Available online at www.sciencedirect.

com

ScienceDirect
Underground Space 3 (2018) 45–60
www.elsevier.com/locate/undsp

A holistic approach for the investigation of lining response

to mechanized tunneling induced construction loadings
Vojtech Ernst Gall a,⇑, Ahmed Marwan a, Mario Smarslik b, Markus Obel b, Peter Mark b,
Günther Meschke a
a
Institute for Structural Mechanics, Ruhr University Bochum, Bochum, Germany
b
Institute of Concrete Structures, Ruhr University Bochum, Bochum, Germany

Received 7 August 2017; received in revised form 20 November 2017; accepted 5 January 2018
Available online 16 February 2018

Abstract

Design methods for segmental tunnel linings used in mechanized tunnel constructions typically employ numerical bedded beam mod-
els and/or classical analytical solutions for the determination of structural forces (i.e. moments and shear and axial forces) and simple
load spreading assumptions for the design of the reinforcement in joint areas. However efficient such methods may be, many physical
details are often overlooked and/or oversimplified in the process of reducing the actual structure to a structural beam model, e.g. ana-
lytically derived loadings are employed, the grouting and ground reactions are reduced to a spring bedding, and the confinement due to
grouting at the longitudinal joint is largely not considered in reinforcement design. Such a design process is not able to account for, or
predict, the susceptibility of tunnel linings to often observed damages that, although they may not be structurally relevant, lead to ser-
viceability or durability issues, such as crack development or chipping at the segment corners. Numerical methods, such as the Finite
Element Method, provide an opportunity to model the segmental tunnel lining and its response to the entire TBM construction process
and to explicitly model the crack development within individual segments using modern methods to model the discontinuities in struc-
tures. In this contribution, a holistic modeling procedure for the representation of the tunnel lining within the tunneling process is pro-
posed and compared to traditional lining models. A 3D process oriented Finite Element model is used to calculate the predicted forces on
the tunnel lining and the obtained results are compared with those generated by traditional methods. Subsequently, the predicted defor-
mations are then transferred to a detailed segment model in which the nonlinear response of the segment at the longitudinal joint is mod-
eled using an interface element based approach to simulate concrete cracking.
Ó 2018 Tongji University and Tongji University Press. Production and hosting by Elsevier B.V. on behalf of Owner. This is an open access article
under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

Keywords: Mechanized tunneling; Segmental tunnel linings; Finite element method

1. Introduction problem, whether continuum based (Ahrens et al., 1982) or

beam based (Schulze and Duddeck, 1964a), may also be
The most commonly used structural models for deter- used for this purpose, and provide a valuable reference
mining structural ring forces in tunnel linings in modern point for the evaluation of obtained structural forces using
tunnel engineering practice are numerical bedded beam other methods. Analytical models often used to generate a
models (FHWA, 2009; German Tunnelling Committee preliminary design concept (Blom, 2002). However,
(DAUB), 2013; Städing, 2007). Analytical solutions to this because pronounced simplifications are involved in numer-
ical or analytical beam models, many common models
neglect or are unable to account for certain design details,
⇑ Corresponding author. such as the allowable crack width in the lining segments.
E-mail address: vojtech.gall@rub.de (V.E. Gall).

https://doi.org/10.1016/j.undsp.2018.01.001
2467-9674/Ó 2018 Tongji University and Tongji University Press. Production and hosting by Elsevier B.V. on behalf of Owner.
This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
46 V.E. Gall et al. / Underground Space 3 (2018) 45–60

These simplifications made in the construction of structural Within the context of this contribution two goals are
models for tunnel linings occur primarily on two levels. addressed. The aim is to firstly examine the validity of
First, the interactions between the lining and the surround- using a full scale 3D process oriented tunnel model for
ing structures are simplified. In bedded beam models extracting the structural forces needed for tunnel lining
(Schulze and Duddeck, 1964a), the structural interaction design, and, secondly, to introduce a new holistic modeling
between the tunnel lining and the grout and soil is reduced concept in which the segment response to tunnel loadings is
to being represented by an elastic bedding, whereas in con- investigated on multiple simulation scales.
tinuum models often only elastic ground deformations are
considered (Ahrens et al., 1982). Loadings are typically 2. Commonly used analytical methods for the determination
derived from either undisturbed in-situ stresses, or from of lining forces and deformations
pre-existing analytical solutions. Secondly, the structure
of the lining is simplified, as the lining and corresponding The first solutions to the tunnel lining problem were
segments themselves become idealized as beam structures. introduced in the early part of the 20th century. As a result,
Such an idealization produces a model which only yields language barriers inhibited a great deal of communication
stress resultants and deformations to be used as design between scientific communities (e.g. between english and
parameters, rather than more precise local stresses and german speakers). This problem continues to this day as,
strains. Even though these parameters may be sufficient typically, german authors and english authors continue to
to determine the structural stability of the tunnel lining, favor referencing different sources, e. g. for the continuum
other factors that may lead to serviceability or durability solution, german speaking communities reference Ahrens
issues, such as local chipping or cracking of the segments, et al. (1982), whereas english speaking communities tend
cannot be predicted or accounted for. to reference other sources, e.g. Wood (1975). Although
In this contribution a holistic multi-level method for the authors of this paper have attempted to include as
the simulation of segmental tunnel segments that many sources as possible, the sources given are (naturally)
addresses the issues mentioned above is proposed. In biased towards german-language literature, however, a
order to accurately evaluate the interactions of the tunnel good overview of german structural models for tunneling
lining with the surrounding ground and structures,‘‘ekate linings written in the english language is provided in
”, a 3D process-oriented finite element (FE) simulation Duddeck and Erdmann (1985).
tool (Nagel et al., 2009) based on the FE code KRATOS The analytical models for determining lining forces are
(Dadvand et al., 2010) is used. This model explicitly generally continuum models, e.g. Schmid (1926), Voellmy
accounts for the advancement and ring building process (1937), Wood (1975), Ahrens et al. (1982), and Einstein
during the construction of a machine driven tunnel. It is and Schwartz (1979), in which the tunnel lining (or lining
therefore able to account for the effects of time- like structure), is assumed to be bedded within an elastic
dependent processes, such as grout hardening, and how domain, but unidimensional beam models, e.g. Schulze
these influence the deformations and structural forces and Duddeck (1964a), Windels (1967), and Hain and
experienced by the tunnel lining. In order to evaluate Horst (1970), in which the differential equations for a 2D
the reliability of this model, the moments, axial forces, bedded beam are explicitly solved, exist as well. The differ-
and radial deformations are compared to those derived ent methods for representing the lining and ground, includ-
from other structural models, i.e. an analytical continuum ing methods to account for the lack of support at the
model, an analytical bedded beam model and a numerical tunnel crown, are depicted in Fig. 1.
beam-and-spring model, which are known to show differ- Continuum solutions for the tunnel lining problem,
ent results due to varying underlying assumptions derived from the theory of elasticity, were first proposed
(Kämper et al., 2016; Smarslik et al., 2017; Zhao et al., by Schmid (1926) and Voellmy (1937) and later models
2017). In order to investigate the detailed response of are modifications thereof. Both of these solutions, how-
the individual segments to radial tunnel loadings, the ever, rely on significant simplifications. In Schmid (1926),
radial deformations resulting from the 3D simulation are the lining is assumed to be very thick and therefore very
applied to a detailed FE segment model in a displacement stiff, and in Voellmy (1937) the tangential transfer of forces
controlled loading process. Specifically, the segment between the lining and ground is neglected. A full solution
response in ring direction is investigated. The segment to the problem was formulated by Ahrens et al. (1982).
model is constructed using a mesh in which non-zero This solution includes the tangential contact forces
thickness interface elements placed between standard geo- between the lining and ground and includes modifications
metrically linear finite elements (i.e. bulk elements) in to the solution to take into account the weak bedding at
order to account for cracking in the concrete segment, the crown of the tunnel for shallow tunnels (i.e. if the depth
as per Zhan (2016). The geometry of the segment is explic- of the tunnel is less than two times its diameter). Additional
itly modeled, and therefore the correct stress distribution modifications to account for segmentation of the lining
and cracking response in the radial and circumferential have been proposed in Blom (2002), and a method based
directions are captured. This provides a crack width which on discontinuous slender arches has been proposed in
may be used as a serviceability parameter for design. Zhang et al. (2017).
 V.E. Gall et al. / Underground Space 3 (2018) 45–60 47

account for a lack of ground support and to prohibit arti-

ficial bracing effects in case of crown slacking.
In the following, two analytical methods used for com-
parison with the numerical solutions are presented; first,
a continuum model by Ahrens et al. (1982) and second, a
bedded beam model as proposed by Schulze and
Duddeck (1964a,b). For the sake of clarity the assumptions
included in these models are briefly described. For a more
detailed comparison of such models, Erdmann and
Duddeck (1983) should be referenced.

