Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
I: Model
for Cemented and Uncemented Sands and Gravels
Juan M. Pestana, M.ASCE1; and Lynn A. Salvati, M.ASCE2
Abstract: A simple formulation is presented that predicts the nonlinear small strain behavior of cemented and uncemented granular soils.
Its performance is evaluated through the comparison of model predictions to results from laboratory tests. A companion paper evaluates
the performance of this model implemented in a site response analysis code through comparison with the measured response at two sites.
The formulation for the maximum shear modulus, Gmax, which is selected through the evaluation of existing formulations and data, is
presented with the hysteretic model developed to describe the shear modulus reduction and damping increase with increasing strains. Few
parameters are needed to predict the small strain response, and correlations between model parameters and index properties of granular
materials are presented when possible. The model, SimSoil, is shown to capture the cyclic response for sands and gravels with varying
densities over a wide range of pressures measured in laboratory tests, including cases when cementation is present.
DOI: 10.1061/共ASCE兲1090-0241共2006兲132:8共1071兲
CE Database subject headings: Shear modulus; Constitutive models; Cementation; Granular materials.
Introduction medium dense to dense sand with gravel and stiff clay, was
230– 300 m deep. The 1989 Loma Prieta earthquake and the 1994
Site response analysis is an essential element in geotechnical Northridge earthquake provided more examples of deep, stiff soil
earthquake engineering. The equivalent linear procedure in the deposits significantly amplifying ground motion. These events
program SHAKE 共Schnabel et al. 1972兲 has been used in practice have shown the need for methods to predict the response at deep,
to approximate the nonlinear properties of soil and perform site stiff sites, since many cities are located on such deposits. There-
response analyses for over 30 years. However, as the field of fore a simple formulation is developed that can describe the maxi-
mum shear modulus, shear modulus reduction, and damping of
earthquake engineering progresses, the limitations of SHAKE
granular materials over a wide range of densities and pressures
become more restrictive. As a result, the use of nonlinear site
that is appropriate for use in nonlinear site response analysis. This
response analyses has increased. However, to obtain the benefits
formulation provides more control over the shear modulus reduc-
from performing a nonlinear analysis, a model that can accurately
tion and damping curves than hyperbolic models employed in
represent the nonlinear behavior of the soil must be used. nonlinear site response analysis codes such as DESRA 共Lee and
Most response analyses of stiff soil sites do not include depths Finn 1978兲, while requiring only a few parameters unlike the
exceeding 50 m. As an example, the 1997 Uniform Building models employed in nonlinear site response codes such as
Code only considers the top 30 m 共⬃100 ft兲 when developing SUMDES 共Li et al. 1992兲.
characteristic soil profiles for stiff materials, following the re- The first section of the paper describes the selected formula-
commendations by Borcherdt 共1994兲. Similar to deep, soft clay tion for the maximum shear modulus, Gmax. The model predic-
deposits, deep deposits of primarily dense granular material can tions are compared with available data for several soils over a
significantly amplify ground motions. The 1967 Caracas, Venezu- wide range of conditions, including very high pressures, and data
ela earthquake, which provided undisputed evidence of the effect for cemented sands. In the second section of the paper, the hys-
of “local soil conditions” on structural response, was also the first teretic formulation that describes the reduction in shear modulus
event to focus attention on the amplification potential of stiff and increase in damping with increasing shear strain is detailed.
soils. Much of the damage from that event occurred in the Palos The model predictions are again validated with laboratory test
Grandes area, where the thickness of alluvium, which consisted of data for sands and gravels, cemented and uncemented, and corre-
lations between model parameters and index properties of the
1
Associate Professor, Dept. of Civil and Environmental Engineering, materials are developed.
Univ. of California, Berkeley, CA 94720.
2
Clare Boothe Luce Assistant Professor, Dept. of Civil Engineering
and Geological Sciences, 156 Fitzpatrick Hall, Univ. of Notre Dame, Small Strain Stiffness
Notre Dame, IN 46556.
