Tunnelling and Underground Space Technology 77 (2018) 216–226

Contents lists available at ScienceDirect

Tunnelling and Underground Space Technology

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

Seismic response of shield tunnel subjected to spatially varying earthquake T

ground motions
⁎
Yu Miaoa, Erlei Yaoa, Bin Ruana, , Haiyang Zhuangb
a
School of Civil Engineering and Mechanics, Huazhong University of Science and Technology, Wuhan 430074, PR China
b
Institute of Geotechnical Engineering, Nanjing Tech University, Nanjing 210009, PR China

A R T I C LE I N FO A B S T R A C T

Keywords: The subway shield tunnel of Sanyang Road in Wuhan, China, is a Yangtze River-crossing tunnel project with a
Generalized response displacement method mega diameter of 15.7 m, which crosses several soil layers with sharply different properties. To investigate the
Multi-support excitation seismic characteristics of this special subway tunnel under spatially varying earthquake ground motions
Spatially varying ground motions (SVEGMs) with high efficiency, the concept of response displacement method was utilized and realized with a
Seismic behaviors
refined free-field model and a simplified soil-tunnel model. Two sets of borehole accelerograms, denoting near-
Identical support excitation
field and far-field ground motions, respectively, were selected and taken as the reference to generate fully non-
stationary SVEGMs. A lot of responses along the longitudinal direction under multi-support excitation (MSE),
including intersegment opening width, inner forces, and acceleration, were observed and compared with those
under identical support excitation (ISE). The comparative results showed that the responses of subway shield
tunnel under MSE are generally higher than those under ISE. The longitudinal distributions of intersegment
opening width and bending moment response of lining are dominantly affected by the site conditions. Excitation
methods, spectral characteristic and PGA of input earthquake can also have effect on the longitudinal dis-
tribution of tunnel response. The results can contribute to revealing the difference between seismic response
characteristics of shield tunnel under MSE and those under ISE and predicting the underlying location of seepage
or severe damage of shield tunnel. The analysis procedure established in this paper possesses highly potential for
application to the aseismic design of practical shield tunnel project.

1. Introduction seismic damage based on a dynamical finite element analysis method.

Due to the severe damage of tunnels occurred in Kobe (1995), Chi-
Chi (1999), Koceali (1999) and Wenchuan (2008) earthquakes, the
importance of seismic analysis of tunnel has drawn more and more
attention. A series of investigations have been performed by a number
of researchers for evaluating the seismic behaviors of tunnels during the
past decades. Hashash et al. (1998) conducted three-dimensional (3D)
analyses of the dynamic response of the immersed tubes in the San
Francisco Bay Area. It has been concluded that the arrival time delay of
the ground motion has a significant influence on the axial deformation
of tunnel. Lee and Trifunace (1979) studied the response of under-
ground circular tunnel subjected to incident SH-waves and obtained an
analytical solution. Park et al. (2009) developed a longitude displace-
ment profile-based method for simulating the tunnel response under
SVEGM and performed a series of pseudo-static 3D finite element
analyses. It has been found that the SVEGM causes longitudinal bending
of the tunnel and can induce substantial axial stress on the tunnel. Chen
et al. (2012) investigated the influence of the depth of a tunnel on its
Bao et al. (2017) clarified sufficiently the seismic behavior and floata-
tion mechanism of the large metro tunnel structure in liquefied soil
deposits by using an effective stress-based soil-water fully coupling fi-
nite element-finite difference method. Tsinidis (2017) carried out a
numerical parametric study on diverse soil-rectangular tunnel systems
to shed light on critical response characteristics of rectangular tunnels
subjected to transversal ground shaking. Shahrour et al. (2010) con-
ducted an elastoplastic finite element analysis of the seismic response of
tunnels constructed in soft soils and the results showed that the plastic
deformations induce an important reduction in the seismic-induced
bending moment in the tunnel, while the soil dilatancy moderately
affects the bending moment in the liner. Amorosi and Boldini (2009)
compared the results of a set of analyses aimed at studying the seismic
transversal response of a shallow tunnel and proposed that adopting
more advanced constitutive models for both soil and tunnel lining,
capable of reproducing more realistically their behavior under dynamic
conditions, is needed. Chen et al. (2017) conducted a time history
analysis to explore the influence of SVEGM on seismic responses of the

⁎
Corresponding author.
E-mail address: ruanbin666@hust.edu.cn (B. Ruan).

