You are on page 1of 357

Soilmodelling

DavidMuirWood
UniversityofDundee,Scotland
SouthEastAsia
OctoberNovember2010

1. Introductiontomodelling:soilbehaviour(SBCSSM1,GM1,2)
2. Elasticmodelling(SBCSSM2,GM3)
3. Themostwidelyusedsoilmodel:MohrCoulomb(GM3)
4. ApplicationofMohrCoulombmodel(exercise)(GM7)
5. Camclay(SBCSSM5,GM3)
6. Camclaygraphicalcalculations(exercise)(SBCSSM5)
7. Camclay:complianceformulation(GM3)
8. Stiffnessformulation:MohrCoulomb,Camclay(GM3)
9. Selectionofsoilparameters(exercise)(GM3)
10. MohrCoulombimproved:strength,criticalstates,particlebreakage
11. Camclayimproved:nonlinearity,structure
12. Conclusion

CUP (1990)

Spon (2004)

CUP (2009)

SBCSSM: Soil behaviour and critical state soil mechanics.


Cambridge University Press 1990: ISBN 0-521-33782-8
GM: Geotechnical modelling. Spon Press 2004: ISBN 0-41923730-5
1D: Soil mechanics: a one-dimensional introduction.
Cambridge University Press 2009: ISBN 978-0-521-74132-3


1.Introductiontomodelling:
soilbehaviour
(SBCSSM1,GM1,2)

Soilmodelling:SouthEastAsia:OctoberNovember2010

1. Introductiontomodelling:soilbehaviour
DavidMuirWood
d.muirwood@dundee.ac.uk

Models:
Scientific understanding proceeds by way of constructing and
analysing models of the segments or aspects of reality under study.
The purpose of these models is not to give a mirror image of reality,
not to include all its elements in their exact sizes and proportions, but
rather to single out and make available for intensive investigation
those elements which are decisive. We abstract from non-essentials,
we blot out the unimportant to get an unobstructed view of the
important, we magnify in order to improve the range and accuracy of
our observation. A model is, and must be, unrealistic in the sense in
which the word is most commonly used. Nevertheless, and in a
sense, paradoxically, if it is a good model it provides the key to
understanding reality.
(Baran and Sweezy, 1968)

Everything engineers (and scientists)


do is concerned with modelling!

You have been engaged in modelling


for years perhaps unconsciously!

examples of models in civil and


geotechnical engineering

behaviour of mild steel


in uniaxial tension

idealisation of behaviour of mild steel


forming basis for plastic design of steel
structures
structural engineering

geological model

particles modelled as equivalent spheres


sieving sedimentation

classification model - particle sizes

theoretical model

useful for confirming results of other modelling


if boundary conditions are somewhat similar

Bjerrum vane correction factor


effects of rate, anisotropy,

empirical model

numerical model
model of dock with
free field boundaries
behind walls
dock structures under seismic loading

model of dock with adjacent docks included


Mair and Muir Wood (2001)

numerical modelling
for example, finite element, finite difference
equilibrium

compatibility of deformations

stresses

strains
stress:strain relationship
constitutive model

lectures primarily concerned with introduction to various aspects


and possibilities of constitutive modelling

shear
stress

shearstrain

observed and idealised shearing behaviour of soil


for settlement and bearing capacity calculations
constitutive models trying to reproduce more of the
actual nonlinearity of pre-failure soil response

stress and strain variables (1)


concentrate on axisymmetric conditions of triaxial test
volume changes are important in soils
and affect mechanical properties
choose volumetric strain increment:

p = a + 2 r
axial strain
increment

radial strain
increment

stress and strain variables (2)


in developing constitutive models concept of work will be important
need to link strain increment and stress variables
work conjugate pairs
choose volumetric stress variable: effective mean stress

p' = ('a +2'r ) / 3


volumetric work work done in changing size is:

Wp = p' p

stress and strain variables (3)


choice of second pair of variables to
describe change of shape somewhat
arbitrary

for convenience choose deviator stress


from triaxial test
distortional stress variable: deviator stress

q = 'a ' r = F / A

area A

stress and strain variables (4)


work conjugate pair for change in shape
distortional strain increment:

q = 2( a r ) / 3
distortional work work done in changing shape is:

Wq = q q
q = a for constant volume deformation p = 0

stress and strain variables (5)


confirm that total work done in strain increment is:
W = 'a a + 2'r r = p' p + q q = Wp + Wq

convenient transformation matrices


p' 1 / 3 2 / 3 'a

=
1 'r
q 1

p 1
2 a

q = 2 / 3 2 / 3
r

'a 1 2 / 3 p'
=

'r 1 1 / 3 q

1 p
a 1 / 3

r 1 / 3 1 / 2 q

stress and strain variables (6)


define stress ratio: = q/p'
mobilised friction

' a ' r
3
=
sin 'm =
' a + ' r 6 +
6 sin 'm
=
3 sin 'm
for conditions of triaxial compression

distortion: change in shape: q, q

distortion: change in shape: q, q

compression: change in size: p', p

Pore pressure parameter


total and effective stresses
undrained deformation

total stress is external perturbation arbitrary


effective stress is soil response constrained by constitutive
properties p' indicates desire of soil to change in volume
when sheared (or not) dilatancy

Pore pressure parameter


pore pressure increment: u = p p'
slope of effective stress path: a = p'/q
then: u = p + aq
logical definition of pore pressure parameter linking pore pressure
change with total stress change (compare Skempton 1, 3)
a is not a soil property function of history etc.
subtraction of arbitrary total stress change p helps in interpretation of
pore pressure change

Soil behaviour
particle continuum duality
laboratory element testing
stiffness

Particle-continuum duality

for analysis we need to treat soil as a continuum (stress, strain)


but its properties emerge from its particulate nature

particle-continuum
duality

Leighton Buzzard sand


(picture width 37 mm)
Leda clay (picture width 13m)

particle-continuum
duality

photoelastic discs
force chains
fabric
Drescher and de Josselin de Jong (1972)

particle-continuum duality

Discrete Element Modelling


of particle interactions
homogenisation procedure required to produce continuum quantities
stress
strain

Laboratory element testing

general stress state has 6 degrees of freedom

we may calibrate models using triaxial test data but soil elements in
the ground and in numerical modelling will certainly break away from
axial symmetry
beware of unintended responses in uncalibrated regions of stress space

Laboratory element testing

axial symmetry
triaxial apparatus
centreline of circular loaded area
widely available source of data of stress:strain response

Laboratory element testing


plane strain occurs more frequently
no strain in long direction

plane strain with fixed principal axes


fairly easy to apply in laboratory

Laboratory element testing

true triaxial apparatus


3 degrees of freedom
no rotation of principal axes

Laboratory element testing


Simple shear apparatus
plane strain with rotation of principal axes
no horizontal strain

development of shear stress on ends of sample?


questions of uniformity of stress state/strain state

Laboratory element testing


direct shear box
inhomogeneous deformation
strength information
pedagogic illustration of dilatancy
use for subsequent modelling

Laboratory element testing


torsional hollow cylinder apparatus
4 degrees of freedom
but radial non-uniformity
average stress quantities

with equal internal/external pressures


b = (2-3)/(1-3) = sin2
constraint on stress space exploration

Stiffness

secant stiffness Gs = / tells us where we are now but gives no


indication of how we got there
tangent stiffness Gt = / tells us about current incremental
response
beware: use of term stiffness does not imply elasticity

Stiffness
progressive yielding of steel cantilever as analytical illustration
of distinction between tangent and secant stiffness

elastic-plastic model for steel

plastic region penetrates from


edges towards centre of beam

Stiffness: steel cantilever

tip displacement log scale

tangent stiffness falls as yielding spreads


tangent stiffness zero when full plastic hinge has formed
secant stiffness remains positive and non-zero

Stiffness: steel cantilever


load line

deflection line
stiffness good compliance bad!
load line limited by collapse load of cantilever
deflection line unlimited (ignore geometry changes)
many to one mapping we can always map from deflection
to load but not always from load to deflection

stiffness

Stiffness
secant stiffness

variation of stiffness is due to plasticity


hence need for constitutive models
Quiou sand (LoPresti et al., 1997)

Summary
modelling is all around us
care in selection of strain increment and stress variables
pore pressure parameter as a variable
particle-continuum duality
laboratory element testing not just axial symmetry
but more general stress states not easily accessible
tangent and secant stiffness
stiffness/compliance formulation?


2.Elasticmodelling
(SBCSSM2,GM3)

Soilmodelling:SouthEastAsia:OctoberNovember2010

2. Elasticmodelling
DavidMuirWood
d.muirwood@dundee.ac.uk

Elasticity
Hooke's law
elastic behaviour in triaxial compression:
drained/undrained
measurement of elastic properties with different devices
elastic anisotropy
elastic nonlinearity hyperelasticity
worked example

IfIhaveseenfurtheritisonlybystandingontheshouldersofgiants.(Newton)
HookewasNewtonspredecessorattheRoyalSocietyofLondon.
Hookewasaverysmallman.

Hookes law
A description of helioscopes and some other instruments. London 1676.
Tofillthevacancyoftheensuingpage,Ihavehereaddedadecimateofthecentesmeof
theInventionsIintendtopublish,thoughpossiblynotinthesameorder,butasIcan
getopportunityandleasure;mostofwhich,Ihope,willbeasusefultoMankindasthey
areyetunknownandnew.
2.ThetrueMathematicalandMechanichalformofallmannerofArchesforBuilding,
withthetruebutmentnecessarytoeachofthem.AProblemwhichnoArchitectonick
Writerhatheveryetattempted,muchlessperformed.
abcccddeeeeefggiiiiiiiillmmmmnnnnnooprrsssttttttuuuuuuuux.

3.ThetrueTheoryofElasticityorSpringiness,andaparticularExplicationthereofin
severalSubjectsinwhichitistobefound:Andthewayofcomputingthevelocityof
Bodiesmovedbythem.
ceiiinosssttuu

Hookes law
Tofillthevacancyoftheensuingpage,Ihavehereaddedadecimateofthecentesmeof
theInventionsIintendtopublish,thoughpossiblynotinthesameorder,butasIcan
getopportunityandleasure;mostofwhich,Ihope,willbeasusefultoMankindasthey
areyetunknownandnew.
2.ThetrueMathematicalandMechanichalformofallmannerofArchesforBuilding,
withthetruebutmentnecessarytoeachofthem.AProblemwhichnoArchitectonick
Writerhatheveryetattempted,muchlessperformed.
abcccddeeeeefggiiiiiiiillmmmmnnnnnooprrsssttttttuuuuuuuux.
Utpendetcontinuumflexile,sicstabitcontiguumrigiduminversum.
(Ashangstheflexiblechain,so,inverted,standstherigidarch.)

Utpendetcontinuumflexile,sic
stabitcontiguumrigiduminversum.
(Ashangstheflexiblechain,so,
inverted,standstherigidarch.)

StPeters:Poleni:1748

Hookes law
Tofillthevacancyoftheensuingpage,Ihavehereaddedadecimateofthecentesmeof
theInventionsIintendtopublish,thoughpossiblynotinthesameorder,butasIcan
getopportunityandleasure;mostofwhich,Ihope,willbeasusefultoMankindasthey
areyetunknownandnew.
3.ThetrueTheoryofElasticityorSpringiness,andaparticularExplicationthereofin
severalSubjectsinwhichitistobefound:Andthewayofcomputingthevelocityof
Bodiesmovedbythem.
ceiiinosssttuu
Uttensiosicvis.(Astheextensionsotheforce.)
stress
force

extension
strain

Hookes law
principal stresses and strain increments:
x
1 'x

y = 1 ' y
E 1 '
z

symmetric

x
' x
1

E
y
' y =
1
' (1 + )(1 2 )

1
z

Hookes law

uniaxial tension
Youngs modulus E = (P/A)/(l/l)
Poissons ratio = (d/d)/(l/l)
direct observation of elastic constants

Hookes law
axial symmetry:
a 1 1 2 'a

r E 1 'r
'a
1 2 a
E

1 r
'r (1 + )(1 2 )

compliance/stiffness matrices not symmetric because stress


and strain variables not work conjugate

Hookes law
separate change of size and change of shape
work conjugate stress and strain variables
p 1 / K
0 p'

q = 0 1 / 3G q

bulk modulus:

E
K=
3(1 2 )

p' K 0 p

q 0 3G q

shear modulus:

E
G=
2(1 + )

change of size and change of shape uncoupled


(off-diagonal zeros)

Hookes law
only 2 independent elastic properties for isotropic material

9KG
E=
G + 3K

3K 2G
=
2(G + 3K )

Stiffness in conventional drained triaxial compression


total stress path in conventional compression
(constant cell pressure):

p = ( a + 2 r ) / 3
if

r = 0 then

q = a r
q
=3
p

p 1 / K
0 p'
=


q
0 1 / 3G q

p' K 0 p

q 0 3G q

Stiffness in drained triaxial compression


q
= 3G
q
q
=E
a
p

G
=
q K
p
a

= 1 2

Stiffness in undrained triaxial compression

no coupling between change of size and change of shape


compression and distortion for isotropic elastic soil
hence, in undrained test pure distortion p'=0 and pore
pressure parameter a = 0 for isotropic elastic soil
conventional triaxial compression q/p = 3 and u = p = q/3
pore pressure from total mean stress alone
effective stress path independent of total stress path

