You are on page 1of 15

SOILS AND FOUNDATIONS Vol. 14, No. 3, Sept.

1974
Japanese Society of Soil Mechanics and Foundation Engineering

STRENGTH CHARACTERISTICS OF DRY SAND


SUBJECTED TO REPEATED LOADING

SHosuKE ToK1* and SmoERU KITAoo**

ABSTRACT
Some results of research on the strength of dry sand are presented, based on experiments
carried out over a period of several years on the geotechnical properties of dry sand subjected
to repeated loading. The dependency of the dynamic strength of sand on the number of cyclic
loading, static stress state of sand prior to the application of cyclic stress, initial and at failure
void ratios of the sand, etc., were investigated, and comparisons were also made between the
dynamic strength and the static strength. The test results indicate that dry sand shows a
higher strength against dynamic load than that against static load, but in view of deformaw
tion, sand is less stable against repeated loading.

Key words: angle of in tern al friction, earthquake, repeated load, sand, stress-strain curve,
triaxial compression test
IGC: D6/D7

INTRODUCTION
Extensive research has been made on the dynamic properties of soil, and as a result the
mechanical behavior thereof has become increasingly clarified. However, exact predictions of
strength parameters of soil under dynamic stress conditions such as earthquakes or traffic loads
are not yet available for computation of bearing capacity of ground and for analysis of slope
stability. Hence, at present we are forced to assume that a soil shows the same strength under
both static and dynamic stress conditions.
Among design engineers, it is more or less accepted that a design based on the static
strength of soil will be on the safe side, since soil is in general more resistant against dynamic
stress than against static stress, except in the special case of liquefaction. This seems to lie
in the background of the above mentioned design practice. Another practice in design is to
assume a lower factor of safety in earthquakes than that under static stress conditions for eco-
nomical reasons., including considerations for rare outbreaks of strong earthquakes. Thus,
this means that a soil is assumed to have rather higher strength during earthquakes than
under static stress conditions.
However, many structures designed by conventional methods with seismic forces in mind
have suffered from earthquakes. It is, therefore, necessary to develop a functional theory for
strength characteristics of soil under dynamic stress conditions, and to accumulate empirical
facts to fortify the theory, in order to make it possible to set forth a reasonable design from
a point of view of economy and safety.
In this paper, the test results are presented and discussed with special regards to the strength
* Associate Professor of Civil Engineering, Hokkaido University, Sapporo, Hokkaido.
** Professor of Civil Engineering, Hokkaido University, Sapporo, Hokkaido.
Written discussions on this paper should be submitted before July 1, 1975.

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


NII-Electronic Library Service
26 TOKI AND KIT AGO

characteristics of dry sand, and the authors hope to offer some contributions to research on
strength characteristics of sand layers subjected to repeated loading as an earthquake.

TEST APPARATUS AND SOIL SAMPLES


A brief explanation of the vibratory triaxial compression test apparatus used for the present
research is as follows; static cell pressure C1ss is applied by water. Static axial stress 0' 1s is
applied through a bellofram cylinder by oil, where the pressure is accurately rated in advance.
Sinusoidally alternating dynamic loads are generated by cammed pulsators and are superim-
posed on the static stresses in both axial and radial directions. The dynamic stress was prevent-
ed from flowing back to static stress sources by non-return valves.
Since the two pulsators are connected by a clutch, the phase difference of repeated loading
in axial and radial directions can be changed as 0°, 90° and 180°. The cycle .of the repeating
load can continuously be changed from. 0. 5 to 5. 0 Hz. The dynamic stresses were measured
by a stress transducer and were automatically recorded.
Axial deformation of a test sample was measured by both dial gauge and LVDT. The
volume change during tests was in most cases measured by an air manometer. The discrepancy,
between computed void ratio from direct measurement by caliper and that from axial .strain
and volume change was less than 3 %. The layout of the test apparatus is given in the.
authors' previous paper (Toki, Kitago, 1973).
The Toyoura standard sand, under 74µ, washed in water, and oven dried, was used in the
present test. The sand has the specific· gravity of 2. 65, 10 % and 60 % diameter of 0.177 mm
and O. 205 mm, respectively, the uniformity coefficient of 1. 2, the maximum void ratio of 0. 953,
and the minimum void ratio of 0. 645. The sand consists of subangular quartz particles.
The dimension of test sample was 50 mm in diameter, and 120 mm in height. In order to
minimize friction at both ends of a test sample, a thin rubber membrane coated with silicone
grease was laid on the pedestal and the cap.

METHOD OF STRESS APPLICATION


The choice of type of load application for the study of dynamic properties of sand depends
on the assumption of stress state in soil mass subjected to cyclic stress. In this paper, the soil
element in the ground, which is in a statically isotropic or anisotropic stress condition, is as-.
sumed to be subjected to shear stress which periodically alternates its direction. These stress
states are approximately simulated within a test sample of the vibratory triaxial compression
test apparatus used in this experiment.
Fig. 1 (a) shows the method of application of alternating shear stress T'n to the test sample.
which was initially in static equilibrium under a deviatoric stress of (0' 18 -0' 88). The static axial
stress, 0' 15 , and the static radial stress, 0'88 , were first applied to the test sample, and then cyclic
stress O'n with a phase difference of 180° was overlaid on the static stress 0' 18 and C1ss· The

o;s
b
1

O 0.5T 1.0T 1~t 2.0T


<a> (b) (C)