https://doi.org/10.1016/j.tust.2018.04.006
Received 28 November 2017; Received in revised form 1 March 2018; Accepted 6 April 2018
0886-7798/ © 2018 Elsevier Ltd. All rights reserved.
Y. Miao et al.
long immersed tunnel. Results showed that SVEGM enlarges seismic
responses of the immersed tunnel significantly and considering wave
passage effect and incoherence effect simultaneously makes the seismic
responses increase more sharply than considering the latter only. Yan
et al. (2016), Yuan et al. (2016) and Yu et al. (2016, 2017) conducted a
series of shaking table tests on long immersed tunnels subjected to non-
uniform seismic loadings. The results indicated that tunnel responses
under non-uniform earthquake excitation are much higher than those
under uniform earthquake excitation and the effect of spatial distribu-
tion of earthquake excitation should be considered in the design of
immersed tunnels.
Subway tunnel is an extremely long structure occupying large space.
The SVEGMs have a significant effect on the response of long structures
(Zerva, 2009), because of which commentaries and provisions on
seismic analysis of spatially extended structures are specified in many
design standards (CEN, 2004; CEN, 2005; Ministry of Housing and
Urban-Rural Development of the People’s Republic of China, 2010;
Ministry of Transport of the People’s Republic of China, 2008). There-
fore, the aseismic design of subway tunnel should account for the ef-
fects of ground motion spatial variability. Spatial variation of ground
motions results from: (i) wave passage effects, which is the difference in
arrival times of the seismic waves at different stations; (ii) incoherence
effects due to wave scattering or extended source effects; and (iii) local
site effects. A lot of contributions have been devoted to the aseismic
analyses of large-scale structures subjected to SVEGM, such as bridges
(Lou and Zerva, 2005; Dumanogluid and Soyluk, 2003; Apaydin et al.,
2016; Zhang et al., 2013; Liang et al., 2017) and pipelines (Soliman and
Datta, 1996; Zerva, 1993). However, studies about the influence of
SVEGM on the response of subway tunnel are still lacking.
Usually, the ISE of earthquake was utilized to study the seismic
response of large-scale subway tunnels and the spatial variance of
earthquake motions was neglected. To investigate the influence of
SVEGM on the response characteristics of subway shield tunnel, a series
of seismic analyses of shield tunnel under MSE and ISE of earthquake
were conducted through a case study in ABAQUS and the results were
compared. Besides, the previous numerical methods for 3D large-scale
underground structures are time-consuming and may cost a great deal
of computing resources. Therefore, the concept of response displace-
ment method (Ministry of Housing and Urban-Rural Development of
the People’s Republic of China, 2014) was adopted and realized to
proceed with the study. This method is named generalized response
displacement method (GRDM) and can facilitate the simulation with
higher efficiency and be more applicable to numerical analysis of large-
scale underground structures. Moreover, the effect of excitation
methods, spectral content and intensity of input motion as well as site
conditions on the response characteristics of shield tunnel was also
studied. The results can help to predict the underlying location of
seepage or severe damage of shield tunnel under earthquake and thus
the analysis procedure established in this paper possesses highly po-
tential for application to practical shield tunnel project. The metho-
dology for the problem, including ground motion selection, simulation
of SVEGMs and the time-history analysis method we used, is presented
in Section 2. In Section 3, the setup of a numerical model is presented as
well as some information about the reference engineering. Subse-
quently, Section 4 offers the dynamic responses of tunnels under MSE
and ISE. Finally, concluding remarks are given in Section 5.
2. Methodology
2.1. Ground motion selection
According to certain criteria, the input motions on bed rock were
selected based on the records from Kiban-Kyoshin Network (KiK-Net) in
Japan. The KiK-Net consists of more than 600 vertical arrays with an
uphole/downhole pair of strong motion seismometers. These arrays
were built by the National Research Institute for Earth Science and
217
Tunnelling and Underground Space Technology 77 (2018) 216–226
Disaster Resilience in Japan (NIED) after 1995 Kobe earthquake. More
than a million accelerograms recorded between 2009 January and 2014
June by all 697 KiK-Net sites during 5007 earthquakes events were
collected and then a ground motion data base was constructed. The
accelerograms encompassed an ample range of peak ground accelera-
tion (PGA) (from 0.1 to more than 1000 Gal), of epicentral distances
(from 1 km to more than 1000 km), of moment magnitudes (from 2 to
9), and of borehole depths (from 100 m to more than 3000 m). Based on
this data base the selection was carried out.
We utilized three conditions to conduct the selection: (i) borehole
depth ≥100 m; (ii). PGA ≥ 100 Gal; (iii). Magnitude = 6.5–7.5 M. We
selected the ground motions through two steps: Firstly, we picked a
total of 59 horizontal borehole accelerograms according to the condi-
tions above mentioned. Secondly, these accelerograms are assigned into
two groups by epicentral distance: 0–60 km and 120–180 km. There are
21 accelerograms with epicentral distance from 0 to 60 km and 7 ac-
celerograms with epicentral distance from 120 to 180 km. Then, we
calculated the Fourier spectrum of every ground motion and the aver-
aged one for each group, and selected the accelerogram which has the
smallest difference with the average Fourier spectrum as the horizontal
input ground motion. Make sure the selected accelerogram is the most
representative in this group.
2.2. Simulation of ground motions
Hao et al. (1989) proposed a method for generating spatially cor-
related ground motions in the process of which only the correlation
between present ground motion and previous ones was considered and
the Cholesky decomposition of power spectral density function (PSDF)
matrix was used correspondingly. Qu and Wang (1998) and Qu and
Wang (1998) extended Hao’s method by taking the correlation between
present ground motion and all the other ones into account. According to
Qu’s method, 1D-nV (one-dimensional, n-variate) stochastic process can
be expressed as:
n N −1
x j (t ) = ∑ ∑ ajm (ωk ) cos[ωk t + θjm (ωk ) + φmk ]
m=1 k=0 (1)
where N is the number of frequency interval. ωk is the kth value of ω .
φmk are random phase angles uniformly distributed between 0 and 2π.
And
ajm = 2 Δω |ujm (iωk )| (2)
Im[ujm (iωk )]
θjm = arctan
Re[ujm (iωk )] (3)
with Im and Re denoting the imaginary and the real part of a complex
number, respectively.Δω is frequency bandwidth. ujm are the elements
of the lower matrix obtained by using Cholesky decomposition of PSDF
matrix. i = −1 .
In the present study, the square root decomposition proposed by Wu
et al. (2011) and Wu et al. (2018) was used to decompose the PSDF
matrix. To incorporate full non-stationarity similar to the reference
accelerogram, the reference earthquake time history was divided into a
few of segments each having a unique, but stationary, PSDF (Rodolfo