Stiffness in undrained triaxial compression


in undrained test

q = a

hence

q
q
=
= 3G
q a

external view of undrained stiffness in terms of total stress increments:

0 p
p 1 / K u

=

q
1 / 3G u q

0
distortional stress q = a r = 'a 'r and hence Gu = G

Stiffness in undrained triaxial compression


0 p
p 1 / K u


q = 0
1 / 3G u q


p = 0 (undrained)
p arbitrary (external change of total stress)
hence Ku =

Eu
1
= u =
Ku =
3(1 2 u )
2

Stiffness in undrained triaxial compression


0 p
p 1 / K u


q = 0
1 / 3G u q


Eu
E
Gu =
=G=
E u = 3G
2(1 + )
2(1 + u )
u = 1/2
drained and undrained elastic properties
cannot be chosen independently

Elastic stiffness from different devices

oedometer
one dimensional compression

E oed

(
'z
1 )
=
=E
= K + 43 G
(1 + )(1 2 )
z

and Ko = /(1-)

Elastic stiffness from different devices

pressuremeter
direct measurement of shear modulus G

Elastic stiffness from different devices

plate loading test

4G
=
R (1 )

composite stiffnesses do not reveal individual elastic properties

elasticity
simple models assume isotropic linear elasticity
two parameters: E, or G, K
convenient but not necessary
explore possibility of anisotropy
and nonlinearity

anisotropic elasticity
many soils deposited over areas of large lateral extent
implied symmetry: all horizontal directions equivalent:
B, C, D, E
cross anisotropy transverse isotropy
(anisotropy of real soils, after real stress/strain histories
will be more complex)

different modes of shearing


cross anisotropy

xx 1 / E h


yy hh / E h
/ E
zz = vh v
yz
0

0
zx

0
xy

hh / E h
1/ Eh
vh / E v

vh / E v
vh / E v
1/ Ev

0
0
0

0
0
0

0
0
0

0
0
0

1 / G vh
0
0
1 / G vh
0
0

5 elastic properties
triaxial can only find 3: Ev, vh, Eh/(1-hh)
hh? Gvh? could use bender elements

xx

yy

zz
0
yz

0
zx
2(1 + hh ) / E h xy
0
0
0

xx
1/ 2

yy
* / 2

*/
1
zz

yz E *
0

0
zx

xy

* / 2
1/ 2
*/
0

0
*/
0
*/
1
0
0
2(1 + *) /

0
0

0
0

0
0

xx

yy

zz
yz

2(1 + *) /
0
zx
0
2(1 + *) / 2 xy
0
0
0
0

0
0
0
0

Graham & Houlsby 3 parameter cross anisotropy


axisymmetric triaxial conditions p
1 * +4 * +2 2
K* = E *
9(1 + *)(1 2 *)
2 2 * 4 * +
G* = E *
6(1 + *)(1 2 *)
1 * + * 2
J = E*
3(1 + *)(1 2 *)

3G * J p'
1

= 3K * G * J 2 J K * q

coupling of compression
and distortion

xx
1/ 2

yy
* / 2

*/
1
zz

yz E *
0

0
zx

xy

* / 2
1/ 2
*/
0
0
0

0
*/
0
*/
1
0
0
2(1 + *) /
0
0

0
0

xx

yy

zz
yz

2(1 + *) /
0
zx
0
2(1 + *) / 2 xy
0
0
0
0

Graham & Houlsby 3 parameter cross p' K *


=
anisotropy
axisymmetric triaxial conditions
forcing particular link between cross
anisotropic elastic parameters

q J

0
0
0
0

J p

3G * q

coupling of compression
and distortion

effective stress path in undrained


triaxial compression
2 = 1.52 = Eh/Ev
Winnipeg clay (Graham & Houlsby)

experimental techniques
laboratory geophysics
bender elements
shear wave velocities: time, distance
Vvh = Vhv ? (elastic, symmetry)

alternatively:
small undrained
unload-reload cycles
during drained test
slope of effective
stress path indicates
elastic anisotropy
vertical Eh = Ev, = 1
anisotropy evolves
with stress ratio
Eh 0 at critical state
Hostun sand (Gajo)

Gault clay
evidence of stiffness
anisotropy from bender
elements

Gault clay and Hostun sand


evolution of elastic
anisotropy with stress ratio

cross anisotropy axis of symmetry of anisotropy coincides


with vertical axis of sample
more general anisotropy misalignment of axes sample
distressed boundary measurements may be unrepresentative
it is tempting to assume that things that you choose not to
observe do not exist

empirical description of elastic nonlinearity or


evolution of anisotropy
hypoelasticity
not necessarily conservative energy
production/dissipation in closed cycles
for example, both G and K functions only of p'
hyperelasticity
thermodynamically consistent
define strain energy density function U
or complementary energy V
U
i =
;
i

V
hence stiffnesses
i =
i

complementary energy function (Boyce, 1980)


2

1
1 q
n +1

+
V = p'
(n + 1)K1 6G1 p'
n (1 n )(2 n ) 2
+

p
6G1

= p'n 1 K1

1

3G1

constant volume paths

3G1 p'

1
q
3G1

p 'o
K1 2
= 1 (1 n )

p'
6G1

constant distortional strain paths

q p 'o
=
q o p'

n 1

road sub-base material: n = 0.2, = 0.3 (K1/G1 = 2.17)

Boyce complementary energy function

contours of constant volumetric strain and


constant distortional strain

Cam clay has K dependent on p'


constant is thermodynamically unacceptable
Houlsby strain energy function
U = p'r e

p /

[ +

p'

= p'

3
2

q2


p

q

constant distortional strain paths: = constant


constant volume paths
asymptotic to

q 2 = 6p'i (p'p'i )

= 3 / 2

Houlsby strain energy function

contours of constant volumetric strain and


constant distortional strain

art in choosing energy functions to give desirable properties


when differentiated twice (stiffness or compliance)
varying slope of undrained path may come from nonlinearity
or evolving anisotropy (or both)
subtle tests to distinguish

even more difficult to discover hyperelastic energy functions


to describe evolving elastic anisotropy
nonlinear elasticity or plasticity? (irrecoverable strains)

Summary
generalised Hookes law one-to-one link between
stress and strain
elastic matrices symmetric with work-conjugate
variables
two independent elastic properties for isotropic
material
deduction of elastic properties from standard tests
evolving anisotropy - nonlinearity

Example: Plane strain isotropic elastic element


under general stress path
Take y direction as plane strain direction:
y = 0 = (y - x - z)/E
hence y = (x + z)

z z

x
x
y=0
y

y=0
y
x
x

z z

Example: Plane strain isotropic elastic element


under general stress path
y = (x + z)
General plane strain stress path: x = z
then y = (1 + ) z

x = z
y=0
y

y=0
y
x = z

Example: Plane strain isotropic elastic element


under general stress path
General plane strain stress path: x = z
strain in z direction z = (z - x - y)/E
z = (1 + )(1 - [1 + ]) z/E

z z

x = (1 + )([1 - ] - ) z/E

x = z
x

x = z
x
y=0

y=0

z z

Example: Plane strain isotropic elastic element


under general stress path
z = (1 + )(1 - [1 + ]) z/E
x = (1 + )([1 - ] - ) z/E
One-dimensional compression (oedometer): x = 0
= / [1 - ]
stiffness: Eoed = z/ z = E(1 - )/ (1 + ) (1 - 2)
For example: = 0.25, Ko = / [1 - ] = 1/3;
Eoed /E = 1.2

Exercise: Plane strain isotropic elastic element


under general stress path
Assume = 0.25;
1. Plot relationship between z stiffness z/ z E and
2. Plot relationship between x stiffness z/ x E and


3.Themostwidelyused
soilmodel:MohrCoulomb
(GM3)

Soilmodelling:SouthEastAsia:OctoberNovember2010

3. Themostwidelyusedsoilmodel:MohrCoulomb
DavidMuirWood
d.muirwood@dundee.ac.uk

Elastic-perfectly plastic Mohr-Coulomb model


Mohr-Coulomb familiar as a strength/failure model
limiting stress ratio characterised by frictional strength
convert to stress ratio M

'a '1 1 + sin '


=
=
'r '3 1 sin '

q 3('a 'r ) 6 sin '


M= =
=
p' 'a +2'r
3 sin '
combine with isotropic elastic stiffness model
aim: to produce elastic-plastic stiffness matrix
strain increment stress increment

Mohr-Coulomb failure; elastic strains


failure locus divides stress plane into
elastic and inaccessible regions
plasticity only occurs for stress states
on failure locus: q/p'=M
stress increments implying q/p'>M not
possible
assume strain increment divided into elastic and plastic components
= e + p
elastic strain accompanies any change in stress
e
e
0 p'
p' K 0 p p 1 / K

e or
=
=
e

q 0 3G q q 0 1 / 3G q

stiffness

compliance

What are the deformations at yield/failure?


work conjugate
strain increments
plotted on local axes
parallel to
corresponding stress
axes
assume that yielding mobilises a plastic mechanism
mechanism defines ratio (not magnitude) of plastic strain increments
pp
M *

= M * or
=
p
p

q
1
q
pp

where is an arbitrary scalar multiplier

Stiffness matrix: consistency condition


for plastic strains, stress
change lies along failure locus
Mp' + q = 0

e
e
p
p

p
'
K
0

p
'
K
p
'
0
K
0
K
K
0
0
K

p p p p p p p 0 M *
= = =p =e = e p

G q3G
q q3G
q
q
q 0q 3G 0
0 q 30G03

q 0q 3G 1

K 0 p

( M 1)

0 3G q
=
K 0 M *

( M 1)
0 3G 1

combine and rearrange

Stiffness matrix which stiffness? elastic/plastic?


stiffness formulation: calculate stress increments
from total strain increments
MM * K 2
p' K 0
1

q 0 3G KMM * +3G 3MGK

3M * GK p

9G
q

elastic predictor and plastic corrector


implementation:
if know
elasticwhether
prediction
produces
a stress
state
we do not initially
a strain
increment
will
be lying
in
inaccessible
elastic
or plasticregion beyond failure locus, plastic correction
required to bring stress state back to failure locus
if plastic: p'

M * p
1
3GK

q KMM * +3G M MM * q

Plastic dissipation? Associated flow?


if plastic:

M * p
p'
1
3GK

q KMM * +3G M MM * q
confirm that q = Mp'
asymmetric unless M = M* (associated flow)
plastic work Wp = p'pp + qqp =(M M*)p'qp
plastic work Wp = 0 for M = M* (associated flow)
physically unreasonable
(but basis of bearing capacity calculations etc)

Mohr-Coulomb model: typical response

typical response in constant p' test


dilation/compression depends on sign of M*

standard elastic-perfectly
plastic Mohr-Coulomb model
non-associated plastic flow
M* M

simplicity
sharp stiffness changes
tangent stiffness either
elastic or zero
continuing volume change

Mohr-Coulomb model: fitting to data?


standard model available in all
geotechnical finite element programs
subjectivity in choice of parameters:
stiffness G, K;
strength M;
dilatancy M*
exercise

Mohr-Coulomb model: system/element response

system response for example, footing shows progressive


yielding even though model for individual soil elements is
elastic-perfectly plastic
exercise response of two element box model

Summary
elastic-perfectly plastic Mohr-Coulomb model is most
widely available constitutive model for soils
associated plastic flow is not physically reasonable
choice of soil parameters requires engineering
judgement (exercise)

active: x = Ka z
z

x = z
z > x
x > z
passive: x = Kp z
x
Mohr-Coulomb failure

x = Ka z
z

x = z

stress path: x = z
x = Kp z
x
initial stress state: Ka < Ko < Kp
general stress path: x = z

(z constant)

x = Ka z

x = z

zo + z

stress path: x = z
x = Kp z

zo
Ko zo

active failure when Ko zo + z = Ka (zo + z)


z /zo = (Ko Ka)/(Ka - )

(for > Ka)

x = Ka z

x = z

stress path: x = z
x = Kp z

zo
Ko zo

passive failure when Ko zo + z = Kp (zo + z)


z /zo = (Kp Ko)/( - Kp)

(for > Kp)

x = Ka z

x = z

stress path: x = z
x = Kp z

zo
Ko zo

strains to failure: z = (1 + )(1 - [1 + ]) z/E


x = (1 + )([1 - ] - ) z/E

Exercise: Plane strain isotropic elastic Mohr-Coulomb


element under general stress path
Assume = 0.25, = 30
1. Plot relationship between normalised strain to (active or
passive) failure Ez/zo and initial state Ko and stress path
2. Plot relationship between normalised strain to (active or
passive) failure Ex/zo and initial state Ko and stress path
What happens if Kp > > Ka ?


