You are on page 1of 16

HOSTED BY Available online at www.sciencedirect.

com

ScienceDirect
Soils and Foundations 60 (2020) 592–607
www.elsevier.com/locate/sandf

Technical Paper

U-shaped tunnel lining design using the Hyperstatic Reaction Method


– Influence of the invert
Dianchun Du a, Daniel Dias b,c,⇑, Ngocanh Do d, Tronghung Vo d
a
School of Civil Engineering, Southeast University, Nanjing 211189, China
b
Univ. Grenoble Alpes, CNRS, Grenoble INP, 3SR, F 38000 Grenoble, France
c
Antea Group, Antony, France
d
Department of Underground and Mining Construction, Faculty of Civil Engineering, Hanoi University of Mining and Geology, Viet Nam

Received 21 January 2019; received in revised form 20 December 2019; accepted 9 February 2020
Available online 16 April 2020

Abstract

A numerical approach, the Hyperstatic Reaction Method (HRM), is presented in this paper to study the behavior of two kinds of
U-shaped tunnel linings (with and without an inverted arch). The developed numerical HRM model is validated by a comparison with
the results obtained from finite difference numerical models using FLAC3D. The results of this paper show that the behavior of U-shaped
tunnel supports could be effectively estimated by the proposed HRM model. Using the HRM model, the effect of the shear springs and
the toe constraint on the behavior of U-shaped tunnel linings without an invert are studied. In addition, the results of two kinds of
U-shaped tunnels obtained by the HRM are compared with those of a circular-shaped kind. A parametric investigation is conducted
considering the changes in the stiffness of the springs, the loads on the supports, the geometries, and the bending stiffness of the support
structures.
Ó 2020 Production and hosting by Elsevier B.V. on behalf of The Japanese Geotechnical Society. This is an open access article under the CC BY-
NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

Keywords: U-shaped tunnel; Lining design; Hyperstatic Reaction Method; Soil/structure interaction

1. Introduction As mentioned in the works of some authors (Hoek and


Brown, 1980; Brown et al., 1983; Panet, 1995; Oreste,
Tunnels are popular in the construction of urban sub- 2007), simplified analytical methods, such as the
ways, railways, and expressways. In particular, U-shaped convergence-confinement method, can be used for the
tunnels (with and without an inverted arch) are widely used dimensioning of support structures. However, these meth-
in practice. However, there is no effective way to study the ods can only be used in the case of circular tunnels. In addi-
dimensioning of the tunnel supports employed at the pre- tion, only the mean displacements and the mean loads of
liminary stages of the U-shaped tunnel construction. the supports were given in the above works; the internal
Therefore, it is necessary to find an effective way to verify forces that are used to determine the correct dimensioning
the influence of the various calculation parameters on the of the support structures were not supplied. To obtain the
dimensioning of U-shaped tunnel supports. axial forces and bending moments of the support structures
of circular tunnels, a simplified approach was presented
(Einstein and Schwartz, 1979).
Peer review under responsibility of The Japanese Geotechnical Society. The tunnel support design of U-shaped tunnels was
⇑ Corresponding author at: Univ. Grenoble Alpes, CNRS, Grenoble studied by some authors, using different methods, such as
INP, 3SR, F-38000 Grenoble, France. the topology optimization method (Yin and Yang, 2000)
E-mail address: daniel.dias@anteagroup.com (D. Dias).

https://doi.org/10.1016/j.sandf.2020.02.004
0038-0806/Ó 2020 Production and hosting by Elsevier B.V. on behalf of The Japanese Geotechnical Society.
This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
D. Du et al. / Soils and Foundations 60 (2020) 592–607 593

Notation

C/D cover-to-diameter of the tunnel ps;lim maximum shear reaction pressure


c cohesion values of the ground plim maximum reaction pressure offered by the
E elastic modulus of the ground ground
E0 initial elastic modulus of the ground q displacement vector of the nodes
Es elastic modulus of the concrete q1, q2, q3, . . ., qn+1 the sub-vectors composed of the three
EJ, EA bending and normal stiffness of the support displacements of each node
F external force vector of the nodes qh horizontal loads
F1, F2, F3, . . ., Fn+1 sub-vectors of the external forces in qv vertical loads
the global Cartesian reference system R tunnel radius
Fi,x, Fi,y axial force and transversal force of node i Rc crown radius
Gs shear modulus of the ground Req equivalent radius of the tunnel
H overburden of the tunnel Ria radius of the invert arch
h initial node of element i t support thickness of the lining
i number of the generic node ui axial displacement of node i
j final node of element i vi transversal displacement of node i
K global stiffness matrix X, Y global Cartesian coordinates
K0 lateral earth pressure coefficient x, y local Cartesian coordinates
Ks bulk modulus of the ground yc height of the invert arch center
K* modification values of elements along the diago- zmin smallest dimension of an adjoining zone in the
nal of the global stiffness structure matrix normal direction
ki local stiffness matrix of the ith element under the ai , aiþ1 angles of element i and element i + 1 between
 global Cartesian coordinates the local and the global reference systems
ki local stiffness matrix under the local Cartesian b dimensionless factor
reference system c unit weight of the ground
ki,a, ki,b, ki,c, ki,d sub-matrices of ki d support deformation
kn stiffness of the normal Winkler-springs h angle between the vertical axis of the cross-
k n;i stiffness of the normal spring connected to node section of the tunnel and the element axis
i hi rotation of node i
k s;i stiffness of the shear spring connected to node i ki transformation matrix
Lh vertical length from the point of the tunnel g* apparent stiffness of the ground
cross-section to the spring line (x-axis) g0 initial stiffness of the ground
Li length of elementi gn,0 initial normal ground stiffness
Lp sidewall length gn normal spring stiffness
Mi bending moment of node i gs shear spring stiffness
mx scale parameter in the x-direction v Poisson’s ratio of the lining
my scale parameter in the y-direction vs Poisson’s ratio of the ground
p reaction pressure u friction angle of the ground
pn;lim maximum normal reaction pressure

