Plastic Yield Lode
Plastic Yield Lode
Plastic Yield Lode
J. P. Bardet
Department of Civil Engineering, University of Southern California, Los Angeles, CA 90089-0242
Experimental investigations indicate that the third stress invariant; Lode angle cy affects significantly the behavior of pressure sensitive materials. The present communication presents a formulation to account for cy in isotropic pressure-sensitive elastoplastic materials. Seven Lode dependences are reviewed. A new one, referred to as LMN, in proposed to generalize Lade and Duncan, and Matsuoka and Nakai failure surfaces. The formulation is general enough to introduce CY into the isotropic elastoplastic modes which are only developed in terms of first and second-stress invariants. As an illustration, several Lode dependences are introduced into Roscoe and Burland model. The performance of the modQiied model is estimated by comparing experimental and analytical results in the case of true triaxial loadings on normally consolidated clay.
Introduction pendences found in the literature and by a new one referred to as LMN dependence. LMN dependence is shown to describe the failure surfaces of Lade and Duncan (1975) and Matsuoka and Nakai (1974). The last section applies the general formulation to introduce (Y into the yield and plastic potential functions of the Roscoe and Burland (1968) model. The performance of the modified model is estimated by comparing experimental and analytical results in the case of true triaxial loadings on normally consolidated clay.
Many conventional plasticity models for pressure-sensitive materials, such as Roscoe and Burland (1968), have been formulated in terms of the first and second-stress invariants without considering the third invariant, sometimes referred to as a Lode angle o(. This simplifying assumption disregards the experimental observations made during true axial testing which indicate that the third stress invariant affects significantly the behavior of pressure sensitive soils, rocks, and concretes (Chen and Saleeb, 1982; Lade and Duncan, 1975; Matsuoka and Nakai, 1974). The objective of the present communication is to propose a formulation which accounts for 01 in isotropic pressure-sensitive elastoplastic materials. The present work reviews, compares, and generalizes the past work of Argyris et al. (1974), Dafalias and Herrmann (1986), Eekelen (1980), Lade (1975), Matsuoka and Nakai (1974), William and Warnke (1975) and Zienkiewicz and Pande (1977). The formulation aims at introducing the effects of cyinto the isotropic elastoplastic models which are developed in terms of only first and second-stress invariants. After reviewing the fundamental definitions of isotropic plasticity, the contribution of 01on the direction and amplitude of incremental plastic strain is isolated from other plastic contributions. A particular class of yield and plastic potential functions is introduced for pressure-sensitive materiaIs. This class of functions is illustrated by seven examples of a-de-
Elastoplastic
Constitutive
Equations
The sign convention of soil mechanics is preferred to the one of solid mechanics in order to eliminate redundant minus signs for pressure-sensitive materials that experience mostly compressive stress and strain. Hereafter, stresses and strains are assumed to be positive in compression and negative in tension. 2.1 Stress-Strain Relationships. According to one of the fundamental assumptions of the flow theory of plasticity, the strain increment is the sum of elastic and plastic strain increments:
(1)
Contributed by the Applied Mechanics Division of THE AMERICAN SOCIETY OF MECHANICAL ENGINEERS for publication in the JOURNAL OF APPLIED MECHANICS. Discussion on this paper should be addressed to the Technical Editor, Leon M. Keer, the Technological Institute, Northwestern University, Evanston, IL 60208, and will be accepted until two months after final publication of the paper itself in the JOURNALOF APPLIED MECHANICS. Manuscript received by the ASME Applied Mechanics Division, Mar. 14, 1988; final revision, Oct. 25, 1988.