4.Applicationof
MohrCoulombmodel
(exercise)(GM7)

Soil modelling: South East Asia: October-November 2010

4. Application of Mohr-Coulomb model (exercise)


David Muir Wood
d.muirwood@dundee.ac.uk

Mohr-Coulomb model: system/element response

system response for example, footing shows progressive


yielding even though model for individual soil elements is
elastic-perfectly plastic
exercise response of two element box model

numerical modelling

for example, finite element, finite difference


equilibrium

compatibility of deformations

stresses

strains

stress:strain relationship

constitutive model
lectures primarily concerned with introduction to various aspects
and possibilities of constitutive modelling

a.

v
h

h
h
v

two element box model


as analogue of footing

b.

single element

equilibrium
kinematic compatibility
stress:strain response

p. 1/1

active failure
v > h

single element

=0

=1

vertical
stress v
initial stress h = Kov
h = v
passive failure h > v
horizontal stress h
stress path characterised by = h /v

p. 2/1

single element

=0

= /(1) = 1

= /(1),
h

= 1,
h = v

v
vertical
stress v

= 0, h

= 0, v
= /(1),
v

a.
horizontal stress h

stress paths

b.
0 strain increment

elastic strains: effect of

p. 3/1

two-element model

Mohr-Coulomb
failure

=
A

vertical
stress v

=0

a.
h
P

b.
P

Mohr-Coulomb
failure

hA = -hP

hP

horizontal stress h

kinematic compatibility
equal & opposite horizontal strains in elements A, P

p. 4/1

kinematic compatibility
equal & opposite horizontal strains in elements A, P
element P: v = 0; equivalent to compression element
loaded horizontally P = 0
for element P, h /h = E/[(1 + )(1 )]
we can deduce the stress path direction A for element
A
A v /h = E/[(1 + )(1 )] =
A E/(1 + )[A (1 ) ]
A = /[2(1 )] (which is half the elastic Ko value)

p. 5/1

Mohr-Coulomb
failure

=
A

vertical
stress v

=0

a.
h
P

b.
P

Mohr-Coulomb
failure

hA = -hP

hP

horizontal stress h

equal & opposite horizontal strains in elements A, P


which element reaches failure first?

p. 6/1

v
A

element A
yields
element P
yields

P
h

element P reaches failure first


h remains constant
A for element A changes from A = /2(1 ) to
A = 0

p. 7/1

element P
yields
v

element A
yields

element A reaches failure first


A = h /v now changes and remains constant at
A = Ka

stress v can continue to increase until element P


reaches failure

p. 8/1

element P
yields
v

element A
yields

element A reaches failure first


compatibility of horizontal strains: compression of
element P matches extension of element A
hA = ehA + phA = ehP
vA = evA + pvA = evA phA

p. 9/1

Use = 0.25, = 30 , E/b = 200


Calculate Ka , Kp and slope of stress path
A = /[2(1 )]
Plot initial normalised stress states v /b = 1,
h /b = Ko for Ko = (1 + Ka )/2, 1 and (1 + Kp )/2
Plot stress paths for elements A and P for each initial
stress state. For each value of Ko discover which
element reaches failure first and calculate the
corresponding vertical strain in element A (normalised
with E/b )
Plot the stress paths followed as the second element
heads for failure
Calculate the normalised vertical strain in element A at
which the second element reaches failure
Plot the normalised stress:strain response of the
footing for each initial state

p. 10/1

Soil modelling: South East Asia: October-November 2010

4. Exercise: Two element box model


David Muir Wood (University of Bristol/Dundee)

Introduction
A two element box model (Fig 1a) provides a simple analogue of a foundation. We can use this simple
model to demonstrate the three constraints that have to be satisfied in any boundary value problem:
equilibrium, kinematic compatibility, and constitutive response. Equilibrium is straightforward: forces
must balance. Kinematic compatibility means that gaps should not open up in our model. Constitutive
response indicates the relationship between stress changes and strain changes for the material: we will
assume an isotropic elastic response up to Mohr-Coulomb failure. But first we will analyse the response
of a single element - like a laboratory test - and present the response graphically.
All stresses are effective stresses.
1. Single element: elastic response: plane strain
Plane strain testing in the laboratory is not particularly common but this is what we need for our single
element test (Fig 1b). The nearest equivalent to the standard triaxial test will be a plane strain test in
which the lateral stress is kept constant but we will look more generally at tests in which the vertical
stress and lateral stress are increased in constant proportion: h = v (Fig 2).
Hookes Law for the plane strain direction of zero strain tells us that:
1
[y (v + h )]
E

(1)

y = (v + h ) = (1 + )v

(2)

y = 0 =
so that:

Hookes Law for the vertical direction tells us that:


v =

1
1
[v (y + h )] = (1 + )(1 )v
E
E

(3)

and for the horizontal direction:


h =

1
1
[h (y + v )] = (1 + )( )v
E
E

(4)

Then for any value of not only can we plot the stress paths in the h : v effective stress plane, but we
can also plot the horizontal and vertical strain increments for the several paths (Fig 3b). Evidently = 0
corresponds to a test with constant lateral stress; = 1 is a sort of isotropic plane strain compression
test; = /(1 ) corresponds to one-dimensional compression h = 0.
Mohr-Coulomb failure implies a limiting ratio of vertical and horizontal effective stresses. We know the
values of this limiting ratio as the active and passive earth pressure coefficients - depending on whether
the vertical or the horizontal stress is the greater.
v > h :

v
1 + sin
= Kp =
h
1 sin

(5)

a.

v
h

h
h
v

b.

Figure 1: (a) Two element box model as analogue of footing; (b) single plane strain element
active failure
v > h

=0

=1

vertical
stress v
initial stress h = Kov
h = v
passive failure h > v
horizontal stress h

Figure 2: Single element: stress paths


v < h :

v
1 sin
= Ka =
h
1 + sin

(6)

These limiting ratios can be plotted on the effective stress plane (Fig 2).
The elastic response is (assumed to be) independent of the initial stress state. We can characterise the
initial stress state by the coefficient of earth pressure at rest Ko (where Ka Ko Kp and hence deduce
the dependency of the vertical (or horizontal) strain at failure on the value of Ko and on the direction
of the stress path .
2. Two element box model: elastic response, Mohr-Coulomb failure
The two element model (Fig 1a) provides a simple analogue of the bearing capacity problem. There is an
initial surcharge stress b on the surface of both elements. The initial horizontal stress in each element
(which must from equilibrium be identical) can be chosen as h = Ko b where Ka Ko Kp .
The right-hand passive element P (Fig 1a) is just a single element being loaded horizontally with P = 0,
so we can plot (we have already plotted) the stress path for this element (but with vertical and horizontal
stresses interchanged) and indicate the strains in the horizontal direction of passive loading (which was
the vertical direction in the previous section) (Fig 4a, b).
The left-hand active element A (Fig 1a) is being loaded vertically by the footing. Equilibrium requires
that the horizontal stress in element A should always be the same as the horizontal stress in element P.
We can indicate this constraint in the stress plane (Fig 4a). Consideration of the nature of the problem

=0

= /(1) = 1

= /(1),
h

= 1,
h = v

v
vertical
stress v

= 0, h

= 0, v
= /(1),
v

a.
horizontal stress h

b.
0 strain increment

Figure 3: Single element: (a) stress paths and (b) stress:strain response
shows that the horizontal stress must increase as the footing load is increased so that A > 0.
Kinematic compatibility requires that no gap should open up between the two elements. The horizontal
compressive strain in element P resulting from the increase of horizontal stress must be equal to the
horizontal tensile strain in element A resulting from the increase of vertical stress. These two constraints
together allow us to deduce the relevant value of A for the loading of element A and hence the value of the
vertical strain in element A and the vertical stress: vertical strain response of our analogue foundation.
The logic is illustrated in Fig 4b. The value of A for the active element is unknown but it can be
calculated from (3) and (4), recalling that element P is being loaded horizontally. Equating appropriate
strains and noting that h /v = A we find that:

A =
(7)
2(1 )
which is exactly half the value for one-dimensional compression: for h /v = 0, h /v = /(1 ).
We are free to choose our initial value of Ko : its value will control which element reaches failure first.
Suppose the right-hand passive element reaches failure first so that h = Kp b (Fig 5a). The horizontal
stress cannot increase any further so that from this moment onwards the left-hand active element is
being loaded with A = 0 - and we can again use the stress path and strain information from the
previous section to deduce the overall footing response. The horizontal strain will continue to increase in
both elements. In the passive element, which has reached failure, there will be no further elastic strains
and the vertical and horizontal strains will be linked by the angle of dilation . Let us, for simplicity,
assume that the failure occurs at constant volume. Then vertical and horizontal strains will be equal
and opposite: the horizontal squeezing of the element will translate directly into a vertical stretching.
The three sections of the vertical stress:vertical strain response for element A are shown schematically
in Fig 6a, b.
Or suppose that the left-hand active element reaches failure first (Fig 5b). The ratio of stresses on this
element must then remain constant but the magnitude of the vertical stress will continue to increase
until the passive element also reaches failure. The stress path followed by the passive element does not
change direction so the continuing loading of the footing can be followed by applying the equilibrium
constraint of equal horizontal stresses in the two elements (Fig 5b). The active element will experience
both elastic strains (along a path with A = Ka ) and plastic strains sufficient to give a total horizontal
strain matching the strain in the passive element. Compatibility of horizontal strains requires that

Mohr-Coulomb
failure

=
A

vertical
stress v

=0

a.
h
P

b.
Mohr-Coulomb
failure

hA = -hP

hP

horizontal stress h

Figure 4: Box model: (a) stress paths and (b) relationship between vertical stress and horizontal strain between
the two elements

= Ka after
yield at A
v

= 0 after
yield at P

a.

b.
P

P
h

Figure 5: Box model: stress paths with (a) passive element P, and (b) active element A reaching failure first

element A
yields
element P
yields

a.

b.

P
v

element P
yields
v

element A
yields

d.

c.
P

Figure 6: Box model: (a)(c) stress paths and (b)(d) stress:strain response for (a) (b) element P reaching yield
first; (c) (d) element A reaching failure first

elastic compression of element P should be equal and opposite to combined elastic and plastic extension
of element A:
hA = ehA + phA = ehP
(8)
so that phA = ehP ehA . Then the constant volume condition at failure implies that:
vA = evA + pvA = evA phA = evA + ehP + ehA

(9)

We can discover the dependence of the strain to yield (when the first element reaches the Mohr-Coulomb
failure condition) and the strain to eventual perfectly plastic failure (when the second element has also
reached the Mohr-Coulomb failure condition) on the initial stress state Ko . The three sections of the
vertical stress:vertical strain response for element A are shown schematically in Fig 6c, d.
3. Graphical and calculation exercise
1. Choose a normalised Youngs modulus E/b = 200, Poissons ratio = 0.25 and angle of friction
= 30 .

2. Calculate Ka , Kp and slope of stress path A = /[2(1 )].


3. Plot initial normalised stress states v /b = 1, h /b = Ko for values of Ko = (1 + Ka )/2, 1 and
(1 + Kp )/2.
4. Plot the stress paths for elements A and P for these three initial stress states. For each value of
Ko discover which element reaches failure first and calculate the corresponding vertical strain in
element A (normalised with E/b ).
5. Plot the stress paths followed as the second element heads for failure.
6. Calculate the normalised vertical strain in element A at which the second element reaches failure.
7. Plot the normalised stress:strain response of the footing for each initial state. (Shown schematically in Fig 6.)
8. Challenge What is the effect of anisotropy of the elastic response on these several plotted relationships?
9. Challenge What is the effect of a non-zero angle of dilation > 0 on the strains that occur before
and during full failure of the footing ?
References
Muir Wood, D (2004) Geotechnical modelling, Spon Press (section 7.5.1)
Nordal, S (1983) Elasto-plastic behaviour of soils analysed by the finite element method. Doctor of
Engineering thesis, Norges Tekniske Hogskole, Trondheim (Chapter 3)


5.Camclay
(SBCSSM5,GM3)

Soil modelling: South East Asia: October-November 2010

5. Cam clay
David Muir Wood
d.muirwood@dundee.ac.uk

Py2
load
P

Py1

a.
b.
l
P

extension l

c.
l1

l2

loading and unloading of annealed copper wire


linear (elastic?) response when load less than previous
maximum load
nonlinear response when load exceeds previous
maximum load (yield load)
cycle of loading and unloading that exceeds past
maximum load leaves permanent (plastic) extension of
wire - but also enlarged elastic (stiff) region

p. 2/2

Py2
load
P

Py1

a.
b.
l
P

extension l

c.
l1

l2

penalty for increased elastic region (Py2 Py1 ) is


permanent increase in length of wire (2 1 ):
hardening law
elastic (recoverable) deformation occurs whenever the
load changes
change in load causing yield produces both elastic
(recoverable) and plastic (irrecoverable) deformations

p. 3/2

Py2

increase in yield load

Py1

plastic deformation = separation of


unloading-reloading lines
unloading-reloading lines
l1

l2

increase in permanent extension

elastic properties do not change when yielding occurs


(assumption)
concept of unloading-reloading lines for loads below the
current yield load
plastic deformation is equal to separation of
unloading-reloading paths at any load

p. 4/2

q
elastic region
p'
Cam clay
elastic-plastic model for soil
first question: are plastic strains occurring?
assume elliptical yield locus bounding elastic region

p. 5/2

Cam clay
effective stress plane
q

p'
v

unloading-reloading line

compression plane
p'

volumetric changes are


important: develop
model in parallel in
effective stress plane
(p , q) and compression
plane (p , v)
bulk modulus K = vp /,
shear modulus G:
isotropic nonlinear
elasticity
v = v ln p : unloadingreloading line in compression plane for all
stresses within current
yield locus

p. 6/2

p'

p'o

assume that shape of yield locus is not affected by prior


loading history
current yield locus characterised by size po at
intersection with p axis
second question: what is the penalty for increase in po ?
(hardening law)

p. 7/2

q
B
A

p'oA

p'

p'oB

penalty for increase in


po ? (hardening law)
volumetric hardening
model: change in po
linked only with change
in volume

iso-ncl
v
url A

pp

po
v po

url B

p'

p. 8/2

isotropic normal
compression line: p = po

B
A

p'oA

p'

p'oB

plastic volumetric strains:

p
po
p
p = v po = v p
elastic volumetric
strains:

ep = v p
p

iso-ncl
v

v = vp =

url A

v(ep

+ pp )

p
p

isotropic normal compression line (iso-nc):


v = N ln p

url B

p'

p. 9/2

=M

yl B
yl A
p'
oA
p'

p'oB

geometry of yield locus fixed: soil property M


yield locus expands to accommodate the change in
stress: size po
po not affected by stress increment causing expansion

p. 10/2

=M

=0
N
v

yield loci
always pass
through origin

p'
N
v

iso ncl

iso ncl

slope

v
=0
url

url

=0
=M

slope
=M

p'

geometric
similarity
of
intersection
points
of
paths
with
q/p = =
constant

p'=1

ln p'

p. 11/2

=M

=0
p'

