You are on page 1of 59

NATIONAL UNIVERSITY OF

SINGAPORE
DEPARTMENT OF CIVIL
ENGINEERING

CRITICAL STATE SOIL


MECHANICS
SHORTCOMINGS OF MOHR-COULOMB
• Stress path below the yield surface (or failure envelope) is
vertical - WRONG. Methods A, B etc.. options can correct
for strength but not stress path.
• After yielding, soil behaviour is also wrongly modelled.
• If angle of dilation = 0 deg, 700
600

Deviator stress (kPa)


5deg

then no post-yield hardening. 500


dilation
angle

• If angle of dilation > 0, then


3deg
400 dilation
angle

300 2deg

post-yield hardening is constant 200


dilation
angle
Measured

(undrained condition). Soil will 100


curve

0
not fail. 0 5 10 15

• Shear behaviour cannot be Shear strain (%)

combined with consolidation/


compression behaviour.
The common e-p’ curve for a typical soil is a
statement of elastic-plastic behaviour.
Hardening Rule
Many materials e.g. steel, copper and
soil, can resist increasing stresses after
yielding.
E.g. copper yielded at point B but can still
sustain increasing stress from B to C.
Compression curve for soil: on the VCL, p’ p’
can still be increased. Known as strain-
hardening. In plasticity theory, hardening
is represented by allowing the yield VCL
surface to expand (i.e. the yield function
to change) after yielding. Cannot be
modelled by Mohr-Coulomb. In
modelling, normally achieved by changing
the values of parameters which control RL1 RL2 RL3

the size of the yield surface, or hardening


parameters. e
Mohr-Coulomb cannot model this.
Strain-Hardening Cap Models
• Developed to address above shortcomings of
Mohr-Coulomb.
• A large class encompassing many, many models
e.g. Drucker-Prager Cap Model, Cam Clay (all
variants), Plaxis soft soil and hardening Soil
Model etc.
• Impossible to cover all. In any case, quite
pointless, because the parameters for some of
the models are not easily measured using
standard laboratory tests.
• We will use the Cam Clay model as an illustration
of this class of models (because it is the simplest).
Cam Clay Family of Models
• Built around the following ideas:
– Continuous shearing will finally lead to a state of
ultimate failure known as “critical state”. Known
as critical state models.
– Isotropic compression, e.g. in triaxial
consolidation, produces an e-ln p’ curve similar to
1-D consolidation, with an isotropic pre-
consolidation pressure.
– Pre-consolidation pressure controls the size of the
yield locus i.e. a hardening parameter.
Critical State (State of Ultimate Failure)
Name given to a state of ultimate failure wherein the soil
has reached a steady state such that, with further
increase in shear strain:
(a) Effective normal stresses (or mean effective
stress) no longer changing;
(b) Void ratio no longer changing;
(c) Shear stresses (or deviator stress) no longer
changing.
It is an end point (ultimate failure) for soil behaviour. In
critical state soil mechanics, yielding is NOT failure,
which refers to ultimate failure.
First developed by Roscoe, Schofield and Wroth in the
1950s.
Introduction

Critical state soil mechanics is the name given to a


framework of looking at soil behaviour, which was
pioneered by Roscoe and his co-workers at
Cambridge University in the 1950s. It probably is
the first self-consistent attempt to characterize
stress-strain behaviour. Much of the experimental
evidence and findings in critical state soil
mechanics were obtained from triaxial tests and
this needs to be remembered at later stages.
Triaxial Test Apparatus
Triaxial Test
Schematic
Water level

sa =s1 h

specimen

sr =s3 u =gwh
The triaxial test is one of the most widely used laboratory tests for soil
behaviour. In this test, the soil specimen is subjected to adjustable
confining pressure (often constant during the conventional triaxial test) as
the behaviour of soil is dependent on mean stress level. A higher-than-zero
pore pressure (termed “back pressure”) may also be applied to the
specimen. The usual measured quantities are

(a) Radius stress sr


(b) Axial stress sa
(c) Pore pressure u
(d) Axial deformation l, giving axial strain a = -l/l
(e) Change in volume V, giving volumetric strain v = -V/V
(f) Radial deformation r, giving radial strain r = -r/r

For items (d) to (f), our convention is “compressive is positive”.

The conventional stress parameters in triaxial tests are sa and sr or their


effective stress equivalence whilst the corresponding strain parameters are
a and v (or u for undrained tests). For study on stress-strain behaviour,
it is generally found that the stress path parameters p and q are more
suitable because they are independent of orientation of the face.
3 Shear Strain

We have already defined p, q and v as three stress-strain parameters.


