Numerical modelling of the strain
localization in soils and rocks
Collin F., Levasseur S., B. Pardoen, F. Salehnia
R. Chambon, D. Caillerie, P. Bésuelle
P. Kotronis, G. Jouan, M. Soufflet
INTRODUCTION
Shear banding occurs frequently (at many scales) and is the source of
many soil and rock engineering problems:
natural or human-made slopes or excavations, unstable rock masses,
embankments or dams, tunnels and mine galleries, boreholes driven for
oil production, repositories for nuclear waste disposal
Failure in soils and rocks is almost always associated with fractures and/or shear bands developing in the geomaterial.
Introduction Experiment Theory Numerical Conclusions
In geomaterials, the understanding of failure processes is more complex by the fact that soils and rocks are multiphase porous materials where different multiphysical processes take place.
2
INTRODUCTION
In situ observation of shear banding
In situ observations of shear banding and/or faulting are made frequently
at many scales
Large scale: railway tracks after an earthquake in Turkey
Introduction Experiment Theory Numerical Conclusions 3
In situ observation of shear banding
Bierset (Belgium) 1998 – Courtesy C. Schroeder
Human-made slope along E42 exit road
Introduction Experiment Theory Numerical Conclusions 4
In situ observation of shear banding
Fractures observed during the construction of the connecting gallery at the URL in Mol. Vertical cross section through the gallery showing the fracturation pattern around it, as deduced from the observations (from Alheid et al. 2005)
Nuclear waste disposal
Introduction Experiment Theory Numerical Conclusions 5
Outline:
Introduction
Experimental observations
Theoretical tools
Numerical models
Conclusions
6
Experimental observations
Introduction Experiment Theory Numerical Conclusions
• Biaxial test
• Axisymetric triaxial test
• True triaxial test
To better understand the development of the shear band, experiments are necessary, which are no more element tests as far as the behavior becomes heterogeneous.
Different teams have performed experimental works devoted to the study of strain localization:
• Desrues and co-workers
• Finno and co-workers
• Vardoulakis and co-workers
• …
7
Experimental observations: triaxial test
Triaxial test:
In triaxial tests (and more generally in axi-symmetric tests), the localization zone may remain more or less hidden inside the sample (need for special techniques to see the process)
Introduction Experiment Theory Numerical Conclusions 8
Experimental observations: triaxial test
Localized rupture in sandstone samples under different confining pressures (Bésuelle et al., 2000)
Introduction Experiment Theory Numerical Conclusions 9
Experimental characterisation of the localisation phenomenon inside a Vosges sandstone in a triaxial cell
P. BESUELLE, J. DESRUES, S. RAYNAUD, International Journal of Rock Mechanics & Mining Sciences 37 (2000) p. 1223-1237
Experimental observations: triaxial tests
Desrues, J. et al. (1996). Géotechnique 46, No. 3, 529–546
Tomodensitometry:
Localization pattern observed in sand sample during axisymetric triaxial test
Introduction Experiment Theory Numerical Conclusions 10
Experimental observations: triaxial tests
Tomodensitometry:
Localization pattern observed in sand sample during axisymetric triaxial test
Introduction Experiment Theory Numerical Conclusions 11
Experimental observations: triaxial test
Increment 4-5
3D digital image correlation applied to X-ray micro tomography images from triaxial compression tests on argillaceous rock LENOIR N , Bornert M, DESRUES J, BESUELLE P, VIGGIANI G Strain vol:43 No 3 pp.193-205
Introduction Experiment Theory Numerical Conclusions 12
Experimental observations: biaxial test
Biaxial test:
As in triaxial tests (and more generally in axi-symmetric tests), the localization zone may remain more or less hidden inside the sample, most of the experimental campaigns on localization have been performed in biaxial apparatus
Introduction Experiment Theory Numerical Conclusions 13
s1 s1
e , s33
Experimental observations: biaxial test
Introduction Experiment Theory Numerical Conclusions
Experimental
set-up
&
a typical test
14
Experimental observations: biaxial test
Localization and Peak
1-2 2-3 3-4 4-5 5-6
Introduction Experiment Theory Numerical Conclusions 15
Outline:
Introduction
Experimental observations
Theoretical tools
Numerical models
Conclusions
16
Theoretical concepts
Experimental evidence:
Introduction Experiment Theory Numerical Conclusions
Initial state Homogeneous strain field Localized strain field
17
Theoretical concepts
Theoretical background
Introduction Experiment Theory Numerical Conclusions
Following the previous works by (Hadamard, 1903), (Hill, 1958) and (Mandel, 1966), Rice and co-workers (Rice, 1976, Rudnicki et al., 1975) have proposed the so-called Rice criterion.
