Oct 04, 2015
Numerical Methods of GeotechnicsNumerical Methods of Geotechnics
Prof. Minna KarstunenUniversity of Strathclyde
Tentative schedule: Jan 16-20, 2012
Introduction to numerical modelling & finite elements
Linear elasticity
Basic concepts of plasticity and Mohr Coulomb model
Non linear finite elements and solution techniques
Applied theory: Introduction to PLAXIS
T01 Tutorial: Soil testing tool, Mohr-Coulomb
Applied theory: Shallow foundations
Applied theory: Structural elements and interfaces
T02 Tutorial: Shallow foundation
Drained/undrained analysis
Consolidation analysis
Applied theory: Soil parameters for drained and undrained analysis
Applied theory: Slope stability and phi-c reduction
T03 Tutorial: Consolidation and phi-c reduction
Tentative schedule: Jan 16-20, 2012
Critical state models
Applied theory: Soil parameter for critical state models
T04 Tutorial: Analysis of an embankment (inc. stability)
Applied theory: Analysis of an embankment
T05 Tutorial: Boston embankment (I) Hardening Soil Model and Small Strain Stiffness
Applied theory: Soil parameters for Hardening Soil model
Applied theory: Excavations
Tentative schedule May 28-May 21, 2012
T06 Tutorial: Excavation in Limburg
Anisotropy, bonding and creep
Applied theory: Numerical modelling of ground improvement
T07 Tutorial: Boston embankment (II)
Assessment:
Coursework (100%) : Independent numerical analysis Part 1A: Identify a research paper with suitable numerical analysis
Date of submission April 2, 2012
Part 1B: Numerical analysis and report Part 1B: Numerical analysis and report
Date of submission June 15, 2012
Muir Wood, D. Geotechnical Modelling. Spon Press, 2004.
Potts, D. & Zdravkovic L. Finite element analysis in geotechnical engineering-Theory. Thomas Telford,1999.
Potts, D. & Zdravkovic L. Finite element analysis in geotechnical engineering-Application. Thomas Telford,1999.
Potts, D., Axelsson K., Grande, L. Schweiger, H. & Long, M. Guidelines for the use
Recommended reading:
Potts, D., Axelsson K., Grande, L. Schweiger, H. & Long, M. Guidelines for the use of advanced numerical analysis. Thomas Telford, 2002.
Azizi, F. Applied analysis in geotechnics. E & F. Spon, 2000.
Muir Wood, D. Soil behaviour and critical state soil mechanics. Cambridge University Press,1990.
Zienkiewicz & Taylor. The Finite Element Method available through various publishers
PLUS selected research papers available
Main aims of the module Give a comprehensive understanding of the role of soil
modelling and numerical analysis in practical geotechnical context.
The focus is on: The selection of appropriate soil model considering a particular
application and information available, application and information available, Interpretation of values for soil parameters for numerical analysis, Idealisation and modelling of geotechnical problems with 2D finite
element code PLAXIS Appreciation on the limitations of finite element modelling.
At the end of the course you will be competent (but not expert) on finite element modelling, and its opportunities and limitations, in geotechnical context.
Introduction to numerical modelling and finite elementsmodelling and finite elements
Real problem Idealized problem (conceptual model)
Mathematical
Introduction
Linear Momentum Balance
Strain displacement equation
Terzaghis principle
Continuity equation
ij j ib 0, + =
( )ij j i i j1 u u2 , , = += +
pTv
=
Real problem Idealized problem (conceptual model)
Mathematical
280
320
360
400
440m a.s.l.
411+100 412+000 413+000
North portal(Lleida)
Rele
vant
phenom
ena
Mathematical Model (PDE)Results
Solution
Analytical Numerical
Continuity equation
Mechanical constitutive law
Darcys law
vt
=
= D
v k H= T
f
H r=
g
280
320
360
400
440m a.s.l.
