Professional Documents
Culture Documents
FTD
FTD
FTD
1.1 Introduction
Numerical solution schemes face several difficulties when implementing constitutive models to
represent geomechanical material behavior. There are three characteristics of geo-materials that
cause particular problems. One is physical instability. Physical instability can occur in a material
if there is the potential for softening behavior when the material fails. When physical instability
occurs, part of the material accelerates and stored energy is released as kinetic energy. Numerical
solution schemes often have difficulties at this stage because the solution may fail to converge when
a physical instability arises.
A second characteristic is the path dependency of nonlinear materials. In most geomechanical
systems, there are an infinite number of solutions that satisfy the equilibrium, compatibility and
constitutive relations that describe the system. A path must be specified for a “correct” solution to
be found. For example, if an excavation is made suddenly (e.g., by explosion), then the solution
may be influenced by inertial effects that introduce additional failure of the material. This may
not be seen if the excavation is made gradually. The numerical solution scheme should be able to
accommodate different loading paths in order to properly apply the constitutive model.
A third characteristic is the nonlinearity of the stress-strain response. This includes the nonlinear
dependence of both the elastic stiffness and the strength envelope on the confining stress. This can
also include behavior after ultimate failure that changes character according to the stress level (e.g.,
different post-failure response in the tensile, unconfined and confined regimes). The numerical
scheme needs to be able to accommodate these various forms of nonlinearity.
The difficulties faced in numerical simulations in geomechanics – physical instability, path depen-
dence, and implementation of extremely nonlinear constitutive models – can all be addressed by
using the explicit, dynamic solution scheme provided in FLAC 3D. This scheme allows the numerical
analysis to follow the evolution of a geologic system in a realistic manner, without concerns about
numerical instability problems. In the explicit, dynamic solution scheme, the full dynamic equa-
tions of motion are included in the formulation. By using this approach, the numerical solution is
stable even when the physical system being modeled is unstable. With nonlinear materials, there is
always the possibility of physical instability (e.g., the sudden collapse of a slope). In real life, some
of the strain energy in the system is converted into kinetic energy, which then radiates away from
the source and dissipates. The explicit, dynamic solution approach models this process directly,
because inertial terms are included – kinetic energy is generated and dissipated.
In contrast, schemes that do not include inertial terms must use some numerical procedure to treat
physical instabilities. Even if the procedure is successful at preventing numerical instability, the
path taken may not be a realistic one. The numerical scheme should not be viewed as a black
box that will give “the solution.” The way the system evolves physically can affect the solution.
The explicit, dynamic solution scheme can follow the physical path. By including the full law of
motion, this scheme can evaluate the effect of the loading path on the constitutive response.
The explicit, dynamic solution scheme also allows the implementation of strongly nonlinear consti-
tutive models because the general calculation sequence allows the field quantities (forces/stresses
and velocities/displacements) at each element in the model to be physically isolated from one
another during one calculation step. The general calculation sequence of FLAC 3D is described in
Section 1.1 in Theory and Background. The implementation of elastic/plastic constitutive models
within the framework of this scheme is discussed in Section 1.3.
The mechanical constitutive models available in FLAC 3D range from linearly elastic models to
highly nonlinear plastic models. The basic constitutive models are listed below. A short discussion
of the theoretical background and simple example tests for each model follow this listing.
There are 15 basic mechanical constitutive models provided in FLAC 3D Version 5.0, arranged into
null, elastic and plastic model groups:
Null model group
(1) null model
A null material model is used to represent material that is removed or excavated.
(See Section 1.4.1.)
Elastic model group
(2) elastic, isotropic model
The elastic, isotropic model provides the simplest representation of material
behavior. This model is valid for homogeneous, isotropic, continuous materi-
als that exhibit linear stress-strain behavior with no hysteresis on unloading.
(See Section 1.5.1.)
(3) elastic, orthotropic model
The elastic, orthotropic model represents material with three mutually perpen-
dicular planes of elastic symmetry. For example, this model may simulate
columnar basalt loaded below its strength limit. (See Section 1.5.2.)
(4) elastic, transversely isotropic model
The elastic, transversely isotropic model gives the ability to simulate layered
elastic media in which there are distinctly different elastic moduli in directions
normal and parallel to the layers. (See Section 1.5.3.)
Plastic model group
(5) Drucker-Prager model
The Drucker-Prager plasticity model may be useful to model soft clays with
low friction angles. However, this model is not generally recommended for
application to geologic materials. It is included here mainly to permit com-
parison with other numerical program results. (See Section 1.6.1.)
(6) Mohr-Coulomb model
The Mohr-Coulomb model is the conventional model used to represent shear
failure in soils and rocks. Vermeer and de Borst (1984), for example, report
laboratory test results for sand and concrete that match well with the Mohr-
Coulomb criterion. (See Section 1.6.2.)
All constitutive models in FLAC 3D share the same incremental numerical algorithm. Given the stress
state at time t, and the total strain increment for a timestep, t, the purpose is to determine the
corresponding stress increment and the new stress state at time t + t. When plastic deformations
are involved, only the elastic part of the strain increment will contribute to the stress increment. In
this case, a correction must be made to the elastic stress increment as computed from the total strain
increment in order to obtain the actual stress state for the new timestep.
Note that all models in FLAC 3D operate on effective stresses only; pore pressures are used to convert
total stresses to effective stresses before the constitutive model is called. The reverse process occurs
after the model calculations are complete.
All stress increments in this section are corotational stress increments. For simplicity of notation,
however, the superscriptˇwill not be used. The stresses at time t + t are computed as “new stress
values,” indicated by a superscript N in this section. However, in large-strain mode, those values
must be incremented by the stress-rotation correction (see Eq. (1.100) in Theory and Background).
We define [σ ] as a generalized stress vector of dimension n with components σ i , i = 1, n and []
as a generalized strain-increment vector with components i , i = 1, n. The components of the
generalized stress- and strain-increment vectors may consist of the six components of the stress- and
strain-increment tensors or other appropriately defined combinations of variables, giving a measure
of stress and strain increments in specific constitutive model contexts.
As a notation convention in this section, we use the subscript n to refer to the range of generalized
components from i = 1 to i = n (e.g., f (σ n ) is used to represent f (σ 1 , σ 2 , ..., σ n )).
f (σ n ) = 0 (1.1)
where f , the yield function, is a known function that specifies the limiting stress com-
bination for which plastic flow takes place. (This function is represented by a surface in
the generalized stress space, and all stress points below the surface are characterized by
elastic behavior.)
(2) the relation expressing the decomposition of strain increments into the sum of elastic and
plastic parts
p
i = ei + i (1.2)
(3) the elastic relations between elastic strain increments and stress increments
p ∂g
i = λ (1.4)
∂σ i
f (σ n + σ n ) = 0 (1.5)
This equation provides a relation for evaluation of the magnitude of the plastic-strain
increment vector.
Substitution of the expression for the elastic-strain increment derived from Eq. (1.2) into the elastic
relation Eq. (1.3) yields, taking into consideration the linear property of the function Si ,
In further expressing the plastic strain increment by means of the flow rule, Eq. (1.4), this equation
becomes
∂g
σ i = Si ( n ) − λSi (1.7)
∂σ n
In the special case where f (σ n ) is a linear function of the components σ i , i = 1, n, Eq. (1.5) may
be expressed as
f (σ n ) + f ∗ (σ n ) = 0 (1.8)
where, as a notation convention, f ∗ represents the function f minus its constant term,
For a stress point σ n on the yield surface, f (σ n ) = 0, and Eq. (1.8) becomes, after substitution of
the expression Eq. (1.7) for the stress increment, and further using the linear property of f ∗ ,
∗
∗ ∂g
f Sn ( n ) − λf Sn =0 (1.10)
∂σ n
i = σ i + σ i
σN (1.11)
σ Ii = σ i + Si ( n ) (1.12)
Note that the term Si ( n ) in Eq. (1.12) is the component i of the stress increment induced by the
total-strain increment n , in case no increment of plastic deformation takes place. This justifies
the name of “elastic guess” for σ Ii .
From the definition Eq. (1.12), it follows, using the preceding arguments, that
f (σ In ) = f ∗ Sn ( n ) (1.13)
Hence, an expression for λ may be derived from Eqs. (1.9), (1.10) and (1.13):
f (σ In )
λ= (1.14)
f Sn (∂g/∂σ n ) − f (0n )
Using the expression of the stress increment, Eq. (1.7), and the definition of the elastic guess,
Eq. (1.12), the new stress may be expressed from Eq. (1.11) as
∂g
σN
i = σ Ii − λSi (1.15)
∂σ n
For clarity, recall that, in these last two expressions, Si (∂g/∂σ n ) is the stress increment obtained
from the incremental elastic law, where ∂g/∂σ i is substituted for i , i = 1, n.
1.3.3 Implementation
In FLAC 3D, an elastic guess σ Ii , i = 1, n for the stress state at time t + t is first evaluated by
adding to the stress components at time, t, increments computed from the total-strain increment for
the step, using an incremental elastic stress-strain law (see Eq. (1.12)). If the elastic guess violates
the yield function, Eq. (1.15) is used to place the new stress exactly on the yield curve. Otherwise,
the elastic guess gives the new stress state at time t + t.
If the stress point σ Ii , i = 1, n is located above the yield surface in the generalized stress space, the
coefficient λ in Eq. (1.15) is given by Eq. (1.14), provided the yield function is a linear function of
the generalized stress vector components. The equation Eq. (1.15) is still valid, but λ is set to zero
in case σ Ii , i = 1, n is located below the yield surface (elastic loading or unloading).
The implementation of each of the 15 basic constitutive models in FLAC 3D is described separately
in the following sections. All models except the null and elastic models potentially involve plastic
deformations. Note that a wide range of material behavior may be obtained from these 15 models
by assigning appropriate values to the model parameters. Recommendations for applying these
models are given in Section 3.6 in the User’s Guide.
A null material is used to represent material that is removed or excavated from the model. The
stresses within a null zone are automatically set to zero:
σijN = 0 (1.16)
The null material may be changed to a different material model at a later stage of the simulation.
In this way, backfilling an excavation can be simulated, for example.
The models in this group are characterized by reversible deformations upon unloading; the stress-
strain laws are linear and path-independent. The elastic models include both isotropic and anisotro-
pic elastic models. Two anisotropic elastic models that correspond to two cases of elastic sym-
metry are available: the orthotropic and transversely isotropic models. For further information on
anisotropic formulations, see Lekhnitskii (1981).
In this elastic, isotropic model, strain increments generate stress increments according to the linear
and reversible law of Hooke:
where the Einstein summation convention applies, δij is the Kroenecker delta symbol, and α2 is a
material constant related to the bulk modulus, K, and shear modulus, G, as
2
α2 = K − G (1.18)
3
The orthotropic model accounts for three orthogonal planes of elastic symmetry. Principal coordi-
nate axes of elasticity, labeled 1 , 2 , 3 , are defined in the directions normal to those planes.
The incremental strain-stress relations in the local axes have the form
⎡ 1 ⎤
⎧ ⎫ − νE122 − νE133 ⎧ ⎫
⎪ 11 ⎪ ⎢ Eν121 ⎥⎪ σ 11 ⎪
⎪
⎪ ⎪ ⎢−E 1
− νE233 ⎪ ⎪
⎪
⎪ 22 ⎪ ⎪
⎪ E2 ⎥⎪
⎥⎪ σ 22 ⎪
⎪
⎬ ⎢ ⎪ ⎪
1
⎨ ⎢ − ν31
− νE322 1 ⎥ ⎨ σ 33 ⎬
33 ⎢ E1 E3 ⎥
⎢ ⎥ ⎪ σ 12 ⎪ (1.20)
⎪
⎪ 2 ⎪
12 ⎪ = ⎢
1
⎥⎪ ⎪
⎪
⎪ ⎪ ⎪ ⎢ G12
⎥⎪⎪ ⎪ ⎪
⎪
⎩ 2 13 ⎪
⎭ ⎣ 1 ⎦⎩⎪ σ 13 ⎪
⎭
2 23 G13
σ 23
1
G23
ν21 ν12
=
E1 E2
ν31 ν13
= (1.21)
E1 E3
ν32 ν23
=
E2 E3
In addition to those nine properties, the user prescribes the orientation of the local axes by giving
the dip and dip direction of the (1 , 2 ) plane, and the rotation angle between the 1 axis and the
dip-direction vector (defined in positive sense from the dip direction vector). Default values for all
properties are zero.
In the FLAC 3D implementation of this model, the local stiffness matrix [K ] is found by inversion
of the symmetric matrix in Eq. (1.20). Using [σ ] and [ ] to represent the incremental stress
and strain vectors present in the right and left members of Eq. (1.20):
In FLAC 3D, the global stiffness matrix [K] is calculated by applying a transformation of the form
where [Q] is a suitable 6 × 6 matrix involving direction cosines of local axes in global axes (Q is
derived from the relations σ ij = cik σkl cj l , where cij is direction cosine j of local axis i).
In particular, if the local axes are obtained from the global axes by positive rotation through an
angle θ about the common 3 ≡ 3 axis, we have
⎡ ⎤
cos2 θ sin2 θ − sin θ cos θ
⎢ sin2 θ cos2 θ + sin θ cos θ ⎥
⎢ ⎥
⎢ 1 ⎥
⎢ ⎥ (1.25)
Q = ⎢ 2 sin θ cos θ −2 sin θ cos θ cos2 θ − sin2 θ ⎥
⎣ ⎦
cos θ − sin θ
sin θ cos θ
The matrix for rotation about the 1 ≡ 1 or 2 ≡ 2 axis may be obtained by cyclic permutation of
indices.
The transversely isotropic model takes a plane of isotropy into consideration. Let the axis of
rotational symmetry, normal to the plane of isotropy, correspond to the local 3 axis. This axis is a
principal direction of elasticity. Also, any two perpendicular directions 1 , 2 , which are principal
directions of elasticity, can be selected in the isotropic plane. With this convention, the transversely
isotropic model may be considered as a particular case of the orthotropic model for which
E1 = E2
G13 = G23
ν13 = ν23
E1
G12 = (1.26)
2(1 + ν12 )
⎡ 1 ⎤
⎧ ⎫ − Eν − Eν ⎧ ⎫
⎪ 11 ⎪ E σ 11 ⎪
⎪
⎪ ⎪
⎪ ⎢−ν 1
− Eν ⎥⎪⎪ ⎪
⎪
⎪
22 ⎪ ⎪ ⎢ E E ⎥⎪⎪
⎪ σ 22 ⎪
⎪
⎪
⎨ ⎬ ⎢ ν ⎥⎨ ⎬
33 ⎢ − E − Eν 1
E ⎥ σ 33
⎢ ⎥ (1.27)
⎪
⎪ 2 12 ⎪ ⎪ =⎢ 1 ⎥⎪ σ 12 ⎪
⎪
⎪ ⎪⎪ ⎢ ⎥⎪⎪ ⎪
⎪
⎦⎪ ⎪
G
⎩ 2 13 ⎪
⎪ ⎭ ⎣ 1 ⎪
⎩ σ 13 ⎪
⎭
G
2 23 1 σ 23
G
The model involves the five independent elastic constants E, E , ν, ν and G . The shear modulus,
G, is calculated by the code from the relation G = E/2(1 + ν) (see Eq. (1.26)). In addition to
those five properties, the user prescribes the orientation of the isotropic plane by giving its dip and
dip direction. Default values for all properties are zero.
The cross shear modulus, G13 , for anisotropic elasticity must be determined. Lekhnitskii (1981)
suggests the following equation, based on laboratory testing of rock:
Ex Ey
Gxy = (1.28)
Ex (1 + 2νxy ) + Ey
All plastic models potentially involve some degree of permanent, path-dependent deformations
(failure) – a consequence of the nonlinearity of the stress-strain relations. The different models
in FLAC 3D are characterized by their yield function, hardening/softening functions and flow rule.
The yield functions for each model define the stress combination for which plastic flow takes place.
These functions or criteria are represented by one or more limiting surfaces in a generalized stress
space, with points below or on the surface being characterized by an incremental elastic or plastic
behavior, respectively. The plastic flow formulation in FLAC 3D rests on basic assumptions from
plasticity theory that the total strain increment may be decomposed into elastic and plastic parts,
with only the elastic part contributing to the stress increment by means of an elastic law (see the
discussion in Section 1.3.2). In addition, both plastic and elastic strain increments are taken to
be coaxial with the current principal axes of the stresses (only valid if elastic strains are small
compared to plastic strains during plastic flow). The flow rule specifies the direction of the plastic
strain increment vector as that normal to the potential surface; it is called associated if the potential
and yield functions coincide, and nonassociated otherwise. See Vermeer and de Borst (1984) for a
more detailed discussion on the theory of plasticity.
For the Drucker-Prager, Mohr-Coulomb, ubiquitous-joint, strain-hardening/softening and bilinear-
hardening/softening-ubiquitous-joint models, a shear yield function and a nonassociated shear flow
rule are used. For the double-yield and Cysoil models, shear and volumetric yield functions,
nonassociated shear flow and associated volumetric flow rules are included. The Chsoil model
is a simplified version of the Cysoil model that provides built-in friction hardening and dilation
hardening/softening laws, and does not include a volumetric cap. In addition, the failure envelope
for each of the above models is characterized by a tensile yield function with associated flow rule.
The modified Cam-clay model formulation rests on a combined shear and volumetric yield function
and associated flow rule.
The two types of Hoek-Brown model in FLAC 3D provide different formulations to represent yielding.
The basic Hoek-Brown model uses a nonlinear shear yield function and a plasticity flow rule that
varies as a function of the stress level. For the modified Hoek-Brown model, plastic flow is handled
in a manner similar to that in the Mohr-Coulomb model in which a dilation angle is specified. Also,
a tensile yield function similar to that used with the Mohr-Coulomb model is included with the
modified Hoek-Brown model.
Note also that all plasticity models are formulated in terms of effective stresses, not total stresses.
The plasticity models can produce localization (i.e., the development of families of discontinuities
such as shear bands in a material that starts as a continuum). Note that localization is grid-dependent
since there is no intrinsic length scale incorporated in the formulations. This is an important consid-
eration when creating a grid for a plasticity analysis, and is discussed more fully in Section 3.10.3
in the User’s Guide.
As discussed in Section 1.3.3, in the numerical implementation of the models, an elastic trial (or
“elastic guess”) for the stress increment is first computed from the total strain increment using
the incremental form of Hooke’s law. The corresponding stresses are then evaluated. If they
violate the yield criteria (i.e., the stress-point representation lies above the yield function in the
generalized stress space), plastic deformations take place. In this case, only the elastic part of the
strain increment can contribute to the stress increment; the latter is corrected by using the plastic
flow rule to ensure that the stresses lie on the composite yield function. This section describes the
yield and potential functions, flow rules and stress corrections for the different plasticity models in
FLAC 3D.
The failure envelope for this model involves a Drucker-Prager criterion with tension cutoff. The
position of a stress point on this envelope is controlled by a nonassociated flow rule for shear failure,
and an associated rule for tension failure.
1.6.1.1 Generalized Stress and Strain Components
The generalized stress vector [σ ] involved in the definition of the Drucker-Prager model has two
components (n = 2): the tangential stress, τ , and mean normal stress, σ , defined as
1
τ= sij sij (1.29)
2
σkk
σ = (1.30)
3
where the Einstein summation convention applies, and [s] is the deviatoric-stress tensor. The com-
ponents of the associated generalized strain increment vector [] are the shear-strain increment,
γ , and volumetric-strain increment, , introduced as
γ = 2eij eij (1.31)
= kk (1.32)
The incremental expression of Hooke’s law, in terms of the generalized stress and stress increments,
has the form
τ = Gγ e (1.33)
σ = K e (1.34)
S1 (γ e , e ) = Gγ e
(1.35)
S2 (γ , ) = K
e e e
(1.36)
The failure criterion used for this FLAC 3D model is a composite Drucker-Prager criterion with
tension cutoff as sketched in the (τ, σ ) representation of Figure 1.1. The failure envelope f (τ, σ )
= 0 is defined, from point A to B on the figure, by the Drucker-Prager failure criterion f s = 0, with
f s = τ + qφ σ − kφ (1.37)
f t = σ − σt (1.38)
where qφ , kφ and σ t are positive material constants, and σ t is the tensile strength for the Drucker-
Prager model. Note that, for a material whose property qφ is not equal to zero, the maximum value
of the tensile strength is given by
kφ
t
σmax = (1.39)
qφ
A f =s
0 τ
kφ B f t=0
C σ
σt
k φ / qφ
The potential function g(τ, σ ) = constant is composed of two functions, g s and g t , used to
describe shear and tensile plastic flow, respectively. The function g s corresponds in general to a
nonassociated law, and has the form
g s = τ + qψ σ (1.40)
where qψ is a constant, equal to qφ if the flow rule is associated. The function g t corresponds to an
associated flow rule, and is given by
gt = σ (1.41)
The flow rule is given a unique definition by application of the following technique. A function
h(τ, σ ) = 0, which is represented by the diagonal between the representation of f s = 0 and f t = 0 in
the (τ, σ ) plane (see Figure 1.2), is defined. The function is selected with its positive and negative
domains, as indicated on the figure, and has the form
h = τ − τ P − a P (σ − σ t ) (1.42)
τ P = kφ − qφ σ t (1.43)
a = 1 + qφ2 − qφ
P
(1.44)
+ τ
f s=
- 0
domain 1
+
h=0 -
domain 2
+
-
t
σ
f =0
Figure 1.2 Drucker-Prager model – domains used in the definition of the
flow rule
An elastic guess violating the composite yield function is represented by a point in the (τ, σ )-plane,
located either in domain 1 or 2, corresponding to positive or negative domains of h = 0, respectively
(see Figure 1.2). If the stress point falls within domain 1, shear failure is declared, and the stress
point is placed on the curve f s = 0 using a flow rule derived using the potential function g s . If
the point falls within domain 2, tensile failure takes place, and the new stress point satisfies f t = 0
using a flow rule derived using g t .