18
Theoretical concepts
Introduction Experiment Theory Numerical Conclusions
1 00n s s
:C Ls
1 0 0 0: ( ) : 0n C L g n C L
det( ) 0nCn
19
Static condition:
Kinematic condition:
Constitutive law:
When it is assumed that C 1=C 0=C , no trivial solution if and only if:
LLL 01
ngLL 01
i
j
j
i
x
u
x
uL
2
1
Theoretical concepts
Introduction Experiment Theory Numerical Conclusions 20
11 11 12 11
22 21 22 22
12 12 12
0
0
0 0 2
C C L
C C L
G L
s
s
s
1 00ij ij jns s
1 0 1 0
11 11 1 12 12 2
1 0 1 0
21 21 1 22 22 2
0
0
n n
n n
s s s s
s s s s
1 0
ij ij i jL g nL
If C 1=C 0=C :
1 2
1 2
11 1 1 12 2 2 12 1 2 2 1
12 1 2 2 1 21 1 1 22 2 2
0
0
C C n n
n C C n
g n g n G g n g n
G g n g n g n g n
Static condition:
Constitutive law in principal axis:
Kinematic condition:
Combining the three previous relationship yields:
Theoretical concepts
Introduction Experiment Theory Numerical Conclusions 21
4 2
1 3 5
4
1 0a z a z an
4 3 2
1 2 3 4 5
4
1 0a z a z a z a z an
2
1 2
2 2
11 1 12 2 1 12 1 2 12 2 1
2 2
21 1 2 12 2 1 22 2 12 1
0
0
C C
C g C
n G n g n n G n n g
n n G n n n G n g
det( ) 0nCn When it is assumed that C 1=C 0=C , no trivial solution if and only if:
4 4 2 22
11 12 1 22 12 2 11 22 12 12 12 1 22 0C G n C G n C C C G C n n
For a constitutive law written in cartesian axis:
2
1
nz
n
Theoretical concepts
Introduction Experiment Theory Numerical Conclusions
ni
qf
2dB2dB
2dB
s1s
1
s2
s2
s1s1
s2
s2
22
Extension to multiphysical context, mainly in hydro mechanical coupling:
Loret and co-workers (Loret et al., 1991) showed that for hydromechanical problems the condition of localization depends only on the drained properties of the medium In coupled problems much more complex localization pattern can be obtained, at least theoretically (Vardoulakis, 1996)
Theoretical concepts
Introduction Experiment Theory Numerical Conclusions 23
Which information can provide these theoretical tools ?
For element test, the tools allow you to check if and when the constitutive model is able to predict the localization direction observed at the laboratory. For boundary value problems, they provide you the stress state when bifucation may arise and the direction of potential bifurcation (fracturation). Be aware that the Rice criterion is a local one ! This criterion could be activated for any constitutive model, if you make the connection in the ELEMB2 and POSPEC routines and some additional state variables have to be defined.
