You are on page 1of 11

ISSN 0016-8521, Geotectonics, 2016, Vol. 50, No. 3, pp. 336–346. © Pleiades Publishing, Inc., 2016.

Prediction of Swelling Rocks Strain in Tunneling1


D. Parsapour and A. Fahimifar
Department of Civil and Environmental Engineering,
Amirkabir University of Technology, Hafez Ave., Tehran, Iran
e-mail: fahim@aut.ac.ir
Received September 8, 2015

Abstract—Swelling deformations leading to convergence of tunnels may result in significant difficulties


during the construction, in particular for long term use of tunnels. By extracting an experimental based
explicit analytical solution for formulating swelling strains as a function of time and stress, swelling strains are
predicted from the beginning of excavation and during the service life of tunnel. Results obtained from the
analytical model show a proper agreement with experimental results. This closed-form solution has been
implemented within a numerical program using the finite element method for predicting time-dependent
swelling strain around tunnels. Evaluating effects of swelling parameters on time-dependent strains and tun-
nel shape on swelling behavior around the tunnel according to this analytical solution is considered. The
ground-support interaction and consequent swelling effect on the induced forces in tunnel lining is consid-
ered too. Effect of delay in lining installation on swelling pressure which acting on the lining and its structural
integrity, is also evaluated. A MATLAB code of “SRAP” is prepared and applied to calculate all swelling anal-
ysis around tunnels based on analytical solution.

Keywords: tunnel, swelling strain, analytical solution, time, lining, SRAP


DOI: 10.1134/S0016852116030092

INTRODUCTION high repair costs in a number of tunnels surrounded by


Tunneling in swelling rocks may result in many dif- such rocks which are capable of swelling. Swelling is
ficulties, including distribution of additional stresses usually attributed to the anhydrite change into gypsum.
in tunnel support system over an extended period of It is markedly the time-dependent process which might
time, which may lead to failure. Therefore, during the take several decades to make the swelling occur and get
design phase of tunnel, it is of paramount importance completed in nature. Serafeimidis and Anagnostou [20]
to predict the swelling response of surrounding proposed a simplified model for anhydrite hydration,
ground. For example, the swelling phenomena in a which defines the governing process and duration of
large number of tunnels in Southwestern Germany swelling process. In general, two types of swelling
have caused strong pressure against lining and high mechanisms are recognized in rocks: mechanical and
floor heave in tunnels and in some cases swelling pres- physico-chemical swelling [3]. Swelling behavior
sures lead to the destruction of the tunnel lining [5]. depends on crucial factors, such as material properties,
Swelling mechanism is a combination of physico- boundary conditions, negative pore pressure and stress
chemical reaction involving water and stress relief. The distribution of the rock mass surrounding the tunnel
physico-chemical reaction in presence of water is usu- [6]. The swelling phenomenon in tunneling is treated as
ally a major contribution but it can only take place a humidity-mechanical coupled process, i.e. the stress
simultaneously with or following stress relief. As a redistribution, as well, water vapor diffusion around the
result of this mechanism, the volume of surrounding tunnel is taken into account [21].
rock mass can increase with time and the rock mass
moves inward over time. The rate of inward movement Up to now, no common opinion is offered about
and amount of deformation depend upon rock type, the applicability of constitutive laws and analyses or
stress distribution, and tunnel geometry [3]. dimensioning methods with which the swelling phe-
Swelling can take place in soils and rocks where clay nomena of rocks can be described. However, various
minerals, anhydrite and pyrite/marcasite are present. opinions still prevail among the experts on suitability
Anhydritic clay stones are among the most problematic of different design concepts for tunnels in swelling
rocks in tunneling. Many cases have been recorded rocks. This is to be recognized, although significant
about the precedency of serious damages to lining and efforts have been made in recent past by many
researchers that have cited in the paper. To choose an
1 The article is published in the original.
appropriate tunnel lining in swelling rocks, the tunnel

