THE NATIONAL UNIVERSITY OF

SINGAPORE
DEPARTMENT OF CIVIL &
ENVIRONMENTAL ENGINEERING

CE6101 CONSTITUTIVE
BEHAVIOUR OF SOILS
1
 Learning Objectives and Main Topics
Learning objective: to familiarise students with the fundamental concepts in
constitutive models, as mathematical descriptions of mechanical behaviour of soils.
At the end of this module, students are expected to be able to use constitutive models
in finite element software and assign material parameters in an informed and
knowledgeable manner.
Main topics:
 (a) Description of elastic material behaviour.
 (b) Principle of effective stress and how it is expressed in a finite element context.
 (c) Stress paths.
 (d) Basic concepts of plasticity, including yielding criterion and flow rule.
 (e) The Mohr-Coulomb model as an example of an elastic perfectly plastic model.
 (f) Critical state soil mechanics - concept of the critical state as a state of ultimate
failure.
 (g) Stress dilatancy: volumetric-shear coupling in soil plasticity.

 (h) Original and modified Cam Clay models as examples of strain hardening
plasticity models.
2
Why
Geotechnical
Engineering and
Constitutive
Modelling?
 Lessons from the
Nicoll Highway
Collapse
 PE (Geo) & AC (Geo)
 MSc (Geotec)
3
An excerpt from the COI report (2005):

4
1. Introduction

Geotechnical design approaches can often be classified into 3 types:

(a) Factor of Safety approach. Classical approach, still widely used in
geotechnical design.
 A failure mechanism is first postulated.
 Using some accepted (on what criteria? Average or most pessimistic?) shear
strength parameters, the resistance of the geosystem to the postulated
failure mechanism is computed and compared with the loads causing failure.
 Ratio of resistance : loads is the factor of safety.
Increasingly being replaced by limit state approach with partial factors of
safety.
Drawbacks:
• not clear what the factor of safety actually factors against, e.g. how
much of the factor of safety safeguards against uncertainties in soil
parameters, groundwater conditions, loading conditions,
displacement control etc.
• working load behaviour of the geosystem not computed since it is
calculated for the failure condition. In some cases e.g. foundations,
the factor of safety is partially used to limit deformation. However,
deformation and ground movement is not directly predicted.

5
(b) Limit state approach e.g. Eurocode. Developed to
rectify the first drawback of the FoS approach, by clarifying
the aim of the factors.
 Failure of the geosystem (limit state) is explicitly
considered by the engineer.
 The engineer chooses a set of conditions e.g.
groundwater (or pore pressure) conditions, strength
parameters etc. which are considered to be the most
pessimistic for a worst-case scenario.
 The geosystem is then designed to be just stable (i.e.

factor of safety = 1) under this worst-case scenario. The

main advantage of this approach is that the contribution of
each factor e.g. strength, loading, groundwater, to the
margin of safety is explicitly reflected. However, it still has
the drawback of not being able to predict working load
behaviour.
6
(c) Performance-based approach.
 Objective: to rectify the second drawback of the factor of

safety approach.
 Behaviour of the geosystem at working load condition is

evaluated and compared with established criteria.

 Made possible by the advent of computational hardware and

algorithms (e.g. finite element analysis) which allows

complex continuum problems to be solved.
 To take advantage of this approach, we need to understand

the constitutive behaviour of the soil in the pre-failure

state and methods of describing it.
Complications of geotechnical design cf. structural design:
 Natural material with natural variation;
 Non-linear inelastic behavior;
 Continuous medium (not beams and columns);
 Pore water interaction.

7
2 Fundamental Concepts of Continuum Mechanics
2.1 Stresses and Strains
 FEM analysis: assume that ground is a continuous medium, or
continuum.
 Externally, the loads on a continuum are often described in terms of
forces and tractions.
 Internally, loading effects => stresses within the continuum.
 Consider a continuum subjected to a set of external loads. Suppose that
we want to evaluate the stresses at a point A within this continuum, then
what we need to do is to take an imaginary cut through the continuum at
and around point A and then evaluate the forces acting on an
infinitesimally small area around it.

Fy
A
Fx
Fz

8
Define a horizontal plane around A, then force acting through A can
be resolved into a normal force Fy and two mutually orthogonal
shear forces Fx and Fz. Then we can define the normal and shear
stresses on A as follows:

 - Fy 
 y  Limit  
A  0  A  (1)

 - Fx 
 yx  Limit   (2)
 A  0  A 

 - F 
 yz  Limit  z  (3)
A  0   A 

9
Define the other components of stresses x, z and xz by taking cuts in the
appropriate vertical planes. Furthermore, taking moment equilibrium of an
infinitesimal cube around A leads to
yx
xy dx dy= yx dy dx → xy = yx
Similarly yz = zy dy
dx
xy (5)
zx = xz (6)

Thus six independent stress components ie. x, y , z, xy, yz and xz are sufficient
to completely define the state of stress at a point in 3-dimensional space.
Externally, the effects of loads on a continuum are manifested in its deformations
and displacements. Internally, these effects are expressed in terms of strains. The
strains are normally defined by

x = - vx/x (7)
y = - vy/y (8)
z = - vz/z (9)
xy = -vx/y - vy/x (10)
xz = -vx/z - vz/x (11)
yz = -vy/z - vz/y (12)
10
  Strains
Externally, the effects of loads are manifested in its deformations and
displacements. Internally, these effects are expressed in terms of strains.
Consider the

vx vx + dvx

dx

rectangular section undergoing longitudinal deformation in the X-direction. Since strain is the
change in length normalized by the original length,
x = dvx/dx
In soils, conventional to adopt positive signs for compressive stresses and strains since soil
can only withstand compressive stresses and strains. Also, deformation can change in many
directions (not just the X-direction), we therefore use the general definition of strain as
x = - vx/x (7)
the partial derivatives ∂/∂x indicating that we only take into account the change of length with
respect to the X-coordinates only (and not Y- and Z-co-ordinates). Similarly, the normal strains
in the Y- and Z-direction y and z are given by
y = - vy/y (8)
z = - vz/z 11
(9)
 The shear strains are defined by the change of shape or distortion
of the body. Consider the same rectangular section being distorted
as shown: -dvx

dy 
dx
-dvy


The shear strain is defined as the total angle of distortion (in rad),  + , which, if small is
approximated by
 +  = -vx/y - vy/x ( and  in radians) or
xy = -vx/y - vy/x (10)
We can also derived similar expressions for the shear strains in the XZ- and YZ-planes
as follows:
xz = -vx/z - vz/x (11)
yz = -vy/z - vz/y (12)