2.1. Continuum model

The model assumes a perfectly circular thin ring of a

given thickness bedded in an infinite elastic half space.
Loading assumptions on the continuum ring are derived
from the in-situ stress state as given in Duddeck (1980).
The assumed loadings are taken to be the in situ stress in
the ground. When formulated in polar coordinates, the
loadings can be expresses as a series in function of the rota-
tion angle /.
The radial (pr ) and tangential (pt ) loadings acting on the
ring are:
X
1
pr ¼
prn cosðn/Þ; ð1Þ
n¼1
X1
pt ¼
ptn cosðn/Þ: ð2Þ
n¼1

If deformations are assumed to be small, each term of

the series can be independently associated with a corre-
sponding deformation mode of the lining. In Ahrens
et al. (1982), the linearly increasing portion of the in-situ
loading due to the weight of the soil across the height of
Fig. 1. Representation of the tunnel lining and ground in analytical the opening is neglected (see Fig. 2). Uplift of the tunnel
continuum (left) and beam models (right), acc. Putke (2016). is therefore also not taken into account. The loadings are
based on the two largest remaining terms of the series.
These are:
The solution for the elastically bedded ring are based on
either geometrically linear or geometrically non-linear
beam theories. These solutions do not explicitly account
for the surrounding medium and therefore loadings, as well
as the bedding stiffness must be taken from other sources.
This may lead to difficulties, as further simplifications to
the system are introduced. Common loading assumptions
are those given in Duddeck (1980), although other previous
assumptions (Houska, 1960; Terzaghi, 1946) exist in the lit-
erature. The solution reacts sensitively to the selected bed-
ding stiffness, i.e. the so-called ‘‘equivalent elastic bedding”
(Erdmann, 1983). In Schulze and Duddeck (1964a),
Windels (1967), and Hain and Horst (1970), for example,
the chosen stiffnesses all differ. In Schulze and Duddeck
(1964a) the linear problem is solved and in Windels
(1967) the problem was modified to take into account sta-
bility issues by using second-order analysis techniques.
Additionally, in the case of shallow tunnels (i.e. depth Fig. 2. Loading assumptions used for continuum models, specifically
6 2D) the bedding in the crown area is often neglected to (Ahrens et al., 1982).
48 V.E. Gall et al. / Underground Space 3 (2018) 45–60

pr ¼ p
r0 þ p
r2  cosð2/Þ; ð3Þ Herein ()0 denotes a derivative with respect to /; v and w
pt ¼
pt2  sinð2/Þ; ð4Þ represent the tangential and radial displacements, respec-
tively. The loadings pr and pt yield:
where:  
1 þ K0 3 þ K0
p0 ¼ c H þ 0:5R  0:3R ; ð14Þ


pr0 ¼ 0:5c½H þ ðH þ RÞK 0 ; ð5Þ 2 1 þ K0
 
p
r2 ¼ p
t2 ¼ 0:5c½H  ðH þ RÞK 0 : ð6Þ R H þ 0:5R
p ¼ c ð1  K 0 Þ  0:3ð3 þ K 0 Þ cosð2/Þ; ð15Þ
2 R
Here c corresponds to the unit weight of the ground, H to
pr ¼ p0 þ p; ð16Þ
the depth of the tunnel crown, and K 0 to the horizontal
earth pressure coefficient. R H þ 0:5R
pt ¼ c ð1  K 0 Þ sinð2/Þ; ð17Þ
Often, the resulting tangential lining load, ptn in the 2 R
above equation, is neglected. This is based on the assump- varying slightly from those used in Ahrens et al. (1982).
tion that grouting affects the frictional bond between the The stiffness of the radial bedding is derived from the
ground and the ring. Within the solutions given here, a full notion of an axially symmetric, inward ground movement.
shear bond between the soil and the lining ring is assumed Moreover, a scalar factor adjusts its extent according to the
and tangential loadings are accounted for. properties of the loading case and the specific structure
The solution given by Ahrens is based on the kinematic under investigation (i.e. segmented or continuous lining).
assumption that the displacement of the ground and that of Doing so, the following expression is obtained:
the lining at the tunnel opening correspond to each other
k r ¼ f  Es =R; ð18Þ
and that the resistance force of the lining plus the resisting
force of the soil at the tunnel opening must coincide with where Es describes the bulk modulus of the ground and f a
the in situ stresses. The forces thus stay in equilibrium. scaling factor which is typically adapted to be between 0.7
The solutions for the radial displacements (w) and the and 3 (Schulze and Duddeck, 1964a).
structural forces (N ; M), given as a series and as a function For shallow tunnels (H 6 2D) in soft ground, a lack of
of the in-situ earth pressures, result as follows: bedding at the crown in the area (50 < / < þ50 ) is
X
1 assumed by not accounting for the reaction forces due to
R bedding at these locations (see Figs. 2 and 3). Typically,
N¼ ðn
ptn þ

prn Þ cosðn/Þ; ð7Þ
i¼1
n2  1 to determine the suitability of continuum vs. bedded beam
X1
R2 models for the evaluation of structural forces, it is recom-
M¼ ð

ptn þ n

prn Þ cosðn/Þ; ð8Þ mended that continuum models be used for deep tunnels
nðn2  1Þ
i¼1
(i.e. overburden > 3D) and that bedded beam models be
X
1
R4 applied for shallow tunnels (overburden < 2D) (Duddeck,
w¼ ð

ptn þ n

prn Þ cosðn/Þ: ð9Þ
i¼1 EInðn2  1Þ2 1980; Erdmann, 1983). For areas inbetween, both models
may serve.
E denotes the Young’s modulus of the beam and I its For a comprehensive description of the bedded beam
moment of inertia. The solution given by Ahrens, as with model, the reader is referred to Schulze and Duddeck
that used for this paper, only considers the first and third (1964a).
term in this series.

2.2. Analytical bedded beam model

The solution of the elastically bedded beam in an elastic

continuum is based on circular beam theory. Under plane
strain assumptions the shear forces and torsional moments
can be neglected. The basic equations describing the ring
read:
Force Balance:
RN 0  M 0 þ R2 pt ¼ 0; ð10Þ
00
M þ RN þ R ðpr þ k r wÞ ¼ 0;
2
ð11Þ

Constitutive Equations:
EI 00
M¼ ðw þ wÞ; ð12Þ
R2
EA 0 M
N¼ ðv þ wÞ þ : ð13Þ
R R Fig. 3. Partially bedded elastic ring acc. Ahrens et al. (1982).
 V.E. Gall et al. / Underground Space 3 (2018) 45–60 49

2.3. Numerical beam models

Numerical models of bedded beams represent the typi-

cally used structural models for the design of segmental
tunnel linings in tunneling practice (German Tunnelling
Committee (DAUB), 2013; FHWA, 2009). Individual seg-
ments are modeled by beams and the segment and ring cou-
pling mechanisms are represented by non-linear springs.
Usually, single radially oriented springs with a ‘‘com-
pression only” feature idealize the lateral ground bedding
while the tangential friction is neglected. This results in a
more pronounced bending behavior within the segments
and comparatively reduced axial forces. Assuming a spac-
ing of the single springs of d and a segment width b, the sin-
gle spring parameters C b;r follow from Eq. (18) to:
ES
C b;r ¼ f bd: ð19Þ
R