d=2G-3B 18~~
t?l
hrjdakk f
Z& da,,
(2)
where the shear modulus G and the bulk modulus B may be stress or strain dependent. Adopting the conventional formulation of the flow theory of plasticity, the plastic strain increment is: Transactions of the ASME
1990
d+= <i
where the symbol any scalar 2:
n,,dakk>mij
the following if z>O if ~50 of the plastic potential -ag aa0 function
(3)
of
(4) function
2.3 Istropic Elastoplastic Constitutive Relationships. The hypothesis of isotropy is a convenient assumption which simplifies the mathematical formulation of plasticity. By definition, isotropic elastoplastic models have yield and plastic potential functions which depend only on three stress invariants, instead of six independent stress components
g(a,,) =g(Z, J, 4,
invariants
(14) and
gradient
where I, J, and (Yare the first and second-stress Lode angle, respectively, I=
akk
mij =
(5)
(15)
aakl
ag ag -auk/
C
The normalized gradient nij of the yield function f(o& is similarly defined by using the flow function in equation (5). According to the terminology of plasticity, the flow rule is referred to as associative when nti and mti coincide. The tensors mij and nij are normalized so that they have only directional effects on the plastic flow, the amplitude of which is directly related to the plastic modulus H. The modulus His generally calculated by enforcing the stress state to stay on the yield surface f(qij) = 0 during plastic flow (consistency condition). The strain mcrement resulting from a given stress increment is obtained by adding equations (2) and (3).
The deviator
stress is
(16)
and the third stress invariant S is
(17)
de, =
where the elastoplastic
cijkl = z
cijk/
dUk/,
matrix is
+$kIrnu.
(6)
cy is preferred to S since cu can be represented in the principal stress space. According to equation (15), CI varies between 7r/6 and +*/6. The differentials of I, J, and (Y with respect to stress components aO are ar = 6, ao, aJ -= aao 3
constitutive
&,dk,+&
(&kh,,+$fijk)
(7)
2.2 Inverted Stress-Strain Relationship. In many instances, such as the displacement formulation of nonlinear finite element methods, stresses need to be computed as results of strain loading. Therefore, equation (6) must be inverted to provide the stress increments resulting from given strain increments.
25
(18)
(8)
constitutive where S, is (19) and has the following properties sp,, = 0
(9)
matrix Etik, is
(10)
s,;i, = 0 The inversion of equation (6) poses no problem as long as the denominator of equation (9) is different from zero, which is generally satisfied except for some particular stress states. Equation (9) may be rewritten in a simpler form Dukt = D,/ - f where Mjj =
Eoklmkl Nj = E,,klh (12) Mipkl (11)
(20)
SCS,,= 1 S, is undefined for (Y= f 7r/6 when the gradient of (Yis equal to zero in (18). By invoking the chain rule of differential calculus, the gradient of isotropic yields functions is a linear combination of f,, fJ and f,, the differential of the yield function with respect to I, J and CX,respectively: 1
IJ
.s.. Jzf
(21)
The terms on the right-hand side of equation (21) defines the volumetric, deviatoric, and Lode contributions, respectively, on the gradient of yield function. The direction and amplitude of Lode contribution are characterized by S,j and f, 7, respecnorm is
As represented in Fig. 1, x is the angle between the radial direction and the gradient of the yield function. According to (21) and (26), x characterizes the amplitude of the Lode contribution acting in the direction of S,,. As it will be ilIustrated in a later section, (25) is useful to introduce the effects of (Yinto the isotropic yield functions f (I, J) that are defined only in terms of first and second invariants. 3.2 Simple Examples of Yield Function for Pressure-Sensitive Materials. Drucker-Prager, Tresca, and Mohr-Coulomb provide three examples of yield functions that satisfy (25). In the principal stress space, the Drucker-Prager surface is a cone with a circular deviatoric section centered on the hydrostatic axis
Fig. 1 Typical failure surface for isotropic pressuresensitive shown in principal stress space and on deviatorfc plane
material
(28)
where a, and b, are two material parameters. Since the yield function is independent of CX, Lode dependence of Druckerthe _f* =dG=/W and n,, may be calculated (22) from (5), (21), and (22). N,, becomes (23) Prager is p,((y)= 1. (29) cone with a regular depends on CY: (30) Its Lode de-
The extended Tresca surface is an angular hexagonal cross-section; its yield function
The definitions and results established for yield functions in (21), (22), and (23) pertain to plastic potential functions. The modulus K of (13) is related to the plastic modulus H
parameters.
K =H f f*g*
3 Lode Dependence
9Bf,g, i- GfJgJ f
ag a I J*
VJ
r
(31) of
- Isin+ - 3ccos+,
(32)
The following section introduces a general class of yield functions that can be used to account for (Y in isotropic elastoplastic materials. After reviewing the yields functions of Diucker-Prager, Tresca, and Mohr-Coulomb requirements are specified for lode dependences to give differentiable and convex yield functions. Five examples of Lode dependences are also presented.
where 4 is the friction angle and c is the cohesion. function of Mohr-Coulomb can be rewritten:
The yield
Jf3(Z, J, a)=*
P3(ol)
+Z(P-
l)-2cJ(l-2@(1 is
+P).