Fig. 1. Method of stress application and stress state within test sample

NII-Electronic Library Service


DYNAMIC STRENGTH OF SAND 27
stress system is illustrated by Mohr's stress circle shown in Fig. 1 (b), which indicates that
on the 45° plane within test sample, shear stres~ changes its value periodically under constant
normal stress. This system of stress corresponds to the stress state wit.hin a soil mass ~hich is
initially in equilibrium under static shear stress and then is subjected to such a periodicall;y:
alternating stress as an earthquake.
Fig. 1 (c) shows a special case for the stress state given in Fig. 1 (a), in which the axial
static stress <1is is equal to radial static stress <188 • · In this case, the diameter of the stress
circle changes its value between 0 and <1D, and the shear Stress generated Within the test sample
consists only of a dynamic stress component. The shear stress on the 45° plane is given by
the ordinate of a point between A and B in Fig. 1 (c), in accordance with the change of the
direction of the major principal stress with time. This simulates the stress state on a horizonfal
layer in the ground during an earthquake.
10.0
Dynamic stress <:ID is raised step by step under constant
static stresses of <1 18 and <1 88 until failure is observed.
The strength of dry sand, in general, shows little deperld- >
J.LJ.656
-,. ..•
I
I
o;s =2.0
I
l 119.(.J)

•-- •
ency on strain rate unless it is extremely high (Whitman tf I
8.0
and Healy, 1962). The relationship between the frequency I:
of dynamic stress cycle and the yielding stress aDY from I
7.0 -~-o . e=0'.879 0 0
o--o-
D-test, which will be described later, is shown in Fig. 2. I
It can be concluded from this figure that the yielding 6.0 I I
stress r.1DY is practically constant for a range of loading I I
frequency from 0. 5 to 5 Hz, which corresponds to the strain 0 05 1.0 2.0 3.0 4.0 5.0
rate of 60 to 600 % per minute. Thus, in this experiment, f (Hz)

each test was carried out at a frequency of cyclic stress Fig. 2. Independency of cyclic stress
at failure to loading frequency
of 2Hz.

TYPES OF TESTS
The tests carried out in this study, in which the method of application of static and cyclic
stresses was changed, consist of the following types :
1) S-test : conventional static triaxial compression test under constant confining pressure of
qss•
2) D-test : dynamic deviator stress is applied to a test sample which is initially in static equi;
librium without static deviator stress. In Fig. 1 (a), at the outset of the test, a test sample is
subjected to an allround pressure of <1 18 =<1 8s, and then a certain level of cyclic stress <1n=<1n 1
is applied for a certain number of cycles, which is hereafter referred to as N. The level of
cyclic stress is then raised to r.1D 2 for the same number of cycles, and this procedure is repeated
until the yielding stress <lny is reached in 7 to 12 steps of dynamic stress increment. The num-
ber of stress cycles N was 500, except for the case in which the effect of N was investigated,.
3) SD-test : this was designed to investigate the effect of static stress conditions before ap-~
plication of cyclic stress on the strength characteristics of dry sand subjected to dynamic loads.
In this test, varying initial static deviator stress was applied, and then dn was raised step by
step up to yielding .stress <1Dy for a particular number N of 1, 000 .
. Another type of test was undertaken to measure the shear strength of dry sand during vi-
bration, using. triaxial compression test apparatus installed on a vibrating table.

RESULTS OF D-TESTS
General Relationships between Dynamic Stress and Deformation
Stress-strain relationships in D-test differ in several points from those in a conventional
S-test. In D-test, firstly, axial strain of a test sample consists of two components, i.e., shear

NII-Electronic Library Service


.28 TOKI AND KIT AGO

component and volumetric component. The former, referred to as tzn, is the axial component
of strain due to deviator stress, periodically changing its magnitude in accordance with the
alternating change of the direction of principal stress by 90°, and the latter, referred to as tzp
is the axial componet of strain due to the volume change of the test sample accumulated dur-
ing dynamic loading.
In Fig. 3, which shows one example of observed t1n vs. tzn and ezp, it can be seen that
ezn increases almost linearly with increasing t1n up until failure stress <Jny is reached, and from
then on it increases rapidly. On the other hand, ezp does not show a clear break as observed
in ezn, although ezp maintains a linear relationship with t1n at its lower level.
Furthermore, comparing ezp and ezl) to the axial strain ezs from S-test, ezp and ezn for the
dynamic stress level of t1n, which is equal to static failure stress (t1 1s-t1as) / of S-test, were
smaller than ezs at failure.

8.0
'} eo= o.e1s
~ (T35: 1.0 (K%.,?)
6.0 45°
~ fr
.... 4.0 4(f
l::S o ¢is
I

~ 2.0 • lo
ct
35° 0
$
<tbs
0.6 0.7 0.8 0.9 0.6 0.9
eo
1.0 2.0 3.0 4.0 5.0
Ezs,Ezo,EzP (a) (b)
Fig. 3~ Comparison of stress-strain behavior Fig. 4. Angles of internal friction in S, SD, D, and
in D-tests to that in S-tests DS-tests for initial void ratio and void ratio at
failure

Comparison of Internal Angle of Friction cf>n from D-test with that cf>s from S-test
Fig. 4 (a) shows the internal angles of friction by various methods of test. In this section,
discussion will be limited to the comparison of c/>ni cf>ns, and ¢ 8• In D-test, the static confinM
ing pressure t1 88 for a certain value of initial void ratio was varied in four steps between 0. 5
and 2. 0 kg/cm 2 •
As shown in Fig. 4(a), c/>n is larger than c/> 8 by about 3. 5° (8 %) for dense sand, and by about
7. 5° (20%) for loose sand, respectively. This increment of strength can be considered to be due
to densification of sand by repeated loading. Hence the rate of strength increase is larger in
sand with a larger initial void ratio than that with a smaller one.
Since considerable change of void ratio takes place during application of cyclic stress t1n far
up to t1ny, it may be inappropriate to study the dynamic strength of dry sand, based on the
comparison of cf>n to c/> 8 determined from test samples with the same initial void ratio. This
is because cf>n in Fig. 4(a) includes strength increase due to densification, and therefore it will
become possible to evaluate the true effect of dynamic stress system on the strength of dry
sand without the influence of sand densification only by comparing t1ny and (t1 18 -t1 38 ) 1 of the
test sample with the same void ratio at failure. For this purpose, computation was made to
find void ratio ec at the time of application of t1ny, and cf>n was plotted in Fig. 4(b) against
e0, c/> 8 being plotted against eo. Fig. 4(b) shows that c/>n is still larger than c/>8 even when it
was compared on the basis of void ratio at failure. If the peak deviator stress, or the stress
at 15 % axial strain is taken to compute cf>n, as is done in S-test, however, it would become
much larger than that shown in Fig. 4. In any case, cf>D computed from cfny was larger than
<:fas·

