HYPERBOLIC PARAMETERS FOR C O M P A C T E D SOILS

By Marco D. Boscardin, 1 Member, ASCE, Ernest T. Selig, 2

Fellow, ASCE, Ruey-Shyan Lin, 3 and Gwo-Rong Yang*

ABSTRACT: Finite element methods are being used more and more during design,
in cases involving compacted soil-structure interaction. In general, it is not prac-
tical to conduct extensive tests to obtain the compacted soil properties required by
the finite element methods during the design phase. Alternatively, if finite element
approaches are used for developing design tables, soil properties representative of
typical soil types and compaction specifications are required. To provide the needed
design soil parameters for a wide variety of soil conditions, laboratory testing is
carried out on three soils: a sand, a silt, and a clay. These soils are prepared at
density states ranging from loose to 95% of the maximum from the standard com-
paction tests [American Society for Testing and Materials (ASTM) D698, Amer-
ican Association of State Highway and Transportation Officials (AASHTO) T-99].
The tests used to obtain the soil parameters are triaxial compression, isotropic
compression, and one-dimensional compression. A consistent set of soil design
parameters are obtained by fitting test results to a hyperbolic soil model repre-
senting Young's modulus and bulk modulus as functions of stress.

INTRODUCTION

Hyperbolic stress-strain equations have been found to be useful tools to

model the nonlinear stress-strain behavior of many soils prior to failure (Duncan
et al. 1980). Such equations are commonly employed in conjunction with
finite element methods to investigate soil-structure interaction. A problem
typically encountered when dealing with soil-structure interaction in a system
where compacted soils play a significant role is that the properties of the
compacted soil are not specifically known prior to analysis and design, and
may not be determined until construction is well underway. This situation
creates difficulties in cases where knowledge of the soil behavior and the
soil-structure interaction is necessary for efficient and rational design of
structural elements. One approach to the problem is to determine hyperbolic
parameters for various classes of soils over a range of compaction levels for
use in design.
In this paper, the results of a study investigating the hyperbolic parameters
of three compacted soils are reported. The three soils include a well-graded
sand, a sandy silt, and a lean clay. These soils were each compacted to
several different densities ranging from the dumped condition to 95% stan-
dard proctor density. The moisture content during compaction was the stan-
dard proctor optimum moisture content and the samples were maintained at
that water content throughout the testing. The aim was to select soils and
compaction conditions that would encompass a wide range of conditions typ-
ically encountered.
Triaxial shear and isotropic compression tests were conducted to obtain a
'Proj.
2
Mgr., GEI Consultants, Inc., 123 Sheep Davis Road, Concord, NH 03301.
3
Prof., Dept. of Civ. Engrg., Univ. of Massachusetts, Amherst, MA 01003.
4
Grad. Asst., Dept. of Civ. Engrg., Univ. of Massachusetts, Amherst, MA.
Grad. Asst., Dept. of Civ. Engrg., Univ. of Massachusetts, Amherst, MA.
Note. Discussion open until June 1, 1990. To extend the closing date one month,
a written request must be filed with the ASCE Manager of Journals. The manuscript
for this paper was submitted for review and possible publication on June 2, 1989.
This paper is part of the Journal of Geotechnical Engineering, Vol. 116, No. 1,
January, 1990. ©ASCE, ISSN 0733-9410/90/0001-0088/$ 1.00 + $.15 per page.
Paper No. 24242.
88
 TABLE 1. Soil Properties
Compaction
ASTM D698
Max dry
Classification
Liquid Plasticity density Optimum
Soil type uses AASHTO limit index (mg/m3) moisture
(D (2) (3) (4) (5) (6) (7)
Well-graded sand SW A-l-a Not 2.21 7.4
plastic
Sandy silt ML/CL-ML A-4 20 4 1.91 12.1
Silty clay CL A-6 32 15 1.65 21.0
Note: 1 lb/cu ft = 0.016 Mg/m3.

complete set of hyperbolic parameters. These parameters were evaluated by

comparing a finite element simulation of one-dimensional compression to
one-dimensional compression test results for the same soils. Further details
of the study are given in Lin (1987) and Yang (1987).