or numerical modeling using a bedded-beam-spring model In recent decades, numerical simulations have attracted
(which can also be called the hyperstatic reaction method) the attention of tunnel designers thanks to the increase in
(Barpi et al., 2011). However, the topology optimization the power of computers. Numerical methods were success-
method (Yin and Yang, 2000) can only provide the tunnel fully used for dimensioning support structures by former
deformation; it cannot be used to calculate the lining researchers (Oreste, 2007; Do et al., 2013a, 2013b). How-
forces. Barpi et al. (2011) presented a method for the study ever, those numerical methods are basically complex and
of U-shaped tunnels based on a fuzzy approach and the time-consuming. To overcome such limitations, a way to
bedded-beam-spring model. However, there are some dif- calculate the lining forces should be found that involves
ferences between the present method and the one in less computational effort than the one necessary for the
Barpi et al. (2011). Indeed, the tunnel invert shape was numerical model. The Hyperstatic Reaction Method
modeled as a straight line by Barpi et al. (2011). The stiff- (HRM) meets this requirement.
ness of the springs is linear and depends only on the equiv- The HRM, proposed and developed by many research-
alent radius Req of the tunnel, the elastic modulus E, and ers (Duddeck and Erdmann, 1985; Leca and Clough, 1992;
the Poisson coefficient vs of the ground. Oreste, 2007), is a Matlab-based ‘‘numerical” method
594 D. Du et al. / Soils and Foundations 60 (2020) 592–607

which can overcome the limitations of a finite element or a be noted that the ground reaction caused by the deforma-
finite difference numerical analysis, e.g., the model bound- tion of the supports is a function of the structural stiffness.
ary effect and the mesh dependency. Due to a large number This reaction is determined using the geometrical and
of connections, namely, the independent Winkler springs mechanical characteristics of the support structures. The
used to simulate the interaction between the support struc- active loads are independent of the displacements that
tures and the ground surrounding the tunnel, the method is develop in the supports and at the ground-support
called hyperstatic. Considering the mechanical characteris- interface.
tics of the ground and the geometry of the tunnel lining, a Although many existing models using the HRM have
specific code (called FEMSUP) was proposed (Oreste, been proposed and improved, most of them are mainly
2007) to perform calculations using the HRM approach. applicable to circular tunnels. In fact, U-shaped tunnels,
However, the solution of Oreste (2007) only considered such as horseshoe-shaped tunnels (with and without an
the normal deformations and ignored the tangential defor- inverted arch) are also often encountered in practice.
mations between the supports and the ground. Based on Therefore, for an analysis of the behavior of U-shaped tun-
the work of Oreste (2007), Do et al. (2014a) extended the nel supports, this paper presents an improved numerical
HRM approach to study the behavior of segmental tunnel approach to the HRM developed on the basis of the mod-
supports, considering the normal and tangential deforma- els proposed by Oreste (2007) and Do et al. (2014a). Two
tions between the supports and the ground. different cases of springs, only normal springs and normal
Three parameters, namely, the active loads, the stiffness springs plus shear springs, are considered here to simulate
of the structures, and the pressure-displacement relation the external load scheme. Varying loads and two kinds of
between the structures and the ground, have to be consid- U-shaped tunnels (with and without an inverted arch) are
ered in the analysis of the stress state of the supports when considered in this study (Fig. 1). In addition, the effect of
dealing with the Hyperstatic Reaction Method. It should the toe constraint conditions on the behavior of

Fig. 1. Calculation scheme of two different U-shaped support structures with HRM: (a) U-shaped tunnel with inverted arch and scheme of ground-
support interaction through springs connected to support nodes and (b) U-shaped tunnel without inverted arch. In the figure, qv: vertical loads, qh:
horizontal loads, kn: stiffness of the Winkler springs, Ria: radius of the inverted arch, Rc, crown radius, Lh, vertical length from the point of the tunnel
cross-section to the spring line (x-axis), and EJ and EA: bending and normal stiffness of the supports.
D. Du et al. / Soils and Foundations 60 (2020) 592–607 595

