Geosciences Journal

Vol. 20, No. 4, p. 517  528, August 2016

DOI 10.1007/s12303-015-0050-1
ⓒ The Association of Korean Geoscience Societies and Springer 2016

Semi-analytical solution for time-dependent deformations in swelling rocks

around circular tunnels

}
Davoud Parsapour Department of Civil and Environmental Engineering, Amirkabir University of Technology, Hafez Ave.,
Ahmad Fahimifar* Tehran, Iran

ABSTRACT: Underground structures in swelling rocks bear time- processes are fundamentally different. In both cases, how-
dependent swelling effects. Upon excavation of underground spaces, ever, swelling depends not only on material characteristics
the state of stresses and displacements change compared with the
but also on the availability of water (Schadlich et al., 2012).
initial state. The stress and displacement variations depend on tunnel
advance and the rheological behavior of surrounding rock mass. The swelling rocks make large heaves in tunnels which are
The swelling behaviour of rocks is known as a complicated phe- often observed in situ. The relationship between the swelling
nomenon and very important task in tunnel design process. The deformation and humidity diffusion are discussed at meso-
paper aim is to predict the time-dependent displacement of the tunnel scopic level. The invert deformations are larger than that of
after the excavation has stopped or after installing the support system. sidewalls interpreted here as a consequence of different
The assumption is made that the medium around tunnel is both humid boundary conditions (Tang and Tang, 2012).
isotropic and homogenous. The section of tunnel is also assumed
circular which in view of stress field, is excavated in Hydrostatic Several time independent constitutive models have been
and non-hydrostatic conditions. A semi-analytical model on the basis developed for the swelling of rocks (e.g., Grob, 1972; Ein-
of experimental results is adopted to evaluate the ground swelling stein et al., 1972; Wittke and Pierau, 1979). Einstein (1994)
strains as a function of time and stress. Given the current assump- examined the available swelling models in the literature and
tions and conditions, a semi-analytical solution is derived to pre- concluded that, although the swelling deformation due to
dict the time-dependent displacements and lining pressure for a chemical changes may be adequately predicted using available
circular shaped tunnel section in a swelling ground. Eventually,
the model is loaded by an axisymmetric far-field pressure and the models, but time-dependent swelling models need to be devel-
effect of the tunnel face on the lining pressure is also considered. oped from a fundamental knowledge of the swelling mech-
On the whole, the comparison between modeling results and actual anism. Lo-Yuen (1981) and Lo-Hefny (1996) developed rheological
data, proved an accordance between them. As the results indicate, models for prediction of swelling properties of rocks sur-
such parameters as, time dependent modulus of swelling and swell- rounding tunnels using closed-form solutions. Field monitoring
ing strain coefficient as well as the initial aspect ratio, play a prom- of tunnels proved that these models simulate the trend of
inent role in controlling the swelling deformations. The set up time
of lining is also considered as an impressing factor controlling the swelling behavior well and presently, they are being used in
swelling pressure. In mediums of high swelling potential, the tem- the design of underground openings. However, Lo and Yuen
porary support system installation has to be made to allow some did not consider the effect of stress on the swelling in their
ground deformations prior to the permanent support (lining) gets model, and Lo-Hefny assumed that only the radial pressure
installed which causes the swelling deformation to get relieved. on the tunnel lining has an effect on the swelling deformations.
The omission of any stress effect makes the prediction con-
Key words: swelling, semi-analytical solution, circular tunnel, time-
dependent, displacement servative. It is also necessary to predict the long-term swell-
ing deformation that is observed in the field. Hawlader et al.
1. INTRODUCTION (2003) developed a constitutive model for the swelling rock
that considers three-dimensional stress effects and anisotro-
When water infiltrates the swelling rock mass, chemical pic swelling. The anisotropic and three-dimensional stress-
processes are initiated within the rock matrix which can result dependent swelling model cannot be presented in a closed-
in large volume increase, and typically large heave defor- form solution, and a numerical method must be implemented.
mations occur in the tunnel invert, with no or a rather flexible Hawlader et al. (2005) used a finite element algorithm,
invert lining. In the case of a rigid support, large swelling incorporating the new constitutive model for the numerical
pressure develops in the tunnel lining. The most prominent analysis.
rock types with swelling behavior are clay stone and anhy- Safe design of civil engineering structures such as tunnels
drite-bearing rocks. Even though their macroscopic swell- in swelling rocks requires a reliable determination of the lateral
ing behavior may appear similar, but the underlying chemical swelling pressure. Swelling of a tunnel is a complex problem
which arises from a wide range of variability. So its effects
*Corresponding author: [email protected] on tunneling design and analysis have always been a challenge.
518 Davoud Parsapour and Ahmad Fahimifar

Finite element-based analyses and other numerical strate- strain decreases with the logarithm of stress, as follows:
gies are also routinely employed to assess the degree of safety

of swelling problems. The numerical methods, encompasses a  zs = –kq Log  -----z , (1)
variety of tunnel geometry and rock properties such as lat-
0

eral earth pressure, but pose trouble in modeling and vali- where  zs is the maximum value of normal swelling strain,
dation of results. However, the results can vary significantly kq is dimensionless coefficient of the final normal swelling
and their applicability relies strongly on the judgments and strain, and z is the principal normal stress corresponding
expertise of the designers. Yet, there has been no simple ana- to the direction of swelling strain, and 0 is the least prin-
lytical solution to estimate the swelling deformations and cipal normal stress at which swelling doesn’t occur (the
lining pressure induced by swelling in tunnels. The present upper limit of stress). Wittke-Gattemann (1998) expanded
work moves from a numerical approach recently proposed Grob’s mathematical model as in Equation (2).
by different authors and, with the help of classical tools of
calculus variations, introduces an exact solution to the prob- 0 z  0
lem of swelling displacements for tunnels with circular cross 
 z
sections.  Log  ----- c  z  0
Generally, the characterization of swelling rocks is car-

