FLAC Training Course

Basic Concepts and Recommended Procedures

for
Geotechnical Numerical Analysis
related to
Nuclear Waste Isolation
August 7-11, 2006
Instructors:

Dr. Roger Hart

Yanhui Han
 Training Schedule
August 7, 2006 (morning)

09:00-12:00 Overview on Numerical Modeling for

Nuclear Waste Isolation
- Introduction and overview by IAEA
- Problems related to repository design and engineering
- Participant perceptions (each participant provides
her/his perspective on numerical modeling in the context
of their national program ~ 10-15 min. per participant)
 Training Schedule
August 7, 2006 (afternoon)

01:00-02:00 Overview of Itasca and Itasca Software

Introduction to FLAC
- Overview of capabilities in geo-engineering analysis
and design
- Modeling features specific to waste isolation studies
02:00-03:00 Introduction to the FLAC Graphical Interface
- Menu-driven versus command-driven operation
03:00-03:15 Break
03:15-05:00 FLAC Theoretical Background
- Explicit finite-difference solution
 Overview of Itasca

Consulting Services and Software for the

Mining, Civil, Petroleum, and Waste Isolation Industries
 Itasca office locations

plus software agents in 13 countries

 Itasca codes
1. FLAC – two dimensional continuum, with joints
2. FLAC3D – three dimensional continuum, with joints
3. PFC3D – three dimensional DEM* spheres + clumps
4. PFC2D – two dimensional DEM disks + clumps
5. UDEC – two dimensional DEM polygonal bodies
6. 3DEC – three dimensional DEM polyhedral bodies

All codes use an explicit, dynamic solution scheme, even to simulate

quasi-static problems. All include coupled fluid and thermal modes,
and include many nonlinear constitutive models.

All codes treat interactions between separate objects as boundary

conditions; there is no concept of a “joint element”. Thus, even for the
continuum codes, the “DEM scheme” is used for interactions.

* DEM (distinct/discrete element method)

 All Itasca codes …
… contain a built-in programming language, called FISH, that
allows users to:

• add new plots or printout options

• control a simulation (and the conditions) automatically
• access and modify most of the internal variables & properties
• set up special in situ conditions & boundary conditions
• add coupling between codes, or between physical entities.

Also, all codes can accept user-written constitutive

(stress/strain) models, written in C++ or FISH (FLAC only).
Many users have written their own models. Several models are
available that have been written by others.
 User support
1. Extensive manuals, with many examples and useful FISH
functions, are provided, both on CD and in hard-copy.
2. Hundreds of references to papers describing applications of
all codes are available on the Itasca web site
(www.itascacg.com).
3. Worked examples are provided and updated regularly on
the web site; a new site provides a repository for new
constitutive models.
4. Latest code updates may be downloaded from the web.
5. International code-user symposia are held regularly.
6. Rapid answers to users’ queries are provided, both by
telephone and email (many hundreds of such questions are
handled every year).
7. Consulting agreements may be set up for more extensive
help with setting up models and interpreting the results.
 What is FLAC?

FLAC is a general-purpose code that can simulate a full range of

nonlinear static & dynamic problems, with coupled fluid flow, heat
flow and structural interaction. Any geometry can be represented,
and the boundary conditions are quite general.
FLAC simulates the behavior of nonlinear continua by the
generalized finite difference method (arbitrary element shapes),
also known as the finite volume method.
FLAC solves the full dynamic equations of motion even for quasi-
static problems. This has advantages for problems that involve
physical instability, such as collapse, as will be explained later. To
model the “static” response of a system, damping is used to absorb
kinetic energy.
FLAC & FLAC3D Basic Features

➢ Nonlinear, large-strain
simulation of continua
➢ Explicit solution scheme,
giving stable solutions to
unstable physical processes
➢ Interfaces or slip-planes are
available to represent
distinct interfaces along
which slip and/or separation
are allowed, thereby
simulating the presence of
faults, joints or frictional
boundaries

Displacements resulting
from construction of a
shallow tunnel

Advanced, Two and Three Dimensional

Continuum Modeling for Geotechnical Analysis
of Rock, Soil, and Structural Support
FLAC & FLAC3D Basic Features

➢Built-in material models:

•"null" model,
•three elasticity models
(isotropic, transversely
isotropic and orthotropic
elasticity),
•eight plasticity models
(Drucker-Prager, Mohr-
Coulomb, strain-
hardening/softening,
ubiquitous-joint, bilinear
strain-hardening/softening
ubiquitous-joint, double-
yield, modified Cam-clay, and
Hoek-Brown) Braced excavation
➢User-defined models written
in FISH (FLAC only)
➢Continuous gradient or
statistical distribution
Advanced, Twoof any
and Three
Dimensional
property may be specified
Continuum Modeling for Geotechnical Analysis
of Rock, Soil, and Structural Support
FLAC & FLAC3D Basic Features

➢Built-in programming
language (FISH) to add user-
defined features
➢FLAC and FLAC3D can be
coupled to other codes via
TCP/IP links
➢Convenient specification of
boundary conditions and
initial conditions

Model grid for service tunnel

connecting two main tunnels

Advanced, Two and Three Dimensional

Continuum Modeling for Geotechnical Analysis
of Rock, Soil, and Structural Support
FLAC & FLAC3D Basic Features

➢Water table may be defined for

effective stress calculations
➢Groundwater flow, with full
coupling to mechanical
calculation (including negative
pore pressure, unsaturated flow,
and phreatic surface conditions)
➢Structural elements,such as
tunnel liners, piles, sheet piles,
cables, rock bolts or geotextiles,
that interact with the
surrounding rock or soil, may be
modeled

Excavation supported by shotcrete

wall, tiebacks and soilnails

Advanced, Two and Three Dimensional

Continuum Modeling for Geotechnical Analysis
of Rock, Soil, and Structural Support
FLAC & FLAC3D Basic Features

➢Automatic 3D grid generator

(FLAC3D) using pre-defined
shapes that permit the creation
of intersecting internal regions
(e.g., intersecting tunnels)
➢Full graphical user interface in
FLAC; partial gui in FLAC3D (for
plotting and file handling)
➢Extensive plotting features –
contours, vectors, tensors, flow,
etc.)
➢Graphical output in industry-
standard formats includes
PostScript, BMP, JPG, PCX, DXF
(AutoCAD), EMF, and a clipboard Sequential excavation and support
for a shallow tunnel
option for cut-and-paste
procedures

Advanced, Two and Three Dimensional

Continuum Modeling for Geotechnical Analysis
of Rock, Soil, and Structural Support
FLAC & FLAC3D Optional Features

➢Optional modules include:

• thermal, thermal-mechanical, and
thermal-poro-mechanical analysis
including conduction and advection;
• visco-elastic and visco-plastic (creep)
material models;
• dynamic analysis capability with quiet
and free-field boundaries, and
• user-defined constitutive models
written in C++
• two-phase fluid flow (FLAC only)

Liquefaction failure of
a pile-supported wharf

Advanced, Two and Three Dimensional

Continuum Modeling for Geotechnical Analysis
of Rock, Soil, and Structural Support
 FLAC Version 5 & FLAC3D Version 3
Features
1. Hysteretic damping – more realistic and more efficient than
Rayleigh damping for dynamic analysis
2. Built-in Hoek-Brown constitutive model
3. Thermal advection (convection) logic for thermal / fluid-flow
analysis
4. Network key license version
5. More efficient calculation of fluid-flow / mechanical analysis
(FLAC)
6. New structural element types: liner elements, rockbolt elements,
strip elements (FLAC)
7. Increased calculation speed (10-20% faster) due to optimization to
calculation cycle and updated compiler (FLAC3D)
8. New MOVIE facility in AVI or DCX format (FLAC3D)
9. Optional hexahedral-meshing preprocessor (3DShop) to facilitate
creation of complex meshes (FLAC3D)
 New Features in FLAC Version 5.1
Pre-release available August 2006
Official release in early 2007

1. Speedup of double-precision version by converting to Intel Fortran

compiler.

2. Automatic re-meshing logic.

3. Parallel processing on multiprocessor computers

(e.g., dual processors or dual core processor)
 New Features in FLAC3D Version 3.1
Pre-release available now
Official release in November 2006

1. Parallel processing on multiprocessor computers

(e.g., dual processors or dual core processor)

2. New structural element type “Embedded Liner” provides shear/slip and

normal interaction with the grid on both sides of the liner (e.g., to simulate
buried sheet pile walls)

3. New Mixed Discretization scheme for tetrahedral elements “Nodal Mixed

Discretization” provides more accurate solution of plasticity problems using
tetrahedral grids.

4. 64 bit version of FLAC3D*

5. Help File containing Command Reference, FISH Reference and Example

Applications.*

6. Tunnel extrusion grid generator tool.*

*not yet available

MODELLING-STAGE TABS
Finite Difference Formulation
of FLAC
 BASIS OF FLAC

FLAC solves the full dynamic equations of motion even for

quasi-static problems. This has advantages for problems that
involve physical instability, such as collapse, as will be
explained later.

