You are on page 1of 16

SOILS Vol. 15, No. 1, Mar.

1975
Japanese Society of Soil Mechanics and Foundation Engineering

UNDRAINED DEFORMATION AND LIQUEFACTION


OF SAND UNDER CYCLIC STRESSES

KENJI IsHIHARA *, FuMIO T ATSUOKA ** and SusuMu YASUDA***

ABSTRACT
The key findings in the previous studies on the deformation of sand are first reviewed to
the necessary extent. Then, in summary form, five fundamental postulates are introduced as
the bases on which a model of undrained deformation of sand under cyclic loading is to be
established. The procedures for assessing pore pressures, shear strains and consequent
occurrence of liquefaction during cyclic loading are illustrated on the diagrams based on
the above postulates as well as actual data obtained in the static triaxial tests. The und·
rained performance of sand predicted by the proposed model were compared with several
cyclic triaxial test results conducted under various conditions. These include stress and
strain controlled tests; static and dynamic tests; and the tests cycled with constant and
erratically changing stress amplitude. The comparison disclosed the relevance and appli·
cability of this method of approach.

Key words: dilatancy, liquefaction, repeated load, sand, stress path, triaxial compression
test
IGC: D6/D7

INTRODUCTION
Increasing awareness of danger caused by liquefaction of sandy ground during earthquakes
has aroused the need for investigating into the mechanism in which loss of strength is
realized as a result of progressive development of pore pressures within saturated sands.
This aspect of the problem was studied by Shibata~Yukitomo-Miyoshi (1972), who for·
mulated the law of pore pressure increase under cyclic stresses based on the assumption
that the residual pore pressure in each cycle is all equal until the onset of liquefaction. By
analyzing effective stress paths, Tanimoto (1971) proposed that liquefaction occurs at the
instant when the effective stresses come to satisfy the failure criterion specified by the true
angle of internal friction of the sand. Somewhat formal procedure was suggested by Iwa·
saki (1971) in which the feature of pore pressure development and consequent liquefaction
was resembled to the process leading to fatigue failure of metals. Although these studies
have added to depth in the understanding of the mechanism in which liquefaction is induced,
it was felt necessary to base the studies on the more fundamental characteristics of soil
deformations. On the other hand, detailed studies have been underway in the laboratory
of University of Tokyo concerning the deformation characteristics of a sand using static
* Associate Professor, Department of Civil Engineering, University of Tokyo, Bunkyo·ku, Tokyo.
** Research Engineer, Structure and Bridge Division, Public Works Research Institute, Anagawa.
Chiba.
*** Graduate Student, Department of Civil Engineering, University of Tokyo, Bunkyo-ku, Tokyo.
Written discussions on this paper should be submitted before January 1, 1976.

This is an Open Access article under the CC-BY-NC-ND license.


NII-Electronic Library Service
30 ISHIHARA ET AL.

triaxial shear test apparatus. These studies have disclosed four important aspects which are
considered most fundamental in dealing with sand deformations: dilataricy law, shear strain
law, yield condition, and anisotropy in yielding. These studies are described in consider-
able detail in a series of the previous papers (Tatsuoka-Ishihara, 1973, 1974a, 1974b). As
a result of these investigation it was found that, with the addition of one more postulate
specifying onset condition of liquefaction, the above studies can be integrated into a model
by which pore pressure, liquefaction as well as stress-strain relations can be described in an
unified fashion. It is the objectives of this paper to present this aspect of the investigation.

APPARATUS AND MATERIALS


All tests were conducted using triaxial shear test apparatus with samples 5 cm in diameter
by 10 cm in height. In static tests, Fuji river sand, and in dynamic tests, Niigata sand
were used, and their physical properties are shown in Table 1.

Table 1. Physical properties of sands

D1o(mm) G* emax emin Uc**

Fuji river sand 0.22 2.68 1.08 0.53 2.21 Subangular

Niigata sand 0. 13 2.67 0. 99 0.55 2.34 Subangular


-
* Specific gravity ** Uniformity coefficient

Loose samples are obtained by spooning freshly boiled sand into the mold filled with de-
aired water. A dense state of packing was achieved by mechanical vibration and tamping.
The stress parameters used here are the mean effective principal stress, p' = (<1a' +2<1/) /3
and the shear stress, q=(Ja' -(Jr', where aa' and a/ are the axial and radial effective stres-
ses, respectively. The shear strain r is given by ea-er and volume change V by (ea+ 2er) /3
where ea and er are the axial and radial strains, respectively. r is defined as positive
when a sample is shorten in length.

WORK HARDENING OF SAND


Typical behavior of sand in an undrained triaxial compression test is shown in Fig. 1.
A loose sample having a void ratio of 0. 749 was subjected to several cycles of unloading and
reloading before it was sheared excessively. Both shear strain and effective stress path
indicate that the reloading process traces the same path as that along which unloading has
been executed, if the stress ratio is kept below a certain limit. When we confine our
attention to the behavior below this limit, it is known that, whenever plastic yielding has
occurred, the effective stress is reduced and residual pore pressure develops. Consequently,
if the pore pressure were to develop in any stage of cyclic deformation, it ought always
t9 be associated with the occurrence of deformation which is plastic in nature. It is on
the basis of this point of view that the following presentation is undertaken regarding the
mechanism of development of dynamic pore pressure.