In the circumferential direction, the segments are con- Fig. 5. Depiction of the longitudinal joint including joint contact area and
nected by longitudinal joints, in which the mechanical sealing band (black strip).
properties in the radial and circumferential directions are
idealized by displacement springs and the rotational prop- i.e. moment rotation relationship ensures that the resultant
erties by rotational springs. of the contact forces between two joints does not lie outside
Fig. 5 illustrates the basic geometrical parameters of the of the joint contact area.
longitudinal joint and shows the typical stretched shape of Depending on the ratio of bending moment to axial
the contact area which results in a biaxial load distribution forces - thus the eccentricity of the axial forces over the
in the segment. These joints are represented by blue spring width bf of the joint - the following stages arise (German
symbols within the rings in Fig. 4. The joints account for a
Tunnelling Committee (DAUB), 2013):
stiff axial contact and a soft rotational resistance. The
spring stiffnesses are determined according to concrete
1. A closed joint (M 6 Nlf =6):
hinge theory (Leonhardt and Reimann, 1966) and its asso-
ciated application to longitudinal joints (Janßen, 1983). bf l2f Ec
Here, the spring stiffness is iteratively modified based on C M;I ¼ ; ð20Þ
12
the actual ratio of axial force to the bending moment,
rather than assuming a constant value for the axial forces. 2. An open joint (M > Nlf =6):
Doing so, the spring behavior is iteratively updated. The
 2
criterion of maximum eccentricity, as in Janßen (1983), is 2M
9bf lf Ec M Nl  1
inherently checked by a plastic plateau of the bending, C M;II ¼
f
; ð21Þ
8N
with

bf : width of the joint,

lf : height of the joint,
Ec : Young’s modulus for concrete,
N: axial force in the joint,
M: bending moment in the joint.

If M P 0:28Nlf , the joint is considered to plastify.

In order to account for the reaction of the entire lining
system, two rings, as shown in Fig. 6, are simulated. Thus,
the stiffening effects of shear connections by dowels, cam-
and-pot connections, or other systems can be suitably cap-
Fig. 4. Beam model of two coupled rings with lateral bedding and tured using springs. Here, common cam-and-pot connec-
individual crown modeling. tions (Behnen et al., 2015; Girmscheid, 2008; Putke et al.,
50 V.E. Gall et al. / Underground Space 3 (2018) 45–60

3. Finite element model of the advancement process

3.1. Components of the Finite Element Model

The finite element model for shield driven tunneling has

been implemented in the object-oriented finite element
framework KRATOS (Dadvand et al., 2010) and is
denoted as ekate (Enhanced Kratos for Advanced Tun-
neling Engineering). A detailed description is given in
Meschke et al. (2011), Nagel et al. (2009), and Nagel and
Meschke (2010) and only a brief description is given here.
The model has recently been extended using BIM concepts,
Fig. 6. Numerical beam model of two rings with seven even-sized by incorporating it into the newly proposed Tunnel Infor-
segments and shear connections. mation Model (TIM) (Alsahly, Gall, et al., 2016a). The pri-
mary design goal is to provide an efficient yet realistic
2015) are used. Their local mechanical behavior is idealized simulation environment for all processes and their interac-
in a three-step manner, namely a primary slip stage, in tions during machine driven tunnel construction. This
which the rings are free to move, a second stiffening stage, model includes all relevant components of the mechanized
in which the rubber stripping along the cam-and-pot is tunneling process and incorporates them into the tunnel
compressed, and finally a direct concrete-to-concrete con- model as sub-models. These are the partially or fully satu-
tact of a high stiffness and final crushing. The simplification rated ground, the tunnel boring machine (TBM), the tunnel
of this behavior by a multi-linear force-displacement rela- lining, hydraulic thrust jacks and the tail void grouting. All
tionship is shown in Fig. 7. It is applied here to describe main components are able to interact with each other by
the behavior of the shear springs. means of algorithmic coupling. Fig. 8 shows the represen-
As a result, numerical beam models provide a more tation of the main components within the finite element
accurate idealization of the coupled ring behavior com- model.
pared to the simplified analytical models. However, in In what follows, a brief description is given of these
order to address the still inherent simplifications in bedding components and their features:
and tangential friction and to capture the complex loading
scenarios that can develop during construction stages, even 1. Ground model and soil: The ground model is formulated
more realistic, holistic approaches must be considered. within the framework of the theory of porous media and
accounts for the coupling between the deformations of
the solid phase and the two fluid pressures exerted by
the incompressible liquid and compressible air phases.
Fluid flow through the pores is described using Darcy’s
law in combination with the concept of relative perme-
abilities, using the soil-water characteristic curve accord-
ing to VAN GENUCHTEN. The material behavior of the soil
skeleton is modeled by means of a nonlinear elasto-
plastic constitutive law such as Drucker-Prager or the

Fig. 7. Ring coupling mechanism for cam-and-pot connections; (a)

conceptual section through a cam-and-pot system, (b) multilinear Fig. 8. (a) Main components involved in mechanized tunneling: (1) soil,
approximation of the shear load transfer at the ring joint. (2) shield, (3) lining, (4) grout, (5) hydraulic jacks and steering.
 V.E. Gall et al. / Underground Space 3 (2018) 45–60 51

Clay and Sand Model (CASM) (Yu, 1998). Quadratic 3.2. Kinematics of the shield machine
and linear approximations are used for the approxima-
tion of the displacements and the pore fluid and air In order to guarantee the proper advancement of the
pressures, respectively. A detailed description of the machine, the following forces, as shown in Fig. 9, are
multi-phase model for partially saturated soils and its explicitly included. In order to account for the correct
spatial and temporal discretization is given in Nagel advancement process of the machine, regardless of the
and Meschke (2007). path, and independently of the mesh, a remeshing scheme
2. Tunnel Boring Machine (TBM): The TBM is modeled as as well as a steering algorithm have recently been incorpo-
an independent, deformable body. It is connected along rated into the capabilities of the tunnel model (Alsahly,
the shield skin to the soil by means of a frictional con- Stascheit, et al., 2016b). The explicit forces that are consid-
tact condition. The weight of the structural and machin- ered are, as denoted in the figure, the following:
ery parts of the machine are explicitly accounted for.
3. Tunnel Lining: The segmented tunnel lining is repre- Fs : The heading face support is adapted to the require-
sented by volume elements that are activated in a step- ments posed by the specific TBM. Depending on
wise manner during the process simulation. Different the type of face support in hydro- and earth-
levels of detail can be accounted for in the lining. In pressure balance shields the pore water pressure, the
the case that the lining is modeled as a continuous ring, total or effective stresses, mechanical pressures or a
the segmentation of the lining is accounted for by means combination of these can be prescribed. The forma-
of a homogenized stiffness reduction according to Blom tion of a filter cake during standstill of the machine
(2002). The tunnel lining tube may also be modeled as a can be accounted for by the application of a fluctuat-
fully segmented model in which the individual segments ing combination of mechanical and liquid pressures.
interact by means of a contact algorithm (Marwan et al., As air pressures in the ground for partially saturated
2017). In this contribution it is modeled as a continuous conditions can be prescribed independently, the
ring in order to better compare it with the analytical model is also able to account for temporary face sup-
models. The tunnel lining tube is used as counter- port by means of compressed air (Nagel et al., 2008).
bearing for the hydraulic jacks thrusting forward the Fsw : The weight of the shield machine, the additional load-
shield machine. Structural forces are extracted from ings from the engine, and the lining erector and the
the continuum ring by resolving the stresses in the lining cutter-head are explicitly accounted for. The weight
to a plane lying along the radius of the tunnel. In doing of the machine causes the machine to dip downwards
so, it is assumed that the center of the lining represents during an excavation step, which in turn effects the
the neutral axis. jack thrust, as these are necessary to correct the
4. Grouting: A two-phase (hygro-mechanical) formulation downward movement.
similar to the one used for the ground model is used
to model the pressurized grouting mortar that is used
to fill the gap between lining and ground. Here, the
grouting pressure is applied as a pore water pressure act-
ing on the boundary of the fresh mortar. Stiffening of
the grouting mortar is accounted for and coupled with
the grout pore water dissipation process, which results
from its infiltration into the surrounding ground.
5. Hydraulic Jacks and the advancement process: The
hydraulic jacks, represented by geometrically non-
linear truss elements, are elongated by means of pre-
scribed strains induced in the elements in order to
advance the machine. The front surface of the last acti-
vated lining segment is used as the counter-bearing for
the hydraulic jacks thrusting the shield machine for-
ward. The elongation of each jack element is controlled
by a steering algorithm that allows for counter-steering
against weight-induced dropping of the TBM to keep
the path of the machine on the prescribed tunnel align-
ment. The simulation of the advancing process for arbi-
trary alignments requires a continuous adaption of the
finite element mesh in the vicinity of the tunnel face in
conjunction with the steering algorithm for the TBM Fig. 9. Modeling of interactions between soil and TBM in the simulation
advance and appropriate algorithms for the transfer of model ekate: F s : heading face support, F sw : weight of the shield machine,
internal variables. F th : thrust forces, F cs : frictional contact between shield skin and soil.
52 V.E. Gall et al. / Underground Space 3 (2018) 45–60