Since there are two stress parameters there should also
be two strain parameters. Since we have already defined a volumetric
strain v, which is a counterpart of p, we now need a shear strain s,
that is a counterpart of q. We start from the equation for work done W
due to a small increment in strain.

W = sx’x+sy’y+sz’z+xygxy +yzgyz +zxgzx (1)

Under triaxial conditions, xy = gxy = yz = gyz = zx = gzx = 0

Furthermore, sx’ = sz’ = sr’ and sy’ = sa’. The same holds for x, y and z.
Thus Eq. 1 reduces to

W = sa’a+ 2sr’r (2)

For energy to be conserved,

W = p’v + qs (3)


We know that, in a triaxial condition,

p'  13 s a '  2s r ' (4)

q = s a’ - s r ’ (5)

and v = a+ 2r (6)

Substituting Eqs. 4 – 6 into Eq. 3 and equating the latter to Eq. 2,


leads to

 s  23  a   r  (7)

It should be noted that this definition of s only holds under triaxial
conditions, under general conditions, the expression is much more
complicated.
Stress History
Soil behaviour is very dependent on previous stress history,
which refers to the changes in stress that the soil has been
subjected to in its past. For instance, certain types of
anisotropic soil behaviour is now recognized to be a result of
stress history. Stress history effects are often difficult to
quantify but can be related to the following aspects of the
structure of a soil:
• Bonding (not for remoulded soils);
• Particle alignment and arrangement;
• Packing.
The first two aspects are difficult to quantify and often
even more difficult to measure. For the purpose of
establishing a first-order representation of stress history
effects, the first two will be ignored.
Particle packing density can therefore be regarded as a
first-order approximate representation of stress history
effects, which also has the advantage that it is a scalar
parameter. Several simple measures of volumetric packing
already exist. These include

Weight of water Ww
• Water content w
Weight of solids Ws

• Void ratio Volume of voids


e
Volume of solids
In general,

Sr e = Gs w
In which Gs = specific gravity of soil particle material ~ 2.6 –
2.7.
For saturated soils, Sr = 1, so that e = Gs w

Volume of soil
Specific volume v
Volume of solid
= Volume of solid  Volume of voids
Volume of solid
Volume of voids
=1 +
Volume of solid
=1+e
We also know that volumetric strain v = -V/V
= -(Vsv)/(Vsv)
Since Volume of solids Vs is constant, Eq. (16) reduces to
v = -v/v
= - (1+e)/(1+e)
= -e/(1+e)
Common Types of Triaxial Tests and their Stress Paths
Drained Tests
Recall: sr = 0,
hence q = sa and p = sa/3 q
 q/p = 3.
Hence, total stress path (TSP) is a ESP
straight line with slope of 3. TSP

Effective stress path (ESP)


depends on the back-pressure ub.
If ub = 0, then ESP is coincident
with TSP. ub
If ub = constant and > 0,
then p’ = p - ub

The ESP is then also a straight line p, p’


with slope of 3, but laterally offset to p’ = p-ub p
the left of the TSP by the amount
equal to ub.

Thus for triaxial compression


tests, the TSP and ESP are as
follows:
For triaxial extension tests: sr = 0,
hence q = sa and p = sa/3
 q/p = 3;
 but sa < 0  q < 0, so that the TSP and ESP still
have slopes of 3 but are headed in the opposite
direction, as shown below:

ESP
TSP

-q
For triaxial compression tests by reduction of cell pressure, sa =
0 and sr < 0.
Hence q = -sr > 0; p = 23 sr = - 23 q  slope of TSP and ESP
each has a slope of 3/2.

TSP
ESP

3
2 ub

p, p'
For constant-p compression test, sa > 0, sr = -0.5sa. Hence q =
1.5sa and p = 0.

ESP TSP

ub

p' p p, p'
Undrained Test

• The TSP for undrained tests is same that those of


the corresponding drained tests since total stress is
controlled by the researcher.
• For undrained test, water cannot flow in or out of
the sample. Hence, pore pressure has to change
to prevent tendency for volume change.
• The change in pore u is dependent upon the
behaviour of the soil and is not totally under the
control of the researcher.
• As a result, effective stress s' = s - u is also not
determinate without knowing the response of the
soil sample.
State Variables and State Space

Triaxial test results are typically plotted in the conventional format, i.e.
q = sa’ - sr’ against a, and v against a.