 zs = –kq  0 . (2)
ried out using the classic odometer, which offers the mea- 
 c
 Log  ----
surement of the axial swelling deformation and the swell  -   C
 0 z
pressure under a condition of zero lateral deformation. Based 
on the results of Huder-Amberg tests (1970), Grob (1972)
proposed a mathematical model to relate the final swelling In which, c is the least principal normal stress at which
strain to stress. Although Wittke-Gattemann (1998) expanded the reduction of swelling strain occurs (lower limit of the
Grob’s mathematical model, but they are computing the swelling stress).
final swelling strains at equilibrium (t = ∞) only. To obtain For the purpose of obtaining a proper time-dependent ana-
a proper time-dependent analytical solution on the basis of lytical solution, considering real behavior of swelling rocks,
real behavior of swelling rocks, Fahimifar and Parsapour a mathematical relation was derived between the progressive
(2010) derived a closed-form solution between the progres- swelling strains and the time considering the experimental
sive swelling strains and the time based on the experimental results from the Huder-Amberg tests as in Equation (3)
results from the Huder-Amberg tests (1970). This model is (Fahimifar and Parsapour, 2010).
capable of predicting the swelling strain of tunnel walls at
a given time with an acceptable approximation. The existing 0 z  0

experimental results prove that this solution overestimates  z
 Log  ----- c   z  0
t
– ------
the time-dependent swelling effects. The three-dimensional  q
 
ZS  t  = –kq   1 – e    0 . (3)
stress effects on the swelling have not been considered in the 
closed-form solutions, and thus, this is a cause of discrepancy,  c
 Log  ----
In which:  -   C
and therefore, the swelling displacements of the tunnel walls  0 z
cannot be predicted directly. This paper describes the develop- 
ment and implementation of a closed-form approach to pre- ZS  t  : Principal swelling strain at time (t) due to stress relief
dict both the time-dependent deformation and lining pressure of t: Time (hours)
swelling rocks in circular tunnels, incorporating three dimen- kq: Coefficient of the swelling strain (dimensionless)
sional stresses (considering face effect). A proper estimation z: Total Principal normal stress in swelling direction
of swelling in tunnels could be adopted implementing the 0: The least total principal normal stress at which swell-
least geotechnical data and data relevant to tunnel geometry. ing strain will cease (the upper limit of stress)
The swelling deformation is calculated based on the time- c: The least total principal normal stress at which the
dependent swelling strain model developed by Fahimifar swelling strain begins to decrease (the lower limit of stress)
and Parsapour (2010). Finally, two case studies concerning q: Time dependent modulus of swelling (hours)
circular tunnels, on account of their occurrence and impor- In Equation (3) parameters kq, 0, c and hq are obtained
tance in civil engineering, are analyzed and discussed. from the odometer swelling tests, and the stress, z, can be
calculated by stress analysis around tunnel using analytical
2. MODEL DESCRIPTION or numerical approaches. The swelling displacement of the
tunnel wall can be predicted using Equation (3). It describes
Based on the results of Huder-Amberg tests (1970), Grob the development of time-dependent displacement due to swell-
(1972) proposed a mathematical model for relating the final ing considering viscoelastic behavior. Computation of stress
swelling strain to stress. According to this model, swelling distribution around tunnel due to excavation (introduced by
 Semi-analytical solution for time-dependent deformations in swelling rocks 519

z in Eq. 3) is based on elastic medium formulation.

The current model considers the tunnel excavation as cir-
cular in an elastic time-dependent swelling material. The
following assumptions are made in this regard:
(1) The tunnel cross section is circular
(2) The surrounding ground is homogeneous, isotropic,
and behaves linearly elastic.
(3) The gravity effect on the roof of the tunnel is neglected.
(4) The excavation (i.e., the removal of material) is assumed
to take place instantaneously.
Regarding the above assumptions, the model can be con-
sidered as axisymmetric, but two cases are defined depend-
ing on the selected section: Fig. 2. (a) Geometry; (b) Stress components at point A after exca-
The sections near the tunnel face selected within dis- vation in polar coordinate (Farzi, 2012).
tances up to four times of the tunnel radius (Panet, 1995).
4 2
The sections far from the tunnel face selected at distances 1–k a a
r  =  –  ----------   1 – 3  ----4 + 2  ----2   sin2  , (4)
more than four times of the tunnel face (Panet, 1995).  2  r r 
Figure 1 presents the tunnel face and a section far from
where r,  and r are stress components around the tun-
the tunnel face.
nel, r and  represent the radial and tangential positions of
For far face sections (Case 2) stress distribution around a
a rock element from the tunnel center, respectively, s is the
circular tunnel can be presented by Kirsch solution (1898).
initial magnitude of stress, a is the tunnel radius, k is the initial
For case 1 (near face sections) an analytical solution for stress
stress ratio and l is a face position parameter. Panet and Guenot
analysis around circular tunnels considering the face effect
(1982) proposed an empirical relation for considering the
was presented by Farzi (2012), which it’s relevant figure (Fig.
effect of tunnel face advancement:
2) and equation, as follows:
2
X
1+k a
2
1–k
4
a
2
a   x  = 0.28 + 0.72 1 –  ----------- , (5)
 r = -----------  1 – ----2  + ----------  1 + 3  ----4 – 4  ----2   cos2  , X+x
2 r 2 r r
2 4
where x is distance of the considering section from tunnel
1+k a 1–k a
  = -----------  1 +  ----2  – ----------  1 + 3  ----4   cos2  , face, X = 0.84a, and a is tunnel radius. The l parameter varies
2 r 2 r between 0.28 and 1 ( = 1 belongs to the far-sections from

Fig. 1. Sections near and far from the tunnel face.