(33)
Therefore,
3.1 A General Class of Yield Functions for Pressure-Sensitive Material. Figure 1 shows a principal stress view of typical yield surfaces that has been proposed for pressuredependent soil, rock, and concrete based on true triaxial experiments (Lade and Duncan, 1975; Matsuoka and Nakai, 1974). In general, such yield functions depend only on Z and the ratio J/p (cu) where p is a function of CY.They belong to the following class of isotropic yield functions:
P3(4
= (I +p)cos~r+&-
Ah
I)sincY to the friction angle 4: /3 is related p=3 - sin@ 3 + sin4
(34)
f(Z, J, cr)=f
where the Lode dependence p(a) is arbitrarily set to one when cy= 7r/6. The yield surfaces f (I, J/p ( (Y) ) = 0 have deviatoric sections in planes of constant Z that look similar but do not necessarily coincide. The deviatoric sections are circular when the yield function does not depend on QL,i.e., p(a) = 1. They are not circular any longer when p varies with CX,however, they remain periodic curves of period 2~/3 since u varies between - 7r/6 and 7r/6. It can also be shown that the yield functions satisfying equation (25) give
When 4 varies between 0 deg and 90 deg, fi varies between 1 to 0.5. As shown in Fig. 2(a), the deviatoric sections of the Mohr-Coulomb surface are piecewise linear. They coincide with Tresca sections when /3 is equal to one and are triangular when @ is equal to 0.5. The angle x of Fig. 1 varies IinearIy as a function of CX,but changes abruptly when Q is equal to *r/6; the yield function of Mohr-Coulomb is not differentiable at Q = f 7r/6. 3.3 Requirements for Lode Dependence for Pressure-SenBesides Drucker-Prager, Tresca, and Mohrsitive Materials. Coulomb, several types of failure surfaces have been proposed for pressure sensitive materials. They generally satisfy (25) and obey the requirements for aspect ratio, differentiability, and convexity. 3.3.1 RequirementforAspect aspect ratio states that: P and
f, = - JtanXfJ
where tanx is related to cy to pu, the differential
Ratio,
The requirement
on
tanX=&.
P
p 0
= 1
(37)
1990
Transactions
of the ASME
Fig. 2 of i3
MohrCoulomb
where fl varies between 0.5 and 1. Equation (37) controls the general shape for the deviatoric sections of yield surfaces. This requirement comes from the experimental results (Lade and Duncan, 1975; Matsuoka and Nakai, 1974): 0.5sJ(<
J(n/6)
1 at Z constant
As shown in Fig. 3(a), the elliptical dependence provides convex cross-sections for all values of fl. Due to its convexity for all values of /3, William and Warnke dependence was used to account for cxin modeling the elastoplastic behavior of concrete (Lin et al., 1987). When /3 tends to 0.5, the deviatoric section becomes triangular, as for Mohr-Coulomb. Argyris 1974) proposed a simpler expression than William and Warnke:
Pfsff) =
where J(r/6) and J( - ~16) are the values of J that correspond to failure in compression with two identical minor principal stresses and to failure in tension with two identical major principal stresses, respectively.
23
1 +P+(Pl)sin3a
3.3.2 Requirement for Differentiability. tions of (25) are differentiable at a= &r/6 =o, provided
However, as shown in Fig. 3(b), the deviatoric section becomes nonconvex when p becomes less than 7/9, which corresponds to a friction angle 4 = 22 deg. Argyris dependence was later generalized by Eekelen (1980) who introduced an additional parameter n:
p,(a) = 2( 1 + pn + (1 - P)sin3a).
that p(cr) and f(Z, J) are differentiable.
(43)
3.3.3 Requirement on Convexity. The yield surfaces of (25) are convex in deviatoric planes (Lin and Bazant, 1986; Jiang and Pietruszczak, 1988) when
Equation (42) is a particular case of (43) when n is equal to - 1. Eekelen dependence is plotted in Fig. 4(a) for the value n = - 0.299. The surface loses its convexity for a value of p between 0.7 and 0.6. 3.5 LMN Dependence. Based on the failure surfaces of Lade and Duncan (1975) and Matsuoka and Nakai (1974), the following Lode dependence is proposed: (44)
d*
-&P+-P:.
