Tunnelling and Underground Space Technology 62 (2017) 1–22

Contents lists available at ScienceDirect

Tunnelling and Underground Space Technology

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

Response characteristics of rectangular tunnels in soft soil subjected

to transversal ground shaking
Grigorios Tsinidis
Research Unit of Soil Dynamics and Geotechnical Earthquake Engineering, Department of Civil Engineering, Aristotle University of Thessaloniki, PO BOX 424, Thessaloniki
GR-54124, Greece

a r t i c l e i n f o a b s t r a c t

Article history: A numerical parametric study was conducted on diverse soil-rectangular tunnel systems, aiming to shed
Received 24 August 2016 light on critical response characteristics of rectangular tunnels subjected to transversal ground shaking.
Received in revised form 9 November 2016 Salient parameters that affect the dynamic response, such as: (i) the soil-tunnel relative stiffness and
Accepted 10 November 2016
interface properties, (ii) the shape, dimensions and burial depth of the tunnel section, (iii) the soil deposit
Available online 22 November 2016
characteristics, and (iv) the input motion characteristics, were accounted for in this study. This paper
summarizes the key findings of this investigation, focusing on the complex deformation modes of the
Keywords:
tunnels during shaking, the dynamic earth pressures and the soil dynamic shear stresses developed
Rectangular tunnels
Dynamic analysis
around the tunnel, and the dynamic lining forces. The numerical results indicated a combined racking-
Racking rocking deformation pattern for the tunnels during shaking, while inward deformations of the slabs
Rocking and the side-walls were also observed for flexible tunnels, when soil inelasticity was encountered. To
Interface conditions quantify the racking deformation of rectangular tunnels, a series of numerical racking ratio - flexibility
Dynamic earth pressures ratio (R-F) relations were developed and compared with existing analytical and empirical ones. The rock-
Soil dynamic shear stresses ing response of rectangular tunnels was quantified by means of dimensionless relations (h/cff-F), similar
Dynamic lining forces to the R-F relations. The soil-tunnel relative stiffness, the interface characteristics and the soil yielding
affected significantly the above relations, as well as the dynamic earth pressures, the soil dynamic shear
stresses and the dynamic forces developed on the lining during shaking. The presented results lead to a
better understanding of the seismic response of rectangular tunnels in soft soil, while the proposed rela-
tions contribute towards the improvement of the R-F analysis method.
Ó 2016 Elsevier Ltd. All rights reserved.

1. Introduction against ground shaking and earthquake induced ground failures

Tunnels constitute crucial components of the transportation
and utility networks in an urban area. Although these types of
structures behaved better than above ground structures during
recent strong earthquakes, post-earthquake observations have
demonstrated that they may undergo extensive deformations or
even collapse under strong ground shaking, especially when seis-
mic design provisions are not encountered (e.g. Sharma and Judd,
1991; Wang, 1993; Iida et al., 1996; Power et al., 1998).
The seismic response of tunnels is quite distinct compared to
that of above ground structures, as the kinematic loading induced
by the surrounding ground prevails over inertial loads stemming
from the oscillation of the structure itself (Wang, 1993; Hashash
et al., 2001).
The response of tunnels and embedded structures (e.g. culverts,
underground stations, underground reservoir structures, etc.)
E-mail address: gtsinidi@civil.auth.gr
has been a subject of intense research by a series of experimental
(Chou et al., 2010; Shibayama et al., 2010; Chian and
Madabhushi, 2012; Cilingir and Madabhushi, 2011a, 2011b,
2011c; Lanzano et al., 2012; Chen et al., 2013; Tsinidis et al.,
2015; Abuhajar et al., 2015a, 2015b; Ulgen et al., 2015;
Hushmand et al., 2016; Tsinidis et al., 2016a, 2016b), numerical
(Hashash et al., 2005; Huo et al., 2005; Anastasopoulos et al.,
2007, 2008; Amorosi and Boldini, 2009; Kontoe et al., 2011,
2014; Bilotta et al., 2014; Baziar et al., 2014; Lanzano et al.,
2015) and analytical (Bobet et al., 2008; Park et al., 2009; Bobet,
2010) studies, during the recent years. However, several issues
related to the dynamic response of rectangular tunnels, including
(i) the complex deformation modes of tunnels mobilized during
ground shaking, (ii) the amplitude and distribution of the seismic
earth pressures on the tunnel side-walls and slabs, and (iii) the soil
dynamic shear stresses mobilized around the perimeter of the tun-
nel during shaking, are still under investigation. Hence, conven-
tional design specifications are based primarily on simplified

http://dx.doi.org/10.1016/j.tust.2016.11.003
0886-7798/Ó 2016 Elsevier Ltd. All rights reserved.
2 G. Tsinidis / Tunnelling and Underground Space Technology 62 (2017) 1–22
methods (e.g. Wang, 1993; Penzien, 2000; Hashash et al., 2001;
ISO, 2005; FHWA, 2009), the implementation of which, may lead
to substantial differences in the seismic design for underground
structures (Pitilakis and Tsinidis, 2014).
This study aims to shed light on the above open issues, by
means of an extended numerical parametric study, which was con-
ducted on a series of idealized soil-rectangular tunnel configura-
tions. The analyses are focused in the transversal direction, as
this commonly related to the maximum stress state of the lining.
Crucial parameters that affect the dynamic response of the soil-
tunnel system, such as: (i) the soil-tunnel relative stiffness, (ii)
the soil-tunnel interface properties, (iii) the shape, dimensions
and burial depth of the tunnel, (iv) the soil properties and response
and (v) the input motion characteristics, are accounted for in this
study. The presented results, in terms of (i) racking deformations
and rotation of the tunnel sections (due to rocking), (ii) dynamic
earth pressures and soil dynamic shear stresses developed around
the tunnel, and (iii) dynamic internal forces, lead to a better under-
standing of the dynamic response of rectangular tunnels in soft
soil.
2. Outline of the parametric analysis
The problem, examined herein, is schematically illustrated in
Fig. 1a. A concrete rectangular tunnel (width: a
height: b) is
embedded in a homogeneous and uniform soil deposit, resting on
elastic bedrock. The distance between the invert slab of the tunnel
and the bedrock is larger than the largest dimension of the tunnel.
Hence, the potential effect of the bedrock on the tunnel response is
minimized.
The size of the tunnel, the mechanical properties of which are
summarized in Table 1, ranged, so as to model diverse typologies
(e.g. culverts, subway tunnels, underground arteries, etc.), as dis-
played in Fig. 1b. In particular, the tunnel width, a, ranged between
2.0 m and 18.0 m, while the tunnel height, b, ranged between 2.0 m
and 10.0 m. Hence, the corresponding aspect ratios, k = a/b, varied
between 0.5 and 3.0. It is noted that the aspect ratios discussed
throughout the paper are computed based on the distances of
the centroids of the structural elements.
An internal column was considered at the middle of the span
(i.e. central column) for tunnels with aspect ratio, k, greater than
2 (e.g. tunnels with large spans), as this concept is more rational
from a static design viewpoint. To investigate the effect of this col-
umn on the dynamic response of the tunnel, additional analyses
were conducted for the 12
6 (m) tunnel (k = 2), excluding the
internal column (single box analyses), and their predictions were
compared to those of the double box analyses (i.e. analyses with
the internal column).
The burial depth of the tunnels, h, varied between 3 m and
12 m, to investigate the response of both shallow and deep tunnels.
(a)
h
t
b Tunnel
H=40 m
a
Soil
L=140 m
Fig. 1. (a) Schematic illustration of the problem and (b) dimensions of the investigated tunnels.
Table 1
Lining mechanical properties.
Elastic modulus, Poisson Density,
E (GPa) ratio, v q (t/m3)
Concrete C30/37 32 0.2 2.5
Hence, the overburden depth ratio (defined as h/a) varied between
0.17 and 6.0.
The majority of analyses were carried out assuming a visco-
elastic homogeneous soil deposit, with soil properties, correspond-
ing to soil classes B and C, according to the Eurocode 8 (CEN, 2004).
Table 2 summarizes the mechanical properties of the investigated
soil deposits. The selection of a homogeneous, isotropic and visco-
elastic soil was made, in order to generalize the conclusions
regarding the tunnels’ response (e.g. deformations modes). Addi-
tionally, this assumption facilitated the direct comparison of the
numerical predictions with key results and references that are
commonly adopted in the seismic analysis of tunnels (Wang,
1993; Penzien, 2000; Anderson et al., 2008).
A sensitivity analysis was conducted, to investigate the effects
of crucial soil parameters, such as the soil’s Poisson ratio, v, and
soil’s damping, D, on the response of the tunnels, as discussed in
the following sections. Additionally, a series of nonlinear analyses
were performed for 6
6 (m) square tunnels, assuming diverse
strength properties for the surrounding ground (Table 2), aiming
to investigate the effect of soil yielding on response characteristics
of the tunnel.
For the sake of simplicity, the lining thickness, t, was assumed
to be constant for all structural elements (i.e. slabs, side-walls
and columns), while it was deliberately chosen and changed for
each soil-tunnel configuration, to achieve flexibility ratios, F, vary-
ing between 0.2 (i.e. rigid tunnel) and 10 (i.e. quite flexible tunnel).
A series of static frame analyses were performed, following Wang
(1993), to facilitate the selection of the lining thickness for the var-
ious flexibility ratios. For the computation of the flexibility ratios,
the following relation was used:
F ¼ ðGs
aÞ=ðS
bÞ ð1Þ
where Gs is the soil shear modulus and S is the required force to
cause a unit racking deflection of the structure. It is worth noticing
that some of the adopted lining thicknesses are unrealistically large
or small for practical applications. However, the selection was made
in order to create an abacus of cases that can describe a wide range
of lining flexibilities, including extreme scenarios.
3. Numerical analysis
The analyses were carried out under plane strain conditions,
using the finite element code ABAQUS (ABAQUS, 2012). The soil
(b) 6 × 6 (m) 12 × 6 (m) 12 × 6 (m)
18 × 6 (m) 3 × 6 (m) 9 × 6 (m)
10 × 10 (m)
2 × 2 (m)
 G. Tsinidis / Tunnelling and Underground Space Technology 62 (2017) 1–22
Table 2
Soil properties.
EC8 soil class Vs (m/s) Poisson ratio,
B1 - sand B 400 0.3
B2 - clay B 400 0.3
C1 - sand C 250 0.3
C2 - clay C 250 0.3
Elastic rock A 1000 0.3
a
Only for the elasto-plastic analyses.
Soil: plane strain elements
Dashpots
Fig. 2. Numerical model of a shallow 18
8 (m) tunnel (h = 3 m).
grid dimensions and boundary conditions were selected, so as to
assure ‘free-field’ and ‘quasi transparent’ conditions at the side
boundaries and the base of the numerical model, respectively. A
sensitivity analysis was conducted, indicating that a soil grid of a
total length of 140 m, with a depth of 40 m, was adequate.
The soil was meshed with quadratic plane-strain elements,
while the tunnel lining was simulated with beam elements
(Fig. 2). The element size was selected, so as to ensure the efficient
reproduction of all the waveforms of the whole frequency range
under study (i.e. frequencies f = 0.2–10 Hz). More refined models,
used during preliminary sensitivity analyses, had a negligible effect
on the computed results, increasing significantly the computa-
tional cost.
Kinematic tie constraints were introduced at the side bound-
aries of the numerical models, forcing the opposite vertical sides
to move simultaneously. The base boundary was simulated as elas-
tic bedrock, by introducing dashpots, following the Lysmer and
Kuhlemeyer scheme (Lysmer and Kuhlemeyer, 1969). The dash-
pots coefficients, c, were defined as the product of the mass den-
sity, q, and shear wave velocity, Vs, of the underlying bedrock,
including also the ‘effect’ area, A, of each dashpot, to maintain pro-
portional results for any horizontal element size:
c ¼ q
Vs
A
To investigate the crucial effect of interface conditions on the
soil-tunnel system response, the analyses were conducted, by
assuming, either full bonding between the soil and the tunnel ele-
ments (i.e. no-slip tie conditions), or full-slip interface conditions.
In the latter case, a finite sliding hard contact algorithm was imple-
mented (ABAQUS, 2012) with the tangential behavior of the inter-
face being modelled using the penalty friction formulation,
introducing the Coulomb friction model (friction coefficient,
l = 0). With reference to the normal interface behavior, two series
of analyses were performed. In the first series of analyses, the
potential separation between the soil and the tunnel elements
was precluded (i.e. tensile stresses are transmitted to the soil ele-
3
v Density, q (t/m3) Damping, D (%) Cohesion, c (kPa)a Friction angle, u (°)a
1.9 5 10 44
1.9 5 50 27
1.8 5 1 36
1.8 5 10 25
2.2 – – –
Tunnel: beam elements,
Soil-tunnel interface Displacement constraints
a (t)
ments). In the final series of analyses the potential separation
between the soil and the tunnel elements was permitted (i.e. no
tensile stresses are transmitted to the soil elements). It is worth
noticing that the above conditions constitute upper and lower limit
cases for the actual interface conditions.
The tunnel lining was assumed to behave in an elastic fashion.
Hence, a linear elastic model was implemented for the simulation
of the lining response. The model was calibrated adopting the
mechanical properties presented in Table 1.
The soil response under seismic shaking was simulated in two
ways. For the elastic analyses, a visco-elastic model was imple-
mented, introducing the soil shear modulus and viscous damping.
For the elasto-plastic analyses, a non-associated Mohr-Coulomb
yield criterion was combined with the above visco-elastic model.
Hence, the permanent deformations of the soil were encountered
in the elasto-plastic analyses. To implement the latter model in this
study, the soil shear modulus and damping were set equal to those
of the visco-elastic model. In both cases, the viscous damping was
introduced in the form of the frequency dependent Rayleigh type.
The double frequency approach was implemented for the calibra-
tion of the Rayleigh parameters, which were properly tuned for
the dominant frequencies of the soil deposits.
The analyses were carried out in two steps; initially the gravity
ð2Þ
loads were introduced, within a static step. Subsequently, the
dynamic loading was introduced, within an implicit dynamic step.
In particular, the shaking ground motions were introduced at the
base boundary through the dashpots in terms of acceleration time
histories.
Three real acceleration time histories with diverse frequency
characteristics (Table 3) were used as input motions, to investigate
the effect of the motion characteristics on the tunnels response.
The records from the 1976 Friuli earthquake and the 1999 Athens
earthquake were recorded on rock outcropping conditions. The
Takatori record from the 1995 Kobe earthquake was selected as
one of the most devastating seismic motions ever recorded. The
majority of analyses were performed scaling the records to a peak
4 G. Tsinidis / Tunnelling and Underground Space Technology 62 (2017) 1–22

