Professional Documents
Culture Documents
Estimation of The Small-Strain Stiffness of Clean and Silty Sands Using Stress-Strain Curves and CPT Cone Resistance
Estimation of The Small-Strain Stiffness of Clean and Silty Sands Using Stress-Strain Curves and CPT Cone Resistance
2009
Japanese Geotechnical Society
ABSTRACT
The initial, linear elastic range of a soil stress-strain curve is often deˆned by the small-strain elastic modulus E0 or
shear modulus G0. In the present study, simpler and eŠective methods are proposed for the estimation of the small-
strain stiŠness of clean and silty sands; these are based on triaxial compression test results and the CPT cone resistance
qc. In the method based on stress-strain curves obtained from triaxial compression tests, an extrapolation technique is
adopted within the small-strain range of a transformed stress-strain curve to obtain estimates of the small-strain elastic
modulus. Calculated small-strain elastic modulus values were compared with the values measured using bender ele-
ment tests performed on clean sands and sands containing nonplastic ˆnes. The results showed that the method
proposed produces satisfactory estimates of the small-strain elastic modulus for practical purposes. In the CPT-based
method, two G0-qc correlations available in the literature were evaluated. For isotropic conditions, both correlations
produced reasonably good estimates of G0 for clean sands but overestimated it for silty sands. A G0-qc correlation
which is proposed takes into account the eŠect of silt content of the sand and stress anisotropy.
Key words: cone resistance, horizontal eŠective stress, hyperbolic stress-strain curve, modulus degradation, silt con-
tent, small-strain elastic modulus, triaxial tests (IGC: D6/E2)
545
ness from its initial maximum value G0. However, the niques such as well logging tests would be possible, iden-
degradation of the elastic modulus obtained experimen- tiˆcation of absolute values of soil density requires a
tally for real soils under static or quasi-static loading is calibration procedure.
not linear. In order to describe more realistically the Empirical equations can be used to estimate G0 if the
modulus degradation relationship, Fahey and Carter relevant soil parameters are known. Most of the empiri-
(1993) and Lee and Salgado (2000) proposed modiˆed cal equations proposed for the estimation of G0 are based
hyperbolic models for 2D and 3D conditions, respec- on either soil properties or in situ test results such as the
tively, as follows: SPT blow count number N or the CPT cone resistance qc
(Hardin and Black, 1966; Robertson and Campanella,
Ø »
g
G t
=1-f (6) 1983; Yu and Richart, 1984; Rix and Stokoe, 1991). The
G0 tmax property-based empirical equations are generally ex-
pressed as:
=« -f Ø » $ Ø I »
g ng
G J- J 2 I 20 1
1 (7)
Ø »
n
G 0 J - J 2max 20 10
G0 s?m
=CF(e) (9)
pA pA
where J2, J20, and J2max are the second invariants of the
deviatoric stress tensor at the current, initial, and failure where pA=reference stress=100 kPa; C and n are non-
states, respectively; I1 and I10 are the ˆrst invariants of the dimensional material constants; F(e) is the function of
stress tensor at the current and initial states, respectively; the void ratio; s?m is the mean eŠective stress. An example
and ng is the material constant. The parameter f in Eqs. of a property-based empirical equation is that suggested
(6) and (7) has the same role as Rf in Eq. (4). The by Hardin and Black (1966) and given by:
parameter g determines the shape of the degradation
Ø » Ø »
ng
G0 (eg-e)2 s?m
curve as a function of stress level. G0 is used in Eqs. (6) =Cg (10)
and (7), whereas Gi is used in Eq. (4). As mentioned pA 1 + e pA
previously, Gi is obtained from the transformed hyper- where Cg, eg and ng are the intrinsic soil variables that de-
bolic curve shown in Fig. 1. If the stress-strain curve of pend only on the nature of the soil and e is the initial void
Eq. (3) and the modulus degradation relationship given ratio.
by Eq. (4) in fact represented the actual stress-strain Application of Eq. (10) is limited as values of the in-
response of soils, then Gi would be the same as G0. As will trinsic soil variables are likely to be unknown for in situ
be discussed later, however, Gi is signiˆcantly diŠerent soils. In such cases, the use of in situ test results would be
from G0. more appropriate since estimation of these variables is
not necessary. This approach, however, requires the de-
velopment and validation of correlations between G0 and
SMALL-STRAIN MODULUS in situ test measurements for various soil conditions.