Example of EDZ around a cavity
Introduction Experiment Theory Numerical Conclusions 24
Skeleton mechanical behaviour
Linear elasticity : E0 et n0
Associated softening plasticity (decrease of cohesion) :
Drucker Prager criterion : 0tan
3
2
3ˆ
ss
cImIIF
sin3
sin2
m c = c0 f(g p)
pR
p
pR
p
pR
pp
si
sif
gg
ggg
gg
2
2
0)1(1)(
Theoretical concepts
Introduction Experiment Theory Numerical Conclusions 25
Softening behaviour : localization effects are very important
0.0
0.5
1.0
1.5
2.0
2.5
3.0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5
Vertical displacement [mm]
Axia
l lo
ad [M
N]
Sample with defect
Perfect sample
NL in the global curve
Bifurcation
Theoretical concepts
Introduction Experiment Theory Numerical Conclusions 26
Theoretical concepts
Introduction Experiment Theory Numerical Conclusions 27
Softening behaviour : localization effects are very important
Bifurcation analysis thanks to the Rice criterion (Acoustic tensor)
4 4 3 2
1 1 2 3 4 5det ( ) 0n n a z a z a z a z a
Theoretical concepts
Introduction Experiment Theory Numerical Conclusions 28
Plastic point Bifurcation dir. Bifurcation cones
Theoretical concepts
Introduction Experiment Theory Numerical Conclusions 29
Plastic point Bifurcation dir. Bifurcation cones
Theoretical concepts
Introduction Experiment Theory Numerical Conclusions 30
The Rice criterion provides us the information on when and how localization may appear. Do we have any problem to model such phenomenon with classical finite element method ? Let’s consider the modelling of a biaxial with a defect triggering the localization, first without any hydromechanical effect.
Bottom-left defect
Smooth and rigid boundary
Theoretical concepts
Introduction Experiment Theory Numerical Conclusions 31
50 elements 200 elements 300 elements
The post peak behaviour depends on the mesh size !
Example of EDZ around a cavity
Introduction Experiment Theory Numerical Conclusions 32
Cylindrical cavity without retaining
Anisotropic initial state of stress
Geometrical dimensions : Internal radius 3 m
Mesh length 60 m
Choice :
Symetry of the problem is assumed
894 elements – 2647 nodes – 7941 dof
Let’s consider now a coupled modelling:
Example of EDZ around a cavity
Introduction Experiment Theory Numerical Conclusions 33
MPap
MPa
MPa
w
yy
zzxx
7.4
64.11
74.7
'
''
s
ss
MPap
MPa
MPa
w
yy
xx
7.4
4.15
5.11
s
s
'
'
0
11.5 1
15.4 1
4.7 1
0
xx xx rw w
yy yy rw w
w
xx yy w
t T
tbS p MPa
T
tbS p MPa
T
tp MPa
T
t T
p
s s
s s
s s
T = 1.5 Ms (17 jours)
ttotal = 300 Ms (9.5 ans)
246 él.
Example of EDZ around a cavity
Introduction Experiment Theory Numerical Conclusions 34
Coupled modelling – Comparison Coarse mesh / Refined mesh
Deviatoric strains
Example of EDZ around a cavity
Introduction Experiment Theory Numerical Conclusions 35
• Localization study : Acoustic tensor determinent
• Mesh dependency of the results for classical FE
• Non-uniqueness of the results in both cases
The numerical modelling of strain localization with classical FE is not adequate. We need another numerical model to fix this mesh dependency problem !
Outline:
Introduction
Experimental observations
Theoretical tools
Numerical models
Conclusions
36
Numerical models
Introduction Experiment Theory Numerical Conclusions 37
• Classical FE formulation: mesh dependency
• Different regularization methods
Gradient plasticity Non-local approach Microstructure continuum Cosserat model Second gradient local model
Mainly for monophasic materials !
Enrichment of the law
Enrichment of the kinematics
Numerical models
Introduction Experiment Theory Numerical Conclusions 38
In second gradient model, the continuum is enriched with microstructure effects. The kinematics include therefore the classical one but also microkinematics (See Germain 1973, Toupin 1962, Mindlin 1964).
Let us define first the classical kinematics:
Numerical models
Introduction Experiment Theory Numerical Conclusions 39
Here is the enrichment:
Numerical models
Introduction Experiment Theory Numerical Conclusions 40
• The internal virtual work (Germain, 1973)
• The external virtual work (simplified)
• The virtual work equations can be extended to large strain problems
Numerical models
Introduction Experiment Theory Numerical Conclusions 41
• Balance equations
• Boundary conditions
written in the current configuration
Three constitutive equations needed !