520 Davoud Parsapour and Ahmad Fahimifar

the tunnel face). In order to consider the face effect, other t

– -------
approximations have been proposed (e.g., Panet, 1995; Car-  i  t  = · i  t  dt =  1 – e
qi
 ×  c ,
kqi × Log  ----
 -
  


ranza-Torres, 2000). swelling 0












coefficient
time dependent term
In the case of hydrostatic stress field, if the stress distri- stress term
i = r, . (10)
bution is symmetrical with respect to the axis of tunnel which
is specified perpendicular to r   plane, the stress compo-
Integrating the radial strain (Eq. 10), the radial displacement is
nents do not depend on the angleq and are merely functions
computed as:
of radius r (Gnirkand and Johnson, 1964). This implies that
 
the stresses are principal stresses and there is no shear stress Ur  t  =  r dr =   r  t  – r0 . dr . (11)
(r = 0). The tangential and radial stresses in hydrostatic r r

stress field (k = 1) for the near-sections, by assuming the face Applying Equation (10) into Equation (11), one gets:
as providing a fictitious internal support pressure, is equal to:
t
– --------
Ur  t  =  1 – e 
qr
 × kqi
2
 
 3 =  r =   1 –  a----2 


swelling
 r






 time dependent term
coefficient


  1 +  a---- . (6)
2
 =  =    c –
× a Log  ---- r0 dr .
 1 
 r
2 - Log  ------
- (12)
  0  0 
 = 0 










 r stress term

The analysis of stresses and displacements near the tunnel Therefore, by replacing r and r0 by Equations (4) and (7)
face is a three-dimensional problem. In the practice of tunnel respectively, the swelling displacement of the tunnel wall is
support design, 3D problem has been approached by regarding calculated as:
t
a section behind the face as a 2D plane-strain problem (Sulem – --------
 1 + k  a
2
Ur  t  =  1 – e   k   Log  ----
- ----------- 1 –  ----2 
qr
et al., 1987). The perturbations in the stress field caused by the   qr r 0  2  r
presence of tunnel reach in situ stress level for distances far
from the tunnel center. Hence, the “far-field” stresses that is 1–k a a
4 2

the definition of “stresses at infinity,” is expressed as follows: + ----------  1 + 3  ----4 – 4  ----2   cos2  
2  r r 
 lim  =  =   1---------- +k 1–k
- + ----------  cos2   1+k 1–k
 r r  2   Log  -----  ----------- + ----------  cos2   dr .
r0
2 (13)
  0  2 2 
 1+k 1–k
 limr   =  =  0 =   ----------- – ----------  cos2   r>>a .
  2 2  (7) For sections far from the tunnel face ( = 1) and in hydro-
 static stress field (k = 1), Equation (13) becomes as:
 limr   r = r 0 = – 1---------
–k
-  sin2 
 2
t
– --------
  a   2
Ur  t  =  1 – e   k   Log  ----
-  1 – ---- – Log  ----- dr.
qr
  qr r   0  r2    0
3. PREDICTION OF SWELLING DISPLACEMENTS
FOR UNLINED TUNNELS (14)
Solving Equation (14), radial swelling of the tunnel wall
Based on Equation (3), the constitutive equation of swell-
is expressed as follows:
ing strains in polar coordinate is given as: t t
– -------- – --------
Ur  t  = aLog4  kqr  1 – e   0.6a  k  1 – e  .(15)
qr qr
1 
· i  t  = ------   kqi Log  -----i  – i  t   i = r  , (8)   qr
 
qi 0
The derivation of radial swelling displacement (Eq. 15) is
where the dot denotes the derivative with respect to time. provided in Appendix A. It is obvious that Equation (15)
The differential form of swelling strain is expressed as fol- depends on the amount of swelling properties of rock and radius
lows: of the tunnel. In this regard, the swelling convergence of the
tunnel (Cs(t)) shall be calculated as:
  d i  t  i  t   d
Log  -----i  = -----qi-  ---- i  t  + ---------
- = ----------  qi  ---- + 1 . (9)
 0 kqi dt kqi kqi  dt  Cs  t  = 2  Ur  t   , (16)
Integrating Equation (9) with respect to time, the swelling where Ur(t) is the displacement of tunnel wall in a specific
strain is computed as: period of time. It is discernible that the amount of conver-
 Semi-analytical solution for time-dependent deformations in swelling rocks 521

gence has an increasing trend due to the swelling of tunnel wall. Using Equation (22), the time-dependent radial displace-
Thus, considering Equation (15), the swelling convergence ments of a tunnel driven in a non-hydrostatic stress field
of the tunnel wall is obtained as: can be computed.
t
– -------
Cs  t  = 2Ur  t   1.2a  kqr  1 – e .
qi
(17) 4. PREDICTION OF SWELLING DISPLACEMENTS
 
FOR LINED TUNNELS
If the tunnel is assumed to be circular and driven in a
non-hydrostatic stress field (k  1), for the section far Tunneling through swelling ground can mobilize high lin-
enough from the tunnel face (l = 1), Equation (13) becomes ing pressures over the time. For this purpose, the authors
as: developed a solution for evaluation of rock-support inter-
t action in circular tunnels driven in a hydrostatic stress field.
– -------
 1+k a
2
Ur  t  =  1 – e qi  k qr   Log  -----  -----------  1 – ----2 
 