Table 3
Input motion characteristics.

A/A Earthquake name Country Station name Magnitude Mw PGA (g) Record at Vs (m/s) EC8 soil class
EQ1 Athens, 1999 Greece Kipsely District 6.04 0.12 934 A
EQ2 Friuli, 1976 Italy Tolmezzo-Diga Ambiesta 6.4 0.34 1030 A
EQ3 Kobe, 1995 Japan Takatori 6.9 0.61 <180 D

Athens (1999) Friuli (1976) Takatori (1995)

(a) 0.4 0.4 0.4

0.2 0.2 0.2

A (g)

0 0 0

−0.2 −0.2 −0.2

−0.4 −0.4 −0.4

0 5 10 15 20 0 5 10 15 20 0 10 20 30 40
t(s) t(s) t(s)
(b) 5
(c) 0.8
Athens (1999)
3.75 Friuli (1976) 0.6
PSA/PSAo

PSD (m)
Takatori (1995)
2.5 EC8 (type A) 0.4

1.25 0.2

0 0
0 1 2 3 4 0 1 2 3 4
T (s) T (s)

Fig. 3. (a) Time histories of the input motions used in this parametric study (scaled to a PGA = 0.35 g), (b) acceleration response spectra of the input motions and (c)
displacement response spectra of the input motions.

ground acceleration PGA = 0.35 g (for outcropping conditions).
Additional analyses were performed within a sensitivity analysis
using the Friuli and the Takatori records scaled to PGAs = 0.1 g
and 0.6 g. Fig. 3a illustrates the scaled time histories of the input
motions (PGA = 0.35 g), whereas in Fig. 3b and c the corresponding
acceleration and displacement response spectra are compared. It is
worth noticing the significant difference of the acceleration and
the displacement response spectra of the Takatori record compared
to the other records.
4. Discussion of the results
Representative results of the parametric study, including: (i)
deformation modes of the soil-tunnel systems, (ii) racking ratio -
flexibility ratio (R-F) relations, (iii) normalized rotation - flexibility
ratio (h/cff-R) relations, (iv) dynamic earth pressures and soil
dynamic shear stresses mobilized around the tunnel, and (v)
dynamic lining forces, are presented and discussed in the following
sections.
4.1. Dynamic deformed shapes of the tunnel sections
The deformation patterns of the tunnel sections due to ground
shaking (i.e. ‘dynamic’ deformed shapes) were evaluated by sub-
tracting the deformations caused by the gravity loads from the
total deformations computed for the sections. Fig. 4 illustrates typ-
ical dynamic deformed shapes of shallow tunnels (e.g. 6 m
6 m;
12 m
6 m; 18 m
6 m) embedded in elastic soil (soil type B1).
The deformed shapes, which are computed for the shaking motion
EQ2 (PGA = 0.35 g) under full bonding conditions between the soil
and the tunnel elements, correspond to the time step of maximum
racking distortion of the tunnel sections. In line with recent exper-
imental findings (Cilingir and Madabhushi, 2011b; Tsinidis et al.,
2015, 2016a), a complex racking-rocking response is identified
for the tunnels during shear shaking. In particular, the rigid tunnels
(i.e. F = 0.2) display a rocking mode of vibration combined with a
reduced racking distortion. For F = 1.0 (i.e. the tunnel shares the
same stiffness with the surrounding ground), the tunnels are sub-
jected to pure shearing, exhibiting no rocking. The flexible tunnels
(i.e.F > 1.0) are subjected to significant racking distortion combined
with increased rocking. Naturally, the racking distortion is ampli-
fied with the increase of the flexibility of the lining, while rocking
is following a similar trend.
Regardless of their shape and burial depth, the tunnels exhibit a
particular rotation trend. More specifically, the tunnels are rotating
around the centroid of their sections. For shaking motion towards
right, rigid tunnels (F < 1.0) are subjected to a clockwise rotation,
while flexible tunnels (F > 1.0) are subjected to a counterclockwise
rotation (Fig. 5a). The exact opposite trends are observed for shak-
ing motion towards left (Fig. 5b). These observations are consistent
for all the investigated cases (e.g. simple or double boxes of various
dimensions and aspect ratios, k, different input motion characteris-
tics, etc.), given that the soil behaves in an elastic fashion and a
perfect bonding condition is considered between the soil and the
tunnel elements.
The consideration of a full-slip interface condition and the asso-
ciated geometrical nonlinearities along the interface (i.e. sliding
and/or separation), were found to affect the deformation modes
of the tunnel sections, causing in some cases a slight movement
of the rotation center from the tunnel section centroid. Fig. 6 illus-
 G. Tsinidis / Tunnelling and Underground Space Technology 62 (2017) 1–22 5

(a) ‘Pure’ racking

F = 0.2 F = 1.0 F = 2.0 F = 5.0 F = 10.0

(b)

F = 0.2 F = 1.0 F = 2.0 F = 5.0 F = 10.0

(c) ‘Pure’ racking

F = 0.2 F = 1.0 F = 2.0 F = 5.0 F = 10.0

Increasing racking distortion

Shaking motion towards left

Fig. 4. Dynamic deformed shapes of tunnels embedded in elastic soil (soil type B1), computed at the time step of maximum racking distortion of the sections for shaking
motion EQ2 (PGA = 0.35 g) and for full bonding between the soil and the tunnel elements, (a) 6
6 (m) tunnel, (b) 12
6 (m) tunnel and (c) 18
6 (m) tunnel (shaking
motion towards left, deformations scale
300).

F < 1.0 F = 1.0 F > 1.0

(a)

(b)

Fig. 5. Schematic representation of dynamic deformed shapes of tunnels in elastic soil, as affected by the lining flexibility (the deformed shapes correspond to a full bonding
condition between the soil and the tunnel elements), (a) motion towards right and (b) motion towards left.

trates dynamic deformed shapes of representative coupled soil-
tunnel configurations, as affected by the soil-tunnel interface prop-
erties. In particular, Fig. 6a portrays deformed shapes of 6
6 (m)
shallow tunnels, embedded in elastic soil (soil type B1), computed
for the shaking motion EQ2 and full-slip interface conditions, pre-
cluding the potential separation between the soil and the tunnel
elements. The deformed shapes are similar to those computed for
full bonding conditions (e.g. pure racking deformation for F = 1.0,
coupled racking-rocking response for F – 1.0). However, the rack-
ing distortion and the rotation of the tunnel sections are somehow
different compared to the predictions of the full bonding condition
analyses. These deviations are further discussed in the following
sections.
Fig. 6b and c illustrate dynamic deformed shapes of representa-
tive soil-tunnel configurations computed for full-slip interface con-
ditions, allowing in this case for the potential separation between
the soil and the tunnel elements. The deformation patterns are
different compared to the previous cases, especially for the rigid
tunnels (F < 1.0). In particular, due to the non-linear phenomena
that are taking place along the soil-tunnel interface (e.g. sliding
and/or separation), the racking deformations of the tunnels are
reduced compared to the full bonding condition analyses results.
Additionally, the above non-linear phenomena may result in an
inversion of the rotation of tunnel sections during shaking, com-
pared to the full bonding condition analyses (e.g.
Fig. 6c, F = 0.20). It is worth noticing the slight rocking response
of tunnels, sharing the same stiffness with the surrounding soil,
in cases where a full-slip interface condition is assumed
(Fig. 6c, F = 1.0). These observations regarding the rotational trends
of the tunnel sections are more evident for higher aspect ratios (e.g.
k > 2.0) and for reduced embedment (i.e. shallow tunnels). The
increase of the burial depth and the associated increase on the con-
fining pressures acting on the tunnel sections result in a reduction
of the sliding and separation phenomena along the soil-tunnel
interface, as theoretically expected. Hence, the deformation pat-
terns of deep tunnels, computed for full-slip interface conditions,
are closer to those computed for full bonding conditions at the
soil-tunnel interface.
Fig. 7 illustrates the effect of soil yielding on the deformed
shapes of 6
6 (m) tunnels embedded in diverse elasto-plastic soil
deposits (Table 2). In addition to the complex racking-rocking dis-
tortion mentioned above, inward deformations of the side-walls
and the slabs (deformations towards the cavity) are also identified,
as a result of the yielding of the surrounding ground. This observa-
6 G. Tsinidis / Tunnelling and Underground Space Technology 62 (2017) 1–22

(a)
‘Pure’ racking

F = 0.2 F = 1.0 F = 2.0 F = 5.0 F = 10.0

(b)

F = 0.2 F = 1.0 F = 2.0 F = 5.0 F = 10.0

(c)
F = 0.2 F = 1.0 F = 2.0 F = 5.0 F = 10.0

Separation

Increasing racking distortion

Shaking motion towards left

Fig. 6. Effect of soil-tunnel interface conditions on the dynamic deformed shapes of tunnels in elastic soil (soil type B1), as computed at the time step of maximum racking
distortion of the sections for the shaking motion EQ2 (PGA = 0.35 g), (a) 6
6 (m) tunnel - full-slip conditions, precluding the potential separation between the soil and the
tunnel elements, (b) 6
6 (m) tunnel - full-slip conditions, allowing for the potential separation between the soil and the tunnel elements and (c) 12
6 (m) tunnel - full-slip
conditions, allowing for the potential separation between the soil and the tunnel elements (shaking motion towards left, deformations scale
300).