2 3
U-shaped tunnel linings without an inverted arch is stud- k 1;a k 1;b 0 0 0 0
ied. The results of the two kinds of U-shaped tunnels 6 k 1;c k 1;d þ k 2;a 0 0 0 7
6 k 2;b 7
obtained by the HRM are compared with those of a 6 7
6 0 k 2;c k 2;d þ k 3;a k 3;b 0 0 7
circular-shaped kind. It should be noted that the necessary 6 7
6 0 k 3;d þ k 4;a ::: 0 7
time for each calculation, using the HRM calculation code, 6 0 k 3;c 7
6 7
is very short (less than 4 s on an Intel CPU 2 GHz PC). For 4 0 0 0 ::: ::: k n;b 5
this reason, the HRM method is particularly suitable for 0 0 0 0 k n;c k n;d
analyses which necessitate a lot of calculations, like proba- 2 3 2 3
q1 F1
bilistic studies. 6 7 6
Comparing the results obtained by the HRM method 6 7 6
q2 F2 77
6 7 6 7
with the ones obtained by the finite difference program 6 7 6
q3 F3 7
6
6
7¼6
7 6
7 ð2Þ
FLAC3D, the HRM improved in this paper is validated. 6 :::
7 6 ::: 7
7
Therefore, the proposed method could be used to effec- 6 7 6 7
4 ::: 5 4 ::: 5
tively design U-shaped tunnel linings.
qnþ1 F nþ1
with ki,a, ki,b, ki,c, and ki,d: the sub-matrices of ki, which
2. Hyperstatic reaction method represent 3  3 sub-matrices of the local stiffness matrix
k i of the ith element under the global Cartesian coordinates,
The Hyperstatic Reaction Method (HRM), as a q1, q2, q3, . . ., qn+1: the sub-vectors composed of the three
numerical method, is suitable for designing tunnel linings displacements of each node, and F1, F2, F3, . . ., Fn+1: the
(Fig. 1). In this method, the interaction between the sub-vectors of the external forces in the global Cartesian
ground and the tunnel supports is built through springs reference system. q and F can be found in
connected to the nodes of the elements and the active The local stiffness matrix k i of the ith element under the
loads (qv and qh in Fig. 1). As described in Oreste global Cartesian reference system is obtained by
(2007) and Do et al. (2014a), the tunnel lining can be 
represented by some mono-dimensional elements (with ½ki ¼ kTi  ½ki  ki ð3Þ
three degrees of freedom) (Fig. 2). Those elements, which
are able to develop bending moments, axial forces, and where ai is the angle of inclination of element i with respect

shear forces, are connected to the ground mass through to the horizontal (Fig. 1a), k i is the local stiffness matrix
springs distributed over the nodes (Fig. 1). The basic under the local Cartesian reference system, and ki is the
idea of the HRM method is to obtain the displacement transformation matrix.
components of the nodes. Then, the internal forces and 2 EA 3
0 0  EA 0 0
the displacements of each element can be known after Li Li
6 7
obtaining those unknown displacements. An evaluation 6 0 12EI 6EI
0  12EI 6EI 7
6 L3i L2i L3i L2i 7
of the unknown displacements is achieved through the 6 7
6 0  2EI 7
definition of the global stiffness matrix of the entire  6 6EI 4EI
0 6EI
Li 7
6 7
2 Li 2
Li Li
structure and of its connections to the surrounding ½ki ¼ 6 7 ð4Þ
6  EA 0 0 EA
0 0 7
environment. 6 Li Li 7
6 7
The global stiffness matrix, K, can be assembled by the 6 0  12EI  6EI 0 12EI
 6EI 7
6 L3i L2i L3i L2i 7
local stiffness matrices of each element according to 4 5
Huebner et al. (2001). The unknown parameters of the 0 6EI
L2
2EI
Li
0  6EI
L2
4EI
Li
i i
problem are the displacement components q of the nodes 2 3
of those elements, which can be evaluated as follows: cos ai sinai 0 0 0 0
6 7
6 sinai cos ai 0 0 0 07
K q¼F ð1Þ 6 7
6 0 0 1 0 0 07
6 7
ki ¼ 6 7 ð5Þ
6 0 0 0 cos ai sinai 07
6 7
6 7
4 0 0 0 sinai cos ai 05
0 0 0 0 0 1
The components of matrix k i are shown in
It should be noted that the stiffness of the structural ele-
ments is modified in the corresponding directions due to
Fig. 2. Beam-type finite element with reference to the local Cartesian
the presence of normal and shear node springs (Fig. 1a).
coordinates. In the figure, h: initial node, j: final node, uh, uj: axial
displacement, vh, vj: transversal displacement, h: rotation, x and y: local The modification of these elements along the diagonal of
Cartesian coordinates, and Li: distance between the terminal connecting the global stiffness structure matrix is given by the follow-
nodes. ing equations:
596 D. Du et al. / Soils and Foundations 60 (2020) 592–607
a a i p
K3i2;3i2 ¼ K3i2;3i2 þ kn;i  cos2
iþ1
þ  þ ks;i 1 E0
2 2 2 gn;0 ¼ b   ð11Þ
a 1 þ vs R
iþ1 a i p
 sin2 þ  ð6Þ with b: dimensionless factor, vs: Poisson’s ratio of the
2 2 2
a ai p ground, E0: initial elastic modulus of the ground, and R:
K3i1;3i1 ¼ K3i1;3i1 þ kn;i  sin2
iþ1
þ  þ ks;i tunnel radius (radius of the crown tunnel).
2 2 2
a ai p  Factor b is difficult to assess, as ground stiffness is not a
iþ1
 cos2 þ  ð7Þ material constant. The value of b has been taken as being
2 2 2
equal to 1 in conventional studies (Möller, 2006;
K3i1;3i2 ¼ K3i1;3i2 þ ðkn;i  ks;i Þ
a Mashimo and Ishimura, 2003, 2005; Plizzari and Tiberti,
iþ1 a i p a
iþ1 a i p
 sin þ   cos þ  ð8Þ 2006; Arnau and Molins, 2010; Do et al., 2014b). In this
2 2 2 2 2 2 paper, the value of b is also taken as being equal to 1.

