Академический Документы
Профессиональный Документы
Культура Документы
Engineering Geology
journal homepage: www.elsevier.com/locate/enggeo
Technical note
a r t i c l e
i n f o
Article history:
Received 3 November 2014
Received in revised form 28 July 2015
Accepted 1 August 2015
Available online 5 August 2015
Keywords:
Soil liquefaction potential
Deterministic assessment
Probabilistic assessment
Liquefaction triggering correlation
Expanded SPT-based database
a b s t r a c t
A new calibrated empirical liquefaction triggering correlation equation with ve parameters determined by trial
and error is derived from the expanded SPT-based database of the Idriss and Boulanger (2010) and Xie (1984)
case histories. The new calibrated correlation equation is reasonably conservative for the expanded SPT-based
database and is insensitive to reasonable variations in the components of the liquefaction analysis framework
adopted in this study. In addition, a probabilistic version of the new calibrated correlation equation expressed
in the form of a mapping function that relates the liquefaction probability to a nominal safety factor obtained
using the weighted maximum likelihood technique is veried using a weighted empirical probability approach
and Monte Carlo simulations. Liquefaction, transition and non-liquefaction zones for clean sands are dened
based on different probability contours, i.e., empirical correlations with different liquefaction probabilities.
Liquefaction triggering correlations for liquefaction probabilities of 4%, 15% and 25% represent the very low,
low, and moderate probability events, respectively, of a liqueable soil being mistaken for a non-liqueable
soil in the liquefaction potential evaluation.
2015 Elsevier B.V. All rights reserved.
1. Introduction
A certain amount of soil liquefaction has often been observed
during major seismic events. Soil liquefaction has caused serious
damage to buildings and infrastructure as well as loss of life, especially when coastal natural or backlled soils have liqueed, as in
the 1964 Niigata earthquake (Ohsaki, 1966), the 1995 HyogokenNanbu earthquake (Sitar, 1995), the 2011 Christchurch earthquake
(Yamada et al., 2011) and the 2011 Great East Japan Earthquake
(Bhattacharya et al., 2011). Therefore, soil liquefaction is one of the
primary seismic geological disasters that has attracted a great deal
of attention in the area of evaluating the seismic safety of civil
projects. These cases of liquefaction raise concerns regarding the
liquefaction of saturated soil.
The most commonly accepted and widely used method for evaluating soil liquefaction potential is the simplied procedure pioneered by
Seed and Idriss (1971). The simplied procedure is generally expressed
in terms of a deterministic model and was developed using empirical
http://dx.doi.org/10.1016/j.enggeo.2015.08.002
0013-7952/ 2015 Elsevier B.V. All rights reserved.
306
model for the Idriss and Boulanger (2004, 2008) liquefaction triggering
correlation equation. It is expressed by Eq. (1):
ln F s 0:13
P L 1
0:13
"
#
!
N1 60CS
N 1 60CS 2
N1 60CS 3
N1 60CS 4
2:67lnCSR
0:13
14:1
126
23:6
25:4
PL
1
h
i
1 exp a ~F s b
1
where is the cumulative standard normal distribution function; PL is
the liquefaction probability; FS is the anti-liquefaction safety factor
dened as the ratio of the cyclic resistance ratio (CRR) to the cyclic stress
ratio (CSR); and (N1)60CS is the equivalent SPT blow count for clean sand.
In this paper, unless otherwise stated, all of the equations are for
(N1)60CS.
Fig. 1 shows the liquefaction triggering correlations for liquefaction
probabilities of 15%, 50%, and 85% in the Boulanger and Idriss (2012)
model and points from the case histories in the Idriss and Boulanger
(2010) database. The input parameters are treated as random variables
in the Boulanger and Idriss (2012) model, and, therefore, the parameters of the CRR and the CSR and the FS are also random variables. Unlike
Boulanger and Idriss (2012), Juang et al. (2012, 2013) express PL as a
function of the ~F S , i.e., a xed value dened as the ratio of the nominal
~ to the nominal cyclic stress ratio (C ~SR).
cyclic resistance ratio (C RR)
Therefore, Eq. (1) is equivalent to Eq. (2), which was given by Juang
et al. (2013):
0
12 0
13 0
14 1
~
~1
~1
~1
N
N
N
N
B 6 1 60CS
60CS
60CS
60CS
A @
A @
A C
@
C
B 4
C
14:1
126
23:6
25:4
P L B
C
B
A
@
i
~
2:67 ln C SR e =0:13
0
2
2
where and e are the uncertainties of (N1)60CS and lnCSR, respectively.