• Local second gradient models: we add the kinematical constraint:
this implies:
the virtual work equation reads
Numerical models
Introduction Experiment Theory Numerical Conclusions 42
Numerical models
Introduction Experiment Theory Numerical Conclusions 43
• Local second gradient models
balance equations
boundary conditions
Numerical models
Introduction Experiment Theory Numerical Conclusions 44
How do we introduce an internal length scale in second grade model ?
Let’s take a simple example of a 1D-bar in traction:
Numerical models
Introduction Experiment Theory Numerical Conclusions 45
if u’<elim
if u’>elim
General differential equation of the problem
Where N1= N - M’ = Cst, K = Cst, A = A1 if u’ <elim and A = A2 if u’ >elim
General solution of the problem
Numerical models
Introduction Experiment Theory Numerical Conclusions 46
Let’s take another example: thick-walled cylinder problem (elastic second gradient model)
General differential equation of the problem
Balance equation
General solution
Numerical models
Introduction Experiment Theory Numerical Conclusions 47
• Local second gradient model : additional assumption * *
ij ijv F
* 2 *
*i i
ij ijk ext
j j k
u ud W
x x xs
Introduction of Lagrange multiplier field :
** *
* *iji i
ij ijk ij ij ext
j k j
vu ud v d W
x x xs
* 0i
ij ij
j
uv d
x
Local quantities
Finite element formulation of a second grade model
Numerical models
Introduction Experiment Theory Numerical Conclusions 48
Local Second gradient Finite element
Numerical models
Introduction Experiment Theory Numerical Conclusions 49
• Biaxial compression test
Strain rate : 0.18% / hour
No lateral confinement
Globally drained (upper and lower drainage)
Bottom-left defect
Smooth and rigid boundary
Numerical models
Introduction Experiment Theory Numerical Conclusions 50
• First gradient law :
E = 5800 MPa n = 0,3
= 25° Y = 25°
Linear elasticity : E0 and n0
Associated softening plasticity (decrease of cohesion) :
Drucker Prager criterion : 0tan
3
2
3ˆ
ss
cImIIF
sin3
sin2
m c = c0 f(g p)
pR
p
pR
p
pR
pp
si
sif
gg
ggg
gg
2
2
0)1(1)(
c0 = 1 MPa = 0,01 gR = 0,015
Numerical models
Introduction Experiment Theory Numerical Conclusions 51
• Second gradient law : Linear relationship deduced from Mindlin
D = 20 kN
Numerical models
Introduction Experiment Theory Numerical Conclusions 52
First modelling: no HM coupling (no overpressure)
Before After
Bifurcation directions (Regularization : Second gradient)
Numerical models
Introduction Experiment Theory Numerical Conclusions 53
Before After
Plastic loading point
First modelling: no HM coupling (no overpressure)
(Regularization : Second gradient)
Numerical models
Introduction Experiment Theory Numerical Conclusions 54
First modelling: no HM coupling (no overpressure)
Before After
Velocitiy field (Regularization : Second gradient)
Numerical models
Introduction Experiment Theory Numerical Conclusions 55
Initiation of localization (Directional research)
Numerical models
Introduction Experiment Theory Numerical Conclusions 56
Non uniqueness of the solution
Initiation of localization (Directional research)
(Regularization : Second gradient)
Numerical models
Introduction Experiment Theory Numerical Conclusions 57
Non uniqueness of the solution
Initiation of localization (Directional research)
(Regularization : Second gradient)
Numerical models
Introduction Experiment Theory Numerical Conclusions 58
Non uniqueness of the solution
Initiation of localization (Directional research)
(Regularization : Second gradient)
Sieffert et al., 2009
Numerical models
Introduction Experiment Theory Numerical Conclusions 59
• Main assumptions
– Quasi static motion
– Fully saturated
– Incompressible solid grains
• Aims
– Equations written in the spatial configuration
– Full Newton Raphson method
Our goal is to extend the second gradient formulation for multiphysics conditions. In the following, we focus on the hydromechanical model but the same procedure can be applied for TM, THM or THMC problems.
