Tunnelling and Underground Space Technology 64 (2017) 164–176

Contents lists available at ScienceDirect

Tunnelling and Underground Space Technology

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

Prediction of tunnel lining forces and deformations using analytical

and numerical solutions
Chenyang Zhao a, Arash Alimardani Lavasan a,⇑, Thomas Barciaga a, Christoph Kämper b, Peter Mark b,
Tom Schanz a
a
Ruhr-Universität Bochum, Chair of Foundation Engineering, Soil and Rock Mechanics, Universitätsstr. 150, 44780 Bochum, Germany
b
Ruhr-Universität Bochum, Institute of Concrete Structures, Universitätsstr. 150, 44780 Bochum, Germany

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

Article history: Structural design of linings requires a reliable prognosis of lining forces and deformations. In engineering
Received 15 October 2015 practice, both analytical and numerical solutions are popular to be employed to predict the system
Received in revised form 5 December 2016 behavior. This paper employs the commonly accepted analytical solutions to calculate the lining forces
Accepted 23 January 2017
and deformations for both shallow and deep tunnels, the results are compared to the numerical results
Available online 17 February 2017
for corresponding equivalent boundary conditions, initial conditions and identical material properties.
Afterward, more sophisticated constitutive models for soil/structure elements in conjunction with more
Keywords:
realistic construction aspects are taken into account. The comparison of the results of analytical and
Lining forces
Analytical solution
numerical solutions highlights the differences between these two well accepted methods as well as
Numerical analysis the effect of considering realistic features in numerical simulations. Moreover, the lining forces and defor-
Tunnel design mations obtained from plain strain condition are compared to the 3D numerical results. The results show
Soil-structure interaction that the analytical bedding model is able to reasonably predict the lining behavior for both shallow and
Arching effect deep tunnels even if the soil is assumed to be an elastic material. In numerical solutions, lining forces and
deformations depend to a large extent on the applied soil constitutive model and construction method.
The face support pressure, backfill grouting and arching effect cannot be captured appropriately in plain
strain condition, which leads to the discrepancy between the model responses obtained from 2D numer-
ical/analytical solutions and realistic 3D simulations.
Ó 2017 Elsevier Ltd. All rights reserved.

1. Introduction
Proper design of lining segments plays a pivotal technical and
economical role in mechanized excavation for both shallow and
deep tunnels. For the conventionally driven open face tunneling
technique, stress release of soil domain after excavation has an
important impact on the structural design of concrete lining.
Nowadays, tunnel boring machines (TBM) have been widely
applied in tunnel construction, especially in urban areas where
the most important aim is to minimize the soil deformation. The
behavior of lining segments is affected by the complex construc-
tion features, for example the sequential excavation process and
backfill grouting. Therefore, developing a framework to accurately
predict the lining forces and deformations before tunnel construc-
tion is essential for the purpose of structural safety and optimum
design. Additionally, the cost of tunnel construction depends to a
⇑ Corresponding author.
E-mail address: arash.alimardanilavasan@rub.de (A. Alimardani Lavasan).
large extent on the cost of lining, this also contributes to the
importance of structural design of linings on short and long terms
stability of the tunnel.
Axial forces, bending moments and radial displacements are the
most significant responses of lining segments during construction,
which strongly depend on the confining pressure due to the
surrounding soil stresses. As soil deformation and soil-lining
interaction induce the variation of soil stresses applied on lining,
analysis of tunnel lining and its interaction with soil becomes even
more complex because of the dependence of such interaction on
the construction technology and schemes (El-Nahhas et al., 1992).
In order to reliably predict the lining forces and deformations,
finite element method (FEM) analysis has become a popular tool
which can simulate staged construction procedures and reproduce
the soil and structure behavior by using appropriate constitutive
models. In engineering practice, the numerical modeling often
relies on the two dimensional (2D) analysis, since it is
straightforward and cost-effective. For the numerical simulation
of mechanized tunneling process in plain strain condition, it