SOILS INVESTIGATED

Three soils are examined in this study: a well-graded sand (SW), a sandy
silt (ML) and a lean clay (CL). These provide a minimum number of soils
indicative of a wide range of types that might be used in construction, for
example, during installation of buried pipes and culverts. General infor-
mation concerning these soils is presented herein and in Table 1.
The well-graded sand is a clean gravelly coarse to fine sand, with all
particles less than 10-mm diameter. This soil is generally considered a high-
quality granular backfill material and is classified as SW in the Unified Sys-
tem [American Society for Testing and Materials (ASTM) D2487] or A-l-
a in the American Association of State Highway and Transportation Officials
(AASHTO) system (ASTM D3282).
The sandy silt is a mixture of fine sand and silt with relatively low plas-
ticity. This soil is generally regarded as a lower-quality backfill material than
the SW soil, but is still adequate for many cases. This soil is classified as
CL-ML/ML or A-4.
The lean clay was created by mixing 50% silica flour and 50% kaolinite,
all finer than the no. 200 sieve. This is the most plastic soil of the three. It
is generally regarded as the lowest-quality backfill material of the three from
the standpoint of strength and deformation. This soil is classified as CL or
A-6.

HYPERBOLIC MODELS

One method of simulating the nonlinear, inelastic, and stress-level-depen-

dent stress-strain behavior of soil is to use a nonlinear incremental approach
where each increment treats the soil as linear material. The stress-strain re-
lationship is assumed to be governed by a generalized form of Hooke's law
with the elastic parameters based on the current stress state (Duncan and
Chang 1970).

89
 Previous work (Kondner 1963; Kondner and Zelasko 1963) has shown that
the stress-strain behavior of many soils can be modeled, up to failure, by
hyperbolas with equations of the form

(a, - 03) = (1)

1 e
— +
_
Ei (0-1 Cr 3 ) u l t

where (o^ - cr3) = principal stress difference; (at - a3)ult = asymptotic value
of the stress difference at large axial strain; e = axial strain; and Et = initial
tangent modulus.
The ultimate stress difference, (CTJ - CT3)UU, is defined in terms of the actual
failure stress difference by the failure ratio, Rf, as

* /= ^ ^ (2)
(cr, - cr3)ult
The variation of the initial tangent modulus in response to changes in con-
fining pressure can be represented using a power law approach as suggested
by Janbu (1963):

Ei = Kpj^j (3)
where pa = atmospheric pressure, used to make K and n nondimensional;
K = modulus number; and n — modulus exponent.
The failure stress difference is a function of the confining stress, cr3. One
approach is to represent the shear strength-confining stress relationship us-
ing a straight line Mohr-Coulomb failure envelope with an expression of the
form
2c cos <b + 2cr3 sin (b
(a, - as)/ = / . ' (4)
1 — sin 4>
where § = slope of the linear failure envelope, termed angle of internal
friction; and c = intercept of the linear failure envelope, termed cohesion
intercept. However, Mohr-Coulomb failure envelopes are not always linear
and so a best-fit straight line must be used or the expression must be re-
written so that <|) can vary with <x3. One approach is to let cf) vary with 0-3,
as given by the following equation (Duncan et al. 1980):

4. = <|>0- A* log (-) (5)

where §„ = value of <f> when cr3 = pa; and A(j> = reduction in $ for an order-
of-magnitude increase in CT3.
If Eq. 1 is differentiated with respect to e, and Eqs. 2, 3, 4, and 5 are
substituted into the resulting expression, the following equation for the tan-
gent modulus, E„ or the instantaneous slope of the stress-strain curve is
derived (Duncan et al. 1980):
(6)
E,=
Rf(l - sin 4>)(g! - g3)
2c cos 4> + 2cr3 sin 4> O
90
Thus, the parameters describing Young's modulus, based on a hyperbolic
fit to the stress-strain curve, are K, n, c, §0, and Rf.
The second soil property needed for input into the generalized form of
Hooke's law is bulk modulus. Bulk modulus, B, is defined as

