You are on page 1of 12

High Overburden Stress Effects in Liquefaction Analyses

Ross W. Boulanger, M.ASCE1

Abstract: A reevaluation is presented of two factors that can strongly affect the estimation of liquefaction resistance for clean sands
under high effective overburden stresses (␴ ⬘v ): the relation used to normalize penetration resistances to a ␴ ⬘v of 1 atm 共i.e., C N ), and the
adjustment factor for the effects of ␴ ⬘v on cyclic resistance ratio 共i.e., K ␴ ). These two factors have been investigated in a number of ways
and several relations exist for each of them. An improved C N relation is developed based on cone penetration theory and validation against
calibration chamber test data for both cone penetration and standard penetration tests. A relative state parameter index (␰ R ) is shown to
Downloaded from ascelibrary.org by Tongji University on 11/19/18. Copyright ASCE. For personal use only; all rights reserved.

provide a consistent theoretical framework for interrelating the penetration and cyclic loading resistances. It is subsequently shown that
the C N and K ␴ relations are interrelated through the sand properties and relative density (D R ) in ways that have compensating effects on
the predicted cyclic resistance. The derived relations provide an improved representation of the effects of high ␴ ⬘v levels, and reduce the
conservatism that results when some established relations are extended to ␴ ⬘v levels higher than they were calibrated for.
DOI: 10.1061/共ASCE兲1090-0241共2003兲129:12共1071兲
CE Database subject headings: Liquefaction; Sand; Cone penetration tests; Standard penetration tests; Stress.

Introduction related to penetration testing at large depths 共e.g., possible use of


a downhole CPT to achieve large depths, energy transmission in
Effective overburden stress (␴ ⬘v ) is accounted for in semiempir- SPT tests, etc.兲 are beyond the scope of this paper.
ical cone penetration test 共CPT兲- and standard penetration test
The effect of ␴ ⬘v on penetration resistances 共CPT or SPT兲 in
共SPT兲-based liquefaction analysis procedures for clean sand
clean sand has been recognized in semiempirical liquefaction
共⬍5% passing the No. 200 sieve兲 through the normalization of
penetration resistances to an effective overburden stress of 1 atm evaluations since the development of the simplified procedure
共i.e., C N ) and the adjustment factor for the effects of overburden 共Seed and Idriss 1971兲. Data on the C N factor for normalizing
stress on cyclic resistance ratio 共i.e., K ␴ ). Overburden stress also SPT N values to an effective overburden stress of one atmosphere
affects the adjustment factor for the effects of static shear stress has come from calibration chamber tests 共e.g., Marcuson and Bie-
on cyclic resistance ratio 共i.e., K ␣ ), which is applicable beneath ganousky 1977a,b兲 and reviews of field data 共e.g., Skempton
slopes. Seed 共1983兲 introduced the K ␴ and K ␣ factors as a means 1986兲, and various expressions that approximate those data have
for extending the SPT-based liquefaction correlations to greater been widely adopted 共e.g., Liao and Whitman 1986兲. The avail-
depths and from level ground conditions to sloping ground con- able SPT data are, however, limited to ␴ ⬘v / P a less than about 5.5,
ditions. These factors are applied as and thus the extension of C N to larger ␴ ⬘v / P a values requires
CRR⫽CRR␴⫽1,␣⫽0 K ␴ K ␣ (1) extrapolation of purely empirical expressions, sometimes tem-
pered by engineering judgment. The C N factor for CPT tip resis-
where ␣⫽static horizontal shear stress ratio (␣⫽␶ s /␴ ⬘v0 ); tance (q c ) is much better defined, being able to draw upon large
␶ s ⫽static horizontal shear stress; ␴ ⬘v 0 ⫽vertical effective consoli- data sets of calibration chamber studies and rigorous numerical
dation stress; CRR␴⫽1,␣⫽0 ⫽cyclic resistance ratio for ␴ ⬘v 0 / P a analyses. For example, Salgado et al. 共1997a,b兲 obtained excel-
⫽1 and ␣⫽0 as obtained through a semiempirical correlation for
lent agreement between numerical analyses and results of over
the earthquake magnitude and other conditions under consider-
400 calibration chamber tests, with the data extending to equiva-
ation; and P a ⫽atmospheric pressure. The effects of ␴ ⬘v on K ␣ are
discussed in Boulanger 共2003兲 and Idriss and Boulanger 共2003兲, lent ␴ ⬘v / P a values of about 7. The extension of CPT C N values to
while the present paper focuses on how ␴ ⬘v affects liquefaction larger ␴ ⬘v / P a values also requires extrapolation, but in this case
evaluations via the C N and K ␴ relations. Together, the C N and K ␴ the extrapolation benefits from the use of a well-calibrated theo-
relations can be critical for liquefaction evaluations at high over- retical solution that captures the essential physics of the problem.
burden stresses 共e.g., ␴ ⬘v / P a ⬎4), such as can be encountered be- Both the SPT and CPT data show a clear dependency of C N on
neath large earth dams or embankments. Other important issues relative density (D R ), with C N values at ␴ ⬘v / P a ⬎1 being progres-
sively larger with increasing D R . Furthermore, the C N factors for
1
Professor, Dept. of Civil and Environmental Engineering, Univ. of SPT and CPT penetration resistances are often assumed to be
California, Davis, CA 95616. equal because of the similarities in the experimental results and
Note. Discussion open until May 1, 2004. Separate discussions must the expectation that they are affected by ␴ ⬘v in similar ways.
be submitted for individual papers. To extend the closing date by one The overburden correction factor (K ␴ ) for cyclic resistance
month, a written request must be filed with the ASCE Managing Editor. ratio continues to be a subject of study and review 共e.g., Seed and
The manuscript for this paper was submitted for review and possible
Harder 1990; Pillai and Byrne 1994; Harder and Boulanger 1997;
publication on December 27, 2001; approved on June 25, 2002. This
paper is part of the Journal of Geotechnical and Geoenvironmental Hynes and Olsen 1998兲. For example, the experimental data in
Engineering, Vol. 129, No. 12, December 1, 2003. ©ASCE, ISSN 1090- Fig. 1 共Vaid and Sivathayalan 1996兲 clearly shows that K ␴ is
0241/2003/12-1071–1082/$18.00. dependent on D R , with K ␴ values at ␴ ⬘v / P a ⬎1 becoming smaller

JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / DECEMBER 2003 / 1071

J. Geotech. Geoenviron. Eng., 2003, 129(12): 1071-1082


clean sands are shown to support the use of the relative state
parameter index (␰ R) for this purpose.
3. Derive relations between CRR and ␰ R that are applicable to
field conditions with ␴ ⬘v / P a ⫽1 as embodied in current semi-
empirical liquefaction correlations.
4. Extend the derived ‘‘field’’ CRR– ␰ R relations to higher over-
burden stresses. A convenient implementation of this step
was achieved by the introduction of state normalized pen-
etration resistances.
5. Compare the ␰ R -based approach to current practice, and then
evaluate further simplified relations that can be used to ap-
proximate the results of the ␰ R -based approach.
The modifications to practice suggested herein are both simple to
implement and reduce the conservatism imposed at high overbur-
Downloaded from ascelibrary.org by Tongji University on 11/19/18. Copyright ASCE. For personal use only; all rights reserved.

den stresses by some of the current approaches.