http://dx.doi.org/10.1016/j.tust.2017.01.015
0886-7798/Ó 2017 Elsevier Ltd. All rights reserved.
 C. Zhao et al. / Tunnelling and Underground Space Technology 64 (2017) 164–176
normally takes into account the tunnel construction process
including TBM excavation, backfill grouting and lining installation
as well as the soil-lining interaction. Oreste (2007) developed a
special code within the FEM framework using hyperstatic reaction
method to consider the actual geometry of the lining support and
the horizontal loads that are different from the vertical ones, it is
therefore able to analyze the mass-structure interaction in detail.
Möller and Vermeer (2008) applied FEM to simulate the conven-
tionally driven Steinhaldenfeld tunnel and Heinenoord slurry
shield tunnel, and studied the influences of constitutive model
and applied construction method on lining forces and ground
deformation. Zhang et al. (2015) analyzed the influence of multi-
layered soil formation on tunnel lining behavior by employing
FEM, and the results show a good agreement between the numer-
ical model responses and the real measurements.
In the realistic tunneling process, the excavation procedure
changes the primary stress field at the tunneling face (ITA, 1988).
Furthermore, the soil stresses on lining segments are influenced
by the sequential excavation process and 3D arching effect of the
soil towards the end of tunnel. Despite of the popularity of 2D
numerical analysis, its deficiencies, that face support in front of
TBM, the sequential excavation process and the inclination of the
tunnel cannot be modeled, are inevitable. Within this framework,
the use of 3D FEM analysis is essential if one wants to correctly
evaluate the influence of staged excavation process on lining struc-
ture responses. Hudoba (1997) studied how the lining structures
react under static loading of the surrounding soil during tunneling
process using both 2D and 3D computing models. Galli et al. (2004)
modeled tunnel excavation and lining installation in both 2D and
3D models, they showed that 3D discretization of soil-tunnel sys-
tem is essential to analyze the soil deformation and stresses in
the lining elements.
Although FEM analysis is a powerful tool in simulations of engi-
neering problems, uncertainty in model responses is unavoidable
due to the complex tunnel construction procedures, limitations
of employed FEM techniques and the insufficiency of constitutive
models themselves. Additionally, it costs many resources to obtain
the input parameters to be used in numerical model, while running
the sophisticated model is time-consuming, especially for complex
3D tunneling model. For the preliminary design of lining segments,
analytical solutions can be used to give a good insight into the
dominant processes. The analytical solutions are developed on
the basis of 2D idealization of tunnel construction. An overview
of the contributions to analytical solutions is given in Table 1.
The list and the following brief discussion are, of necessity, incom-
plete. Schmid (1926) was probably the first who proposed an ana-
lytical solution for thick lining segments in contact with elastic
Table 1
Brief overview of the contributions to structural design models of linings.
Reference Description
Schmid (1926) First to analyze the elastic continuum with
considering the soil-lining interaction
Voellmy (1937) The continuum model, omission of the tangential
components of soil pressure
Bull (1944) The bedding model for shallow tunnel, tedious
calculation
Engelbreth (1961) The continuum model with closed form
Schulze and Duddeck The bedding model with complete and closed
(1964a) solution
Windels (1967) The continuum model with complete solution
Ahrens et al. (1982) The exhaustive and theoretically complete solution
Bakker (2003) Unidimensional model without considering the
soil-lining interaction
Kim and Eisenstein Using correction factors considering the non-linear
(2006) ground behavior
165
soil. A thinner lining is assumed by Voellmy (1937), however, his
results do not comply with the equilibrium of forces since they
neglected the tangential components of soil pressure. In 1964,
Schulze and Duddeck (1964a) published a complete and closed
solution which was applicable for shallow tunnels. After that,
Windels (1967) proposed a complete solution on the circular tun-
nel in an elastic soil domain by also taking into account the geo-
metric nonlinearity and an approximation for the lining
deformations. In 1982, an exhaustive and theoretically complete
solution was finally achieved by Ahrens et al. (1982). Additionally,
Molins and Arnau (2011) and Arnau and Molins (2011) conducted
experimental and analytical study of the structural response of
segmental lining based on an in situ loading test. Their approach
is able to asses a realistic consideration of the soil-structure
interaction.
In cases where the lining responses from analytical and numer-
ical solutions differ expensively, Behnen et al. (2015) pointed out
that a design strategy based on simple or comprehensible analyti-
cal models in combination with the valuable knowledge of experi-
enced engineers should be preferred instead of relying on the
complex computational models. In this study, different commonly
accepted analytical methods are compared firstly in order to eval-
uate their applicability. Afterward, the FE-models for tunneling
simulation and corresponding hierarchical modeling strategies
are illustrated. By doing so, the results of analytical and numerical
solutions are compared for equivalent boundary conditions, initial
conditions and identical material properties. Subsequently, more
sophisticated constitutive models for soil/structure elements in
conjunction with more realistic construction aspects are taken into
account. The comparison of the results of analytical and numerical
solutions highlights the differences between these two well
accepted methods as well as the effect of considering realistic fea-
tures in numerical simulations. Additionally, the difference
between lining forces and deformations calculated via 2D and 3D
FE-models is also discussed.
2. Analytical solutions
An overview of different contributions to structural design
models for tunnel linings is given in Table 1. General agreement
of these analytical models lies on the following basic assumptions:
(1) the analytical solutions are sufficient to consider only a cross-
section, which means plain strain condition is assumed; (2) the
cross section of the tunnel is assumed to be circular; (3) the soil
stresses on the lining segments are assumed to be equal to the pri-
mary stresses in the undistributed ground; (4) there is a bond
between the lining and the ground, it takes into account the soil-
lining interaction; and (5) the material behavior of soil and lining
is generally assumed to be elastic. In Germany two main categories
of structural models for mechanized tunneling have been proven
and established for usual applications: the continuum models
(see Fig. 1(a) and (b)) preferred for deep tunnels and the bedding
models (see Fig. 1(c)) preferred for shallow tunnels.
Commonly, the analytical continuum model consists of a
homogenous elastic circular ring embedded in a plane 2D-
continuum (see Fig. 1(a)). Herein, the idealized primary stress state
is obtained from equilibrium of vertical and horizontal forces
induced by earth pressures. The vertical component of lining load
(rv ) is modeled as an uniformly distributed load on top and bot-
tom of the tunnel. This lining load is depth independent and deter-
mined based on the soil stress at the depth of tunnel axis. The
horizontal load (rh ) is also applied as a constant pressure, its mag-
nitude is defined as the vertical earth pressure multiplied by the
coefficient of lateral earth pressure at rest (K 0 ), namely
rh ¼ K 0 rv . To compute the internal forces on lining, it is necessary
166 C. Zhao et al. / Tunnelling and Underground Space Technology 64 (2017) 164–176
Fig. 1. Schematic illustration of different analytical solutions: (a) the continuum model with Cartesian coordinates, (b) the continuum model with polar coordinates and (c)
the bedding model.
to transform the vertical and horizontal earth pressures into radial
and tangential load components (see Fig. 1(b)). Therefore, the Four-
ier transformation can be employed to respectively split the radial
and tangential forces acting on the circular ring and the contin-
uum. Generally, two different analytical solutions can be distin-
guished: (a) full bond takes into account both, radial and
tangential bond between the soil and lining; (b) tangential slip
assumes only the radial bond between soil and lining, and the tan-
gential load component does not affect the lining behavior. In con-
crete lining design the model of the circular ring embedded in an
elastic continuum serves as an adequate analytical model for deep
tunnels. It takes into account the arching effect due to a fully
embedded ring and allows for tensile forces in springs to minimize
loads on the lining. Consequently lower deformations and lining
forces occur which constitute the minimum level of lining forces
at all. In case of shallow tunnels the idealizations, more or less sim-
plifications, of the continuum model (e.g. soil-structure interac-
tion) strongly underestimate deformations and lining forces so