B=- (7)
Ae.vol
where Ao-m = the change in the mean normal stress; and Aevol = change in
volumetric strain. For the conventional triaxial compression test with a con-
stant confining pressure and beginning with a hydrostatic stress state, Eq. 7
may be expressed as

B=^ ^
•J^vol

which is a secant modulus definition whose value depends on the selection

of the data points used in its determination. Duncan et al. (1980) offer meth-
ods to select the data points that will be representative of the soil. The vari-
ation of B with confining stress,CT3,can then be represented using a power
law approach by

B= *»PJ-J (9)

where Kb = bulk modulus number; and m = bulk modulus exponent.

The limitations of this approach are: (1) That the conventional triaxial test
typically generates volumetric strain data only for saturated samples, which
precludes representing compacted samples in the as-compacted or partially
saturated states; and (2) that the bulk modulus employed is a secant rather
than a tangent value.
To overcome the aforementioned limitations, alternative methods of rep-
resenting volumetric strain as a function of mean stress were investigated
(Domaschuk 1969; Jennings 1964; Domaschuk and Valliappan 1975). This
led to the use of an isotropic compression test with the results represented
by a hyperbolic equation of the form
B,eVl
(10)
6
vol

e,

where 5, = initial tangent bulk modulus; e„ = asymptotic value of the vol-

umetric strain at large stresses; and a,„ = mean confining stress.
Differentiating Eq. 10 with respect to volumetric strain gives the following
expression for the tangent bulk modulus, B,:

B, = B, 1 +• (11)
(B,-e„).
The values for 5, and e„ are the intercept and the inverse of slope, respec-
tively, of a plot of o-,„/evoi versus am.
91
TESTING PROGRAM

Thirty-two triaxial compression tests were performed. Each soil was tested
at its standard proctor optimum water content (hereafter denoted wopt) in the
dumped (noncompacted) state, and at densities corresponding to 85% and
95% standard proctor maximum dry density (MDD). The dumped state cor-
responded to 61%, 49%, and 45% standard proctor maximum dry density
for the SW, ML, and CL soils, respectively. Confining pressures were 35
(5), 103 (15), 207 (30), and 310 (45) kPa psi. An axial compression rate
of 0.5 mm/min was employed during the shear phase of the tests.
The triaxial samples were 71 mm in diameter and 157 mm in length. The
dumped samples were prepared by pouring a known weight of soil at wopt
into a mold lined with a membrane. No tamping was employed and height
of fall was controlled to yield the loosest sample reasonably possible. The
85% and 95% MDD samples were prepared by tamping a known weight of
soil at wopt into a membrane-lined mold in six layers of equal height and
weight. The procedure was similar to that suggested by Ladd (1978).
Eighteen isotropic compression tests were performed. Each of the three
soil types were tested at five densities: dumped, 80% MDD, 85% MDD,
90% MDD, and 95% MDD; with the moisture content at wopt in each case.
Two dumped samples were tested for each soil type. Volume change was
measured at 10-kPa increments of isotropic stress up to 350 kPa. The iso-
tropic samples were 38 mm in diameter and 76 mm in length. The specimens
were prepared in the same fashion as the specimens for the triaxial testing.
The isotropic stress was provided by a triaxial cell with cell pressure in-
creased without piston loading. Volume changes of the soil samples were
determined to the nearest 1 mm3 by monitoring the changes in volume of
the cell fluid. Corrections were made for the expansion of the cell and sys-
tem as the confining pressure was increased. Each pressure increment was
maintained until volume change readings stabilized and then the next incre-
ment was applied.
Nine one-dimensional compression tests were performed. Each of the three
soil types was tested at wopt in the dumped state and at 85% MDD and 95%
MDD. The specimens were prepared by tamping the soil into a 155-mm
diameter rigid ring 51-mm deep. Three lifts of equal height and equal weight
of soil were used.
The one-dimensional compression specimens were loaded continuously at
a rate of 0.5 mm/min and deformation readings were recorded at each 35
kPa (5 psi), applied up to 345 kPa (50 psi), and at each 69 kPa (10 psi)
applied thereafter to a maximum of 690 kPa (100 psi).

