Continuum Theory of Ductile Rupture by

Void Nucleation and Growth: Part 1 —

Yield Criteria and Flow Rules for Porous
Ductile Media
A. L. GURSON Widely used constitutive laws for engineering materials assume plastic incompress-
Division of Engineering, ibility, and no effect on yield of the hydrostatic component of stress. However, void
Brown University, nucleation and growth (and thus bulk dilatancy) are commonly observed is some
Providence, R. I.
Currently, Structures Dept. processes which are characterized by large local plastic flow, such as ductile fracture.
Avco Corp. The purpose of this work is to develop approximate yield criteria and flow rules for
Wilmington, Mass. porous (dilatant) ductile materials, showing the role of hydrostatic stress in plastic
yield and void growth. Other elements of a constitutive theory for porous ductile ma-
terials, such as void nucleation, plastic flow and hardening behavior, and a criterion
for ductile fracture will be discussed in Part II of this series.
The yield criteria are approximated through an upper bound approach. Simplified
physical models for ductile porous materials (aggregates of voids and ductile matrix)
are employed, with the matrix material idealized as rigid-perfectly plastic and obeying
the von Mises yield criterion. Velocity fields are developed for the matrix which con-
form to the macroscopic flow behavior of the bulk material. Using a distribution of
macroscopic flow fields and working through a dissipation integral, upper bounds to
the macroscopic stress fields required for yield are calculated. Their locus in stress
space forms the yield locus. It is shown that normality holds for this yield locus, so a
flow rule results. Approximate functional forms for the yield loci are developed.

1 Introduction and growth. This constitutive theory will then be used in a model
1 of the ductile fracture process developed in Part II.
It has been observed [1-6] that ductile fracture in metals can A plastic constitutive theory can be constructed from the
involve the generation of considerable porosity, via the nuclea- following components: First, a function of stress which defines
tion and growth of voids. Constitutive theories which take ac- the combinations of stress for which plastic yield takes place (a
count of porosity are therefore a desirable component of a mathe- yield criterion) is needed. The next component is a flow rule,
matical model of the ductile fracture process. Previously de- which defines the ratio of the strain components as a function
veloped plasticity models, such as that due to von Mises, predict of the stress state at yield. This can often be put in terms of
plastic incompressibility and therefore could not show the the normal to the plastic potential, another function of stress.
dilatancy evident in porous ductile materials. In this and a com- (In many cases, the yield function can be used as a plastic
panion paper (Part II), an approximate plastic constitutive potential, see [7].) To relate the increment of plastic flow to the
theory will be developed which takes account of void nucleation increment in stress, a consistency relation and some hardening
assumptions are needed. When void nucleation as well as void
growth takes place, a nucleation criterion must be added to
complete the constitutive description.
'Numbers in brackets designate References a t end of paper.
In this paper, the yield criterion and flow rule for porous
Contributed by the Materials Division for publication in the JOURNAL OF ductile materials are investigated. Approximate yield criteria
ENGINEEBING MATEBJALS AND TECHNOLOGY. Manuscript received by the
Materials Division, September 19, 1975; revised manuscript received June 1,
are derived using simple rigid-plastic material models and the
1976. Paper No. 76-Mat-CC. upper bound theorem of plasticity. (Because void growth and

2 / J A N U A R Y 1977 Transactions of the ASME

Copyright © 1977 by ASME

Downloaded From: http://materialstechnology.asmedigitalcollection.asme.org/ on 09/13/2013 Terms of Use: http://asme.org/terms

 ductile fracture involve large amounts of plastic flow, a rigid- Mises equations are used to characterize the yield and flow be-
nlastic idealization is taken as acceptible.) A normal flow rule havior of the matrix, which is itself incompressible. An ap-
is established, with the approximate yield functions serving as proximate form is then assumed for the microscopic velocity
nlastic potentials. (For a discussion of the normal flow rule and field in the matrix, which allows the voids to change volume while
its application to the specific forms of the plastic potential de- maintaining matrix incompressibility. This velocity field must
veloped here, see the appendix.) The subjects of continuing obey compatibility, and meet kinematic boundary conditions on
nlastic flow, hardening, void nucleation at second phases, and the surface of the unit cube which correspond to prescribed
predictions of ductility (strain to fracture) will be investigated macroscopic rates of deformation. An upper bound inequality
in the subsequent paper. is then used to calculate upper bounds to the macroscopic stresses
Along with the observations cited above, some recent theo- required to sustain plastic flow. The locus of upper bound
retical studies [8-10] indicate that when voids are present in macroscopic stresses for a given void geometry (size, shape, dis-
ductile materials, the hydrostatic component of stress can cause tribution, etc.) and a range of macroscopic rate of deformation
macroscopic dilatation and affect plastic yield. Because the ma- fields form an upper bound yield locus for that unit cube. Given
terial surrounding the void is idealized as incompressible, the certain restrictions on the approximate microscopic velocity
dilatation is due completely to void growth. The approximate field, a maximum plastic work principle can be established for
yield functions (usable as plastic potentials) developed here will the upper bound yield locus. (Bishop and Hill establish one for
contain terms which are functions of hydrostatic stress, and thus the true yield locus). A functional form which gives a good fit
allow plastic dilatancy through the normality rule. to the upper bound yield locus can be called the approximate
The general physical model used here is a " u n i t " cube not (upper bound) yield function.
unlike t h a t of Bishop and Hill [11], only of a porous material For purposes of analysis, the void-matrix aggregate is idealized
(aggregate of voids and ductile matrix) rather than .a poly- as a single void in a rigid-plastic cell; the void volume fraction
crystalline aggregate. The cube is by definition large enough (/) of the cell equalling t h a t of the aggregate. (In this way, some
to be statistically representative of the properties of the ag- account is taken of the interaction of neighboring voids.) The
gregate. As Berg [12] has pointed out, the Bishop and Hill cell is presumed to behave under loading as the aggregate would,
analysis can be extended to apply to a porous material. This exhibiting void growth when undergoing yield with a positive
allows proof of a macroscopic maximum plastic work principle hydrostatic component of macroscopic stress. Two void geom-
for the aggregate, if the same principle applies locally to the etries are considered; the long circular cylinder and the sphere.
matrix material. Throughout this paper, the adjective "macro- The outer cell wall is idealized as geometrically similar to, and
scopic" refers to average values of physical quantities (stress, centered around the void. These geometries were chosen be-
dissipation, velocity, etc.) which represent the aggregate be- cause they resemble m a n y of the void shapes seen experimentally,
havior. "Microscopic" refers to pointwise quantities, such as the they provide the expected isotropy (transverse directions for the
stress or velocity fields in the ductile matrix. cylinder, total for the sphere), and because their symmetry
The general method used here to calculate an upper bound properties significantly aid the analysis. See Figs. 1, 2, and 3.
yield function for a void-matrix aggregate is as follows: The von These simple cell models will not, of course, behave exactly like

