You are on page 1of 13

Available online at www.sciencedirect.

com

ScienceDirect
Underground Space 2 (2017) 60–72
www.elsevier.com/locate/undsp

A coupled elastoplastic and visco-plastic damage model for hard


clay and its application for the underground gallery excavation
Hanbing Bian a,b,⇑, Xiaotian Zhang c, Jianfu Shao c,d
a
Department of Geotechnical Engineering, School of Civil Engineering, Tongji University, Shanghai 200092, China
b
LEM3 UMR CNRS 7239, Université de Lorraine, Ile du Saulcy, 57045 Metz, France
c
College of Civil and Transportation Engineering, Hohai University, Nanjing 210098, China
d
Polytech’Lille, Université de Lille 1, Cité scientifique, 59655 Villeneuve d’Ascq, France

Received 22 January 2017; received in revised form 14 March 2017; accepted 27 March 2017
Available online 1 April 2017

Abstract

The numerical modelling of the excavation of an underground gallery in hard clay has been discussed in current article. A constitutive
model is proposed to describe poromechanical behaviour of the hard clay. The main features of the hard clay observed in laboratory and
in-situ experimental investigation have been taken into account in the proposed constitutive model, in particular the plastic deformation,
the visco-plastic deformation, the damage, etc. The influence of the initial in-situ stress and the pore pressure has been taken into con-
sideration. The numerical modelling of the underground excavation has been implemented by using a fully coupled hydro-mechanical
finite element calculation code. The performance of the model is examined by comparing numerical simulations with in situ measure-
ments. The proposed model and the calculation procedure for the modelling of the excavation of an underground gallery have the capac-
ity to reproduce well the excavation damaged/distributed zone and other main features and phenomena observed during the excavation
process. However, the in-situ observed convergence could not be reproduced correctly. More effort on the discontinuous problem should
be made for the reproduce the observed convergence.
Ó 2017 Tongji University and Tongji University Press. Production and hosting by Elsevier B.V. on behalf of Owner. This is an open access article
under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

Keywords: Full coupled hydro-mechanical processes; Elasto-plastic-damage model; Visco-plastic model; Underground excavation

Introduction the coordination of the French ‘‘Agence Nationale de ges-


tion des Déchets RAdioactifs” (ANDRA) for several years
In the context of feasibility study for nuclear waste stor- in the Meuse/Haute-Marne underground research Labora-
age, and its stability during the excavation and construc- tory (called URL-Bure). The current work is performed
tion phase of the underground facilities, and finally the also in the framework of excavation damage/distributed
durability and safety of the underground repository, exten- zone in cooperation with ANDRA. The current research
sive experimental investigations and numerical simulations includes two underground experimental galleries, GED
have already been and have to be performed in order to and GCS, as illustrated in Fig. 1 (ANDRA, 2012). The
evaluate thermal-hydro-mechanical and chemical distur- two galleries located at the same formation, the hard clay
bances in engineering and geological barriers. A series of so called Callovo-Oxfordian argillite, which has been con-
research works have been conducted for this purpose under sidered as the future geological barrier for the high level
long-lived radioactive waste. As shown in Fig. 1, the two
⇑ Corresponding author at: LEM3 UMR CNRS 7239, Université de galleries are quasi perpendicular one against the other.
Lorraine, Ile du Saulcy, 57045 Metz, France. The initial in-situ stresses are: the vertical stress is about
E-mail address: hanbing.bian@univ-lorraine.fr (H. Bian).

http://dx.doi.org/10.1016/j.undsp.2017.03.002
2467-9674/Ó 2017 Tongji University and Tongji University Press. Production and hosting by Elsevier B.V. on behalf of Owner.
This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
H. Bian et al. / Underground Space 2 (2017) 60–72 61

Fig. 1. Locations of the studied two galleries: GED and GCS in the underground laboratory of ANDRA (ANDRA, 2012).

12.7 MPa; the horizontal stress parallel to the direction of desaturation in the surrounding host geological formation.
the gallery GED is about 12.4 MPa; while the horizontal The whole analysis of the excavation should be taken in the
stress in the direction of gallery GCS is about 16.12 MPa. full coupled hydro-mechanical framework. And the hard
The initial in-situ pore pressure is about 4.7 MPa. clay should be considered as porous media saturated by
The host formation of the two galleries, Callovo- either liquid water and/or dry air. In fact, the excavation
Oxfordian argillite, and here after will be named as hard process of the underground gallery is not only the redistri-
clay in the context. According to extensive experimental bution of the initial in-situ stresses but also the unloading
data (Chiarelli, 2000; Jia, Song, Duveau, Su, & Shao, of the initial pore pressure. In our cases, the initial in-situ
2007), the principal mechanical behaviour of such hard pore pressure with an order of 4.7 MPa could not be
clay is characterized by important plastic deformation, neglected. During the last thirty years, many advances have
which is tightly coupled with induced damage due to the been performed for unsaturated soils and clays (Alonso,
growth of microcracks. Concerning plastic description, Gens, & Josa, 1990; Cui & Delage, 1996; Fredlund &
some fundamental features should be taken into account, Rahardjo, 1993; Gens & Alonso, 1992; Matyas &
for example, non-linear failure surface, strain hardening Radhakrishna, 1996; Swoboda & Yang, 1999; Thomas,
and softening, transition from plastic compressibility to Rees, & Sloper, 1998). And also a few works have been
dilatancy, and sensitivity to water content. Therefore, it is conducted on experimental study and on constitutive mod-
needed to develop a coupled elastoplastic damage model elling of unsaturated rocks (Jia et al., 2007; Ju, 1989; Shao,
to describe the fundamental mechanical behaviour of this Jia, Kondo, & Chiarelli, 2006a; Shao, Jia, Kondo, &
material. However, comparing to the induced damage the Chiarelli, 2006b). The aim of this paper is to present a cou-
plastic deformation is generally the dominant mechanism. pled elastoplastic and visco-plastic damage model for the
In additions, the hard clay shows important visco-plastic hard clay in saturated and unsaturated conditions for the
strain with times. The obvious evidence is the continuous unloading process. The model is formulated in the frame-
convergences of the galleries after the excavation. In order work of unsaturated porous media (Coussy, 2004). The
to take account of this phenomenon, an elastoplastic and current paper is organised as following: after a brief intro-
visco-plastic damage model is necessary. duction, the constitutive model for the hard clay in the
The influence of the pore water on the mechanical beha- framework of unsaturated porous media specific for the
viour of the hard clay could not be neglected (Jia et al., underground excavation has been presented and discussed.
2007). Furthermore, the excavation and the eventual After the validation of the proposed model, it has been
ventilation in the gallery after excavation will induce the used to study the excavation of the two experimental
62 H. Bian et al. / Underground Space 2 (2017) 60–72