Fth : The jack thrust force results from the jacks pushing settlements are tied regardless of the FE discretization
off the previously installed lining ring. This force is (Fig. 10). The weight and the stiffnesses of buildings, how-
generated by straining the jack elements (i.e. applying ever, must be modified due to their discretization.
a controlled displacement) and produces a reaction The buildings are assumed to be masonry constructions,
force against the machine and the lining. Through in which the main load bearing structure in each building
inclusion of the steering algorithm, the eccentric consists of an outer masonry wall with window openings.
application of the jack forces as a result of necessary Additional structural components of the buildings are
corrections or a curvature of the alignment is assumed to be their roofs, plate foundations, and their
included in the simulation. floor plates. In order to derive a replacement stiffness for
Fcs : The forces resulting in the contact of the shield with the building, the bending stiffnesses of the buildings are
the soil are modeled using a contact algorithm. taken to be equivalent to that of a beam with a rectangular
Because of the taper of the TBM, stiffening elements cross-section. A lower bound solution can be given if it is
like pressure wall and the variation of the thickness assumed that only the walls perpendicular to the tunnel
along the shield skin are accounted for in the geomet- axis produce an effective bending stiffness, and an upper
rical representation of the TBM. Additionally, the bound can be defined if all the above mentioned structural
possible flow of process liquids (support fluid and components are included in a shear-stiff manner. These
grouting mortar) along the shield skin is taken into bounds are given as follows:
account in the model by means of a finite difference Lower Bound:
scheme along the element vertices of the shield. The
bh3
possible existence of a pressurized liquid film between EI l ¼ E aw ; ð22Þ
shield and ground is therefore explicitly considered 12
within the contact formulation (Nagel and Upper Bound:
Meschke, 2011). Although the frictional forces can  3  X n  3 
bh bi hi
be taken into account between the soil and the skin, EI u ¼ E aw þ z s A þ
2
Ei þ zsi Ai ;
2
ð23Þ
in this contribution it is assumed that the frictional 12 i¼1
12
forces are zero due to the lubricating effects of the here E; b; h, and A are the Young’s Modulus, width, height
grout. and cross-sectional area of the outer masonry walls, and
the subscript i denotes the same for the floor, roof, and
foundation plates; aw is a window factor applied to reduce
3.3. Incorporation of surrounding structures and buildings the masonry wall stiffness due to the windows, and zs and
zsi denote the vertical distance between the neutral axis of
During urban tunnel drives, particular interest must be the entire structure to that of the individual walls and
paid to controlling surface settlements in order to minimize floors/roof/foundation, respectively.
damage to existing structures (Mark et al., 2012; The simulation model ekate can also be modified to
Neugebauer et al., 2015; Schindler et al., 2016). The tunnel incorporate pile structures into simulations. The piles are
and surface structures mutually affect each other; the tun- represented as embedded beam structures which are con-
nel may cause settlement related damages to a building, nected to the ground through a modified contact algorithm
and the building may affect surface settlements. Addition- as per Ninić et al. (2014) (Fig. 11). In this contribution,
ally, the building may affect the tunnel. Any underground however, because all the modelled buildings rest on plate
components of the building, such as pile foundations, gar- and strip foundations, only the modeling of surface struc-
ages, or basements, can affect the stresses around a tunnel
tube, and the weight of the buildings can result in asymmet-
ric loadings of the tunnel lining. It is therefore important
to, if a detailed investigation is desired, model the affect
of surface or subsurface constructions. However, the
detailed discretization of additional structures, in particu-
lar for large models with several surface or sub-surface
structures, results in high computational costs. For this
reason, and because the later investigated scenario includes
loads due to buildings on plate foundations, a newly devel-
oped technique is introduced in this section to reduce the
modeling cost of buildings.
The buildings are modeled using volume elements. In
order to allow an accurate orientation of the structures
independent of the ground mesh, a node to volume
Lagrange tying algorithm is used to impose a deformation Fig. 10. Replacement stiffness for the incorporation of buildings in the
constraint in which the buildings’ bases and surface finite element model.
 V.E. Gall et al. / Underground Space 3 (2018) 45–60 53

Table 2
Grouting parameters used the process oriented FE model. The subscript g
denotes ‘‘grout.” The grout is modeled using a time-dependent stiffness.
Stiffness values of the grout, Eg , are provided for different times. Eg;1 is the
final grout stiffness. The grouting pressure applied behind the shield
machine immediatley after machine advancement is provided at tunnel
crown (rpress;0 ), its spring-line rpress;90 ;0h , and its invert, rpress;180 ;0h .
Grout
Parameter Value Unit
tg 0.2 [m]
cg 24 [kN=m3 ]
m 0.3 [–]
Eg;2h 15 [MN=m2 ]
Eg;12h 44 [MN=m2 ]
Eg;24h 60 [MN=m2 ]
Eg;1 120 [MN=m2 ]
rpress;0 ;0h 327 [kN=m2 ]
rpress;90 ;0h 380 [kN=m2 ]
rpress;180 ;0h 433 [kN=m2 ]

Fig. 11. Representation of pile foundations by means of embedded beam

Table 3
elements in contact with the surrounding ground acc. Ninić et al. (2014).
Lining parameters used. For further explanation of the geometrical
parameters see Fig. 5. Here the subscript c refers to concrete and Router to
tures, specifically of their weight, are of primary impor- the outer radius of the lining.
tance for lining loadings. Pile foundations are therefore Lining
not present in the investigated scenario.
Parameter Value Unit
Router 5.3 [m]
4. Investigated scenario
h 0.5 [m]
bf 1.00 [m]
A tunnel with a diameter of 10.6 m and an overburden lf 0.25 [m]
of 21.2 m (i.e. 2D), has been chosen as a basis for the com- b 1.00 [m]
parison of the investigated lining models. It is assumed that Segmentation 7+0 [–]
cc 24 [kN=m3 ]
the tunneling process takes place under drained ground
Ec 36,000 [MN=m2 ]
conditions and that the water table is located at the ground mc 0.25 [–]
surface. Hence, three simultaneously acting load cases are
assumed, i.e. the earth pressure, the water pressure, and Table 4
the dead loads. The weight of the buildings has not been TBM shield parameters used in the process oriented FE model. DTBM and
included as loading component in any of the numerical LTBM are the shield diameter and length, respectively. The shield is modeled
beam models or analytical models because of the crown as a perfect cylinder and therefore the conicity of the shield is taken to be
0. The subscript st denotes ‘‘steel.”
depth of 20 m. The building loads have, however, been
included in the 3D numerical continuum model. The TBM
results, as shown in Section 5 show that, even under these Parameter Value Unit
assumptions, the predicted loadings and deformations are DTBM 10.7 [m]
comparable.The parameters used for the simulation of LTBM 15 [m]
the ground, grouting, TBM and lining are summarized Conicity 0 [cm]
cst 78.5 [kN=m3 ]
within Tables 1–4. E 210,000 [MN=m2 ]
Pst
P jacks 30,000 [kN]
Table 1
Ground parameters used. The subscript s denotes ‘‘soil” and K 0 is the
horizontal earth pressure coefficient.
5. Comparison of predicted lining forces and deformations
Ground
Parameter Value Unit For the following discussion, the structural models will
cs 21 [kN=m3 ] be referred to as follows:
c0s 11 [kN=m3 ]
/s 35 [ ] AK: 2D analytical elastic continuum model
cs 0 [MN=m2 ]]
Es 120 [MN=m2 ]
AB: 2D analytical bedded beam model
ms 0.3 [–] NB: 3D numerical bedded beam model
K0 0.4 [–] NK: 3D process-oriented FE model
54 V.E. Gall et al. / Underground Space 3 (2018) 45–60

