalent linear viscoelastic model

The equivalent linear viscoelastic model is developed from the

viscoelastic theory and uses the viscoelastic Kelvin model to reflect the
hysteretic properties of the soil under cyclic loading [41]. Its stress–­
strain relationship is
τ = Gγ + ηG γ̇ (10)

Where, G is the shear modulus; ηG is the shear viscosity coefficient; τ and

Fig. 1. 1D free field model. γ, respectively, are expressed as the shear stress and shear strain; γ̇ is the

3
Y. Zhai et al. Structures 58 (2023) 105488

creep rate. process considering soil nonlinearity is given in the paper, as shown in
According to the concept of damping ratio, the shear viscosity co­ Fig. 3. The specific process is as follows:
efficient ηG can be expressed as:
(1) An equivalent linearization procedure (DeepSoil) is used to solve
ηG = 2Gλ/ω (11)
the 1D soil model in the frequency domain to obtain the
maximum dynamic shear strains in soil layers at different depths.
Where, λ is the damping ratio and ω is the circular frequency.
The normalized dynamic shear strain is solved by substituting
The function of shear modulus G and damping ratio of soil is a
into Eq. (14). Then the equivalent shear modulus and equivalent
function of shear strain, which can be used in the following form:
damping ratio of the soil for different soil layers and depths are
( ′ )n
k2 σ obtained by combining the dynamic shear modulus, dynamic
G= pa 3 (12)
1 + k1 γd pa shear strain and damping ratio relationship curves of the soil
material.
k1 γd (2) The soil-structure analysis model is established based on Abaqus.
λ = λmax (13)
1 + k1 γ d Using the equivalent shear modulus of the soil at different depths
obtained above, the material parameters of the soil layer analysis
Where, σ′3 denotes the circumferential pressure; k1, k2 and n are material model are set. Determine the Rayleigh damping coefficients of
parameters determined experimentally; γ d is the normalized shear soil materials at different depth locations based on the model
strain. first-order natural frequency and the predominant frequency of
( ′ )n− 1 the input ground motion fourier amplitude spectrum [40]. Set
σ
γ d = 0.65γdmax 3 (14) corresponding MC model parameters based on soil properties in
pa Abaqus software
(3) The ground motion response at different soil depth locations is
Where, γdmax indicates the maximum dynamic shear strain.
obtained by solving the 1D soil model in frequency domain
For three-dimensional working conditions, the modulus matrix can
analysis. The infinite element-substructure ground motion input
be expressed in terms of the bulk modulus K and shear modulus G.
method is used to obtain the seismic equivalent inertia force of
Similar to equation (11), the bulk viscosity coefficient of the material ηK
the soil nodes on the outside of the model. Based on Abaqus
can be expressed as:
software, after completing the static analysis of the model, the
K = [2(1 + μ)]/[3(1 − 2μ)] (15) equivalent effects of the nodes obtained above can be used as
input for seismic loads to complete the soil structure seismic
ηK = 2Kλ/ω (16) response analysis considering soil nonlinearity.