Note. Discussion open until January 1, 2007. Separate discussions
must be submitted for individual papers. To extend the closing date by Previous Investigations
one month, a written request must be filed with the ASCE Managing
Editor. The manuscript for this paper was submitted for review and pos- As the database of laboratory test results has increased over the
sible publication on July 11, 2003; approved on September 29, 2005. This years, many empirical relationships have been developed to pre-
paper is part of the Journal of Geotechnical and Geoenvironmental dict the maximum shear modulus, Gmax. For sand, Gmax is prima-
Engineering, Vol. 132, No. 8, August 1, 2006. ©ASCE, ISSN 1090- rily controlled by the density and confining pressure, which has
0241/2006/8-1071–1081/$25.00. been represented in formulations for Gmax dating from the 1960s,
e.g., Hardin and Richart 共1963兲; Seed and Idriss 共1970兲. If aniso-
tropy is not considered, a general formulation for Gmax can be Fig. 2. 共a兲 Effect of particle angularity on Gmax for selected uniform
proposed cohesionless soils; 共b兲 effect of density on Gmax for selected sands
Gmax
pat
冋 册
= Gb · f 1关e兴·f 2
p
pat
共1兲
and gravels
where p = mean effective stress; pat = atmospheric pressure; have problems describing the shear modulus of sands that can
e = void ratio; and Gb = material constant. All of the formulations have high formation void ratios, such as Dog’s Bay sand.
developed from 1960 to 2000 that were reviewed 共Salvati 2002兲
use a power law to describe the effect of confining pressure on
Selected Formulation
Gmax. Formulations that prescribe the exponent for the power law
are in a well-bounded range of 0.4–0.6 with the exception of the Fig. 1共a兲 compares predictions of Gmax using three different
value of 0.33. The 0.33 value suggested by Pestana and Whittle power-law exponents with measured values of Gmax for Monterey
共1995兲 and later used by others 共Assimaki et al. 2000; Kausel and sand. The power-law exponent, n = 0.5 tends to fit the data for
Assimaki 2002兲 is based on the Hertzian contact theory for elastic uniform sands better than n = 0.4, which was suggested by Iwasaki
compression of spheres 共Mindlin and Deresiewicz 1953兲. The ef- and Tatsuoka 共1977兲, or n = 0.33 共Pestana and Whittle 1995兲.
fect of void ratio is not as uniformly described, however. It is Based on its wide range of applicability 共many other materials
desirable for Gmax to approach zero as the void ratio increases to examined were not included here for conciseness兲 a power-law
a limiting value. Some of the relations do not uphold this, and exponent n = 0.5 can be used as a default value in the Gmax rela-
tionship for sands, unless there is data that indicates a different Jamiolkowski et al. 共1991兲, f 1关e兴. The f 2关e兴 function does not
value would be more appropriate. Gravels and materials with a describe the trend of the data as well and f 3关e兴 does not match
significant amount of gravel tend to have higher values of n the data at higher void ratios as well. Using the formulation
共n = 0.5– 0.75兲, as shown in Fig. 1共b兲. f 1关e兴 = e−1.3, all of the different materials can be described with a
To isolate the effect of void ratio on Gmax, only values of Gmax material constant, Gb, that is within a well-bounded range. Fur-
measured at a mean effective stress of within 10% of 101 kPa thermore, the value of Gb seems to be well-correlated with the
共1 atm兲 were analyzed initially. In Fig. 2共a兲, the test data is shown angularity of the material for fairly uniform sands; the sands with
with three functions used to describe the effect of void ratio on more angular grains tend to have higher values of Gb. Sands and
Gmax: f 1关e兴 = e−1.3 共Jamiolkowski et al. 1991兲, f 2关e兴 = 共1 + e兲 / e gravels, which are very well graded 共Cu ⬎ 7兲 tend to have lower
共Pestana and Whittle 1995兲, and f 3关e兴 = 共2.17− e兲2 / 共1 + e兲 共Hardin values of Gb 共Gb = 200– 400兲. In Fig. 2共b兲, reported Gmax values at
and Richart 1963兲. Both f 2关e兴 and f 3关e兴 are shown with typical all measured confining pressures are shown with the formulation
values of the material constant for those formulations, whereas proposed by Jamiolkowski, f 1关e兴 = e−1.3, after normalizing the data
f 1关e兴 is shown with a range of Gb values from 400 to 800. The by the confining pressure and Gb for the material. Almost all of
trend of the data is best matched by the formulation presented by the data is within the 20% error bars shown. Based on this ex-
Gmax
pat
= Gbe−1.3 冉 冊
p
pat
n
共2兲 where ␦b = 1 − p / pb; pb = pref共1 / e兲1/c; and pref, , c, and
Kb = material constants. The equivalent stress on the limiting com-
pression curve 共LCC兲 at a given void ratio is pb, and the dimen-
sionless measure of distance between the current mean effective
Gb values for the proposed formulation were determined for many
stress and the stress on the LCC is given by ␦b.