To model the “static” response of a system, a

relaxation scheme is used in which damping absorbs kinetic
energy. This approach can model collapse problems in a more
realistic and efficient manner than other schemes, e.g.,
matrix-solution methods.
 A SIMPLE MECHANICAL ANALOG

m
u, u , u
F(t)

Newton´s Law of Motion

du
F = ma = m
dt
For a continuous body, this can be generalized as
dui  ij
 = + g i
dt x j
where  = mass density,
xi = coordinate vector (x,y)
ij = components of the stress tensor, and
gi = gravitation
 STRESS-STRAIN EQUATIONS

In addition to the law of motion, a continuous

material must obey a constitutive relation -
that is, a relation between stresses and strains.
For an elastic material this is:

In general, the form is as follows:

where
 A GENERAL FINITE-DIFFERENCE FORMULA

In the finite difference method, each derivative in the previous equations

(motion & stress-strain) is replaced by an algebraic expression relating
variables at specific locations in the grid.

The algebraic expressions are fully explicit; all quantities on the right-hand
side of the expressions are known. Consequently each element (zone or
gridpoint) in a FLAC grid appears to be physically isolated from its neighbors
during one calculational timestep.

(The time-step is sufficiently small that information

cannot propagate between adjacent elements
during one step)

This is the basis of the calculation cycle:

 Basic Explicit Calculation Cycle

For all gridpoints (nodes)

velocities Equilibrium Equation nodal forces
(Equation of Motion)
du i ij
 = + gi
dt x j
Gauss´ theorem Fi = ijn jL

For all zones (elements)

Stress - Strain Relation

strain rates (Constitutive Equation) new stresses

e.g., elastic
FLAC’s grid is internally composed of triangles. These are
combined into quadrilaterals. The scheme for deriving
difference equations for a polygon is described as follows:

Overlaid Triangular element Nodal force vector

Elements with velocity vectors
 FLAC:
For all gridpoints...

Once all stresses have been calculated, gridpoint forces

are derived from the resulting tractions acting on the
sides of each triangle. For example,

Then a “classical” central finite-difference formula is used

to obtain new velocities and displacements:

(… in large strain mode)

 FLAC:
For all elements...

Gauss’ theorem, f
 nifdS = 
S A x
i
dA

is used to derived a finite difference formula for elements of arbitrary shape.

u (i b ) nodal velocity
b
S

a
u (i a ) nodal velocity

For a polygon the formula becomes f 1

  f n i S
x i A S

This formula is applied to calculating the strain increments, eij, for a zone:
u i

1
x j 2A S
 (u (i a ) + u (i b ) )n jS

1  u i u j 
eij =  +  t
2  x j x i 
 Overlay & Mixed-Discretization Formulation of FLAC:

+ /2 =

Each is constant-stress/constant-strain:

Volume strain averaged over . Deviatoric strain evaluated for

and separately
(Mixed discretization procedure)
Solution is “Updated Lagrangian” (grid moves with the material), and
explicit (local changes do not affect neighbors in one timestep )
 Methods of solution in time domain
numerical grid displacement
F
u
stress

u force
F
x
EXPLICIT IMPLICIT
All elements:
Assume (u) element
F = f (u, )
(nonlinear law)
are fixed F = Ku
All nodes: Assume (F) global
u  =  F t
m
are fixed mu+ Ku = F
Repeat for Correct if
n time-steps x min
t  Solve complete set of equations
Cp for each time step
No iterations
within steps p-wave speed Iterate within time step if
nonlinearity present
Information cannot physically
propagate between elements during
one time step
 Methods compared

Explicit, time-marching Implicit, static

1. Can follow nonlinear laws without 1. Iteration of the entire process is

internal iteration, since necessary to follow nonlinear laws
displacements are “frozen” within
constitutive calculation.

2. Solution time increases as N3/2 for 2. Solution time increases with N2 or

similar problems. even N3.

3. Physical instability does not cause 3. Physical instability is difficult to

numerical instability. model.

4. Large problems can be modeled 4. Large memory requirements, or disk

with small memory, since matrix is usage.
not stored.

5. Large strains, displacements and 5. Significantly more time needed for

rotations are modeled without extra large strain models.
computer time.
 Strengths & Limitations
The explicit solution scheme used in FLAC enables the following
problems to be solved most efficiently:

1. Strongly nonlinear systems, with extensive yield and large

strain.
2. Systems in which localization occurs.
3. Systems that embody complex interactions, or which need
special user-defined conditions or material models.

Disadvantages are:
1. Slow execution (compared to – say – finite elements) for
linear (or well-behaved) systems.
2. Slow execution if there are great contrasts in material
stiffnesses or element sizes.
 DYNAMIC RELAXATION

In dynamic relaxation gridpoints are moved according to

Newton’s law of motion. The acceleration of a gridpoint is
proportional to the out-of-balance force. This solution scheme
determines the set of displacements that will bring the system
to equilibrium, or indicate the failure mode.

There are two important considerations with dynamic relaxation:

1) Choice of timestep
2) Effect of damping
 TIMESTEP

In order to satisfy numerical stability the timestep must satisfy the

condition: xmin
t 
Cp

where Cp is proportional to 1 /mgp. For static analysis, gridpoint

masses are scaled so that local critical timesteps are equal ( t = 1 )
which provides the optimum speed of convergence. Nodal inertial
masses are then adjusted to fulfill the stability condition:

Note that gravitational masses are not affected.

 DAMPING

Velocity-proportional damping introduces body forces that can

affect the solution.

Local damping is used in FLAC --- The damping force at a

gridpoint is proportional to the magnitude of the unbalanced
force with the sign set to ensure that vibrational modes are
damped:
 LOCAL DAMPING

• Damping forces are introduced to the equations of motion:

t
ui = Fi − | Fi | sgn (ui )
m

where Fi is the unbalanced force

• The damping force, Fd is:

Fd = − Fi sgn( ui )

• In FLAC the unbalanced force ratio (ratio of unbalanced force, Fi , to the

applied force magnitude, Fm) is monitored to determine the static state.

• By default, when Fi / Fm < 0.001, then the model is considered to be in an

equilibrium state.
 STATIC ANALYSIS

FLAC is a dynamic solution method that provides a static

solution (with the effect of inertial forces minimized) provided
the unbalanced force ratio reaches a small value (~ 0.001 or
less).

This is comparable to the “level of residual error” or “convergence

criterion” defined for matrix solution methods used in many finite
element programs. In FLAC, the level of error is quantified by the
unbalanced force ratio. In both FLAC and FE solutions, the static
solution process terminates when the error is below a desired value.
The collapse load can be determined from either :

1. A load-controlled test, i.e., apply a constant force and calculate the

solution. (stable or unstable?) Iterate until the difference between the stable
and unstable load is smaller than a selected tolerance.

2. A velocity-controlled test, i.e., apply a small constant velocity until an

unstable state is reached.

load

settlement
 Training Schedule
August 8, 2006 (morning)

09:00-10:00 Numerical Analysis of Continuum and Discontinuum

Mechanics
- DEM versus continuum analysis numerical methods
Introduction to Material Models to Simulate Geological
Materials
- Characteristics of soil and rock
- Constitutive models to represent continuum and
discontinuum behavior
- Selecting appropriate material models and properties
10:00-10:15 Break
10:15-12:00 Introduction to Material Models to Simulate Geological
Materials (continued)
 Training Schedule
August 8, 2006 (afternoon)

01:00-03:00 Model Building – Grid Generation

- Grid building/altering/shaping tools; adding interfaces
Model Building – Basic Material Models
- Assigning materials and properties in a FLAC model
03:00-03:15 Break
03:15-05:00 Model Building – Boundary Conditions / Initial Conditions
- Applying boundary and initial conditions
Model Building – Solution
- Solving for equilibirum and monitoring model response
Model Building – Result Interpretation
- Plotting unbalanced force, gridpoint velocities,
plasticity indicators
 DISCONTINUUM ANALYSIS TECHNIQUES
NUMERICAL SCHEMES TO MODEL CONTACTS
OR INTERFACES BETWEEN DISCRETE BODIES

Discrete Element Methods Continuum Methods

(DEM) For example:
Various DEM schemes exist.
... main differences are associated with:
• Finite Elements with Joints
Rigid
Contacts
Deformable
• Finite Differences with interfaces
Solid Rigid
Materials Deformable
• Limit Equilibrium Methods
Static
Solution
Dynamic
 Definitions

The name “Discrete Element Method” (DEM)

should be applied to a method only if it:
1. allows finite displacements and rotations of
discrete bodies; including complete detachment
2. recognizes new interactions (contact)
automatically as the calculation progresses

The name “Distinct Element Method” is used for a

DEM that uses an explicit dynamic solution to
Newton’s laws of motion.
• A discrete element code will embody an efficient
algorithm for detecting and classifying contacts. It
will maintain a data structure and memory
allocation scheme that can handle many hundreds
or thousands of discontinuities or contacts.

• Finite element codes for modeling “discontinua”

are often modified continuum programs, which
cannot handle general interaction geometry (e.g.
many intersecting joints). Their efficiency may
degenerate drastically when connections are
broken repeatedly.
 Overview of DEM & explicit, dynamic method

The formulation of is very simple. For example, for a ball impacting a wall,

(all contacts, in general)

If x  R, F ( t ) = 0
mass m One time step,  t
If x  R, F (t )
= ( R − x ) kn
(t )

R
(all particles, in general)
F
u u ( t +t / 2) = u ( t −t / 2) + F ( t )  t / m
x x ( t +t ) = x ( t ) + u ( t +t / 2)  t Full dynamic equations
(integration of Newton’s 2nd law)

unknowns knowns } Explicit solution scheme

(central difference – 2nd order accurate)

Three consequences of this formulation are as follows …

1. Treating each body as discrete (DEM) allows discontinuous
material (such as a rock mass) to be modeled easily.

2. Full dynamic equations of motion allow the evolution of unstable

systems to be simulated realistically.