NII-Electronic Library Service


DYNAMIC STRENGTH OF SAND 29

The above mentioned fact that c/>n was larger than cf>s when compared even for the void
ratio around yielding, indicates that the strength increase of dry sand under dynamic stress
condition may be considered to be affected by factors other than the decrease of void ratio.
The difference between the ordinates of the chain-dotted curve and solid curve in Fig. 4(b)
gives the strength differences under dynamic and static stress conditions of dry sand with the
same void ratio. This increment of strength may partly be attributed to the effect of strain
rate. The strain rate used in D-test, however, according to Casagrande et al. (1949) and Lee
et al. (1969), does not seem to be of such a magnitude as to give an appreciable effect on cf>D
as is seen in Fig. 4 (b). Eventually the strength increase of dry sand under dynamic stress
conditions should largely be attributed to the particular structure of the sand, brought about
by subjecting to many times of repeated loading, during which a considerable decrease of void
ratio takes place. This structure is quite different from that of the test sample which was
prepared for S-test, but has the same void ratio to that of D-test at failure. It can, therefore,
be concluded that the higher value of c/>n than c/>s cannot be attributed to the substantial dif-
ference of strength of the sand under static and dynamic loading, but it is largely attributable
to the difference of state of the test sample, such as sand structure, due to the different stress
history in both tests, in spite of their same void ratios at failure.
In order to clarify the true effect of the dynamic stress application on the strength of dry
sand, excluding the effect of void ratio and sand structure, it is necessary to carry out both
tests and to compare them, using the test samples which have the same stress histories with
each other until the yielding stress is reached. For this purpose, a modified S-test referred to
as DS-test, was carried out, in which the cyclic stress was applied in the same manner as in
D-test until the yielding stress dny was reached and then it was terminated to transfer to
S-test. Fig. 4(a) and (b) show the values of c/>ns thus measured, using four steps of cell
pressure, and it can be seen from Fig. 4(a) that c/>ns is larger than c/>s by 6 to 11° (14 to 32
%). It is obvious from Fig. 4(b), in which cf>ns was plotted against the void ratio at the end
of cyclic loading, i.e., at the start of the subsequent S-test, that the increase of friction angle
is not entirely attributable to the decrease of void ratio of the test sample by cyclic loading,
since c/>ns is still larger than c/> 8 by 5° even when compared on the basis of void ratio ec and
initial void ratio e0 for S-test.
Fig. 4(b) shows that cf>n is smaller than cf>ns by about 2. 5 to 3. 0°. This means that the
yielding strength determined by D-test is smaller than that· determined by S-test on a test
sample which has the same stress history as that in D-test at the start of S-test. As far as
the comparison of experimental results of D-test to those of S-test is concerned, the sand
sample showed a higher strength against repeated loading than against static loading. However
the yielding strength of the sand against repeated loading became lower than the static stren-
gth, when compared under the conditions where the test samples for both tests were free from
the effect of void ratio and sand structure.

Effect of the Num.ber of Cyclic Load Application on Dynamic Yield Strength


The test results that the dynamic strength of dry sand is larger than its static strength, as
already mentioned, may be attributed to a re-formation of the sand structure caused by many
times applications of repeated loading (Lambe, 1958; Seed et al., 1958). Fig. 5 is a plot of dny
from D-test against the number of cyclic loading N for each step of dn. The points on ab-
scissa C were obtained in such a manner that dn was so raised in every cycle as to reach <1Dy
in about 500 cycles. Their ordinates are smallest among those in the figure. As seen in· the
figure, the dynamic strength of the sand increases with increasing number of dynamic stress
cycles, and its increment amounts to more than 200% for the range of N shown in the figure.
Void ratio of test sample decreases with increasing N, but from the test results in which N
was varied up to 10, 000 under equal intensity of dD, it was observed that the void ratio attains

NII-Electronic Library Service


:.30 TOKI AND KITAGO

an almost constant value in a comparatively


. 100 I J= 3.ot~) • .- small number of N. For example, the de-
1· ~ii crease of void ratio due to the increase of N
~
".'"" 8.0
71 • *.
~.v
I
~.
.,;;.;.-rs-
.--·..Q.--·--r
o· beyond 250 or 500 was negligibly small from
a practical point of view (Toki and Kitago,
·~.
6.0
/i .
4.0 -,,oL-
I
o~·

.
~·--~ -
I

10'"'~"
I ·cJ,s~·

-·-·o-·-
. eo~__2.:.~10--0-·-o
. ~ . I tO
I

.. '>="
1974).
In Fig. 5, however, the strength of loose
sand was increased by 5° as expressed in fric-
4 tion angle when N was increased from 250
20
-~>---1
0 c
· I
to 3, 000, and the difference of void ratios at
failure for both tests was estimated to be less
than O. 05 to 0. 07. The increment of dynamic
100 250 500 1.000 3.000 10000
N
strength was more than anticipated from the
Fig. 5. Increase of cyclic stress at failure with decrease of void ratio, judging from the rela-
increasing number N of cyclic loading
tion between dynamic friction angle ¢ D and
void ratio eo shown in Fig! 4(b). A similar trend was observed in dense sand.
Thus, the re-formation of ~and structure, including particle crushing and change of the property
.of the grain surface resulting from many times applications of the cyclic stress, can be considered
.as the controlling factor of the increase of dynamic strength with increasing N, other than
the effect of void ratio change.
Since the dynamic strength of dry sand depends on N, as mentioned above, the test results
shown in Fig.. 4(a) and (b) cannot necessarily be concluded to be the general relationship be-
tween the strength of dry sand under dynamic and static stress systems. The plots in Fig. 4
are from tests where N =500, and therefore they cannot be valid for other number of stress
·.cycles or for other types of stress application. One example for the latter case was such that
the observed ¢D ·from a modified D-test, in which cyclic stress aD was continuously raised,, :not
in steps, so as to attain- a yielding stress aDY by 120 repetitions of stress application, was found
to be equal to or even a little smaller than c/Js from S-test. .