Where, μ denotes the Poisson’s ratio, Combining Eqs. (11) and (16) 3.4. Seismic response verification analysis of layered half-space soils
yields that ηK/ηG = K/G, that is, the model damping matrix is propor­
tional to the stiffness matrix, and its scale factor is 2λ/ω. To verify the validity of the infinite element-substructure ground
vibration input method in this study, a four-layer foundation soil model
3.2. Mohr-Coulomb model is established based on Abaqus. The 1D soil analysis software DeepSoil
solution is used as the reference solution, and the pulse wave and the
The MC model is an ideal elastic–plastic model that is widely used in measured wave are used as the ground motion input respectively for the
geotechnical engineering [42]. The model uses a shear stress yielding verification analysis.
criterion with the criterion equation:
3.5. Seismic response analysis of elastic half-space soil
f = τn − C − σn tanϕ = 0 (17)
Among them The distribution of model soil layers and the arrangement of obser­
vation points are shown in Fig. 4. The size of the 2D model is 100m80m,
1( )
the layered soil are set as miscellaneous fill, clay, sand and weathered
τn = Rcosϕσ n = σ x + σy − Rsinϕ (18a,b)
2 rock from top to bottom in order. Soil material parameters: density ρ1 =
Where, φ denotes the angle of internal friction, c denotes the soil ρ2 = 1900 kg/m3, ρ3 = ρ4 = 2000 kg/m3; Shear wave speed cs1 = 100 m/
cohesion, τn is the normal positive stress, σn is the normal positive stress, s, cs2 = 150 m/s, cs3 = 200 m/s, cs4 = 300 m/s; Poisson’s ratio v1 = v2 =
and R denotes the radius of the Mohr stress circle. 0.20, v3 = v4 = 0.30. The Abaqus finite element analysis model is
Using the stress invariants I1, J1 and the generalized shear stress established, CINEP4 infinite element unit is used on the outside, CEP4R
azimuth θ, the above equation can be expressed as: units are used for the internal soil material, the model is divided into
√̅̅̅̅̅ 1386 various units and 1462 nodes. The incident wave R(t) is a single
1 √̅̅̅̅̅ π J2 π pulse wave (the displacement peak is 0.1 cm, and the pulse period T0 =
f (I1 , J2 , θ) = I1 sinϕ + J2 sin(θ + ) + √̅̅̅ cos(θ + )sinϕ − Ccosϕ = 0
3 3 3 3 0.2 s), expressed in the form of equation (20), Fig. 5 shows the pulse
(19) wave acceleration, velocity and displacement time history curves.
[ ( ) ( ) ( ) ( )
In general, the size of the shear expansion angle of the soil is related t t 1 t 1 t 3
R(t) = 0.016 G − 4G − + 6G − − 4G −
to the surrounding pressure and pore ratio. For the soil with friction T0 T0 4 T0 2 T0 4
angle less than 30◦ , the shear expansion angle takes the value of 0◦ . (
t
)]
+ GG − 1
T0
3.3. Seismic response analysis process based on soil viscous-elastic–plastic (20)
model
where, G(t) = (Tx0 )3 H(x), H(x) is the Heaviside step function.
On the basis of the infinite element-substructure ground motion The velocity and displacement responses of the two observation
input method established above, combined with the soil viscous-elas­ points A and B are given in Figs. 6 and 7, respectively, and it can be seen
tic–plastic model model, a soil-structure seismic response analysis that the displacement and velocity time curves of the observation points

4
Y. Zhai et al. Structures 58 (2023) 105488

Fig. 3. Process of nonlinear seismic response analysis of layered soil.

at different soil depth positions are in better agreement with the refer­
ence solution. Besides, unlike the homogeneous foundation, the waves
are reflected and transmitted at the interface of different soil layers. The
peak displacement response is 0.34 cm at the free surface observation
point B (the peak displacement response under homogeneous condition
is 0.2 cm), which indicates that the layered soil has a significant
amplification effect on the seismic wave input.

3.6. Nonlinear seismic response analysis of layered soil

Under seismic loads, the mechanical properties of the foundation soil

usually undergo nonlinear changes. The ground motion records of
Northridge-01 was selected, and the acceleration time curve and fourier
amplitude spectrum of the ground shaking records are given in Fig. 8
and Fig. 9, respectively. The seismic waves were held for 40 s with a time
interval of 0.01 s. The seismic wave amplitude was halved as the input
wave, and the model was analyzed for seismic response based on the
Fig. 4. Soil layer distribution.
above established nonlinear seismic response analysis procedure for
layered soils.
Using the computational analysis model in Section 4.1. The rela­
tionship curves of dynamic shear modulus, dynamic shear strain and

Fig. 5. Time history curve of pulse wave.

5
Y. Zhai et al. Structures 58 (2023) 105488

Fig. 6. Velocity response of observation point. Fig. 9. Fourier amplitude spectrum of seismic wave.

Fig. 7. Displacement response of observation point. Fig. 10. Dynamic shear modulus ratio-shear strain curve.

Fig. 8. Time history curve of seismic wave.

Fig. 11. Damping ratio-shear strain curve.

damping ratio for each soil material layer are shown in Fig. 10 and
domain to obtain the maximum dynamic shear strain of soil layers at
Fig. 11, respectively.
different depths. Then the equivalent shear modulus and equivalent
According to the established nonlinear seismic response analysis
damping ratio of the soil at different soil layers and different depth lo­
process for layered soils. Firstly, the equivalent linearization procedure
cations are obtained by combining the dynamic shear modulus, dynamic
(DeepSoil) is used to analyze the 1D soil layer model in frequency
shear strain and damping ratio relationship curves of the soil material.