∂g s
=1
∂τ
(1.45)
∂g s
= qψ
∂σ
Substitution of ∂g s /∂τ and ∂g s /∂σ for γ e and for e , respectively, in Eq. (1.35) gives
∂g s ∂g s
S1 , =G
∂τ ∂σ
(1.46)
∂g s ∂g s
S2 , = Kqψ (1.47)
∂τ ∂σ
Using f = f s (see Eq. (1.37)), Eqs. (1.14) and (1.15) then become
τ N = τ I − λs G (1.48)
σ N = σ I − λs K qψ (1.49)
and
f s (τ I , σ I )
λs = (1.50)
G + Kqφ qψ
The new stress tensor components may be expressed in terms of the generalized stresses using a
rescaling technique. Formally, we may write Eq. (1.48) as
τN = μ τI (1.51)
where μ is a known proportionality factor (μ = 1 − λs G/τ I ). From the definition Eq. (1.29) of τ
in terms of the tensor components sij , we then have
τN
sij N
= sij I
(1.53)
τI
Finally, new stress components may be calculated from Eqs. (1.49) and (1.53) using the definition
∂g t
=0 (1.55)
∂τ
∂g t
=1 (1.56)
∂σ
∂g t ∂g t
S1 , =0
∂τ ∂σ
(1.57)
∂g t ∂g t
S2 , =K (1.58)
∂τ ∂σ
Eqs. (1.14) and (1.15) then give, with f = f t , as given by Eq. (1.38):
τN = τI (1.59)
σ N = σ I − λt K (1.60)
and
σI − σt
λt = (1.61)
K
τN = τI (1.62)
σN = σt (1.63)
In this mode of failure, sij N = sij I , and we may write, from the definition of deviatoric stresses
and Eq. (1.63),
In the implementation of the Drucker-Prager model in FLAC 3D, an elastic guess, σijI , is first computed
by adding to the old stress components, increments calculated by application of Hooke’s law to the
total strain increments, ij (see Section 1.5.1). Deviatoric stresses, sijI , are evaluated from σijI ,
and the elastic guess (τ I , σ I ) is derived from sijI and σijI using Eqs. (1.29) and (1.30).
If this guess violates the composite yield criterion (see Eqs. (1.37) and (1.38)), then either h(τ I , σ I )
> 0 or h(τ I , σ I ) ≤ 0. In the first case, shear failure takes place, and τ N and σ N are evaluated
from Eqs. (1.48) and (1.49), using Eq. (1.50). New deviatoric stress components are derived using
Eq. (1.53) and, finally, new stress components are calculated from Eq. (1.54). In the second case,
tensile failure occurs, and new stress components are evaluated from Eq. (1.65).
If the point (τ I , σ I ) is located below the composite failure envelope, no plastic flow takes place for
this step, and the new stresses are given by σijI .
t
Note that the tensile strength default value is zero for a material with qφ = 0, and is σmax otherwise
t
(see Eq. (1.39)). This last value will be substituted for σ if the assigned value for the tensile
t . There is no tensile softening in this model.
strength exceeds σmax
6
qφ = √ sin φ (1.66)
3(3 − sin φ)
6
kφ = √ c cos φ (1.67)
3(3 − sin φ)
6
qφ = √ sin φ (1.68)
3(3 + sin φ)
6
kφ = √ c cos φ (1.69)
3(3 + sin φ)
In the special case qφ = 0, the Drucker-Prager criterion degenerates into the von Mises criterion,
which corresponds to a cylinder in the principal stress space. The Tresca criterion is a special case
of the Mohr-Coulomb criterion for which φ = 0. It is represented in the principal stress space by a
regular hexagonal prism. The von Mises cylinder circumscribes the prism for
qφ = 0 (1.70)
2
kφ = √ c (1.71)
3
-σ 3
Drucker-Prager qφ > 0 σ3
σ2 =
σ1 =
qφ Von Mises qφ = 0
3 kφ
-σ 2
-σ 1
Figure 1.3 Drucker-Prager and von Mises yield surfaces in principal stress
space
-σ 3
Mohr-Coulomb φ > 0 σ3
σ2 =
σ1 =
tφ Tresca φ = 0
co
3 C
-σ 2
-σ 1
The failure envelope for this model corresponds to a Mohr-Coulomb criterion (shear yield function)
with tension cutoff (tension yield function). The position of a stress point on this envelope is
controlled by a nonassociated flow rule for shear failure, and an associated rule for tension failure.
1.6.2.1 Generalized Stress and Strain Components
The Mohr-Coulomb criterion in FLAC 3D is expressed in terms of the principal stresses σ1 , σ2 and
σ3 , which are the three components of the generalized stress vector for this model (n = 3). The
components of the corresponding generalized strain vector are the principal strains 1 , 2 and 3 .
1.6.2.2 Incremental Elastic Law
The incremental expression of Hooke’s law in terms of the generalized stress and stress increments
has the form
where α1 and α2 are material constants defined in terms of the shear modulus, G, and bulk modulus,
K, as
4
α1 = K + G
3
(1.73)
2
α2 = K − G
3
The failure criterion used in the FLAC 3D model is a composite Mohr-Coulomb criterion with tension
cutoff. In labeling the three principal stresses so that
σ1 ≤ σ2 ≤ σ3 (1.75)
this criterion may be represented in the plane (σ1 , σ3 ) as illustrated in Figure 1.5. (Recall that
compressive stresses are negative.) The failure envelope f (σ1 , σ3 ) = 0 is defined from point A to
B by the Mohr-Coulomb failure criterion f s = 0 with
f s = σ1 − σ3 Nφ + 2c Nφ (1.76)
ft = σ3 − σ t (1.77)
where φ is the friction angle, c is the cohesion, σ t is the tensile strength, and
1 + sin(φ)
Nφ = (1.78)
1 − sin(φ)
Note that the tensile strength of the material cannot exceed the value of σ3 corresponding to the
intersection point of the straight lines f s = 0 and σ1 = σ3 in the f (σ1 , σ3 ) plane. This maximum
value is given by
c
t
σmax = (1.79)
tan φ
B C
A
+
-
The potential function is described by means of two functions, g s and g t , used to define shear plastic
flow and tensile plastic flow, respectively. The function g s corresponds to a nonassociated law and
has the form
g s = σ1 − σ3 Nψ (1.80)
1 + sin(ψ)
Nψ = (1.81)
1 − sin(ψ)
g t = −σ3 (1.82)
The flow rule is given a unique definition by application of the following technique. A function
h(σ1 , σ3 ) = 0, which is represented by the diagonal between the representation of f s = 0 and f t
= 0 in the (σ1 , σ3 )-plane (see Figure 1.6), is defined. The function is selected with its positive and
negative domains, as indicated on the figure, and has the form
h = σ3 − σ t + a P (σ1 − σ P ) (1.83)
a = 1 + Nφ2 + Nφ
P
(1.84)
σ P
= σ Nφ − 2c Nφ
t
I!
domain 2
domain 1
fJ=0
-
I
An elastic guess violating the composite yield function is represented by a point in the (σ1 , σ3 )-
plane, located either in domain 1 or 2, corresponding to negative or positive domains of h = 0,
respectively (see Figure 1.6). If the stress point falls within domain 1, shear failure is declared, and
the stress point is placed on the curve f s = 0 using a flow rule derived using the potential function
g s . If the point falls within domain 2, tensile failure takes place, and the new stress point conforms
to f t = 0 using a flow rule derived using g t .
Note that by ordering the stresses as in Eq. (1.75), the case of a shear-shear edge is automatically
handled by a variation on this technique. The technique, applicable for small-strain increments,
is simple to implement: at each step, only one flow rule and corresponding stress correction is
involved in case of plastic flow. In particular, when a stress point follows an edge, it receives stress
corrections alternating between two criteria. In this process, the two yield criteria are fulfilled to
an accuracy which depends on the magnitude of the strain increment. Results obtained for the
oedometric test are presented as a validation of this approach in Section 1.6.2.6.
1.6.2.4 Plastic Corrections
∂g s
=1
∂σ1
∂g s
=0 (1.85)
∂σ2
∂g s
= −Nψ
∂σ3
Substitution of ∂g s /∂σ1 , ∂g s /∂σ2 and ∂g s /∂σ3 for 1e , 2e and 3e , respectively, in Eq. (1.74)
gives
∂g s ∂g s ∂g s
S1 ( , , ) = α1 − α2 Nψ
∂σ1 ∂σ2 ∂σ3
∂g s ∂g s ∂g s
S2 ( , , ) = α2 (1 − Nψ ) (1.86)
∂σ1 ∂σ2 ∂σ3
∂g s ∂g s ∂g s
S3 ( , , ) = −α1 Nψ + α2
∂σ1 ∂σ2 ∂σ3
Eqs. (1.14) and (1.15) then yield, using f = f s (see Eq. (1.76)):
and
f s (σ1I , σ3I )
λs = (1.88)
(α1 − α2 Nψ ) − (−α1 Nψ + α2 ) Nφ
∂g t
=0
∂σ1
∂g t
=0 (1.89)
∂σ2
∂g t
=1
∂σ3
∂g t ∂g t ∂g t
S1 , , = α2
∂σ1 ∂σ2 ∂σ3
∂g t ∂g t ∂g t
S2 , , = α2 (1.90)
∂σ1 ∂σ2 ∂σ3
t
∂g ∂g t ∂g t
S3 , , = α1
∂σ1 ∂σ2 ∂σ3
σ1N = σ1I − λt α2
σ3N = σ3I − λt α1
and
σ3I − σ t
λt = (1.92)
α1
α2
σ1N = σ1I − (σ3I − σ t )
α1
α2
σ2N = σ2I − (σ3I − σ t ) (1.93)
α1
σ3N = σ t
After evaluation of σ1N , σ2N and σ3N , the stress-tensor components are evaluated in the system of
reference axes, assuming that the principal directions have not been affected by the occurrence of
a plastic correction.
1.6.2.5 Implementation Procedures
In the implementation of the Mohr-Coulomb model in FLAC 3D, an elastic guess (σijI ) is first com-
puted by adding to the stress components, increments calculated by application of Hooke’s law to
the total strain increments ij (see Section 1.5.1). Principal stresses σ1I , σ2I , σ3I and corresponding
directions are then calculated.
If the stresses σ1I , σ2I , σ3I violate the composite yield criterion (see Eqs. (1.76) and (1.77)), then
either h(σ1I , σ3I ) ≤ 0 or h(σ1I , σ3I ) > 0. In the first case, shear failure takes place, and σ1N , σ2N and
σ3N are evaluated from Eq. (1.87), using Eq. (1.88). In the second case, tensile failure occurs, and
new principal stress components are evaluated from Eq. (1.93).
If the point (σ1I , σ3I ) is located below the representation of the composite failure envelope in the
plane (σ1 , σ3 ), no plastic flow takes place for this step, and the new principal stresses are given by
σiI , i = 1,3.
The stress tensor components in the system of reference axes are then calculated from the principal
values by assuming that the principal directions have not been affected by the occurrence of a plastic
correction.
In FLAC 3D, the default value for the tensile strength, σ t , is zero. This value is set to σmax t (see
t
Eq. (1.79)) if the value assigned to the tensile strength exceeds σmax . By default, if tensile failure
occurs in a zone, the value assigned for the tensile strength in this zone remains constant. Alter-
natively, if the property keyword brittle is set true, then tensile strength is set to zero for that zone
when tensile failure occurs. This simulates instantaneous tensile softening.
1.6.2.6 Oedometer Test
11 = 0
22 = vt/L (1.94)
33 = 0
where v is the constant y-component of the velocity applied to the sample (v < 0), and L is the
height of the sample.
Assuming zero initial stresses, the principal directions of stresses and strains are those of the
coordinate axes. For simplicity, we consider a sample of unit height L = 1.
v v
x, z
In the elastic range, application of Hooke’s law gives, using 22 = vt at time t:
σ11 = α2 vt
σ22 = α1 vt (1.95)
σ33 = σ11
f 1 = σ22 − σ11 Nφ + 2c Nφ
f 2 = σ22 − σ33 Nφ + 2c Nφ (1.96)
At the onset of yield, f 1 = f 2 = 0 and, using Eqs. (1.95) and (1.96), we find
2c Nφ
t= (1.97)
−v(α1 − α2 Nφ )
p
11 = 11
e
+ 11
p
22 = 22
e
+ 22 (1.98)
p
33 = e
33 + 33
p
e
11 = −11
p
e
22 = vt − 22 (1.99)
p
e
33 = −33
The flow rule for plastic flow along the edge of the Mohr-Coulomb criterion corresponding to
σ11 = σ33 has the form (e.g., see Drescher 1991)
p ∂g 1 ∂g 2
11 = λ1 + λ2
∂σ11 ∂σ11
∂g 1 ∂g 2
p
22 = λ1 + λ2 (1.100)
∂σ22 ∂σ22
∂g 1 ∂g 2
p
33 = λ1 + λ2
∂σ33 ∂σ33
g 1 = σ22 − σ11 Nψ
g 2 = σ22 − σ33 Nψ (1.101)
p
11 = −λ1 Nψ
p
22 = λ1 + λ2 (1.102)
p
33 = −λ2 Nψ
p
11 = −λ1 Nψ
p
22 = 2λ1 (1.103)
p
33 = −λ1 Nψ
The stress increments, derived from Hooke’s law, are given by the relations
σ11 = α1 11
e
+ α2 (22
e
+ 11
e
)
σ22 = α1 22 + α2 211
e e
(1.104)
σ33 = σ11
e = e .
where we have used the symmetry condition 11 33
The parameter λ1 can now be determined by expressing the condition that during plastic flow, f 1
= 0. Using Eq. (1.96), this condition takes the form
Substitution of Eq. (1.105) in Eq. (1.106) yields, after some manipulations, the expression
λ1 = vtλ (1.107)
where
α1 − α2 Nφ
λ= (1.108)
(α1 + α2 )Nφ Nψ − 2α2 (Nφ + Nψ ) + 2α1
The FLAC 3D simulation is carried out using a single zone of unit dimensions. Several properties
are used in conjunction with the Mohr-Coulomb model:
The velocity components are fixed in the x-, y- and z-directions. A velocity of magnitude 10−5
m/steps is applied to the top of the model in the negative y-direction, for a total of 1000 steps.
The stress and displacement components in the y-direction are monitored and compared to the
analytic prediction obtained from Eqs. (1.95), (1.97) and (1.105), using Eqs. (1.107) and (1.108).
Two runs are carried out using the data file in Example 1.1 with values of 10◦ and 0◦ for the
dilation parameter. The corresponding project file is “oedometer-mohr.f3prj,” located in folder
“datafiles\constitutive models\oedometer-mohr.” The match is very good, as can be seen in Fig-
ures 1.8 and 1.9, where numerical and analytic solutions coincide at the precision of the plot
resolution.
def d_sigy
global a_sy = 0.0
global pnt = zone_head
local c_k = z_prop(pnt,’bulk’)
local c_g = z_prop(pnt,’shear’)
global e1 = c_k + 4. * c_g /3.
local e2 = c_k - 2. * c_g /3.
local sf = z_prop(pnt,’friction’) * degrad
local nf = sin(sf)
nf = (1. + nf) / (1. - nf)
local sp = z_prop(pnt,’dilation’) * degrad
local np = sin(sp)
np = (1. + np) / (1. - np)
local rl = (e1-e2*nf)/((e1+e2)*nf*np-2.*e2*(nf+np)+2.*e1)
global dsigy = vyv * (e1+2.*rl*(e2*np-e1))
global stepl = -2.*z_prop(pnt,’cohesion’)*sqrt(nf)/((e1-e2*nf)*vyv)
end
def esigy
while_stepping
if step < stepl then
a_sy = a_sy + e1 * vyv
else
a_sy = a_sy + dsigy
end_if
global n_sy = z_syy(pnt)
end
; --- boundary conditions ---
fix x y z
ini yvel @vyv range y 0.9 1.1
; --- histories ---
his nstep 50
his add gp ydisp 0 1 0
his add @n_sy
his add @a_sy
save ini
;
; --- dilation 10
pro dil 10
@d_sigy
step 1000
save dil0
; --- dilation 0
res ini
pro dil 0
@d_sigy
step 1000
save dil10
ret
A
eeee e
eUly
e
2p©2ev ey
e)uev
b
e©ppp
Ga2ga2p©2e©2,©p,2©eAU
ss^
l2e
eMHv3h
l0e eMHv3h
bel©el
eee0
^
v ey
e)uev
b
U
A
eeee e
eUly
e
2p©2ev ey
e)uev
b
e©ppp
Ga2ga2p©2e©2,©p,22eAU
ss^
l2e
eMHv3h
l0e eMHv3h
bel©el
eee0
^
v ey
e)uev
b
U
brittle If true, then the tension limit is set to 0 in the event of tensile failure.
The default is false.
bulk elastic bulk modulus, K
cohesion cohesion, c
density mass density, ρ
dilation dilation angle, ψ
friction friction angle, φ
shear elastic shear modulus, G
tension tension limit, σ t
Note that the default tension limit is zero for a material with no friction, and is c/ tan φ otherwise.
The value assigned for the tension limit remains constant when tensile failure occurs if brittle is
false, and is set to zero when tensile failure occurs if brittle is true.
This model accounts for the presence of an orientation of weakness (weak plane) in a FLAC 3D
Mohr-Coulomb model. The criterion for failure on the plane, whose orientation is given, consists
of a composite Mohr-Coulomb envelope with tension cutoff. The position of a stress point on the
latter envelope is controlled again by a nonassociated flow rule for shear failure, and an associated
rule for tension failure.
In this numerical model, general failure is first detected, and relevant plastic corrections are applied,
as indicated in the FLAC 3D Mohr-Coulomb model description. The new stresses are then analyzed
for failure on the weak plane, and updated accordingly. The FLAC 3D Mohr-Coulomb model was
addressed earlier; developments related to plastic flow on the weak plane will be discussed in this
section.
1.6.3.1 Definitions
The weak-plane orientation is given by the Cartesian components of a unit normal to the plane in
the global x-, y-, z-axes. A local system of reference axes is defined with x and y in the plane
and z pointing in the direction of the unit normal [n]. (The x -axis points downward in the dip
direction, and the y -axis is horizontal. The local system is right-handed.)
Using a matrix notation, we define [C] as the rotation tensor whose three columns are the direction
cosines of x , y and [n]. The stress components in the local axes σi j are computed using the
transformation
In turn, the global stress components may be obtained from the local components using the reverse
transformation,
The magnitude of the tangential traction component on the weak plane, referred to as τ , is defined
as
τ= σ1 3 2 + σ2 3 2 (1.111)
In what follows, we use σijO to refer to the stress components obtained at the end of the current
timestep, after application of the plastic corrections related to general failure only (as opposed to
failure on the weak plane).
The generalized stress vector used to describe weak-plane failure has four components (n = 4):
σ1 1 , σ2 2 , σ3 3 and τ . The components of the corresponding generalized strain vector are 1 1 ,
2 2 , 3 3 and γ .
1.6.3.3 Incremental Elastic Law
The incremental expression of Hooke’s law in terms of the generalized stress and strain increments
has the form (see Section 1.5.1)
τ = 2Gγ e (1.116)
where α1 and α2 are material constants defined in terms of the shear modulus, G, and bulk modulus,
K, as
4
α1 = K + G
3
(1.117)
2
α2 = K − G
3
Note that the elastic strain increments in the preceding formula may be expressed, according to
Eq. (1.2), as differences between total and plastic strain increments. Taking into consideration
that the plastic increments may contain a known contribution associated with general failure (as
opposed to weak plane failure), the derivation leading to the expressions Eqs. (1.14) and (1.15) for
the new stresses may be followed, provided we interpret σ Ii as the generalized stress component,
σOi , obtained at the end of the current timestep, after application of the plastic corrections relating
to general failure only.
1.6.3.4 Composite Failure Criterion and Flow Rule
The weak-plane failure criterion used in the FLAC 3D model is a composite Mohr-Coulomb crite-
rion with tension cutoff expressed in terms of (σ3 3 ,τ ), as illustrated in Figure 1.10. (Recall that
compressive stresses are negative.) The failure envelope f (σ3 3 , τ ) = 0 is defined from point A to
B by the Mohr-Coulomb failure criterion f s = 0, with
where φj , cj and σjt are the friction, cohesion and tensile strength of the weak plane, respectively.
Note that, for a weak plane with nonzero friction angle, the maximum value of the tensile strength
is given by
cj
σjt max = (1.121)
tan φj
A f =s
0 τ
Cj B f t=0
C σ3 3
σ jt
Cj / tan φ j
The potential function is composed of two functions (g s and g t ) used to define shear and tensile
plastic flow, respectively. The function g s corresponds to a nonassociated law, and has the form
where ψ is the weak-plane dilation angle. The function g t corresponds to an associated flow rule
and is written as
gt = σ3 3 (1.123)
The flow rule is given a unique definition by application of the following technique. A function,
h(σ3 3 , τ ) = 0, which is represented by the diagonal between the representation of f s = 0 and f t =
0 in the (σ3 3 , τ ) plane (see Figure 1.11), is defined. The function is selected with its positive and
negative domains, as indicated on the figure, and has the form
An elastic guess violating the composite yield function is represented by a point in the (σ3 3 , τ )-
plane located either in domain 1 or 2, corresponding to positive or negative domains of h = 0,
respectively (see Figure 1.11). If in domain 1, shear failure is declared and the stress point is placed
on the curve f s = 0 using a flow rule derived using the potential function g s . If in domain 2, tensile
failure takes place and the stress point conforms to f t = 0 using a flow rule derived using g t .
f s=0 τ
domain 1
h=0
domain 2
σ3 3
f t=0
∂g s
=0
∂σ1 1
∂g s
=0
∂σ2 2
(1.126)
∂g s
= tan ψj
∂σ3 3
∂g s
=1
∂τ
Substitution of ∂g s /∂σ1 1 , ∂g s /∂σ2 2 , ∂g s /∂σ3 3 and ∂g s /∂τ for 1e 1 , 2e 2 , 3e 3 and γ e ,
respectively, in Eq. (1.118), gives
∂g s ∂g s ∂g s ∂g s
S1 ( , , , ) = α2 tan ψj
∂σ1 1 ∂σ2 2 ∂σ3 3 ∂τ
∂g s ∂g s ∂g s ∂g s
S2 ( , , , ) = α2 tan ψj
∂σ1 1 ∂σ2 2 ∂σ3 3 ∂τ
(1.127)
∂g s ∂g s ∂g s ∂g s
S3 ( , , , ) = α1 tan ψj
∂σ1 1 ∂σ2 2 ∂σ3 3 ∂τ
∂g s ∂g s ∂g s ∂g s
S4 ( , , , ) = 2G
∂σ1 1 ∂σ2 2 ∂σ3 3 ∂τ
Eqs. (1.14) and (1.15) then yield, using f = f s (see Eq. (1.119)):
τ N = τ O − λs 2G (1.129)
and
f s (σ3O 3 , τ O )
λ =
s
(1.130)
2G + α1 tan ψj tan φj
The new shear stress components on the weak plane may be derived from τ N and τ O , using the
relations
τN
σ1N 3 = σ1O 3
τO
τN
σ2N 3 = σ2O 3 (1.131)
τO
The stress corrections for shear failure may thus be expressed as (see Eqs. (1.128) and (1.131))
τN − τO
σ1 3 = σ1O 3
τO
τN − τO
σ2 3 = σ2O 3
τO
∂g t
=0
∂σ1 1
∂g t
=0
∂σ2 2
(1.133)
∂g t
=1
∂σ3 3
∂g t
=0
∂τ
∂g t ∂g t ∂g t ∂g t
S1 ( , , , ) = α2
∂σ1 1 ∂σ2 2 ∂σ3 3 ∂τ
∂g t ∂g t ∂g t ∂g t
S2 ( , , , ) = α2
∂σ1 1 ∂σ2 2 ∂σ3 3 ∂τ
(1.134)
∂g t ∂g t ∂g t ∂g t
S3 ( , , , ) = α1
∂σ1 1 ∂σ2 2 ∂σ3 3 ∂τ
∂g t ∂g t ∂g t ∂g t
S4 ( , , , )=0
∂σ1 1 ∂σ2 2 ∂σ3 3 ∂τ
Eqs. (1.14) and (1.15) then give, with f = f t as given by Eq. (1.120):
σ1N 1 = σ1O 1 − λt α2
σ3N 3 = σ3O 3 − λt α1
and
σ3O 3 − σjt
λ =
t
(1.136)
α1
Finally, the stress corrections for tensile failure may be expressed, after substitution of Eq. (1.136)
for λt in Eq. (1.135), in the form
α2
σ1 1 = −(σ3O 3 − σjt )
α1
α2
σ2 2 = −(σ3O 3 − σjt ) (1.137)
α1
In large strain, the orientation of the weak plane is adjusted, per zone, to account for rigid-body
rotations and rotations due to deformations. The corrections applied to the global components of
the unit normal to the weak plane are computed, at each step, as averages over all tetra involved in
the zone.