tion is in line with recent findings of Cilingir and Madabhushi
(2011b), who derived, during dynamic centrifuge testing,
deformed shapes of flexible square model tunnels in dry sand,
using particle image velocimetry methods. Generally, the lining
inward deformations are more important for flexible tunnels (i.e.
F > 2.0). Additionally, the yielding of the surrounding ground
affects the rocking response of rectangular tunnels (e.g.
Fig. 7c, F = 2.0), while the rotation center may be shifted from the
section’s centroid. A rocking response may also be identified for
tunnel sections sharing the same stiffness with surrounding
ground (Fig. 7c and d, F = 1.0).
The lining rigidity, as well as the soil-tunnel interface condi-
tions, may affect the yielding of the surrounding ground (Huo
et al., 2005; Tsinidis et al., 2015, 2016b, 2016c). These effects are
verified in Fig. 8, where distributions of the ‘dynamic part’ of the
soil plastic strains, computed around 6
6 (m) tunnels for diverse
soil deposits and various soil-tunnel interface conditions, are dis-
played. The increase of the lining flexibility results in an increase
of the soil plastic strains around the tunnel. This observation veri-
fies the increased inward deformations of the slabs and side-walls
of flexible tunnels, mentioned above. With reference to the effect
of interface conditions; for full-slip interface conditions (Fig. 8d),
the computed soil plastic strains are concentrated near the corners
of the tunnel. On the contrary, the soil plastic strains computed
under full bonding conditions between the soil and the tunnel ele-
ments are distributed more uniformly around the slabs and the
side-walls.
It is reminded that the investigated soil-tunnel interface charac-
teristics constitute upper or lower limits for the actual interface
behavior. Hence, the presented deformation patterns should be
interpreted as limit cases for the actual deformation mechanisms.
Additionally, the soil separation (i.e. detachment) from the tunnel
lining is meaningless for cohesionless soils, while it may become
possible for cemented soils or undrained clays. In the idealized
cases studied herein, these phenomena exist, due to the simulation
of the soil as an elastic continuum.
4.2. Racking and R-F relations
The preliminary seismic analysis of rectangular tunnels is often
performed by implementing the simplified analysis method pro-
posed by Wang (1993). The method, referred hereafter as R-F
method, prescribes a simple static frame analysis for the evalua-
tion of the tunnel response, where the structural racking distortion,
Dstr, is modelled as an equivalent static load or pressure imposed
on the frame (Fig. 9). This distortion is evaluated by the free-field
ground racking distortion, Dff, which is properly adjusted, through
the so-called racking ratio, R, so as to account for the soil-tunnel
interaction effects, as follows:
R ¼ Dstr =Dff ! Dstr ¼ R
Dff ð3Þ
The racking ratio is usually correlated with the flexibility ratio.
Several analytical or empirical R-F relations may be found in the
literature (e.g. Wang, 1993; Penzien, 2000; Anderson et al.,
2008). Following this context, the results of this parametric study
were used for the development of numerical R-F relations. In par-
ticular, the racking ratios were evaluated at the time step of max-
imum racking distortion of the tunnel section, using the horizontal
deformations computed directly from the dynamic analyses, as dis-
played in Fig. 10. Hence, the effect of the rotation of the tunnel sec-
tion on the horizontal deformations (i.e. u2 in Fig. 10) and, thus, on
the computed racking ratios, R, was explicitly accounted for. The
presentation of the racking distortion of the tunnels through
dimensionless R-F relations facilitates the investigation of the
 G. Tsinidis / Tunnelling and Underground Space Technology 62 (2017) 1–22 7

(a)
‘Pure’ racking

F = 0.2 F = 1.0 F = 2.0 F = 5.0 F = 10.0

(b)

F = 0.2 F = 1.0 F = 2.0 F = 5.0 F = 10.0

(c)

F = 0.2 F = 1.0 F = 2.0 F = 5.0 F = 10.0

(d)

F = 0.2 F = 1.0 F = 2.0 F = 5.0 F = 10.0

Separation

Increasing racking distortion

Shaking motion towards left

Fig. 7. Dynamic deformed shapes of 6
6 (m) square tunnels embedded in diverse elasto-plastic soil deposits, computed for shaking motion EQ2 (0.35 g) at the time step of
maximum racking distortion, (a) soil type B1, full bonding condition at the interface, h = 3 m, (b) soil type C1, full bonding condition at the interface, h = 3 m, (c) soil type C2,
full bonding condition at the interface, h = 3 m and (d) soil type B1, full slip condition allowing for the potential separation between the soil and the tunnel elements, h = 3 m
(motion towards left, deformations scale
300).

effect of salient parameters on the racking response of these types
of structures, while it allows for direct comparisons with the exist-
ing literature.
Fig. 11 illustrates numerical R-F relations computed for diverse
tunnel sections embedded in elastic soil (soil type B1). The results
highlight the effects of critical parameters, namely (i) the aspect
ratio, k, and the burial depth, h, of the tunnel, (ii) the soil deposit
properties, (iii) the soil-tunnel interface characteristics, and (iv)
the input motion characteristics, on the computed R-F relations.
In particular, Fig. 11a presents R-F relations computed for a full
bonding condition between the soil and the tunnel elements.
Fig. 11b portrays R-F relations computed for full-slip interface con-
ditions, precluding potential separation, while in Fig. 11c, R-F rela-
tions, computed for full-slip interface conditions, allowing for
potential separation, are illustrated. Naturally, the racking ratio,
R, increases with the increase of the flexibility ratio, F (i.e.
increased racking deformations for flexible tunnels). Additionally,
the computed racking ratios are increased with the decrease of
the tunnel embedment (i.e. higher racking ratios for tunnel burial
depth h = 3 m) and the increase of the aspect ratio, k. The racking
ratios computed for diverse soil deposits (e.g. soil classes B or C)
and input motions (scaled up to the same PGA = 0.35 g), assuming
a full bonding condition at the interface, are almost identical.
Hence, the effects of the above parameters on the computed rela-
tions are negligible. This is actually expected due to the elastic
response of the idealized soil-tunnel configurations. It is worth
noticing that if the strain compatible soil properties (i.e. soil equiv-
alent properties) were encountered in the analyses, the computed
racking ratios would have been affected by the input motion char-
acteristics (due to the effect of the input motion on the soil equiv-
alent properties). In addition, the flexibility ratios would have been
different for each input motion case, as they depend on the soil
stiffness (Eq. (1)). Hence, a range of flexibility ratios, as well as a
range of racking ratios, would have been computed for each soil-
tunnel system, considering different input motions (Hashash
et al., 2010). The shear stiffness of the soil was purposely kept con-
stant for the various shaking motions in the herein study, in order
to investigate the effects of other critical parameters on the racking
response of the tunnels.
The consideration of the frictionless soil-tunnel interface and
the associated non-linear phenomena (e.g. sliding and/or separa-
tion), amplify the effects of the soil properties and input motion
characteristics on the racking response of the tunnels. This is evi-
dent in Fig. 11b and c, where the values of the racking ratios, com-
puted for the same flexibility ratio and various assumptions
regarding the above properties (i.e. soil properties and input
motion characteristics), are more scattered. Racking ratios are
more scattered for the cases, where separation between the soil
8 G. Tsinidis / Tunnelling and Underground Space Technology 62 (2017) 1–22

(a)

F = 0.2 F = 1.0 F = 2.0 F = 5.0 F = 10.0

PE, plastic strain

0.0006 0.00045 0.0003 0.00015 0.000

(b)

F = 0.2 F = 1.0 F = 2.0 F = 5.0 F = 10.0

PE, plastic strain

0.0019 0.00142 0.00095 0.00047 0.000

(c)

F = 0.2 F = 1.0 F = 2.0 F = 5.0 F = 10.0

PE, plastic strain

0.0023 0.00172 0.00115 0.000575 0.000

(d)

F = 0.2 F = 1.0 F = 2.0 F = 5.0 F = 10.0

PE, plastic strain

0.001 0.00075 0.0005 0.00025 0.000

Fig. 8. Soil plastic strain distributions (dynamic part) around 6
6 (m) square tunnels, embedded in diverse elasto-plastic soil deposits, computed at the time step of
maximum racking distortion (EQ2, 0.35 g), (a) soil type B1, full bonding condition at the interface, h = 3 m, (b) soil type C1, full bonding condition at the interface, h = 3 m, (c)
soil type C1, full bonding condition at the interface, h = 12 m and (d) soil type B1, full slip conditions allowing for the potential separation between the soil and the tunnel
elements, h = 3 m.

and the tunnel elements is permitted, while the deviation is higher
for tunnels with larger aspect ratios (i.e. k > 2.0).
Fig. 11d compares ‘average’ R-F relations (blue1 dashed lines in
Fig. 11a–c) computed for diverse soil-tunnel configurations, assum-
1
For interpretation of color in Figs. 11, 16 and 21, the reader is referred to the web
version of this article.
ing various soil-tunnel interface characteristics. Clearly, a full bond-
ing (i.e. perfect bonding) condition between the soil and the tunnel
elements results in higher racking ratios compared to a full-slip
interface condition. This comes in accordance with recent findings
of Tsinidis et al. (2015, 2016a, 2016b). Actually, the consideration
of a full-slip interface allows the tunnel section to rotate more easily
within the soil, reducing its racking distortion. In line with recent
 G. Tsinidis / Tunnelling and Underground Space Technology 62 (2017) 1–22 9

(a) structure =R× free-field

(b) structure =R× free-field
P P

Concentrated Pseudo-
Force triangular
distribution

Fig. 9. Simplified frame analysis models, (a) concentrated force, (b) triangular distribution (modified after Wang (1993)).

u1 u2 u 3
A C

uC uD Soil
R st
‘free field’
ff u A uB
B D u2
Tunnel
Fig. 10. Computation of racking ratio, R, and effect of rotation of the tunnel section on the horizontal deformations computed by the numerical analysis; u1: rigid body
movement of the tunnel section, u2: deformation due to the rotation of the tunnel section, u3: deformation due to the racking distortion of the tunnel section.

findings of Debiasi et al. (2013), the effect of interface conditions is
more evident for shallow tunnels (e.g. tunnels with h/a < 1.0). Theo-
retically, an increase in the burial depth (increase of the depth ratio,
h/a) results in an increase of the confining pressure around the tun-
nel, hence leading to a more ‘rigid’ interface.
Given that the perfect bonding between the soil and the tunnel
elements results in higher racking ratios, R (i.e. conservative design
concept), the following discussion is focused on the effects of crit-
ical parameters (e.g. existence of an internal column, aspect ratio,
k, and burial depth, h, of the tunnel section, input motion ampli-
tude etc) on the R-F relations computed for this particular interface
condition.
Fig. 12a compares average R-F relations computed for 12
6
(m) single and double box tunnels, highlighting the effect of the
internal column on these relations. The thickness of the lining is
properly adjusted in each case, so as to model the desirable flexibil-
ity ratio, F. In this context, the lining thickness of the double sec-
tion is smaller compared to that of the one-barrel tunnel. The
comparisons indicate a perfect much between the two cases.
Therefore, the R-F relations computed herein may be used for both
single and double box tunnels, given that the flexibility of the tun-
nel section (expressed by the flexibility ratio, F) is properly esti-
mated (e.g. through a static frame analysis, as proposed by Wang
(1993)).
Fig. 12b illustrates the effect of the size of the tunnel section on
the R-F relations computed for a full bonding interface condition.
The results are in good agreement with each other, for the majority
of the cases. A slight difference (7–9%) between the computed
racking ratios of the shallow rigid tunnels (F = 0.20) is only
observed. This allows for further generalization of the relations
based on the aspect ratio, k.
The effect of the input motion amplitude on the numerical R-F
relations, computed for full bonding between the soil and the tun-
nel elements, is illustrated in Fig. 12c. The results refer to Friuli
(EQ2) and Takatori (EQ3) records scaled up to diverse amplitudes.
For the perfect bonding interface condition and the elastic soil
response examined herein, the computed R-F relations are in very
good agreement. There are some minor differences for the shallow
and flexible tunnels (e.g. h = 3 m, F > 5.0), which in any case do not
exceed 5–6%. In this context, the numerical R-F relations may be
generalized for different input motion amplitudes. This observa-
tion comes in accordance with recent findings of Debiasi et al.
(2013).
Fig. 13 illustrates the effect of the aspect ratio, k, on the numer-
ical R-F relations. The results, which are plotted for both the exam-
ined burial depths, are also compared with available analytical and
empirical R-F relations (Wang, 1993; Penzien, 2000; Anderson
et al., 2008). For flexible tunnels (F > 1.0), the racking ratio, R, is
increasing with the increase of the aspect ratio, k. The exact oppo-
site trend is observed for rigid tunnels (F < 1.0). As mentioned
above, racking ratio is decreasing with the increase of the tunnel’s
burial depth. All the available relations underestimate the racking
response of rigid tunnels compared to the numerical results, with
the underestimation being generally higher for lower aspect ratios.
With reference to the flexible tunnels, the NCHPR611 relation
(Anderson et al., 2008) either over-predicts or under-predicts the
racking response compared to the numerical results. These finding
comes in accordance with recent results of Hashash et al. (2010)
and Debiasi et al. (2013). Similar conclusions are drawn by com-
paring the numerical results with Wang’s (1993) results. Finally,
Penzien’s solution (Penzien, 2000), referring to no-slip interface
conditions, seems to provide a maximum envelope for the racking
ratios.
As mentioned above, a sensitivity analysis was conducted, aim-
ing to investigate the effects of soil damping and Poison ratio, as
well as soil yielding, on the racking response of rectangular tun-
nels. The effect of soil Poisson ratio on the numerical R-F relations
is illustrated in Fig. 14a. The relations are computed for diverse
soil-tunnel configurations and soil Poisson ratios, ranging between
0.1 and 0.45. In line with the existing literature (Penzien, 2000;
Debiasi et al., 2013), the increase of the soil Poisson ratio leads to
a decrease of the racking ratio, R.
Fig. 14b compares R-F relations computed for square tunnels
embedded in soil with varying damping ratio (i.e. D = 0, 5, 20%).
The comparisons indicate a minor effect for a wide range of flexi-
bility ratios. The deviations between the racking ratios, computed
for diverse damping ratios, are more evident for flexible tunnels
(F > 5.0).
The effect of the yielding of the surrounding ground on the
racking ratios of rectangular tunnels is highlighted in Fig. 14c,
10 G. Tsinidis / Tunnelling and Underground Space Technology 62 (2017) 1–22

