IJMGE

Int. J. Min. & Geo-Eng.

Vol.48, No.2, December 2014, pp.147-157.

Numerical Modelling of the Segmental Lining of Underground

Structures
1*
Akbar Salemi , Farhang Sereshki 1 , Morteza Esmaeili 2

1
Department of Mining, Petroleum and Geophysics, Shahrood University, Shahrood, Iran
2
School of Railway Engineering, Iran University of Science and Technology, Tehran, Iran

Received 19 August 2014; Received in revised form 23 Sep 2014; Accepted 15 Dec 2014
*Corresponding author: akbar.salemi@gmail.com

Abstract
There are several methods for analysing the behaviour of underground structures under different
loading conditions. Most of these methods have many simplifications; therefore, in some cases, the
results are too conservative and a very high safety factor, usually of more than 2 is needed. On the
other hand, for stability analysis and the designing of support systems, these methods consider
segmental lining and its joints as a uniform lining or a lining with pin connections. In this study,
numerical modelling of the segmental lining of a tunnel was analysed using a sensitivity analysis of
the static modelling. The numerical results were obtained by using a finite difference method
(FLAC2D). Using this form of analysis, a new simple methodology was introduced so that more
reliable results can be obtained. By comparing the frame analysis results obtained by the SAP2000
software with those obtained by the proposed method, it was concluded that the suggested method can
be used as a simple and reasonable approach for the segmental lining of underground structures such
as tunnels.

Keywords : frame analysis, underground structures, numerical modelling, segmental tunnel lining,
sensitivity analysis.

1. Introduction
Tunnelling projects are often deemed as some construction and eventually reduce the costs of
of the most challenging and costly tunnelling, more attention has been focused on
underground projects. This has prompted the design of this type of lining.
many to research ways for improving the The first concrete segmental support on
constructability and stability of tunnels and to record appears to date back to 1903 and was
reduce the high costs involved in such developed by the British contractor McAlpine
underground structures. In recent years, by in Glasgow, Scotland. Since the use of pre-
using segmental lining to improve the speed of cast concrete segmental lining for the eastward

147
Salemi et al./ Int. J. Min. & Geo-Eng., Vol.48, No.2, December 2014

extension of the London Transport Central considers the segment as a curved or straight
Line in 1937, a large number of different beam. Furthermore, the longitudinal and
concrete linings have been developed [1]. circumferential joints are assumed as a
The effect of joints between segmental rotational and shear spring, respectively [6].
linings is one of the most influential factors For the usual cases where loads and the
affecting the stresses induced in tunnel lining, structure do not change in the longitudinal
which is often ignored in tunnel design. If the direction, the three-dimensional behaviour of
joints in the lining are not considered, a the segments has no significant influence on
conservative simplification will result in the the system. This means that for these types of
overestimation of the bending moments, load configurations, two-dimensional analyses
whilst the normal forces will remain are sufficient. For special cases such as
unaffected by this assumption [2]. openings in the lining, various loads on the
The most important problem of segmental rings (e.g., swelling in partial areas), varying
lining design theory is whether the design bending conditions for the rings (e.g., if the
model can reflect the actual stresses of the grouting of the tail gap had not been done
segments. Commonly used methods for design properly at one ring), etc., internal forces and
calculations are classified by joint evaluations deformations of the lining can be only
as follows: i) the usual calculation; ii) the predicted by 3D analysis [6]. In 3D
modified usual calculation; iii) the ring with calculations, the stiffness coefficients of the
multiple hinged joints calculation; iv) the longitudinal and circumferential joints were
beam-spring model calculation. determined by laboratory tests using the actual
i) The usual calculation method assumes joints or theoretical calculations. The lateral
that the segmental ring is a ring with uniform earth pressure coefficient and the coefficient
bending rigidity and ignores the decrease of of subgrade reaction were determined from
rigidity at segment joints. In this method, previous case studies using similar ground
neither the segment (longitudinal) nor the ring conditions [7].
(circumferential) joints are considered [3]. In this study, the induced bending moments
ii) In the modified usual calculation were calculated from simple static analyses
method, a coefficient of the effective ratio of using FLAC2D , on account of the progressive
bending rigidity (   1 ) for evaluating the reduction of some parts of the lining
rigidity of joints, a bending rigidity as EI , a corresponding to the segments' connections.
Finally, by conducting the sensitivity analysis,
transfer ratio of bending moment  , the a model was chosen so that the bending
moment of the main section as (1   ) M and moment of lining in all joint locations were
the moments of the joints as (1   )M are close to zero. As a case study, the results of
calculated. Then, the value of  and  are the segmental lining modelling of the Tabriz
primarily determined by experiences that Urban Railway (TUR) were compared with
consider the joint performance test results and the frame analysis results obtained by the
records of other projects, which are random SAP2000 software. This comparison showed
that the suggested method can be used as a
and uncertain [4].
iii) The ring with multiple hinged joints simple and rational method for the lining of
calculation method is utilized under good segmented tunnels.
ground conditions. The longitudinal joints are
treated as hinges and their influence is 2. Analytical methods for designing of
exaggerated; however, the influence of segmental lining with equal segments
From a mechanical point of view, the ground
circumferential joints is not considered [5].
pressure on the longitudinal joints can be
iv) The beam-spring model calculation
assumed as the compressive axial forces and
method assumes the segmental ring as a ring
bending moment (REF). The most important
with rotational and shear springs. In this
point is that the amounts of bending moment
method, the reduction of bending rigidity and
in the vicinity of joints will be reduced due to
the splice effects of staggered arrangement are
the reduction of the moment of inertia. Thus,
evaluated by using a special model that
148
 Salemi et al./ Int. J. Min. & Geo-Eng., Vol.48, No.2, December 2014