It is often preferable, to re-plot the triaxial findings in the form of


(1) q vs p (or p’) stress space in the form of a stress path. This implicitly
assumes that p’ and q give a fairly complete description of the current
stress state of the soil. Note that this plot does not tell us anything about
the volume change or dilatancy of the soil. [We have already learned this
previously]
(2)Specific volume v (or void ratio e) vs p’ in compression space.
Essentially a compression (e-log p’) plot. Implicitly assumes that v gives
a fairly good first approximate account of the particle arrangement of the
soil.

Taken together, p’, q and v give a fairly complete description of the state of
the soil, that is, they are state variables.
We can also combine (a) and (b) and have a plot of p’ vs v vs q, which is a 3-
dimensional plot of the entire state space. We can then regard (a) and (b)
as two projections of the state information onto the p’-q space and v-p’
spaces.
1000

800
Deviator stress, q (kN/m2)

tension cut-off line

600

400

200
fitted yield locus

0 200 400 600 800 0 0.04 0.08 0.12 0.16 0.2

Mean normal effective stress, p' (kN/m2) Deviatoric strain, s


3.6
(a) (b)

CID50
3.4 CID100
CID250
Specific volume, 

CIU50
3.2 CIU100
CIU250
isotropic compressin curve
3 primary yield points
gross yield points

2.8

2.6 (c)
3-Dimensional p’-q-v space
Ideal v-p' curve
3.6

3.5
q
3.4
Specific volume v

3.3

3.2 V
3.1

2.9
0.00 100.00 200.00 300.00 400.00 500.00 600.00
p' (kPa)

P’
Another view of 3D p’-q-v space
State Paths for Isotropic Compression and Unloading
The state path for isotropic compression and unloading of a clay
looks like the below after projection onto the stress and
compression spaces.

Stress paths
After isotropic compression After isotropic unloading
Normally consolidated (NC) clay Over-consolidated (OC) clay
q

p’
V V
Compression paths

NCL NCL

URL

p’

The main feature of the compression plot is that the compression of clay
is non-linear and irreversible (=> plastic volumetric change along the
virgin compression or normal compression line NCL).

It is now well-known that some of the curvature of the v-p’ plot can be
removed by plotting it in v-ln p’ (or e-lnp’) space.
V V

NCL
NCL

URL URL

Actual plot Idealised plot ln p’


ln p’
Altogether much of the curvature may be removed by plot v against ln p’
instead of p’, nonetheless some of the curvature still remains. In addition,
some hysteresis still remains between the unloading and reloading segments
of the unloading-reloading line (URL).
To simplify the situation in our working model, we will assume that the NCL
and URL are both straight lines in v-ln p’ (or e-ln p’) space, and that there is
no hysteresis.
The Normal compression line (NCL) is therefore given by
the expression
v = V -  ln p’
This is equivalent to the equation for specific volume in 1-D
compression, that is
v = V0 – Cc lg p’

Thus, we can look at  and Cc as merely different


expressions of the compression index.  is applicable to
isotropic compression and natural log (loge) scale whereas
Cc is applicable to 1-D compression and common log
(log10) scale.

Common assumption: that the NCL is a unique line in


compression space, then V and V0 are constants.
The Unloading-reloading line (URL) is given by the
expression
v = V -  ln p’
For 1-D swelling and recompression, we normally assume
v = V1 – Cr lg p’
 and Cr are the recompression indices.

The parameters V and V1 are not constants since their


values depend on the point at which unloading takes place.
Typical Behaviour of Clay in Drained and Undrained Triaxial Tests

• Figure shows the


results of 4 tests.
• Tests 1 and 3 are on
normally consolidated
(NC) clay and is
obtained by
consolidating the clay
along the VCL.
• Tests 2 and 4 are on
heavily over-
consolidated (HOC)
clay specimens which
are obtained by
allowing the samples
to swell back under a
reduced effective
stress after
consolidation.
• All four samples have
approximately the
same water content.
Drained Test
on NC clay
=>
volumetric
compression

Test 1 is a drained test whereas Test 3 is an undrained test


Undrained
test:
volumetric
compression
is prevented,
so that the
tendency to
compress is
translated
into excess
pore pressure
increase.
HOC test:
drained
sample shows
volumetric
dilatancy.
Undrained test:
volumetric
change is
prevented, so
that the
tendency to
dilate is
translated into
negative
excess pore
pressure.

Hence the change in pore pressure (or excess pore pressure) mirrors the
prevented volumetric change for both NC and OC clay.
Compression
paths: In p’-
v space.
This is the plan view of
the p’-q-v state path
projected onto the
“floor of the p’-q-v
space.