PARAMETER DETERMINATION

Triaxial test results are shown in Figs. 1, 2, and 3, and the failure strains
and the corresponding axial stresses are listed in Table 2. The trends are as
expected.
Triaxial data were used to determine Young's modulus parameters Rf, K,
n, c, <$>0, and A<)> for the hyperbolic model by methods described by Selig
(1988). The general approach was adapted from Duncan et al. (1980). The
data plotted in the form of e/to-j — a3) versus e were fit using a least-squares
straight-line approach. These best-fit hyperbolic parameters were used in turn

92
 1600 1600 I I I
/.'•. o, = 310kPa
a Test Data t (45 psi)
- Hyperbolic Model
L1200 1200 1200 f
if , 207 (30)

6 800 800 800 f

y., 104(15)

400
a . = 207 kPa (30 psi)
400 400 -
104(15) * . 35 (5)
35(5)
n i i i
0 10 20 0 10 20 0 10 20
Axial Strain, Percent Axial Strain, Percent Axial Strain, Percent
(a) 6 1 % T-99 (b) 85% T-99 (c) 95% T-99

FIG. 1. Triaxial Test Results for SW Soil

1600 1600 1600

- a Test Data
-Hyperbolic Model
1200 1200 1200

O3 = 310kPa(45psi)
800 800 800
O3 = 207kPa(30psi) :
207,0 (30)
104(15)-
35 (5) •>. i 400 k ? . .104(15) -
400 400

0 10 20 0 10 20 0 10 20
Axial Strain, Percent Axial Strain, Percent Axial Strain, Percent
(a) 45% T-99 (b) 85% T-99 (c) 95% T-99

FIG. 2. Triaxial Test Results for ML Soil

1600 i i I 1600
» Test Data
- Hyperbolic Model
•1200 200 1200

* 800 800 800

O3=310kPa(45psi)-,
207 (30) -. I-
Og = 207kPa(30psi)-, 104(15), \ /
400 104 (15) • 400 400
35 (5) v j l i j
35(5)-

•S*T*"i i
0 10 20 0 10 20 0 10 20
Axial Strain, Percent Axial Strain, Percent Axial Strain, Percent
(a) 45% T-99 (a) 45% T-99 . (a) 45% T-99

FIG. 3. Triaxial Test Results for CL Soil

93
 TABLE 2. Trlaxial Test Results
AXIAL STRAIN AT DEVIATOR STRESS AT
FAILURE (%) FAILURE (kPa)
Confining Pressure (kPa) Confining Pressure (kPa)
Soil type (%) T-99 34 104 207 311 35 104 207 311
(D (2) (3) (4) (5) (6) (7) (8) (9) (10)
sw 95 2 2.9 2.6 3.6 255 656 1,056 1,470
85 3 5.8 15.2 8.4 118 336 651 978
61 20 20 20 35 186 386
ML 95 7.4 9.6 12.4 9.8 190 387 618 751
85 16.3 20 20 20 136 311 510 695
49 20 20 20 53 150 282
CL 95 14.3 19.5 20 188 239 326
85 20 20 20 20 158 222 299 313
45 20 20 20 56 110 215
Note: 1 psi = 6.9 kPa.