336
PREDICTION OF SWELLING ROCKS STRAIN IN TUNNELING 337

engineer has to quantify the swelling strains, timely ceased at a specific level of compressive stress.
and correctly. Y. Huder and G. Amberg [14] proved experimentally
Several time-independent constitutive models dependence of swelling on stress distribution in marl
have been developed for swelling of rocks [7, 10, 23]. and clay-stone rocks. Y. Huder and G. Amberg
H.H. Einstein [8] examined the available swelling applied an axial load to a cylindrical rock sample in
models in literature and concluded that, although the oedometer. Then, the sample is unloaded and subse-
swelling deformation due to chemical changes may quently watered, and finally strains due to swelling are
be adequately predicted using available models, but measured. Strain magnitudes are plotted for different
time-dependent swelling models need to be devel- normal stresses on a logarithmic scale. The plot fits
oped from a fundamental knowledge of swelling well with a straight line approximation. The results of
mechanism. An overview of the theoretical models Y. Huder and G. Amberg [14] tests show the relation
and experimental investigations of swelling rock can between normal stress and normal swelling strain.
be found in [1]; the authors of [16, 17] developed rhe- F.T. Madsen [18] has introduced modification to the
ological models for prediction of swelling properties original procedure of Y. Huder and G. Amberg test,
of rocks surrounding tunnels using closed-form solu- basically removing loading/unloading cycle.
tions. Field monitoring of tunnels showed that these Based on the results of Y. Huder and G. Amberg
models simulate the trend of swelling behavior well test, H. Grob [10] proposed a mathematical model for
and they are being used in design of underground relating final swelling strain to stress. According to this
openings. However, K.Y. Lo and C.M.K. Yuen [16] model, swelling strain decreases with logarithm of
did not consider the effect of stress on swelling in stress, and maximum swelling occurs when there is a
their model, K.Y. Lo and C.M.K. Hefny [17] minimum normal stress acting on the specimen. This
assumed that only the radial pressure on tunnel lining relationship is expressed in Eq. (1) below:
has an effect on swelling deformation. The exclusion
of any stress effect makes the prediction conservative. ε ∞zs = K q log(σ z σ 0 ), (1)
It is also necessary to predict long-term swelling
deformation that is observed in the field. Y.R. Kiehl
[15] developed a three-dimensional constitutive law where ε ∞zs is the maximum value of normal swelling
on the basis of the results obtained from multi-axial strain, Kq is dimensionless coefficient of final normal
laboratory swelling tests. According to the swelling swelling strain, and σ z is a principal normal stress
law, the total strain rate consists of elastic, time-
dependent swelling and viscoplastic components. corresponding to direction of swelling strain, and σ 0
This model has been implemented within a numeri- is a least principal normal stress at which swelling
cal program using finite element method. doesn’t occur (the upper limit of stress). P. Wittke-
Gattemann [24] expanded Grob’s mathematical
On the basis of laboratory tests, A. Hefny et al. [13] model as in Eq. (2):
proposed a rheological model consisting of three Kel-
vin units connected in series, in order to model the
non-linear long term swelling behavior of Queenston ⎧0, σ z ≥ σ0

shale. The time dependency was simulated in the ε ∞zs = −K q ⎨log(σ z σ 0 ), σ c ≤ σ z ≤ σ o. (2)
model using rate parameters, and stress dependency ⎪log(σ c σ 0 ), σ z ≤ σ c
was handled through the use of stress dependent ⎩
deformation moduli for Kelvin unit.
In which, σ c is the least principal normal stress that
B.C. Hawlader et al. [11] developed a constitutive causes reduction of swelling strain to occur (lower
model for rock swelling that considers three-dimen- limit of swelling stress).
sional stress effects and anisotropic swelling. The aniso-
tropic and three-dimensional stress-dependent swelling Figure 1 shows the linear decrease of swelling
model cannot be presented in a closed-form solution, strains in no rmal direction of a rock specimen with
and a numerical method must be applied. B.C. Haw- logarithm of stress, in which the slope line denoted as
lader et al. [12] used a finite element algorithm, incor- Kq, and it physically defines the swelling potential. For
porating new constitutive model, for numerical analy- small stresses, swelling is limited by a minimum prin-
sis. In this paper, an experimental based analytical solu- cipal normal stress, σ c , then swelling decreases lin-
tion is used to compute time-dependent swelling early for stresses more than σ c , and eventually swelling
strains. The application of this model is also presented strains cease at a maximum normal stress σ 0 .
to evaluate swelling strain around tunnels.
PROPOSAL OF A TIME-DEPENDENT
SETTING THE FINAL SWELLING STRAIN ANALYTICAL SOLUTION
Increase in volume of a swelling rock depends upon Mathematical models presented in Eqs. (1) and
stress distribution. This is why that swelling will be (2), are limited to computing final swelling strains at

GEOTECTONICS Vol. 50 No. 3 2016


338 PARSAPOUR, FAHIMIFAR

0.7 εzs(t) ε∞zs


1.0
0.6 σz = 0
Test results dε
0.5 W. Gatterman 0.8 dt
Test results
0.4
ε∞zs , %

Proposed model
0.6

εzs(t), %
0.3 σz = 16.2 KPa
σz = 32.2 KPa
0.2 0.4 σz = 65 KPa
0.1 σ0 σz = 130 KPa
σc
0.2
0 σz = 260 KPa
1 10 100 1000
σz = 520 KPa
σz, KPa
0 5 10 15 20 25 30
Time, days
Fig. 1. Final swelling strain vs. principal normal stress,
Expanded model (Eq. (2)) [24]. Fig. 2. Experimental results for swelling strains vs. time [14].