K3i2;3i1 ¼ K3i2;3i1 þ ðkn;i  ks;i Þ The tangential stiffness (gs) is one-third of the normal stiff-
a a i p a a i p ness when considering the shear springs (Möller, 2006;
iþ1 iþ1
 sin þ   cos þ  ð9Þ Mashimo and Ishimura, 2003, 2005; Plizzari and Tiberti,
2 2 2 2 2 2
2006; Arnau and Molins, 2010; Do et al., 2014b).
where i denotes the number of the generic node, k n;i repre-
As the shear spring stiffness is not easy to estimate, a
sents the stiffness of the normal spring connected to node i,
simple relationship between normal stiffness gn and shear
k s;i is the stiffness of the shear spring connected to node i,
stiffnessgs is adopted (Plizzari and Tiberti, 2006; Arnau
and ai and aiþ1 are the angles of element i and element
and Molins, 2010).
i + 1 between the local and the global reference systems
(Fig. 1a), respectively. ‘‘K*” means the modification values 1
gs ¼ gn ð12Þ
of these elements along the diagonal of the global stiffness 3
structure matrix. The formulas for k n;i and k s;i will be
Maximum normal reaction pressure pn;lim depends on
shown in Section 3.
cohesion c, friction angle u, Poisson’s ratio vs of the
Details of the mathematical formulas for this method
ground, and the active loads.
are given in Oreste (2007) and Do et al. (2014a). It should
be noted that the stiffness of the shear springs is disre- 2  c  cosu 1 þ sinu qh þ qv vs
pn;lim ¼ þ   ð13Þ
garded when only normal springs are considered. Once vec- 1  sinu 1  sinu 2 1  vs
tor q of the nodal displacement from Eq. (2) is obtained,
Maximum shear reaction pressure ps;lim is estimated by
the strain and stress at the nodes of the supports can be
calculated. qh þ qv
ps;lim ¼  tanu ð14Þ
2
3. Ground-support interaction and active loads
Then, the normal stiffness k n;i and shear stiffness k s;i of
each spring are given by the following formulas:
As mentioned above, the structure supports interact  
with the surrounding mass in two ways: through springs  ðLi1 þ Li Þ
kn;i ¼ gn;i  1
connected to the nodes of the structures and through the 2
applied active loads. Two different cases of springs, only !
pn;lim pn;lim ðLi1 þ Li Þ
normal springs and normal springs plus shear springs, are ¼  1  ð15Þ
considered in this work. dn;i pn;lim þ gn;0  dn;i 2
This relation between reaction pressure p and support  
ðLi1 þ Li Þ
deformation d is taken as being equal to (Oreste, 2007) ks;i ¼ gs;i  1
  2
plim !
p ¼ plim  1  ð10Þ ps;lim ps;lim ðLi1 þ Li Þ
plim þ g0  d ¼  1  ð16Þ
ds;i ps;lim þ gs;0  ds;i 2
where plim: the maximum reaction pressure offered by the
ground and g0: the initial stiffness (when the value of d is where Li is the length of element i (Fig. 1a). More details on
close to 0) of the ground. The apparent stiffness g* of the the HRM can be found in the work of Do et al. (2014a).
ground is given by the p/d ratio. It should be noted that when the lining moves towards
The values of maximum reaction pressure plim are esti- the tunnel in practice, the normal springs will be deacti-
mated considering the effect of the confining pressure, fric- vated. Therefore, only compressive loads are allowed by
tion angle u, and the cohesion values c of the ground (Do the normal springs in this method. In contrast, shear
et al., 2014a, 2013b). springs allow both positive and negative stresses along
This non-linear relationship will also be applied to both the tangential direction.
the normal ground stiffness and the tangential ground stiff- Generally, the active loads can be estimated backwards
ness. Initial normal ground stiffness gn,0 can be estimated as on the basis of in situ tests using back-analysis procedures.
per Möller (2006) as In this paper, linear variable loads applied on the tunnel
D. Du et al. / Soils and Foundations 60 (2020) 592–607 597

supports are considered (Fig. 1). These variable loads are


calculated at every point of the tunnel cross-section. In
the two different U-shaped tunnel HRM models proposed
here (Fig. 1), the vertical ground loads on the supports can
be estimated as
qv ¼ cðH þ Lh Þ ð17Þ
where Lh: the vertical length between the point of the tun-
nel cross-section and the spring line (x-axis), H: the over-
burden of the tunnel, and c: the unit weight of the ground.
However, the horizontal loads working on the lining can
be estimated as
qh ¼ K 0 qv ð18Þ
Fig. 4. Support geometries considered in the FEMSUP calculation code:
where K0: the lateral earth pressure coefficient. (a) Geometry1 (U-shaped tunnel with inverted arch) and (b) Geometry2
Fig. 3 provides a flow chart of the calculation process by (U-shaped tunnel without inverted arch). In this figure, Ria: radius of the
the HRM. inverted arch, Rc: crown radius, Lp: sidewall length, yc: height of the
A comparison between the results obtained for both inverted arch center, h: angle between the vertical axis of the cross-section
of the tunnel and the element axis, which can be used to determine the
kinds of U-shaped tunnels is also presented in this paper. coordinates (h is 00 at the initial node and is 900 at the connection of the
sidewall and the crown), and X and Y are the global Cartesian
4. FEMSUP calculation code and dimensioning of the coordinates.
supports

