INTERNATIONAL JOURNAL FOR NUMERICAL AND ANALYTICAL METHODS IN GEOMECHANICS. VOL. 5 .

295-311 (1981)

PROCEDURE, SELECTION AND APPLICATION

OF PLASTICITY MODELS FOR A SOIL
C. S. DESAI.'
Department of Civil Engineering,
Virginia Polytechnic Institute and State University,
Blacksburg, Virginia, U.S.A.

H.v. PHAN?
Weidlinger Associates, Menlo Park, C A , U.S.A.

AND
s. STURES
Department of Civil and Environmental Engineering, University of Colorado, Boulder, CO, U.S.A.

SUMMARY
A five-step procedure involving mathematical formulation, identification and determination of parameters
and verification is presented for development and selection of appropriate and reliable constitutive law(s)
for geologic media. Comprehensive analyses are performed toward determination of an appropriate law
for a (artificial) soil. The most suitable model is obtained by critical evaluation of four different plasticity
models; here verification and comparisons of predictions with observations from laboratory tests, and with
those from two boundary value problems are used as the basis of the selection. The model thus selected is
found to be appropriate for applications to relevant practical problems.

INTRODUCTION
Recognition of the importance of constitutive modelling for geologic media has spurred
significant recent activity towards development and implementation of advanced constitutive
laws. This has brought a growing awareness of the fundamental theory of constitutive laws and
the theoretical aspects needed to establish a basis for handling complex factors that were not
included in the simplified models of the past.
The theory of plasticity with new developments to account for special features of geologic
media such as volume changes under shear has been one of the major recent directions towards
development of models for soils and rocks. In plasticity, yield criteria such as von Mises,
Mohr-Coulomb and Drucker-Prager have been considered as well as recent modifications such
as critical state and cap models. In view of the development and availability of a large number of
models, it becomes necessary to compare and evaluate them so as to select an appropriate model
for a given medium.
The objectives of this paper are to present a procedure for evolving reliable constitutive
model(s) for geological materials, and for selecting appropriate model(s) for a given class of
problems. A number of plasticity models such as Drucker-Prager, critical state and cap models

* Professor.
t Research Engineer.
$ Assistant Professor.

0363-9061/81/030295-17$01.70 Received 25 February 1980

@ 1981 by John Wiley & Sons, Ltd. Revised 10 May 1980
296 C . S . DESAI. H. V. PHAN AND S. STURE

are considered for a (artificial) soil, and the proposed procedure is applied in order to select the
most appropriate model; the selected model is intended for application to relevant problems for
which the evaluation is performed.

PROCEDURE
Development of a viable constitutive law for a given medium is considered to involve five steps:4
1. mathematical formulation;
2. identification of significant parameters;
3. determination of parameters from appropriate tests;
4. verification with respect to laboratory test data under various stress paths and physical
conditions; and
5 . verification and evaluation with respect to boundary value problems.
Since the theory and mathematical basis and the required parameters for the Drucker-
Prager,6 critical state. and cap model. are well known, steps 1 and 2 are stated only briefly
here, whereas steps 3 to 5 are given more detailed consideration.

Soil
The soil considered is an artificial soil made of sand, clay and oil; it has been used in a number
of current research project^.^.'.^.'^ The main aim in using oil as a binder has been to eliminate
significantly the influence of moisture changes during laboratory tests with boundary value
problems. The soil consists of 50% Florida Zircon sand and 50% Fire clay; 10% of No. 5 SAE
mineral oil is added to the mixture. Since the saturation of soil with oil was quite low, the
laboratory tests for constitutive laws were assumed to be fully drained. The artificial soil was
highly compressible and exhibited small amounts of cohesion. Figure 1 shows grain-size
distribution for the soil; the maximum and minimum densities were found to be 2.65 gm/cm3
and 1.OO g/cm3, respectively.