(a) 6x6 (m), h=3m 12x6 (m), h=3m (single box) 18x6 (m), h=3m (double box) 18x6 (m), h=12m (double box)
2.8 2.8 2.8 2.8

2.1 2.1 2.1 2.1

1.4 1.4 1.4 1.4

R

R
0.7 0.7 0.7 0.7

0 0 0 0
0 2.5 5 7.5 10 0 2.5 5 7.5 10 0 2.5 5 7.5 10 0 2.5 5 7.5 10
F F F F

(b) 6x6 (m), h=3m 12x6 (m), h=3m (single box) 18x6 (m), h=3m (double box) 18x6 (m), h=12m (double box)
2.8 2.8 2.8 2.8

2.1 2.1 2.1 2.1

1.4 1.4 1.4 1.4

R

R
0.7 0.7 0.7 0.7

0 0 0 0
0 2.5 5 7.5 10 0 2.5 5 7.5 10 0 2.5 5 7.5 10 0 2.5 5 7.5 10
F F F F

6x6 (m), h=3m 12x6 (m), h=3m (single box) 18x6 (m), h=3m (double box) 18x6 (m), h=12m (double box)
(c) 2.8 2.8 2.8 2.8

2.1 2.1 2.1 2.1

1.4 1.4 1.4 1.4

R

R
0.7 0.7 0.7 0.7

0 0 0 0
0 2.5 5 7.5 10 0 2.5 5 7.5 10 0 2.5 5 7.5 10 0 2.5 5 7.5 10
F F F F
EQ1, Soil B1 EQ2, Soil B1 EQ3, Soil B1
EQ1, Soil C1 EQ2, Soil C1 EQ3, Soil C1 Average

6x6 (m), h=3m 12x6 (m), h=3m (single box) 18x6 (m), h=3m (double box) 18x6 (m), h=12m (double box)
(d) 3 3 3 3

2.25 2.25 2.25 2.25

1.5 1.5 1.5 1.5

R

0.75 0.75 0.75 0.75

0 0 0 0
0 2.5 5 7.5 10 0 2.5 5 7.5 10 0 2.5 5 7.5 10 0 2.5 5 7.5 10
F F F F
Interface (μ=0, separation) Interface (μ=0, no separation) Perfect bonding (μ=∞, no separation)

Fig. 11. R-F relations computed for diverse soil-tunnel configurations, assuming (a) a perfect bonding condition at the soil-tunnel interface, (b) full-slip interface conditions
precluding potential separation between the soil and the tunnel elements, (c) full-slip interface conditions, allowing for the potential separation between the soil and the
tunnel elements, (d) comparison of average F-R relations (PGA = 0.35 g).

where R-F relations, computed by elasto-plastic analyses, are plot- shaking motion characteristics on the soil yielding response and
ted. Recent experimental and numerical studies (e.g. Lanzano et al., thus on the computed R-F relations (e.g. EQ2, EQ3 in soil type C2).
2012; Tsinidis et al., 2015, 2016a, 2016b, 2016c; Cilingir and
Madabhushi, 2011a, 2011b, 2011c), have demonstrated that the 4.3. Rocking and h/cff-F relations
yielding of the surrounding ground results in stress redistributions
within the soil that may cause a permanent response for the soil- To quantify the rocking response of rectangular tunnels, mobi-
tunnel configuration. The R-F relations are found to be affected sig- lized during ground shaking, an average rotation of the tunnel sec-
nificantly by this permanent response. Naturally, the effect is more tion, h, was computed for all the examined cases (Fig. 15), as
evident for the soil deposits with reduced strength characteristics follows:
(e.g. soil type C2 in Fig. 14c), due to the associated increasing yield- 

h ¼ max 2 ðtÞ  u2 ðtÞ =a
ulw rw
ð4Þ
ing phenomena in these cases. It is worth noticing the effect of the
 G. Tsinidis / Tunnelling and Underground Space Technology 62 (2017) 1–22 11

12x6 (m), h=3m Burial depth h=3 m Burial depth h=12 m

(a) 3
(b) 2 2

2.25 1.5 1.5

R 1.5 1 1

R
0.75 0.5 0.5

0 0 0
0 2.5 5 7.5 10 0 2.5 5 7.5 10 0 2.5 5 7.5 10
F F F
Single box Double box 2 x 2 (m) 6 x 6 (m) 10 x 10 (m)

6x6 (m), h=3m 12x6 (m), h=3m (single box) 18x6 (m), h=3m (double box)
(c) 2.8 2.8 2.8

2.1 2.1 2.1

1.4 1.4 1.4

R

R
0.7 0.7 0.7

0 0 0
0 2.5 5 7.5 10 0 2.5 5 7.5 10 0 2.5 5 7.5 10
F F F
EQ2 − 0.1g EQ2 − 0.7g EQ2 − 0.35g EQ3 − 0.35g EQ3 − 0.6g

Fig. 12. R-F relations computed for a perfect bonding interface condition and elastic soil response, (a) effect of the central column, (b) effect of the size of the tunnel section
and (c) effect of input motion amplitude.

Burial depth h=3 m Burial depth h=12 m

2.6 2.6

1.95 1.95

1.3 1.3
R

0.65 4 1 vs F 0.65
R R 2F 1 F
3 4vs F
0 0
0 2.5 5 7.5 10 0 2.5 5 7.5 10
F F

λ=0.5 λ=1.0 λ=1.5 λ=2.0 λ=3.0

Wang (1993) Penzien (2000), No slip Anderson et al. (2008)

Fig. 13. Numerical R-F relations, computed for various aspect ratios, k, plotted against existing literature.

2 ðtÞ; u2 ðtÞ are the vertical displacement time histories of the side-
ulw rw
walls, computed by the dynamic analysis, and a is the width of the
tunnel. Eq. (4) implies that the tunnel rotates around the geometri-
cal centroid of the section. As mentioned above, this assumption is
valid for elastic response of the soil-tunnel configuration, while its
validity decreases when non-linearities of any type are encountered
(e.g. geometrical non-linearities at the soil-tunnel interface, materi-
als non-linearity). Indeed, for the cases, where the non-linearities
are accounted for, the center of rotation of the tunnel section does
not necessarily coincide with its centroid. However, the two points
are still very close to assume that the effect on the computed rota-
tion, h, is reduced. Hence, the above hypothesis (i.e. rotation around
the centroid of the section) seems to be fair enough.
The above rotations were normalized by ‘average’ ground shear
strains ðcff Þ, computed at ‘free-field’ conditions and at the same
depth with the tunnel. This was performed, so as to be possible
to compare the rotational response of the tunnels, as computed
for different soil properties and input motion characteristics. The
above results were then used to plot h/cff-F relations, similar to
R-F relations. This section summarizes representative h/cff-F rela-
tions computed for diverse soil-tunnel configurations, revealing
the effects of critical parameters on the rocking response of rectan-
gular tunnels. For the sake of comparison, the sign of the computed
rotations is properly adjusted, as discussed in the following
paragraphs.
Fig. 16 portrays h/cff-F relations computed for diverse soil-
tunnel systems and various interface conditions. Fig. 16a plots
h/cff-F relations computed for a full bonding interface condition;
Fig. 16b portrays h/cff-F relations computed for full-slip interface
conditions precluding the potential separation between the soil
12 G. Tsinidis / Tunnelling and Underground Space Technology 62 (2017) 1–22

(a) λ=1, h=3m λ=1, h=12m λ=2, h=3m

2.6 2.6 2.6

1.95 1.95 1.95

1.3 1.3 1.3

R

R
0.65 0.65 0.65

0 0 0
0 2.5 5 7.5 10 0 2.5 5 7.5 10 0 2.5 5 7.5 10
F F F
v = 0.1 v = 0.3 v = 0.45

(b) λ=1, h=3m (c) Soil type B Soil type C

1.8 2.6 2.6

1.35 1.95 1.95

1.3 1.3
R

R
0.9
R

0.45 0.65 0.65

0 0 0
0 2.5 5 7.5 10 0 2.5 5 7.5 10 0 2.5 5 7.5 10
F F F
Elastic soil . Elastic soil
D=0%
Elasto−plastic, Soil B1 (EQ2) Elasto−plastic, Soil C1 (EQ2)
D=5% Elasto−plastic, Soil C2 (EQ2)
Elasto−plastic, Soil B2 (EQ2)
D = 20 % Elasto−plastic, Soil C1 (EQ3)
Elasto−plastic, Soil C2 (EQ3)

Fig. 14. Effects of (a) soil Poisson ratio, (b) soil damping ratio and (c) soil yielding on the R-F relations computed for a full bonding condition at the soil-tunnel interface.

u2lw t u2rw t
Fig. 15. Definition of tunnel section rotation, h.
and the tunnel elements, while in Fig. 16c relations computed for
full-slip interface conditions allowing for potential separation are
presented. The physical meaning of negative h/cff values in the
graphs is that the tunnels in these cases are rotating in the opposite
direction compared to tunnels with the positive h/cff values (set
clockwise). Generally, the computed rotation is higher for shallow
tunnels (i.e. h = 3 m). In other words, the rocking response of rect-
angular tunnels is decreasing with increasing embedment. Addi-
tionally, the rotation decreases with the increase of the tunnel
aspect ratio, k, as theoretically expected.
The h/cff ratios computed for diverse soil conditions and input
motion characteristics, assuming full bonding interface conditions,
are less scattered for the rigid or relatively rigid tunnels (i.e.
F < 2.0). On the contrary, the results are more scattered for flexible
tunnels (i.e. F > 5.0). This is attributed to the increased lining curva-
ture of the flexible tunnels during shaking. Actually, a flexible tun-
nel is not rotating as a rigid body in the soil. It is worth noticing
that for F = 1.0 no rocking response is observed for the tunnel sec-
tions, thus h/cff = 0.
The consideration of a frictionless interface between the soil
and the tunnel elements results in an increase of the rotation of
the tunnel (i.e. higher h/cff ratios compared to the results of the
perfect bonding interface analyses, Fig. 16c). The h/cff ratios, com-
puted for the same flexibility ratio and for diverse soil types and
input motion characteristics, are more scattered compared to the
results of the full bonding condition analyses. This observation is
more evident for the flexible tunnels with larger aspect ratios,
k > 2.0, while it becomes more important for the cases, where sep-
aration between the two media is permitted (Fig. 16c). As men-
tioned in Section 4.1, the consideration of a full-slip interface
may result in an inversion of the rotation of the tunnel section
compared to the one predicted by a perfect bonding condition
analysis. This observation, refers mainly to rigid or relatively rigid
tunnels (i.e. F < 1.0 or 1.0) with aspect ratios, k > 2.0. Additionally,
tunnels, which share the same stiffness with the surrounding
ground (F = 1.0), exhibit a complex racking-rocking response (i.e.
h/cff – 0) and not a pure racking response (i.e. h/cff = 0), as in the
case of a full bonding condition analysis (Fig. 16b and c).
Fig. 16d highlights the effect of the soil-tunnel interface charac-
teristics on the rocking response of the tunnels, by plotting ‘aver-
age’ h/cff-F relations, computed for diverse soil-tunnel
configurations (blue lines in Fig. 16a–c). The comparisons verify
the higher rocking response of rectangular tunnels, when a full-
slip interface condition is considered.
The effects of salient parameters (e.g. internal column, aspect
ratio, k, and burial depth, h, of the tunnel section, input motion
characteristics etc) on the h/cff-F relations computed for perfect
bonding between the soil and the tunnel elements are highlighted
and discussed in the ensuing.
Fig. 17a compares h/cff-F relations computed for 12
6 (m) sin-
gle and double box tunnels, highlighting the effect of the internal
column on these relations. Similar to the R-F relations, the results
are well compared. Hence the relations computed herein may be
used for both single and double box tunnels, given that the flexibil-
ity of the tunnel is properly evaluated.
 G. Tsinidis / Tunnelling and Underground Space Technology 62 (2017) 1–22 13