The small-strain shear modulus G0 is a state soil varia-
ble that is observed for strains in the range of 10-6 to 10-5
for sands. Within this small-strain range, soil behaves as ESTIMATION OF ELASTIC MODULUS BASED ON
a linear-elastic material and strains are recoverable. The TRANSFORMED STRESS-STRAIN CURVES
corresponding Young's modulus is denoted by E0. As G0 Transformed Stress-Strain Curves
is a state soil variable, its value is constant for a given soil In this section, a methodology to estimate the small-
condition, regardless of the nature of the loading type strain elastic modulus based on conventional triaxial
(Shibuya et al., 1992). compression stress-strain curves is proposed. For this
There are a number of ways to estimate G0 for a given purpose, triaxial compression test results by Salgado et
soil and stress state. These may be grouped into in situ al. (2000) and Lee et al. (2004) are adopted. The drained
tests, laboratory tests, and empirical equations (Yu and triaxial compression tests were performed under static
Richart, 1984; Baldi et al., 1989; Viggiani and Atkinson, conditions using a CKC automatic triaxial testing system
1995). While the laboratory tests allow accurate determi- (Chan, 1981). The soil samples were prepared with the
nation of the shear wave velocity Vs, these are still consi- slurry deposition method of Kuerbis and Vaid (1988). All
dered costly for use in routine projects as they require triaxial tests were performed for isotropically-consolidat-
complex testing systems speciˆcally designed to measure ed soil samples at strain rates that were slow enough to al-
the wave propagation characteristics or other small-strain low full dissipation of pore pressures during loading. The
soil properties. In experimental approaches using shear soil samples used in the triaxial compression tests were
wave velocity measurements, G0 is determined from the mixtures of Ottawa sand and nonplastic silt (silt contents
following relationship: sco equal to 0, 2, 5, 10, 15, and 20z by weight) prepared
in a wide range of relative densities ( DR=10 to 100z).
G0=r(Vs )2 (8)
Conˆning stresses (s?c ) ranging from 100 to 500 kPa were
where r=total mass density of the medium through used in these tests.
which the shear wave propagates. Values of r are known G0 was measured with bender element (BE) tests. The
with certainty in laboratory tests. In the ˆeld, while rela- BE tests were performed using a wave generator and
tive evaluation of soil density using geophysical tech- receiver (the bender elements) attached at the base
548 LEE ET AL.
Table 1. Basic properties of Ottawa sand with diŠerent silt contents (after Salgado et al., 2000)
pedestal and top platen of the triaxial apparatus (Salgado E0 versus Ei calculated following Duncan and Chang's
et al., 2000; Lee et al., 2004). The BE tests were per- procedure. As shown in Fig. 4, the initial tangent elastic
formed after consolidation for a number of triaxial sam- modulus obtained from Duncan and Chang's procedure
ples. G0 values for these samples were obtained using Eq. is signiˆcantly smaller than the measured small-strain
(8). The shear wave velocity from the test was measured elastic modulus. The degree of underestimation depends
from the eŠective length of the test sample (i.e., the dis- on the relative density of the sand. Figure 5(a) shows Ei
tance between tips of the bender elements) and the travel /E0 ratios as a function of DR. From Fig. 5(a), it is seen
time of the shear wave identiˆed from the ˆrst arrival of that the denser the soil, the higher the values of Ei/E0.
the signal generated by the source bender element. Ac- This indicates that, as the soil becomes more dilative, the
cording to Viggiani and Atkinson (1995), possible errors underprediction of the small-strain modulus by Duncan
in G0 values from the BE tests could be up to 15z, in ex- and Chang's (1970) procedure becomes less pronounced.
treme cases. The errors are mainly due to deviation from Such tendency can be more clearly observed in Fig. 5(b).