Saragoni and Hart, 1974). Note that the number of segments is arbi-
trary, but in the present research three segments are used. For each
segment, the target PSDF was estimated using periodogram method and
used in conjunction with coherency model to establish cross-PSDF
matrix. Herein, the Sobczyk model which is applicable to describe the
coherency loss at the base rock was used as shown in Eq. (4).
γjm (iω) = |γjm (iω)| exp(−iωdjm cos(α )/ Vapp)
2
= exp(−βωdjm /Vapp)·exp(−iωdjm cos(α )/ Vapp) (4)
The phase difference spectrum was simulated for each segment ac-
cording to the distribution of that of the counterpart in the original
Y. Miao et al.
w1 Overlap w2 Overlap w3
1 tunnel project and site conditions is shown in Fig. 3. It is noted that part
0 3.2. Numerical model of free-field
T1 T2 T3
t Free-field model was established according to the site profile subtly,
Fig. 1. Trapezoid windows.
earthquake record and then corresponding phase angle was synthesized
and employed to replace the random phase angle in Eq. (1). As the
phase difference spectrum having the identical distribution as that of
the reference record was used, the similar amplitude variation with the
record can be achieved. For every simulation point, three subprocesses
corresponding to different segments can be generated through Eqs.
(1)–(3). The final fully non-stationary stochastic process can be gener-
ated through Eq. (5) using trapezoid windows which are shown in
Fig. 1. Herein, the trapezoid window was used to avoid abrupt variation
at the joint between two windowed subprocesses. Note that the overlap
can be set arbitrarily and herein it is set to T2/10 .
P 3.2.2. Boundary conditions and input earthquake motions
x j (t ) = ∑ wl (t ) al (t )
l=1 (5)
where P is the number of segments. wl (t ) denotes the lth trapezoid
window andal (t ) is the lth subprocess corresponding to the lth segment.
2.3. Generalized response displacement method
In this section, the GRDM for analysis of soil-tunnel system is in-
troduced briefly. The GRDM is based on the concept of response dis-
placement method described in Code for seismic design of urban rail
transit structures, GB50909-2014 (Ministry of Housing and Urban-
Rural Development of the People’s Republic of China, 2014) and a time
history analysis method including two steps: Firstly, a two-dimensional
(2D) free-field finite model is established and the displacement re-
sponse on the element nodes corresponding to the position of subway
tunnel to ground motions is extracted. Secondly, a 3D simplified soil-
tunnel model is constructed where soil around the tunnel is denoted by
using soil springs to consider the soil-structure interaction. A series of
joint springs are set between lining segments to account for the relative
motion between adjacent lining segments and the transmission of force
in tunnel. The extracted displacement time histories are input into the
simplified soil-tunnel model to calculate the subway tunnel response.
In fact, GRDM divides a seismic analysis of structures or systems
into two steps in general. Utilizing a 2D free-field model and a sim-
plified model saves calculation resources and leads to a dramatic re-
duction of simulation time. Therefore, comparing to those 3D solid
model-based methods, the GRDM can be employed to conduct the
seismic analysis of a more refined or a larger model. Combining an
example of application, the detail of GRDM for soil-tunnel model is
introduced in Section 3.3.
3. Numerical model
3.1. Yangtze tunnel project—sanyang road tunnel
The Yangtze River tunnel project of Sanyang road in Wuhan City is
1.3 km from the No. 2 Yangtze River Bridge. It is a river-crossing tunnel
project and consists of urban road and rail transit line 7. The diameter
of the tunnel is 15.7 m, ranking first in China and third in the world.
The inner and outer diameters of the liner are 13.9 m and 15.2 m, re-
spectively. The burial depth of tunnel is up to 24.1–55.2 m. Fig. 2
218
Tunnelling and Underground Space Technology 77 (2018) 216–226
illustrates the location of the tunnel. The profile of the Sanyang road
of the tunnel lies on the bed rock, while the rest of tunnel is in the silty-
fine sand.
3.2.1. Setup of the numerical model
considering the distribution of soil layers thoroughly. The horizontal
length and height of the model are 2.6 km and 150 m, respectively.
Four-node bilinear plane strain quadrilateral reduced integration ele-
ments (CPE4R) were dominated in the model, while a few of three-node
linear plane strain triangle elements (CPE3) were also used for transi-
tion. In terms of mesh, the variation of wavelength along the site depth
was taken into account, and the maximum size of the mesh (hmax) for
each soil layer was set to 1/8 to 1/10 of the corresponding wavelength
(Liu and Liao, 1989). Besides, under the condition above mentioned,
the discretization should also account for the position of subway tunnel
to facilitate the extraction of displacement response. Therefore, the
mesh size in the 2D free-site model covers a horizontal dimension of
0.45–2.0 m and a vertical dimension of 0.45–1.2 m, respectively.
For the boundary conditions, the viscous-spring artificial boundary
(Zhang et al., 2016) composed of linear springs and dampers were used
at the two lateral sides and the bottom side to absorb the energy of
scattering waves. A figure for regional boundary conditions is shown in
Fig. 4.
After wave selection, Niigata and Iwate earthquake ground motions
each having a horizontal component and a vertical component were
finally selected as input motions, named as E1 and E2 respectively, as
shown in Fig. 5(a). Herein, the Niigata earthquake represents near-field
earthquake, while Iwate the far-field one. Fig. 5(b) displays the Fourier
amplitude spectra of the horizontal ground motions for each group
(grey lines), the corresponding average spectra (dark lines) and the
Fourier amplitude spectra of the horizontal component of Niigata and
Iwate earthquake (dashed red lines). Some key information is shown in
Table 1.
For either E1 or E2, a total of 1301 acceleration time histories were
generated. The samples of synthesized SVEGMs and corresponding
PSDFs for Niigata earthquake are shown in Fig. 6 and those for Iwate
earthquake in Fig. 7. The PSDFs of simulation matches those of target
very well, which demonstrate the validness of the generated ground
motions. The earthquake motions derived from synthesized SVEGMs
and the original record were applied for MSE and ISE, respectively. In
this study, the PGAs of input motions were adjusted to 0.2 g and 0.4 g,
corresponding to severe and mega earthquake, respectively and com-
bining with different excitation methods, a total of eight study cases
were set as shown in Table 2.
3.2.3. Constitutive model
The modified Davidenkov viscoelastic dynamic constitutive model
was employed to describe the dynamic characteristics of soil (Zhao
et al., 2017), which can be expressed by
|γ −γc | ⎞ ⎤
τ −τc = Gmax ·(γ −γc )·⎡1−H ⎛ ⎜ ⎟
⎢ ⎝ 2na ⎠ ⎥ (6)
⎣ ⎦
1−R
(2na γ0 )2B = (γex ± γc )2B ·⎛ ⎞
⎝ R ⎠ (7)
1
τex ± τc A
R = ⎜⎛1− ⎞
⎟
⎝ G ·(γ
max ex ± γ )
c ⎠ (8)
where τ and γ are shear stress and strain, respectively; Gmax is the initial
Y. Miao et al. Tunnelling and Underground Space Technology 77 (2018) 216–226

