Você está na página 1de 25

Rock Mech. Rock Engng.

(1993) 26 (4), 307--331 Rock Mechanics


and Rock Engineering
9 Springer-Verlag1993
Printed in Austria

A Model for Swelling Rock in Tunnelling


By
G. Anagnostou

Swiss Federal Institute of Technology, Zurich, Switzerland

Summary
In this paper, the phenomenon of swelling in tunnelling will be treated as a
hydraulic-mechanical coupled process. This approach allows one to model the
observed floor heaves realistically, i.e. without the prediction inevitable in the
previous models of movements at the tunnel crown and walls. Furthermore, the
development of heave and pressure over the course of time can be studied. The
absence of deformations above the floor level is here interpreted as a consequence
of the hydraulic boundary conditions. Besides the importance of seepage flow, the
influence of rock strength is illustrated. Swelling rock is considered as an elasto-
plastic material. This allows one to predict the often large heaves of a tunnel floor
as observed in situ. According to the numerical results, the area of practically
relevant swelling strains extends as far as the plastic zone.

I. Introduction

The phenomenon of the heave of a tunnel floor, or of damage to the invert


arch, has been well-known since the beginnings of railroad-tunnelling in
the middle of the 19th century (cf. e. g. Pressel and Kauffmann, 1860). This
phenomenon has been associated with the volume increase of the rock
mass due to water uptake. Swelling rock contains certain clay minerals, as
well as, in some cases, anhydrite. Geological formations where swelling-
related difficulties can occur are widespread, e. g. in Switzerland (jurassic
claystones and tertiary marlstones), in southern Germany (gypsum-
keuper), and in eastern France (jurassic claystones). A comprehensive
review of the empirical findings connected with this study is given else-
where (Anagnostou, 1991). It is necessary, however, to mention some
features of the observed phenomenon in order to distinguish swelling from
other phenomena in tunnelling, in particular squeezing.
In contrast to squeezing, the presence of water, as well as a specific
mineralogical composition is necessary for swelling. Since squeezing
(depending on rock strength and overburden) may take place in principle
308 G. Anagnostou

in any rock, i. e. in rocks containing swelling minerals as well, one cannot


exclude the possibility of simultaneous occurrence of swelling and
squeezing. It should be stressed, however, that simultaneous swelling and
squeezing is a rather rarely occuring phenomenon which takes place only
in weak rock. The typical pattern of deformations observed in tunnels in
swelling rock consists of -- sometimes very large -- floor heaves and an
absence of comparable deformations at the walls and the crown. Usually,
swelling rock shows a high strength during excavation. The walls and the
crown remain stable over many years.
It should be noted, that, in laboratory, one measures swelling pres-
sures which are often higher than the vertical stress corresponding to the
overburden in the original location of the specimen. Also, in situ observa-
tions clearly imply that swelling pressure, in contrast to other kinds of rock
pressure, can be higher than the primary vertical stress, especially when the
tunnel is shallow (Terzaghi, 1946).

Design and Analysis Methods


A discussion of the different constructive counter-measures proposed can
be found in Kovfiri et al. (1981, 1987, 1988). The development of floor
heave can be prevented either through installing a rigid invert arch, or
systematic anchoring. According to new methods applied in tunnelling
over the past 15 years, one arranges a deformable layer between the invert
arch and the excavation boundary. Thus, both floor heaves and swelling
pressures are allowed to occur, but in a controlled way, i.e. without
impairment to stability and serviceability.
The various constructive counter-measures require a proper estimation
of stresses and deformations. For example, when dimensioning a lining,
predictions of the stress resultants are necessary. These result from the
complex interaction of rock mass and lining, for the latter is partially
loaded by the rock (floor) and partially embedded in it (walls and crown).
Furthermore, the design of a deformable layer requires knowledge of the
characteristic line of swelling rock, i. e. the relationship between support
pressure and floor heave.
The first analytical approach to the swelling problem is found in a
publication by Wiesmann (1914), where the stability of tunnel mason-
ry-work in connection with the construction of the lower Hauenstein
tunnel in Switzerland is studied. Since then, various analysis-methods have
been proposed. The simplified methods of Einstein et al. (1972), Grob
(1972) and Kovfiri et al. (1987, 1988) are based upon a priori assumptions
concerning the stresses or strains along the vertical symmetry axis beneath
the tunnel floor. These methods represent simple design tools, and they are
not intended to explain the observed phenomena, or to predict the
complete deformation- and stress field. On the contrary, they presuppose
knowledge of the long-term stress- or strain-distribution. The problems that
arise from this prerequisite can be circumvented by treating the swelling
process on a continuum-mechanical basis (see Wittke and Rissler, 1976;
A Model for Swelling Rock in Tunnelling 309

Gysel, 1977, 1987; Fr6hlich, 1986). In such an approach, complete stress-


strain relations are formulated. These relations, together with the equi-
librium and the compatibility equations, represent a system of equations
which is solved for given boundary and initial conditions.

2. Outline of the Proposed Model

The degree of refinement of a model should depend on which questions


one intends to answer. Our procedure starts, therefore, with the simplest
model, and subsequently refines it with as few additional elements as
possible, but with enough to account for the essential effects.

2.1 Theoretical Framework

According to the simplest continuum-mechanical model, swelling is


considered as a pure stress-analysis problem. Stress-analysis proves,
however, to be insufficient for a realistic modelling of the observed
behaviour because it predicts swelling not only in the tunnel floor, but also
at the crown and walls (cf. Kovfiri et al., 1987, 1988). Note that in the
simple case of a deep, cylindrical tunnel in a homogeneous swelling rock
mass, no difference between floor and crown concerning geometry, initial-
and boundary conditions exists. It should be pointed out, furthermore, that
all existing continuum-mechanical models fail to explaha the absence of
deformations above the floor level on the basis of tunnel shape, even when
considering a tunnel with a plane floor.
An improved computational model for tunnels in swelling rock should
take the movement of water into consideration. In this case, the seepage
flow equations must be considered simultaneously to the equations of
stress analysis. In such a coupled hydraulic-mechanical model, the
displacement field depends on the hydraulic potential field and, conse-
quently, on the hydraulic boundary conditions as well. In general, these are
different for the floor (e. g. free water) and for the crown (e. g. an imper-
vious boundary). This asymmetry of the hydraulic boundary conditions
opens up for us the possibility of modelling floor heaves without any of the
movements at the walls and crown inevitable in previous models.
Furthermore, due to consideration of the seepage flow, the time-devel-
opment of swelling deformations can be modelled.
In the mathematical formulation, the pore water pressure p, the flux
vector q~ and the water content m appear as additional independent vari-
ables besides the displacement vector uk, the strain tensor ei~ and the stress
tensor o-~j. These variables are connected by (cf. Zienkiewicz and Taylor,
1989) the kinematic relations
1
e~j = 5- (~ uj + 8j u3, (1)
310 G. Anagnostou

where geometrical linearisation, i. e. infinitesimal strains, are assumed; the


equilibrium equations
gk crkj + bj = 0, (2)
where bj denotes the body force vector; the stress-strain relations, schemati-
cally given by
d a~j = f ( d ~u, au, dp, p ) , (3)
where time-independent material behaviour is assumed; the equation of
mass conservation for water
rn
- 3, q; ; (4)
P0
furthermore, the equation of motion of water, for example, Darcy's Law
q~ = - k c~ O, (5)
where k and @ denote the hydraulic conductivity and the hydraulic head,
respectively (@ = z + p / g p ) ; and, finally, an additional equation which
expresses the relationship between water content, strain components and
pore water pressure:
a m = f ( e , j , dp, cru, p ) . (6)
Note that (3) allows for the consistent modelling of elementary swelling
processes. According to this equation, the gradual development of swelling
strain under a constant state o f stress, as well as of swelling pressure in a
constrained specimen is caused by a gradual change of the pore water
pressure (cf. Terzaghi, 1925).