underground galleries in underground laboratory of et al., 2007). It is worth mentioning here that the descrip-
ANDRA. By the discussion on the numerical results, serval tion of the poroelastic behaviour is not unique. For exam-
conclusion could be made with the proposed numerical ple, Fredlund and Rahardjo (1993) and Alonso et al.
model. (1990), they suggested to using two independent variables
in place of the concept of effective stress. In our work,
The constitutive model for hard clay during excavation the equivalent pore pressure is used to represent the effects
of liquid pressure and its dependency on water saturation
As presented in the section precedent, an elastoplastic degree. The equivalent pore pressure is expressed in the fol-
and visco-plastic damage model is necessary for the lowing form:
description of unsaturated poromechanical behaviour of dp ¼ dpgz  S lq ðpcp Þdpcp ð4Þ
the hard clay. The outlines of the model are briefly pre-
sented in this section. Generally, for clay and especially The function S lq ðpcp Þ represents the isothermal water
for hard clay, the anisotropic properties could be observed. retention curve of the studied hard clay. Avec this formu-
However, in current research, for simplicity reasons, the lation (Eq. (4)), the equivalent pressure is then introduced
hard clay is assumed as an isotropic porous medium satu- in the incremental form of Biot’s poroelastic relation. It
rated by liquid water (noted by index lq) and a gas mixture can be seen that, this is highly nonlinear problem.
(noted by index gz). The gas mixture is a perfect mixture of
dry air (noted by index da) and vapour (noted by index va). Damage characterization
However, in current research, we don’t distinguish the
water vapour from the dry air, and the phase change Based on the experimental investigation, in spite of the
between liquid water and water vapour has not been taken plastic mechanism, the damage is also a main facture for
into consideration. It is also supposed that the pore air the failure of the hard clay, especially under the unloading
pressure pgz is supposed as equal to the atmosphere pres- process. The obvious evident is the so called EDZ (excava-
sure (patm ). As a result, the capillary pressure in increment tion damaged/distributed zone). So, a coupled elastic-
form is defined as follow: plastic-damage model will be used in the present study
for the description of the unloading behaviour of the hard
dpcp ¼ dpgz  dplq ð1Þ clay. The constitutive equation for the damaged hard clay
The basic mechanical behaviour of argillite, as illus- then can be written as follows:
trated by Jia et al. (2007) and Zhou, Jia, and Shao (2007) dr ¼ cðxÞ : ðde  dep Þ  bdpd ð5Þ
is characterized by a unified elastic–plastic and visco-
plastic model. The assumption of small strains is then In Eq. (5), b is Biot coefficient. d represents the second
adopted, thus the total strain can be decomposed into an order unit tensor. And the fourth order tensor cðxÞ repre-
elastic part eeij and plastic part epij :
sents the damaged elastic matrix. With the elastic matrix
deij ¼ deeij þ depij ð2Þ given as:
In fact, as illustrated in the section precedent, the stud- cðxÞ ¼ 2GðxÞK þ 3KðxÞJ ð6Þ
ied hard clay has a time dependant deformation properties.
And the major time dependant deformation occurs at the where isotropic symmetric fourth order tensors K, J are
beginning of loading variation. During the unloading of defined by:
the excavation process, the quantity of the time dependent
dd dik djl þ dil djk
deformation could not be neglected. It is essential to taking J¼ ; K ¼ I  J ; I ijkl ¼ ð7Þ
the time dependant deformation into consideration. For 3 2
the purpose to distinguish the time-independent deforma- I is the symmetric fourth order unit tensor. And dam-
tion from that of the time dependent behaviour, it is
aged bulk modulus and damaged shear modulus are given
assumed that the plastic strain epij in the equation above
as:
can then be further decomposed into two parts: instanta-
neous plastic strain erp vp KðxÞ ¼ K 0 ð1  xÞ; GðxÞ ¼ G0 ð1  xÞ ð8Þ
ij and visco-plastic strain eij . It is
worth mentioning that, for the sake of simplicity, an where K 0 and G0 are effective drained bulk and shear mod-
isothermal condition and small deformation assumption uli of intact material.
are adopted in the present study. As a result, the total In practice, the microcracks developed in hard clay are
strain can be decomposed as: distributed in some preferential orientations and then lead
deij ¼ deeij þ derp to an induced anisotropic damage of materials. As the
ij þ deij ð3Þ
vp