Boulanger and Idriss (2012) suggested that should be taken as a normal random variable with a zero mean and a standard deviation
between 0.15 (1)60CS and 0.2 (1)60CS, while e is taken as a normal random variable with a zero mean and a standard deviation between 0.15
and 0.25.
Because of the complexity of the Boulanger and Idriss (2012) probabilistic model, Juang et al. (2012, 2013) derived a simplied probabilistic model in terms of PL and the nominal safety factor ( ~F s ) for the Idriss
and Boulanger (2004, 2008) liquefaction triggering correlation equation
Fig. 1. The liquefaction triggering correlation equations of Boulanger and Idriss (2012) for
liquefaction probabilities of 15%, 50%, and 85% compared with data from the Idriss and
Boulanger (2010) case history database.
Revised using the Idriss and Boulanger case history database (2010) and Eq. (1).
where a and b are the model parameters. The values of a and b are 7.545
and 0.952 (Juang et al., 2012) and 7.612 and 0.898 (Juang et al., 2013),
respectively.
The Idriss and Boulanger (2004, 2008) liquefaction triggering correlation equation is equivalent to Juang et al. (2012 and 2013) models
with liquefaction probabilities of 41.0% and 31.5% when ~F s 1:0 is
substituted into Eq. (3). Therefore, Juang et al. (2012 and 2013) models
corresponding to the Idriss and Boulanger (2004, 2008) liquefaction
triggering correlation equation might be reasonable to use in common
civil projects with low to medium potential risk levels but are dangerous
and unacceptable for enormously complex civil projects that have extremely high risk potential. Therefore, it is necessary to re-calibrate
the CRR model using the expanded SPT-based case history database
such that it has a very low probability of mistaking liqueable soil for
non-liqueable soil when the liquefaction potential of soil sites is
evaluated for projects with different potential risk levels.
2. A new calibration of the empirical liquefaction triggering
correlation equation
2.1. The liquefaction case history database
Both the Idriss and Boulanger (2010) and the Xie (1984) case history
databases are used in this study. The total number of cases in the updated Idriss and Boulanger (2010) database is 230 of which 115 cases
have surface evidence of liquefaction, 112 cases have no surface
evidence of liquefaction, and 3 cases are on the border between liquefaction and non-liquefaction. The total number of case histories in the
Xie (1984) database (see Table 1) is 155 of which 94 cases have surface
evidence of liquefaction, and 61 cases have no surface evidence of
liquefaction.
The Xie (1984) database includes the SPT-based case history data
from the Chinese code for the Seismic Design of Buildings (TJ11-74)
and from the 1975 Haicheng and 1976 Tangshan earthquakes. It
includes data from seven strong earthquakes that occurred in China between 1962 and 1976 (Xie, 1979; Seed et al., 1983, 1985). The results of
evaluating the soil liquefaction potential using Xie's empirical equation
(Xie, 1979; Ishihara, 1985) have been incorporated into the Chinese
code (TJ11-74), and the results of Seed's simplied procedure are
generally in excellent agreement with them (Seed et al., 1983, 1985).
Additionally, the reliability of Xie's empirical equation has been veried
using case history data from the 1975 Haicheng and 1976 Tangshan
earthquakes (Xie, 1984), and the updated Xie (1984) equation was
later incorporated into updated versions of the Chinese code (GBJ1189, GB5001-2001, GB5001-2010).