The FEMSUP calculation code is able to consider the


support geometries (Fig. 4) and the types of constraints
at the bases of the tunnel supports (Fig. 5), which makes
this code available for the dimensioning of the preliminary
supports. Due to the symmetry of the problem, only half of
the tunnel is considered here.
An analysis for the verification of the preliminary sup-
port is given here as an example (Toulon Tunnel, France)
using the FEMSUP calculation code. To reduce the length Fig. 5. Type of constraints: (a) horizontal hinged-roller, (b) fixed hinge,
of this paper, only the clamped constraint condition (c) horizontal clamped-roller, (d) clamped, and (e) vertical clamped-roller.
(Fig. 5d, which corresponds to the worst case of internal ((a)-(d) are for the final node of Geometry2 and (e) is for the final node of
forces (Fig. 7), is adopted as a reference case to investigate Geometry1 and for the initial nodes of both U-shaped tunnels).
the tunnel lining behavior in the next sections.
The geometrical and geomechanical parameters of the
tunnel, the ground, and the adopted support structure
Table 1
are shown in Table 1. Using the reference case with the
Parameters of the Toulon Tunnel.
clamped constraint condition (Fig. 5d), comparisons of
Parameter Symbol Value Unit
Properties of tunnel
Crown radius Rc 6.02 m
Invert arch radius Ria 20.22 m
Height of invert arch center yc 15.00 m
Height of sidewall Lp 3.80 m
Scale parameter in x-direction mx 1.00 –
Scale parameter in y-direction my 1.00 –
Support thickness t 0.22 m
Poisson’s ratio v 0.15 –
Elastic modulus of concrete Es 35,000 MPa
Properties of ground
Cohesion c 50 kPa
Unit weight c 22 kN/m3
Poisson’s ratio vs 0.3 –
Internal friction angle u 20 °
Elastic modulus E 300 MPa
Lateral earth pressure factor K0 0.5 –
Overburden H 40 m
Fig. 3. Flow chart of calculation process by HRM.
598 D. Du et al. / Soils and Foundations 60 (2020) 592–607

the numerical results obtained from the two different U- forces are obvious at the tunnel sidewall, and the results
shaped tunnels, with only normal springs and with normal of Geometry1 are higher than those of Geometry2. The
springs plus shear springs, are presented in Fig. 6. normal forces obtained from Geometry1 are clearly lower
Fig. 6 shows that the maximum absolute values of the than those obtained from Geometry2 at the tunnel crown
bending moments and the shear forces are located at the and the sidewall. In contrast, the radial displacements
sidewall for both kinds of U-shaped tunnels. The normal obtained from Geometry1 are higher than those obtained
forces of Geometry1 in the case of considering only normal from Geometry2 at the tunnel crown and the sidewall.
springs are smaller than those in the case of considering An analysis is also conducted to show the effect of the con-
both normal springs and shear springs. These normal straint conditions on the tunnel lining behavior of Geome-
forces are slightly higher at the tunnel’s inverted arch. In try2 (Fig. 7).
contrast, the normal forces of Geometry2 in the case of From Fig. 7, it can be seen that the effect of the con-
considering only normal springs are only smaller at the straint conditions on the structural forces and the displace-
tunnel crown than those in the case of considering both ments at the tunnel crown can be disregarded. However,
normal springs and shear springs. they affect the bending moments, the shear forces, and
The radial displacements obtained in the case of Geom- the radial displacements at the tunnel sidewall. Constraint
etry2 are smaller than those of Geometry1. The maximum (d) yields the largest bending moment (Fig. 7a) because of
radial displacements of Geometry1 are located at the center the restrained rotation at the bases of the tunnel lining
of the tunnel crown and at the sidewall and inverted arch (Fig. 7f). On the other hand, the smallest bending moments
connection. The same maximum radial displacements for are found in the cases of constraints (a) and (b) where the
Geometry2 can be observed at the center of the tunnel bases of the tunnel lining are free for rotation (Fig. 7f). The
sidewall. maximum shear forces are observed at the tunnel sidewall
In addition, Fig. 6 also shows that the structural forces in the case of constraint (d). It should be noted that the
and displacements of Geometry2 experience no significant radial displacements of the tunnel lining are clearly differ-
change when considering the shear springs, which means ent for those four constraint cases at the tunnel sidewall.
that the spring type has little effect on the internal forces In addition, it can also be found that Geometry2 has
and displacements of Geometry2. almost the same shear displacements under those four con-
Furthermore, Fig. 6 shows that Geometry1 and Geom- straint conditions, which means that the effect of the con-
etry2 provide almost the same values for the bending straint types on the shear displacements can be ignored.
moments and the shear forces at the tunnel crown. Never- Considering the influence of the normal springs and the
theless, the differences in bending moments and shear shear springs, comparisons of the results between circular-

Fig. 6. Comparisons of results obtained from two different U-shaped tunnels with only normal springs and with normal springs plus shear springs: (a)
bending moment, (b) normal force, (c) shear force, and (d) radial displacement (positive towards the inside of the tunnel).
D. Du et al. / Soils and Foundations 60 (2020) 592–607 599

Fig. 7. Effect of constraint conditions on tunnel lining behavior of Geometry2: (a) bending moment, (b) normal force, (c) shear force, (d) radial
displacement, (e) shear displacement, and (f) rotations.

shaped and two kinds of U-shaped tunnels obtained by the from the circular tunnel are between the ones of the Geom-
HRM calculations are also shown in Fig. 8. The parame- etry1 and Geometry2 cases. The normal forces of the
ters used in the comparison are the same as those presented Geometry2 case are higher than those of the Geometry1
in Table 1 and Geometry2 in Fig. 4 under the reference case case and the circular case at the tunnel crown and the side-
whereby the clamped constraint condition (Fig. 5d) is wall. However, the bending moments and the shear forces
considered. not only show similar trends, but also have approxima-
It can be seen from Fig. 8 that the internal forces and the tively the same values at the tunnel crown and the sidewall
displacements obtained from the three different geometries for Geometry1 and Geometry2. The trends of the internal
have similar trends at the tunnel crown, and that the results forces and the displacements obtained from the circular
of the normal forces and the radial displacements obtained tunnel are smoother than those of the Geometry1 and
600 D. Du et al. / Soils and Foundations 60 (2020) 592–607

Fig. 8. Comparisons of the results between circular-shaped and two different U-shaped tunnels obtained by HRM: (a) bending moment, (b) normal force,
(c) shear force, and (d) radial displacement (positive towards the inside of the tunnel).