First taking into account rotations due to deformations, the corrections in local axes [n]d to the
weak-plane normal may be expressed as
nd1 = −1 3
nd2 = −2 3
nd3 = 0 (1.138)
The corresponding global components nd1 , nd2 , nd3 are obtained using the transformation
In the implementation of the ubiquitous-joint model in FLAC 3D, stresses corresponding to the elastic
guess for the step are first analyzed for general failure, and relevant plastic corrections are made, as
described in the FLAC 3D Mohr-Coulomb model. The resulting stress components, labeled as σijO ,
are then examined for failure on the weak plane.
Local stress components σ1O 1 , σ2O 2 , σ3O 3 and τ O are first evaluated using the transformation
Eq. (1.109) and the expression Eq. (1.111). If the stresses σ3O 3 and τ O violate the composite
yield criterion (see Eqs. (1.119) and (1.120)), then either h(σ3O 3 , τ O ) > 0 or h(σ3O 3 , τ O ) ≤ 0. In
the first case, shear failure takes place, and the stress corrections σ1 1 , σ2 2 , σ3 3 σ1 3 and
σ2 3 are evaluated from Eq. (1.132), using Eq. (1.130). In the second case, tensile failure occurs
and the stress corrections σ1 1 , σ2 2 and σ3 3 are evaluated from Eq. (1.136). The tensor of
stress corrections is expressed in the system of reference axes using the transformation Eq. (1.110),
and added to [σ ]O to yield new stress values.
If the point (σ3O 3 , τ O ) is located below the representation of the composite failure envelope in the
plane (σ3 3 , τ ), no plastic flow takes place for this step, and the new stresses are given by σijO .
In large-strain mode, the unit normal to the weak plane is adjusted per zone to account for body
rotations (see Eqs. (1.138) and (1.139)).
Note that the default value for the weak-plane tensile strength is zero if φj = 0, and σjt otherwise
max
t
(see Eq. (1.121)). This last value is substituted for σ if the value assigned for the weak plane tensile
strength exceeds σjt .
max
This model is based on the FLAC 3D Mohr-Coulomb model with nonassociated shear and associated
tension flow rules, as described earlier. The difference, however, lies in the possibility that the
cohesion, friction, dilation and tensile strength may harden or soften after the onset of plastic yield.
In the Mohr-Coulomb model, those properties are assumed to remain constant. Here, the user can
define the cohesion, friction and dilation as piecewise-linear functions of a hardening parameter
measuring the plastic shear strain. A piecewise-linear softening law for the tensile strength can
also be prescribed in terms of another hardening parameter measuring the plastic tensile strain. The
code measures the total plastic shear and tensile strains by incrementing the hardening parameters
at each timestep, and causes the model properties to conform to the user-defined functions.
The yield and potential functions, plastic flow rules and stress corrections are identical to those of
the Mohr-Coulomb model, as discussed in Section 1.6.2.
1.6.4.1 Hardening Parameters
The two hardening parameters for this model (κ s and κ t ) are defined as the sum of some incremental
measures of plastic shear and tensile strain for the zone, respectively. The zone-shear and tensile-
hardening increments are calculated as the volumetric average of hardening increments over all
tetrahedra involved in the zone.
The shear-hardening increment for a particular tetrahedron is a measure of the second invariant of
the plastic shear-strain increment tensor for the step, given as
1 p p p p p
κ = √ (1 s − ms )2 + (ms )2 + (3 s − ms )2
s
(1.141)
2
p
where ms is the volumetric plastic shear strain increment,
p 1 p p
ms = (1 s + 3 s ) (1.142)
3
The tetrahedron tensile-hardening increment is the magnitude of the plastic tensile-strain increment,
p
κ t = |3 t | (1.143)
The plastic-strain increments involved in the preceding formula may be derived from the definition
Eq. (1.4) of the flow rule, using Eqs. (1.85) and (1.89). They have the form
p
1 s = λs
p
3 s = −λs Nψ
p
3 t = λt (1.144)
User-defined functions for zone yielding parameters can be determined by back-analysis of the
post-failure behavior of a material. Consider a one-dimensional stress-strain curve, σ versus ,
which softens upon yield and attains some residual strength (Figure 1.12).
yield
σ
e=ee e=ee+ep
The curve is linear to the point of yield; in that range, the strain is elastic only: = e . After yield,
the total strain is composed of elastic and plastic parts: = e + p . In the softening/hardening
model, the user defines the cohesion, friction, dilation and tensile strength variance as a function
of the plastic portion, p , of the total strain. These functions, which could in reality be sketched as
indicated in Figure 1.13, are approximated in FLAC 3D as sets of linear segments (Figure 1.14).
C φ
ePS ePS
(a) (b)
Figure 1.13 Variation of cohesion (a) and friction angle (b) with plastic strain
C φ
ePS ePS
(a) (b)
Hardening and softening behaviors for the cohesion, friction and dilation in terms of the shear
parameter ps (see Eq. (1.141)) are provided by the user in the form of tables. Each table contains
pairs of values: one for the parameter, and one for the corresponding property value. It is assumed
that the property varies linearly between two consecutive parameter entries in the table. Softening
of the tensile strength is described in a similar manner, using the parameter pt (see Eq. (1.142)).
The criterion for failure in the matrix used in this model is sketched in the principal stress plane
(σ1 , σ3 ) in Figure 1.15. (Recall that compressive stresses are negative and, by convention, σ1 ≤
σ2 ≤ σ3 .)
The failure envelope is defined by two Mohr-Coulomb failure criteria: f2s = 0 and f1s = 0 for
segments A − B and B − C, and a tension failure criterion f t = 0 for segment C − D.
The shear failure criterion has the general form f s = 0. The criterion is characterized by a cohesion,
c2 , and a friction angle, φ2 , for segment A − B, and by a cohesion, c1 , and a friction angle, φ1 ,
for segment B − C. The tensile failure criterion is specified by means of the tensile strength, σ t
(positive value); thus we have
f s = σ1 − σ3 Nφ + 2c Nφ (1.145)
f t = σ3 − σ t (1.146)
where
1 + sin(φ)
Nφ = (1.147)
1 − sin(φ)
Note that the tensile cap acts on segment B − C of the shear envelope and, for a material with
nonzero friction angle φ1 , the maximum value of the tensile strength is given by
c1
σ t max = (1.148)
tan φ1
σ3
σ1
=
σ 3
f t =0 c2/tan φ2
C
f =0
s
1
D
B 1 σt c1/tan φ1
N φ1
σ1
f 2 =0
s
1
Nφ 2
A
In the model formulation, elastic guesses for the stresses are first evaluated for the step using total
strain increments. Plastic yielding is detected if the corresponding stress point (σ1I , σ3I ) lies outside
the failure surface representation in Figure 1.15. In this case, a stress correction must be applied to
the elastic guess. It is determined by allowing plastic flow to occur, in order to restore the condition
f2s = 0, f1s = 0 or f t = 0, depending on the position of the stress point above A − B, B − C or
C − D. (Bisectors are used at B and C to delimit the domain of failure attached to a particular
segment of the yield surface.)
The usual assumption is made: that total strain increments can be decomposed into elastic and
plastic parts. The flow rule for plastic yielding has the form
p ∂g
i = λ (1.149)
∂σi
where i = 1,3. The potential function, g, for shear yielding is g s . This function corresponds to the
nonassociated law,
g s = σ1 − σ3 Nψ (1.150)
where ψ, the dilation angle, is equal to ψ2 for failure along A − B, and ψ1 along B − C, and
1 + sin(ψ)
Nψ = (1.151)
1 − sin(ψ)
The potential function for tensile yielding is g t . It corresponds to the associated flow rule,
g t = σ3 (1.152)
The plastic strain increments for shear failure have the form
p
1 s = λs
p
2 s = 0 (1.153)
p
3 s = −λs Nψ
where
I − σ I N + 2c N
σ1 3 φ φ
λs = (1.155)
(α1 − α2 Nψ ) − (−α1 Nψ + α2 ) Nφ
and, by definition:
4
α1 = K + G
3
(1.156)
2
α2 = K − G
3
In turn, the plastic strain increments for tensile failure have the form
p
1 t = 0
p
2 t = 0 (1.157)
p
3 t = λt
σ1 = −λt α2
σ3 = −λt α1
where
σ3I − σ t
λ =
t
(1.159)
α1
1.6.5.2 Failure Criterion and Flow Rule for the Weak Plane
The stresses, corrected for plastic flow in the matrix, are resolved into components parallel and
perpendicular to the weak plane, and tested for ubiquitous-joint failure. The failure
criterion is
expressed in terms of the magnitude of the tangential traction component, τ = σ1 3 2 + σ2 3 2 ,
and the normal traction component, σ3 3 , on the weak plane.
The failure criterion is represented in Figure 1.16 and corresponds to two Mohr-Coulomb failure
criteria (f2s = 0 for segment A − B, and f1s = 0 for segment B − C) and a tension failure criterion
(f t = 0 for segment C − D). Each shear criterion has the general form f s = 0, and is characterized
by a cohesion and a friction angle cj , φj , equal to cj2 , φj2 along segment A − B, and cj1 , φj1 along
B − C. The tensile criterion is specified by means of the tensile strength, σjt (positive value). Thus,
we have
Note that for a weak plane with nonzero friction angle φj1 , the maximum value of the tensile strength
is given by
cj1
σjt max = (1.162)
tan φj1
A
τ
f2 s=
0
B
f1 s= f t =0
0 Cj2
Cj1 C
φj1 φj2
D σ3'3'
σtj
Yield is detected, and stress corrections applied, using a technique similar to the one described in
the matrix context.
Here, the flow rule for plastic yielding has the form
p ∂g
3s3 = λ
∂σ3 3
∂g
γ ps = λ (1.163)
∂τ
p 2 p 2
γ ps
= 1s3 + 2s3 (1.164)
The potential function, g, for shear yielding on the weak plane is g s . It corresponds to the nonas-
sociated law,
where ψj , the dilation angle, is equal to ψj2 for failure along A − B, and ψj1 along B − C.
The potential function g for tensile yielding on the weak plane is g t . It corresponds to the associated
flow rule,
g t = σ3 3 (1.166)
The local plastic strain increments for shear failure are such that
p
3s3 = λs tan ψj
γ ps = λs (1.167)
τ = −λs 2G (1.168)
where
τ O + σ3O 3 tan φj − cj
λ =
s
(1.169)
2G + α1 tan ψj tan φj
and the superscript O indicates values obtained just before detection of failure on the weak plane.
The plastic corrections for the local shear stress components on the weak plane are derived by
scaling:
σ1O 3
σ1 3 = τ O
τ
σ2O 3
σ2 3 = τ (1.170)
τO
In turn, local plastic strain increments for tensile failure have the form
p
3t3 = λt
γ pt = 0 (1.171)
σ1 1 = −λt α2
σ3 3 = −λt α1
where
σ3O 3 − σjt
λ =
t
(1.173)
α1
In large strain, the orientation of the weak plane is adjusted per zone to account for rigid-body
rotations, and rotations due to deformations. The corrections are identical to those described in
Section 1.6.3.6.
1.6.5.4 Hardening Parameters
In the bilinear strain-hardening/softening ubiquitous-joint model, some or all of the zone yielding
parameters (i.e., cohesion, friction, dilation and tensile strength) for the matrix and joint are modified
automatically, after the onset of plasticity, according to piecewise linear laws specified on input in
terms of a range of values for the hardening parameters. (See Section 1.6.4.3.) One table number
must be specified in the PROPERTY command for each softening parameter. (If no table property
number is specified, the parameter is taken as constant.) The corresponding table data contain pairs
of values for the parameter and the property between which a linear variation is assumed. The last
property value is used for values of the hardening parameter beyond the last one specified in the
table.
Four independent hardening parameters are used in this model:
(1) κ s measures the matrix plastic shear strain and is used to update the matrix cohesion,
friction and dilation;
(2) κ t measures the matrix plastic volumetric tensile strain and is used to update the matrix
tensile strength;
(3) κjs estimates the joint plastic shear strain and controls the joint cohesion, friction and
dilation update; and
(4) κjt evaluates the joint plastic volumetric tensile strain and controls the joint tensile strength
update.
The parameters are defined as the sum of incremental measures of plastic strain for the zone. The
zone-hardening increments are calculated as the volumetric average of hardening increments over
all tetrahedra involved in the zone.
The shear-hardening increment for a tetrahedron is the square root of the second invariant of the
incremental plastic shear-strain deviator tensor for the step. For the matrix, it is given as
1 p p p p p
κ = √ (1 s − ms )2 + (ms )2 + (3 s − ms )2
s
(1.174)
2
p
where ms is the volumetric plastic shear strain increment,
p 1 p p
ms = (1 s + 3 s ) (1.175)
3
and the plastic strain increments are given by Eq. (1.153), using the expression Eq. (1.155) for λs .
For the joint, the formula is
1 p p p
κjs = (3s3 )2 + (1s3 )2 + (2s3 )2 (1.176)
3
where the plastic strain increments are given by Eq. (1.167) (see Eq. (1.164)), using the form
Eq. (1.169) for λs .
The tetrahedron tensile-hardening increment is the plastic volumetric tensile-strain increment.
For the matrix, we have
p
κ t = 3 t (1.177)
where the plastic strain increment is given by Eq. (1.157), using the form Eq. (1.159) for λt .
For the joint, the expression is
p
κjt = 3t3 (1.178)
where the plastic strain increment is given by Eq. (1.171), using the form Eq. (1.173) for λt .
The implementation of the bilinear model in FLAC 3D proceeds as indicated above. An elastic guess,
σijI , is first computed using stress increments for the step evaluated by application of Hooke’s law to
the total strain increments, ij . Principal stresses are calculated, ordered such that σ1I ≤ σ2I ≤ σ3I ,
and tested for failure in the matrix using the yield criteria defined by Eqs. (1.145) and (1.146). In
principle, matrix failure is declared if the representation of the stress point (σ1I , σ3I ) falls outside
the yield surface in Figure 1.15. In this case, stress corrections are applied to the principal values of
the elastic guess, which depend on the position of the stress point above A − B, B − C or C − D.
(Bisectors are used at B and C to delimit the domain of failure attached to a particular segment of
the yield surface.)
The stress corrections for shear failure in the matrix are given by Eqs. (1.154) and (1.155), where
the parameters of cohesion, c, friction, φ, and dilation, ψ, have value c2 , φ2 , ψ2 for failure along
A − B, and c1 , φ1 , ψ1 for failure along B − C.
The stress corrections for tensile failure in the matrix are given by Eqs. (1.158) and (1.159).
The stress tensor components in the system of reference axes, σijO , are then calculated from the
corrected principal values by assuming that the principal directions have not changed during plastic
flow.
Local traction components on the weak plane are defined as σ3 3 and τ , with σ3 3 being the normal
component, and τ = σ1 3 2 + σ2 3 2 being the magnitude of the tangential traction component.
(The local axis 1 is in the dip direction, 2 the strike, and 3 the normal to the weak plane.) These
stresses are resolved from σijO and examined for ubiquitous-joint failure using the yield criteria
Eqs. (1.160) and (1.161). In principle, ubiquitous-joint failure is declared if the representation of
the stress point (σ3O 3 , τ O ) falls outside the yield surface in Figure 1.16. In this case, local stress
corrections, which depend on the position of the stress point in the vicinity of A − B, B − C or
C − D, are applied. (Bisectors are used at B and C to delimit the domain of failure attached to a
particular segment of the yield surface.)
The stress corrections for shear joint failure are given by Eqs. (1.168) to (1.170), where the param-
eters of cohesion, cj , friction, φj , and dilation, ψj , have values cj2 , φj2 , ψj2 for failure along A − B,
and cj1 , φj1 , ψj1 for failure along B − C.
The stress corrections for tensile joint failure are given by Eqs. (1.172) and (1.173).
Finally, the local stress components are resolved back into global axes.
In large-strain mode, the unit normal to the weak plane is adjusted per zone to account for body
rotations.
After determination of the new stresses for the step, the hardening parameters are incremented using
Eqs. (1.174), (1.176), (1.177) and (1.178). These parameters are then used to determine new values
of cohesion, friction, dilation and tensile strength for the matrix and the joint from the available
input tables.
It is assumed that the tensile strength of the material can never increase. Also, for material with
t
friction, the value will not exceed the maximum value σmax or σjt (see Eqs. (1.148) and (1.162)).
max
Note that, by default, the yield model is linear in both the matrix and the joint, in which case only
sections 1 (where f1s = 0) and 3 (where f t = 0) of the yield curve are recognized (even if properties
are assigned for section 2, where f2s = 0). To activate the bilinear laws, the property bimatrix and/or
the property bijoint must be set to 1.
Also, if the friction angles for sections 1 and 2 become equal, the model will be considered as linear,
and section 2 will be ignored (for the matrix and/or the joint, as appropriate). Section 2 will also be
ignored if the intersection of sections 1 and 2 corresponds to a stress point that violates the tensile
criterion.
1.6.5.6 MODEL Command and PROPERTY Keywords
dj2table number of table relating joint dilation, ψj 2 , to joint plastic shear strain
djtable number of table relating joint dilation, ψj 1 , to joint plastic shear strain
dtable number of table relating matrix dilation angle, ψ1 , to matrix plastic
shear strain
f2table number of table relating matrix friction angle, φ2 , to matrix plastic
shear strain
fj2table number of table relating joint friction angle, φj 2 , to joint plastic shear
strain
fjtable number of table relating joint friction angle, φj 1 , to joint plastic shear
strain
fr2 matrix friction angle, φ2
friction matrix friction angle, φ1
ftable number of table relating matrix friction angle, φ1 , to matrix plastic
shear strain
jc2 joint cohesion, cj 2
jcohesion joint cohesion, cj 1
jd2 joint dilation angle, ψj 2
jddirection dip direction [degrees] of weakness plane
jdilation joint dilation angle, ψj 1
jdip dip angle of weakness plane
jf2 joint friction angle, φj 2
jfriction joint friction angle, φj 1
jnx x-component of unit normal to weakness plane
jny y-component of unit normal to weakness plane
jnz z-component of unit normal to weakness plane
jtension joint tension limit, σjt
tjtable number of table relating joint tensile limit, σjt , to joint plastic tensile
strain
ttable number of table relating matrix tensile limit, σ t , to matrix plastic
tensile strain
Table 1.1 lists the properties by matrix and joint failure segments.
Note that the default tension limits for the matrix and weakness planes are the same as those in the
ubiquitous-joint model.
The following properties can be printed, plotted or accessed via FISH.
es plastic accumulated matrix plastic shear strain
Permanent volume changes caused by the application of isotropic pressure are taken into account in
this model by including, in addition to the shear and tensile failure envelopes in the FLAC 3D strain-
softening/hardening model, a volumetric yield surface (or “cap”). For simplicity, the cap surface,
defined by the “cap pressure” pc > 0, is independent of shear stress; it consists of a vertical line on
a plot of shear stress versus mean stress. The hardening behavior of the cap pressure is activated
by volumetric plastic strain, and follows a piecewise-linear law prescribed in a user-supplied table
(like the one described in Section 1.6.4.3). The tangential bulk and shear moduli evolve as plastic
volumetric strain takes place according to a special law defined in terms of a factor, R, assumed to
be constant, and defined as the ratio of elastic bulk modulus to plastic bulk modulus.
Only two additional material parameters and a table are required, in addition to those associated
with the strain-softening model:
(1) the initial value of pc , which corresponds to the maximum mean pressure that the material
has experienced in the past;
(2) the value of R, greater than unity, which controls the slope of the stress-strain curve on
volumetric unloading (the “swelling” line, in soil mechanics terms); and
(3) the table representation of the “hardening curve,” which relates cap pressure, pc , to plastic
volume strain, epv .
Hence, any laboratory-determined hardening behavior may be modeled within the constraints im-
posed by a two-parameter model.