5.1. Deformations The results of model NK deviate noticeably from those

of the others in terms of a vertical translational deforma-
The final radial deformations in [mm] for the investi- tion, specifically at the invert. This effect is mainly induced
gated scenario as predicted by all model variations are by the inherent earth pressure approach, in which the undis-
summarized in Fig. 12. The deformations are displayed turbed stress state of the ground prior to the excavation is
from 0° to 360°, as measured clockwise starting at the tun- explicitly considered. Consequently, the resultant force
nel crown. The solid black circle represents the undeformed after excavation at the invert is larger than at the crown,
reference configuration. which causes an uplift of the tunnel. This effect is the pri-
All deformations display a similar tendency of the lining mary feature that leads to differences in the predictions
cross-section to ovalize and the displacements agree well at between the model NK and the other model predictions,
and around the tunnel springline, both qualitatively and which consider horizontal and vertical loadings to be in
quantitatively. The most significant differences occur at equilibrium. It should also be noted that the apparent kink
the tunnel crown and invert. While the analytical model in the representation of the deformations in model NK is a
AK considers a full 360 of radial bedding, which is active result of the plotting method which leads to distortions at
both in tension and compression, the consideration of a the center of the graph. Unlike the deformations of N-BR2
compression only bedding (NB) or a non-bedded crown at the tunnel invert, which in fact do have a kink, the slope
(AB) lead to significantly greater deformations at the of the deformations predicted by model NK are continuous.
crown and represent a more realistic ground-structure- Segmentation of the lining is exclusively accounted for
interaction. The only model that simulates the surrounding in model NB. All other models assume a continuous ring.
soil and resulting soil reactions, NK, like models NB and Nevertheless, the results are in good agreement with the
AB, predicts greater radial displacements than model AK analytical solutions and only display noticeable differences
and is additionally able to capture the relaxation of the at the crown (N-BR1) or at the invert (N-BR2) as a result of
ground due to excavation underneath the lining. the position of the longitudinal joints or rotational springs
in the consecutive rings (see Fig. 6). Rotating rings by a
half segment with each TBM advance results in a staggered
joint configuration. Therefore, within ring 1, a longitudinal
joint is located directly at the crown (0 ) and within ring 2,
a joint is located directly at the invert (180 ). Due to the
inward movement of the tunnel lining the radial spring
becomes ineffective, and is not able to transfer the acting
forces. This, in combination with the joint position, leads
to the differences between the lining deformations predicted
by model NB and those of the other models. Furthermore,
the interaction between the two coupled rings has an addi-
tional stiffening effect as a result of their staggered joint
configuration. This is most apparent when comparing the
respective predicted deformations at the tunnel springline.
Here the segmentation has no apparent influence and dis-
placements are in good agreement with those models
assuming a rigid ring. Neglecting a ring interaction would
lead to significantly greater radial displacements (Kämper
et al., 2016) which would need to be taken into account
in the design of the longitudinal joints.
In summary, the incorporation of a more realistic
ground behavior by permitting only the transfer of com-
pressive forces between the lining and the surrounding soil
and considering relaxation effects as well as a detailed sim-
ulation of longitudinal joints by rotational springs lead to
higher deformations of the lining. Conversely, a simplified
ground-structure-interaction and a consideration of stag-
gered joints in combination with shear load transfer mech-
anism between rings reduce deformations.

5.2. Structural forces

Fig. 12. Comparison of lining deformations predicted by the investigated
models, as given in the legend. Note that the subscripts R1 and R2 refer to The predicted axial forces and bending moments are dis-
the 1st and 2nd rings of the numerical beam model. played in Fig. 13. The same notation as Fig. 12 is
 V.E. Gall et al. / Underground Space 3 (2018) 45–60 55

model NB exhibits the smallest variations in bending

moments and predicts almost constant axial forces. More-
over, allowable eccentricities of the axial forces at the lon-
gitudinal joints are only ensured if a realistic approach for
the modeling of the longitudinal joints is incorporated, in
which the joints soften under increasing rotations. In this
framework, the analytical models provide conservative esti-
mates of bending moments and therefore serve well for
conceptional design studies.
Second, displacement induced bedding at the tunnel
crown that arises from uplift effects are not captured by
uniform loading assumptions that simply balance the equi-
librium of vertical forces at the tunnel base. In contrast, as
indicated by the bending line computed with the most elab-
orate model NK, additional bending occurs due to non-
uniform redistributions of reaction forces at the tunnel
crown. Therefore, especially if complex loading scenarios
or shallow undercuttings of buildings occur, it may be
favorable to turn to full-scale FE models.

6. Investigation of detailed segment response

Existing segment design guidelines e.g. German

Tunnelling Committee (DAUB) (2013), JSCE-Tunnel
Engineering Committee (2007), and ITA Working Group
No. 2 (2000) (AFTES, 1997; FHWA, 2009), either directly
provide or refer to a maximum allowable crack width as a
design parameter, but do not present any methods for
investigating or predicting crack development. Such crack-
ing phenomena are often observed locally at segment
extremities, such as the corners of the longitudinal joints
and along the center of a segment (Sugimoto, 2006). Chip-
ping at the segment corners is often attributed to mishan-
dling of segments or to radial loadings experienced by the
lining structure (Sugimoto, 2006), whereas cracking along
the center of the segment is attributed to the action of
the thrust forces on the ring within the shield machine
(Conforti et al., 2016). An additional important loading
case that may result in concrete cracking is the local load-
ing of the segment immediately beneath the thrust jack
(Tiberti et al., 2015). These loading scenarios, i.e. longitudi-
nal jack thrust loading, radial loading due to ground loads
and/or grout pressures, and mishandling of the segments,
can therefore all result in concrete cracking associated with
Fig. 13. Comparison of moment and axial force distributions as predicted
serviceability issues. Within this contribution however,
by the investigated models, as given in the legend. Note that the subscripts
R1 and R2 refer to the 1st and 2nd rings of the numerical beam model. only the segment response to radial ground loadings expe-
rienced outside of the shield machine are investigated in
detail. Specifically, the effect of reinforcement diameter
employed. The results are characterized by the degree to on the load spreading conditions immediately below the
which each modeling approach assigns more or less bend- longitudinal joint are investigated.
ing stiffness to the tunnel lining. Longitudinal joint design is based on unconfined load
Two major issues become apparent when comparing the spreading tests (Janßen, 1983; Leonhardt and Reimann,
structural forces predicted by the various models. First, a 1966). These tests generally do not include the influence
simplified stiff modeling of the segments, in which a contin- of the confinement at the outer surface of the segment.
uous ring without joints is adopted, provokes an increase in Attempts have been made to describe nonlinear segment
bending related loadings of the ring while simultaneously behavior under tunnel loads using complex numerical mod-
reducing corresponding axial forces. Thus, the segmented els (Delgado, 2012), but often more attention is paid to the
56 V.E. Gall et al. / Underground Space 3 (2018) 45–60