The assumptions are the same as the calculation processes
  a 0  2  r 
made for unlined tunnels. The face effect is also considered.
4 2 In this case, the tunnel is supported in each section even
1–k a a
+ ----------  1 + 3 ----4 – 4 ----2   cos2    those are adjacent to the tunnel face. Assuming that the sup-
2  r r 
port is a closed ring made of elastic material (such as concrete
or shotcrete linings), the stiffness of the support system (Ks)
 1+k 1–k
 Log  -----  ----------- + ----------  cos2   dr . (18) is assumed to be constant. In all cases, the support is assumed
 0  2 2 
to act over the entire surface of the tunnel. In other words,
the shotcrete and concrete linings are closed rings; the steel
Equation (18) can be written as follows:
sets are complete circles; and the mechanically anchored
t
– ------- rock bolts are installed systematically, in a regular pattern
Ur  t  =  1 – e qi  k qr  a U  r  dr .
 
(19) that completely surrounds the tunnel.
 
The parameter t0 is the elapsed time until the support struc-
In which: ture is installed. Before this period of time, the tunnel was
 1+k a2 1 – k a4 a2 unlined (t  t0). At a time t = t0, the lining is installed. For
Ur  t  = Log  -----  -----------  1 – ----2  + ----------  1 + 3 ----4 – 4 ----2   cos2   (t  t0), the radial displacement of tunnel wall can be cal-
0  2  r  2  r r   
culated from Equation (15). For (t > t0), the ground pressure
 1+k 1–k acting on the lining is related to the tunnel closure. Thus, it
 Log  -----  ----------- + ----------  cos2   . (20) can be written as:
  2
0 2 
Ur  t  Ur  t0 
Equation (19) can’t be analytically integrated; so it is solved Ps = – Ks  -----------
- – -------------- , (23)
by numerical integration. Numerical integration is the approx-  a a 
imate computation of integral using numerical techniques.
where Ks is the stiffness of the support system (see Eq. 25).
The method for numerical integration considered for the cur-
Excavation of the tunnel and installation of support system
rent model is Simpson’s rule on a closed interval [a,10a] (Davis
is assumed to be implemented instantaneously. Thus, no dis-
et al., 1984). More accurate approximation could be made
placement will occur before the support installation ( Ur  t 0  = 0.
by breaking up the interval [a,10a] into some number “n” of
In this case, the stress distribution around the circular tunnels
subintervals, computing an approximation for each subinterval,  a 2

equal to the total sum of the lining pressure (Ps  ----2  and the 2fic-
and then adding up the whole results. The Compound Simp- r
titious lining pressure due to the face effect (  1 –  a----2  ):
son’s rule is then yielded: r
b – a fa b–a fb 
-  --------- + n – 1 f  a + i -----------  + --------- . (21)
b 2 2 2
a a a
a f  r  dr  ----------
n  2  i = 1  n  2   3 =  r =   1 –  ----2  + PS  ----2  =  – ----2   – PS 
 r r r
The upper limit of the integration (Eq. 19) is considered  2 2 2
. (24)
  a  a a
ten times of radius (b = 10a) and the error in approximating  1 =  =   1 +  ----2  –PS  ----2  =  + ----2   – PS 
swelling displacement for upper limit more than 10a is neg-  r r r
ligible. Comparing the coefficients of Equations (19) and
(21), the following equation is derived: Stiffness of the support system can be evaluated from the
t following equation (Brady and Brown, 1985):
– --------
Ur  t    1 – e  k
qr 2 2
  qr Ec  a – b 
KS = ----------------------------------------------------------
2 2
-, (25)
 1 +  c    1 – 2c a + b 
9a U  a  9a U  10a 
 ------  ------------ +  n – 1 U  a + i ------  + ------------------ . (22) where Ec and c are elastic properties of the support, a is the
n  2 i = 1   n   2 
522 Davoud Parsapour and Ahmad Fahimifar

5. APPLICATION OF THE MODEL TO CASE

RECORDS

Two case studies are presented in order to show the accu-

racy and validity of the proposed semi-analytical solution.

5.1. Case a

The semi-analytical procedure presented in this paper was

used to analyze a tunnel in a swelling rock in southern Ontario.
Hawlader et al. (2005) developed a finite element algorithm
incorporating a new constitutive model for rock swelling
considering three-dimensional stress effects and anisotropic
swelling. A numerical procedure was used to analyze the
Fig. 3. The stress distribution around the circular lined tunnel. Heart Lake storm sewer tunnel in southern Ontario and, the
tunnel advance monitored over a long period of time (Lo and
outer and b is the inner radii of the support, respectively (it Yuen, 1981; Lo and Lukajic, 1984). The tunnel with 1.5 km
was assumed that the support is a ring made of elastic material). long and final diameter of 3 m was evaluated considering
The hoop and radial stresses calculated from Equation (24), is three construction methods: (i) cut and cover (232 m), (ii) drill
presented in Figure 3. and blast (183 m), and (iii) machine boring (1050 m). Lo and
Substituting Equation (24) into Equation (3), radial dis- his coworkers (Lo et al., 1979; Lo and Yuen, 1981) reported
placement at the tunnel wall (r = a) is expressed as follows: the site geological conditions and rock properties. The rock
t
– -------- 2 properties, in situ stresses, and the lining properties applied
Ur  t  =  1 – e
qr
  k   Log  1 – a----b dr , in the analysis are listed in Table 1. The numerical analysis
 qr a
(26)
  r2 
mainly applied for the machine-bored section at CH 136 +
where the undetermined parameter b is given by 20 (Lo and Yuen, 1981). Analysis and field measurements
(Lo et al., 1979) proved a slight in situ stress change due to
 – P excavation beyond a radial distance of 5a, where a is the radius
b = -----------------S . (27)
 of excavation. As the depth of the tunnel at CH 136 + 20 is
For computation of the lining pressure, Equation (27) can considerably greater than 5a, so the tunnel is considered deep
be rewritten as: and merely one quarter of the tunnel section needs to be numer-
ically analyzed.
PS =  – b  . (28) Hawlader et al. (2003) modeled the swelling of rock under
three-dimensional stress using the following equation:
Solving Equation (26), it is obtained as:
t
 t
 .  i  t  = mi log  --t-0 ,
s
– --------
 1 – b  (31)
Ur  t  =  1 – e   k  a  b Log ---------------------
qr
– Log  1 – b 
  qr   1 + b  
where  i  t  is the swelling strain in principal direction at
s