Swelling of Sulfatic Argillaceous Rocks

Besides clay-minerals, swelling rock also often contains anhydrite


(CaSO4). When anhydrite and water are within a closed system (i. e. a
system without mass exchange with its surroundings), anhydrite dissolves
in water, and gypsum begins to precipitate. The volume of gypsum is
about 61% higher than the volume of the original anhydrite. Therefore,
the swelling of sulphatic rock has often been attributed to the hydration
of anhydrite (the so-called "anhydrite theory"). In an open system (like
the one present in situ), however, a great variety of processes are
possible. Depending on the water circulation, and on the reaction
kinetics, either a leaching of anhydrite or a hydration may occur. In the
first case, solid matter is transported away by the water. In the second
case, the solid matter volume increases by about 60%. Provided that
porosity remains constant, hydration causes a volumetric strain, the
magnitude of which depends on the volume fraction of the original
anhydrite. It is also possible, however, that the formation of gypsum
causes a gradual filling up of the pores, and thus a sealing of the rock.
In this case, a lower, or even zero volumetric strain would occur.
A Model for Swelling Rock in Tunnelling 311

Consequently, the properties of the CaSO4-H20 system do not allow


for a definite statement concerning the contribution of anhydrite to
swelling. It should be mentioned that the common laboratory tests are not
suitable for investigating this problem, for conditions are different in situ
from those in the laboratory as concerns water circulation. Furthermore,
since these conditions possibly differ even from tunnel to tunnel, it is ques-
tionable whether a generally valid mechanism exists at all.
A purely hydraulic-mechanical model, as the one proposed in this
paper, is inadequate if swelling is caused by a hydration of anhydrite. A
consideration of the anhydrite-water-gypsum system requires a coupled
hydraulic-mechanical-chemical model, i.e. a substantial enlargement of
the theoretical framework of this paper. In applying the model proposed
here also for tunnelling in sulphatic argillaceous rocks, one does not,
however, contradict the facts as they stand: Empirical evidence for the
validity of the anhydrite theory is not available; only a few in situ investi-
gations exist, and their results are either not clearly interpretable, or they
contradict the anhydrite theory.

2.2 On the Stress-Strain Relations

Biot's (1941) poroelasticity theory provides the simplest model for a


hydraulic-mechanical process. Numerical simulations show, however, that
the modelling of larger floor heaves is not possible on the basis of elasticity
theory. Neglecting the limits to rock strength leads to inadmissible stress
fields, as well as to a severe underestimation of swelling deformation. Due
to the low drained strength of argillaceous rocks, strength can be achieved
over the course of swelling, and plastic deformations will occur (cf.
Lombardi, 1984; Bellwald and Einstein, 1987). The development of plastic
strains during swelling can also be elucidated by means of simple
oedometer tests. As several authors have noted, when a swollen specimen
is re-loaded into an oedometer, it becomes apparent that the axial strain is
partially irreversible (cf. e. g. Holtz and Gibbs, 1956).
For these reasons, swelling rock is modelled here -- as in Bellwald
and Einstein (1987) -- as an elasto-plastic material. The simplest possible
model is the elastic perfectly-plastic material with a Mohr-Coulomb yield
criterion. This model is widely used in engineering and contains easily
interpretable parameters. It should be noted, however, that even this simple
material requires the determination of no less than seven material
constants, when strength anisotropy, non-associated flow rule, and tensile
strength are taken into account (see section 3.2).
In assuming the standard, linearly elastic, perfectly plastic model,
however, one cannot describe some important features of actual behaviour.
Laboratory tests show that the swelling strain perpendicular to the bedding
plane is considerably higher than the strain parallel to it (see e. g. Fr6hlich,
1986). Oedometer tests show, furthermore, that the steady-state axial
strains of non-sulphatic argillaceous rocks are proportional to the
respective stress logarithms (see e. g. Terzaghi, 1925). From an engineering
312 G. Anagnostou

point of view this relationship is interesting, since it suggests that a


significant reduction of the necessary support pressure can be achieved by
allowing a small floor heave (Kov~tri et al., 1987, 1988). Even though
empirical findings concerning the steady-state relation between floor heave
and support pressure are not yet available, in situ tests show that support
pressure reduces the rate of floor heave considerably (Kovari et al., 1987,
1988; Fecker and Wullschl~iger, 1991).
The stress-strain relations of the common elasto-plastic material are
here therefore extended to include the swelling anisotropy and the loga-
rithmic relationship between strain and stress. Due to the reversible char-
acter of the microscopic mechanisms underlying swelling (Madsen, 1979),
an additional anisotropic and non-linear term will be introduced in the
elastic part of the constitutive equations (see section 3.1).

3. Constitutive Relations

3.1 Elastic Stress-Strain Relationships

To take the example of a rock element, in an arbitrary reference configu-


ration, the stress state is given by o-u0 and the pore water pressure is given
by P0. We require equations which will give the strains and the change of
water content caused by a transition from {0.u0,P0} to another state
{ 0.u0, P}. Within the elastic domain, this transition is reversible. An arbi-
trary path leading from { 0.uo, P0 } to { 0.u, P } can, therefore, be chosen. For
reasons of simplicity, we choose a way from { 0.u0, P0 } to { ~j, p }, which
consists of different segments, so that, along each segment, either the stress
state changes, or the pore water pressure changes. Subsequently, we will
describe the path chosen, as well as the assumptions concerning material
behaviour along different segments of this path. Lastly, the final equations
will be given, together with a short discussion on the relationship of these
equations to established models. For the detailed derivation, the reader is
referred to Anagnostou (1991).
Figure 1 shows the projection of the chosen path onto the { 0.1a,p } ,
and { o12, p },-plane (OABC). Along OA, the pore water pressure changes
from P0 to atmospheric pressure, and the normal stresses change by the
same amount. Along AB, the stresses change by a,v - au0 + 6 u (p - P0),
whereas the pore water pressure remains equal to atmospheric pressure.