3. Explicit solution scheme makes the task of handling

nonlinearity trivial. Examples of nonlinearities are: (a) contact
making & breaking; (b) softening material behavior (rock-like); e.g.,

The explicit scheme

m force
uses a time step so t 
small that information k
cannot propagate
between neighbors in OUTPUT
one step.

Thus, each element is isolated

during one step, enabling displacement

INPUT
 COMPUTATION CYCLE IN THE DEM

All the contacts

CONSTITUTIVE
kn
u
ks
 F u
s

n
Fs ALL THE BLOCKS

Fn = Fn − k n un n

c
Fs = Fs − k s us Fi
Fs = minFn , Fs sgn(Fs )
zone
ALL THE BLOCKS
c node
Fi At the element

MOVEMENT
xi 1  du i du j 
 ij = + t
+

2  dx j dx i 
M
MOVEMENT

At the centroid  ij = C( ij ,  ij ,... )

c
Fi =  F i At the node
M = e ijx i Fj Fie =   ijn jds
t = t + t
z
 i = Fi / m
u c

 = M / I Go to Fi = F + Fi
i
e

 i = Fi / m
u
 What is the applicability of each code?

In general, if there are few discontinuities in the application,

FLAC or FLAC3D may be used.

If the application contains many discontinuities, UDEC or

3DEC should be used, because these codes allow easy
specification of multiple joint sets.

For granular materials or solids that may fracture, PFC is

the best choice.

Note that all Itasca codes may be coupled – e.g., a FLAC

model may contain regions represented by PFC.
 Characteristics of soil & rock

1. Behavior changes in character, according to stress state (e.g axial splitting

in unconfined test; shear bands when confined).

2. Memory of previous stress or strain excursions, in both magnitude and

direction. (c.f. - moving yield surfaces, evolving anisotropic damage tensors,
Kaiser effect)

3. Dilatancy that depends on history, mean stress and initial state.

4. Continuously nonlinear stress-strain response, with ultimate yield, followed

by softening or hardening.

5. Hysteresis at all levels of cyclic loading/unloading.

6. Transition from brittle to ductile shear response as the mean stress is

increased.

continued …
 Characteristics of soil & rock … continued

7. Dependence of incremental stiffness on mean stress and history.

8. Induced anisotropy of stiffness and strength with stress & strain path.

9. Nonlinear envelope of strength.

10. Spontaneous appearance of microcracks and localized macro-

fractures in rock, and shear bands in soil.

11. Spontaneous emission of acoustic energy.

It would take a complicated constitutive law to reproduce all of
these phenomena.

If such a model existed (which it doesn’t ), very many material

parameters or internal state variables would be needed.
(For example, some existing laws have 20 parameters, and/or
families of yield surfaces involving perhaps 100 state variables).
 What mechanisms should be included in a model?

Only include those things that actually impact the relevant behavior
of the system (i.e., things that are important to successful design).
The following examples illustrate the modeling approach for several
particular requirements:

Collapse or ultimate failure of the system:

- use elastic/plastic law (no effect of moduli); try FLAC/Slope
Monotonic loading; displacements are important:
- use simple hardening law (yield stress increases with strain)
Cyclic loading; damping is important:
- use hysteretic damping option in FLAC/FLAC3D
Cyclic loading; volume-change is important (e.g.,liquefaction):
- use empirical void-collapse scheme in FLAC/FLAC3D
General loading paths; several nonlinear effects important
- must consider complex constitutive model, OR …
… use a micromechanical approach, in which complexity arises
automatically from the interaction of many simple objects
(emergent behavior*).

Note that all 11 of the characteristics of soil & rock (listed earlier)
are reproduced by a micromechanical model consisting of an
assembly of frictional and/or bonded particles.
(Calibration is needed to match the observed magnitude of each effect)

* Often, a collection of simple objects exhibits complex

behavior at the system level. This is an example of “emergent
behavior” (e.g., see “Emergence” by Steven Johnson, Scribner
2001).
In this case it is not necessary to invent complex constitutive
laws – just create a system of the appropriate micro-elements,
and the complex behavior will emerge “automatically.”
Constitutive Models for FLAC and FLAC3D
Built-in Models User-defined Models*
Elasticity models: Elasticity models:
Isotropic Hyperbolic elastic
Transversely isotropic Duncan-Chang, 1980
Orthotropic
Plasticity models:
Plasticity models: NorSand
Drucker-Prager Jardine et al., 1986
Mohr-Coulomb Manzari-Dafalias, 1997
Ubiquitous-joint Kleine et al., 2006
Strain-hardening/softening Concrete hydration
Bilinear strain-hardening/softening/ubiquitous-joint vonWolffersdorff hypo-plastic
Double-yield
Modified Cam-clay Dynamic Liquefaction models:
Hoek-Brown UBCSAND
UBCTOT
Dynamic Liquefaction models:
Wang, 1990
Finn (Martin et al., 1975) model Roth et al.,2001
Bryne, 1991 model
Andrianopoulos, 2005
Creep models: Creep models:
Viscoelastic Minkley viscoplastic
Burger’s substance viscoelastic Hein-crushed salt
Two-component power law
Salzer creep
Reference creep formulation (WIPP) Lubby2 creep
Burger-creep/Mohr-Coulomb viscoplastic
Two-component power law/Mohr-Coulomb viscoplastic
WIPP-creep/Drucker-Prager viscoplastic *partial list of models created by
Crushed-salt
or developed for code users
 FLAC CONSTITUTIVE MODELS
Model Representative material
Null void
Elastic homogeneous, isotropic continuum;
linear stress- strain behavior
Anisotropic thinly laminated material exhibiting
elastic anisotropy
Drucker-Prager limited application; soft clays with
low friction
Mohr-Coulomb loose and cemented granular materials
soils, rock, concrete
Strain-hardening/softening granular materials that exhibit nonlinear
Mohr-Coulomb material hardening or softening
Ubiquitous-joint thinly laminated material exhibiting
strength anisotropy (e.g., slate)
Bilinear strain-hardening/ laminated materials that exhibit non-
softening ubiquitous-joint linear material hardening or softening
lightly cemented granular material in
Double-yield which pressure causes permanent
volume decrease
Modified Cam-clay materials for which deformability and shear
strength are a function of volume change
Hoek-Brown * isotropic rock material
*new in FLAC 5
Example application
holes, excavations, regions in which
material will be added at later stage
manufactured materials (e.g. steel)
loaded below strength limit; factor of
safety calculation
laminated materials loaded below
strength limit
common model for comparison to
implicit finite-element programs
general soil or rock mechanics
(e.g., slope stability and underground
excavation)
studies in post-failure (e.g., progressive
collapse, yielding pillar, caving)
excavation in closely bedded strata
studies in post-failure of laminated
materials
hydraulically placed backfill
geotechnical construction on soil
geotechnical construction in rock
 CONSTITUTIVE MODELS
FOR CONTINUUM ELEMENTS

•NULL all stresses are zero: for use as a void - e.g., for excavated regions
•ELASTIC isotropic, linear, plane strain or plane stress
•ANISOTROPIC elastic,assumes that the element is transversely anisotropic:

g b

−b planes are planes of symmetry. The , b axes may be at any angle f to the x, y axes:

y
 b

f x
 FLAC PLASTICITY MODELS
Drucker-Prager
Mohr-Coulomb
Ubiquitous-Joint
Strain-Hardening-Softening
Double-Yield
Modified Cam-clay
Hoek-Brown
1. All models are characterized by yield functions, hardening/softening functions and flow rules.

2. Plastic flow formulation is based on plasticity theory that total strain is decomposed into elastic
and plastic components and only the elastic component contributes to stress increment via the
elastic law. Also, elastic and plastic strain increments are coaxial wuth the principal stress axes.

3. Ducker-Prager, Mohr-Coulomb, Ubiquitous Joint and Strain-Softening models have a shear yield
function and non-associated flow rule.

4. Drucker-Prager, Mohr-Coulomb, Ubiquitous Joint and Strain-Softening models define the tensile
strength criterion separately from the shear strength and use an associated flow rule.

5. All models are formulated in terms of effective stresses.

6. Double-yield and modified Cam-clay models take into account the influence of volumetric change
on material deformability and volumetric deformation (collapse).