According to Xie (1984), most of the sands in the database are ne
and silty. Unfortunately, the database lacks plausibility for the ne content of the sands, the effective overburden stress, and the moment magnitude. In this study, the ne content (FC) is set to 5%. During the
calculation of the total and effective overburden stress in the CSR and
the CRR equations for the Xie (1984) case histories, the weight densities
of the soils above and below the ground water level are set to 20 kN/m3
and 18.5 kN/m3, respectively. In the original case history data, the
magnitudes of the earthquakes were measured using the Richter scale
(Xie, 1984). The moment magnitudes are available only for the 1975
Haicheng earthquake (Mw = 7.0) and the 1976 Tangshan earthquake
(Mw = 7.6); the other ve earthquakes are considered in terms of
their moment magnitudes in this study.
307
Table 1
Basic information from the Xie (1984) liquefaction case history database.
No
1
2
3
4
5
6
7
Earthquakes
1962 Heyuan
1966 Xingtai
1969 Yangjiang
1969 Bohai
1970 Tonghai
1975 Haicheng
1976 Tangshan
6.4
6.7
6.4
7.4
7.8
7.3
7.8
PGA/g
0.20
0.1790.526
0.0990.108
0.0940.140
0.10.676
0.0890.190
0.0860.797
CSR
0.156
0.0950.330
0.0570.088
0.0570.11
0.1280.523
0.0610.16
0.0650.740
No. of cases
L
NL
0
9
3
3
17
7
55
1
3
1
0
14
5
37
(N1)60CS
6.5
1.154.41
0.94.5
2.02.5
0.55.4
313
1.320
0
02.77
01.0
02.0
02.35
12
0.435.9
7.41
8.6718.08
2.683.53
6.8010.47
4.9749.5
6.3417.79
1.3466.07
Note: PGA represents the horizontal peak ground surface acceleration, L denotes liquefaction cases, and NL denotes non-liquefaction cases.
a5
a1
a2
a3
a4
CRR exp
CRR exp
3:1
12:6
130:8
25:4
27:3
5
"
#
N1 60CS
N1 60CS 2
N1 60CS 3
N1 60CS 4
CRR exp
3:6
9:4
135:1
24:2
26:7
6
Fig. 2a and b show the case history data of the Idriss and Boulanger
(2010) and Xie (1984) databases with the liquefaction triggering correlations given in Eqs. (5) and (6) in terms of CSR and (N1)60CS, respectively. Only one liquefaction case history lies below the calibrated
liquefaction triggering correlation given in Eqs. (5) or (6) (see Fig. 2a
or b).
In the same way, for the combined database of Idriss and Boulanger
(2010) and Xie (1984), the calibrated liquefaction triggering correlation
is expressed by Eq. (7):
"
where a1a5 are the ve parameters. These ve parameters were determined by trial and error using the following two principles:(1) the liquefaction triggering correlation for (N1)60CS 15 is generally
controlled by the lower bound of the liquefaction case history data;
and (2) the liquefaction triggering correlation for (N1)60CS 25 is generally controlled by the upper bound of the non-liquefaction case history
data. Therefore, the probability of mistaking a liquefaction case history
for a non-liquefaction case history is reduced, and the liquefaction triggering correlations will be reasonably conservative (see Figs. 2 and 3).
For the Idriss and Boulanger (2010) and Xie (1984) databases, the
calibrated liquefaction triggering correlations on the basis of the
CRR exp
#
N1 60CS
N1 60CS 2
N1 60CS 3
N1 60CS 4
3:6 :
9:7
135:1
24:7
27:5
7
Fig. 3 shows CSR and (N1)60CS for the case histories in the combined
database and the differences between the three liquefaction triggering
correlations (Eqs. (5), (6) and (7)) and those published by Youd et al.
(2001) and by Idriss and Boulanger (2004, 2008). It is notable that the
liquefaction triggering correlation given in Eq. (7) for the combined
database envelops the lower bound of the liquefaction triggering
correlations given in Eqs. (5) and (6) and those published by Youd
Fig. 2. The distribution of the case history data with the liquefaction triggering correlation from: (a) the Idriss and Boulanger (2010) case history database; and (b) the Xie (1984) case
history database.