Overburden Normalization of Penetration


Resistance

Cone Penetration Test C N Relation Based


Fig. 1. Comparison of K ␴ relations with data from reconstituted on Calibrated Penetration Theory
Fraser delta sand specimens 共Vaid and Sivathayalan 1996兲 and vari- Salgado et al. 共1997a兲 compared q c values predicted by a numeri-
ous field samples 共Seed and Harder 1990兲 cal analysis method to the measured values for about 400 calibra-
tion chamber tests on clean sands. The numerical analyses were
based on computing the cylindrical cavity expansion limit pres-
sure (p l ) and then performing a stress rotation analysis to obtain
with increasing D R , and that K ␴ is different for simple shear the cone penetration resistance (q c ). The stress–strain behavior
versus triaxial loading conditions. Further complicating the issue was represented using a form of the Hardin equation for small-
is the wide scatter in published K ␴ data for different soils 共includ- strain shear modulus 共Hardin and Black 1968兲, a modified form of
ing reconstituted specimens and field samples兲, which translates the K – G model 共Naylor et al. 1981兲 for describing the nonlinear-
into uncertainty in the appropriate selection of K ␴ for projects ity of the tangent shear and bulk moduli, and the relation by
where testing of field samples is not feasible. Bolton 共1986兲 for describing the peak friction and dilatancy
The fact that C N and K ␴ are both dependent on D R is clearly angles. Details of the theory and its implementation are presented
consistent with critical state concepts wherein the behavior of in Salgado et al. 共1997a兲. The numerical analyses were shown to
sand depends on its state 共via ␴ ⬘v and D R ) rather than ␴ ⬘v or D R provide excellent agreement with the experimental data, provid-
alone. With regard to K ␴ , Pillai and Muhunthan 共2001兲 observed ing reasonable validation of the theory for predominantly silica
that the cyclic resistance ratio 共CRR兲 of clean sand was approxi- sands of medium compressibility. Salgado et al. 共1997b兲 subse-
mately constant for a given value of state parameter 共␰兲 and that quently developed relations between q c , D R , ␴ ⬘v , and the lateral
K ␴ depends on the critical state parameters of the sand. Boulanger earth pressure coefficient at rest (K 0 ) for typical, upper bound,
共2003兲 noted the practical difficulties in determining critical state and lower bound sets of soil properties. The results of those nu-
properties for most site-specific projects, and instead proposed a merical analyses are closely approximated as
relative state parameter index (␰ R ) derived from the relative dila-
tancy index (I RD) of Bolton 共1986兲 and embodying critical state
concepts in an approximate, yet practical, manner. Boulanger
qc
Pa
⫽C 0 C 1 冉 冊冉 冊
␴ ⬘v
Pa
m
K0
0.45
m⫺0.077
(2a)

共2003兲 showed that the effect of ␴ ⬘v and D R on the static shear m⫽0.7836⫺0.5208D R (2b)
stress ratio correction factor (K ␣ ) could be reasonably and ratio-
nally represented by ␰ R . In the present study, the ␰ R index is C 0 ⫽25.7⫹39.7D R ⫹212.3D R2 (2c)
shown to also provide a rational basis for interrelating CRR, D R , C 1 ⫽1.0 typical soil property set
and ␴ ⬘v 共as typically represented through C N and K ␴ ).
The purpose of this study is to reevaluate overburden stress C 1 ⫽0.64 lower bound property set (2d)
effects 共e.g., C N and K ␴ relations兲 on liquefaction analysis proce-
C 1 ⫽1.55 upper bound property set
dures for clean sands using a consistent theoretical framework for
interrelating the penetration and cyclic loading resistances. The with K 0 ⫽0.45 being considered as reasonable for most normally
steps involved in this study were as follows. consolidated sands. The resulting relations between q c , ␴ ⬘v , and
1. Derive a C N relation that is based on penetration theory and D R are plotted in Fig. 2 for the typical soil property set and K 0
validated against calibration chamber test data for both CPT ⫽0.45.
and SPT tests. The overburden normalization 共or correction兲 of q c to ␴ ⬘v / P a
2. Establish a fundamental framework that: 共1兲 describes how ⫽1 is performed as
D R and ␴ ⬘v affect CRR; 共2兲 is consistent with the theory
q C1 ⫽C N q C (3)
underlying the above derivations for CN; and 共3兲 is simple to
implement in practice. Laboratory test data on reconstituted where the C N can be obtained from Eq. 共2兲

1072 / JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / DECEMBER 2003

J. Geotech. Geoenviron. Eng., 2003, 129(12): 1071-1082


Downloaded from ascelibrary.org by Tongji University on 11/19/18. Copyright ASCE. For personal use only; all rights reserved.

Fig. 2. Relations between relative density, confining stress, and cone


penetration test tip resistance for clean sand with typical properties
共after Salgado et al. 1997b兲

C N⫽ 冉 冊
Pa
␴ ⬘v
m

(4)

The CN relation in Eq. 共4兲 is independent of the material property


set 共as represented by C 1 ) and matches the widely used relation
by Liao and Whitman 共1986兲 Fig. 3. Overburden normalization exponent m from nonlinear regres-

冉 冊
sion on Marcuson and Bieganousky’s 共1977a,b兲 standard penetration
0.5
Pa test calibration chamber test data: 共a兲 m versus D R and 共b兲 m versus
C N⫽ (5) N1
␴ ⬘v
at a D R of 54%.
the data set兲. The resulting nonlinear regression for N was first
weighted by the inverse of the number of data points in each bin,
Standard Penetration Test C N Relation Based
thereby giving equal weight to the different bins. The regression
on Reevaluation of Calibration Chamber Test Data
was further weighted by 1/N 1.5 to account for the nonuniform
Standard penetration test calibration chamber test data by Marcu- variability in N; weights of 1/N and 1/N 2 were also evaluated, but
son and Bieganousky 共1977a,b兲 were reevaluated using a 1/N 1.5 appeared to be more closely proportional to the inverse
weighted least-squares nonlinear regression that better defines the variance in N across the entire data set. The results of the regres-
effects of D R on C N . The reevaluated data were for Platte River sion are summarized in Fig. 3 showing m versus both D R and N 1
sand 共median grain diameter D 50⫽2.0 mm, coefficient of unifor- 共all plotted for the mean D R within each data bin兲.
mity c u ⫽5.5), standard concrete sand (D 50⫽0.49 mm, c u ⫽2.2), The variation of m with D R and N 1 is consistent for the three
and Reid Bedford model sand (D 50⫽0.25 mm, c u ⫽1.6). For the sands, with m decreasing by more than a factor of about 2 from
Reid Bedford model sand, the reevaluated data were limited to the loosest to densest conditions. Some of the variability in the m
cases where the specimens were placed by rotating rainer, sub- values can be attributed to the experimental difficulty in achieving
merged, and normally consolidated 共the data for other conditions repeatable test data 共specimen D R and N values兲, the use of only
were too limited for this regression兲. Standard penetration tests three different confining stresses, and the sensitivity to small ex-
were generally performed at three different confining stresses perimental variations when N values are in the single digit range.
(␴ ⬘v / P a ⫽0.68, 2.77, and 5.44兲 and three different relative densi- The weighting scheme used in regression had relatively small
ties (D R from ⬇25 to ⬇96%兲, and thus the data for each of the effects on the majority of the m values 共except for two cases兲 and
sands were grouped into bins of common ␴ ⬘v / P a and D R 共allow- little effect on the average trend in Fig. 3. Overall, the results are
ing for the experimental variations兲. The nonlinear regression very consistent and produce an average m – D R relation 关solid line
equation follows from the common normalization in Fig. 3共a兲兴 that is very close to the previously derived CPT
relation 关dashed line in Fig. 3共a兲兴. The average m – N 1 relation, as
N 1 ⫽C N N (6)
shown in Fig. 3共b兲, is similarly very close to that produced by
where C N was assumed to have the same form as Eq. 共4兲. Then, combining the CPT relation with a common D R – N 1 correlation
both N 1 and the exponent m were expressed as third-order poly- 共these types of correlations are discussed later兲. The agreement
nomials of D R , with the constraints that N 1 always increases and between the SPT and CPT relations would be even better if the
m always decreases with increasing D R 共over the range of D R in one m value for Reid Bedford model sand at a mean D R of 30%

JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / DECEMBER 2003 / 1073

J. Geotech. Geoenviron. Eng., 2003, 129(12): 1071-1082


Downloaded from ascelibrary.org by Tongji University on 11/19/18. Copyright ASCE. For personal use only; all rights reserved.