(29) time t; m is the swelling potential under three dimensional

i
The derivation of Equation (29) is provided in Appendix stress conditions; and to is reference time which is the time
B. Applying Equation (29) into Equation (23) leads to: from which swelling begins. The physical characteristics of
t the model can be found in Hawlader et al. (2003). Hawlader
– --------
   – b  = –Ks  kqr   1 – e  et al. (2005) analyzed Heart Lake tunnel implementing the
qr
 mentioned model in numerical analysis.
 1 – b  To analysis the Heart tunnel, it was assumed that the tunnel
 b Log --------------------- – Log  1 – b  . (30)
  was driven in a hydrostatic initial stress field ((σx0 + σy0)/2).
 1 + b  The plane strain condition was used in both analyses. Based
Equation (30) is the proposed formula for evaluation of the on the Equation (3), the swelling parameters of the rock sur-
time-dependent lining pressure with considering the face rounding the tunnel (kqr and hq) were obtained from the best-
Ps
- varies between (1)
effect. Based on this equation, the ---- fitted curve of the Hawlader et al. (2003) three-dimensional
and . Solving Equation (30), b' parameter is obtained and swelling model. So the values of kqr = 1.05% and q = 375
then using Equation (28) and Equation (29), the lining pres- hours were used in the analytical solution.
sure and the radial displacement over a course of time can Once the lining is in contact with rock, the time dependent
be computed, respectively. swelling of the surrounding rock generates axial thrust and
 Semi-analytical solution for time-dependent deformations in swelling rocks 523

Table 1. Geometry and parameters used in the analyses

Numerical Analysis, Hawlader et al. (2005) Present work
Geometry
Outer radius of tunnel, a (m) 1.675 1.675
Thickness of concrete lining (m) 0.3 0.3
Shape ofthe tunnel Circular Circular
Initial stresses
Vertical stress, σy0 (MPa) 0.435 2.828 (average of σy0 and σx0)
Horizontal stress, σx0 (MPa) Koσy0 2.828 (average of σy0 and σx0)
Initial stress ratio (Ko = σx0/σy0) 12 1
Rock properties
Elastic
Elastic modulus, Er (MPa) 12 400 12 400
Poisson’s ratio, r 0.15 0.15
Time-dependent swelling
Minimum threshold stress, σc (MPa) 0.001 0.001
Reference time, t0 (days) 10 0
Free swelling potentials (%)
mx(0) 0.10 mavg(0) = 0.2067, using curve fitting,
my(0) 0.42 kqr = 1.05% and q = 375
mz(0) 0.10
Concrete lining
Elastic modulus, Ec (MPa) 28000 28000
Poisson’s ratio, c 0.2 0.2
90, (tl–t0) used in the closed-form solution for swelling
Time of lining installation, tl (days) 100
deformation before lining installation time

bending moment within it. The non-uniform swelling of rock Comparison of the semi-analytical and numerical results for
under anisotropic stress distribution causes bending moment axial thrust is presented in Figure 4. The axial thrust is mainly
of the lining. A value of 12 for the initial stress ratio (K0 = compressive and touches the maximum at the crown. The
σx0/σy0) were used in numerical analysis (Lo and Yuen, 1981). radial displacement of the tunnel with time is shown in Figure
The proposed model was used for isotropic stress condition 5. The results of numerical analysis proved the inward move-
and the average of initial horizontal stress to vertical stress ment of spring line with time, while the crown moves upward
was assumed as an initial stress condition. Moreover, in the case and the proposed model predicted uniform radial deforma-
of hydrostatic stress, neither shear nor moment will develop tion. The predicted results by (Hawlader et al., 2005) agree
in the support system (Carranza-Torres and Labuz, 2006). well with the records of tunnel advance. Comparison of the
The axial thrust force, in this case, can be calculated as: results between the analysis and the presented closed-form
solution proved a proper agreement between the semi-ana-
a+b
N = ps  ----------- . (32) lytical and the numerical solution for the maximum values
 2 
of axial thrust and radial displacement over the time (see
The results of the proposed solutions for swelling displace- Table 3). Table 3 summarizes the results of total swelling
ments and lining pressure after lining installation are shown displacement in the Heart tunnel.
in Table 2.