to generate arithmetic plots of (o^ -CT3)versus e that were compared to the

original data. The parameters were then adjusted, as needed, to improve the
fit to the individual arithmetic plots. In addition, the fitted curves for the
various compaction levels and soil types were compared, and then param-
eters adjusted to yield final values that would give a best fit across com-
paction levels, confining pressures, and soil types as a group.
Nine sets of parameters, shown in Table 3, were developed corresponding
to the three soils, each at three densities. Intermediate values for 80 and
90% MDD were obtained by nonlinear interpolation of the individual pa-
rameters plotted as a function of soil type and percent compaction. These
parameters were used to calculate the triaxial stress-strain curves that are
compared to the measured data in Figs. 1,2, and 3. The computed curves
generally fit the data well up to failure.
Typical isotropic compression test results are presented in Fig. 4. These
data were used to determine parameters Bt and e„ in Eq. 10 by methods
described by Selig (1988). The data plotted in the form of o-m/evol versus am
were fit using a least-squares straight-line approach. These best-fit hyper-
bolic parameters were used in turn to generate arithmetic plots of um versus
evoi that were compared to the original data. The parameters were then ad-
justed, as needed, to improve the fit to the individual arithmetic plots. In
addition, the fitted curves for the various compaction levels and soil types
were compared and then parameters adjusted to yield final values that would
give best fit across compaction levels and soil types as a group.
The data from the isotropic compression tests were also plotted in log-log
form, as shown for one test in Fig. 5, to obtain bulk modulus parameters
Kb and m for Eq. 9. The best-fit parameters represent curve A. The values
are listed in Table 4 for the three soils at the various compaction levels.
One-dimensional compression test results are shown in Fig. 6. As noted
earlier, the main reason for the one-dimensional compression tests is to pro-
vide independent checks on the validity of the hyperbolic model and the
parameters developed. This goal was accomplished by using the hyperbolic
94
 TABLE 3. Young's Modulus Parameters

Soil type T-99 (%) K n Rf c (kPa) +.0 A+O

0) (2) (3). (4) (5) (6) (7) (8)
SW 95 950 0.60 0.70 0 48 8
90 640 0.43 0.75 0 42 4
85 450 0.35 0.80 0 38 2
80 320 0.35 0.83 0 36 1
61 54 0.85 0.90 0 29 0
ML 95 440 0.40 0.95 28 34 0
90 200 0.26 0.89 24 32 0
85 110 0.25 0.85 21 30 0
80 75 0.25 0.80 17 28 0
49 16 0.95 0.55 0 23 0
CL 95 120 0.45 1.00 62 15 4
90 75 0.54 0.94 48 17 7
85 50 0.60 0.90 41 18 8
80 35 0.66 0.87 35 19 8.5
45 16 0.95 0.75 0 23 11
Note: 1 psi = 6.9 kPa.

Isotropic Stress, kPa

200 400

I
"•••».
• • • • B B B B H

50 J I I L J I I L

FIG. 4. Volumetric Strain versus Isotropic Stress

95
 3.0

en
o
• = Isotropic compression data
1.5 A = Power law model best fit to get K b and m
B = Power law model K b and m from trial and
error to fit 1-D Compression test data
C = Hyperbolic model fit

-1.0 0.0 1.5

Log0m/Pa

FIG. 5. Comparison of Bulk Modulus Models for SW Soil at 95% T-99 Density

model in a nonlinear soil-structure interaction finite element program (McVay

1982) to simulate one-dimensional compression. The soil model parameters
determined from the triaxial and isotropic compression tests were used as
input to the model, and the resulting predictions of the load-deformation
behavior of the soil were compared to the measured behavior.
The one-dimensional compression predictions using the power law with
curve-A parameters gave strains that were much larger, often an order of
magnitude larger, than the experimental values. Thus, the power-law param-
eters Kh and m were adjusted by trial and error to get a good fit to the one-
dimensional compression data. The resulting parameters are listed in Table
4 under the designation curve B. However, these values give a poor fit to

TABLE 4. Bulk Modulus Parameters from Tests

POWER LAW MODEL

T-99 Curve A Curve B a Hyperbolic Model

Soil type (%) Kb m Kb m Bt/Pa «K

(D (2) (3) (4) (5) (6) (7) (8)

SW 95 250 0.80 250 0.0 66.8 0.02
85 90 1.02 130 0.10 12.7 0.06
61 35 1.59 35 0.82 1.3 0.19
ML 95 110 0.60 185 0.00 48.3 0.06
85 35 0.49 45 0.25 17.9 0.24
49 15 1.40 15 0.94 1.3 0.38
CL 95 50 0.60 80 0.20 22.1 0.13
85 25 1.05 30 0.55 5.2 0.20
45 15 1.77 15 1.02 0.4 0.49
'Trial-and-ei Tor approach.

