Professional Documents
Culture Documents
ON SAND LIQUEFACTION
Downloaded from ascelibrary.org by Indian Inst Of Technology Library, Kharagpur on 10/28/16. Copyright ASCE. For personal use only; all rights reserved.
INTRODUCTION
1347
on the soil skeleton (effective stresses) are the same, then the soil exhibits
the completely identical mechanical properties no matter how much the pore
pressure would be; and (2) Henry's law about the dissolution of air in water
and Boyle's law in classic physics. In other words, under the same tem-
perature, the volume of air dissolved in a certain volume of water is the
same, and the product of the volume of air and the value of pressure existing
on the water is constant. The back-pressure technique has been playing a
very important role in geotechnical laboratory tests. In the mid-1960s, with
the increasing research interests in dynamic properties of soil, typically in
the liquefaction of saturated sand, much stricter control of the degree of
saturation for sand test samples is challenged. The adoption of the back-
pressure technique sometimes appears to be a natural way. The purpose of
the present work is to investigate the possible effects of the method on the
resistance of liquefaction of a sand. However, in this paper, it will show
that the use of back-pressure technique in liquefaction test of the sand dis-
torts the test results. Needless to say, the so-called back pressure in this
paper can also be viewed as the hydrostatic pore pressure existing initially
in sand before the application of dynamic loading. For example, the case
could be easily found in practice, such as a sand layer embedded within the
region of a reservoir. Due to the restriction of the present research to the
laboratory investigation here, the term back pressure is used throughout this
paper to refer to this hydrostatic pressure.
The test material is a fine sand taken from the drill holes at the site of
Tong Jaizhi Power Station. It has a coefficient of uniformity C„ = 3.7, me-
dian size of particle D50 = 0.0995 x 103 m, and specific gravity Gs = 2.74.
A cyclic triaxial apparatus is used with sizes of cylindrical samples of 0.05
m in diameter and 0.12 m in height. The dry unit weight controlled for the
tested sand samples is yd = 1,620.0 kg/m 3 , which is corresponding to 60%
in relative density. Fig. 1 illustrates the pressure cell of the cyclic triaxial
apparatus used.
In Fig. 1, No. 1 is the sand sample in latex membrane; No. 2 is the applied
all-around pressure; No. 3 is compression air; No. 4 is the porous stones;
No. 5 is the lower platen; No. 6 is the upper platen; No. 7 is the axial piston;
No. 8 is the axial displacement transducer; No. 9 is the static axial pressure;
No. 10 is the applied dynamic loading; and No. 11 is the pore pressure
transducer.
The latex membrane is sealed at both ends onto the lower platen and upper
platen by O-rubber rings, respectively. Porous stones with filter paper are
used to separate the upper platen as well as the lower platen from the sand
sample. The pore pressure transducer is located very close to the lower end
of the sample, which has a size 0.003 m in diameter and 0.008 m in length.
The effect of the stiffness of the pore pressure measuring system on the
readings of pore pressure is restricted to a very small and negligible level
by making the measuring tunnel of the pore pressure transducer very short
and mainly embedded in the very rigid base of the pressure cell. The pore
pressure coefficient B is obtained by measuring the ratio of the increment
1348
11
FIG. 1. Pressure Cell of Cyclic Triaxial Apparatus
of the pore pressure Au and the increment of the all-around pressure Ao-3.
The axial piston is the exact same size as the upper platen, through which
the axial static pressure and the dynamic force are applied. The ambient
pressure is provided by compression air. The fluctuation of the ambient pres-
sure during the excitation of dynamic loading can therefore be brought down
to a negligible value since the great compressibility of air provides the ad-
vantage for easy up-down movement of the axial piston. The effect of air
penetrating into the tested sand sample is found insignificant if the latex
membrane is not defective. It can be regarded that both axial load and am-
bient pressure are controlled separately. All-around pressure is given by ap-
plying the same amount of the axial pressure and ambient pressure at the
same time. Dynamic exciting load is applied from the top of the sample.