Fig. 2. Location of Sanyang road tunnel.

Elevation (m)
0 South 0
-30 River bed -30
-60 -60
-90 -90
-120 -120
-150 -150
0 325 650 975 1300 1625 1950 2275 2600
Distance(m)
Banket Silty clay Medium-coarse sand Silty-fine sand Bed rock Tunnel bottom

Fig. 3. Site profile of the Sanyang shield tunnel project.

y beam elements each having a length of 1 m. The segment rings are

y linear elastic Timoshenko’s beams that allow for transverse shear de-
formation. The horizontal shear strain of the cross-section is constant,
Spring Mesh which means that the cross section after the deformation is still flat and
Boundary

not distorted. The density, elastic modulus and Poisson ratio of C60
x concrete lining segment are 2650 kg/m3, 3.6 × 10 4 MPa and 0.2, re-
Damper spectively.
Spring Damper x Four tridirectional soil springs were set around each lining segment
evenly to denote the soil-structure interaction as shown in Fig. 9(a). A
total of 5200 soil springs were set in the model. According to Ministry
of Housing and Urban-Rural Development of the People’s Republic of
China (2014), the stiffness of soil springs can be determined for dif-
ferent soil types:
Fig. 4. Schematic diagram of working boundary setting. Kt = 3Gu (9)

shear modulus; A, B, and γ0 are dimensionless fitting parameters; na is K1 = βKt (10)

the coefficient controlling the scale of hysteresis loop; τc and γc denote
the shear stress and strain at the last stress reversal point, respectively;
and the τex and γex are the shear stress and strain at the last extreme
value point, respectively. Through resonant column tests of soils in the
engineering site and curve fitting, the parameters of this model for
different soils are obtained. The material parameters for soils are shown
in Table 3.
The relation between normalized shear modulus (G/Gmax) and shear
strain (γ) and that between damping ratio (λ) and shear strain obtained
from the resonant column test of soil samples around the Sanyang Road
tunnel site are shown in Fig. 8.
3.3. Numerical model of soil-tunnels system
The simplified soil-tunnel model was 2.6 km in length and estab-
lished by using beam elements and springs. Every lining segment was
2 m in width and constructed by utilizing two 3D hollow linear B31
where Kt, Kl are spring coefficients per unit length perpendicular to and
parallel to soil layer, respectively; Gu denotes the shear modulus of
foundation soil corresponding to the maximum strain amplitude of
earthquake vibration; β is a constant which can be set to 1/3.
Considering the influence of the longitudinal joints of the tunnel
gap, the lining segments were connected by three joint nonlinear
springs to represent the effect of bolts, including an axial spring, a
tangential spring and a bending resistance spring. Note that the tan-
gential spring was assumed to be rigid. In other word, the tangential
relative motion and torsion between adjacent lining segments is not
considered. Through the simplification, the models of compression,
tension and rotation stiffness of joint spring are shown in Table 4.where
Ec denotes the elastic modulus of C60 concrete for the shield tunnel; Ac
the ring cross section area; and ls the length of a segment for the shield
tunnel; n the number of joint bolts at the intersegment; Es, As and lb are
the elastic modulus, cross section area and length of a joint bolt, re-
spectively; αs is the ratio of the post-yield tension stiffness over the

219
Y. Miao et al. Tunnelling and Underground Space Technology 77 (2018) 216–226

0.1 E1-EW 100

Acceleration(g)
E1
0.0
-0.1
0.1 10-1
E1-UD

Amplitude(m/s)
0.0
-0.1
40 10-2
0 20
100
0.1 E2-NS E2
Acceleration(g)

0.0
-0.1
10-1
0.1 E2-UD
0.0
-0.1
10-2 -1
0 35 70 10 100 101
t (s) Frequency(Hz)
(a) Selected ground motions (b) Fourier amplitude spectra
Fig. 5. The selected acceleration time histories and corresponding Fourier spectra.