THEORETICAL BACKGROUND AND STEPS

Steps 1 and 2
As stated earlier, details relevant to steps 1 and 2 for the plasticity models are available in

various r e f e r e n c e ~ ; , ~ , ~for
* ~the
* sake of completeness, brief details are presented below.
Drucker-Prager model. The Drucker-Prager failure surface, ff, is given by
fi = J J 2 D - a J 1- K = O (1)
where J 1 = u1+ u2 + u3is the first invariant of the stress tensor,
J -1
2D - 6{(c1 -r 2 ) + ((+2 - c 3 ) + (Pl- a3)2)12

is the second invariant of the deviatoric stress tensor and (Y and k are the material parameters,
and u ~u 2,. u3 are the principal stresses. The expressions for (Y and k for plane strain idealization
are given below:
 PLASTICITY MODELS FOR A SOIL 297

Gravel Sand Fines

Coarse to Fine Silt Clay

medium
I

*I S
g p
R
U.S. standard sieve sizes
sI
p
8 8
I

tJ
g P

Grain diameter, mm

Figure 1. Particle size distribution for artificial soil

where c =cohesive strength and 4 =angle of internal friction. Behaviour prior to yield is
assumed to be elastic.
Critical state model. The modified cam clay9 model was used. The yield surface according to
this model is given by (Figure 2(a))

f c = e2 - p o e + - ?M=qO2 (3)

where p = (u, + ~2 + ( ~ 3 ) / 3 is the mean pressure, po = initial mean pressure, M = slope of the
critical state line and q = deviatoric stress (v,- u3). The hardening rule is defined as a function of
the plastic volumetric strain, E :, as
298 C . S . DESAI, H. V. PHAN AND S. STURE

Figure 2. (a) Critical state model; (b) e-log p relation from consolidation test

where A and K = compression and swelling indices, respectively, Figure 2(b), and eo = initial
void ratio. Behaviour prior to yield is assumed to be elastic.
C a p model. Figure 3 shows schematic representation of the concept of the model for a sand
proposed by DiMaggio and Sandler.' The failure surface, ff, and the yield or cap surface, f c , are
given by
ff = J:S - A + C e-BJ1 =0 (5)
and

where JzD is the second invariant of the deviatoric stress tensor, A, B and C are material
parameters found from conventional triaxial tests and p e and p i are defined in Figure 3(a). The
plastic volumetric strain is expressed as a function of the hydrostatic pressure as
E$ = W(eDP'- 1) (7)
where W and D are material parameters obtained from hydrostatic compression tests.

Step 3: determination of constitutive parameters

The material parameters for the Drucker-Prager, critical state and cap models can be
determined from conventional triaxial, consolidation and hydrostatic tests. However, for
 PLASTICITY MODELS FOR A SOIL 299

Sandler

Figure 3. Cap model and modified cap model: (a) cap models; (b)plastic volumetric strain us mean pressure relation

general and more precise characterization, it is desirable to simulate all possible stress paths and
loading conditions. Then it becomes necessary to use advanced equipment such as a truly triaxial
device.
A comprehensive series of tests were performed for the artificial soil by using the truly triaxial
(IT)or multiaxial device. This device permits tests with cubical samples 4 X 4 X 4 inches
(10 x 10 x 10 cm) and under a wide range of stress paths, Figure 4. The stress paths under which
tests were performed were: hydrostatic compression (HC), conventional triaxial compression
(CTC), conventional triaxial extension (CIE), reduced triaxial extension (RTE), triaxial
compression (TC), triaxial extension (TE) and simple shear (SS). In addition, the samples were
tested under three initial densities of about 1.70,1-80and 2.00 g/cm3. Two of the typical test
data under CTE and SS stress paths are shown in Figures 5 and 6.
Details of the foregoing tests have been reported p r e ~ i o u s l y ; ~ ~Reference
~-~ 13 describes
the procedure used to evolve the cap model for the soil. In this paper, the aim is to compare and
evaluate a number of models and then to adopt an appropriate model for implementation and
application to boundary value problems.
Figure 3(a) shows the modified cap model found suitable for the soil. For the modified cap
model, the failure surface was found to be
ffm = J i g - A - /3J1+ C eBl = 0 (8)
3 00 C.S. DESAI, H. V. PHAN AND S. STURE