N
v

N
v

iso ncl

iso ncl

slope

v
=0
url

url

=0
=M

slope
=M

p'

p'=1

anisotropic
consolidation
lines slope
in semilogarithmic
compression
plane
unloadingreloading
lines
have
slope in
ln p , v plane

ln p'

p. 12/2

Cam clay state boundary surface


in p, q, v space

each yield locus is associated with


an unloading-reloading line
state boundary surface encloses all
permissible states

qp
p

pp

M
p'o

p'

third question: what is mechanism of plastic


deformation? (relative amounts of plastic distortional
and plastic volumetric strain: pq /pp )
assume normality of plastic strain increments to yield
locus at current stress
= 0, pq /pp = 0; = M , pq /pp =

p. 13/2

Cam clay

volumetric hardening model: penalty for increased size


of elastic region is permanent change in volumetric
packing
adopt a graphical approach to the explanation of the
way in which the model operates in p , q effective stress
plane and in p , v compression plane
behaviour bounded in these two planes: stresses and
volume cannot change indefinitely (failure, limits to
density of packing)
compute stress:strain response: distortional strain is
unbounded and can increase indefinitely

p. 14/2

B
A

graphical construction
of the Cam clay
stress:strain response

p'

iso-ncl
A
B

p'

increment AB in
drained compression
test q/p = 3
stress increment
defines intersection
with new yield locus
and po
project
down
to
unloading-reloading
line in compression
plane

p. 15/2

iso-ncl

A
-vp

p'

separation of urls produces irrecoverable volume


change v p
plastic volumetric strain pp = v p /v

p. 16/2

1
q

A
p'
normality to yield locus gives D = pp /pq and hence
plastic distortional strain

p. 17/2

qe

3G

qp
B

elastic distortional strain increment is eq = q/3G


elastic and plastic components summed to give total
distortional strain
then repeat for next increment

p. 18/2

=M

p'
as stress ratio = q/p increases, so ratio of plastic
strain increments pp /pq reduces
yield locus can only expand if plastic volumetric strain
occurs
as M , soil tends to state of perfect plasticity
critical state: continuing shearing with no further
change in stresses or volume

p. 19/2

Cam clay drained compression test (1)

B-C-D-E-F: slope of yield locus becomes flatter


ratio distortional/volumetric strain becomes larger

F: no plastic volumetric strain (normality)

q : end of test!

Cam clay drained compression test (2)


how did we reach B? normal compression to A?
behaviour beyond B not affected by previous history

Cam clay drained compression test (3)


how did we reach B? overconsolidation to A?
behaviour beyond B not affected by previous history

Cam clay drained compression test (4)


heavily overconsolidated KP: yielding at Q with pp < 0
q

yield locus shrinks: softening stress:strain response

elastic: q > 0
Q-R-S-T: slope of yield locus becomes flatter:
ratio distortional/volumetric strain becomes larger:
T: no plastic volumetric strain (normality)
q : end of test!

elastic: p' > 0

Cam clay ambiguity of response


softening regime: yielding with > M
same stress change associated either with plastic (QR) or elastic
(QR') strain

bifurcation of response
numerical ambiguity

Physical consequence of uncertainty of response


elastic unloading

plastic softening confined to thin localised region


Taylor, 1948

Cam clay hardening and softening response

yl B
yl A

P
p'

stress increment PQ
inside yield locus yA
purely elastic

url A Q
url B

ep = p /vp

P
iso-ncl
p'

p. 20/2

yl B
yl A

R
P

stress increment PR
p'

p = 0 ep = 0

expands yield locus


from yA to y B:
po = poB poA

url A
P

url B
R

iso-ncl

plastic volumetric strain


pp = ( )po /vpo

p'

p. 21/2

yl B
yl A

R
P
p'

stress increment PS
pP S = pP Q and
poP S = poP R

url A Q
url B
S

P
R

iso-ncl

project down to
unloading-reloading
lines in compression
plane
v = 0: undrained

p'

p. 22/2

yl B
yl A

R
P
p'

stress increment PS:


v = 0: undrained
elastic + plastic
volumetric strain = 0

url A Q
url B
S

ep + pp = 0

P
R

p
vp

po
+ ( ) vpo

=0

iso-ncl
p'

p. 23/2

yl B
yl A

R
P
p'

stress increment PS:


v = 0: undrained
elastic + plastic
volumetric strain = 0

url A Q
url B
S

plastic compression
po > 0

P
R

iso-ncl

elastic expansion p < 0

p'

p. 24/2

q
B
yl A

yl B
A
p'o

p'

graphical construction of
Cam clay stress:strain
response
increment AB in
undrained compression
test v = 0

iso-ncl
v
url A
url B

volume constraint
defines intersection with
new unloading-reloading
line and po

A
B

project up to yield locus


in effective stress plane
p'

p. 25/2

url A

iso-ncl

url B
A

-vp
B

p'

separation of urs produces irrecoverable volume


change v p
plastic volumetric strain pp = v p /v
elastic volumetric strain ep = pp

p. 26/2

B
yl A
A

D
yl B

p'

normality to yield locus gives D = pp /pq and hence


plastic distortional strain

p. 27/2

qe

3G

qp
B

elastic distortional strain increment is eq = q/3G


elastic and plastic components summed to give total
distortional strain

p. 28/2

yl B
yl A

-p'
B
esp

p
tsp

p'

pore pressure from total and effective stress paths


pore pressure parameter a = q/p
then repeat for next increment

p. 29/2

Cam clay undrained compression

undrained stress increments in effective stress


plane and compression plane
plastic hardening AB: < M, p' < 0, p'o > 0
plastic softening QR: > M, p' > 0, p'o < 0

perfect plasticity G: = M, p' = p'o = 0

Cam clay undrained compression

step through undrained test on


normally compressed soil
conventional compression q = 3p

Cam clay undrained compression

step through undrained test on lightly


overconsolidated soil
conventional compression q = 3p

Cam clay undrained compression

step through undrained test on


heavily overconsolidated soil
conventional compression q = 3p

Weald clay undrained response division of pore pressure


6.Camclaygraphical
calculations(exercise)
(SBCSSM5)

Soil modelling: South East Asia: October-November 2010

6. Cam clay: triaxial tests


Graphical construction and direct calculation
David Muir Wood
University of Dundee

Graphical construction: drained test


Figure 1 can be used to follow the operation of the Cam clay model using essentially entirely graphical
construction. The test is constructed in four linked plots. Figures 1a, b are the two finite plots: the
effective stress plane (a) and the compression plane (b). Each yield locus in the effective stress plane is
linked with a specific unloading-reloading line in the compression plane, which intersects the isotropic
normal compression line (iso-ncl) at the corresponding value of isotropic preconsolidation pressure p0o ,
which is the mean stress where the yield locus intersects the p0 axis. With the known constraints of
a particular test (constant cell pressure, drained, for example) the path of the test can be definitively
tracked in these two plots.
Figure 2 shows a detail of one step of a conventional drained triaxial compression test. The effective
stress path is constrained to climb at gradient q/p0 = 3 in the effective stress plane. This path thus
fixes the points A and B at which it crosses two successive yield loci. Direct vertical projection down to
the compression plane at the same values of mean effective stress then fixes the points A and B on the
volumetric path from the intersections with the corresponding unloading-reloading lines.
The stress-strain response produces relationships which are unbounded: there is no limit, in principle, to
the shear strain that can be imposed. First we calculate the plastic volumetric strain. The vertical separation, at constant mean stress, of the unloading-reloading lines through A and B gives the irrecoverable
change in specific volume, v p (Fig 3). The plastic volumetric strain is then:
pp

v p
=
v

(1)

The plastic distortional strain is then calculated from the condition of normality to the yield locus applied
at the current stress point in the effective stress plane (Fig 4). The direction of the dotted arrow, drawn
perpendicular to the yield locus at point A indicates the ratio of plastic distortional strain increment to
plastic volumetric strain increment, 1/D:
pq
1
=
D
pp

(2)

We can calculate the elastic distortional strain from the change in distortional, deviator stress, q (Fig
5):
eq =

q
3G

We can sum the plastic and elastic components to find the total distortional strain increment:

(3)

Figure 1: Cam clay: drained test

100

b.

150 p'

50

100

150

p'

d.

c.

2.1

Cam
clay: drained test
200
graphical construction

2.2

2.2

2.3

2.3

2.1

2.4

iso-ncl

2.5
v

2.4

2.5
v

200

50

50

50

100

a.

100

B
A
p'

iso-ncl

A
B

p'
Figure 2: Drained test: stress increment

iso-ncl

A
-vp

p'
Figure 3: Plastic volumetric strain increment

1
q

A
p'
Figure 4: Normality to yield locus

qe

3G

qp
B

q
Figure 5: Elastic and plastic distortional strain increments
q = eq + pq

(4)

and we can project across to the strain diagrams, Figs 1c, d, to plot the corresponding points on the
distortional stress:distortional strain plot (Fig 1c) and the volume:strain plot (Fig 1d).
(In fact, the inclusion of the elastic distortional strain is equivalent to imposing a skew or shear on the
distortional stress:distortional strain plot, since the elastic distortional strain is directly proportional to
q. So, for the purposes of this exercise, we can simply work in terms of plastic strain and recall that for
completeness we would need to add the elastic strains. Alternatively, we need to assume a value of shear
modulus: for example, G = 2MPa? All the other constitutive parameters for Cam clay are implicit in
the plots of Figs 1.)

Graphical construction: undrained test


Figure 6 can be used to follow the operation of the Cam clay model for an undrained test, again using
essentially entirely graphical construction. The test is constructed in four linked plots. Figures 6a, b are
the two finite plots: the effective stress plane (a) and the compression plane (b). Each yield locus in the
effective stress plane is linked with a specific unloading-reloading line in the compression plane. With
the known constraints of a particular test (constant cell pressure, undrained, for example) the path of
the test can be definitively tracked in these two plots.
Figure 7 shows a detail of one step of a conventional undrained triaxial compression test. The path is
constrained to remain at constant volume in the compression plane (Fig 7b). This path thus fixes the

points A and B at which it crosses two successive unloading-reloading lines. Direct vertical projection
up to the effective stress plane at the same values of mean effective stress then fixes the points A and B
on the effective stress path from the intersections with the corresponding yield loci.
The stress-strain response produces relationships which are unbounded: there is no limit, in principle, to
the shear strain that can be imposed. First we calculate the plastic volumetric strain. The vertical separation, at constant mean stress, of the unloading-reloading lines through A and B gives the irrecoverable
change in specific volume, v p (Fig 8). The plastic volumetric strain is then:
pp =

v p
v

(5)

The plastic distortional strain is then calculated from the condition of normality to the yield locus applied
at the current stress point in the effective stress plane (Fig 9). The direction of the dotted arrow, drawn
perpendicular to the yield locus at point A indicates the ratio of plastic distortional strain increment to
plastic volumetric strain increment, 1/D:
pq
1
=
D
pp

(6)

We can calculate the elastic distortional strain from the change in distortional, deviator stress, q:
eq =

q
3G

(7)

We can sum the plastic and elastic components to find the total distortional strain increment (Fig 5):
q = eq + pq

(8)

and we can project across to the strain diagram, Fig 6c to plot the corresponding point on the distortional
stress:distortional strain plot.
(In fact, once again, the inclusion of the elastic distortional strain is equivalent to imposing a skew or
shear on the distortional stress:distortional strain plot, since the elastic distortional strain is directly
proportional to q. So, for the purposes of this exercise, we can simply work in terms of plastic strain and
recall that for completeness we would need to add the elastic strains. Alternatively, we need to assume
a value of shear modulus: for example, G = 2MPa? All the other constitutive parameters for Cam clay
are implicit in the plots of Figs 6.)
There is by definition no volume change during an undrained test. The fourth plot for this test is a plot
of pore pressure against distortional strain (Fig 6d). The increase in pore pressure is the increase in
separation of the total stress path (which is arbitrary, but might be a conventional triaxial compression
path imposed with constant radial stress and hence with slope q/p = 3) and the effective stress path
(which was deduced in Fig 7a). It is instructive to distinguish the components of pore pressure change
that result from the change in total mean stress and the change in effective mean stress (Fig 10):
u = p p0

(9)

and the plot of pore pressures can separate the development of the two components indicating clearly
the pore pressure that comes from suppressed volume change.

Figure 6: Cam clay: undrained test

2.1

2.2

2.3

v
2.4

2.5

b.

50

iso-ncl

100
p'

150

50

100

p'

150

d.

c.