411+100 412+000 413+000
North portal(Lleida)
Quaternary Middle Eocene Early EoceneColluvion Limestone Claystone & Siltstone
Marl Anhydritic-Gypsiferous Claystone
Mathematical Model (PDE)
Solution
Analytical Numerical
ResultsQuaternary Middle Eocene Early Eocene
Colluvion Limestone Claystone & Siltstone
Marl Anhydritic-Gypsiferous Claystone
Introduction
A rigorous solution must satisfy the following Equilibrium Compatibility Compatibility Stress-strain relationship Boundary conditions
Empirical methods
Analytical methods
Numerical methods:
Introduction
The idealized problem can be solved by different methods:
Outside the scope of this course
Best alternative, but it is generally quite difficult for actual problems to obtain an analytical solution
Finite Difference Method (FDM) Finite Element Method (FEM) Boundary Element Method (BEM) Discrete Element Method (DEM)
Equations in integral form
Equations in differential form
Methods of analysis (1)
Closed form Constitutive behaviour linear elastic
Simple Limit equilibrium (e.g. compatibility not satisfied, soil rigid
with a failure criterion) Stress field (e.g. compatibility not satisfied, soil rigid with a
failure criterion) Limit analyses (e.g. lower bound does not satisfy
compatibility, upper bound not equilibrium, both assume soil ideally plastic with associated flow rule)
Settlement calculations (immediate, consolidation, creep) stresses from elastic theory, principle of superposition
Settlement calculations
Both assume elastic stresses & principle of superposition
The bearing capacity theory was first introduced for plane straincondition by Terzaghi and then developed by Meyerhof (1953), Vesic(1979) and Brinch-Hansen (1968) for more general conditions. The ultimate bearing capacity of a shallow foundation is expressedas:
1q B N D N c N' = + +
Bearing capacity
f f d cq B N D N c N2' = + +
where N, Nd and Nc are the so-called bearing capacity factors.
Formula is for general bearing capacity failure, so it does notconsider local failure or punching failure
Assumes principle of superposition, so in some cases error 25%(but on the safe side) Soil assumed isotropic and homogeneous
Errors increase when foundations subjected to moment loading.
Bearing capacity
Meyerhof (1953) and Vesic (1979) account for that by considering it tobe caused by a vertical load with a particular eccentricity e, and theeffective footing area.
Methods of analysis (2)
Numerical Beam-spring approaches
Investigation of soil-structure interaction Soil modelled by springs or elastic interaction factors, so Soil modelled by springs or elastic interaction factors, so
need to make gross assumptions Only a single structure can be analysed Solutions include forces and movement of the structure,
but do not provide information on global stability or movements of the adjacent soil
Often neglect shear stresses
Methods of analysis (3)
Numerical FEM- Finite Element Method (PLAXIS, SIGMA-W,
SAFE, ABAQUS, DIANA etc.) Simulation of the BVP attempting to satisfy all
theoretical considerations Accuracy depends on the ability of constitutive model to
represent real soil behaviour and the correctness of the idealization
User defines geometry, construction procedure, soil parameters and boundary conditions
Effect of time on pore pressure development can be modelled, including coupled consolidation
Methods of analysis (3)
Numerical FDM- Finite Difference Method (FLAC)
Same as above, but equations in differential form Solution explicit, so results heavily dependent on step-
size No matrices are formed, so e.g. not possible to check
for convergence (you get a results but have no way of assessing its accuracy)
Finite Difference MethodFLAC Program
u&
v&
Typical FLAC Model
Typical FLAC Model
Slope at Failure
Methods of analysis (3)
Numerical BEM- Boundary Element Method (mainly research
codes) Only surfaces are meshed, so less unknowns than in Only surfaces are meshed, so less unknowns than in
FEM/FDM Mathematically more complex than FEM, but
computationally more efficient Great for linear elastic materials Not possible to use advanced soil models
Methods of analysis (3)
Numerical DEM- Discrete (or distinct) Element Method
(PFC2D, PFC3D, UDEC, 3DEC) Modelling of blocks or granular assemblies by Modelling of blocks or granular assemblies by
considering individual particles/clusters Particle contact properties need to be calibrated, as they
cannot be directly measured
Discrete Methods
Discrete Methods
Discrete Methods
Notation and FEM fundamentalsNotation and FEM fundamentals
Problem
PDE1f
2f
solutionexact PDE3f
Introduction
Solution
Function
Approximation
Function (approximation) Cordao-Neto
Approximation
Numerical methods are approximationsHow good are numerical models or approximations?