that this model might be inappropriate. For more details, one is
referred to Ahrens et al. (1982).
In general, the bedding beam model employs linear or nonlinear
springs whose stiffnesses are assumed to be equal to the prede-
fined subgrade moduli (K s ¼ Es =r, where Es is the constrained mod-
ulus of the surrounding soil and r is the radius of the tunnel). These
moduli are specifically obtained with respect to the alternative
modeling approach. For instance, in case of deep tunnels a reason-
able mechanical model assumes fully embedded circular rings and
elastic springs. Generally, the bedding model is derived from a cir-
cular ring embedded in an elastic domain assuming tangential slip
for soil-structure interaction. Accordingly, fully embedded crowns
lead to tensile stresses in the elastic springs (crown of the tunnel).
Consequently, the lining is locally unloaded. In practice this is
related to the soil arching effect that results in lower bending
moments. However, the arching effect is normally neglected in
case of shallow tunnel (the overburden is less than twice the diam-
eter of the tunnel). To ensure appropriate bending moments the
crown is modeled to be unbedded in the analytical method (see
Fig. 1(c)). In mechanized tunneling, the unbedded crown happens
when the ring leaves the TBM or due to loosening of the surround-
ing soil during excavation (Schulze and Duddeck, 1964a). After
that, the crown, due to inward deflections, is subjected to the soils
total weight but not embedded anymore. This results in larger
bending moments in the crown. Hereby, the size of the unbedded
zone (about 100°) is characterized by transition of inwards to out-
wards radial deformations and the positions of longitudinal joints
of a segmental lining. The analytical bedding model in this paper is
based on the original equations for elastically embedded tubes
given by Voellmy (1937). These equations have been modified
and transferred to derive internal forces and radial deformations
by Schulze and Duddeck (1964a).
Duddeck and Erdmann (1985) pointed out that it is rather diffi-
cult to decide whether continuum or bedding model is more
appropriate for the individual tunnel under consideration. They
recommended that the design engineer may choose either the con-
tinuum model (for moderately deep tunnel) or the bedding model
(for shallow tunnel without reduction of the ground pressure at the
crown) for the structural design of linings. In this paper, the contin-
uum model and bedding model are both employed to calculate the
lining axial forces, bending moments and radial displacements. It
should be noted that the analytical solutions are not able to cap-
ture the effects of TBM excavation, grout injection and three
dimensional arching effect. Generally the idealization of loads
around the lining of the bedding model is comparable to the
approach of the continuum model. A deeper discussion can be
found in Schulze and Duddeck (1964b) and Schulze and Duddeck
(1964a). In both analytical models a homogenous ring does not
represent a segmental ring perfectly but delivers satisfactory inter-
nal forces for practical applications aiming to estimate range and
trend.
As mentioned before, the analysis of lining forces and deforma-
tion depends on the loading action and soil stress condition. If the
soil stresses would not depend on the lining deformation and there
would not be any interaction, it would be relatively easy to calcu-
late the stresses on lining segments and henceforth the axial forces,
bending moments and lining deformations. Bakker (2003) assumes
that the vertical deformation of tunnel which is needed to derive
vertical equilibrium of the system, gives a stress increase at tunnel
crown that is equal to the stress relieve at the tunnel invert. Even
though it is known that in practice that the stress relieve at the
invert might be higher than the stress increase at the crown.
Within this framework, Bakker (2003) calculates axial forces (N),
bending moments (M) and radial displacements (ur ) based on the
strength of materials:
ðrv þ rh Þ ðrv  rh Þ
N¼ rþ rcosð2UÞ ð1Þ
2 2
ðrv  rh Þ 2
M¼ r cosð2UÞ ð2Þ
4
ð rv þ rh Þ r 2 ð rv  rh Þ r 4
ur ¼   cosð2UÞ ð3Þ
2 E l Al 12 El I l
where rv and rh are respectively vertical and horizontal stresses at
the depth of tunnel center, r is the radius of tunnel, U is the
 C. Zhao et al. / Tunnelling and Underground Space Technology 64 (2017) 164–176
orientation angle representing the position of observation points
(see Fig. 1(a)), El Al and El Il are normal stiffness and flexural rigidity
of linings respectively.
It should be noted that Bakker’s method assumes that soil stres-
ses are known and their values are constant during tunnel excava-
tion. The lining segments are regarded to behave as a linear elastic
material. This simple form of lining forces and deformations gives a
straightforward insight into the mechanism of load-bearing behav-
ior, one can easily distinguish the key parameters that govern the
lining forces and radial displacements. Moreover, if the soil and lin-
ing properties are known, the preliminary design can also be con-
ducted. In this study, this simplified analytical solution is
presented as the basic model in the section of results and
discussions.
3. Numerical solution
3.1. 2D FEM model
The commercially available FE-code PLAXIS (version 2013) is
used in this paper to simulate the mechanized excavation of syn-
thetic shallow and deep tunnels. Due to the symmetry of the
model, only half of the model is discretized to optimize the compu-
tational cost of the numerical solution. The diameter of tunnel (D)
is assumed to be equal to 8.5 m for both shallow and deep tunnels.
By conducting a series of trial analyses, the appropriate size of the
model domain and appropriate mesh discretization were deter-
mined in a way that model responses are not influenced by bound-
ary conditions and domain discretization (Zhao et al., 2015). The
geometry and mesh discretization for shallow and deep tunnels
are respectively shown in Figs. 2(a) and 3(a). As can be seen, it is
assumed that the crown of shallow and deep tunnels are located
at the depth of 1D and 5D, respectively.
The concrete lining segments are modeled by using structural
triangular elements which obey isotropic linear elastic constitutive
behavior. The material properties for lining elements are given in
Table 2.
The interaction between soil and structure elements is simu-
lated with reference to in-built interface elements (Brinkgreve
et al., 2013). The mechanical behavior of the soil-structure contact
elements is modeled by Mohr-Coulomb (MC) model, the material
strength of contact elements is determined by considering a
Fig. 2. Geometry of shallow tunnel for (a) 2D model and (b) 3D model.
167
reduction of the adjacent soil strength. Here the shear strength of
the contact elements is 40% less than that of embedding soil.
In the 2D plain strain condition, the simulation of staged exca-
vation, face support pressure and backfill grouting are not straight-
forward. Since the soil/structure elements in the direction of
tunnel axis cannot be associated to the progressive excavation,
only soil excavation and lining installation are modeled in 2D sim-
ulation. Moreover, this tunneling problem is conducted in the nor-
mally consolidated and fully dry soil deposit. By doing so, it is
sufficient to conduct the comparison of lining forces and deforma-
tions obtained by analytical and numerical solutions. However, the
analytical solution appears to be insufficient for realistic structural
design due to the fact that many other aspects that play an impor-
tant role in structural design of linings, are neglected. These factors
have been taken into account in the 3D numerical model to study
their influences on the model responses.
3.2. 3D FEM model
As seen in Figs. 2(b) and 3(b), the 3D models are generally the
extrusion of 2D model in the direction of the tunnel axis. The width
of single precast concrete lining segments are assumed to be equal
to 1.5 m. Thus, TBM advances 1.5 m in each step of the staged exca-
vation. To avoid the collapse of the soil body in front of TBM, face
support pressure is applied (Anagnostou and Kovári, 1994; Broere,
2001). This face support is numerically simulated through a depth
dependent laterally distributed load on the front soil elements. The
value of face pressure is determined with respect to the vertical
stress arisen from the weight of soil deposit, and its gradient is
12 kPa/m related to the unit weight of bentonite suspension. The
grouting pressure is applied as depth dependent distributed load
at the tail of the TBM (Bezuijen and Talmon, 2008; Talmon and
Bezuijen, 2009) (see Fig. 4). The value of grouting pressure is rec-
ommended to be higher than the face pressure at the same depth
(Teil 5, 2012), in this research the value of grouting pressure is
defined by increasing the face pressure for 50 kPa at tunnel crown.
The grouting pressure linearly increases from tunnel crown to
invert with gradient of 15 kPa/m which is determined based on
the unit weight of the grout material.
The staged excavation process is modeled via a step-wise
procedure, where the progressive advancement of TBM is
achieved by a sequential of 1.5 m soil excavation in each step.
168 C. Zhao et al. / Tunnelling and Underground Space Technology 64 (2017) 164–176