Table 2. Predicted swelling displacements and axial thrust forces of Heart Lake tunnel with time
t (years) b' (Eq. 30) Urs(t) (mm) (Eq. 29) Ps (KPa) (Eq. 28) Ps/s Thrust force (KN/m)
1 0.1400 0.6817 2431.65 0.860 3708.27
2 0.1062 0.7082 2527.22 0.894 3854.01
5 0.0933 0.7185 2563.69 0.907 3909.63
10 0.0927 0.7193 2565.39 0.910 3912.22
524 Davoud Parsapour and Ahmad Fahimifar

5.2. Case b

The presented solution is applied to the exploration gal-

lery that was driven in the course of the Freudenstein tunnel
project on the new railway line from Mannheim to Stuttgart.
This gallery was driven starting at the alignment of the Freuden-
stein tunnel and runs parallel to it in the unleached gypsum
Keuper. In this exploration gallery, four support systems were
tested. In case of the resisting principle (Fig. 6), stiff rein-
forced concrete linings with differing thickness were used
(average thickness 0.6 m), which were designed for the swell-
ing pressure. The properties of the support are presented in
Table 4. By means of a monitoring program, the behavior
of the rock and the respective support systems were observed
over a period of more than nine years (Fecker and Wullschläger,
1991; Fecker, 1992).
A three dimensional constitutive law for swelling rock was
presented by Wittke-Gattermann (2003). It describes the
development of time-dependent strains and stresses due to
swelling. This constitutive model was implemented within
Fig. 4. Development of axial thrust in concrete lining with time a numerical computation program according to the finite
(Heart Lake tunnel) (positive compression). element method to evaluate the Freudenstein tunnel. A
good agreement between the numerical solutions and the
monitoring results was achieved and the result is compared
to show the accuracy and validity of the proposed semi-
analytical solution.
The unit weight of the rock mass and depth of the gallery
are 0.023 MN/m3 and 87 m, respectively. To analyze the tunnel,
it was assumed that the tunnel is driven in a hydrostatic ini-

Fig. 6. Freudenstein tunnel, exploration gallery, resisting principle

support system (Wittke-Gattermann, 2003).

Table 4. Properties of the support

Fig. 5. Predicted radial displacement of Heart Lake tunnel with a (m) b (m) Ec (MPa) υc Ks (MPa)
time (negative inward). 3.65 3.05 25000 0.2 4842.20

Table 3. Predicted wall displacement of Heart Lake tunnel

2
a 1+
Elastic displacements (Kirsch equations) Ure =  ----  ------------ = 0.44 mm
r E 
Ure  t   0.6a   1 – e –--------
t 
Swelling displacements when support installed (t = 90 days) qr  k qr = 2.25 mm
 
Swelling displacements after 10 years Equation (29), (see Table 2) = 0.72 mm
Total displacement (mm) after 10 years (Proposed model) Ure + Urs  t  + Equation (29), = 3.41 mm
Maximum Total displacement (mm) after 10 years (Numerical analysis,
= 3.67 mm
Hawlader et al., 2005)
 Semi-analytical solution for time-dependent deformations in swelling rocks 525

Table 5. Swelling and elastic parameters of the rock (Wittke-Gat- tial stress field (s = 2 MPa) (Wittke-Gattermann, 2003). The
termann, 2003) elastic properties of the rock mass and the swelling param-
Elastic Swelling eters obtained from the measurement results are also given
E (MPa)  q (Days) kq in Table 5.
4000 0.25 11000 0.1 Figure 7 and Table 6 show the comparison of the displace-
ments measured with sliding micrometers in the rock mass
below the gallery and the calculated time-dependent swelling
displacements for a section in the resisting part. As the
Figure 7 shows, a proper agreement is observed between the
measured and the semi-analytical results.
The magnitude of the maximum lining pressure calculated
after nine years agrees well with the values obtained from
the measurement results (Fig. 8 and Table 7).

6. SUMMARY AND CONCLUSIONS

The present study deals with an exact solution for the pre-
diction of the time-dependent swelling displacements and
the lining pressure in circular tunnels presented by using
classical tools of calculus variations. The three dimensional
Fig. 7. Time-dependent swelling displacements underneath the invert. stresses (effect of the tunnel face) and anisotropy of the ini-
Table 6. Comparison of swelling displacement for 3 years
Time (days) 100 120 200 240 300 365 (1 year) 480 730 (2 years) 1095 (3 years)
Swelling displacement (mm) (Eq. 29) 1.495 1.695 2.493 2.791 3.237 3.630 4.293 5.107 5.946
Measurement (mm) 1.634 1.728 2.497 2.748 3.102 3.427 3.994 4.812 6.260
Numerical analysis (mm)
1.317 1.445 2.011 2.379 2.893 3.371 4.022 5.394 6.855
(Wittke-Gattermann, 2003)

Fig. 8. Time-dependent lining pressure in center of invert.

Table 7. Comparison between lining pressure in center of invert results for 9 years
12 24 36 72 108 Ps/σ
Time (months) 0 3 6 9
(1 year) (2 years) (3 years) (6 years) (9 years) max
b' (Eq. 30) 1 0.841 0.754 0.677 0.615 0.458 0.369 0.240 0.182
Ps (KPa) (Eq. 28) 0 317.2 491.36 646.56 770.6 1083.8 1262 1520.6 1635.4 0.82
Measurement (KPa) 0 327.47 379.89 414.29 450.88 645.49 824.84 1369.01 1870.11 0.94
Numerical analysis (KPa)
0 82.19 179.23 262.64 344.36 651.10 887.14 1429.23 1787.36 0.89
(Wittke-Gattermann, 2003)
526 Davoud Parsapour and Ahmad Fahimifar