Recall:
• For a drained test, the ESP is constrained to move in a certain way in stress space;
in this case, with a slope of 3:1.
• For undrained test, the ESP is not directly under external control. However, specific
volume v is constrained to remain constant in compression space. In other words,
the state path must move in such a way that v remains constant; that is the state
path is a horizontal straight line in compression space.
Critical State Line (CSL): a line defining the
states of the soil at the critical state i.e.
state of ultimate failure.

In compression space (i.e. p’-v space), the


state-paths of the 4 tests appears to converge
towards a single unique curve is obtained in p’-v
space, which appears to similar in shape as the
NCL. This is the projection of the Critical State
Line (CSL) onto the p’-v space (i.e. the “floor” of
the p’-q-v space).
Shape of the CSL when projected onto ln(p’) – v
or ln(p’) – w space. Recall: v = 1 + e = 1 + Gsw
Hence, v and w are linearly related.
Data from
Roscoe and his
co-workers
shows the states
of the soil
samples heading
towards a straight
CSL in w-ln p’
space, which is
approximately
parallel to the
NCL. Thus this
curve can be
given the relation
v =  - lnp’
in which  is a
constant for a
given soil.
Critical State Line (CSL) when plotted in p’ – q
space i.e. the wall of the p’- q - v space
In stress space, a single unique straight line is obtained for each type of soil. Similar
to the Mohr-Coulomb failure envelope but C = 0 => c’ = 0
This line passes through the Origin and has a slope which is
denoted by the parameter M (Roman for ), that is, it is
represented by the equation
q = Mp’

At failure, if c’ = 0, then s a 1  sin '



s r 1 - sin '
Substituting Eq. 24 into Eqs. 4 and 5 and dividing leads to
6 sin '
M
3 - sin '
Impt! This expression of M is unchanged from Mohr-Coulomb
because it only depends on the definitions of p’ and q.
M is somewhat different for different types of soils but often lies
between 0.8 and 1.2. [Cohesion c’ is theoretically 0 at critical
state or ultimate failure – no more cementation.]
3-D view of the Critical State Line. How does CSL
look like in 3D p’ – q – v space?

Red line is the


CSL in 3D p’-q-v
space.
Green line is
projection of CSL
on to the p’-v
“floor”.
Blue line is
projection of CSL
onto the p’-q
“wall”.
How to distinguish critical state (State
of Ultimate Failure)?
Critical state is the name given to a state of ultimate failure
wherein the soil has reached a steady state such that, with
further increase in shear strain:
(a) Effective normal stresses (or mean effective stress) no
longer changing;
(b) Void ratio (or specific volume) no longer changing;
(c) Shear stresses (or deviator stress) no longer changing.
It is an end point (ultimate failure) for soil behaviour. In
critical state soil mechanics, yielding is NOT failure, which
refers to ultimate failure.
First developed by Roscoe, Schofield and Wroth in the
1950s.
Data from Roscoe et al.’s early work verifying critical
state.
Since the stress and compression
plots are projections of a 3-D state
space on 2-D stress and compression
spaces, these unique lines are
projections of a unique curve in 3-D
state space; this unique curve in 3-D
space is known as the Critical State
Line (CSL).

When the soil reaches critical state:


Even if we continue to shear the
specimen, the shear strain may be
increasing; but there is no further
change of state,
=> p’, q and v are all not changing
Thus the critical state line represents
the ultimate failure of the soil (cf. peak
strength which is not critical state).

This is our definition of failure.


Conventional and critical state soil mechanics
e v

Lg p’ Ln p’
 q

s’ p’
Regimes of Soil Behaviour in CS Framework – soil
behavior on the “wet” and “dry” side of critical.

“Wet”
side of
critical
“Dry”
side of
critical
“Wet”
side of
critical
“Dry”
side of
critical
Method A issue!

Since state paths never cross the CSL, CSL provides clear divider
of patterns of behaviour, as follows:
Method A issue!
Soils wetter than critical. Positive pore pressure or
volumetric compression. Drained tests give higher strength.
Thus, for loading tests, short-term stability is critical.
Generally applicable to normally consolidated (nc) clay and
lightly overconsolidated (loc) clays with OCR < ~2. This is
where “Method A” issues are likely to arise!!

Soils drier than critical. Negative pore pressure or


volumetric expansion. Drained test give lower strength.
For loading tests, long-term stability critical. Applicable to
hoc clay with OCR > ~2.