Dynamic Strength of Dry Sand from a Viewpoint of Deformation


The dynamic strength of dry sand as determined from D-test was concluded to be greater
than the strength from S-test. It often happens, however, that subsoil of loose sand shows a
·.considerable decrease of void ratio when sub-
Jected to such a cyclic loading as an earthqua- s.o~---------/
. ke, and an excessive settlement of the ground eo=0.879
takes place to give serious damage to struc-
.tures. Hence, D and S-tests should be com-
pared from a viewpoint of deformation during
.cyclic loading.
In case of D-test, the volume· change due
to cyclic loading. takes place. almost isotropi-
.cally, and the axial plastic strain ezp indicates
.the axial compop.ent of volume change• ·as o;s . OOv . I --o-- S-test
1.0 4.05(K9/cm~l _ _ D-test
mentioned befqre. · In practice, however, the t5 5.95 I

.settlement of ground due to a decrease of void 2.0 7.90 i


!
.ratio· takes place one-dimensionally in vertical 5.0 10.0 . 15.0
:·direction. Thus the vertical strain ezv due to Ezs, Ezv+ Eio/2 (°lo)
void ratio change in actual ground can become · Fig. 6. Comparison of strain in D-test due to
equal to the .volumetric strain ev. volume decrease and shear deformation to
In Fig. 6, a comparison is shown for. loose · that in S-test

NII-Electronic Library Service


DYNAMIC STRENGTH OF SAND 31

·sand between the axial strain of S-test, and the sum of E:zv and half of the shear component
ezD of the axial strain of D-test. Here, half of E:zD was assumed as the amplitude of shear
strain which changes its magnitude periodically. (ezv+tzD/2) can be considered as the sum of
the volumetric and shear component of the strain created in ground due to cyclic loading.
It should be noted in Fig. 6 that although <1DY is always larger than the static strength (<1u ·
· -<1 88 ) 1 , the total strain (tzv+eZD/2) for the lower level of <JD, is considerably larger than that
in S-test, in contrast to the fact that the total strain in D-test for the higher level of <1D
around (<1is-<1ss) 1 is smaller than that of S-test.
As seen in Fig. 7, except for the case of the smallest initial void ratio, total dynamic axial
strain amounts to two to three times the static axial strain at the same stress level, which was
1/1. 5 times the static strength, i. e., the dynamic stress level was at a safety factor of ·i. 5. In
Table l, the computed values of strains for the three stress conditions are shown and it can
be seen that the total dynamic strain at the stress level of <1D=2/3(<1 18 -<1 8s) 1 amounts to 2. 6
and 4. 4 % for e0 =0. 815 and 0. 879, respectively, although it is much smaller than the static
&train at failure. However, if this amount of strain is induced in a sand deposit 5 m thick,
the settlement of the ground surface would exceed the allowable magnitude for a spread foun-
dation.

J!4.0 • fi~
Table 1. Average axial strain in S and D-tests for
~3.o:~~:g
e three confining pressures (%)
w 2.0
+
~t.0-
0..___...__ _.__ _,
0.60 0.70 0.80 0.90 tzs in S-test at 2/3· (dis-<f 88 ) 1 0.69 0.61 0.92 1. 85.
eo
Fig. 7. Ratio of strain in (tzv+tzD/2) at <fD=2/3·(d 18 -<fss)f 0.58 1.06 2.59 4.35
D-test to that in S-test
both at the stress level Failure strain t 81 in S-test 4. 40 4. 80 8. 50 13. 00
of two thirds of static
strength
From D-test, which was intended to simulate the stress state in the ground at an earthquake,
:it was concluded that the dynamic strength of dry ~and is larger than the static strength.
However, from a viewpoint of deformation, the· sand behavior is less stable against dynamic
stress than against static stress, ·and this trend is remarkable in loose sand.

RESULTS OF SD-TESTS
General Relationship between Stress State and Deformation .
Fig. 8 gives the observed stress-strain relationships of SD-test. in which the ratio Rm of
mobilized stress to static failure stress at the start of cyclic loading was varied. The ratio Rm
is defined as
(1)
where (<1is-<1ss) 1 is the conventionally determined static strength of the sand. In this test, a
deviator stress Rm· (<JIS-<1 88 ) 1 is first statically applied to the test sample, and then the dynamic
stress of <1D is added. ,
In Fig. 8, it can be seen that for a lower range of <JD the axial strain E:zp of SD-test, whic}J.
..starts from the point on the cur:ve of S-test, is larger than that of. S-test at the same stress
level, whereas for a higher range of <Jn the stress-strain curves of SD-test are )ocated above
·that of S-test, and its yielding stress is larger than that of S-test.
It can further be deduced from Fig.. 8 that the slope of the. <Jn vs. ezp curve changes its
.value according to V!lrying (Rm. For a sm~ller Rm,, the increment of e:zp caµsed by increasing db