commonly used test sands and gravels based on published data, Jovicic and Coop 共1997兲 performed bender element tests on
and the values are listed in Tables 1 and 2. Dog’s Bay sand and Ham River sand isotropically compressed to
very high pressures. When the compression curves are plotted in
Shear Stiffness at High Pressures log e − log p space, as in Fig. 3, it can be seen that both of these
materials have reached the LCC during compression. Therefore
The previous section discussed the elastic shear modulus, which the maximum shear modulus cannot be determined simply using
describes deformation in the soil that is solely due to the elastic the elastic formation, explaining the nonlinearity of the plot of
compression of the soil skeleton. However, as the confining stress Gmax in log G-log p space. Using the compression curves that
increases, particle rearrangement begins, and plastic strains begin were presented by Jovicic and Coop 共1997兲 and the data pre-
to accrue. Then, as the increase in stresses continues, the particles sented by Pestana 共1994兲, c, pref, and were determined for both
begin to crush and the strains become increasingly dominated by sands and are shown in the figures. Kb was determined from the
the plastic contribution. Using the formulation discussed above to reported response at low stress levels. Then the compression
describe the elastic shear modulus of the soil within the frame- model was used to predict Kmax values with increasing confining
work of the compression model developed by Pestana and Whittle pressure for the lowest and highest initial void ratios tested for
共1995兲 the plastic and elastoplastic stiffness can be described as each of these sands. The Gmax values are calculated from the Kmax
well as the elastic stiffness. The functions used to describe the values based on the Poisson’s ratio. The two sets of predictions
effect of confining pressure and void ratio on the elastic bulk 共for each void ratio兲 are shown in Fig. 3: one in which the Pois-
modulus, K, have been modified from those used by Pestana and son’s ratio is assumed to be constant, and a second in which the
Whittle 共1995兲, as shown below, based on the formulation pre- Poisson’s ratio,, is assumed to vary with the void ratio as shown
sented in the previous section in the equation below
Fig. 3. Comparison of measured and predicted Gmax values over a large stress range
Previous Formulation
Pestana and Whittle 共1999兲 used the perfectly hysteretic formula-
tion in the generalized constitutive model, MIT-S1, to predict the
modulus reduction and damping curves. In the perfectly hysteretic
formulation proposed by Hueckel and Nova 共1979兲 the isotropic
tangential moduli are a function of the most recent stress reversal
state. The concept was developed to describe unload-reload
cycles for clay, and based on this relation the reversibility of
strain was guaranteed over a stress loop between two stress rever-
sal points. The modulus reduction at a constant mean effective
stress in MIT-S1 is described by
Gtan 1
= 共6兲
Gmax 共1 + s兲共1 + ss兲
Fig. 4. Effect of light cementation on confining stress dependence of
Gmax: 共a兲 model predictions; 共b兲 model predictions and measured where and s = constants for a given material; and
values for Ottawa sand s = dimensionless stress measure. s is given by
:˙ = 再 艌0 loading
⬍0 unloading
冎 共8兲
This formulation describes the unload-reload cycles in monotonic Fig. 5. Translation of hysteresis loop
loading well, and can predict reasonable behavior for cyclic tests
on soils. However, when describing the response to cyclic load-
ing, the parameters do not provide adequate control of the modu- is capped so that it cannot exceed 0.005 to limit its effect in the
lus reduction and damping curves, since the parameters do not intermediate to large strain range. Masing rules can be used to
provide independent control of the response at larger strains. describe the hysteresis loops, and the loading/unloading condition
Therefore it is difficult to model the hysteretic behavior of a wide is the same as given above.