12
 Elastic Stress-Strain Relations
Ratio of the stress experienced by the material at a point to the strain that it incurs
is known as a modulus.
Different moduli can be defined depending upon the loading and boundary
conditions. Important to ensure that we are using the appropriate modulus. The
most commonly encountered modulus is the Young’s modulus E, which is the
modulus manifested by a specimen subjected to a normal stress on one plane,
while the other two orthogonal planes are unconstrained. If an elastic body is
compressed in the x-direction by a normal stress x then it experiences the
following strains
x = x/E (13)
y = -x/E (14)
z = -x/E (15)

x

y = z = 0 13
If the same elastic body is stressed in the y-
direction by a normal stress y then it
experiences the following strains

y = y/E (16)
x = -y/E (17)
z = -y/E (18)

Similarly, the effects of z can be expressed as

z = z/E (19)
x = -z/E (20)
y = -z/E (21)14
 + +

x = x/E x = -y/E x = -z/E

y = -x/E + y = y/E + y = -z/E
z = -x/E z = -y/E z = z/E

x = x/E-y/E-z/E
→ = y/E-x/E-z/E
y

z = z/E-x/E-y/E
15
The combined effects of x, y and z on x can be expressed by
superposition Eqs. 13, 17 and 20, which leads to

x = x/E-y/E-z/E (22)

Similarly for the other two normal strain components

y = y/E-x/E-z/E (23)
z = z/E-x/E-y/E (24)

The effect of a shear stress is to cause a shear strain, ie.

xy = xy/G (25)

yz = yz/G (26)
xz = xz/G (27)

where G is the shear modulus and is related to the Young’s modulus

E by
E (28)
G
2(1   )
16
 Relationship between G and E (just for interest)
Consider the square element distorted by shear stress 
 to the parallelogram as shown:
c0  2a c 2  2a 2 1  cos  90  2   2a 2 1  sin 2   a c0

For small  , sin2 2 (in radians) c a
c= 2a 1  2  2a 1    using binomial expansion
1
2

and ignoring higher order terms.

The diagonal extensional strain
ε3 =   c  c0   
c0
1
The shear strain  =2 . Hence,  3    1
2 3
By the generalized Hooke's Law:
 3  1  1  
3   
E E E
or
  1    E
   G 17
2 E  2 1   
Eqs. 22 to 27 (generalized Hooke’s Law) can be written in matrix form as
follows:
  x   1/E  / E  / E 0 0   x 
0
     / E 1/E  / E 0 0 0   y 
 y 
  z    / E  / E 1/E 0 0 0   z 
   
 xy   0 0 0 1/G 0 0   xy 
(29)
 yz   0 0 0 0 1/G 0   yz 
    
 zx   0 0 0 0 0 1/G   zx 
We can also express the stresses in terms of strains as follows:

 x  1-   0 0 0  x 
    1-  0 0 0   
 y   y
 z    1- 0 0 0   z 
   A    

 xy   0 0 0 0.5- 0 0   xy  (30)
 yz   0 0 0 0 0.5- 0   yz 
     

 zx   0 0 0 0 0 0.5-   zx 
18
 E
in which A
(1 - 2 )(1   )
Eq. 30 is also written in matrix notation as
 = De  (31)
where De is the elastic stress-strain matrix.
Sometimes it is more convenient to express De in terms of the bulk modulus
K, where
E
K
3(1 - 2 ) (32)

Expressing De in terms of K and G leads to the matrix form

K  43 G K  23 G K  23 G 0 0 0
 
K  3 G K  43 G K  23 G
2
0 0 0
 K  23 G K  23 G K  43 G 0 0 0 (33)
D  
e

 0 0 0 G 0 0
 0 0 0 0 G 0
 
 0 0 0 0 0 G 
19
 Types of Two-Dimensional Problems
Finite element analyses are often conducted in 2- rather than 3-dimensions,
owing to limitations in computational resources, time and voluminous data input
needed for 3-dimensional analyses. There are two common types of 2-
dimensional problems, namely plane stress and plane strain.
 Plane Stress Analysis
A plane stress analysis is one in which the stresses on two of the faces is zero
e.g. thin plates. If we consider the anti-plane faces (i.e. the faces which is
facing out of and into the page) as the stress-free faces, then we effectively set
 z = yz = zx = 0 (44)

 Setting z = 0 in Eq. 30 leads to

εz =  (45)
(εx + εy)
1 

Substituting Eq. 45 into the first two rows of Eq. 30 leads to

2 2
x = A {(1-) εx +  εy - εx - εy}
1  1  (46)

which simplifies into

20
 x = E
 [εx + εy]
1  2

Similarly, y is given by
E
y =
1  2 [εx + εy]
so that the [De] matrix in two-dimensional plane stress condition can be written
as
 x  1  0   x 
  E 
   

 y  = 1  
2
1 0  y 
 xy   0 0 1  2    xy 
 

Plane stress analyses are often applied to problems relating to thin plates and
structural members wherein out-of-plane stresses cannot build up. Not often
used geotechnical engineering.

21
  Plane Strain Analysis
A plane strain condition is one in which the out-of-plane strain components are all
zero, i.e.

εz = yz = zx = 0
Substituting these conditions into Eq. 30 leads to
 x  1    0   x 
  = E   1   
 y  0   y 
 xy 
1  2 1    
  xy 
   0 0 1  
2   

z = E
And [εx + εy] =  (x + y)
1  2 1   

Plane strain conditions are commonly encountered in problems which involved great
lengths in the out-of-plane direction, e.g. tunnels, retaining walls. Many
geotechnical problems are assumed to fall into this category (or are they really??).

22
 Constrained
Modulus x
Cross-sectional area

The constrained modulus D

A

is commonly used in soil

mechanics. It is the ratio
between the stress and strain
in one direction (say X) when

x
the strains in the other two
orthogonal directions (say Y

Cross-sectional area
and Z) are constrained to be
zero e.g. oedometer.

A
23
Obtained by setting y = z = 0 in Eqs. 23 and 24, which
leads to
z = y =  x/(1-) (39)
Substituting Eq. 39 into Eq. 22 leads to
  2  2   x  1    2 2   x  1   1  2  
x  x 1         
E  1  1   E  1   E  1  

x/x = D = E (1   ) (40)
(1   )(1  2 )

The constrained modulus is the reciprocal of the coefficient

of volume change mv, which is usually used in the 1-D
compression and consolidation theory.