96
 Vertical Stress, kPa
„0 100 200 300 400 500 600 700

^^
10
^ ~ ^ ) 8 5 % T-99

. 20 3 Dumped -

in -Y\
-
i-^4 SW
H—H ML
o 8—9 CL
| 30 - *t

40

50 i I I 1 1 1

FIG. 6. One-Dimension Compression Test Results

Vertical Stress, kPa

200 400 600

FIG. 7. One-Dimension Compression Predictions for SW Soil at 85% T-99 Den-

sity

97
 Vertical Stress, kPa

400 600 80
ocQ*~-~- 200
i i i
o T&E Power Law
Q Hyperbolic Model
X Best Fit Power Law
e Test Data

s?
Vertical Strain,
o

-
\ . o

- -
?n

FIG. 8. One-Dimension Compression Predictions for ML Soil at 85% T-99 Den-

sity

the isotropic compression data (Fig. 5).

Example finite element simulations of one-dimensional compression using
trial-and-error power-law bulk modulus, best-fit power-law bulk modulus,
and hyperbolic bulk modulus parameters together with the hyperbolic Young's
modulus parameters are shown in Figs. 7, 8, and 9 for comparison with the
test data. Note that neither the hyperbolic bulk modulus nor the hyperbolic

Vertical Stress, kPa

400 800

FIG. 9. One-Dimension Compression Predictions for CL Soii at 85% T-99 Density

98
 800

sw
ML , 95% T-99
CL
600

/ , 85% T-99
/ / ' . 95% T-99

/ ,,-'> y 85% T-99

r.S/ , / . . . Dumped

d? ^ ..-'' y Dumped

200

0 200 400 600 800

Normal Stress, kPa

FIG. 10. Failure Envelopes for the Various Soil and Compaction Conditions

Young's modulus parameter determinations involved the use of trial-and-

error procedures to fit the one-dimensional compression curves.

MODEL TRENDS

Figs. 10-13 demonstrate the consistency of the soil parameters by show-

ing trends of predicted behavior as a function of soil type and compaction
level. The patterns of systematic differences that emerge are reasonable and
consistent.
Failure envelopes for the various conditions examined are presented in
Fig. 10. The need for nonzero values of c, §0, and A§ for the CL soil is
apparent from the curved envelopes with intercepts for this soil. Fig. 11(a)
presents shear strength as a function of compaction level and soil type for
a normal stress of 345 kPa (50 psi). Fig. 11(b) shows secant Young's mod-
ulus at 50% of the failure stress, for a confining stress of 104 kPa (15 psi),
as a function of compaction level and soil type. Secant bulk modulus, at a
mean stress of 207 kPa (30 psi), and secant constrained modulus (slope of
axial stress versus axial strain in a one-dimensional compression test), at a
vertical stress of 345 kPa (50 psi), as functions of compaction level and soil
type are presented in Figs. 12 and 13, respectively. Figs. 10-13 illustrate
the substantially different strengths and stiffnesses that the same compaction
level produces in the various soil types.
The last item to be discussed is the issue of the modeling of the bulk
modulus according to the approaches noted earlier. Figs. 7 , 8 , and 9 indicate
that simulations based on the hyperbolic bulk' modulus parameters fit the
one-dimensional compression test data better than simulations based on bulk
modulus parameters developed using a power-law modulus approach. The
99
 I
Normal Stress Normal Stress
300 h
<3„= 345 kPa <V = 207 kPa
100

SW • /
200 ML
CL 7
/

i
/
50

£ 1001-
c/>

J_
"0 50 100 0 50 100

Percent T-99 Dry Density Percent T-99 Dry Density

(a) Shear Strength (b) Secant Modulus

FIG. 11. Shear Strength and Young's Modulus as Function of Compaction Level
and Soil Type

power-law approach appears to underestimate the bulk modulus greatly at

low confining stress and thus overestimate deformation dramatically. A trial-
and-error approach can be used to yield power-law bulk modulus parameters
that do model one-dimensional compression relatively well. However, this
approach requires judgment on the part of the individual selecting the trial
parameters because the trial-and-error approach does not lead to a unique
set of Kb and m.
An example of a bulk modulus versus confining stress curve based on the
trial-and-error parameters presented in Table 4 is represented by Curve B in
Fig. 5. In this figure, it is apparent that the trial-and-error Kb and m define
a load-deformation curve rather different than the measured bulk modulus
data, even though the trial-and-error parameters will fit the one-dimensional