emphasis of the present work focuses on the numerical


The poroelastic behaviour is described using the modelling of the excavation process, for the reason of sim-
extended Biot’s theory for partially saturated porous media plicity, an isotropic damage characterized by a scalar vari-
(Biot, 1955, 1973; Bishop & Blight, 1963; Coussy, 2004; Jia able x is adopted in the present work. Due to the different
H. Bian et al. / Underground Space 2 (2017) 60–72 63
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
 !
failure modes, the damage variable is decomposed in two p f c ðxÞ
F p ¼ q  gðhÞaðc Þ c1 þ c1 þ 4c2 c3 
p 2
¼0 ð14Þ
parts: xc for the compressive damage and xt for the tensile f c ðxÞ 2c2
one. The global damage variable is simply defined as:
rffiffiffiffiffiffiffiffiffiffiffi
krþ k trðrpl Þ 3 trðrpl Þ
x ¼ ð1  at Þxc þ at xt ; at ¼ ð9Þ p¼ ;q ¼ s : s; s ¼ rpl  d ð15Þ
krk 3 2 3

The coefficient at determines contribution of the differ- In the equations describing the yield surface, s is the ten-
ent effects of tensile and compressive damage. rþ is the pos- sor of deviatoric effective stress; f c represents the uniaxial
itive part of stress tensor. As defined in Eq. (9), it can be compression strength of intact hard clay which can be
seen that for a uniaxial compression test at = 0 and in a determined by using the peak stress of a uniaxial compres-
uniaxial tensile tests at = 1. sion tests; three constants c1 ; c2 ; c3 define the curvature of
The experimental investigations show that in tensile yield surface (Fig. 6) and can be determined by plotting
regime, the damage evolution is inherently induced by ten- the peak stress versus the mean stress obtained from classi-
sile strains while in compressive regime, the material dam- cal triaxial compression tests in p  q plan. For taking the
age development is related to frictional sliding along micro- damage effects on the mechanical strength, it is supposed
crack surfaces. In view of this fact, it is assumed that the that the uniaxial compression strength of hard clay
damage variations, under tensile and compressive loading decreases with the induced damage in a linear manner.
conditions, are respectively controlled by two different The following linear relation is then proposed:
driving forces Y tx and Y cx , respectively for the damage in  
hx  x0 i
traction and in compression, as: f c ðxÞ ¼ 1  f c0 ð16Þ
1  x0
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
X3 2
Yx ¼ 2
t
i¼1
hei i ð10Þ Here f c0 denotes the uniaxial compression strength of
undamaged hard clay. And x0 represents the damage
Z rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi  
2 de threshold for which the damage has the non-negligible
Yx ¼
c
de : de; de ¼ de  tr d ð11Þ
3 3 influence on the plastic mechanism. In fact, Eq. (16) allows
the proposed model to describe the softening behaviour of
where ei (i = 1, 2 and 3) denotes the three principal strain hard clay due to the presence of the micro-cracks during
components. The bracket hxi means that only the positive excavation. In the laboratory test, it seems the plastic
value will be considered: hxi = 0 if x 6 0 and hxi = x if mechanism is dominant, thus in current study, we given a
x > 0. Inspired by the damage model proposed by relative high value for x0 as 0.97.
Mazars (1984), the evolutions of tensile damage and com- The plastic hardening law, which controls the evolution
pressive damage are determined respectively by the follow- of yield surface, is defined as the function of the generalized
ing criteria: plastic shear strain cp which has the tendency to increase.
 
1 Based on the previous works (Chiarelli, Shao, & Hoteit,
f tx ¼ xt  1  ¼0 ð12Þ 2003) and its representative experimental data, the follow-
expðBt Y tx Þ
  ing plastic hardening law is used, as:
1
f cx ¼ xc  1  ¼0 ð13Þ cp
expðBc Y cx Þ aðcp Þ ¼ a0 þ ð1  a0 Þ ð17Þ
B þ cp
The parameter Bc controls the kinetic of compressive
Z rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
damage and can be determined by using a uniaxial com- 2 p trðdep Þ
pression test. The parameter Bt characterizes the evolution cp ¼ de : dep ; dep ¼ dep  d ð18Þ
3 3
of tensile damage and can be identified from a uniaxial ten-
sion test. The hardening function of a varies from 0 to 1. For
example, when a ! 1, that means the failure surface is
Plastic characterization reached. The model parameter B, which can be determined
by calculating aðcp Þ versus cp on stress stain curves in uni-
For the laboratory tests, it seems that the plastic defor- axial compression test, controls the kinetic of the plastic
mation of hard clay depends on the confining pressure hardening. The constant a0 defines the initial plastic
(Jamet, Millard, & Nahas, 1984; Sfer, Carol, Gettu, & domain of the hard clay, as observed from the laboratory
Etse, 2002). Thus, a non-linear yield surface which depends tests the value could be set as 0.15.
on the mean stress is generally necessary. As inspired by the The geomaterials generally need a non-associated plastic
previous works, such as that of Pietruszczak, Jiang, and flow rule. Among various existing models, the following
Mirza (1988) and Zhou, Bian, Jia, and Shao (2013), a plastic potential proposed by Shao et al. (2006a,b) and
curved yield surface maybe adequate for the studied hard Jia, Bian, Su, Kondo, and Shao (2010) has been adopted
clay (as illustrated in Fig. 2): here, as:
64 H. Bian et al. / Underground Space 2 (2017) 60–72