24
 Concept of Effective Stress
 Saturated soil comprises 2 phases, i.e. pore water
and soil skeleton.
 Effective stress: characterizes the contact forces
between the soil grains in the soil skeleton [“soil
skeleton” → the assemblage of soil grains].
 The higher the effective stresses σx’, σy’ and σz’,
the higher are the contact forces between soil
particles and thus, the higher is the frictional
resistance these soil particles have against relative
sliding → higher shear strength. Higher effective
(normal) stress → higher shear strength. Good
news! 25
 Saturated Soil
 Interaction between pore water and soil skeleton =>
interesting behaviour.
 Soil skeleton can take both normal and shear stresses.
Pore water can carry normal but not shear stresses.
Shear behaviour of soils depends on the inter-particle
friction and interlocking (see above). Both are related
to the macroscopic normal stress on the soil skeleton.
 To determine the macroscopic normal stress on
the soil skeleton - discount the pore pressure
contribution from the external normal stress.
 Consider a beaker full of soil + water.
 Vertical equilibrium ->
 Fi + u(A-Ac ) = P
Fi = vertical components of inter-particle contact
forces,
u = pore water pressure,
Ac = total inter-particle contact area.

26
Basis of Effective Stress Principle
•The Principle of Effective Stress is one of
 Fi + u(A-Ac ) = P the fundamental equations in soil
Generally Ac << A. Also P = σA mechanics.
Where σ is called “total stress” •The effective stress ’ is much better for
quantifying soil behaviour because it
so directly quantifies the inter-particle
 Fi + uA ~ A contact forces and discounts the pore
 Fi/A + u =  water pressure.
 Fi/A =  - u
Hence the total interparticle
contact force per unit area
Fi/A is the difference between
the total stress and pore
pressure.
Let ' =  Fi/A, then
' + u =  or ’=  - u
' = effective stress. ***********
27
 Karl Terzaghi – first proposed
the effective stress principle
• From Wikipedia:
• Karl von Terzaghi
(October 2, 1883 –
October 25, 1963)
was an Austrian civil
engineer, geotechnical
engineer and
geologist known as
the "father of soil
mechanics".

28
 2.3 Effective Stresses
The effective stress equations for continuum are
x’ = x – u (34)
y’ = y – u (35)
z ’ = z – u (36)
xy’ = xy (37)
yz’ = yz (38)
zx’ = zx (39)
Water can only carry normal stresses (i.e. pressure) but not shear stresses
In incremental terms: Δx’ = Δx – Δu (34)
Δy’ = Δy – Δu (35)
Δz’ = Δz – Δu (36)
Δxy’ = Δxy (37)
Δyz’ = Δyz (38)
Δzx’ = Δzx (39)
29
 Drained and Undrained Conditions – Two Extremes
in Pore-Water-Soil-Skeleton Interaction

 We will deal only with saturated soil in this

module.
 In saturated soil, all of the voids between soil
grains are completely filled by pore water. In
general, the pore water may flow relative to the
soil skeleton and the interaction between the two
phases can be quite complex. However, we can
define two extremes in this interaction, namely
drained and undrained conditions
30
 Drained
Condition
Consider a saturated soil body
submerged in water. If soil
body is relatively permeable
and the pore water can drain
into and out of the soil body,
then the pore pressure within
the soil body must be in
equilibrium with that in the
surrounding water, even when
∞
the soil skeleton is deformed.

E.g. if the surrounding water is in hydrostatic equilibrium, then the pore pressure would also
be in hydrostatic equilibrium i.e. excess pore pressure (i.e. that component of pore pressure
which is not in equilibrium with the ambient pressure) = 0. Termed drained condition. Hence,
drained condition => pore pressure distribution remains constant, usually at a known (e.g.
hydrostatic condition) or controlled (e.g. pore pressure control in a drained triaxial test) state
when the soil skeleton is deformed or loaded, i.e. no excess pore pressure i.e. Δu = 0 => Δx’
= Δx and same for y and z. In a drained test, we control the pore pressure at the drainage
boundaries, the soil controls the volumetric change.
31
 Undrained Condition
“Undrained” condition. Another extreme condition, not in terms of (no) excess pore
pressure, but in terms of (no) relative movement between soil skeleton and pore water
phase at all points within the soil body. Whereas drained condition is one in which the
pore water is allowed to come into complete equilibrium with the surrounding water,
“undrained condition” constrains the pore water to move together with the soil skeleton
i.e. no flow of water relative to the soil => no drainage, hence the name “undrained”
condition.
Note: drained condition is a “final” condition if drainage can occur and time is allowed
for pore pressure to equilibrate, undrained condition is a “non-equilibrium” pore
pressure condition. Given time, pore pressure will generally come to equilibrium. Thus,
undrained condition is usually a short-term condition. It will occur if the deformation
occurs so fast that the pore water has no time to drain and equilibrate (i.e. if excess
pressure has no time to dissipate). Best way to achieve undrained condition is not to
allow time for pore water to drain. However, this may not be practical in many
situations. For example, in triaxial tests, there is a limit to how fast one can practically
strain a specimen. In many “undrained” triaxial tests, even though the pore water
drainage valve is turned off, flow of pore water between one point within the sample to
another can still occur. If this happens, undrained condition is not strictly achieved.
Important: in undrained condition, there is no volume change i.e. volumetric strain32
increment Δεv~0
 Volumetric and Normal Strains
 Consider a cubical element with dimensions dx, dy and dz,
then
 original volume = dx dy dz
dz
 Suppose this element undergoes axial compressions -dvx, - dx
dvy and -dvz, such that the magnitudes of dvx, dvy and dvz
are much smaller than those of dx, dy and dz. This is
known as the “small strain” assumption. dy
 new volume = (dx-dvx)(dy-dvy)(dz-dvz)
 = dxdydz-dxdydvz-dydzdvx-
dxdzdvy+dxdvydvz+dydvxdvz+dzdvxdvy-dvxdvydvz
 = dxdydz (1+ x + y + z + higher order terms in x, y and
dx-dvx
z)
 Neglecting the higher order terms (since  is small) leads to dz-dvz

 new volume=original volume(1+x + y + z)

 change in volume/original volume = (x + y + z) dy-dvy

 Change in volume/original volume is the volumetric strain

v, thus
33
 v = (x + y + z)
Drained and Undrained Effective Stress Analyses
 An effective stress analysis is one in which the effective stress, pore pressure
and, of course, the total stress, are explicitly computed, i.e. soil skeleton and
pore water are separately characterized. An effective stress analysis needs
effective stiffness and strength parameters to characterize the stiffness and
strength of the soil skeleton. In here, we will only consider the effective stress
stiffness parameters. One can characterize the behavior of the soil skeleton in
terms of an effective stress-strain matrix De’, which is given by