I I I I I I I I

Mean Isotropic Stress "m = 207 kPa

ASW
BML
• CL

£ 10 h

I I I I •~fz I I I I

Percent T-99 Dry Density

FIG. 12. Secant Bulk Modulus as Function of Compaction Level and Soil Type

100
 At Vertical Stress c v =345 kPa

ASW
B ML
9CL

0 50 100
Percent T-99 Dry Density

FIG. 13. Secant Constrained Modulus as Function of Compaction Level and Soil
Type

compression curves well. Thus, the trial-and-error power-law parameters yield

simulations that match the one-dimensional test results very well, but involve
compensating errors that cannot be relied upon to compensate in a similar
fashion to render correct solutions for other boundary conditions and stress
ranges. It should also be noted that the hyperbolic approach to modeling the
bulk modulus-confining stress relationship relates the bulk modulus to the
mean confining stress, whereas the power law approach as typically used
relates the bulk modulus solely to the minor principal stress. The former is
believed to be more realistic than the latter.

SUMMARY AND CONCLUSIONS

The primary goal of this study is to develop soil stress-strain parameters

for general classes of compacted soils for design involving finite element
analysis of soil-structure interaction problems. Typically, the particular soil
to be used as backfill is not known until construction, and even then nec-
essary laboratory tests for properties will generally not be performed. As a
consequence, designers are faced with the task of estimating appropriate soil
parameters based on soil classifications and specified levels of compaction.
To this end, three soils (SW, ML, and CL) were investigated and the hy-
perbolic parameters for these soils at various compacted states were deter-
mined. The suggested Young's modulus parameters are presented in Table
3 and the suggested bulk modulus parameters are presented in Table 5. These
suggested parameters were based on best-fit curves through the data that
were adjusted to give consistent behavior across soil types and compaction
levels.
As part of the study, a method for determining bulk modulus parameters
for partially saturated soils was developed. Volume change data are required
for determination of bulk modulus. However, compacted soils are generally
unsaturated and volume change data are typically not available from triaxial
tests on unsaturated soils. Hence, the method proposed by Duncan (1980)
cannot be used. An alternative method employs bulk modulus data deter-

101
 TABLE 5. Suggested Bulk Modulus Parameters
Hyperbolic Model
Soil type T-99 (%) Bt/Pa £u

(1) (2) (3) (4)

sw 95 74.8 0.02
90 40.8 0.05
85 12.7 0.08
80 6.1 0.11
6,1 1.7 0.23
ML 95 48.3 0.06
90 18.4 0.10
85 9.5 0.14
80 5.1 0.19
49 1.3 0.43
CL 95 21.2 0.13
90 10.2 0.17
85 5.2 0.21
80 3.5 0.25
45 0.7 0.55

mined from isotropic compression tests where the cell fluid flowing into or
out of the cell is monitored and, after correction for cell and system expan-
sion, is used to determine the volume change of the partially saturated spec-
imens in response to pressure changes. The nonlinear bulk modulus versus
confining pressure data is then fitted to a hyperbolic relationship for use in
the general model. For the soils examined, this approach gave better agree-
ment with hydrostatic and one-dimensional compression test data than the
power law approach.
An important condition examined in this study that has not been addressed
in the literature reviewed by the writers was the case of soils with little, if
any, compaction. Such conditions may be encountered in practice in hard-
to-reach zones, such as under pipe haunches. Their effect on soil-structure
interaction and structure performance can become important questions for
evaluation.
The study is limited to three "typical soils," each tested only at its re-
spective standard proctor optimum water content. However, these soils do
cover the range from those considered among the best to those considered
among the poorest of materials commonly used as backfill. The parameters
developed give behavior that is consistent across soil types and compaction
levels. Thus, the data presented herein should provide a basis for reasonable
estimates of behavior of a variety of compacted soils.
Further work is needed to investigate the effect on the model parameters
of the following:

1. Variation of compaction water content from standard-proctor optimum.

2. Change in water content after compaction.
3. Variation in soil composition within the same classification group.
4. Shear-induced volume change as exhibited in the triaxial test.
102
ACKNOWLEDGMENTS

Support for this work was provided in part by the American Concrete Pipe
Association. Atef Haggag, University of Massachusetts graduate student, as-
sisted in the interpolation to get the soil parameters, and Atis Liepins of
Simpson, Gumpertz, and Heger, Inc., participated in the final selection of
recommended values.