Introduction
It is difficult to distinguish difference between the circle and the approximation if N = 16.
N=5 N=8 N=16
Cordao-Neto
Vectors, Matrices & Notation (see also separate handout)
Stress and strain vectors
y
x
'
'
'
Tzxyzxyzyx
zx
yz
xy
z ]'''[''
=
=
Tzxyzxyzyx ddddddd ][ =
Vectors, Matrices & Notation
Stress-strain relationship
dDd ]['=
whereT
zxyzxyzyx ddddddd }'''{' =
and [D] is the elastic or elasto-plastic matrix (6 x 6 matrix in 3D).
Vectors, Matrices & Notation
Differential matrix for 3D continuum:
00
00
00
y
x
continuum:
=
0
0
0
00
xz
xz
yz
z
Vectors, Matrices & Notation
Nabla operator is a vector defined as:
=
xIt is an operator because it assumes a meaning onlywhen applied to a matrix or vector, such as:
=
z
ywhen applied to a matrix or vector, such as:
z
w
yv
x
u
w
v
u
zyxvT
+
+
=
=
Vectors, Matrices & Notation
Gradient:
The gradient of a scalar function f(x,y,z) is a vector whose components are the partial derivatives of f with
= fx
f
fpartial derivatives of f with respect to x, y, and z
=
z
fyff
Soils are multiphase materials whichare commonly assumed to have twophases:
General Formulation (3D)
pores1. Solid particles
The resolution of equilibrium equations involves the static equilibrium of soilparticles, subjected to self weight, external loads and buoyancy due to thepresence of wetting fluid.
The simultaneous fulfilment of static and hydraulic equations is commonlyreferred to as fully coupled problem.
grains1. Solid particles 2. Water (saturated soils) or air (dry
soils)
soil
H Balance Equations (global)
Constitutive Equations (related to the behaviour of the materials)
General Formulation (3D)
x
y
zdx
dy
dz
the materials)
General Formulation (3D)
Equation of motion (which reduces to Newtons second law if the quasi-static problem is approached (no time derivatives) are considered) :
Linear Momentum Balance
0=+ b where b is a vector of body forces
zyyzzxxzyxxy
zzzyzx
yyzyyx
xxzxyx
bzyx
bzyx
bzyx
===
=+
+
+
=+
+
+
=+
+
+
;;
0
0
0
General Formulation (3D) Small deformation formulation (1)
Assumed that all elongations, distortions and the rates of change of displacements are small. All higher order terms are ignored and it is assumed that strains are linear function of displacements and their derivatives.
+
=
+
=
+
=
=
=
=
+
=
z
u
x
w
yw
z
v
x
v
yu
z
w
yv
x
u
x
u
x
u
zxyzxy
zyx
i
j
j
iij
21
;21
;21
;;
21
General Formulation (3D) Small deformation formulation (2)
We normally use the Cartesian notation:
wvu
z
u
x
w
yw
z
v
x
v
yu
z
w
yv
x
u
zxyzxyxy
zyx
+
=
+
=
+
==
=
=
=
;;2
;;
General Formulation (3D) Terzaghis principle
= + where is the vector (1,1,1,0,0,0)T and the pore water pressure.
Continuity equation
When a fluid is flowing through the soil mass, the mass conservation law has to be fulfilled. Such a conservation equation (the continuity equation) reads
pTvt
=
within the hypothesis of incompressible fluid and without sources or sinks. v is the flow velocity
is the volumetric strain, given by the trace of the strain tensor (=x+y+z). p
General Formulation (3D)
Mechanical problem
A further set of equations are the constitutive equations (or constitutivelaws) for the mechanical and hydraulic problem.