K' 43 G' K' 23 G' K' 23 G' 0 0 0  (40)
K' 2 G' K' 4 G' K' 2 G' 0 0 0 
 3 3 3 
K' 23 G' K' 23 G' K' 43 G' 0 0 0 
D ' 
e

 0 0 0 G' 0 0 
 0 0 0 0 G' 0 
 
 0 0 0 0 0 G' 

 in which K’ and G’ are the effective bulk and shear moduli, respectively [i.e.
moduli of the soil skeleton]. 34
Relationships between Effective and Undrained Modulus
for Undrained Situations
Total E and  under undrained conditions denoted by Eu and u (undrained modulus and
Poisson’s ratio). Different from effective Young’s modulus E’ and Poisson’s ratio ν’ which
characterize the soil skeleton (i.e. without the pore water). Relationship between the
undrained and effective moduli:
For undrained situation, v = x + y + z = 0 (40a)
The incremental forms of Eqs. 22 – 24 for total stress in undrained situation are as follows:
Δεx = Δx/Eu -u Δy/Eu -u Δz/Eu (22a)
Δεy = Δy/Eu -u Δx/Eu - u Δz/Eu (23a)
Δz = Δz/Eu - u Δx/Eu - u Δy/Eu (24a)
Adding Eqs. 22a, 23a & 24a:
(x + y + z) (1 - 2u)/Eu = 0 (40b)
Since (x + y + z) and Eu are both non-zero, therefore 1 - 2u = 0  u = ½.
Pore water has no shear rigidity => Gu = G’ => Eu = E'
1  u 1  '
(1   ' ) Eu
E’ =
1.5 35
 Drained Analysis
 In a drained analysis, it is easy to separate the
effects of the soil skeleton and pore water
since the pore pressure distribution does not
change and any changes to the total stress must
be accommodated by the soil skeleton (i.e.
through effective stress). Hence the stress-
strain matrix is just given by Eq. 40.
 i.e Δu = 0 => Δσ’ = Δσ – Δu = Δσ

36
 Undrained Analysis
For undrained problems, strain components of pore water and those of skeleton are same =>
strain of the soil skeleton and pore water can be represented by a single strain vector, say ε.
Pore pressure vector u
 x   'x  u 
    '  u 
 y  y  
  z    'z  u 
u    -       D -D ' 
e e
 
 xy   'xy  0
 yz   'yz  0
     
But  
 zx   zx 
' 0
 x   K  43 G K  23 G K  23 G 0 0 0    x  Hence
 
 y
K  2 G
 3 K  43 G K  23 G 0 0 0    y 
 z   K  23 G K  23 G K  43 G 0 0 0   z 
       D 
e
 x   'x  u 
 xy   0 0 0 G 0 0   xy      '  u 
 yz   0 0 0 0 G 0   yz   y   y   
       z    'z  u 
 zx   0 0 0 0 0 G   zx  u    -       D e
- De '  
 xy   'xy  0
 'x   K'  43 G' K'  23 G' K'  23 G' 0 0 0    x   yz   'yz  0
 '   K'  2 G' K'  4 G' K'  2 G' 0 0 0          
 y   3 3 3  y   zx   'zx  0
 'z   K'  3 G' K'  3 G' K'  3 G' 0 0 0    z 
2 2 4
     D '
e
  
 'xy   0 0 0 G' 0 0   xy 
 'yz   0 0 0 0 G' 0   yz  37
    
 'zx   0 0 0 0 0 G'  zx 
in view of Eqs. 31 & 34 – 39 and the fact that the soil skeleton undergoes the
same strains as the pore water phase. In other words,

K - K' 43 (G - G' ) K - K' 23 (G - G' ) K - K' 23 (G - G' ) 0 0 0  x  u 

K - K' 2 (G - G' ) K - K' 4 (G - G' ) K - K' 2 (G - G' ) 0 0 0    u 
 3 3 3  y   
K - K' 23 (G - G' ) K - K' 23 (G - G' ) K - K' 43 (G - G' ) 0 0 0  z  u 
    
 0 0 0 G - G' 0 0   xy  0
 0 0 0 0 G - G' 0   yz  0
    
 0 0 0 0 0 G - G'   zx  0 

Note that the “effective” strain (i.e. strain of the soil skeleton) is the same as
the “total” strain (i.e. strain of the soil), in an undrained situation. This is
because the pore water in the voids deforms together with the soil skeleton. This
concept may be important in solving practice, assignment and exam
problems(!).

38
Since the shear stresses xy, yz and zx are entirely arbitrary and not
necessarily zero,
G – G’ = 0  G = G’ (43)
Thus, the total shear moduli is the same as the shear modulus of the
soil skeleton or the effective shear modulus, because pore water
has no shear rigidity.
Adding up the first 3 equations of Eq. 42 and simplifying leads to
(K – K’) (x + y + z ) = u (44)

Volumetric strain of soil v = x + y + z. Thus, Eq. 44 can be re-written

as
(K – K’) v = u

Let Ku = K – K’, then Ku v = u.

We note that v = -V/V where V is the volume of the soil. V is the

change in volume and is the sum of the changes in volume of the soil
grains and water. But soil grains (NOT soil skeleton) are much less
compressible than water and can usually be considered to be
incompressible.
39
Let εv = volumetric strain of soil (or soil skeleton) in undrained condition.
εw = volumetric strain of water
εg = volumetric strain of individual soil grain added up.
Then, εg << εw << εv . Hence when we compare εw and εv , we often
Can assume that εv ~ 0.

If soil grains are incompressible, then

V = Vw (45)

where Vw = change in volume of water.