Geometry2 cases at the tunnel sidewall and the inverted response of the mechanical system that is composed of a
arch. solid continuum and a framework of load-carrying mem-
bers can be simulated. The support is represented by a col-
5. Validation of the HRM method lection of structural elements (shells), each of them being
associated with a set of nodes. It interacts with the solid
A comparison between the HRM and a 2D numerical continuum by means of links which connect the structural
model is shown in this section, which makes it possible to nodes to the soil mass nodes. In the present comparison,
estimate the accuracy achieved by means of the present the ground is assumed to follow a linear elastic-perfectly
improved HRM. The 2D numerical simulation was done plastic constitutive relation based on the Mohr-Coulomb
using the finite difference program FLAC3D (Itasca, 2009) shear failure criterion under plane-strain conditions.
under the hypothesis of plane strain conditions. The mate- Parameters from the Toulon Tunnel project in the city
rials are represented by polyhedral elements within a three- of Toulon are adopted (Oreste and Dias, 2012). The
dimensional grid that is adjusted by the user to fit the shape parameters of the tunnel, the ground, and the adopted sup-
of the object to be modeled (Itasca, 2009). The structural port structure are shown in Table 1.

Fig. 9. Numerical geometry model of U-shaped tunnel without inverted arch considered in this paper.
D. Du et al. / Soils and Foundations 60 (2020) 592–607 601

Fig. 10. Numerical model: (a) two-dimensional numerical model, (b) U-shaped tunnel without inverted arch, and (c) vertical displacements (m) in
FLAC3D.

The 2D numerical model of the U-shaped tunnel with-


out an inverted arch (taking Geometry2 in Fig. 4 as an
example) is shown in Figs. 9 and 10. The dimensions of
the mesh elements increase with the distance to the tunnel.
The FLAC3D model is 60 m long in the z-direction, 240 m
wide in the x-direction, and consists of 6082 zones and
12,362 grid points. Liner elements are embedded along
the tunnel boundary and attached to the surrounding
ground zones. A continuous support is simulated for a
comparison with the results of the HRM in this work.
The apparent stiffness of a zone in the direction normal
to the surface is estimated as (Itasca, 2009)
 
K s þ 43 Gs
max ð19Þ Fig. 11. Flow chart of calculation process by FLAC.
Dzmin

where Ks and Gs: the bulk and shear moduli, respectively, at the tunnel supports through the continuous media,
and zmin: the smallest dimension of an adjoining zone in and the arching effects can then be simulated.
the normal direction. Similar results are obtained by HRM and FLAC for all
The numerical simulation is conducted according to the the five constraint conditions applied at the bases of the
following steps, as shown in Fig. 11. tunnel lining (see Fig. 5). Considering the clamped con-
Comparing the two methods, HRM and FLAC3D, it is straint condition in Fig. 5d as an example, the results
found that the lining is divided into a number of elements obtained from the two methods are shown in Fig. 12.
in both methods. The main difference lies in the fact that Fig. 12 illustrates the structural forces and the displace-
the presence of the ground is considered through the exter- ments of the tunnel supports. While no slip is assumed in
nal loads (caused by the ground gravity). These loads act the FLAC3D model, which means a completely shear
directly and explicitly on the nodes of the beam elements resistance, the HRM model considers both cases of only
through the springs in the HRM method. In contrast, the normal springs (Case 1) and normal springs plus shear
loads in the numerical method using FLAC3D are applied springs (Case 2). It can be seen that the normal forces
602 D. Du et al. / Soils and Foundations 60 (2020) 592–607

(between FLAC and HRM with normal and shear


springs) is equal to 12.3% and is located at the tunnel
sidewall. This difference could be related to the different
ways the ground surrounding the tunnel is considered.
However, the bending moments in the case of considering
both normal and shear springs are slightly smaller than
those in the case of considering only normal springs along
the whole tunnel. The bending moments calculated by the
HRM have almost the same trend as those calculated
with FLAC3D for the whole tunnel. Similarly, the results
of the radial displacements obtained from FLAC3D show
the same trend as the results obtained from the HRM
along the whole tunnel.
Based on the above-mentioned results, it is therefore
reasonable to conclude that the support displacements
and the structural internal forces obtained from the
HRM are basically in good agreement with the ones
obtained from the numerical FLAC3D with respect to the
positions along the tunnel boundary.

6. Results and parametric analysis

Taking Geometry2 in Fig. 4b with the clamped con-


straint condition (Fig. 5d) as the reference case, a paramet-
ric analysis is developed to study the effect of the different
levels of stiffness of the springs, different loads on the sup-
port, different geometries, and different levels of bending
stiffness of the support structures on the structural forces
and displacements. From Eq. (11), it can be found that
the stiffness of the springs is related to the ground elastic
modulus and the tunnel radius. Details of the cases
Fig. 12. Comparison of structural forces and displacements in U-shaped adopted for the parametric analyses are shown in Table 2.
tunnel supports (Geometry2 in Fig. 4) between HRM and FLAC3D Two values of lateral earth pressure factor K0 are consid-
model: (a) normal force, (b) bending moment, and (c) radial displacement. ered (K0 = 0.5 and 1.5) and the tunnel dimension is chan-
ged to the range of 2–5 m. Four different types of ground
are considered and the parametric analyses are done using
in the case of considering both normal and shear springs the FEMSUP code.
(Case 2) are larger than those in the case of considering The values of the bending moments, normal forces, and
only normal springs (Case 1) at the tunnel crown, but normal displacements at the tunnel sidewall and the
smaller at the tunnel sidewall. Moreover, it can be clearly tunnel crown have been determined. The results obtained
seen that FLAC3D yields slightly lower normal forces for the four different types of ground are illustrated in
than HRM at the tunnel crown, but slightly higher nor- Figs. 13–16. Some interesting conclusions are obtained
mal forces at the tunnel sidewall. The maximum difference from the parametric analyses.