(a) 6x6 (m), h=3m 12x6 (m), h=3m (single box) 18x6 (m), h=3m (double box) 12x6 (m), h=12m (double box)
0.9 0.9 0.9 0.9

0.45 0.45 0.45 0.45

θ/γff

θ/γff

θ/γff

θ/γff
0 0 0 0

−0.45 −0.45 −0.45 −0.45

0 2.5 5 7.5 10 0 2.5 5 7.5 10 0 2.5 5 7.5 10 0 2.5 5 7.5 10
F F F F

6x6 (m), h=3m 12x6 (m), h=3m (single box) 18x6 (m), h=3m (double box) 12x6 (m), h=12m (double box)
(b) 1.5 1.5 1.5 1.5

0.75 0.75 0.75 0.75

θ/γff

θ/γff

θ/γff

θ/γff
0 0 0 0

−0.75 −0.75 −0.75 −0.75

0 2.5 5 7.5 10 0 2.5 5 7.5 10 0 2.5 5 7.5 10 0 2.5 5 7.5 10
F F F F

6x6 (m), h=3m 12x6 (m), h=3m (single box) 18x6 (m), h=3m (double box) 12x6 (m), h=12m (double box)
(c) 1.5 1.5 1.5 1.5

0.75 0.75 0.75 0.75

θ/γff

θ/γff

θ/γff

θ/γff
0 0 0 0

−0.75 −0.75 −0.75 −0.75

0 2.5 5 7.5 10 0 2.5 5 7.5 10 0 2.5 5 7.5 10 0 2.5 5 7.5 10
F F F F

EQ1, Soil B1 EQ2, Soil B1 EQ3, Soil B1

EQ1, Soil C1 EQ2, Soil C1 EQ3, Soil C1 Average

(d) 6x6 (m), h=3m 12x6 (m), h=3m (single box) 18x6 (m), h=3m (double box) 18x6 (m), h=12m (double box)
1.26 1.26 1.26 1.26

0.63 0.63 0.63 0.63

θ/γff

θ/γff

θ/γff

θ/γff

0 0 0 0

−0.63 −0.63 −0.63 −0.63

0 2.5 5 7.5 10 0 2.5 5 7.5 10 0 2.5 5 7.5 10 0 2.5 5 7.5 10
F F F F
Interface (μ=0, separation) Interface (μ=0, no separation) Perfect bonding (μ=∞, no separation)

Fig. 16. h/cff-F relations computed for diverse soil-tunnel configurations, assuming (a) a full bonding condition at the interface, (b) full-slip interface conditions precluding
potential separation between the soil and the tunnel elements, (c) full-slip interface conditions, allowing for potential separation between the soil and the tunnel elements
and (d) comparison of average h/cff-F relations (PGA = 0.35 g).

Fig. 17b compares h/cff-F computed for square tunnels of vari-
ous sizes embedded in soil type C1, assuming a perfect bonding
condition at the soil-tunnel interface. The increase of the size of
the tunnel section results in an increase of its rotation. This obser-
vation is generally more evident for shallow flexible tunnels (e.g.
F > 5.0) embedded in looser soil deposits. Indeed, the deviations
are considerably reduced for higher embedment of the tunnel
section.
Fig. 17c illustrates the effect of the input motion amplitude on
the h/cff-F relations computed for diverse soil-tunnel configura-
tions. The presented results, which refer to the Friuli (EQ2) and
Takatori (EQ3) records scaled to diverse amplitudes, are in rela-
tively good agreement.
Fig. 18 plots h/cff-F relations computed for various aspect ratios,
k, and both burial depths (h = 3, 12 m), assuming a perfect bonding
interface condition. Regardless of the flexibility of the tunnel, the
rotation increases with the decrease of the aspect ratio, k. Once
again, it is observed that shallow tunnels exhibit a higher rocking
response.
Similar to the racking ratios, the h/cff-F ratios decrease with
increasing soil’s Poisson ratio and increasing damping of the soil
deposit (Fig. 19a and b). These observations are more important
for flexible tunnels (i.e. F > 5.0) Moreover, the effect of soil
yielding on the computed h/cff-F ratios is significant, especially
in cases of soil deposits with low strength properties
(Fig. 19c).
14 G. Tsinidis / Tunnelling and Underground Space Technology 62 (2017) 1–22

12x6 (m), h=3m Burial depth h=3 m Burial depth h=12 m

(a) 1.2 (b) 1.4 1.4

0.6 0.7 0.7

θ/γff

θ/γff
θ/γff

0 0 0

−0.6 −0.7 −0.7

0 2.5 5 7.5 10 0 2.5 5 7.5 10 0 2.5 5 7.5 10
F F F
Single box Double box 2 x 2 (m) 6 x 6 (m) 10 x 10 (m)

6x6 (m), h=3m 12x6 (m), h=3m (single box) 12x6 (m), h=12m (single box)
(c) 1 1 1

0.5 0.5 0.5

θ/γff

θ/γff

θ/γff
0 0 0

−0.5 −0.5 −0.5

0 2.5 5 7.5 10 0 2.5 5 7.5 10 0 2.5 5 7.5 10
F F F
EQ2 − 0.1g EQ2 − 0.7g EQ2 − 0.35g EQ3 − 0.35g EQ3 − 0.6g

Fig. 17. h/cff-F relations computed for a perfect bonding condition and elastic soil response, (a) effect of the central column, (b) effect of the size of the tunnel section and (c)
effect of input motion amplitude.

Burial depth h=3 m Burial depth h=12 m

1.2 1.2

0.6 0.6
θ/γff

θ/γff

0 0

−0.6 −0.6
0 2.5 5 7.5 10 0 2.5 5 7.5 10
F F
λ=0.5 λ=1.0 λ=1.5 λ=2.0 λ=3.0

Fig. 18. Numerical h/cff-F relations computed for various aspect ratios, k.

4.4. Dynamic earth pressures
This section summarizes the key findings of this study, regard-
ing the dynamic earth pressures developed around rectangular
tunnels during ground shaking. The effects of critical parameters,
namely the soil-tunnel relative flexibility and interface characteris-
tics, the aspect ratio, k, and burial depth, h, of the tunnel, the input
motion characteristics, and the soil properties and yielding
response, on the dynamic earth pressures magnitudes and distri-
butions, are highlighted and discussed. It is noted that the follow-
ing results refer to the dynamic part of the earth pressures (i.e.
‘dynamic’ earth pressures). The ‘dynamic’ earth pressures should
always be added to the geostatic pressures, to evaluate the total
earth pressures developed around the tunnel.
Fig. 20a portrays representative dynamic earth pressures distri-
butions computed along the perimeter of diverse tunnels in elastic
soil (soil type B1). The results, which are computed for shaking
motion EQ2 (PGA = 0.35 g), assuming a full bonding interface con-
dition, refer to the time step of maximum racking distortion of the
tunnel sections. In line with recent findings of Tsinidis et al.
(2016a, 2016b), the dynamic earth pressures are distributed in
an anti-symmetric fashion along the slabs and the side-walls, even
for the cases of double boxes (e.g. 12
6 m), while the magnitudes
rise with the increase of tunnel’s burial depth. The slight decrease
of the earth pressures near the corners of the tunnel sections
should be attributed to the local curvature of the linings at these
locations, caused to maintain the right angle of the concrete cor-
ners. These curvatures are amplified for flexible tunnels, due to
the higher flexibility of the lining in these cases. The observed
anti-symmetric response indicates that, at a specific time incre-
ment, half length of the slabs and the side-walls is subjected to
active earth pressures, while the other half is subjected to passive
earth pressures. This observation is consistent with the rotation of
the tunnel section around its centroid. It is worth noticing the dif-
ference in the sign of the pressures distributions computed around
rigid (e.g. F < 0.2) and flexible tunnels (F > 1.0) for the same input
 G. Tsinidis / Tunnelling and Underground Space Technology 62 (2017) 1–22 15

λ=1, h=3m λ=1, h=12m λ=2, h=3m

(a) 1.5 1.5 1.5

0.75 0.75 0.75

θ/γff

θ/γff

θ/γff
0 0 0

−0.75 −0.75 −0.75

0 2.5 5 7.5 10 0 2.5 5 7.5 10 0 2.5 5 7.5 10
F F F
v = 0.1 v = 0.3 v = 0.45

(b) λ=1.0, h=3m (c) Soil type B Soil type C

1.2 2.6 2.6

1.3 1.3
0.6
θ/γff

θ/γff
θ/γff

0 0
0
−1.3 −1.3

−0.6 −2.6 −2.6

0 2.5 5 7.5 10 0 2.5 5 7.5 10 0 2.5 5 7.5 10
F F F
Elastic soil . Elastic soil
D=0%
Elasto−plastic, Soil B1 (EQ2) Elasto−plastic, Soil C1 (EQ2)
D=5% Elasto−plastic, Soil C2 (EQ2)
Elasto−plastic, Soil B2 (EQ2)
D = 20 % Elasto−plastic, Soil C1 (EQ3)
Elasto−plastic, Soil C2 (EQ3)

Fig. 19. Effects of (a) soil Poisson ratio, (b) soil damping ratio and (c) soil yielding on the h/cff-F relations computed for a full bonding condition at the soil-tunnel interface.

motion and time increment. This observation is consistent for all
the investigated cases and is attributed to the effect of the lining
rigidity on the rotation mechanism of the tunnel sections, men-
tioned above (e.g. Fig. 5). The shape of the earth pressures distribu-
tions is not affected by the size of the tunnel. The magnitude, on
the other hand, increases with increasing size of the tunnel section.
Generally, the dynamic earth pressures developed along the side-
walls are higher compared to those developed along the slabs.
Additionally, the dynamic earth pressures developed along the roof
slab are reduced compared to those acting on the invert slab (i.e.
foundation slab). This observation is more evident for shallow tun-
nels (e.g. h = 3 m), as the overburden soil weight and the associated
earth pressures on the roof slab, are highly reduced in these cases.
The deviations between the dynamic earth pressures computed
along the slabs and the side-walls increase for increasing aspect
ratios, k.
The soil-tunnel relative flexibility (expressed by the flexibility
ratio in Fig. 20a) seems to play an important role in the develop-
ment of the dynamic earth pressures. Indeed, the dynamic earth
pressures are highly reduced for the cases, where the tunnel shares
the same stiffness with the surrounding ground (i.e. F = 1.0). Addi-
tionally, higher dynamic earth pressures are developed on flexible
tunnels (e.g. on the corners) compared to the rigid ones. This seem-
ingly irregular behavior is explained as follows: for a given soil-
tunnel configuration, the increase of the lining flexibility results
to an increase of its racking distortion, which in addition to the
rocking response, leads to higher earth pressures on the corners
of the slabs and the side-walls (i.e. higher resistance of the sur-
rounding ground to the increased lining deformations). In other
words, the dynamic earth pressures near the corners reach the
‘passive state’. Similar conclusions were drawn by Tsinidis et al.
(2016a).
The internal column in cases of double boxes ‘interrupts’ the
dynamic earth pressures distributions along the roof and the
inverted slabs. This is clearly observed by comparing the dynamic
earth pressures distributions computed along the slabs of 12
6
(m) tunnels (single and double boxes in Fig 20a). However, the
distributions predicted around double boxes still remain
anti-symmetric, while the dynamic earth pressures magnitudes
computed near the corners are close to those computed for single
boxes at the same locations. The effect of the internal column on
the distributions is generally more important for flexible tunnels
(F > 1.0), while it becomes negligible for rigid tunnels (e.g.
F = 0.20), due to the increased rigidity of the lining in the latter
cases.
For the elastic soil response and the full bonding interface
condition discussed herein, the effects of soil properties and
input motion characteristics on the distributions and the
magnitudes of the dynamic earth pressures are negligible. As
discussed in the ensuing, these parameters become more
important when geometrical and/or material non-linearities are
encountered.
Fig. 20b illustrates representative dynamic earth pressures dis-
tributions, computed around diverse tunnel sections in elastic soil
(soil type B1), for full-slip interface conditions, precluding the
potential separation between the soil and the tunnel elements.
Although, the anti-symmetric response still exists, the distribu-
tions are somehow more complex compared to those computed
for full bonding interface conditions. In particular, higher dynamic
earth pressures are ‘concentrated’ near the corners of the tunnels.
Actually, the dynamic earth pressures magnitudes computed at
these locations may be as 50% higher compared to those estimated
for full bonding interface conditions. The effects of the aspect ratio,
k, burial depth, h, and soil-tunnel relative flexibility, are similar to
16 G. Tsinidis / Tunnelling and Underground Space Technology 62 (2017) 1–22

