Rock Engineering

Practice & Design

Lecture 9:
Deformation Analysis
and Elasto-
Elasto-Plastic Yield

1 of 40 Erik Eberhardt – UBC Geological Engineering ISRM Edition

Author’s Note:
The lecture slides provided here are taken from the course
“Geotechnical Engineering Practice”, which is part of the 4th year
Geological Engineering program at the University of British Columbia
(V
(Vancouver, Canada).
C d ) The
Th course covers rock k engineering
i i andd
geotechnical design methodologies, building on those already taken
by the students covering Introductory Rock Mechanics and Advanced
Rock Mechanics.
Mechanics
Although the slides have been modified in part to add context, they
of course are missing the detailed narrative that accompanies any
l
lecture. It is also
l recognizedd that
h these
h lectures
l summarize,
reproduce and build on the work of others for which gratitude is
extended. Where possible, efforts have been made to acknowledge
th vvarious
the ri us ssources,
urc s with
ith a list of
f references
r f r nc s being
b in provided
pr vid d att the
th
end of each lecture.

Errors, omissions, comments, etc., can be forwarded to the

author at: erik@eos.ubc.ca

2 of 40 Erik Eberhardt – UBC Geological Engineering ISRM Edition

Analysis in Geotechnical Engineering

LIMIT EQUILIBRIUM
(infinite slope,
method of slices,
slices etc.)
etc )

DISCONTINUUM
(distinct element,
element etc
etc.))
CONTINUUM
(finite element,
finite difference, etc.)

3 of 40 Erik Eberhardt – UBC Geological Engineering ISRM Edition

Numerical Modelling
Numerical methods of stress and deformation analysis fall into two
categories:

incl. boundary-element method

Integral only problem boundary is defined &
Methods discretized
Pro: more computationally efficient
Con: restricted to elastic analyses

incl. finite-element/-difference &

distinct-element methods
Differential
D fferent al problem domain is defined &
Methods discretized
Pro: non-linear & heterogeneous
material properties accommodated
C
Con: l
longer solution
l ti run times
ti

4 of 40 Erik Eberhardt – UBC Geological Engineering ISRM Edition

Numerical Analysis – Differential Methods
Differential
ff l methods
h d are more difficult
d ff l and d time consuming than
h boundary
b d
analyses (BEM), both in terms of model preparation and solution run times. As
such, they require special expertise if they are to be carried out
successfully.
… finite-difference method
continuum
… finite-element method

… discrete-element method
discontinuum
… distinct-element method

New Considerations (relative to Boundary Methods):

• Division of problem domain (i.e. meshing efficiency & element types).
• Selection of appropriate
pp p constitutive models (i.e. stress-strain response
p of
elements to applied forces).
• Determination of material properties for selected constitutive models (generally
derived from lab testing with scaling to field conditions).
• Limiting boundary conditions and special loading conditions.

5 of 40 Erik Eberhardt – UBC Geological Engineering ISRM Edition

Numerical Analysis – Differential Methods
Continuum Methods Discontinuum Methods
 Rock/soil mass behaviour  Rock mass represented as a
represented
p as a continuum. assemblage
g of distinct interactingg
 Procedure exploits approximations to blocks or bodies.
the connectivity of elements, and  Blocks are subdivided into a
continuity of displacements and deformable finite-difference mesh
stresses between elements.
elements which follows linear or non
non-linear
linear
stress-strain laws.

Stead et al. (2006)

6 of 40 Erik Eberhardt – UBC Geological Engineering ISRM Edition

Numerical Analysis – Continuum Methods

Commercial Software:
FLAC (Itasca) - http://www.itascacg.com/
Phase2 (Rocscience) - http
http://www.rocscience.com/
//www.rocscience.com/
DIANA (TNO) - http://www.tnodiana.com/
ELFEN (Rockfield Software Ltd.) - http://www.rockfield.co.uk/
VISAGE (VIPS Ltd.) - http://vips.co.uk/
PLAXIS (PLAXIS BV) - http://www.plaxis.nl/
SVS lid (S
SVSolid (Soil
il Vision
Vi i S Systems
t m Ltd
Ltd.)) - http://www.soilvision.com/
http:// il i i m/
ANSYS (ANSYS, Inc.) - http://www.ansys.com/

7 of 40 Erik Eberhardt – UBC Geological Engineering ISRM Edition

Continuum Analysis – The Basics
Continuum methods act to divide the soil/rock continuum into a
set of simple sub-domains called “elements”. These elements
can be of any y geometric
g shape
p that allows computation
p of the
solution (or its approximation), or provides the necessary
relation to the values of the solution at selected points, called
“nodes”.