V  Vw  Vw Vw e
 v =   =    n w (46)
1 e
w
V V Vw V

in which Vw = volume of water,

w = volumetric strain of the pore water,
e = void ratio, and
n = porosity.
Sub. Eq. 46 into Eq. 44, u = Ku n w = Kw w (47)
Where Kw = bulk modulus of water.
 Ku = Kw/n = Kw (1+e)/e = K – K’ (48)
 K = K’ + Kw/n (49) 40
 Sub. Eq. 49 into Eq. 33 gives
 Kw 4 Kw 2 Kw 2 
 K'  3 G’ K'  3 G’ K'   3 G’ 0 0 0
 n n n 
K K K
 K' w  23 G’ K' w  43 G’ K' w  23 G’ 0 0 0
 n n n 
D   K' K w  2 G’ K' K w  2 G’ K' K w  4 G’
e
0 0 0

 n
3
n
3
n
3
 0 0 0 G’ 0 0
 
 0 0 0 0 G’ 0
  (50)
 0 0 0 0 0 G’

for undrained loading. Thus the undrained stress strain relation of soil can
be expressed in terms of its effective moduli and the bulk modulus of water.
This is the approach used in Plaxis, SageCrisp etc. i.e. add a very large
modulus Kw/n into some of the terms in the De matrix. In other words, even
in an undrained analysis, one can still input K’ and G’. In general, Kw/n is
much higher than K’ and its actual value is usually not important. Thus, we
usually set Kw or Kw/n to about 10,000 times the K’ value.
41
 Total and Effective Stress Undrained Analysis
 Previous section shows how an undrained FE analysis can be conducted using effective
stress parameters.
 Approach 1 – add bulk modulus of the pore water (in the form of large number
compared to the modulus of the soil skeleton) to the effective stress-strain matrix. This
stress-strain matrix describes behavior of a nearly incompressible material, but the soil
skeleton and pore water behaviors are separately characterized using their respective
parameters namely K’, G and Kw. 2 parameters needed.
 Approach 2- in terms of total stress, using “undrained” parameters. In total stress
analysis, soil body is considered as a single-phase medium; the soil skeleton and pore
water are not separately characterized. The most common way to reflect
incompressibility of soil body → Poisson’s ratio νu ~ 0.5. The other parameter can be
either G or Eu of the soil body. In undrained condition, the Young’s modulus of the soil
body (i.e. soil skeleton + pore water) as a whole is usually denoted by Eu, the undrained
Young’s modulus → only 1 parameter needed.
 Total stress analysis (i.e. Approach 2) requires one less parameter than effective stress
analysis, but most modern geotechnical softwares use effective stress analysis. This is
because, in an effective stress analysis, one can evaluate the effective stress and pore
pressure separately, and of course the total stress. In a total stress analysis, only the total
stress is evaluated, the effective stress and pore pressure cannot be separately evaluated.
42
2.4 Stress Path Parameters

In the previous section, we saw that a complete description of

the state of stress requires six stress components . Very often,
we want to express or study how the stress changes over the
course of the event in question. This can be done by plotting
stress paths, which are plots of one stress component or
parameter against another. A very simple way of visualizing
stress path is to consider the changes to a Mohr-Circle in the
course of a typical triaxial test as shown below:

43
Triaxial Test
Apparatus
Triaxial Test
Schematic
Water level

a =1 h

specimen

r =3 u =wh

In a conventional triaxial compression test, σr = σ3 constant while σa = σ1 increases.

44
 

Stress path

O 3 1 

At the start of the test, 1 = 3 and the Mohr-circle is a point on the

-axis. As the test progresses, 1 increases while 3 remains constant;
thus the Mohr circle moves to the right as shown. From the start to
the end of the test, the sample progresses through an infinite number
of states, each of which can be represented by a Mohr-Circle. If we
are going to sketch all these circles, the stress space will be so
45
congested that the trend cannot be detected.
One way to solve this problem is to just plot the path traced out by the
topmost point of the circles as they change. This path can be considered
as a stress path (see figure above).

In a triaxial test, 1 = a and 3 = r; thus the co-ordinates of the

topmost point of a Mohr-circle can be expressed as follows:

x-coordinate s = (1 + 3)/2= (a + r)/2 (51)

y-coordinate t = (1 - 3)/2= (a - r)/2 (52)

The co-ordinates (s, t) can therefore be regarded as stress path parameters.

[In fact, their use was first suggested by MIT researchers Lambe, Whitman,
Ladd et al., and thus they are commonly known as the MIT stress path
parameters].
Effective Stress: s’ = (’1 + ’3)/2 = (1 –u + 3 - u)/2 = s – u
t’ = (’1 - ’3)/2 = (1 –u - 3 + u)/2 = t
t’ = t simply means radius (or diameter) of effective stress Mohr Circle =
radius (or diameter) of total stress Mohr Circle.
46
 Incremental Forms
We can express the stress path parameters in incremental
forms as follows:
Incremental total stress parameters:
Δs = ½ (Δσ1 + Δσ3)
Δt = ½ (Δσ1 - Δσ3) (52a)
Incremental effective stress parameters:
Δs’ = ½ (Δσ1’ + Δσ3’)
Δt’ = ½ (Δσ1’ - Δσ3’) (52b)
= ½ (Δσ1 -Δu - Δσ3 + Δu) = ½ (Δσ1 - Δσ3) = Δt

47
 Linear Elastic Material
Undrained Triaxial Test Stress Path of a Linear Elastic Material
The total stress path can be readily deduced by setting Δσ3 = 0, which
leads to
Δt/Δs = 1.
→ triaxial test total stress path is a straight line with gradient = 1.
For effective stress path, consider the generalized Hooke’s Law in
effective stress incremental form:
Δε1 = Δ’1/E’ -’ Δ’2/E’ -’ Δ’3/E’
Δε2 = Δ’2/E’ - ’ Δ’1/E’ - ’ Δ’3/E’
Δ3 = Δ’3/E’ - ’ Δ’1/E’ - ’ Δ’2/E’

Adding up the LHS and RHS and setting Δ’2= Δ’3:

Δεv = Δε1+Δε2 +Δ3 =1  2 '   '1  2 '3  = 0 => Δ’1 = -2Δ’3
E'
3
Δs’ = ½ (Δσ’1 + Δσ’3) = -½ Δσ’3 & Δt = ½ (Δσ1’ - Δσ3’) =   '3
2
Hence Δt/Δs’ = 3
48
Alternatively, use Skempton’s pore pressure equation.
 Sir Alec Westley Skempton (after
Wikipedia)
Sir Alec Westley Skempton FRS FREng (4
June 1914 – 9 August 2001)[1] was an English
civil engineer internationally recognised,
along with Karl Terzaghi, as one of the
founding fathers of the engineering discipline
of soil mechanics.[2] He established the soil
mechanics course at Imperial College London,
where the Civil and Environmental
Engineering Department's building was
renamed after him in 2004,[3] and was
knighted in the 2000 New Year Honours for
services to engineering. He was also a notable
contributor on the history of British civil
engineering. 49
50
51
 Alternatively, use Skempton’s Pore
Pressure Equation:
Skempton’s pore pressure equation: is a way of describing the excess
pore pressure in a triaxial test in terms of total stress changes.
We know from Skempton’s pore pressure equation
Δu = B {Δσ3 + A (Δσ1 - Δσ3)}
For a saturated soil, B = 1. Setting B = 1 and A = 1/3 for elastic saturated
material:
Δu = Δσ3 + 13 (Δσ1 - Δσ3) (52c)
Δσ1’ = Δσ1 – Δu = 2 3 (Δσ1 - Δσ3) = 4
3 Δt (52d)
Δσ3’ = Δσ3 – Δu = - 2 3 Δt (52e)
Δs’ = ½ (Δσ1’ + Δσ3’) =Δt /3
52
=> Δt/Δs’= 3
s-t stress paths for undrained
triaxial test – elastic material
t

ESP TSP

3
1

1 1

s’ s s, s’
u
53
 Elastic Plastic Material
Hence, for an isotropic linear elastic material, the effective stress path is a
straight line with a gradient of 3. Once the material becomes plastic, then the
gradient changes. In practice, this can provide a fairly accurate way of
identifying the yield point (see figure below).

54
Undrained Plane Strain
Biaxial Test on a Linear
Elastic Material
In a plane strain test, Δεz = 0 => Eq. 24 gives
Δσz ’ = ν’ (Δσx ’ + Δσy ’) (52f)
Sub Eq. 52f into Eqs. 22 & 23 gives
Δεx = (Δσx’ – ν’Δσy ’ – ν’2Δσx ’ – ν’2Δσy ’)/E’ (52g)
Δεy = (Δσy ’ – ν’Δσx ’ – ν’2Δσx ’ – ν’2Δσy ’)/E’ (52h)
In an undrained test, Δεv = Δεx + Δεy = 0 =>
(1 – ν’ – 2ν’2) (Δσx’ + Δσy’)/E’ = 0 (52i)
Since (1 – ν’ – 2ν’2)/E’ ≠ 0 all the time, Δσx’ + Δσy’ = 0
Thus, Δs’= ½ (Δσx’ + Δσy’) = 0
Hence, the effective stress path for undrained biaxial test on a linear
elastic material is a vertical straight line.
55
 Plane strain condition
 Quite commonly assumed in
geotechnical analysis.
 z = yz = zx = 0
 If there is no shear stresses as
well i.e. xy = 0, then only σx
and σy active. This is the plane
t
strain equivalent of the triaxial
test – sometimes called plane ESP
strain biaxial test. TSP

 Total stress path same as

triaxial, but undrained 1
effective stress path is
1
vertical.
s’ s s, s’
u
56
The stress path parameters s & t are widely used in practice as they are quite
intuitive.
However, they are used less in constitutive relationships as s and t do not take
into account the changes in the intermediate principal stress 2.

57
 Cambridge Stress Path Parameters p’ and q
An alternative way to define stress path parameters is through the way in
which stress affects the volumetric strain and strain energy of the material.
Summing Eqs. 22 – 24,
E
x + y + z = (1-2)(x + y + z)/E = v K
= (x + y + z)/(3K) = p/K (54) 3(1 - 2 )
in which v is the volumetric strain and p is the mean normal stress,
defined by p = (x + y + z)/3. (55)
In incremental form: Δv = Δp/K.
In effective stress: Δv = Δp’/K’.
In the special case where the stress is isotropic, i.e. x = y = z, then p =
x = y = z .
Volumetric strain at a point is independent of the direction of orientation
the x, y & z axes i.e. a strain “invariant”.
→ p (= KΔεv) is also invariant with respect to orientation and is thus known
as a stress invariant. Using a stress invariant to define stress path has the
important advantage that it can be treated as a scalar, without having to
define a direction. In contrast stress components e.g. σx, σy are dependent
on direction. 58
As shown in Eq. 54, the effect of p is to cause a volumetric strain; this
is true not only in linear elastic materials but indeed in most materials
including soils. In other words, the application of p = x = y = z
would only lead to volumetric strain.
Conceptually, we can consider all deformations to be sum of
volumetric change + distortion or shear (i.e. change of shape)
In most soils (and indeed most materials), failure does not occur by
pure compression but by shear or distortion {and hence the
importance of shear strength!}. Volumetric compression does not
usually lead to a failure situation wherein strains can increase
virtually indefinitely without increase in stress. Since we know
that the effect of p cannot usually cause failure, p is a useful stress
path parameter, since this implies that the parts of the stress
components which constitute p cannot cause failure.

59
Most stress path plots are 2-dimensional and therefore requires 2
parameters to define. We have now selected p as a parameter
and we know that the effect of this parameter is volumetric strain.
A useful complementary parameter is one which does not cause
the volumetric to change but instead the shape of the medium to
distort, at constant volume. This can be achieved by
considering the energy components. For a material subjected to
stress σ and undergoing strain ε, the strain energy is given by U
= ½ σε. σ

For 3-D stress-strain situations, the specific

strain energy (i.e. strain energy per unit ½ σε
volume) of a linear elastic medium is given by ε

U  12 ( x x   y y   z z   x x   y y   z z )
Substituting Eqs. 22 – 27 into Eq. 56 leads to
1 2 2 2  1 2 2 2
U  ( x   y   z ) - ( x y   y z   z x )  ( xy   yz   zx )
2E E 2G 60
The energy associated with the volumetric change of the medium is given
by
Uv = pv /2 = p2/(2K) = 1
   2  x y
(58)
z 
18K
Thus the specific strain energy Ud that causes distortion to the medium
without inflicting volumetric change (known as distortional energy) is given
by
Ud = U - Uv