The process of dynamic axial displacement of the sample is obtained from
1349
"f,/0 "id Od
1350
±0.02 g and then kept in glass cups. Two sets of the samples with degrees
of saturation of 100% and 99.5%, respectively, are obtained by soaking those
sand samples in their cups with distilled water for 2 hours and then putting
them into a vacuum pump to deair with a vacuum pressure of —0.74 m
mercury for 3 hours, and 1/2 hours, respectively. Another set of samples,
with 9 9 . 1 % saturation, is obtained by only soaking them in their cups with
distilled water for 2 hours—no vacuum pressure is applied. The last set of
samples, with 97.8% saturation, are simply obtained by soaking them in
their cups with common running water for about 20 min. All the samples
are formed in three layers in an open brass sample mold, each layer of sand
is uniformly compacted. A very thin and good-quality latex membrane is
used to keep the sample in shape and separate it from the air pressure. The
four sets of sand samples have the pore pressure coefficients B = 1.0, 0.928,
0.857, and 0.714, respectively. The net weight of dry sand used for each
sample is obtained by weighing the remaining part of the sand left after the
sample is formed in a calibrated bottle, then deducting that value from the
total weight of the dry sand used. The initial volume of each sample before
consolidation is obtained by carefully measuring the average height and di-
ameter of the sample prior to the setting of the triaxial compression cell.
The volume of each sample after consolidation is obtained by deducting the
amount of drainage water from the initial volume of the sample. The initial
dry unit weight, as well as the dry unit weight after consolidation, are then
1351
0.28
0.24
0.20
calculated. The initial liquefaction criterion is used (i.e., the residual pore
pressure is equal to the all-around pressure). The liquefaction resistance curves
\<jd/2/a'x versus log AO of the tested sand are given in Fig. 2 for the dif-
ferent degrees of saturation.
The pore pressure coefficients B are obtained by measuring the value of
pore pressure produced by the all-around pressure 0-30
The relationship between pore pressure coefficient B and degree of satu-
ration is associated with the density of the sample, the compressibility of
soil skeleton, and the compressibility of voids (water-air system) (Skempton
1954). Theoretically speaking, this relationship can be easily derived. How-
ever, the compressibility of the soil skeleton is quite unpredictable because
the stress-strain curve of the soil is not a straight line. This curve would
also be obviously different even if the same test conditions are maintained
but only, a new soil sample is used. A good suggestion is to use the reloaded
modulus of the soil, which will give a comparatively stable value for the
compressibility of the soil corresponding to the quite small value of applied
loads. If it is possible, the best way to get the degree of saturation with
respect to the different pore pressure coefficients B is no doubtfully to mea-
sure the water content of the typical samples directly. However, this is not
the case for the present problem, because the very close values of degree of
saturation for all test samples cannot be discriminated clearly by directly
measuring the water contents of samples. Therefore the theoretical relation-
ship is adopted with a slight empirical modification. That part of the work
is presented in Appendix I.
The corresponding approximate degrees of saturation for the four curves
shown in Fig. 2 are about 100%, 99.5%, 99.1%, and 97.8%, respectively,
which are calculated in Table 1 by using the method in Appendix I.
1 U i
1
FIG. 3(a). Typical Oscillogram Recording: ad = 34.34 kPa, ed = 19.00%, N = 6.5,
P = 0.0 kPa
is connected to the sample by small brass tubes, which can apply an incre-
ment of static water pressure to the sample through the hole for the pore
pressure transducer at the lower platen as shown in Fig. 1. Conventionally,
on such equipment, back pressure is applied by following these steps: the
all-around pressure is firstly applied on the testing sample by equally pro-
viding the same amount of ambient pressure and axial pressure through a
separate controlling system, then the sample is allowed to consolidate for a
certain time period by releasing the two drainage valves that connect to the
top and bottom of the sample, respectively. After the consolidation proce-
dure is completed, the two drainage valves are to be shut off and the ap-
plication of back pressure begins. Only a small amount of back pressure can
be applied for each time. To apply each small amount of back pressure AP,
the increments of the same amount of axial pressure and ambient pressure
will also need to be applied onto the sample simultaneously with the back
pressure. After the application of back pressure is completed, the back pres-
sure valve should be shut off before the dynamic loading can be applied.
This often causes a slight change of the applied back pressure due to the
possible uncontrolled water flow that can happen at the same instant. There-
fore, it is realized that the following two problems are to be solved before
carrying out the cyclic back-pressure tests: (1) Sand samples should not undergo
the disturbance that is different from the disturbance caused in the usual
consolidation process; and (2) the accumulation process of the dynamic pore
pressure inside the testing sample should not be affected by the external
systems that are irrelevant to the cyclic loading. It was found that in liq-
uefaction tests, a slight disturbance on the tested sample leads to much higher
liquefaction resistance than the sand possesses (Miura and Toki 1982; Mori
et al. 1978). For those reasons, the back-pressure device on this equipment
cannot be used because the water flow into the sample from the back-pres-
sure system is not fully controllable, and it is also impossible to apply the
back pressure and the required equal amount of increment of the axial pres-
sure and the ambient pressure onto the sample absolutely simultaneously.