Fig. 3. Geometry of deep tunnel for (a) 2D model and (b) 3D model.

Table 2 3.3. Hierarchical modeling strategy

Input parameters of the used linear elastic constitutive model of the plate elements
for representing the tunnel lining and TBM.
In numerical simulations of tunneling construction, the ground
Parameter Lining TBM Unit deformations and structural forces are affected by large numbers of
Thickness (d) 0.40 0.35 [m] factors due to the complexity of ground behavior and excavation
Elastic modulus (E) 30,000 210,000 [MPa] procedures. It is a challenge to choose an appropriate constitutive
Unit weight (c) 24 38 [kN/m3] model to reproduce the soil behavior by considering the realistic
Poisson’s ratio (m) 0.10 0.30 [–]
heterogeneity, anisotropy, sensitivity and loading history (OCR)
of soil deposit. Normally the soil is assumed to be homogeneous
and isotropic material, and both elastic and plastic behaviors are
In this sequence, the excavated soil is deactivated and the support taken into account. While the soil stiffnesses may vary due to dif-
pressures as well as the newly installed lining segments are ferent stress levels which are related to the confining pressure due
activated. By using the similar approach introduced in Zhao to depth of the soil deposit. In the plastic domain around the exca-
et al. (2015), the first six excavation steps represent the advance- vation zone, considering the hardening/softening behavior is more
ment of 9 m long TBM-shield, the shell elements are activated crucial. Additionally, soils show different stiffnesses in unloading/
with assigned TBM material. Afterward, the installation of lining reloading conditions compared to the primary loading (Schanz,
proceeds by assigning the lining material to the corresponding 1998). The non-linear elasto-plastic models were employed in
shell elements. Möller and Vermeer (2008), it was found that soil deformations
To observe the final lining forces and deformations after tunnel are highly influenced by the constitutive model, while structural
excavation, the synthetic monitoring section is defined as the forces are less influenced by the choice of the constitutive model.
vertical cross section which is 50 m far away from the boundary Vakili et al. (2014) investigated the influences of soil constitutive
(see Figs. 2(b) and 3(b)). Lining axial forces, bending moments model on the numerical results of mechanized tunneling and sug-
and radial displacements are extracted when TBM advances gested to be careful in employing the appropriate constitutive
for 110 m to ensure that present excavation process has no model for reliable results. On the other hand, the soil deformations
influence on the lining forces and deformations at the monitoring and structural forces are also highly influenced by the construction
section. procedures and the soil-structure interaction. TBM excavation may