lining system (Arnau and Molins, 2012), the load transfer

mechanism between the joints (Cavalaro and Aguado,
2012), or to the improper installation of the segments
(Cavalaro et al., 2011) than is paid to crack development
within the segment under normal loadings. Additionally,
when design loadings are investigated, the reinforcement
is rarely explicitly modeled. The stresses in the segment
are analysed assuming a uniform concrete cross section,
and then this stress distribution is used for reinforcement
design. This procedure does not accurately account for
the influence that the reinforcement bars have on the stres- Fig. 15. Degenerated 3D solid element characterized by its ‘‘base” surface,
ses in the segment. In this section, a methodology is pro- an ‘‘apex” point (Node-4), and the projection of the apex point onto its
posed in which the radial displacements generated by the base (point 40 ) along N.
full 3D FE model above are applied to a nonlinear compos-
ite FE model of the tunnel segment in order to investigate 2016). This ensures that the elements are very thin, there-
the influence of the reinforcement bars on the pre- and fore the strain in the ISE’s is almost exclusively related to
post-cracking behavior of the longitudinal joint. the (regularized) unbounded strain ^e, defined as:
1
e  ^e ¼ ðsut  nÞ :
s
ð24Þ
6.1. Composite FE model of the reinforced segment h
The displacement jump, sut, is determined from the rela-
In order to model the reinforced concrete, cohesive tive displacement of the apex node with respect to its pro-
interface elements as proposed by Manzoli et al. (2012), jection on the base (see Fig. 15).
in conjunction with standard small-strain triangular Finite The inelastic material behavior of the ISE’s is formu-
Elements (i.e. Bulk elements) and a penalty based tying lated in terms of damage mechanics as:
algorithm between standard truss elements and the bulk
elements are used. The interface solid elements (ISE’s) 1 s
r ¼ ð1  dÞCe : e  ð1  dÞ Ce : ðsut  nÞ ; ð25Þ
are used to simulate the cracking behavior, and consequent h
loss of strength, of the concrete and the standard elements where d is the scalar damage variable and Ce denotes the
control the compressive behavior of the structure. The elastic stiffness tensor. A Poisson’s ratio of m ¼ 0 is chosen
ISE’s are formulated as per Zhan and Meschke (2016). to decouple the stresses parallel and normal to the interface
The truss elements that are used to represent the reinforce- surface. The loading criterion, f ðr; aÞ is defined in terms of
ment bars are tied to the bulk elements in order to enable the equivalent stress r~ and the displacement-like internal
the reinforcement to bridge cracked areas. The interface parameter a:
elements are of a discrete thickness and therefore the FE
~  tðaÞ 6 0:
f ðr; aÞ ¼ r ð26Þ
mesh is processed such that each standard element is sur-
rounded by interface elements. The mesh of the truss ele- The softening behavior of interface is determined based on
ments, because of the tying algorithm, is independent of the fracture energy of concrete (Zhan, 2016):
that of the others (see Fig. 14).  
a
tðaÞ ¼ f t exp  ; ð27Þ
wref
6.1.1. ISE element
The ISE elements used to model the cracks can, in the where f t represents the tensile strength of the concrete. The
general 3D case, be represented as shown in Fig. 15. internal parameter a is defined based on the maximum
Here h is the height of the element, which is usually value of equivalent separation experienced during the load-
taken to be a value of approximately 1/1000 of the average ing history:
side length of the element’s base (Zhan and Meschke, a ¼ maxð~uÞ  ~u0 : ð28Þ
~u0 corresponds to the limit state of the elastic interface:
f t hf 
~u0 ¼ ¼ t  0; ð29Þ
K int E
here K int represents the ‘‘rigid” elastic stiffness of the equiv-
alent interface behavior. The scalar damage variable dðaÞ is
obtained by comparing the secant stiffness K sec with the
elastic stiffness of the equivalent interface behavior:
Fig. 14. Mesh composition: ISE’s are dark grey, bulk elements are light
K sec ht
grey, and the truss is represented by the horizontal line. The ’s mark the dðaÞ ¼ 1  ¼1  : ð30Þ
location of the truss nodes. K int E ða þ ~u0 Þ
 V.E. Gall et al. / Underground Space 3 (2018) 45–60 57

Damage is only allowed to occur if the ISE is in tension, Table 5

i.e. if ðsutn  ~
u0 Þ P 0. Hence, no damage occurs if the Properties of linear elastic (bulk) and interface solid elements (ISE) used to
model plain concrete in the presented simulations.
interface is in compression. For this reason the elastic
properties of the ISE must be chosen in such a manner as Property Bulk ISE Unit
to match those of the bulk elements. More information Young’s modulus, E 36,000 36,000 [N=mm2 ]
concerning the ISE element formulation can be found in Poisson ratio, m 0.2 0.0 [–]
Tensile strength, f t – 3 [N=mm2 ]
Zhan (2016) and Zhan and Meschke (2016). Fracture energy, Gf – 0.1 [N/mm]

6.1.2. Rebar element

In order to incorporate the reinforcement into the The chosen reinforcement layout does not correspond to
model, the nodes of the truss elements are tied to their pro- a specific tunnel project, but, has been chosen so as to rep-
jected point within the bulk elements by enforcing the fol- resent an exemplary reinforcement layout (Putke et al.,
lowing condition: 2016). Reinforcement for the splitting stresses and a stan-
dard cage reinforcement around the perimeter of the seg-
!
truss  NðnN ; trussÞ  uBulk ¼ 0;
uNi ð31Þ ment are modeled. The geometry is presented in Fig. 17.
All rebars are modeled as linear elastic truss elements with
where uNi
truss are the displacements at a truss node, ubulk the
displacements of all the nodes of the bulk element in which a Young’s Modulus of E ¼ 200; 000 N=mm2 , and have a
the investigated node of the truss are embedded, and standard reinforcement diameter of £ ¼ 10 mm. It is
NðnN ;truss Þ is the shape function matrix evaluated at the pro- assumed that rebar cages are placed at 10 cm intervals.
jected point of the truss node within the bulk element. To investigate the effect of increasing reinforcement on
the crack distribution of the segment, a second example
6.2. Transfer of predicted loading to the segment has been calculated using a reinforcement diameter of
£ ¼ 20 mm. Although such a large reinforcement diame-
In order to investigate the effect of the ground loadings ter may be uncommon, this comparison serves to exhibit
at the segment level, and specifically the effects of the load- the effects of increasing reinforcement.
ings on the longitudinal joint, the radial deformations
obtained (see Fig. 12) by the 3D FE simulation are applied 6.3. Predicted crack width
to a finely meshed plane-strain model of a lining segment.
The geometry of the lining segment can be seen in Fig. 16. Figs. 18–20 display the predicted crack distribution for a
The investigated segment is assumed to be at the tunnel plain concrete segment (PC), a standard reinforced seg-
crown. This placement of the segment has been chosen ment, and an over-reinforced segment, respectively. The
because most observed segment damages in TBM driven unreinforced segment is intended as a reference case, and
tunnels tend to occur at the crown (Sugimoto, 2006). The the over-reinforced segment is intended to show the non-
reaction of the adjoining segments are taken into account linear effects of increasing reinforcement. The predicted
by the linear elastic segments at both ends of the modeled crack widths are shown in Table 6.
segment. Interface elements are placed between the Even though the plain concrete segment does not repre-
modeled segment and the adjacent segment to enable the sent a realistic design case, it presents interesting results
opening of the longitudinal joint. and provides a reference case for analysis. What can be
The simulation performed above, although realistic, seen in the unreinforced segment is that the primary failure
does not represent a distinct tunnel project. For this mechanism corresponds to ‘‘chipping” at the segment cor-
reason, no ‘‘real” concrete properties exist for the presented ner, and not to a splitting stress failure due to load spread-
simulation. Instead the standard concrete material proper- ing, as is the case which is designed for. The shape, or
ties according to Table 5 have been chosen. rather size, of the load transfer area between two segment
joints inhibits the formation of stresses in the corner of

Fig. 16. Setup for the numerical simulation. All units are given in [mm]. Fig. 17. Layout for the reinforcement bars. All units are given in [mm].
58 V.E. Gall et al. / Underground Space 3 (2018) 45–60