equilibrium (t = ∞). Thus, determining swelling time. If isotropic swelling is assumed, it can be
strains during time is not possible. A. Fahimifar and adopted from Fig. 3:
D. Parsapour [9] employed a closed-form solution to
achieve a time-dependent swelling strain formulations d ε zs (t ) e zs ( ∞ ) − ε zs (t )
= . (3)
based on experimental results considering Wittke- dt C
Gattemann’s model. The calculation process of solu-
Equation (3) was also presented by Y.R. Kiehl [15].
tion is described in the following sections.
Comparison between Eq. (3) and Kiehl’s swelling law,
For the purpose of obtaining a proper time-depen- yields C = ηq , that ηq is time-dependent modulus of
dent analytical solution, on the basis of real behavior
of swelling rock, a mathematical relation is derived swelling.
between the progressive swelling strains and time Incorporating Eq. (2) into Eq. (3) and setting C = ηq
using experimental results from Huder-Amberg tests yields Eq. (4):
[14], as is shown in Fig. 2. The analytical solution is
also illustrated in Fig. 3 [9]. d ε zs (t )
Swelling strains in one-dimenional condition are dt
represented as a function of two variables of stress and ⎧− ε zs (t ) , σz ≥ σ0 (4)

= ⎨K q log(σ z σ 0 ) − ε zs (t ) , σ c ≤ σ z ≤ σ o ,
1
ηq ⎪
⎩K q log(σ c σ 0 ) − ε zs (t ) , σ z ≤ σ c
C Equation (4) is solved using Bernoulli’s bivariate
solution [25] and final relationship for swelling strains
zs ( t )

as a function of time and principal normal stress is


derived as follows [9]:
zs – εzs

dε εzs(t)
dt
ε∞zs

ε zs (t )
εzs(t), %

⎧0, σz ≥ σ0
ε∞zs

( ) ⎪
(−t ηq ) (5)
= 1−e K q ⎨log(σ z σ 0 ), σ c ≤ σ z ≤ σ o .
εzs(t)

⎪log(σ c σ 0 ), σ z ≤ σ c

In which:
Time ε zs (t ): Principal swelling strain at time (t) due to
stress relief; t : Time; Kq : Coefficient of swelling strain
Fig. 3. Mathematical model of swelling strain vs. time [9]. (dimensionless); σ z : Total Principal normal stress;

GEOTECTONICS Vol. 50 No. 3 2016


PREDICTION OF SWELLING ROCKS STRAIN IN TUNNELING 339

σ 0 : Least total principal normal stress at which swell- prediction, but the present model gives a quick esti-
ing strain ceases (upper limit of stress); σ c : Least total mation of swelling effects in a tunnel. For design pur-
principal normal stress at which swelling strain begins poses, like what is implemented in the current paper,
to decrease (lower limit of stress); η q : Time dependent any adoption between two numerical and closed-form
modulus of swelling. solutions depends on the nature of project, availability
of a complete characterization of rock mass and geo-
The convergence of tunnel contour is the most logical conditions. It goes without saying that numer-
important effect arises from swelling, particularly ical solution has to consider anisotropic and three-
during and after construction. According to the exper- dimensional stress-dependent swelling models. It
iments used in Grob’s model expansion (Eq. (2)), would be expedient to use the present solution for pre-
strains in rock mass surrounding a tunnel are calcu- liminary design consideration and then, a numerical
lated in radial direction, and tangential strains are method to examine the selected lining for finalizing
practically zero. This behavior is similar to actual design. For underground structures in rocks of high
deformation behavior of tunnels after excavation, in swelling potential, the present model could remark-
which maximum principal stresses are tangent to tun- ably facilitate design measures to be adopted.
nel perimeter, and as a result, swelling strains in this
direction are negligible. Such a condition may be sim-
ulated in oedometer testing apparatus, in which lateral VALIDATION OF PROPOSAL SOLUTION
deformations of rock sample are prohibited by steel
A. Fahimifar and D. Parsapour [9] used the experi-
sample holder. Thus, by replacing stress σ z with the mental results of odeometer test done by Y. Huder and
radial stress σ r in Eq. (5), swelling deformations in G. Amberg [14] and verified the analytical solution.
tunnel invert and sidewalls are predicted. In Eq. (5), Experimental results from Y. Huder and G. Amberg
parameters Kq, σ 0 , σ c and η q are obtained from oedometer tests on samples of swelling rocks were used
oedometer swelling tests, and stress, σ z , can be calcu- to examine the results obtained using proposed analyti-
lated using analytical or numerical approaches. cal model. Swelling strains calculated using proposed
model in this paper, were also compared with the results
The proposed model is based on a series of labora- of rheological model of A. Hefny et al. [13] and also four
tory experiments ran by Y. Huder and G. Amberg, and investigations and testing programs on rock cores recov-
thus, it is expected that the corresponding results are ered at different depths and locations in Queenston for-
conditioned by the results and duration of conducted mation at the Niagara area [13]. Using measured results
swelling tests. However, it is shown that the model is of investigations and testing programs in Queenston
capable of predicting swelling deformation of tunnel formation, as shown in Fig. 4, and also by using try and
walls with an acceptable approximation. Though the
model allows one to calculate swelling strains at a error method, approximate values as η q = 75 hours,
given time, there are still few restrictions due to the Kq = 0.23% and σ 0 = 600 KPa are allocated to the rock.
following issues: Proposed model illustrates appropriate correlation in
—The model is clearly dependent upon swelling representing swelling strains within period of time, in
tests duration and the amount of maximum swelling particular; it characterizes adequately the strains for
strain measured. higher values of time as representatively shown in Fig. 4.
This figure proves that two models have different slopes
—No scale factors are considered between labora- in long term. This could result in substantial differences
tory and in-situ behavior. when very long periods, as decades, are considered.
—Definition of “long term” referred to tunnel ser-
vice life is different from that of “long term” for labo- Parameter, ηq , in Eq. (5) which was defined as time
ratory tests. dependent modulus of swelling, impresses rate of
swelling strains in rocks. The effects of this parameter
However, it is important to denote the following was studied by A. Fahimifar and D. Parsapour [9] and
points to distinguish necessity and merits of the pre-
sented analytical solution: it is observed that with a high value of 1/ηq , rock mass
reaches its ultimate swelling strain in a shorter time
(a) The proposed model, in closed-form solution and rate of swelling strain raises accordingly. In gen-
for the case of supported tunnel, represents significant eral, it may be implied that, for surrounding rocks with
advances, as it lets calculation of lining pressure build-
up with time. Besides, the model includes stress higher values of 1/ηq , tunnel convergence needs to be
dependency in simulation of time dependent behavior. controlled rapidly. It may also be concluded that, in a
The present model is capable of taking into account rock mass with specific geomechanical properties,
the state of initial stress prior to excavation, time value of ηq does not have any effect on final value of
dependent properties of rock, and time lapse occur- swelling strain, however, it characterizes the rate or
ring between tunnel excavation and lining installation. time of reaching ultimate value of swelling strain.
(b) Although anisotropic and three-dimensional Previous arguments dealt with a model extraction
stress-dependent swelling models can provide better and data verification following it, the new proposed