Table 2
Details of cases adopted for parametric analyses.
Description/Case 1 2 3 4 5 6 7 8
Properties of tunnel A A B B C C D D
Elastic modulus Es (MPa) of ground 50 50 150 150 500 500 1500 1500
Friction angle u (°) 28 31 35 45
Cohesion c (MPa) 0.005 0.02 0.5 0.75
Lateral earth pressure factor K0 0.5 1.5 0.5 1.5 0.5 1.5 0.5 1.5
Unit weight c (kN/m3) of ground 22
Young’s modulus Es (GPa) of liner 35
Tunnel crown radius Rc (m) Changed for three cases: R1 = 2.0 m, R2 = 3.5 m, and R3 = 5.0 m
Cover-to-diameter C/D values Five values were considered: 2, 4, 10, 20, and 40
Support thickness t (m) Two cases were considered: t1 = 0.20 m and t2 = 0.50 m
D. Du et al. / Soils and Foundations 60 (2020) 592–607 603

Fig. 13. Structural forces and displacement considering different levels of stiffness of the springs, different loads on the support, different geometries, and
different levels of bending stiffness of the support structures (Case 1): (a) bending moments at point B in Fig. 13b, (b) bending moments at point A, (c)
normal force at point B, (d) normal force at point A, (e) normal displacement at point B, and (f) normal displacement at point A.

1. The higher the ground quality (the higher the stiffness of 3. The larger the loads (the cover-to-diameter C/D
the springs), the lower the bending moments, normal ratio) on the support, the higher the normal forces,
forces, and normal displacements. the bending moments, and the normal
2. The thicker the tunnel support (the higher the bending displacements.
stiffness of the support structures), the higher the normal 4. The larger the radius of the tunnel, the higher the nor-
forces and the bending moments, but the lower the nor- mal forces, the bending moments, and the normal
mal displacement. displacements.
604 D. Du et al. / Soils and Foundations 60 (2020) 592–607

Fig. 14. Structural forces and displacement considering different levels of stiffness of the springs, different loads on the support, different geometries, and
different levels of bending stiffness of the support structures (Case 3): (a) bending moments at point B, (b) bending moments at point A, (c) normal force at
point B, (d) normal force at point A, (e) normal displacement at point B, and (f) normal displacement at point A.

5. The absolute values of the structural forces and dis- ground quality, the higher the effect of the support
placements increase with the increase in the cover-to- thickness on the normal forces (Fig. 13c and d and
diameter of C/D ratio. Fig. 16c and d).
6. For the four types of ground, the values of the normal 7. For a weak ground (ground type A), the larger the tun-
forces at the tunnel sidewall are higher than those of nel radius, the higher the absolute values of the normal
the normal forces at the tunnel crown; the higher the displacements at the tunnel sidewall (Fig. 13e).
D. Du et al. / Soils and Foundations 60 (2020) 592–607 605

Fig. 15. Structural forces and displacement considering different levels of stiffness of the springs, different loads on the support, different geometries, and
different levels of bending stiffness of the support structures (Case 5): (a) bending moments at point B, (b) bending moments at point A, (c) normal force at
point B, (d) normal force at point A, (e) normal displacement at point B, and (f) normal displacement at point A.

8. For strong ground (ground type D), the larger the tun- to be known, the structural internal forces and displace-
nel radius and the thinner the tunnel thickness, the ments can then be estimated. It is possible, therefore, to
higher the absolute values of bending moments at the conduct a detailed design of the support structure of a
tunnel crown (Fig. 16b). U-shaped tunnel and then decide the necessary reinforce-
ments, concrete quality, etc. According to the proposed
In this parametric analysis, once the ground parameters, design figures, it is also possible to verify the effect of the
the tunnel dimensions, the thickness of the supports, and various calculation parameters on the structural behavior
the values of the cover-to-diameter C/D ratio are assumed of a U-shaped tunnel support.
606 D. Du et al. / Soils and Foundations 60 (2020) 592–607

Fig. 16. Structural forces and displacement considering different levels of stiffness of the springs, different loads on the support, different geometries, and
different levels of bending stiffness of the support structures (Case 7): (a) bending moments at point B, (b) bending moments at point A, (c) normal force at
point B, (d) normal force at point A, (e) normal displacement at point B, and (f) normal displacement at point A.

7. Conclusions of the structural forces and support displacements around


the tunnel supports, the present method has been found
The Hyperstatic Reaction Method has provided the suitable for estimating the tunnel support behavior. How-
opportunity to study the behavior of two kinds of U- ever, the disadvantages of the HRM approach are that it
shaped tunnel linings (with and without an inverted arch). is difficult to accurately evaluate: (1) the stiffness of the
The calculation time can be shortened (4 s per calculation springs and (2) the active load applied to the lining by
on an Intel CPU 2 GHz PC) with this method. In terms the soil mass.
D. Du et al. / Soils and Foundations 60 (2020) 592–607 607