APPENDIX I. REFERENCES

Domaschuk, L. (1969). "A study of bulk and shear moduli of a sand." / . Soil Mech.
and Found. Engrg. Div., ASCE, 95(2), 561-581.
Domaschuk, L., and Valliappan, P. (1975). "Nonlinear settlement analysis by finite
element." / . Geotech. Div., ASCE, 101(7), 601-614.
Duncan, J. M., and Chang, C. Y. (1970). "Nonlinear analysis of stress and strain
in soils." J. Soil Mech. and Found. Engrg., ASCE, 96(5), 1629-1653.
Duncan, J. M., et al. (1980). "Strength, stress-strain, and bulk modulus parameters
for finite element analyses of stresses and movements in soil masses." Report No.
UCB/GT/80-01, Univ. of California, Coll. of Engrg., Berkeley, Calif.
Janbu, N. (1963). "Soil compressibility as determined by oedometer and triaxial tests."
Proc. European Conf. on Soil Mech. and Found. Engrg., Wiesbaden, Germany,
1, 19-25.
Jennings, P. C. (1964). "Periodic response of a general yielding structure." J. Engrg.
Mech. Div., ASCE, 90(2), 131-166.
Kondner, R. L. (1963). "Hyperbolic stress-strain response: Cohesive soils." J. Soil
Mech. and Found. Engrg. Div., ASCE, 89(1), 115-143.
Kondner, R. L., and Zelasko, J. S. (1963). "A hyperbolic stress-strain formulation
of sands." Proc. 2nd Pan-Am Conf. on Soil Mech. and Found. Engrg., Brazil,
1, 289.
Ladd, R. S. (1978). "Preparing test specimens using undercompaction." Geotech.
Testing J., 1(1), 16-23.
Lin, R-S. (1987). "Direct determination of bulk modulus of partially saturated soils."
Masters Project Report No. ACP87-341P, Univ. of Massachusetts, Amherst, Mass.
McVay, M. C. (1982). "Evaluation of numerical modeling of buried conduits," dis-
sertation presented to the University of Massachusetts, at Amherst, Mass., in par-
tial fulfillment of the requirements for the degree of Doctor of Philosophy.
Selig, E. T. (1988). "Soil parameters for design of buried pipelines." Proc. Pipeline
Infrastructure Conf, ASCE, 99-116.
Yang, G. W. (1987). "Hyperbolic Young's modulus parameters for compacted soils,"
Master's Project Report No. ACP87-342P, Univ. of Massachusetts, Amherst, Mass.

APPENDIX II. NOTATION

The following symbols are used in this paper:

B = bulk modulus;
B, — initial tangent bulk modulus;
B, = tangent bulk modulus;
c = cohesion intercept;
E = Young's modulus;
E> = initial tangent Young's modulus;
E, = tangent Young's modulus;
K = modulus number;
Kb = bulk modulus number;
m = bulk modulus exponent;
103
 n = modulus exponent;
Pa = atmospheric pressure;
Rf = failure ratio;
Ae v o l = volumetric strain increment;
Ao-,„ = mean stress increment;
A4> = reduction in <>j per log cycle change in <x3;
e = normal strain;
e„ = asymptotic value of volumetric strain at large stresses;
e
vol = volumetric strain;
«"l = major principal stress;
0-3 = minor principal stress;
-e-

= angle of internal friction; and

$ 0 = 4> whenCT3= pa.

Subscripts
/ = failure;
i = initial tangent;
t = tangent; and
ult = ultimate or asymptotic value.

104