GEOTECTONICS Vol. 50 No. 3 2016


340 PARSAPOUR, FAHIMIFAR

Time, days
0 100 200 300 400 500 600 700
0.7
Measured
Proposed model σz = 20 KPa
0.6 Hefny et al., 1996

0.5
Swelling strain, %

σz = 100 KPa
0.4

0.3

0.2 σz = 1000 KPa

0.1 σz = 3000 KPa

Fig. 4. Typical measured [13] and calculated swelling strains vs. time under different applied stresses.

model is used for a tunnel analysis implementing vari- chanical parameters including material properties,
ous tunnel geometrry, so as the model becomes appli- model dimensions, boundary conditions, and rock
catory and above all, the effect of various factors such overburden were chosen the same and equal for all tun-
as tunnel section geometry and lining installation time nels. 2D stress analyses were performed by numerical
are examined. analysis of tunnels, i.e. finite element method under
stress ratio (horizontal to vertical in-situ stress) as K =
THE ROLE OF TUNNEL SHAPE 0.7. In-situ vertical stress is assumed as the overburden
ON SWELLING DEFORMATION pressure (according to tunnel depth), and in-situ hori-
A. Fahimifar and D. Parsapour [9] evaluated the zontal stress is obtained from stress ratio (K0). Numeri-
tunnel geometry effect on swelling strains in unsup- cal analysis was just used for obtaining stress distribu-
ported tunnels. Here, in addition to a brief history of tion around each tunnel and swelling strains were calcu-
their works in 2010, further studies as well as new tun- lated by analytical solution (Eq. (5)).
nel analysis are provided too. Besides, the ground-
support interaction and the effect of support installa- Swelling strain and stress were calculated as a
tion time on tunnel deformation and support’s inter- function of time, using the programming capabilities
nal forces induction are considered. of MATLAB R2013a code [19] and, was entitled with
To evaluate the effect of tunnel geometry on swelling “Swelling Rock Analysis Program” or “SRAP”.
deformations, five different types of tunnel shapes were SRAP can closely model the events and subsequent
considered in analyses, as shown in Fig. 5. All geome- effects (swelling stress and strain) of tunnel construc-
tion processes. This approach provides mathematical
relations for lining pressure, stress, thrust force and
Type 1 Type 2 Type 3
moment through the rock and lining interface for a
R4775 mm

R4775 mm
R4775 mm

8598.84 mm
1423.84 mm 7175 mm
7175 mm
7475 mm

8175 mm

range of isotropic swell and deformation behavior


85 mm

parameters and also provides an explicit strain/stress


R8567.
m
15 m

versus time solution.