as shown in Figure 1, these regions can be (several times of 10-3 ); thus, the joints do not
modelled in the form of the reduced thickness change the axial stiffness modulus of the lining.
of lining corresponding to the contact area of Under these conditions, the bending stiffness
the lining segments [3]. modulus can be calculated by Equation 1 and the
corresponding moment of inertia of the lining
can be defined as Equation [2].
2
4 e
3
I  Ij   (2)
 n  12
where Ij is the moment of inertia in joints and
n is the number of existing joints in a ring.
This formula was accepted for n>4. It is also
assumed that the joints of ring are not opened
in the inner or outer face of the contacts. Thus,
if the moment of the inertia of joints is
3
Fig. 1. A typical model of a segmental ring with considered as I  0.57 e , then the equal
12
uniform and regular segments (modified from [3])
moment of inertia can be expressed as:
In design practice, there are many known  4  e
2 3

structural systems for calculating internal I   3     (3)

force within the tunnel lining. In the simplest   n   12
of these, a rigid bedded ring is used. This where, for  = 0.5 and n= 6, it can be shown
model does not take the behaviour of the joints 3
into account. For an uncoupled system of that I  0.57 e . This moment of inertia can be
12
hinged rings, the estimated bending moments 3
are too high and should give conservative defined as I  e for a cylindrical shell.
12
results. Therefore, the bending stiffness of the
Muir Wood [3] developed a very easy to segmental lining must be lower than that of
use empirical formula for the estimation of the the continual lining with the same properties;
effects of the longitudinal joints of uncoupled also, the bending moment will be decreased at
rings in a calculation with a homogenous rigid the joints [3].
ring by reducing the bending stiffness of the On the other hand, for four or fewer
lining. The maximum bending moments segments, Muir Wood (1975) suggested that
calculated with this approach are close to the existence of joints will not affect the
those of uncoupled hinged rings. In the rigidity of the lining. However, for a lining
relations proposed by Muir Wood, the with multiple segments, the stiffness at the
theoretical axial stiffness modulus for a joint may be appreciably less than that of the
cylindrical shell is defined by the following lining; thus, reducing the bending moment in
equation [3]. the lining is an approach for considering the
 joints.
E E (1) In the equations suggested by Muir Wood
 (1   )   m [3], it is assumed that the joints at all rings are
where, according to Figure 1, e is the thickness regular (i.e., the longitudinal joints are in a
of the ring, E is equilibrium modulus, is line), whereas this is not truly the case and a
reduced thickness corresponding to the joint on the nth ring increases the bending
longitudinal joints,  is the central angle related moment on the same location of the joints, i.e.,
to each segment, ß is the average circular width on the adjacent rings ((n-1) and (n+1) rings).
of each longitudinal joint, R is the radius of the Thus, non-consecutive joints decrease the
ring and Em is elasticity modulus of ring lining deformation. Generally, the bending
material. stiffness of the lining is exceeded at levels
The dimensionless coefficient ß is very small high above its actual value and the Muir Wood