2 (39)
P
The convexity requirement is useful to satisfy the stability postulate of Drucker (1951). The following sections present five Lode dependences which satisfy the requirement for aspect ratio and differentiability and occasionally the requirement for convexity.
where 3.4 Additional Examples of Lode Dependence. The first dependence is based upon a cubic polynomial that is calibrated to satisfy the requirements of (37) and (38). This cubic dependence is: p,(a)=l-2(~)3(l-il)(~-a)2(~+~) (40) if ~50 0= if cr>O. > (45) This dependence is hereafter referred to as LMN, after the first letter of Lade, Matsuoka, and Nakai. 0 varies continuously with cy since (45) gives 0 = 7r/6 for cr = 0. The deviatoric section is shown in Fig. 4(b). By comparing Figs. 3(a) and 4(b), it is concluded that LMN and elliptical dependences are similar. In summary, Lode contribution can be accounted for in isotropic elastoplastic constitutive models by using cubic,
Figure 2(b) shows ~~(0) for various values of p between 0.5 and 1. The deviatoric cross-sections are smooth but do not satisfy the convexity requirement of (39) for p smaller than 0.85, which corresponds to a friction angle += 13.8 deg according to (35). x varies continuously with CY but not linearly as in the case of Mohr-Coulomb. Instead of a cubic dependence, William and Warnkle (1975) proposed an elliptical dependence: 2( 1 - /3)cos(*/6
P&4 =
- Q) + (20 - 1)2/4( I- /32)cos2(7r/6 - (Y)+ /3(5@- 4) 4(1- fi2)cos2(7r/6 - (u) + (2fi - 1)2 (41)
SEPTEMBER
Fig.
Fig. 4
Argyris, elliptical, Eekelen or LMN dependence. Cubic, Argyris, and Eekelen functions have simple mathematical expressions but lose convexity for values of parameter 0, typical to frictional materials. Elliptical and LMN dependences, which have a more complicated analytical expression than the other dependences, satisfy the convexity requirement and are almost identical for all values of parameter p.
It can be shown that J=p(cu)J,,,,, is the acceptable (49) when p(a) is given by (44) and J,,,,X is
solution
of
II = l 4 Application of LMN Dependence to Failure Surfaces 27One of the advantages of LMN over elliptical dependence is that it describes exactly the failure surfaces of Lade-Duncan and Matsuoka-Nakai. 4.1 Application to Lade-Duncan Duncan (1975) proposed the following P-l),z,=o Surface. Lade failure criterion: and (46)
4 is the friction angle measured during triaxial compression, i.e., for (Y= n/6. When fi tends toward one, which corresponds to 4 = 0 deg, vI is equal to 27. When fi tends toward 0.5, which corresponds to 4 = 90 deg, 7, is infinite. It is worth noting that Lade and Duncan give a larger value for J in tensile failure (CY - p/6) than Mohr-Coulomb. = 4.2 Application to Matsuoka-Nakai Surface. Based also upon experimental results, Matsuoka and Nakai (1974) proposed a failure surface that slightly differs from a Lade-Duncan surface z,z where v2 is a dimensionless second stress invariant z2=1 2
l/*z3 = 0,
where ?, is a dimensionless material constant, Z is the first stress invariant and Z3 is the third stress invariant Z, = det(u,,) = ~1~2~3. (Jl, a29 and u3 are the principal between Z,, I, J, and (Y:
I3 =
stresses.
material
2 -J3sin3cu 36
1 1 - -ZJ* + -Z3. 3 27
(48) cubic
(z*-u#7~). equation:
(53)
By using (46) and (48), the ratio J/Z obeys the following equation: $(g3sin3cy(g*+ (k-i) =O.
(49)
$(g3sin3a-
(1-i)
(g*+
(k-i)
=O.
