Sunteți pe pagina 1din 8

The 12th International Conference of

International Association for Computer Methods and Advances in Geomechanics (IACMAG)


1-6 October, 2008
Goa, India

Elastoplastic Soil Models for Numerical Analysis of Underground


Constructions
Y. Hejazi, D. Dias, R. Kastner
INSA-Lyon, LGCIE, F-69621 ; France

Keywords: Elostoplastic models, underground constructions, modelling


ABSTRACT: The constitutive model frequently used in numerical calculations of tunnel execution is linear-elastic
perfectly plastic with a Mohr-Coulomb failure criterion (MC). Generally, this does not leave to realistic soil
movement around the excavated tunnel. In this paper, constitutive models representing the behaviour of an
overconsolidated clay with different levels of complexity are implemented in FLAC. The study is based on the
results of triaxial tests representing London Clay (Gasparre, 2005). The mechanical parameters calibrated from
these tests for the different constitutive models adopted in this study are then applied to the excavation of a
shallow tunnel. The influence of the chosen model and the cover depth over a tunnel are highlighted on soil
displacements. It is concluded that it is unsatisfactory to simulate soils by the MC model when modeling
underground structures and it is therefore necessary to adopt advanced models incorporating non linear behavior
under small-strain.

1 Introduction
In urban zones, the influence of underground works on neighbouring buildings has an important economic and
environmental aspect. The volume loss caused by tunnelling propagates toward the surface creating ground
movements that can more or less affect existing buildings. As a first step, soil movements are estimated by
empirical (OReilly and New, 1982 ; Peck, 1969), analytical (Sagaseta, 1987 ; Panet, 1995) or numerical (Mroueh,
1998 ; Oteo and Sagaseta, 1982 ; Swoboda et al., 1995) methods. The interaction between existing structures
and underground works is a complex phenomenon in which the behaviour of the surrounding ground is one of the
main aspects of a tunnel excavation to be taken into account. Consequently, a realistic ground model is crucial in
order to estimate the magnitudes and distribution of the strains. The constitutive model frequently used during the
numerical simulation of an underground excavation work is linear-elastic perfectly-plastic with a Mohr-Coulomb
failure criterion. In general, its use leads to settlement troughs shallower and wider than those observed
experimentally (Bolton et al., 1994; Addenbrooke et al., 1997; Oettl et al., 1998; Man and Herle, 2005). To take
into account some of the fundamental aspects of the soil behaviour such as dilatancy before failure, variation of
modulus according to stress state, and a stiffness modulus in unloading different from that in loading, it is
necessary to use an elastoplastic model with at least isotropic hardening. In order to model the behavior of
overconsolidated clays during tunnel excavations, Boh et al. (2002); Man and Herle (2005) and Addenbrooke
et al. (1997) simulated the surrounding soil using different models with different complexity levels including
softening (post peak), non-linear pre-failure and anisotropy. They showed the importance of taking into account
the small-strain non-linearity of the soil in numerical modelling.
In this paper, constitutive models with increasing levels of complexity are adopted in order to simulate the
execution of a shallow tunnel in plane strain. First, we briefly describe the selected models that have been
calibrated on basis of the results of extensive in-situ and laboratory tests on an overconsolidated clay (Gasparre,
2005). The impact of a constitutive model for the surrounding soil on the prediction of soil displacements is
highlighted as well as the limits of the simplest constitutive model. As model performance is not only related to the
chosen model, but also to the input parameter used, some investigations are therefore presented on the
significance of different input parameters (for the most complicated model used in this paper) and on the potential
of default parameters.

2 Constitutive models
2.1 Mohr-Coulomb (MC)
To model the linear-elastic perfectly plastic behaviour, the MC model was used with its five parameters: elastic
parameters, E (Youngs modulus) and (Poissons ratio), and plastic parameters, (angle of friction), c
(cohesion) and (angle of dilatance).

3968

2.2 Mohr-Coulomb incorporating variable parameters with depth (MJanbu)


Based on engineering experience and physical observations on geotechnical structures, soil characteristics are
improved with depth. Implemented in the present paper within FLAC (Itasca, 1994), the MJanbu model permits to
use improved soil characteristics with depth. This improvement is considered in the MJanbu model by increasing
soil stiffness with depth. In order to formulate the constitutive model in generalized three-dimensional finite
difference code, the Janbus equation (1963) is slightly modified. The Youngs modulus takes therefore the
formulation
m

P'
E
= K E .
Pa
Pa

(1)

in which Pa, is the atmospheric pressure (used to insure that the constant KE is dimensionless), P, the mean
effective pressure, m and KE, are the model parameters. Consequently, the MJanbu model adds two parameters
to the MC model.