NII-Electronic Library Service


32 TOKI AND KIT AGO

is small, and its curve crosses the curve of s~


8.0....-----.----------------
...•.... .() test before the failure in S-stst is reached,
·············<>-··············· showing a higher dynamic strength, whereas
/I ..........<>-············ for a larger Rm a considerable amount of ezp
~ 6.0 _, _ _ __,,~
...·· ..•·· --·-@ takes place for a smaller change of an·
"'e . -41-·-· - - - · - I Timmerman et. al. (1969) has already point-
cf. ,...()·· I .------· -=~-i;l"'iF-=-~-:-:--=--~-
:e.... / - : S - test I ed out that the strain caused by cyclic load-
ing is somewhat affected by the static state
of stress of the test sample prior to the ap-
plication of cyclic stress. In the following
Rm paragraphs discussions will be made on the
0 0.0
11'1---~--+-----+--- • 0.326
() 0.500
-+-----I effect of the static stress system on the strength
(!) 0.572 of dry sand subjected to cyclic stress.
Q 0.667
\!) 0.776
~ 0.850 Comparison of <Psn in SD-test with <Ps
5.0 10.0 15.0
As can be seen in Fig. 8, the peak point is
Ezs, Ezp (°lo) not observed in the stress-strain curves of
Fig. 8. Stress-strain behavior with varying mo· SD-test. Hence, the following stresses for
bilized static stress ratio Rm in SD-test the two states of strain are assumed as the
strength in SD-test :
(a) the. stress at 15 % of axial strain, <lnF, and
( b) the stress at the same strain as that at failure of S-test, <lns·
The angles of internal friction computed from <JnF and <lns are referred to as <PsnF and
<Psns, respectively.
In Fig. 9, <PsnF and <Psns are plotted against e0 for the case of
Rm=2/3, and <Ps in S-test is also shown for comparison. The
figure indicates that <Psns is almost equal to ¢ 8 , except for the
case of a larger e 0 , and <PsnF is always larger than <Ps by about 2°.
Since, in general, the magnitude of initial void ratio of a test
sample and the behavior of volume change at failure are closely
related to the strength of a sand, the void ratio change of test
0.6 0.7 0.8 0.9 sample in SD-test, first of all, will be discussed. For Rm=2/3,
~
Fig. 9. Values of angle the volume of a test sample with the larger initial void ratios
of internal friction (0. 878, O. 813) continues to decrease during the application of
in SD-test cyclic stress <Jn until the axial strain of 15 % is reached, whereas
the volume of a test sample with the smaller inital void ratios (O. 732, 0. 656) starts to increase
immediately after the application of <Jn. This type of dilation is due to such a sand structure
as is formed in dense sand by an application of static stress of Rm= 2/3, and readily dilates at
the stress increment which follows (Toki and Kitago, 1973). It can, therefore, be considered
that the larger values of <Psns for a larger initial void ratio than that of <Ps is above all due to
decrese of the void ratio of the test sample. However, in spite of the fact that the void ratio
for dense sand, due to the dilatation at the application of asns or <1snF, is rather larger than
that of S-test at failue, the <Psns or <PsnF is equal to or even larger than ¢ 8 • As mentioned
before the strength increase of sand in D-test was largely ascribed to the re-formation of sand
structure accompanied by void ratio decrease, and a similar trend is seen in SD-test, indicating
that the effect of work hardening due to application of cyclic loading exceeds that of void
ratio increase.
As described above, the dynamic strength of dry sand in SD-test at Rm=2/3, was equal to or
higher than the static strength. On the other hand, the axial strain in SD-test at a stress
level of <1sDF• however, is larger than that of S-test at failure, e.g., the former amounts to

NII-Electronic Library Service


DYNAMIC STRENGTH OF SAND 33

about 1. 2 times the latter for loose sand, and about 3. 4 times for dense sand. aesides, it can
clearly be seen in Fig. 8 that ezp in SD-test is larger than that in S-test for a lower range of
rJn, and that the ratio of €zp in SD-test at the stress level of 5/6 of static strength to the axial.
strain of S-test at the same stress level is calculated to be about 1. 3 for Rm=2/3, regardless.
of the values of eo.
The safety factor of a. sand '!ayer, subjected to such a cyclic stress as was applied in SD-test,
therefore, may be concluded to be greater than that based on the strength from S-test. If
deformation is taken into consideration, however, its safety factor against cyclic stress may be
smaller than that against static loading.
Effects of Static Stress Condition on the Dynamic Strength in SD-test
In the preceding section stress-,.strain behavior of dry sand at a specific value of Rm of 2/3
was presented, and in this section the effects of the degree of static stress mobilization prior
to the application of cyclic loading on the strength of dry sand will be discussed.
In Fig. 10, the ratio of stress ((rJ 18 -rJ 88) +rJsnF) at €zp=l5% in SD-test with varying Rm to
failure stress (d18 -dss) 1 in S-test is plotted against Rm- The figure indicates that the stress in
SD-test at 15% axial strain for a fower range of Rm is larger than the static failure stress, but
that it decreases considerably with increasing Rm to become equal to the static strength at Rm,
=0. 7 to 0. 8, and for a higher Rm than O. 8 it becomes even smaller than the latter.

2.0 l•)
eo 03s
e (Kll,{.J)
,::- <!>0819 1.0
2.0----.----

-J
trI 1.8 -0- ,, 2.5

.
ct0.815 2.0

-t:::>"'
()
Q0.732 1.5

~
1.0
%
eQ656
~
~1.5
... ~ 2.5
$ 1.6
t::) 0~ I \ ct-,
+ ~ \:5 o-~ \
~ 1.4 \ ~ ~i
1. ~ iJ'l 1Q
0 0 ,_'jJ_
\O,
·tf \ ~ . \~ ()

.
.:::::;. 1.2 a: ' \
o,.
~"4. Q r,...-_,____,,,,,_._...._ 'q
'() eooas \.
@""~ 0.5 0 0.656 1.0 (~ \\
1.0
.~~·~ <» 0.732 1.5 ~
•0.879 2.5

0.8
I 0 0.5 1.0
0 0.2 0.4 0.6 0.8 1.0 Rm
Rm Fig. 11. Relationship between remnant
Fig. 10. Ratio of strength in SD-test strength ratio Rr and mobilized
to that in S-test with varying Rm static stress ratio Rm