O~x L%1 01 ib~12


A B " A B
Fig. 1. Projection of the path into the (O-ll, p)- and (0-12,p)-plane
A Model for Swelling Rock in Tunnelling 313

Consequently, AB represents a common drained state transition. The last


segment (BC) is similar to OA; one changes the pore water pressure from
atmospheric to its final value p, while imposing an equal change on the
normal stresses.
The volume change of the solid matter, as well as the strains that occur
in the course of a state transition like OA or BC, are assumed to be propor-
tional to the change of pressure. Due to an alteration of state under atmos-
pheric pore water pressure (i. e. along the segment AB), the volume of the
solid matter is assumed to change proportionally to the stress change. The
strains that occur along the segment AB are considered as consisting of two
parts. One part depends linearly on the stress change following generalized
Hooke's law. The other one is attributed to swelling, and it depends non-
linearly on the stresses. The exact form of this non-linear stress-strain
relation is obtained by means of an energy-based analogy between the
macroscopically observed swelling and the underlying, microscopic mecha-
nisms, i. e. the interactions of the clay platelets. In this way, the following
equations result, giving the strains and change of water content as func-
tions of stresses and pore water pressure:

eq -- 2 # A a q - S i - f - ~ Ala~k -- B f l q l n flq%so' (7)

m A ak'k BIn flq ~5


cq (ais - ~ o ) + A * (p - Po); (8)
Po
- -

3 K flij O'ijO
t

where

e,} = gq + <s (P - P0), o-~ = o-q + 6q p, (9)

I -fl 2#
flq - 3 6q + / 3 ni n;, K =/t. + --5-- (10)

In these equations, 2 and # are Lam6's constants, B and fl are two swelling-
related parameters, and A* and cq represent seven independent material
constants, which are related to solid grain compressibility, to water
compressibility, and to rock porosity. In the case of incompressible solid
grains, the constants cq are equal to zero (Anagnostou, 1991). ni denotes the
unit vector directed perpendicular to the bedding plane. The incremental
stress-strain relations are obtained from (7):

d crqt t = Dqk
EL
z d ekl ; (11)
where
d ~ jt t = d crqt - Dqk
EL
l c~l dp, (12)

B ( 2 6 q + 2/-tflil) (26kl + 2/.z&,)


D~5L, = s 6k, + 2 #6i~ 6j, - (13)
314 G. Anagnostou

In the special case where B = 0 (i. e. without the swelling term) and cU
= 6ij c (e = a material constant), Eqs. (7) and (8) describe an isotropic,
linearly elastic material. One can easily verify that, in this case, the
proposed equations are identical to the well-known equations of Blot
(1941), and a/y to Zienkiewicz's (1985) effective stress tensor.
According to (12), the tensor a~}' is identical with the c o m m o n
effective stress tensor %~ [defined by (9)], only in the ideal case of incom-
pressible solid grains (i. e. c~j = 0). For a given compressibility of solid
grains, the higher the overall stiffness of the rock, the bigger the deviation
between cry} and a~}'. Parametric studies have revealed that the influence of
solid grain compressibility is quantitatively negligible as concerns the
swelling phenomenon. Neglecting the compressibility of the solid grains
leads, furthermore, to a substantial simplification of the constitutive equa-
tions; the material obeys then the principle of effective stress which is
commonly used in geotechnical engineering. Accordingly, the use of the
following simplified equations is suggested:

{ ~ t} ~ijO'iS" (14)
g'J - 2lit Acri~ - 5U-f~ Aaik -- B fli:ln fl~j~ruo, ,

m
- eke + A (p - P0)- (15)
P0
The constant A in (15) represents the volume of water that one can inject
into the rock by increasing the water pressure by one unit, while keeping
the rock volume constant. A is, therefore, identical with the c o m m o n
specific storage coefficient (see Marsily, 1986). When an alteration in state
takes place quickly (e. g. during the excavation of a tunnel), water content
remains constant, whereas pore water pressure changes. Stress-strain rela-
tions for such undrained processes, as well as the induced excess pore
water pressure, can be obtained from (14) and (15) by setting rn = 0.
Besides the c o m m o n constants of poroelasticity (~, #, A), only two
additional constants appear in the proposed equations: B relates to the
amount of swelling strain, and fl to the swelling anisotropy. Subsequently,
we show some properties of the non-linear term in (14), and discuss the
experimental determination of the swelling parameter B and the swelling
anisotropy factor ft.

The Non-linear Term


Provided that the normal stresses acting parallel to the bedding plane are
equal, the following equations give the strains resulting from the non-linear
term in (14):
e(q) = lan[4"~kke(q),F-'(q)= tip ~k~(q), (16)

+
(q) -- B In , (17)
A Model for Swelling Rock in Tunnelling 315

where ft, = (1 + 2 fl)/3 and tip = (1 - fl)/3 ; the subscripts n and p denote
the direction perpendicular and parallel to the bedding, respectively. Equa-
tions (16) and (17) are graphically represented in Fig. 2. For a given stress
change, the strain components are proportional to the swelling parameter
B ; the higher the value of this constant, the greater the contribution of the
non-linear term to strain. The ratio ~(pq)//~(q) is governed by the anisotropy
factor ft. Furthermore, the volumetric strain tr, k k relates logarithmically to a
linear combination of the normal stress parallel and perpendicular to the
bedding plane (fin or" + 2 tip or/). In the borderline case of isotropic
material (fl = 0), this linear combination is equal to the mean stress (O-k'k/3),
and the volumetric strain relates logarithmically to the first stress invariant.
In the other borderline case of markedly anisotropic behaviour (fl --- 1), the
non-linear term furnishes a strain only perpendicular to the bedding plane.
This strain relates logarithmically to the normal stress perpendicular to the
bedding plane.

'.(q)

~ rctan(B)
~rctan(.~n B)
Atlanta m In /
~ ~n(Yno+2[3pO'po/
1
Fig. 2. The non-linear term of the elastic equations

Experimental Determination of B and fl

The constants B and fl can be determined by means of a single test,


consisting of two phases. In the first phase, one prevents the development
of any strain, and measures the swelling pressure which develops over the
course of time. When the steady state of phase 1 is reached, the specimen is
unloaded slowly and isotropically. Such a test can be performed in a true
triaxial apparatus or in a triaxial apparatus with controlled lateral strain.
At the beginning of the test, stress state and pore water pressure are
given by ~j0 = 0 and P0 > 0, respectively. In the course of phase 1, the pore
water pressure increases from P0 to atmospheric pressure (Pl = 0). Since
strain development is prevented, the effective stresses remain constant (14)
and consequently the total stresses experience the same change as the pore
water pressure. At the end of phase 1, the stress state is given by ~v1
= 8,7p0. Accordingly, the steady-state swelling pressure under volume
constancy is equal to the negative pore water pressure at the beginning of
the test. Note that, in contrast to existing models, swelling pressure does
not represent here a material constant, but is related to Poo i. e. to an initial
316 G. Anagnostou

condition. Furthermore, according to the proposed model, the same


swelling pressure develops parallel to the bedding plane as well as perpen-
dicular thereto. This contradicts some laboratory results (e. g. Froehlich,
1986). An analysis of swelling pressure tests has revealed, however, that the
observed anisotropy of swelling pressure can be explained by the different
stiffness of the oedometer apparatus in the radial and axial direction (see
Anagnostou, 1991).
In phase 2, drained isotropic unloading takes place down to a small
hydrostatic pressure O-rain(e. g. 0.01 MPa). During this phase, stress state is
given by aij = 8ij a, where cr represents a scalar with the dimension of a
stress (O-min> a--> P0). According to the assumed Mohr-Coulomb yield
condition, yielding cannot occur during phase 2, for the stress state is
continuously hydrostatic. The elastic stress-strain relations (14) can,
therefore, be used. Furthermore, the contribution of the linear term to the
volumetric strain is small compared to the contribution of the non-linear
term. By neglecting the linear term in (14), one obtains the following
approximate expressions for the strains that occur in the course of phase 2:

e~ -= - B 1 + 2fl in a _ B_l~_fl i n c r (18)