Variant I Variant II Variant III

Soil Soil Soil

Lining Lining Shield tail Lining Shield tail

Distributed load

Fig. 4. Different variants used to model the tunnel construction process considering the shield excavation, backfill grouting and lining installation.
 C. Zhao et al. / Tunnelling and Underground Space Technology 64 (2017) 164–176 169

change the ground pressure distribution before lining installation, Table 3

and the grouting injection keeps the support firm and reduces the Input parameters of the used soil constitutive model  the Hardening Soil model.

soil deformations. This plays an important role in the accurate pre- Parameter Description Value Unit
diction of system behavior. In this study, applying the hierarchical u 0
Friction angle 35 [°]
modeling strategy aims to evaluate the effects of constitutive mod- w0 Dilatancy angle 5 [°]
els, construction procedures and the soil-structure interaction. c0 Cohesion 10 [kN/m2]
Eref Secant stiffness in triaxial test 35,000 [kN/m2]
50

3.3.1. Hierarchical constitutive models for granular material Eref Tangent stiffness for oedometer loading 35,000 [kN/m2]
oed

According to Kim and Eisenstein (2006), a good estimation of Eref Unloading-reloading stiffness 100,000 [kN/m2]
ur

lining forces and deformations should take into account the elastic
deformation as well as plastic behavior of the soil. In this study, an
elasto-plastic constitutive model is employed to reproduce the soil
behavior. However, in the existing analytical solutions, the soil is
assumed to be elastic material for the purpose of simplicity. In
order to evaluate the applicability of analytical methods, an elastic
constitutive model is also employed in FE-model. To be specific,
three hierarchical constitutive models are employed. Namely (a)
Linear elastic model with constant stiffness (LE), (b) Modified lin-
ear elastic model with stress dependent stiffness (LEM), and (c)
Hardening soil model (HS). Generally, LE model is sufficient to con-
duct the comparison of numerical model and the analytical solu-
tions. Since LE model is not sufficient to capture the realistic
plastic behavior of soil, employing a sophisticated constitutive
model such as HS model is essential. HS model is used for different
types of soils in the framework of isotropic hardening plasticity.
Moreover, LEM model as an intermediate model, is employed to
consider the fact that soils have larger elastic stiffnesses under
higher stress levels.
In LE model, there are only two constitutive parameters, namely
Young’s modulus E and Poisson’s ratio m. In this study, the values of
elastic parameters are assumed to be E = 35,000 kN/m2 and
m ¼ 0:3.
In HS model, the shear failure obeys Mohr-Coulomb (MC) fail-
ure criterion and plasticity is governed by double hardening rules
on cone-cap surfaces (Schanz, 1998; Schanz et al., 1999). The soil
strength is described by three non-linear stiffnesses corresponding
to the stress levels as: the primary loading stiffness (Eref
50 ), the
oedometer loading stiffness (Eref
oed ) and purely elastic unloading/
reloading stiffness (Eref
ur ). Vakili et al. (2013) indicated that the best
consistency between stress paths obtained by MC and HS models
in triaxial soil element test is observed when the Young’s modulus
in MC model is assumed to be equal to Eref
50 in HS model. As both MC
and LE models have linear stress-strain relationship in the elastic
domain, it is supposed that the stress paths obtained by LE and
HS models are comparable when elastic parameter of LE model is
same as Eref ref
50 in HS model. Within this framework, the value of E50
in HS model is assumed to be equal to the value of elasticity
parameter E in LE model, soil properties for HS model are given
in Table 3.
As mentioned before, in HS model, the soil stiffnesses
(Eoed ; E50 ; Eur ) are defined to be stress dependent. Accordingly, the
stiffness increases with depth where the confining stress is higher.
Such stress dependency of the stiffness is well proven in the liter-
ature. Several studies have indicated that the tunneling problems
mainly deal with unloading-reloading stiffness (Eur ) as the main
deformations are imposed to the system due to soil excavation
(unloading) or the grout injection and face stabilizing process
(reloading). Since the linear elastic model includes a single stiffness
for all loading conditions, the elastic modulus in modified linear
elastic model (LEM) is assumed to be correlated to Eur in hardening
soil at each depth. Accordingly, the profile of stiffness (Eur ) over
depth (z) is retrieved at geostatic stress condition (before the exca-
vation) for HS model and assigned to the elastic soil stiffness (E) as
a linear function of depth (z), namely E ¼ f ðzÞ. After that, this depth
dependent soil elastic stiffness is defined in the LEM model.
pref Reference stress 100 [kN/m2]
m Exponent power 0.7 [–]
Rf Failure ratio 0.9 [–]
mur Poisson ratio 0.2 [–]
c Soil unit weight 17 [kN/m3]
3.3.2. Hierarchical tunnel construction procedures
Kasper and Meschke (2004) reviewed the development of
numerical simulation of mechanized tunneling and generated a
3D model which takes into account the soil domain, ground water,
shield machine, face support, hydraulic jacks, tail void grouting,
tunnel lining installation and soil-structure interaction. Since lin-
ing forces and deformations are affected by each component of
tunnel construction method, uncertainty of model response
increases inevitably by considering all these components. Möller
and Vermeer (2008) found that appropriate method of modeling
grout injection yields more realistic soil deformation and structural
forces. Within this framework, the 3D FE-model in this study takes
into account the sequential excavation process, face support, grout
injection, lining installation and soil-structure interaction.
It should be mentioned that many researchers refer to the
importance of 3D discretization for tunneling that accounts for
construction details, however, analytical solutions or 2D simula-
tions of tunneling process are still the most popular methods used
for design purpose. Thus, comparison of the results of these meth-
ods is essential to have a better insight into the realistic design
concept. As seen in Fig. 4, three hierarchical construction proce-
dures are described as follows: In variant I, only soil excavation
and lining installation are simulated. This simplified numerical
simulation is comparable to the conventional open face tunneling.
In plain strain condition, the model responses of analytical and
numerical solutions can be evaluated under equivalent initial con-
ditions and boundary conditions. While in 3D simulations, it is
available to study the effect of staged excavation. In variant II, lin-
ing segments are installed right after the TBM excavation. This kind
of numerical model is similar to the present mechanized tunneling
technique. One can study the influence of stress release induced by
unloading (before lining installation) on the lining forces and
deformations. In variant III, tunneling process with TBM excava-
tion, grout injection and final lining installation are considered.
This is the most sophisticated construction method in this paper,
which aims to model the tunneling process as realistic as possible,
where the influence of backfill grouting on lining forces and
deformations can be analyzed. Moreover, Table 4 shows the 3D
simulation scenarios conducted in the study which aim to evaluate
the effects of hierarchical construction procedures. For the 2D
Table 4
Combination of different constitutive models and construction procedures.
Construction procedure Constitutive model
LE LEM HS
Variant I LE-1 LEM-1 HS-1
Variant II LE-2 LEM-2 HS-2
Variant III LE-3 LEM-3 HS-3
170 C. Zhao et al. / Tunnelling and Underground Space Technology 64 (2017) 164–176
FE-model, as the staged excavation process cannot be simulated in
correct manner, only variant I is employed.
3.3.3. Hierarchical soil-lining contact properties
Since lining forces and deformations highly depend on the
surrounding soil stresses, stress redistribution induced by
tunnel construction plays an important role in the structural
design of linings. The soil-lining interaction can be illustrated
by the concept of ground response curves. This concept was
initiated by Fenner (1938) to illustrate the soil pressure on
lining as a function of inward deformation. At present the soil
response curves are mainly used to demonstrate the soil
pressures on lining as a function of installation procedures
and soil-lining relative stiffness (Möller and Vermeer, 2008).
In open face tunneling, the soil stress distribution is varying
with the progressive inward displacements, when the lining is
installed, further cutting and creep behavior make the lining
deform until a quasi final state where the support pressures
of lining meet the released soil stresses. Normally the soil
deformations are relatively large in open face tunneling and
yield small soil pressures. Conversely in closed face tunneling,
support is firm and there is little soil deformation which leads
to high soil pressures on tunnel lining.
In the numerical model, the soil-structure contact elements
are composed of 12-node interface elements. Interface elements
consist of pair of nodes, compatible with side of soil element or
plate element. It is available to have differential deformations
of the plate element and surrounding soil element. As the TBM/
lining and soil have different roughness, the strength properties
of interface can be manually defined to describe the roughness
in soil-structure interaction. To study the influence of interface
properties on lining forces and deformations, the 2D shallow
tunnel model (2D-HS-1) is taken as an example. MC model is
employed with different elastic and plastic parameters to simu-
late the soil-structure interaction. Young’s modulus (E) controls
the elastic deformation, cohesion (c0 ) friction angle (u0 ) dominate
the shear behavior of contact elements. It is noted that dilatancy
is not considered in the interface behavior. Accounting for the
possible elastic or plastic deformation, different scenarios are
shown in Table 5. No reduction means the contact elements have
the same strength properties as the surrounding soil (for elastic-
ity parameter E ¼ Eref
50 ). However, the relative deformation and
slippage between soil and lining elements are allowed. Only
plastic parameters are reduced in scenario 2, both elastic and
plastic parameters are reduced in scenario 3. In case of no contact
element (scenario 4), the plate and the surrounding soil elements
are fully connected and no contact element exists. The effect of
contact elements is elaborated by comparing the model responses
of scenarios 3 and 4. Based on the comparison of lining behavior
of scenarios 1 and 2, the effect of plastic parameters can be
obtained. By analogy, the influence of elasticity parameters on
lining forces can be evaluated by conducting simulations for
scenarios 2 and 3. It is should be noted in this research, if
there is no special description on the soil-structure interaction,
interface type of scenario 3 is applied.
Table 5
Combination of different interface properties (m = const. = 0.2).
Interface type Elastic parameter Plastic parameters
E [kN/m2] c0 [kN/m2] u0 [ ]
Scenario 1 – No reduction 35,000 10 35
Scenario 2 – Reduced c0 ; u0 35,000 6 22.8
Scenario 3 – Reduced E; c0 ; u0 21,000 6 22.8
Scenario 4 – No contact element – – –
4. Results and discussions
4.1. Analytical and simplified numerical solutions
Lining axial forces, bending moments and radial displacements
of shallow and deep tunnels calculated by analytical methods are
shown in Figs. 5 and 6. The basic model is conducted based on
the method proposed by Bakker (2003). The bedding model
employs the analytical method given by Schulze and Duddeck
(1964a), the continuum model applies the analytical solution pro-
posed by Ahrens et al. (1982). To evaluate the difference between
numerical and analytical models in the same sophistication
category, the simplest 2D and 3D models based on linear elastic
model that does address the tunnel construction details are
chosen in this section. This comparison is shown in Figs. 5 and 6.
Furthermore, the analytical deviation values of lining forces and
deformations in accordance with numerical scenario 3D-LE-1 are
given in Tables 6 and 7. The deviation of model X in accordance
with model Y is defined as: Dev. (i)= ðiðXÞ  iðYÞÞ=iðYÞ, where i is
{N max ; M max ; Mmin ; ur;max ; ur;min }.
The basic model neglects the interaction between soil and lin-
ing, the soil stresses acting on the lining do not change with respect
to the soil deformation. Consequently, according to the comparison
of 2D FE-model (2D-LE-1) and analytical methods, the basic model
generally over-predicts the axial forces, bending moments and
radial displacements for both shallow and deep tunnels. While
the predictions of bending moments and radial displacements
obtained by the bedding model and continuum model are compa-
rable. For the axial forces, the bedding model provides higher val-
ues due to omitting the tangential bond in the applied bedding
model. Thus, the axial forces obtained by bedding model could
be the upper limit of analytical results (Schulze and Duddeck,
1964a; Duddeck and Erdmann, 1985). The discrepancy between
the analytical and numerical results is mainly due to assuming a
constant equally distributed load in the analytical model (see
Fig. 1(a)). Therefore, in the upper section of the lining the applied
pressures are higher than those in the numerical model, while they
are smaller in the lower section. This results in larger axial forces
are obtained via FE-model in the lower section of the lining seg-
ments (see Fig. 5(a)). Nevertheless, in case of deep tunnels, the
bedding model predicts larger axial forces compared to those of
2D-LE-1 (see Fig. 6(a)). This is because stress release induced by
unloading, which is neglected in analytical models. The effect of
stresses release is more significant at deeper levels of soil domain.
According to Fig. 5(c), inward deformations at both crown and
invert of lining are obtained for all analytical methods. However,
uplift of the tunnel structure is observed in the numerical simula-
tion, this is due to high stress release rate during excavation and
low surcharge above the tunnel, which can be more realistic for
shallow tunnels. On the contrary, Fig. 6(c) shows no uplift of the
tunnel structure due to higher surcharge above the tunnel. It can
be concluded that the differences between analytical and numeri-
cal solutions are mainly related to stress distribution and deforma-
tion mode around the tunnel.
Additionally, the comparison between the level of axial forces in
the lining segments for shallow and deep tunnels (see Figs. 5(a)
and 6(a)) indicates that the axial forces depend upon (a) the tunnel
confining pressures that mainly induce compression in the lining
and (b) the deformations in lining element that can induce com-
pression or tension. As the lining is subjected simultaneously to
both of these components, the superposition is valid due to the
elastic behavior of the lining elements. Although a relative tensile
component of load can be generated in the lining when it is sub-
jected to uplift, the compression in the lining segments induced
by the confining stress is much more dominant in this study.
Accordingly, the resultant component of the axial force is still
 C. Zhao et al. / Tunnelling and Underground Space Technology 64 (2017) 164–176 171

Fig. 5. Analytical and numerical (Variant I) lining forces and deformations for shallow tunnels; (a) axial forces N, (b) bending moments M, (c) radial displacements ur .