•Nomenclature*
2/yy = macroscopic dilatational stress, cylin-
(To = microscopic equivalent tensile yield stress drical model
d, (Tij = microscopic stress tensor 2u = macroscopic dilatational stress, spherical
S, Sij = microscopic deviatoric stress tensor model
e. in = microscopic rate of deformation tensor T, Tis = normalized macroscopic stress
V, Vi = microscopic velocity field = intermediate parameters
x, g
X, Xi = microscopic position vector, cartesian $ = yield function
coordinates f^eqv = an empirical coefficient
Zi, Ziij = macroscopic stress tensor cosh, sinh = hyperbolic cosine and sine
M« = macroscopic rate of deformation tensor dm, Am, nm = coefficients in arithmetic series
i\EU = macroscopic deviatoric rate of deforma- m = index and exponent
tion tensor a = angle of rigid-plastic boundary
V = volume of body V, Vi = macroscopic velocity boundary conditions
<S, Svoid = outer surface of body, void surface VN, ViN = V, normalized by b
n, Hi = unit normal vector on «S T = dilatation function
W = macroscopic dissipation Ton — normalized dilatation stress, cylindrical
f - void volume fraction model
undefined parameters Bi — macroscopic yield function coefficients,
f - parameter used to minimize W; angle of functions of /
rigid-plastic boundary 2s n, &s = normal and shear stresses on a plane of
w Vn , Vis = macroscopic parameters, used to meet zero extension
macroscopic boundary conditions P, L — void radius and spacing, used in one of the
9, 7 = angles of rotation about the (3) axis, y references
also used as an index s, v = superscripts on v and c, indicating shape
E', E, E?3 = macroscopic rate of deformation param- and volume change
eters (see equations (3.3)) r,6,<t> = spherical coordinates
a, b = inner and outer radius of void-matrix ha = geometric parameter
solid angle
r =
model
radius of a point inside the matrix
a=
TH = normalized dilatation stress, spherical
H, A = geometric parameters model
macroscopic equivalent tensile stress 8 = indicates small variation

Journal of Engineering Materials and Technology JANUARY 1977 / 3

Downloaded From: http://materialstechnology.asmedigitalcollection.asme.org/ on 09/13/2013 Terms of Use: http://asme.org/terms

 *" 2

Fig. 1 Void—matrix aggregate, random void shapes and orientations,

macroscopic and microscopic tensor quantities
UNIT SPHERE
PRINCIPAL AXIS SYSTEM '
Fig. 3
aggregates with random distributions of voids. Therefore, the
upper bound yield loci obtained for the cell models are in a strict
•sense only estimates (bounding properties unknown) of the yield mination of functional forms for the upper bound yield functions,
loci for more random aggregates. thus giving a clear (if approximate) picture of the role of hydro-
The approximate velocity fields used in the upper bound cal- static stress in the yield and flow of porous ductile materials.
culations can be simple or complex, depending on the degree of Some of the previous work done in this field is as follows:
approximation desired. T h e simplest fields need only meet those McClintock [8] developed a fracture criterion based on his analy-
conditions discussed previously. More complex fields may con- sis of a cylindrical cavity in an infinite matrix subject to axial
tain extra factors which can be adjusted to achieve better upper and transverse stresses. An exponential dependence of the void
bounds. The simplest velocity fields allow the rigorous deter- growth rate on biaxial stress was found. Rice and Tracey [9]
considered a spherical void in an infinite matrix, and found an
exponential dependence of void growth rate on triaxial stress.
Kahlow and Avitzur [10] studied the problem of the critical
pressure needed to prevent void growth during axially sym-
metric deformation. Their model was a closed cylinder of ductile
material with a cylindrical hole at the center, and their approach
was in some ways similar to t h a t used here.
There are several studies which motivated the examination of
more complex microscopic flow fields. Nagpal, McClintock,
Berg, and Subudhi [13] presented a plane strain slip line solution
for a band of evenly spaced holes under varying ratios of shear
and normal traction. Needleman [14], and Haward and Owen
[15] presented finite element studies of large flow in two-dimen-
sional models of porous materials, using elastic-plastic constitu-
tive descriptions for the matrix material. Specifically, these
studies suggest t h a t part of the matrix might not attain plastic
yield for some types of aggregate yield. These "non-plastic"
areas are idealized as rigid here. Their shapes are approximated
as shown in Fig. 4.

2 General Theory
T h e general model considered here is a " u n i t " cube of porous
material of volume V, large enough to be statistically representa-
tive of the properties of the aggregate (Fig. 1). The matrix ma-
terial is a homogeneous, incompressible, rigid-plastic, von Mises
material. Its yield and flow relations are

Fig. 2 Long circular cylindrical void in a matrix of rigid—perfectly

plastic von Mises material
— SijStj = 2
<T 0 (a), ««(e) - i 2 •
- foe«v
3
(«*I«*I)1/8
(6)

4 / J A N U A R Y 1977 Transactions of the ASME

Downloaded From: http://materialstechnology.asmedigitalcollection.asme.org/ on 09/13/2013 Terms of Use: http://asme.org/terms

 equation (2.3). Where n is a surface normal vector, positive
outward from the void surface,

X
% = — fet/dV + \ \ (<Mi + vjm)dS. (2.4)
J Vmatrix *//?void

T h e last term above includes the dilatational part of E;J, and is

zero when there is no porosity. The boundary conditions on the
outer surface which must be met by the «; field are expressed in
terms of the 77,;. I t is important to note the 2?,-,- are average
quantities (equation (2.2)), and can represent many different
boundary distributions of «,-.
The velocity field must also meet the constraints of incom-
pressibility and continuity in the matrix. Velocity fields which
involve matrix separation are therefore excluded.
Among the infinity of incompressible Vi fields which meet the
above conditions, the actual v, field is characterized by its gen-
CYLINDRICAL MODEL eration of the minimum of the dissipation W;
RIGID WEDGE
PLANE STRAIN
l
y XSii
W = {i)indV (2.5)

where the terms in the integrand are related to w,- by equations

(2.1). (This is proven later on; see equation (2.23).) All physic? 1
quantities associated with the actual Vi field will be labeled with
the superscript "A," since they are also actual solutions. Quanti-
ties associated with other v; fields are approximate solutions. An
important property of <TijA is that it is an equilibrium stress field.
I t is expected that ViA is homogeneous of degree one in the Ei$.
Bishop and Hill postulate t h a t no correlation exists between
(TijA and position or displacement (velocity) over any plane sec-
SPHERICAL MODEL tion through the unit cube. They also establish a maximum
RIGID CONE plastic work principle on the microscopic level, which also ap-
AXISYMMETRIC FLOW
plies to Sij(i) as defined in equation (2.1):
Fig. 4
(sy(e) - Siy(e*))e,-3- ^ 0, E * i* (2.6)

Using the noncorrelation postulate (above) and the principle of

virtual work, Bishop and Hill are able to prove that, when
l/dvi diij
tkk 0 (c) (2.1) S j j A is defined as the area average of <r;/A over the appropriate
2 Ida;,' dxi face of the unit cube, S 4 is the work conjugate of E:
where <7o is the equivalent tensile yield stress in the matrix, SJ.-
WA 2yA£,... (2.7)
the microscopic deviatoric stress field, e^- is the microscopic
rate of deformation field, Vi is the microscopic velocity field, and
1
A L **dS =