Fig. 2. The illustration of the yield surface, the damaged yield surface and the plastic potential.

 
p þ c3 f c ðxÞ illustrated in (d) in Fig. 3, not only the values but also
Qp ¼ q  ggðhÞðp þ c3 f c ðxÞÞ ln ¼0
I0 the direction of the principal stresses have been changed.
ð19Þ In fact, during the excavation, each point has its unique
loading path. It is well known that the geomaterial is load-
The variable I 0 defines geometrically the intersection of ing path dependant material. Even at the beginning and the
the plastic potential (the blue one in Fig. 2) and the p axis. end of load process, the same materials have the same
The boundary between compaction and dilatancy domains stress states, if they submit with different loading paths;
is defined by the condition @Qp =@p ¼ 0 (the violet one in the final deformations will not be the same. In our model,
Fig. 2), which can be written as follows: we use the function gðhÞ for taking into consideration of
this effect. This function is given as:
F s ¼ q  gðp þ c3 f c ðxÞÞ ¼ 0 ð20Þ
pffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffi
ð 1 þ a  1  aÞK
The parameter g defines the slope of the boundary gðhÞ ¼ pffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ð21Þ
K 1 þ a  1  a þ ð1  KÞ 1  a sin 3h
between compaction and dilatancy domain in the p  q
plan (the violet one in Fig. 2). It can be determined by
In function gðhÞ, the parameter K is defined as the ratio
reporting onto p  q plane the stress state at the volumetric
of the peak strength in unloading triaxial test to the loading
transition point from compaction to dilatancy.
triaxial test. As illustrated in Fig. 4, when the Lode’s angle
It is also worth mentioning that, in the yield surface and
equal to 30°, which represents the unloading triaxial test,
the plastic potential, the variable gðhÞ in function of the
the peak strength is relatively low comparing to the normal
Lode’s angle has been integrated for taking the influence
triaxial test with the Lode’s angle as to 30°. The param-
of loading path. During the excavation of underground
eter a in Eq. (21) controls the couverture of the function
gallery, the loading paths are not as simple as usually sup-
gðhÞ. It can be determined by best fitting of the laboratory
posed. In fact, each point has a different loading path. For
tests with different loading path (different Lode’s angles).
example, we suppose an excavation in a homogenous, iso-
tropic material with an initial in-situ stresses parallel to
axes of the coordinate system, as illustrated in Fig. 3. Sup- Time-dependant plastic deformation
pose the excavation direction is in the in-plan direction and
parallel to the third principle stress. We chose three repre- The laboratory experiments and the in-situ observation
sentative points for the explication of different loading show that important creep deformation occurs in the hard
path. Due to the excavation, at Point A, as illustrated in clay, especially the convergence of the underground open-
(c) in Fig. 3, the vertical stress decreases to zero, while ing in the unloading process (excavation). In the classic
the stress in the horizontal direction increases. For that modelling, the instantaneous plastic deformation and the
at Point B, in the contrary, the vertical stress increases time depend plastic deformation are generally treated
while the stress in the horizontal direction decreases to separately. Two different yield surfaces and corresponding
zero, as illustrated in (b) in Fig. 3. While for Point C, as plastic potentials are then necessary for describing instanta-
H. Bian et al. / Underground Space 2 (2017) 60–72 65

Fig. 3. The loading paths for three different points during the excavation: (a) initial state; (b) the stress state of Point B after excavation; (c) the stress state
of Point A after excavation; (d) the stress state of Point C after excavation.

hardening function. In place of aðcp Þ the function of


visco-plastic hardening function avp ðcp Þ is used, and given
as:
cp
avp ðcp Þ ¼ ð23Þ
Bvp þ cp
The parameter Bvp controls the evolution of visco-plastic
hardening law. If avp ¼ a, the yielding surface for the time
dependent plastic deformation is coincides with the yield
surface for the instantaneous plastic strain failure surface
defined as in Eq. (14), thus there is no anymore time depen-
dant plastic strain. In fact, the development of the time
dependent plastic strain is also the process of the yield sur-
face defined in Eq. (22) approaches to that defined in Eq.
Fig. 4. The function gðhÞ varies with Lode’s angle. (14). The visco-plastic potential, similar to that of the yield-
ing surface, also keeps the same form as that of time-
neous plastic behaviour and the time dependant plastic independent plasticity:
strains, respectively. In the current study, a unified  
p þ c3 f c ðxÞ
approach is adopted (Zhou et al., 2007). It is assumed that Qvp ¼ q  ggðhÞðp þ c3 f c ðxÞÞ ln ¼0
I0
the total plastic strain of the material composed of two
parts: time-independent (instantaneous) plastic strain erp ð24Þ
which is generated by the stress change, and time- A visco-plastic flow rule, which is proposed or the hard
dependent (creep) plastic strain evp which is generally due clay basing on the experimental results, is given in the fol-
to material viscosity and temporal evolution of mechanical lowing form as:
properties related with various physical-chemical processes.  n
F vp @Qvp
The two plastic strains are described by the same plasticity e_ vp ¼ AðT Þ ð25Þ
ij
f c ðxÞ @rij
framework but in two different time scales. The loading sur-
face associated with the visco-plastic flow (creep) is con- The parameter n describes the shape of the curve of
trolled by the delayed hardening law avp ðcp Þ. Thus, the creep. The function AðT Þ defines the fluidity of material
visco-plastic yield function is written as: which depends on the temperature T :
 