(a) 50 2x2 (m), h=3m

150
3x6 (m), h=3m
150
6x6 (m), h=3m
300
6x6 (m), h=12m

25 75 75 150
σ (kPa/m)

σ (kPa/m)

σ (kPa/m)
σ (kPa/m)
0 0 0 0

−25 −75 −75 −150

−50 −150 −150 −300

A B C D A A B C D A A B C D A A B C D A
12x6 (m), h=3m (single box) 12x6 (m), h=3m (double box) 18x6 (m), h=3m (double box) 18x6 (m), h=12m (double box)
150 150 150 300

75 75 75 150
σ (kPa/m)

σ (kPa/m)
σ (kPa/m)

σ (kPa/m)
0 0 0 0

−75 −75 −75 −150

−150 −150 −150 −300

A B C D A A E B C F D A A E B C F D A A E B C F D A

(b) 150 6x6 (m), h=3m 12x6 (m), h=3m (single box)
300
18x6 (m), h=3m (double box)
300
18x6 (m), h=12m (double box)
400

75 150 150 200

σ (kPa/m)

σ (kPa/m)
σ (kPa/m)

σ (kPa/m)

0 0 0 0

−75 −150 −150 −200

−150 −300 −300 −400

A B C D A A B C D A A E B C F D A A E B C F D A

(c) 150
6x6 (m), h=3m 12x6 (m), h=3m (single box)
150
18x6 (m), h=3m (double box)
300
18x6 (m), h=12m (double box)
400

75 75 150 200
σ (kPa/m)

σ (kPa/m)
σ (kPa/m)

σ (kPa/m)

0 0 0 0

−75 −75 −150 −200

−150 −150 −300 −400

A B C D A A B C D A A E B C F D A A E B C F D A

(d) 80 Soil B1
80
Soil B2
80
Soil C1
80
Soil C2

40 40 40 40
σ (kPa/m)

σ (kPa/m)

σ (kPa/m)

σ (kPa/m)

0 0 0 0

−40 −40 −40 −40

−80 −80 −80 −80

A B C D A A B C D A A B C D A A B C D A
D C D F C
F=0.2
F=1.0
F=2.0
F=5.0
F=10.0
A B A E B
Fig. 20. (a–c) Distributions of dynamic earth pressures developed along the perimeter of diverse tunnel sections at the time step of maximum racking distortion, (a) elastic
soil (soil type B1), full bonding conditions at the soil-tunnel interface, (b) elastic soil (soil type B1), full-slip conditions precluding potential separation between soil and tunnel
elements, (c) elastic soil (soil type B1), full-slip conditions permitting potential separation between soil and tunnel elements and (d) Dynamic earth pressure distributions
developed along the perimeter of 6
6 (m) shallow tunnels (h = 3 m), as computed for diverse elasto-plastic soil deposits and full bonding conditions at the soil-tunnel
interface (shaking motion EQ2, PGA = 0.35 g).

those observed and discussed for the full bonding interface the potential slippage phenomena along the interface, leading to
conditions. On the other hand, the effect of the input motion char- slightly different distributions of the dynamic earth pressures for
acteristics becomes more important, as these characteristics affect the various shaking motions investigated herein.
 G. Tsinidis / Tunnelling and Underground Space Technology 62 (2017) 1–22
Similar observations are made for the cases, where a full-slip
interface condition is adopted, allowing for the potential separa-
tion between the soil and the tunnel elements (Fig. 20c). The
non-linear phenomena that are taking place along the interface
(i.e. slippage and/or separation) result in more complex dynamic
earth pressures distributions compared to the perfect bonding con-
dition analyses. It is worth noting that the presented results refer
to the dynamic earth pressures computed at the soil elements,
adjacent to the tunnel lining elements. Hence, non-null earth pres-
sures may be computed even for the locations, where separation
between the two media (i.e. soil and tunnel) is predicted by the
analysis. This is attributed to the simulation of the soil as a
continuum.
Fig. 20d illustrates representative dynamic earth pressures dis-
tributions computed along the perimeter of 6
6 (m) shallow tun-
nels (h = 3 m), embedded in diverse elasto-plastic soil deposits,
highlighting the effect of soil yielding on the development of the
dynamic earth pressures. In line with recent experimental and
numerical findings of Cilingir and Madabhushi (2011a; 2011b)
and Tsinidis et al. (2016a; 2016b), the yielding of the soil and the
associated stress redistributions lead to post-shaking residual
earth pressures on lining. This result in more complex forms for
the dynamic earth pressures distributions around the tunnel com-
pared to those computed evaluated for elastic soil response. Soil
strength controls the development of soil plastic straining around
the tunnel and thus the dynamic earth pressures. Hence, the com-
plexity of the distributions increases with the decrease of the soil
strength. Additionally, input motion characteristics affect the soil
yielding and thus the computed dynamic earth pressures.
4.5. Soil dynamic shear stresses
The effects of salient parameters on the soil dynamic shear
stresses mobilized around rectangular tunnels during ground shak-
ing are discussed in this section.
Fig. 21a portrays representative soil dynamic shear stresses dis-
tributions computed around diverse tunnel sections embedded in
elastic soil (soil type C1). The results, which are computed for full
bonding interface conditions and for the EQ2 shaking motion,
scaled to PGA = 0.35 g, refer to the time step of maximum racking
distortion of the sections. The presented distributions are com-
pared to the soil dynamic shear stresses computed at the same
depth with the tunnel at the soil ‘free-field’ (purple dashed line
in Fig. 21a). Naturally, the soil shear stresses are higher along the
invert slabs, due to the higher confining pressure at these locations.
The soil shear stresses along the roof slabs are quite reduced in
cases of shallow embedment (e.g. h = 3 m), while they are increas-
ing significantly with the increase of the tunnel’s burial depth. The
soil shear stresses along the side-walls are gradually increasing
with depth, moving from the lower magnitudes of the roof slab
to the higher magnitudes of the invert slab. The distributions of
the shear stresses are not affected significantly by the size of the
tunnel. The magnitudes, however, are affected to some extend. In
particular, the shear stresses magnitudes increase with the increas-
ing size of the tunnel section. Similar to the dynamic earth pres-
sures, the effects of the soil deposit properties and input motion
characteristics on the computed distributions are minor for the
purely elastic soil-tunnel configurations discussed herein (i.e. elas-
tic soil response and full bonding interface conditions).
With reference to the effect of the soil to tunnel relative flexibil-
ity, the dynamic soil shear stresses distributions are more uniform
around rigid tunnels (i.e. F < 1.0), while they become more complex
in the case of flexible tunnels (i.e. F > 1.0), due to the increased
inward deformations of the lining in the latter cases. In particular,
for flexible tunnels, the dynamic soil shear stresses are decreasing
near the corners, while they are increasing significantly at the mid-
17
dle sections of the slabs and the side-walls. The ‘free-field’ soil
shear stresses are generally closer to the dynamic shear stresses
computed along the perimeter of tunnels that share the same stiff-
ness with the surrounding soil (F = 1.0).
The effect of the internal column (e.g. in case of double box tun-
nels) on the soil dynamic shear stresses developed along the roof
and the invert slabs is highlighted in Fig. 21a, by comparing the
computed distributions for the 12
6 (m) tunnels (single and dou-
ble boxes). Similar to the dynamic earth pressures, the distribu-
tions are ‘interrupted’ by the internal column in the cases of the
double boxes, with the effect being more important for the flexible
tunnels (F > 1.0).
Fig. 21b draws dynamic soil shear stresses distributions com-
puted around diverse tunnel sections embedded in elastic soil,
assuming full-slip interface conditions. The potential separation
between the two media (i.e. soil and tunnel) is precluded in this
case. It is noted that the results refer to the dynamic shear stresses
computed at the soil elements (adjacent to the tunnel) and not to
the shear stresses developed along the soil-tunnel interface (which
in case of full-slip conditions will be null). The soil shear stresses
are reduced significantly along the middle sections of the slabs
and side-walls compared to the full bonding interface condition
analyses, while they increase near the corners of the tunnel sec-
tions. The concentration of soil shear stresses near the corners
comes in accordance with the increasing confining pressures at
these locations, computed for the specific soil-tunnel interface con-
ditions (i.e. Fig. 20b). Generally, the computed magnitudes are
higher along the middle sections of the slabs and the side-walls
of flexible tunnels (e.g. F > 1.0), due to the increased inward defor-
mations of the linings in theses cases. Similar conclusions were
drawn for the cases, where separation between the soil and the
tunnel elements was permitted (Fig. 21c). Given, the geometrical
non-linearities that may take place along the interface, the effects
of soil properties and input motion characteristics became more
important in these cases compared to the full bonding conditions
analyses.
Fig. 21d illustrates the effect of soil yielding on the soil dynamic
shear stresses distributions developed along the perimeter of 6
6
(m) shallow tunnels (h = 3 m) embedded in diverse elasto-plastic
soil deposits (Table 2). The soil yielding and the associated stress
redistributions around the tunnel alter the distributions of the soil
dynamic shear stresses computed around the tunnels, compared to
those predicted for elastic soil response. In particular, the distribu-
tions become more complex, especially in cases of soil deposits
with reduced strength characteristics, where the yielding response
of the soil is amplified.
4.6. Dynamic internal forces
This section discusses the effects of critical parameters on the
dynamic forces developed on the linings of rectangular tunnels
subjected to transversal ground shaking.
4.6.1. Dynamic bending moment
Fig. 22a portrays dynamic bending moment distributions com-
puted along the perimeter of diverse tunnel sections embedded in
elastic soil (soil type B1). The results, which are computed for the
EQ2 shaking motion, scaled to a PGA = 0.35 g and for perfect bond-
ing interface conditions, refer to the time step of maximum racking
distortion. The dynamic bending moments are distributed in an
anti-symmetric fashion along the slabs and the side-walls, simi-
larly to the dynamic earth pressures distributions. Naturally, the
magnitude of the dynamic bending moment increases with the
increase of the lining rigidity, while the increase of the tunnel flex-
ibility and the associated increased curvature of the lining during
shaking, result in more ‘curved’ bending moment distributions.
18 G. Tsinidis / Tunnelling and Underground Space Technology 62 (2017) 1–22

(a) 2x2 (m), h=3m 3x6 (m), h=3m 6x6 (m), h=3m 6x6 (m), h=12m
80 130 130 200

60 97.5 97.5 150

τ (kPa/m)

τ (kPa/m)

τ (kPa/m)

τ (kPa/m)
40 65 65 100

20 32.5 32.5 50

0 0 0 0
A B C D A A B C D A A B C D A A B C D A
12x6 (m), h=3m (single box) 12x6 (m), h=3m (double box) 18x6 (m), h=3m (double box) 18x6 (m), h=12m (double box)
130 130 130 250

97.5 97.5 97.5 187.5

τ (kPa/m)

τ (kPa/m)

τ (kPa/m)

τ (kPa/m)
65 65 65 125

32.5 32.5 32.5 62.5

0 0 0 0
A B C D A A E B C F D A A E B C F D A A E B C F D A

(b) 6x6 (m), h=3m 12x6 (m), h=3m (single box) 12x6 (m), h=3m (double box) 18x6 (m), h=12m (double box)
150 200 200 300

112.5 150 150 225

τ (kPa/m)
τ (kPa/m)

τ (kPa/m)
τ (kPa/m)

75 100 100 150

37.5 50 50 75

0 0 0 0
A B C D A A B C D A A E B C F D A A E B C F D A

(c) 6x6 (m), h=3m 12x6 (m), h=3m (single box) 18x6 (m), h=3m (double box) 18x6 (m), h=12m (double box)
150 150 250 300

112.5 112.5 187.5 225

τ (kPa/m)

τ (kPa/m)
τ (kPa/m)
τ (kPa/m)

75 75 125 150

37.5 37.5 62.5 75

0 0 0 0
A B C D A A B C D A A E B C F D A A E B C F D A

(d) Soil B1 Soil B2 Soil C1 Soil C2

110 110 110 110

82.5 82.5 82.5 82.5

τ (kPa/m)