1.6.6.1 Incremental Elastic Law
In the FLAC 3D implementation of this model, principal stresses σ1 , σ2 , σ3 are used. The principal
stresses and principal directions are evaluated from the stress tensor components, and ordered so
that (recall that compressive stresses are negative)
σ1 ≤ σ2 ≤ σ3 (1.179)
The corresponding principal strain increments e1 , e2 , e3 are decomposed:
p
ei = eie + ei i = 1, 3 (1.180)
where the superscripts e and p refer to elastic and plastic parts, respectively, and the plastic compo-
nents are nonzero only during plastic flow. (Note that extensional strains are positive.) It is assumed
that the plastic contributions of shear, tensile and volumetric yielding are additive, so we may write
p ps pt pv
ei = ei + ei + ei (1.181)
where the superscripts ps, pt and pv stand for plastic shear, plastic tensile and plastic volumetric
strain. By convention, in this section, the symbol e is used to refer to the minus volumetric strain
increment (e1 + e2 + e3 ) with plastic part ep and elastic part ee . The symbol epv refers
pv pv pv
to minus the value of the plastic volumetric strain (e1 + e2 + e3 ).
The incremental expression of Hooke’s law in terms of principal stress and strain has the form
where α1 = Kc + 4Gc /3, α2 = Kc − 2Gc /3 and Kc and Gc are the current tangential bulk and
shear moduli defined according to the following considerations.
Consider an isotropic compression test with increasing pressure, pc . As the material becomes more
compact, its plastic stiffness (dpc /depv ) usually increases. It seems reasonable that the elastic
stiffness will also increase, since the grains are being forced closer together. A simple rule is
adopted in this model whereby, under general loading conditions, the incremental elastic stiffness,
Kc , is a constant factor, R, multiplied by the current incremental plastic stiffness. The values of
bulk and shear modulus, K and G, supplied by the user are taken as upper limits to Kc and Gc ,
and it is assumed that the ratio Kc /Gc remains constant and equal to K/G. Using incremental
notation, this law is defined by the relations
pc
Kc = R Kc := min(Kc , K)
epv
(1.183)
Kc
Gc = G
K
where the factor R is given, and pc /epv is the current slope of the table of pc values.
The type of behavior exhibited by the double-yield model is illustrated in Figure 1.17, which shows a
nonlinear volumetric loading curve with several unloading excursions; these excursions are elastic,
with slope related by R to the plastic stiffness at the point of unloading.
Main
Loading
Pressure
Path
Unloading
e
- Volumetric Strain
The shear and tensile yield functions, referred to as f s and f t , have the form
f s = σ1 − σ3 Nφ + 2c Nφ (1.184)
f = σ − σ3
t t
(1.185)
where
and φ is the friction angle, c is the cohesion, and σ t is the tensile strength.
The volumetric yield function f v is defined as
1
fv = σ1 + σ2 + σ3 + pc (1.186)
3
where pc is the cap pressure.
The shear potential function, g s , corresponds to a nonassociated flow rule; the tensile and volumetric
potential functions, g t and g v , correspond to associated laws. They have the form
g s = σ1 − σ3 Nψ
g t = −σ3 (1.187)
1
g v = σ1 + σ2 + σ3
3
where
The shear and volume yield surfaces can harden (or soften), and the tensile yield surface can soften,
according to hardening rules that are specified by lookup tables (see Section 1.6.4.3). Entry to the
tables is by hardening parameters that record some measure of accumulated plastic strain. In shear
and tension, the hardening parameter incremental forms are
1
1 ps ps 2 1 ps 2 1 ps ps 2 2
e ps
= e1 − em + em + e3 − em
2 2 2
pt
ept = e3 (1.188)
where
ps ps ps
em = 1/3 e1 + e3
ps pt
ej , j = 1,3 and e3 are plastic shear and tensile strain increments in the principal directions.
pv pv pv
epv = |e1 + e2 + e3 | (1.189)
pv
where ej , j = 1,3 are plastic volumetric strain increments in the principal directions.
These hardening parameters are used in the tables to determine new values of friction, cohesion,
dilation, tensile strength and cap pressure. The current bulk and shear moduli are also calculated
from the table values as per Eq. (1.183).
Let the superscript I be used to represent the elastic guess obtained by adding to the old stresses
σijO , elastic increments computed using the total strain increments. In principal axes, we then have
In the FLAC 3D implementation, shear yield is detected if f s (σ1I , σ3I ) < 0, volumetric yield if
fv (σ1I , σ2I , σ3I ) < 0, and tensile yield if f t (σ3I ) < 0. Corresponding plastic corrections are evalu-
ated using the following techniques.
We first consider the case where tensile failure is not detected for the step, but both shear and
volumetric yield conditions are exceeded. Using Eqs. (1.180) and (1.181), the principal strain
increments may be expressed as
ps pv
ei = eie + ei + ei i = 1, 3 (1.191)
ps ∂g s
ei = λs
∂σi
pv ∂g v
ei = λv (1.192)
∂σi
where i = 1,3. Using Eq. (1.187), these expressions become, after differentiation:
ps
e1 = λs
ps
e2 = 0
ps
e3 = −λs Nψ (1.193)
and
pv 1 v
e1 = λ
3
pv 1
e2 = λv
3
pv 1
e3 = λv (1.194)
3
e1e = e1 − λs − λv /3
e3e = e3 + λs Nψ − λv /3
With these expressions for the elastic strain increments, Hooke’s incremental equations yield (see
Eq. (1.182))
where σiI , i = 1,3 are the initial trial stresses in Eq. (1.190), and σiN = σiO + σi , i = 1,3 are the
new principal stresses for the step.
To determine the multipliers λs and λv , we require that if shear and volumetric yielding occur, the
new stresses lie on both yield surfaces and we must have f s (σ1N , σ3N ) = 0 and f σ (σ1N , σ2N , σ3N )
= 0. Substituting Eq. (1.196) for σi , i = 1,3 in Eqs. (1.184) and (1.186), and solving for λs , we
obtain
f sI − f vI (1 − Nφ )
λs = (1.197)
α1 − α2 Nψ − α2 Nφ + α1 Nφ Nψ − K(1 − Nφ )(1 − Nψ )
Hence,
f vI
λv = − λs (1 − Nψ ) (1.198)
K
In these equations, the notation f I stands for the function f evaluated for the initial trial stresses.
Eqs. (1.197) and (1.198) can now be used to evaluate the new stresses from Eq. (1.196); these
stresses simultaneously satisfy both yield conditions and both flow rules.
If the element is only yielding in shear, then
λv = 0
f sI
λs = (1.199)
α1 − α2 Nψ − α2 Nφ + α1 Nφ Nψ
λs = 0
f vI
λv = (1.200)
K
Eq. (1.199) or Eq. (1.200) may be used in Eq. (1.196), as appropriate, to compute new stresses.
We now consider the case where tensile failure is detected by the condition f t (σ3I ) < 0. If
volumetric failure is not detected, we use the same technique and stress corrections as described
in the Mohr-Coulomb model. If volumetric failure is detected in addition to tensile failure, then
either f s (σ1I , σ3I ) ≤ 0 or f s (σ1I , σ3I ) > 0. We begin by assuming that all three yield conditions
are exceeded. We assume that the plastic contributions of shear, volumetric and tensile yielding are
additive:
ps pv pt
ei = eie + ei + ei + ei i = 1, 3 (1.201)
pt ∂g t
ei = λt i = 1, 3 (1.202)
∂σi
pt
e1 = 0
pt
e2 = 0 (1.203)
pt
e3 = −λt
Using the same reasoning as above, and Eqs. (1.193) and (1.194) for the shear and volumetric flow
rule, we obtain
The multipliers λs , λv and λt are determined by solving the system of three equations, f s (σ1N , σ3N )
= 0, f v (σ1N , σ2N , σ3N ) = 0 and f t (σ3N ) = 0. This gives
f tI (1 + 2Nφ ) + 3f vI − 2f sI
λs =
α2 − α1
f vI −3(1 + Nφ )f tI − 6f vI + 3f sI
λv = + (1.205)
K α2 − α1
−f [Nψ (1 + 2Nφ ) + 2 + Nφ ] − 3(1 + Nψ )f vI + (1 + 2Nψ )f sI
tI
λt =
α2 − α1
σ1N = σ t Nφ − 2c Nφ
σ2N = −3pc − σ t (1 + Nφ ) + 2c Nφ (1.206)
σ3N = σt
If only tensile and volumetric yield are detected, then λs = 0 in Eq. (1.204). The constants λv and λt
are determined by requiring that conditions f v (σ1N , σ2N , σ3N ) = 0 and f t (σ3N ) = 0 both be fulfilled.
After some manipulation, we obtain
α1 f vI + Kf tI
λv =
K(α1 − K)
f vI + f tI
λ =
t
(1.207)
α1 − K
3f vI + f tI
σ1N = σ1I −
2
3f + f tI
vI
σ2N = σ2 −
I
(1.208)
2
σ3N = σt
Hardening and softening behaviors for the cohesion, friction and dilation in terms of the shear
parameter eps (see Eq. (1.188)) are provided by the user in the form of tables. Softening of the
tensile strength is described in a similar manner using the parameter ept (see Eq. (1.188)). In turn,
the variation of cap pressure is specified in a table in terms of the parameter epv (see Eq. (1.189)).
Each table contains pairs of values: one for the parameter, and one for the corresponding property
value. It is assumed that the property varies linearly between two consecutive parameter entries in
the table.
In the implementation of the double-yield model in FLAC 3D, new stresses for the step are computed
using the current values of the model properties. In this process, an elastic guess, σijI , is first com-
puted by adding to the old stress components, increments calculated by application of Hooke’s law
to the total strain increment for the step. Principal stresses σ1I , σ2I , σ3I , and corresponding princi-
pal directions, are calculated and ordered. If these stresses violate the composite yield criterion,
corrections are applied to the elastic guess (as described in Section 1.6.6.4) to give the new stress
state. The stress tensor components in the system of reference axes are then calculated from the
principal values by assuming that the principal directions have not been affected by the occurrence
of a plastic correction.
Plastic strain increments are evaluated from Eqs. (1.193), (1.194) and (1.203), using relevant ex-
pressions of λs , λt and λv for the mode of failure taking place. Zone hardening increments are
then calculated as the surface average of values obtained from Eqs. (1.188) and (1.189) for all
triangles involved in the zone. The hardening parameters are updated, and new zone properties for
cohesion, friction, dilation, tensile strength and cap pressure are evaluated by linear interpolation
in the tables. New elastic constants are derived from the cap pressure table using Eq. (1.204). All
these properties are stored for use in the next step. The hardening or softening lags one timestep
behind the corresponding plastic deformation. In an explicit code, this error is small because the
steps are small.
For a material with friction, the maximum value of the tensile strength is evaluated from Eq. (1.79),
using the new cohesion and friction angle. This value is retained by the code if it is smaller than
the tensile strength updated from the table.
1.6.6.6 Choice of Volumetric Properties
The “hardening curve” and ratio, R, of elastic bulk modulus to plastic bulk modulus are volumetric
properties that may be derived from the results of a triaxial test in which axial stress and confining
pressure, p, are kept equal. This test (for which dep = depv ) is recommended because it is best to
determine the parameters related to a particular mode of failure from a test that only involves that
failure mode.
p
tan -1
( (
hKc
h+Kc dp
dp
Kc= dee
tan -1Kc
dpc
h= dep
e
dep de
de dep
R=
dee
Consider the experimental graph of minus mean stress (pressure) versus minus volumetric strain, for
an increasing stress level with a small unloading excursion, obtained from such a test and presented
in Figure 1.18. The volumetric strain increment de at a point of the main loading path (assuming
that we are above any initial pre-consolidation stress level) is composed of an elastic part, dee , and
a plastic part, dep . (Recall that, in this section, de, dee and dep refer to minus the value of the
volumetric strain.) The observed tangent modulus may be expressed as
dp hKc
= (1.209)
de h + Kc
dpc
h=
dep
dp
Kc = e (1.210)
de
With the preceding notation convention, the volumetric property, R, may be defined as
R = Kc / h (1.211)
dp Kc
= (1.212)
de 1+R
Kc
R= −1 (1.213)
dp/de
Values for dp/de and Kc can be estimated from main loading and unloading increments on the
graph. Hence, R can be calculated from Eq. (1.213). Note that in the context of this model, the
ratio R is assumed to be constant. Using Kc = Rh and h = dpc /dep in Eq. (1.209), we can write,
after some manipulation,
dpc 1+R dp
=h= (1.214)
dep R de
From this, it follows that values of pc for a particular ep can be obtained to the first approximation
by multiplying the value p on the graph corresponding to e = ep by the ratio (1 + R)/R. For
example, if R = 5, then the graph curve must be scaled by a factor of 1.2 to convert it to table values,
assuming no over-consolidation.
To be sure that input parameters are reasonable, a single-element test should be done with FLAC 3D,
exercising the double-yield model over stress paths similar to those of the physical tests and plotting
similar graphs.
As an illustration, the data file in Example 1.2 exercises the double-yield model for a material
that exhibits a response eleven times stiffer upon unloading than upon loading. The corresponding
project file is “double-yield.f3prj,” located in folder “datafiles\constitutive models\double-yield.”
The loading tangent modulus dp/de observed in the physical test was constant and equal to 10
MPa. The slope of unloading increments corresponded to a value Kc = 110 MPa. To define the
volumetric properties of the numerical model, we substitute those values in Eq. (1.213) and find that
R = 10. As can be seen from Eq. (1.214), the hardening curve has a constant slope corresponding
to dpc /dep = h = 11 MPa. The hardening table is derived from this result, assuming no over-
consolidation.
Note that the input value for bulk modulus, K, must be higher than Kc (see Eq. (1.204)). The input
shear modulus controls the ratio of G/K. In this example,
Kc
Gc = G = 50.77 MPa
K
3Kc − 2Gc
v= = 0.3
2(3Kc + Gc )
Results of the numerical test are presented in the plot of minus vertical stress versus minus vertical
strain in Figure 1.19. The loading slope is 10 MPa, and the unloading slope is eleven times stiffer,
as expected.
A
xxf
xxA
b
x
MYdMxN x:
xvuxN
l
xM1YY
C-Ma-MYdMxdMGdYGdCx/
bdxb xx
xd
lxbMxb
xxx©
o
A
N x:
xvuxN
l
Figure 1.19 Single-element test in which unloading is eleven times stiffer than
loading
The maximum elastic moduli K and G should be estimated for the maximum pressure likely to
be produced in the model. They should not be set larger than this because FLAC 3D does its mass
scaling (for a stable timestep) on the basis of the moduli; setting them too high will give rise to
sluggish response (e.g., the model may be slow to converge to a steady-state solution). The elastic
moduli also act as a limit on plastic moduli.
If a material to be modeled has experienced some initial compaction (i.e., it is over-consolidated),
then pc may be set to this pre-consolidation pressure. In this case, ep must also be set, so that it is
consistent with pc and the given table (use PROPERTY ev plastic to set ep ).
The modified Cam-clay model is an incremental hardening/softening elastoplastic model. Its fea-
tures include a particular form of nonlinear elasticity and a hardening/softening behavior governed
by volumetric plastic strain (“density” driven). The failure envelopes are self-similar in shape, and
correspond to ellipsoids of rotation about the mean stress axis in the principal stress space. The
shear flow rule is associated; no resistance to tensile mean stress is offered in this model. See
Roscoe and Burland (1968) and Wood (1990) for detailed discussions on the modified Cam-clay
model.*
1.6.7.1 Incremental Elastic Law
The generalized stress components involved in the model definition are the mean effective pressure,
p, and deviatoric stress, q, defined as
1
p = − σii
3
q = 3J2 (1.215)
where the Einstein summation convention applies, and J2 is the second invariant of the effective
deviatoric-stress tensor [s]:
1
J2 = sij sij (1.216)
2
The incremental strain variables associated with p and q are the volumetric strain increment, p ,
and shear strain increment, q , and we have
p = −ii
2
q = 3J2 (1.217)
3
where J2 stands for the second invariant of the incremental deviatoric-strain tensor [e]:
1
J2 = eij eij (1.218)
2
* For convenience, we drop the qualifier “modified” in the following discussion. Also, recall that all
models are expressed in terms of effective stresses. In particular, all pressures referred to in this
section are effective pressures.
In the FLAC 3D plastic flow formulation, it is assumed that both plastic and elastic principal strain-
increment vectors are coaxial with the current principal stress vector. The generalized strain incre-
ments are then decomposed into elastic and plastic parts so that
p
p = pe + p
p
q = qe + q (1.219)
V
v= (1.220)
Vs
v
p = − (1.221)
v
And the new specific volume, v N , for the step may be calculated as
The incremental expression of Hooke’s law in terms of generalized stress and strains is
p = Kpe
q = 3Gqe (1.223)
√
where q = 3J2 , and J2 stands for the second invariant of the incremental deviatoric-stress
tensor:
1
J2 = sij sij (1.224)
2
In the Cam-clay model, the tangential bulk modulus, K, in the volumetric relation Eq. (1.223)
is updated to reflect a nonlinear law derived experimentally from isotropic compression tests.
The results of a typical isotropic compression test are presented in the semi-logarithmic plot of
Figure 1.20:
vλ
normal
consolidation line
vκA
A
vκB
κ B
1
swelling lines λ
1
ln p1 ln p
Figure 1.20 Normal consolidation line and unloading-reloading (swelling)
line for an isotropic compression test
As the normal consolidation pressure, p, increases, the specific volume, v, of the material decreases.
The point representing the state of the material moves along the normal consolidation line, defined
by the equation
p
v = vλ − λ ln (1.225)
p1
where λ* and vλ are two material parameters, and p1 is a reference pressure. (Note that vλ is the
value of the specific volume at the reference pressure.)
An unloading-reloading excursion, from point A or B on the figure, will move the point along an
elastic swelling line of slope κ, back to the normal consolidation line where the path will resume.
The equation of the swelling lines has the form
p
v = vκ − κ ln (1.226)
p1
* λ is used by Wood (1990) to define the slope of the normal consolidation line. It should not
be confused with the plastic (volumetric) multiplier, λv , used in the plasticity flow rule given in
Section 1.6.7.3.
where κ is a material constant, and the value of vκ for a particular line depends on the location
of the point on the normal consolidation line from which unloading was performed (i.e., vκA for
unloading from point A, and vκB for unloading from point B in Figure 1.20).
The recoverable change in specific volume, v e , may be expressed in incremental form after
differentiation of Eq. (1.226):
p
v e = −κ (1.227)
p
vp e
p = p (1.228)
κ
In the Cam-clay model, it is assumed that any change in mean pressure is accompanied by elastic
change in volume according to the preceding expression. Comparison with Eq. (1.223), therefore,
suggests the following expression for the tangent bulk modulus of the Cam-clay material:
vp
K= (1.229)
κ
Under more general loading conditions, the state of a particular point in the medium might be
represented by a point, such as A, located below the normal consolidation line in the (v, ln p) plane
(see Figure 1.21). By virtue of the law adopted in Eq. (1.226), an elastic path from that point
proceeds along the swelling line through A.
normal
A consolidation line
Δvp
vcA
Δvp
Δve
A'
A'
vc
A A'
ln pc ln pc ln p
The specific volume and mean pressure at the intersection of swelling line and normal consolidation
line are referred to as (normal) consolidation (specific) volume and (normal) consolidation pressure
(vcA and pcA , in the case of point A). Consider an incremental change in stress bringing the point from
state A to state A . At A , there is a corresponding consolidation volume, vcA , and consolidation
pressure, pcA . The increment of plastic volume change, v p , is measured on the figure by the
vertical distance between swelling lines (associated with points A and A ), and we can write, using
incremental notation,
pc
v p = −(λ − κ) (1.230)
pc
After division of the left and right member by v, we obtain, comparing with Eq. (1.221),
p λ − κ pc
p = (1.231)
v pc
Hence, whereas elastic volume changes take place whenever the mean pressure changes, plastic
volume changes occur only when the consolidation pressure changes.
The yield function corresponding to a particular value pc of the consolidation pressure has the form
where M is a material constant. The yield condition f = 0 is represented by an ellipse with horizontal
axis pc and vertical axis Mpc in the (q, p)-plane (see Figure 1.22). Note that the ellipse passes
through the origin; hence, the material in this model is not able to support an all-around tensile
stress.
The failure criterion is represented in the principal stress space by an ellipsoid of rotation about
the mean stress axis (any section through the yield surface at constant mean effective stress, p, is a
circle).
The potential function g corresponds to an associated flow rule, and we have
g = q 2 + M 2 p(p − pc ) (1.233)
e
lin
te
ta
plastic dilation als
itic
epp < 0 cr
p plastic compaction
qcr = M c
2
epp > 0
pc pc p
pcr =
2
The flow rule used to describe plastic flow has the form
p ∂g
p = λv
∂p
p ∂g
q = λv (1.234)
∂q
p
p = λv ca
p
q = λv cb (1.235)
where
ca = M 2 (2p − pc )
cb = 2q (1.236)
The elastic strain increments may be expressed from Eq. (1.219) as total minus plastic increments.
In further using Eq. (1.235), the elastic laws in Eq. (1.223) become
p = K(p − λv ca )
q = 3G(q − λv cb ) (1.237)
Let the new and old stress states be referred to by the superscripts N and O, respectively. Then, by
definition, for one step:
pN = pO + p
q N = q O + q (1.238)
p N = pI − λv Kca
q N = q I − λv 3Gcb (1.239)
where the superscript I is used to represent the elastic guess obtained by adding to the old stresses,
elastic increments are computed using the total strain increments:
pI = pO + Kp
q I = q O + 3Gq (1.240)
The parameter λv may now be defined by requiring that the new stress point be located on the yield
surface. Substitution of pN and q N (as given by Eq. (1.239)) for p and q in f (q, p) = 0 gives, after
some manipulation (see Eq. (1.232)),
where
a =(MKca )2 + (3Gcb )2
b = − Kca caI + 3Gcb cbI (1.242)
c =f (q I , pI )
Of the two roots of this equation, the one with the smallest magnitude must be retained.
Note that at the critical point corresponding to pcr = pc /2, qcr = Mpc /2 in Figure 1.22, the
normal to the yield curve f = 0 is parallel to the q-axis. Since the flow rule is associated, the plastic
volumetric strain-rate component vanishes there. As a result of the hardening rule Eq. (1.231),
the consolidation pressure, pc , will not change. The corresponding material point has reached
the critical state in which unlimited shear strains occur with no accompanying change in specific
volume or stress level.