R13
1000 mm

8950 mm 8950 mm 8950 mm


In the analyses, surrounding rock mass is assumed
to be isotropic and elastic .The basis of formulation in
Type 4 Type 5
MATLAB is described below:
R4775 mm

R4775 mm
2375 mm 7175 mm
2375 mm 7175 mm

9550 mm
9550 mm

Total strain ε ltot (t ) is the total sum of elastic strain


ε1el (t ) and swelling strain ε1swelling (t ) :
8950 mm 8950 mm

Fig. 5. Different tunnel geometries. {ε tot


1 (t )} = {ε1el (t )} + {ε1swelling (t )} . (6)

GEOTECTONICS Vol. 50 No. 3 2016


PREDICTION OF SWELLING ROCKS STRAIN IN TUNNELING 341

The elastic strain is based on three dimensional X, m


Hooke’s law, so the swelling strain according to Eq. (5) –20 –10 0 10 20
is as follows: 40

ε1tot (t ) = 1 σ1 (t )
E 30
t (7)
− ν ( σ 2 (t ) + σ 3 (t )) + ε 1swelling
∫ ( σ1 (t ) , ε1
swelling
)
(t ) dt.

Ht
E 20
0

In which, E and ν are elastic modulus and Pois-

Y, m
H
son’s ratio of rock mass, respectively.
10
According to Eq. (7), principal normal stress σ1(t) R1

Hm
can be calculated as a function of time as in Eq. (8): a
R2
b 0
⎛ t

( )
W
σ1 (t ) = E ⎜ ε1tot (t ) − ε1swelling σ1 (t ) , ε1swelling (t ) dt ⎟

Hb
⎜ ⎟ (8)
⎝ 0 ⎠ –10
+ ν ( σ 2 (t ) + σ 3 (t )) . –14
Mesh generation and numbering of elements and Bl Br
nodes were carried out using automatic mesh genera- B
tion capabilities of ANSYS15 [2]. The finite element
model is based on finely divided mesh with isoparamet-
ric Q4 elements (containing four integration points) and Fig. 6. General geometry model and ANSYS mesh gener-
ation
plane strain condition [4]. Typical geometrical model
and finite element mesh are shown in Fig. 6.
Mesh generation was undertaken to ensure that data for SRAP, to perform stress analysis of surround-
aspect ratios of all elements to be as close as to unity, ing rock mass.
particularly in proximity of tunnel boundaries. Geo- Swelling analysis of surrounding rock mass was per-
metrical characteristics of the model and tunnels cross formed by SRAP. Subsequently, total strains were com-
sections are listed in Table 1. puted both in horizontal and vertical directions (radial
The geomechanical properties of surrounding rock strain) for the sidewalls (Point ‘a’) and invert (Point ‘b’)
mass are listed in Table 2. These values are chosen of all section shapes. All tunnels with specific section
based on practical projects and previous swelling tests. were considered unsupported (without lining) in the
ANSYS program is only used for modeling of tunnel analyses. The results were plotted as strain-time curves,
geometry and its surrounding rock mass. The results as illustrated in Figs. 7 and 8. As is observed, type 5, i.e.
from ANSYS including mesh specification, such as the inversed U-shaped tunnel, has the largest deforma-
numbering of elements and nodes, are used as input tion in both sidewalls and invert. On the other hand,

Table 1. Geometric properties of the model and parameters of tunnels sections (meters)
W R2 R1 B Br Bl Ht Hm Hb H Type

8.95 – 4.775 40 20 20 30 7.475 12.525 50 1


8.95 13.150 4.775 40 20 20 30 8.175 11.825 50 2
8.95 8.568 4.775 40 20 20 30 8.599 11.40 50 3
8.95 4.775 4.775 40 20 20 30 9.550 10.45 50 4
8.95 – 4.775 40 20 20 30 9.550 10.45 50 5

Table 2. Geomechanical properties of surrounding rock mass


E (MPa) ν K0 γ (KN/m3) Kq ηq (hours) σ 0 (MPa) σ c (MPa)

4000 0.25 0.7 22.00 0.5% 500 4.25 0.005

GEOTECTONICS Vol. 50 No. 3 2016


342 PARSAPOUR, FAHIMIFAR

1.6 1.6
1.4 1.4
1.2 1.2
Total strain, %

Total strain, %
1.0 1.0
0.8 0.8
0.6 0.6
0.4 0.4
0.2 0.2

0 1 2 3 4 5 6 0 1 2 3 4 5 6
Time, mounths Time, mounths
Type 1 Type 3 Type 5 Type 1 Type 3 Type 5
Type 2 Type 4 Type 2 Type 4

Fig. 7. Horizontal (radial) strain curves for tunnel sidewalls Fig. 8. Vertical (radial) strain curves for tunnel invert
(Point ‘a’). (Point ‘b’).