P o ' ep = Po'
O"
ek~ = - B In - - , (19)
P0

According to (19), the swelling parameter B is given by the slope of the ekk
-a-line in a diagram with logarithmic stress axis (Fig. 2). On account of
(19), the swelling parameter B can be expressed as a function of P0 (which
is measured in phase 1) and of ekk. m,x (which represents the volumetric
strain at the end of phase 2, i. e. at cr = amin):
j ~ __ ~ k k, m a x

The following relation can be derived from (18):

I_Ep
e~,
fl - . (21)
1+2E
Accordingly, the swelling anisotropy factor depends only on the ratio of
the strains parallel and perpendicular to the bedding plane. One can easily
verify that (21) also holds for the strains which occur in a free swelling test.
Consequently, the anisotropy factor can also be determined by means of
such a simple test. Equations (20) and (21) are graphically represented in
Fig. 3 a and 3 b, respectively.
A Model for Swelling Rock in Tunnelling 317

dvk (MPa)=
0,10 0.2 1,0

0,08 0.5 0,8


1.0
0,06 0,6
2.0
B 5.0
0,04 0,4

0,02 0,2

0,00 0,0 . , . , 9 , .

0,0 0,1 0,2 0,3 0,0 0,2 0,4 0,6 0,8 1,0
~kk at O'rnin=-0.01 MPa Ep/C. n

a b
Fig. 3. Graphs for the determination of the material constants B and fl

3.2 Elasto-Plastic Stress-Strain Relationships


For stress states on the yield surface and loading, the c o m m o n incremental
elasto-plastic relations hold (cf. e. g. Zienkiewicz and Taylor, 1989):

d a,.5. = DiPL, d ek,; (22)


where
eL c~g 3 f Dfq~,
D~s~, 3a,',n 3a;q (23)
3 f DfqL,.s 3g

and Di~z is given by (13). The symbols f and g denote the yield function
and the plastic potential function, respectively. Assuming a Mohr-
Coulomb yield criterion, a different strength in the bedding plane and a
limited tensile strength, five material constants are necessary to specify the
yield surface: The uniaxial tensile strength ~ the angle of friction ~, the
cohesion c, and the corresponding values ~a and ca in the bedding plane.
Furthermore, a non-associated flow rule is assumed. As an additional
parameter we have the angle of dilatancy ~ (~a in the bedding plane). If
~ = 0, plastic dilatancy does not occur, whereas the normality rule holds if

The yield function is not continuously differentiable at points on the


edges of the yield surface. At these points, we proceed according to the
method of Koiter (1953). It should be mentioned, that the integration of the
elasto-plastic stress-strain relations (22) is not trivial, due to the stress-
dependence of Di~1 [see Eq. (13)], as well as due to the singularities of the
yield surface. This problem can be solved by approximating the yield
318 G. Anagnostou

surface with an arbitrary number of hyperplanes. This linearization allows


for a semianalytical, numerically stable integration of the constitutive equa-
tions, which can be made as accurate as desired.

3.3 Seepage Flow Equations


Intact swelling rock shows a low hydraulic conductivity, due to its high
clay content and low porosity. For example, Vardar and Fecker (1984) have
measured conductivity values less than 10 -12 m / s e c in gypsum-keuper
specimens. Furthermore, intact swelling rock is assumed to be 100% satu-
rated, as air enters into the pores of a saturated porous medium when the
pore water pressure becomes lower than a material-dependent negative
value. The finer the pores are, the lower this value will be. A number of
laboratory investigations have shown that swelling soil or rock remains
practically saturated up to very high negative pore water pressures (see
e. g. Chenevert, 1969).
In a rock mass, however, various discontinuities may exist -- natural
joint sets or randomly oriented fissures which either pre-date the tunnel
excavation or are induced by it (e. g. due to blasting). The discontinuities
contribute to the rock mass permeability, provided that they are water-satu-
rated. When they are unsaturated, however, the seepage flow takes place
through the low-permeable matrix. This feature has some interesting conse-
quences. For example, consider the case of a lined tunnel, and assume that,
due to circulation in the longitudinal direction, water under atmospheric
pressure is available within the interface between lining and rock. In this
case, infiltration and water uptake in rock would occur far more quickly
beneath the floor than above the crown, for discontinuities beneath the
floor will fill up gradually with water, and thus contribute increasingly to
the rock mass permeability.
Clearly, it is impracticable to numerically model all individual fissures
separately, as well as the matrix between them. We therefore retain the
model of a porous continuum for the rock mass. The simplest model is
assumed, i. e. Darcy's law (5). An additional equation is necessary to take
into account the conductivity reduction imposed by fissure-desaturation.
Here, the simplified relation of Desai and Li (1983) will be assumed
(Fig. 4): At positive water pressures, the rock mass is 100% saturated, and
its conductivity shows the maximum value kmax. This value is furnished

!~kkrnax
rain
Prnin0 I saturated
discontinuities ~
Fig. 4. Dependence of conductivity on pore water pressure
A Model for Swelling Rock in Tunnelling 319

mainly by the discontinuities and it corresponds to the conductivity, which


is measured in situ, e. g. by a packer test. At negative pore water pressures,
the contribution of joints and fractures to rock mass conductivity decreases
due to air-entry. For the sake of simplicity, the non-linear and hysteretic
behaviour (Marsily, 1986) is ignored here. At pore water pressures lower
than Pmin, the saturation degree of the discontinuities is so low that their
contribution to rock mass permeability vanishes; the conductivity shows
then its minimum value kmin. Obviously, kmi n cannot be higher than the
permeability of intact rock (e. g. as measured in the laboratory on rock
specimens), for the seepage flow will be interrupted by the partially satu-
rated discontinuities.