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
 ! f
p f c ðxÞ AðT Þ ¼ A0 exp  ð26Þ
F vp ¼ q  gðhÞavp ðc Þ c1 þ c1 þ 4c2 c3 
p 2
¼0 RT
f c ðxÞ 2c2
The parameter f controls the influence of temperature
ð22Þ
on the viscosity of material. The average temperature of
the formation where the galleries are excavated is about
It can be seen that, the yield surface for the time depen- 20 °C, and is quasi constant. The parameter f then also
dant plastic deformation is quasi the same as that for the could be considered as a constant. The parameter A0 con-
instantons plastic deformation. The only difference is the trols the initial value of function AðT Þ. And R is a constant
66 H. Bian et al. / Underground Space 2 (2017) 60–72

of model associated with the temperature. Using triaxial Table 1


test for different confining pressure and a creep test, all Representative values of model parameters for hard clay.
parameters of the proposed model can be identified. Model Parameters Value
Elasto-plastic Young’s modulus E ðMPaÞ 4000
Coefficient of Poisson v 0.3
Validation of the proposed model Uniaxial compression strength f c0 (MPa) 17.5
Parameter c1 0.15
Parameter c2 0.25
For the proposed model, including elastic, plastic, visco- Parameter c3 0.04
plastic and damage mechanism, there are totally 13 param- Parameter g 2.5
eters. There are two elastic parameters, Young’s modulus Parameter B 2.3E4
and coefficient of Poisson. They can be determined directly Visco-plastic Parameter A0 1.0E12
from the triaxial test. There are totally 6 parameters for the Parameter n 15
plastic mechanism. The first one is the uniaxial compres- Parameter Bvp 0.01
sion strength f c . As its name, this parameter can be deter- Damage Parameter Bt 40
mined via the uniaxial compression tests. The three Parameter Bc 5
parameters c1 , c2 and c3 controlling the curvature of the
plastic yielding surface can be determined by plotting the
Table 2
failure stresses of a serious triaxial test with different con- Typical value of coupling parameters for hard clay.
fining pressure in the p  q plan, as illustrated in Fig. 2.
Parameters Value
By best fitting, the three parameters then could be deter-
mined. The parameter g defines the slope of Eq. (20). It Physique Volume density q ðkg=m3 Þ 2470
Porosity / 18
also can be defined by the volumetric deformation of the
spacemen during triaxial tests, as in the classic plastic Hydro-mechanical Biot’s coefficient b 0.6
Hydraulic conductivity k ðm=sÞ 1.0E13
mechanic course. The parameter B controls the plastic
hardening can be determined by best fitting the axial and
lateral deformation during the triaxial tests. The advantage
of the current model is that the yield surface and the plastic
potential for the visco-plastic mechanism are the same as
that for plastic mechanism. So only three parameters are
need for describing the visco-plastic deformation. The
parameter A0 and n control the velocity of the visco-
plastic strain. And parameter Bvp controls the approaching
velocity of the visco-plastic yield surface to the plastic yield
surface. The three parameters are determined via the creep
tests on the hard clay. For the damage mechanism there are
only two parameters Bt and Bc , respectively controlling the
damage evolution velocity in traction and compression.
The two parameters are determined via a uniaxial traction
test and a uniaxial compression test respectively. All the Fig. 5. Numerical simulation of a triaxial test on hard clay with a
total 13 parameters are regrouped in Table 1. confining pressure as 12 MPa, comparing with the experimental results
The underground excavation concerns not only the in- realised at LML and LAEGO.
situ stresses but also the pore water pressure. The excava-
tion process should be conducted in the fully coupled
hydro-mechanical framework. Certain coupling parame- together with the numerical results in Fig. 6. It is worth
ters are needed, as listed in Table 2. mentioning that with the different loading path, the peak
By using the determined parameters (Table 1), a triaxial strengths of the two triaxial tests are different. For the
compression test with a confining pressure as 12 MPa has loading triaxial test, the softening process is very clear both
been done. The numerical results are given in Fig. 5. The in the experimental and numerical results. However, in the
laboratory results realised in both Laboratoir de Mécani- unloading triaxial test, there is no softening phenomenon
que de Lille (LML) and Laboratoire Environnement Géo- observed. The presented model can reflex well the influence
mécanique et Ouvrages (LAEGO) are comparing with the of loading path with the function given in Eq. (21).
numerical results. It seems that the numerical model can Again, with the same parameter (Table 1) three creep
represent relatively well the main features of the hard clay tests realised at LAEGO are simulated. The numerical
in the triaxial tests. With the same parameters (Table 1), results and the experimental results are given in Fig. 7. It
another triaxial test with also a confining pressure as is worth mentioning that, for the three tests, the same
12 MPa is simulated. However, this time, it is unloading parameters are used; the only differences are the boundary
test. The experimental results realised by LML are given conditions. Comparing with the experimental results, the
H. Bian et al. / Underground Space 2 (2017) 60–72 67

ε 1% 0 ε 3%
-1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6

-3

Numerical_Simualtion
Numerical_Simualtion -6
Numerical_Simualtion
Experiment
Experiment -9
Experiment

-12

-15

σ1− σ3 (MPa)
-18

Fig. 6. Numerical simulation of a triaxial test with unloading path on hard clay comparing with the experimental results.