Xi is the position of a material point in cartesian coordinates. Using the principle of virtual work, they are then able to prove
The macroscopic rate of deformation is defined, as in Bishop a maximum plastic work principle on the macroscale for S' 1 :
and Hill [11], in terms of the velocity field on the surface of the
unit cube. (2V - WW, 2 0, (2.8)
A
where £ * is related to E*, and E* differs from E. The yield
En v-2js (VPIJ vpii)dS. (2.2) locus of S ' 1 thus has the properties of convexity and normality.
Equations (2.7) and (2.8) can then be combined to show
V is the volume of the unit cube, S is its outer surface, and n is
the unit outward normal on 5. Using the Gauss theorem and
equation (2.1c), it can be shown t h a t h~ZijAEij = 0 by normality
A
dW
E<, ~ f iitdV, = \: I" CitjdV + fcadV .: 2,- (2.9)
dEiS
Jr ' \_Jvmatrix J Vvoid .
iijA was derived from ViA, the actual velocity field, and gave
(2.3)
the actual dissipation through equation (2.5). When an ap-
Note t h a t in inside the void can be derived from a «; field ex- proximate field vi is used, equation (2.5) defines an approximate
tended from the matrix vt, the extended Vi field being arbitrary dissipation. The u,- fields considered here have the functional
except for the following: I t must equal the matrix vt field a t the form
void surface, and the combined (matrix and void) in field must
Vi = Vi(E, f, x). (2.10)
be continuous and have continuous first derivatives over the
entire volume V. (These conditions are required by the Gauss There may be additional dependence on other parameters qh
theorem.) As seen below, the Vi field inside the void influences 52, . . ., as in
the results only through its value at the void surface. This equals
Vi = V((E, /, x, qh qi...) (2.11)
the matrix Vi at the void surface, so in effect, the void Vi field does
not influence the results at all. with the qi's chosen to minimize the dissipation. For these
The Gauss theorem can now be applied to the last term in optimal values,

Journal of Engineering Materials and Technology JANUARY 1977 / 5

Downloaded From: http://materialstechnology.asmedigitalcollection.asme.org/ on 09/13/2013 Terms of Use: http://asme.org/terms

 g< = ?.-(E, / ) (2.12) to prove convexity from equation (2.17). Convexity can be
used, however, to help judge the value of approximate yield
so the form of equation (2.11) reduces to t h a t of equation (2.10). functions generated with this formulation. Since the object is
All of the forms of Vi considered here are homogeneous of degree to approximate the actual yield and flow behavior (which obeys
one in the components of E, i.e., convexity via equation (2.8)), an approximation which violates
for c = constant, Vi{cEn, cEn, ) = cVi(Eu, Em, ) (2.13) convexity should be considered a bad approximation.
To prove that the approximate yield locus lies outside the
e and W are then also homogeneous of degree one in the En, actual yield locus in stress space, write the principle of maximum-
giving plastic work in the following form:
dW
W = En. (2.14)
dEti Sij(iA))eijdV } 0, (2.22)
M i )
vfr
Define the approximate macroscopic stress needed tp cause
where the s;y are of the form in equation (2.1), and both E and
yielding (via the flow field vi) in a way analogous to equation
E A are compatible with the same macroscopic rate of deforma-
(2.9):
tion E. Using equation (2.5), the principle of virtual work, and
d
S.v =
dW_
dEa vj.
1 C
v
...
ski{i)
<ki
j ^ -
"En
dV, (2.15) the result t h a t both S and S x are work conjugates to E, equation !
(2.22) becomes

(St, - 2 , - / ) f t y ^ 0.
using the normality of s«(e) (see equation (2.6)) to set one part (2.23)
of the integrand to zero. This gives, with equation (2.14)
Since En is an outward normal to both the 2,-y and 2 y A yield
loci, this proves that the 2,7 surface always lies on or outside
W. (2.16)
the 2 , 7 A surface; S is an upper bound approximation to S A .
The most important properties established for the approximate
Thus, Sjy as defined above is a work conjugate to E,j, as is 2 ; / 1 .
macroscopic yield stress are its upper bound relationship to
By analogy with equation (2.9), normality is thus established for
S' 4 , the normality and convexity (given the conditions described
the approximate yield locus (the locus of stress states 2,7 de-
above) properties of its yield locus, and equations (2.15) and
fined by equation (2.15), for all possible directions of Eij).
(2.21) which give its relationship to E. Subsequent sections of
A maximum plastic work principle has been established for this paper are devoted to the solution of equations (2.15) and
J*A, giving both convexity and normality. I t is desirable to (2.21) for various types of vi fields, ranges of E, ranges of void
examine under what conditions this might also be established volume fraction /, and the two void geometries discussed earlier.
for £ . Consider two approximate stress fields 51 and £ * , cor- Varying the E field results in the generation of an approximate
responding to E and E*, respectively, through equations (2.15), yield locus for a particular void geometry, volume fraction, and
(2.10), (2.2), and (2.1). Write the following: flow field type. With some success, approximate functional forms
are derived for these yield loci.
deu
(2y - 2t,*)Etl s,,(t)
Vjy
dik
3a Long Circular Cylindrical Voids—Fully Plastic
s*i(e*) EijdV. (2.17) Flow
dEis*
This void geometry is meant to represent one limit of random
If this could be proven non-negative, a maximum plastic work void shape. Long, roughly cylindrical voids are seen to appear at
principle would result. Consider the case of e not only homo- the necks of tensile bars after large deformation (see, for ex-
geneous of degree one, b u t linear in E. Then, ample, reference [3]). They might also result from long cylindrical
inclusions (e.g., sulfides in steels) which decohere from the matrix
deu* + din
En — Eij = ikU (2.18) after straining, or on a larger scale, from drilled holes in a homo-
dEa' dEtj geneous material. The centered and geometrically similar matrix
With maximum plastic work proven on the microscopic level, displays the transverse isotropy expected of an aggregate with
the proof is complete. an isotropic matrix and a void distribution which is random in
Equation (2.18) applies to one class of velocity fields used later the transverse directions.
on. A second class, to which it does not apply, is of the form For the type of flow field considered here, in which all of the
matrix material is in the plastic state, a simplified form of
Vi(t, f, X, \p) (2.19) equation (2.2) is used as the boundary conditions of iv-
where »,-, and thus s and W, are homogeneous and linear in the
i\\s = EikXk\n (cartesian coordinates) (3.1)
Eij for a fixed value of ip. \p is an additional parameter equivalent
to qi in equation (2.12), and has the effect of making vi homo- The boundary values of «,- are thus uniquely defined by the Eij.
geneous of degree one, but no longer linear, in the Eij. The ap- In other formulations, the boundary distribution of w; will be
proximate dissipation thus has the form varied, within the constraints of equation (2.2) and certain
(2.20) geometric approximations, to achieve the best upper bound solu-
W = W(t, J, >A(E, /)).
tion.
Equation (2.15) establishes normality, and gives The approximate velocity field is constructed in a manner
similar to t h a t used by Rice and Tracey [9]; components rep-
dip
2 - — ' — (2 21) resenting shear and dilatation are constructed separately, each
dEa dip ah™ - satisfying compatibility and incompressibility. These com-
ponents are determined to within macroscopic parameters, which
Because ^(E, / ) is determined by minimizing W with respect to
are in turn determined by boundary conditions of the form of
equation (3.1). When the form of a component is not completely
dW determinable from symmetry and incompressibility, a general
— = 0 (b) (2.21) form is constructed in accord with a linear viscous (or equivalent-
Unfortunately, the dependence of ip on E makes it impossible ly, an incompressible elastic) model. This approach leads to