2.3 Fahey & Carter (1993)


Many authors have studied the behaviour of soils using high precision triaxial tests. They obtained a reversible
behaviour, high stiffness for strains less than 10-5 and showed that the shear modulus was constant under very
small-strain (strains between 10-6 and 10-5)
for granular materials: Bard (1993); Viggiani and Atkinson (1995) and Biarez et al. (1999);
for clays: Houlsby and Wroth (1991); Burland (1989) and Atkinson and Sallfors (1991).
This effect is referred to as small-strain stiffness. For strains higher than 10-5 a rapid drop of small-strain stiffness
is measured. Therefore, considering a constant modulus at higher strain level would not be an appropriate
approximation for a detailed analysis of settlements since strains around a tunnel (under its lifetime) are generally
between 10-4 and 10-2 (e.g. Al Hafez, 1997; Almeida e Sousa et al., 2001; Gomes Correia, 2004; Jardine et al.,
1986). The strain levels obtained here are far below those obtained by conventional laboratory testing, which
requires special measuring devices such as dynamic methods or local strain gauges.
The starting point for the non-linear model adopted in this paper (Fahey & Carter model) is so-called small strain
stiffness or initial tangent stiffness G0 (Gmax is also used in the literature for this value). This parameter may be
obtained from in situ measurement of seismic shear wave velocity, using a seismic CPT for example (Fahey,
1999). For uncemented cohesionless soils, the value of G0 proposed by Fahey and Carter depends (by the
intermediate of the parameter C) on the confining stress raised to some power

P'
G0
= C.
Pa
Pa

ng

(2)

In order to insure G0 always greater than a minimum value (equals to

C.Pa ), this equation was replaced by that

proposed by Bourgeois et al., 2005. It takes the form

< P' >


G0

= C.1 +
P
Pa
a

ng

(3)

where ng is a constant parameter and <P> is the positive value of P defined by the relation

< P' > =

P '+ P '
2

(4)

The stress-strain model used is that described by Fahey and Carter (1993). The shear stiffness of the model is
based on hyperbolic equations, which may be written in the form

max

Gt
=
g
G0


1 f (1 g )

max
f and g in this expression are model parameters, whereas, tmax is given by

3969

(5)

max =

3(P ' sin '+ c' cos ')


3 sin '

(6)

t is the half difference between the maximum and minimum principal stresses.

1 ' 3 '

(7)

The bulk modulus (K) depends mainly on the amount of confining stress (Naylor et al., 1981). Based on the
discussion of the K-G model by Naylor et al. (1981) the tangent bulk modulus Kt can be represented by the
following equation

K t = DS .(P') k .(Pa )
n

1 nk

(8)

where DS is a material constant that can be calculated from the initial values of bulk modulus and confining stress
and nk, can be taken as 0.5.
Figure 1 gives an indication of the range of shapes of modulus degradation curves (G/G0 - /max) that can be
obtained by using different values of f and g. Values of g less than 1 allow the G/G0 ratio reduce more rapidly with
increasing shear stress than occurs with the hyperbolic relationship. Conversely, values of g greater than 1 give a
slower rate of degradation initially, with high values of g giving an extended elastic. values of f greater than 1 have
no physical meaning.

Figure 1. Shear modulus degradation curves (Fahey, 1999).

2.4 Fahey & Carter with softening after the peak (FaheySoft)
The FaheySoft model is an extension of the Fahey & Carter model. All the features of the Fahey & Carter model
are therefore included in the FaheySoft model. FaheySoft constitutes a coupling between the Fahey & Carter and
the Strain hardening/softening model pre-implemented in FLAC. Once the plastic strains (eps) are reached, the
strength of the soil in Fahey & Carter starts to degrade. Figure 2 shows an example of this evolution for the angle
of friction ().

Figure 2. Hardening/softening behaviour (Itasca, 1994).