tial stress on the solution were considered. A number of for- Huder, J. and Amberg, G., 1970, Quelling in Mergel, Opali-mudston
mulae for the proposed models were driven to predict the und Anhydrite. Schweizerische Bauzeitung-88. Jahrgang Heft
displacements of lined and unlined circular tunnels and the 43, Oct. 22, p. 975–980.
Einstein, H., 1994, Comments and recommendations on design and
pressure on the lining. The effect of lining stiffness and the analysis procedures for structures in argillaceous swelling rock.
delay in installation of lining was also considered. Then, a International Journal of Rock Mechanics and Mining Sciences
comparison made between the semi-analytical solution and and Geomechanics, 31, 535–546.
the results of two case records and a numerical analysis. The Kirsch, G., 1898, Die theorie der elastizitaet und die deduerfnisse der
proposed models show a proper agreement with the field festigkeitlehre. Zeitschrift des Vereines deutscher Ingenieure, 42,
measurements. 797–807.
Lo, K.Y., Devata, M., and Yuen, C.M.K., 1979, Performance of a
shallow tunnel in a shaly rock with high horizontal stresses. In:
REFERENCES Jones, M.J. (ed.), Proceedings of the 2nd International Sympo-
sium on Tunnelling, London, March 12–15, p. 1–12.
Brady, B. and Brown, E., 1985, Rock mechanic for underground min- Lo, K.Y. and Yuen, C.M.K., 1981, Design of tunnel lining in rock for
ing, first edition. George Allen & Unwin, London, 628 p. long-term time effects. Canadian Geotechnical Journal, 18, 24–39.
Carranza-Torres, C., 2000, Application of the convergence-confine- Lo, K.Y. and Lukajic, B., 1984, Predicted and measured stresses and
ment method of tunnel design to rock masses that satisfy the displacements around the Darlington intake tunnel. Canadian
Hoek-Brown failure criterion. Tunneling and underground Space Geotechnical Journal, 21, 147–165.
Technology, 15, 187–213. (Personal communication with Hoek) Lo, K.Y. and Hefny, A., 1996, Design of tunnels in rock with long
Carranza-Torres, C. and Labuz, J., 2006, Class Notes on Underground term time-dependent and nonlinearly stress-dependent deforma-
Excavations in Rock. Topic 8, Department of Civil Engineering tion. Tunnelling Association of Canada Annual Publication, p.
University of Minnesota, UE-T8, p. 1–6. 179–213.
Davis, P.J. and Rabinowitz, P., 1984, Methods of Numerical Integra- Panet, M. and Guenot, A., 1982, Analysis of convergence behind the
tion (2nd edition). Academic Press, New York, 624p. face of a tunnel. Proceedings of the 3rd International Symposium,
Einstein, H.H., Bischoff, N., and Hofmann, E., 1972, Behaviour of invert Brighton, June 7–11, p. 197–550.
slabs in swelling shale. Proceedings of the International Sympo- Panet, M., 1995, Le calcul des tunnels par la methode convergence-
sium on Underground Openings, Lucerne, Switzerland, Zurich, confinement. Presses de l‘Ecole Nationale des Pontset Chauss-
Sept. 11–14, p. 296–319. ees, Paris, July 1–3, p 17–24.
Fahimifar, A. and Parsapour, D., 2010, Analytical Model for Predic- Tang, S.B. and Tang, C.A., 2012, Numerical studies on tunnel floor
tion of Strains for Tunneling in Swelling Grounds. Proceeding of heave in swelling ground under humid conditions. International
American Society of Civil Engineering Conference, GeoFlorida, Journal of Rock Mechanics & Mining Sciences 55, 139–150.
Feb. 20–24, p. 370–380. Schädlich, B., Marcher, T., and Schweiger, H.F., 2012, Application of
Farzi, S., 2012, Evaluation of tunnel face effect on initial stability of a constitutive model for swelling rock to tunneling. Geotechnical
the tunnel. M.Sc. Thesis, Department of Civil and Environmental Engineering Journal of the Southeast Asian Geotechnical Society
Engineering, Amirkabir University of Technology, Tehran, 105 p. & Association of Geotechnical Societies in Southeast Asia, Vol.
Fecker, E., 1992, Untersuchungen von schwellvorgängen und erpro- 43, ISSN: 0046-5828.
bung von auskleidungskonzeptenbeimfreudensteintunnel. Proceed- Sulem, J., Panet, M., and Guenot, A., 1987, An analytical solution for
ing 10, National Rock Mechanic Symposium, Geotechnik Special time-dependent displacements in circular tunnel. International
Edition, p. 16–20. Journal of Rock Mechanics and Mining Science & Geomechan-
Fecker, E., and Wullschläger, D.J., 1991, Geotechnischemebeinrich- ics Abstracts, 24, 155–164.
tungen in der untersuchungsstreckeu1 des Freudensteintunnels. Wittke, W. and Pierau, B., 1979, Fundamentals for the design and
In: Ingenieurbauwerke Nr. 7, Elite Trust Reg. Vaduz, Dec. 7–11, construction of tunnels in swelling rock. Proceedings of the 4th
p. 194– 213. International Congress on Rock Mechanics, Rotterdam, Sept. 2–
Gnirk, P.F. and Johnson, R.E., 1964, The deformational behavior of 8, Vol. 2, p. 719–729.
a circular mine shaft situated in a viscoelastic medium under Wittke-Gattermann, P., 1998, Verfahrenzurberechnung von tunnels in
hydrostatic stress. Proceedings of the 6th Symposium Rock quellfähigemgebirge und kalibrierung an einemversuchsbauwerk.
Mechanic, Rolla, Oct. 28–30, p. 231–259. Geotechnik in Forschung und Praxis. VerlagGlückauf GmbH,
Grob, H., 1972, Schwelldrucim Belchentunnel. Proceedings of the Essen, 280 p.
International Symposium on Underground Openings, Lucerne, Wittke-Gattermann, P., 2003, Dimensioning of tunnels in swelling
Switzerland. Zurich, Sept. 11–4, p. 99–119. rock. ISRM–Technology roadmap for rock mechanics, South Afri-
Hawlader, B.C., Lee, Y.N., and Lo, K.Y., 2003, Three-dimensional can Institute of Mining and Metallurgy, Sept. 8–12, p 229–237.
stress effects on time-dependent swelling behaviour of shaly
rocks. Canadian Geotechnical Journal, 40, 501–511. Manuscript received May 29, 2015
Hawlader, B.C., Lo, K.Y., and Moore, I.D., 2005, Analysis of tunnels Manuscript accepted July 27, 2015
in shaly rock considering three-dimensional stress effects on
swelling. Canadian Geotechnical Journal, 42, 1–12.
 Semi-analytical solution for time-dependent deformations in swelling rocks 527