7. Hoek-Brown incorporates a nonlinear failure surface with a plasticity flow rule that varies with
confining stress.
 CONSTITUTIVE MODELS — DRUCKER-PRAGER

•Drucker-Prager elastic/plastic with non-associated flow

rule: shear yield stress is a function of
isotropic stress

t
A

kf ft=0
B

C 
t
kf /qf

Drucker-Prager Failure Criterion in FLAC

 CONSTITUTIVE MODELS — MOHR-COULOMB
•Mohr-Coulomb elastic / plastic with non-associated flow rule: operates on
major and minor principal stresses

3

ft=0
B
C c
2c t tan f
A Nf
1

Mohr-Coulomb Failure Criterion in FLAC

t
(for constant n)
shear
stress
slope = G
shear strain g
CONSTITUTIVE MODELS – UBIQUITOUS-JOINT MODEL

•Ubiquitous-Joint Model uniformly distributed slip planes embedded in a

Mohr-Coulomb material

element 
n


rigid-plastic, dilatant
tmax = c j + n tanf
Mohr-Coulomb

Note:  rotates with the element in large-strain mode

A t

cj f t=0
B

 tj C 22
cj
tan f j
CONSTITUTIVE MODELS — STRAIN-SOFTENING / HARDENING
•Strain-softening / hardening identical to the Mohr-Coulomb model except that f, C and 
are arbitrary functions of accumulated plastic strain (gp )*

t
gp

f
g
produces

gp v


g

gp
Input by user Output

 
1

 g p =  (e11
P (d )
) + (e22P(d ) ) + (e12P )
2 2 2 2
 CONSTITUTIVE MODELS
BILINEAR STRAIN-HARDENING/SOFTENING MODEL
• Bilinear model a generalization of the ubiquitous-joint model. The failure envelopes for
the matrix and joint are the composite of two Mohr-Coulomb criteria with
a tension cut-off. A non-associated flow rule is used for shear plastic flow
and an associated flow rule for tensile-plastic flow.

t
3
A

Cj2
c2
Cj1 C
B c1 tanf2 fj2
Nf1 1 C D t tanf1
A
Nf2
1
1 D fj1 3’3’
jt

FLAC bilinear matrix failure criterion FLAC bilinear joint failure criterion
 CONSTITUTIVE MODELS – DOUBLE-YIELD MODEL

• Double-yield model extension of the strain-softening model to simulate

irreversible compaction as well as shear yielding.
 CONSTITUTIVE MODELS - MODIFIED CAM-CLAY MODEL

• Modified Cam-Clay model incremental hardening/softening elastic-plastic model,

including a particular form of non-linear elasticity and
a hardening/softening behavior governed by volumetric
plastic strain (“density” driven).

v
q
N
normal
consolidation line
plastic
vl
dilation
A
vk A
pc − e p  0 plastic compaction
qcr = M
vkB k
1
B 2 − e p  0
l
swelling lines 1

ln p1 ln p pc pc p
pcr =
2
Normal consolidation line and swelling line Cam-Clay failure criterion in FLAC
for an isotropic compression test
 CONSTITUTIVE MODELS – HOEK-BROWN MODEL

• Hoek-Brown model empirical relation that is a nonlinear failure surface which

represents the strength limit for isotropic intact rock and
rock masses. The model also includes a plasticity flow
rule that varies as a function of confining stress.
BUILT-IN MATERIAL MODELS
FLAC Interface Model
 FLAC (OR CONTINUUM CODE)

Use for problems at either end of the joint-density spectrum

single or isolated discontinuities multiple, closely-packed blocks

“interface” “ubiquitous jointing”

problems
 INTERFACES

• Interfaces represent planes on which sliding or separation can occur:

- joints, faults or bedding planes in a geologic medium

- interaction between soil and foundations

- contact plane between different materials

• To join regions that have different zone sizes

• Elastic-plastic Coulomb sliding:

- tensile separation of the interface, and

- axial stiffness to avoid inter-penetration

 INTERFACE MECHANICS

Each node on the surface of both bodies owns a length, L, of interface for the purpose of converting
from stress to force. L is calculated in the following way

A1 B1 C1 D1 Body 1

B2 C2 E2
Body 2
A2 D2

LB2 LC2 LB1 LD2 LC1 LD1

LINEAR MODEL
n= -Knun [Kn]=stress/disp
t = -Ksus
t = max (tmax, t) sgn (t)
tmax= ntan f+c
Fn = nL
Fs = t L
 INTERFACE ELEMENTS
PROCEDURE
1. Form interface using grid generation commands

2. Null out region

bside
(i3, j3) (i4, j4)
(i2, j2)
(i1, j1)
aside

3. Move grid halves together

4. Declare interface
int n aside from i1, j1 to i2, j2 bside from i3, j3 to i4, j4
5. Input the interface properties
int n ks =... kn = ... fric =... coh =...
 INTERFACE PROPERTIES
kn : normal stiffness [stress/displacement]
ks : shear stiffness [stress/displacement]
cohesion : cohesion [stress]
friction : friction angle [degrees]
dilation : dilation angle [degrees]
tbond : tensile strength [stress]

If the interface is used to attach two sub-grids,it is necessary to declare it glued.

Properties estimation
• Sub-grids attached:
- declare glued
- set kn and ks = 10 *

• Geologic joints
- shear tests; considering “scale effect”
- kn and ks for rock mass joints, can vary between 10-100 MPa/m
for joints with soft clay in-filling, to over 100 GPa/m for tight joints
in basalt or granite.
 INTERFACE CONDITIONS

1. Glued Interface --- No slip or separation is allowed, but elastic displacement, defined by
kn and ks, occurs.

2. Unbonded Interface --- Slip occurs as defined by Coulomb shear-strength criterion

(and including dilation at onset of slip). The interface has zero tensile bond strength.

3. Bonded Interface ---- It a tensile bond (tbond) strength is specified, the interface acts
as if glued while the normal stress is below the bond strength. If magnitude of normal
stress exceeds bond strength, the bond “breaks” (tbond is set to zero) and the interface
behaves as an unbonded interface.

A shear bond strength is also specified when tbond is set, in which case the bond will break
if either the shear stress exceeds the shear bond strength (sbratio*tbond) or the normal
stress exceeds the normal bond strength (tbond). The interface then reverts to unbonded.
(By default, sbratio = 100.)

If bslip=on is specified, slip (defined by the Coulomb criterion) can occur even though
the interface is still bonded. Dilation is suppressed in this case.
INTERFACE MODEL

Create interface
and assign properties
Key Features of FLAC for Grid Generation

1. FLAC is command-driven.
2. GIIC Build tools provide
mouse-driven facilities for
grid generation from
templates.
3. FISH tools in the FISH
Library are used to create
complicated grid shapes.
 Geometry – grid setup
1. Always start with a coarse grid*; work out the bugs,
and increase the grid density only as much as
necessary (are results converging?).
2. Avoid badly-shaped zones, and sudden jumps in zone
widths.
3. Avoid high aspect ratios in regions of high strain
gradients.
4. Make sure the boundaries are far enough away to
avoid influencing the results.
5. Try to avoid triangular zones at free surfaces,
especially if performing large-strain plasticity analysis.

* For dynamic analysis, the zone size should be small

enough to model wave propagation accurately.
 Boundary conditions
There are two main classes of boundary conditions: velocity
conditions and stress conditions (although there are additional
conditions in dynamic simulations).

Both can be activated with the APPLY command: e.g.,

APPLY XVEL=1.0 I=1 J=1,5 ; FLAC … velocity
APPLY SXX=-1e5 J=21 ; FLAC … stress
APPLY SXX=-1e5 RANGE Z=19.9 20.1 ; FLAC3D
For historical reasons, the velocity conditions can also be
set with a FIX command and an INI command: e.g.,
FIX X I=1 J=1,5
INI XVEL=1.0 I=1 J=1,5
The latter 2 commands achieve the same effect as the
first APPLY command above.
Boundary locations

stress

displacement

Extreme grids – “tunnel”

sizes are the same
ATTACH - accuracy
Load applied here

Note smooth displacement contours

 Grid Generation

Create Mesh

Alter Mesh to Fit Shape

Material Models and Properties
Boundary and Initial Conditions

Histories, Tables, FISH Library

Global Settings
Solution

Result Interpretation - Plotting

 Training Schedule
August 9, 2006 (morning)

09:00-10:00 Introduction to FISH

- FISH variables, arithmetic, systax ad data types
- Writing FISH functions
- Simple exercises using the FISH Editor & Library
10:00-10:15 Break
10:00-12:00 Factor of Safety Calculation
- Implementation of the strength reduction method in FLAC
- Application of FLAC for factor-of-safety calculations
 Training Schedule
August 9, 2006 (afternoon)

01:00-03:00 Soil/Rock – Structure Interaction

- Beams, liners, cables and rockbolts
- 2D/3D equivalence
03:00-03:15 Break
03:00-05:00 Simulating Support for Underground Excavations and
Emplacement Drifts
- Using interface elements for tunnel liner and rock
interaction
FISH - The programming language
of FLAC
 FISH (1)

FISH is a compiler. Functions are entered via a data file and

are translated into a list of instructions stored in the memory
space of the code.

Variable names and values are available for monitoring and

changing at any time.
 FISH (2)
Common ways to use FISH:

1. Special-purpose operations; e.g., grid generation, profile of

material properties, automation of input commands, plot or
print user-defined variables.

2. Use as a HISTORY variable.

3. Automatic execution during stepping; e.g., use as a servo-control for

numerical test (with WHILE_STEPPING command).

4. Drive a data file; e.g., change parameters while calculation

progresses (using COMMAND statements).

5. Use as a constitutive model function; e.g., apply a user-written

constitutive model.
 FISH Variables, Functions and Operations

FISH handles definitions of:

• User-defined variables or function names.

• Pre -defined scalar variables.
• Grid variables (e.g., stresses, properties).
• Intrinsic functions.
• Tables, general memory access.

