MAY 1975, ow
JOURNAL OF THE
GEOTECHNICAL
ENGINEERING DIVISION
FUNDAMENTALS OF LIQUEFACTION
‘uNDER CycLic Loabinc
By Geotfrey R. Martin," W. D. Liam Finn,?
‘and H. Bolton Seed,’ Members, ASCE
lwnooucron
pressure builds up to a value equal
‘sand, having
may be only
for the purposes of
‘recognized that other criteria may be necessary
Wote.—Diseassion open until October 1,
ied with the Editor of Te
anaas MAY 1975, ors
‘The mechanism of the progressive increase of pore-water pressure during
undrained cyeie loading is investigated and quantitative relationships between
Yolume reductions during drained cyclic loading and corresponding pore-water
increases under undrained conditions are established, These relationships
Permit the theoretical computation of the progressive increase of pore-water
Pressure during, undrained cyclic loading vsing basic physical parameters of
The sund, and thus, they also provide means for coupling such pore-water pressure
“dynamic response analyses of horizontal deposits subjected to
‘earthquakes,
Prewous Lioveracron STUDIES,
Considering the id of horizontal sand
formation, the most significant deformations and stre
by an earthqual
bedrock. Asa
toa sequence
to the earthquake are represented by the principal
K,0,q. in which K, = the coefficient of earth pres
‘effective sires. During the earthquake, the element wil
Shear sttesses, yyy the magnitude of which will vary
to cycle.
1s based on approximations or simulations of the preceding
have been the principal means of studying liquefaction. Compre
faction
potential have been pres
‘The cyclic triaxial
various factors contro
form of such a test,
fan ambient press
under undrained conditions. Stress conditions on a plane
‘sample approximate those on the horizontal planes shown i
end of each s
progressively with increasing number of cycles. The rate of pore-water pressure
Gevelopment accelerates as liguefaction is approached, at which point strain
amplitudes repidly increase, The nature of pore-water pressure increases due
to nonuniform cyclic deviator stresses simulating the irregular loading from
actual earthquakes also has been studied (13,14).
itations to the eyci
(zero lateral deformation), and a eyclic
‘with drainage prevented and with zero lateral deformation co
pore-water pressure
lie fluctuationsin pore-water
magnitude of the buildup of pore-water pressure
for both triaxial and simple shear tests have been.‘ows LIQUEFACTION 425
well documented (4,23), For simple shear tests with a uniform cyclic shear
stress, results may
Torsion tests G3,
‘examine the undrained beh
Practical applications of
iustrated by Seed and Idi
e oe
Pen to
Tv
(1 Cycle Looe cui
3. 1 —Idealized Field Loading Conditions
of sh
‘sand as the pore-waler pressures progressively increase26 MAY 1975, ors
‘are not taken into account. Obviously an analysis where progressive pore-water
response analysis is more
later herein.
‘Ghaboussi and Wilson (7,8) have presented a generalized numerical method
for the coupled dynamic anal ied je medium,
gation theory developed
he method of analysis, por:
were introduced ig pseudocyclic and permanent void ratio changes.
However, a method for assigning such volume changes was not established.
‘Vou Cuan Crunacrensres noe Dranen Cyouc Lownie
Considerations of volume changes in drained cyclic tests and their relationship
to pore-water pressure increases during undrained cyclic loading are restricted
‘conditions, which best simulate field deformations induced
been studied by Silver and
conditions are induced b izontal motion between the cap and base.
‘Youd (25) has shown th ral boundaries of the sample deform reasonably
uniformly for shear strains as large as 5% and although the absence of comple-
“hear stresses produces stress concentrations within the sample, Lucks,
7) have shown that approx 70% of the sample has a uniform stress
‘equal to volumetric strains.
icates that for a given density and number of cycles, volumetric
ress for a given cyclic shear strain
‘changes during cyclic loading are
amplitudes, as for a constant shear strain amplitude the lat
increasing vertical stress. Tests also"show that for a given density and number
of cycles, volumetric strains are reasonably proportional to the cyclic shear
‘strain amplitude for strain amplitudes less than 0.3%. By measuring
ing from the applied shear strains, changes in shear modulusLIQUEFACTION a
amplitude) with number of cycles may be
juring cyclic torsion
increase and shear moduli decrease with
independent of frequency in the range, 0.2 Hz to 2 He.