The new stress components, σijN , are expressed in terms of old and new generalized stress values,
using the expressions
where
qN
sij N = sij I (1.244)
qI
The size of the yield curve is dependent on the value of the consolidation pressure, pc (see
Eq. (1.232)). This pressure is a function of the plastic volume change, and varies with the specific
volume as indicated in Eq. (1.231). The consolidation pressure is updated for the step, using the
formula
p v
pcN = pc (1 + p ) (1.245)
λ−κ
p
where p is the plastic volumetric strain increment for the step, v is the current specific volume
and λ and κ are material parameters, as given previously in Eq. (1.225).
1.6.7.5 Initial Stress State
The Cam-clay model in FLAC 3D is only applicable to material in which the stress state corresponds
to a compressive mean effective stress. This model is not designed to predict the behavior of
material in which this condition is not met. In particular, the initial state of the material (just before
application of the Cam-clay model) must be consistent with this requirement. The initial stress
state may be specified using the INITIAL command, or may be the result of a run in which another
constitutive model has been used. An error message will be issued if the Cam-clay model is applied
to a zone where the initial effective pressure, defined as p0 , is negative.
1.6.7.6 Over-consolidation Ratio
The over-consolidation ratio, R, is defined as the ratio between the zone initial pre-consolidation
pressure (a material property) and initial pressure, p0 :
pc0
R= (1.246)
p0
In the implementation of the Cam-clay model in FLAC 3D, an elastic guess, σijI , is first computed
after adding to the old stress components, increments calculated by application of Hooke’s law to
the total strain increment for the step.
Elastic guesses for the mean pressure, pI , and deviatoric stress, q I , are calculated using Eq. (1.215).
If these stresses violate the criterion for yield and f (q I , pI ) > 0 (see Eq. (1.232)), plastic deforma-
tion takes place and the consolidation pressure must be updated. In this situation, a correction must
be applied to the elastic guess to give the new stress state. New stresses pN and q N are first evaluated
from Eq. (1.239) using the expression for λ corresponding to the root of Eqs. (1.241) and (1.242)
with smallest magnitude. Note that in this version of the code, the expressions in Eq. (1.236) for
ca and cb are evaluated using the elastic guess; the error associated with this technique is expected
to be small, provided the steps are small. New stress tensor components in the system of reference
axes are hence evaluated using Eqs. (1.243) and (1.244).
Volumetric strain increment, p , and mean pressure, p, for the zone are computed as average
over all involved tetrahedra (see Eqs. (1.215) and (1.217)). The zone volumetric strain, p , is
incremented, and the zone specific volume, v, is updated, using Eq. (1.222). In turn, the new zone
consolidation pressure is calculated from Eq. (1.245), and the tangential bulk modulus is updated
using formula Eq. (1.229).
If a nonzero value for the Poisson ratio property is imposed, a new shear modulus is calculated from
the expression G = 1.5(1 − 2ν)K/(1 + ν). Otherwise, G is left unchanged as long as the condition
0 ≤ ν ≤ 0.5 is satisfied; if it is not, G is assigned a value of ν = 0 or ν = 0.5, as appropriate.
The new values for the consolidation pressure and shear and bulk moduli are then stored for use in the
next timestep. The material properties thus lag one timestep behind the corresponding calculation.
In an explicit code, this error is small because the steps are small.
1.6.7.8 Determination of the Input Parameters
Frictional constant M – M is the ratio of q/pcr at the critical state line. Therefore, a series of
triaxial tests (drained or undrained with pore-pressure measurement) can be used to obtain this
constant. These tests should be carried out to large strains to ensure that the final values of pcr and
q are close to the critical state line. The slope of a best-fitting line of q vs pcr will be the parameter
M.
M is related to the effective stress friction angle, φ , of the Mohr-Coulomb yield function. However,
since the Cam-clay critical state line is dependent on the intermediate stress σ2 , while Mohr-
Coulomb is not, the relation between M and φ will be different for different values of σ2 at
yield. (This condition is similar to the relation between Mohr-Coulomb and Drucker-Prager yield
functions.) For triaxial compression tests,
6 sin φ
M= (1.247)
3 − sin φ
6 sin φ
M= (1.248)
3 + sin φ
The slopes of the normal consolidation and swelling lines (λ and κ) – Ideally, these two param-
eters should be obtained from an isotropically loaded triaxial test (q = 0), with several unloading
excursions. The slope of the normal compression line in a v versus ln p plot will be the parameter
λ. The slope of an unloading excursion in the same plot will be the parameter κ.
These two parameters can also be derived from an oedometer test, making certain assumptions. Let
σv and σH be the vertical and horizontal stresses in an oedometer test. In most oedometer apparatus
it is not possible to measure the horizontal stresses, σH , so the mean stress, p = (σv + 2σH )/3,
is not known. However, experimental data show that the ratio of horizontal to vertical effective
stresses, K0 , is constant during normal compression. Since p = σv (1 + 2K0 )/3 along the normal
compression line, the slope of v vs ln p will be equal to the slope of e versus ln σv , where e is the
void ratio = v − 1.
The compression index, Cc , is calculated as the slope of e vs log10 (σv ). So the parameter λ will be
λ = Cc / ln(10) (1.249)
Experimental data show that along a swelling line in an oedometer test, K0 is not constant, so an
estimate of κ based on the swelling coefficient, Cs , will only be an approximation.
κ ≈ Cs / ln(10) (1.250)
p
v = vλ − λ ln (1.251)
p1
The specific volume intercept, , at the critical state line for p = p1 , is given by
= vλ − (λ − κ) ln(2) (1.252)
In a soil, the undrained shear strength, cu , is uniquely related to the specific volume, vcr , by the
equation
Mp1 − vcr
cu = exp (1.253)
2 λ
Thus, the value of for a given p1 , and therefore vλ , can be calculated if the undrained shear
strength for a particular specific volume, vcr , along with the parameters M, λ and κ, is known.
Pre-consolidation pressure, pc0 – The pre-consolidation pressure determines the initial size of the
yield surface in the equation
If a sample has been submitted to an isotropic loading path, pc0 will be the maximum past mean
effective stress. If the sample has followed other non-isotropic paths, pc0 has to be calculated from
the previous maximum p and q, using the Eq. (1.254).
The maximum vertical effective stress can be calculated from an oedometer test using Casagrande’s
method (for details, see Britto and Gunn 1987). Some hypothesis has to be made about the maximum
horizontal effective stress. A common hypothesis is Jaky’s relation:
σh max
Knc = 1 − sin φ (1.255)
σv max
where Knc is the coefficient of horizontal (σh max ) to vertical σv max stress at rest for normally
consolidated soil. For example, if a soil with an effective friction angle of 20◦ has experienced a
maximum vertical effective stress, σv max = 1 MPa. Then, using Jaky’s relation,
Substituting these two values in the yield function Eq. (1.232), we obtain the pre-consolidation
pressure,
2
qmax
pc0 = pmax + = 1.026 MPa
M 2 × pmax
Initial values for specific volume, v0 , and bulk modulus, K – Given an initial effective pressure
p0 , the initial specific volume v0 must be consistent with the choice of parameters κ, λ, p1 and pc0 .
The initial value v0 is calculated to correspond to the value of the specific volume corresponding to
p0 on the swelling line through the point on the normal consolidation line at which p = pc0 . From
Figure 1.7, it follows that
pc0 pc0
v0 = vλ − λ ln + κ ln (1.256)
p1 p
vλ
normal consolidation line
swelling line
v0
pc0
κ ln
p0
pc0
vλ-λ ln
p1
The initial value of the bulk modulus, K, may in turn be evaluated using Eq. (1.229), which gives
v0 p0
K= (1.257)
κ
In FLAC 3D, the default values for the properties v0 and K are evaluated using Eqs. (1.256) and
(1.257) when the first step command is issued.
Note that the mean effective pressure, p, must be initialized corresponding to the initial stress state
before stepping begins with the Cam-clay model. This can be done with a FISH function (e.g., see
the function camclay ini p in Example 1.3).
Maximum value of the elastic parameters K and G – In the Cam-clay model, the value of the bulk
modulus, K, changes as a function of the specific volume and the mean stress:
vp
K= (1.258)
κ
The input values of Kmax and G are used in the mass scaling calculation performed in FLAC 3D to
ensure numerical stability. This calculation is done once every time a STEP command is issued.
These input values should be chosen to give an upper bound to the sum (K + 4/3G), as evaluated
by the model between two consecutive STEP commands. Values should not, however, be set too
high or the model may be slow to converge. They should be selected based on the stress level in
the problem.
G or ν – The modified Cam-clay model in FLAC 3D allows the user to specify either a constant shear
modulus or a constant Poisson’s ratio.
If no Poisson’s ratio is specified, a constant shear modulus equal to the input value is assumed.
Then the Poisson’s ratio will vary as a function of the specific volume and the mean stress:
vp
3 − 2G
ν= κ
vp (1.259)
6 κ + 2G
If a nonzero Poisson’s ratio is specified, the shear modulus will vary at the same rate as the bulk
modulus in order to maintain a constant Poisson’s ratio:
vp
3 (1 − 2ν)
G= κ
(1.260)
2 (1 − 2ν)
The following data file in Example 1.3 exercises the Cam-clay model for a normally consolidated
material subjected to several load-unload excursions in an isotropic compression test. The results
are shown in Figures 1.24, 1.25 and 1.26. The corresponding project file is “cam-clay.f3prj,” located
in folder “datafiles\constitutive models\cam-clay.”
endif
global lnp = ln(sp)
global svol = z_prop(p_z,’cv’)
global mk = z_prop(p_z,’bulk’)
global mg = z_prop(p_z,’shear’)
global cpc = z_prop(p_z,’mpc’)
end
; ... loading-unloading excursions ...
def trip
local i
loop i(1,5)
command
ini xvel -0.5e-4 range x 0.9 1.1
ini yvel -0.5e-4 range y 0.9 1.1
ini zvel -0.5e-4 range z 0.9 1.1
step 300
ini xvel mul -.1 yvel mul -.1 zvel mul -.1
step 1000
ini xvel mul -1. yvel mul -1. zvel mul -1.
step 1000
end_command
end_loop
end
; --- histories ---
hist nstep 20
hist add unbal
hist add fish @path
hist add fish @sp
hist add fish @lnp
hist add fish @sq
hist add fish @svol
hist add fish @mk
hist add fish @mg
hist add gp zdisp 0 0 1
; --- test ---
@trip
;
save camiso
return
A
,rp,0N 0v
0fl0N
g
0pperr
8G,/G,rp,0p,:pr:p50H
.0 0CdN4o
g0i80 i
000S
N 0v
0fl0N
g
A
,rp,0! 04
0"l0!
g
0pperr
8G,/G,rp,0p,:pr:p50H
50
0C#!$o
g0S0
0C#!$o
! 04
0"l0!
g
A
,rp,0> 0H
0<l0>
g
0pperr
8G,/G,rp,0p,:pr:p50(d
90%0C+>-o
,g.5p5ur,00Ai700Sg8rp/ur.
/00C+>-o
,gerrrur,00Ai700,gerrrur,
g0i80 i
000S
> 0H
0<l0>
g
d
The Hoek-Brown failure criterion is an empirical relation that characterizes the stress conditions
that lead to failure in intact rock and rock masses. It has been used very successfully in design
approaches that use limit equilibrium solutions, but there has been little direct use in numerical
solution schemes. Alternatively, equivalent friction and cohesion have been used with a Mohr-
Coulomb model that is matched to the nonlinear Hoek-Brown strength envelope at particular stress
levels (e.g., see “HOEK.FIS” in Section 3 in the FISH volume in the FLAC 7.0 manual). Numerical
solution methods require full constitutive models, which relate stress to strain in a general way; in
addition to a failure (or yield) criterion, a “flow rule” is also necessary, in order to provide a relation
between the components of strain rate at failure. There have been several attempts to develop a full
constitutive model from the Hoek-Brown criterion: Pan and Hudson (1988), Carter et al. (1993)
and Shah (1992), for example. These formulations assume that the flow rule has some fixed relation
to the failure criterion, and that the flow rule is isotropic, whereas the Hoek-Brown criterion is not.
In the formulation described here, there is no fixed form for the flow rule: it is assumed to depend
on the stress level, and possibly some measure of damage.
In what follows, the failure criterion is taken as a yield surface, using the terminology of plasticity
theory. Usually, a failure criterion is assumed to be a fixed, limiting stress condition that corresponds
to ultimate failure of the material. However, numerical simulations of elastoplastic problems allow
continuing the solution after “failure” has taken place, and the failure condition itself may change
as the simulation progresses (by either hardening or softening). In this event, it is more reasonable
to speak of “yielding” than failure. There is no implied restriction on the type of behavior that is
modeled: both ductile and brittle behavior may be represented, depending on the softening relation
used.
1.6.8.1 The General Formulation
The “generalized” Hoek-Brown criterion (Hoek and Brown, 1980 and 1998), adopting the conven-
tion of positive compressive stress, is
σ a
3
σ1 = σ3 + σci mb +s (1.261)
σci
where σ1 and σ3 are the major and minor effective principal stresses, and σci , mb , s and a are
material constants that can be related to the Geological Strength Index and rock damage (Hoek
et al. 2002). For interest, the unconfined compressive strength is given by σc = σci s a , and the
tensile strength by σt = - s σci / mb . Note that the criterion (Eq. (1.261)) does not depend on the
intermediate principal stress, σ2 . Thus, the failure envelope is not isotropic.
Assume that the current principal stresses are (σ1 ,σ2 , σ3 ), and that initial trial stresses (σ1t ,σ2t , σ3t )
are calculated by using incremental elasticity:
where E1 = K + 4G/3 and E2 = K − 2G/3, and (e1 , e2 , e3 ) is the set of principal strain
increments. If the yield criterion (Eq. (1.261)) is violated by this set of stresses, then the strain
increments (prescribed as independent inputs to the model) are assumed to be composed of elastic
and plastic parts:
p
e1 = e1e + e1
e2 = e2e (1.263)
p
e3 = e3e + e3
Note that plastic flow does not occur in the intermediate principal stress direction. The final stresses
f f f
(σ1 ,σ2 , σ3 ) output from the model are related to the elastic components of the strain increments.
Hence:
f p p
σ1 − σ1 = E1 (e1 − e1 ) + E2 (e2 + e3 − e3 )
f p p
σ2 − σ2 = E1 e2 + E2 (e1 − e1 + e3 − e3 ) (1.264)
f p p
σ3 − σ3 = E1 (e3 − e3 ) + E2 (e1 − e1 + e2 )
f p p
σ1 = σ1t − E1 e1 − E2 e3
f p p
σ2 = σ2t − E2 (e1 + e3 ) (1.265)
f p p
σ3 = σ3t − E1 e3 − E2 e1
p p
e1 = γ e3 (1.266)
p
where the factor γ depends on stress and is recomputed at each timestep. Eliminating e1 from
Eq. (1.265):
f p
σ1 = σ1t − e3 (γ E1 + E2 )
f p
σ2 = σ2t − e3 E2 (1 + γ ) (1.267)
f p
σ3 = σ3t − e3 (γ E2 + E1 )
σ
f a
f f
F = σ1 − σ3 − σci mb 3 + s = 0 (1.268)
σci
f f
By substituting values of σ1 and σ3 from Eq. (1.267), Eq. (1.268) can be solved iteratively for
p
e3 , which is then substituted in Eq. (1.267) to give the final stresses. The method of solution is
described later, but first the evaluation of γ is discussed.
1.6.8.2 Flow Rules
We need to consider an appropriate flow rule, which describes the volumetric behavior of the
material during yield. In general, the flow parameter γ will depend on stress, and possibly on
history. It is not meaningful to speak of a “dilation angle” for a material when its confining stress
is low or tensile, because the mode of failure is typically by axial splitting, not shearing. Although
the volumetric strain depends in a complicated way on stress level, we consider certain specific
cases for which behavior is well-known, and determine the behavior for intermediate conditions by
interpolation. Three cases are considered:
Associated Flow Rule
It is known that many rocks under unconfined compression exhibit large rates of volumetric expan-
sion at yield, associated with axial splitting and wedging effects. The associated flow rule provides
the largest volumetric strain rate that may be justified theoretically. This flow rule is expected to
apply in the vicinity of the uniaxial stress condition (σ3 ≈ 0). An associated flow rule is one in
which the vector of plastic strain rate is normal to the yield surface (when both are plotted on similar
axes). Thus,
p ∂F
ei = −γ (1.269)
∂σi
where the subscripts denote the components in the principal stress directions, and F is defined by
Eq. (1.268). Differentiating this expression, and using Eq. (1.266),
1
γaf = − (1.270)
1 + aσci (mb σ3 /σci + s)a−1 (mb /σci )
σ1
γrf = (1.271)
σ3
γcv = −1 (1.272)
1
γ = 1 σ3
(1.273)
1
γaf + ( γ1cv − γaf ) σ3cv
The equations presented above are implemented in a constitutive model named hoekbrown. One
difficulty with the failure criterion (Eq. (1.268)) is that real values for F do not exist if σ3 <
−sσci /mb . During an iteration process, this condition is likely to be encountered, so it is necessary
that the expression for F , and its first derivatives, be continuous everywhere in stress space. This
is fulfilled by adapting the composite expression
sσci σ a f
f f
if σ3 ≥ − then F = σ1 − σ3 − σci mb 3 + s = 0 (1.274)
mb σci
sσci σ a f
f f 3
if σ3 < − then F = σ1 − σ3 + σci mb +s =0 (1.275)
mb σci
p
To initialize the iteration, a starting value for e3 is taken as the absolute maximum of all the
strain increment components. This value, denoted by e1 , is inserted into Eq. (1.267), together
with the value for γ found from the flow-rule equations, and the resulting stress values inserted into
Eqs. (1.274) and (1.275). The resulting value of F is denoted by F1 . Taking the original value of F
as F0 (and the corresponding plastic strain increment of zero as e0 ), we can estimate a new value
of the plastic strain increment, using a variant of Newton’s method:
F1 e0 − F0 e1
e2 = (1.276)
F1 − F0
From this, we find a new value of F (call it F2 ) and, if it is sufficiently close to zero, the iteration
stops. Otherwise, we set F0 = F1 , F1 = F2 , e0 = e1 and e1 = e2 and apply Eq. (1.276)
again.
Tests show that the iteration scheme converges for all stress paths tried so far, including cases
in which s = 0 (material with zero unconfined compressive strength), which led to problems in
previous implementations. For high confining stresses, the iteration converges in one step, but at
low confining stresses, up to ten steps are necessary (the limit built into the code is presently 15).
In the Hoek-Brown model, the material properties σci , mb , s and a are assumed to remain constant,
by default. Material softening after the onset of plastic yield can be simulated by specifying
that these mechanical properties change (i.e., reduce the overall material strength) according to a
softening parameter. The softening parameter selected for the Hoek-Brown model is the plastic
p p
confining strain component, e3 . The choice of e3 is based on physical grounds. For yield near
the unconfined state, the damage in brittle rock is mainly by splitting (not by shearing) with crack
p
normals oriented in the σ3 direction. The parameter e3 is expected to correlate with the microcrack
damage in the σ3 direction.
p p
The value of e3 is calculated by summing the strain increment values for e3 calculated by
Eq. (1.276). Softening behavior is provided by specifying tables that relate each of the proper-
p p
ties σci , mb , s and a to e3 . Each table contains pairs of values: one for the e3 value, and one
for the corresponding property value. It is assumed that the property varies linearly between two
consecutive parameter entries in the table.
A multiplier, μ (denoted as multable), can also be specified to relate the softening behavior to the
confining stress σ3 . The relation between μ and σ3 is also given in the form of a table. (See Cundall
et al. (2003) for an application of softening parameters.)
Triaxial compression tests are performed on models composed of Hoek-Brown material in FLAC 3D,
to verify the stress and strain paths that develop. The triaxial load conditions are illustrated in Fig-
ure 1.27. The triaxial tests are performed on a sample of Hoek-Brown material with properties of
mb = 5, s = 1, a = 0.5, σci = 1.0 and σ3cv = 1.5, and with elastic properties of E = 100 and
ν = 0.35. Compression loading tests are performed under two loading conditions: σ3 /σci = 0 and
1.0. The analytical solutions for stress and strain during compression loading are presented by the
plots shown in Figures 1.28 and 1.29. Example 1.4 lists the input commands for this test. The corre-
sponding project file is “hoek-brown.f3prj,” located in folder “datafiles\constitutive models\hoek-
brown.”
A single-zone model to simulate the triaxial loading tests is constructed in FLAC 3D. The FLAC 3D
results are compared to the analytical solutions in Figures 1.30 through 1.33. The solutions compare
within 1%.