initial tension stiffness for the ring joint bolts; r is the average of inner
and outer radius of lining segments; t is the wall thickness of the lining
segments; φ is a rotation angle for the neutral axis location of tunnel
segment cross section to the horizontal axis; αθ is the ratio of the post-
yield bending stiffness over the initial bending stiffness. The number of
joint springs is up to 3897 and the properties of joint nonlinear spring
are shown in Tables 5 and 6.
Based on the results of the nonlinear seismic response analysis of the
2D free-field, the horizontal and vertical displacement responses at the
element nodes (e.g. nodes A, B, C in Fig. 9(b)) corresponding to the
intermediate point of the segments (e.g. nodes A′, B′, C′ in Fig. 9(c))
were extracted and input to the fixed end of soil spring (Geng et al.,
2013) to calculate the tunnel response. Note that, to conduct tri-di-
rectional input, the horizontal displacement response is input in
transverse and longitudinal direction and the vertical one in vertical
direction (Anastasopoulos et al., 2007).
4. Results and discussion
A series of seismic responses of subway tunnel under MSE are
analyzed and compared with those under ISE, including intersegment
opening width, inner forces, and acceleration responses.
4.1. Opening width
The envelopes of intersegment opening width are displayed in
Fig. 10 where the opening width is the maximum value around each
intersegment. As can be seen, the opening width under ISE is lower than
that under MSE. And the distribution of opening width under ISE and
MSE are different. The ground motion with spatial variability gives rise
to differential motion between lining segments, especially the wave
propagation effect can cause bend of the whole tunnel and the coher-
ency effect and random phase may facilitate the deformation. There-
fore, comparing to ISE, MSE brings about higher opening width of
subway tunnel and different distributions. Moreover, under any
circumstance concerned, the opening width is extremely lower in the
distance range of approximate 1400–2000 m due to the strong con-
straint of hard rock on lining. Still, the opening width of tunnel under
ISE is lower than that under MSE in this range. At the interface between
soft soil and bed rock, a sharp and dramatic variation of opening width
shows up due to the difference of material properties on both sides of
the interface which may cause seepage or severe damage.
It is also found that the opening width under near-field earthquake
(E1) is over 2 times higher than that under far-field earthquake (E2).
This is because that the power contained in the near-field earthquake is
significantly larger than that contained in the far-field one (see Figs. 6
and 7. In the cases of different epicentral distances, the distributions of
opening width are still different due to the diverse spectral contents.
Besides, note that as PGA increases, the opening width response
increases and the distribution of opening width along the longitudinal
direction also changes. This is because that the free-field model is a
nonlinear system, including constitutive model and boundary condition
nonlinearity. The variation of spatial distribution of intersegment
opening width can be attributed to the evolution of soil properties as
PGA increases and the scattered soil layer distribution. Different soil
types exhibit different evolution characteristics as shear strain ampli-
tude increases: soft soil is damaged or enters plastic state first and then
the hard soil. Thus, the output of displacement under mega earthquake
(PGA = 0.4 g) is different from that under severe earthquake
(PGA = 0.2 g), which results in the different distribution of opening
width above mentioned.
Among all influence factors, the site conditions or soil layer dis-
tribution dominantly affect the longitudinal distribution of interseg-
ment opening width of tunnel, because the opening width is always
extremely low in the range of bed rock for all cases performed. The
excitation method, spectral content, and intensity of input motion can
also impact the longitudinal distribution of intersegment opening
width.

Table 1
Information of the selected ground motions.
Earthquake name Station Direction PGA Arias intensity (m/s) Predominant period Significant duration D5–95 (s) Hypocentral distance
(g) Tp (s) (km)

NGNH291103120359-E1 NGNH29 EW 0.106 0.1091 0.08 9.65 18.07

UD 0.066 0.0571 0.1 11.01
IWTH041104072332-E2 IWTH04 NS 0.116 0.1483 0.2 21.19 135.13
UD 0.094 0.1027 0.14 21.54

220
Y. Miao et al. Tunnelling and Underground Space Technology 77 (2018) 216–226

Acceleration Acceleration
0.1 E1-EW-Record
Target

(g)
0

PSDF(10-3m2/s3)
Simulation
E1-EW
-0.1
1
0.1 E1-EW-Simulation
(g)

0
-0.1
0
0 20 40 0 5 10 15 20 25
2
0.1
Acceleration

E1-UD-Record
Target
0 Simulation

PSDF(10-3m2/s3)
(g)

-0.1 E1-UD
1
0.1 E1-UD-Simulation
Acceleration

0
(g)

-0.1
00 5 10 15 20 25
0 20 40
t (s) Frequency(Hz)
Fig. 6. Samples of SVEGMs and PSDFs for Niigata earthquake.

1
Acceleration Acceleration

0.1 E2-NS-Record
Target
(g)

0 Simulation
PSDF(10-3m2/s3)

E2-NS
-0.1
0.5
0.1 E2-NS-Simulation
0
(g)

-0.1
0.0
0 35 70 0 5 10 15 20 25
1
Acceleration

0.1 E2-UD-Record Target

0 Simulation
(g)

PSDF(10-3m2/s3)

E2-UD
-0.1
0.5
Acceleration

0.1 E2-UD-Simulation
0
(g)

-0.1
0.0
0 35 70 0 5 10 15 20 25
t (s) Frequency(Hz)
Fig.7. Samples of SVEGMs and PSDFs for Iwate earthquake.

Table 2 Table 3
Matrix of analyses performed. Values of fitting parameters of the modified Davidenkov model.
No. of cases Input motion PGA/g Excitation method Soil types Density Poisson’s ratio Initial shear Fitting parameters
(kg/m3) modulus
Case 1 E1 0.2 MSE (MPa)
Case 2 E1 0.4 MSE A B γ0 (×10–4)
Case 3 E1 0.2 ISE
Case 4 E1 0.4 ISE Banket 1780 0.492 52 1.02 0.35 4.0
Case 5 E2 0.2 MSE Silty clay 1880 0.491 53 1.12 0.41 9.0
Case 6 E2 0.4 MSE Silt-fine 1895 0.493 62 1.10 0.37 6.5
Case 7 E2 0.2 ISE sand
Case 8 E2 0.4 ISE Medium- 2000 0.486 157 1.20 0.37 7.4
coarse
sand
Bad rock 2522 0.427 2042 1.30 0.40 10.0

221
Y. Miao et al. Tunnelling and Underground Space Technology 77 (2018) 216–226

1.0 20 Table 4
Parameters of joint spring and corresponding calculation equations.

0.8 Banket
16 Item Equation

(%)
Silty clay Compression stiffness Kc = EcAc/ls
Silty-fine sand Initial tension stiffness Kt1 = n EsAs/lb
0.6 12
G/Gmax

Damping ratio
Post-yield tension stiffness Kt2 = αsKt1
Medium-coarse sand
Initial bending stiffness cos3φ πr 3tEc
Bed rock K θ1 =
cosφ + (π / 2 + φ)sinφ ls
0.4 8 Post-yield bending stiffness Kθ2 = αθKt1

0.2 4
Table 5
Parameters of the axial spring.
0.0 -4 -3
0 Compression Initial Post-yield Tensile Tensile Yield
10 10 10 -2 10 -1 10 0 stiffness Kc (N/ tension tension yield ultimate deformation/
Shear strain m) stiffness stiffness force Ny force Nm ultimate
Kt1 (N/m) Kt2 (N/m) (kN) (kN) deformation
Fig. 8. G/Gmax ∼ γ and λ ∼ γ curves of the soil. δy/δm

4.72E11 1.17E10 1.17E8 2.74E4 3.24E4 2.3/60.6

4.2. Inner force

The inner force responses were also observed, including axial tensile
stress, axial compression stress, and longitudinal bending moment as
shown in Fig. 11(a)–(c), respectively. Generally, the inner force re-
sponse under MSE is also higher than that under ISE, due to the addi-
tional inner force resulting from spatial variation of ground motion. The
inner force responses under near-field earthquake is still greater than
those under far-field one. Moreover, different excitation methods can
also lead to different longitudinal distribution of inner forces. Under the
same input motion with different PGAs, the distribution formations of
inner force responses differ to some extent, and different earthquakes
result in remarkably different distributions. Each distribution curve of
inner force response under ISE exhibits non-uniform fluctuation,

Fig. 9. Soil-tunnel finite element model.