3 Parameter calibrations
The model calibration is based on various laboratory and on site tests. These tests were carried out in order to
characterize an overconsolidated clay (Gasparre, 2005). The selected samples were tested during the execution
of Terminal 5 of Heathrow airport (London). Investigation of the geological formation of this particular zone
allowed Gasparre (2005) to distinguish the following lithological units: a layer of gravel up to 2.5 m depth, below
which a layer of London Clay (in which three different lithological units were distinguished) has been observed. In
order to simplify the model, we took into account one layer with average value parameters. Numerical models of
drained triaxial tests corresponding to different depths were formulated in order to calibrate the parameters of
each model used for this study. The Fahey & Carter and FaheySoft parameters under small strain were obtained
from the results of bender element measurements. Figure 3 (a and b) shows an example of the calibration
results. Figure 3a shows an example of the stress-strain relationship whereas figure 3b explains the degradation
of the shear modulus according to the axial strain level. These figures show the numerical calibration results
which correspond to the experimental data. The simulation by MJanbu is not represented here because its
parameters (KE and m) were chosen to obtain identical results with the MC model. The calibrated parameters for
the different constitutive models of the two layers as well as the lining (assumed elastic) are shown in table 1.

3970

Table 1. Calibrated parameters for the different constitutive models.


Lining
(elastic)
25
0.2
6
510

Gravel
(MC)
17
0
35
0
0.3
3
5010

KE

m
DS
C
f
g
nk
ng

Deviatoric stress (kPa)

E, kN/m

Clay
(MJanbu)
19.5
7.6
22
4.14
0.22
-

Clay
(MC)
19.5
7.6
22
4.14
0.22
3
2710
-

Clay
(FaheySoft)
19.5
9.7
25.7
3.73
-

136.28

Clay
(Fahey & Carter)
19.5
7.6
22
5.7
-

0.65
-

300
200
0.5
1.97
0.5
0.55

300
200
0.5
1.97
0.5
0.55

450

40

400

35

350

30

300

25

250

20

Experiment (Gasparre, 2005)

200
150
100

MC

15

Fahey & Carter

10

FaheySoft

50

0
0

0,01

0,02

0,03
Axial strain

Shear modulus (MPa)

Parameter/
Unit
, kN/m3
2
c, kN/m
, ()
, ()

0,04

0,05

0,06

0,0001

0,0010

0,0100

0
0,1000

Axial strain

(b)

(a)

Figure 3. Parameter calibrations a) Stress-strain curves; a) Stiffness-strain curves.

4 Finite difference model


A plane strain model was created with the finite difference code FLAC2D in order to simulate the excavation of a
shallow tunnel in an overconsolidated clay (Fig. 4). The initial horizontal to vertical effective stress ratio (K0)
equals to 0.4 and 1.15 for the gravel and the clay, respectively. The extension of the mesh is 100 m wide and the
tunnel centre is located at H = 53.75 m above the substratum. The tunnel is assumed to be circular, its diameter
is D = 7.7 m. the cover height adopted for the reference calculation is 2D. As can be seen from figure 4, the
model is fixed in horizontal direction on both left and right sides. The bottom part of the mesh was pinned. All the
calibrations have been done under drained conditions. It is recognized that for a quantitative estimate of
deformations due to tunnelling in fine-grained soils, coupled consolidation analysis should be used. The drained
analysis however appears to be sufficient to evaluate long-term response in a qualitative manner.

73 m

100 m

Figure 4. model geometry and mesh.

3971

The numerical simulation of the excavation uses the procedures proposed by Panet (1995). It is thus performed
in two phases:
Deactivating the excavated ground and simultaneously applying a rate of stress relaxation (d) to the
excavation circumference;
Activating the support system and applying the total relaxation.
The value of d adapted for the reference study is 35%. Moreover, additional calculations were then carried out to
deduce the influence of various parameters:
The soil constitutive model (MC, MJanbu, Fahey & Carter and FaheySoft);
The various parameters of the Fahey & Carter model;
The cover height (variable from 1D to 5D).

5 Results and discussions


5.1 Reference calculation (H=2D and d = 35%)
The final surface settlement troughs and the normalised ones by the maximum surface settlement (over the
tunnel centreline) are presented on figure 5.