Fig. 6. Analytical and numerical (Variant I) lining forces and deformations for deep tunnels; (a) axial forces N, (b) bending moments M, (c) radial displacements ur .

Table 6
Comparison of the analytical and numerical results for shallow tunnels.

Models Deviation values for each model

N max (%) M min (%) Mmax (%) urmin (%) urmax (%)
Basic model 10.7 >100 >100 32.7 26.3
Bedding model 14.7 41.1 24.7 77.6 72.9
Continuum model 27.6 63.4 63.0 91.4 83.4
2D-LE-1 0.3 11.8 26.6 2.1 8.2
3D-LE-1 0 0 0 0 0

Table 7
Comparison of the analytical and numerical results for deep tunnels.

Models Deviation values for each model

N max (%) M min (%) Mmax (%) urmin (%) urmax (%)
Basic model 9.4 >100 >100 >100 >100
Bedding model 13.1 69.1 >100 18.3 50.6
Continuum model 19.8 50.4 55.9 33.9 33.3
2D-LE-1 15.9 95.9 >100 16.3 71.9
3D-LE-1 0 0 0 0 0
172 C. Zhao et al. / Tunnelling and Underground Space Technology 64 (2017) 164–176
compressive (positive). Nevertheless, the tensile force component
induced by the tunnel uplift at the tunnel crown and invert causes
a significant decrease in the compression.
According to the Table 6, the 2D and 3D numerical lining forces
and deformations are comparable, especially the maximum axial
forces are almost identical. While the bedding model and contin-
uum model under-estimate the lining forces and deformations.
Based on the deviation values, the bedding model is more appro-
priate than continuum model for lining structural design of shal-
low tunnels. For the deep tunnels, Table 7 shows that there is
huge discrepancy between 2D and 3D numerical results. This is
due to the fact that three dimensional arching effect is more dom-
inant in deep tunnels than in shallow tunnels. The detailed expla-
nations will be discussed in Section 4.2. According to the
comparison of analytical and numerical results, although the axial
forces of bedding model is closer to the 3D numerical results than
the continuum model, neither bedding model nor continuum
model can adequately predict the lining forces and deformations
in deep tunnels. Furthermore, it should be noted that, the maxi-
mum axial forces obtained by continuum model are comparable
to those of 3D FE-model (3D-LE-1) in both shallow and deep tun-
nels, the deviation value is around 10%. Therefore, the basic model
can be applied to have a preliminary view of the lining axial forces.
4.2. Effect of 2D and 3D simulations with various constitutive models
Figs. 7–9 show the numerical model responses with the three
hierarchical soil constitutive models introduced above. Only soil
excavation and lining installation are simulated in the numerical
model (Variant I). The 3D model responses calculated with HS
model (3D-HS-1) are taken as the reference lining behavior which
is used to calculate the deviation values for the other models.
This construction procedure (Variant I) indicates that unloading
process dominates the systems mechanical behavior. This is con-
sistent with the result that lining forces and radial displacements
are almost same in case of employing LEM and HS models in which
same elasticity moduli are employed. When using LE model, lower
stiffnesses around the excavation domain induce larger soil defor-
mations. Consequently, radial deformations of lining segments are
more apparent in case of employing LE model (see Fig. 9).
Tunnel excavation changes the primary stress field into a three
dimensional pattern at the tunnel face. When the tunnel head
advances far away from the monitoring section, the stress field
returns to an essentially two dimensional system. To study the
Fig. 7. Effect of constitutive model on lining axial forces obtained from 2D and 3D analyses (Variant I); (a) 2D shallow tunnels, (b) 3D shallow tunnels, (c) 2D deep tunnels, (d)
3D deep tunnels.
influence of simplifying the 3D tunneling problem into 2D plain
strain model, comparison between the 2D and 3D model responses
is presented in Figs. 7–9. As seen, the 2D model responses agree
reasonably with the 3D model output for shallow tunnel. Axial
forces and bending moments predicted by the bedding model are
very close to the more realistic 3D model response (3D-HS-1),
the maximum deviation value is about 10%. This indicates that
the bedding model is feasible in the preliminary structural design
of lining segments. For deep tunnels, in either 2D or 3D models,
the model responses are almost same when employing HS or
LEM models, while LE model leads to higher axial forces, bending
moments and radial displacements due to lower soil stiffnesses.
This proves that the elastic parameter is the key constitutive
parameter that governs the lining forces and deformations. In
other words, the analytical solutions can be trustful if a modifica-
tion on elastic modulus is applied.
According to Fig. 7, for the simplified construction method
(Variant I), the axial forces distribution pattern obtained in 3D
model is rotated slightly to the lower part of the tunnel, the 3D
FE-model predicts higher axial forces and lower bending moments
compared to those of 2D model. As the 2D plain strain model over-
estimates the arching mechanism, the higher lining forces in 3D
model might be due to more realistic three dimensional soil arch-
ing effect. Zhao et al. (2015) also found that 2D simulation under-
estimates the lining axial forces compared to those of 3D model,
and attributed this phenomenon to the arching effect. To be more
specific, there is no arching action in the soil domain before exca-
vation. However, it becomes gradually influential after soil excava-
tion and lining installation. In this process, the stress distribution
remains unaffected. Thus, less soil deformation around the newly
installed lining segments is induced which leads to larger axial
forces and smaller bending moments. After tunnel construction,
the system reaches an equilibrium state. However, the realistic
step-wise arching the progresses with tunnel advancement is sim-
ulated as an immediate artificial support in the 2D FE-model
(Möller and Vermeer, 2008). It means there would be same soil
deformations and stress state variation along the tunnel, which
does not depend on the progressive excavation. Within this frame-
work, the 3D model is more realistic to simulate tunnel excavation
via the sequential excavation process. In other words, the 2D FE-
model is inadequate for the structural design of the deep tunnel
without simulating the arching effect in right manner. Although
the analytical bedding solution is based on the 2D idealization,
the predicted lining axial forces at tunnel’s side are comparable
 C. Zhao et al. / Tunnelling and Underground Space Technology 64 (2017) 164–176 173

Fig. 8. Effect of constitutive model on lining bending moments obtained from 2D and 3D analyses (Variant I); (a) 2D shallow tunnels, (b) 3D shallow tunnels, (c) 2D deep
tunnels, (d) 3D deep tunnels.