149
Salemi et al./ Int. J. Min. & Geo-Eng., Vol.48, No.2, December 2014

formula cannot predict this. However, this lining with an emphasis on the effects of the
approach is nonetheless useful for gaining a number and orientation of longitudinal joints.
first idea of the forces in the lining;  and ß Additionally, they proposed a simple design
coefficients can be determined by experiences methodology for determining the stresses
of past similar projects or laboratory models. induced in jointed linings, which can be
However, Muir Wood did not take into carried out without incorporating the joints in
consideration certain structural effects such as the analysis. In Hefny et al.'s model, an
the joint arrangement. Lee and Ge [8] did, equivalent tunnel is defined as an unjointed
however, consider these structural effects. tunnel that has a lining thickness with the
They found that the joint stiffness ratio and the same maximum stresses as those found in a
radius of the tunnel had a large influence on jointed tunnel lining. The thickness of
the effective rigidity ratio, whereby either a unjointed tunnel lining is varied to achieve the
reduction in the joint stiffness ratio or the maximum stresses were induced in the jointed
radius of tunnel will result in a substantial tunnel lining. These maximum stresses are
drop of effective rigidity ratio. This result matched with the maximum stresses obtained
implies that the bending moment of the lining in the both conditions of critical and most
also decreases. If the number of joints were to favourable orientation of the joints. [9].
increase, the effective rigidity ratio and the However, this design method requires more
bending moment of the lining will decrease iteration to compare the jointed and unjointed
substantially, due to the increasing flexibility linings.
of the lining. Other factors such as joint
distribution do not affect the bending moment 3. Geological and geotechnical characteristics
in the lining very much. Though Lee and Ge of the Tabriz Urban Railway (TUR) site
took into consideration some of the effects The case study of this research is the Tabriz
ignored by Muir Wood, their design method Urban Railway (TUR) twin tunnels, which is
often requires rather long iterations in addition located in Tabriz city, in the centre of East-
to approximations in the assumed loadings. Azerbaijan province (Fig. 2). This province is
Therefore, in continuation of these studies, located in the northwest of Iran, i.e., N 38°, E
Hefny et al. [9] investigated the factors 46° [10].
affecting the stresses induced in the tunnel

Fig. 2. The location of Tabriz city [11]

150
 Salemi et al./ Int. J. Min. & Geo-Eng., Vol.48, No.2, December 2014

Line 1 of the TUR tunnels for a total length The groundwater level in the tunnel section
of 7.2 km was excavated by two TBM-EPB under consideration in this study was at a 21.5
machines. The depth of overburden in the m depth, i.e., about 7 m below the tunnel
cross section of the tunnels was about 7 m and bottom [14]. Therefore, in the all models and
consists of three layers: gravelly sand, analyses of this research, the groundwater and
silty/clayey sand and clay (respectively, its effects on the tunnels were neglected. The
starting from the surface). The average values geometry of the tunnels' cross-section is
of the geotechnical parameters for these three described in Figure 3.
soil layers are listed in Table 1 [12].