range of soils well. Since the objective of this paper is to describe a simple model
that can be used to evaluate the site response at deep sites due to
vertically propagating shear waves, the formulation can be sim-
New Formulation
plified for this purpose. The analytical solutions, which can be
A new formulation was developed to describe the nonlinear re- used to determine shear modulus reduction and damping curves,
sponse of soils. This formulation, which is shown below, provides are developed and shown below for the case in which pore pres-
better control of the shear modulus reduction curves without an sure changes are not considered. For one-dimensional 共1D兲 site
increase in the number of parameters. It also provides improved response analyses, s = 冑2 / p, where = horizontal shear stress;
prediction of damping at very low strains, which is a difficulty and p = mean effective stress, which is considered to be constant.
with hysteretic models, such as the one described above. These The factor of 冑2 is a result of the transformed variables that are
improvements result in the ability to model a much wider range of used by Pestana 共1994兲 and the MIT family of models 共Potts and
materials with more accuracy than with the previous formulation. Zdravkovic 1999兲. To find the damping and shear modulus reduc-
Also, the new formulation provides better predictions of the shear tion curves, it is convenient to translate the hysteresis loop so that
modulus and damping at high pressures, which are necessary to it begins at the origin as shown in Fig. 5. For the backbone curve,
model deep deposits accurately the reversal state is set at zero 共rev = 0 and rev = 0兲, and Eq. 共9兲
Gtan 1 can be rewritten as
= 共9兲
Gmax 1 + ss0.75
* + s s + a s
2 2 Gtan 1
= 共10兲
Gmax 1 + s共c*兲 + sc + 2a共c兲2
0.75
where s and a = material parameters, which describe the non-
linearity. The parameter, a, controls the intermediate to large where c = 冑2 / p, and * is limited to a maximum value of 0.005/c.
strain behavior; s controls the small to intermediate strain be- The definition of tangent shear modulus, Gtan = d / d␥, is used to
havior. The definition of s is given above, and the parameter s* solve for the shear strain, which is given in Eq. 共11兲
冋 册
冦 冧
s共c*兲1.75 s 2a
c1 c + + 共c兲2 + 共c兲3 if 共c兲 艋 0.005
1.75 2 3
冋 册
␥= 共11兲
s共c*兲1.75 s 2a
c1 c + + 共c兲2 + 共c兲3 + s共c*兲0.75c** otherwise
1.75 2 3
where c1 = p / 冑2Gmax and ** = − *. Since the hysteresis loop is translated and the reversal state is set at zero, the shear strain calculated
in Eq. 共11兲 is taken as the double amplitude shear strain. The equivalent viscous damping ratio 共兲 is a function of the ratio of energy
dissipation 共⌬W兲 in one hysteretic cycle to the maximum stored energy 共W兲. For models in which the shear strain is expressed in terms
of shear stress, damping can be evaluated as 共Ishihara 1996兲
冕
冤 冥
a
2 ␥共兲d
1 ⌬W 2 0
= = 1− 共12兲
4 W a␥共a兲
冉 冊再 冋冉 册冎
= 共13兲
2
1−
2c1
␥
c s共c兲2 2a共c兲3
2
+
6
+
12
冊
+
s共c*兲1.75 *
1.75
冉
2.75
+ ** + 冊
s共c*兲0.75c2**
2
otherwise
With Eqs. 共11兲 and 共13兲 the damping curves can be developed, range of pressures and/or void ratios. However, if the shear modu-
and the secant shear modulus needed for the shear modulus re- lus is not measured at small enough strains to be considered Gmax,
duction curves is given by Gsec = a / ␥a, as shown in Fig. 5. then the Gb value described earlier can be used to calculate Gmax
Fig. 6共a兲 shows the influence of s on the shear modulus reduc- and the model predictions of shear modulus versus shear strain
tion and damping curves. As s increases, the shear modulus can be used to compare with the available data. Then an appro-
decreases and the damping increases for a given strain. In many priate value of s can be determined simply by comparing the
of the following figures, the ratio of shear modulus to a reference
SimSoil model predictions to the measured data. The parameter
shear modulus, Gref, is shown. Gref is the shear modulus at a shear
a is selected in a similar manner to s, if there is data at large
strain of 0.0001%. The influence of a is similar to s, but the
enough strains to compare to the predictions.