308
histories is reduced with respect to the liquefaction triggering correlations given in Eqs. (5) and (6).
2.3. The sensitivity of the case history data to components of the liquefaction
analysis framework
Fig. 3. The distribution of the data in the combined Idriss and Boulanger (2010) and Xie
(1984) case history databases compared with the liquefaction triggering correlations
given in Eqs. (5), (6) and (7) and those published by Idriss and Boulanger (2004) and
Youd et al. (2001).
Fig. 4. The distribution of the data in the combined Idriss and Boulanger (2010) and Xie (1984) case history databases using different models of rd compared with the calibrated liquefaction
triggering correlation given in Eq. (7): (a) Cetin et al. (2004); (b) Kishida et al. (2009); and (c) Youd et al. (2001).
the liquefaction case history data and accounts for the effects that FC has
on the CRR and the SPT blow count. The effect of using the equivalent
clean sand adjustment (N1)60 from the procedures described by
Youd et al. (2001) and the Idriss and Boulanger (2010) procedure is analyzed. The ne content correction is equivalent to a null adjustment for
FC 5% and reaches a limiting value for FC 35% (Youd et al., 2001) or
FC 50% (Idriss and Boulanger, 2010). Because the Xie (1984) database
lacks data on the ne contents of the sands, Fig. 5 illustrates the case history data of only sands with FC N 15% in the Idriss and Boulanger (2010)
database using the equations proposed by Idriss and Boulanger (2010)
and Youd et al. (2001). Zero liquefaction case histories fall below the
liquefaction triggering correlation given in Eq. (7) for the FC given by
different equations. These results indicate that the position of the
calibrated liquefaction triggering correlation given in Eq. (7) is not
sensitive to the two clean sand adjustment procedures for the Idriss
and Boulanger database (2010) used.
Both the expression for the magnitude scaling factor (MSF) derived
by Idriss (1999) and the values of rd depend on the moment magnitude
Mw. Therefore, variations in the earthquake magnitude can affect the
MSF and the rd and, therefore, the CSR. Fig. 6 illustrates the Xie (1984)
liquefaction case history data for the 1975 Haicheng earthquake and
the 1976 Tangshan earthquake by comparing the Richter magnitude
ML and the moment magnitude Mw to the liquefaction triggering correlation given in Eq. (7). The comparisons indicate that the use of the Richter magnitude ML instead of the moment magnitude Mw shifts the case
history data slightly upward. However, the amount of liquefaction case
history data below the liquefaction triggering correlation line remains
the same (only one point) whether the Richter magnitude ML or the
moment magnitude Mw is used. This result indicates that the distribution of the data in the Xie (1984) database can be reasonably well
interpreted using the calibrated liquefaction triggering correlation
given in Eq. (7) and that the position of the calibrated liquefaction triggering correlation given in Eq. (7) is not sensitive to the two magnitude
scales used for the data in the Xie (1984) database.
3. A probabilistic version of the new calibrated liquefaction
triggering correlation equation
Deterministic liquefaction triggering correlations only give a yes or
no answer to the question of whether soil liquees. However, liquefaction triggering correlations for general project sites cannot and should
not envelop all of the liquefaction case histories (Idriss and Boulanger,
2010) shown in Fig. 3. Therefore, statistical methods should be used to
evaluate the uncertainty in liquefaction triggering correlations.
In this section, probability contour curves for the new calibrated
liquefaction triggering correlation given in Eq. (7) for the combined
database were determined using the weighted maximum likelihood
technique that was initially adopted for a probabilistic liquefaction
309
analysis by Cetin et al. (2004) and veried using the weighted empirical
probability approach and Monte Carlo simulations (MCS).
3.1. The weighted maximum-likelihood technique and the empirical
probability approach
The logistic regression model used in the weighted maximumlikelihood technique was suggested by Juang et al. (2002) and Chen
and Li (2006); in this model, the probability of liquefaction (PL)
corresponding to the nominal safety factor ( ~F s ) is:
PL
1 ~F s =a
b ;
1
;
La; bjD
@
A
b>
b >
>
>
: j1
;
:i1 1 ~F =a ;
1 ~F s =a
s
Fig. 5. The distribution of the data in the Idriss and Boulanger (2010) case history database for sands with FC N 15 % reprocessed using the two different equivalent clean sand adjustment
procedures to determine (N1)60 compared with the calibrated liquefaction triggering correlation given in Eq. (7): (a) Idriss and Boulanger (2010); and (b) Youd et al. (2001).