Some remarks
FISH is case-insensitive. All characters after a semi-colon (;) are ignored.
If parameters are to be passed to functions, then they must be set beforehand
by using the SET command.
If a number is expected in FLAC, it can be substituted by FISH symbols.
As soon as a variable is mentioned in a valid FISH program line, it is
globally recognized both in FLAC commands and FISH code.
 FISH Control Statements (1)

DEF <name>
To define a FISH function ...
END

Conditional statements
CASE_OF expr IF expr 1 test expr 2
··· ···
CASE n ELSE
··· ···
END_CASE END_IF

‘Looping’ statements

LOOP var (expr1, expr2) LOOP WHILE expr1 test expr 2

··· ···
END_LOOP END_LOOP
 FISH Control Statements (2)

Sectioning statements

SECTION
···
END_SECTION

COMMAND
···
END_COMMAND

EXIT
EXIT SECTION
 FISH Specification Statements

WHILESTEPPING (execution of the function at every FLAC step)

WHILE_STEPPING

CONSTITUTIVEMODEL (the function is taken to be a new constitutive model)

CONSTITUTIVE_MODEL

INT (change the type of the associated variable)

FLOAT
STRING

ARRAY var(n1, n2) (definition of an array)

 FISH Functions

Mathematical functions Message functions

atan atan2 cos in out
exp tan ln
log sin sqrt Random generator
abs max min grand urand
sgn
Logical operators
Type conversion and not or
float int string
type Others
fc_arg get_mem lose_mem
Tables
xtable ytable table

Memory Access
imem fmem
 FISH Editor

The FISH Editor allows you to create and edit text files that contain FISH functions.

FISH functions defined in this way can be executed using the UTILITY/FishLib tool
if they are stored within the /flac/gui/fishlib directory. You can also run FISH
functions directly using the Run/Execute menu item from the FISH Editor.
To automate the execution of FISH functions, special comment lines are included in
the file. There are four types of input field:

1. Name: This is the name of the primary FISH function to run.

(A file can have more than one FISH function.)

2. Diagram: This is the name of an optional file name of an image (GIF/JPG)

that shows what the FISH function does.

3. Input: This contains the input values for the function.

4. Note: This contains notes and comments that describe the FISH function.
 FISH Input Parameter Data

The Input/Define parameters menu item brings up a dialog that allows input
parameters to be defined. These will have values requested when the FISH
function is run either through the FISH Library (using the UTILITY/FishLib tool),
or executed here.

The input parameters for FISH functions are entered as a comment string of the
form:

;Input: name/type/value/description

in which

1. Name - FISH variable name.

2. Type - int/float/string corresponding to data type:

integer, floating-point or string.

3. Value - Default value for parameter.

4. Description - Helpful string describing what the parameter is.

 FISH Notes

The Input/Define notes menu item brings up a text area where comments
can be saved.

Here you can describe the FISH function and these comments will be shown when
you try to execute the function from either the FISH Editor or the FISH Library
(using the UTILITY/FishLib tool).

These lines are added to a FISH file as comments prefixed by [Note:]

 FISH Library

The directory flac/gui/fishlib/ contains files with FISH functions accessed from the
UTILITY/FishLib tool.

These FISH functions have special comment lines included to allow the GIIC to identify
input parameters, notes and diagrams.

The directory structure inside flac/gui/fishlib/ is mirrored in this tool as a tree structure.
 Application of the shear strength
reduction method in design:
using numerical solutions for factor of
safety
 Factor of Safety (FS) in
Geotechnical Engineering

structural mechanics approach

load capacity
FS =
acting load

load = force, moment, pressure

 Footings

1
qu = cN c + qN q + g BN g
qu 2
Q FS bc =
qQ
Q
qQ =
B
bearing capacity theory
q
FoS calculation independent of
B load capacity and acting load calculation
 Retaining Walls
M r = W rw
Mr M o = P rp
FS o =
Mo
Fr = W tan  A
F
FS s = r Fs = P
Fs 1
W qu qu = cN c + qN q + g BN g
FS bc = 2
P qmax
W M
q max = +
b S

1
B P = g H 2Ka
2
FoS calculation independent of earth pressure theory
load capacity and acting load calculation bearing capacity theory
 Slopes
qu
FSbc = q FS = ?
q

1
qu = cN + qN + g BN g*
*
c
*
q
2 load ?

bearing capacity theory unit weight

FS calculation independent of gf
load capacity and acting load calculation FSg =
g
Structural Mechanics Approach
Mr Fr
FS o = FS s =
Mo Fs

specified failure mode

statically determinate cases
global equilibrium
Strength Reduction Approach

t c + tan  c tan  c cd
FSt = = = = =
t d cd + tan d cd tan d tan  tan  d
τ

φ c
cd σ
φd
Alternative Strength Reduction
c tan 
FS c =  = const FS = c = const
cd tan d
τ τ

φd
φ c
c
cd σ σ
φ φ

tanφ and c varying nonlinearly

 Method of Slices
(Limit Equilibrium)
statically undeterminate cases
strength reduction c tan 
FSt = =
cd tan  d

specified failure mode

slices
global equilibrium
 Numerical-modeling approach -
• A “full” solution of the coupled stress/displacement,
equilibrium and constitutive equations is made with
codes like FLAC.

• Given a set of properties, the system is either found

to be stable or unstable.

• By performing a series of simulations, with various

properties, the factor of safety can be found that
corresponds to the point of stability.

• This approach is much slower, but much more general,

than the limit-equilibrium solution. Only in the past few
years has it become a practical alternative to the limit
equilibrium method (as computers have become faster).
 What is a “full numerical solution” and how does it
differ from the limit equilibrium method?

Numerical solution Limit equilibrium

Satisfied only for specific

Equilibrium Satisfied everywhere
Computed everywhere using
Stresses
field equations
Deformation Part of the solution
Yield condition satisfied
everywhere; failure surfaces
Failure
develop “automatically” as
conditions dictate
The “mechanisms” that
Kinematics develop satisfy kinematic
constraints
objects (slices)
Computed approximately on
certain surfaces
Not considered
Failure allowed only on
certain pre-defined surfaces;
no check on yield condition
elsewhere
Kinematics are not
considered – mechanisms
may not be feasible
A single numerical simulation – with given properties –
will show either failure or stability (like a single
physical model).

How do we get a factor of safety?

Several simulations are performed, with different

properties.

The strength properties (e.g., cohesion and friction)

are reduced by “trial” values of the factor of safety, as
follows -
1 tan f
f
-1
C trial
= trial
C trial
= tan { trial }
F F
How can the exact value of F trial be found quickly,
with the smallest number of simulations?

Dawson et al (Géotechnique 49, 1999) give the basis

for the bracketing method of finding numerically the
factor of safety. In essence, the interval between
trial
values of F giving failure and stability is
repeatedly halved. The process quickly converges,
and is stopped when the interval becomes small
(e.g., < 0.005).

In more detail, the scheme implemented in the code

FLAC is as follows …
Steps in the strength-reduction solution scheme for FLAC

1. Determine the “characteristic response time” of the system

in terms of steps needed for equilibrium – call it Nc.
2. Set F=1.0, and keep halving it until lower bound (first
stable case) is found – call it Fs.
3. Keep doubling F until upper bound (first unstable case) is
found – call it Fu.
4. Set F = (Fu+Fs)/2, and determine if stable (then set Fs=F)
or unstable (then set Fu=F).
5. If Fu-Fs < 0.005, then stop, else go to 4.
 How is instability (failure) determined?

The condition of “stability” or “instability” is determined with

a program-specific method.

For example, with implicit, matrix-solution finite element

codes, the condition of instability is often based on the
non-convergence of the system of equations (see Griffiths
and Lane, 1999).

In FLAC, instability is determined by monitoring the kinetic

energy in the model. The change in kinetic energy is
measured by the “unbalanced force ratio.”
 Definition of stability/instability in FLAC

Steps to test for stability/instability:

1. Do up to Nc steps. Record unbalanced force ratio, Ru.

2. If Ru falls below 0.001 during stepping, exit as stable.
3. If (Ru – Ru(old)) / Ru < 0.1, exit as unstable.
4. If total iterations (steps 1-3) > 6, exit as unstable.
5. Go to 1.

During the whole process, the following information is displayed

• the number of calculation steps completed in 1 – as a % of Nc,

• the number of completed solution cycles (steps 1-3),
• the current values of Fu and Fs (brackets).
How good is the scheme? We can compare it with exact solutions.
The following example, solved analytically by Chen (1975), has a
factor of safety of 1.0.

This example was set up with FLAC, using two different

grids. The results are …
 Non- Associated
associated flow rule
Coarse grid (20x20) 0.99 1.03
Fine grid (40x40) 0.97 1.01
Note that the exact solution assumes an associated flow rule (dilation
angle equals friction angle)
JOB TITLE : (*10^1)

FLAC (Version 4.00)

LEGEND
1.400
9-Oct-01 18:09
step 18546
-1.167E+00 <x< 2.217E+01
-5.167E+00 <y< 1.817E+01
1.000
Boundary plot

0 5E 0
Max. shear strain-rate
0.00E+00 0.600
5.00E-08
1.00E-07
1.50E-07
2.00E-07
2.50E-07
0.200
3.00E-07
3.50E-07

Contour interval= 5.00E-08

Velocity vectors
-0.200
Max Vector = 2.294E-07

0 5E -7
Itasca Consulting Group, Inc.
Minneapolis, Minnesota USA
0.200 0.600 1.000 1.400 1.800 2.200
(*10^1)

Velocity field at failure – fine grid & associated flow rule

 LOS ANGELES ALAMEDA CORRIDOR

Practical application
of the strength
reduction method

10-mile long corridor

33-foot deep, 51-feet wide
freight rail connection
from ports of Long Beach
and Los Angeles to rail hub
in downtown Los Angeles

Design by URS Corporation

 LOS ANGELES ALAMEDA CORRIDOR

Trench walls are 3-foot diameter cast-in-place reinforced concrete piles,

4 feet on center with shotcrete on inside, and supported by pre-cast concrete
struts at top.
Stage 3 of construction is critical because potential for “kick-out” failure
governs required pile length.