type 4 (circular tunnel) has the least amount of defor- mum and maximum bending moments occur in tun-
mation in sidewalls and invert. Strains in sidewalls and nel types 4 and 5, respectively. Critical combination of
invert of all tunnels reach maximum value, within internal forces occurs in end sides of invert. To evalu-
approximately 3 months after excavation. ate the structural integrity of concrete lining due to
Swelling behavior in each tunnel invert, is evalu- swelling deformations, the force-moment interaction
ated using SRAP. In Fig. 9, the swelling zones in invert diagram was constructed. Through the graph of inter-
have been displayed by shaded zones. All tunnels have nal forces, Mmax and P, structural integrity of concrete
been considered unsupported in analyses. Minimum lining was examined at two locations, i.e. at two ends
strain for beginning of swelling, in these analyses, was of tunnel invert. As illustrated in Fig. 10, except for
selected as 0.2% (an average of about 1.5 mm). Type 5 tunnel type 5, the support system will be adequate.
(inversed U-shaped tunnel) has the largest shaded As swelling strains in rocks are highly time-depen-
zones corresponding to a large vertical displacement in dent, the SRAP can be used to evaluate the effect of
invert, and type 4 (circular tunnel) has the least shaded (lining) support installation time on radial swelling
zone corresponding to a small vertical displacement in pressure development acting on the support. It is
invert. It is also evident that, swelling strains in the cir- shown that small strains have a strong influence on the
cular tunnel are minimum. swelling stress [22]. If a period of time is allowed to
The effect of swelling strain on tunnel lining struc- lapse between excavation of tunnel and lining installa-
ture which originates from final swelling deforma- tion, so lining pressure would significantly reduce.
tions, can also be evaluated in SRAP, for internal Thus, considering type of surrounding rock mass and
forces; including thrust force and bending moment. method of tunnel excavation, maximum possible
The lining consists of 30 cm thick concrete, reinforced elapsed time between excavation and installation of
with two rows of ∅20@100 × 100 mm steel bars. lining can be used to permit swelling deformations of
Mechanical properties of the lining used in the analy- rock mass to occur. Thus, the pressure acting on sup-
ses are listed in Table 3. port, would decrease. In this regard, tunnel type 3 with
The results obtained from SRAP, are shown below an arch invert was analyzed with different values of
in Fig. 10, as a pair of maximum bending moment and time lapse before lining installation. The results are
thrust force in the same location in the lining. Mini- illustrated in Fig. 11, indicating maximum radial

Table 3. Mechanical properties of lining


Concrete properties Steel bar properties

f c' (MPa) ν γ c (KN/m3) Ec (MPa) Gc (MPa) bar type f y (MPa) f u (MPa)

24 0.2 24 24500 10200 AIII 400 600

GEOTECTONICS Vol. 50 No. 3 2016


PREDICTION OF SWELLING ROCKS STRAIN IN TUNNELING 343

swelling pressure in invert and sidewalls at elapsed Type 1 Type 2


times of zero, 1, 2, 7, 14, and 28 days between excava- 9 9
tion face and lining installation time. 8 8
7 7
Figure 11, evidently illustrates that with delay in 6 6
lining installation time, swelling pressure is signifi- 5 5
cantly reduced. It could be implied that, maximum 4 4

Y, m

Y, m
radial swelling pressure in invert is reduced by 30% in 3 3
one day, and 50% in two days. 2 2
1 1
For better understanding of effect of time lapse on 0 0
internal forces reduction before lining installation, –1 Swelling zone
–1 Swelling zone

design analysis of lining for tunnel types 1 and 5 –2 –2


which have greater internal forces (according to –3 –3
–4 –2 0 2 4 –4 –2 0 2 4
Fig. 10), were re-analyzed, with respect to time X, m X, m
effect. As shown in Figs. 12 and 13, for type 1, the Type 3 Type 4
interaction points for elapsed times (between excava- 9 9
tion face and lining installation time) of 1, 2, 7, 14, 8 8
and 28 days as computed by SRAP are all within the 7 7
envelope. For type 5, the interaction points for 6 6
elapsed times of 7, 14, and 28 days all lie within the 5 5
4 4

Y, m

Y, m
envelope. However, the interaction points for zero, 1, 3 3
and 2 days of elapsed times are outside the envelope 2 2
which clearly reveals time effect of swelling on struc- 1 1
tural integrity of lining. 0 0 Swelling zone
Swelling zone
–1 –1
–2 –2
SUMMARY AND CONCLUSIONS –3 –3
–4 –2 0 2 4 –4 –2 0 2 4
Time-dependent deformation is a significant issue X, m X, m
in design of underground structures in swelling rocks, Type 5
specifically for long term performance of tunnels. The 9
8
deformation level due to swelling depends on stress 7
distribution and swelling potential of surrounding rock 6
mass. Since, experience about swelling rocks around 5
tunnels clearly shows that swelling deformations need 4
Y, m

to be considered in design of these structures, so addi- 3


tional work to incorporate further details of constitu- 2
tive response and construction process is clearly war- 1
0
ranted. –1 Swelling zone