6 / JANUARY 1977 Transactions of the ASME

Downloaded From: http://materialstechnology.asmedigitalcollection.asme.org/ on 09/13/2013 Terms of Use: http://asme.org/terms

 more macroscopic parameters than can be determined from have a functional form very close to the result of a much simpler
boundary conditions alone, so other conditions, leading to refined calculation, which is described below.
velocity fields and lower values of W, are devised. This calcula- Suppose t h a t in addition to d , the coefficients C3, d, and C6
tion is carried out in reference [17], and draws on an example in are also set to zero. Then,
reference [16]. ri
The general results are as follows: vr = C%r cos (26) H 7
- E33 - + Vt\z cos (y),
T Z
C,
Vr = (C ir 3 + Or + C3r-1 + C4r-3) cos 2d + Vwz cos 7 + ve = - ( V sin (2d) - V*2z sin (7), v, = E33 • z + &r cos (y).
(3.7)
— Ex (3.2) Equating v a t r = 6 with the boundary conditions in equation
(3.4) gives
= ( - 2 C i r 3 - dr + dr*) sin 20 - V&z sin y,
X
Vz = E33z + (C3r + CV - 1 ) cos y C, = E\ C6 = Vn, and C7 = - Ekkb>,

d through Ci and V 32 are the macroscopic parameters referred

to above. The angle 6 is referred to transverse axes in which i„ = E< cos (26) - JSbV"2 - - E33
En is zero, and the angle y is referred to axes in which Eu is
zero. (This is done to facilitate the application of boundary con-
ditions—again, the details are in reference [17].) Vii is the shear- 6 09 = - E< cos (26) + Ebh-i - - E33,
ing velocity per unit axial length. In Fig. 2, the axes are such
that En = 0. Quantities in the axes where Ea = 0 (reached by a
«« = E33, ire = - E' sin (29), e„ = - (V*s + V^) cos (y),
rotation about the (3) axis) are denoted b y the superscript "*".
Boundary conditions are applied at the outer boundary (r
= b) in terms of the new macroscopic variables defined below. « « » = - - (V& + V&) sin (7), = - E?3 sin (7). (3.8)
V*3 is a normalized shear velocity parallel to the (3) axis, and
is in the same spirit a t V3i* (equation (3.2)).
Equations (3.8) describe an approximate E field in which the
deviatoric part has been set equal to the deviatoric part of E,
E' = - {En - En) (a), E = - Ehh ^ - (En + En + E33) (b), and a dilatational part added on which is derived from Ekk,
z z z
E33, and matrix incompressibility. This e field will now be used
1 to solve for X via equation (2.15).
p* _ p* (Vi + V&) (c) (3.3)
The dissipation can be written as follows, using equations (2.5)
and (2.1)
The velocity field on the outer surface is, in terms of these
macroscopic quantities;

vr\s = E'b cos (20) + - (En + En)b + V3*2 • z • cos (y),

z
W = —
r/A o-o • (iiiii^dV. (3.9)

Using equations (3.8), and referring all quantities to the same

axes;
v,\s = V&b cos (y) + E33 • z,
iuiu = 2 i " + En* + Ea* + ifiE'EX'1 + 2Ei\->
vd\s = - E'b sin (26) - V3% sin (y) (3.4)
3 j.
These are three equations in the seven unknowns Ci through C^. E33* + Ei3* + E3? + E^ + B322 (a) (3.10)
Some methods for determining the four additional equations
needed for a solution are discussed below. where
First, consider setting Ci equal to zero. Ci approaches zero
when / is very small in the previously cited elastic example [16]. (E&f = El3* + E13\b), tx = cos (26),
Also, it seems reasonable t h a t v should approach linearity in r
for r> >a "a" is the void radius). This can be true only if G\ is 6 measured c.c.w. from (2) axis (c), X a — (d)
zero. For these reasons, the approximation bl

Ci = 0 I t is assumed t h a t cr0 is constant in the matrix when equation

(3.5)
(2.15) is applied.
is adopted. T h e locus of S in stress space which is related to the entire
I t is reasonable to expect that a good approximate flow field range of E (the yield locus) will be expressed approximately in
will behave in a manner closely resembling the actual flow field. terms of the macroscopic equivalent tensile stress, and the
Note now t h a t the void surface is free of both normal and shear macroscopic transverse stress:
tractions. Because the matrix is a von Mises material, the related
shear deformation rates are also zero a t the void surface (see Seqv = ( " S i j S i y 1, = - (SJJ — Sn) 2 + ~ 233
equation (2.16))
ire\r-a = €«|r-a = 0 (3.6) 3 3
+ - (Sl22 4- S212) + -1 (S r 3S 7 3 + ~23~^J3y),
(Of course, an equivalent statement for normal components of 2 2
traction and £ does not apply.) y = 1, 2 and S 7 T , = S u -f- S J 2 (3.11)
A fourth and final condition results from minimizing the
macroscopic dissipation, W. These procedures are carried out in Note t h a t summation over a Greek index is over one and two.
detail in reference [17]. Equations (3.4), (3.5), and (3.6) are The stress components in equation (3.11), when calculated via
used to reduce the number of undetermined d to one. This last equation (2.15), are # expressed as volume integrals which contain
value is determined by finding its value which minimizes W. the components of E and the angular and radial terms /x and \ .
The details of this calculation are omitted here. The results T h e general method of solution (described in reference [17]) was

Journal of Engineering Materials and Technology JANUARY 1977 / 7

Downloaded From: http://materialstechnology.asmedigitalcollection.asme.org/ on 09/13/2013 Terms of Use: http://asme.org/terms

 CYLINDRICAL VOIDS, FULLY PLASTIC FLOW
E33=0, PLRNE STRAIN FLOW
. APPROXIMATE FUNCTIONAL FORM
* = Teqv + 2 / COSh
(f'~) 1-/2 = 0 (3.18)

+ NUMERICAL CALCULATIONS
which is the yield function. I t is shown graphically in Fig. 6
(axisymmetric flow) for a range of values of / . Note that when /
or T~,y are zero, the yield function takes on a form of the von
Mises yield function. This is quite reasonable, since the matrix
material is a von Mises material.
When the conditions of equations (3.6) and minimization of
W are not discarded, the approximate flow field becomes too
complicated for analytical analysis (see reference [17]). The in-
tegrals which define the components of 2 can be carried out
numerically, and the result is a class of yield functions which con-
form closely to the following analytic form:

$ = CWTeqv + 2 / c o s h
(T'") " l - / 2 = o (3.19)

^.00 1.50 2.00 2.50

Ceqv is an empirical function of / which varies with the direction
0.5*(TU+T22)
of E. (The subscript " e q v " indicates it is a coefficient of T 8 qv)
Fig. 5
Its upper and lower limits are

<?oqv c^ (1 + 3 / + 24/ 6 ) 2 for E» = 0, plane strain,

to expand the integrands into polynomials in ,u, integrate over
M, and then perform the radial integration. I t was found t h a t Ceqv = 1 for En = En, axial symmetry. (3.20)
for this simple flow field (equation (3.8)) the relation between
One possible contradiction which comes to mind is the case
2(, qv and S 7 T was invariant to E, to within first order in /x.
where plain strain and axial symmetry occur simultaneously,
Numerical integration of several specific cases showed that the
approximation was a good one. The analysis is outlined below leading to two separate values of C0qv. This can be resolved
for the simplest case; axisymmetric deformation. using the normal flow rule and equation (3.19):
Becasue ar0 is a constant, use the following normalization:
E33 = 0 -> T3i = - Tyy, Tu. = Tn -» r 3 S = Tn (3.21)
T = (3.12)
Co symmetry —> Ty = 0 for i 7^ j , Cartesian coordinates, Ta = Tit
- Ta —> Teqv = 0- The coefficient of C 6qv in equation (3.19) is
Due to axial symmetry, the stress system is
zero in this case, so Ceqv need not be well defined and there is no
^ U ^ J 22, Teqv =
1^33 — J^ll |, j " Y y 2 7',,. (3.13) contradiction.
In Fig. 5, equation (3.19) is compared to numerical solutions
Using equation (2.15),
of equations (2.15) for plane strain. In Fig. 6, the yield loci which
result from application of equation (3.19) to three different
J 33 —
dV, types of E fields are shown. Gev, for the intermediate curves
{E33 = 1, E' = 1) was found by solving equations (2.15) for zero
dilatation.
\-2 ( £2^-2 + . E^ dV (3.14)
3b Long Circular Cylindrical Voids—Plane Strain
The following changes of variable are used: Flow With Rigid Section
The geometry considered here is the same as in Section "da.
drdddz,

d\ for axial symmetry, x = -EX-1, g s - Ess1 (3.15)