URS Corporation
 LOS ANGELES ALAMEDA CORRIDOR

FLAC analysis for

Factor of Safety includes
soil-structure interaction
producing factor of 1.3.

Limit-equilibrium method
predicts failure, which
would result in over-
design of pile length by
up to 8 feet.

FLAC analysis resulted

in cost savings of several
million dollars.
URS Corporation
What are the advantages of using a numerical FoS solution?
1. Any failure mode develops naturally – no need to specify a
range of trial surfaces in advance.
2. There are no restrictions on geometry – all situations (slopes,
footings, tunnels, etc) are modeled in the same way.
3. No artificial parameters (e.g., functions for inter-slice force
angles) need to be given as input.
4. Multiple failure surfaces (or complex internal yielding) evolve
naturally, if the conditions give rise to them.
5. Structural interaction is modeled realistically – as fully-coupled
deforming elements, not simply as equivalent forces.
6. Kinematics is respected!

weak
planes
Stable with FLAC
Unstable by LE solution
What are the disadvantages?

- only one: Speed. A limit equilibrium program typically executes in

a fraction of a second. A numerical solution for FoS using a coarse
grid often takes less than a minute. A medium-grid solution may take
several minutes, and is usually quite accurate. A fine-grid solution
may take an hour or two.

Thus, there is no real drawback, given that most problems can

be solved in a few minutes.

One further perceived problem …

Programs that perform full numerical solutions are often
difficult to use – FLAC is no exception!
To make numerical solution for factor of safety easily accessible, a
new program – FLAC/Slope has been produced. This is as easy
(or perhaps, easier) to use than limit equilibrium programs such as
XSTABL or SLOPE/W.

FLAC/Slope has a simple graphical interface that is oriented to

setting up slope stability cases. It uses FLAC as the “computational
engine” but the user is completely insulated from it. Point-and-click
operations are all that are needed – for example:

Select slope type; then

enter dimensions
 FLAC/Slope allows -

• Several types of slope: e.g., benched slope, dam, embankment

• Several layers, and fairly general layer geometry
• Library of material properties – built-in and user-defined
• Water table – specified as an arbitrary surface
• Structural reinforcement: e.g., geo-grids, soil nails, rockbolts
• Weak plane, modeled as an interface
• Surface loads
• Regions can be excluded from the FoS calculation
• Instant comparison of runs using different parameters, and even
comparison of results from different projects
• Hard-copy reports and plotting in several formats
Various parameters may be included or excluded from variation
during the FoS solution. The following factors are included by
default: material friction and cohesion.
The following items can also be included:

1. material tension – tensile strength of materials

2. interface strength – cohesion and friction of interfaces
3. reinforcement grout strength – cohesion and friction of
reinforcement grout (soil/reinforcement interface) *
4. group – regions of space included in, or excluded from,
the scope of the parameter-variations *
* new in FLAC/Slope 5
Note that, at present, only Mohr-Coulomb material can be assigned
in FLAC/Slope. (Mohr-Coulomb and ubiquitous-joint materials
can be assigned in FLAC.)
 Structural elements in FLAC
• General Application :
– Design and analysis of structural support to stabilize a rock or soil
mass.
• Types of elements available:
– Beam
– Liner
– Cable
– Pile
– Rockbolt
– Strip
– Support Member
• Geometry:
– Linear element with 2 end nodes
 Structural elements in FLAC
There are seven types of structural element available in FLAC:

1. Beam elements. These allow bending, and are connected to the

grid (soil) either at nodes (rigid connection) or via interfaces, which
allow separation & slip. Plastic yield occurs as a function of axial
thrust, plastic moments can be specified.
2. Liner elements.* Similar to beams, and also include bending
stresses in yield criterion.
3. Cable elements. No bending resistance. Cable nodes are slaved to
grid motion in the normal direction, and via shear springs & slip
elements in the shear direction. Yield may occur axially. Cable
nodes may also be connected rigidly to gridpoints.
4. Pile elements. Bending resistance is included. Connection to the
grid is via yielding springs in both the normal and shear directions.

* new in FLAC 5.0

 Structural elements in FLAC
There are seven types of structural element available in FLAC:

5. Rockbolt elements.* Similar to piles, and also can account for

change in confining stress, strain-softening of grout, and tensile
rupture of element.
6. Strip elements.* Similar to rockbolts, but cannot sustain bending.
Shear behavior at strip/soil interface is defined by nonlinear shear
failure envelope that varies as a function of confining stress.
7. Support members. Simple 1D nonlinear spring elements that link
points on free surfaces (used in mining to represent – for example –
wooden props).

* new in FLAC 5.0

 Structural elements in FLAC
• Formulation:
– Each element type is characterized by a combination of
a) structural behavior
b) medium/structure interaction.
– The structural element logic is implemented in the framework of
FLAC two-dimensional Lagrangian, Explicit Finite-Difference
scheme, which uses a Dynamic Relaxation Method to solve static
problems.
 Beam Elements
Structural behavior:
• 3 degrees of freedom per node (2 translations
+ 1 rotation)
• Constant axial force, F; constant shear force,
T; linear moment, M
• Linear axial displacement, cubic deflection.
• Axial peak and residual strengths
• Can be joined together and/or the grid
• Nodal behavior may also include plastic
hinges.

Applications:
Modeling of structural support in which
bending resistance is important, including
sheet piles, support struts in an open-cut
excavation.
 Beam Properties
Structural:
• Cross-sectional area [or height and width, or radius]
• Elastic modulus
• Moment of inertia
• Axial peak and residual yield strengths

Optional:
• Plastic moment
• Density
• Thermal expansion coefficient
 Liner Elements
Structural behavior:
• similar to beam elements
• bending stresses are included in the yield
criterion

Applications:
Modeling of structural support in which
bending resistance, limited bending
moments and yield strengths are important,
such as concrete or shotcrete tunnel linings

Typical moment-thrust diagram for liner elements

 Liner Properties
Structural:
• Cross-sectional area [or height and width, or radius]
• Elastic modulus
• Moment of inertia
• Cross-sectional shape factor
• Thickness
• Axial peak and residual yield strengths

Optional:
• Density
• Thermal expansion coefficient
 Cable Elements
Structural behavior:
• One degree of freedom per node (axial
translation).
• Can also fail in tension and compression,
no flexural resistance.
Medium/structure interaction:
• Can be point-anchored or grouted so that
the cable element develops forces along its
length resisting relative motion between
cable and grid.
• May be pre-tensioned, if desired.

Applications: supports for which tensile

capacity is important, including
rock bolts, cable bolts and
tie-backs.
 Grout behavior accounted for in Cables

Conceptual Model

Constitutive Model
 Cable Properties
Structural:
• Elastic modulus
• Tensile yield
• Compressive yield

Grout:
• Stiffness
• Cohesive strength
• Frictional resistance

Optional:
• Density
• Thermal expansion coefficient
 Rockbolt Elements

Structural behavior:
• Similar to pile elements
• Can also account for:
- effect of changes in confining stress
- strain-softening of grout
- tensile rupture of element.

Applications: rock reinforcement in which

nonlinear effects of confinement,
grout bonding or tensile rupture
are important.
 Rockbolt Properties
Structural:
• Cross-sectional area [or radius]
• Elastic modulus
• Moment of inertia [automatic calculation for radius]
• Yield strength
• Tensile failure strain
Medium/structure interaction, Shear and Normal:
• Stiffness
• Cohesive strength
• Frictional resistance
• Exposed perimeter
Optional:
• Plastic moment
• Density
• Tables relating cohesion and friction to shear displ.
• Thermal expansion coefficient
Structural Boundary Conditions

• Free/fixed velocities (translation and rotation)

• Applied forces and moments
• Pin connection
 2D / 3D Equivalence in FLAC
Two dimensional modeling of structural features:
FLAC structural elements can be used to model structures which are either ‘long’ or
‘short but regularly spaced’ in the out of plane direction.

1. The element structural behavior is formulated in plane stress. To model ‘long’

features, the stiffness property should be divided by (1-  2) to account for plane-
strain conditions.

2. Reducing 3D problems with regularly spaced beams, liners, cables, piles,

rockbolts or support involves averaging the effect of 3D over the structure
spacing, S. FLAC uses a “linear scaling of properties” method proposed by
Donovan et al. (1984)* to distribute effects of elements over a discrete spacing.