UNDRAINED STRESS PA TH
By means of tamping or vibration, a sand can be brought to any desired state of pack-
ing between the maximum and minimum densities. When a sample having been compacted
to any one of these densities is subjected to stresses, it exhibits plastic yielding from the
very begining of loading, and therefore each of initial states having different densities can
be considered to define a virgin-state. Thus, there could exist an infinite number of vir-

NII-Electronic Library Service


UNDRAINED DEFORMATION AND LIQUEFACTION 31

"'E' 5 1-----+-----+-----l-
..I( e=0·749
en
6
d 4 1-----1------r----

(/)
IJl
3
(LI
!.....
~~
(/)

L 2
0
(LI
..c (a)
I.fl

12
0 1 3 4
effective mean principal stress
p (kg/en{)

e= 0.749
P~=2.250 kg;crrf
"E 4 1--------+----r----+-----+---_,.,=-r----1
~
Ol
.:><::

IJl
I.fl
(LI

.!::; 2
Ul

L (b)
d
(LI
..c
Ul

6
2 4 6 8 10 12 14
shear strain r (°lo)
Fig. 1. Undrained test on a loose sample cycled with varying
amplitude of stress ratio
gin-states in a sand. Now, prepare a sample having one of these initial densities and sub-
ject it to a stress system consisting of effective mean principal stress, p', and shear stress,
q. Then, some changes will occur in the density or void ratio, e, of the sample. If these
changes are plotted in a space having p', q and e as cordinates, it is possible to construct
a vector which defines a state path in p'-q-e space. Now, prepare several samples with a
given initial density which has been compressed under a given value of p'. Perform a set
of triaxial shear tests all on the samples of this state by employing different state paths.
Repeat the same set of triaxial tests on the samples now prepared under different values of
p' but with the same initial density. Then, we have a number of curved vectors in p' -

NII-Electronic Library Service


32 ISHIHARA ET AL.

q-e space all rising from points lying on a compression curve in p' -e plane. These curved
vectors defining state paths do not generally lie on one common surface in p' -q-e space.
However, if another curve on p'-e plane termed base line is properly chosen in the imme·
diate vicinity of the compression curve, and if the aforementioned curved vectors are slightly
modified so that they rise from the base line, a surface can be defined in p' -q-e space such
that all the curved vectors defining state paths lie on this surface. This surface may be
called state surface. The state surface thus constructed can exist for each of given initial
states of packing of sand and, therefore, there exists an infinite number of such surfaces
stratified in p' -q-e space. These findings are described in detail in the previous paper
(Tatsuoka-Ishihara, 1973). It is very important to recall that a state surface specifies the
route or path in. p'-q-e space along which stresses must be changed in order for the defor·
mation to be plastic. The plastic yielding in sand can occur only when stresses are changed
along the paths lying on the state surface, and all other paths away from it are associated
merely with elastic deformations.
In the case of undrained tests in which volume change is inhibited, the state paths along
which stresses are to be altered are the curved loci formed by the intersection of e= const.
plane with the state surface. Shown in Fig. 2 are families of state paths(or stress paths)
derived in this way from both triaxial compression and extension: on loose samples. Note
that there are many state paths for a given void ratio, because of the multiplicity of the
state surface. The undrained stress. paths derived above are based on the concept of the
state surface, but for practical purposes·, they may be obtained by carrying out a series of
triaxial undrained tests with different initial confining pressures. Also shown in Fig. 2
are the unloading stress paths which are due to elastic recovery of the deformation.

3 12
I - : loose sample
loose sample/ e=0.72-0.78
e=o.n-o.7a I
10
2
'Eu c
0
Ol
-"' 'iii "E' 8
ID ~

~
O"
01
-"'
0
u O"'
U)
U)
(lJ
6
~
L
tl (lJ

L
0 L
0 tl
(lJ
L
4
.c CJ
Vl
c (lJ
0 .c
(/)
-1 ~ ------t-i.........,,.'---f-·~
2x 2
(lJ

-21..-.---'---...;....i.--~"-----'
0 3
effective mean principal stress effective mean principal stress
p (kg /cm') p' (kg/cm')
· Fig. 2. Typical undrained stress paths Fig. 3. Yield loci for loose
for loose samples (Fuji river sand) samples

YIELD CONDITIONS
When stresses are changed during cyclic loading with irregular amplitude, it is highly
necessary to have criterions by which to decide, in any stage of the given stress history,
on whether the current deformation concerned is plastic or merely elastic. This criterion

NII-Electronic Library Service


UNDRAINED DEFORMATION AND LIQUEFACTION 33

is furnished by a family of yield loci which are depicted in p' -q space. Some studies have
been made by Tatsuoka and Ishihara (1974, a) to define families of yield loci for different
densities of the sand; The results of the tests performed exclusively in the domain of
triaxial compression have shown that, for any given density of a sand, yield loci in p' -q
space can be uniquely determined irrespective of the stress history to which the sample has
previously been subjected. Some of the yield loci obtained for the yielding in triaxial
compression are shown in Fig. 3. Several test results have also shown that the yield loci
on the side of triaxial extension have similar characteristics. When one looks at the yield
loci in Fig. 3, it is realized that, for the effective mean principal stresses below about 2. 0
kg/cm 2 , the yield loci are almost straight lines radiating outwards from the origin. The
approximation in this regard will be conveniently used in the later section.
The yield loci are interpreted as the curve at which yielding begins whenever stresses
are altered across it, and therefore, when stresses are changed in the lower side of the cur-
rently effective locus, the associated deformation is elastic.