Fig. 5. Use of relative dilatancy index (I RD) to predict dilatancy


angle in drained triaxial compression tests 共after Bolton 1986兲

Fig. 5, where the difference between peak and critical state fric-
tion angles (␾ ⬘peak⫺␾ cs
⬘ ) in drained triaxial compression was rea-
Fig. 4. Recommended overburden normalization factor (C N ) for sonably estimated as 3I RD for 12 sands.
clean sands and regression results for two sands from Marcuson and The I RD relation in Eq. 共7兲 can be used to derive a correspond-
Bieganousky 共1977b兲 ing state parameter 共␰兲 共Been and Jefferies 1985兲, where ␰ is the
difference between the current void ratio 共e兲 and the critical state
void ratio (e cs) for the current value of p ⬘ . The definition of ␰ is
was omitted; This m value corresponds to the lowest mean N 1 shown on Fig. 6, along with the critical state line produced from
value, only 2.5, and hence it is the most susceptible to experimen- the I RD relation with Q⫽10 共i.e., critical state corresponds to
tal errors 共including the integer reporting of N values兲. Given the I RD⫽0). The critical state line shows a notable increase in slope
similarity of results in Fig. 3 and the experimental limitations on the e – ln p⬘ plot at p ⬘ / P a values of about 10–20. This curva-
noted above, the expression for m in Eq. 共2b兲 was adopted as ture of the critical state line has been observed for many sands,
providing a satisfactory C N approximation for both CPT and SPT reflecting the onset of significant particle crushing, and its posi-
tests. tion is controlled by the parameter Q. The value of ␰ can be
The C N relation from Eqs. 共2b兲 and 共4兲 is plotted in Fig. 4 normalized by the difference in the maximum and minimum void
along with the SPT regression results for the Platte River and ratios (e max⫺emin) to arrive at a relative state parameter (␰ R ) that
standard concrete sands for comparison. The importance of ac- provides improved correlations to the shear behavior of sand
counting for D R clearly increases with increasing confining stress 共Konrad 1988兲. This leads to
共or depth兲, with C N values for D R of 40 and 80% differing by a
factor of 1.62 at ␴ ⬘v / P a ⫽10.

Relating Cyclic Resistance Ratio to D R and ␴ v⬘


through ␰ R

Relative State Parameter Index, ␰ R


A key feature of the CPT penetration theory by Salgado et al.
共1997a兲 is that volume change characteristics and peak friction
angle relations are based on Bolton’s 共1986兲 relative dilatancy
index (I RD)


I RD⫽D R Q⫺ln
100p ⬘
Pa 冊⫺R (7)

where p ⬘ ⫽mean effective normal stress; Q⫽empirical constant;


and R⫽fitting parameter that was taken as 1.0 for the test data that
was evaluated. The value of Q determines the value of p ⬘ at
which dilatancy is suppressed 共i.e., I RD⫽0). Bolton indicated that
Q depends on the grain type, with Q⬇10 for quartz and feldspar,
8 for limestone, 7 for anthracite, and 5.5 for chalk. The I RD index
is necessarily empirical in nature, but has a functional form that is
reasonable for describing the combined effects of relative density
Fig. 6. Critical state line and state parameter from Bolton’s 共1986兲
and confining stress on the dilatancy or volume change character-
I RD relation
istics of clean sands. An example of Bolton’s work is shown in

1074 / JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / DECEMBER 2003

J. Geotech. Geoenviron. Eng., 2003, 129(12): 1071-1082


1

冉 冊
␰ R⫽ ⫺D R (8)
100p ⬘
Q⫺ln
Pa

where this ␰ R is an empirical index that has a functional form


consistent with critical state concepts, just like the empirical I RD
index from which it is derived. The particular advantage of ␰ R for
this study is that it provides a consistent link to the CPT penetra-
tion theory of Salgado et al. 共1997a兲.
It is worth noting that the C N relation given by Eqs. 共2b兲 and
共4兲 was derived from the CPT analyses of Salgado et al. 共1997b兲
with Q⫽10. This value of Q represents a reasonably typical value
for many sands, while values of 9–11 might provide better fits
with the shear behavior of specific silica sands at high confining
Downloaded from ascelibrary.org by Tongji University on 11/19/18. Copyright ASCE. For personal use only; all rights reserved.

stresses. Additional CPT analyses were therefore performed using


the same typical property set as Salgado et al. 共1997b兲 but with
Q⫽9 and 11. The resulting effects on q c could be largely ap-
proximated by setting C 1 ⫽0.61 for Q⫽9 and C 1 ⫽1.64 for Q
⫽11. The accompanying C N values at ␴ ⬘v / P a ⫽10 were only
about 5% smaller for Q⫽9 and about 5% greater for Q⫽11.
Consequently, the adopted C N relation in Eqs. 共2b兲 and 共4兲 holds
reasonably well for this range of Q values.
Fig. 7. Cyclic resistance ratio versus ␰ R for reconstituted specimens
of Fraser delta sand
Evaluating Cyclic Resistance Ratio –␰ R Relations
for Reconstituted Specimens
where ␰ R1 ⫽value of ␰ R for the same D R at ␴ ⬘v / P a ⫽1. The K ␴
The effect of D R and ␴ ⬘v on the CRR of clean sand would intu- curves derived in this manner are shown in Fig. 8, and are in
itively be related to the sand’s position relative to critical state or excellent agreement with the raw experimental results previously
some other reference state, and in fact Pillai and Muhunthan presented in Fig. 1.
共2001兲 showed CRR to be approximately constant for a given
value of state parameter 共␰兲. Vaid et al. 共1981兲 related the CRR of
sand to its dilation angle during drained loading, and since dila- Deriving Cyclic Resistance Ratio–␰ R Relations
tion angle can be related to ␰, it follows that CRR can be related for In Situ Conditions
to ␰. Relating cyclic behavior to a critical state framework is both
rational and desirable, but its implementation in engineering prac- Cyclic resistance ratio–␰ R relations for in situ conditions can be
tice is hampered by the difficulty in defining an appropriate criti- expected to be significantly different from those for laboratory
cal state or other reference line without recourse to advanced
sampling and testing of representative field samples. Conse-
quently, the ␰ R index provides a practical substitute for ␰ in rep-
resenting the combined influence of D R and ␴ ⬘v on sand behavior.
The relation between CRR and ␰ R is first illustrated for labo-
ratory tests on reconstituted clean sand. The data used herein are
from Vaid and Sivathayalan 共1996兲 and Vaid and Thomas 共1995兲,
and includes simple shear and triaxial tests on clean Fraser delta
sand at D R of 31, 40, 59, and 72% and consolidation stresses
(␴ ⬘v / P a for simple shear, ␴ 3C
⬘ / P a for the isotropically consoli-
dated triaxial tests兲 of 0.5, 1, 2, and 4. These data were the basis
for the K ␴ relations shown previously in Fig. 1 and correspond to
ten uniform cycles of loading to cause a single amplitude shear
strain of 3.75% 共or 2.5% axial strain for triaxial兲. The data are
replotted in Fig. 7 as CRR versus ␰ R for two cases, Q⫽9 and
Q⫽10, and with the assumption that K 0 ⫽0.45 in the simple
shear tests 共for obtaining p ⬘ ). The simple shear and triaxial test
data collapse onto single CRR– ␰ R relations, with the quality of fit
being better for Q⫽9 than for Q⫽10. In either case, these data
suggest that ␰ R can reasonably represent the combined effects of
D R and ␴ ⬘v on CRR. Furthermore these CRR– ␰ R relations were
fit with regression equations, from which K ␴ relations can be
calculated as