3. The three dimensional effects associated with the flow of soil through a row of
piles can also be accounted for, in an approximate manner, by calibration of
properties associated with the normal component of the pile medium/structure
interaction.
*
Property Scaling – 2D/3D Equivalence
Property Scaling – 2D/3D Equivalence
Property Scaling – 2D/3D Equivalence
 Property Scaling – 2D/3D Equivalence
When the spacing keyword is specified in FLAC Version 5, structural
element properties are scaled automatically to account for regular
spacing. Gravity loads and pre-tensioning values are also scaled.
Actual structural element forces and moments will automatically be
printed and plotted, accounting for spacing.
Note, any loads or pre-tensioning that are applied to structural elements
(e.g., pre-loaded struts) using the STRUCT node n load command
should be scaled by dividing by S.
 Training Schedule
August 10, 2006 (morning)

09:00-10:00 Coupled Modeling - Introduction to Effective Stress and

Groundwater Analysis
- Effective stress calculation
- Governing equations for transient fluid flow and coupled
analysis
- Recommended approaches for coupled calculations
- Two-phase flow analysis
10:00-10:15 Break
10:15-12:00 Coupled Modeling – Introduction to Effective Stress and
Groundwater Analysis (continued)
 Training Schedule
August 10, 2006 (afternoon)

01:00-03:00 Coupled Modeling – Introduction to Thermal and

Thermal-mechanical Analysis
- Governing equations for thermal, thermal-mechanical and
poro-thermal-mechanical analysis
- Procedures for performing thermal and thermal-mechanical
calculations
- Constitutive models in coupled analyses
- Thermal loading and boundary conditions
03:00-03:15 Break
03:15-05:00 Coupled Modeling – Introduction to Thermal and
Thermal-mechanical Analysis (continued)
 Groundwater flow and
consolidation

FLAC models the flow of groundwater through a permeable solid,

such as soil.

The modeling of flow may be done:

- by itself, independent of the usual mechanical calculation of FLAC
- in parallel with the mechanical modeling, so as to capture the effects
of fluid / solid interaction.
 Unsaturated flow
The code handles both fully saturated flow
as well as flow in which a phreatic surface develops.

In the latter case:

- pore pressures are zero above the phreatic surface.
- a cubic law is used that relates the apparent permeability
to the saturation in the unsaturated region
- flow due to capillary pressures in the unsaturated region
is not included.

More complicated behavior can be implemented

using the two-phase flow option (described later)
 Consolidation
This type of behavior involves two mechanical effects:

First, changes in pore pressure cause changes in effective stress,

which affect the response of the solid — for example,
a reduction in effective stress may induce plastic yield.

Second, the fluid in a zone reacts to mechanical volume changes

by a change in pore pressure.
 Mathematical Formulation

Transport Law Compatibility Equation

 1  ui u j 
qi = −kij ( P −  w g k xk )  ij =  + 
x j mobility coefficient
2  x j xi 
Balance Laws
 q  ij dui
= − i + qv +  s gi =  s
t xi x j dt
wet density
Constitutive laws
P K w    
t
= 
n  t
−  
t 
d
dt
(  ij +  P ij ) = H ( ij ,  ij , k )
fluid bulk modulus Biot coefficient

s 1    
=  − 
t n  t t 
porosity
 Total versus Effective Stress Formulations

In FLAC, equilibrium is expressed using total stress:

 ij
+  s gi = 0 d : material dry density
x j n : porosity
s : saturation
 s = d + n  s   w w : fluid density

By definition of effective stress:

 ij =  ij − p ij
Substitution of the last 2 equations in the first, gives:
p xg
 ij p f f= − l l
+  d gi − (1 − n ) − ng w =0 w g g
x j xi xi
g w = w g
Solid weight
Drag
Buoyancy
(seepage force)
 Groundwater Modeling Approaches (1)

Choose simplest model consistent with mechanisms.

In order of complexity we can have:

1. No fluid-mechanical interaction (pore pressure distribution is

needed to compute correct effective stresses).

2. Flow calculation is used to obtain pore pressure distribution

(medium can be saturated or partially saturated with phreatic
surface).
 Groundwater Modeling Approaches (2)

3. Pore pressure generated by mechanical deformation.

There is no flow, the analysis can be static or dynamic
(e.g., undrained pore pressure buildup or liquefaction).

4. Coupled mechanical deformation and fluid flow.

a) time scale not important
b) time scale is important
 Groundwater Modeling in FLAC (1)

1. Effective Stress Calculation

- pore pressure fixed
- no groundwater flow
- specify pore pressures with the INITIAL pp command
or WATER table command (in non CONFIG gw mode)
- in non CONFIG gw mode, wet and dry densities of material
are supplied by the user
- in CONFIG gw mode:
- wet and dry densities are calculated by FLAC
- SET flow off
- set WATER bulk = 0
- if pore pressures changed instantaneously (e.g., dewatering),
use CONFIG ats to automatically adjust existing total stresses
 Groundwater Modeling in FLAC (2)

2. Flow-Only Calculation
- specify CONFIG gw and SET mech off
- pore-pressure distribution and phreatic surface location will be calculated
- specify correct permeability, but low fluid bulk modulus if only steady-state
condition is required
 Groundwater Modeling in FLAC (3)

3. Mechanical Generation of Pore Pressures with No Flow

- specify CONFIG gw and SET flow off
- specify correct fluid-bulk modulus; permeability is not required
 Groundwater Modeling in FLAC (4)

4. Coupled-Flow and Mechanical Calculation

- specify CONFIG gw and SET flow on
- specify realistic fluid-bulk modulus and permeability
- for basic-flow logic:
- use SET nmech SET ngw (default: nmech=1 ngw=1)
- SET force SET sratio (default: force=0 sratio=10-3)
- SET step SET clock (default: step=100000 clock=1440 min.)
- SOLVE auto on age
 When to Use Fast-Flow Schemes

* Consider flow incompressible if Kw >>> K + 4G/3

 Common Fluid-Flow Boundaries

Impermeable Boundary
- default conditions
- pore pressure free to vary
- saturation free to vary

Free Surface
- pore pressure fixed to zero (FIX pp)
- saturation free to vary if pore pressure fixed at zero

Applied Pore-Pressure Boundary

- pore pressure fixed (FIX pp)
 Permeability of Porous Medium (1)

Darcy’s Law expressed in terms of pressure is

dP
q=k
dx
where q is the specific discharge (in units of velocity - e.g., ft/s or m/s)
dP/dx is the pressure gradient (e.g., in psf/ft or Pa/m)
k is the mobility coefficient (e.g., in ft4/lb-sec or m2/Pa-sec )

The more usual expression of Darcy’s Law is

dh
q = KH
dx
where h is the head (e.g., in ft or m)
KH is the hydraulic conductivity (e.g., in ft/s or m/s).
 Permeability of Porous Medium (2)
Since P=gwh
(where g is the gravitational acceleration and w is the mass density of water),
KH
k=
g w
Another constant that is sometimes used is intrinsic permeability,
k, which is related to k and K by
K H
k= = k
g w

where  is the dynamic viscosity (e.g., units of lb-s/ft2 or Pa-s).

The units of k are [length]2 (e.g., ft2 or m2).

 Bulk Modulus of Water

P
Kw =
V / V

Kw = 4.18x107 psf (or 2.0 GPa) for pure water

Steady-State Flow
(a) fully saturated - solution-time independent
(b) partially saturated - solution time reduced by lowering Kw
(If Kw is too low, results are “erratic.” Set K w  0.3 z  w g )

Transient Flow
(a) flow-field solution (high modulus)
(b) phreatic surface migration (low modulus)
(c) use SET funsat algorithm to alternate solutions automatically
 Groundwater - tips
A fully coupled simulation with FLAC (e.g. a consolidation process)
using the basic fluid-flow scheme can be very time-consuming. The
FLAC manual provides detailed suggestions about various
approximations that can be made to reduce the solution time. The
important factors to consider are:

1. The ratio between the required simulated time and the

characteristic time of the diffusion process in the system.
2. The nature of the imposed perturbation (fluid or mechanical).
3. The ratio of fluid to solid stiffness.
 Increasing time step
When the fluid modulus is much larger than the bulk modulus of the
solid material, the timestep is small, and the simulation time long
for the basic fluid-flow scheme. It is possible to reduce (artificially)
the fluid modulus, without affecting the results; the allowed
reduction factor (for given error) depends on the problem
constraints, but in almost all cases the following upper limit of fluid
modulus gives minimal error:

K w  20n( K + 43 G)

It can often be reduced further. Rapid, partial simulations can be

made to assess the error introduced by various reduction factors.
 Caution!
If the timestep is small, and there are many steps, it may be
necessary to use the double precision version of FLAC 4.0. The
regular version uses single precision, which corresponds to an
accuracy of 1 part in 106. If – for example – a million timesteps are
executed, then accumulated quantities (such as pore pressure
increments) may be lost.

Note that the regular version of FLAC 5.0 uses double precision.
Unsaturated Flow in Porous Media
• The two phases are air and water
• Flow of air is neglected

Assumptions
• air phase is continuous
• pore air pressure is atmospheric
 Applications

• Slope stability
• Groundwater contamination control
• Design of hydraulic structures
References
Richards (1931), Philip et al. (1989)
van Genuchten (1982), Fredlund (1987), Forsyth (1995)
 Capillarity: micro-viewpoint
• Differences in molecular attraction between immiscible constituents
• Wetting (w) and non wetting (a) fluids
• Sharp interfaces:
– Surface tension
– Curvature

Capillary pressure

Pc = Pa − Pw
Capillary tube
 Capillary pressure

Pc depends on
• saturation
• geometry of the void space
• nature of solid and liquid

Micro-observation Macro-observation
 Saturation
Definitions:
• Fluid phase saturation
Vw Va
Sw = Sa = S w + Sa = 1
nV nV
• Residual saturation

S rw
• Effective saturation

S w − Srw
Se =
1 − Srw
 a=0.336 (clay)

Pc/P0 a=0.6 (sand) Cc/(P0tanf )

saturation saturation

van Genuchten relation Capillary-induced cohesion

Yield criterion for partially saturated soil:

Steady unsaturated flow around a
drift

movie
 Conclusions

1. Water is deflected from the drift roof and “drip lobes” are formed in which saturation
and flow velocity are increased (compared to initial steady state).