Exp_90% Exp_75% Exp_50%


12000 Deformation Axial
Num_90% Num_75% Num_50%

9000

6000

3000

Time(day)
0
0 25 50 75 100 Fig. 8. The boundary conditions for the simulations of the excavation of
two galleries (GED and GCS).
Fig. 7. Numerical simulation of three creep tests comparing with the
experimental results.

excavation of the underground gallery is a 3D problem.


numerical model reproduces relatively well the main fea- However, for the reason of simplicity and reducing the cal-
tures of the hard clay for the time dependant deformation. culation time, a 2D plan strain model will be used in cur-
All the three tests are simulated with the same parameters. rent study. As we suppose that the material is isotope
Thus, the proposed model and the parameters are validated and the boundary conditions could be simplified as sym-
in the same time. This set of parameters will be used later metric, only a quarter of the 2D model is then needed.
for the simulation of the excavation of the underground The geometry and the boundary conditions of the two gal-
galleries. leries are given in Fig. 8. As an academic study, the dimen-
sion of the calculation domain is selected as 50 m. The
Numerical modelling of the excavation of underground radius of the excavation is 2.6 m. The vertical in-situ stress
gallery and discussion is 12.7 MPa. The maximum horizontal in-situ stress is
16.12 MPa, and the minimum value is 12.4 MPa. The ini-
Presentation tial in-situ pore pressure is 4.7 MPa. The excavation is
not only the redistribution of the in-situ stress but also
In this section, the excavation of the gallery GED and the unloading of the initial pore pressure. The excavation
GCS in the underground laboratory of ANDRA has been process in current study is simulated as the reduction of
numerically studied via the proposed model presented in the initial in-situ stress and the initial in-situ pore pressure
the precedent section. The two galleries locate in the forma- to the atmosphere pressure at the excavation surface.
tion of hard clay with an average thickness about 150 m at However, unloading of the initial in-situ stress at studied
a depth of about 500 m. This type of gallery will later be the excavation plan is not linear in function of time. In fact,
used as the radioactive waste repository. In reality, the its variation is so complicate that cannot be described with
68 H. Bian et al. / Underground Space 2 (2017) 60–72

a simple equation, especially in a two dimensional simula- the reference case will be given. And then each case will
tion. As illustrated in Fig. 9, the considered plan is located be compared and discussed in the following.
at Point B. The excavation is in the red arrow direction.
Before arriving the considered point B, when the excava- Reference case (Gallery GED)
tion front arrives at the point A, the unloading process is
already trigged at point B. However unloading rate is rela- In this section, only some representative numerical
tively slow and the quantity is small. With the excavation results of the reference case (GED) are presented. The dis-
front approaches to the point B, the unloading rate tribution of the displacement in the studied cross-section at
increases rapidly, as illustrate in Fig. 10. When the excava- the end of excavation (28 day) is presented in Fig. 12. It is
tion front coincides with the point B, the initial in-situ observed that the maximum convergence occurs in the hor-
stresses are not totally unloaded. Due to the structure effect izontal direction with an order of 2 cm. This is quasi the
with the hard clay behind point B, certain in-situ stress same order observed in-situ. However, the convergence at
could not be reduce to its minimum value, zero. In fact, vault in the vertical direction predicted by the numerical
the in-situ stress at point B will arrives to its final value model is about 1.7 cm. The corresponding value of the
when the excavation front arrive point C. The total unload- in-situ observation is about 5 cm.
ing process of the in-situ stress is illustrated in Fig. 10 by The distribution of the pore water pressure at the end of
the unloading coefficient in function of time. Here, the excavation is given at Fig. 13. With the excavation process,
excavation is supposed as a constant velocity of 2 m/day the pore pressure at the excavation surface reduces to the
and the total excavation process considered lasts for atmosphere pressure. As its small hydraulic permeability,
28 day. the distributed zone is limited at the two diameter of the
However, the unloading of the initial pore pressure is
quite different comparing with the initial in-situ stress. As 5 Pw (MPa)

the permeability of the hard clay is relatively small


(Table 2), the excavation process and structure effect has 4
small influence on the unloading process of the pore pres-
sure. The pore pressure reduction occurs only when the
3
excavation front approach, and reaches its final value
immediately. The unloading of pore pressure used in cur-
2
rent study is given in Fig. 11.
The configuration of gallery GED is set as the reference
case. The influence of initial in-situ stress is considered by 1

the configuration of gallery GCS. Firstly, the results of Time (day)


0
0 5 10 15 20 25 30

Fig. 11. Illustration of unloading process of the pore pressure in function


of excavation time.

Fig. 9. Illustration of the excavation process of the underground gallery.