6
Y. Zhai et al. Structures 58 (2023) 105488

The Rayleigh damping coefficients of soil materials at different depth

locations were determined based on the model first-order natural fre­
quency and the predominant frequency of the input ground motion
Fourier amplitude spectra. Accordingly, the material parameters of
Abaqus soil layer analysis model are set. The schematic diagram of the
analysis of the nonlinear ground motion response considering the soil
layer is given in Fig. 12 (Rayleigh damping formula reference [40]). It
can be seen that the dynamic response of the soil is consistent at the
same elevation under the seismic load.
The velocity and displacement responses of observation points A and
B are given in Figs. 13–16, respectively. Under the action of the
measured seismic waves, the velocity and displacement time curves of
the observed points are in good agreement with the reference solutions
after considering the nonlinearity of the soil body. It indicates that the
nonlinear seismic response analysis method of soil based on infinite
element-substructure input method in this study has good accuracy.
After considering the nonlinear of the soil, the variation curve of the
model PGA peak with soil depth is shown in Fig. 17. As the soil depth
Fig. 13. Velocity response of point A.
increases, the seismic wave PGA peak shows alternating changes of
decreasing and increasing, where the maximum PGA peak also appears
structure as the research object, the seismic response of the soil-structure
at the surface location of the foundation. The curves of the peak
integral system is analyzed by using the amplitude modulated
displacement of the model with the depth of the soil layer are given in
Northridge-01 and Hollister-01 ground motion records as horizontal
Fig. 18, and the rate of the peak displacement with depth is not the same
design seismic waves, respectively. The following five working condi­
in different depth ranges of the soil layer. Among them, the peak
tions (h = 4 m, 8 m, 12 m, 16 m, 20 m) are set to study the influence law
displacement changes faster in the depth range of soil layer 1, while it
of burial depth on the dynamic response of underground structure.
changes slower in the depth range of soil layer 4. Overall, when the
nonlinear of the soil is considered, the maximum errors of acceleration
and displacement for each soil depth are 3.2 % and 2.8 %, respectively,
4.1. Project overview
compared with the reference solution, and the calculation accuracy
meets the engineering requirements.
A single-story double-span subway station is used as the research
object, and the calculated width of the model is taken as 4 times of the
4. Study of factors influencing the dynamic response of
station, the geometric model of soil-structure is established as shown in
underground structure
Fig. 19 and Fig. 20. The burial depth of the structure is h = 4.0 m, the
span is D = 15.5 m, and the height is H = 7.0 m. According to the ground
To explore the effect of burial depth on the seismic performance of
survey data, the engineering site is divided into five soil layers in the
underground structures. In this section, taking an underground station
depth range of 0–50 m. The contact between soil and structure adopts

Fig. 12. Schematic diagram of ground motion response analysis of layered soils.

7
Y. Zhai et al. Structures 58 (2023) 105488

Fig. 14. Displacement response of point A.

Fig. 17. PGA variation curve with soil depth.

Fig. 15. Velocity response of point B.

Fig. 18. Displacement peak value changing with soil depth.

dynamic interaction [43,44], the units do not intrude into each other,
the tangential direction adopts friction contact, the normal direction
adopts hard contact. Referring to similar engineering [43], in this study,
the friction coefficient is taken as μ = 0.4.

4.2. Calculation parameters

4.2.1. Physical parameters

Model material parameters: The model foundation soil layers from
top to bottom are miscellaneous fill, sandy soil 1, sandy soil 2, clay and
weathered rock, the material parameters of each soil layer are shown in
Table 1. The relationship curves of dynamic shear modulus, dynamic
shear strain and damping ratio for each soil layer are shown in Fig. 4 and
Fig. 5, respectively. The internal structure of the station adopts C35
concrete, elastic modulus is 3.15 × 104 MPa, density is 2.5 × 103 kg/m3,
middle column adopts C40 concrete. Considering the longitudinal
equidistant (5.0 m) distribution of the middle columns in the structure,
Fig. 16. Displacement response of point B.
the equivalent post-elastic modulus is taken as 6.5 × 103 MPa and the
density is 5 × 102 kg/m3.

4.2.2. Ground motion parameters

For the ground motion input of the layered soils, this study takes the
free surface acceleration time as the reference and inverts it to the
bedrock surface to obtain the bedrock input wave. Northridge-01 (Port
Hueneme station) and Hollister-01 (Hollister City Hall station) of the

8
Y. Zhai et al. Structures 58 (2023) 105488

Fig. 19. Geometric model of soil-structure (h = 4 m).