Cam clay: undrained test


200
graphical construction

50

100

100

50

a.

50

100

100
a.

50

A
0

50

2.5

100

150 p'

200

iso-ncl

v
2.4

2.3

A
B

2.2
b.
2.1

50

100

150

200
p'

Figure 7: Undrained test: stress increment

2.4
iso-ncl

2.3

-vp
B

2.2
50

100

p' 150

Figure 8: Plastic volumetric strain increment

50

D
A

0
0

50

100

150

p'

Figure 9: Normality to yield locus


q

-p'
B

50

esp

tsp

0
0

50

100

150

p'

200

Figure 10: Pore pressure generation: tsp = total stress path; esp = effective stress path
Calculation: drained test
Of course, we can perform analytically the same operations that we have done graphically: and having
the graphical experience allows us to see exactly how we should proceed with the analytical calculations.
We can set these up in a spreadsheet but can also perform the calculations by handand it is important
to use hand calculation in order to check that the spreadsheet is indeed doing what is intended.
1. We must evidently know the values of soil parameters: G, , , M , N .
2. We need initial conditions for the increment of the test: current effective stresses p0A , qA and specific
volume vA , and size of the yield locus (preconsolidation pressure) p0oA . In general the current effective
stress state can lie either on or inside the current yield locus: let us assume that our soil is normally
compressed so that the current stress state lies on the current yield locus. This implies that the current
conditions must satisfy the equation:
p0o = p0 +

q2 1
M 2 p0

(10)

In fact, the way that this volumetric hardening model is formulated means that Cam clay also imposes
a link between the current specific volume and the current effective stresses and size of the yield locus.
The geometry of the compression plane (Figs 1b, 11) gives:

v=N

ln p0o

p0o
+ ln 0
p

(11)

3. We will assume that we are conducting a conventional drained test with constant lateral stress (cell
pressure) so that:

iso-ncl

A
-vp

ln p'
Figure 11: Plastic change in specific volume: semilogarithmic compression plane
q
=3
p0

(12)

and we can start the calculation process by choosing a value for the increment q and hence the new
value of qB .
4. From (12) we can calculate the corresponding increment p0 and hence the new value of p0B .
5. From (10) we can calculate the size of yield locus p0oB required to accommodate this new stress state.
6. From (11) we can calculate the new specific volume vB . Through steps 4-6 we have established the
position of the new point in the effective stress and compression planes (Figs 1a, b) as was demonstrated
graphically in Fig 2.
7. Now we have to calculate the strain increments. We can divide the change in specific volume into two
parts: one due to the change in p0o , the plastic part; and one due to the change in p0 , the elastic part.
The plastic part, v p is the volume separation of the unloading-reloading lines at the start and end of
the increment: the geometry of the compression plane shows that this is (Figs 3, 11):
v p = ( ) ln (p0oB /p0oA )

(13)

The plastic volumetric strain increment is then:


pp =

v p
v

(14)

8. The plastic strain increment is normal to the yield locus at the current stress, so the plastic dilatancy,
D, is given by:
pq
1
2
=
p =
2
D
p
M 2
where stress ratio = q/p0 . Hence we can calculate the plastic distortional strain increment, pq .

(15)

Table 1: Cam clay: drained test calculation


q

p0

p0

= q/p0

p0o

v p

pp

1/D = pq /pp

pq

eq

eq

9. The elastic distortional strain increment, eq , is calculated using the shear modulus G for the soil:
eq =

q
3G

(16)

The plastic and elastic components are added to give the total distortional strain increment, q .
10. The calculation process is repeated for a new stress increment, starting at step 3. A table illustrating
the sequence of calculation steps may be helpful (Table 1).

Calculation: undrained test


We can follow a similar process to develop the analytical result for the undrained testguided again by
the graphical experience.
1. We must evidently know the values of soil parameters: G, , , M , N .
2. We need initial conditions for the increment of the test: current effective stresses p0A , qA and specific
volume vA , and size of the yield locus (preconsolidation pressure) p0oA . These are linked as described for
the analysis of the drained test.
3. From our graphical experience (Fig 7), we start the calculation process by choosing a value for the
change in mean effective stress p0 and hence the new value of p0B .
4. The specific volume remains constant so that we can calculate the new preconsolidation pressure p0oB
from the geometry of the compression plane (11).
5. Hence we can calculate the new value of qB from (10), and deduce the corresponding increment
q. Through steps 3-5 we have established the position of the new point in the effective stress and
compression planes (Figs 6a, b) as was demonstrated graphically in Fig 7.
6. Now we have to calculate the strain increments. We divide the change in specific volume into two
parts: one due to the change in p0o , the plastic part; and one due to the change in p0 , the elastic part.
The plastic part, v p is the volume separation of the unloading-reloading lines at the start and end of
the increment: the geometry of the compression plane is shown in (Figs 8, 11). The plastic change in
volume can be calculated from (13) and the plastic volumetric strain from (14). The elastic volumetric
strain is exactly equal to the plastic volumetric strain but of opposite sign.
7. The plastic strain increment is normal to the yield locus at the current stress, so the plastic dilatancy,
D, is given by (15). Hence we can calculate the plastic distortional strain increment, pq .
8. The elastic distortional strain increment, eq , is calculated using the shear modulus G for the soil.
The plastic and elastic components are added to give the total distortional strain increment, q .
9. We will assume that we are conducting a conventional undrained compression test with constant
lateral total stress (cell pressure) so that:

Table 2: Cam clay: undrained test calculation


v

p0

p0

p0o

= q/p0

v p

pp

1/D = pq /pp

q
=3
p

pq

eq

eq

p u

(17)

and we can calculate the increment in pore pressure from the known changes in total and effective mean
stress (9).
10. The calculation process is repeated for a new stress increment, starting at step 3. A table illustrating
the sequence of calculation steps may be helpful (Table 2).

Cam clay: graphical and calculation exercise


1. Use the plots provided (Figs 1, 6) to follow the progress of a drained and undrained triaxial compression
test on Cam clay. Guess reasonable values for any soil parameters that are required.
2. Follow the steps through the calculation procedure for conventional drained and undrained tests
in order to produce numerical values to compare with the graphical results. Estimate values for soil
parameters consistent with the geometry of the plots in Figs 1, 6. (Values implied in these figures may
not be the same for the two tests.) You do not need to use the same step sizes implied by the intervals
between yield loci in these figures.


7.Camclay:compliance
formulation
(GM3)

Soil modelling: South East Asia: October-November 2010

7. Cam clay: compliance relationships


David Muir Wood
d.muirwood@dundee.ac.uk

p'
M2
2
p'o M 2
flow
rule

yield
locus

pp

M 2 2

p
2
q

p', q p'o pp qp
hardening
rule

pp

p'o
v p 'o

p, q

p', q pe, qe
elastic
properties
ep

eq


vp'

0 p'

1 q

3G

Cam clay calculation process

Cam clay: convert graphics into equations


qp

=M

pp

=0
p'o

p'
N

N
iso ncl

iso ncl

slope

v
=0
url
url

=0

slope

=M

=M
p'

p'=1

ln p'

p. 2/

=M

yl A
p'
oA
p'
elliptical yield locus
q2

p
(p

p
)=0
o
2
M

or

p
M2
= 2

po
M + 2

soil parameter M current size (history) po

p. 3/

=M

yl B
yl A
p'oB

p' p'oA

q2

p
(p

p
)=0
o
2
M

incremental form: p , q po
po

= (2p

2q q
+ 2
p
M p

p
po )

p. 4/

q
B
A

volumetric hardening law


p'oA

p'

p'oB

p
o
pp =
v po

iso-ncl
v




q
2
2 p
=
(M ) + 2
2
2
v(M + )
p
p

url A

url B

p'

p. 5/

qp
p

pp

M
p'o

p'

mechanism of plastic deformation: flow rule


assume normality of plastic strain increments to yield locus
at current stress
M 2 2
pp
p =
2
q

p. 6/

decomposition of strain increment into elastic and plastic


components = e + p
elastic compliance
!
!
!

0
ep
p

vp
=
1
eq
q
0 3G
plastic compliance
!
p
p

= 2
p
vp (M + 2 )
q

M 2 2

4 2
M 2 2

p
q

normality (flow rule) symmetric plastic compliance


relationship

p. 7/

Cam clay undrained compression


p'o
p'
p

0
vp'
vp'o
p'
2

2
2
p'

p'


p'i M
> M, p' > 0 < M, p' < 0

undrained stress path cannot be the same as the yield locus


M i2
2
2
2

balance of elastic expansion/plastic compression: < M


balance of elastic compression/plastic expansion: > M

Cam clay undrained compression

p'o
p'
p

0
vp'
vp'o

initial stress A inside current yield locus

elastic response: p'o = 0


hence: p' = 0: constant p'
until yield at B
(isotropic elastic model)

Cam clay undrained compression

reminder: pore pressure has external and internal components


u = p p' = p + aq

(pore pressure parameter a)

2
a
M 2 2 2 2

a is not a soil constant depends on history


8.Stiffnessformulation:
MohrCoulomb,Camclay
(GM3)

Soil modelling: South East Asia: October-November 2010

8. Stiffness formulation: Mohr-Coulomb, Cam clay


David Muir Wood
d.muirwood@dundee.ac.uk

General elastic-perfectly plastic model


define boundary of elastic region: yield surface f () = 0
divide strain increment into elastic and plastic parts:
= e + p
inside yield surface response is elastic: = De
inaccessible

yield = failure

elastic

stress

p. 1/

define boundary of elastic region: yield surface f () = 0


when the soil is yielding the stress change must lie along
the yield surface (consistency): f () = 0 and
f T
f =
= 0

inaccessible

yield = failure

elastic

stress

p. 3/2

describe mechanism of plastic deformation via a plastic


potential g() such that:
g
=

where is a scalar multiplier which will define the


magnitudes of the plastic strain increments
g
= D = D( ) = D D

combine with consistency condition to deduce:


e

f T
D
g
f T
D

p. 4/2

... and thus the stiffness relationship linking stress


increments with increments of total strain:

T
f
D g
D

= D ep
= D
g
f T
D

which has the form of an elastic predictor and a plastic


corrector
is a given strain increment elastic or plastic?
guess elastic (predictor)
if resulting stress unacceptable then apply plastic correction

p. 5/2

example: elastic perfectly plastic Mohr-Coulomb model


qp
p
q

q
inaccessible

pp

M
M*

elastic
p'

p'

isotropic elastic D =

K 0
0 3G

p. 6/2

example: elastic perfectly plastic Mohr-Coulomb model


qp
p
q

q
inaccessible

pp

M
M*

elastic
p'

p'

yield surface f () = q M p = 0

p. 7/2

qp
p
q

q
inaccessible

pp

M
M*

elastic
p'

plastic potential

p'

g() = q M p + k = 0

where k fits the plastic potential to the current stress state

p. 8/2

qp
p
q

q
inaccessible

pp

M
M*

elastic
p'

p'

plastic potential
g() = q M p + k = 0
!
!
!
p
p
g/p
M
=
=
p
q
g/q
1
plastic mechanism

pp

=
M
pq

p. 9/2

initial
stress
plastic

q = Mp'

elastic
p'

consistency requires M p + q = 0: stress change must


lie along failure locus for plastic strains
!
!
!
p
K 0
ep
=
q
0 3G
eq
!"
!
!#
K 0
p
M
=

0 3G
q
1

p. 10/2

q = Mp'

initial
stress
plastic

elastic
p'

deduce =




M 1

M 1

K 0
0 3G
!
K 0
0 3G

p
q

M
1

p. 11/2

final result
p
q
1

KM M + 3G

"

K 0
0 3G

M M K 2 3M GK
3M GK
9G2

!#

p
q

elastic predictor and plastic corrector which is asymmetric


unless M = M (associated flow)

p. 12/2

if plastic strains occurring


!
p
3GK
=
KM M + 3G
q

1
M
M MM

p
q

check that q = M p
plastic work W p = p pp + qpq = (M M )p pq
plastic work W p = 0 for M = M - physically unreasonable
- in practice M < M

p. 13/2

q
a.

b.
q

M* > 0

p'

dilation
q

p
c.