Table 1. Geotechnical parameters for three layers of the studied section [13]

Thickness of layer  C  E
Soil Class.
(m) (Kn/m3 ) (kPa) (deg.) (MPa)

6 Gravelly Sand 20.0 8 23 23

24 Silty/Clayey Sand 18.0 5 35 23
10 Clay 15.0 34 19 23

Fig. 3. Geometry of the tunnels

4. Numerical modelling of segmental lining 4. model solving and its establishment

with non-uniform segments prior to starting tunnelling;
According to numerical modelling principles, 5. beginning tunnels' excavation and
the modelling steps taken by FLAC2D include exerting the appropriate stress relaxation;
[15]: 6. installing the tunnel lining, so that as
1. the boundary limitation around the shown in Figure 4, the lining may consist of
tunnels by assuming the plane strain condition 5+1 segments (1 corresponds to the key
and selecting the model dimensions as 40×150 segment) with a 30 cm thickness [16].
m2 ; The original idea of using the numerical
2. defining the constitutive model and its modelling of segmental tunnel lining was that
properties, where the Mohr-Coulomb material the induced bending moment in a jointed
model and elastic beam element have been lining is reduced in the joints location. Thus,
used to consider the plastic behaviour of the segments could be simulated by a reduced
alluvial ground material and the behaviour of thickness of the lining at these locations. This
tunnel lining, respectively; is clearly shown in Figure 4 below.
3. drawing of the tunnels' geometry;

151
Salemi et al./ Int. J. Min. & Geo-Eng., Vol.48, No.2, December 2014

Fig. 4. The original idea of using segmental lining modelling with non-uniform segments

In the numerical modelling of TUR tunnels frame (as the element number applied in FLAC)
lining, the lining consisted of six non-uniform was used as concrete beams. To appropriately
segments (5+1), as shown in Figure 4. Two consider the surrounding ground effects on the
series models with 15 and 10 cm of mesh tunnel lining in the SAP2000 model (according
dimensions were created by FLAC2D . The to Fig. 7), the engineering and design guidance
reason for making a fine mesh model (10 cm of tunnels and shafts (EM 1110-2-2901) were
meshing) was to control the deflection effect, used, the tangential and radial springs were
i.e., the reduction of beam elements bending applied and the soil-structure interactions were
moment in the joints. restricted. The stiffness of radial and tangential
Finally, one may observe that deflection springs was obtained from Equations 4 and 5,
had no effect on the bending moment respectively [17].
reduction of these beam elements. In each
Em b
series of modelling, after finding the joint kr  (4)
locations on the beam elements, the thickness (1   m )
of these elements was reduced by 1/2, 1/4, 1/8, G kr
1/16, 1/32 and 1/64 times of the original kt  k r m  (5)
thickness. The effect of this process on the Em 2(1  r )
induced bending moment in the lining could where, Er and Gr are the elasticity and shear
then be estimated, i.e., the best state may modulus, respectively, Vm is the Poisson ratio
happen when the bending moment of the of the surrounding ground materials and b is
lining in all joint locations were close to zero. the tunnel length in the longitudinal direction
The charts in Figures 5 and 6 show the or ring width. In this study, b was taken as 1 m
induced bending moments in the mentioned
for all models. The parameter  is an arc
status for models with 15 and 10 cm mesh angle of the beam element in the radian.
dimensions, respectively. As shown in these After exerting the loads due to
charts, a reduced thickness of about 1/8 to
overburdening and ground convergence on the
1/16 had the best situation in the non-uniform
tunnel lining, the under tension springs were
segmental lining simulation for models with omitted. Finally, the induced bending moment
15 cm meshing. In the models with 10 cm on the segmental lining due to the in situ
meshing, this reduced thickness was 1/16. stresses of the ground is shown in Figure 8.
However, in this series of models, the induced Comparing Figure 6 (concerned with segment
bending moments, instead of six points (joint t/16 in the graph) and Figure 8 provided a
locations) were close to zero in an additional
good resemblance between the bending
point. To remove this difficulty, the same
moment results obtained from the SAP2000
problem was modelled by SAP2000 structural and FLAC2D models.
analysis software.
In the modelling of tunnel lining, a 24-piece
152
 Fig. 5. The induced bending moment in different elements of segmental lining in models with 15 cm meshing