20

Offset from tunnel centreline (m)


40
60
80

100

0
Normalized surface settlement

Surface settlement (m)

0,000
-0,001
-0,002
-0,003
-0,004

MC
MJanbu
FaheySoft
Fahey & Carter

-0,005
-0,006
-0,007

20

Offset from tunnel centreline (m)


40
60
80

100

0,0
-0,2
-0,4
-0,6
-0,8
-1,0

(a)

(b)

Figure 5. Reference calculation a) Surface settlements; b) Normalized surface settlements.


The impact of constitutive model can be seen in the differences between the responses for movements (Fig. 5).
In spite of satisfactory wide mesh extension, surface settlements at 100 m from the left axis were observed when
using the MC model, which appears unrealistic. This problem can be explained by the fact that the stiffness
modulus in the MC model is constant, even in the small-strain zones.
No significant difference between the curves observed by the Fahey & Carter model and those of the FaheySoft
model, which means that the soil behavior undertook under elastic strains (small-strain of about 1.210-4 in our
case) and then only small plastic strains occurred. Despite its fairly simple formulation, the MJanbu model, which
implies stiffness depending on the depth, approached also the experimental observations (a deep and narrow
settlement trough, e.g. Addenbrooke et al., 1997 and Mair et al., 1993).
Normalizing the surface settlement troughs by the maximum settlement, one can observe that the MC model
shifts away from the results of other models which are in good agreement.
-0,012

-0,010

Horizontal displacement (m)


-0,008
-0,006
-0,004

-0,002

0,000
20

0
-10
MC
MJanbu
FaheySoft
Fahey & Carter

-20

Depth (m)

10

-30
-40
-50
-60

Figure 6. Reference calculation horizontal displacements.


Another difference between the predictions carried out by the MJanbu model and the other models adopting
stiffness as a function of stress-state of the soil is present from the horizontal displacements. Indeed, the MJanbu

3972

model provides inclinometer measurements close to those measured by the MC model (Fig. 6). Fahey & Carter
and FaheySoft models greatly limit these displacements (the inclinometer measurement by MC is about 3.4 times
that measured by the FaheySoft model).

5.2 Impact of covering depth


In order to evaluate the influence of the cover depth above a tunnel on the adjacent soil displacements,
simulations were carried out for d = 35% and a variable cover depth from 1D to 5D (Figs. 7). The settlement
presented on figure 7a corresponds to surface settlements over the tunnel centerline and the lateral
displacements (Fig. 7b) are those obtained at the tunnel centerline level (sec A-A, Fig. 4).
1

Covering depth (D)


3

-0,001
Horizontal displacement (m)

-0,002
Surface settlement (m)

Covering depth (D)


3

0,000

-0,003
-0,004
-0,005
-0,006

MC
MJanbu
Fahey & Carter
FaheySoft

-0,007
-0,008
-0,009

-0,005
-0,010
-0,015
-0,020
-0,025

(a)

(b)

Figure 7. Impact of covering height on a) surface settlements; b) horizontal displacements.


For the MC model, settlements increasing with depth were observed, which appears to be unrealistic. In fact, the
arch effects around a tunnel limit the propagation of movements towards the surface. Therefore the opposite
phenomenon should have been noticed. Contrary to the MC model, the other constitutive models predict
settlements that decrease with depth. These settlements became constant after 3D for Fahey & Carter and
FaheySoft models and 4D for MJanbu model.
Figure 7b shows lateral displacements at the tunnel centerline level (sec A-A, Fig. 4) for the different constitutive
models used in the present paper and a variable covering depth (1D to 5D). We obtained minimum horizontal
displacements near the surface, these displacements became greater in depth. The slope of the MC is the
highest, however, the other three curves are approximately parallel and maintaining a significant different of about
130% between MJanbu and the two other models.

6 Reliability of the Fahey & Carter constitutive model


In order to investigate the relative importance of the elastic parameters used in the Fahey & Carter and
FaheySoft models, an analysis was performed using first of all the calibrated parameters of these constitutive
models. Compared hereto are series of analyses where the value were individually varied by 25%.
Ref-25%

Ref+25%

Ref-25%

15

C
Ds

ng
10
nk
f

Ds
f

Ds

nk

nk

5
0

-10

ng
C

-5

g
C

-10

-20

ng

10

20
Variation of i (%)

Variation of maximum settlement (%)

Ref+25%
30

nk

ng
Ds

-15

(a)

(b)

Figure 8. Sensitivity analysis for the tunnel excavation a) maximum settlement; b) inflection point (i).