SHEAR STRAIN
Several series of triaxial compression and extension tests in which sand samples were
subjected to complex sequences of stress changes have indicated that the shear strains occur-
ring with plastic yielding are a function of present state of stresses and density alone and
determined irrespective of any stress history the sample has ever sustained insofar as the
amplitude of stress changes is kept ·within
prepeak level (Tatsuoka-Ishihara, 1973).
loose sample
e = o.n - 0.78 It was also shown that, if those points in
6 p' -q space which yield the same shear
number show
shear strain, 5 strain are linked together, we can define
5 (°lo) r 3
what may be called equi-r-lines. Shown
in Fig. 4 with data points are typical ex-
4 2
amples of the equi-r-lines for loose samples
of a sand obtained by several tests in which
c different stress paths were employed.
0
·~
2
O"
r:
0..
O"
(a)
'ta
(/) E
;,
tf)
g] 1 0
.u
(ii
....
I...
Vi '/A
1n
0
(ii
..c.
tf)

c
0 p'
'(ii
-0.5 c mean principal stress
(!)
-2 -1 x
(!).
.:e-
O"
(b)
loading
--3 o l'.1r= canst.test -2 ~
II

• p:::const. ,, -5
s"11s I

4 e "'IA I
- -"'04:::const. ,,
iR
(])
~
I
0 undrained "
-5 .__ failure line _ __,
_ _.___........._ _..___._.....__........_ .... I
Vi I
I
0 2 3 4 5 6 '/
effective mean principal stress P(kg/cm~) '(A ¥e
shear strain
Fig. 4. Equi-r-lines for loose samples (Fuji Fig. 5. Implication of equi-r-
0

river sand) lines

NII-Electronic Library Service


,34 ISHIHARA ET AL.

The equi-r-lines shown in Fig. 4 suggest that they may be approximated with a good
.accuracy by a set of straight lines radiating outwards from the origin in p' -q space. This
approximation helps greatly to simplify the procedure. When this type of contour lines is
made available, it is an easy task to evaluate shear strains in any stage of complex time
history of stress alteration. Shown in Fig. 5 are schematic pictures to illustrate the pro-
cedures. Two adjacent equi-r-lines on which shear strains rA and rs are to be produced
.are shown in Fig. 5(a). It is to be recalled that, according to the definition of the equi-
r-lines, the shear strain that are produced when stresses are changed from point A to B
through an arbitrary path must be equal to the shear strain which is developed during the
:stress change from A' to B'. This implies the fact that the shear strain in the sample of
given density is a function of stress ratio, q/p', alone and independent of p'. Consequently,
the plot in p' -q space can be converted to the stress ratio-shear strain relation as shown
in Fig. 5(b), in which the elastic rebound curve resulting from unloading is also indicated.
Using these two sorts of diagrams interchangably, the shear strains during cyclic loading
-can be evaluated.

DIRECTIONAL INDEPENDENCY OF YIELD CONDITION


In a series of triaxial drained tests in which samples of sand were sheared in triaxial
-compression after being subjected to a small cycle of stress change in triaxial ex ten ion, it
was demonstrated that the deformation characteristics in the triaxial compression were
.almost the same as that which the samples would exhibit if they were not subjected to a
previous cycle in extension. Rephrased, a memory of previous stress history experienced
.during the low-amplitude cycle in extension does not appear and the sample shows yielding
from the very begining of loading in triaxial compression as if it were in a virgin state.
Similar behavior was also observed in the triaxial extension tests which were conducted
.after samples were subjected to a small cycle in triaxial compression. Also in the case of
undrained tests, it was shown that the current behavior of sand during straining in one
·direction is not affected to any significant extent by a previous cycle having been applied
in the other direction. Thus, it may be mentioned that the current deformation charac-
teristics as a virgin sample is not wiped off, whether the sample is loose or dense, . even
.after it has sustained some stress history in an opposite direction, insofar as its amplitude
was kept within some limit. The test results and related discussions leading to the above
consequence are described in detail in the previous paper, (Tatsuoka-Ishihara, 1974 b).