Discussion
In general, rock joint permeability is not constant, but depends on joint
aperture. Assuming a cubic relationship between conductivity and aperture
(cf. e.g. Marsily, 1986), a joint closure by 50% causes a conductivity
reduction by about an order of ten. It is therefore still possible that rock
mass permeability significantly decreases in the course of swelling, due to
the gradual closing of joints and fissures. On the other hand, swelling
causes an increase in porosity, and thus an increase in rock matrix permea-
bility. The validity of the empirical formula of Kozeny-Carman presup-
posed, an approximately cubic relationship between conductivity and volu-
metric strain can be obtained. Depending on the initial porosity, a volu-
metric strain of 10% causes a conductivity increase by a factor of between
10 and 100. Consequently, conductivity changes by orders of ten are thor-
oughly possible, merely as a consequence of volumetric strains as small as
a few percent, or to the closing of existing joints by fractions of a ram.
Since the assumed constitutive relations do not account for such changes
in rock permeability, predictions of swelling duration are extremely
uncertain.
The modelling of a fissured rock mass as a porous continuum with an
"equivalent" conductivity is problematic when considering a transient
process, for hydraulic head within the matrix is, in general, different from
the head within neighbouring discontinuities (see Marsily, 1986). Due to
these local head disequilibria, Barenblatt et al. (1960) introduced the
"double-porosity" concept. Barenblatt et al.'s seepage flow equations
express very clearly the decisive role of the matrix with regard to stora-
tivity, as well as the decisive role of the discontinuities with regard to per-
meability. Due to this considerable conceptual strength, the modelling of
swelling rock as a double porosity medium possibly represents a worth-
while pointer for future development.
320 G. Anagnostou

4. Governing System of Equations

The equations presented in sections 2 and 3 constitute a system of partial


differential equations for the unknown fields eij (xk, t), aq (xk, t), m (xk, t),
ui (x~, t), qi (x~, t) and p (xk, t), or @(xk, t). Through eliminating stresses,
strains, water content and fluxes, one obtains the following system of four
differential equations for the three displacement fields and the hydraulic
head:
~ Oijkm 6~k tim - - Po g 3j C~ + bj = 0 (j = 1, 2, 3) (24)

3k fik + pog A C]) - 3i k 3~ ~ = 0. (25)

Equation (24) results from the equilibrium condition, whereas Eq. (25)
expresses the mass conservation for water. For problems with arbitrary
geometry, initial conditions and boundary conditions, the above system of
equations can be numerically solved by the finite element method (Zien-
kiewicz and Taylor, 1989). The corresponding matrix equations are:

du df { d~O}
K dt - dt + L--~- . (26)

d 6]) ~LT dU ~
H cI)+ S T ~ q- [ d tJ" (27)

The terms which are outside the brackets in (26) and (27) represent the
matrix equation of stress analysis and of transient seepage flow analysis,
respectively. The terms inside the brackets express the
hydraulic-mechanical coupling. To numerically solve the differential equa-
tions, the constitutive relations were implemented in HYDMEC, a
computer-code developed by the author.

5. Modelling of Swelling in Situ

Subsequently, we study the problem of a deep, cylindrical, unlined tunnel


in a homogeneous swelling rock mass (Fig. 5). Only some important points
will be discussed in this paper. For comprehensive parametric studies refer
to Anagnostou (1991, 1992). The computations presented here have been
carried out on the Cray-YMP of the Swiss Federal Institute of Technology,
Zurich.
The system used for the numerical solution consists of a thick-walled
ring (hatched area in Fig. 5). Assuming a horizontally layered rock mass
and an initial stress field with horizontally and vertically oriented principal
axes, the system is symmetric concerning the z-axis. Only half of the ring
has, therefore, to be discretized by finite elements.
A Model for Swelling Rock in Tunnelling 321

~3m

;ystern

Fig, 5. Geometric lay-out of deep cylindrical tunnel (MPa)

5.1 Material
The assumed material constants are summarized in Table 1; the idealized
material shows a marked swelling- and strength-anisotropy, i. e. swelling
occurs only perpendicular to the bedding plane (fl = 1), and yielding takes
place solely in the bedding plane. The material does not exhibit plastic
dilatancy, i.e. volumetric strains are caused solely by swelling. The
swelling rock is treated here as a no-tension material. However, these
parameter assumptions do not qualitatively affect the results presented in
this paper.

T a b l e 1. Assumed parameters

Swelling parameter B variable


Swelling anisotropy factor fl 1.oo
Swelling pressure G~ 4 MPa
Lam6's constant Z 3000 MPa
Lam6's constant # 1500 MPa
Tensile strength f 0 MPa
Yielding (bedding plane):
Cohesion ca variable
Angle of friction G variable
Angle of dilatancy r 0~
Storativity factor A 4.10-6 MPa
Conductivity constants: kmin/kmax variable
Pmin - 10-3 MPa
-

Unit weight ?" 25 kN/m 3

The constant Pmin indicates how rapidly the hydraulic conductivity


decreases due to air entry (Fig. 4). Here, we assume that the contribution of
fissures vanishes abruptly at p = 0. This represents the borderline case for
unsaturated flow. Nevertheless, due to numerical considerations, a non-
zero, but very small value for Pmi, has been chosen (see Table 1). The
assumed value of the specific storage coefficient A corresponds to a water
compressibility of 2500 -1 MPa -1 and a porosity of 1%. The influence of A
322 G. Anagnostou

is not significant, because the time delay is caused primarily by the strong
deformability of the swelling rock.

5.2 I n i t i a l C o n d i t i o n s

The initial vertical stress increases linearly with depth according to a total
unit weight of 25 k N / m L The initial state of stress is assumed to be hydro-
static. Numerical simulations revealed that the coefficient of lateral stress
is not important when the bedding is horizontal and the material markedly
anisotropic (fl = 1; yielding only in the bedding plane). For other material
behaviour, the initial lateral stress has a quantitative influence on the
predicted deformation field. The results presented in this paper retain,
however, their validity in a qualitative manner for these cases as well.

Initial Pore Water Stress in Situ


The initial pore water pressures can be computed, e.g. based upon the
position of a known water table. The natural water content of jurassic or
tertiary swelling rock is, however, very low. Due to their low porosity and
low permeability, these argillaceous rocks seem to be totally dry during the
excavation period. Usually, one cannot discover a water table, which could
serve as a clue in formulating the required initial conditions. Subsequently,
we show how an assumption concerning the initial pore water pressures
can be made in a consistent manner.
According to the present model, the swelling pressure that develops in
an oedometer (under volume constancy) is equal to the pore water pressure
at the beginning of the test (see section 3.1). This pressure is related, but
not identical to the in situ initial pore water pressure as the sampling
causes stress state changes. Through computing the sampling-induced
change of pore water pressure, one obtains a relationship between swelling
pressure and in situ initial pore water pressure.
The sampling-induced change in pore water pressure can be computed
based upon (14) and (15), if we assume that sampling is an undrained
reversible process, and that the initial vertical stress is a principal stress. In
this case, the following non-linear equation expresses the relationship
between initial vertical stress av0, coefficient of lateral stress s initial in
situ pore water pressure P0 and swelling pressure Gk:
1+2s
-r 3 + (1 + g a ) ( ~ - po)

- BKln o-~k = 0. (28)