This technique allows:

… accurate representation
p of complex
p
geometries and inclusion of dissimilar
materials.
… accurate representation
p of the
solution within each element, to bring
out local effects (e.g. stress or
strain concentrations).

8 of 40 Erik Eberhardt – UBC Geological Engineering ISRM Edition

Numerical Analysis – Continuum Methods
In geotechnical
I t h i l engineering,
i i th
there are two
t key
k continuum-based
ti b d
differential approaches used (to find an approximate solution to a set of
partial differential equations):
Fi it Diff
Finite-Difference Fi it El
Finite-Element
t
- method is an approximation to the - method is an approximation to the
differential equation solution of the differential equation
- solves a problem on a set of points - solves a problem on the interiors of
that form a grid the grid cells (elements) and for the
grid points
- easier to implement, but
approximation between grid points can - can more easily handle complex
be problematic. geometries

FLAC – Fast Lagrangian Analysis of Continua (by Itasca)

finite difference

Phase2 ((by
y RocScience))
finite element
9 of 40 Erik Eberhardt – UBC Geological Engineering ISRM Edition
Difference Between Differential Methods
Point-wise
P i t i A Approximation:
i ti In
I the
th finite-difference
fi it diff method,
th d the
th problem
bl
domain is represented by an array of gridpoints for which every
derivative in the set of governing equations is replaced by an algebraic
expression
p written in terms of
f the field
f variables ((e.g.
g stress or
displacement); these variables are undefined within the elements.

10 of 40 Erik Eberhardt – UBC Geological Engineering ISRM Edition

Steps in a FEM Solution

Division of the problem domain into parts

(both to represent the geometry
as well as the solution of the problem)

Seek
S k an approximate
i solution
l i forf each h part
(using a linear combination of nodal values
and approximation functions)

Assemble the p parts and solve for the whole

(by deriving the algebraic relations among the nodal values
of the solution over each part)

11 of 40 Erik Eberhardt – UBC Geological Engineering ISRM Edition

Basic Formulation of FEM Equations
FEM
E does not solve l for a single
l element,
l it is assembled
l and solved
l as
a whole (FDM, on the other hand, sweeps through a mesh and solves
implicitly, element by element).

Compatibility Material Behaviour Equilibrium

matrix that relates matrix off materiall

the strains inside the behaviour or the
constitutive matrix interpolation of
element with the
for the element displacements
nodal displacements
across element
(i.e. through shape
 = B·ae  = D·B·ae functions matrix)

strains stress solved for

nodal strain
displacements

12 of 40 Erik Eberhardt – UBC Geological Engineering ISRM Edition

Basic Formulation of FEM Equations
Piece-wise
Pi i approximation:
i ti Th finite-element
The fi it l t method
th d has
h a central
t l
requirement that the field quantities (stress, displacement) vary
throughout each element in a prescribed fashion using specific functions.
As such,, the problem
p m domain
m is represented
p byy an array
y of
f small,
m ,
interconnected subregions.

Compatibility Material Behaviour Equilibrium

ković (1999)
Pottts & Zdravk
13 of 40 Erik Eberhardt – UBC Geological Engineering ISRM Edition
Finite-
Finite -Element Element Types

Element dimensions are 1D

for lines, 2D for areas, and
3D for volumes. The number
of nodes between vertices
per edge are none for
linear, one for quadratic,
q
and two for cubic elements.

u ffirst-order (constant
element strain)

x
u

(linear second-order
strain) element x

14 of 40 Erik Eberhardt – UBC Geological Engineering ISRM Edition

Finite-
Finite -Element Matrix Assembly

Potts & Zdravković (1999)

15 of 40 Erik Eberhardt – UBC Geological Engineering ISRM Edition

 Numerical Problem Solving
3. Choose constitutive
model & material
2. Mesh properties
y = 600 m

1 B
1. Build
ild geometry
t

x = 800 m

4 D
4. Define
f boundary
b d &
initial conditions

6. Visualize & interpret

5. Compute

16 of 40 Erik Eberhardt – UBC Geological Engineering ISRM Edition

Problem Solving: 2-
2-D or 3-
3-D?
Many geotechnical problems can be assumed to be plane strain
(2-D assumption) without significant loss of accuracy of the
solution.