The mechanical constituve law can be simply considered as a law which correlates strain and effective stress tensor.
= D Hydraulic problem
The link between velocity of fluid flowing through the soil mass (v) and hydraulic head (H) is established via the Darcys equation, as a function of soil permeability k:
v k H= Note that Bernoullis equation defines the hydraulic head (neglecting the dynamic term).
General Formulation (3D)
Equation Name Mathematical form NEquationsN
Unknowns
Linear Momentum Balance 3 6 ()
Strain displacement equation 6 3 (u)( )ij j i i j1 u u2 , , = +
0=+ b
Terzaghis principle 6 6 ()
Continuity equation 1 1 ()
Mechanical constitutive law 6 6 ()
Darcys law 3 + 3 (v)
Total 25 25
= + pTvt
=
= D v k H=
What we do?
System the Partial Differential
equations (PDEs)+
Boundary Conditions
System of Equations in Integral form
System of algebraic equations
Finite Element Method
We solve this systemDomain + Boundary
Different techniques can be used to obtain the system of algebraic equations:
Weighted Residual Method
Variational Method
Virtual Work Principle (very common in structural problems)
systemDomain + Boundary
Typical 2D meshElements connected together in nodes
Primary variables (displacements) are calculated in nodes
Footing width = B
Node
Gauss point
Typical 2D mesh
Footing width = B
Node
Displacements interpolated within element to calculate strains at Gauss points
Constitutive law is used to relate strains Node
Gauss point
used to relate strains to stresses
Forces acting at nodes are calculated (inc. body forces and surface tractions)
b: body forcesDifferential Equation of Equilibrium
=+
+
=+
+
0
0
yyxy
x
xyx
byx
byx
x 0
2D FEM - Mechanical Problem
x
y
xy
=
0
0
x y
y x
=
TLLLL
T + b = 0LLLL
=
+
00
0
0
y
x
xy
y
x
bb
x
y
y
x
=+
+
=+
+
0
0
yyxy
x
xyx
byx
byx
( ), ux yv
= =
u u
u
V
x
y
2D FEM - Mechanical Problem
Displacements
u
v
= =
u
3
i i 1 1 2 2 3 3i 1
u u N u N u N u N=
= = + +3
= = + +e
3
i i 1 1 2 2 3 3i 1
v v N v N v N v N=
= = + +
1
1
3 21 2
3 21 2
3
3
e e e
u
v
N uN NuN vN Nv
u
v
= = = =
u N a
xux
=
yx
y
=x
v
yu
xy
+
= u
u
V
x
y
2D FEM - Mechanical Problem
Deformations (1)
0x
(Deformation Vector)
(DisplacementVector ) Note. The unknowns of theproblems are the nodal
displacements
e{0 ex
e e
y
xy
x
u u
v vy
y x
= = = =
B
N aL L LL L LL L LL L L
{3 x 2 2 x 6 6 x 1
3 x 6e
e e
B
N a123LLLL
uu
V
x
y
2D FEM - Mechanical Problem
Deformations (2)
{
0
0e
x
e e
y
xy
x
u u
v vy
y x
= = =
B
N aL LL LL LL L
B is the matrix that relates nodal displacements to strains
ey x
31 2
31 2
1 1 2 2 3 3
00 0
0 0 0e e
NN Nxx x
NN Ny y y
N N N N N Ny x y x y x
= =
N BLLLL
e e B a
Constitutive laws expresses the relationship between stress () and the strain () vectors in a matrix equation of the form:
= D
where D is often referred to as the material stiffness matrix.For the case of linear plane strain isotropic elasticity, the matrix is:
2D FEM - Mechanical Problem
Stressese e
B a Deformations Mechanical law Stresses
For the case of linear plane strain isotropic elasticity, the matrix is:
D = +
Ev v
v v
v vv( )( )1 2 1
1 01 0
0 0 1 22
where E is the Young's modulus and is the Poisson's ratio. Because the coefficients are constants (for the case where geometry changes are assumed to be small) the resulting FE equations are linear and can be solved in a one step.