1 + 22,s + 2 f l ( 1 - •)
P 0 + Go 3

In the special case with a hydrostatic initial stress field (s -- 1) and zero water
compressibility (A = 0), the following simple relation results from (28):

a ~ = a~o + Po = a/o. (29)


A Model for Swelling Rock in Tunnelling 323

Accordingly, the sampling causes a decrease in the pore water pressure by


an amount which is equal to the initial stress. Furthermore, the swelling
pressure developing in the oedometer is equal to the initial effective
vertical stress cr]0. Consequently, one obtains the initial in situ pore water
pressure by subtracting initial vertical stress from swelling pressure. In
general, the initial stress field is, however, not hydrostatic and the water
shows a low compressibility. Thus, P0 should be computed by using (28).
We performed a sensitivity study, where all constants appearing in (28)
were varied within the limits relevant to practical applications. This study
revealed that the ratio of swelling pressure to initial vertical effective stress
varies between 0.70 and 1.30. For practical purposes, therefore, simplified
Eq. (29) can be considered as an accurate enough approximation.
In the examples presented in this paper, a maximum swelling pressure
of 4 MPa is assumed. This pressure corresponds to the laboratory values
for strongly swelling rock. At the level of the tunnel axis, the total vertical
stress amounts to 2.5 MPa. From (29), one thus obtains a value of - 1 . 5
MPa for the initial pore water pressure, respectively of - 1 5 0 m for the
initial hydraulic potential at the origin of the co-ordinate system. For the
sake of simplicity, a uniform initial potential field is assumed here~

5.3 Boundary Conditions


Since the tunnel is unlined, the tractions along the inner boundary of the
system are equal to zero. The outer boundary is assumed to be fixed. Alter-
natively, tractions corresponding to the initial stress field could be
prescribed there. Due to the chosen size of the ring, the influence of this
boundary condition is, however, not significant.
The hydraulic boundary conditions have a decisive influence upon the
predicted deformations. As we will see, they influence the deformation
field in a qualitative manner, i. e. different hydraulic boundary conditions
lead to qualitatively different deformations of the excavation boundary. In
spite of their significance, assumptions concerning the hydraulic boundary
conditions are, however, generally less influential than assumptions about,
e. g. the mechanical boundary conditions. It should be noted, furthermore,
that the hydraulic boundary conditions do not necessarily remain constant
during the lifetime of a tunnel. For example, it is possible that water is
present on the tunnel floor only occasionally, or that a water level shows
seasonal fluctuations. Due to the uncertainties bound to the choice of
hydraulic boundary conditions and due to the anticipated importance
these conditions have, we performed numerous computer simulations with
different modelling assumptions. In the present paper, we limit our
discussion to certain important cases.

Outer Boundary
Concerning the outer boundary of the system, two kinds of boundary
conditions are possible. According to a first assumption, the excavation
324 G. Anagnostou

causes a disturbance of the potential field, which vanishes in a greater


distance from the tunnel. In this case, the hydraulic potential along the
outer boundary is fixed and equal to the initial potential. According to the
other model, the undisturbed rock mass, i. e. the rock mass far away from
the tunnel, is considered to be practically impermeable, whereas a limited
zone surrounding the opening displays a higher permeability due to the
excavation-induced generation of fissures, or due to the opening of pre-
existing fissures. In this case, a no-flow boundary condition holds at the
outer boundary.
A numerical study revealed that the effect of the far field boundary
condition on the predicted deformations is not significant compared to the
influence of the hydraulic conditions at the excavation boundary.
Furthermore, as expected, this effect decreases with increasing outer radius
of the system. Except for the case with an impervious excavation
boundary, the influence of the far field boundary conditions is of limited
importance.
The numerical results presented in this paper have been obtained by
assuming an impervious outer boundary. However, they are, at least in a
qualitative manner, valid as well for the other possible boundary condi-
tions.

Excavation Boundary

Often, swelling rocks seem to be completely dry during excavation, due to


their very low natural water content. Therefore, it is justifiable to ask after
the origin of the water causing swelling. An evaluation of numerous cases
reveals that, mostly, this water originates either from adjacent aquifers, or
from the tunnel portals (for a review see Anagnostou, 1991). It reaches the
initially dry tunnel sections either due to accidents (e. g. leakage from
drainage pipes, unsealed boreholes etc.), or through seeping in the longi-
tudinal direction due to the tunnel gradient. This seepage takes place either
on the tunnel floor, or within the cleft between the invert arch and the rock,
or within a narrow loosened rock mass beneath the floor. In all these cases,
water infiltrates into the rock mass practically only from the tunnel floor,
whereas walls and crown appear to be dry. The existing observations
clearly suggest that the boundary conditions at the lower and at the upper
part of the excavation boundary are different.
The availability of free water on the tunnel floor can be taken into
account by the boundary condition O = z, i. e. the pore water pressure at
the boundary is atmospheric. Concerning the appropriate hydraulic
boundary condition at the crown and walls, two models, corresponding to
two different kinds of boundary conditions, are possible.
Model 1 takes into account that the rock is in contact with humid air
at the excavation boundary. In this case, the boundary condition is, in
general, non-linear and of a mixed type. Following the numerical study of
Anagnostou (1992), the use of a simplified condition is suggested in which
a suction is prescribed with a value which corresponds to the relative air
A Model for Swelling Rock in Tunnelling 325

humidity h~ The relationship between suction and relative humidity is


expressed by the following equation (Chenevert, 1969):
RT
p = - - In hr. (30)
v
where R --- the gas constant (0.083 1 a t m / m o l ~ T = the absolute
temperature in ~ and v = the molar volume of water (0.018 I/too0.
In Model 2, the tunnel walls and crown are treated as impervious
boundaries (Q = 0).
5.4 Numerical Results
In the first computational step, the tunnel excavation is simulated as an
undrained process. In the second step, the steady state is computed. Alter-
natively, a transient analysis can be carried out, i. e. (26) and (27) are inte-
grated iteratively in time.

Tunnel Excavation
As a consequence of the tunnel excavation, the water content remains
constant, but the pore water pressures change around the opening. Due to
the constraint of the volumetric strains, the predicted excavation-induced
displacements are very small. Figure 6 shows the contour lines of the
computed excess pore-water pressures.

-0,20

0,00 -0,40

-I ,00

+(

.1,00

-0,40
-0,20
Fig. 6. Excavation-induced excess pore-water pressure (MPa)

One recognizes that the pore-water pressures decrease beneath the


floor and above the crown, whereas an increase/ in the pressures is
326 G. Anagnostou

predicted in the area beside the opening. Because of the swelling an-
isotropy, the volumetric strain mainly depends on the effective stress
component perpendicular to the bedding plane, i.e. on the vertical
effective stress. Since volumetric strains do not occur in an undrained
process, the vertical effective stresses remain constant. The total vertical
stresses decrease, however, beneath the floor and above the crown due to
the excavation. Since the effective vertical stress has to remain constant,
the pore-water pressure must decrease. The increase in pore-water pres-
sures beside the tunnel can be explained in a similar way.