Fig. 9. Effect of constitutive model on lining radial displacements obtained from 2D and 3D analyses (Variant I); (a) 2D shallow tunnels, (b) 3D shallow tunnels, (c) 2D deep
tunnels, (d) 3D deep tunnels.

to the 3D model responses. This can be interrelated to the insignif-
icant influence of arching on the tunnel sides. Despite of the good
agreement on maximum axial forces, the bedding model is conser-
vative in predicting the bending moments and radial
displacements.
4.3. Effects of staged construction procedure
As mentioned before, reliable prediction of lining forces and
deformations should take into account both the elastic and plastic
ground deformation. In order to study the influence of tunnel con-
struction procedure on the model responses, elasto-plastic HS
model is employed in the FE-model. The applied construction
methods have been introduced in Section 3.3.2, and their influence
on lining forces and deformations is shown in Figs. 10 and 11. The
deviation values of different methods are given in Tables 8 and 9,
here the model responses obtained by HS model in 3D simulation
(3D-HS-3) are assumed to be the reference values. Lining forces
and deformations highly depend on the applied construction pro-
cedures. In 3D numerical simulation of shallow tunnel, when the
TBM is simulated in excavation step prior to an installation of lin-
ings (Variant II), both axial forces and bending moments increase
in comparison to Variant I. This is because the stiffer TBM shield
induces smaller deformation and higher loads on the TBM shield.
Moreover, when the backfill grouting is applied after shield exca-
vation and before lining installation (Variant III), less lining axial
forces are generated. This is due to the fact that grouting pressure
is applied on the soil elements, the soil body can deform to reach
an equilibrium. In this process, a slight stress release/redistribution
takes place that results in the difference of lining forces and defor-
mation using difference construction procedures. However, the
influence of this stress redistribution/release around the shield
can be more pronounced in terms of ground movements due to
tunneling. It can be concluded that Variant III not only accounts
for the details of the construction, it more realistically captures
the stress release around the shield.
As seen in Figs. 10(b) and 11(b), by simplifying construction
assumptions (Variants I and II) the bending moments are signifi-
cantly under-predicted for both shallow and deep tunnels. Despite
of using advanced constitutive model in 2D and 3D models along
with Variant I, the accuracy of the results are in the range of ana-
lytical bedding model for both shallow and deep tunnels. However,
the results obtained from Variant II are in a better agreement with
most sophisticated Variant III. According to Table 8, although the
174 C. Zhao et al. / Tunnelling and Underground Space Technology 64 (2017) 164–176

Fig. 10. Effect of staged construction on lining forces and deformations for shallow tunnels; (a) axial forces N, (b) bending moments M, (c) radial displacements ur .

Fig. 11. Effect of staged construction on lining forces and deformations for deep tunnels; (a) axial forces N, (b) bending moments M, (c) radial displacements ur .

Table 8
Comparison of lining forces and displacements in shallow tunnels for different numerical construction variants.

Models Deviation values for each model

N max (%) M min (%) Mmax (%) urmin (%) urmax (%)
Bedding model 4.6 69.0 65.6 5.0 16.7
2D-HS-1 4.5 61.0 59.1 54.4 16.4
3D-HS-1 2.7 69.2 66.7 52.1 19.4
3D-HS-2 7.5 14.5 19.7 15.9 25.0
3D-HS-3 0 0 0 0 0

Table 9
Comparison of lining forces and displacements in deep tunnels for different numerical construction variants.