CRR共 ␰ R 兲 Fig. 8. Overburden correction factor (K ␴ ) derived from laboratory


K ␴⫽ (9) cyclic resistance ratio–␰ R relation for reconstituted Fraser delta sand
CRR共 ␰ R1 兲

JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / DECEMBER 2003 / 1075

J. Geotech. Geoenviron. Eng., 2003, 129(12): 1071-1082


can then be mapped onto a corresponding CRR– ␰ R relation by
mapping q c1 to ␰ R . This mapping only requires the selection of a
C 1 value for Eq. 共2兲, after which each q c1 value corresponds to a
unique D R value, which in turn produces a unique ␰ R value 共with
Q⫽10 for consistency with the CPT analyses兲. It is implicitly
assumed that the functional form of Eq. 共2兲 holds for field condi-
tions as well as for calibration chamber testing conditions 共i.e.,
reconstituted specimens versus natural deposits兲, and that the ef-
fects of grain characteristics, fabric, age, stress–strain history, and
cementation can be represented through C 1 . The choice of C 1
could potentially be based on a material-specific calibration or
through independent measurements of intrinsic soil properties.
However, for calibration to the semiempirical liquefaction corre-
lations, C 1 was based on the most likely value of D R that corre-
Downloaded from ascelibrary.org by Tongji University on 11/19/18. Copyright ASCE. For personal use only; all rights reserved.

sponds to a CRR of 0.6 共roughly the limit at which liquefaction


might be triggered兲 when ␴ ⬘v / P a ⫽1; i.e., D R,lim⫽the D R value
corresponding to CRR⫽0.60 at ␴ ⬘v / P a ⫽1.
The specification of D R,lim has an important effect on the sub-
sequent analyses and so its selection warrants some discussion.
First, the CPT-based semiempirical liquefaction correlations in
Fig. 9 curve nearly vertical at q c1 / P a values of 160 and 184,
Fig. 9. Semiempirical cone penetration test-based liquefaction which correspond to D R of 71–77% using the CPT relations with
correlations the typical property set (C 1 ⫽1). The SPT-based correlation by
Seed et al. 共1985兲 curves nearly vertical at (N 1 ) 60⫽30, for which
the corresponding D R,lim may be estimated using the common
correlation
tests on freshly reconstituted specimens because CRR is strongly
affected by factors such as fabric 共e.g., Mulilis et al. 1977兲, age 共 N 1 兲 60
共e.g., Seed 1979; Tatsuoka et al. 1988兲, K 0 共e.g., Ishihara et al. ⫽C d (12)
共 DR兲2
1977兲, overconsolidation 共e.g., Ishihara and Takatsu 1979兲, prior
strain history 共e.g., Finn et al. 1970; Suzuki and Toki 1984兲, and Skempton 共1986兲 suggested C d values between 55 and 65 for
cementation. Similarly, it follows that there will be no single clean sands, which would give D R,lim of 68 –74%. Cubrinovski
CRR– ␰ R relation that applies to all in situ conditions. and Ishihara 共1999兲 provide a more exhaustive and complete
Consequently, the approach adopted herein is to derive repre- evaluation of C d and its dependence on basic soil characteristics.
sentative in situ CRR– ␰ R relations from existing penetration test- They suggest C d of 38 –91 for clean sands 共or 29–70 for their
based semiempirical liquefaction correlations. This will require standard SPT energy ratio of 78%兲, corresponding to D R,lim of
relating ␰ R to penetration resistance, which will also not be a 58 – 89%. Direct measurements of CRR and D R,lim from cyclic
unique relation because of the effects that grain characteristics, triaxial tests on sand samples obtained by frozen sampling tech-
fabric, age, K 0 , overconsolidation, prior strain history, and ce- niques were shown to compare well with SPT-based correlations
mentation can have on penetration resistance 共e.g., Marcuson and in Japan 共Yoshimi et al. 1989; Hatanaka et al. 1995兲, and these
Bieganousky 1977a; Seed 1979; Mitchell 1986; Joshi et al. 1995兲. suggest that CRR⫽0.6 corresponds to D R ⬇85% and (N 1 ) 78
Despite this nonuniqueness, CRR– ␰ R relations derived by this ⬇30. Allowing for SPT energy corrections, these data would give
approach are expected to provide a reasonable and rational basis C d ⬇54, so that D R ⬇75% at (N 1 ) 60⬇30. The specification of
for evaluating the relative effects of ␴ ⬘v on CRR and retain a D R,lim for either SPT or CPT relations is complicated by numer-
consistency with the liquefaction correlations to which the find- ous factors, including the fact that the semiempirical correlations
ings of this study will ultimately be applied. embody the influence of field conditions beyond the local material
Two correlations between CPT q c1 and CRR for clean sands in 共or element兲 behavior and the fact that CRR is not solely depen-
M ⫽7.5 earthquakes are shown in Fig. 9: one by Robertson and dent on D R . Such uncertainties are embodied in probabilistic
Wride 共1998兲, which can be approximated as methods where the positions of CPT or SPT liquefaction curves
depend on the probability of liquefaction. A lower probability of
0.63348
CRR␴⫽1,␣⫽0 ⫽e ⫺3.2467⫹0.014226q C1N ⫹ for q C1N liquefaction shifts the curves to the right, such that the inferred
162.86⫺q C1N D R,lim would be larger. Consequently, the subsequent derivations
in this paper were performed with two D R,lim values to bracket the
⭐160.0 (10)
likely range of values that might be used with different semi-
and the other by Idriss 共personal communication, 2003兲, which empirical liquefaction correlations, and to illustrate the effects of
can be approximated as the D R,lim selection: CPT derivations are presented for C 1 ⫽1.0

冋 冉 冊 冉 冊 冉 冊 册
2 3 4 and C 1 ⫽0.817, which correspond to D R,lim of 75 and 85% at
q C1N q C1N q C1N q C1N
CRR␴⫽1,␣⫽0 ⫽exp ⫹ ⫺ ⫹ ⫺3 q c1 / P a ⫽174, respectively, and SPT derivations are presented for
540 67 80 114 C d ⫽53.3 and 41.5, which correspond to D R,lim of 75 and 85% at
(11) (N 1 ) 60⫽30, respectively.
where q c1N ⫽q c1 / P a . It is assumed that these semiempirical liq- The semiempirical liquefaction correlations were thus mapped
uefaction correlations pertain exactly to ␴ ⬘v / P a ⫽1 and to nor- onto their corresponding CRR– ␰ R relations in Fig. 10. The five
mally consolidated sands with K 0 ⫽0.45. Each CRR– q c1 relation CRR– ␰ R relations in Fig. 10 are reasonably consistent, with in-

1076 / JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / DECEMBER 2003

J. Geotech. Geoenviron. Eng., 2003, 129(12): 1071-1082


Downloaded from ascelibrary.org by Tongji University on 11/19/18. Copyright ASCE. For personal use only; all rights reserved.