'53

Figure 4. Stress paths simulated in truly triaxial device: (a) stress paths in octahedral plane; (b) stress paths in triaxial
plane

whereas the yield surfaces and the volumetric strain-mean pressure relations were found to be
the same as those in equations (6) and (7) respectively.
It may be noted that an additional term &TI in equation (8)was needed to characterize the
observed behaviour of this soil compared with the original cap model, equation ( 5 ) .The physical
meaning of this term can be explained from the fact the failure surface, equation (8), consists of
an initial section of Drucker-Prager surface connected through a transition curve to the

10 o3 = 10 psi
Yo = 1.96 g/cni

1 2 3 4 5 6 7 8 9 10
extension compress ion L ( %)
Figure 5. Typical stress-strain test data under CTE stress path
 PLASTICITY MODELS FOR A SOIL 301

15 10 5 0 5 10
extension c(%) compression

Figure 6. Typical stress-strain test data under SS stress path

subsequent Drucker-Prager surface at higher loading. This indicates that at higher pressure, the
behaviour of this soil is influenced by mean pressure. On the other hand, the original cap model,5
equation ( 5 ) , for a sand involved an initial Drucker-Prager curve connected to the subsequent
von Mises curve.

Values of the parameters

Based on the comprehensive series of tests with the 'IT device, the following parameters were
obtained for the Drucker-Prager, critical state, cap and modified cap models:
E = 4,000 psi (27,600 kPa); v = 0.35;
c = 0.00; C#J = 35.00 deg;
A =o.ll; ~ = 0 * 0 0 1 ;
A = 5.60 psi (38.64 kPa); p = 0-11;
C = 5.60 psi (38.64 kPa); B = 0.062 psi-' (0.009 kPa-');
R = 2.00; W = 0.18; D = 0-05 psi-' (0.0072 kPa-I).
Here R = ratio of the semi-major and semi-minor axes of the yield ellipse, fc, or cap, Figure 3(a).
For the boundary value problem in verification 2, considered subsequently, a value of the
ratio of initial horizontal to vertical stress, K , equal to 1 was used. This value was based on an
approximate analysis of stresses in the soil with specific number of roller-passes used to compact
the ~ 0 i l . l ~
302 C . S . DESAI, H. V. PHAN A N D S. STURE

Step 4: verification: laboratory tests

This step can be performed, (a) by comparing predictions from a constitutive model with
observed stress-strain data and/or, (b) by comparing predictions from a solution procedure
incorporating the model with observed stress-strain test data. A typical result for the first
approach is shown in Figure 7 which compares observed failure envelope with that predicted
from the modified cap model, equation (8).
Figure 8 shows a finite element layout for a single cubical element 4 x 4 x 4 inch (10.16 X
10.16 x 10.16 cm) which simulated the cubical sample in the truly triaxial device. By taking
advantage of the symmetry, a one-element mesh involving a quarter of the cube was used. Then
various loading (stress) paths were simulated in the three-dimensional finite element procedure;
a brief description of the procedure is given below.