222
Y. Miao et al. Tunnelling and Underground Space Technology 77 (2018) 216–226

Table 6 responses of subway tunnel along longitudinal direction for all cases
Parameters of the torsion spring. performed. For both horizontal and vertical direction, the acceleration
Initial Post-yield Yield Ultimate Yield Ultimate response under MSE is higher than that under ISE in general. The
bending bending moment moment rotational rotational longitudinal distribution of acceleration response under MSE is dis-
stiffness stiffness My (kN Mm (kN m) angle angle tinguished from that under ISE, which may result in different damage
K θ1 (kN m/ K θ2 (kN m/ m) θ y (rad) θm (rad) locations. Under ISE, lining segments tend to vibrate uniformly, while
rad) rad)
in the case of MSE the spatial variability of ground motion leads to non-
3.89E11 4.38E9 6.90E4 8.74E4 1.78E−4 4.4E−3 uniform vibration due to: (i) the wave propagation causing phase dif-
ference and thus different displacement amplitude at different locations
at the same time; (ii) the coherency effect and random phase angle
implying the effect of soil inhomogeneity on the response distribution. resulting in the variation and randomness of ground motion and then
Thus, all the excitation methods, epicentral distance (or spectral con- the difference of response at different locations. Therefore, the long-
tent), intensity of ground motion as well as site conditions can affect the itudinal distributions of acceleration response under MSE and ISE are
distribution of inner force response. However, from Fig. 11(c), it can be different. An AF envelope combining results under MSE and those
found that considering diverse excitation methods, epicentral distances, under ISE is recommended to be used in the aseismic design of subway
and PGAs, the bending moment response is still always very low in the shield tunnel.
range of bed rock, because the hard rock constraints the deformation of All AF curves exhibit significant fluctuations, implying the effect of
lining. Therefore, the distribution of bending moment response is the variation of soil properties on the acceleration response. Under
especially related to the site conditions. different intensities of input motions, both horizontal and vertical ac-
celeration responses of subway shield tunnel exhibit different dis-
4.3. Acceleration response tribution characteristics. It is also found that under circumstance of
different epicentral distance, the distributions of acceleration response
Fig. 12 shows the amplification factors (AFs) of peak acceleration are different due to the diverse dynamic characteristics and/or spectral

6
Maximum Intersegment

E1-0.2g ISE
opening width(mm)

MSE
4

0
0 650 1300 1950 2600
Distance (m)
3
Maximum Intersegment

E2-0.2g ISE
opening width(mm)

MSE
2

0
0 650 1300 1950 2600
Distance (m)
8
Maximum Intersegment

E1-0.4g ISE
opening width(mm)

6 MSE

0
0 650 1300 1950 2600
Distance (m)
3
Maximum Intersegment

E2-0.4g ISE
opening width(mm)

MSE
2

0
0 650 1300 1950 2600
Distance (m)

Fig. 10. Intersegment opening width of subway tunnel.

223
Y. Miao et al. Tunnelling and Underground Space Technology 77 (2018) 216–226

1.2 0.9
ISE ISE
MSE

Tensile stress(MPa)
MSE

Tensile stress(MPa)
0.8 0.6

0.4 0.3
E1-0.2g E2-0.2g
0.0 0.0
0 650 1300 1950 2600 0 650 1300 1950 2600
Distance (m) Distance (m)
1.5 1.2
ISE ISE

Tensile stress(MPa)
MSE
Tensile stress(MPa)

MSE
1.0 0.8

0.5 0.4

E1-0.4g E2-0.4g
0.0 0.0
0 650 1300 1950 2600 0 650 1300 1950 2600
Distance (m) Distance (m)

(a) Tensile stress response

1.0 0.75
Compression stress(MPa)

Compression stress(MPa)
ISE ISE
0.75 MSE MSE
0.5
0.5
0.25
0.25
E1-0.2g E2-0.2g
0.0 0.0
0 650 1300 1950 2600 0 650 1300 1950 2600
Distance (m) Distance (m)
2.0 1.5
Compression stress(MPa)

Compression stress(MPa)

ISE ISE
1.5 MSE MSE
1.0
1.0
0.5
0.5
E1-0.4g E2-0.4g
0.0 0.0
0 650 1300 1950 2600 0 650 1300 1950 2600
Distance (m) Distance (m)

(b) Compression stress response

0.9 0.6
ISE ISE
MSE MSE
Bending moment
Bending moment

(kN m×104)
(kN m×104)

0.6
0.3
0.3

E1-0.2g E2-0.2g
0 0.0
0 650 1300 1950 2600 0 650 1300 1950 2600
Distance (m) Distance (m)
1.5 1.0
ISE ISE
Bending moment
Bending moment

(kN m×104)

MSE
(kN m×104)

MSE
1.0
0.5
0.5

E1-0.4g E2-0.4g
0.0 0.0
0 650 1300 1950 2600 0 650 1300 1950 2600
Distance (m) Distance (m)

(c) Bending moment response

Fig. 11. Inner force responses of subway tunnel.