All of those may produce irrational disturbances on the tested sample.
In the present work, the pore pressure that is spontaneously generated
during the application of all-around pressure is used as the back pressure.
1353
FIG. 3(b). Typical Oscillogram Recording: ad = 39.24 kPa, <.d = 14.50%, N = 6.5,
P = 98.1 kPa
.SL
2 0,'r
0.38
P (KPa)
0.34 A 382.4
o 298.3
0.30 X 196.2
0.26 • 98.1
0.22
^ ^ - A 0.0
"V*^. ~~^*—~, _
0.18
^ 1 -•
0.14 - I *
3 5 10 20 50 100 200
The following procedures are used to obtain the back pressures. For the first
set of samples with zero back pressure, 98.1 kPa of ambient and axial pres-
sures are applied onto the samples simultaneously, then a 15 min consoli-
dation is provided by releasing the drainage valves. The sample gains an
equal amount of 98.1 kPa effective axial stress a[c and effective ambient
1354
J
3C
FIG. 5(a). Effective Dynamic Stress Path: ov = 34.34 kPa, N = 6.5, P = 0.0 kPa
FIG. 5(b). Effective Dynamic Stress Path: <jd 63.77 kPa, N = 10.0, P =
392.4 kPa
1356
compared with the corresponding diagrams with lower or zero back pressure
as showed in the Fig. 5(a) and Fig. 3(a). However the densities of all sam-
ples are strictly controlled during tests. Because the spontaneously rising
pore pressures are used as the back pressures, there is little possibility to
change the densities of samples significantly. The differences of the lique-
faction resistance can only be possibly ascribed to the back pressure itself.
Since the application of back pressure does not change the values of the
static effective stress states, how can the liquefaction resistance be affected
by the back pressure? According to Terzaghi's effective stress principle, the
stress states of the soil skeleton determines the mechanical properties of soil
element. It is usually understood that no matter what the current general
stress states (CT1,a2,o-3) would be for a soil element, if the current effective
stress states (ST[,<T2,<J'3) are the same in values and directions for the soil
element, then the soil element can be considered to be in the same current
stress state and it will display the identical properties (the same stress-strain
relationship, the same strength, etc.) if the same kind of loading is applied.
However, although the same effective stress states are kept for all back-
pressure tests before applying dynamic loading, the mechanical properties
relating to liquefaction of this sand has been apparently changed as in Figs.
4, 5(a), and 5(b). It contradicts Terzaghi's effective stress principle. This
means that the back-pressure technique based on Terzaghi's effective stress
principle is invalid in liquefaction tests. Although this method maintains the
same static effective stress state for all tests with different values of back
pressures, it seems to have changed the liquefaction potential of those sam-
ples. The extra hydrostatic pressure on the samples tends to give additional
confinement to the movement of the sand particles. Therefore when the same
value of dynamic axial load is applied, this dynamic load appears not to
cause the same movement of sand particles. As a result, the increment of
pore pressure in the sample caused in each cycle of the dynamic loading
will be different, and thus leads to different dynamic effective stress paths.
How does the extra hydrostatic pressure (back pressure) provide the addi-
tional confinement to the samples and consequently strengthen the samples?
A possible explanation can be given according to the microstates of the ef-
fective stresses. The microforce conditions inside samples have been changed
by back pressure. Those force conditions are referred to as the actions of
microforces among particles. They include the values of the interparticle
forces; the way the interparticle forces transfer (through what kind of con-
tacts—which can be point, line, or surface contacts); and the way the in-
terparticle forces interact (frictional actions or push actions). Application of
back-pressure does not change the macroeffective stress states that are mea-
sured on the surfaces of samples, but it may change those microconditions
of the interna] particle forces. It may be possible that back pressure affects
the liquefaction properties of samples by changing those conditions, at the
same time the resultant forces on the soil skeleton are still kept unchanged.
From this point of view, Terzaghi's effective stress principle should have
additional conditions for the interparticle actions of soil. The effective stress
principle will only be valid if all those additional conditions are also kept
the same. For the tests herein, if those additional conditions were kept the
1357
FIG. 6. Types of Contacts for Two Particles: (a) Surface Contact; and (b) Point
or Line Contact
same too, the application of back pressure might have no influence on the
properties of liquefaction of the sand.