the segment. This leads to a type of shearing strain devel-

oping between the strained and unstrained areas of the seg-
ment at bottom corner of the joint that in turn leads to
crack development. Interestingly, because of the effect of
the confinement due to the grout, this effect is not observ-
able along the outer edge of the segment.
The conventionally reinforced segment (£ ¼ 10 mm)
displays, as would be expected, a much smaller cracked
Fig. 18. Predicted crack distribution for an unreinforced concrete
zone. The observed failure phenomenon corresponds to a
segment; intended as a reference case (cracked area in red; 200 fold splitting type failure as would be expected due to uncon-
magnification of displacements). fined load spreading. The presence of the reinforcement
bars effectively redistributes the stresses so that cracking
at the corner is prevented. It should be noted that a diagonal
crack is present immediately at the contact area of the lon-
gitudinal joint, but does propagate to the inner edge of the
segment. Additionally, the calculated crack width (see
Table 6) is within the allowable range i.e. 60.3 mm
(Deutsches Institut für Normung, 2011). The strongly rein-
forced segment displays a significantly different crack pat-
tern, as only the ‘‘chipping” type of failure is observed.
Because the reinforcement bars are significantly stiffer (4
times stiffer than in the previous example), and because
the reinforcement bars are able to sustain compressive
loads, the reinforcement redistributes the stresses to the
stiffest area of the segment, which occurs at its inner corner.
Because of the minimal concrete cover along the inner rein-
forcement, the resulting maximum crack width is slightly
Fig. 19. Predicted crack distribution in the tunnel segment assuming a
above the allowable crack widths of 0.3 mm as given in
rebar diameter of £ ¼ 10 mm (cracked area in red; 200 fold magnification
of displacements). Deutsches Institut für Normung (2011). In reality, however,
because the reinforcement is distributed across a certain
width, this crack width would most likely be somewhat
smaller than that predicted, and therefore most likely lie
within the allowable range. What the presented examples
show is that the investigated segment is structurally stable
with respect to the applied loads. Cracking is local and does
not propagate throughout the segment in a manner so as to
compromise the load bearing capacity of the structure. The
expected crack width, however, is not always within an
acceptable range, depending on the choice of reinforcement.
The design case for which the splitting reinforcement is
designed is based on predicted load spreading in an uncon-
fined loading case (German Tunnelling Committee
(DAUB), 2013; Deutsches Institut für Normung, 2001).
The loading scenario that occurs beneath longitudinal joints
of tunnel linings is, however, not unconfined, as the grout-
Fig. 20. Predicted crack distribution in the tunnel segment assuming a
rebar diameter of £ ¼ 20 mm (cracked area in red; 200 fold magnification
ing confines another area of the joint. The load eccentricity
of displacements). coupled with the one-sided confinement leads to the often
observable ‘‘chipping” phenomenon. This may only be true
at the investigated location. The moment at the springline
of the tunnel lining is typically negative, and therefore the
segment bends in the opposite direction than it does at the
Table 6
tunnel crown. Consequently, a different failure mode would
Predicted crack widths for the unreinforced segment (PC) and the two
variations of reinforced segments (RC). be expected. Additionally, the reinforcement has a signifi-
cant effect on crack distribution as a more strongly
PC RC £ ¼ 10 mm RC £ ¼ 20 mm
reinforced segment leads to a different failure pattern than
Crack [mm] 0.57 0.18 0.38
a more weakly reinforced segment.
 V.E. Gall et al. / Underground Space 3 (2018) 45–60 59

7. Conclusions Behnen, G., Nevrly, T., & Fischer, O. (2015). Soil-structure interaction in
tunnel lining analyses. Geotechnik, 38(2), 96–106.
Blom, C. (2002). Design philosophy of concrete linings for tunnels in soft
A new multi-scale method for the evaluation of segmen- soils (Ph.D. thesis). Delft University.
tal lining response to mechanized tunnel-induced construc- Cavalaro, S., & Aguado, A. (2012). Packer behavior under simple and
coupled stresses. Tunnelling and Underground Space Technology, 28,
tion loadings has been presented. Two analytical models 159–173.
and an industry-standard numerical beam model, are com- Cavalaro, S., Blom, C., Walraven, J., & Aguado, A. (2011). Structural
pared with a full-scale process-dependent 3D FE simula- analysis of contact deficiencies in segmented lining. Tunneling and
Underground Space Technology, 26, 734–739.
tion with respect to their ability to generate design Conforti, A., Tiberti, G., & Plizzari, G. A. (2016). Combined effect of high
relevant structural forces, i.e. moments and axial forces. concentrated loads exerted by TBM hydraulic jacks. Magazine of
It is shown that the choice of loading assumptions is critical Concrete Research, 68, 1122–1132.
Dadvand, P., Rossi, R., & Oñate, E. (2010). An object-oriented environ-
to the calculations. Using a full 3D model for mechanized ment for developing finite element codes for multi-disciplinary appli-
tunneling, in which all components are accounted for, may cations. Archives of Computational Methods in Engineering, 17,
offer the most accurate forces for design, as it provides real- 253–297.
Delgado, O. A. (2012). Structural response of precast concrete segmental
istic spatial-temporal distributions of the loading acting on tunnel linings (Ph.D. thesis). Universitat Politècnica de Catalunya.
the lining segments. Additionally, when using such a Deutsches Institut für Normung. (2001). DIN 1045-1: Tragwerke aus
model, no loading assumptions are required for the analy- Beton, Stahlbeton und Spannbeton. Teil 1: Bemessung und Konstruktion.
Berlin, Germany: Beuth Verlag GmbH (in German).
sis of the lining segments. The loadings obtained from the Deutsches Institut für Normung. (2011). DIN EN 1992-1-1: 2011-01:
3D tunnel model are, in the form of displacements, trans- Eurocode 2 - Bemessung und Konstruktion von Stahlbeton und Spann-
ferred to a high resolution non-linear composite FE model betontragwerken - Teil 1-1: Allegemeine Bemessungsregeln und Regeln
für den Hochbau. Berlin, Germany: Beuth Verlag GmbH (in German).
of a single lining segment in order to investigate the nonlin- Duddeck, H. (1980). Empfehlungen zur Berechnung von Tunneln im
ear structural response of the segments, particularly with Lockergestein. Die Bautechnik, 10, 349–356.
respect to the load-transfer mechanism between segments Duddeck, H., & Erdmann, J. (1985). On structural design models for
tunnels in soft soil. Underground Space, 9, 246–259.
at the longitudinal joints. In doing so, it is observed that Einstein, H. H., & Schwartz, C. W. (1979). Simplified analysis for tunnel
the primary variable which controls design at this level is supports. Journal of Geotechnical and Geoenvironmental Engineering,
the developed crack width and that, depending on the level 105, 499–517.
Erdmann, J. (1983). Vergleich ebener und Entwicklung räumlicher Berech-
of reinforcement, the observed failure mechanism of the nungsverfahren für Tunnel (Ph.D. thesis). TU Braunschweig (in German).
segment does not always correspond with that to be Erdmann, J., & Duddeck, H. (1983). Statik der Tunnel im Lockergestein -
expected from the design process. The introduced multi- Vergleich der Berechnungsmodelle. Bauingenieur, 58, 407–414 (in
German).
level modeling technique presents a modeling methodology FHWA. (2009). Technical manual for design and construction of road tunnel
in which not only structural factors, but also serviceability - civil elements. Technical Report Report No. FHWA-NHI-10-034 U.
factors, such as crack development in linings, can be S. Department of Transportation Federal Highway Administration.
German Tunnelling Committee (DAUB). (2013). Recommendations for the
designed for. The proposed technique therefore provides design, production and installation of segmental rings. Technical Report
a blueprint for a more holistic segment design process. Deutscher Ausschuss für unterirdisches Bauen e. V. (DAUB).
Girmscheid, G. (2008). Baubetrieb und Bauverfahren im Tunnelbau. Berlin:
Ernst & Sohn (in German).
Acknowledgement Hain, H., & Horst, H. (1970). Spannungstheorie 1. und 2. Ordnung fur
beliebige Tunnelquerschnitte unter Berücksichtigung der einseitigen
Bettungswirkung des Bodens. Strasse Brücke Tunnel, 22, 85–94 (in
Financial support for this work was provided by the German).
German Research Foundation (DFG) in the framework Houska, J. (1960). Beitrag zur Theorie der Erddrücke auf das Tunnel-
of subprojects B1, B2, C1 and D3 of the Collaborative mauerwerk. Schweizerische Bauzeitung, 78, 607–609 (in German).
ITA Working Group No. 2, I.T.A. (2000). Guidelines for the design of
Research Center SFB 837. This support is gratefully shield tunnel lining. Tunnelling and Underground Space Technology 15,
acknowledged. 303–331 (in German).
Janßen, P. (1983). Tragverhalten von Tunnelausbauten mit Gelenktübbings
(Ph.D. thesis). TU Braunschweig (in German).
References JSCE-Tunnel Engineering Committee. (2007). Standard Specifications for
Tunneling-2006: Shield Tunnels. Japan Society of Civil Engineers.
AFTES. (1997). Recommendations for the design, sizing and construction of Kämper, C., Putke, T., Zhao, C., Lavasan, A. A., Barciaga, T., Mark, P.,
precast concrete segments installled at the rear of a tunnel boring & Schanz, T. (2016). Vergleichsrechnungen zu Modellierungsvarianten
machine (TBM). Association Francaise des tunnels et de l’espace für Tunnel mit Tübbingauskleidung. Bautechnik, 93, 421–432 (in
souterrain (version 1 ed.). German).
Ahrens, H., Lindner, E., & Lux, K. H. (1982). Zur Dimensionierung von Leonhardt, F., & Reimann, H. (1966). Betongelenke. Der Bauingenieur,
Tunnelausbauten nach den Empfehlungen zur Berechnung von Tun- 41, 49–56 (in German).
neln im Lockergestein (1980). Bautechnik, 8, 260. Manzoli, O., Gamino, A., Rodrigues, E., & Claro, G. (2012). Modeling of
Alsahly, A., Gall, V. E., Marwan, A., Ninić, J., Meschke, G., Vonthron, interfaces in two-dimensional problems using solid finite elements with
A., & König, M. (2016a). From building information modeling to real high aspect ratio. Computers and Structures, 70–82.
time simulation in mechanized tunneling. In Proceedings of the world Mark, P., Niemeier, W., Schindler, S., Blome, A., Heek, P., Krivenko, A.,
tunneling congress. & Ziem, E. (2012). Radarinterferometrie zum Setzungsmonitoring
Alsahly, A., Stascheit, J., & Meschke, G. (2016b). Advanced finite element beim Tunnelbau. Bautechnik, 89, 764–776 (in German).
modeling of excavation and advancement processes in mechanized Marwan, A., Alsahly, A., Gall, V., & Meschke, G. (2017). Computational
tunneling. Advances in Engineering Software, 100, 198–214. modelling for segmental lining installation in mechanized tunneling. In
Arnau, O., & Molins, C. (2012). Three dimensional structural response of G. Hofstetter, K. Bergmeister, J. Eberhardsteiner, G. Meschke, & H.F.
segmental tunnel linings. Engineering Structures, 44, 210–221. Schweiger (Eds.), Proceedings of the 4th international conference on
60 V.E. Gall et al. / Underground Space 3 (2018) 45–60