2D FEM - Mechanical Problem
Stresses and element matrix
= LLLL
= D ;= D LLLL e e e= N a
{e e e e e
e
= =
B D N a DB aLLLL
1
2
3
fy1
fx1
fx3
fx2
fy2
fy2
( )T T Ti i iN d N d N d = + b tL
e e e
e e
T e T Ti i
e e e
d N d N d = +
=
K f
B D B a b t
K a f
14424 43 14 4 4424 4 4 43
6 x 3 3 x 3 3 x 6 6 x 6
eT e=B D B K
Equation of equilibrium applied to the element as a whole
2D FEM - Mechanical Problem
e e e=K a f
1
2
3
fy1
fx1
fx3
fx2
fy2
fy2
11
11
22
22
33
33
;
x
y
xe e
y
x
y
fufvfufvfufv
= =
a f
Local matrix (element)
Body forces and surface tractions applied to the element may be generalized into a set of forces acting at the nodes.
Footing width = B
Node
Gauss point
=
Assembly of the global matrix
2D FEM- Solution
The global stiffness matrix generally contains many terms that are zero. If node numbering is efficient these are clustered in a band along the diagonal. In this case it is necessary to store only element within the bandwidth saves considerably on amount of storage.
Ns s
0
0
N
s s
0
2D FEM- Solution
PLAXIS uses skyline procedure and some of the zero terms within the band are not stored.
Skyline profileBand profile
To determine the nodal displacements corresponding to the applied forces it is necessary to solve the global stiffness equation ( KU = P ) subject to the appropriate boundary conditions.
FEM- Solution
Forces are related to displacement via stiffness equations which are solved within the FE code to find within the FE code to find values of nodal displacements
Elements for 2D analysis
(a) Triangular elements
(b) Lagrange elements
A polynomial interpolation function is used to describe the displacements within each element (higher order polynomial yields more accurate results)
Elements for 3D analysis
(a) Tetrahedron (b) Hexahedron (c) Pentahedron
Some practical aspectsSome practical aspects
FE meshes for geotechnical analysis
Typically the aspect ratio of the elements should be smaller than 3
Typically for triangular elements in internal angles should be 15 in internal angles should be 15 < < 165
Fine mesh for area with high gradients
FE meshes for geotechnical analysis
Summary of FE method FEM is a computational procedure that may be used to obtain
approximate solutions to mathematical problems
Governing mathematical equations (generally continuous) are approximated by a series of algebraic equations involving approximated by a series of algebraic equations involving quantities that are evaluated at discrete points within the region of interest
Linear elasticityLinear elasticity
Constitutive equations
Relate generally stress and strain, i.e. not just in 1D or for a special stress path
A variety of constitutive equations available and important to adopt a model that is most appropriate for the particular analysis to be appropriate for the particular analysis to be carried out. Choice depends on:
material (clay, sand) information available (in situ & lab. tests) type of problem (embankment, excavation..) knowledge of the user
Constitutive equations (cont.)
The simplest: linear elastic model (Hookes law)a
xx b yy
xx
xx2b
In the case of a uniaxial tensile stress applied to an elastic bar, the strains produced in two perpendicular directions are related by the expression:
where is an elastic constant known as Poisson's ratio.
2a
vx
y=
Constitutive equations (cont.)In 3D the stress-strain equations take the form:
yy
)(1 zyxx vvE =
xx
zz
)(1)(1
yxzz
zxyy
vvE
vvE
E
=
=
Constitutive equations (cont.)
If a shear strain in applied to an elastic material, a shear strain is produced.
xy =
yx
xy
Gxy
xy
=
yx
xy xy
where G is the shear modulus.