Reusnowsn serween Vote Reduction Dunina Dram Gveus Tess AN
one: Wares Presoune neneast1w Unonanen Tests
‘rain contacts
Consider now a cycle of loading having the same shear strain a
‘ndrained conditions x
slip deformation mus
Fees to the more incompressible
erement of recoverable vo
Ae, ‘grain contacts. For volumetric compatibility
at the end of the load cycle, we must have:
‘change in volume of voids = net change in volume of sand structure. . . . - (D)
If Au = the increase in residual pore pressure for the cys
le; and B,a8 MAY 1875 os
@b)
wheres E, is generally of the onder
‘of magnitude ofthe moduli, the water
pressible, and this, under conditions
For s
of 108 pf. C
may be assur
‘of zero volume change:
legree of saturation, S,
r values of S, = 9%, and thus, from
cycle would be
reduced considerably. during undrained
cyclic tests is apparer are to be obtained.
Tn the remainder of the paper complete saturatio
of Eq, 3 for pore-water pressure calculations is developed further.
‘Avazyss oF Voue Cuants Dunia Daaneo Nowraron Creu Loanwe
During nonuniform cyclic toa that occuring during earthquakes or
during stress controled cyclic I the volumetric strain increments per
cycle, Ae,,, need to be compi nonuniform sequence of cyclic strain
‘amplitudes. While volume change characteristics during uniform strain amplitude
Sequences have been well established by the investigations of Silver and Seed
(25,26), a method for computing volume changes during nonuniform sequences
using basic phy
density of 43% and were normally loaded to a vertical stress
‘of 4,000 psf. Following tests to establish the volume-change characteristics
for uniform cyclic shear strain amplitude cycles (frequency = 1 Hz), a series
im shear strain amplitudes was carried out and @
‘method of correlation attempted.
strain curves for uniform shear strain amplitude cycles are
ly the curve for 0.396 shear strain amplitude was establishedFIG. 4—Volumetric Strain Curves for
Constant Cylie Shear Strain Amplitude
Toate
FIG, 6.—Ineremental Volum
Curves
tle Strain
) J
the (N+ Ith eycle may be
tude.
determined480 MAY 1975, ors
by finding the point of intersection of the horizontal line corresponding to
&,g= OAM with the 7 = 0.2% curve shown in Fig. 4 and me val
of the volumetric strain increment, Ae,q appropriate to a further cycle
on the 0.2% shear strain curve, ie., Atyq = 0.06%. Thus,
Volumetric strain may be envisaged to represent a strain history parameter.
"To enable such a procedure o be conveniently used for numerical calculations,
f changes
corresponding volumetric strain increment/cycle also is doubled to sati
observation that volumetric strain is proportional to shear strain amplitude.
fortant feature of the curves shown in Fig. 5 is that for a given
‘the change in volumetric strain, Ae,, = Oat
ide, as given by the intersection Of the curves with the
‘axis, Therefore, for example, if the volumetric strain reaches 0.87%
les of 0.19%
ly as described subsequently and is indicative of the asymptotic
e curves shown in Fig. 4
“To permit the family of curves shown in Fig. 5to be utilized by a computer,
they were fitted to an analytical function of the form
‘The evaluation of the four constants, Cy, Cy» Cy, and Cy, using data plot
stant strain amplitude cyclic tests, completely defines
ing. The constant, C, = the slope
of the ey = 0 line, while the remaining three constants may be computed by
constant €y cont
hree resulting simultaneous equations.
‘eq and y are expressed as a percentage.
“Fest results for the nonuniform cyclic shear strain amplitude sequences are
imetrc strain function given by Eq. 4, volumetric
ere predicted by means of a computer program.