3973

The relative influence of parameter variation is presented in figure 8 as percentage of maximum settlement and
coordinate of the inflection point (i) variations for the reference calculation (H=2D and d = 35%). In figure 8b, the
inflection point is the parameter of the empirical surface settlement representing by the Gaussian equation:

y2
S = S max . exp 2
2i

(9)

Variation of maximum horizontal displacement (%)

where S, is the settlement of a point y from the tunnel centreline, Smax, is the maximum settlement.
Practically, parameters f and g do not have influence on the maximum surface settlement, but they lead to
decrease the value of the inflection points. Whereas, the parameters controlling the initial stiffness (especially the
parameters controlling shear stiffness: C and ng) have a significant influence on both: surface settlement and
inflection point.
Values of the coefficient DS smaller than the reference permit to decrease the inflection point as well as to
increase surface settlements. Moreover, shear stiffness parameters have the same effect on both surface
settlements and inflection point (a compromise must be done on what is more important between these two
parameters).
Ref+25%

30

Ref-25%

ng

20
10
f

Ds

nk

0
Ds

nk

-10
-20

ng

-30

Figure 9. Sensitivity analysis for the tunnel excavation horizontal displacement.


The impact of varying the different Fahey & Carter parameters on horizontal displacements can be seen on figure
9. This figure demonstrates that initial shear stiffness parameters have a strong effect on the prediction of the
horizontal displacement around an underground construction.

7 Conclusions and perspectives


The parametric analysis carried out in the present paper permits highlighting the influence of the choice of a
constitutive soil model on the simulation of an underground work. It appears to be unsatisfactory to simulate soils
by the constitutive model usually used (MC) when modeling underground structures. This can be overcome by
using stiffness-dependence on the depth for example (the MJanbu model) which gives better predictions of
surface settlement trough. Moreover, it was clearly shown in this article that displacements around a tunnel are
strongly influenced by soil parameters under very small-strain (C, ng, Ds and nk in the Fahey & Carter constitutive
model). Therefore, the use of a soil model adapted to the excavation problem of the underground work (including
non-linearity pre-failure and high stiffness under very small-strain) considerably improves the prediction of
displacements around this construction.
Further complementary calculations should be done to study the influence of the soil relaxation rate and other
constitutive models derived from the Cam-Clay model. Moreover, further comparison with experimental site
results obtained from a real excavation would permits to highlight the importance phenomena to be taken into
account during tunneling simulations.

8 References
Addenbrooke T.I., Potts D.M., Puzrin A.M. (1997) The influence of pre-failure soil stiffness on the numerical analysis of tunnel
construction. Gotechnique 47-3, 693-712
Al Hafez T. (1997) Proprits gotechniques des terrains du tunnel sous la Manche. Interprtation des mesures de
dformations du revtement avec calculs comparatifs. Thse de doctorat, Ecole Centrale de Paris, France, 193p
Almeida e Sousa J., Marques F., Lemos L.L. (2001) Tnel de trmino da Estao Alameda II. Anlise do comportamento.
Revista Portuguesa de Geotecnia, SPG 93, 5-32
Atkinson J.H., Sallfors G. (1991) Experimental determination of soil properties. Proc. 10th ECSMFE, Florence, volume 3, 915956
Bard E. (1993) Comportement des matriaux granulaires secs et avec liant hydrocarbon. Thse de doctorat, Ecole Centrale