M* < 0

compression

Mohr-Coulomb: typical response in tests with p = constant


dilation/compression depends on sign of M

p. 14/2

p
plastic potential g() = 0

elastic
stress
hardening yield surface f(, ) = 0

elastic-hardening plastic model


natural extension of perfectly plastic model
yield surface now function not only of stresses but also of
hardening parameter(s) : f (, ) = 0

p. 15/2

p
plastic potential g() = 0

elastic
stress
hardening yield surface f(, ) = 0

consistency condition requires stress to remain on


expanding yield surface
f T
f
f =
+
= 0

p. 16/2

divide strain into elastic and plastic parts = e + p


elastic model = De
plastic mechanism defined by plastic potential g()
g
=

g
= D = D( ) = D D

combine consistency and flow rule


e

f T
f T g
f =
+
=0
p

p. 17/2

define a hardening function H


f T g
H=
p

and deduce as before:


=

f T
D
f T
g
D

general stiffness relationship is

g f T
D D
= D ep

= D
g
f T
D

+ H

elastic predictor and plastic corrector

p. 18/2

example: elastic-hardening plastic Cam clay model


qp
p

pp

M
p'o

p'

yield function and plastic potential assumed identical


2
q
g() = f (, po ) = 2 p (po p ) = 0
M

single hardening parameter po

p. 19/2

yield function and plastic potential assumed identical


2
q
g() = f (, po ) = 2 p (po p ) = 0
M

plastic strain increments given by:


!
!
g
p
p
p
=
=
p
g
q
q

2p po
2q
M2

p. 20/2

N
v

hardening rule linked with


linear relationships in semilogarithmic compression plane
po /pp
po /pq

p' = 1

ln p'

p' = p'o

vpo /( )
0

pure volumetric hardening

p. 21/2

compute hardening function function H





 vpo

f po g

2p po
H = p = p
po p p

and full elastic-plastic stiffness relationship (K = vp /):


p
q

K 0
0 3G

h
K

2 6GKq(2p po )
2

K (2p po )
2
M

6GKq(2p po )
36G2 q 2

M2
M4

i
vp po (2p po )
2
12Gq 2

(2p po ) + M 4 +

p
q

elastic predictor (is the resulting stress state within the yield
locus?) and plastic corrector

p. 22/2

increasing overconsolidation ratio

M=1.2
1
normally consolidated

a.
0

0.05

0.1

0.05
normally consolidated

b.
-0.05

increasing
overconsolidation ratio
0

0.05

0.1

Cam clay: drained compression with p = 0

p. 23/2

100

100
a.

b.

increasing overconsolidation ratio

q: kPa

q: kPa

50

50
increasing
overconsolidation
ratio

0
0

50
p': kPa

100

0
0

40
u: kPa
20

0.05

0.1

c.

0
-20
-40

increasing overconsolidation ratio

0.05

0.1

Cam clay: undrained compression

p. 24/2

Cam clay
stiffness formulation always works: strain increment
stress increment
elastic prediction - followed by plastic correction if
predicted elastic stress state violates the yield condition
however, compliance formulation may be less
cumbersome: stress increment strain increment
but note that this breaks down if soil wants to soften and is unsure whether to unload plastically or elastically

p. 25/2

Sand
to first order yielding depends on
mobilised friction (at low
stresses)
volume changes primarily linked
with distortional dilatancy rather
than mean stress compression

Fuji River sand (Tatsuoka, 1972)

explore possibilities for


distortional hardening MohrCoulomb model

example: add prefailure nonlinearity to Mohr-Coulomb


model: elastic-hardening plastic model
qp

failure: q - pp' = 0
q

inaccesssible

plastic potentials
M
pp

elastic + plastic
p

yield
loci

yield: q - yp' = 0
elastic

p'

p'

frictional yield function


f (, ) = f (p , q, y ) = q y p = 0

single hardening parameter y

p. 26/2

Dilatancy

loose

loose packings contract, dense packings expand


on shearing

dense

Dilatancy

shear box
direct observation of dilatancy
link mobilised friction Q/P and
dilatancy y/x

Ottawa sand (Taylor, 1948)

Dilatancy
Taylor flow rule
dense

loose

incremental work: W = Py + Qx
dissipate in friction: W =Px
y Q
tan =
= = tan m
x P

dilatancy = friction constant


sawblade analogy
constant friction c on
inclined surfaces
mobilised friction m = c +

Dilatancy
tan = tan m = tan m tan c
= tan c : critical state friction shearing with no volume change
Ottawa sand: 0.49, c 26
analogy with triaxial test
x qp

evidence for link between dilatancy and friction

y pp

detail of link less certain

Q q
P p'
y
Q
=
x
P

flow rule stress-dilatancy law for sand model

pp
qp

=M

q
=M
p'

example: add prefailure nonlinearity to Mohr-Coulomb


model: elastic-hardening plastic model
qp

failure: q - pp' = 0
q

inaccesssible

plastic potentials
M
pp

elastic + plastic
p

yield
loci

yield: q - yp' = 0
elastic

p'

p'

nonassociated flow - normality to yield function gives


excessive dilation (and W p = 0)

p
g() = q M p ln r
p

plastic mechanism
!
p
p
=
p
q

g
p
g
q

M
1

p. 27/2

hyperbolic distortional hardening rule


1

mobilised friction

shear strain
0

0.1

0.2

y
pq
=
p
a + pq

or incrementally
(p y )2 p
y =
q
ap

p. 28/2

define a hardening function H


f T g
H=
p

general stiffness relationship is

g f T
D D
= D ep
= D
f T
g
D

+ H

elastic predictor and plastic corrector

p. 1/

isotropic elastic properties:


D=

K 0
0 3G

partial differentials required for construction of stiffness


matrix:
!
!
p
y
y /p
0
=
=
p
p

y /q
(p y )2 /ap

f
=

f /p
f q

y
1

f
= p
y

p. 29/2

hardening function:


f y T g
2

H=
=
p
0
(

)
p
y /ap
p
y

M y
1

2
(

)
y
p
=p
(ap )

and stiffness matrix:


! "
!
p
K 0
=
q
0 3G
K 2 y (M y ) 3GK (M y )
3GKy
9G2

3G Ky (M y ) + p (p y ) / (ap )

elastic predictor and plastic corrector

p
q

p. 30/2

drained and undrained compression tests


1

mobilised friction

1
0

q/p'i

shear strain
0

0.1

M = 0.8, p = 1

0.2

M = 1, p = 1

expansion
0

M = 0.8, p = 1.0
0.2

M = 1.0, p = 1.0
M = 1.2, p = 1.0

compression

M = 1.2, p = 1
0
0

p'/p'i

0.05

volumetric strain

but no softening
continuing volumetric deformation (unless M = p )

p. 31/2

compare elastic-perfectly plastic Mohr-Coulomb model


!
3GK
1
M
3G + KM M
M MM
and elastic-hardening plastic Mohr-Coulomb model
!
3GK + KH
3GK (M y )
3GKy
3GKy (M y ) + 3GH
3G Ky (M y ) + H

hardening function
H 0;

M y ;

2
(

)
p
y
H = p
(ap )

M (M y )

expanding yield locus; variable dilatancy

p. 2/


9.Selectionofsoilparameters
(exercise)
(GM3)

Soilmodelling:SouthEastAsia:OctoberNovember2010

9. Practicalexercise:choiceofsoilparameters
DavidMuirWood
d.muirwood@dundee.ac.uk

triaxial compression of normally compressed Weald clay


elastic-perfectly plastic (EPP)
Mohr-Coulomb model is very
widely available but requires
judgement for selection of
stiffness and strength
Cam clay (CC) is quite widely
available and provides more
control over nonlinearity
hardening Mohr-Coulomb (MCH)
is non-standard but demonstrates
that there may be several routes to
fitting available data

triaxial compression of Hostun sand


use spreadsheet to find optimum sets
of parameters for [elastic-perfectly
plastic Mohr-Coulomb,] Cam clay
and elastic-hardening plastic MohrCoulomb models

stress ratio q/p

1.5

0.5

observation
0
0

0.05

0.1

0.15

0.2

0.15

0.2

shear strain eq

shear strain eq
0

0.05

0.1

there should be many right answers


there are many ways of obtaining
matches of similar quality

it would be more realistic and


more challenging to try to match
several sets of data simultaneously

volumetric strain ep

0.005
0.01
0.015
0.02
0.025
0.03
0.035
0.04
0.045
0.05

observation

data from supposedly the


same test in different
apparatus are not the same
what accuracy should be
expected in comparing
simulation and
observation?

advanced soil model


requires critical state data
difficulty in obtaining
reliable void ratio
information
model prefers a slightly
higher critical state line to
optimise fit

Summary
Mohr-Coulomb model is widely available but
unsuitable to describe prefailure nonlinearity
elastic-hardening plastic models are more
useful but one simulated test is insufficient
for decision on optimum model
desirable to match range of data for stress
paths relevant to application
experimental data are not infallible


10.MohrCoulombimproved:
strength,criticalstates,
particlebreakage

Soilmodelling:SouthEastAsia:OctoberNovember2010

10.MohrCoulombimproved:strength,critical
states,particlebreakage
DavidMuirWood
d.muirwood@dundee.ac.uk

a.MohrCoulombimproved:strengthandcriticalstates

standardelasticperfectly
plasticMohrCoulombmodel
nonassociatedplasticflow

simplicity
sharpstiffnesschanges
tangentstiffnesseither
elastic orzero
continuingvolumechange

elastichardeningplasticMohrCoulombmodel
nonassociatedflowsteadyfallinstiffnesscontinuingvolumechange

standardelasticperfectlyplastic
MohrCoulombmodel
availableinallnumericalanalysis
programs
subjectivityinselectingvaluesofsoil
parameters stiffness,strength,
dilatancy

SevernTrentsand
buildonMohrCoulombmodel
describejourneyfrominitialelastic
responsetoultimatecriticalstate
includenonlinearity,peakstrength
andsoftening
simplicity?
adequatecomplexity?

SevernTrentsand:strength
error:strengthasanindependent quantity
infactdependent ondensityandstresslevel
introducecriticalstateline locusof
asymptoticstates(largestrain,endoftest)
criticalstateline

stateparameter =volumedistance
fromcriticalstateline

'loose'
specific
volume

functionofdensityand stresslevel

'dense'
meanstress

moreusefulthanvoidratioalone
indicatingeffectofdensityand stress
Been&Jefferies

SevernTrentsand:strength
peak
strength

whatispeakstrength?
propertyofthesoilwhichchanges
withstresslevel,density
dataconfirmlinkbetweenstrength
andstateparameter

stateparameter

'loose'
specific
volume

SevernTrentsandisjusttheMohr
Coulombmodelwithcurrent
strengthdependentoncurrent
stateparameter

'dense'
meanstress

Been&Jefferies

SevernTrentsand:dilatancy
dilatancy:volumechangeduring
shearing

2.0

q/p'

dense

1.5

constant strength

1.0

loose

'dense'sandexpands

0.5

'loose'sandcontracts

0.0

a.
0

volumetric

dilatancydependsondensity

10

unlimited
contraction

loose

10

15

20

shear strain: %
shearstrain

dilatancyvariesduringtest
-4

unlimited dilation
b.

-8

whatdowemeanby 'dense' and


'loose'?

20

shear strain: %

volumestrain
strain: %
0

15

dense

Benahmed

SevernTrentsand:dilatancy
criticalstateline

dilatancy

'loose'
specific
volume

'dense'

'loose'

'dense'

stateparameter

meanstress

dataconfirmlinkbetweendilatancyandstateparameter
ifsoilisnot atcriticalstatewhenitisbeingsheared( 0):
thenvolumechangesoccurtowardsthecriticalstate:dilatancy
'loose': >0:contraction
'dense': <0:dilation

Been&Jefferies

MohrCoulombmodelwithstrengthdependentonstatevariable
ratio currently mobilisedstrength
currently availablestrength

distortionalhardening
monotonicincreaseofratioofmobilised
toavailablestrength(/p)with
distortionalstrainqp

1
monotonic
relationship
shear strain

hyperbolichardeninglaw:simple

mobilised
strength'mob

available strength':
varieswithstateparameter

conventional drained triaxial compression tests


3

variation of current
peak strength

increasing
initial
density

mobilised
strength
q/p' 2

hence change in available strength

a.

0.1

0.2

0.3

shear strain

0.2

volumetric b.
strain 0.1

-0.2

increasing
initial
density

model automatically homes in on critical


state
softening emerges without being
described mathematically

0
-0.1

volume change accompanies shearing


hence change in state variable

1
0
0

different initial density (state variable)

0.1

0.2

0.3

shear strain

peak strength is moving target reached at


infinite distortional strain then identical
with critical state strength

Soil modelling: South East Asia: October-November 2010

10.Mohr-Coulomb improved: strength, critical


states, particle breakage
David Muir Wood
d.muirwood@dundee.ac.uk

b. Mohr-Coulomb improved:
non-monotonic loading

p. 2/1

G/Gmax
0.5

end of elastic
region?

typical experimental observation

stiffness
0

0.0001

0.001
0.01
shear strain %

0.1

stiffness falls steadily with


monotonic straining

20 %

is there an elastic region?

damping
10

damping
Quiou sand: resonant column and torsional
0.0001

0.001

0.01
shear strain %

0.1

shear tests (after LoPresti et al, 1997)

p. 3/1

perfectly plastic
stress

tangent stiffness Gt/G

elastic

secant stiffness Gs/G


strain

damping

0.5

tangent stiffness Gt/G


0

secant
stiffness

0.001

0.01

0.1

10

strain

elastic-perfectly plastic Mohr-Coulomb model


implied secant stiffness and damping variation

p. 4/1

1.0

secant stiffness
Gs/G

mobilised friction
0.5

shear strain

0
0

0.1

tangent stiffness
Gt/G

damping

0.0
0.2

0.001

0.01

0.1

qp/a

10

hyperbolic hardening
implied stiffness and damping variation

p. 5/1

kinematic
yield surface
q

failure
elastic region
stress path

elastic
region

b.

kinematic
yield surface

p'

a.