Constitutive equations (cont.)
Four elastic parameters are commonly used : Young's modulus, E Shear modulus, G, Poisson's ratio Bulk modulus, K
An elastic material is fully specified, however, whenvalues of two of these parameters are given.
)21(3)1(2 vEK
v
EG
=
+=
2D elastic analysis
x
z
zz = 0
P
To carry out FE analysis of 2D problems, it is necessary to specify the condition in the third dimension. The plane straincondition is most commonly used in soil mechanics.
yP
y
x
2D elastic analysis (cont.)
In plane strain condition the out-of plane strain is set to zero and Hookes law gives:
x
y
( )( )
xyxy
xyy
yxx
G
vvvv
E
vvvv
E
=
++
=
++
=
)1()1)(21(
)1()1)(21(
Drained and Undrained Analysis
u= =0
Dissipation of u with time
u=0=
1. Undrained 2. Consolidation
3. Drained
Drained and Undrained Analysis (cont.)(a) The shear modulus is identical for drained and undrained loading.
(b) The drained and undrained Young's moduli are related by the expression:
G G Gu = =
Note that for most soils the value of generally lies in the range 0.3 to 0.35 for sands and 0.2-0.3 for clays. Youngs modulus values, however, may vary substantially between different materials and stress levels.
uEvE )1(32
+=
y x
y
xyyx
yz
zy
Isotropic elasticity in 3D
)'''''(1)'''''(
'
1
)'''''('
1
vv
vvE
vvE
zxyy
zyxx
=
=
=
x
yz
x
z
xz zx
zy
'
'
'
)'''''('
G
G
G
vvE
zxzx
yzyz
xyxy
yxzz
=
=
=
=
y x
y
xyyx
yz
zy
Cross-isotropic elasticity (around y-axis)
'
'
'
'
'
'
'
'
'
'
'
'
'
'
'
'
'
'
'
'
'
zhhvh
z
v
hh
v
yx
v
hhy
z
v
vhy
v
hh
h
xx
vv
Ev
EEv
Ev
Ev
E
+=
+=
=
Sampling direction
vh
x
yz
x
z
xz zx
zy
'
'
'
'
'
'
'
'
hh
zxzx
vh
yzyz
vh
xyxy
h
zy
v
hhx
h
vhz
G
G
G
EEE
=
=
=
+=
General 3D elasticity would require the specification of 21 elastic constants!
Needs 5 elastic
constants!
Oedometer test
Soil
Load
yVertical
strainyy
E'
oed'/1 Emv =
1
yy
Primarycompression
E'oed,ur
1
E'oed
Unload/reloadpath
Lateral strains are zero, therefore measure modulus is not Youngs modulus. Based on Hookes law:
=
+ E v E
v voed
( )( )( )
11 2 1
Triaxial test (see also Muir Wood 1990)
E'ur
Deviatorstress
q qf
E'501 1
Axial strain, a
qf / 2
50
Shearing with constant cell pressure
Secant or Tangent E?
Es
'1
Et1
'1
Youngs modulus E
1
Es1
1
Poissons ratio 5.0'1
Non-Linear Elasticity
Non linear elasticity
')1(')1('' pededp
eddpK +=+==
1
ln(p')eBulk modulus K
)1('
dee
dK
v
=+== 1
')'1(2)'21(3
' KG
+
=
Shear modulus G
5.0'1
Elasticity vs. Plasticity
In elasticity, there is a one-to-one relationship between stress and strain. Such a relationship may be linear or non-linear. An essential feature is that the application and removal of a stress leaves the material in its original condition
Elasticity vs. Plasticity
for elastic materials, the mechanism of deformation depends on the stress increment
for plastic materials which are yielding, the mechanism of (plastic) deformation depends on the stress
reversible = elastic irreversible = plasticreversible = elastic irreversible = plastic
Next lecture will look at the Basic Concepts of Plasticity and Mohr Coulomb model