1 Coef

0.8

0.6

0.4

0.2

Time (day)
0
0 5 10 15 20 25 30

Fig. 10. Illustration of the unloading coefficient in function of excavation Fig. 12. The distribution of the displacement at the end of excavation of
time. GED (28 day).
H. Bian et al. / Underground Space 2 (2017) 60–72 69

Fig. 13. The distribution of the pore pressure at the end of excavation of Fig. 14. The distribution of the parameter of plastic hardening at the end
GED (28 day). of excavation of GED (28 day).

excavation. It is worth mentioning also, in the current con-


figuration, an excess pore pressure of about 1 MPa was
observed at the vault of the gallery. In fact, this phe-
nomenon has been observed also in the underground labo-
ratory. That is mainly due to the small permeability of the
hard clay and the excavation induced contraction in that
region.
In Fig. 14, the distribution of the plastic hardening
parameter at the end of excavation is given. The plastic
hardening is mainly concentrated at the vault of the gallery
and limited in the zone of 2 diameters of the excavation. It
is worth mentioning that, the distribution shape and the
extended distance of the plastic hardening parameter is
similar to that observed in the underground laboratory.
The distribution of damage at the end of excavation has
been given in Fig. 15. It can be observed that, the damage
distribution shape is the same as that of the plastic harden-
ing. It is also similar to the observed EDZ of the gallery
GED. However, it is worth mentioning that, the damage Fig. 15. The distribution of damage at the end of excavation of GED
(28 day).
here is limited as a relative threshold of damage has been
used in Eq. (16). Even though, the main feature of the dam-
age around the gallery due to the excavation has well gallery. The maximum convergence is predicted by the
reproduced. model is about 1.5 cm. And small difference could be dis-
tinguished in the horizontal and vertical direction: the con-
Influence of the initial in-situ stresses (Gallery GCS) vergence in the vertical direction is greater than that in the
horizontal direction. However, the measured values in the
The configuration of the gallery GCS is the same as that underground laboratory are more important than the
for gallery GED, the only difference between them are the numerical predicted values. And it also seems that the hor-
excavation direction as illustrated in Fig. 8. For the GED, izontal convergence is more important than that of vertical
the excavation direction is parallel to the smaller initial in- one. The numerical predicated convergences are exactly
situ stress; while the gallery GCS, the excavation direction contrary to the measured one.
is parallel to the greater initial in-situ stress. However, as The distribution of the pore water pressure around gal-
the small difference, the results are totally different. As lery GCS at the end of excavation is given at Fig. 17. Sim-
shown in Fig. 16, the distribution of the displacement ilar to that in the gallery GED, the pore pressure at the
around the gallery GCS after the excavation. The distribu- excavation surface reduces to the atmosphere pressure.
tion of the displacement is quasi uniform around the Comparing to that of GED, the distribution of the pore
70 H. Bian et al. / Underground Space 2 (2017) 60–72

Fig. 16. The distribution of displacement at the end of excavation of GCS Fig. 18. The distribution of the plastic hardening at the end of excavation
(28 day). of GCS (28 day).

Fig. 19. The distribution of the damage at the end of excavation of GCS
Fig. 17. The distribution of pore water pressure at the end of excavation (28 day).
of GCS (28 day).

Discussion

pressure perturbed by the excavation is quasi uniform. A From the results, it can be seen that, the proposed
small excess pore pressure is also observed, with an order model, including elastoplastic and viscoelastic damage
of 0.1 MPa. The location of the excess pore pressure is at model, gives satisfactory results for the pore pressure
the horizontal direction, not as that of GED at the vault. (including the excessive pore pressure), the plastic harden-
In Fig. 18, the distribution of the plastic hardening at ing and damage (the excavation damaged/distributed
the end of excavation is given. The plastic hardening is zone). However, the predicted convergence (displacement)
mainly concentrated in the horizontal direction in a limited neither quantitatively nor qualitatively satisfies the
zone of 2 diameters of the excavation. Comparing with the observed one. In fact, the predicted convergence is exactly
in-situ observation, the numerical model gives a satisfac- contrary to that measured one. This, however, is not error.
tory prediction. The distribution of damage around GCS The proposed model is, in the contrary, rather precise. The
at the end of excavation has been given in Fig. 19. The dis- excavation induced damage zone and the excess pore
tribution shape of the damage is similar to the observed pressure have been successfully predicted. The disagree in
EDZ in the underground. the displacement between the measured value and the
H. Bian et al. / Underground Space 2 (2017) 60–72 71

redistribution of the pore pressure. The initial in-situ stress


has significant influence on the excavation damage/dis-
tributed zone and the pore pressure redistribution. The
in-situ measure convergence, which including the contribu-
tion of the discontinuous (crack width), could not be cor-
rectly reproduced with current continuous model.

Acknowledgements

This work has been partially supported by the ANDRA,


which is gratefully acknowledged.