4
CYLINDRICAL VOIDS, FDLLY PLASTIC FLOW
AXISYMMETRIC FLOW
PLANE STRAIN FLOW
Equations (3.14) can now be written I E33=E'=1. CTQV= ( U - F ) / (TEQV (TGH = 0) ) ) * K ;

'ill :: :KITI

T.V = g^E x-2(x2 + g)-i/3dx, i4Hiif ^ : : : : : : ^>.

~~~~~-^^" ***N>.
• • • . \ "v\
sit
J2_ (X3 g)-ll*dx
Tyv — + - ^ ^ ''^ A

These integrals can be found in many standard reference tables -••\ ^

(e.g., reference [21]), The results are
VI
^
Ti, - Tyy = in | —
5) J
? + E)
(3.17) \lo I 1Q k
005

jg g 1
i? and 3 can be eliminated between equations (3.17) to give an t'.oo 1.50 2.00 2.50
equation in yeqV, Tyy, and /, which is the approximate yield 0.5XIT11+T22)
function [(22], [17]). The result is Fig. 6

8 / JANUARY 1977 Transactions of the ASME

Downloaded From: http://materialstechnology.asmedigitalcollection.asme.org/ on 09/13/2013 Terms of Use: http://asme.org/terms

 (TO is both an index and an exponent.) Apply incompressibility,
specialized to plane strain.
v2
Vr.r + - (Va,a + Vr) = 0 - - (TO + 1) - . (3.23)
r
Equations (3.22) thus become
vr = amrm cos (nma), va (TO + 1) — r" sin (nma). (3.24)
nm
The rigid-plastic boundary requires t h a t vr and va be independent
of r a t a = ((7r/2) — ip). Terms with TO = 0 obviously meet this
condition. Because sin (x) and cos (x) are never both zero for
the same value of x, meeting this condition for m p* 0 requires
that the coefficient of either sin (nma) or cos (nma) vanish for
each value of TO. This is fiulfilled nontrivially for am y* 0 only
when TO = — 1. Equations (3.24) thus become
1
vT — an cos (n<,a) + a-i? --1 cos (n-ia), n-i =

Fig. 7 Cylindrical model, rigid wedge, plane strain flow, quarter

section
;-?]
- sin (noa) (3.25)

The next step is to determine no, a0, and a_i in terms of Vi and
The flow field, however, has an important difference. For the
Vi. The process is simplified by the following changes of variable:
case of plane strain (ESs = 0), part of the matrix remains rigid
while the other part undergoes plastic flow to accommodate the a a-x
V* = Vib-\ A ° A
macroscopic deformation. Also, the macroscopic boundary con- V\= Vib-\ A A 1
^ W' - VW
ditions now fall into the class of equation (2.2), rather than the
simpler class of equation (3.1). (3.26)
Plane strain deformation of a matrix containing cylindrical
The boundary conditions then give, using equations (3.25)
cavities has been studied via elastic-plastic finite elements [14,
15] and (rigid-perfectly plastic) slip line theory [13], all of which
support the idea of part of the matrix remaining rigid for some / at ViNb -> ViN = A„ViN + A-iVi" (a)
and E. The finite element solutions suggest that a radial plane
('•-)•
might be a good approximation to the rigid-plastic boundary, a t lr = b. vr = ViNb cos {\p) -*
and that the rigid region is symmetric around the principle axis
along which there is the largest absolute strain rate. r
2
In this section, primary consideration is given to the plane
strain component of z. If desired, other components as derived
n
in section 3a could be added on. A quarter section of the model
is shown in Fig. 7; wedges of rigid material are symmetric about
cos (i/0 = An cos
° (I - * )_(6) (3.27)
the (2) axis. (This model can therefore be called the "wedge" va = ViNb sin (i/0
model.) An approximate velocity field is constructed which
allows for the rigid-plastic boundary, and is of the form of equa-
sin {\p) =
A
° • r /,r / \ (c)
tion (2.19). xp is the angle of the rigid-plastic boundary to the — sin n0 I - — W I
(2) axis, and takes on its optimum value (i^opt) when W is mini- 2
i0 L V /
mized for a given E. (The minimization is carried out numerical- For a given \p, equations (3.27 b) and (c) can be solved numerical-
ly.) Stresses are calculated via equations (2.21). ly for ra0 and Ao. Note t h a t A0 is a function of \p only, and not of
This formulation leads to some interesting numerical results, Vj and V % . A-i is, from equation (3.27a), expressible in terms
but does not lead to concise derived functional forms like equa- of Ao, V\, and VN2.
tions (3.18) and (3.19). Because of this, much detail of the type Using X as defined in equation (3.10d), v can now be written as
given in Section 3a will not appear here. At the end of the
section, an empirical functional form for the yield criterion will vr = b[A0ViN cos (iioa) + A - i F ^ X " ! ' 2 cos (n-i«)],
be presented which has some success in fitting the numerical
data. va = — 6 — ViN sin (noa). (3.28)
n0
Consider the model in Fig. 7. Boundary velocity is specified
at a = 0 and 7r/2, in terms of the boundary velocities Vi and Vi. For a given \p, v, and va are linear functions of V^ and VN2.
The general procedure is to find the microscopic velocity field Noting t h a t the rigid section has the velocity field
in terms of Vh Vi, and \p, then to use equation (2.2) to obtain
Vi and Vi for a given E and \p. \j/ is then varied to locate ip0pu vr = ViN sin (a), va = ViN cos (a), (3.29)
As before, a yield locus results when S is calculated (equation
(2.21)) for a range of E. a linear homogeneous relationship between V w and E (for con-
The microscopic velocity field is constructed as follows: Start s t a n t \p) can be obtained via equation (2.2). (The surface in-
with general series solutions which obey the necessary symmetry tegral in equation (2.2) is, of course, taken over both the rigid
conditions about the (1) axis ( a = 0). and the plastic sections.) V ff , v, e, and thus W are derivable from
E for a given \p. A numerical procedure then locates Wain and
Vr amrm cos (n m a), va = dmrm sin (nma) (3.22) lAopt.
sum over the TO I t is assumed t h a t the yield function for this type of flow

Journal of Engineering Materials and Technology JANUARY 1977 / 9

Downloaded From: http://materialstechnology.asmedigitalcollection.asme.org/ on 09/13/2013 Terms of Use: http://asme.org/terms

 C Y L I N D R I C A L V O I D S , PLANE STRAIN FLOW competing solution. Because the curves in Fig. 8(6) intersect, a
FLOW WITH RIGID SECTION preferred yield locus over all stress space would consist of seg-
+ NUMERICAL CALCULATIONS
LINEAR INTERPOLATION ments of the wedge and fully plastic yield functions.
EXTRAPOLATION The flow direction predicted (via normality) by the actual
yield locus should obey the following symmetry conditions. For
zero biaxial stress (Tyy = 0), there should be no dilatation
(Eyy — 0), and for an axisymmetric stress state, the flow
should also be axisymmetric. (In plane strain, the latter cor-
responds to reqv = 0. See equation (3.21).) These two conditions
on the yield loci are satisfied by the fully plastic model, but not
by the inherently asymmetric wedge model. Therefore, if the
approximate fully plastic yield function dominates (is closer to
the origin) at Tyy = 0 and at Tgqv = 0, the models gain credibili-
ty.
This is indeed the case over the entire range of / when the yield
locus which resulted from the refined fully plastic flow field
(equation (3.19)) is used, but is not the case for the unrefined
TJ.OO 1.50 2.00 2.50 yield locus (equation (3.18)). The importance of a refinement