A
e
L rM
r)-rL
w
N
A
,rp,0! 0L
0Ul0!
g
0,rrrr
:G,8G,rp,0p,AprAp,0)
Ae
A
,pS00 00C
o
ppS00 00CN*)L.3o
e
A
nSen
! 0L
0Ul0!
g
Figure 1.31 Triaxial compression test – lateral strain versus axial strain
(σ3 /σci = 0)
A
e
F rA
r(-rF
w
N
A
,rp,0! 0F
0Ul0!
g
0,rrrr
:G,8G,rp,0p,3pr3p.0(
Ae
A
,pS00 00C
o
ppS00 00CN*(F.Lo
e
A
nSen
! 0F
0Ul0!
g
Figure 1.33 Triaxial compression test – lateral strain versus axial strain
(σ3 /σci = 1.0)
define _input
; --- To reconstruct the compression analytical curves: ---
global _sig_conf = 0.0 ; negative is compression
global _max_eyy = -6.0e-2 ; <-- maximum ’driving’ strain
; (contraction negative)
;
; Plastic properties
global _sig_ci = 1.0 ; <-- enter UCS as positive always
global _mb = 5.0
global _s = 1.0
global _a = 0.5
global _sig3_cv = 1.5 ; <-- enter UCS as positive always
;
global _sig_tm2 = - _s*_sig_ci/_mb
;
; Elastic properties
global _young = 100
global _poiss = 0.35
global _bulk = _young/3.0/(1-2*_poiss)
global _shear = _young/2.0/(1+_poiss)
;
; Loading
global _cyc = 20000 ; <-- number of steps in which load is to be applied
global _delta_u = _max_eyy * 1.0
global _y_vel = 0.5*_delta_u / _cyc
global _minus_y_vel = -_y_vel
;
end
@_input
;
gen zone brick size 1 1 1
model mech hoekbrown
prop shear=@_shear bulk=@_bulk
prop hbsigci=@_sig_ci hbmb=@_mb hbs=@_s hba=@_a
prop hbs3cv=@_sig3_cv
prop dens = 1.0
;
define locptrs
global _zp1 = zone_head
global _gp11 = gp_near(0, 0, 0)
global _gp12 = gp_near(0, 1, 0)
global _gp21 = gp_near(1, 0, 0)
global _gp22 = gp_near(1, 1, 0)
end
@locptrs
;
define record_variables
;
global _disp_0 = 0.5*(gp_xdisp(_gp11) + gp_xdisp(_gp12))
global _disp_1 = 0.5*(gp_xdisp(_gp21) + gp_xdisp(_gp22))
global _eps_xx = -(_disp_0 - _disp_1)/1.0
;
global _disp_0 = 0.5*(gp_ydisp(_gp11) + gp_ydisp(_gp21))
global _disp_1 = 0.5*(gp_ydisp(_gp12) + gp_ydisp(_gp22))
global _eps_yy = -(_disp_0 - _disp_1)/1.0
;
global _sig_zz = z_szz(_zp1)
global _sig_xx = z_sxx(_zp1)
global _sig_yy = z_syy(_zp1)
;
global record_variables = 1.0
;
end
;
apply yvel @_y_vel range y 0.9 1.1
apply yvel @_minus_y_vel range y -0.1 0.1
his add id 1 fish @record_variables
hist nstep 1000
his add id 11 fish @_eps_xx
his add id 12 fish @_eps_yy
his add id 13 fish @_eps_yy
his add id 21 fish @_sig_xx
his add id 22 fish @_sig_yy
his add id 23 fish @_sig_zz
save ini
;
; unconfined
set @_sig_conf = 0.0
apply sxx=@_sig_conf
apply szz=@_sig_conf
ini sxx = @_sig_conf
ini syy = @_sig_conf
ini szz = @_sig_conf
step @_cyc
save conf0
;
; confined
restore ini
set @_sig_conf = -1.0 ; negative is compression
apply sxx=@_sig_conf
apply szz=@_sig_conf
ini sxx = @_sig_conf
ini syy = @_sig_conf
ini szz = @_sig_conf
step @_cyc
save conf1
ret
p
atable number of table relating a to e3
bulk elastic bulk modulus, K
p
citable number of table relating σci to e3
density mass density, ρ
p
hb e3plas accumulated plastic strain, e3
hb ind number of iterations
hba Hoek-Brown parameter, a
hbmb Hoek-Brown parameter, mb
hbs Hoek-Brown parameter, s
hbs3cv Hoek-Brown parameter, σ3cv
hbsigci Hoek-Brown parameter, σci
p
mtable number of table relating mb to e3
multable number of table relating a multiplier to σ3
shear elastic shear modulus, G
p
stable number of table relating s to e3
The Hoek-Brown criterion is used for plastic yielding when the minor principal stress, σ3 , is
compressive. The criterion is based on a nonlinear relation between major and minor principal
stresses, σ1 and σ3 , as shown previously in Eq. (1.261), and repeated here:
σ3 a
σ1 = σ3 + σci mb +s (1.277)
σci
where σci is the unconfined compressive strength of the intact rock, and mb , s and a are material
constants that can be related to the Geological Strength Index and rock damage (Hoek et al. 2002).
The Hoek-Brown envelope is extended, for σ3 tensile, by a combination of the Mohr-Coulomb
envelope (tangent to Hoek-Brown at σ3 = 0) and a tensile cutoff at σ3 = −sσci /mb .
The numerical implementation of the Hoek-Brown model uses a linear approximation, whereby
the nonlinear failure surface is continuously approximated by the Mohr-Coulomb tangent, at the
current stress level, σ3 . The logic is similar to that adopted in the FISH version of the model (see
“HOEK.FIS” in Section 3 in the FISH volume in the FLAC 7.0 manual). The current tangent
Mohr-Coulomb criterion is
σ1 = σ3 Nφc + 2cc Nφc (1.278)
where
1 + sin φc φc
Nφc = = tan2 ( + 45◦ ) (1.279)
1 − sin φc 2
The current (apparent) value of cohesion, cc , and friction, φc , are calculated using
φc = 2 tan−1 Nφc − 90◦ (1.280)
σ ucs
cc = c (1.281)
2 Nφc
where
σ3 !a−1
Nφc = 1 + amb mb +s (1.282)
σci
σ3 !a
σcucs = σ3 (1 − Nφc ) + σci mb +s (1.283)
σci
The Mohr-Coulomb envelope extension in the region of tensile σ3 is accounted for in the logic by
considering the following cap in Eqs. (1.282) and (1.283) (recall that compressive stress is positive):
σ3 = max(σ3 , 0) (1.284)
The tensile yield logic is the same as the one used for the strain-hardening/softening model; it is
described in Section 1.6.4.3.
The plastic strain increment for shear yielding is defined using the current Mohr-Coulomb flow
rule:
p ∂g 1 p
eij = ep − evol δij i = 1, 3 (1.285)
∂σij 3
where ep is the plastic flow increment intensity, g is the plastic potential function, and
p ∂g ∂g ∂g !
evol = ep + + (1.286)
∂σ11 ∂σ22 ∂σ33
g = σ1 − σ3 Nψc (1.287)
1 + sin ψc
Nψc = (1.288)
1 − sin ψc
∂g
The direction of plastic flow ( ∂σij
in Eq. (1.285)) is expressed using the plastic potential function.
The plastic flow increment intensity, ep , is derived from the current tangent Mohr-Coulomb yield
criterion, Eq. (1.278). There are three choices of flow rule for the model:
1. Input a dilation angle (ψc is a constant value) specified by property keyword hbpsi and
hb doption = 0;
2. Specify associated plastic flow (ψc is set equal to φc ) by setting hb doption = −1;
3. Set dilation as a fraction of the friction angle (ψc is set to a constant times φc ); hb doption
is a positive fraction.
Strain hardening/softening behavior can be prescribed for the Hoek-Brown properties, mb , s, a,
and σci . The input is via tables, and in terms of an evolution parameter. There are two choices for
evolution parameter: plastic shear strain (property keyword hb poption is set to 1), or plastic strain
in the direction of the least compressive principal stress, σ3 (property keyword hb poption is set to
0). The evolution parameter can be monitored via the property hb plas.
A simple regularization technique can be selected to address the issue of grid dependency on
softening behavior. To activate this technique, the grid zone size used to calibrate model properties
with experimental data is assigned to the property hb len. This property is the calibration length.
The input softening rate is then adjusted automatically to account for a different zone size used in
the full FLAC 3D model. Note that this technique is experimental and should be used with caution.
1.6.9.2 Triaxial Compression Test
The triaxial compression tests performed in Section 1.6.8.5 for the Hoek-Brown model described
in Section 1.6.8 are repeated for the modified Hoek-Brown model. Two compression loading tests
are performed: one test at zero confining stress, σ3 /σci = 0, and the other test at high confining
stress, σ3 /σci = 1.
The data file shown in Example 1.4 is used for these tests, with the Hoek-Brown model and
properties replaced by the modified Hoek-Brown model and properties. The corresponding project
file is “mhoek-brown.f3prj,” located in folder “datafiles\constitutive models\mhoek-brown.” For
the zero confining stress case, we specify an associated flow rule; this is done with the command
PROPERTY hb doption = -1. The following commands are specified for this case.
model mhoekbrown
prop bulk @ bulk shear @ shear
prop hbsigci=@ sig ci hbmb=@ mb hbs=@ s hba=@ a
prop hb do -1
For the higher confining stress case, we need to specify a dilation angle that is consistent with the
limiting constant-volume stress, σ3cv , = 1.5, chosen for this test in Section 1.6.8.5. We linearly
interpolate a value for dilation corresponding to the current confining stress level of σ3 = 1, relative
to a nonassociated zero dilation at σ3cv , = 1.5. The current dilation, ψc , is then taken to be a fraction
of the current friction angle, φc , using the linear interpolation:
ψc σ3
= 1 − cv = 0.333 (1.289)
φc σ3
The following commands are used to apply the modified Hoek-Brown model for this case:
model mechanical mhoekbrown
prop bulk @ bulk shear @ shear
prop hbsigci=@ sig ci hbmb=@ mb hbs=@ s hba=@ a
prop hb do 0.333
The FLAC 3D results for the zero confining stress case are compared to the analytical solution (from
Section 1.6.8.5) in Figures 1.34 and 1.35, and the results for the high confining stress case are
compared in Figures 1.36 and 1.37.
A
e
F rP
r(-rF
w
D
A
,rp,0! 0L
0Nl0!
g
0,rrrr
prGppG,rp,07Pr9P,90;
Ae
A
,pS00 00C
o
ppS00 00CD*)L.Ao
e
A
nSen
! 0L
0Nl0!
g
Figure 1.35 Triaxial compression test – lateral strain versus axial strain
(σ3 /σci = 0) – modified Hoek-Brown model
A
e
L rM
r)-rL
w
z
A
,rp,0! 0L
0Nl0!
g
0,rrrr
prGppG,rp,07Pr9P,90;
Ae
A
,pS00 00C
o
ppS00 00CD*)L.Ao
e
A
nSen
! 0L
0Nl0!
g
Figure 1.37 Triaxial compression test – lateral strain versus axial strain
(σ3 /σci = 1.0) – modified Hoek-Brown model
* If hb doption = 0, hb dilation = min(hbpsi, ψc ). Thus, hb dilation may not always be equal to the
specified value of hbpsi.
The double-yield model (see Section 1.6.6) is a shear and volumetric hardening/softening model
for the simulation of soil behavior. One flexible feature of the model is the capability for adding
user-defined hardening/softening laws, which are then communicated to the model by means of
tables. As mentioned by Davis and Sepulvadurai (2002), the general concept of hardening is very
useful whenever there is a need for model responses that are more detailed than possible with perfect
plasticity. In particular, a closed yield surface would be almost useless without volumetric (cap)
hardening.
The motivation for closing the yield surface by a cap on the mean stress axis is to permit plastic
behavior in response to an isotropic stress increase. This plasticity effect accounts for grain crushing
and rearrangement, and is particular to soils. In the double-yield model, the cap is a plane, normal to
the mean stress axis in stress space. The impact of this particular shape on the coefficient of lateral
earth pressure, K0 , as predicted by the model in uniaxial compression tests, has been considered
to be somewhat restrictive by some users. The Cysoil model is a modification of the double-yield
model that addresses this issue by accounting for a cap with an elliptic shape in the (p , q) plane.
The ratio of axes of the ellipse, α, determines the value of K0 , and is a material property for the
model, which can be chosen to match a known value in uniaxial compression.
In addition, when subjected to deviatoric loading, soils usually exhibit a decrease in stiffness,
accompanied by irreversible deformation. In most cases, the plot of deviatoric stress versus axial
strain obtained in a drained triaxial test may be approximated by a hyperbola. This feature has
been used by Duncan and Chang (1970) to formulate their well-known “hyperbolic soil” model.
The hyperbolic soil model of Duncan and Chang is a nonlinear elastic model that has been shown
to exhibit some drawbacks. These drawbacks include, for example, difficulty in detecting and
characterizing unloading/reloading and, in specific cases, producing a nonphysical bulk-modulus
value that can lead to an erroneous energy generation in the model. Because the Cysoil model
is formulated in the theory of hardening plasticity, it allows for an alternative expression of the
hyperbolic behavior (based on friction hardening), which is capable of addressing some of these
problems.*
When tested under drained triaxial conditions, soils generally exhibit shear-induced volume changes
that are strongly dependent on soil density. Typically, there is a tendency for the soil to contract
under small shear strains, and to dilate under larger strains, unless it is very loose (Byrne et al.
2003). In particular, when fluid fills the pores, it is this tendency of the soil skeleton to contract and
dilate that controls its liquefaction response. Also, the shear-stress/shear-strain response of loose
soil may exhibit a softening response under undrained conditions. It is the existence of a peak in
shear strength which may lead to instability (static liquefaction) during a monotonic load-controlled
process (Boukpeti 2001). Shear-induced volume changes can be accounted for in the Cysoil model
by means of a dilation hardening/softening law.
* A simplified version of the Cysoil model, called the Chsoil model, also addresses the difficulties of
the Duncan and Chang model and is provided as an alternative to the Duncan and Chang model.
See Section 1.6.11.
The elastic behavior is expressed using Hooke’s law. The incremental expression of the law in
terms of principal stress and strain* has the form
where α1 = K e + 4Ge /3, α2 = K e − 2Ge /3, and K e and Ge are current, tangent elastic bulk
and shear modulus, respectively. Some useful relations between K e , Ge , Young modulus, E e , and
Poisson’s ratio, ν, are listed for reference:
Ee Ee
Ke = Ge =
3(1 − 2ν) 2(1 + ν)
(1.291)
Ke 2(1 + ν)
=
Ge 3(1 − 2ν)
* Principal stress and strain components, represented by the symbol σi and ei (i = 1,3), respectively,
are positive in extension. Also, effective stresses are denoted by a prime. The principal effective
stresses are σ1 , σ2 , σ3 and, by convention, σ1 < σ2 < σ3 (i.e., σ1 is the most compressive stress).
Shear Yield Criterion and Flow Rule – Shear yielding is defined by a Mohr-Coulomb criterion. The
yield envelope is expressed in a form consistent with the cap formulation:
f = Mp − q + Nc (1.292)
where c is cohesion, M = 6 sin φm /(3 − sin φm ), and N = 6 cos φm /(3 − sin φm ). p is the mean
effective stress, p = −(σ1 + σ1 + σ1 ) / 3, and q is a measure of shear stress, defined as
g = M ∗ p − q ∗ (1.294)
where
In these equations, δ ∗ = (3 + sin ψm )/(3 − sin ψm ), and M ∗ = 6 sin ψm /(3 − sin ψm ). Also,
the mobilized dilatancy angle, ψm , is given in terms of plastic shear strain, γ p , by means of a
user-defined table. If no table is provided, it is assumed that dilation is constant, and equal to the
input value of dilation property.
Volumetric Cap Criterion and Flow Rule – Yielding on the cap is associated; the criterion is
q2
fc = + p2 − pc2 (1.296)
α2
where α is a dimensionless parameter, defining the shape of the elliptical cap in the (p , q) plane,
and pc is cap pressure. The hardening curve relating cap pressure, pc , to cap plastic volumetric
f t = σ t − σ3 (1.297)
The tensile strength, σ t , is given in terms of the plastic tensile-strain measure, ept , as defined in
Section 1.6.4.1, and input by means of a user-defined table. If no table is provided, it is assumed
that tensile strength is constant, and equal to the input value of the tensile strength property.
Hardening Parameters – The evolution parameters for shear, cap and tensile yielding are indepen-
dent. The parameter for shear yielding, γ p , is defined incrementally as
1
1 ps ps 2 1 ps 2 1 ps ps 2 2
γ =p
e1 − em + em + e3 − em (1.298)
2 2 2
where
ps 1 ps ps
em = e1 + e3
3
ps
and ej , j = 1, 3 are the principal plastic shear strain increments.
The evolution parameter for cap yielding is the modulus of plastic volumetric strain, ep , and its
increment is defined as (see Section 1.6.6.3)
p p p
ep = |e1 + e2 + e3 | (1.299)
p
where ej , j = 1, 3 are principal plastic strain increments from yielding on the cap.
The evolution parameter for tensile yielding is the modulus of plastic tensile strain, ept . The
increment of plastic tensile strain is defined as (also see Section 1.6.4.1)
pt
ept = e3 (1.300)
pt
where e3 is the increment of tensile plastic strain in the direction of the major principal stress
(recall that tensile stresses are positive).
1.6.10.3 Customizing the Cysoil Model
Cap Hardening – Soil stiffness usually increases in a nonlinear fashion as a function of isotropic
pressure. A cap-hardening table is used to specify a power law behavior. In most experimental
cases, soil volumetric behavior in an isotropic compaction test can be captured by a power law of
the form
m
dp p
= Kref
iso
(1.301)
de pref
where e is (minus) volumetric strain, Krefiso is the slope of the laboratory curve for p versus e at
reference effective pressure, pref , and m is a constant (m < 1). A typical graph, representative of
this law, is sketched with a small unloading excursion in Figure 1.38:
For input into the Cysoil model, a relation between cap pressure, pc , and plastic volumetric strain,
ep , is required. This relation can be obtained from the following considerations. For isotropic
compression dp = dpc , and we can write
dpc de dp
= (1.302)
dep dep de
Now, the total strain increment has elastic and plastic contributions: de = dee + dep . Substitution
of this expression for de in Eq. (1.302) gives
The Cysoil model assumes that the ratio of elastic modulus, K e , to hardening modulus, H , is equal
to a constant, R = K e / H , where, by definition, K e = dp /dee and H = dpc / dep . For isotropic
compression, dp = dpc , and we can write
dep
R= (1.304)
dee
Finally, using Eq. (1.304) in Eq. (1.303) gives, after some manipulation,
dpc 1 + R dp
= (1.305)
dep R de
dpc 1 + R iso p m
= Kref (1.306)
dep R pref
iso 1
1 + R Kref p 1−m
pc = pref (1 − m) e (1.307)
R pref
This formula is used to generate the input table of pc in terms of ep . The law has 4 parameters:
ref , m and R. Note that, since the Cysoil model assumes that K = R × dpc /de , by virtue
iso , p
Kref e p
m
p
K =
e
(1 + R)Kref
iso
(1.308)
pref
Friction Hardening – For most soils, the plot of deviatoric stress versus axial strain obtained in a
drained triaxial test may be approximated by a hyperbola. The model is supplemented by a friction
strain-hardening table to capture this hyperbolic behavior. For friction-hardening behavior, we adopt
the following (hyperbolic) incremental law, similar to the one implemented in the UBCSAND model
(Byrne et al. 2003):
Gp
d(sin φm ) = d(γ p ) (1.309)
p
2
sin φm
G = βG
p e
1− Rf : φm ≤ φf (1.310)
sin φf
In this formula, Ge is the elastic tangent shear modulus, φf is the ultimate friction angle, Rf (the
failure ratio) is a constant, smaller than 1 (0.9 in most cases), used to assign a lower bound for Gp ,
and β is a calibration factor.
The elastic tangent shear modulus is a function of p , and we have
m
p
G =
e
Geref (1.311)
pref
where Geref is elastic tangent shear modulus at reference effective pressure, pref , and m is a constant
(m ≤ 1), taken as 1 for this discussion. After substitution of Eq. (1.311) in Eq. (1.310), the resulting
expression in Eq. (1.309), and rearranging terms, we obtain
1−m
pref p d(sin φm )
d(γ ) = e
p
2 (1.312)
Gref pref sin φm
1− sin φf Rf
1−m
pref p sin φf 1
γ = e
p
−1 (1.313)
Gref pref Rf 1 − sin φm
sin φf Rf
pref sin φf 1
γ = e
p
−1 (1.314)
Gref Rf 1 − sin φm
R
sin φf f
This expression is used to generate the model input table of friction in terms of plastic shear strain.
As demonstrated in the example below, the use of this hardening law for modeling primary loading
in a triaxial test will produce a hyperbolic curve of deviatoric stress versus axial strain. The law
has 5 parameters: Geref , pref , Rf , φf and β.
where
sin φf − sin ψf
sin φcv = (1.317)
1 − sin φf sin ψf
and φf and ψf are ultimate (known) values of friction and dilation, respectively.
A table of dilation value versus plastic shear strain is produced for input in FLAC 3D, based on the
last two equations and the assumed relation between φm and γ p reported in Eq. (1.314).
Note that the parameters involved in the selected shear and volumetric hardening laws are not all
independent: they must be chosen consistent with the model assumptions. One assumption is that
the ratio K e / Ge be constant, and equal to the ratio of input upper-bound values K and G. Assuming
that the reference pressure is the same in Eqs. (1.308) and (1.311), it follows that the exponent m
should be the same in both laws. Also, we must have
K (1 + R)Kref
iso
= (1.318)
G Geref
iso and Ge
Thus, the parameter R should be consistent with the choice for Kref ref (assuming these
were selected independently). The consistency condition gives
e
K Gref
R= iso
−1 (1.319)
G Kref
Obviously, the derived value for R should be larger than or equal to zero.
1.6.10.5 Isotropic Compression Tests
Isotropic compression tests on dense, medium and loose sand are simulated using the Cysoil model
with the cap-hardening law described by Eq. (1.307), input in the form of a table. Example 1.5
lists the input data file. The corresponding project file is “cysoil-iso.f3prj,” located in folder
“datafiles\constitutive models\cysoil-iso.” The Cysoil model properties for the tests are listed
in Table 1.2:
The initial stress state is isotropic in each zone; the magnitude of the confining stress is 10 kN/m2 .
The sand is normally consolidated for the tests (the initial cap pressure is equal to 10 kN/m2 ). The
input values of bulk and shear modulus are set to high values, consistent with the given Poisson’s
ratio, ν, listed in Table 1.2. (These input values are used as the upper bound for tangent bulk and
shear elastic modulus calculated by the code.) The cap behavior is assigned using a different table
for each zone, consistent with the formula (Eq. (1.307)) and the data in Table 1.2. The base of the
zones is fixed in the axial (y-) direction, confining velocities of magnitude 10−6 m/step are applied
at the top and lateral sides of the zone for a total of 2500 steps. Five unloading/reloading excursions
are also included.
A plot of vertical (axial) stress versus axial strain for the test is shown for dense, medium and loose
soil cases in Figure 1.39. The plot shows the power law and stiffer behavior achieved by the denser
soil, as expected from the model. A plot of elastic bulk modulus versus axial strain for the test is
shown in Figure 1.40. The bulk modulus is seen to remain constant during unloading/reloading;
also, the value is higher for higher strain levels, consistent with the dependency of the property on
plastic deformation.
A
ss6 sf
s&
lauls6 sf
s<©s6
0
sgl,aa
51l31laulsul9ua9aSs-U
2,aas2 ss
s,
2g0gaaGTals827s2u0aS..Tau
2gaas2 ss
sg
2l0aS3aTals827s2u0a,S,Tau
2uaas2 ss
su
i i!"k
m i
/Nm
2u0ugSlTals827s2u0agS3Tau
0s2,aus2
ssslu
6 sf
s<©s6
0
U
Figure 1.39 Axial stress (in kN/m2 ) versus axial strain for dense, medium and
loose sand
A
,rp,0d 0F
0-l0d
g
0.,err
/G,9G,rp,0p,8pr8r50k
er,0
%
e0CvdYo
.r,0
%
.0CvdYo
pr,0
%
p0CvdYo
g0ierp0i
000,p
#$!i
$$i!"u
o i
mk/
d 0F
0-l0d
g
Figure 1.40 Bulk modulus (in kN/m2 ) versus axial strain for dense, medium
and loose sand
define setup
global _time = clock
; --- dense ---
global _kid = 4e4
global _eurd = 12e4
; --- medium ---
global _kim = 3e4
global _eurm = 9e4
; --- loose ---
global _kil = 2e4
global _eurl = 6e4
;
global _em = 0.5
global _Pref = 100.
global _nu = 0.2
global _pc0 = 10.