Fig. 20. Local geometric model of soil structure (h = 8 ~ 20 m).

Table 1
Property parameters of each soil layer.
No Soil layer Severe γ(kN/m3) Soil depth Shear wave speed Poisson’s ratio Cohesion Friction angle φ(◦ )
h (m) cs (m/s) λ c (KPa)

Soil layer 1 Mixed fill 18.1 0–4 138 0.35 10.5 18

Soil layer 2 Sandy soil 1 19.4 4–8 162 0.32 3.0 25
Soil layer 3 Sandy soil 2 19.8 8–16 185 0.30 3.0 28
Soil layer 4 Clay 20.1 16–26 248 0.33 18.3 22
Soil layer 5 Weathered rock 23.6 26–50 450 0.25 65 25

same site were selected as the free surface horizontal design seismic strain and dynamic shear modulus and damping ratio of soil is
waves, and their acceleration peaks were amplitude modulated by Amax substituted to obtain its equivalent shear modulus and equivalent
= 0.1 g. The acceleration curve and response spectrum after amplitude damping ratio, which the soil layer parameters in Abaqus time domain
modulation are shown in Fig. 21 and Fig. 22, respectively. Fig. 23 shows analysis are defined. The soil Rayleigh damping coefficient is deter­
the Fourier amplitude spectrum of seismic waves, it can be seen that the mined based on the first-order natural frequency of the model and the
predominant frequency of Fourier amplitude spectra of the two ground predominant frequency of the input ground motion Fourier amplitude
motion records are 1.27 Hz and 1.81 Hz, respectively. spectrum. Accordingly, the Abaqus finite element analysis model is
established as shown in Fig. 25. The internal structure and soil part are
4.3. Soil-structure finite element analysis model simulated by solid unit CEP4R unit, and the external side is simulated by
CINEP4 infinite element unit, the unit feature length is 0.5 m ~ 1 m, and
Taking Northridge-01 seismic wave incidence as an example, the the model is divided into 6720 various units and 6971 nodes. Consid­
variation curves of peak PGA and maximum dynamic shear strain with ering static and dynamic loads, the bottom of the model is fully con­
soil depth for the soil model are given in Fig. 24. In this study, the strained during static analysis, while the side is constrained in a normal
maximum dynamic shear strain of soil at different depths is calculated direction. In dynamic analysis, infinite element dynamic boundary
using DeepSoil software, the relationship curve between dynamic shear conditions are used.

9
Y. Zhai et al. Structures 58 (2023) 105488

Fig. 21. Time history curve of seismic wave.

Fig. 24. Curve of PGA and dynamic shear strain.

Fig. 22. Fourier amplitude spectrum of seismic wave.

Fig. 25. Finite element analysis model of soil-structure.

Fig. 23. Fourier amplitude spectrum of seismic wave.

4.4. Dynamic response analysis of underground structures

Taking the Northridge-01 seismic wave incidence as an example, the

relative displacement time curves of the top and bottom of the structure
at different burial depths are given in Fig. 26. It can be seen that the Fig. 26. The curve of relative displacement.
change trend of the displacement response curve of the structure is more
consistent under different working conditions. Fig. 27 shows the peak

10
Y. Zhai et al. Structures 58 (2023) 105488

Fig. 27. Curve of displacement peak value. Fig. 29. Bending moment time history curve (A).