contents between E1 and E2. For E1 in horizontal direction, the AFs characteristics above mentioned was also observed in the axial com-
under MSE and ISE in the range of bed rock are especially higher due to pression stress response shown in Fig. 11(b). The amplification degree
the filtering effect of this engineering site; however, this phenomenon for E1 and E2 in vertical direction is generally larger than that in
was not exhibited in the AF curves for horizontal component of E2. The horizontal direction and no extremely remarkable amplification effect

224
Y. Miao et al. Tunnelling and Underground Space Technology 77 (2018) 216–226

2 3
E1-0.2g ISE E1-0.2g ISE
Horizontal MSE Vertical MSE

AF

AF
1

0 0
0 650 1300 1950 2600 0 650 1300 1950 2600
Distance(m) Distance(m)
2 3
E2-0.2g ISE E2-0.2g ISE
Horizontal MSE Vertical MSE

2
AF

AF
1

0 0
0 650 1300 1950 2600 0 650 1300 1950 2600
Distance(m) Distance(m)
2 3
E1-0.4g ISE E1-0.4g ISE
Horizontal MSE Vertical MSE

2
AF

AF

0 0
0 650 1300 1950 2600 0 650 1300 1950 2600
Distance(m) Distance(m)
2 3
E2-0.4g ISE E2-0.4g ISE
Horizontal MSE Vertical MSE

2
AF

AF

0 0
0 650 1300 1950 2600 0 650 1300 1950 2600
Distance(m) Distance(m)
Fig. 12. Envelopes for AF of acceleration response for all cases.

is observed in the distance range of bed rock; by contrast, a lot of ex-
treme values are presented near the interface between soft soil and bed
rock. Therefore, for the acceleration response of tunnel, the influence of
distribution of soil layers is not dominant any longer and all the ex-
citation method, epicentral distance, intensity of input motion and site
conditions can clearly affect the longitudinal distribution of accelera-
tion response.
5. Conclusion
characteristics of 2.6 km-long shield tunnel across the Yangtze River,
China, under SVEGMs, which is performed using the GRDM. The in-
tersegment opening width, inner force, and acceleration responses of
tunnel under MSE were studied and the results were compared with
those under ISE. The influence of excitation method, epicentral dis-
tance, and intensity of input motion as well as site conditions on the
longitudinal seismic characteristics of subway shield tunnel was also
qualitatively studied. The following conclusions of practical sig-
nificance emerge from our study:

This paper investigates the longitudinal nonlinear seismic response

• The seismic responses, including intersegment opening width, inner
225
Y. Miao et al.
force, and acceleration responses, of subway shield tunnel under
MSE are generally larger than those under ISE. Different support
excitation methods can result in different longitudinal distributions
of opening width, inner force, and acceleration responses. Utilizing
the result under MSE or the combination of the results under MSE
and those under ISE in aseismic analysis and design of subway shield
tunnel is more reliable.
• The spatial distribution of intersegment opening width and bending
moment of tunnel are highly related to site conditions. Spatial in-
homogeneity of soils can cause dramatic changes and extreme va-
lues in the intersegment opening width and bending moment along
the tunnel longitudinal which may result in seepage or severe da-
mage of tunnel.
• Although the peak acceleration is adjusted to the identical value, the
near-field earthquake can bring about more severe damage to
subway tunnel. The frequency content or spectral characteristics of
earthquake itself can also affect the trend of the response distribu-
tion.
•
Due to the evolution of soil properties and the site inhomogeneity,
the intensity of input motion can also affect the distribution of
seismic responses of subway shield tunnel.
Acknowledgements
This study was financially supported by the National Key Research
and Development Program of China (No. 2016YFC0800206), the
National Natural Science Foundation of China (Nos. 51778260,
51378234, 51278213).
References
Amorosi, A., Boldini, D., 2009. Numerical modelling of the transverse dynamic behavior
of circular tunnels in clayey soils [J]. Soil Dyn. Earthq. Eng. 29 (6), 1059–1072.
Anastasopoulos, I., Gerolymos, N., Drosos, V., et al., 2007. Nonlinear response of deep
immersed tunnel to strong seismic shaking. J. Geotech. Geoenviron. Eng., ASCE 133
(9), 1067–1090.
Apaydin, N.M., Bas, S., Harmandar, E., 2016. Response of the Fatih Sultan Mehmet
Suspension Bridge under spatially varying multi-point earthquake excitations [J].
Soil Dyn. Earthquake Eng. 84, 44–54.
Bao, X., Xia, Z., Ye, G., et al., 2017. Numerical analysis on the seismic behavior of a large
metro subway tunnel in liquefiable ground [J]. Tunn. Undergr. Space Technol. 66,
91–106.
CEN, 2004. Eurocode 8: Design provisions of structures for earthquake resistance, Part 1:
General rules, seismic actions and rules for building, EN 1998-1:2004. European
Committee for Standardization, Brussels.
CEN, 2005. Eurocode 8: Design provisions of structures for earthquake resistance, Part 2:
Bridges, EN 1998-2:2005. European Committee for Standardization, Brussels.
Chen, Z.Y., Liang, S.B., He, C., 2017. Influence of Incoherence and wave passage effects
on seismic performances of long immersed tunnel [J]. Int. J. Comput. Methods 2017,
1840012.
Chen, C.H., Wang, T.T., Jeng, F.S., et al., 2012. Mechanisms causing seismic damage of
tunnels at different depths [J]. Tunnel Undergro Space Technol. Incorporat.
Trenchless Technol. Res. 28 (1), 31–40.
Dumanogluid, A.A., Soyluk, K., 2003. A stochastic analysis of long span structures sub-
jected to spatially varying ground motions including the site-response effect [J]. Eng.
Struct. 25 (10), 1301–1310.
Geng, P., He, C., Yan, Q.X., 2013. The current situation and prospect of seismic analysis
methods for tunnel structure [J]. China Civil Eng. J. 46 (S1), 262–268 (in Chinese).
Hao, H., Oliveira, C.S., Penzien, J., 1989. Multiple-station ground motion processing and
simulation based on SMART-1 array date [J]. Nucl. Eng. Des. 111 (3), 293–310.
226
Tunnelling and Underground Space Technology 77 (2018) 216–226
Hashash, Y.M.A., Tseng, W.S., Krimotat, A., 1998. Seismic soil-structure interaction
analysis for immersed tube tunnels retrofit [J]. Geotech. Spec. Publ. 2 (75),
1380–1391.
Lee, V.W., Trifunace, M.D., 1979. Response of tunnels to incident SH waves [J]. J. Eng.
Mech. Div.-ASCE 105 (4), 643–659.
Liang, F.Y., Jia, Y.J., Sun, L.M., et al., 2017. Seismic response of pile groups supporting
long-span cable-stayed bridge subjected to multi-support excitations [J]. Soil Dyn.
Earthquake Eng. 101, 182–203.
Liu, J.B., Liao, Z.P., 1989. Elastic wave motion in discrete grids (Ⅱ): comparison of
common finite element models [J]. Earthq. Eng. Eng. Vib. 9 (2), 1–12 (in Chinese).
Lou, L., Zerva, A., 2005. Effects of spatially variable ground motions on the seismic re-
sponse of a skewed, multi-span, RC highway bridge [J]. Soil Dyn. Earthq. Eng. 25
(7–10), 729–740.
Ministry of Housing and Urban-Rural Development of the People’s Republic of China,
2010. Code for seismic design of buildings, GB 50011-2010. Chinese Architectural
Industry Press, Beijing (in Chinese).
Ministry of Housing and Urban-Rural Development of the People’s Republic of China,
2014. Code for seismic design of urban rail transit structures, GB50909-2014. China
Project Press, Beijing (in Chinese).
Ministry of Transport of the People’s Republic of China, 2008. Guidelines for seismic
design of highway bridges, JTG/T B02-01-2008. China Transport Press, Beijing (in
Chinese).
Park, D.H., Sagong, M., Kwak, D.Y., Jeong, C.G., 2009. Simulation of tunnel response
under spatially varying ground motion [J]. Soil Dyn. Earthq. Eng. 29, 1417–1424.
Qu, T.J., Wang, Q.X., 1998. Simulation of spatially correlative time histories of multi-
point ground motion, Part II: Application of fundamental formulas [J]. Earthq. Eng.
Eng. Vib. 18 (2), 25–32 (in Chinese).
Qu, T.J., Wang, Q.X., 1998. Simulation of spatially correlative time histories of multi-
point ground motion, Part I: Fundamental formulas [J]. Earthq. Eng. Eng. Vib. 18 (1),
8–15 (in Chinese).
Rodolfo Saragoni, G., Hart, G.C., 1974. Simulation of artificial earthquakes [J]. Earthq.
Eng. Struct. Dyn. 2 (3), 249–267.
Shahrour, I., Khoshnoudian, F., Sadek, M., et al., 2010. Elastoplastic analysis of the
seismic response of tunnels in soft soils [J]. Tunnel. Undergr. Space Technol.
Incorporat. Trenchless Technol. Res. 25 (4), 478–482.
Soliman, H.O., Datta, T.K., 1996. Response of overground pipelines to random ground
motion [J]. Eng. Struct. 18 (7), 537–545.
Tsinidis, G., 2017. Response characteristics of rectangular tunnels in soft soil subjected to
transversal ground shaking [J]. Tunn. Undergr. Space Technol. 62 (1), 1–22.
Wu, Y.X., Gao, Y.F., Li, D.Y., 2011. Simulation of spatially correlated earthquake ground
motions for engineering purposes[J]. Earthq. Eng. Eng. Vibr. 10 (2), 163–173.
Wu, Y.X., Gao, Y.F., Zhang, N., et al., 2018. Simulation of spatially varying non-gaussian
and nonstationary seismic ground motions by the spectral representation method. J.
Eng. Mech. ASCE 144 (1), 04017143.
Yan, X., Yuan, J.Y., Yu, H.T., et al., 2016. Multi-point shaking table test design for long
tunnels under non-uniform seismic loading [J]. Tunnel. Undergr. Space Technol.
Incorporat. Trenchless Technol. Res. 59, 114–126.
Yu, H.T., Yuan, Y., Xu, G.P., et al., 2016. Multi-point shaking table test for long tunnels
subjected to non-uniform seismic loadings – Part II: Application to the HZM im-
mersed tunnel [J]. Soil Dyn. Earthquake Eng.
Yu, H.T., Yan, X., Bobet, A., et al., 2017. Multi-point shaking table test of a long tunnel
subjected to non-uniform seismic loadings[J]. Bull. Earthq. Eng. 1–19.
Yuan, Y., Yu, H.T., Li, C., et al., 2016. Multi-point shaking table test for long tunnels
subjected to non-uniform seismic loadings – Part I: Theory and validation [J]. Soil
Dyn. Earthquake Eng.
Zerva, A., 1993. Pipeline response to directionally and spatially correlated seismic ground
motions [J]. J. Pressure Vessel Technol. ASME 115, 53–58.
Zerva, A., 2009. Spatial Variation of Seismic Ground Motions: Modeling and Engineering
Applications. CRC Press, Taylor and Francis Group.
Zhang, D.Y., Li, X., Yan, W.M., et al., 2013. Stochastic seismic analysis of a concrete-filled
steel tubular (CFST) arch bridge under tridirectional multiple excitations [J]. Eng.
Struct. 52 (9), 355–371.
Zhang, X.L., Li, X.J., Chen, G.X., et al., 2016. An improved method of the calculation of
equivalent nodal forces in viscous-elastic artificial boundary [J]. Chinese J. Theor.
Appl. Mech. 48 (5), 1126–1135 (in Chinese).
Zhao, D.F., Ruan, B., Chen, G.X., et al., 2017. Validation of the modified irregular loading-
reloading rules based on Davidenkov skeleton curve and its equivalent strain algo-
rithm implemented in ABAQUS [J]. Chinese J. Geotech. Eng. 39 (5), 888–895 (in
Chinese).