INTERNATIONAL SOCIETY FOR

SOIL MECHANICS AND

GEOTECHNICAL ENGINEERING

This paper was downloaded from the Online Library of

the International Society for Soil Mechanics and
Geotechnical Engineering (ISSMGE). The library is
available here:

https://www.issmge.org/publications/online-library

This is an open-access database that archives thousands

of papers published under the Auspices of the ISSMGE and
maintained by the Innovation and Development
Committee of ISSMGE.
 7/4

Some Effects of Sample Disturbance on Soft Clay

Quelques Effets de la Perturbation des Echantillons dans une Argile Molle

J.H. ATKINSON The City University, London, United Kingdom

L.M. KUBBA

SYNOPSIS Sampling of soil causes distortions and stress changes which m a y alter the mechanical behaviour of the soil.
The effect of sampling on the strength of soil has been the subject of previous research but less is known about the
effect on stiffness and deformation. The paper gives the results of tests on virgin soil and on samples with different
degrees of disturbance in which stress-strain behaviour was carefully observed. Various procedures for minimising
sampling disturbance and for reconsolidating samples before testing are discussed.

The response of a soil sample to a given increment of

DISTURBANCE DUE TO SAMPLING
load (i.e. its stiffness or its stress-strain behaviour)
Soil samples are most often obtained by taking tubes from
depends on the fundamental properties of the soil but it
bore holes or by hand cutting blocks from test pits. In
depends too on the current state of the soil and also,
either case the total stresses change from those in the
possibly, on the previous loading history. If sampling
ground, to zero as the sample is trimmed, and finally to
has the effect of changing the state of the soil as well
those applied by the testing machine with corresponding
as changing its history without changing the fundamental
changes of pore pressure. This sequence of total stress
properties it m a y be possible to recover the virgin
changes may or may not change the behaviour of the sample
behaviour by a suitable reconsolidation procedure before
but, in conventional practice, it is unavoidable. During
the test increment is applied but, if the properties are
tubing, where a smooth tube is pushed into the ground,
s omehow altered by sampling, it is difficult to see now
soil must be displaced to make way for the wall of the
the true v irgin behaviour of soil can be determined from
tube and soil inside the tube becomes distorted to a
laboratory samples. Thus, it is of importance to examine
degree determined largely by the ratio t/d of the
not only the magnitude and causes but also the nature of
thickness t of the wall of the tube to its diameter d.
sampling disturbance.
For a carefully hand cut block there need be no
d istortion as soil is cut away from around the block and
a hand cut block may be regarded as the same as a tube
sample taken with a tube with infinitely thin walls
LABORATORY TEST PROCEDURES
(i.e. t/d = 0 ) . If any property of a soil sample is
different to that of the soil in the ground the sample is
The investigation was carried out using spestone kaolin
disturbed and the properties of soil of most interest to
(LL = 51 PL = 30) in a conventional strain controlled
the geotechnical engineer are strength, stiffness and rate
triaxial machine modified to allow stress controlled
of drainage. The effect of sampling disturbance on the
loading and in a special stress controlled triaxial
strength of soil was examined by Skempton and Sowa (1963)
apparatus. Details of the apparatus and test procedures
and in this paper we will examine some effects of sampling
are given by Kubba (1980). Virgin samples were prepared
disturbance on stiffness and deformation.
from slurries by anisotropic consolidation to a stress
2 -2
A sample of soil which is completely undisturbed is known state a* = 276 kN/m and a/ = 172 kN/m with a value
v n
as a v irgin sample for it has the same properties as the
of K 0 ■ 0.63 corresponding to one-dimensional compression.
virgin gound. A virgin sample of an artifical soil m a y be
Perfect samples were prepared from virgin samples without
made in the laboratory by one-dimensional consolidation of
removing the sample from the apparatus by reducing the
a slurry and a virgin sample made in this way may be used
total stresses to zero for a period of 1 2 hours with no
as a standard for assessing the effects of various degrees
drainage allowed and then replacing the original total
of disturbance. A virgin sample unloaded undrained to a
stresses. V irgin and perfect samples were all nominally
state of zero total stress and reloaded undrained without
36 m m dia. by 76 m m long. Samples for tubing were 100 m m
removal from the apparatus is known as a perfect sample
dia, and were prepared by the same anisotropic consolid­
for it has suffered no disturbance other than the
ation procedure as that used to make the virgin samples
unavoidable cycle of total stress associated with all
and tube samples were taken from the 1 0 0 m m dia. Sample
practical sampling procedures. Tube samples ma y be taken
using 38 m m dia. tubes with different wall thicknesses.
from a virgin sample using tubes with various values of
Virgin samples, perfect samples and tube samples were
t/d and a perfect sample is the same as a tube sample for
subjected to a number of different tests but, for the
which t/d ■ 0. By comparing the properties of virgin
present, we will consider only the states of the samples
samples, perfect samples and tube samples it is possible
immediately after sampling and initial recompression and
to separate the unavoidable effects of the total stress
their stress-strain behaviour in stress controlled un­
changes f r o m the avoidable effects of distortion caused
drained triaxial compression tests. We will not discuss
by tubing.
the states of the samples at failure as these states were
not well defined in the stress controlled tests carried
out.