‘The measured and computed volumetric strains were found to be in remarkably
‘good agreement, as shown, indicating the success of the computational method
‘adopted. Tests C and D show that once large volumetric strains have occurred,
Tow cyclic shear strain amplitudes have no effect in increasing the volumetric
strain
CCaueuanon oF Sean Srman AneL Tunes Dunaa Stas Conrmots> LonDwa
To calculate volume changes during stress controlled cyclic simple shear tests,
the calculation of the corresponding cyclic shear strain amplitudes is necessary
For drained cyclic tests, the shear modulus is a function of both the shearFIG, &.—Comperison between Analyt
‘cl and Exparimental One-Dimensional
Unloading Curves (1 psf = 47.9 N/m*)
tHonahips (1 paf =
FIG. 9.—Computed Results for Undrained Strees Controlied Cyclic Simple Shear
‘Teste (1 pat = 47:8 N/m?)
i»
FIG, 10.—Computed Results for Number of Gyles Causing Liqueectionon
loading, as previously described,
shear modulus for any cycle
«2
strain amplitude and
land thus a method for pre
43 contro i test sequence is necessary.
he eycic shear stress amplitudes during the tests described
previously enabled the changes in shear stress amplitude with shear strain
amplitude and volumetric
ddata revealed that the effect of the number of cycles on the shear modulus
into account by expressing the shear stress-shear strain amplitude
fa function of the accumulated volumetric strain, €,4. For a given,
train, the smoothed shear stress-shear strain amplitude curves were
7. By assuming that the
ly general fr ing their use during undrained
‘carves shown in Fig, 7 were fitted to a hyperbolic
in which the parameters, @ and b= functions of ¢,,, The parameters, a and
», were in turn fitted to functions of the form
a=A,- 6
YO A Ase °
and b= B,-——*— . ”
B+ Bye
‘The evaluation of the six constants, A,
plotied from say three constant strain amplitude cy:
le behavior of sa
0.
, iche
‘The shear mod any cycle ‘can be evaluated from the ratio
of shear stress to shear strain amplitude as expressed by the functions shown
in Fig. 7. The changes in shear modulus measured during the experimental
the function given by Eq. 5, the changes
of strain history in increasing the modulus is i
and b on the in amplitude plot have approximately the same amplitude,
but the shear modulus is approx 80% greater for point b than for point a
‘Ana.vets oF One-Divensona. Untoanwa CnanAcrensmies
‘The use of the pore-water pressure prediction theory requires a knowledge
of the recoverable deformation characteristics of sand during one-dimensional
unloading from a. given i
fof sand in one-dimensional compression has been studied by Hendronor LIQUEFACTION 43
‘The volumetric or vertical strains on loading to a vertical stress, 4, may be
or slip strain component,
‘component, €,,, resulting
1s shown in Fig.
il vertical stresses, o'» are geometrically simi
thus they may be related by an expression of the fOrM, €4,/€ ye = (6/0,
in which ¢,,, = the magnitude of recoverable strain on unloading from a st
turn, may be related to a7, by a curve of the form,
min Fig. 8. Therefore, the general form of the unloading
senna (2) =
‘The unloading modulus, B,, at any stress, 0}, is given by:
‘The controlling parameters, m, n, and k,, may be obtained readily from a series
of say three unloading curves from different initial vertical stresses ,,
Pone- Waren Prescune Wengases Dura Unoranen Srmess
Covrnouso Creuc Tests
eyelic simple shear
{0 compare pore-wat
s-been well documented (4 interest
predict ial expres-
sions developed with experimental results for the particular case of a sequence
of uniform shear stress cycles. The basic physical parameters
the test program on the crystal silica sand were used in all computations
‘The procedure adopted for each cycle of the calculations is as fc
1, The available initial conditions prior to each loaé
the vertical effective stress,
‘and corresponding pore:
cycle are the shear
‘and the accumulated
pressure u from the
previous loading
2. he stress-strain amplitude curves shown in Fig. 7, the shear strain
amplitude corresponding to the applied shear stress amplitude may be computed
using Eq. 5.