153
Salemi et al./ Int. J. Min. & Geo-Eng., Vol.48, No.2, December 2014

Fig. 6. The induced bending moment in different elements of segmental lining in models with 10 cm meshing
Salemi et al./ Int. J. Min. & Geo-Eng., Vol.48, No.2, December 2014

Fig. 7. The beam-spring model applied in the SAP2000 model containing soil-structure interaction (taken from [17])

5. Comparing the behaviour of continual

and segmental linings
In this part of the paper, the behaviour of both
continual and segmental linings duos to the
distribution of induced axial and shear forces
and bending moments are compared, using the
same model with 10 cm meshing. The results
of the static analysis of these two models are
shown in Figures 9 to 16 below.
Comparing the above results shows some
differences between these two modelling
results; therefore, the effect of joints must be
considered in the analyses. To observe the
further differences in the results of these two
modelling schemes, the graphs shown in
Fig. 8. The induced bending moment on the segmental
lining model in S AP20005. Comparing the behaviour
Figures 17 and 18 must be considered, as they
of continual and segmental linings illustrate the different induced bending
moments in the two models.

Fig. 9. The induced axial force on the continual Fig. 10. The induced axial force on the segmental
lining lining

154
 Salemi et al./ Int. J. Min. & Geo-Eng., Vol.48, No.2, December 2014

Fig. 11. The induced bending moment on the Fig. 12. The induced bending moment on the
continual lining segmental lining

Fig. 13. The induced shear force on the continual Fig. 14. The induced shear force on the segmental
lining lining

Fig. 15. The induced axial strain on the continual Fig. 16. The induced axial strain on the segmental
lining lining

segmental lining lining

2.00E+06
Axial Force (N)

1.50E+06

1.00E+06

5.00E+05

0.00E+00
1

2
10

19

28

37

46

55

64

73

82

91

10

10

11

12

13

14

15

16

17

18

19

19

20

21

22

23

24

25

26

Elements

Fig. 17. The induced axial force variation graphs in the beam elements of continual and segmental linings

155
Salemi et al./ Int. J. Min. & Geo-Eng., Vol.48, No.2, December 2014

segmental lining lining

3.00E+05

2.00E+05
Moment (N.m)

1.00E+05

0.00E+00

-1.00E+05

-2.00E+05
1

2
10

19

28

37

46

55

64

73

82

91
10

10

11

12

13

14

15

16

17

18

19

19

20

21

22

23

24

25

26
Elements

Fig. 18. The induced bending moment variation graphs in the beam elements of continual and segmental linings

According to these graphs (Figures 17 and tunnels with various cross sections, its ability
18), the axial force values in different parts of to model the different segmental
the linings did not change more, but the arrangements, the lack of need for expensive
corresponding bending moment values laboratory tests or other experimental data, the
changed considerably. capability of using different loading, as well as
other facilities of modelling. For example, by
6. Conclusion comparing the results of continual and
The results obtained in this paper show that segmental linings, it can clearly be shown that
the structural analyses of segmental lining the bending moment values in different parts
require the consideration of joints' behaviour. of the tunnel lining have changed
Some of the suggested methods and models considerably, while the axial force values have
for considering joint behaviour ignore some of not changed. Finally, it must be noted that all
the joint effects, while others require of the models in this research were applied to
laboratory results or experiments gained from TUR tunnels site conditions; in other cases, a
past similar projects for their completion. wider scope may need to be applied to
Therefore, they may not provide trustworthy research such as ignoring the circumferential
results with sufficient accuracy. In addition, in joint effects and considering the effects of
nearly all of the previous analytical methods, consecutive rings on each other, disregarding
the segments have been assumed in a similar the special nature of segment junctions, as
way. However, 3D modelling may be the best well as other modelling simplifications, some
method for estimating the longitudinal and of which may considerably affect modelling
circumferential joints' effect on segmental results.
lining. This method is very complex, however,
as well as time consuming and requires some References
real properties such as the stiffness of joints,
which require expensive laboratory tests or [1] Whittaker, B. N., Frith, B. C. (1990). Tunneling −
results gained from ambiguous experiments design, stability and construction, The Institution
conducted for past similar projects. Therefore, of Mining and Metallurgy, London, Ch. 14, pp.
in this paper, two dimensional numerical 331-350.
modelling of the segmental lining with non- [2] Kasper, T., Edvardsen, C., Wittneben, G.,
uniform segments was proposed by employing Neumenn, D. (2008). Lining design for the district
a sensitivity analysis of the induced bending heading tunnel in Copenhagen with steel fiber
moments in the tunnel lining. Comparing the reinforced concrete segments , Journal of
FLAC modelling results with the structural Tunneling and Underground Space Technology,
analysis, accomplished by using SAP2000 Elsevier, Vol. 23, pp. 574-587.
software, clearly confirmed the validity and [3] Muir Wood, A.M. (1972). The circular tunnel in
accuracy of the proposed method. The primary elastic ground, Proc. Int. Symp. Underground
advantages of this method may be its Openings, Lucerne, pp. 356- 369.
simplicity, good efficiency and flexibility for