Severn-Trent sand: add kinematic hardening


elastic region of high stiffness carried round with recent
stress history: boundary of elastic region is the yield surface
plastic hardening stiffness depends on separation of the
yield surface and bounding surface
Mohr-Coulomb: hierarchical development of models

p. 6/1

e=0.83
e=0.87
e=0.92
e=0.94

- batr06 (model)
- batr05 (model)
- alert51 (model)
- batr02 (model)

q (kPa)

q (kPa)

100

100

increasing initial
state parameter
0

0
0

100

p' (kPa)

200

10
axial strain (%)

15

Severn-Trent sand calibrated against triaxial test data for


Hostun sand
effect of different density/stress level automatically
described
Gajo & Muir Wood, 1999

p. 7/1

Modelling non-monotonic loading


experiment

q (kPa)

simulation

q (kPa)

100

25 cycles

100
89 cycles

0
0

200

p' (kPa)

400

200

p' (kPa)

400

undrained cyclic loading: pore pressure rise leading to liquefaction


q (kPa)

q (kPa)

100

100

0
0

2
axial strain (%)

2
axial strain (%)

p. 8/1

Severn-Trent sand with kinematic hardening


simulate cyclic undrained loading leading to liquefaction
model fails after 25 cycles
actual soil (Hostun sand) fails after 89 cycles
number of cycles to liquefaction is not a particularly reliable
parameter to use for model calibration
significant difference: samples which liquefy in one or two
cycles and those which survive for many cycles
character of cyclic pore pressure build-up reproduced in
model

p. 9/1

B
D

C
a.

b.
G

Masing rule for strain reversals: double scale on reversal


remember history until previous path regained
stiffness always jumps at reversal (= experimental
observation)

p. 10/1

60
1.0
Gs/Gmax
0.8

hyperbolic
typical relationship

hyperbolic

damping

40

damping
ratio %

0.6
typical relationship
0.4

20

stiffness
0.2
0.0001 0.001

0.01
0.1
shear strain %

10.0

10

0
0.0001 0.001 0.01
0.1
shear strain %

10

0.04
/Gmax

hyperbolic

0.02

0.000

4
6
shear strain %

10

hardening law secant stiffness and damping

p. 11/1

increasing density (reducing state parameter)


200

a.

0.02
00

200

20
pore pressure
00

100

0.02

100

0.02

0.02

0.02

100

50
100
200
100

00

200
100

b.

50
00

50 z

100

0.02

00

pore pressure
0.02

100

00
100

0.02

100
100

100

00

50 z

50

100

0.02

pore pressure

00

0.02

c.

100
0.02

00

100
0.02

Severn-Trent: kinematic hardening: undrained strain cycles

p. 12/1

messages
possible to develop elegant models which reproduce
desirable mechanical characteristics
especially effects of density and stress level
mathematical complexity not essential
build up from well known model: Mohr-Coulomb

p. 13/1

Soil modelling: South East Asia: October-November 2010

10.Mohr-Coulomb improved: strength, critical


states, particle breakage
David Muir Wood
d.muirwood@dundee.ac.uk

c. Mohr-Coulomb improved: particle


breakage

p. 2/

Chattahoochee River sand


100

triaxial
compression

80
% finer
60
40
20
0

compression
62.1MPa
0.02

initial
grading

0.05 0.1 0.2


0.5
particle diameter: mm

evolving particle size distribution: particle breakage


triaxial compression with confining pressure 62.1MPa
(Vesic & Clough, 1968)

p. 26/8

100

% finer

grading state index IG


A
limiting
grading

50
D
d=dmin
0

current
grading

d=dmax

B
C
particle diameter: d (log scale)

IG = area ABC/area ABD (current and limiting


gradings)

limiting grading might be fractal (Appolonian)...


...scaling factor for calculation of IG (area ABD = B )
(other definitions possible)
for linear grading IG = [ln (dmax /dmin )]/2B

p. 14/3

Grading state index IG: critical states


original
material

specific
volume

onset of
crushing

critical state line?


deduced from tests with
increasing stress levels
IG increasing (irreversible)
changing material
crushing
complete

mean stress: log scale

Grading state index IG: critical states


critical state line
before crushing: IG = 0

specific
volume
IG increasing:
evolving
critical state
line

reinterpretation:
critical state
surface

limit of critical state lines


during crushing: 0 < IG < 1

critical state line


crushing
exhausted: IG = 1

IG
mean stress: log scale

critical state and isotropic compression lines


specific
volume
v

isotropic compression line


through current state

critical
state line

1
p' (logarithmic scale)

vc = v + (
v v) exp (p /pcs )

ensure realistic values at low and high stress

p. 20/3

critical state surface


2.4
specific
volume v
2.0
1.6
1.2
0
0.5
grading state
index: IG

1 10-3

105

109

10
mean stress p'
log scale

specific volume as combined function of grading IG and


mean stress p
critical state line changes as particles removed

p. 22/3

agglomerated particles: DEM: (Cheng, 2005)


100

% passing by volume

Influence of axial strain:

6.66 mm

389 agglomerates

80
60

originalPSD
Original
distribution
20MPa
20
MPa
before
compression
shear
axial
20
MPastrain
axial strain
= 0.3
= 0.3
axial
20
MPastrain
axial strain
= 0.5
= 0.5 crushing
axial
20
MPastrain
axial strain
= 0.6
= 0.6
axial
20
MPastrain
axial strain
= 0.7
= 0.7
(Cheng, 2004)

40
20
0
100

Size, log d (m)

1000

evolving particle size distribution through breakage of


contact bonds within agglomerates
isotropic compression to 20MPa (negligible change)
shearing (axial strains indicated)
dmax somewhat constant

p. 32/8

particle crushing: critical state surface


critical state line
before crushing
IG = 0

Voids ratio, e

2.7
2.5

fresh samples

specific
volume

Final states of p' -constant tests:


Dense
Loose
Loose, over-compressed

2.3
Loose
2.1
1.9
Dense
1.7

loose: normal
compression and
unloading

1.5 Loose, over-compressed


1.3
1.1
0.9
0.7

IG increasing:
evolving critical
state line

precompressed samples
1

10
Pressure, p' (MPa)

100

IG

mean stress
logarithmic scale

loci of end points on critical state surface


precompression leads to lower critical state specific
volume (higher IG )
(Cheng)

p. 33/8

effect of addition of fine particles?


1.2
void ratio
1.0
emax
0.8
emin

0.6
0.4

20

fine particles filling


gaps - natural result
of crushing

40
60
80
fines content: %

100

fine particles pushing


larger particles apart unnatural?

effect on emax and emin


all aspects of behaviour linked with void ratio range
affected
for example: location of critical state line

p. 29/8

DEM: removal of particles: gradings


100
% finer

RD=10 after
removal

50

10
20

5
RD

10
20
50
100
particle diameter (mm)

gradings used for discrete element modelling (solid


curves)
grading reached by removal of particles from initial
grading with RD = 10 (dotted curve)

p. 17/8

DEM: grading and critcal states


1.25
specific
volume

RD=dmax/dmin

RD
2
5

1.2

10
20

1.15
0

2
mean stress (MPa)

broadening grading lowers critical state line


broader gradings pack more efficiently

p. 21/3

Severn-Trent sand model


specific
volume

critical state line &


state parameter

critical state
critical
state
line

stress:dilatancy
relationship
expansion

0
plastic dilatancy

mean effective stress


available
strength

critical state

strength dependent
on state parameter
0
state parameter

contraction

1
mobilised strength
available strength

monotonic hardening
towards current strength
plastic distortional strain

p. 19/3

Severn-Trent sand
extended Mohr-Coulomb model
model built round critical state line as divider of
response
adequate complexity - effects of density, strain softening
simple assumed relationships
(use as basis for extended model)
many such models exist - aesthetic judgement mathematical expediency

p. 21/4

particle crushing: effect of increasing IG


specific
volume

state parameter

critical
state line
falls
mean effective stress

available
strength

critical state

mobilised strength
available strength

strength
unchanged
0
state parameter

critical state

stiffness
unchanged

stress:dilatancy
unchanged
expansion

plastic distortional strain

contraction
0
plastic dilatancy

lowering of critical state line (first order )


strength unchanged (first order )
stiffness unchanged (first order )
dilatancy unchanged (first order )
slope of critical state line unchanged (first order )
few data - often from artificial mixtures not naturally
crushed or eroded materials

p. 31/8

how best to crush particles...

compression produces particle breakage ...


... but shearing better

p. 34/8

ring shear apparatus: Dogs Bay sand


100
% finer

estimated shear strain


11100%
730%
104%

10

52%

compression
initial

0.001

0.01
0.1
particle size: mm

evolution of particle size distribution: constant after


about 730%? (definition of strain in ring shear?)
double logarithmic axes
(after Coop et al., 2004)

p. 35/8

IG 1 inevitably?
0.8

compression
800kPa

relative
breakage
Br
0.4

650-930kPa
248-386kPa

60-97kPa

1?
grading
state index
IG
0.1?

10

1000
100000
shear strain: %

relative breakage Br IG
different normal stresses
crushing does not continue indefinitely
final grading depends on stress level
Dogs Bay sand: ring shear tests (Coop, 2004)

p. 36/8

particle crushing
normalised
deviator stress particle breakage
(yield criterion)

current strength
4
P

3
1
mobilised strength
(yield criterion)
normalised mean stress

crushing yield locus: normalised stresses (Hardin)


h
i
p 1 + 12 (
/M )3 pc = 0
normality of crushing strains

rotation of principal axes certainly damaging

p. 38/8

particle crushing
specific
volume
v

isotropic compression line


through current state
critical
state line

vmin

-vcIG

1
p' (logarithmic scale)

energy to create new material surface through crushing


magnitude of crushing volumetric strain
critical state line falls as IG rises:
v IG
newly crushed material steals some void ratio from
remaining soil
compression line falls somewhat

p. 39/8

particle crushing: simulations


AF

1.5

1.9
F

stress ratio
q/p'
1.0

1.7
specific
volume v

critical state
line IG=0

1.5

0.5

critical state
line IG=1

a.
c.
1.3

0.0
0.0

0.2

0.4 0.6 0.8


axial strain

1.0

volumetric
strain 0.04 contraction

1.0
IG
0.8

0.02
0.00
0.0
-0.02

0.2

0.4 0.6 0.8


axial strain

1.0
b.

100
1000
10000
mean effective stress p' (kPa)
no crushing for test AF

0.6
0.4

-0.04
0.2

-0.06
-0.08

expansion

AF

d.

0.0 A
100
1000
10000
mean effective stress p' (kPa)

precompression: A

p. 40/8

particle crushing: simulations


AF
ABAF

1.5

1.9

stress ratio
q/p'
1.0

1.7
specific
volume v

critical state
line IG=0

1.5

0.5

critical state
line IG=1

a.
c.
1.3

0.0
0.0

0.2

0.4 0.6 0.8


axial strain

1.0

volumetric
strain 0.04 contraction

1.0
IG
0.8

0.02
0.00
0.0
-0.02

0.2

0.4 0.6 0.8


axial strain

1.0
b.

100
1000
10000
mean effective stress p' (kPa)
no crushing on unloading

0.6
0.4

-0.04
0.2

-0.06
-0.08

expansion

ABAF
AF

d.

B
0.0 A
100
1000
10000
mean effective stress p' (kPa)

precompression: A, ABA

p. 41/8

particle crushing: simulations


AF
ABAF

1.5

1.9

stress ratio
q/p'
1.0 ACAF

1.7
specific
volume v

critical state
line IG=0

C evolution of
critical state on
path ACAF

1.5

0.5

critical state
line IG=1

a.
c.
1.3

0.0
0.0

0.2

0.4 0.6 0.8


axial strain

1.0

volumetric
strain 0.04 contraction

1.0
IG
0.8

0.02
0.00
0.0
-0.02

0.2

0.4 0.6 0.8


axial strain

-0.04

ACAF

-0.06
-0.08

expansion

ABAF
AF

1.0
b.

100
1000
10000
mean effective stress p' (kPa)
no crushing on unloading

0.6
0.4
0.2

d.

B
0.0 A
100
1000
10000
mean effective stress p' (kPa)

precompression: A, ABA, ACA

p. 42/8

particle crushing: simulations


AF
ABAF

1.5

1.9

stress ratio
q/p'
1.0 ACAF
ADAF

1.7
specific
volume v

critical state
line IG=0

B
C

1.5

0.5

critical state
line IG=1

a.
c.
1.3

0.0
0.0

0.2

0.4 0.6 0.8


axial strain

1.0

volumetric
strain 0.04 contraction
0.02

1.0
IG
0.8

ADAF

0.00
0.0
-0.02

0.2

0.4 0.6 0.8


axial strain

-0.04

ACAF

-0.06
-0.08

evolution of
critical state on
path ADAF

expansion

ABAF
AF

1.0
b.

100
1000
10000
mean effective stress p' (kPa)
no crushing on unloading

0.6
0.4
0.2

d.

B
0.0 A
100
1000
10000
mean effective stress p' (kPa)

precompression: A, ABA, ACA, ADA

p. 43/8

particle crushing: simulations


AF
ABAF

1.5

1.9

stress ratio
q/p'
1.0 ACAF
AEAF
ADAF

1.7
specific
volume v

critical state
line IG=0

evolution of
critical state on
path AEAF

C
D

1.5

0.5

E
critical state
line IG=1

a.
c.
1.3

0.0
0.0

0.2

0.4 0.6 0.8


axial strain

volumetric
strain 0.04 contraction
0.02

1.0

AEAF

IG
0.8

ADAF

0.00
0.0
-0.02

0.2

0.4 0.6 0.8


axial strain

-0.04

ACAF

-0.06
-0.08

1.0

expansion

ABAF
AF

1.0
b.

100
1000
10000
mean effective stress p' (kPa)
no crushing on unloading
E
D

0.6
0.4
0.2

d.