τ (kPa/m)

τ (kPa/m)

τ (kPa/m)

55 55 55 55

27.5 27.5 27.5 27.5

0 0 0 0
A B C D A A B C D A A B C D A A B C D A

D C D F C
F=0.2
F=1.0
F=2.0
F=5.0
F=10.0
Soil free field
A B A E B
Fig. 21. (a–c) Distributions of soil dynamic shear stresses developed along the perimeter of diverse tunnel sections at the time step of maximum racking distortion, (a) elastic
soil (soil type B1), full bonding conditions at the soil-tunnel interface, (b) elastic soil (soil type B1), full-slip conditions precluding potential separation between soil and tunnel
elements, (c) elastic soil (soil type B1), full-slip conditions permitting potential separation between soil and tunnel elements, and (d) Soil dynamic shear stresses distributions
developed along the perimeter of 6
6 (m) shallow tunnels, as computed for diverse elasto-plastic soil deposits and full bonding conditions at the soil-tunnel inteface
(shaking motion EQ2, PGA = 0.35 g).
 G. Tsinidis / Tunnelling and Underground Space Technology 62 (2017) 1–22 19

(a) 2x2 (m), h=3m 3x6 (m), h=3m 6x6 (m), h=3m 6x6 (m), h=12m
50 400 700 1600

25 200 350 800

M (kNm/m)

M (kNm/m)

M (kNm/m)

M (kNm/m)
0 0 0 0

−25 −200 −350 −800

−50 −400 −700 −1600

A B C D A A B C D A A B C D A A B C D A
12x6 (m), h=3m (single box) 12x6 (m), h=3m (double box) 18x6 (m), h=3m (double box) 18x6 (m), h=12m (double box)
1500 800 1200 2600

750 400 600 1300

M (kNm/m)

M (kNm/m)

M (kNm/m)

M (kNm/m)
0 0 0 0

−750 −400 −600 −1300

−1500 −800 −1200 −2600

A B C D A A E B C F D A A E B C F D A A E B C F D A

(b) 6x6 (m), h=3m 12x6 (m), h=3m (single box) 18x6 (m), h=3m (double box) 18x6 (m), h=12m (double box)
700 1500 800 1600

350 750 400 800

M (kNm/m)

M (kNm/m)

M (kNm/m)

M (kNm/m)
0 0 0 0

−350 −750 −400 −800

−700 −1500 −800 −1600

A B C D A A B C D A A E B C F D A A E B C F D A

(c) 6x6 (m), h=3m 12x6 (m), h=3m (single box) 18x6 (m), h=3m (double box) 18x6 (m), h=12m (double box)
700 800 800 1000

350 400 400 500

M (kNm/m)

M (kNm/m)

M (kNm/m)

0 0 0 M (kNm/m) 0

−350 −400 −400 −500

−700 −800 −800 −1000

A B C D A A B C D A A E B C F D A A E B C F D A

(d) Soil B1 Soil B2 Soil C1 Soil C2

700 700 400 400

350 350 200 200

M (kNm/m)

M (kNm/m)

M (kNm/m)

M (kNm/m)

0 0 0 0

−350 −350 −200 −200

−700 −700 −400 −400

A B C D A A B C D A A B C D A A B C D A

D C D F C
F=0.2
F=1.0
F=2.0
F=5.0
F=10.0
A B A E B

Fig. 22. (a–c) Dynamic bending moment distributions developed along the perimeter of diverse tunnel sections at the time step of maximum racking distortion, (a) elastic
soil (soil type B1), full bonding conditions at the soil-tunnel interface, (b) elastic soil (soil type B1), full-slip conditions precluding potential separation between the soil and
tunnel elements, (c) elastic soil (soil type B1), full-slip conditions permitting potential separation between the soil and tunnel elements, and (d) Dynamic bending moment
distributions developed along the perimeter of 6
6 (m) shallow tunnels, as computed for diverse elasto-plastic soil deposits and full bonding conditions at the soil-tunnel
interface (shaking motion EQ2, PGA = 0.35 g).

For the shallow tunnels (i.e. h = 3 m), the dynamic bending roof slab corners, ranging between 0.6 and 0.8 of the maximum
moment is maximized at the bottom corners of the tunnel section bending moment. The distributions become more symmetric along
(i.e. invert slab corners), with the dynamic bending moment at the the perimeter of deep tunnels (i.e. h = 12 m), with the dynamic
20 G. Tsinidis / Tunnelling and Underground Space Technology 62 (2017) 1–22

(a) 2x2 (m), h=3m 3x6 (m), h=3m 6x6 (m), h=3m 6x6 (m), h=12m
50 200 280 420

25 100 140 210

N (kN/m)

N (kN/m)

N (kN/m)

N (kN/m)
0 0 0 0

−25 −100 −140 −210

−50 −200 −280 −420

A B C D A A B C D A A B C D A A B C D A
12x6 (m), h=3m (single box) 12x6 (m), h=3m (double box) 18x6 (m), h=3m (double box) 18x6 (m), h=12m (double box)
400 400 500 800

200 200 250 400

N (kN/m)

N (kN/m)

N (kN/m)

N (kN/m)
0 0 0 0

−200 −200 −250 −400

−400 −400 −500 −800

A B C D A A E B C F D A A E B C F D A A E B C F D A

(b) 6x6 (m), h=3m 12x6 (m), h=3m (single box) 18x6 (m), h=3m (double box) 18x6 (m), h=12m (double box)
100 300 500 800

50 150 250 400

N (kN/m)

N (kN/m)

N (kN/m)

N (kN/m)
0 0 0 0

−50 −150 −250 −400

−100 −300 −500 −800

A B C D A A B C D A A E B C F D A A E B C F D A

(c) 6x6 (m), h=3m 12x6 (m), h=3m (single box) 18x6 (m), h=3m (double box) 18x6 (m), h=12m (double box)
400 500 1000 1200

200 250 500 600

N (kN/m)

N (kN/m)

N (kN/m)

N (kN/m)

0 0 0 0

−200 −250 −500 −600

−400 −500 −1000 −1200

A B C D A A B C D A A E B C F D A A E B C F D A

(d) Soil B1 Soil B2 Soil C1 Soil C2

260 260 200 260

130 130 100 130

N (kN/m)

N (kN/m)

N (kN/m)

N (kN/m)