Numerical models
Introduction Experiment Theory Numerical Conclusions 60
• Classical poromechanics field equations
Saturated porous medium
Balance of linear momentum for the mixture
* * *
ij ij mix i i i id g u d t u ds e
ij j in ts
'
ij ij ijps s
Boundary condition
Terzaghi’s postulate
Numerical models
Introduction Experiment Theory Numerical Conclusions 61
• Classical poromechanics field equations
Fluid mass balance
*
* * *
i
i
pM p m d Q p d q p d
x
( )i w w i
i
pm g
x
w ww
pM
k
i iq m nBoundary condition
Darcy’s law
Storage law
Numerical models
Introduction Experiment Theory Numerical Conclusions 62
• Classical poromechanics field equations
Balance of momentum for the fluid phase
Mass balance equation for the solid
Viscous drag force :
Numerical models
Introduction Experiment Theory Numerical Conclusions 63
• Coupled local second gradient model
Second gradient effects are assumed only for solid phase
For the mixture, there are stresses which obey the Terzaghi postulate and double stresses which are only the one of the solid phase
Boundary conditions for the mixture are enriched
Numerical models
Introduction Experiment Theory Numerical Conclusions 64
• Coupled local second gradient model
* 2 *
* * *i iiij ijk mix i i i i i
j j k
u ud g u d t u T Du d
x x xs
*
* * *
i
i
pM p m d Q p d q p d
x
Numerical models
Introduction Experiment Theory Numerical Conclusions 65
• Coupled local second gradient model
** *
*
* * *
iji i
ij ijk ij ij
j k j
imix i i i i i
vu ud v d
x x x
g u d t u T Du d
s
* 0i
ij ij
j
uv d
x
*
* * *
i
i
pM p m d Q p d q p d
x
Numerical models
Introduction Experiment Theory Numerical Conclusions 66
Equations are assumed to be met at time t
We are looking for the values of the different fields at time: t+t=t1
using a full Newton Raphson method and an implicit scheme for the rate :
Finite element formulation of the coupled local second gradient model
Numerical models
Introduction Experiment Theory Numerical Conclusions 67
• Field equations at time t+t
R, S and W : Residuals of the balance equations
Numerical models
Introduction Experiment Theory Numerical Conclusions 68
•Linearization of field equations Auxiliary linear problem
*
( , ) ( , )
T
x y x yU E dU d R S W
R, S and W : Residuals of the balance equations
1 1 2 2
( , ) 1 2
1 2 1 2 1 2
11 11 12 22
11 22 11 22
1 2 1 2
x y
du du du du dp dpdU du du dp
x x x x x x
dv dv dv dvdv dv d d
x x x x
Numerical models
Introduction Experiment Theory Numerical Conclusions 69
(4 4) (4 2) (4 8) (4 4) (4 4)(4 3)
(2 4) (2 2) (2 3) (2 8) (2 4) (2 4)
(3 2) (3 8) (3 4) (3 4)(3 4) (3 3)
(8 4) (8 2) (8 3) (8 8) (8 4) (8 4)
(4 4) (4 2) (4 3) (4 8) (4 4) (4 4)
(4 4
1 0 0 0
1 0 2 0 0 0
0 0 0 0
2 0 0 0 0
3 0 0 0 0
4
x x WM x x xx
x x x x x x
MW x WW x x xx x
x x x x x x
x x x x x x
x
E K I
G G
K KE
E D
E I
E
) (4 2) (4 3) (4 8) (4 4) (4 4)0 0 0 0
x x x x xI
E1, E2, E3, E4 and D : see monophasic local sec. Gradient model
G1 and G2 : related to gravity volume force
KWW : Classical flow matrix
KMW and KWM : Coupling terms including large strain effect
Numerical models
Introduction Experiment Theory Numerical Conclusions 70
Isoparametric Finite Element :
8 nodes for macro-displacement and pressure field 4 nodes for microkinetic gradient field 1 node for Lagrange multipliers field
Numerical models
Introduction Experiment Theory Numerical Conclusions 71
1 1
*
1 1
*
detT T T
node node
T
node node
U B T E T B J d d dU
U k dU
•FE element discretization of linear auxiliary problem
Local stiffness matrix
1 1
* *
1 1
*
detT T T
ext node
T
node HE
R S W P U B T J d d
U f
s
Elementary out of balance forces
Numerical models
Introduction Experiment Theory Numerical Conclusions 72