Model I
In Model 1, the tunnel floor is covered by water under atmospheric
pressure, but the crown and walls are in contact with air (see section 5.3).
The assumed relative humidity amounts to 90%.
Subsequently, only steady-state results are discussed. The absolute
value of hydraulic conductivity is, therefore, not relevant. The relation
between conductivity and pore water pressure is, nevertheless, important
due to the inhomogeneity of the hydraulic potential field. The constant Pm~n
being fixed (section 5.1), this inter-relationship is governed by the ratio
krnin/kmax (Fig. 4). The choice of a value for the ratio kmin/kraax represents a
difficult task. However, the numerical results are not sensitive to the value
of kmin/kmax, provided that this value is less than 10 .2 (see Anagnostou,
1991, 1992). Values of kmin/km,x in this range are thoroughly realistic due to
the very low permeability of intact swelling rock, and due to the dominant
contribution of discontinuities to rock mass permeability.
Figure 7 shows the predicted deformations of the excavation
boundary (two-fold magnification), the distribution of vertical strain, the
plastic zone and the change of water content. The computed heave of the
floor amounts to about 0.45 m, whereas the crown moves less than one cm.
The high negative pore-water pressures over there increase the effective
stresses and thus prevent swelling. Furthermore, significant strains are
predicted up to a depth of about one diameter beneath the floor. It is inter-
esting to note that the area with large deformations extends as far as the
plastic zone. As expected, the change of water content is highest at the
tunnel axis and decreases with increasing distance from the tunnel.
Figure 8 shows the influence of swelling intensity (represented by the
constant B) and of rock strength (represented by the cohesion ca) on the
floor heave. An additional axis gives the values of the maximum swelling
strain (i. e. the strain in the free swelling test), which correspond to the
values of the swelling parameter B. In the case of an elasto-plastic material
without swelling (B = 0), very low heaves of less than one cm result. By
considering the swelling rock as an elastic (but no-tension) material, we
obtain the heaves given by the deepest curve in Fig. 8. Even for a high
swelling parameter of 0.05, corresponding to a free swell of 30%, the
predicted heaves amount to a few cm. Neither swelling alone (curve with ca
= ~ ) , nor plasticity alone (value for B = 0) explains floor heaves of
A Model for Swelling Rock in Tunnelling 327

several dm often observed in situ. It is the combined effect of swelling and


limited rock strength which accounts for the large heaves of the tunnel
floor.
We obtained these results by assuming a relative humidity of 90% at
the upper part of the excavation boundary. The numerical results are,
however, not sensitive to variation in relative humidity in the range of up to
about 97% (see Anagnostou, 1991, 1992). This interesting feature is a
consequence of the non-linearity of the seepage flow equations. At a

0,5 ~-.~ / = 0.1 MPa


g 0,4

32 0,2

0,1
13_ ~ ea=~ _

0,0 , v~v~..,,.........~.
. . , . , . , .
Vertical Change of 0,00 0,01 0,02 0,03 0,04 0,05
Strain ez Water Content Swelling Parameter B
5 10 15%
" -- -- >10%

5- 10% o ~o% 20~176 30O/o


............................5% Max, Swelling Strain
Fig. 7 Fig. 8

F i g . 7. D e f o r m e d b o u n d a r y , v e r t i c a l s t r a i n e~, p l a s t i c z o n e a n d c h a n g e o f w a t e r c o n t e n t
( B = 0 , 0 3 ; Ca = 2 0 ~ ; ca = 0 . 1 0 M P a ; kmin/kmax< 10 - 2 )
F i g . 8. D e p e n d e n c e of floor heave on rock cohesion and on swelling parameter (,~20~
/c~in//c~"x< 10-2)

relative humidity of 100%, the air suction is equal to zero (30); concerning
the boundary conditions, a difference does not exist between floor and
crown. In this case, almost equal deformations will be, therefore, predicted
at floor and crown.

Model 2
In contrast to Model 1, a no-flow boundary condition is prescribed at the
tunnel crown and walls. In this case, the steady-state hydraulic potential
field is nearly uniform with a value corresponding to a water table at about
the floor level. Therefore, the effective stresses above the crown do not
differ significantly from those beneath the floor, and consequently not only
floor heaves, but also crown settlements are predicted.
328 G. Anagnostou

A transient analysis shows, however, that there is a time lag between


swelling at the crown and swelling at the floor (Fig. 9). Besides the time
itself, the conductivity k is the only parameter containing the dimension
"time". The displacements can, therefore, be represented as functions of
the product kt. The results of Fig. 9 hold for a constant hydraulic conduc-
tivity (k = kmin = kmax), i. e. unsaturated seepage flow has not been taken
into account. Consequently, the time lag in Fig. 9 is due solely to the
greater distance which water must traverse to reach the crown area.

_ _F~or(t~=J. . . . . . . . . . . . . .
0,30' -
E _ C:rown~=~)_ . . . . . . . . . . .
v

0,20
o

0,10

0'0010-s 10-s 10.4 10"a 10.2 10.1 10 o 101

t k (m)
F i g . 9. T i m e - d e v e l o p m e n t o f f l o o r h e a v e a n d c r o w n s e t t l e m e n t ( B = 0.03 ; q5a = 3 0 ~ ; G = 0 . 1 0
MPa)

Parametric studies with the non-linear seepage flow equations (Fig. 4)


have been presented by Anagnostou (1992). The time needed to achieve a
steady state at the floor area is governed by the saturated conductivity kmax,
whereas the development of swelling above the floor is governed almost
exclusively by the minimum conductivity kmin. Assuming a low, but never-
theless realistic value for the ratio of minimum to maximum conductivity,

Deformed
Boundary z,~

Vertical Plastic
Strain ez Zone
5 10
m i n d
15%

F i g . 10. D i s p l a c e m e n t s , v e r t i c a l s t r a i n a n d p l a s t i c z o n e a t kmax t = 0.1, 0.2 a n d 10 ( B = 0 . 0 3 ;


qba = 3 0 ~ ; ca = 0 . 1 0 M P a ; kmin/kmax= 10 - 6)
A Model for Swelling Rock in Tunnelling 329

the time lag between floor and crown deformations becomes so large that,
when crown movements begin to appear, floor swelling has long been
completed. Figure 10 shows the computed deformations (five-fold magnifi-
cation) of an unlined tunnel, the vertical strain along the axis of symmetry
and the plastic zone at three different time-points. We see how the swelling
starts from the floor and progressively proceeds to the deeper rock area.
We see, furthermore, that the development of swelling strains takes place
simultaneously with a yielding of the rock mass.

6. Concluding Remarks

In this paper, we modelled the swelling phenomenon in tunnelling as a


coupled hydraulic-mechanical process, and the swelling rock as an elasto-
plastic material. The swelling phenomenon can be seen, therefore, from a
new perspective. The proposed model represents an extension of the
conventional framework, and hence allows for a more consistent interpre-
tation of the observed phenomena.