INST ABILITY CONDITION


The effective stress path in cyclic undrained triaxial compression test shown in Fig. 1 (a)
indicates that, when the stress ratio is raised to a great value, a very unusual behavior
.appears: the magnitude of pore pressure which develops in response to the subsequent un~
loading becomes surprisingly large, leading to instability of the sample. In order to look
into this aspect more closely, some additional tests were conducted of which the results
are demonstrated in Fig. 6. In one of the triaxial compression tests, the application of
shear stress causes the stress path to go first counterclockwise as shown in Fig. 6 (a), but
later to bend at point 3 over in the direction of a constant q/p' ratio path. Upon removal
of shear stress at point 4, the sample tends to develop a considerable pore pressure, but
much more radically as soon as the sample is brought into the state of triaxial extension.
In accord to this, shear strains behave as shown by point 4 to 5 to 7 in Fig. 6 (b). The
similar trends are also noted in the undrained path for a triaxial extension test as indicated
by points 1, 2', 3',. to 6 in Fig. 6. Also shown in Fig. 6 is an undrained stress path for

NII-Electronic Library Service


UNDRAINED DEFORMATION AND LIQUEFACTION 35

-.e::0.756
o-o e:0.774
~ A-6 e:0.757
2.0
u
01 2.0 ·a.
.:.:; (b) ~ ________ ..________.,.4
c:r
0

<fl
<fl
-·-..·· ·-·-·-----~~
2 i.o
.... 1.0
(I.I
<fl
<fl
Vi (lJ
1..
I
L
5 Ul
d
OJ 5
L
<fl 0 -1.0 9.0

..- e:0.756
o--o e= 0.774
t:r-A e = o.757

t--t-----1.Qt+-------l

1.0 2.0 3.0


effective mean principal stress
P (kg/cm')
Fig. 6. Undrained test results illustrating onset condition of liquefaction

still another test in which triaxial compression is stopped midway at point 2 and brought
into extension. In this case, the sample does not develop appreciable pore pressure. It is
of interest to notice in Fig. 6 that, whether samples develop excessive pore pressures or
not during unloading depends largely on the maximum shear stress from which stresses
have been removed. Careful .observation of Fig. 6 as well as Fig. 1 seems to suggest that
there is some threshold value in the stress ratio under which stress removal is involved
merely in usually defined elastic rebound, but beyond which excessive pore pressure develops
upon removal of loads. Close scrutiny of a number of test data at hand, together with
the results in Figs. 1 and 6 has indicated that the stress ratio corresponding to a point at
which stress paths turn its direction in p' -q space may logically be chosen as the thresh·
old separating two different behaviors which would be manifested in the subsequent
unloading and loading process in the other direction. As explained before, removal of
loads from stress ratios below this threshold would bring only elastic recovery and even if
the stress ratio exceeds again that previously applied, the deformation would remain of
plastic work-hardening type. Hence, unless the stress ratio goes beyond the threshold, it
may be said that the deformation produced through any stress history, be it simple or
complex cycles, could be of plastic work-hardening type, thereby preserving the characte·
ristics as a solid body. Conversely, once the stress ratio greater than this is applied to the
sample, excessive pore pressures develop during subsequent unloading, and loading in the
other direction, and if cyclic stress application is continued further the effective confining
pressure would become equal to zero, bringing the sample to a liquefied state. Although
the passage of the threshold stress ratio does not always lead to complete liquefaction, it
is necessary for a sample to go at least once through this critical value in order to be taken
to a completely liquefied state. In the sense as described above, the critical stress ratio
may be considered as a threshold at which the behavior of sand as a solid is lost and trans·
formed into that of a liquefied state. For this reason, the angle defined by the threshold
value may be called an angle of phase transformation. In the following presentation, it

NII-Electronic Library Service


36 ISHIHARA ET AL

will be assumed that liquefaction could occur when the stress ratio. becomes greater than
the value corresponding to the angle of phase transformation. Note that this definition of
liquefaction is congruous with the definition of initial liquefaction given by Lee and Seed
(1967). In the case of loose samples made of Fuji river sand, the angle of transformation
as read off from Figs. l(a) and 6(a) gives the stress ratios of 1. 30 for triaxial compression
and -0. 91 for extension. These values are slightly smaller in magnitude than the stress
ratios corresponding to failure lines, i.e., 1. 55 in triaxial compression and -1. 05 in triaxial
extension.

UNDRAINED DEFORMATION MODEL UNDER CYCLIC STRESSES


The basic characteristics as mentioned above permit a deformation model to be established
for the saturated sand subjected to cyclic stresses in undrained conditions. By means of
this model, it is, therefore, possible to follow successively how pore pressures build up with
time and when liquefaction can occur in the course of load cycles changing with irregular
amplitude. The fundamental postulates on which the model is based can be itemized as
follows.

Pore Pressure Characteristics (Postulate 1)


Whenever. there is a tendency for a sample to cause irrecoverable volume change, impo-
sition of undrained condition reacts so as to develop residual pore pressures. The pore
pressures that develop in each cycle after previously applied stresses are exceeded are
determined by the undrained stress path as shown in Fig. 2 which are derived from the
state surface. When effective confining pressures are reduced after each cycle with con-
comitant increase in residual pore pressure, it is assumed that a changeover takes place
from the state surface currently concerned to that lying in the next neighborhood, and
hence, the undrained stress path to be followed in the immediate subsequent cycle is re·
presented by another stress path located inside adjacent to the current one. Elastic recovery
of effective stresses during unloading and reloading is specified by the stress path determined
from experiments under undrained conditions. On the basis of the test results, it will be
assumed that the unloading paths are all identical both in triaxial compression and extension.