… in plane strain, one

dimension must be considerably
longer than the other two;
… strains along the out-of-
plane
l direction
di ti can be b assumed
d
to be zero;
… as such, we onlyy have to
solve for strains in one 2-D
plane.
Potts & Zdravković (1999)

17 of 40 Erik Eberhardt – UBC Geological Engineering ISRM Edition

Problem Solving: Meshing
The intention of grid generation is to fit the model grid to the physical
domain under study. When deciding on the geometric extent of the grid
and the number of elements to specify, the following two aspects must be
considered:
1. How will the location of the grid boundaries influence model results?

2. What density of zoning is required for an accurate solution in the

region of interest?

Do’s and Don’ts

- The density of elements should be highest
in regions of high stress or strain gradients.
- The greatest accuracy is achieved when
the element’s aspect ratio is near unity;
anything above 10:1 is potentially inaccurate
(5:1 for FDM).
- The ratio between adjacent elements
should not exceed 4:1 (using a smooth
transition when
h zoning from
f a fine
f to coarse
mesh).

18 of 40 Erik Eberhardt – UBC Geological Engineering ISRM Edition

Material Properties
A key advantage of differential methods over integral methods is
that by discretizing the problem domain, the assignment of varying
material properties throughout a heterogeneous rock mass is
permitted.
d

Material properties
p p
required by the chosen
constitutive stress-strain
relationship are generally Mat 3
derived from laboratory (
(pre-historic
hi i
testing programs. gneissic
slide block)
Laboratory values should
be extrapolated
p to closely
y Material 1 Material 2
(weathered Matt 4
M
correlate with the actual (gneiss)
(glacial soils)
gneiss)
in situ conditions.
… multiple material zonation used in a model of
an unstable slope in the Örztal
Ö Alps, Austria.

19 of 40 Erik Eberhardt – UBC Geological Engineering ISRM Edition

Problem Solving: Constitutive Models
During deformation, solid materials undergoes irreversible strains
relating to slips at grain/crack boundaries and the opening/closing
of pore space/cracks through particle movements. Constitutive
relations act to describe, in terms of phenomenological laws, the
stress-strain behaviour of these particles in terms of a collective
behaviour within a continuum.

   

   
elastic rigid – perfectly elastic - perfectly elastic - plastic
plastic plastic (strain
hardening/softening)

20 of 40 Erik Eberhardt – UBC Geological Engineering ISRM Edition

Constitutive Models
“Most fundamental ideas of science are essentially simple and may,
as a rule, be expressed in a language comprehensible to everyone”.
Einstein
… the more complex the constitutive model, the more the number of
input parameters it requires and the harder it gets to determine these
parameters without extensive
extensive, high quality (and of course
course, expensive)
laboratory testing;

… as such, one should always begin by using the simplest model that can
represent the
h key
k behaviour
b h of
f the
h problem,
bl andd increase the
h
complexity as required.

“Everything should be made as simple as possible…

but not simpler”.
Ei t i
Einstein

21 of 40 Erik Eberhardt – UBC Geological Engineering ISRM Edition

General Statement of Constitutive Laws

A general mathematical function for constitutive equations of

laws can be expressed as:
. .
f (,,,) = 0

… where
h  denotes
d n t s st ss  denotes
stress, d n t s (infinit
(infinitesimal)
sim l) st
strain,
in and
nd
the overdot denotes the rate of change with respect to time.

For equations that are homogeneous in time

time, i.e.
i e the stresses
and strains at constant external loads do not change with time,
the constitutive laws can be expressed in matrix notation as:

{d}= [C]·{d} {d}= [D]·{d}

[the stress-strain form] [the strain-stress or inverse form]

22 of 40 Erik Eberhardt – UBC Geological Engineering ISRM Edition

Constitutive Models for Geomaterials

 Linear elastic (isotropic) 

 Linear elastic (anisotropic)

 Viscoelastic
- influence of rate of deformation E

 Elastic-perfectly plastic 
- von Mises
- Drucker-Prager
g
- Mohr-Coulomb

 Elasto-plastic
- strain
t i h hardening
d i
- strain softening

23 of 40 Erik Eberhardt – UBC Geological Engineering ISRM Edition

The Elastic Compliance Matrix - Isotropy

Isotropy assumptions used for rock:

H d
Hudson & Harrison
H i (1997)

24 of 40 Erik Eberhardt – UBC Geological Engineering ISRM Edition

Constitutive Models for Geomaterials

 Linear elastic (isotropic)

 Linear elastic (anisotropic)