References

Alonso, E. E., Gens, A., & Josa, A. (1990). A constitutive model for
partially saturated soils. Géotechnique, 40(3), 405–430.
ANDRA (2012). Etude de la fracturation induite par le creusement et de
l’EDZ pour une galerie de conception souple orientée suivant la
contrainte horizontale majeure. Expérimentation OHZ (GCS-GET)
(Technical report).
Biot, M. A. (1955). Theory of elasticity and consolidation for a porous
anisotropic solid. Journal of Applied Physics, 26, 182–185.
Biot, M. A. (1973). Non linear and semilinear rheology of porous solids.
Journal of Geophysical Research, 78(23), 4924–4937.
Bishop, A. W., & Blight, G. E. (1963). Some aspects of effective stress in
saturated and partly saturated soils. Géotechnique, 13(3), 177–197.
Chiarelli, A. S. (2000). Experimental investigation and constitutive modeling
of coupled elastoplastic damage in hard claystones (Doctoral thesis).
University of Lille.
Chiarelli, A. S., Shao, J. F., & Hoteit, N. (2003). Modeling of elastoplastic
damage behaviour of a claystone. International Journal of Plasticity,
19, 23–45.
Coussy, O. (2004). Poromechanics. John Wiley & Sons.
Fig. 20. The observed macrocracks on the wall of underground gallery
Cui, Y. J., & Delage, P. (1996). A vanishing diffusion processs in
(ANDRA, 2012). unsaturated soil. International Journal of Non-Linear Mechanics, 33(6),
1027–1037.
Fredlund, D. G., & Rahardjo, H. (1993). Soil mechanics for unsaturated
soils. New York: John Wiley & Sons.
predicted one is mainly due to the marcocracks induced by Gens, A., & Alonso, E. E. (1992). A framework for the behavior of
unsaturated expansive clays. Canadian Geotechnical Journal, 29,
the excavation. The measured convergence includes not 1013–1032.
only the continuous displacement (that can be predicted Jamet, P., Millard, A., & Nahas, G. (1984). Triaxial behavior of a micro-
by the proposed model) but also the width of the cracks concrete complete stress-strain for confining pressures ranging from 0
to 100 MPa. Proc. of Int. Conf. on Concrete under Multiaxial
(current could not be predicted by the proposed model), Conditions (vol. 1, pp. 133–140). Toulouse: Presses de l’Université
as shown in Fig. 20, the observed macrocracks on the wall Paul Sabatier.
of underground gallery after excavation (ANDRA, 2012). Jia, Y., Bian, H. B., Su, K., Kondo, D., & Shao, J. F. (2010). Elastoplastic
damage modeling of desaturation and resaturation in argillites.
In fact, the model proposed in the present study is in the International Journal for Numerical and Analytical Methods in Geome-
framework of the continuous media. The crack (the initia- chanics, 34, 187–220.
tion and the propagation of the crack) is a moving bound- Jia, Y., Song, X. C., Duveau, G., Su, K., & Shao, J. F. (2007).
Elastoplastic damage modelling of argillite in partially saturated
ary condition problem. This could be solved by either the condition and application. Physics and Chemistry of Earth, 32,
XFEM (extended finite element method) or other numeri- 656–666.
cal method, such as meshless method. The current model Ju, J. W. (1989). On energy based coupled elastoplastic damage theories:
Constitutive modeling and computational aspects. International Jour-
can predict both the plastic hardening and damage cor- nal of Solids and Structures, 25(7), 803–833.
rectly. The two parameters, then can served as the criteria Matyas, E. L., & Radhakrishna, H. S. (1996). Volume change character-
for the crack initiation and propagation. All these will be istics of partially saturated soils. Géotechnique, 18, 432–448.
Mazars, J. (1984). Application de la mécanique de l’endommagement au
discussed further in another work. comportement non linéaire et à la rupture du béton de structure (Ph.D.
thesis). Université Paris 6.
Conclusions Pietruszczak, S., Jiang, J., & Mirza, F. A. (1988). An elastoplastic
constitutive model for concrete. International Journal of Solids and
Structures, 24, 705–722.
In the present work, a coupled elastoplastic and visco- Sfer, D., Carol, I., Gettu, R., & Etse, G. (2002). Study of the behaviour of
plastic damage model has proposed and has been used concrete under triaxial compression. Journal of Engineering Mechanics,
128, 156–163.
for the numerical simulation of the excavation of two Shao, J. F., Jia, Y., Kondo, D., & Chiarelli, A. S. (2006a). A coupled
underground galleries. The proposed model can predict elastoplastic damage model for semi-brittle materials and extension to
well the excavation damaged/distributed zone and the unsaturated conditions. Mechanics of Materials, 38, 218–232.
72 H. Bian et al. / Underground Space 2 (2017) 60–72

Shao, J. F., Jia, Y., Kondo, D., & Chiarelli, A. S. (2006b). A coupled Zhou, H., Bian, H. B., Jia, Y., & Shao, J. F. (2013). Elastoplastic damage
elastoplastic damage model for semi-brittle materials and extension to modeling the mechanical behavior of rock-like materials considering
unsaturated conditions. Mechanics of Materials, 38, 218–232. confining pressure dependency. Mechanics Research Communications,
Swoboda, G., & Yang, Q. (1999). An energy based damage model of 53, 1–8.
geomaterials – I. Formulation and numerical results. International Zhou, H., Jia, Y., & Shao, J. F. (2007). A unified elastic - plastic and
Journal of Solids and Structures, 36, 1719–1734. viscoplastic damage model for quasi brittle rocks. International Journal
Thomas, H. R., Rees, S. W., & Sloper, N. J. Y. (1998). Three dimensional of Rock Mechanics and Mining Science, 45(8), 1237–1251.
heat, moisture and air transfer in unsaturated soils. International Journal
for Numerical and Analytical Methods in Geomechanics, 22, 75–95.

You might also like