3974

de Paris, France
Biarez J., Bougriou Z., Fayad T., Hammoud I., Liu W., Gomes Correia A. (1999) Les modules de 10-5 10-1 pour les sols
remanis et non remanis, pour les fondations des voies ferres et les routes. Xth ECSMGE Geotechnical Engineering
for Transportation Infrastructure, The Netherlands, volume 3, 1737-1742
Boh J., Herle I., Man D. (2002) Stress and strain dependent stiffness in a numerical model of a tunnel. Proc. 2nd Int.
Conference on Soil Structure Interaction in Urban Civil Engineering. Zurich, Switzerland, 357-364
Bolton M.D., Dasari G.R., Britto A.M. (1994) Putting small-strain non-linearity into Modified Cam Clay model. Proc. 8th Int. Conf.
on Computer Methods and Advances in Geomechanics, Morgantown, West Virginia, 537-542
Bourgeois E., Coquillay S., Mestat Ph. (2005) Exemples dutilisation dun modle lastoplastique avec lasticit non linaire
pour la modlisation douvrages gotechniques. Bulletin des laboratiores des Ponts et Chausses, Juillet-Aout-Septembre,
Rf. 4551, 67-84
Burland J.B. (1989) Small is beautiful the stiffness of soils at small strains, Ninth Laurits Bjerrum memorial lecture. Canadian
Geotechnical Journal 26, 499-516
Dias D. (1999) Renforcement du front de taille des tunnels par boulonnage : Etude numrique et application un cas rel en
site urbain. PhD Thesis, INSA de Lyon, France, 320p
Fahey M. (1999) Soil stiffness values for foundation settlement analysis. Proc. 2nd Int. Conf. on Pre-Failure Deformation
Characteristics of Geomaterials. Torino. Vol. 2, 1325-1332
Fahey M., Carter J.P. (1993) A finite element study of the pressuremeter test in sand using a nonlinear elastic plastic model.
Canadian Geotechnical Journal. Vol. 30, 348-362
Gasparre A. (2005) Advanced laboratory characterisation of London clay. PhD Thesis, Imperial College London, 598p
Gomes Correia A. (2004) Caractersticas de deformabilidade dos solos que interessam funcionalidade das estruturas.
Revista Portuguesa de Geotecnia, SPG 100, 103-122
Houlsby G.T., Wroth C.P. (1991) The variation of shear modulus of a clay with pressure and overconsolidated ratio, Soil and
Foundations 31 (3), 138-143
Itasca Consulting group (1994) FLAC Users manual
Janbu N. (1963) Soil compressibility as determined by oedometer and triaxial tests. European conf. on soil mechanics and
foundation engineering. Wiesbaden, Germany, Vol. 1, 19-25
Jardine R.J., Potts D.M., Fourie A.B., Burland J.B. (1986) Studies of the influence of non-linear stress strain characteristics in
soil-structure interaction. Gotechnique 36 (3), 377-396
Mair R.J., Taylor R.N., Bracegirdle (1993) Subsurface settlement profiles above tunnels in clay. Gotechnique, 43-2, 315-320.
Man D., Herle I. (2005) Numerical analyses of a tunnel in London clay using different constitutive models. Proc. 5th Int.
Symposium TC28 Geotechnical Aspects of Underground Construction in Soft Ground, Amsterdam, The Netherlands, 595600.
Mroueh H. (1998) Tunnels en site urbain : modlisation numrique et interaction creusement, ouvrage existants. Thse de
doctorat, Universit des Sciences et Technologies de Lille, Lille, France
Naylor D.J., Pande G.N., Simpson B., Tabb R. (1981) Finite Elements in Geotechnical Engineering, Pineridge Press, Swansea,
Wales, UK
Oreilly M.P., New B.M. (1982) Settlements above tunnels in the United Kingdom their magnitudes and prediction. Tunnelling
82, London, IMM, 173-181
Oettl G., Strak R.F., Hofstetter G. (1998) A comparison of elastic-plastic soil models for 2D FE analyses of tunnelling.
Computers and Geotechnics, 23, 19-38
Oteo C.S., Sagaseta C. (1982) Prediction of settlements due to underground openings. Int. Symp. On numerical Models in
Geomechanics, Zurich, 653-659
Panet M. (1995) Le calcul des tunnels par la mthode convergence-confinement. Presses de lENPC, Paris
Peck R.B. (1969) Deep excavation and tunnelling in soft ground, State of the art report. 7th Int. Conference on Soil Mechanics
and Foundation Engineering, Mexico, 225-290
Sagaseta C. (1987) Evaluation of surface movements above tunnels, a new approach. Colloque interaction sol/structure, Paris,
Press ENPC 445-452
Swoboda G., Mertz W., Schmid A. (1989) Three dimensional numerical models to simulate tunnel excavation. Numerical
Models in Geomechanics NUMOG III. Elsevier. 581-586
Viggiani G., Atkinson J.H. (1995) Stiffness of fine-grained soil at very small strains. Gotechnique 45 (2), 249-265

3975

S-ar putea să vă placă și