Fig. 11. Cyclic resistance ratio versus q c1 / P a for clean sand at dif-
Fig. 10. Derived cyclic resistance ratio–␰ R relations for field condi- ferent overburden stresses using Robertson and Wride’s 共1998兲 cor-
tions relation as baseline

dividual variations that represent both the unique characteristics


of their respective sources as well as any inherent variability in
The CRR– q c1 – ␴ ⬘v relations shown in Fig. 11 can also be ex-
the CRR– ␰ R relation for different sands. These CRR– ␰ R rela-
pressed in terms of implied K ␴ relations as
tions can be used to derive a corresponding consistent set of K ␴
relations, as will be illustrated in the following section. CRR
K ␴⫽ (13)
CRR␴⫽1

Extending Field Cyclic Resistance Ratio–␰ R where CRR and CRR␴⫽1 are for the same q c1 value. Identical K ␴
Relations to Higher Overburden Stresses relations can also be obtained directly from the CRR– ␰ R relation
using Eq. 共9兲. K ␴ relations derived in this way depend on the
The effect of overburden stress on a liquefaction analysis is illus- baseline liquefaction correlation, D R , and the parameters C 1 and
trated by tracking its effects on both penetration resistance and Q, as illustrated by the examples in Fig. 12. K ␴ values cannot be
CRR. The effect of ␴ ⬘v on q c is given by the relations in Eqs. 共2兲 derived if CRR␴⫽1 is undefined, which happens for D R ⭓71%
and 共3兲, such that the corresponding q c1 values are properly in- with the Robertson and Wride baseline correlation and C 1 ⫽1.
dependent of ␴ ⬘v ; Note that if C N is truly dependent on D R , then Consequently, K ␴ factors are often insufficient for representing
the use of a D R -independent C N relation 关such as in Eq. 共5兲兴 the increase in q c1,lim with ␴ ⬘v 共i.e., the rightward shift of curves in
would result in calculated q c1 values that are not actually inde- Fig. 11兲. Despite this limitation, the K ␴ relations in Fig. 12 are in
pendent of ␴ ⬘v . The effect of ␴ ⬘v on CRR is accounted for by reasonable agreement with the available experimental data 共e.g.,
assuming that the CRR– ␰ R relation is applicable for all values of Fig. 1兲 given the scatter that is undoubtedly related to numerous
␴ ⬘v 共as was previously shown to be an acceptable approximation factors such as grain characteristics, fabric, age, cementation, and
stress–strain history.
in Fig. 7兲. Thus, for clean sand at a given D R , a change in ␴ ⬘v
The ␰ R -based approach is most easily implemented by intro-
causes a change in ␰ R 关via Eq. 共8兲兴, which results in a change in
ducing the concept of state normalized penetration resistances
CRR 共Fig. 10兲. The combined effects of ␴ ⬘v are illustrated in Fig.
11 showing how the Robertson and Wride 共1998兲 CRR– q c1 rela- q C1␰ ⫽C ␰ C N q C (14)
tion would be modified for ␴ ⬘v / P a ⫽0.25, 4, and 10 for the case
with C 1 ⫽1.0. Very similar results are obtained with C 1 ⫽0.817 N 1␰ ⫽C ␰ C N N (15)
because the C 1 parameter affects both the derivation of the where q c1␰ and N 1␰ ⫽penetration resistances normalized 共or cor-
CRR– ␰ R relation and its mapping back onto the CRR– q c1 plot rected兲 to the same ␰ R at ␴ ⬘v / P a ⫽1; and C␰ ⫽state normalization
for different overburden stresses. Increasing ␴ ⬘v not only causes a factor. For a sand to have the same ␰ R at the reference stress of
decrease in CRR for a given q c1 共shifting the curves downwards兲, ␴ ⬘v / P a ⫽1, its D R would be smaller 共relative to its in situ value兲
but also increases the limit at which triggering can develop 共shift- by the amount
ing the curves to the right兲. This increase in q c1,lim 共where CRR
1 1
exceeds 0.6兲 with ␴ ⬘v occurs because the limiting state at which
冉 冊 冉 冊
⌬D R,CS⫽ ⫺
triggering of liquefaction can occur is actually tied to a limiting 1⫹2K 0 ␴ V⬘ 1⫹2K 0
Q⫺ln 100 Q⫺ln 100
value of ␰ R . With ␰ R,lim being a constant, it follows that an in- 3 Pa 3
crease in ␴ ⬘v causes an increase in both D R,lim and q c1,lim . (16)

JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / DECEMBER 2003 / 1077

J. Geotech. Geoenviron. Eng., 2003, 129(12): 1071-1082


Downloaded from ascelibrary.org by Tongji University on 11/19/18. Copyright ASCE. For personal use only; all rights reserved.

Fig. 12. K ␴ relations derived from cyclic resistance ratio–␰ R rela-


tions for two semiempirical liquefaction correlations for clean sand Fig. 13. State normalization factor C ␰ for penetration resistance in
clean sand

where ⌬D R,CS⫽difference between the critical state D R values


(D R,CS in Fig. 6兲 for the in situ ␴ ⬘v and reference stress ␴ ⬘v / P a
⫽1. The C ␰ for the CPT follows from Eq. 共14兲 CRR␴⫽1 will be multiplied by the K ␴ relation proposed by
Harder and Boulanger 共1997兲, which can be closely approximated
q C1␰ as
C ␰⫽ (17)

冉 冊
q C1
␴ V⬘
which using Eq. 共2兲 gives K ␴ ⫽1⫺C ␴ ln (20)
Pa
25.7⫹39.7共 D R ⫺⌬D R,CS兲 ⫹212.3共 D R ⫺⌬D R,CS兲 2 with C ␴ ⫽0.185. Method A is compared to the ␰ R -based approach
C ␰⫽
25.7⫹39.7D R ⫹212.3D R2 in Fig. 14, using the Robertson and Wride 共1998兲 CRR– q c1 cor-
(18) relation as the baseline for ␴ ⬘v / P a ⫽1. The ␰ R -based approach is
shown for two C 1 values 共0.817 and 1.0兲 with Q⫽10, and the
Similarly, the C ␰ for the SPT follows from Eqs. 共12兲 and 共15兲 calculated CRR are plotted against the measured 共not corrected兲
共 D R ⫺⌬D R,CS兲 2 q c values for ␴ ⬘v / P a ⫽1, 4, and 10. This comparison illustrates
C ␰⫽ (19) how Method A results in substantial conservatism compared to
D R2
the ␰ R -based approach at high overburden stresses 共i.e., the
The resulting C ␰ curves from the above two relations are shown CRR– q c curves are shifted down and to the right兲. The biggest
in Fig. 13. The CRR for any ␴ ⬘v can then be directly obtained by factor in the differences between these two approaches is, in fact,
evaluating the liquefaction correlation 关e.g., Eqs. 共10兲 or 共11兲兴 the C N relations. The extrapolation of the Liao and Whitman
using q c1␰ or N 1␰ instead of q c1 or N 1 , respectively. This imple- 共1986兲 expression to high overburden stresses is what causes the
mentation of the ␰ R -based approach: 共1兲 is mathematically greater shift of the CRR– q c curves to the right and makes it
equivalent to using derived CRR– ␰ R relations; 共2兲 eliminates the appear that even the densest possible sands are liquefiable. The
need to explicitly establish the CRR– ␰ R relation; 共3兲 can be used ␰ R -based curve for C 1 ⫽0.817 is to the left of the curve for C 1
in conjunction with any liquefaction correlation; and 共4兲 elimi- ⫽1 because the lower C 1 value produces higher D R values for
nates the use of a K ␴ factor. The approach does, however, require the same q c1 value, which produces C N values closer to unity.
numerical iteration because the penetration resistances, measured The difference in C N relations also affects the relative shifting of
or corrected, are nonlinearly dependent on D R . the CRR– q c curves downwards, such that the conservatism of
Method A at lower q c values is due to both the K ␴ and C N rela-
tions.
Simplified Implementation and Comparison
to Practice Simplified Approximation of ␰ R -Based Method
The ␰ R -based approach requires an iterative estimation of D R
Comparison of ␰ R -Based Method with Current Practice
that, while not mechanistically difficult, does add another element
One adaptation of current practice, hereafter called Method A, is of complexity to the liquefaction analysis. Consequently, a sim-
presented for comparison against the ␰ R -based approach. The plified approximation 共hereafter called Method B兲 that uses
measured penetration resistance will be corrected using the C N D R -independent C N and K ␴ relations is also introduced. The key
relation by Liao and Whitman 共1986兲 in Eq. 共5兲 and the observation that makes this simplification possible is that both C N