A similar trend of dynamic strength decrease with increasing Rm can be seen for rJsns· In
Fig. 11, the remnant strength ratio· R 7 , defined as the ratio of rJ sDs to the static remnant
~trength of (r1 18 - d 8s)f' (I-Rm), is plotted against Rm. It is seen in the figure that Rr is larger
than unity for a lower range of Rm, but it decreases linearly with increasing Rm. This means
that a test sample subjected to a higher static deviator stress prior to cyclic loading, easily at-
tains an axial failure strain in S-test by a cyclic loading of lower intensity. In other words,
a sand in a stress state of high Rm readily fails by application of a comparatively low dn.
Rmo defined as Rm where Rr is equal to unity, means that at this value of Rm failure is
caused by an application of rJn, whose magnitude is equal to the remnant strength. Therefore,
if Rm is greater than Rmo• failure takes place on application of lower rJn than the remnant
strength, and it may be said that in such a stress state a sand is less resistant to dynamic

NII-Electronic Library Service


34 TOKI AND KITAGO

1.0--------.......-
E
a:
I
5io.a . ----i--=-..;;::::

ts'
I

~o.s .. --~-,,......,,--~
I:)
II
a: 0.4
Ois
1.0(K%
O"sos O'"SDF Ois
0 $ 1.0(~
<D 1.5 () <I> 1.5
Q e 2,0 0.2 .. -----1~ 9 e 2.0
0.4 • <:>· 2.5
• G> 2S
0.6 0.7 0.8 0.9 0.6 0.7 0.8 0.9
eo eo
Fig. 12. Limiting mobilized static Fig. 13. Remnant strength ratio
stress ratio Rmo versus ini- Rr in case of Rm =0. 83
tial void ratio e0

stress than to static stress.


In Fig. 12, Rmo, computed on <lsnF and <lsns from data similar to those given in Fig. 11,
.are plotted against eo. The values of Rmo on <lsnF are comparatively large and are scarcely re·
lated to eo or <188 , whereas those on <lsns are in between 0. 5 arid 0. 8, and are smaller for
smaller values of e0 and <185 • It follows, therefore, that the dynamic strength thus deter•
mined in the case of a smaller <1 88 and e0 , becomes smaller than the static strength even by
the low mobilization of static strength prior to the cyclic loading.
A foundation designed with a safety factor of 3. 0 has an Rm of 0. 33, and it is clear from
Figs. 10 and 11 that the failure stress in a dynamic stress state is even higher than the static
:strength. In the design of a slope, however, it is common practice for economical reasons to
give a comparatively low factor of safety of 1. 2 for earthquakes, corresponding to Rm=O. 83.
In such a stress state, the decrease of strength against cyclic loading may be anticipated to
result in failure.
In Fig. 13, Rr computed for both cases of <1SDF and <1sns at a static stress mobilization Rm
-of O. 83, which corresponds to a safety factor of 1. 2, is plotted against eo, and the figure indi·
.cates how the reduction of dynamic strength takes place for the lower values of eo and <lss at
.comparatively high Rm. As seen in the figure, Rr for d SDF is not in regular relation to e0 and
.o'88 , and their values do not seem to be too small. Rr for dsns, however, assumes a smaller
value with decreasing e0 and <188 , as it was for Rm 0 , e.g., <lsns at e0 =0. 656 and <185 =1. Okg/cm 2
amounts to only one third of the remnant strength. It may naturally be anticipated that the
dynamic strength will be further reduced for lower values of 0'88 than 1. 0 kg/cm 2, and that a
foundation designed with a lower value of safety factor will be less stable against cyclic loading.

Reasoning for Strength Decrease in SD-test


The reasons for the dynamic strength decrease, as mentioned in the preceding paragraph, at
both lower e0 and 0' 8s, and higher Rm, will be discussed here. Fig. 14 shows the observed re-
lations between void ratio e and stress of ((au-O'ss) +<JsvF) both at axial strain of 15 % in
SD-test, with a varying Rm under the test conditions of e0 of 0. 879 and 0. 656, and 0' 88 of 1. 0
and 2. 5 kg/cm 2 , respectively. In case of loose sand with the initial void ratio eo of 0. 879, the
test samples decrease their void ratio due to cyclic loading except for a few points on the extreme
right part of the curves for larger values of Rm. On the other hand, in case of dense sand

NII-Electronic Library Service


DYNAMIC STRENGTH OF SAND 35

with the initial void ratio e0 of 0. 656, their void ratios at axial strain of 15% are larger than
their initial void ratios, including those points on the extreme left part of the curves for the
smallest value of Rm. In other words, a loose sand 120
~ '~ °" ~

'~
under such a stress system as mentioned above, de- NE-
creases its volume with decreasing Rm, whereas a .;;: o- .:JI <;>~/"'\. •Q 0.879
i
.dense sand increases its volume with increasing Rm. ~ 8.0 --<'->,~·