0 0 0 0

−130 −130 −100 −130

−260 −260 −200 −260

A B C D A A B C D A A B C D A A B C D A

D C D F C
F=0.2
F=1.0
F=2.0
F=5.0
F=10.0
A B A E B
Fig. 23. (a–c) Dynamic axial force distributions developed along the perimeter of diverse tunnel sections at the time step of maximum racking distortion, (a) elastic soil (soil
type B1), full bonding conditions at the soil-tunnel interface, (b) elastic soil (soil type B1), full-slip conditions precluding potential separation between soil and tunnel
elements, (c) elastic soil (soil type B1), full-slip conditions permitting potential separation between soil and tunnel elements, and (d) Dynamic axial force distributions
developed along the perimeter of 6
6 (m) shallow tunnels, as computed for diverse elasto-plastic soil deposits and full bonding conditions (shaking motion EQ2,
PGA = 0.35 g).
 G. Tsinidis / Tunnelling and Underground Space Technology 62 (2017) 1–22
bending moment at the roof slab corners being almost equal to the
dynamic bending moment computed at the invert slab corners. The
size of the tunnel affects significantly the magnitude of the
dynamic bending moment. The same conclusion is drawn for the
aspect ratio, k. In particular, higher aspect ratios (larger spans)
result in an increase of the dynamic bending moment. Clearly,
the existence of an internal column (e.g. 12 m
6 m double box
in Fig 22a) affects the bending moment distributions along the
slabs, reducing significantly the dynamic bending moment at the
corners.
Fig. 22b illustrates dynamic bending moment distributions
computed along the perimeter of diverse tunnel sections in elastic
soil (soil type B1), for full-slip interface conditions. The potential
separation between the two media (i.e. soil and tunnel) is pre-
cluded herein. As mentioned in Section 4.1, the full-slip condition
analyses predict lower racking distortion for the tunnel sections
compared to the full bonding condition analyses. Hence, the corre-
sponding dynamic bending moment is found to be reduced for full-
slip interface conditions. Generally, the dynamic bending moment
is further reduced for the cases of full-slip conditions, where the
potential separation between the two media is permitted
(Fig. 22c).
The effect of soil yielding on the dynamic bending moment dis-
tributions, computed along the perimeter of 6
6 (m) shallow tun-
nels (h = 3 m), embedded in diverse elasto-plastic soil deposits
(Table 2), is portrayed in Fig. 22d. Similar to the dynamic earth
pressures and the soil dynamic shear stresses, and in line with
recent experimental and numerical findings of Lanzano et al.
(2012) and Tsinidis et al. (2015, 2016a, 2016b, 2016c), the yielding
of the surrounding soil leads to post-shaking residual values on the
lining dynamic bending moment, which subsequently lead to more
complex forms for the dynamic bending moment distributions
around the tunnel, compared to those predicted for elastic soil
response. The effects of the soil shear strength, lining flexibility
and input motion characteristics on the yielding of the surrounding
ground and thus on the computed lining dynamic bending moment
are important.
4.6.2. Dynamic axial force
Representative dynamic axial force distributions, computed
along the perimeter of various tunnel sections embedded in elastic
soil (soil type B1), are present in Fig. 23a. The results are plotted for
full bonding interface conditions and refer to the time step of max-
imum racking distortion. Similar to the dynamic bending moment,
the dynamic axial forces are distributed in an anti-symmetric fash-
ion along the slabs and the side-walls. The effect of the tunnel
rigidity on the dynamic axial forces is less important compared
to the dynamic bending moment, especially for the smaller tunnel
sections. The magnitude of the dynamic axial force increases with
the increase of the rigidity of the tunnel, as well as with the
increase of the burial depth and size of the tunnel section. For
the shallow tunnels, the dynamic axial force is maximized at the
bottom corners of the tunnel section, with the dynamic axial force
at the top corners, ranging between 0.5 and 0.75 of the maximum
axial force. The dynamic axial force distributions computed along
the linings of deep tunnels are more symmetric, with the dynamic
axial force at the roof slab corners being almost equal to the one
computed at the invert slab corners.
Similar to the dynamic bending moment, the existence of an
internal column (e.g. 12 m
6 m double box in Fig. 23a) affects
the axial force distributions along the slabs, reducing significantly
the axial force magnitude at the corners.
The effect of the soil-tunnel interface characteristics on the
dynamic axial force is very important. Both the distributions and
the magnitudes are affected by the different interface characteris-
tics, as portrayed in Fig. 23b and c. Although it is not easy to draw
21
clear conclusions in these cases, the dynamic lining forces com-
puted for full-slip interface conditions are significantly reduced
(compared to full bonding interface conditions), while they are dis-
tributed in a more ‘uniform’ fashion along the slabs and the side-
walls.
Fig. 23c illustrates typical dynamic axial force distributions
computed along the perimeter of shallow tunnels, embedded in
diverse elasto-plastic soil deposits. Similar to the dynamic bending
moment, the yielding of the surrounding ground results to post-
shaking residual axial forces on the lining, thus diversifying the
dynamic axial force distributions compared to those computed
for full bonding interface conditions. The effect is more evident
for soil deposits with low shear strength properties, where soil
yielding is generally amplified.
5. Conclusions
A numerical parametric study was conducted, aiming to inves-
tigate critical aspects of the transversal seismic response of rectan-
gular tunnels embedded in soft soil. The key findings of this study
are summarized in the following:
 A combined racking-rocking deformation pattern was identified
for the rectangular tunnels during ground shaking. Additionally,
inward deformations of the slabs and the side-walls were
observed when soil inelasticity was encountered, with this
observation being more evident for flexible tunnels. The soil-
tunnel relative flexibility and interface conditions, as well as
the yielding of the surrounding ground were found to affect
the above deformation patterns. For elastic soil response, the
racking distortion was amplified for full (i.e. perfect) bonding
between the soil and the tunnel elements. For these conditions
(i.e. elastic soil response and full bonding interface conditions),
the rotation pole of the tunnels coincided with the centroid of
the sections. Non-linearities of any kind (e.g. geometrical or
material non-linearities) resulted in a slight movement of the
rotation pole from the tunnel section centroid, affecting also
the magnitudes of both the racking distortion and rotation of
the tunnel sections.
 A series of numerical R-F relations were developed and com-
pared to the existing literature. The racking ratios, R, were found
to increase with the increase of the tunnels’ burial depth. Addi-
tionally, the racking ratios, R, of flexible tunnels (F > 1.0)
increased with the increase of the aspect ratio, k, while the
exact opposite trend was observed for rigid tunnels. The R-F
relation proposed by Penzien (2000) for no-slip interface condi-
tions was found to be a maximum envelope for the numerical R-
F relations.
 The rocking response of rectangular tunnels was quantified by
means of dimensionless relations (i.e. h/cff-F) similar to the R-
F relations. Generally, the rotation of the tunnel section
increased for lower aspect ratios, k.
 For elastic soil response and perfect bonding interface condi-
tions, the dynamic earth pressures were distributed in anti-
symmetric fashion along the slabs and the side-walls. Higher
dynamic earth pressures were developed around flexible tun-
nels, while the consideration of a full-slip soil-tunnel interface
resulted in concentrations of the dynamic earth pressures near
the corners of the tunnel. Additionally, the soil yielding, pre-
dicted by the elasto-plastic analyses, resulted in complex distri-
butions of the dynamic earth pressures around the tunnel.
 The distributions of the soil dynamic shear stresses were
affected significantly by the soil-tunnel interface characteristics
and relative flexibility. For full-slip interface conditions and
elastic soil response, the soil dynamic shear stresses were
22 G. Tsinidis / Tunnelling and Underground Space Technology 62 (2017) 1–22
reduced at the middle sections of the slabs and the side-walls,
while they were increased significantly near the corners of the
tunnel. The consideration of a full bonding interface condition
resulted in a significant increase of the soil shear stresses along
the middle sections of the slabs and side-walls. Similar to the
dynamic earth pressures, the yielding of the surrounding
ground resulted in more complex forms of the soil shear stres-
ses distributions around the tunnels.
 The dynamic bending moments, as well as the dynamic axial
forces, computed for full bonding interface conditions and elas-
tic soil response, were distributed in an anti-symmetric fashion
along the slabs and the side-walls. The consideration of a full-
slip interface resulted in a reduction of the dynamic bending
moment magnitudes, compared to those predicted for full
bonding conditions. The effect of interface conditions was par-
ticularly important for the dynamic axial forces. Finally, the
yielding response of the surrounding ground was found to affect
significantly the dynamic lining forces distributions, with the
effect being more evident on flexible tunnels.
The presented results lead to a better understanding of the
dynamic response of rectangular tunnels in soft soil, while the pro-
posed relations contribute towards the improvement of the R-F
analysis method, which is often used in tunnelling design practice.
Acknowledgements
The Author gratefully acknowledges Prof. Kyriazis Pitilakis
(Aristotle University of Thessaloniki) for the fruitful discussions
he had with during this study.
References
ABAQUS, 2012. ABAQUS: theory and analysis user’s manual, version 6.12. Dassault
Systèmes SIMULIA, Providence, RI, USA.
Abuhajar, O., El Naggar, H., Newson, T., 2015a. Seismic soil-culvert interaction. Can.
Geotech. J. 52, 1–19.
Abuhajar, O., El Naggar, H., Newson, T., 2015b. Experimental and numerical
investigations of the effect of buried box culverts on earthquake excitation. Soil
Dyn. Earthq. Eng. 79, 130–148.
Amorosi, A., Boldini, D., 2009. Numerical modeling of the transverse dynamic
behavior of circular tunnels in clayey soils. Soil Dyn. Earthq. Eng. 59 (6), 1059–
1072.
Anastasopoulos, I., Gerolymos, N., Drosos, V., Kourkoulis, R., Georgarakos, T.,
Gazetas, G., 2007. Nonlinear response of deep immersed tunnel to strong
seismic shaking. J. Geotech. Geoenviron. Eng. 133 (9), 1067–1090.
Anastasopoulos, I., Gerolymos, N., Drosos, V., Georgarakos, T., Kourkoulis, R.,
Gazetas, G., 2008. Behavior of deep immersed tunnel under combined normal
fault rupture deformation and subsequent seismic shaking. Bull. Earthq. Eng. 6
(2), 213–239.
Anderson, D.G., Martin, G.R., Lam, I., Wang, J.N., 2008. NCHPR611: Seismic Analysis
and Design of Retaining Walls, Buried Structures, Slopes and Embankments.
National Cooperative Highway Research Program, Transportation Research
Board, Washington, DC, USA.
Baziar, M.H., Moghadam, M.R., Kim, D.-S., Choo, Y.W., 2014. Effect of underground
tunnel on the ground surface acceleration. Tunn. Undergr. Space Technol. 44,
10–22.
Bilotta, E., Lanzano, G., Madabhushi, S.P.G., Silvestri, F., 2014. A numerical Round
Robin on tunnels under seismic actions. Acta Geotech. 9 (4), 563–579.
Bobet, A., Fernandez, G., Huo, H., Ramirez, J., 2008. A practical iterative procedure to
estimate seismic-induced deformations of shallow rectangular structures. Can.
Geotech. J. 45, 923–938.
Bobet, A., 2010. Drained and undrained response of deep tunnels subjected to far-
field shear loading. Tunn. Undergr. Space Technol. 25 (1), 21–31.
CEN, 2004. Eurocode 8: Design of structures for earthquake resistance - Part 1:
General rules, seismic actions and rules for buildings. EN1998-1: 2004,
European Committee for Standardization, Brussels.
Chen, G., Wang, Z., Zuo, X., Du, X., Gao, H., 2013. Shaking table test on seismic failure
characteristics of a subway station structure in liquefiable ground. Earthq. Eng.
Struct. Dyn. 42, 1489–1507.
Chian, S.C., Madabhushi, S.P.G., 2012. Effect of buried depth and diameter on uplift
of underground structures in liquefied soils. Soil Dyn. Earthq. Eng. 41, 181–190.
Chou, J.C., Kutter, B.L., Travasarou, T., Chacko, J.M., 2010. Centrifuge modeling of
seismically induced uplift for the BART Transbay tube. J. Geotech. Geoenviron.
Eng. 137 (8), 754–765.
Cilingir, U., Madabhushi, S.P.G., 2011a. A model study on the effects of input motion
on the seismic behavior of tunnels. Soil Dyn. Earthq. Eng. 31, 452–462.
Cilingir, U., Madabhushi, S.P.G., 2011b. Effect of depth on the seismic response of
square tunnels. Soils Found. 51 (3), 449–457.
Cilingir, U., Madabhushi, S.P.G., 2011c. Effect of depth on the seismic response of
circular tunnels. Can. Geotech. J. 48 (1), 117–127.
Debiasi, E., Gajo, A., Zonta, D., 2013. On the seismic response of shallow-buried
rectangular structures. Tunn. Undergr. Space Technol. 38, 99–113.
FHWA (Federal Highway Administration), 2009. Technical manual for design and
construction of road tunnels-Civil elements. Publication No. FHWA-NHI-10-
034, Department of transportation, Federal Highway Administration,
Washington D.C., U.S.
Hashash, Y.M.A., Hook, J.J., Schmidt, B., Yao, J.I.-C., 2001. Seismic design and analysis
of underground structures. Tunn. Undergr. Space Technol. 16 (2), 247–293.
Hashash, Y.M.A., Park, D., Yao, J.I.C., 2005. Ovaling deformations of circular tunnels
under seismic loading, an update on seismic design and analysis of
underground structures. Tunn. Undergr. Space Technol. 20 (5), 435–441.
Hashash, Y.M.A., Karina, K., Koutsoftas, D., O’Riordan, N., 2010. Seismic design
considerations for underground box structures. In: Proceedings of Earth
Retention Conference - Earth Retention Conference 3. ASCE, Bellevue, WA,
USA, pp. 620–637.
Huo, H., Bobet, A., Fernández, G., Ramírez, J., 2005. Load transfer mechanisms
between underground structure and surrounding ground: evaluation of the
failure of the Daikai station. J. Geotech. Geoenviron. Eng. 131 (12), 1522–1533.
Hushmand, A., Dashti, S., Davis, C., McCartney, J.S., Hushmand, B., 2016. A centrifuge
study of the influence of site response, relative stiffness, and kinematic
constraints on the seismic performance of buried reservoir structures. Soil
Dyn. Earthq. Eng. 88, 427–438.
Iida, H., Hiroto, T., Yoshida, N., Iwafuji, M., 1996. Damage to Daikai subway station.
In: Special issue on geotechnical aspects of the January 17 1995 Hyogoken-
Nanbu earthquake. Soils and Foundations, pp. 283–300.
ISO (International Organization for Standardization), 2005. ISO 23469: Bases for
design of structures - Seismic actions for designing geotechnical works.
International Standard ISO TC98/SC3/WG10. Geneva, Switzerland:
International Organization for Standardization.
Kontoe, S., Zdravkovic, L., Potts, D., Mentiki, C., 2011. On the relative merits of
simple and advanced constitutive models in dynamic analysis of tunnels.
Geotechnique 61 (10), 815–829.
Kontoe, S., Avgerinos, V., Potts, D.M., 2014. Numerical validation of analytical
solutions and their use for equivalent-linear seismic analysis of circular tunnels.
Soil Dyn. Earthq. Eng. 66, 206–219.
Lanzano, G., Bilotta, E., Russo, G., Silvestri, F., Madabhushi, S.P.G., 2012. Centrifuge
modeling of seismic loading on tunnels in sand. Geotech. Test. J. 35 (6), 854–
869.
Lanzano, G., Bilotta, E., Russo, G., Silvestri, F., 2015. Experimental and numerical
study on circular tunnels under seismic loading. Eur. J. Environ. Civ. Eng. 19 (5),
539–563.
Lysmer, J., Kuhlemeyer, R.L., 1969. Finite dynamic model for infinite media. J. Eng.
Mech. Div., ASCE 95 (4), 859–878.
Park, K.H., Tantayopin, K., Tontavanich, B., Owatsiriwong, A., 2009. Analytical
solution for seismic-induced ovaling of circular tunnel lining under no-slip
interface conditions: a revisit. Tunn. Undergr. Space Technol. 24 (2), 231–235.
Penzien, J., 2000. Seismically induced racking of tunnel linings. Earthq. Eng. Struct.
Dyn. 29, 683–691.
Pitilakis, K., Tsinidis, G., 2014. Performance and seismic design of underground
structures. In: Maugeri, M., Soccodato, C. (Eds.), Earthquake Geotechnical
Engineering Design, Geotechnical Geological and Earthquake Engineering 28.
Springer international publishing, Switzerland, pp. 279–340.
Power, M., Rosidi, D., Kaneshiro, J., Gilstrap, S., Chiou, S.J., 1998. Summary and
evaluation of procedures for the seismic design of tunnels. Final Report for Task
112-d-5.3(c). National Center for Earthquake Engineering Research, Buffalo,
New York.
Sharma, S., Judd, W.R., 1991. Underground opening damage from earthquakes. Eng.
Geol. 30, 263–276.
Shibayama, S., Izawa, J., Takahashi, A., Takemura, J., Kusakabe, O., 2010. Observed
behavior of a tunnel in sand subjected to shear deformation in a centrifuge.
Soils Found. 50 (2), 281–294.
Tsinidis, G., Pitilakis, K., Madabhushi, G., Heron, C., 2015. Dynamic response of
flexible square tunnels: centrifuge testing and validation of existing design
methodologies. Geotechnique 65 (5), 401–417.
Tsinidis, G., Pitilakis, K., Madabhushi, G., 2016a. On the dynamic response of square
tunnels in sand. Eng. Struct. 125, 419–437.
Tsinidis, G., Rovithis, E., Pitilakis, K., Chazelas, J.L., 2016b. Seismic response of box-
type tunnels in soft soil: experimental and numerical investigation. Tunn.
Undergr. Space Technol. 59, 199–214.
Tsinidis, G., Pitilakis, K., Anagnostopoulos, C., 2016c. Circular tunnels in sand:
dynamic response and efficiency of seismic analysis methods at extreme lining
flexibilities. Bull. Earthq. Eng. 14 (10), 2903–2929. http://dx.doi.org/10.1007/
s10518-016-9928-1.
Ulgen, D., Saglam, S., Ozkan, M.Y., 2015. Dynamic response of a flexible rectangular
underground structure in sand: centrifuge modeling. Bull. Earthq. Eng. 13,
2547–2566.
Wang, J.N., 1993. Seismic Design of Tunnels: A Simple State of the Art Design
Approach. Parsons Brinckerhoff Inc., New York.