(54)
1990
Transactions
of the ASME
It can be shown that J=p(a)J,, is the acceptable solution of (54) when p(u) is given by (44) and J,,,, is the same as for Mohr-Coulomb:
0.2
J,,,~ = 18++-4
The constant q2 is related to fl and $J through:
(55)
0.1
(56)
0.0
In summary, LMN dependence describes the failure surfaces of Lade-Duncan and Matsuoka-Nakai by using two different values of 0. In contrast to the Lade-Duncan surface, the surfaces of Matsuoka-Nakai, and Mohr-Coulomb coincide at Q= &t/6.
-0.1
-0.2
Application
of Lode Dependence
to the Roscoe-BurFig. 5
0.0
0.i
0.4
land Model
One of the main applications of Lode dependence is to introduce the effects of (Y into isotropic elastoplastic models that are formulated only in terms of first and second-stress invariants. Following the suggestion of Dafalias and Hermann (1986), the modified yield functions f (Z, J, CY) are obtained by replacing J by J/p( (Y) in the yield function f(Z, J) as indicated in (25). The following section illustrates the introduction of o( into a particular elastoplastic model.
I/IO
0.6
O.s
1 .b
fJ=
-= --
54J
Mt~ta)I
(63)
5.1 Generalization of Stress-Strain Relationships. Among all the material models of soil mechanics, the Roscoe and Burland (1968) model is certainly the most commonly used to describe the behavior of normally and overconsolidated clays. This model was developed based upon triaxial tests. In spite of the appellation triaxial, these latter tests apply axisymmetric stress states to the soil samples 022=(r33, and uij=O if i#j. (57)
Since the evolution of the yield surface during plastic flow is controlled by only one internal variable; G, the consistency condition of plasticity theory implies that
gduij 1,
Since f is isotropic,
+ $dG
= 0.
(64)
According to (57), stresses are described by two independent components, such as mean pressure p and deviator stress q 1 P=-(%+2%) 3 When (57) applies, invariants Z and J q=u,,-J33. to the stress
(64) becomes
fdZ+f&J+f,da+s
By using (3), the plastic volumetric
dG=O.
strain is
q=&J
(59)
Therefore,
and
dp,= (Y=
~16 1 -?r/6 if q>O if q<O.
(f,dZ+fdJ+fada)z,
IJ
(67)
(60)
yield surface
f(p,
s 4, <)=*+p-poexp
(61)
where M, X, and K are three material constants. The constants p0 and e, are the initial values of mean pressure and void ratio, respectively. The only internal variable of the model is the plastic volumetric strain 4 that is referenced to the initial state characterized by p. and eo. By using (35) and (59), equation (61) is generalized in terms of stress invariants
Equations (67) generalizes the incremental stress-strain relationships of Roscoe and Burland (1986) by accounting for a. The yield and plastic potential functions can be modified by selected any type of Lode dependence p(a) defined in the previous sections. Hereafter, their Lode dependences are of the same type and satisfy
g1=f*
gj
=fP
(68)
However, g, may be set different from f, by selecting two different values of constants fl and fl,,, respectively. The flow rule is associative when /3 and &, are equal. Otherwise, it is nonassociative. 5.2 Theory Versus Experiment for Loading at Constant I and cr. In the case of loading at constant Z and cr, the second stress invariant Jis gradually increased until the specimen fails: z=zo,
f(Z, J, a, G)=
27J*
M%(a)*1
+ Z - Zoexp
where IO= 3 po. The yield surface of (62) is plotted in Fig. 5 for two different values of fi. The differentials of yield function with respect to Z, J, and I$ are: Journal of Applied Mechanics
dZ= 0,
and
dcr=O.
(69)
SEPTEMBER
The incremental strain can be integrated analytically when J are constant, the stress is varies from 0 to J,,,,,. Since Z and CY a linear function of J: sin(i?-a) if i=j= if i#j and the direction of Lode contribution is constant: if i=j=
>
1,2,3
(74)
s,=
L
0
1> 2 13
2 sin3++$$
if i#j (75)
strains
Such stress-controlled loadings were carried out in the laboratory by Nakai et al. (1986). With the mean of a true triaxial apparatus, they applied five different values of cx ((Y= 30 deg, 15 deg, 0 deg, - 15 deg, and -30 deg) to cubical material samples made of normally consolidated Fujinomori clay. Theory and experiment can be compared by examining (1) failure stress states and (2) stress-strain responses. The values of the model constants are provided by Nakai et al. (1986): h v=O. Thevalue = 0.0508, *=0.0112,+=33.7degand 1 +e, of p=O.688 for yiid function corresponds to 4= 33.7 deg. The value of &, for plastic potential function is chosen to be equal to 0.688 and 1 in the case of associative and nonassociative flow rules, respectively. 5.2.1
(76) where
Comparison
The maximum
Stresses.