1-D wave propagation, wave interference at the caps, and According to Bolton (1986), the dilatancy of sands,
diŠerent time delays between the generation of the electri- which is essentially controlled by relative density and con-
cal signal and its transformation into a mechanical pulse ˆning stress can be quantiˆed using the dilatancy index IR
(Salgado et al., 2000). As indicated by Arulnathan et al. as follows:
(1998), these factors sometimes compensate each other,
while sometimes they do not. Therefore, it can be said
that values of G0 from the BE tests considered in this
«
IR=ID Q-ln Ø 100s?mp
pA » $-R (11)
paper may include errors of up to 15z, while the actual where ID=relative density as a number between 0 and 1;
errors are likely to be smaller due to the self-compensat- pA=reference stress=100 kPa; s?mp=mean eŠective stress
ing eŠects (Salgado et al., 2000). Other basic properties of at peak strength (in the same units as pA ); and Q and R=
the test sands used in the triaxial and bender element tests intrinsic soil variables. Values of Q and R for the clean
are given in Table 1. and silty sands tested are given by Salgado et al. (2000)
Figure 2 shows the stress-strain, modulus degradation, and Lee et al. (2004). As shown in Fig. 5(b), the degree of
and transformed stress-strain curves obtained from underprediction of the small-strain modulus obtained
drained triaxial compression tests performed on two with Duncan and Chang's procedure becomes less
clean sand (i.e., sco=0z) samples with DR=38 and 63z pronounced with increasing IR.
at s?3 =400 kPa (tests were also performed for other sands The diŠerence between Ei and E0 can be attributed to
with diŠerent conˆning stress). In Fig. 2(b), the secant the inability of the hyperbolic relationship to represent
elastic modulus E was normalized with respect to E0 (the the rate of modulus degradation in the small-strain range.
small-strain elastic modulus obtained from the bender The transformed stress-strain relationship shown in Fig.
element tests). As expected, the denser sand sample 1 is given by:
shows higher strength and lower modulus degradation
e
rate at a given stress level. The curves shown in Fig. 2(c) = a + be (12)
s
were plotted until the peak strength was reached (axial
strain levels of around 15.7 and 4.7z were observed for The initial tangent elastic modulus is then deˆned as
the sand samples with DR=38 and 63z, respectively). (Tatsuoka et al., 1993):
Figure 3 shows measured peak deviatoric stresses s?d, p
d ( e /s ) 1 Rf
versus those calculated using Eqs. (1) and (3) by Kondner lim = b= = (13)
eª / de slim sf
(1963) and Duncan and Chang (1970), respectively. As
shown in Fig. 3, both approaches are in reasonably good
e 1
agreement with the measured values of s?d, p. lim =a= (14)
eª 0 s E0
Determination of Small-Strain Elastic Modulus from where s and e are the axial stress and strain; slim is the
Transformed Stress-Strain Curves limit axial stress as deˆned in Fig. 1; and sf is the axial
Figure 4 shows measured small-strain elastic modulus stress at failure. If the stress-strain curves of the soils fol-
SMALL-STRAIN ELASTIC MODULUS 549
Ø »
0.389
G0 pA
= G1 (16)
qc qc
where G1=correlation parameter=50; and pA=reference
stress=100 kPa=1 bar. Originally, Eq. (16) was
proposed in terms of the SPT blow count N by Imai and
Tonouchi (1982). It was later modiˆed by Robertson and
Campanella (1983) by converting N to qc. The relation-
ship between qc and G0 of Rix and Stokoe (1991), on the
other hand, can be given in a normalized form as follows:
Ø »
- 0.75
G0 qc pA
=G2 (17)
qc pA s?v0
where s?v0=initial vertical eŠective stress; and G2=corre-
lation parameter=290; and pA=reference stress=100
kPa. The units in Eqs. (16) and (17) are the same as the
units of pA.
Fig. 10. Values of the correlation parameters G1 and G2 for sand con-
taining diŠerent silt contents
Fig. 12. Values of G0/qc for the conditions given in Table 2, as well as
those from Eq. (17)
Ø »
- 0.75
250 100 125 100 G0 qc pA
5 30, 70 500 200 250 200 = G3 (20)
qc pA s?m0
750 300 375 300
250 100 125 100
10 30, 70 500 200 250 200 G3=110・e-0.23・s +160
co
(21)
750 300 375 300
where sco=silt content in z; s?m0=in situ mean eŠective
250 100 125 100
15 30, 70 500 200 250 200 stress and pA=reference stress=100 kPa=1 bar.
750 300 375 300 Note that Eqs. (20) and (21) apply only to recently
250 100 125 100 deposited, uncemented clean and silty sands with proper-
20 30, 70 500 200 250 200 ties similar to the ones considered in this study. It is well
750 300 375 300
documented in the literature that in situ G0-qc correla-
tions depend on many factors such as sand compressibili-
SMALL-STRAIN ELASTIC MODULUS 555