0-4 Observed

75 &--A Predicted

OY I I

0 50 100 150 200 250 300 350 400

J1 (@a)

Figure 7. Comparisons of predictions of modified cap model with observations

Figure 8. One-element mesh for cubical specimen

 PLASTICITY MODELS FOR A SOIL 303

Finite element procedure. The foregoing models were incorporated in a numerical solution
procedure based on a three-dimensional finite element idealization intended for general
soil-structure interaction. The computer code includes provision for material and geometric
non-linearities, and the non-linear effects caused by interfaces and cracking, and uses variable
node, 8-to 21 -node, hexahedral isoparametric
Figures 9(a) to (d) show comparisons between finite element predictions and observations
from four typical stress paths: HC, CTC, CTE and SS. It can be seen that the agreements for the
H C and CTC stress paths are excellent, whereas those for the CTE and SS stress paths are not
that satisfactory. It may be mentioned that at this time, no constitutive law would be capable of
predicting all stress paths with equal degree of agreement. It is hoped that future and continuing
research would produce improved and general models.

Step 5: verification: boundary value problems

The foregoing finite element procedure was used to predict observed behaviour of two
boundary value problems simulated in laboratory tests.

a - 0, (psi)
A 3

(bl

Figure 9. Comparisons of finite element predictions with stress-strain data for various stress paths: (a) hydrostatic
compression test; (b) conventional triaxial compression test; (c) conventional triaxial extension test; (d) simple shear test
3 04 C . S . DESAI. H. V. PHAN AND S. STURE

c3(%) -3. -2. - 1 . 1. 2. 3. 4. E1"L2(%)

Extension Compression

c3(% j
-6. -4. -2. 2. 4. 6. 8. L,(%)
Id)

Figure 9. Continued

Verification 1: model strip footing on artificial soil

The laboratory model consisted of a rectangular box, Figure 10(a), 34.5 inches (90-00cm)
long, about 12.00 inches (30.50 cm) high and 4.50 inches (I 1.50 cm) thick. The box was filled
with the artificial soil up to a height of 8.00 inches (20.00 cm). A footing, made of plastic, 3.00
inches (7.60 cm) wide and 0.75 inches (1-90cm) thick and 4.5 inches (11.50 cm) long, was
placed at the centre of the box. Vertical incremental loads were applied on the footing, and the
vertical displacements of the footing were measured. The displacement patterns in the soil at the
inside vertical face of the box were also obtained by taking photographs after different load
increments. For this purpose, a grid was designed between the inside face of the wall and the soil.
Figure 10(b)shows the finite element mesh for the half of the footing-soil system. The vertical
boundaries and the centre line were assumed to be smooth, allowing only vertical movements,
while the boom boundary was assumed to be rough. The problem was solved by using the
three-dimensional procedure by assuming that the displacements perpendicular to the vertical
side of the box were zero; this leads to the plane strain idealization. It may be noted that this
 PLASTICITY MODELS FOR A SOIL 305

Footing

4.5

P l a s t i c Box

34.5 inch
(0)

Smooth Boundary

(b) \ Rough Boundary

Figure 10. Details of laboratory test set-up and finite element mesh: (a) dimensions of test box; (b) finite element mesh

problem can be solved by using a two-dimensional plane strain procedure. However, since the
three-dimensional code was available and required for the other aspects (see next example) of
the it was adopted herein.
Figure 11 shows comparisons between the predictions from the Drucker-Prager, critical
state, cap model, modified cap model and the experimental behaviour; the theoretical limit load

Theoretical Lim. 14.7Opsi

Vertical Disp1acernent.w ( i n c h )
- : experimental
0-0 : Drucker-Prager Model
t-+ : C r i t i c a l S t a t e Model
0- o : Cap Model, DiMaggio e t a1
8-6 : Cap Model, present a n a l y s i s

Figure 11. Comparison between predictions and observations

306 C. S. DESAI. H. V. PHAN A N D S. STUR2

of 14.70psi (101 kN/mZ) is based on limit equilibrium consideration. The Drucker-Prager

model yields poor correlation. The critical state model shows relatively better correlation.
However, the cap models show excellent correlation with the experimental data. The cap
models were found to be capable of predicting the hardening behaviour at higher loads whereas
the other two models indicate a limiting condition.
The artificial soil exhibited significant compressibility and volume change under shear and
hardening behaviour at higher loads. The comparisons in Figure 11 show that the cap model is
capable of accounting for these aspects of the soil; one of the reasons for this can be that the cap
model includes provision for the volume change characteristic exhibited by the soil.