Jmax= 3&ZMg
Figure 6 compares the experimental and theoretical failure stresses obtained for five different values of cy. LMN and elliptical dependences provide the best agreement with experimental observations. The Mohr-Coulomb dependence gives conservative failure stresses whereas cubic, elliptical, Argyris, and Eekelen dependences produce nonconservative failure stresses. The original Roscoe-Burland model that does not account for cy(/3 = 1) clearly overestimates the material strength when LY different of 7r/6. is 5.2.2
Response.
increment
where the shear modulus G is constant (Roscoe and Burland, 1968): G=The plastic strain increment
&+--h--K
since Z is constant
l+eo -I.
K
(72)
fJ
%dJ
1 +eo
(73)
In (76), p and tanx depend on fl and flp, respectively. The stress-strain responses of (76) have been specified for the material constants of Nakai et al. (1986) and plotted in Figs. 7, 8, and 9 for various values of a. The material responses are reported as in Nakai et al. (1986) by plotting the ratio between the major and minor principal stresses, i.e., u1/u3, versus major, intermediate, and minor principal strains, noted t,, t2, and e3, respectively. The volumetric strain E,= t, + t2 + t3 is also plotted versus the major principal strain t,. In the triaxial compression test of Fig. 7(a), all Lode dependences give identical results since p(7r/6) = 1 and p,(r/6) = 0 for any values of p and 6,. The stress-strain response is identical to the one predicted by the original Roscoe-Burland model. In the tension test of Fig. 7(b), all Lode dependences with 0 = 0.688 give the same stress-strain response independently of Pp, since p( - 7r/6) =/3 and p,( - 7r/6)=0 for any 0,. The stress-strain response of the original Roscoe and Burland model that is plotted in a dashed line in Fig. 7(b) largely overestimates the material strength. Similar results are obtained for the loadings corresponding to cy= f 15 deg which are shown in Figs. 7(c) and 9(d). Figure 8 compares the associative and nonassociative responses resulting from LMN dependences during the particular loading at CX=O deg; the intermediate principal strain Edis simulated more accurately with a nonassociative flow rule. Figure 9 compares the responses of six different Lode dependences for p=O.688 and &= 1. Figure 9 indicates major differences for the maximum value of a,/a3, but not for the intermediate strain c2, which is to be expected since pa = 0 when /3,= 1. In summary, it is concluded that both LMN and ellipitical dependences give identical stress-strain responses and failure stresses for the generalized Roscoe and Burland model. The Mohr-Coulomb dependence gives the most conservative failure stress and the softest stress-strain response. The other Lode dependences-cubic, Argyis, and Eekelen-can be used alternatively to LMN or elliptical dependences, when the friction angle is small or when yield and plastic potential functions are not required to be convex. Transactions of the ASME
Fig. 7(a)
6 60!= 5-
Fig. 7(c)
-30
\ \
\
0 -I
-a,,
I I I I
CY= -15
\ \
a,;
/ I
4-
A A E3
\ \ \\
E,
2-
I-10
I 10
ooA+
Fig. 7 Comparison between theory and experiment: principal strains versus principal stress ratio and volumetric strain in true triaxial tests for (a) a=30 deg, (b) CI= -30 deg, (c) a=15 deg, and (do U= -15 deg (experimental data after Nakai et al., 1966)
6 cl!= 0 (T -I
Fig. 6 Comparison between experiment and theory for associative and nonassociative flow rules: principal strains versus principal stress ratio and volumetric strain in true triaxial tests for u=O deg (experimental data after Nakai et al., 1966).