 Viscoelastic
- influence of rate of deformation

 Elastic-perfectly plastic
- von Mises
- Drucker-Prager
g
- Mohr-Coulomb

 Elasto-plastic
- strain
t i h hardening
d i
- strain softening

25 of 40 Erik Eberhardt – UBC Geological Engineering ISRM Edition

Constitutive Models for Geomaterials

 Linear elastic (isotropic)

 Linear elastic (anisotropic)

 Viscoelastic
- influence of rate of deformation

 Elastic-perfectly plastic
- von Mises
- Drucker-Prager
g
- Mohr-Coulomb

 Elasto-plastic
- strain
t i h hardening
d i
- strain softening

26 of 40 Erik Eberhardt – UBC Geological Engineering ISRM Edition

Plasticity: An Introduction
Elastic
El ti materials
t i l have
h a unique
i stress-strain
t t i relationship
l ti hi given
i b
by
the generalized Hooke’s law. For many materials, the overall
stress-strain response is not unique. Many states of strains can
correspond to one state of stress and vice-versa.
vice versa Such materials
are called inelastic or plastic.

… when load is increased, material

behaves elastically up to point B,
and regains its original state upon
unloading.
… if the material is stressed
beyond point B up to C, and then
unloaded, there will be some
permanent or irrecoverable
deformations in the body, and the
material is said to have undergone
plastic deformations.

27 of 40 Erik Eberhardt – UBC Geological Engineering ISRM Edition

Plasticity & Yield
Solution
- Initial stress state.
When run elastically, yield and/or failure Step 1
within the model are not
considered/enabled. For a slope stability
analysis this state is typically only applied
analysis,
as an intermediary step to initialize the in
situ stress conditions in the slope.

The elastic
Th l ti model
d l may then
th be
b changed
h d to
t
an elasto-plastic one to determine yielding
Solution
within the slope (and the resulting - Disequilibrium condition.
Step 2
displacements that arise). All plastic
models
d l potentially
t ti ll iinvolve
l some degree
d of
f
permanent, path-dependent deformations.
Once an element has reached it’s yield
state, further increases in stress must be
supported
t d b
by neighbouring
i hb i elements
l t ((exceptt
in the case of strain hardening), which in
turn may yield, setting off a chain reaction
leading to localization and catastrophic
f il
failure.

28 of 40 Erik Eberhardt – UBC Geological Engineering ISRM Edition

 Shear Strength Reduction

2nd strength
reduction

1stt “ l
“failure”
”

Displacement
strength
reduction
Slope)
ca – FLAC\S
(Itasc

Calculation steps (time)

The shear
Th h strength
h is reduced
d d untill collapse
ll
Strength reduction: occurs, from which a factor of safety is
produced by comparing the estimated shear
cmob=c/F φmob = φ/F strength of the material to the
reduced/increased
d d/i d shear
h strength
h at failure.
f il

29 of 40 Erik Eberhardt – UBC Geological Engineering ISRM Edition

Understanding Shear Strength Reduction
mesh
dependency

Eberhardt (2008)

modelling
“failure”

30 of 40 Erik Eberhardt – UBC Geological Engineering ISRM Edition

Case History: Cause of Fatal Landslide
Lutzenberg (2002) – 3 deaths

slide surface

Although this slide occurred during

a period of heavy rain, the
investigation uncovered the presence
of a broken water pipe suspiciously
located in the head of the slide.
This led to questions of whether

04)
the pipe ruptured during “aa

ey et al. (200
rainfall-triggered landslide”, or
whether non-critical slope
movements caused the pipe to
rupture and the leaking pipe was

Valle
responsible for the landslide.

31 of 40 Erik Eberhardt – UBC Geological Engineering ISRM Edition

 Case History: Cause of Fatal Landslide

In-situ mapping and

testing for shear
strength properties.
properties
04)
ey et al. (200
Valle

32 of 40 Erik Eberhardt – UBC Geological Engineering ISRM Edition

 Case History: Coupled H
H--M FEM Analysis

Dry/Unsaturated Slope

Model agrees with

fi ld observations
field b i

no

Saturated Slope
p

Model agrees with

field observations
Failure due to heavy
precipitation event likely;
no yes
no needd for
f leaking
l ki water t
pipe to induce failure.