Following the expansion of Wittke-Gattermann’s –2


model with considering stress in surrounding rock mass –3
of tunnel, behavior of swelling rock can be predicted –4 –2 0 2 4
X, m
with proper approximation. Stress in rock surrounding
tunnel can be calculated analytically or numerically,
and swelling potential of rock Kq can be determined Fig. 9. Swelling zones in tunnels’ invert.
experimentally. Considering time-dependent charac-
teristics of rock formation with high swelling potential,
an analytical model and a computer program of lining interface for a range of isotropic swell and defor-
“SRAP”, presented in this paper, can be used to com- mation behavior parameters and also provides an
pute swelling strain and stress with regard to time. The explicit strain/stress versus time solution.
presented analytical solution and corresponding
approach, for analysis of tunnels in swelling rock, con- Finally, based on the results presented in this
sider time-stress dependent behavior of swelling rock paper, in rocks with high potential of swelling, Kq, and
and taking into account initial stress distribution prior low time modulus of swelling ηq , it is better to use sec-
to excavation, time dependent properties of rock, and tions with shapes more or less similar to circular
time lapse occurring between tunnel excavation and lin- shapes (3 and 4 sections). Such tunnel sections, not
ing installation. This approach also includes interaction only insures the least swelling zone around tunnel, but
between rock and tunnel structure. also a delay of one or two days in support installation,
The solution provides expressions for lining pres- causes 30 and 50% reduction in swelling pressure,
sure, stress, thrust force and moment along rock and respectively, reduces support internal forces as well.

GEOTECTONICS Vol. 50 No. 3 2016


344 PARSAPOUR, FAHIMIFAR

5000

4000
Force-moment diagram
3000 Type 1
Thrust force, KN

2000 Type 2
Type 2 Type 5 Type 3
1000 Type 4 Type 3
Type 1 Type 4
0
200 400 600 800 1000 1200 1400 Type 5
–1000

–2000
3.6 m
–3000 Bending moment, KN-m

Type 1 Type 2 Type 3 Type 4 Type 5


P, KN 677.9 790.0 834.3 543.7 676.8
Mmax, KN-m 800.0 767.9 670.0 293.4 1329.7

Fig. 10. Force-moment interaction diagram of linings (the ends sides of invert) for various tunnel types.

700
Hor. stress sidewall
Ver. stress invert
600

500
Lininig pressure, KPa

400

300

Invert
200
Sidewall

100

0
2 4 6 8 10 12 14 16 18 20 22 24 26 28
Elapsed time, days

Fig. 11. Maximum radial swelling stress in the invert and sidewalls as a function of time of lining installation.

Unlike the sections with shapes nearly the same as cir- economically remarkable on account of rise in excava-
cular sections, in the case of inversed U-shaped sec- tion volume and consequently the lean concrete vol-
tions, all conditions assumed the same, not only the ume in tunnel invert. Such cases may happen in sub-
swelling zone is bigger, but also there is much more way, transport and railway tunnels. So in current cases,
internal forces in support than that of in nearly circular it is strongly recommended to enhance invert curve-
shapes. It is recommended that in NATM excavated ness (sections No. 2 and 3) to the purpose of swelling
tunnels, choosing circular sections are probably less strains decline to desirable degree. Further more, if the

GEOTECTONICS Vol. 50 No. 3 2016


PREDICTION OF SWELLING ROCKS STRAIN IN TUNNELING 345

5000

4000
Force-moment diagram
3000 Day zero
Day 1
Thrust force, KN

2000
Day 2
Day 7
1000 Day 1 Day zero
Day 14 Day 7 Day 2 Day 14
Day 28 Day 28
0
200 400 600 800 1000 1200
–1000

–2000

–3000 3.6 m
Bending moment, KN-m

Fig. 12. Force-moment interaction diagram for lining of tunnel type 1 (the ends sides of invert) at various times of lining instal-
lation.

5000

4000
Force-moment diagram
3000 Day zero
Day 1
Thrust force, KN

2000 Day 2
Day 7
1000 Day zero Day 14
Day 7 Day 2 Day 1 Day 28
Day 14
Day 28
0
200 400 600 800 1000 1200 1400
–1000

–2000
4.78 m
–3000
Bending moment, KN-m

Fig. 13. Force-moment interaction diagram for lining of tunnel type 5 (the ends sides of invert) at various times of lining instal-
lation.

ground characteristics and features, letting no delay in REFERENCES


support installation, it is better to use a temporary 1. G. Anagnostou, E. Pimentel, and K. Serafeimidis,
flexible lining, such as shotcrete and wire mesh or “Swelling of sulphatic claystones-some fundamental
fiber reinforcement shotcrete, makes swelling defor- questions and their practical relevance,” Geomech.
mation to take place prior to installation of final lining. Tunnelling 3, 567–572 (2010).
As a result, methods of tunnel excavation and sup- 2. ANSYS Ver. 15.0. User’s Manual (ANSYS Inc., 2013).
port installation in swelling ground govern rate of tun- 3. M. Barla, Ph. D. Thesis, (Politecnico di Torino, Turin,
nel advance, and this for its part, overrides magnitude 1999).
of swelling deformation and eventually lining perfor- 4. K. J. Bathe, Finite Element Procedures (Prentice-Hall,
mance. Englewood Cliffs, NJ, 1982).