Shear Strain Characteristics (Postulate 2)


Undrained cyclic loadings cause the stress path in e=const. plane to move around in
such a general way that the stress vector goes over successively accross the equi-r-lines,
thereby producing plastic shear strains. It is assumed that the equi-r-lines can be approxi-
mated by straight lines passing through the origin in p' -q space. With the reduction in
effective mean principal stress, p', a changeover is assumed basically to take place from the
current stress ratio-shear strain relation to that corresponding to an adjacent state of pac-
king, but the difference due to this can be practically neglected because the associated
change in void ratio, if existing, is very small. Therefore, for all the states to be realized
in terms of reduction in p', the stress ratio-shear strain relationship will be assumed iden-
tical. It is on the basis of this assumption that the equi-r-lines shown in Fig. 4 can
have practical meanings. Elastic shear strains which occur during unloading and reloading
are assumed to abide by the same rule for both triaxial compression and extension.
Yield Conditions (Postulate 3)
It is assumed that the yield loci can be approximated with sufficient accuracy, regard-
less of the density of sand, by a family of straight lines radiating outward from the origin.
Directional Independency of Yield Condition (Postulate 4)
As long as stresses are changed either within the range of triaxial compression or exten-

NII-Electronic Library Service


UNDRAINED DEFORMATION AND LIQUEFACTION 37

sion, whether yielding is occurring or not can be judged by the yield condition mentioned
above. In cases where both triaxial compression and extension are included in the time
history of stress changes, other criterions become necessary. Then, it is assumed that the
yield condition in either of triaxial compression or extension is independent of the stress
history which the sample has previously sustained in the opposite stress range.

Onset Condition of Liquefaction (Postulate 5)


It was shown in the foregoing that instability of the saturated sand occurs when the
·effective stress ratio, q/p', becomes equal to the tangent of the angle bf phase transfor-
mation. Hence, the pore pressure· build-up during cyclic loading is assumed to continue
until the effective stress ratio touches on the angle of phase transformation, whereupon
liquefaction is initiated.
On the basis of the five basic postulates as mentioned above, it is possible to follow
completely the trace of stress path and, hence, to evaluate the way in which pore pressures
and shear strains are developed within the sample when subjected to cyclic stress histories
with irregular amplitudes. The procedures for doing this are illustrated by a set of
diagrams in Fig. 7. Suppose that a sample has been consolidated to an effective mean
principal stress designated by point 1 in Fig. 7 (b) and, then, a given sequence of shear
stress changes shown in Fig. 7 (a) is applied to the sample. In response to an initial in·
crease in shear stress, the effective stress path runs from point 1 to 2 in Fig. 7(b) along
the virgin curve which is known from Postulate 1. Unloading and reloading cycle from
point 2 through 4 does not produce any irrecoverable change in pore pressure, because the
·events in this process are totally elastic. Begining again with point 4, the stress path
traces the virgin curve up to point 5, thereby creating permanent change in pore pressure.
Unloading from point 5 to 6 involves only elastic event whose stress path is determined
from the elastic rebound curve. As soon as the stress path falls in the domain of triaxial
extension, the sample again begins to exhibit plastic deformation in accordance with the
Postulate 4 stipulating directional independency of yield condition. Those parts of stress
path designated by points 7, 8, 9, 10, through 11 are associated only with elastic defor·
mations, because there is no chance for stress path to go across any fresh yield locus during
this process. Plastic yielding is resumed when the stress path is made to run once again
.along the virgin curve from point 11 on the extensional side. Unloading path from point
12 to 13 produces only elastic deformations as well as the reloading path in the domain of
triaxial compression starting from point 13. At point 14, stress ratio, q/p', becomes equal
to that at point 5, i.e., the maximum stress ratio which the sample has ever experienced
is again reached. Since unerased new yield loci are now trodden until stress point 15 is
reached, plastic yielding takes place, in conformity with Postulate 3, all the way from point
14, guided by the other virgin stress path pertaining to the state surface adjacent to that
previously concerned. Unloading and subsequent reloading from stress point 15 through 17
.are simply an elastic phenomenon. At point 17, the stress ratio previously applied is ex-
ceeded and, therefore, plastic yielding sets in and continues until state of stress designated
by point 18 is finally reached where the stress ratio coincide with the line of phase trans-
formation. At this instant, liquefaction occurs in compliance with Postulate 5. The
subsequent loading may cause the stress path to go approximately as shown in Fig. 7 (b),
but it should be mentioned that considerable difficulty may arise when one tries to seek for
the path after the liquefaction. This task is beyond the scope of the present investigation.
The above procedure permits a complete map of stress path to be described as shown in
Fig. 7 (b), and with this diagram it is now possible to know the time history of stress
ratio, q/p', as demonstrated in Fig. 7(c). As the relationship between stress ratio and
shear strain is already known for both plastic and elastic parts of deformation, the five