~.,,
'•' a- \L-o • o.656

1.
In addition to this, such a dynamic strength charac- ..... J,s -"'o~
teristic, as seen in Figs. 12 and 13, where the smaller ;;~
()l<h .::1.0 (lrg
QI I
<Issi the smaller Rr, corresponds to the general strain 40 -~~ I
behavior of sand that volume change due to the
increment of shearing stress becomes larger for a
t:S·~
- 1~~1
.smaller confining . pressure <1 88 • 0
Hence, it can tentatively be concluded that the 0.6 0.1 0.8 e 0.9 1.0
major factor responsible for the decrease of strength Fig. 14. Relationship between strength and
.against cyclic loading at both smaller eo and dss, and void ratio at an axial strain of 15 %
higher Rm, is the increase of void ratio. for varying Rm
The dynamic strength of dense sand with a comparatively low mobilization of static deviator
:stress as seerr in the left part of the curves in Fig. 14, is, however, higher than the static
-Strength, denoted by the letter D in the figure, of the dense sand with the initial void ratio
-0f 0. 656, in spite of the fact that its void ratio at 15 % axial strain is larger than the initial
void ratio. Namely, the sand exhibits a higher strength in a looser state of void ratio at
.failure under dynamic stress conditions than the static strength in a denser state of initial void·
ratio. In direct contrast, for loose sand the dynamic strength is slightly smaller than the
.static one, denoted by L in the figure, in spite of the d~nser void ratio at failure in the
dynamic test than the initial void ratio of the static test. From these experimental results,
therefore, it must be considered that the volumetric behavior of a sand, characterized by the
mobilization ratio Rm of static deviator stress, initial void ratio e0 , static confining pressure <fss
.and so forth, i.e., whether the sand is densified by applying dynamic stress or is loosened by
.doing so, does not seem to be the controlling factor of the dynamic strength of the sand.
Other controlling factors than densifi.cation and structure change, therefore, must be taken into
.consideration.
The state of stress induced in the test sample can be cited as another factor controlling the
.dynamic strength of sand. In D-test, the shearing stress on any plane within the test sample
.alternates its direction in accordance with the rotation of principal plane by 90°, and it follows
that the test sample is subjected to cyclic deformtion of both compression and extension. In
,contrast to this, in SD-test, the rotation of the principal plane ceases to take place under the
,stress condition of (ais-a 88 ) >an, as seen in Fig. 1 (a) and (b), although it still takes place
when Rm is relatively small. However, the deviator stress at the moment when the maximum
principal stress is acting in radial direction becomes smaller in accordance with increasing value
<Of Rm for the same stress level of an. Therefore, under the stress conditions of a large value
-0f Rm, the irreversible component of axial strain accumulates to a large amount, since the
axial compressive strain created by axial cyclic stress dn is scarcely transferred into the axial
.extension strain by subjecting the test sample, half cycle later, to <Jn in radial direction.
This type of deformation may lead to a lower evaluation of the dynamic strength of a sand
.at a higher Rm in SD-test, in which the dynamic strength was determined with respect to
·deformation. Similar behavior to this can be expected in actual ground.
Since a sand under above mentioned stress condition is unstable against cyclic loading, it
may be required to use underestimated strength parameters to design a foundation on sandy
soil.

NII-Electronic Library Service


TOKI AND KITAGO

ANGLE OF INTERNAL FRICTION OF DRY SAND DURING VIBRATION


One of the methods of investigation, which has long been used for the determination of
strength of a soil under dynamic stress conditions, is to apply vibration to the soil mass to
determine the relationships of strength of the soil to the magnitude of acceleration of vibratory
motion.
Tanimoto (1959) in his review of the papers found that the friction angle of soil under
vibratory stress condition decreases in proportion to the logarithm of acceleration, and proposed
an approximate expression given by Eq. (2) for the relation of the two as follows.
µ=O. 5-0. 4 log 7J ( 2)
where
µ=friction angle ratio, i.e., the ratio of friction angle of a soil under vibratory stress
condition to that under static stress condition,
1/ =acceleration ratio, i.e., the ratio of acceleration of vibration to that of gravity,
and 0.1::;;;77::;;;5.
The present authors carried out an experiment, using a triaxial compression test apparatus,
in order to study how a friction angle of a sand varies in accordance with magnitude of
acceleration to which the sand is subjected. Photo. 1 shows the main part of the testing
apparatus. The shaking table is connected to a cam by a crank rod and a maximum accelera·
tion of 1, 500 Gal under 2 Hz can be given to a test sample in a horizontal direction by chang·
ing the excentricity of the cam. The frequency of the vibratory cycle can continuously be
changed from 0. 5 through 10 Hz. The triaxial compression test apparatus is of a stress con-
trolled type. Axial stress was applied through a bellofram cylinder, and the air pressure act·
ing in it was calibrated in advance. Air pressure was also used for confining test samples in
order to lessen the weight of the pressure cell.
The initial void ratios of test samples were 0. 69 and 0. 88, the static confining pressure, 1. 0
and 4. 0 kg/cm 2 , the frequency of vibration 2, 5 and 10 Hz, and the horizontal acceleration of
the shaking table 250, 620 and l, 200 Gal. The test results were compared to those of static
tests. Fig. 15 shows the plots of friction angle ratio µ versus acceleration ratio 1/ for the
present tests. As seen in the figure, the value of µ ranges from O. 92 to 1. 00, irrespective of
the test conditions of e0 , <1 88 and vibration frequency. The rapid decrease of friction angle
ratio µ with increasing acceleration as in the tests reviewed by Tanimoto, was not observed
in the present test. However, the authors' test results coincide with those of Takeshita et al.
(1965) on a dry sand, using a ·direct shear test apparatus under approximately the same test
conditions as in the present tests. His conclusion was that the angle of internal friction of

0.6 0.8 1.0 2.0


1l
Fig. 15. lndependency of angle of internal Photo. 1. Triaxial compression test
friction to acceleration of vibration assembly on shaking table

NII-Electronic Library Service


DYNAMIC STRENGTH OF SAND 37

dry sand did not depend on the magnitude of acceleration to which the test sample was sub·
jected.
In addition to this, the effect of vibration on the coefficient of deformation of the sand,
and on its volume change was not observed, irrespective of vibration frequency, acceleration
or other experimental conditions.
The reason for the discrepancy of the present test results to those from Tanimoto's empirical
equation shown in Fig. 15, can be attributed to the difference of testing method, since. the
test data in Tanimoto's equation were mostly obtained by indirect determination of strength
parameters based, for example, on the qbservations ·of pull-out resistance of a pile in a sand
container on a shaking table, or on the bearing capacity of model footing on sand at similar
stress conditions to the pile test.
The effect of vibration on strength of soil, however, should be studied on the basis of direct
evaluation of the stress state in soil during vibration. The magnitude of acceleration is not
more than an indication of input of dynamic force which is assumed to act on a soil mass.
The stress state created by the input of acceleration of a certain level is not necessarily the
same for all cases, but depends on the boundary conditions of the soil system. It follows
that the strength of soil under such a condition should not be regarded as true, but as ap•
parent.
One of the reasons for using the term "apparent" strength is that there is a question of
determining the strength of soil under vibration. That is the uncertainty in determining the
magnitude of a dynamic component of the forces acting on a test sample on a shaking table
involved in this type of test. When a soil is resisting, under vibratory stress condition, to
both dynamic and static forces, the static component of the whole system of the acting force
is in many cases measured as the dynamic strength of the soil.
It can, therefore, be considered inappropriate to relate the dynamic strength of soil directly
with the magnitude of acceleration to which soil mass is subjected, although it is very con-
venient to do so in practice. In this experiment, the dynamic stresses within the test sample
were supposed to be inappreciably small, although it was submitted to a considerable level of
acceleration by means of the shaking table. As far as the effect of acceleration to which sand
particles are subjected is concerned, it may be concluded that the internal angle of friction of
a sand depends little on the magnitude of acceleration, and that the strength of a sand under
dynamic stress conditions should be studied on the basis of state of stress within· the soil.