A comparison of two different cases of springs, namely, References


only normal springs and normal springs plus shear springs,
has been made. The conclusion obtained is that the spring Arnau, O., Molins, C., 2010. Experimental and analytical study of the
type has little effect on the internal forces and displace- structural response of segmental tunnel supports based on an in situ
loading test. Tunn. Undergr. Space Technol. 26, 778–788.
ments of Geometry2. The analysis was also conducted to Barpi, F., Barbero, M., Peila, D., 2011. Numerical modelling of ground-
show the effect of the constraint conditions (Fig. 4) on tunnel support interaction using bedded-beam-spring model with fuzzy
the lining behavior of Geometry2. It was found that the parameters. Gospodarka Surowcami Mineralnymi 27, 71–87.
effect of the constraint conditions on the bending moments, Brown, E.T., Bray, J.W., Ladanyi, B., Hoek, E., 1983. Ground response
shear forces, and radial displacements at the tunnel side- curves for rock tunnels. J. Geotech. Eng. 109, 15–39.
Do, N.A., Dias, D., Oreste, P.P., Djeran-Maigre, I., 2013a. 2D numerical
wall should not be ignored. investigation of segmental tunnel lining behavior. Tunn. Undergr.
The results from the present method have been com- Space Technol. 37, 115–127.
pared with the results of the numerical model FLAC3D. Do, N.A., Dias, D., Oreste, P.P., Djeran-Maigre, I., 2013b. Three-
The acceptable differences found between these two models dimensional numerical simulation for mechanized tunneling in soft
ground: the influence of the joint pattern. Acta Geotech. 9 (4), 673–
have allowed the HRM method to be validated.
694.
The parametric analysis permitted the estimation of the Do, N.A., Dias, D., Oreste, P.P., Djeran-Maigre, I., 2014a. A new
U-shaped tunnel support behavior. The effect of four typi- numerical approach to the hyperstatic reaction method for segmental
cal types of grounds, three values for tunnel crown radius tunnel supports. Int. J. Numer. Anal. Meth. Geomech. 38 (15), 1617–
Rc, two values for tunnel support thickness t, and five val- 1632.
ues for the cover-to-diameter C/D ratio, was investigated. Do, N.A., Dias, D., Oreste, P.P., 2014b. The behavior of the segmental
tunnel support studied by the hyperstatic reaction method. Eur. J.
The results presented in the design figures can be used for Environ. Civil Eng. 18 (4), 489–510.
the preliminary estimation of the structural internal forces Duddeck, H., Erdmann, J., 1985. On structural design models for tunnels
induced in the supports of U-shaped tunnels under the in soft soil. In: Underground Space, 9. Pergamon Press, Oxford, UK,
action of active loads. It should be noted that before using pp. 246–259.
the proposed figures for a preliminary U-shaped tunnel Einstein, H.H., Schwartz, C.W., 1979. Simplified analysis for tunnel
supports. J. Geotech. Eng. Div., Am. Soc. Civ. Eng. 105 (4), 499–518.
support design, the external loads applied to the tunnel Hoek, E., Brown, E.T., 1980. Underground Excavations in Ground. The
structure should be known, in terms of vertical load qv Institution of Mining and Metallurgy, London.
and lateral earth pressure coefficient K0. Huebner, K.H., Dewhirst, D.L., Smith, D.E., Byrom, T.G., 2001. The
Compared with other numerical methods, it can easily Finite Element Method for Engineers. John Wiley and Sons Inc, New
York.
be found that:
Itasca, Consulting. Group. Inc., 2009. FLAC fast Lagrangian analysis of
Firstly, the computational efficiency is greatly improved continua. Version 5.0, User’s manual.
when using the proposed method. This makes it possible to Leca, E., Clough, W., 1992. Preliminary design for NATM tunnel support
perform extensive parametric and reliability analyses con- in soil. J. Geotech. Eng. 118 (4), 558–575.
sidering the influence of the parameters on the surrounding Mashimo, H., Ishimura, T., 2003. Evaluation of the load on a shield
ground mass and the tunnel lining. tunnel support in gravel. Tunn. Undergr. Space Technol. 18, 233–241.
Mashimo, H., Ishimura, T., 2005. Numerical modelling of the behavior of
Secondly, different support conditions are also easily shield tunnel support during assembly of a tunnel ring. Earth 60 (80),
implemented using the proposed method. Loads coming 100.
from the surrounding soil mass to the tunnel lining can Möller, S., 2006. Tunnel induced settlements and structural forces in
be modified, which allows consideration of complex load- supports Ph.D. dissertation. Stuttgart University.
ing in terms of the loading direction and the magnitude. Oreste, P.P., 2007. A numerical approach to the hyperstatic reaction
method for the dimensioning of tunnel supports. Tunn. Undergr.
Lastly, all the boundary conditions and loading scenar- Space Technol. 22, 185–205.
ios are explicitly applied in the HRM method. Hence, tun- Oreste, P.P., Dias, D., 2012. Stabilisation of the excavation face in shallow
nel designers could easily control the input parameters tunnels using fibreglass dowels. Rock Mech. Rock Eng. 45 (4), 499–
which are used. 517.
Panet, M., 1995. The calculation of tunnels by the Convergence
Convergence Method (Le calcul des tunnels par la me´thode conver-
Acknowledgements gence confinement). Presses de l’école nationale des Ponts et chaussées,
Paris.
The first author was supported by the China Scholarship Plizzari, G.A., Tiberti, G., 2006. Steel fibers as reinforcement for precast
Council for this research work. The funding is greatly tunnel segments. Tunn. Undergr. Space Technol. 21 (3–4), 438–439.
appreciated. This research was also funded by the Vietnam Yin, L., Yang, W., 2000. Topology optimization for tunnel support in
layered geological structures. Int. J. Numer. Meth. Eng. 47 (12), 1983–
National Foundation for Science and Technology Devel- 1996.
opment (NAFOSTED) under grant number 105.08-
2018.310.

You might also like