423
7/4

at I or S from the state at V is a measure of the sample

The tests were carried out relatively slowly to allow
disturbance and it is convenient, following Okumura
comparison with similar drained tests which formed
(1971), to define a degree of disturbance D = l- (p’/p^)
another part of the research programme (Kubba 1980) and
to allow equalisation of pore pressures and the normal where p ’ is taken as p.* or p' so that D = 0 when p ’ = p'
M *s Y *v
duration of a test was approximately 10 days.
and D = 1 for completely reconstituted samples for which
p T = 0, Table 1 shows measured values of D for kaolin
for perfect samples and for tube samples with different
values of t/d.
CHANGE OF STATE DUE TO SAMPLING

The state of a sample must be described both by the

effective stresses and by its volume. For triaxial
TABLE 1
samples for which = cr* it is convenient to choose the
parameters q : p' : v where q and p ’ are the stress
invariants q* =(a.' -a') and i Thin tube Thick tube
' = "3' ^°1 + 2a3^ and V Sample Virgin Perfect
is the specific volume. 0.039 1 - 0 . 1 1

State
(Fig. 2) V I S I I
s v
Disturbance
D 0 .30 .18 .44 ,61

Ps Pi Pi P

These data show that even perfect sampling causes

appreciable disturbance and that tube sampling causes
further disturbance.
Fig, 1, State Paths During Sampling

RECONSOLIDATION
The state of a sample may be represented by a state path
plotted with a x e s q' : p' and v:lnp' as shown in Fig, 1.
Before testing a sample may be reloaded in the triaxial
Thus Fig. 1 shows effective stresses and the total and
apparatus without drainage or it may be reconsolidated.
effective stresses are related by q ’ = q and p 1 = p - u
In the first case there will be no volume change but the
where u is the pore pressure. The point V represents
effective stresses will differ from those in a virgin
the state of virgin soil and OV is the state path for
sample while, in the second case there will be a change
anisotropic consolidation. If there is no drainage
of volume. Thus, in general, it is impossible to obtain
during sampling and preparation the specific volume of
a disturbed sample with the same effective stress and
saturated soil remains constant and the approximate state
specific volume as a virgin sample, and if soil behaviour
paths may be sketched as in Fig. 1. The point I
depends on its state, soil properties cannot be obtained
represents the state of a sample during trimming when
directly from tests on disturbed samples even if they
the total stresses are zero; the pore pressure is u.
are reconsolidated to the virgin effective stresses.
which has a negative value and the effective stress is
p!. The point I also represents the state of a
saturated sample with an isotropic total stress state
and thus I represents the state of stress at the start
of a conventional unconsolidated test. The point S
represents the state of a sample in the apparatus with
total stresses the same as those in the ground but with
a pore pressure ug which may not be the same as the
pore pressure uv at V. Since q' is independent of the
pore pressure, q^ = q^ but since the pore pressures at
Fig. 2. Reconsolidation of Samples.
V and S may be different we will, in general, have
p:
s * pf*
rv
Comparing the samples at V, at I and at S, the state is
The state of a virgin sample of normally consolidated
different and the previous loading history is different
soil is at V in Fig. 2. The state of an unconsolidated
for each and thus the stress-strain behaviour of each is
disturbed sample with the same total stress as the virgin
likely to be different. In Fig. 1 the point V is shown
sample is at S and the state of an unconsolidated
for normally consolidated soil and the point S represents
disturbed sample with an isotropic stress state is at I,
an overconsolidated state and the difference in behaviour
The state of a disturbed sample reconsolidated to the
of the two samples will be appreciable. If the state at
effective stresses in the ground is at C but its specific
V is overconsolidated the state at S will still, in
volume is less than that of the virgin sample. Disturbed
general, be different but the differences in the stress -
samples at K and L
strain behaviour will be less. The distance of the state