0.5K(T11+T22) procedure is thus underscored. Note that the wedge flow field
Fig. 8(a) was also refined by varying the angle of the rigid-plastic bound-
ary.
As noted previously, no functional form for the yield locus
C Y L I N D R I C A L V O I D S , PLANE STRAIN FLOW emerges from calculations using the wedge flow field. However,
+ DATA POINTS, FLOW WITH RIGID SECTION some success has been achieved in fitting the following empirical
LINEAR INTERPOLATION form to the data:
I 4. FULLY PLASTIC FLOW

$ =0 [Bo + B1T0B + BiTo'a ], T0B = i Tyy

(3.32)
The coefficients Bo, Bu and Bi are constant for a given/. They
were determined by fitting equation (3.32) to three numerical
data points, spaced as evenly as possible over the computed
yield locus. See Fig. 9(a).
For values of / between those for which numerical data are
available, approximate yield curves can be obtained by interpola-
tion of the Bi over /. Fig. 9(b) shows the computed values of
Bo, Bi, and Bi as functions of/, with solid lines showing the linear
interpolation.
In a paper by Nagpal, McClintock, Berg, and Subhuti [13],
1.50 2.00 2.50 plane strain slip line solutions for bands of evenly spaced cy-
0.5H(T11+T22)
lindrical cavities (long axis in the plane strain direction) are de-
Fig. 8(b)
veloped. The cross sections of the cavities considered included
slits at various angles, and circular holes. The bands have zero
extension in the transverse direction, a result of a fracture
field takes a general form similar to that in the previous section: criterion being investigated there. For their cylindrical cavities
with circular cross sections, an analogy can be made to the cy-
$ = TeQv + F (/, Tyy) = 0, (3.30) lindrical cavity models developed here. The slip line solutions
where F is some function. Therefore, for plane strain deforma- are done for various ratios of shear traction to normal traction
tion, (on the band), so with a bit of manipulation, the results gen-
erated here can be compared directly with theirs. The counter-
2V = 0 = \Tm - Z'ul (3.31) clockwise rotation from principle axes to axes in which Bu = 0
is given by
so yield loci for plane strain can be calculated without computing
the Tit.
These yield loci are shown in Figs. 8. In comparing these tan' (0) - - p (3.33)
curves with those obtained for fully plastic flow, it is immediately •"22

apparent that the wedge model gives a more negative slope in Then, where the superscript * denotes quantities in coordinates
the lower range of Tyy. By the normal flow rule, the wedge model reached by that rotation, and subscripts "n" and "s" denote
thus predicts more dilatation in this rs.nge. Considering that the "normal" and "shear,"
effect of a rigid wedge would be to inhibit contraction in the
(1) direction (see Fig. 7), this is reasonable. The wedge yield 2„ = 2j2 = sin2 (0)2 U + cos2 (0)2M
loci have the desirable property of convexity to the origin (see
2 . = 2M = sin (0) cos (0)(2 n - 222) (3.34)
Section 2).
Given the same matrix material and void geometry and the Any comparison of results requires an interpretation of the
same boundary value problem, each type of yield function will geometric parameters in reference [13] in terms of void volume
give a different solution. The correct choice is the solution which fraction /. Their parameters are P, the void radius, and L, half
gives the best upper bound (the lowest dissipation). This choice the intervoid spacing. Extending their one dimensional void
should also reflect which type of microscopic velocity field which array into a two dimensional square array, and considering the
more closely resembles the actual field. Its point in stress space (on two most likely band directions (horizontal and diagonal direc-
the "correct" yield function) will lie closer to the origin than the tions in the square array), one can say that

10 / JANAURY 1977 Transactions of the ASME

Downloaded From: http://materialstechnology.asmedigitalcollection.asme.org/ on 09/13/2013 Terms of Use: http://asme.org/terms

 LTL1NUH1LHL VOiUS, rLHINt S I IIH 1 IN r L U W
CYLINDRICAL VOIDS, PLRNE STRAIN FLOW COMPARISON OF WEDGE OATH TO SLIP LINE MODEL
FLOW WITH RIGID SECTION RIGID UEDGE <D F=0.0178 + F-0.05 » F=D.30
+ NUMERICAL CALCULATIONS
FULLY PLASTIC FLOH
POLYNOMIAL APPROXIMATIONS
OATH FROM SLIP LINE MODEL

1.00 1.50 2.00

(TN/2K)
Fig. 9(a)
Fig. 10

CYLINDRICAL VOIDS, PLANE STRAII 0.208 0.051 ~ 0.05

FLOW WITH RIGID SECTION
0.50 0.294 ~ 0.30 (3.36)
C O E F F I C I E N T S IN YIELD F U N C T I O N
• BO x Bl ® B2 As can be seen from Fig. 10, the wedge yield locus is a much
better match than the fully plastic yield locus to the results of
the slip line model, particularly in slope. (Note that the inter-
pretation of / in terms of P/L is rather arbitrary, and could be
changed considerably.) The slope is the most important factor,
because the direction of E is determined via normality. For this
reason, the similarity in slope of the slip line and wedge yield
loci is particularly satisfying, and helps to justify the develop-
ment of the wedge model.

4a Spherical Voids, Fully Plastic Flow

Si«
This void geometry, shown in Fig. 3, is meant to represent a
limit of void shape different from the long circular cylinder.
o Voids of spheroidal shape can result from decohesion or breakage
QQO
of similarly shaped second phases during deformation [1-6].
»»-» 9 Nucleation without second phases has also been seen to occur at
grain boundary misfits due to straining [18]. Sintered compacts
of metal powder can be specially prepared so that they contain
approximately equiaxed voids [19, 20]. Voids can also be present
by accident, through faulty processing.
Many parts of this analysis are similar to parts of section 3(a).
The spherical geometry is simpler to work with than the cy-
lindrical geometry because there are no preferred axes. A refine-
ment procedure analogous to that carried out for the cylinder
was carried out for the sphere; in contrast, little change in the
calculated yield loci resulted. For this reason, a refinement pro-
cedure will not be discussed here.
Fig. 9(b) Because there are no preferred directions, the approximate
microscopic velocity field will be broken up into two parts (as in
reference [9]); shape change at constant volume (v), and volume
change at constant shape (v"). The total field is then
TTP 2 e irP2 3 /j°Y
/£ (3.35) V = Vs + V" (4.1)
4W ^ ^ ' "i'(ij
The intermediate value is used in the comparisons. The E field calculated from v must, as before, be incompressible,
There are three types of data sets in Fig. 10. The points con- and y must meet external boundary conditions put in terms of
nected by solid lines and denoted by values of P/L are from the the En.
slip line model. The dashed lines and the unconnected points Vi\s = EijXj\s (Cartesian coords.) (4.2)
represent fully plastic and wedge yield functions respectively,
transformed via equations (3.34). Note that This is the same type of boundary condition used for the fully
plastic cylindrical model. It can be met by a simple incompress-
P 3 (P ible flow field which relates v« to E' (the deviatoric part of E)
v = 0.123 = 0.0178