156
 Salemi et al./ Int. J. Min. & Geo-Eng., Vol.48, No.2, December 2014

[4] I.T.A. Working Group No. 2 (2000). Guidelines [10] NGO, (2002). The gazetteer of township in the
for design of shield tunnel lining, International I.R. of Iran-Tabriz township, National
Tunneling Association, Journal of Tunneling and Geographical Organization Inc. Tehran, Iran,
Underground Space Technology, Elsevier, Vol. 209 p.
15, Issue 3, pp. 303-331.
[11] World statesmen (2008),
[5] Huang, Z. R., Zhu, W., Liang, J. H., Lin, J., Jia, R. http://www.worldstatesmen.org/Iran.html.
(2006). Three-dimensional numerical modeling of
[12] TURO, (2002). Geological report of Tabriz
shield tunnel lining, Journal of Tunneling and
urban railway line 1, Tabriz Urban Railway
Underground Space Technology, Elsevier, Vol.
Organization, Tabriz, Iran.
21, pp. 434.
[13] Arthe Civil & Structure B.V. (2005).
[6] Klappers, C., Grübl, F., Ostermeier, B. (2006).
Determination of final geotechnical and geo
Structural analyses of segmental lining − coupled
(hydro) logical parameters for calculations, Report
beam and spring analyses versus 3D-FEM
No.: ACS05203-R06-S03, 26 p.
calculations with shell elements, Journal of
Tunneling and Underground Space Technology, [14] TURO, (2005). Final results of geological and
Elsevier, Vol. 21, Issues 3-4, pp. 254-255. geotechnical investigation on Tabriz urban
railway line 1, Report for the Tabriz Urban
[7] Mashimo, H., Ishimura, T. (2003). Evaluation of
Railway Organization, Tabriz, Iran.
the load on shield tunnel lining in gravel, Journal
of Tunneling and Underground Space [15] Itasca. (1999). FLAC, Fast Lagrangian Analysis
Technology, Elsevier, Vol. 18, pp. 233-241. of Continua, version 4.0.307, User’s manual,
Itasca Consulting Group, Inc. Minneapolis, MN,
[8] Lee, K. M., Ge, X. W. (2001). The equivalence of
http://itascacg.com.
a jointed shield-driven tunnel lining to a
continuous ring structure, Canadian Geotechnical [16] Arthe Civil & Structure B.V. (2006). Design
Journal, Vol. 38 (3), pp. 461-483. and calculation of concrete segments , Report
No.: ACS05203-R09-S04 & ACS05203-R09-
[9] Hefny, A. M., Tan, F. C., Macalevey, N. F.
S05, 57 p.
(2004). Numerical study on the behavior of
jointed tunnel lining, Journal of The Institution of [17] U.S. Army corps of engineers (1997).
Eng., Singapore, Vol. 44 Issue 1 2004, pp. 108- Engineering and design of tunnels and shafts in
118. rock, EM 1110-2-2901, Dep. of the U. S. Army,
Washington, DC 20314-1000, 206 p.

157