-SCALES
_
Coordinate : 1 inch = 2.875 inch
Displacement: 1 inch = 0.266 inch
F

(b)
Figure 12. Computed displacement patterns: (a) Drucker-Prager model, p = 16.0 psi; (b) modified cap model,
p = 15.0 psi
 PLASTICITY MODELS FOR A SOIL 3 07

Figure 13. Observed displacement patterns at three loading steps

308 C. S. DESAI, H. V. PHAN AND S. STURE

Although both the original and the modified cap models yield similar results in this case, the
latter was found to yield better overall predictions for this and the subsequent
Hence, the modified cap model is considered to be most suitable for this soil.
A further evidence for the foregoing conclusion is shown in Figures 12 and 13. Figure 12
shows computed displacements around ultimate loads from the Drucker-Prager and the
modified cap models. The Drucker-Prager model indicates almost uniform movements around
and below the footing and in the regions away from it. It shows significant amount of heave away
from the footing. On the other hand, the cap model shows larger movements in the vicinity and
below the footing with very little movements in the regions away from it. The computed
behaviour is similar to that observed in the experiment, Figure 13. These figures show
photographic records of the displacement patterns at three typical loading stages. It can be seen
that the displacements are essentially concentrated below the footing and its vicinity, and the
regions away experienced very small movements. This indicates that the footing essentially
punches into the soil, and the deformations are absorbed in the soil mass through significant
volumetric changes of the compressible soil.

Smooth
Boundary

Figure 14. Finite element mesh for structure pushed against vertical soil bank: (a) finite element mesh; (b) interface
element
 PLASTICITY MODELS FOR A SOIL 309

Verification 2: structure pushed against a vertical soil bank

Figure 14 shows the finite element mesh for a laboratory test in which a short structure, 6.00
inch (15.25 cm) long and 2 x 2 inches (5.08 x 5.08 cm) cross section was loaded by applying a
lateral load and was pushed against a vertical face bank made of the artificial soil. In order to
allow for slip between the structure and the soil a three-dimensional interface element was
provided between the tool and the soil mass. The test was performed in a soil-bin about 30 f t
(10 m) long, 3.3 ft (1 m) wide and 1.2 ft (0.37 m) d e e ~ .The
~ . geometry
~ of the problem requires
a fully three-dimensional analysis.
The finite element discretization involved zones extending up to 6 to 8 times the width of the
tool in the x, y and z directions. Horizontal displacements were assumed to be zero on the end
vertical boundaries, and the vertical displacements were assumed to be zero at the bottom
discretized boundary.
Figure 15 shows comparisons between the experimental load-displacements curve and that
predicted by using the modified cap model. In view of the complexity of this problem, the
correlation is considered to be highly satisfactory. It demonstrates that for the artificial soil the
modified cap model is appropriate.

CONCLUSIONS
Development and implementation of appropriate constitutive law(s) for geologic media require
significant care and analysis. A reliable constitutive model needs attention to various steps such

D i r e c t i o n o f Movement

Tool 1 . ' '

t
100.0 -
80.0 -
60.0 -
-2 40.0 -
v
7

I I I I I I I c
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0 .7

H o r i z o n t a l Displacement, u ( i n c h )

Figure 15. Comparisons between predictions and observations

310 C. S. DESAI. H. V. PHAN AND S. STURE

as mathematical details, identification of parameters, determination of parameters and satis-

factory verifications with respect to solution of relevant boundary value problems.
A comprehensive analysis and evaluation are presented for a (artificial) geologic medium in
order to select the most suitable model from among three available plasticity laws. After
undergoing the required steps, it was found that a modified cap model was suitable for the soil.
Based on this evidence, the model can be used for practical applications involving such a
geologic medium.