Models Deviation values for each model

N max (%) M min (%) Mmax (%) urmin (%) urmax (%)
Bedding model 4.4 3.2 26.7 >100 >100
2D-HS-1 14.7 60.0 52.2 65.3 >100
3D-HS-1 1.3 74.2 73.9 40.1 98.5
3D-HS-2 7.9 15.4 13.1 6.1 13.9
3D-HS-3 0 0 0 0 0
 C. Zhao et al. / Tunnelling and Underground Space Technology 64 (2017) 164–176
Table 10
Comparison of lining forces and displacements in shallow tunnels for different soil-structure contact properties (2D-HS-1).
Models
N max (%) M min (%)
Bedding model 3.9 28.0
Scenario 1 – No reduction 4.2 5.3
Scenario 2 – Reduced c; u 0.7 2.1
Scenario 3 – Reduced E; c; u 0 0
Scenario 4 – No contact element 7.1 15.4
results obtained from 3D model in conjunction with Variant II
deviate about 20% from the most sophisticated 3D-HS-3 model, it
still offers far better results for lining forces and deformations in
shallow tunnels in comparison with the simple numerical and ana-
lytical models. The differences between analytical and 2D solutions
with 3D model with the same level of simplification can be attrib-
uted to more realistic arching mechanism in three dimensional
models while this influence can be more pronounced for deep
tunnels.
4.4. Effects of soil-lining contact properties
Table 10 shows the influence of contact properties on lining
forces and radial displacements. Description of different scenarios
has been given in Section 3.3.3. HS model is employed in numerical
solution and lining installation (Variant I) is simulated. According
to the description of FE-model in Section 3.1, the stiffness and
shear strength of soil-lining contact elements are assumed to be
40% less than that of the surrounding soil (scenario 3). Within this
framework, the model responses on the basis of scenario 3
(reduced stiffness and shear strength) are defined as reference val-
ues. It is excepted that lining forces and deformations are influ-
enced by both elastic and plastic parameters.
When the soil and lining elements are assumed to be fully con-
nected with no extra contact element (scenario 4), the radial dis-
placements at tunnel side is under-estimated about 15%.
Consequently the maximum axial forces at tunnel side are over-
estimated about 7%. As in scenarios 2 and 3 the contact element
between the soil and lining is assumed to have the same shear
strength as the surrounding soil, almost same lining forces and
deformations are obtained. Accordingly, it can be deduced that
plastic parameters dominate the system behavior when relative
movement between the soil and structure is allowed. Comparing
scenarios 2 and 3 indicates that the elastic modulus plays a less
significant role in the variation of the forces and deformations in
the lining elements.
Additionally, since both scenario 4 (no extra contact element)
and analytical bedding model do not account for the realistic dis-
placements and mechanical properties between the soil and lining,
the forces obtained in the lining elements from these solutions are
in the same level of accuracy. Accordingly, the soil-lining interac-
tion is essential to be taken into account to provide reliable prog-
nosis of lining forces and deformations.
5. Summary and conclusions
Both analytical methods and numerical simulations are carried
out in this paper to predict the lining forces and deformations for
tunnel structural design. Different commonly used analytical solu-
tions are compared to the numerical solutions for equivalent
boundary conditions, initial conditions and identical material
properties. Regarding the capabilities of a numerical model to con-
sider the tunneling process as realistic as possible, the important
aspects which cannot be captured by the analytical methods are
studied. That is, the influence of hierarchical constitutive models,
175
Deviation values for each model
Mmax (%) urmin (%) urmax (%)
18.3 41.8 34.8
2.7 6.2 4.7
0 5.5 2.7
0 0 0
10.2 16.8 13.8
staged construction procedure, and soil-structure interaction prop-
erties. Additionally, the model responses calculated in plain strain
condition are compared to the results of 3D numerical models. The
general conclusions can be drawn as follows:
1. The basic model (Bakker’s method) appears to be too conserva-
tive, the bedding model is more appropriate than the contin-
uum model for the structural design of both shallow and deep
tunnels. Especially the prediction of maximum axial forces is
reliable even if the soil is taken as a linear elastic material.
The discrepancy between analytical solutions and numerical
solutions is mainly due to the assumption of analytical method
that depth dependent soil pressures are replaced by a constant
equally distributed load.
2. The soil elasticity modulus is the key parameter of constitutive
model that governs the lining forces and deformations. Harden-
ing soil model and linear elastic model with depth dependent
stiffness provide almost the same lining forces and
deformations.
3. Construction method plays a significant role in structural
design of linings. The 3D numerical simulation is essential to
model the three dimensional arching effect in right manner,
which can not be adequately captured in plain strain
conditions.
4. It is necessary to model the soil-structure interaction for the
purpose of realistic prediction of lining behavior. Plastic param-
eters are more influential than elastic parameters when the
contact element between soil and lining allows relative
displacements.
5. There is no crucial disagreement between the 2D and 3D
numerical model responses of the shallow tunnels when the
construction details are not taken into consideration. However,
in case of the deep tunnels, the 2D model underestimates the
lining axial forces to an extensive degree due to inappropriate
arching mechanism in 2D plain strain model.
Acknowledgments
This research has been supported by the German Research
Foundation (DFG) through the Collaborative Research Center (SFB
837), and the first author is sponsored through a scholarship by
China Scholarship Council (CSC). These supports are gratefully
acknowledged. Moreover, the authors appreciate the discussion
with Mr. Thomas Putke on the analytical solutions of tunnel lining
forces.
References
Ahrens, H., Lindner, E., Lux, K., 1982. Zur Dimensionierung von Tunnelausbauten
nach den Empfehlungen zur Berechnung von Tunneln im Lockergestein. Die
Bautech. 8, 260–273.
Anagnostou, G., Kovári, K., 1994. The face stability of slurry-shield-driven tunnels.
Tunnel. Undergr. Space Technol. 9 (2), 165–174.
Arnau, O., Molins, C., 2011. Experimental and analytical study of the structural
response of segmental tunnel linings based on an in situ loading test. Part 2:
Numerical simulation. Tunnel. Undergr. Space Technol. 26 (6), 778–788.
176 C. Zhao et al. / Tunnelling and Underground Space Technology 64 (2017) 164–176
Bakker, K., 2003. Structural design of linings for bored tunnels in soft ground. Heron
48 (1), 33–63.
Behnen, G., Nevrly, T., Fischer, O., 2015. Soil-structure interaction in tunnel lining
analyses. Geotechnik 38 (2), 96–106.
Bezuijen, A., Talmon, A., 2008. Processes around a TBM. In: Proceedings of the 6th
International Symposium on Geotechnical Aspects of Underground
Construction in Soft Ground, p. 48–56.
Brinkgreve R., Engin E., Swolfs W., 2013. Reference manual. Delft, Netherlands.
Broere W., 2001. Tunnel face stability & New CPT applications. Ph.D. thesis, Delft
University of Technology.
Bull, A., 1944. Stresses in the linings of shield driven tunnels. J. Soil Mech. Found.
Div. ASCE, 1363–1394.
Duddeck, H., Erdmann, J., 1985. On structural design models for tunnels in soft soil.
Undergr. Space 9, 246–259.
El-Nahhas, F., El-Kadi, F., Ahmed, A., 1992. Interaction of tunnel linings and soft
ground. Tunnel. Undergr. Space Technol. 7 (1), 33–43.
Engelbreth, K., 1961. Correspondence on Morgan, H.D., A contribution to the
analysis of stress in a circular tunnel. Géotechnique 11 (3), 246–248.
Fenner, R., 1938. Untersuchung zur Erkenntnis des Gebirgsdrucks. Glückauf 74 (32),
681–695.
Galli, G., Grimaldi, A., Leonardi, A., 2004. Three-dimensional modelling of tunnel
excavation and lining. Comput. Geotech. 31 (3), 171–183.
Hudoba, I., 1997. Contribution to static analysis of load-bearing concrete tunnel
lining built by shield-driven technology. Tunnel. Undergr. Space Technol. 12 (1),
55–58.
ITA, 1988. Guidelines for the design of tunnels. Tunnel. Undergr. Space Technol. 3
(3), 237–249.
Kasper, T., Meschke, G., 2004. A 3d finite element simulation model for TBM
tunnelling in soft ground. Int. J. Numer. Anal. Methods Geomech. 28, 1441–
1460.
Kim, H., Eisenstein, Z., 2006. Prediction of tunnel lining loads using correction
factors. Eng. Geol. 85 (34), 302–312.
Molins, C., Arnau, O., 2011. Experimental and analytical study of the structural
response of segmental tunnel linings based on an in situ loading test. Part 1:
Test configuration and execution. Tunnel. Undergr. Space Technol. 26 (6), 764–
777.
Möller, S., Vermeer, P., 2008. On numerical simulation of tunnel installation. Tunnel.
Undergr. Space Technol. 23 (4), 461–475.
Oreste, P., 2007. A numerical approach to the hyperstatic reaction method for the
dimensioning of tunnel supports. Tunnel. Undergr. Space Technol. 22 (2), 185–
205.
Schanz, T., 1998. Zur Modellierung des mechanischen Verhaltens von
Reibungsmaterialien. Habilitationsschrift, Mitteilung 45 des Instituts für
Geotechnik, Universität Stuttgart.
Schanz, T., Vermeer, P., Bonnier, P., 1999. The hardening soil model: formulation
and verification. In: Proceedings of 1st International PLAXIS Symposium on
Beyond 2000 in Computational Geotechnics, Balkema, p. 281–296.
Schmid, H., 1926. Statische Probleme des Tunnel- und Druckstollenbaus.
Schulze H., Duddeck H., 1964a. Statische Berechnung schildvorgetriebener Tunnel.
In: Festschrift: Beton- und Monierbau Aktien-Gesellschafte, p. 1889–1964.
Schulze, H., Duddeck, H., 1964b. Stresses in shield driven tunnels. Beton-und
Stahlbetonbau 8, 169–175.
Talmon, A., Bezuijen, A., 2009. Simulating the consolidation of TBM grout at
Noordplaspolder. Tunnel. Undergr. Space Technol. 24, 493–499.
Teil 5, 2012. Tunnelbau. In: Zusätzliche technische Vertragsbedingungen und
Richtlinien für Ingenieurbauten, Bundesanstalt für Strassenwesen.
Vakili, K., Barciaga, T., Lavasan, A., Schanz, T., 2013. A practical approach to
constitutive models for the analysis of geotechnical problems. In: Proceedings
of 3rd International Symposium on Computational Geomechanics (ComGeo III),
p. 738–749.
Vakili, K., Lavasan, A., Datcheva, M., Schanz, T., 2014. The influence of constitutive
modeling in the numerical assessment of mechanized tunneling. In:
Proceedings of 8th European Conference on Numerical Methods in
Geotechnical Engineering.
Voellmy, A., 1937. Eingebettete Rohre. Ph.D. thesis, ETH Zürich.
Windels, R., 1967. Kreisring im elastischen Kontinuum. Der Bauingenieur 42, 429–
439.
Zhang, D., Huang, H., Hu, Q., Jiang, F., 2015. Influence of multi-layered soil formation
on shield tunnel lining behavior. Tunnel. Undergr. Space Technol. 47, 123–135.
Zhao, C., Lavasan, A.A., Barciaga, T., Zarev, V., Datcheva, M., Schanz, T., 2015. Model
validation and calibration via back analysis for mechanized tunnel simulations
– the Western Scheldt tunnel case. Comput. Geotech. 69, 601–614.