Detailed behaviour depends on TSP. In compression tests,


drained strength is higher than undrained strength. Thus,
using undrained strength for foundation design often leads
to conservative values, OK.
Types of axisymmetric stress paths
There are an infinite variety of axisymmetric
total stress paths. One way to think about this
is to consider the rate of change radial stress
to vertical stress. Let q
s r q
K  and =
s v p
• In an isotropic situation, Δσr = Δσv and ΔK = K=-½
1, stress path is horizontal, i.e. Δη = 0. K=0
• In a standard triaxial situation Δσr = 0, ΔK =
0 and Δη = 3, stress path has gradient of 3.
• If we have a situation in which Δσr = -½Δσv
i.e. ΔK = -½, then Δη = ꝏ, stress path is
vertical. Effect pf increasing
• Hence, ΔK increases from -½ to 1, the stress K on stress path
path changes from vertical to a horizontal
stress path.
• In the soil beneath a footing, ΔK usually lies
between 0 and 1, i.e. it is somewhere K=1
p
between standard triaxial and isotropic.
Hence the p-q stress path will have a
gradient somewhere between 3 and 0.
Assumption:
(a) Loading is undrained.
Indicative stress paths
Short-term stress path is
undrained.
beneath pile tip.
(b) Short-term stress path
may not reach failure q
envelope because of
factor of safety being
applied.
(c) Loading is followed by
consolidation. During
consolidation, short-term
stress path migrates
towards long-term stress
path. Consolidation stress
(d) During this process of path
migration, stress path
Long-term
moves further away from stress path for
failure envelope => the 0<K<1
Possible short-
pile becomes safer.
term stress
paths
p, p
In active stress relief tests (i.e. failure induced by reduction in lateral stress, e.g.
deep excavation), drained strength is lower than undrained strength. Thus, the
use of long-term or drained parameters for deep excavation is generally
conservative. [but not always true for loc or nc clays]
Critical and Residual States
• There is some confusion over the meanings of critical and residual
states. The problem is probably compounded by some publications
(e.g. Lambe and Whitman 1973) which explicitly treats the critical
state to be synonymous with the residual state. For example,
excerpts from Lambe and Whitman (1973) state (erroneously) that
• “… the direct shear test is widely used to study the strength of an
overconsolidated clay at very large strains, i.e., the ultimate
strength or residual strength.” p. 302.
• and that
• “… Skempton used the phrase residual strength; we shall retain the
phrase ultimate strength. Roscoe et al. (1958) refers to this
ultimate condition as the critical state.” Footnote on p. 312.

This is extremely unfortunate as the critical state is totally different


from the residual state. It is also potentially very significant because
of the increasing use of the critical state concept in design (e.g.
BS8002: 1994). The table below highlights the differences between
critical and residual states, in terms of philosophy, conditions,
materials, strain required and testing requirements.
Difference Critical State Residual State
Philosophy Interlocking broken, void ratio (or Packing density no longer changing.
specific volume) has reached a Particle alignment is along direction
steady state in drained test, pore of slippage or shearing.
pressure has reached a steady state
in undrained test. Packing density
no longer changing. Particle
alignment is random.
Material All materials, sand silt and clay. Only exist in soils with flat platy
Normally and over-consolidated. particles such as clayey soils
Loose and dense. Do not require containing >50% clay. Usually only
slip plane formation. in stiff, over-consolidated clays
which shear along narrow slip
planes. Sands do not exhibit
residual states.
Strain Usually occur at between 2% to Usually require a few hundreds or
20% shear strain. thousands of % strain to develop.

Test Can be observed and has been Cannot be observed in triaxial tests
observed in triaxial tests. because strain not large enough.
Requires a reversing direct shear
test or a ring-shear apparatus to
achieve the required strain level.
Relevance In all situations apart from slippage Only applicable in situations where
along an old or pre-existing slip an old or pre-existing slip plane from
plane e.g. old landslips. an old landslip is present.
In general, the residual angle of friction is very much lower (quite
often less than half) than the critical state angle of friction, see
Figure below. In fresh ground which has not suffered any previous
landslides, field evidence (e.g. Skempton 1970, Chandler 1974,
Chandler & Skempton 1974) shows that the residual strength has
no relevance, and that the critical state strength is a very good
parameter to use.
One should be very selective when using the
residual strength. As Bolton (1983) noted,
“… The fact that r’ can be very small means
that no engineer can afford to apply it to
every clay design problem simply to be safe:
he would needlessly bankrupt his clients. The
problem of r’ becomes the problem of
identifying sites which might contain the
hidden wounds of a past landslide, or of
locating the ancient slip surface if it exists,
and removing a sample containing it in order
to determine the friction angle that can be
mobilized upon it by conducting a
conventional slow drained shear box test….”.
Remoulded strength from vane
shear
THANK YOU

You might also like