relative displacement curves of the top and bottom of the structure with shear force and bending moment at B show a variation pattern of
the burial depth of the structure. Under the action of horizontal increasing and then decreasing with the increase of the burial depth of
Northridge-01 and Hollister-01 seismic waves, with the increase of the the structure under the static load condition. This is due to the fact that
burial depth of the structure, the difference of the top–bottom the change of burial depth changes the magnitude of the peripheral soil
displacement shows the change law of increasing first and then load on the structure, on the one hand, the increase in the thickness of
decreasing. the overlying soil makes the underground structure subject to increased
The time curves of shear force and bending moment at the bottom A soil shear and dynamic earth pressure, the other hand, as the burial
of the middle column in the structure under the action of Northridge-01 depth increases, it limits the deformation response of the subsurface
seismic wave are given in Fig. 28 and Fig. 29, respectively. It can be seen structure. Under the joint influence of both, the shear force and bending
that at the beginning moment of seismic action, the structural middle moment at B of the sidewall reach the maximum when the burial depth
column is almost free from shear and bending moment, and their shear of the structure is h = 16 m.
and bending moment response increases gradually with the increase of The variation curves of the proportion of peak dynamic shear force
seismic load. The curves of the peak shear force and bending moment of and bending moment (alone dynamic load) caused by seismic action in
the middle column with the burial depth of the structure are given in the total shear force and peak bending moment (static + dynamic load)
Fig. 30 and Fig. 31, respectively. The variation of shear force and with burial depth are given in Fig. 36 and Fig. 37, respectively. In
bending moment at the bottom A of the middle column of the structure general, the percentage of dynamic shear force at the bottom B of the
under the seismic load is approximately the same. With the increase of sidewall decreases gradually with increasing depth of burial in the range
the burial depth of the structure, both show the variation law of of 4–20 m. The curve of dynamic bending moment proportion appears
increasing first and then decreasing. Overall, in the seismic response abruptly at the burial depth h = 16 m, which may be caused by the fact
analysis of the single-story double-span station structure in this study, that the bottom of the structure just crosses the interface between soft
the seismic resistance of the middle column is more unfavorable at the and hard soil layers at this time. At the same depth of burial, the dy­
burial depth of the structure of 8 ~ 12 m. namic bending moment due to seismic loading is larger compared with
The dynamic shear force and dynamic bending moment time curves the shear response. Under the action of Northridge-01 and Hollister-01
at the bottom B of the sidewall caused by the seismic load are given in seismic waves, the peak percentage of dynamic shear force and dy­
Fig. 32 and Fig. 33 for different working conditions. The peak shear namic bending moment reaches the maximum when the structure is
force and bending moment due to static and dynamic loads are given in buried at a depth of h = 4 m, and their values are 16.9 %, 14.1 % and
Fig. 34 and Fig. 35. In the burial depth range from 4 to 20 m, the peak 33.6 %, 28.8 %, respectively.

5. Conclusions

In the seismic response analysis of underground structures, this study

presents a new method for ground motion input of layered foundation.
The new method avoids applying springs and damping elements to the
boundary nodes of layered soil, simplifies the seismic input process of
layered foundations, and has high accuracy, making it easy to promote
and use in engineering. The soil viscoelastic-plastic model is established
based on the soil equivalent linear and MC models, and the seismic
response analysis process of underground structures considering soil
nonlinearity is given. Taking an underground station structure as the
research object, the nonlinear seismic response analysis of the overall
soil-structure model is carried out. Concluded as follow:

(1) The numerical example results show that the error of the new
method is within 5 % compared with the 1D soil analysis software
DeepSoil solution, which indicates that this method has good
accuracy and can well simulate the ground motion response law
Fig. 28. Shear time history curve (A).

11
Y. Zhai et al. Structures 58 (2023) 105488

Fig. 30. The shear peak value.

Fig. 31. The bending moment peak value. Fig. 33. Bending moment time history curve (B).

Fig. 32. Shear time history curve (B).

Fig. 34. Shear peak value (B).

12
Y. Zhai et al. Structures 58 (2023) 105488

(3) In the burial depth range of 4 ~ 20 m, the dynamic shear force

ratio at the bottom of the sidewall gradually decreases with the
increase of the burial depth of the structure, while the dynamic
bending moment ratio shows alternate changes of decreasing and
increasing. Under the same burial depth condition, the dynamic
bending moment caused by the seismic load is larger compared
with the shear response.

6. Funding source declaration

No.

7. Ethics approval

Approved.

Fig. 35. Bending moment peak value (B).

Declaration of Competing Interest

The authors declare that they have no known competing financial

interests or personal relationships that could have appeared to influence
the work reported in this paper.