ACKNOWLEDGEMENT

The research investigations presented herein were supported by a grant from the National
Science Foundation. Assistance of J. V. Perumpral is gratefully acknowledged.

NOTATION
A = parameter in cap model
B = parameter in cap model
C = parameter in cap model
c = cohesion
D = parameter in cap model
E = elastic modulus
e =void ratio
f c = cap or yield surface
ff = failure surface
f f m = modified failure surface
J1= first invariant of stress tensor
JzD =second invariant of deviatoric stress tensor
k = parameter in Drucker-Prager model
M = slope of critical state line
P = applied load
p = pressure
4 =shear stress or stress difference
R = ratio of semi-major to semi-minor axes of yield ellipse
W = parameter in cap model
w = vertical displacement
a = parameter in Drucker-Prager model
00 = parameter in cap model
E, = E~~ = volumetric strain
A = slope of consolidation curve
K = slope of swelling curve
Y = Poissons ratio
4 = angle of friction

REFERENCES
1 . J. T. Christian and C. S. Desai, Constitutivelaws of geologic media, Chapt. 2 in NumericalMethods in Geofechnical
Engineering, Eds. C. S. Desai and J. T. Christian (McGraw-Hill., New York, 1977.
2 . C. S . Desai, Some aspects of constitutive models for geologic media, Proc. 3rd Inf. Conf. on Num. Mefh. in
Geomech.. Aachen, W. Germany (1979).
 PLASTICITY MODELS FOR A SOIL 311

3. C. S . Desai, H. V. Phan and J. V. Perumpral, Mechanics of three-dimensional soil-structure interaction: theory (I)
and verification (II), submitted to J. Engn. Mech. D i n , Proc. A m . SOC.Giu. Eng.
4. C. S. Desai and H. J. Siriwardane, Constitutiue Laws for Engineering Media, Prentice-Hall (Under publication).
5. F. L. DiMaggio and I. S. Sandler, Material model for granular soil, J. Engng Mech. Div., Proc. A m . SOC.Cir. Eng.
197, No. 3 (1971).
6. D. C. Drucker and W. Prager, Soil mechanics and plastic analysis of limit design, Quart. Appl. Math. 10, No. 2
(1952).
7 . J . Mould, Constitutive characterization of granular materials at low effective stress levels, M.S. thesis, Dept of Civil
Engng, Va Tech, Blacksburg, Va. (1979).
8. H. V. Phan, C. S. Desai, S . Sture, and J. V. Perumpral, Three-dimensional geometric and material nonlinear
analysis of some problems in geomechanics, Proc. 3rd Znr. Conf.on Num. Meth. in Geomech. Aachen, W. Germany,
1979.
9. K. H. Roscoe and J. B. Burland, On the generalized stress-strain behaviour of wet clay, in Engineering Plasticity,
eds. Heyman and Leckie, Cambridge Univ. Press, Endland, 1968.
10. I. S., Sandler, F. L. DiMaggio and G. Y.Baladi, Generalized cap model for geologic Materials, J. Geotech. Engng
Div. Proc. A m . SOC.Ciu. Eng. 102, No. 2 (1976).
11. A. N. Schofield, and C. P. Wroth, Crticial State Soil Mechanics, McGraw-Hill, London, 1968.
12. S. Sture and C. S.Desai, Fluid cushion truly triaxial or multiaxial testingdevice, J. Geotech. Testing, ASTM, 2, No.
l(1979).
13. S . Sture, C. S. Desai and R. Janardhanam, Development of a constitutive law for an artificial soil, Proc. 3rd Int.
Conf. on Num. Merh. in Geomech. Aachen, W. Germany (1979).
14. S . Sture, C. S. Desai and J. V. Perumpral, Laboratory testingfor artificial soil, Report to NSF, Dept of Civil Engng.,
Va Tech., Blacksburg, V.A. (1980).