1078 / JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / DECEMBER 2003

J. Geotech. Geoenviron. Eng., 2003, 129(12): 1071-1082


Downloaded from ascelibrary.org by Tongji University on 11/19/18. Copyright ASCE. For personal use only; all rights reserved.

Fig. 15. Comparison of ␰ R -based method 共with Q⫽10) and its sim-
plified approximation 共method B兲, using Robertson and Wride’s
Fig. 14. Comparison of ␰ R -based method 共with Q⫽10) and a cur-
共1998兲 correlation as baseline
rent method in practice 共method A兲, using Robertson and Wride’s
共1998兲 correlation as baseline

and K ␴ are actually dependent on D R , but in ways that produce


opposing effects on the predicted CRR. For example, at ␴ ⬘v / P a smallest m lim of 0.32 occurs for the densest of these D R,lim values
⬎1, an increase in D R causes both an increase in C N and a de- while the largest m lim of 0.46 occurs for the loosest of these D R,lim
crease in K ␴ . A single D R -independent C N relation was therefore values. Note that m and m lim are not equal for equal values of D R
derived that tracked the limiting condition at which liquefaction and D R,lim , respectively, because m lim tracks the effects of ␴ ⬘v on
can be triggered 共i.e., ␰ R,lim). This relation can be reasonably ap- CRR as well as on penetration resistance.
proximated as The second part of Method B is the choice of a compatible

冉 冊
D R -independent K ␴ relation. This K ␴ relation was assumed to
m lim
Pa follow Eq. 共20兲 because its form can reasonably approximate the
C N⫽ (21) range of K ␴ curves in Fig. 12. A single C ␴ value was subse-
␴ ⬘v
quently chosen 共for the already selected m lim) that results in pre-
where m lim depends on the baseline liquefaction correlation and dicted CRR values that remain conservative relative to the
the parameters that relate penetration resistance to D R 共i.e., C 1 for ␰ R -based CRR values, but to the minimum extent possible.
CPT and C d for SPT兲. Values of m lim are summarized in Table 1 A comparison of Method B 共using m lim⫽0.46 and C ␴ ⫽0.19)
for the two CPT liquefaction correlations 共with C 1 ⫽0.817 and and the ␰ R -based method is shown in Fig. 15 for the Robertson
1.0兲 and the one SPT liquefaction correlation 共with C d ⫽41.5 and and Wride 共1998兲 correlation as the baseline and C 1 ⫽1. As was
53.3兲 that were used to derive the CRR– ␰ R relations in Fig. 10. done for Fig. 14, CRR is plotted against the measured 共not cor-
Also given in Table 1 are the D R,lim values that correspond to the rected兲 q c value for ␴ ⬘v / P a ⫽1, 4, and 10. Comparing Figs. 14 and
␰ R,lim at ␴ ⬘v / P a ⫽1 共i.e., where CRR⫽0.6 with ␴ ⬘v / P a ⫽1). The 15, Method B is conservative relative to the ␰ R -based method but

Table 1. Parameters for Method B’s Simplified Approximation of ␰ R -Based Method


Parameter relating D R D R at CRR⫽0.6
Baseline correlation to penetration resistance with ␴ ⬘v / P a ⫽1 m lim for C N C ␴ for K ␴
Robertson and C 1 ⫽1.0 71% 0.46 0.19
Wride 共1998兲
C 1 ⫽0.817 81% 0.40 0.21
Idriss 共personal communication, 2003兲 C 1 ⫽1.0 77% 0.39 0.23
C 1 ⫽0.817 88% 0.32 0.25
Seed et al. 共1985兲a C d ⫽53.3 75% 0.46 0.19
C d ⫽41.5 85% 0.39 0.21
a
As modified at NCEER workshop 共Youd et al. 2001兲.

JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / DECEMBER 2003 / 1079

J. Geotech. Geoenviron. Eng., 2003, 129(12): 1071-1082


Downloaded from ascelibrary.org by Tongji University on 11/19/18. Copyright ASCE. For personal use only; all rights reserved.

Fig. 16. Comparison of ␰ R -based method 共with Q⫽10), its simpli-


fied approximation 共method B兲, and a current method in practice Fig. 17. Comparison of ␰ R -based method 共with Q⫽10), its simpli-
共method A兲, with Idriss’ 共personal communication, 2003兲 correlation fied approximation 共method B兲, and a current method in practice
as baseline 共method A兲, with Seed et al. 共1985兲 standard penetration test corre-
lation as baseline

is less conservative than Method A, particularly in setting the


limit of q c values for which triggering of liquefaction is consid- that both C N and K ␴ are truly dependent on D R , but with oppos-
ered possible. ing effects on the predicted CRR.

Additional Comparisons of ␰ R -Based Method


Discussion
with Methods A and B
Another comparison of Methods A and B against the ␰ R -based The preceding ␰ R -based relations are only applicable to those
method is presented in Fig. 16, this time using the Idriss 共personal conditions for which the underlying theory and correlations have
communication, 2003兲 correlation as the baseline with C 1 been validated by experimental data, namely clean silica sands of
⫽0.817 and Q⫽10. In this case, Method B 共using m lim⫽0.32 and medium compressibility for which Q⬇10 is appropriate. The true
C ␴ ⫽0.25) produces CRR values that are reasonably close to relations between ␰ R and CRR or penetration resistance will not
those of the ␰ R -based method, while Method A (m⫽0.5, C ␴ be unique because of the documented effects of factors such as
⫽0.185) produces excessively conservative values at high over- grain characteristics, fabric, age, stress–strain history, and cemen-
burden stresses. tation. Nonetheless, pending further research on the role of such
Finally, these three methods are compared in Fig. 17 for the factors, the use of representative relations in the ␰ R -based method
SPT-based correlation of Seed et al. 共1985兲, as modified in Youd still provides improved consistency in the evaluation of high ␴ ⬘v
et al. 共2001兲, and with C d ⫽41.5. Again, Method B 共using m lim effects and provides a framework for incorporating future re-
⫽0.39 and C ␴ ⫽0.21) produces CRR values that are conservative search findings.
relative to the ␰ R -based method, but substantially less so than for The application of semiempirical liquefaction analysis meth-
Method A. If C d ⫽53.3 instead, the differences between Method ods to large depths is inherently complicated by the shortage of
B 共using m lim⫽0.46 and C ␴ ⫽0.19) and the ␰ R -based method are case histories for such conditions, and thus considerable engineer-
smaller, and the degree of conservatism in Method A is less 共the ing judgment must be exercised in the procedures being used to
effect of increasing C d is similar to the effect of increasing C 1 extrapolate our shallow-depth experiences. The present study at-
that was shown in Fig. 14兲. tempts to improve that extrapolation by analyzing the various
Inspection of the above results shows that once C N is set in- components of the design process with a consistent theoretical
dependent of D R , then the use of conventional K ␴ relations that basis, but it nonetheless remains an extrapolation with concurrent
depend on D R would result in poorer comparisons to the ␰ R -based unavoidable uncertainties. Improved guidance on this issue re-
method. Considering Figs. 15, 16 and 17, a better match between quires direct physical evidence, such as may be obtained through
Method B and the ␰ R -based method would require K ␴ values to well-instrumented earth dams that experience significant earth-
increase 共at ␴ ⬘v / P a ⬎1) with increasing D R , contradicting the ex- quake shaking or through centrifuge models of liquefaction at
perimental data 共e.g., Fig. 1兲. This observation reflects the fact large depths with in-flight measurements of q c , shear wave ve-