CONCLUSIONS
The conclusions derived from the above mentioned series of experiments on the strength of
dry sand under cyclic loading are as follows :
( 1) The internal angle of friction ef>v determined from D-test in which the ,principal
stresses alternate their directions was larger than the conventionally measured static angle ef>s
by 3. 5° through 7. 5° for dense and loose sands, and for a specific number of stress cycles of
500.
( 2) When the number of stress cycles was varied, the dynamic strength of the sand inc-
reased with increasing number of stress cycles.
These two conclusions cannot be attributed entirely to the decrease of the void ratio, but
rather to the change of sand structure.
( 3) Notwithstanding that the sand exhibited a higher strength against cyclic loading than
against static load, it was less stable against deformation under repeated loading than in static
stress state.
( 4 ) The sand showed a higher strength against cyclic loading than against static load in
the case of lower degree of mobilization of static deviator stress prior to the application of cyclic

NII-Electronic Library Service


38 TOKI. AND KITAGO

loading. The critical degree of the static stress mobilization depends on the initial void ratio
of test sample and the static confining pressure. In case of a higher degree of stress mobili-
zation than the critical one, however, the dynamic strength became smaller than the static
strength, and the .rate of strength decrease became larger with decreasing initial. void ratio
and confining pressure.
·( 5 ) The dynamic strength of the sand from D-test was lower than the static strength of
the sand in DS-test, where the void ratio and structure were exactly the same as those of
test sample reaching failure in D-test.
( 6) The angle of internal friction of the sand subjected to vibration depended little on the
magnitude of acceleration, as far as the effect of accelerration to which sand particles were
subjected was concerned.

ACKNOWLEDGEMENTS
The authors are deeply indebted to Messrs. T. Suzuki, H. Matsumura, H. Kataoka, 0.
Motnose, and M. Chiba, without whose help they could no.t carry out the experiments or treat
the test data. ·
Some of the present research were undertaken by the financial aid of a research grant from
the Ministry of Education.

NOTATION
e, e0 , ec=void ratio, initial void ratio, and void ratio at failure
N =number of stress cycles for a certain magnitude of dD
Rm=mobilized static stress ratio (a 1s-d88 ) / (d 18 -d88 ) I
Rmo =limiting mobilized static stress ratio
Rr=remnant strength ratio dn/(d 1s-dss)r (1-Rm)
tzs=axial strain in S-test
tzn, tzp=shear and volumetric components of axial strain in cyclic loading
<fn, dny=cyclic stress and that at failure (yielding)
<lw 0'88 =static axial and confining stresses
dsnF, O'sns=cyclic stress at 15% of axial strain in SD-test and that at failure strain in S-test
</>=angle of internal friction

REFERENCES
1) Casagrande, A. and Shannon, W. L. (1949) : "Strength of soils under dynamic loads," Transactions,
ASCE, Vol. 114, pp. 755-772.
2) Lambe, T. W. (1958) : "The engineering behavior of compacted clay," Journal of the Soil Mechanics
and Foundations Division, ASCE, Vol. 84, No. SM 2, Proc. Paper 1655.
3) Lee, K. L., Seed, H. B. and Dunlop, P. (1969) : "Effect of transient loading on the strength of sand,"
Proc. of 7th ICSMFE, Mexico, Vol. l, pp. 239-247.
4) Seed, H.B., McNeill, R. L. and Guenin, J. (1958) : "Increased resistance to deformation of clay
caused by repeated loading," Journal of the Soil Mechanics and Foundations Division, ASCE, Vol. 84,
No. SM 2, Proc. Paper 1645, pp. 1-28.
5) Takeshita, S. and Futaba, K. (1966) : "Strength of dry sand in vibration," Proceedings of Japan
Earthquake Engineering Symposium, pp. 67-71 (in Japanese).
6) Tanimoto, K. (1959) : "On the dynamic bearing capacity of a soil," Tsuchi-to-Kiso, JSSMFE, Vol. 7,
No. 5, pp. 14-19 (in Japanese).
7) Timmerman, D. H. and Wu, T. H. (1969) : "Behavior of dry sand under cyclic loading," Journal of
the Soil Mechanics and Foundations Division, ASCE, Vol. 95, No.· SM 4, Proc. Paper 6698, pp. 1097-
1112.

NII-Electronic Library Service


DYNAMIC STRENGTH OF SAND 39

8) Toki, S. and Kitago, S. (1973) : "Compaction and deformation of dry sand under cyclic loading,'~
Jour. of JSSMFE, Vol. 13, No. 3, pp. 77-86 (in Japanese).
9) Toki, S. and Kitago, S. (1974) : "Effects of repeated loading on deformation behavior of dry sand,',
Jour. of JSSMFE, Vol. 14, No. 1, pp. 95-103 (in Japanese).
10) Whitman, R. and Healy, K. A. (1962) : "Shear strength of sands during rapid loadings,,, Journal of
the Soil Mechanics and Foundations Division, ASCE, Vol. 127, No. SM 2, Proc. Paper 3102, pp. 99-
132.
(Received December 20, 1973)

NII-Electronic Library Service

You might also like