• Biaxial compression test
Strain rate : 0.18% / hour
No lateral confinement
Globally drained (upper and lower drainage)
Bottom-left defect
Smooth and rigid boundary
Numerical models
Introduction Experiment Theory Numerical Conclusions 73
• Second gradient law : Linear relationship deduced from Mindlin
• Flow model parameters
D = 20 kN
= 10-19 / 10-12 m2
w= 1000 kg/m³ = 0.15
kw = 510-10 Pa-1
w = 0.001 Pa.s
Numerical models
Introduction Experiment Theory Numerical Conclusions 74
(20 x 10) (30 x 15) (40 x 20)
•Equivalent strain after 0.2 % of axial strain ( = 10-12 m²)
Second modelling: HM coupling
Numerical models
Introduction Experiment Theory Numerical Conclusions 75
•Plastic loading point after 0.2 % of axial strain ( = 10-12 m²)
(20 x 10) (30 x 15) (40 x 20)
Numerical models
Introduction Experiment Theory Numerical Conclusions 76
•Fluid flow after 0.2 % of axial strain ( = 10-12 m²)
Numerical models
Introduction Experiment Theory Numerical Conclusions 77
•Load-displacement curve ( = 10-12 m²)
Numerical models
Introduction Experiment Theory Numerical Conclusions 78
•Load-displacement curve ( = 10-19 m²)
‘Undrained’ behaviour
Second modelling: HM coupling
Numerical models
Introduction Experiment Theory Numerical Conclusions 79
For = 10 -19 m², the behaviour is undrained, we recover the
experimental observation showing that for dilatant material, no localization is possible before cavitation.
Example of EDZ around a cavity
Introduction Experiment Theory Numerical Conclusions 80
MPap
MPa
MPa
w
yy
zzxx
7.4
64.11
74.7
'
''
s
ss
MPap
MPa
MPa
w
yy
xx
7.4
4.15
5.11
s
s
'
'
0
11.5 1
15.4 1
4.7 1
0
xx xx rw w
yy yy rw w
w
xx yy w
t T
tbS p MPa
T
tbS p MPa
T
tp MPa
T
t T
p
s s
s s
s s
T = 1.5 Ms (17 jours)
ttotal = 300 Ms (9.5 ans)
246 él.
Example of EDZ around a cavity
Introduction Experiment Theory Numerical Conclusions 81
Coupled modelling – Comparison Coarse mesh - Refined mesh
Deviatoric strains
Classical FE formulation
Example of EDZ around a cavity
Introduction Experiment Theory Numerical Conclusions 82
Coupled modelling – Comparison Coarse mesh - Refined mesh
Deviatoric strains
Coupled second gradient FE formulation
Example of EDZ around a cavity
Introduction Experiment Theory Numerical Conclusions 83
Coupled modelling
Coupled second gradient FE formulation
François et al., 2012
Example of EDZ around a cavity
Introduction Experiment Theory Numerical Conclusions 84
Coupled modelling
Coupled second gradient FE formulation
François et al., 2012
Example of EDZ around a cavity
Introduction Experiment Theory Numerical Conclusions 85
Coupled modelling
Coupled second gradient FE formulation
François et al., 2012
Example of EDZ around a cavity
Introduction Experiment Theory Numerical Conclusions 86
Coupled modelling
Coupled second gradient FE formulation
François et al., 2012
Example of EDZ around a cavity
Introduction Experiment Theory Numerical Conclusions 87
Coupled modelling
Coupled second gradient FE formulation
François et al., 2012
Outline:
Introduction
Experimental observations
Theoretical tools
Numerical models
Conclusions
88
Conclusions
Introduction Experiment Theory Numerical Conclusions 89
Strain localization in shear band mode can be observed in most laboratory tests leading to rupture in geomaterials.
Complex localization patterns may be the result of specific geometrical or loading conditions.
The numerical modelling of strain localization with classical FE is not adequate. Enhanced models are needed for a robust modelling of the post peak behaviour.
Many experimental works and numerical developments are necessary to improve the prediction of failure in boundary value problems