310
Fig. 6. The distribution of the Xie (1984) liquefaction case history data after being reprocessed using the different earthquake magnitude scales compared with the calibrated liquefaction
triggering correlation given in Eq. (7): (a) Richter magnitude ML; and (b) moment magnitude MW.
1
3:851 :
~
1 F S =0:635
10
Combining Eqs. (7) and (10) allows the probabilistic version of the
liquefaction triggering correlation given in Eq. (7) for the combined
database to be expressed as:
"
#
N 1 60CS
N1 60CS 2
N 1 60CS 3
N 1 60CS 4
CRR ~F S P L exp
3:6
9:7
135:1
24:7
27:5
11
Table 2
Parameters a and b for formula (8) derived with various assumptions of weight
ratios wNL/wL.
wNL/wL
1.0
1.2
1.5
1.8
0.704
0.673
0.635
0.605
3.996
3.921
3.851
3.781
Fig. 7. The empirical relationship between PL and ~F s derived for different values of wNL/wL
using the weighted maximum-likelihood technique and the weighted empirical approach
to determining the calibrated liquefaction triggering correlation given in Eq. (7).
311
Fig. 8. The empirical relationship between PL and ~F s when wNL/wL = 1.5 is derived using
the weighted maximum-likelihood technique, the weighted empirical approach, and
Monte Carlo simulations for the calibrated liquefaction triggering correlation given in
Eq. (7).
4. Conclusions
In this paper, a new calibrated liquefaction triggering correlation
equation was proposed based on the combined databases of Idriss and
Boulanger (2010) and Xie (1984), the sensitivity of the case history
data to the components of the liquefaction analysis framework was investigated, and an empirical equation for determining the liquefaction
probability was expressed in the form of a mapping function that relates
the liquefaction probability PL to the nominal safety factor ~F s . The
primary conclusions were as follows,
(1) The possibility of mistaking liquefaction case history data as nonliquefaction case history data can generally be reduced by using the two
principles for determining the lower boundary of the liquefaction triggering correlation (4) with ve parameters. The equivalent clean sand
can be classied as non-liqueable when (N1)60CS 28 or CSR b 0.06.
(2) The new calibrated liquefaction triggering correlation equation
given in Eq. (7) is reasonably conservative and is insensitive to reasonable variations in the components of the liquefaction analysis framework used in this study.
(3) The use of wNL/wL = 1.5 with the weighted maximum-likelihood
technique is shown to be appropriate for the combined database.
Fig. 9. The liquefaction triggering correlations for liquefaction probabilities of 4%, 15%, 25%,
45%, 75%, and 95% and published by Idriss and Boulanger (2004, 2008), compared with the
combined case history data of Idriss and Boulanger (2010) and Xie (1984).
312
Yamada, S., Orense, R., Cubrinovski, M., 2011. Earthquake news: geotechnical damage due
to the 2011 Christchurch, New Zealand. Int. Soc. Soil Mech. Geotech. Eng. Bull. 5 (2),
2745.
Youd, T.L., Idriss, I.M., Andrus, R.D., Arango, I., Castro, G., Christian, J.T., Dobry, R., Finn,
W.D.L., Harder, L.F., Hynes, M.E., Ishihara, K., Koester, J.P., Liao, S.S.C., Marcuson III,
W.F., Martin, G.R., Mitchell, J.K., Moriwaki, Y., Power, M.S., Robertson, P.K., Seed,
R.B., Stokoe, K.H., 2001. Liquefaction resistance of soils: summary report from the
1996 NCEER and 1998 NCEER/NSF workshops on evaluation of liquefaction resistance
of soils. J. Geotech. Geoenviron. ASCE 127 (10), 817833.