; --- derived quantities ---
global _bulk = 1e15
global _shear= _bulk*(3.*(1.-2.*_nu))/(2.*(1.+_nu))
global _num = 5000.
global _nt = 1
end
@setup
def _endtime
_time = clock - _time
end
define cap_table
def _hist_setup
global _z5 = find_zone(5)
global _z3 = find_zone(3)
global _z1 = find_zone(1)
end
@_hist_setup
def _bulk_current5
global _bulk_current5 = z_prop(_z5, ’bulk_current’)
global _shear_current5 = z_prop(_z5, ’shear_current’)
;
global _bulk_current3 = z_prop(_z3, ’bulk_current’)
global _shear_current3 = z_prop(_z3, ’shear_current’)
;
global _bulk_current1 = z_prop(_z1, ’bulk_current’)
global _shear_current1 = z_prop(_z1, ’shear_current’)
end
set hist_rep 20
hist add id 500 zone syy id 5
hist add id 501 gp ydis 0 5 0
hist add id 502 fish @_bulk_current5
hist add id 503 fish @_shear_current5
def trip
loop i (1,5)
command
ini xv -1e-6 range x=0.9 1.1
ini yv -1e-6 range y=0.9 1.1
ini yv -1e-6 range y=2.9 3.1
ini yv -1e-6 range y=4.9 5.1
ini zv -1e-6 range z 0.9 1.1
step 300
ini xv mul -.1 yv mul -.1 zv mul -.1
step 3000
ini xv mul -1. yv mul -1. zv mul -1.
step 3000
end_command
end_loop
end
step 1000
@trip
@_endtime
list @_time
return
Oedometer test simulations are carried out for different values of the parameter α, to evaluate
the impact of the cap aspect ratio on confining stress when yielding occurs on the cap. The
values of α considered for the tests are 0.5, 1 and 1020 (i.e., a value that is very large compared
to 1, in which case the behavior of the double-yield model is recovered). Input for this test is
listed in Example 1.6. The corresponding project file is “cysoil-oedo.f3prj,” located in folder
“datafiles\constitutive models\cysoil-oedo.”
We use the same setup and properties (apart from α) as in the previous example, except in this
case the simulations are run in plane strain, with fixed lateral boundaries to simulate oedometer test
conditions. Friction is assigned a large value to prevent shear yielding. The ratio, K0 , of confining
stress to vertical stress, σxx / σyy , is plotted versus axial strain for dense, medium and loose sand in
two figures: Figure 1.41 compares predictions for α = 0.5 and α = 1020 (double-yield model); and
Figure 1.42 shows plots for α = 1.0 and α = 1020 (double-yield model).
The results of the oedometer simulations show that, with the given properties, a higher value of K0
is achieved for the Cysoil model than the double-yield model. And for all tests, the lower the aspect
ratio of the cap, the higher the K0 that is achieved. Also, dense, medium and loose sands converge
to the same ultimate K0 value as deformation takes place, and they do so at a faster deformation
rate for the Cysoil model than the double-yield model.
A
,rp,0- 0F
05l0-
g
0S9rrr
/G,9G,rp,0p,:pr:r,0yU
.r0%r0Cm-vo
.p0%r0Cm-vo
.,0%r
0Cm-vo
..0%r0Cm-vo
.S0%r0Cm-vo
.e0%r
0Cm-vo
g0ip0i
000.
r
A
r
- 0F
05l0-
g
U
Figure 1.41 K0 versus axial strain for dense, medium and loose sand
– α = 0.5 (lines) and double-yield model (dash lines)
A
,rp,0- 0F
05l0-
g
0S9rrr
/G,9G,rp,0p,:pr:r,0yU
.r0%r0Cm-vo
.p0%r0Cm-vo
.,0%r
0Cm-vo
..0%r0Cm-vo
.S0%r0Cm-vo
.e0%r
0Cm-vo
g0ip0i
000.
- 0F
05l0-
g
U
Figure 1.42 K0 versus axial strain for dense, medium and loose sand
– α = 1.0 (lines) and double-yield model (dash lines)
title
Oedometer test - dense, medium, loose sand - cysoil and DY
def setup
global _alpha = 0.5 ;0.5 ;1.0 ;1e20
; --- dense ---
global _kid = 4e4
global _eurd = 12e4
; --- medium ---
global _kim = 3e4
global _eurm = 9e4
; --- loose ---
global _kil = 2e4
global _eurl = 6e4
;
global _em = 0.5
def _starttime
global _time = clock
end
def _endtime
_time = clock - _time
end
def cap_table
global _mul = _eur/(3.*(1.-2.*_nu)*_ki) - 1.
global evpmax= 1e-2
global _de = evpmax/float(_num)
global ratr = (1.+_mul)/_mul
xtable(_nt,1) = 0.0
ytable(_nt,1) = 0.0
global coe = 1.-_em
global _mex = 1./coe
global _coep = _Pref*(coe*ratr*_ki/_Pref)ˆ_mex
loop ii (2,_num)
local xval = _de*float(ii-1)
ytable(_nt,ii) = _coep*(xvalˆ_mex)
xtable(_nt,ii) = xval
end_loop
global evp=((_pc0/_Pref)ˆcoe)*(_mul/(1.+_mul))*_Pref/(coe*_ki)
end
; --- dense ---
set @_ki=@_kid @_eur=@_eurd @_nt=1
@cap_table
prop mul=@_mul ev_plastic=@evp cptable=@_nt range y 4 5
; --- medium ---
set @_ki=@_kim @_eur=@_eurm @_nt=2
@cap_table
prop mul=@_mul ev_plastic=@evp cptable=@_nt range y 2 3
; --- loose ---
set @_ki=@_kil @_eur=@_eurl @_nt=3
@cap_table
prop mul=@_mul ev_plastic=@evp cptable=@_nt range y 0 1
fix x y z
ini sxx -10. syy -10. szz -10.
ini yvel -0.125e-6 range y 0.9 1.1
ini yvel -0.125e-6 range y 2.9 3.1
ini yvel -0.125e-6 range y 4.9 5.1
def _hist_setup
global _z15 = z_near(0.5, 4.5, 0.5)
global _z13 = z_near(0.5, 2.5, 0.5)
global _z11 = z_near(0.5, 0.5, 0.5)
;
global _z35 = z_near(2.5, 4.5, 0.5)
global _z33 = z_near(2.5, 2.5, 0.5)
global _z31 = z_near(2.5, 0.5, 0.5)
end
@_hist_setup
def _bulk_current15
global _bulk_current15 = z_prop(_z15, ’bulk_current’)
global _shear_current15 = z_prop(_z15, ’shear_current’)
save ini
; -- alpha = 0.5
@compute_k0
prop alpha=@_alpha range group gcy
@_starttime
step 48000
@_endtime
list @_time
save al1
; -- alpha = 1.0
rest ini
set @_alpha = 1.0
@compute_k0
prop alpha=@_alpha range group gcy
@_starttime
step 48000
@_endtime
list @_time
save al2
; -- alpha = 1.0e20
rest ini
set @_alpha = 1.0e20
@compute_k0
prop alpha=@_alpha range group gcy
@_starttime
step 48000
@_endtime
list @_time
save al3
return
Constant Dilation
Triaxial tests on dense, medium and loose sand are simulated using the Cysoil model with the
friction-hardening law described in Eq. (1.313) (input in the form of tables). The model properties
are listed in Table 1.3. The cap pressure is assigned a value very large compared to the stress level
reached in the simulations, in order to prevent yielding on the cap. The cap α is 1.
Two simulations are run: one without cohesion, and one with cohesion set to 100 kPa. The FLAC 3D
grid consists of one zone with unit dimensions. The initial stress state is isotropic, with mean
pressure equal to 100 kPa. The lateral pressure is kept constant during the test, the base of the
model is fixed in the axial (y-) direction, and an axial velocity of 10−6 m/step is applied at the
top of the model for a total of 50,000 steps. In addition, three unloading/reloading excursions are
performed.
A plot of deviatoric stress versus axial strain for the cohesionless test is shown in Figure 1.43
for dense, medium and loose sand cases. The plot shows the hyperbolic behavior expected from
the model, and the higher failure level achieved by the denser soil. Three unloading/reloading
excursions, also shown on the figure, illustrate the model capabilities. The plot of volumetric strain
versus axial strain for the tests is shown in Figure 1.44. The volumetric behavior is monotonic in
the plot, the dilatant behavior of the dense sand is clearly shown.
The comparable results for the simulation with cohesion set to 100 kPa are shown in Figures 1.45
and 1.46. By specifying the same friction table and zero initial friction, the material behavior is
expected to be elastic for values of maximum shear stress smaller than the cohesion value, and
elastic-plastic at higher shear stress levels. This is evident in Figure 1.45. This simulation is done
to illustrate the working of the cohesion logic; the friction hardening table should be adjusted to
reflect more realistic behavior for a cohesive material.
A
+
a
a aoa
a a va
a
0©20am a6
a3yam
v
a©
p.0G.0©20a20,©p,IGaA
%#
2©a
o v
2vIpu0c©©aa9o8aa1vCCp0c©0
0©ao v
2v©IGpc©©aa9o8aa0vuu1gc©0
1©a
o v
uvCuu©o©2aa9o8aa2vp2Cuc©0
m a6
a3yam
v
Figure 1.43 |σ1 - σ3 | (in kN/m2 ) versus axial strain for dense, medium and
loose sand – constant dilation
A
6
a
a aoa
a
a va
a
©m.©a5 aM
a>ya5
v
am
Gu©gu©m.©a.©pmGp0Ga4
%#
..a
ov
Iv2g..omIaaSo/aa.vCGggom©
©.aov
o1v2.2ComIaSo/aaCv©mC©om2
2.a
ov
o.vG.1.om2aSo/ao1vIG0gomI
l
A
l
5 aM
a>ya5
v
Figure 1.44 Volumetric strain versus axial strain for dense, medium and loose
sand – constant dilation
A
4
a
a aoa
a a va
a
0©20a< aM
a8ya<
v
a©
2©.22.0©20ap,©u,Cga;U
%#
I©a
o ov
uv©1Guc©©aa/o:aagvG0Ggc©0
C©ao ov
IvC0p©c©©aa/o:aauvIp21c©0
u©a
o ov
1v©2p1c©©aa/o:aaCv01G©c©0
< aM
a8ya<
v
U
Figure 1.45 |σ1 - σ3 | (in kN/m2 ) versus axial strain for dense, medium and
loose sand – constant dilation – cohesion = 100
A
M
a
a aoa
a
a va
a
©m.©a> aP
a3ya>
v
am
.mu..u©m.©aGpmIp1ga;A
%#
0.a
oov
o.vmmm2om2aSo9aa.v021gom©
1.aoov
o.vI2Imom2aSo9aa2vg022om2
I.a
oov
o1v©2C1om2aSo9ao©vgG1Com1
l
A
l
> aP
a3ya>
v
A
Figure 1.46 Volumetric strain versus axial strain for dense, medium and loose
sand – constant dilation – cohesion = 100
Example 1.7 lists the data file used to perform the triaxial test. Test results for all three cases are
written to tables. Example 1.8 lists the data file which runs the triaxial tests on dense, medium and
loose sands, and gathers the results for comparison. The project file is “cysoil-dtriax-constdil.f3prj,”
located in folder “datafiles\constitutive models\cysoil-dtriax-constdil.”
title
Drained triaxial test - cysoil
def setup
local _nu = 0.2
;_iChoice = 0 ; 0 = dense, 1 = medium, 2 = loose
caseof _iChoice
case 0
local _gref = 120e3/(2.*(1.+_nu)) ; reference tangent shear mod
local _kref = 120e3/(3.*(1.-2.*_nu)) ; reference tangent bulk mod
local _friu = 40. ; ultimate friction
global _dil = 10.0 ; dilation
global _fnameds = ’dense-ds.log’
global _fnameev = ’dense-ev.log’
case 1
_gref = 90e3/(2.*(1.+_nu)) ; reference tangent shear mod
_kref = 90e3/(3.*(1.-2.*_nu)) ; reference tangent bulk mod
_friu = 35. ; ultimate friction
_dil = 5.0 ; dilation
_fnameds = ’medium-ds.log’
_fnameev = ’medium-ev.log’
case 2
_gref = 60e3/(2.*(1.+_nu)) ; reference tangent shear mod
_kref = 60e3/(3.*(1.-2.*_nu)) ; reference tangent bulk mod
_friu = 30. ; ultimate friction
_dil = 0.0 ; dilation
_fnameds = ’loose-ds.log’
_fnameev = ’loose-ev.log’
endcase
local _Pa = 100. ; reference pressure
global _fric = 0.0 ; initial friction
global _al = 1.0 ; alpha for cap
local _Rf = 0.90 ; failure ratio Rf
local _beta = 0.35
global _yv = -1e-6 ;-1e-9
local _p0 = 100.
; --- derived quantities ---
def _cy_q
local _z = zone_head
global _cy_q = z_prop(_z, ’cy_q’)
global _cy_p = z_prop(_z, ’cy_p’)
global _syy = z_syy(_z)
def trip
local i
loop i(1,3)
command
ini yv -1e-6 range y 0.9 1.1
step 8000
ini yv mul -.1 range y 0.9 1.1
step 6000
ini yv mul -1. range y 0.9 1.1
step 6000
end_command
end_loop
end
step 3000
@trip
ini yv -1e-6 range y 0.9 1.1
step 23000
new
call choice suppress
set @_iChoice = 0 ; dense
call cysoil-dtriax-constdil
new
call choice suppress
set @_iChoice = 1 ; medium
call cysoil-dtriax-constdil
new
call choice suppress
set @_iChoice = 2 ; loose
call cysoil-dtriax-constdil
new
title
Drained triaxial test - cysoil
Dilation Hardening
The triaxial test simulations with friction hardening in Section 1.6.10.7 are repeated, this time also
using a dilation hardening table as described by Eq. (1.315). Initial dilation is zero, and the ultimate
value of dilation is listed in Table 1.3. Example 1.9 lists the data file used to perform the triaxial
test. Test results for all three cases are written to tables. Example 1.10 lists the data file that runs
the triaxial tests on dense, medium and loose sands, and gathers the results for comparison. The
project file is “cysoil-dtriax-dilhard.f3prj,” located in folder “datafiles\constitutive models\cysoil-
dtriax-dilhard.”
The simulation results of deviatoric stress and volumetric strain versus axial strain are plotted in
Figures 1.47 and 1.48, respectively. The non-monotonic volumetric behavior is apparent in the
second plot: all soil types do compact initially, and the denser soil is shown to dilate upon further
shearing.
A
<
a
a aoa
a a va
a
0©20a7 a4
a+ya7
v
a©
p.0G.0©20a20,©p,1©aA
%#
2©a
o v
pvG1IIo©2aa9o8aa1vCCp0c©0
0©ao v
gvI0u1o©2aa9o8aa0vuu1gc©0
1©a
o v
IvpgGgo©2aa9o8aa2vGu©1c©0
l
A
7 a4
a+ya7
v
Figure 1.47 |σ1 - σ3 | (in kN/m2 ) versus axial strain for dense, medium and
loose sand – dilation hardening
A
5
a
a aoa
a
a va
a
©m.©a7 aM
a4ya7
v
am
Gu©gu©m.©a.©pmGp2ma3
%#
..a
ov
o1v©mmGom2aSo/aa0v22©Iom2
©.aov
oCv©IgCom2aSo/ao2v©01©om1
2.a
ov
o.v2m0mom©aSo/ao2v.Gg2om1
7 aM
a4ya7
v
Figure 1.48 Volumetric strain versus axial strain for dense, medium and loose
sand – dilation hardening
title
Drained triaxial test - cysoil
def setup
local _nu = 0.2
;_iChoice = 0 ; 0 = dense, 1 = medium, 2 = loose
caseof _iChoice
case 0
local _gref = 120e3/(2.*(1.+_nu)) ; reference tangent shear mod
local _kref = 120e3/(3.*(1.-2.*_nu)) ; reference tangent bulk mod
local _friu = 40. ; ultimate friction
local _dilu = 10.0 ; ultimate dilation
global _fnameds = ’dense-ds.log’
global _fnameev = ’dense-ev.log’
case 1
_gref = 90e3/(2.*(1.+_nu)) ; reference tangent shear mod
_kref = 90e3/(3.*(1.-2.*_nu)) ; reference tangent bulk mod
_friu = 35. ; ultimate friction
xtable(2,ii) = xval
end_loop
ytable(1,_num+1) = _phic
xtable(1,_num+1) = 0.1
ytable(2,_num+1) = _psic
xtable(2,_num+1) = 0.1
; (initial value of plastic shear strain)
sval = sin(_fric*degrad)
_coe = _sinu/_Rf
global _esp = (_coe/(1.-sval/_coe)-_coe)/_Gi
end
@setup
def _cy_q
local _z = zone_head
global _cy_q = z_prop(_z, ’cy_q’)
global _cy_p = z_prop(_z, ’cy_p’)
global _syy = z_syy(_z)
global _sxx = z_sxx(_z)
global _szz = z_szz(_z)
global _cap_pressure = z_prop(_z, ’cap_pressure’)
global _es_plastic = z_prop(_z, ’es_plastic’)
global _ev_plastic = z_prop(_z, ’ev_plastic’)
global _et_plastic = z_prop(_z, ’et_plastic’)
global _friction_mob = z_prop(_z, ’friction_mob’)
global _dilation_mob = z_prop(_z, ’dilation_mob’)
global _vsi = z_vsi(_z)
global _bulk_current = z_prop(_z, ’bulk_current’)
global _shear_current = z_prop(_z, ’shear_current’)
end
step 50000
new
call choice suppress
set @_iChoice = 0 ; dense
call cysoil-dtriax-dilhard
new
call choice suppress
set @_iChoice = 1 ; medium
call cysoil-dtriax-dilhard
new
call choice suppress
set @_iChoice = 2 ; loose
call cysoil-dtriax-dilhard
new
title
Drained triaxial test - cysoil
return
Tests similar to those in Section 1.6.10.7 are repeated, but this time under undrained conditions.
The model setup and properties are the same as those used in the previous examples. In addition,
the groundwater flow configuration is selected, with flow turned off. Porosity is 0.3, and the fluid
bulk modulus is 2000 kPa for the runs.
The undrained simulation results for deviatoric stress and volumetric strain versus axial strain are
plotted in Figures 1.49 and 1.50, respectively. A softening stress-strain response is observed for
the loose soil in Figure 1.49. Also, while the excess pore pressure to rise initially for all soils (as
indicated in Figure 1.50), it is shown to decrease upon further shearing as a result of dilation taking
place, for the medium and dense soils.
Example 1.11 lists the data file that performs the triaxial case. All three cases are compared in
Figures 1.49 and 1.50 by loading the table results as described in Example 1.12. The correspond-
ing project file is “cysoil-udtriax.f3prj,” located in folder “datafiles\constitutive models\cysoil-
udtriax.”
A
nn
n
n non
n n vn
n
0©20nm n+
n3ynm
v
n©
p.0G.0©20n20,©p,2GnM
%#
2©n
o v
2v©©0uc©©nn9o8nnIv1I2Cc©0
0©no v
gvu2Ipo©2nn9o8nn0v©©gGc©0
1©n
o v
Cv2u2Go©2nn9o8nnuvGpCCc©2
l
A
m n+
n3ynm
v
Figure 1.49 |σ1 - σ3 | (in kN/m2 ) versus axial strain for dense, medium and
loose sand – undrained triaxial tests
A
nn
n
n non
n
n vn
n
©m.©n3 n+
n7yn3
v
nm
Gu©gu©m.©n.©pmGp.gnA
%#
..n
ov
o©v.GGgcm.nSo/nn©vG0..cm.
©.nov
.vGC.1om.nnSo/nn2vG0.Gcm.
2.n
ov
.vg1G2om.nnSo/nnIvI©g©cm.
3 n+
n7yn3
v
Figure 1.50 Volumetric strain versus axial strain for dense, medium and loose
sand – undrained triaxial tests
title
Undrained triaxial test - cysoil
def setup
local _nu = 0.2
;_iChoice = 0 ; 0 = dense, 1 = medium, 2 = loose
caseof _iChoice
case 0
local _gref = 120e3/(2.*(1.+_nu)) ; ref. tangent shear mod
local _kref = 120e3/(3.*(1.-2.*_nu)) ; ref. tangent bulk mod
local _friu = 40. ; ultimate friction
local _dilu = 10.0 ; ultimate dilation
global _fnameds = ’dense-uds.log’
global _fnameev = ’dense-pp.log’
case 1
_gref = 90e3/(2.*(1.+_nu)) ; ref. tangent shear mod
_kref = 90e3/(3.*(1.-2.*_nu)) ; ref. tangent bulk mod
_friu = 35. ; ultimate friction
xtable(2,ii) = xval
end_loop
ytable(1,_num+1) = _phic
xtable(1,_num+1) = 0.1
ytable(2,_num+1) = _psic
xtable(2,_num+1) = 0.1
; (initial value of plastic shear strain)
sval = sin(_fric*degrad)
_coe = _sinu/_Rf
global _esp = (_coe/(1.-sval/_coe)-_coe)/_Gi
end
@setup
config fluid
generate zone brick size 1 1 1
model mech cysoil
prop dens 1000 bulk=@_k shear=@_g friction=@_fric
prop dilation=0.0
prop tension=1e10 cap_pressure=@_cp alpha=@_al
prop es_plastic=@_esp ev_plastic=0.0
prop ftable 1
prop dtable 2
def _cy_q
local _z = zone_head
global _cy_q = z_prop(_z, ’cy_q’)
global _cy_p = z_prop(_z, ’cy_p’)
global _syy = z_syy(_z)
global _sxx = z_sxx(_z)
global _szz = z_szz(_z)
step 50000
new
call choice suppress
set @_iChoice = 0 ; dense
call cysoil-udtriax
new
call choice suppress
set @_iChoice = 1 ; medium
call cysoil-udtriax
new
call choice suppress
set @_iChoice = 2 ; loose
call cysoil-udtriax
new
title
Undrained triaxial test - cysoil
return
cy q deviatoric stress, q
As discussed previously in Section 1.6.10, when soils are subjected to deviatoric loading they usually
exhibit a decrease in stiffness accompanied by irreversible deformation. The well-known Duncan
and Chang model (1970) is commonly used to simulate this hyperbolic stress-strain behavior. The
Duncan and Chang model is relatively easy to use. However, as noted in Section 1.6.10, this model
has several drawbacks. In addition, hyperbolic relations that rely on nonlinear elasticity are known
to have significant limitations (Duncan et al. 1980), including (1) the relations are applicable prior to
failure, but may produce unrealistic behavior of soils at and after failure, (2) the hyperbolic relations
do not include volume change resulting from change in shear stress (the implied dilation relies on
the Poisson’s ratio), and thus may not be able to predict deformations in dilatant soils accurately,
such as dense sands under low confining pressure, and (3) the relations predict an isotropic behavior
in the plane, which is not always realistic (soil strengths in triaxial compression and extension
usually differ in magnitude).