2. A “dry shadow” is formed, sheltered by the drift cavity.

 Rainfall on a Slope
Stable slope with initial
water table
(soil saturation above the
water table is ~ 0.5)

Steady rainfall of 9 inches over

4 days results in slope failure

movie
 Conclusions
1. A coupled analysis evaluates the transient response for the given infiltration rate.

2. The saturation is seen to increase toward a steady value consistent with the higher
magnitude of the rainfall event.

3. The increase in saturation near the slope surface causes a reduction in soil cohesion
and failure of the slope.
 Thermal Analysis in FLAC
FLAC models transient heat transfer in materials, such as soil
and rock, including the development of thermally induced
displacements and stresses.

The modeling of heat transfer may be done:

- by itself, independent of the usual mechanical calculation of
FLAC
- in parallel with the mechanical modeling, so as to capture the
effects of heat transfer / solid interaction.
 Thermal Option
The thermal option includes both conduction and advection
models.

Conduction – Transient transfer of heat based on Fourier’s

law of heat conduction.

Advection – Transient transfer of heat by convection in

porous media, by:
forced convection – heat carried by fluid motion,
and
free convection – fluid motion caused by fluid
density difference due to temperature variation.
 Mathematical Formulation for Conduction

 T
Energy-Balance Equation: −  q + q =
T T

t
v

where qT heat-flux vector [W / m 2 ]

qvT 3
volumetric heat source intensity [W / m ]
T stored heat per unit volume [ J / m3 ]

Thermal constitutive law relates temperature changes to the heat storage ( T ),

so the energy-balance equation can be rewritten as:

T
−   qT + qvT = Cv (1)
t
 Mathematical Formulation for Conduction

Heat Transport Law: qT = −k T T (2)

(Fourier’s Law)
where T temperature [ oC ]
T
thermal conductivity tensor [W / m oC ]
k ij

Substitution of Eq.(2) in to Eq.(1) yields differential equation for heat

conduction.
 Mathematical Formulation for Advection

T
Energy-Balance Equation: −   qT + qvT = cT +  0 cwq w  T (1)
t

where T heat-flux vector [W / m 2 ]

q
qvT volumetric heat source intensity [W / m3 ]
qw fluid specific discharge [ J / m3 ]

 0 , cw fluid density, fluid specific heat

c T effective specific heat cT =  0Cv + nS 0cw

 Mathematical Formulation for Advection

Heat Transport Law: qT = −k T T (k T = k sT + nSk wT ) (2)

(Fourier’s Law)
where T temperature [ oC ]
T
effective thermal conductivity (isotropic) [W / m oC ]
k
k sT , k wT solid and fluid thermal conductivities

Fluid Transport Law: q w = −k( P −  wg  x) (3)

(Darcy Law)
Where k fluid mobility coefficient
w fluid density ρw = ρ0 (1 − b f (T − T0 )
g gravity vector
bf volumetric thermal expansion of the fluid
Substitution of Eq.(2) and Eq.(3) into Eq.(1) gives differential
equation for heat advection.
 Thermal-Mechanical-Pore Pressure Coupling

Thermal-Mechanical Coupling Thermal-Fluid Coupling

• Strain rate due to thermal • Thermal expansion of the fluid

change is subtracted from total and solid induces pore pressure
change which affect effective
strain rate. stress
• Thermal strain change due to
P   T
temperature change has the = M ( − +b )
form: t t t t
 ijT
T
= t  ij  - variation of fluid content
t t
M - Biot modulus
• One way coupling: thermal  - Biot coefficient
changes affect mechanical but
b - Volumetric strain
not vice versa
 Example: Constitutive equations for elastic material

no water with water

 ij  ij 2   ij P  2 
= 2G ( ) +( K − G )  ( kk ) ij +  ij = 2G( ij ) +( K − G)  ( kk ) ij
t t 3 t t t t 3 t

no water, thermal with water, thermal

 ij  ij T  ij P  T
= 2G ( − t
 ij ) + +  ij = 2G ( ij −  t  ij ) +
t t t t t t t
2  T  T
( K − G )  ( kk − 3 t ) ij 2
( K − G )  ( kk − 3 t ) ij
3 t t 3 t t

P   P   T
= M( − ) = M ( − +b )
t t t t t t t
 Thermal Modeling Approaches

FLAC can perform thermal-only and coupled thermal calculations:

1. Thermal conduction-only calculation

2. Thermal advection-conduction calculation

3. Thermal-mechanical calculation

4. Thermal-mechanical-pore pressure coupling

CONFIG thermal command is given in all cases to assign

extra memory for the thermal calculation.
 Thermal Modeling in FLAC (1)

1. Thermal conduction-only calculation

- select SET mech off (to turn off mechanical calculations)

- select thermal model and properties
- set thermal initial and boundary conditions
- choose explicit or implicit solution algorithm (if implicit, select
thermal time step)
- for steady state solution, select SOLVE
- use SOLVE age command to stop calculation at a given thermal
time
 Thermal Modeling in FLAC (2)

2. Thermal advection-conduction calculation

- select advection thermal model (MODEL th_ac)

- forced convection can be modeled with or without the fluid configuration.
- if CONFIG gw is not selected, assign thermal properties with
econduct, espec_heat, f_qx, f_qy, f_t0, lspec_heat, thexp
- free convection can only be modeled with the fluid configuration (CONFIG gw)
- if CONFIG gw is selected, assign thermal properties with
conductivity, f_t0, f_thexp, lconduct, lspec_heat, spec_heat, thexp
- WATER density must also be set in both cases
- fluid and thermal times must be synchronized. Select one time step as an
integer multiple of the other making sure each time step is smaller or equal
to its default value.
- simulation should be run in an alternating sequence of thermal and flow
steps-one step for the module with the larger time step and enough steps of
the other module to close the time gap.
 Thermal Modeling in FLAC (3)
3. Thermal-mechanical calculation

- mechanical effects are assumed to occur instantaneously compared

to diffusion effects
- one directional coupling – temperature changes cause thermal strains
which influence the stresses. Thermal calculation is not affected by
mechanical changes.
- if system is elastic use SET therm on mech off and perform thermal
calculations untill the desired time. Then perform mechanical only
(SET therm off mech on) to find mechanical equilibrium
- if plasticity is involved, alternate between the thermal-only and mechanical-
only mode to account for path-dependency of the solution. Alternatively,
turn both (mechanical and thermal calculations) on and step or solve. By
default for each thermal step one mechanical step is taken. This can be
modified manually, or use CONFIG gw and SET flow off so that
SOLVE auto on age command can be used. This adjusts mechanical
cycles automatically to keep maximum unbalanced force ratio below a
specified value.
 Thermal Modeling in FLAC (4)

4. Thermal-mechanical-pore pressure coupling calculation

- grid must be configured for both thermal and groundwater options

(CONFIG thermal gw)
- coupling is induced by the influence of temperature changes on
volumetric change of the fluid and the grains in a saturated matrix
(assigned by PROPERTY f_thexp and PROPERTY g_thexp)
- thermal and groundwater calculations can not be performed at the
same time. Use SET thermal on flow off and SET flow on thermal
off and alternate calculations.
 Thermal Constitutive models

• Conduction models:
- anisotropic heat conduction: MODEL th_anisotropic
- isotropic heat conduction: MODEL th_isotropic
- general isotropic heat conduction: MODEL th_general
(thermal conductivity of the form
k (T ) = k1 + k2T n
- null model: MODEL th_null
(excavated material and insulators)

• Advection model:
- isotropic advection-conduction: MODEL th_ac
(activates advection logic)
 Common Thermal Boundaries

Adiabatic (insulated) Boundary

- default condition
- temperature free to vary

Fixed Temperature
- temperature fixed to specified value (FIX temp value)

Applied Thermal Boundary

- convective boundary (APPLY convection)
- heat flux boundary (APPLY flux)
- radiation boundary (APPLY radiation)

Interior Heat Source

- hest-generating source (INTERIOR source)
 Training Schedule
August 11, 2006 (morning)

09:00-10:00 Simulation of Deep Underground Isolation of Nuclear

Waste
- Setting up and running a thermal-mechanical analysis of a
waste repository drift
10:00-10:15 Break
10:15-12:00 Simulation of Deep Underground Isolation of Nuclear
Waste (continued)
 Training Schedule
August 11, 2006 (afternoon)

01:00-03:00 Overview of Waste Isolation Problems Analyzed by Itasca

Codes
- Review of Several modeling studies performed using Itasca
codes: UDEC, 3DEC, FLAC3D, PFC
03:00-03:15 Break
03:15-05:00 Overview of Waste Isolation Problems Analyzed by
Itasca Codes (continued)
General Question/Answer Session