NII-Electronic Library Service


38 ISHIHARA ET AL.

-:plastic part hase transformation


----:elastic part q equ i- r-l ines
or
yield loci
q
15

9
14fi I.\
~ I I
I\ I I 21
81 I 13 1 1 16
p'
6 I
I
10\ I
I
I I I

7 11y
12
-q
(aJ
-q failure tine

- : plastic part
----:elastic part o/p
15 21

5 ,J\ virgin
curv~

I
I I
24 \ 9 f \
\~ ',, ,/\ I \I
I
--_.;..~'
6 _\..,10'---+1--'1"""16'-+-_.. ti me ~~~-tr--o-,~~-+,~-~
Is \ 113 I
I
I I I

7 11v 17
I
I
17
12

19
-o/p' ( c) (d)

- : plastic part
----:elastic part
q

Fig. 7. Diagrams illustrating the graphical representation of undrained


deformation model under cyclic stresses

basic postulates as established above also enable a complete picture to be drawn regarding
the stress ratio-versus-shear strain relation. This is presented in Fig. 7 (d). The physical
implications of each branch of curves are exactly the same as those ·of the corresponding
part of the stress path indicated in Fig. 7 (b). It is to be noted that the shear strains after

NII-Electronic Library Service


UNDRAINED DEFORMATION AND LIQUEFACTION 39

stress ratio exceeds the angle of phase transformation are indefinitely large.
If the shear strains are read off from Fig. 7 (d) and plotted against time, we get a time
history of strain changes as shown in Fig. 7 (e). This diagram, together with the given
time histry of sfress changes in Fig. 7 (a), permits a complete picture of stress-strain
relationship to be drawn as demonstrated in Fig. 7 (f).

APPLICATION OF DEFORMATION MODEL


The deformation model established above will be applied to assess the performances which
a sand will manifest when subjected to various types of loading. By comparing the beha-
viors predicted with those actually observed, the relevance and applicability of the model
will be examined.

1.5 ,..----,---,...---,...----. ,,..... 1.0


•start of yieling "E
u
~ 0.5

0
~
~' \\)
-0.5 11'~"' ___ /4
~~,~
11 4
_ 1 O • st9rt ?f
• yielding
"->
0 0.5 1.0 1.5 2.0 0 , 1.0
0.5 1.5 2.0
p
(kg/cm') P ( kg/cm2 )
(a) measured stress path (b) predicted stress path
Fig. 8. Comparison of measured and predicted undrained,:stress paths in
stress-controlled ·static test cycled with irregular amplitude

-o--o- measured curve


---predicted curve
------- virgin curve
• start of yielding

-5.0 -4.0 -3.0

-1.0

Fig. 9. Comparison of measured and predicted shear strains in stress-


controlled static test cycled with irregular amplitude

NII-Electronic Library Service


40 ISHIHARA ET AL.

Stress-Controlled Static Cycles with Irregular Amplitude


A loose sample having a void ratio of 0. 749 was consolidated to p' = 1. 595 kg/cm 2 and
subjected to a sequence of erratic changes in triaxial shear stress. The test results are
shown in Figs. 8 (a) and 9. The stress ratio becomes equal to the angle of phase trans·
formation at point 10 on the side of triaxial extension, whereupon liquefaction is induced.
On the other hand, the stress path and stress ratio-shear strain relationship were assessed
by means of the proposed model using the basic data presented in Figs. 2 and 4 and the
results are shown on Figs. 8 (b) and 9. While coincidence appears good between the
predicted and observed performances, some difference in detail seems to exist. Note, how·

1.0.----....,...---..,.-----.----,-----, I
/ I
20 16 12 8 4 e:0.737 //
//
(b)
(a)
..E 0.5 /;
u //
(/)
(/)
jl
,.f"
~ Qiiii::---------+-+-+Ht1+tt-tt7'-P'-5-+"3'----r.-i
iii
(/) '\_~ ~
""'
(/)

"""'~
QJ
I... L
Vi Cl
(l.J '~
ii -0.5
8 -o.s '\"
'~'
QJ
..!:
(/)
3026 22 . 6':::::--x'2
line of phase ' "~ ~,
transtormati on ~;::_---------'---- --
- 1·0 .....__ _ _Q5,___ _ _t._o_ _---Jt-5_ _____,2_0_ ___,2.s -1.0!:..:='.==~t::====--2:U~ _ __,__ ___.l___J
0 0 Q5 1.0 1.5 2.0
effective mean principal stress effective m.ean principal stress
p' (kg/cm")
p' (kg/ cm2 )
Fig. 10. Comparison of measured and predicted undrained stress paths in stress-controlled
static test .cycled with constant,:a?1plitude

toi------e-=-'-o.-73-7-----r------r----4
~.J limit of shear strain
(Q) before liquefaction

16-o
(b)
~
Q51--...;.._---+-----·t--7'----12-o
limit of shear strain a- -4 0.5 • start of yielding
before liquefaction
o points at which9/p
is maximum

numberscorrespond to
stress points shown ·
in Fig.10 (a)

-1.0l-------·t-------·---r------i-1

-3.0 -2.0 -1.0 0 -:.G -2.0 -1.0 -Q5 0.0


r . (°lo) r (°lo)

Fig. 11. Comparison of measured and predicted shear strains in stre$s-cpntrolled


static test, cycled with constant amplitude

NII-Electronic Library Service


UNDRAINED DEFORMATION AND LIQUEFACTION 41

ever, that as far as the residual pore pressure within each cycle is concerned, the agreement
seeins satisfactory.