= 1
12G
 2 2 2 1 2
2G

 x -  y    y -  z    z -  x    xy   yz2   zx2  
1 2
= q
6G
where q is defined by

q= 1
2
 x -  y 
2
  y -  z 
2
  z -  x   6
2
 2
xy   2
yz  2
zx 
61
Since the energy of distortion Ud [in fact, energy in general] is a
scalar and therefore an invariant, the parameter q is also a
stress invariant. Furthermore, whereas p defines the stress
component which causes pure volumetric change, q defines the
stress component which causes pure distortional change in an
elastic material.
Thus, they form a set of suitable and complementary stress
parameters. Note that, in a isotropic stress state,
x = y = z and xy = yz = zx = 0,
→ p = x = y = z → q = 0.
The parameter q therefore quantifies the amount of deviation
from the hydrostatic stress state and is thus known as the
deviator stress. The (p, q) stress path parameters were first
proposed for soil mechanics study by Roscoe, Schofield, Wroth
et al. in Cambridge and has thus come to be known commonly
as the Cambridge stress path parameters.
62
 Ken Roscoe

63
Andrew Schofield Peter Wroth John Burland
3. Triaxial Test Conditions

64
3. Triaxial Test Conditions

3.1 Total and Effective Stress Paths in an

Undrained Triaxial Test
Water level

Because the behaviour of soils under triaxial test

conditions have been very extensively studied, it is
useful to employ triaxial test conditions to illustrate soil
behaviour. In a triaxial test: a =1 h

specimen
x = z = r (i.e. the radial stress) (62)
y = a (i.e. the axial stress) (63) r =3 u =wh
xy = yz = zx = 0 (64)

Substituting Eqs. 62 – 64 into Eq. 55 and 61 leads to

p = (a + 2r)/3 (65) y
and q = a - r (66)
x
and hence the origin of the term deviator stress to
describe the quantity a - r. z

65
Good guys

Bad guys e.g. τ, t or q

vs bad guys

Good guys e.g. σ’, s’ or p’

 Effective stress path plot – a plot of good guy vs bad guy.
 Good guy – ’, s’, p’. All are ways of representing
effective normal stress.
 Bad guy – , t, q. All these are ways of representing shear
stress.
 Higher effective normal stress => better frictional strength
=> good! [Good guy]
 Higher shear stress => shear failure more likely => bad!
[Bad guy] 66
Bad guys e.g. τ, t or q

Good guys e.g. σ’, s’ or p’

67
In a conventional triaxial compression test, cell pressure r is
kept constant (i.e. r = 0) while a is increased. From Eqs. 65
& 66, we deduce that
p = (a + 2r)/3 (67)
q = a - r (68)

Since r = 0, Eqs. 67 & 68 reduces to

p = a/3 (69)
q = a (70)
 q/p = 3 (71)
or dq/dp = 3  q = 3p + C (72)
where C is a constant of integration. At the start of the test,
a = r = the cell pressure
q = 0 and p = r  C = -3r
q = 3 (p - r)

68
 In other words, the total stress path (TSP) in a conventional triaxial
compression test is a straight line having a gradient of 3 and originating
from the point (r, 0) in p-q stress space as shown below:

q TSP

r p

In parallel with the total stress path parameters, a corresponding set of

effective stress path parameters can also be defined as follows:
p’ = (x’ + y’ + z’)/3
q’ = 1
2
 ' -  '   ' -  '   ' -  '  6
x
2
y y
2
z
2
z
2
x
2 2
xy   yz   zx 

69
Also, in view of Eqs. 34 – 36, we deduce that
x’ - y’ = x - y (75)
y’ - z’ = y - z (76)
z’ - x’ = z - x (77)

Substituting Eqs. 75 – 77 leads to the conclusion

q’ = q (78)

Since q’ is the portion of q which is borne by the soil skeleton,

this physically implies that the pore water cannot sustain stresses
which causes distortion to the shape of the medium, it can only resist
changes to the volume. Stresses which cause distortion must be borne
entirely by the soil skeleton.
Under triaxial conditions,

p’ = (a’ + 2r’)/3 (79)

and q’ = a’ - r’ = a - r = q (80)
Substituting a’ = a – u & r’ = r – u into Eq. 79 leads to
p’ = (a + 2r)/3 - u = p – u (81)

70
Thus Eqs. 80 & 81 tell us that the pore water only sustain stresses which
causes volumetric changes not distortion.
Similarly, in incremental form,

p’ = (a’ + 2r’)/3 = p - u (82)

and q’ = a’ - r’ = a - r = q

In an undrained triaxial test, we only control the total not effective stress; the
effective stress changes depend on the behaviour of the material and the strain
increments it undergoes. To do this, express Eq. 54 in incremental form as follows:
v = p’/K’ (83)
In soils, it is usual to use the incremental form of strains because it is difficult to
define the state of zero strain. Strain history of most soils are unknown.

In Eq. 83, note that v refers to the volumetric strain increment of the soil skeleton
as well as that of the soil; i.e. they are one and the same in undrained condition.
Since the soil skeleton cannot undergo volumetric strain changes during an
undrained test, therefore v ~ 0. K’ is the effective bulk modulus of the soil
skeleton.
Since K’  , p’ = 0. In other words, in an undrained triaxial test on a linear elastic
material, there cannot be any changes in the mean effective stress p’; thus the effective
stress path (ESP) of such a material must be a vertical straight line in stress space

71
 Q ESP TSP

A’ u A
q’=q
=q
3
1

p’ = r p p,p’

Using the figure above, since q’ = q, the effective stress state must have the same
q as the total stress state at any point of the test. For instance, in the figure above,
the co-ordinates (p,q) of the point A on the TSP represents the total stress state at
an instant during the test. The only point on the ESP which has the same value of
q as A is the point A’, which is laterally offset from A. Thus the co-ordinates (p’,q) of
A’ represents the effective stress state at the same instance. Furthermore, since p’
= r = cell pressure which does not change throughout the test, thus p’ = 0.
Moreover, p = p – p0 = p - r. Thus, the excess pore pressure u is given by u =
p - p’ = p - r = the lateral offset between A and A’
In a conventional triaxial test, the sample starts off in a state of isotropic stress
since a = r; thus q0 = 0. Thus q = q – q0 = q. In other words, the change in
deviator stress is the y-coordinate of A (also A’). From the figure above, it can be
deduced that u = q/3. 72
3.2 Skempton’s Pore Pressure Equation Re-visited.
Recall the Skempton’s pore pressure equation for undrained testing:
u = B[3 + A(1 - 3)].
Or u = B[r + A(a - r)] (84)

Recall that you were told that

B=1 for an undrained test on a saturated soil, and

A=1/3 for a linear elastic material in triaxial test. Why???? Here are the
proofs using stress path parameters.