computational methods in tunneling and subsurface engineering (EURO: Schulze, H., & Duddeck, H. (1964a). Spannungen in schildvorgetriebenen
TUN 2017) (pp. 153–160). Tunneln. Beton- und Stahlbetonbau, 59, 169–175 (in German).
Meschke, G., Nagel, F., & Stascheit, J. (2011). Computational simulation Schulze, H. & Duddeck, H. (1964b). Statische Berechnung schildvor-
of mechanized tunneling as part of an integrated decision support getriebener Tunnel. In Beton- und Monierbau Aktiengesellschaft 1889-
platform. Journal of Geomechanics (ASCE), 11, 519–528, Special 1964 (pp. 87–113). Düsseldorf (in German).
Issue: Material and Computer Modeling. Smarslik, M., Putke, T., Marwan, A., Gall, V. E., Meschke, G., & Mark,
Nagel, F., & Meschke, G. (2007). Three-phase modeling in partially P. (2017). Berechnungsmodelle für Bau- und Endzustände von
saturated soils. Proceedings in Applied Mathematics and Mechanics, 7, Tübbingtunneln. In Deutsche Gesellschaft für Geotechnik e. V.
4070009–4070010. (Ed.), Taschenbuch für den Tunnelbau 2018 (pp. 111–146). Ernst &
Nagel, F., & Meschke, G. (2010). An elasto-plastic three phase model for Sohn GmbH & Co. KG (in German).
partially saturated soil for the finite element simulation of compressed Städing, A. (2007). Empfehlungen zu Ausführung und Einsatz unbe-
air support in tunnelling. International Journal for Numerical and wehrter Tunnelinnenschalen: Deutscher Ausschuss für unterirdisches
Analytical Methods in Geomechanics, 34, 605–625. Bauen (DAUB, Arbeitskreis unbewehrte Tunnelinnenschalen, stand:
Nagel, F., & Meschke, G. (2011). Grout and bentonite flow around a 24. april 2000.). Tunnel, 5, 19–28 (in German).
TBM: Numerical simulations addressing its impact on surface settle- Sugimoto, M. (2006). Causes of shield segment damages during construc-
ments. Tunnelling and Underground Space Technology incorporating tion. In International symposium on underground excavation and
Trenchless Technology Research, 26, 445–452. tunnelling (pp. 67–74). Bangkok, Thailand.
Nagel, F., Stascheit, J., & Meschke, G. (2008). A numerical simulation Terzaghi, K. (1946). Rock defects and loads in tunnel supports. In R. V.
model for shield tunnelling with compressed air. Geomechanics and Proctor, & T. L. White (Eds.), Rock tunneling with steel supports
Tunneling, 1, 222–228. (pp. 17–99). Youngstown, Ohio, USA: The Commercial Shearing
Nagel, F., Stascheit, J., & Meschke, G. (2009). A simulation model for and Stamping Co.
shield tunnelling and its interactions with partially saturated soil. Tiberti, G., Conforti, A., & Plizzari, G. A. (2015). Precast segments under
Proceedings of Applied Mathematics and Mechanics, 9, 215–216. TBM hydraulic jacks: Experimental investigation on the local splitting
Neugebauer, P., Schindler, S., Pähler, I., Blome, A., & Mark, P. (2015). behavior. Tunnelling and Underground Space Technology, 50, 438–450.
Präventives Schädigungsmanagement im Tunnelbau: Schutz der Voellmy, A. (1937). Eingebettete Rohre (Ph.D. thesis). Eidgenössische
oberirdischen Bebauung. In: Deutsche Gesellschaft für Geotechnik e. Technische Hochschule Zürich (in German).
V. (Ed.), Taschenbuch für den Tunnelbau 2015 (pp. 318–361). Ernst & Windels, R. (1967). Kreisring im elastischen Kontinuum. Der Bauinge-
Sohn (in German). nieur, 42, 429–439 (in German).
Ninić, J., Stascheit, J., & Meschke, G. (2014). Beam-solid contact Wood, A. M. (1975). The circular tunnel in elastic ground. Géotechnique,
formulation for finite element analysis of pile-soil interaction with 25, 115–127.
arbitrary discretization. International Journal for Numerical and Yu, H. (1998). CASM: A unified state parameter model for clay and sand.
Analytical Methods in Geomechanics, 38, 1453–1476. International Journal for Numerical and Analytical Methods in Geome-
Putke, T. (2016). Optimierungsgestützter Entwurf von Stahlbetonbauteilen chanics, 48, 773–778.
am Beispiel von Tunnelschalen (Ph.D. thesis). Ruhr-Universität Zhan, Y. (2016). Multi-level modeling of fiber reinforced concrete and
Bochum (in German). application to numerical simulations of tunnel lining segments (Ph.D.
Putke, T., Bohun, R., & Mark, P. (2015). Experimental analyses of an thesis). Ruhr University Bochum.
optimized shear-load transfer in the circumferential joints of concrete Zhan, Y., & Meschke, G. (2016). Multilevel computational model for
segmental linings. Structural Concrete, 16, 572–582. failure analysis of steel-fiber–reinforced concrete structures. Journal of
Putke, T., Bergmeister, K., & Mark, P. (2016). Wirtschaftliches Konstru- Engineering Mechanics (ASCE), 142, p. 04016090(1–14).
ieren und Bewehren. In K. Bergmeister, F. Fingerloos, & J. D. Wörner Zhang, J.-L., Vida, C., Yuan, Y., Hellmich, C., Mang, H. A., & Pichler, B.
(Eds.), Beton-Kalender 2016 (pp. 695–739). Ernst & Sohn (in German). (2017). A hybrid analysis method for displacement-monitored seg-
Schindler, S., Hegemann, F., Koch, C., König, M., & Mark, P. (2016). mented circular tunnel rings. Engineering Structures, 148, 839–856.
Radar interferometry based settlement monitoring in tunnelling: Zhao, C., Lavasan, A. A., Barciaga, T., Kämper, C., Mark, P., & Schanz,
Visualisation and accuracy analyses. Visualization in Engineering, 4 T. (2017). Prediction of tunnel lining forces and deformations using
(1), 1–16. analytical and numerical solutions. Tunneling and Underground Space
Schmid, J. (1926). Statische Grenzprobleme in kreisförmig durchörtertem Technology, 64, 164–176.
Gebirge (Ph.D. thesis). Eidgenössische Technische Hochschule Zürich
(in German).