L and v» to Etk:

Journal of Engineering Materials and Technology JANUARY 1977 / 1 1

Downloaded From: http://materialstechnology.asmedigitalcollection.asme.org/ on 09/13/2013 Terms of Use: http://asme.org/terms

 SPHERICAL VOIDS, FULLY PLASTIC FLOW SPHERICAL VOIDS, AXISYMMETRIC FLOW
+ OATH POINTS, FLOW WITH RIGID SECTION
— FIRST ORDER SOLUTION
--• SECOND OROER SOLUTION LINEAR INTERPOLATION
© Ell=E22, E33>0, AXISTMMETFlIC TENSION --- FULLY PLASTIC FLOW. FIRST ORDER SOLUTION
X E11=E22, E33<0, flXISTHHETRiC COMPRESSION

^.OO 0.50 1.00 1.50 2.00 1.50 2.00

T (HYO) (T11+T22+T33)/3
F i g . 11 Fig. 12

»«' = EijXi -> «} = Eh v? = ^ - vf = v$ = 0

6 T*

kkk = 2eee = — 2e$£,

"~ ~ 3\r) I nd& = 0, fi is a solid angle. (4.8)
Ja
ire = ir$ = «8$ = 0. (4.3)
As before, the integrands are expanded into polynomials in ju;
Adding the two components of £ gives the multiple integrals can then be carried out approximately.
Equations similar to (3.18) and (3.19) are the result. To first
e - e ! + £», ^ = E'n + - i*tfejy. (4.4) order in /J, the yield function for the spherical geometry and this
o simple flow field is
In spherical coordinates,
$ = r.J T + 2/ cosh (i Tnn \ - 1 - /« = 0 (4.9)
h = 2 = 2/lM = 2?t /l, , l, = (4-5)
" " ( ~r ) ~ ~ **' ' ' '« °' This yield function is shown in Fig. 11, along with a solution
and in cartesian coordinates, to second order in jj. and some data points resulting from numeri-
cal solution of the stress integrals. It can be seen that the first
order solution is very close to the second order solution. Unlike
ha = (5<y - 3ra,-wy) ( " ) . r2 = si + xl + x32,
the case of the cylinder, this yield function appears rather insensi-
tive to the direction of E. Given the geometric isotropy of the
_ Xi _ cartesian components of unit normal to sphere model, this is not unexpected.
of radius r. (4.6)

Carrying out equation (2.15) for this simple E field, and 4b Spherical Voids—Flow With Rigid Section
separating into deviatoric and hydrostatic components gives
The model used here is a spherical analog of the model used in
Section 3(6), and is limited to axisymmetric deformation. The
Si,'a = y j Si,ii)dV, S nn = l- f su{i)hkidV: (4.7) rigid sections are idealized as truncated circular cones capped
with spherical sections, whose axes coincide with the tensile axis.
Using s„„ = 0 and equation (4.5), one can write \p is the angle between the tensile axis and the cone wall, and is
used to optimize the flow field via minimization of W. See Fig. 4.
i r 3 The form of the calculations for this model is very similar to
2„„ = - I - Srr(i)h„dV. (4.7)
that in Section 3a, as are the solutions. Again, the calculations
do not lead to a derived functional form, but there is some
Equations (4.7) can be solved approximately in a manner very success in fitting the numerical results to an equation similar
similar to that used for the fully plastic cylindrical model. As in form to equation (3.32).
before, <r0 is presumed constant with respect to geometry.
In the model, the (3) axis is the tensile direction, and the (1)
The calculations are done in detail in reference [17], and were and (2) axes are the transverse directions which are equivalent
originally suggested in reference [22]. Some of the intermediate due to symmetry. Fig. 7 can therefore be used to illustrate im-
expressions which are similar to those in the cylindrical example portant quantities, once the index " 3 " replaces the index "2."
are
Because the similarities are so great, the reader should refer
to Section 36 (or reference [17]) for details of the calculation.
«««<; EH'EH' ~2EnnEr, Some differences with the cylindrical model should, however, be
noted. These include the incompressibility equation for the
spherical geometry (with axial symmetry):
= X = X_1 dV = r2 S i n
\b~ ) ' ' Wd<t>dddr a
rUQdr,
+ [2vr — va tan (a) + va,a] = 0. (4.10)

12 / JANUARY 1977 Transactions of the ASME

Downloaded From: http://materialstechnology.asmedigitalcollection.asme.org/ on 09/13/2013 Terms of Use: http://asme.org/terms

 Summary
SPHERICAL VOID MODEL
FLOW WITH RIGID SECTION A method has been developed for calculating approximate
COEFFICIENTS IN YIELD FUNCTION yield loci via an upper bound approach for porous ductile ma-
n BO x Bl o B2 terials. This method was applied to material models with sim-
ra
plified matrix and void geometries, and with two different types
of matrix flow field approximations. The results were two dif-
ferent types of upper bound yield functions. Comparison with
somewhat similar work by other authors was encouraging.
The approximate yield functions developed here show the im-
portant property of plastic dilatation; a property not evident in
ordinary incompressible plasticity formulations. Because the
dilatancy increases with the hydrostatic component of stress, the
yield functions (and flow rules) developed here could lead to
better understanding of plastic behavior in regions of high
hydrostatic stress (e.g., necks in sheets and bars, and near the
tips of cracks and notches).
As examined here, plastic dilatation requires that some
porosity be present. When this is not initially the case, porosity
can sometimes be nucleated during straining at second phases in
a ductile matrix or at grain boundary misfits. Nucleation at
second phases is examined in reference [17], and will be discussed
m the next paper of this series.

Acknowledgment
The author gratefully acknowledges the financial support of
the Energy Research and Development Agency under contract
AT(ll-l)-3084, and the initial support of a National Science
Foundation Traineeship during the course of the work leading to
this paper.
This work is based on the first part of the author's PhD thesis
'0.00 0.05 0.10 0.15 0.20 (1975), Brown University). The contribution of the thesis
F advisor, Professor J. R. Rice, is also gratefully acknowledged.
Fig. 13
References
1 Gurland, J., and Plateau, J., "The Mechanism of Ductile
Rupture of Metals Containing Inclusions," Transactions of the
A.S.M., Vol. 56, 1963, pp. 443-454.
When equations (3.22) are inserted, the terms sin (nma) tan (a) 2 Beachem, C. D., "An Electron Fractographic Study of the
appear. Nontrivial equations result only for special values of nm; Influence of Plastic Strain Conditions Upon Ductile Rupture
where trigonometric identities such as Processes in Metals," Transactions of the A.S.M., Vol. 56, 1963,
pp. 318-326.
sin (2a) tan (a) = 1 — cos (2a), 3 Bluhm, J. I., and Morrissey, R. J., "Fracture in a Tensile
Specimen," in International Conference on Fracture, Sendai,
Japan, 1965.
sin (4a) tan (a) = — 1 + 2 cos (2a) — cos (4a) (4.11) 4 Darlington, H., "Ductile Fracture Under Axisymmetric
a
Stresses in Electrolytic Iron and Spheroidized Low-Carbon
Pply. (In the calculations, nm is taken no larger than 4.) Be- Steel" PhD thesis, Dept. of Metallurgy and Materials Science,
cause of this, v takes on a slightly different form than in the cy- Lehigh University, Bethlehem, Pa., 1971.
lindrical case. 5 Liu, C. T., and Gurland, J., "The Fracture Behavior of
Spheroidized Carbon Steels," Transactions of the A.S.M., Vol. 61,