APPENDIX 1
2 –3
2a r
------------  -------------2-
Ln10 a
A1. DERIVATION OF EQUATIONS FOR UNLINED 1 – ----2
r 1 2 –1 1
SECTION lim -------------------------------- = lim – ------------  2a r  -------------2 = 0 ,
r 1 r   Ln10 a
– ----2 1 – ----2 (A9)
For far-sections from the tunnel face ( = 1) and hydro- r r
static stress field (k = 1), the time-dependent radial swelling 2
a r–a
displacement based on the Equation (14) is expressed as  lim rLog  1 – ----2  – aLog  ---------- = aLog4 . (A10)
r  r  r + a
follows:
t 2
– --------
a
Ur  t  =  1 – e qr   kqr  a Log  1 – ----2  dr .


(A1) Considering Equations (A8)–(A10), Equation (A7) can
   r be written as follows:








For integrating of the term ‘I’, the corresponding stages are I = aLog4 . (A11)
considered:
By substituting Equation (A11) into Equation (A2), time-
t
– -------- dependent radial swelling displacement of the tunnel wall
Ur  t  =  1 – e  k I,
qr
(A2)
  qr (r = a) is expressed as follows:
t t
– -------- – --------
Ur  t  = aLog4  kqr  1 – e qr   0.6a  kqr  1 – e qr  .
2  
a
I = a Log  1 – ----2  dr ,

(A3)    
 r (A12)








2  2
a 2a  1 A2. DERIVATION OF EQUATIONS FOR LINED SEC-
I = rLog  1 – ----2  – ------------ a ---------------------- dr , (A4)
 r  Ln10 2 a
2 TION
a 
r 1 – ----2
 r
The radial displacement at lined tunnel wall (r = a) is








II
expressed as follows (Eq. 26):
2  2
a 2a
I = rLog  1 – ----2  – ------------  II ,
t 2
– --------
(A5) a  .
Ur  t  =  1 – e qr   kqr   Log  1 – ----b
 
 r  Ln10    r2 
dr (A13)
a a









 I
1 1 1 r–a
- dr = ------ Ln  ---------- .
 
II = a ---------------------
- dr = a -----------------
a
2 2
r – a 
2
2a  r + a a For integrating the term ‘I’, the following steps are con-
r  1 – ----2 
2
 r sidered:
(A6)
  a
2

  – ----   – PS 
Substituting Equation (A6) into Equation (A5), Equation  I =  Log  ----------------------------------- 
2 2
r  a 
 1 – ----b
(A5) is expressed as follows:  a -  dr = a Log
 r2 
dr
   ,
  
a
2
r–a


I =  rLog  1 – ----2   – aLog  ---------- . (A7)   – P
  r   r + a  b = -----------------S (A14)
a
 
The following mathematical steps (Eqs. A8–A10) help to 2
a 
I =  Log  1 – ----b

solve Equation (A7). dr
a  r2 
r–a
limr   aLog  ---------- = 0 ,
2  2
(A8) a  2ba
rLog  1 – ----b

 r + a =  ---------------------------------------- , (A15)
 r2  ba 
2
a
a 2

r Ln10 1 – ---------
 Log  1 – a---- 
2
 r 










2
III
  r2  










II
d  ----------------------------
 1 
---  r – a b 

 r – b 
 a
2
 r  a b
II = ------------Ln ------------------------
a b
= – ------------Ln --------------------- ,
limr   rLog 1 – ----2 = lim -------------------------------------
 r  r dr Ln10  r + a b  a
Ln10  r + b 
528 Davoud Parsapour and Ahmad Fahimifar

(A16)
2 
a 
I = rLog  1 – ----b 1 – b
I = a   bLog --------------------- – Log  1 – b  . (A20)
 r2   
a  1 + b 
2
a  Applying Equation (A20) into Equation (A13) and con-
limr   rLog  1 – ----b –  aLog  1 – b   , (A17)
 r2  sidering Equations (23) and (27), it gives:














IV
U r  t 
 PS = – Ks  -----------
- ,
2   a 
a 
d  rLog  1 – ----b 
a   2  r 
2  t

IV = lim rLog  1 – ----b 1 – b

– --------
= lim -------------------------------------------- U r  t  = a  k qr   1 – e qr   bLog --------------------- – Log  1 – b  ,
r  r  r2
dr     1 + b  

2 –3   – P
1 2ba r  b = -----------------S  PS =  – b  ,
------------  ------------------2- 
Ln10 ba 
1 – --------- 2 2
  =  H. (A21)
r = lim
= lim ------------------------------------- –2ba
-.
---------------------------------- (A18)
r 1 r 2
 r – a----b Ln10 Solving the coupled Equation (A21) with respect to b'
– ----2
r  r  parameter, the general relationship, for evaluation of rock-
support interaction in circular tunnel driven in hydrostatic
Substituting Equation (A18) into Equation (A17), it results: stress field, resulted as follows:
t
– --------
III = – aLog  1 – b  .  =   – b  = KS kqr  1 – e 
qr
(A19)

Substituting Equations (A16) and (A19) into Equation 1 – b
 bLog ---------------------
(A15), resulted in – Log  1 – b  . (A22)
  1 + b  