References

[1] Huang Z, Pitilakis K, Tsinidis G, et al. Seismic vulnerability of circular tunnels in

soft soil deposits: The case of Shanghai metropolitan system. Tunn Undergr Space
Technol 2020;98:103341.
[2] Sun Q, Dias D, Luis R. Impact of an underlying soft soil layer on tunnel lining in
seismic conditions. Tunnelling Underground Space Technol 2019;90–:293–308.
[3] Yang L, Luo J, Chen W, et al. Amplification of seismic response in poroviscoelastic
soil layer. Adv Civil Eng 2020;2020(6):1–12.
[4] Messioud B. Harmonic seismic waves response of 3D rigid surface foundation on
layer soil. Earthquake Struct 2019;16(1).
[5] Jingbo L, Hui T, Dongyang W, et al. A new seismic motion input method in soil-
structure dynamic interaction analysis. China Earthquake Eng J 2019;41(01):1–8.
[6] Jingbo L, Xin B, Hui T, et al. Seismic wave input method for soil-structure dynamic
interaction analysis based on internal substructure. Chin Civil Eng J 2020;53(08):
Fig. 36. Proportion of dynamic shear force (B). 87–96.
[7] Zhonghui B, Liaojun Z, Yafei Z, et al. Seismic input method of layered slopes.
J Water Resour Water Eng 2021;32(04):214–20.
[8] Ba Z, Liang J, Lee VW, et al. IBEM for impedance functions of an embedded strip
foundation in a multi-layered transversely isotropic half-space. J Earthq Eng 2017;
50(8–10):1415–46.
[9] Liang J, Han B, Todorovska MI, et al. 2D dynamic structure-soil-structure
interaction for twin buildings in layered half-space I: Incident SH-waves. Soil Dyn
Earthq Eng 2018;102:172–94.
[10] Zhou Z. An analytical solution for 2D dynamic structure-soil-structure interaction
for twin flexible tunnels embedded in a homogeneous half-space. Appl Sci 2021;11.
[11] Jin L, Liang J. 2D dynamic structure-canyon-structure interaction for the buildings
along the Urban River-canyon I: Incident SH-waves in homogenous half-space.
J Earthquake Eng 2020;2:1–19.
[12] Jin L, Zhu J, Zhou W, et al. 2D dynamic tunnel-soil-aboveground building
interaction I: Analytical solution for incident plane SH-waves based on rigid tunnel
and foundation model. Tunnelling Underground Space Technol 2022.
[13] Bazyar MH, Song C. Time-harmonic response of non-homogeneous elastic
unbounded domains using the scaled boundary finite-element method. Earthq Eng
Struct Dyn 2010;35(3):357–83.
[14] Bazyar MH, Song C. Transient analysis of wave propagation in non-homogeneous
elastic unbounded domains by using the scaled boundary finite-element method.
Earthq Eng Struct Dyn 2010;35(14):1787–806.
[15] Birk C, Behnke R. A modified scaled boundary finite element method for
threedimensional dynamic soil-structure interaction in layered soil. Int J Numer
Fig. 37. Proportion of dynamic bending moment (B). Meth Eng 2012;89(3):371–402.
[16] Gao Y, Yanjie X, et al. A simplified high order doubly asymptotic open boundary
for modeling elastic wave propagation in layered foundation. Sci Sin Technol 2016;
of layered soil (linear elastic/nonlinear) with the change of soil 46(5):527–34.
depth. [17] Aslmand M, Kani IM, Birk C, et al. Dynamic soil–structure interaction in a 3D
layered medium treated by coupling a semi-analytical axisymmetric far field
(2) Under the action of seismic load, with the increase of the burial formulation and a 3D finite element model. Soil Dyn Earthq Eng 2018;115:531–54.
depth of the structure, the top–bottom displacement difference [18] Song C. The Scaled Boundary Finite Element Method Introduction to Theory and
shows the change law of increasing first and then decreasing. The Implementation. Chichester: John Wiley & Sons; 2018.
[19] Kazakov KS. Elastodynamic infinite elements based on modified Bessel shape
variation law of the shear force and bending moment response at functions, applicable in the finite element method. Struct Eng Mech 2012;42(3):
the bottom of the columns and side walls in the structure is not 353–62.
the same. [20] Kazakov KS, Stoynova IY. Dynamic thermal wave response and propagation
through building structures using infinite elements in time and frequency domain.
International Conference Functional Materials Author(s) 2017.