Fig. 9 Comparison between experimental and theory for six different types of lode dependence: principal strains versus principal stress ratio and volumetric strain in true triaxial tests for a=0 deg (experimental data after Nakai et al., 1966)
Conclusion
A particular class of yield and plastic potential functions has been introduced in order to account for the effects of Lode angle on isotropic pressure-sensitive materials. This class of functions is illustrated by seven Lode dependences found in the literature and by a new one, referred to as LMN, that is similar to the William and Warnke (1975) dependence. The Journal of Applied Mechanics
LMN dependence gives a convex surface for all values of frictional angle and generalizes the failure surfaces of Lade and Duncan (1975) and Matsuoka and Nakai (1974). The proposed formulation has been applied to introduce the Lode angle (Y into the Roscoe and Burland model. The modified model is capable of reproducing the experimental observations made on true triaxial tests of clay. The formulation is general enough to introduce CY into the isotropic elastoplastic models which SEPTEMBER 1990, Vol. 57 I505
invar-
Acknowledgment
The financial support of the U.S. National Science Foundation (grants CEE-8404315 and MSM 8657999) is acknowledged.
References
in the Finite Element Argyris, J. H., et al., 1974, Recent Development Analysis of Pressure Container Reactor Vessel, Nuclear Engineering and Design, Vol. 28, pp. 42-75. Chen, W. F., and Saleeb, A. F., 1982, ConstirutiveEquarionsforEngineedng Materials: EIosficify ond Modeling, John Wiley and Sons, New York. Dafalias, Y. F., and Herrmann, L. R., 1986, Bounding Surface Plasticity. Mechanics, II Application to Isotropic Cohesive Soils, Journal of Engineering Vol. 112, pp. 1263-1290. Drucker, D. C., 1951, A More Fundamental Approach to Plastic-Stress Strain Relations, Proceedings 1st U.S. National Congress on Applied Mechonrcs, ASME, New York, pp. 487-491. Eekelen, H. A. M., 1980, Isotropic Yield Surface in Three Dimensions for Use in Soil Mechanics, Inl. J. for Numerical and Anolytrcal Methods in Geomechanics, Vol. 4, pp. 89-101. Gudehus, G., 1973, Elastoplastische stoffgleichungen fur trockenen sand, Ingenieur Archiu, Vol. 42, pp. 151-169.
Jiang, J., and Pietruszczak, S., 1988, Convexity of Yield Loci for Pressure Sensitive Materials, submitted to Computers ond Georechnics. Lade, P. V., and Duncan, J. M., Duncan, 1975, Elastoplastic Stress-Strain Theory for Cohesionless Soil, J. Geotechnical Engineering Division, ASCE, New York, Vol. lOI, pp. 1037-1053. Lin, F. B., and Bazant, Z. P., 1986, Convexity of Smooth Yield Surface of Frictional Material, Journal of Engineering Mechanics, ASCE, New York, Vol. 112, pp. 1259-1262. Lin, F. B., Bazant, Z. P., Chern, J. C., and Marchertas, A. H., 1987, Concrete Model with Normality and Sequential Identification, Computers and Stmclures, Vol. 26, pp. 1011-1025. Matsuoka, H., and Nakai, T., 1974, Stress-deformation and Strength Characteristics of Soil Under Three Different Principal Stress, Proceedings Japanese Society of Civil Engineers, Vol. 232, pp. 59-70. Nakai, T., Matsuoka, H., Okuno, N., and Tsuzuki, K., 1986, True Triaxial Tests on Normally Consolidated Clay and Analysis of the Observed Shear Behavior Using Elastoplastic Constitutive Models, Soils ond Foundations. Vol. 26, pp. 67-78. Roscoe, K. H., and Borland, J. B., 1968, On the Generalized Stress-Strain Behavior of Wet Clay, Engineering Plasricify, Cambridge Univ. Press, pp. 535-609. William, K. J., and Warnke, E. P., 1975, Constitutive Model for the Triaxial Behavior of Concrete, Inlernational Association for Bridge and Structure Engineering Proceedings, Bergamo, Italy, Vol. 19. Yang, B. L., and Dafalias, Y. F., and Herrmann, L. R., 1985, A Bounding Surface Plasticity Model for Concrete, Journal of Engineerrng Mechanics, ASCE, New York, Vol. I I I, pp. 359-380. Zienkiewicz, 0. C., and Pande, G. N., 1977, Some Useful Forms of Isotropic Yield Surfaces for Soil andRock Mechanics, Fimle Elements in Geomechanics, G. Gudehus, ed., John Wiley and Sons, New York, pp. 179-190.
508