However, for materials like sand, it is almost impossible for it to satisfy
those additional conditions. Under different values of back pressure, the ways
the interparticle forces act are very different from one another, because of
the irregular shapes of the sand particles. For the sand used in this paper,
the surface of its particles is composed of irregular multiplane faces. If the
plane face of a hemisphere is used to represent one of the faces a sand
particle possesses, the types of contact that may happen between any two
faces of adjacent sand particles are illustrated in Figs. 6(a) and (b). Although
a sand particle may not have the perfect plane face as is shown in those
figures, the explanation here can still be valuable to illustrate the contacts
between two particles.
For the case as in Fig. 6(a), it is a surface contact. When the hydrostatic
pressure is increased by an amount of back pressure P as in Fig. 7(a), then
the interaction force on the contacting face between A hemisphere and B
hemisphere is also increased by the amount of P. Therefore it can be said
that the interparticle force has been changed in value. Such increase of push
action also increases the frictional resistance between the interfaces. For the
case of point or line contact as in Fig. 6(b), there is a clear wedge space
between the two faces of the two particles. Usually this space is very small.
Therefore the influence of the viscous property of water and the molecular
forces in the absorbed water layer cannot be totally ignored. The water con-
strained in this space is not free water. It is known that only free water can
carry the hydraulic pressure equally and simultaneously in all directions pro-
vided that the water is in an enclosed system. Therefore, when the hydro-
static pressure is increased by a back pressure of P as in Fig. 7(b), this kind
iiimu IUHIII
\AU/.
p
3 /
•flttt
(o)
(b) (c)
constrained water, then the forces are not in equilibrium. The resultant force
may compel the two hemispheres to push together or may enlarge the wedge
space between the two faces, as in Fig. 7(c).
However, no matter what case happens, the back pressure has changed
the interparticle forces in value, direction, and ways those microforces trans-
fer and act, as well as changed the arrays of sand particles.
The mechanism of sand liquefaction is due to the sliding of some particles
over others, and therefore causes the shrinkage of the soil skeleton. The
dynamic pore pressure is generated due to the offsetting of such shrinkage
happening in soil skeleton. The changes of the interparticle forces by the
back pressure in the aforementioned cases will drastically change the shrink-
age property of the soil skeleton, therefore leading to different liquefaction
resistances.
CONCLUSIONS
ACKNOWLEDGMENT
This research program is sponsored under the fund offered to the writers
from Natural Science Research Foundation, Academia Sinica. The writers
are very thankful for this support.
1359
Skempton (1954) stated that the following relationship exists for soil:
1
B = (1)
1 + n It
JS _
K„
(2)
Kw
K„, (3)
1 + (1 - Sr)\ ^ - 1
Kn
We also have
Id
n = 1 (4)
Kv
1
1 - - 1 (5)
Kw
l--^-|*.
K„
7..A.
where Ks is obtained by conducting the repeated loading test on three sam-
ples in drained conditions. The repeated all-around pressure is 490.5 kPa.
A specially designed electric capacitance volume change transducer is used
to measure the volume changes. This transducer is made of one string of
Tantalum wire and one string of silver wire as shown in Fig. 8.
The water-level change in the tube will cause the change of the electric
capacitance between those two wires. The signals sensed by this transducer
is amplified by a dynamic strain gauge and is recorded on the oscillograph.
It is found that the volume changes in the first three cycles due to the
1360
Silver wire
Water flow
^
Wire trminals
Sr = I - 0.0097592
2.385857
(H (6)
S0 =
Sr
= 0.9903351 - 0.0996487
2.385857
(H (7)
°rmai y*
ywG,
1361
Miura, S., and Toki, S. (1982). "A sample preparation method and its effect on
static and cyclic deformation-strength properties of sand." Soils and Foundations,
22(1), 61-77.
Mori, K., Seed, H. B., and Chan, C. K. (1978). "Influence of sample disturbance
on sand response to cyclic loading." J. Geotech. Engrg. Div., ASCE, 104(3),
323-339.
Skempton, A. W. (1954). "The pore pressure coefficient A and B." Geotechnique,
London, England, 4, 143-147.
Richart, F. E., Jr., Hall, J. R., Jr., and Woods, R. D. (1970). Vibrations of soils
and foundations. Prentice-Hall, Inc., Englewood Cliffs, N.J., 129-132.
1362