1080 / JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / DECEMBER 2003

J. Geotech. Geoenviron. Eng., 2003, 129(12): 1071-1082


locity, and other soil characteristics. In the meantime, the proce- and K ␴ relations are interrelated through the sand properties and
dures developed herein require a judgment regarding the appro- state in ways that can have compensating effects on the predicted
priate value of D R,lim , as set by C 1 or C d for the CPT and SPT CRR. Subsequently, the appropriate choice of D R -independent
methods, respectively. Within the range of values used herein, it is C N and K ␴ relations 共eliminating the need to estimate D R or ␰ R )
suggested that the smaller values may be most appropriate for can reasonably approximate the effects of ␴ ⬘v on predicted CRR.
fine, uniform sands and the larger values may be most appropriate The ␰ R -based procedures recommended herein are easily imple-
for coarse, well-graded sands. mented in practice, as illustrated in more detail by Boulanger and
On a site specific basis, the geologic characterization provides Idriss 共2004兲. These recommendations are limited to clean sands
essential information that should be considered as part of any of medium compressibility for which the theory and analyses em-
evaluation of liquefaction potential; i.e., the age of the deposit, ployed herein have been validated.
the depositional environment, stress–strain history 共e.g., overcon-
solidation, prior seismic events兲, cementation, and heterogeneity
Acknowledgments
共e.g., can soil conditions beneath the center of a dam be compared
to soil conditions near the toe?兲. For natural deposits, liquefaction The writer is grateful for the many discussions and insights pro-
Downloaded from ascelibrary.org by Tongji University on 11/19/18. Copyright ASCE. For personal use only; all rights reserved.

has been observed primarily at depths shallower than about vided by I. M. Idriss throughout this study and the valuable re-
10–15 m. This aspect of the case history database is believed to view comments of R. Salgado and P. M. Byrne.
reflect the combined influence of several factors: 共1兲 deeper soils
tend to be older and hence more resistance to liquefaction; 共2兲 the
earthquake-induced cyclic shear stress ratio tends to be greatest References
within this depth range; and 共3兲 surface observations of sand boils Been, K., and Jefferies, M. G. 共1985兲. ‘‘A state parameter for sands.’’
and ground deformations are more likely for shallow liquefaction, Geotechnique, 35共2兲, 99–112.
while deeper liquefaction may not clearly manifest itself at the Bolton, M. D. 共1986兲. ‘‘The strength and dilatancy of sands.’’ Geotech-
ground surface in many cases 共e.g., beneath level ground兲. In nique, 36共1兲, 65–78.
contrast, the construction of a large earth dam on a natural deposit Boulanger, R. W. 共2003兲. ‘‘Relating K ␣ to a relative state parameter
of sand can: 共1兲 impose large confining stresses that result in a index.’’ J. Geotech. Geoenviron. Eng., 129共8兲, 770–773.
more contractive stress-strain response for the soils; 共2兲 does not Boulanger, R. W., and Idriss, I. M. 共2004兲. ‘‘State normalization of pen-
impart the benefits that may be associated with large depths in etration resistance and the effect of overburden stress on liquefaction
entirely natural deposits 共i.e., a greater age with associated history resistance.’’ Proc., 11th International Conf. on Soil Dynamics and
Earthquake Engineering and 3rd International Conference on Earth-
and cementation, etc.兲; and 共3兲 can present a mode of slope insta-
quake Geotechnical Engineering, Univ. of California, Berkely, Ca.
bility if the foundation soils were to liquefy. For these reasons,
Cubrinovski, M., and Ishihara, K. 共1999兲. ‘‘Empirical correlation between
liquefaction at large depths beneath earth dams is a major con- SPT N-value and relative density for sandy soils.’’ Soils Found.,
cern, and the potentially beneficial effects associated with certain 39共5兲, 61–71.
geologic conditions need to be directly evaluated from a geologic Finn, W. D. L., Bransby, P. L., and Pickering, D. J. 共1970兲. ‘‘Effect of
site characterization. strain history on liquefaction of sand.’’ J. Soil Mech. Found. Div., Am.
Soc. Civ. Eng., 96共6兲, 1917–1934.
Harder, L. F., Jr., and Boulanger, R. W. 共1997兲. ‘‘Application of K ␴ and
Summary and Conclusions K ␣ correction factors.’’ Proc., NCEER Workshop on Evaluation of
Liquefaction Resistance of Soils, Rep. No. NCEER-97-0022, National
Center for Earthquake Engineering Research, SUNY Buffalo, N.Y.,
The effect of overburden stress on liquefaction evaluations was 167–190.
reevaluated using a theoretical framework that provided consis- Hardin, B., and Black, W. 共1968兲. ‘‘Shear modulus and damping in soils.’’
tency between the different components of the design process. J. Soil Mech. Found. Div., Am. Soc. Civ. Eng., 94共2兲, 353–369.
The cone penetration theory of Salgado et al. 共1997a兲, which had Hatanaka, M., Uchida, A., and Oh-Oka, H. 共1995兲. ‘‘Correlation between
been validated against a large set of calibration chamber test data, the liquefaction strengths of saturated sands obtained by in-situ freez-
was used to develop improved C N relations and to relate the pen- ing method and rotary-type triple tube method.’’ Soils Found., 35共2兲,
etration resistance to a relative state parameter index (␰ R ). Cali- 67–75.
bration chamber test data for SPT tests from Marcuson and Bie- Hynes, M. E., and Olsen, R. 共1998兲. ‘‘Influence of confining stress on
ganousky 共1977a,b兲 were reevaluated using a weighted nonlinear liquefaction resistance.’’ Proc., International Symposium on the Phys-
ics and Mechanics of Liquefaction, Balkema, Rotterdam, The Nether-
regression that better defined the effects of D R on C N and showed
lands, 145–152.
that the C N expression given by Eqs. 共2b兲 and 共4兲 provides a Idriss, I. M., and Boulanger, R. W. 共2003兲. ‘‘Estimating K ␣ for use in
satisfactory approximation for both CPT and SPT tests. Experi- evaluating cyclic resistance of sloping ground. Proc., 8th U.S.–Japan
mental data by Vaid and Sivathayalan 共1996兲 and Vaid and Tho- Workshop on Earthquake Resistant Design of Lifeline Facilities and
mas 共1995兲 on reconstituted sand specimens were used to show Countermeasures Against Liquefaction, Rep. No. MCEER-03-0003,
that CRR could be approximated as a unique function of ␰ R , Multidisciplinary Center for Earthquake Engineering Research,
thereby capturing the combined effects of D R and ␴ ⬘v on CRR. SUNY Buffalo, N.Y., 449– 468.
Relations between CRR and ␰ R for field conditions were subse- Ishihara, K., Iwamoto, S., Yasuda, S., and Takatsu, H. 共1977兲. ‘‘Liquefac-
quently derived from semiempirical liquefaction correlations, and tion of anisotropically consolidated sand.’’ Proc., 9th International
Conf. on Soil Mechanics and Foundation Engineering, Vol. 2,
these CRR– ␰ R relations can be used to calculate the effects of ␴ ⬘v
JSSMFE, Tokyo, Japan, 261–264.
on predicted CRR. A more convenient implementation of the
Ishihara, K., and Takatsu, H. 共1979兲. ‘‘Effects of overconsolidation and
␰ R -based approach was, however, provided by the introduction of K 0 conditions on the liquefaction characteristics of sands.’’ Soils
state normalized penetration resistances, q c1␰ and N 1␰ . Found., 19共4兲, 59– 68.
The resulting ␰ R -based approach was shown to reduce the con- Joshi, R. C., Achari, G., Kaniraj, R., and Wijeweera, H. 共1995兲. ‘‘Effect of
servatism imposed at high overburden stresses by some of the aging on the penetration resistance of sands.’’ Can. Geotech. J., 32,
current C N and K ␴ relations. In addition, it was shown that the C N 767–782.

JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / DECEMBER 2003 / 1081

J. Geotech. Geoenviron. Eng., 2003, 129(12): 1071-1082


Konrad, J.-M. 共1988兲. ‘‘Interpretation of flat plate dilatometer tests in level ground during earthquakes,’’ J. Geotech. Eng. Div., Am. Soc.
sands in terms of the state parameter.’’ Geotechnique, 38共2兲, 263–277. Civ. Eng., 105共2兲, 201–255.
Liao, S. C., and Whitman, R. V. 共1986兲. ‘‘Overburden correction factors Seed, R. B., and Harder, L. F., Jr. 共1990兲. ‘‘SPT-based analysis of cyclic
for SPT in sand.’’ J. Geotech. Eng., 112共3兲, 373–377. pore pressure generation and undrained residual strength.’’ Proc.,
Marcuson, III, W. F., and Bieganousky, W. A. 共1977a兲. ‘‘Laboratory stan- SeedMemorial Symposium, J. M. Duncan, ed., BiTech, Vancouver,
dard penetration tests on fine sands.’’ J. Geotech. Eng. Div., Am. Soc. B.C., Canada, 351–376.
Civ. Eng., 103共6兲, 565–588. Seed, H. B., and Idriss, I. M. 共1971兲. ‘‘Simplified procedure for evaluat-
Marcuson, III, W. F., and Bieganousky, W. A. 共1977b兲. ‘‘SPT and relative ing soil liquefaction potential.’’ J. Soil Mech. Found. Div., Am. Soc.
density in coarse sands.’’ J. Geotech. Eng. Div., Am. Soc. Civ. Eng., Civ. Eng., 97共9兲, 1249–1273.
103共11兲, 1295–1309. Seed, H. B., Tokimatsu, K., Harder, L. F., and Chung, R. 共1985兲. ‘‘The
Mitchell, J. K. 共1986兲. ‘‘Practical problems from surprising soil behav- influence of SPT procedures in soil liquefaction resistance evalua-
ior.’’ J. Geotech. Eng., 112共3兲, 255–289. tions.’’ J. Geotech. Eng., 111共12兲, 1425–1445.
Mulilis, J. P., Seed, H. B., Chan, C. K., Mitchell, J.K., and Arulanandan, Skempton, A. W. 共1986兲. ‘‘Standard penetration test procedures and the
K. 共1977兲. ‘‘Effects of sample preparation on sand liquefaction.’’ J. effects in sands of overburden pressure, relative density, particle size,
Geotech. Eng. Div., Am. Soc. Civ. Eng., 103共2兲, 91–108. ageing and overconsolidation.’’ Geotechnique, 36共3兲, 425– 447.
Naylor, D. J., Pande, G. N., Simpson, B., and Tabb, R. 共1981兲. Finite
Downloaded from ascelibrary.org by Tongji University on 11/19/18. Copyright ASCE. For personal use only; all rights reserved.

Suzuki, T., and Toki, S. 共1984兲. ‘‘Effects of preshearing on liquefaction


elements in geotechnical engineering, Pineridge, Swansea, Wales,
characteristics of saturated sand subjected to cyclic loading.’’ Soils
U.K.
Found., 24共2兲, 16 –28.
Pillai, V. S., and Byrne, P. M. 共1994兲. ‘‘Effect of overburden pressure on
Tatsuoka, F., Kato, H., Kimura, M., and Pradhan, T. B. S. 共1988兲. ‘‘Liq-
liquefaction resistance of sands.’’ Can. Geotech. J., 31, 53– 60.
uefaction strength of sands subjected to sustained pressure.’’ Soils
Pillai, V. S., and Muhunthan, B. 共2001兲. ‘‘A review of the influence of
Found., 28共1兲, 119–131.
initial static shear (K ␣ ) and confining stress (K ␴ ) on failure mecha-
nisms and earthquake liquefaction of soils.’’ Proc., 4th International Vaid, Y. P., Byrne, P. M., and Hughes, J. M. O. 共1981兲. ‘‘Dilation angle
Conf. on Recent Advances in Geotechnical Earthquake Engineering and liquefaction potential.’’ Proc., International Conf. on Recent Ad-
and Soil Dynamics, S. Prakash, ed., Univ. of Missouri Rolla Press, vances in Geotechnical Earthquake Engineering and Soil Dynamics,
Rolla, Mo. Vol. 1, S. Prakash, ed., Univ. of Missouri-Rolla Press, Rolla, Mo.,
Robertson, P. K., and Wride, C. E. 共1998兲. ‘‘Evaluating cyclic liquefac- 161–165.
tion potential using the cone penetration test.’’ Can. Geotech. J., Vaid, Y. P., and Sivathayalan, S. 共1996兲. ‘‘Static and cyclic liquefaction
35共3兲, 442– 459. potential of fraser delta sand in simple shear and triaxial tests.’’ Can.
Salgado, R., Boulanger, R. W., and Mitchell, J. K. 共1997b兲. ‘‘Lateral Geotech. J., 33, 281–289.
stress effects on CPT liquefaction resistance correlations.’’ J. Geotech. Vaid, Y. P., and Thomas, J. 共1995兲. ‘‘Liquefaction and postliquefaction
Geoenviron. Eng., 123共8兲, 726 –735. behavior of sand.’’ J. Geotech. Eng., 121共2兲, 163–173.
Salgado, R., Mitchell, J. K., and Jamiolkowski, M. 共1997a兲. ‘‘Cavity ex- Yoshimi, Y., Tokimatsu, K., and Hosaka, Y. 共1989兲. ‘‘Evaluation of liq-
pansion and penetration resistance in sands.’’ J. Geotech. Geoenviron. uefaction resistance of clean sands based on high quality undisturbed
Eng., 123共4兲, 344 –354. samples.’’ Soils Found., 29共1兲, 93–104.
Seed, H. B. 共1983兲. ‘‘Earthquake resistant design of earth dams.’’ Proc., Youd, T. L. et al. 共2001兲. ‘‘Liquefaction resistance of soils: Summary
Symposium on Seismic Design of Embankments and Caverns, Phila- report from the 1996 NCEER and 1998 NCEER/NSF workshops on
delphia, ASCE, N.Y., 41– 64. evaluation of liquefaction resistance of soils.’’ J. Geotech. Geoenvi-
Seed, H. B. 共1979兲. ‘‘Soil liquefaction and cyclic mobility evaluation for ron. Eng., 127共10兲, 817– 833.

1082 / JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / DECEMBER 2003

J. Geotech. Geoenviron. Eng., 2003, 129(12): 1071-1082

You might also like