GEOTECTONICS Vol. 50 No. 3 2016


346 PARSAPOUR, FAHIMIFAR

5. I. Berdugo, E. Alonso, E. Romero, A. Gens, and 15. J. R. Kiehl, “Ein dreidimensionales Quellgesetz und
M. Albis, “A review of expansive phenomena in seine Anwendung auf den Felshohlraumbau,” Proceed-
Wagenburg North Tunnel,” Rev. Acad. Colomb. Cienc. ings of the 9th National Felsmechanik Symposium,
Exactas, Fis. Nat. 33, 455–468 (2009). Aachen, Germany, 1990, pp. 185–207.
6. M. Bonini, M. Barla, and G. Barla, “FLAC applica- 16. K. Y. Lo and C. M. K. Yuen, “Design of tunnel lining
tions to the analysis of swelling behavior in tunnels,” in rock for long-term time effects,” Can. Geotech. J.
2nd FLAC Symposium on Numerical Modeling in Geo- 18, 24–39 (1981).
mechanics, Lyon, France, 2001 (Balkema, Lisse, 2001), 17. K. Y. Lo and A. Hefny, “Design of tunnels in rock with
pp. 29–31. long term time-dependent and nonlinearly stress-
7. H. H. Einstein, N. Bischoff, and E. Hofmann, “Verh- dependent deformation,” Can. Tunneling, 179–213
alten von Stollensohlen in quellendem Mergel,” in (1996).
Berichte, Internationales Symposium für Untertagebau, 18. F. T. Madsen, “Suggested methods for laboratory test-
Luzern, Switzerland, 1972 (1972), pp. 296–319. ing of swelling rocks,” Int. J. Rock Mech. Min. Sci.
8. H. Einstein, “Comments and recommendations on Geomech. 36, 291–306 (1999)
design and analysis procedures for structures in argilla- 19. MATLAB R2013a Ver. 8.1.0. User’s Manual (Math-
ceous swelling rock,” Int. J. Rock Mech. Min. Sci. works Inc., 2013).
Geomech. 31, 535–546 (1994). 20. K. Serafeimidis and G. Anagnostou, “On the time-
9. A. Fahimifar and D. Parsapour, “Analytical model for development of sulphate hydration in anhydritic swell-
prediction of strains for tunneling in swelling grounds,” ing rocks,” Rock Mech. Rock Eng. 46, 619–634 (2013).
in GeoFlorida 2010: Advances in Analysis, Modeling and 21. S. B. Tang and C. A. Tang, “Numerical studies on tun-
Design, West Palm Beach, USA, 2010 (Curran Assoc., nel floor heave in swelling ground under humid condi-
Inc., Red Hook, NY, 2010), Vol. 1, pp. 370–380. tions,” International Journal of Rock Mechanics and
10. H. Grob, “Schwelldruck im Belechtunnel,” in Berichte, Mining Sciences, Vol. 55, p. 139–150 (2012).
Internationales Symposium für Untertagebau, Luzern, 22. M. R. Vergara, K. Balthasar, and T. Triantafyllidis,
Switzerland, 1972 (1972), pp. 99–119. “Comparison of experimental results in a testing device
11. B. C. Hawlader, Y. N. Lee, and K. Y. Lo, “Three- for swelling rocks,” Int. J. Rock Mech. Min. Sci. Geo-
dimensional stress effects on time-dependent swelling mech. 66, 177–180 (2014).
behaviour of shaly rocks,” Can. Geotech. J. 40, 501– 23. W. Wittke and B. Pierau, “Fundamentals for the design
511 (2003). and construction of tunnels in swelling rock,” in Pro-
12. B. C. Hawlader, K. Y. Lo, and I. D. Moore, “Analysis ceedings of the 4th International Congress on Rock
of tunnels in shaly rock considering three-dimensional Mechanics, Montreux, Switzerland, 1979 (Balkema,
stress effects on swelling,” Can. Geotech. J. 42, 1–12 Rotterdam, 1979), Vol. 2, pp. 719–729.
(2005). 24. P. Wittke-Gattermann, “Verfahren zur Berechnung
13. A. Hefny, K. Y. Lo, and J. A. Huang, “Modelling of von Tunnels in quellfähigem Gebirge und Kalibrierung
long-term time dependent deformation and stress an einem Versuchsbauwerk,” in Geotechnik in For-
dependency of Queenston Shale,” Can. Tunnelling J., schung und Praxis (Glückauf, Essen, 1998), pp. 22–31.
115–146 (1996). 25. D. Zwillinger, Handbook of Differential Equations (Aca-
14. J. Huder and G. Amberg, “Quellung in Mergel, Opali- demic Press, Boston, 1997).
nuston und Anhydrit,” in Vol. 88 of Schweizerische
Bauzeitung, Jahrgang Heft 43 (1970), pp. 975–980. Reviewer: V.G. Trifonov

GEOTECTONICS Vol. 50 No. 3 2016

You might also like