3, From the calculated shear strain amplitude, the increment of volumetric
slip strain, Ae,g, for the cycle can be computed from the curves shown in
Fig. 5 by making use of Eq. 4434 MAY 1975 ows
4, The unloading modulus, E,, corresponding to the initial conditions may
be computed from Ea. 9
'. By substituting for Se,and
pressure, Aw, occurring during
to pore-water pressure u existing
3, the increment of residual pore-water
ad cycle can be determined and added
start ofthe eyele.
By programming this sequence of calculations for a computer, the time history
id pore-water pressu
i
and the
pressure show a
fe as the vertical eff
characteristic rapi
stress approaches zer
ferent relative densities
iso probably due to minor
to obtain the drained
ind Seed and Peacock
as sample uniformity
ite equipment used forthe drained cyclic tests was unsuitable for undrained
‘Nevertheless, the general agreement of the form of computed results
1 shows computed pore-water pressure
ith varying values of o},, The
res and number of cycles to cause
ineach ease are very simil
form Stress Cyces,—The computation method described inthe preceding
can be used equally well to compute pore-water pressure increases resulting
rm sequence of stress pulses, e.g., those occurring during an
‘earthquake. Indeed, this feature would permit the computation method to be
coupled with dynamic response analyses as previously mentioned.
te the effects of nonuniform stress cycles,
to form three different time histories, as shown in Fig.
pressure increases computed for each sequence.
jeance of the order of stress cycles is apparent, For time history
large cycles result in a rapid increase of the volumetric slip strains,
the subsequent volumetric slip strains and corresponding
om
the end of the sequence are still effective in producing pore-water pressure435
ler cycles following a
pressure increases
‘of nonuniform cyclic
at methods used to convert
FIG. 11.—Computed Results Showing Etfect of lie! Vertical Stress on Pore-Watar
Pressure Ineroates, z,,/2, Constant (1 pat = 47.9 N/m*)
Aa
FIG, 12.—Computed Results Showing Effect of Nonuniform Shear Stress Cycles on
Pore-Water Pressure Incranses (1 psf = 47.9 N/m?)
‘ nonuniform cycle sequence to an equivalent number of uniform cycles should
sighting technique.
thie zero stress axis, However, the
{is based on symmetric full cycle calculations coul
calculations,
\d be adapted easly for half-cycle438 May 1975, ors
sand.
1e case of undrained cyclic simple shear
progressive pore-water pre
trated horizontal sand dep
at the computational methods
improving the understanding of
potential during earthquakes.
1. Biot, M. A. “Theory of Propagation of Elastic Waves in a Flid Saturated Porous
1 “Journal of the Accoustical Society of America,“ors LIQUEFACTION a7
snd 3
[ASCE, Vol. 97, No. SMS, Proc.
4, Tokyo, Japan, Des.
‘Saad Liguefction under Random Earthquake Loading
‘World Conference on Earthquake Engineering, Rome,
KL, and Fitton, J. Au, “Factors Affecting the Dynamic Strength of Sol
tion Effects on Solts and Foundations, ASTM STP4S0, American Society for
ing and Material
6.
IH, B,, “Sand Liquefaction under Cylic Louding Simple
‘Soil Mechanics and Foundations Division, ASCE,
o
‘of Soil Liquefaction: Nigta Earthquake,”
‘ASCE, Vel 98, No. SM3,
Procedure for Evaluating Soil Liquefaction
cs and Foundations Division, ASCE, Vol.
pp. 1249127.
faction of Sturated Sands Dating Cyclic Loading.”
Vol 92, No. SMB,
Procedures for Measuring Soil Lian
‘Mechanles and Foundations Divisi
fon Characteristics of Sands Under Cyclic
‘ASCE, Vol.
Journal of
Proc. Paper 8384, Septoe
don Computations by
‘Structural
The following symbols are used inthis paper:
constants;
function of volumetric strain (Eq. 6):
leading;
Ac, increment of recoverable volumetric strain occuring during
‘one-dimensional unloading:
gq = sccumulated volumetric slip strain from cyclic loading
sequence;
«,, = recoverable volumetric strain from one-dimensional load-