B
A
0.0
100
1000
10000
mean effective stress p' (kPa)

precompression: A, ABA, ACA, ADA, AEA

p. 44/8

particle crushing
shearing at constant mean effective stress p
samples with different precompression histories
evolution of grading state index IG
compression plane ln p : v
IG , peak strength , soil feels looser

increases pore pressure generation ...

p. 45/8

particle crushing: undrained compression

150
deviator
stress q: kPa
100

specific
volume
2.0
v

1.8
increasing
initial
1.6
50
specific
volume
1.4
0
0
50
100 150
mean stress p': kPa

initial state

v=2.0

v=1.9
v=1.8

critical state
line IG=1

critical
state line undrained path
for v = 1.7 in
v=1.7
absence of
v=1.6
crushing
v=1.5

1
10
100
1000
mean effective stress p' (kPa)

effective stress paths


paths in compression plane ln p : v
effect of initial specific volume v

p. 46/8

increased fines: decreased permeability


permeability of mixtures of kaolin clay with fine gravel
10%
log10[permeability]
(m/s)
-8

-9
0

clay content

20%
90%
30% 50% 70%
80%
100%
40% 60%

1 void ratio 2

log10[permeability]
(m/s)
clay content
10%
-8
30-100%
20%

-9
1

2
3
clay void ratio

for clay contents above 30% the permeability is determined by the clay matrix
the gravel particles merely occupy space and reduce flow cross-section
(data from Kumar)

p. 47/8

crushing: consequences
crushing lowers critical state line
material feels looser
increased tendency for pore pressure generation
reduced permeability: slower pore pressure dissipation
increased likelihood of liquefaction

p. 48/8

modelling proposals
underpinning constitutive model for distortional
response
characterisation of grading
link between grading and particle removal or particle
breakage
link between grading and critical states
assumption concerning change of specific volume
resulting from erosion (destabilisation) or stolen void
ratio for particle breakage
problem of validation data: importance of
fabric/structure

p. 76/8


11.Camclayimproved:
nonlinearity,structure

Soilmodelling:SouthEastAsia:OctoberNovember2010

11.Camclayimproved:nonlinearity,structure
DavidMuirWood
d.muirwood@dundee.ac.uk

a.Camclayimproved:nonlinearity

plastic
lessstiff
shear
stress

shear
stress
elastic stiff
mean
stress

classicalelasticplasticmodellingofsoil

forexample,Camclay(1963,1968)

shear
strain

stress

yield?

classicalidentificationofyield
fromstress:strainresponse
typical actual response

void
ratio
strain

geometricalconstructionfor
estimationofpreconsolidation
pressure

preconsolidation
pressure

vertical stress
(log scale)

Camclayprovidinginspiration:
searchforCamclaylike yieldloci
egkaolin(AlTabbaa,1984)
200

Cam clay?

q: kPa

150
100
50
p': kPa

0
0

100

200

300

-50
-100
-150
Fig 3: Anisotropic yield locus for one-dimensional stress history
(after Al-Tabbaa, 1984)

400

yieldlocifornaturalclays
0.9
0.8
0.7

q/ vc

0.6
0.5
0.4

Rang de Fleuve
Belfast

0.3

Winnipeg
0.2

St Alban
Lyndhurst

0.1

Mastemyr

0
0

0.1

0.2

0.3

0.4

0.5
p/ v c

collectedbyGrahametal(1988)

0.6

0.7

0.8

0.9

typicalexperimentalobservation:
stiffnessfallssteadilywithmonotonicstraining:
isthereanelasticregion?

limitofelastic
response??

shearstiffnessdegradationdataforQuiousandfromresonant
columnandtorsionalsheartests(afterLoPrestietal,1997)

howdoweobjectivelyidentifyyielding?
occurrenceofirrecoverablestrain?
dissipationofenergyinloading/unloadingcycles?
changeinslopeofstress:strainresponse?
stress

stress

b.

a.
strain

strain

directprobing

70

yieldingofBothkennar
clay:

q: kPa
60

boundariesdeduced
frominspectionof
stress:strainresponse

50

Y1approximately
centredoninsitu
stressstate

30

Y3reflectsnatural
structure damaged
byany irrecoverable
strain illusory
afterSmithetal(1992)

Y3 yield locus

40
Y2 yield locus

20
Y1 yield locus
nonmonotonic
probing damage
toyieldlocus?

10
0
0
-10
-20

10

20

30

40 p': kPa
50

60

kaolinrevisited:
isotropicconsolidationhistories
200

deviator stress q: kPa

a.

150
100
50
0
0
-50

50

100

150

200

250

mean effective stress p': kPa

datafromAlTabbaa(1987)

300

350

400

450

kaolinrevisited:
onedimensionalconsolidationhistories
200

deviator stress q: kPa

b.

150
100
50
0
0
-50

50

100

150

200

mean effective stress p': kPa

datafromAlTabbaa(1987)

250

300

350

400

q
kPa

q
kPa

p' kPa

p' kPa

plasticstrainincrements:approximatenormalitytokinematicyield
loci
q
q
kPa

kPa

p' kPa

kaolin: AlTabbaa,1987

p' kPa

Camclay

elastichardeningplasticmodel
volumetrichardening
associatedflow normality

Camclay
responseindrainedtriaxial
compressiontestswith
constantp'
asymptoticapproachto
criticalstate
effectofoverconsolidation
ratio
sharpdivisionbetween
elasticandplasticresponse

compareresponseofsoilonnonmonotonic
loadingwithcapabilityofsingleyieldsurface
model
elastichardeningplasticmodelexpectselastic
behaviouronreversal,suddendropinstiffness
atyield
soilstypicallyshowhystereticbehaviouron
unloadreloadcycles,steadychangein
incrementalstiffness

compareresponseofsoilonnonmonotonicloading
withcapabilityofsingleyieldsurfacemodel
extensiontosimplemodels
usingkinematichardening
andboundingsurface
plasticity

kinematichardeningextension
yieldlocuscarriedaroundwithstressstate 'bubble'
stronglyinfluencedbyrecenthistory
stiffnessfallsasyield'bubble'approachesbounding
surface controlledbydistanceb
whenloadingwith'bubble'incontactwithbounding
surfacemodelisidenticaltoCamclay

assumerelativesizeRof'bubble'
assumerulefortranslationof'bubble'
assumeinterpolationrulelinkingplasticstiffnesswithb
otherwiseidenticaltoCamclay

new
bounding
surface

bounding
surface

newyield
surface

p
yieldsurface
(bubble)

consistencylaw:stresscannotlieoutsidecurrentyieldsurface: +
yieldsurfaceexpands withboundingsurface asresultofhardeninglawandplastic
strains entirestressspaceisstretched
yieldsurfacetranslatesparalleltovectorb linkingpointswithparallelnormalsonyield
surface(atcurrentstress)andonboundingsurface ensuresthatyieldsurfaceand
boundingsurfaceneverintersect

assumerelativesizeRof'bubble'
assumerulefortranslationof'bubble'
assumeinterpolationrulelinkingplasticstiffnesswithb
otherwiseidenticaltoCamclay

volumetricstrain

kaolin

constantp' cycles
hysteresis
experiment

simulation

distortionalstrain

buildupof
volumetricstrain

q
kPa

p' kPa

migrationof'bubble'duringconstantp' unloading
afteronedimensionalnormalcompression
hardeningof'bubble'andboundingsurface

experiment

simulation

constantq cyclesafteronedimensionalnormalcompression

b.Camclayimproved:structure/bonding

naturalsoilsoftencontainstructure:bondingbetweenparticles:
destroyedwithmechanicalorchemicaldamage
designmodelinwhichyieldsurfacehasincreasedsizeasaresultof
thebondingbetweenparticles
extensionof'bubble'kinematicextensionofCamclay

ratioofsizesofstructure
surfaceandreferencesurface
givesindicationofcurrent
degreeofstructure
yieldsurfaceandbounding(structure)surfaceshaveincreasedsize
asaresultofthebondingbetweenparticles
withplasticstraining(orchemicalweathering)theyieldsurface
graduallyshrinkstotheyieldsurface,forremoulded,structureless
material
extensionof'bubble'kinematicextensionofCamclay
allfeaturesof'bubble'modelretained

reference
surface

structure
surface

p
yieldsurface
(bubble)

consistencylaw:stresscannotlieoutsidecurrentyieldsurface: +
yieldsurfaceexpands withboundingsurfaceandreferencesurface asresultof
hardeninglawandplasticstrains entirestressspaceisstretched

structure
surface

reference
surface

p
yieldsurface
(bubble)

consistencylaw:stresscannotlieoutsidecurrentyieldsurface: +
plasticstrainsdamagethestructureofthesoilaccordingtoadamagelaw structure
surfaceandbubbleshrinkandshifttowardsreferencesurface

structure
surface

p
yieldsurface
(bubble)

yieldsurfacetranslatesparalleltovectorb linkingpointswithparallelnormalsonyield
surface(atcurrentstress)andonboundingsurface ensuresthatyieldsurfaceand
boundingsurfaceneverintersect
carefulconsiderationofgeometricallogic!

addmeasureofstructureorbonding:singlescalarparameterr
'bounding'surfacenowcalled'structure'surface:sizertimes
largerthanareferencesurface
structurelostwheneverplasticstrainsoccur
damagelaw:

k
(r 1) dp
r =

damageplasticstrainincrementdp combinesplasticvolumetric
andplasticdistortionalstrainincrements:
additionalparametertocontroltheirrelative
importance
structureprogressivelydisappears:
r 1asplasticdeformationincreases

logic:structurelesssoilisonewhichhasbeensomechanically
pummelledthatithasnoremainingbondsbetweenparticles=
Camclay
particularformsoflaboratorytesting(triaxialtesting,for
example)maynotbeabletoprovidesufficientdamage
evolutionlawanddefinitionofdamagestrainmayneedtoinclude
somemoresubtlereferencetothenatureofthestrainpath
shearingwithrotationofprincipalaxesislikelytobeespecially
damaging
feasibletointroduceotherevolutionlawswhichrelatechange
(increaseordecrease)ofscalarmeasureofstructurer(ortensor
measureofstructure)tochemicalenvironmentortimeor
temperatureeffects
Camclaycanberegainedbysettingr=1,R=1

hierarchicalextensionof'bubble'
modeltoincludeeffectsof
structure
otherevolutionlaws:relatechange
(increaseordecrease)ofscalar
measureofstructurertochemical
environmentortimeor
temperatureeffects
Camclaycanberegainedbysettingr=1,R=1

Norrkping clay calibration tests

Rouainia & Muir Wood (2000)

Norrkping clay parametric variation


Rouainia & Muir Wood (2000)

Norrkping clay undrained isotropic consolidation


Rouainia & Muir Wood (2000)

Norrkping clay undrained anisotropic consolidation

Rouainia & Muir Wood (2000)

Norrkping clay undrained isotropic overconsolidation


Rouainia & Muir Wood (2000)

simulation

experiment
Bothkennarclay
Gajo&MuirWood,2001

resultsnormalisedbyHvorslevequivalent
consolidationpressurep'e forstructureless soil

HierarchicalextensionsofCamclay
itisrelativelystraightforwardtoaddextrafeaturestoa
soilmodel
advantageinusingwellknownmodelasbasis check
implementation acceptability
extrafeaturesimplyadditionalsoilparametersand
additionalcalibrationtests
seekadequatecomplexityinmodelling matchcomplexity
ofmodeltoavailabilityofdataandneedsofapplication


12.Conclusion

Soil modelling: South East Asia: October-November 2010

12. Conclusion
David Muir Wood
d.muirwood@dundee.ac.uk

volume
strength

critical
state line

density
strength
volume

p'
compression

density

specific
volume
v

loose: v > vf

stress level
strength

dense:
v < vf

strength

critical state
line: v = vf

mean effective stress p'


stress level

critical state soil mechanics


discipline for exploring and interpreting mechanical response of soils
importance of considering stresses and density
compare undrained and drained strengths of loose/soft and dense soils

p. 1/

Modelling should be elegant?


It is more important to have beauty in one's equations than
to have them fit experiment. (Dirac)

Those involved in modelling tend to become more interested


in the process than its purpose:
the stimulation of simulation is greater than the pleasurement
of measurement:
but it makes you go blind.
but Ockhams razor: Pluralitas non est ponenda sine neccesitate
(William of Ockham, ca 1285-1349)

difficulties with numerical modelling (Potts, 2003):


- because there is no standard numerical strategy for
implementation of nonlinear models
- because some constitutive models seem to be unable to give
reasonable predictions

- because, even for apparently simple problems, the results of


numerical modelling can be very dependent on the decisions made
by the user

PLAXIS analyses
same problem different modellers

spread of predictions
(Schweiger, 2003)

useful numerical modelling requires skilled operators who:


- have a detailed understanding of soil mechanics and the
underpinning theory for the numerical algorithms
- understand the limitations of constitutive models
- are familiar with the software that is being used for the
numerical modelling

Concentration on constitutive modelling


- essential component of any numerical modelling
- finite element, finite difference

Similar messages apply to numerical modelling itself


- and to physical modelling

modelling of geotechnical processes (understand soil model)


(also important for physical modelling)

- kinematic soil stiffness stiffness dependent on recent history


- centrifuge modelling

Numerical modelling Step Zero:


write down the answer!

Reflective practice cycle


always start with prediction of what you expect to happen
if observation unexpected, reflection required to improve predictions

You might also like