A simplified version of the Cysoil model, named the Chsoil model, is provided as an alternative to
the Duncan and Chang model that does not have the drawbacks of this model. The Chsoil model is
derived from the same strain hardening/softening logic that exists in the Cysoil model, and therefore
can provide a realistic stress-strain relation at failure and post-failure. The Chsoil model provides
built-in features and does not have a volumetric cap.
The Chsoil model has three specific features:
1. a built-in friction hardening law that uses hyperbolic model parameters as direct input;
2. a frictional Mohr-Coulomb shear envelope; and
3. two built-in dilation laws – one is based on Rowe stress dilatancy theory (Rowe 1962),
and the other is a user-defined dilation hardening/softening law.
The unloading behavior of the Chsoil model is elastic. Reloading is elastic up to the outermost
yield envelope reached previously. Note that in its present form, the Chsoil model is not intended
to simulate cyclic loading.
The elastic behavior of the Chsoil model is expressed using Hooke’s law. The incremental expression
of the law in terms of principal stress and strain has the form
where α1 = K e + 4Ge /3, α2 = K e − 2Ge /3, and K e and Ge are the tangent elastic bulk and shear
modulus, respectively.
In the Chsoil model, the elastic shear modulus depends on the initial value of mean effective stress,
pm . The mean effective stress is specified by the user, using, for example, the expression
pm !n
Ge = Gref pref (1.322)
pref
The parameter pref is the reference pressure, and Gref is the shear modulus number (Gref × pref
is the value of the tangent elastic shear modulus at reference pressure). n is a constant modulus
exponent (n ≤ 1). Also, the tangent elastic bulk modulus is described by the relation
pm !m
K e = Kref pref (1.323)
pref
The parameter Kref is the bulk modulus number, the product (Kref ×pref is the value of the tangent
elastic bulk modulus at reference pressure), and m is a constant modulus exponent (m ≤ 1).
Some useful relations between Kref , Gref , Young’s modulus number, Eref , and Poisson’s ratio at
reference pressure, νref , are listed for reference:
Eref Eref
Kref = Gref =
3(1 − 2νref ) 2(1 + νref )
(1.324)
The user can specify either Eref and νref or Kref and Gref as input properties for the model. If Eref
and νref are specified, Kref and Gref are calculated internally from Eq. (1.324), and the resulting
values are used with Eqs. (1.322) and (1.323). Values of Poisson’s ratio are restricted to positive
values smaller than 0.49. Accordingly, upper and lower bounds for K e are specified internally as
2Ge
< K e < 49.66Ge (1.325)
3
, must be provided as an input property for
As noted above, the initial mean effective pressure, pm
the model. (This value may be evaluated using FISH and stored in the associated zone property
is used to calculate the tangent elastic shear and bulk moduli, according
offset.) The value of pm
to Eqs. (1.322) and (1.323). Ge and K e stay constant in this implementation; the moduli are not
updated automatically in terms of mean effective pressure.
1.6.11.2 Shear Yield Criterion and Flow Rule
f s = σ1 − σ3 Nφm + 2c Nφm (1.326)
1 + sin φm
Nφm = (1.327)
1 − sin φm
where
1 + sin ψm
Nψm = (1.329)
1 − sin ψm
and ψm is the mobilized dilation angle. Several different laws are available in the literature to
characterize ψm .
By default, ψm can be given in terms of plastic shear strain via a user-defined table ofψm versus
γ p . The evolution parameter for shear yielding, γ p , is defined incrementally by
√
dp dp dp
γ p = (ε1 )2 + (ε2 )2 + (ε3 )2 / 2 (1.330)
dp
where εi , i = 1, 3 are the principal, deviatoric, plastic shear-strain increments.
The model parameter dilaw must be set to zero to activate this option. If dilaw = 0 and no table is
provided, it is then assumed that dilation is equal to the input value for ultimate dilation property
set with dilf. Note that dilf is not used if a dilation table is provided.
Two built-in dilation laws are also available. A law based on Rowe (1962) dilatancy theory is used
if dilaw = 1. This is the default setting. Alternatively, a simple step function can be used, in which
ψm is zero for φm < φcv , where φcv is a constant specified by the user, and ψm is equal to the
ultimate dilation value ψf (set by dilf) for values of mobilized friction larger than ψcv . The model
property dilaw must be set to 2 to activate this option. The additional constraint that ψm cannot
exceed φm (to prevent unwanted generation of energy from taking place) is enforced internally by
the code.
1.6.11.3 Tensile Yield Criterion and Flow Rule
The tensile yield function is the same as that used for the Cysoil model, and is of the form
f t = σ t − σ3 (1.331)
The tensile strength, σ t , is given in terms of the plastic tensile-strain measure, ept , and input by
means of a user-defined table. If no table is provided, it is assumed that tensile strength is constant,
and equal to the input value of the tensile strength property.
The evolution parameter for tensile yielding is the modulus of plastic tensile strain, ept . The
increment of plastic tensile strain is defined as
pt
ept = e3 (1.332)
pt
where e3 is the increment of tensile plastic strain in the direction of the major principal stress
(recall that tensile stresses are positive).
For most soils, the plot of deviatoric stress versus axial strain obtained in a drained triaxial test can
be approximated by a hyperbola. The Chsoil model incorporates a friction strain-hardening law
to capture this behavior. In this formulation, the mobilized friction angle, φm , is given in terms
of plastic shear strain measure, γ p , by means of the following differential law, similar to the one
implemented by Byrne et al (2003) in their UBCSAND liquefaction model:
Gp
d(sin φm ) =
d(γ p ) (1.333)
pm
sin φm !2
φm ≤ φf : Gp = Ge 1 − Rf (1.334)
sin φf
In this formula, Ge is the elastic tangent shear modulus, φf is the ultimate friction angle, and Rf
(the failure ratio) is a constant smaller than 1 (0.9 in most cases) used to assign a lower bound for
Gp .
,
According to Eq. (1.322), the elastic tangent shear modulus is a function of pm
n
pm
G = Gref pref
e
(1.335)
pref
After substitution of Eq. (1.335) into (1.334), the resulting expression in (1.333), and rearranging
terms, we obtain
pm d(sin φm )
d(γ p ) = !2 (1.336)
Ge sin φm
1− sin φf Rf
sin φ
pm 1 !
f
γ = e
p
−1 (1.337)
G Rf 1 − sin φm
sin φf Rf
sin φf 1 !
sin φm = 1− e Rf
(1.338)
Rf 1 + γ p pG
m sin φf
This expression is used in the Chsoil model to calibrate mobilized friction in terms of plastic shear
strain. The use of this hardening law for modeling primary loading in a triaxial test produces a
hyperbolic-type curve of deviatoric stress versus axial strain.
1.6.11.5 Over-consolidation
The initial state of a normally consolidated soil or an over-consolidated soil is prescribed by speci-
fying an initial value to the friction property that is equal or larger than the normally consolidated
value, φnc , respectively. For normal consolidation Eqs. (1.326) and (1.327) give
σ1 − σ3
φnc = arcsin (1.339)
σ1 + σ3
The material behavior is considered to be elastic for stress points below the current yield envelope.
The initial value of friction must be smaller than the ultimate value, φf . (An upper bound equal
to φf is set automatically by the model.) An initial value of the evolution parameter, γ p , that is
consistent with the specified initial value of mobilized friction angle is also assigned automatically.
1.6.11.6 Shear-Induced Compaction and Dilation
A certain amount of irrecoverable volumetric strain, ep , is expected to take place as a result of soil
shearing. Also, under small monotonic shear strains, there is a tendency for the soil skeleton to
contract due to grain rearrangements. For larger shear strains, the soil skeleton may dilate if the soil
is dense, as a result of grains riding over each other. A dilation strain-hardening law is incorporated
in the logic to model this behavior.
The shear-hardening flow rule implemented in the Chsoil model has the form
where ep is the plastic volumetric strain increment, γ p is the plastic shear strain increment, and
ψm is the (mobilized) dilation angle.
One possible way to characterize ψm is to adopt an equation based on Rowe stress-dilatancy theory
(Rowe 1962). According to this theory, there is a (constant-volume) friction angle, φcv , below
which the material contracts (i.e., for φm ≤ φcv ), while for higher stress ratios (i.e., for φm > φcv ),
the material dilates. The equation has the form
where
sin φf − sin ψf
sin φcv = (1.342)
1 − sin φf sin ψf
and φf and ψf are ultimate (known) values of friction and dilation, respectively.
Dilation is evaluated in terms of plastic shear strain based on the last two equations, and the assumed
relation between φm and γ p reported in Eq. (1.337). Rowe’s dilation law is selected by setting the
material property dilaw = 1 (default value). A simple dilation law also can be used, in which ψm = 0
for φm < φcv , and ψm = ψf for φm ≥ φcv ; the option is activated by specifying dilaw = 2. Finally,
the user may choose to define an alternative dilation law by creating an input table of dilation values
versus plastic shear strain. The model property dilaw must be set to zero to activate this option.
1.6.11.7 Correlation Between Chsoil Properties and Duncan and Chang Properties
A parallel can be drawn, for particular cases, between properties used in the hyperbolic stress-strain
relations reported in Duncan et al. (1980) for the Duncan and Chang model and the properties of
the Chsoil model. The correlations depend on property input; they are listed in Table 1.4:
1.6.11.8 Calibration of the Chsoil Model to Triaxial Test Results on Nevada Sand
Results of drained triaxial tests at constant mean effective stress, p , on Nevada sand are used to
provide an example of the methodology that can be used to calibrate the properties of the Chsoil
model. The test results are listed in data files “40-40.log,” “40-80.log” and “40-160.log,” and
correspond to values of the mean stress at 40 kPa, 80 kPa and 160 kPa, respectively. The relative
density of the sand for these tests is 40%. The test results consist of three sets of data:
(1) deviatoric stress versus mean effective stress;
(2) deviatoric stress versus axial strain (up to 25% strain); and
(3) volumetric strain versus axial strain (up to 25% strain).
The corresponding project file is “chsoil-dtriax.f3prj,” in the “datafiles\constitutive models\cysoil-
udtriax” folder.
For this calibration exercise, we consider axial strains up to 5%, because strains are not expected to
develop above this level for the intended (static, drained) application. The Nevada sand appears to be
purely frictional in character. Also, two main features characterize the data: (1) hyperbolic behavior
is observed in the plot of deviatoric stress versus axial strain, and (2) bilinear dilatant behavior
is exhibited by the plot of volumetric strain versus axial strain. The Chsoil model capabilities,
including friction hardening and variable dilation, are used to simulate these features.
The elastic tangent shear and bulk moduli are functions of the initial mean effective stress according
to Eqs. (1.322) to (1.324). For friction hardening we use the built-in law, Eq. (1.337). The dilation
law is bilinear; its value is zero below the stress ratio φ = φcv , and a constant, ψf , above it:
ψ =0 φ < φcv
(1.343)
ψ = ψf φ ≥ φcv
The failure ratio, Rf , is 0.99 for this exercise. Eight properties must be defined:
Calibration of the model properties is done in two steps. First, using theoretical considerations
and values recorded in the literature, we derive a first estimate for the property values. Second,
we improve on the estimates by modeling the triaxial experiments numerically and matching the
results obtained in the laboratory.
First Estimates – The value of ultimate friction is derived from a linear fit to plots of deviatoric
stress, q, versus mean effective stress, p , obtained from the laboratory tests. The linear fit to
maximum q at given p provides a value for the maximum stress ratio, a:
q
a= (1.344)
p
From the definition of the purely frictional Mohr-Coulomb criterion, we obtain the relation
3a
sin φf = (1.345)
6+a
from which the ultimate friction angle, φf , can be derived. The estimated value for Dr = 40% sand
is φf = 34◦ .
The value of ultimate dilation angle is derived from the slope, b, of a linear fit to the laboratory curves
of minus volumetric strain versus axial strain. To relate b to ψf , we use, as a first approximation,
the expression for bilinear idealization of triaxial stress results provided by Vermeer and de Borst
(1984):
2 sin ψf
b= (1.346)
1 − sin ψf
adjusted (in the order listed above), and the numerical experiment is repeated until a satisfactory
curve fitting is obtained.
The results of the curve fitting experiment are listed in Table 1.5:
A comparison between numerical predictions using the calibrated properties and laboratory re-
sults is shown in Figures 1.52 through 1.51. Note that the soil-mechanics convention for positive
stress/strain in compression is adopted in these plots. (Dilation is negative.) The comparison is
quite reasonable. The data file for this comparison example is listed in Example 1.13.
A
6
a
a aa
a
a a6Gy=ama
a a pa
a
0=40a> a5
a3ba>
p
a4=====
4=u44u0=40a1I=vIv.a;N
%#
4-a
m4p====m=0agm,aavp©©2=o=4
0-a
=p====o==aagm,aa4p=0v-o=0
--a
=p====o==aagm,aa0p0=24o=0
44-a
4p1411o=4aagm,aavpv4yvo=4
40-a
l
A
0p1-1©o=4aagm,aa4p=1©=o=0
4--a
ypy=42o=4aagm,aa0p4©©0o=0
> a5
a3ba>
p
N
Figure 1.51 Deviator stress versus mean stress for Dr = 40% – comparison
between laboratory (line) and numerical (cross) results
A
6
a
a aa
a
a a6Gy=ama
a a pa
a
0=40a7 a5
a<ba7
p
a4=====
4=u44u0=40a1I=vIv.a;N
%#
44a
m4p====m=0agm,aavp©©2=o=4
04a
=p====o==aagm,aa4p=0v-o=0
-4a
=p====o==aagm,aa0p0=24o=0
444a
4p1411o=4aagm,aavpv4yvo=4
404a
7 a5
a<ba7
p
N
Figure 1.52 Deviatoric stress versus axial strain (in %) for Dr = 40%
– comparison between laboratory (line) and numerical (cross)
results
A
6
a
a aa
a
a a6Gy=ama
a
a pa
a
0=40a> a5
a+ba>
p
a4=====
4=u44u0=40a1I=vIv.a;N
%#
40a
m4p01==o==agm,aa=p====o==
00a
m4p00==o==agm,aa=p====o==
-0a
m4py©==o==agm,aa=p====o==
440a
m4py=2©o==agm,aa4p44©©m=y
400a
m4p-©=2o==agm,aa4p4=21m=y
4-0a
m4p-4©=o==agm,aa4p4=yym=y
> a5
a+ba>
p
N
Figure 1.53 Volumetric strain (in %) versus axial strain (in %) for Dr = 40%
– comparison between laboratory (line) and numerical (cross)
results
Example 1.13 Drained triaxial test at constant mean pressure – Dr = 40 – Chsoil model
new
set fish autocreate off
set fish safe_conversion on
;
def sp
global _nu = 0.35
global _Eref = 1800. ; Young?s mod number
global _Pa = 100. ; reference pressure
global _m = 0.5 ; bulk modulus exponent
global _n = 0.5 ; shear modulus exponent
global _friu = 34. ; ultimate friction
global _dilu = 7.5 ; ultimate dilation
global _ficv = 28. ; transition phase angle
global _Rf = 0.99 ; failure ratio Rf
global _fric = 0. ; initial friction
global _coh = 0. ; cohesion
; (servo)
global _zvel = -0.5e-6
global _gain = 1.
global high_vel = 0.6e-6
; (declaration)
global _p0 = 40.
global _nt = 2
; --- derived quantities ---
global _xvel = -0.5*_zvel
global _yvel = -0.5*_zvel
end
@sp
;
gen zone brick size 1 1 5
model mech chsoil
;
prop young_ref=@_Eref poisson=@_nu p_ref=@_Pa
prop Rf=@_Rf fricf=@_friu m_k=@_m n_g=@_n
prop dens=1000. cohesion=@_coh dilation=0.
prop dilaw=2 fricv=@_ficv dilf=@_dilu
;
prop dtable 2 range id = 1
prop dtable 3 range id = 3
prop dtable 4 range id = 5
prop p_ini=40. range id = 1
prop p_ini=80. range id = 3
prop p_ini=160. range id = 5
;
model mech null range id = 2
model mech null range id = 4
;
ini sxx -40. syy -40. szz -40. range id = 1
ini sxx -80. syy -80. szz -80. range id = 3
ini sxx -160. syy -160. szz -160. range id = 5
;
fix x y z
ini xvel @_xvel range x 0.9 1.1
ini yvel @_yvel range y 0.9 1.1
ini zvel @_zvel range z 0.9 1.1
ini zvel @_zvel range z 2.9 3.1
ini zvel @_zvel range z 4.9 5.1
;
def zn_setup
global _z1 = z_near(0.5,0.5,0.5)
global _z3 = z_near(0.5,0.5,2.5)
global _z5 = z_near(0.5,0.5,4.5)
end
@zn_setup
;
; --- servo for constant mean stress ---
def servo_sig0
while_stepping
local _sig=-(z_sxx(_z1)+z_syy(_z1)+z_szz(_z1))/3.
local _svel=gp_xvel(gp_near(1,1,1))-_gain*(1.-_sig/40.)
if abs(_svel) > high_vel then
_svel=sgn(_svel)*high_vel
end_if
gp_xvel(gp_near(1,0,0))=_svel
gp_xvel(gp_near(1,0,1))=_svel
gp_xvel(gp_near(1,1,0))=_svel
gp_xvel(gp_near(1,1,1))=_svel
gp_yvel(gp_near(0,1,0))=_svel
gp_yvel(gp_near(0,1,1))=_svel
gp_yvel(gp_near(1,1,0))=_svel
gp_yvel(gp_near(1,1,1))=_svel
;
_sig=-(z_sxx(_z3)+z_syy(_z3)+z_szz(_z3))/3.
_svel=gp_xvel(gp_near(1,1,3))-_gain*(1.-_sig/80.)
if abs(_svel) > high_vel then
_svel=sgn(_svel)*high_vel
end_if
gp_xvel(gp_near(1,0,2))=_svel
gp_xvel(gp_near(1,0,3))=_svel
gp_xvel(gp_near(1,1,2))=_svel
gp_xvel(gp_near(1,1,3))=_svel
gp_yvel(gp_near(0,1,2))=_svel
gp_yvel(gp_near(0,1,3))=_svel
gp_yvel(gp_near(1,1,2))=_svel
gp_yvel(gp_near(1,1,3))=_svel
;
_sig=-(z_sxx(_z5)+z_syy(_z5)+z_szz(_z5))/3.
_svel=gp_xvel(gp_near(1,1,5))-_gain*(1.-_sig/160.)
if abs(_svel) > high_vel then
_svel=sgn(_svel)*high_vel
end_if
gp_xvel(gp_near(1,0,4))=_svel
gp_xvel(gp_near(1,0,5))=_svel
gp_xvel(gp_near(1,1,4))=_svel
gp_xvel(gp_near(1,1,5))=_svel
gp_yvel(gp_near(0,1,4))=_svel
gp_yvel(gp_near(0,1,5))=_svel
gp_yvel(gp_near(1,1,4))=_svel
gp_yvel(gp_near(1,1,5))=_svel
global q1 = z_sxx(_z1)-z_szz(_z1)
global p1 = -(z_sxx(_z1)+z_syy(_z1)+z_szz(_z1))/3.
global q2 = z_sxx(_z3)-z_szz(_z3)
global p2 = -(z_sxx(_z3)+z_syy(_z3)+z_szz(_z3))/3.
global q3 = z_sxx(_z5)-z_szz(_z5)
global p3 = -(z_sxx(_z5)+z_syy(_z5)+z_szz(_z5))/3.
global eps_v1 = z_vsi(_z1)*100. ; volumetric strain (%)
global eps_a1 = gp_zdisp(gp_near(1.0,1.0,1.0))*100. ; axial strain (%)
global eps_v2 = z_vsi(_z3)*100. ; volumetric strain (%)
global eps_a2 = gp_zdisp(gp_near(1.0,1.0,3.0))*100. ; axial strain (%)
global eps_v3 = z_vsi(_z5)*100. ; volumetric strain (%)
global eps_a3 = gp_zdisp(gp_near(1.0,1.0,5.0))*100. ; axial strain (%)
end
;
hist add id 9 fish @q3
hist add id 10 fish @p3
hist add id 11 fish @eps_v3 ; vol. strain (%) dilation positive
hist add id 12 fish @eps_a3 ; axial strain (%)
;
hist nstep 1000
step 100000
;
save chsoil-dtriax1
;
return
1.7 References
Roscoe K. H., and J. B. Burland. “On the Generalised Stress-Strain Behavior of ‘Wet Clay’,” in
Engineering Plasticity, pp. 535-609. J. Heyman and F. A. Leckie, eds. Cambridge: Cambridge
University Press (1968).
Rowe, P. W. “The Stress-Dilatancy Relation for Static Equilibrium of an Assembly of Particles in
Contact,” Proc. Roy. Soc. A., 269, 500-527 (1962).
Shah, S. A Study of the Behaviour of Jointed Rock Masses. Ph.D. Thesis, University of Toronto
(1992).
Vermeer, P. A., and R. de Borst. “Non-Associated Plasticity for Soils, Concrete and Rock,” Heron,
29(3), 3-64 (1984).
Wood, D. M. Soil Behaviour and Critical State Soil Mechanics. Cambridge: Cambridge Univer-
sity Press (1990).