Stress-Controlled Static Cycles with Constant Amplitude


Figs. 10 (a) and 11 (a) show the results of a triaxial shear test on a loose sample with a
void ratio of 0. 737. After the pore pressure has built up at approximately a constant rate,
liquefaction is caused at stress point 23, where the stress ratio becomes equal to the angle
of phase transformation. Predicted performances by means of the model using the basic
data in Figs. 2 and 4 are demonstrated in Figs. lO(b) and 11 (b). As far as the stress
path is concerned, agreement is not necessarily good, because of the fact that the stress
path in this case must pass many times through the portion where the curvature of the
virgin curves exerts very delicate influence. This may not, however, underestimate the
suitability of the proposed model, because the more randomly stresses are changed as is
generally the case, the better the agreement is as clearly manifested by the foregoing
example. The stress ratio versus strain plot in Fig. 11 shows that the strains (or equiva-
lently, modulus of deformation) in each cycle remain almost uncha,nged until initiation of
liquefaction, although the modulus should seemingly be reduced due to the successive dec-
rease in effective confining pressure. This contradiction may be resolved by considering the
fact that plastic strains of about the same magnitude but with different sign are made to
occur alternatively within each cycle. This can account for the usual observation that
occurrence of liquefaction is quite abrupt and until then the deformation of soil mass is
very small and almost unvisible.

Strain-Controlled Static Cycles with Constant Amplitude


The pore pressure and stress ratio-strain relationship which are anticipated to develop
under strain-controlled loading conditions can also be assessed by using the procedures
introduced in the foregoing section. This time, a given sequence of strain changes is first
plotted versus time as was done in Fig. 7 (e). Then, with the aid of the basic vergin and
·elastic curves shown on Fig. 4, a hysteresis diagram between stress ratio and strain can be
constructed in the same fashion as illustrated on Fig. 7 (d). This procedure gives a time

1.5 .-----r-----r----r----.. 1.5 r----,----,,;-r-;---r-----,


locus of peak I line of phase
stress points faillirn-1~' J:: ""' transformgUon
~2 -
7,,,./,,,. . . . 'I "
,~{/'
1.0 1.0

6 "\
~
~
g
0.5

I
~, \ \

Io.o ~f-ft--8--9{;*14/.*-~-F----0:---1 ~---fHHt-tf-l<t-+L----i.'2..3


O"

r 0.0 --,/1-J-"1.--1

e: 0.745 I
I
I
-0.5 -0.5 _,.,/
0=0.05 ~ 0.71 °lo •:start of
~ yielding
(Q)
- 1.0 .-::----:"::"-----'-'-"---.......1...----l
0.5 1.5 2.0 o.o 0.5 1.0 1.5 2.0
- p' (kg/cm') - - p (kg/cm')

Fig. 12. Comparison. of measured and predicted undrained paths in strain-controlled


static test cycled with constant amplitude

NII-Electronic Library Service


42 ISHIHARA ET AL.

19
C\lp' 3
12 ,.
1.0 1.0
-6
6
/~/
2 . .
vtr
.0 1 cycle 1 tQC:lj~I/-~

2 cycle •start of yielding


I!. 5 cycle numbers
0.5 correspond to Q5
stress points
shown in Fig 12(b)

1 5 3
r (°lo) I 1.0 '( (°lo)

0.71 °lo
vir..,in curve
----~~//
//j
4
, Double amplitude
-0.5 numbers ,,/ -0.5 ~-~of~.strain _
correspond to / 4
stress points ,,./ I
(a) s I shown in Fig12(a) ,./ 10
_.. 16
'
0.05 °lo
(b)
It
Q.05%

Fig. 13. Comparison of measured and predicted shear strains in strain-


controlled static test cycled with constant amplitude

history of stress ratio as shown on Fig. 7(c). Then, with the help of the basic stress
paths given in Fig. 2, the known time history of stress ratio can be fitted into the map of
stress path as shown in Fig. 7 (b). Figs. 12 and 13 demonstrate the predicted performance
of a loose sample subjected to cycles with constant strain amplitude varying between 0. 05
and 0. 71%. On the other hand, a strain-controlled triaxial shear test was conducted under
the same conditions, and the results are presented on Fig. 12 (a) and 13 (a). It may well
be said that the prediction by means of the model furnishes satisfactory duplication of real
performances. In the previous paper (Tatsuoka-Ishihara, 1974 b) dealing with cyclic drained
tests, it was demonstrated that in the case of strain-controlled tests with a constant ampli-
tude, the shear stress is increased with cycles. As opposed to this, cyclic undrained tests:
under otherwise identical conditions give a gradual decrease in the amplitude of shear
stress as the cycle goes on. This seemingly conflicting behaviors can be explained well by
the proposed model.