effects are seen only at larger strains as shown in Fig. 6共b兲. With
these two parameters, the model can predict the behavior over a
wide range of stresses 关see Fig. 6共c兲兴. Also, since the parameters Model Evaluation
are based only on the material, the effects of varying density can
be described with the same single set of parameters as shown in Fig. 7 shows the model predictions, measured shear modulus, and
Fig. 6共d兲. This hysteretic formulation together with the formula- damping values for Monterey No. 0 sand over a range of pres-
tion for Gmax comprises a model, SimSoil, that can predict the sures. Considering the range in the results reported by Chung
small strain behavior of granular materials. et al. 共1984兲 and Saxena and Reddy 共1989兲, the SimSoil model
fits the data well on average. Comparisons of the model predic-
Parameter Selection tions with available test data for Toyoura sand from Kokusho
The parameters s and a can be determined using modulus re- 共1980兲, as seen in Fig. 8, demonstrate that the effect of confining
duction and/or damping data from any test if the general form of pressure is much greater than the effect of density on the small
the model is used. If the simplified version of the model is used strain nonlinearity, and the SimSoil model captures that trend.
关Eqs. 共11兲–共13兲兴, it is most appropriate to select the parameters The model can match shear modulus reduction and damping data
from resonant column, torsional, or simple shear tests. Normaliz- ranging over almost two orders of magnitude in pressure in the
ing the curves 共Gsec / Gmax versus shear strain兲 simplifies the de- case of Mortar sand 共Laird and Stokoe 1993兲, as shown in
termination of parameters for a material with test data over a Fig. 9共a兲. This formulation can also be used to describe the be-
Fig. 6. 共a,b兲 Effect of material parameters describing the nonlinearity in shear and 共c,d兲 predicted effect of confining pressure and density on the
shear modulus and damping
Cemented Soils
As the cementation level of a material increases, it becomes more
brittle, and therefore begins to degrade at a much lower strain
level than an uncemented material. The increasing cementation
also results in an increase in damping at lower strain levels than
in the uncemented material. These trends are illustrated in the
work done by Saxena et al. 共1988兲 and the model predictions as
shown in Fig. 11. Further evidence of the model’s applicability to
cemented materials can be shown in conjunction with the results
Fig. 7. Comparison of measured and predicted values of shear from Yasuda et al. 共1996兲. In this study both undisturbed and
modulus and damping for Monterey No. 0 sand reconstituted samples of gravel were tested. From the shear
modulus measurements for the gravel samples during loading and
unloading, it is evident that the lower gravel samples have some
level of cementation, most probably due to aging effects. Since
the gradation of the reconstituted sample, R-3, is the most similar
to the lower gravel, it is used for comparison with the undisturbed
Fig. 8. Comparison of measured and predicted values of shear modulus and damping for Toyoura sand
Fig. 10. Empirical correlation for model parameter, s, for poorly
graded granular soils
Fig. 9. Comparison of measured and predicted values of shear Fig. 11. Comparison of measured and predicted values of shear
modulus and damping for 共a兲 mortar sand; 共b兲 rockfill modulus and damping for lightly cemented Monterey No. 0 sand
Acknowledgments
Notation