13
Y. Zhai et al. Structures 58 (2023) 105488

[21] Yang YB, Zhou Z, Zhang X, et al. Soil seismic analysis for 2D oblique incident waves [33] Gao Z, Zhao Mi, Yangzhou Wu, Wang M, Xiuli Du. Parameter design and
using exact free-field responses by frequency-based finite/infinite element method. performance evaluation of tuned mass damper (TMD) for seismic control of
Front Struct Civil Eng 2022. structure considering soil-structure interaction (SSI). Structures 2023;52:1116–29.
[22] Yang YB, Li J, Nie QQ, et al. A hybrid artificial boundary combining perfectly [34] Ungless RF. An infinite finite element. UK: University of British Columbia; 1973.
matched layer and infinite elements for analysing semi-infinite problems. Eng Anal [35] ABAQUS Inc. ABAQUS Theory Manual. Providence: Dassault Systemes Simulia
Bound Elem 2023;155:528–40. Corp, 2010.
[23] Bianchi SD. Combining finite and infinite elements: Why do we use infinite [36] Puri N, Jain A, Mohanty P, et al. Earthquake response analysis of sites in state of
idealizations in engineering? Synthese 2019. haryana using DEEPSOIL Software. Procedia Comput Sci 2018;125.
[24] Sameti AR, Ghannad MA. Equivalent linear model for existing soil-structure [37] Barder JP, Ichii K, Lin CH. EERA: A computer program for equival-ent-linear
systems. Int J Struct Stab Dyn 2016;16(02):1450099. earthquake site response analysis of layered soil deposits. Los Angeles: University
[25] Ahmadi E, Kiani M, Paytam F, et al. Equivalent linearization parameters of soil- of Southern California; 2000.
MDOF structure systems subjected to pulse-like earthquakes. Soils Found -Tokyo [38] Song Z, Wang F, Liu Y, et al. Infinite Element Static-Dynamic Unified Artificial
2018;58(6):1371–82. Boundary. Shock and Vibration, 2018, 2018(PT.7):1-14.
[26] Yang S, Han X, Lei Q, et al. Study on the seismic effect of the interbedded soil layer [39] Du Xiuli X, Chengshun ZX, et al. Time-history analysis method for soil-
in the Yinchuan Alluvial Plain. Adv Civil Eng 2021;22:1–12. underground structure system based on equivalent linear method. Chin J Geotech
[27] Xu C, Zhang Z, Li Y, et al. Seismic response and failure mechanism of underground Eng 2018;40(12):2155–63.
frame structures based on dynamic centrifuge tests. Earthq Eng Struct Dyn 2021;50 [40] Xu Zigang D, Chengshun XX, et al. Comparison of determination methods of site
(7):2031–48. Rayleigh damping coefficients in seismic responses analysis of underground
[28] Tsinidis G, Pitilakis K, Anagnostopoulos C. Circular tunnels in sand: dynamic structures. Rock Soil Mech 2019;40(12):4838–47.
response and efficiency of seismic analysis methods at extreme lining flexibilities. [41] Chandra K, Kiran M, Nath P, et al. A study of local soil effect on the earthquake
Bull Earthquake Eng 2016, 14(10): 2903-2929.54. ground Motion in Bhaktapur City, Nepal using equivalent linear and non-linear
[29] Tsinidis G, Pitilakis K, Madabhushi G. On the dynamic response of square tunnels analysis. Iran J Sci Technol Trans Civil Eng 2022;46(6).
in sand. Eng Struct 2016;125:419–37. [42] Ali S, Côme C, Kamalibandpey, et al. Evaluation of the effect of geomechanical
[30] Argyroudis S, Tsinidis G, Gatti F, et al. Effects of SSI and lining corrosion on the parameters and in situ stress on tunnel response using equivalent mohr-coulomb
seismic vulnerability of shallow circular tunnels. Soil Dyn Earthq Eng 2017;98: and generalized Hoek-Brown criteria. Geosciences 2022;12(7).
244–56. [43] Jiang J, Nggar H, Xu C, et al. Effect of ground motion characteristics on seismic
[31] Yu HT, Yang YS, Yuan Y, Duan KP, Gu Q. A comparison between vibration and fragility of subway station. Soil Dyn Earthq Eng 2021;143:106618.
wave methods in seismic analysis of underground structures. China Earthquake [44] Zhab C, Hao X, Pg C, et al. Seismic demand and capacity models, and fragility
Eng J 2019;41(04):845–52. estimates for underground structures considering spatially varying soil properties.
[32] Ma C, Dechun Lu, Zhao Y, Wang Z, Xiuli D. Performance of an underground Tunnelling Underground Space Technol 2022, 119:104231.
structure seismic mitigation system improved by frictional deformation absorbing
braces. Structures 2022;37:1–16.

14