Stress-Controlled Dynamic Cycles with Irregular Amplitude


By means of a closed loop servo-hydraulic loading system, some dynamic triaxial tests
were carried out on loose samples of sand. Detailed descriptions of the tests and some of
their results were already presented (Ishihara-Yasuda, 1972). In those tests, the time
histories of N-S component of horizontal accelerations recorded at the time of the Niigata
earthquake were converted to the time history of shear stress changes and applied to the·
triaxial samples of Niigata sand. Using the measured pore pressure alteration, the diagrams
of effective stress path were drawn. Fig. 14(a) shows the stress path thus obtained in
the test in which the maximum shear stress was made to occur on the side of triaxial
compression. On the other hand, the predicted stress paths using the proposed model and
effective stress paths obtained by undrained static tests on Niigata sand (Table 1) are·
presented in Fig. 14(b). Fig. 14 seems to offer good agreements in spite of the fact that
the prediction is based on the static behavior while the observation is made on the dynamic
aspect of the sand behavior.

NII-Electronic Library Service


UNDRAINED DEFORMATION AND LIQUEFACTION 43

1.0 1 - - - - - - - - 1 - - - - - - - + - - - + t - - - t - i
Niigata sand

-
'"E
eo= 0.781 • start of yielding
o local max. stress

-u
Ol -
~b
"' overlap ot. and o
~QSl-------+------1--1'---t---tt-+H-t
Ci

l/l
VI
--
O'l
..x:
0.51---------r-----:r:--1:..+----+t--+-t---'-+-+

Cll
L
......

I o~~~~~~~~~~~~:;_~~~~~
VI

a
Cll
oi---o------+--~---o--+---~

..c
l/l
I..
'\
0
<11
.c \\:\
-o.s~-----'<-~---1-----,,£-- I
I
I
I

/,/
<a) ( b)
'\;-----/-// //,/
,,
-1.00 0.5 1.0 1.5 0.5 1.0 1.5
effective mean principal stress effective mean principal stress.
p (kg/cm") p' (kg I en-fl
Fig. 14. Comparison of measured and predicted undrained paths in stress-controlled
dynamic test cycled with irregular amplitude (CM-test with N-S component of
the acceleration in the Niigata earthquake)

CONCLUSIONS
On the basis of the five fundamental postulates, a model was proposed by which it is
possible to assess the development of pore pressures and shear strains in an element of sand
when it undergoes cyclic loading in undrained conditions. The performances of sand
predicted by this method proved relevant and useful to the assessment of dynamic pore
pressure development and consequent occurence of liquefaction for wide ranges of loading
conditions from static to dynamic, stress-controlled to strain-controlled, and from uniform
to irregular cycles of excitation.

ACKNOWLEDGEMENTS
In developing the investigation described herein, efforts expended by many colleagues in
the Soil Mechanics Laboratory of University of Tokyo are immeasurable. The authors
wish to express their gratitude to them. The assistance by Mr. E. Shiba and Mr. M.
Yoshihara in running experiments and coorporation Mr. K. Suga in drawing figures are
particularly to be acknowledged.

NOTATION
e=void ratio
p' =effective mean principal stress= (aa' +2ar') /3 (kg/cm2 )

NII-Electronic Library Service


44 ISHIHARA ET AL.

q=shear stress=O'a' -ar' (kg/cm 2 )


v volumetric strain=ea+2er (%)
ea, e1 =axial and radial strains (%)
r shear strain=E:a-E:r (%)
<J a. O' r =axial and radial stress (kg/cm 2)

REFERENCES
l) Ishihara, K. and Yasuda, S. (1972): "Sand liquefaction due to irregular excitation," Solis and
Foundations, Vol. 12, No. 4, pp. 65-77.
2) Iwasaki, Y. (1971): "Liquefaction of sand interpreted as fatique failure and its predictive method
of analysis, " Proc. 6th Annual Research Convention of Soil Engineering, pp. 523-526.
3) Lee, K. L. and Seed, H.B. (1967): "Cyclic stress conditions causing liquefaction of sand," Proc.
ASCE, Vol. 93, SM L. pp. 47-70.
4) Poorooshasb, H.B., Holubec, I. and Sherbourne, A. N. (1967): "Yielding and flow of sand in
triaxial compression, Part II, " Canadian Geotechnical Journal, Vol. 4, pp. 376-388.
5) Shibata, T., Yukitomo, H. and Miyoshi, M. (1972): "Liquefaction process of sand during cyclic
loading," Solis and Foundations, Vol. 12, No. 1, pp. 1-16.
'6) Tanimoto, K. and Suga, Y. (1971): "Liquefying process of saturated sand subjected to dynamic
loading," Proc. 4th Asian Regional Conf. on Soil Mechanics and Found. Engineering, Vol. i, pp.
235-239.
7) Tatsuoka, F. and Ishihara, K. (1973): "Stress path and dilatancy performance of a sand," Proc.
8th ICSMFE, Moscow, Vol. 1, pp. 419-424.
. .8) Tatsuoka, F. and Ishihara, K. (1974 a): "Yielding of sand in triaxial compression," Soils
and Foundations, Vol. 14, No. 2, pp. 63-76.
9) Tatsuoka, F. and Ishihara, K. (1974 b) : "Drained deformation of sand under cyclic stresses
reversing direction," Soils and Foundations, Vol. 14, No. 3, pp. 51-65.
(Received May 1, 1974)

NII-Electronic Library Service

You might also like