Consider a case of isotropic compression a = r, as in a B-test.

i.e. no ram load being applied, only increment in cell pressure, then
Skempton’s pore pressure equation reduces to u = Br. (86)
Furthermore, substituting Eq. 85 into Eq. 67 leads to
p = r (87)
which allows Eq. 86 to be written as u = Bp = B(p’ + u) (88)
But p’ = 0 in an undrained test on a saturated soil.

Thus
u = B u  B = 1 for an undrained test on a saturated soil. (89)
73
CIU Test Procedure from ISO 17892

74
This is also evident from the figure above, in which both u and
p are represented by the lateral offset between A and A’,
indicating that they must be equal.

Consider a case in which only the ram loading is changed, i.e. a
 0 but r = 0 e.g. in conventional triaxial compression.

Eq. 84 reduces to u = A a in view of B=1 (90)

We note that q = a - r = a, since r = 0 (91)
Which allows Eq. 90 to be written as
u = A q (92)
Since u = q/3, thus A = 1/3 for a linear elastic material

75
 3.3 Stress-strain Behaviours in Triaxial Test
Re-writing Eq. 29 in incremental form:
  x   1/E  / E  / E 0 0    x 
0
     / E 1/E  / E 0 0 0    y 
 y 
  z    / E  / E 1/E 0 0 0    z 
   

 xy   0 0 0 1/G 0 0   xy 
 yz   0 0 0 0 1/G 0    yz 
    
 
 zx   0 0 0 0 0 1/G    zx 
Note that xy = yz = zx = 0, thus xy = yz = zx = 0.
Furthermore x = z = r and y = a.
Consider a case in which only the ram loading is changed,
i.e. a  0 but r = 0.
In this case, x = z = r = 0, and Eq. 93 reduces to
r = z = - a/E, and (94)
a = y = a/E (95)
Furthermore, Eq. 68 reduces to q = a, so that Eq. 95 can be
written as q = E a (96)
Note that the E and  used here relates the incremental total (not effective)
stresses to the strain increments. They are dependent on the interaction
between soil skeleton and pore water and thus on the test conditions. 76
Undrained Test.
The total stress E and  under undrained conditions are often
denoted by Eu and u.
For undrained test, v = x + y + z = a + 2r = 0 (97)
Substituting Eqs. 94 & 95 into Eq. 97 leads to
a (1 - 2u)/Eu = 0 (98)
Since a and Eu are both non-zero, therefore 1 - 2u = 0  u =
½.
From Eq. 43, we know that Gu = G’ (99)
Substituting Eq. 28 into Eq. 99 leads to the condition that
Eu E'

1  u 1  '

Since u = 0.5, thus Eu = 1.5 E’/(1+’) and Eq. 96 becomes

1.5 E'
q   a (100)
1  '
For a linear elastic soil skeleton, E’ and ’ are constant. Thus the
deviator stress vs axial strain curve has the form of a straight line
with slope 1.5E’/(1+’), i.e.
77
 q

a
The stress path plot (p and p’-q) and stress-strain plot (q vs a) are
the normal representations of soil behaviour used in constitutive
modelling.
Do the stress paths and stress-strain relations for a linear elastic
material reflect those of real soil well?

78
 Drained Test
In a drained test, there is no pore pressure change, thus u = 0,
and
r’ = r = 0 and a’ = a and p’ = p (101)
Eq. 96 can be written as q = E’ a (102)
Moreover, v = x + y + z = a + 2r = a’ (1 - 2’)/E’
= a (1 - 2’)/E’ in view of Eq. 101 (103)
but Eq. 91 states that q = a. Thus Eq. 103 can be written
as
v = q (1 - 2’)/E’ (104)
Substituting Eq. 102 into Eq. 104 leads to
v = a (1 - 2’) (105)
Since ’ for a linear elastic material  0.5, v has the same
sign as a.
Is this a good reflection of real soil behaviour???
79
3. Triaxial Test Conditions

80
 Triaxial Test data on Real Soil

Tests 1 & 3 –
Normally
consolidated
clay.
Tests 2 & 4 –
Heavily
overconsolidated
clay.

81
Triaxial Test data on Real Soil

Drained Test. 82
Undrained Test.
Triaxial Test data on Real Soil

Drained Test. 83
Undrained Test.
 Real Soil Behaviour in Triaxial Tests
(a) As shown in the attachments, real soil behaviour differs
from linear elastic model in the following aspects:
Stress (q) – strain (a) curves are non-linear for real soils
whereas Eq. 100 would suggest that q is proportional to
a. This may be corrected by varying the value of G with
a in an appropriate fashion. Substituting Eq. 28 to Eq. 100
leads to
q = 3G a or in differential form,
dq/da = 3G (106)

Thus if we reduce G progressively with a or q, the slope of

the tangent of the q-a curve will reduce accordingly, similar
to that shown by real soil data. A common example of such
an approach is the hyperbolic model by Duncan, Chang,
Korner and others.
84
 (b) In a drained test, the normally consolidated
(nc) clay tends to undergo volumetric compression
whereas the heavily over-consolidated (hoc) clay tends
to dilate. In an undrained test, these tendencies are
manifested as positive and negative excess pore
pressures, respectively. If we examine the ESP (ie. p’-
q) for the undrained tests, we find the following trends:
 In a nc clay, the ESP curves towards the q-axis (ie. p’
reduces) whereas the linear elastic model would
predict a vertical straight line.
 In a hoc clay, the ESP curves away from the q-axis (ie.
p’ increases) whereas the linear elastic model would
predict a vertical straight line.
Both these trends are totally different from that
predicted by the linear elastic model. Can these be
remedied?? To answer this, we examine Eq. 105 for
the drained situation. In this case, the magnitude and
sign of v can be changed by changing ’. However,
this may cause a few problems since ’ should
theoretically lie between 0 and 0.5, which implies that 85
v0.
 If ’ > 0.5, then K’< 0 and it would be hard to envisage a
material with a negative bulk modulus!!! Furthermore, if
’  0.5, K’  !!! Hence, it is often not possible to
vary ’ over a wide range of values. In an undrained
test, Eq. 83 tells us that, as long as v = 0, then p’ = 0
(no change in p’, implying ESP must be a vertical
straight line) regardless of the value of K’. Hence,
the cross-coupling between a with v and p’
cannot be fully modelled even with a non-linear
model which allows K’, G’ or ’ or E’ to be varied with
strains and stresses.
 In order to overcome the above shortcomings of the
elastic (both linear and non-linear) models, elasto-
plastic models have to be used.

86
THANK YOU

87