Failure due to
heavy precipitation
alone unlikely; explore
?
alternative
lt ti scenario.
i

Valley et al. (2004)

33 of 40 Erik Eberhardt – UBC Geological Engineering ISRM Edition

Case History: Coupled H
H--M FEM Analysis

Dry/Unsaturated Slope

Model agrees with

fi ld observations
field b i leaking water pipe

no

Saturated Slope
p

Model agrees with

field observations

no

Leaking Water Pipe

Model agrees
g with
field observations
Valley et al. (2004)
no yes

34 of 40 Erik Eberhardt – UBC Geological Engineering ISRM Edition

Case History: Coupled H
H--M FEM Analysis

Dry/Unsaturated Slope

Model agrees with

fi ld observations
field b i

no

Saturated Slope
p

Model agrees with

field observations

no

Leaking Water Pipe

Valley et al. (2004)

Model agrees
g with
field observations
Leaking water pipe, in
combination with heavy
yes
precipitation event,
cause of failure.

35 of 40 Erik Eberhardt – UBC Geological Engineering ISRM Edition

Good Modelling Practice
The modelling of geomechanical processes involves special considerations
and a design philosophy different from that in other fields of applied
mechanics. This is because situations in earth materials often involve
limited amounts of input data.
data
As such, the model should never be considered as a “black box” that
accepts data input at one end and produces a prediction at the other. The
model should instead be prepared carefully and tested several times in
progression of increasing difficulty to gain a full understanding of the
problem.

Step 1 - Define the objectives of the model analysis.

analysis
In order to Step 2 - Create a conceptual picture of the physical system.
perform a
Step 3 - Construct and run idealized models.
successful
numerical Step 4 - Assemble problem-specific data.
study, several Step 5 - Prepare a series of detailed runs.
steps are
Step 6 - Perform the model calculations.
recommended:
Step 7 - Present results for interpretation.

36 of 40 Erik Eberhardt – UBC Geological Engineering ISRM Edition

Input & Assumptions
The fundamental requirement for a meaningful modelling study
should include the following steps of data collection/evaluation:
– site characterization (geological and hydrogeological conditions);
– groundwater conditions (pore pressure model);
– g
geotechnical parameters
p m (strength,
( g , deformability,
f m y, permeability);
p m y);
– instability mechanisms (kinematics or potential failure modes).

“if you do not know what you are looking for,

you are not likely to find much of value”

R. Glossop, 8th Rankine Lecture, 1968

37 of 40 Erik Eberhardt – UBC Geological Engineering ISRM Edition

Good Modelling Practice

“numerical modelling should not be used as a

substitute for thinking,
g but as an aid to thought
g ”

… results of a survey
of nine commonly
usedd geotechnical
t h i l
modelling programs
and their response to
impossible (e.g. E<0)
and
d iimplausible
l ibl
(Esoil>Erock) input
data.

C ill (1993)
Crilly

38 of 40 Erik Eberhardt – UBC Geological Engineering ISRM Edition

Good Modelling Practice

Definition of Problem

E t bli h Controlling
Establish C nt llin Failure
F il Mechanism
M h ni m

Choice of Appropriate Analysis Method

Definition of Input Parameters

Initial Analysis

D
Detailed
l d Analysis
l

Rigorous Validation

39 of 40 Erik Eberhardt – UBC Geological Engineering ISRM Edition

Lecture References
Crilly, M (1993). Report on the BGS meeting ‘Validation of Geotechnical software for design’. Ground
Engineering 26(9): 19-23.
Eberhardt, E (2008). Twenty-Ninth Canadian Geotechnical Colloquium: The role of advanced
numerical methods and geotechnical field measurements in understanding complex deep-seated rock
slope failure mechanisms. Canadian Geotechnical Journal 45(4): 484-510.
Glossop, R (1968). The rise of geotechnology and its influence on engineering practice. Géotechnique
18: 105-150.
Hudson, JA & Harrison,
Hudson Harrison JP (1997).
(1997) Engineering Rock Mechanics – An Introduction to the Principles .
Elsevier Science: Oxford.
Potts, DM & Zdravković, L (1999). Finite Element Analysis in Geotechnical Engineering: Theory.
Thomas Telford: London.
Stead, D, Eberhardt, E & Coggan, JS (2006). Developments in the characterization of complex
rock slope deformation and failure using numerical modelling techniques. Engineering Geology 83(1-3):
217-235.
y, B,, Thuro,, K,, Eberhardt,, E & Raetzo,, H ((2004).
Valley, ) Geological
g and g
geotechnical investigation
g
of a shallow translational slide along a weathered rock/soil contact for the purpose of model
development and hazard assessment. In Proceedings of the 9th International Symposium on
Landslides, Rio de Janeiro. A.A. Balkema: Leiden, pp. 385-391.

40 of 40 Erik Eberhardt – UBC Geological Engineering ISRM Edition