424
 7/4

have been anisotropically reconsolidated to effective According to the principle of effective stress, strains
stresses larger than those for the virgin sample at V but are related to changes of effective stress and hence the
with the same value of K . Comparing the disturbed stress-strain behaviour of two samples cannot be the
o
same unless their effective stress paths are the same.
samples with the virgin sample those at S and I have the
Thus samples are disturbed, and their undrained stress-
same specific volume but different effective stresses
strain behaviour will not be the same as that of a virgin
while those at C, K and L have the same value of K but
9 o sample if the stress path is different to that for a
different specific volumes. The disturbed sample at L virgin sample.In Fig 3 the paths for perfect samples
is like the sample at V for its state is on the normal initially at I, S and C are clearly different to the
consolidation line and its immediate stress history was path for the virgin sample V but the path for the
anisotropic consolidation along the line OVKL, perfect sample reconsolidated to K appears to be
geometrically similar to that of the virgin sample. In
Fig. 4 none of the paths for tubed samples are similar
to the path for the virgin sample. The paths for all
TEST RESULTS the samples, except that for sample reconsolidated to K
in Fig. 3, are approximately linear indicating that
Fig. 3 shows effective stress paths for undrained these samples behaved as though they were overcon­
triaxial compression tests on a virgin sample and on solidated, It should be noted that the behaviour of the
perfect samples with different degrees of reconsolidation samples reconsolidated to the same effective stresses as
and Fig. 4 shows the corresponding stress paths for the virgin sample was rather different to the behaviour
samples taken with thick and thin walled tubes. of the virgin sample even though, by definition, the
degree of disturbance was zero for each sample. In order
to compare directly the paths for the samples V and K in
Fig. 3 the data may be normalized with respect to the
equivalent stress p^ as discussed by Atkinson and Bransby
(19781. In Fig. 5 the two sets of normalized data fall
close to a unique path indicating that the behaviour of
the two samples was fundamentally the same.

Fig.3. Results of Undrained Triaxial Compression

Tests on Perfect Samples

Fig, 5. Normalized State Paths for Perfect Samples

Fig. 6 shows the change of deviator stress 6q' plotted

against axial strain ea for the virgin sample and for a
sample reconsolidated to K and these are clearly
different

p' KN/ m^

Fig. 4. Results of Undrained Triaxial Compression

Tests on Tubed Samples

425
 7/4

However, even after reconsolidation the effective stress

path of a tubed sample is still different to that of a
virgin sample. Thus, if a perfect sample is recon­
solidated to a state equivalent to that of a virgin
sample - i.e. both states are on the anisotropic normal
consolidation line - the fundamental behaviour of the
perfect sample is the same as that of the virgin sample
although the values for some soil parameters remain
altered. Further work is necessary to determine whether
all the properties of the virgin sample can be recovered
if samples are reconsolidated beyond K to L in Fig. 2
and to determine how far L should be beyond K.

The research reported here is concerned with undrained

loading of normally consolidated soil and virgin samples
Fig. 6. Stress-strain Curves for Virgin and
were prepared under laboratory conditions. Additional
Reconsolidated Perfect Samples
drained triaxial tests (Kubba 1980) show that the
findings for undrained loading are generally applicable
also for drained loading. Further work is necessary
however to investigate the effects of sampling
Since the initial states V and K of both samples are on
disturbance on natural soils which may have features not
the anisotropic normal consolidation line in Fig, 2
properly represented in laboratory consolidated samples.
both are normally consolidated and their behaviour will
be inelastic, but in order to compare stress-strain
curves it is convenient to define an undrained stiffness
modulus at a load factor of 2 where 6q* = 0.5 <5q^,
ACKNOWLEDGEMENTS
Scaling from Fig. 6 for the virgin sample E = 3800
2 u 2 The work described forms part of a programme of research
kN/m and for the reconsolidated sample E
= 1900 kN/m.
u into the stress-strain behaviour of soft clay supported
Thus, although their normalized stress paths are the by the Science Research Council and was carried out in
same the stiffness of the reconsolidated perfect sample the Department of Civil and Structural Engineering,
is appreciably less than that of the virgin sample. It University College,Cardiff. The authors are grateful to
is of course inappropriate to compare the stress-strain Professor K.C, Rockey for provision of laboratory
curve of a virgin sample with those of other disturbed facilities.
samples since their effective stress paths are funda­
mentally different. Stress-strain curves may be norm­
alized with respect to p* and v as discussed by Kubba
(1980) and Fig. 7 shows normalized stress-strain curves REFERENCES
for the virgin sample and for a sample reconsolidated
to K. Atkinson, J.H. and Bransby, P.L. (1978). The Mechanics
of Soils, McGraw-Hill, London.

Kubba, L.M. (1980) The effect of sampling disturbance on

the deformation of clay. Ph.D. Thesis, University
of Wales.

Okumura, T. (1971).The variation of mechanical properties

of clay samples depending on its degree of
disturbance. Proc, 4th Asian Conf. on Soil Mech.
and Foundation Eng. Bangkok, pp. 73-81.

Skempton, A.W. and Sowa, V.A. (1963). The behaviour of

saturated clays during sampling and testing.
Geotechnique 13, pp. 269-290.

Fig, 7, Normalized Stress-strain Curves for

Virgin and Reconsolidated Perfect Samples

DISCUSSION

The experimental results show that even perfect

sampling causes disturbance which alters the initial
state of the sample and the undrained stress path and
that tubing causes additional disturbance, A perfect
sample reconsolidated back to the anisotropic normal
compression line has the same normalized effective
stress path as a virgin sample but the stiffness of '
the perfect sample is considerably less than that of
the virgin sample.

426