Acknowledgements
The author wishes to acknowledge Professor Kalman Kovfiri, without whose
encouragement and support this research work would not have materialized. The
author expresses his gratitude to the following colleagues for valuable discussions
during the course of this work: Christian Amstad, Dr. Thomas Arn, Christian
Gmuender, Dr. Bernhard Graf and Jan G. Korvink.

References

Anagnostou, G. (1991): Untersuchungen zur Statik des Tunnelbaus in quellfahigem


Gebirge. Dissertation 9553, Swiss Federal Institute of Technology Zurich.
Anagnostou, G. (1992). Importance of unsaturated flow in predicting the deforma-
tions around tunnels in swelling rock. In: Mermoud et al. (eds.) Porous or
fractured unsaturated media: Transports and behaviour. Swiss Federal
Institute of Technology of Lausanne, University of Neuchatel, 343--359.
Barenblatt, G. I., Zheltov, I. P., Kochina, I. N. (1960): Basic concepts in the theory
of seepage of homogeneous liquids in fissured rocks [Strata]. PMM 24 (5),
852--864.
Bellwald, Ph., Einstein, H. H. (1987): Elasto-plastic constitutive model. In: Herget,
G., Vongpaisal, S. (eds.) Proc., 6th Int. Congress on Rock Mechanics,
Montreal. Vol. 3, Balkema, Rotterdam, 1489--1492.
Biot, M. A. (1941): General theory of three-dimensional consolidation. J. Appl.
Phys. 12, 155--165.
Chenevert, M. E. (1969): Adsorptive pore pressures of argillaceous rocks. In:
Proc., 11th AIME-Symposium, Rock Mechanics from Theory to Practice,
599--627.
330 G. Anagnostou

Desai, C. S., Li, G. C. (1983): A residual flow procedure and application for free
surface flow in porous media. Adv. Water Res. 6, 27--35.
Einstein, H. H., Bischoff, N., Hofmann, E. (1972): Verhalten von Stollensohlen in
quellendem Mergel. In: Grob, H., Kovfiri, K. (eds.) Int. Symposium on Under-
ground Openings, Lucerne. Swiss Soc. for Soil Mech. and Found. Engng.,
Zurich, 296--319.
Fecker, E., Wullschl~iger, D. (1991): Geotechnische Mel3einrichtungen in der
Untersuchungsstrecke U1 des Freudensteintunnels, MeBergebnisse, ibw Inge-
nieurbauwerke 7, 195--213.
Fr6hlich, B. (1986): Anisotropes Quellverhalten diagenetisch verfestigter Tonsteine.
Ver6ff. Institut f. Bodenmechanik und Felsmechanik, Universit~it Frideri-
ciana, Karlsruhe, 99.
Grob, H. (1972) Schwelldruck im Belchentunnel. In: Grob, H., Kov/tri, K. (eds.)
Int. Symposium on Underground Openings, Lucerne. Swiss Soc. for Soil
Mech. and Found. Engng., Zurich, 99--119.
Gysel, M. (1977): A contribution to the design of a tunnel lining in swelling rock.
Rock Mech. 10, 55--71.
Gysel, M. (1987): Design of tunnels in swelling rock. Rock Mech. Rock Engng. 20,
219--242.
Holtz, W. G., Gibbs, H. J. (1956): Engineering properties of expansive clays. Trans.
ASCE 121, Paper 2814, 641--663.
Koiter, W. T. (1953): Stress-strain relations, uniqueness and variational theorems
for elasto-plastic materials with a singular yield surface. Q. Appl. Mathem. 11,
350--354.
Kov/tri, K., Madsen, F. T., Amstad, Ch. (1981): Tunnelling with yielding support in
swelling rocks. In: Akai, K. (ed.) Proc., Int. Symposium on Weak Rock,
Tokyo. Balkema, Rotterdam, 1019--1026.
Kov/tri, K., Amstad, Ch., Anagnostou, G. (1987): Tunnelbau in quellfahigem
Gebirge. Mitt. Schweizer. Ges. Boden-Felsmechanik 115.
Kovhri, K., Amstad, Ch., Anagnostou, G. (1988): Design/construction methods --
Tunnelling in swelling rocks. In: Cundall et al. (eds.) Key questions in rock
mechanics. Proc., 29th U. S. Symposium. Balkema, Rotterdam, 17--32.
Lombardi, G. (1984): Underground openings in swelling rock. In: Proc., 1st
National Conference on Case Histories in Geotechnical Engineering, Lahore.
Madsen, F. T. (1979): Determination of the swelling pressure of claystones and
marlstones using mineralogical data. In: Proc., 4th Congress ISRM,
Montreux, vol. 1. Balkema, Rotterdam, 237--241.
Marsily, G. (1986): Quantitative hydrogeology. Groundwater hydrology for
engineers. Academic Press, London.
Pressel, W., Kauffmann, J. (1860): Der Bau des Hauensteintunnels auf der Schwei-
zerischen Centralbahn. Bahnmaier's Buchhandlung, Basel.
Terzaghi, K. (1925): Erdbaumechanik auf bodenphysikalischer Grundlage.
Deuticke, Leipzig.
Terzaghi, K. (1946): Rock defects and loads on tunnel supports. In: Proctor, R. V.,
White, T. (eds.) Rock tunneling with steel supports. Commercial Shearing and
Stamping Company, Youngstown, Ohio.
Vardar, M., Fecker, E. (1984): Theorie und Praxis der Beherrschung 16slicher und
quellender Gesteine im Felsbau. Felsbau 2, 91--99.
A Model for Swelling Rock in Tunnelling 331

Wiesmann, E. (1914): l~lber die Stabilit/~t yon Tunnelmauerwerk unter Berticksich-


tigung der Erfahrungen beim Bau des Hauenstein-Basistunnels. Schweizer.
Bauzeitung 64, 27--32.
Wittke, W. (1978): Grundlagen fiJr die Bemessung und Ausftihrung von Tunnels in
quellendem Gebirge und ihre Anwendung beim Bau der Wendeschleife der
S-Bahn Stuttgart. Ver6ff. Institut f. Grundbau, Bodenmech., Felsmech.,
Verkehrswasserbau, RWTH Aachen, vol. 6.
Wittke, W., Rissler, P. (1976): Bemessung der Auskleidung von Hohlr/iumen in
quellendem Gebirge nach der Finite-Element-Methode. Ver6ff. Institut f.
Grundbau, Bodenmech., Felsmech., Verkehrswasserbau, RWTH Aachen,
vol. 2, 7--46.
Zienkiewicz, O. C. (1985) Numerical modelling and geomechanics (soil-rock-
concrete). In: Bazant, Z. (ed.) Mechanics of Geomaterials. J. Wiley, New
York, 471--499.
Zienkiewicz, O. C., Taylor, R. L. (1989): The finite element method. 4th ed.,
McGraw-Hill, London.

Author's address: G. Anagnostou, Swiss Federal Institute of Technology, ETH


H6nggerberg, P. O. Box 133, CH-8093 Zurich, Switzerland.

Você também pode gostar