As before, the T field is calculated via equation (2.21.) The 1968, pp. 156-167.
yield function is assumed to have the form 6 Low, J. R., et al., "Investigation of the Plastic Fracture
of Aluminum Alloys, High Strength Steels," NASA Technical
4> = < v + T<J, TH) = 0, (4.12) Reports nos. 2, 3, 4 NGR 39-087-003, Carnegie-Mellon Univer-
sity, 1972.
where, due to axial symmetry, 7 Hill, R., The Mathematical Theory of Plasticity, The Uni-
versity Press, Oxford, 1950.
8 McClintock, F. A., "A Criterion for Ductile Fracture by
2 V = \TS3 - r „ | , TB= - ^ = - (2Tn + T3S) (4.13) the Growth of Holes," Journal of Applied Mechanics, Vol. 35,
1968, pp. 363-371.
o o 9 Rice, J. R., and Tracey, D. M., "On the Ductile Enlarge-
Values of Teqv versus TH for several values of / are shown in ment of Voids in Triaxial Stress Fields," J. Mech. Phys. Solids,
Fig. 12, and compared with yield functions derived from fully Vol. 17, 1969, pp. 201-217.
10 Kahlow, K. J., and Avitzur, B., "Void Behavior as In-
plastic flow fields. It should be noted that the symmetry argu- fluenced by Deformation and Pressure," report to the American
ments cited in section 36 do not apply in the spherical case. It Iron and Steel Institute. Lehigh University, 1969.
is therefore not expected that the fully plastic yield function will 11 Bishop, J. F. W., and Hill, R., "A Theory of the Plastic
dominate at TH = 0 for all values of /. (It does dominate, how- Distortion of a Polycrystalline Aggregate Under Combined
Stresses," Phil. Mag., Vol. 42, 1951, pp. 414-427.
ever, for very small values of /; this is to be expected.) Also 12 Berg, C. A., "Plastic Dilation and Void Interaction m
note that the yield loci which result from the "rigid cone" model the Proceedings of the Balelle Memorial Institute Symposium on
are convex. Inelastic Processes in Solids, 1969, pp. 171-209.
13 Nagpal, V., McClintock, F. A., Berg, C. A., and Subudhi,
The yield loci derived from the rigid cone model can also be M., "Traction-Displacement Boundary Conditions for Plastic
fit to equation (3.32). Calculated values of B0, Bi, and B2 are Fracture by Hole Growth" in the Intl. Symposium on Founda-
shown in Fig. 13. tions of Plasticity, Vol. 1, (A. Sawczuk, ed.), 1972, pp. 365-385.

Journal of Engineering Materials and Technology JANUARY 1977 / 13

Downloaded From: http://materialstechnology.asmedigitalcollection.asme.org/ on 09/13/2013 Terms of Use: http://asme.org/terms

 <f> = 0

\</> = 0

Fig. 14(6)
Fig. 14(a)

14 Needleman, A., "Void Growth in an Elastic Plastic Me-

dium," Journal of Applied Mechanics, Vol. 39, 1972, pp. 964-970.
w = s«jj„ =. (x„< + -3 S*A7 Via, + ^ iijit, j
15 Haward, R. N., and Owen, D. R. J., "The Yielding of a
Two Dimensional Void Assembly in an Organic Glass", Journal
of Materials Science, Vol. 8, 1973, pp. 1136-1144. = S!/^ y + I 2kkEnn (A.3)
16 Timoshenko, S. P., and Goodier, J. N., Theory of Elastic-
ity, 1934, 1951, p. 78.
17 Gurson, A. L., "Plastic Flow and Fracture Behavior of For the special case of axially symmetric distributions of the
Ductile Materials Incorporating Void Nucleation, Growth, and principle values of £ and E, this becomes
Interaction," PhD thesis, Division of Engineering, Brown Uni- 1
versity, Providence, R. I., 1975. /2 • • V
18 Bauer, R. W., and Wilsdorf, H. G. F., "Void Initiation in W = SeqvJSeav + - 2kkEnn, where E,
Ductile Fracture", Scrivta Metallurqica, Vol. 7, 1973, pp. 1213- o
1220. (A.4)
19 Edelson, B. I., and Baldwin, W. M., "The Effect of Second
Phases on the Mechanical Properties of Alloys", Transactions of The normal flow rule then gives
the A.S.M., Vol. 55, 1962, pp. 230-250.
20 Rostoker, W., and Liu, S. Y. K., "The Influence of Porosity . 3$ 3$
on the Ductility of Sintered Brass", Journal of Materials J.M.- E„„ = A
L.S.A., Vol. 5, 1970, pp. 605-617.
21 C.R.C. Standard Math Tables, The Chemical Rubber Co.,
Cleveland, Ohio.
dSe,
0s*v (A.5)

22 Rice, J. R., 1972 private communication. The flow rule is also valid in terms of T, with the factor cr0 taken
account of in A:
APPENDIX a$ -,
E.„ = A etc. (A.6)
The Normal Flow Rule and Convexity dTe
The normal flow rule arises from the definition of £ (equation Referring to Figs. 11 and 12, the ratios of the strain rate measures
(2.15)) and the fact that the dissipation W is homogeneous of is equal to the slope of the normal to the yield function:
degree one in E (equation (2.14)). These lead to
E,eqv dT..
(A.7)
SZijEi, = 0 (A.1) Enn dTH normal to O

for 5S emanating from 5 on the yield surface, and directed along

the yield surface. This means that the components of E are A very simple interpretation of normality thus results for the
proportional to the components of the normal to the yield surface spherical void model when X and E are axisymmetric.
in stress space, i.e., Suppose that the yield function is constrained to be a function
of the first two stress invariants, as in equations (3.18), (3.19),
En = A (A.2)
32,-J and (3.32) (the approximate yield function for the spherical
A is a macroscopic scalar, determined either by boundary condi- model):
tions in E, or cffi and the hardening behavior of the aggregate. $ = $ ( r e a „ Tu,f) = 0. (A.8)
See Fig. 14(a).
Convexity as well as normality can be proven when a maximum Applying the normal flow rule then gives
plastic work principle (equations (2.8), (2.17), and (2.18)) exists.
The proof is illustrated in Fig. 14(6), where a concavity in the ' d$ 3 Tn 3$
En = A +
yield surface is shown to violate the principle. JTw 2 T„qv dTkk "")
Now; consider the yield functions shown in Figsi 5, 6, 8, 9, 11,
and 12, where the axes are functions of stress other than the
• Eeav ~ A Enn = A (A.9)
tensor components. The questions to be discussed below are dTea dT„
when and how the principles of convexity and normality apply
to yield functions expressed in terms of those functions of stress. The simple interpretation is thus invariant to E when equation
In the general case, the answer lies in the proofs of convexity (A.8) is true. >
and normality in Chapter 2. They will work for any strain rate The approximate yield functions derived for the cylindrical
measure in which W is homogeneous of degree one, when the model are expressed in terms of Teqy and Tyy; Tyy is used in
stress measures are work conjugates to those strain rate measures. place of Tkk because it is a more logical choice as the driving
Acceptible stress and strain rate measures can be found by force for dilation in the' cylindrical geometry. Equation (A.3)
manipulating the expression for W. For example: can apply in this situation when the stress is constrained such

14 / J A N U A R Y 1977 Transactions of the ASME

Downloaded From: http://materialstechnology.asmedigitalcollection.asme.org/ on 09/13/2013 Terms of Use: http://asme.org/terms

 that Tyy is expressible in terms of Tkk, such as for plane stress
where (EK fin) = -Beq -" S v (A.12)
,rpi3 = 0, i = 1, 3). When the yield function takes the form
$ = ${T6m Tyy,f) = 0, (A. 10)
1 d*
so, E», i?v = A
plane strain is also in this category because by normality, S^Y ~ •"•
r
0 -)'
dT,a,

Ea = 0 -» 27, = 0 7*83 — " ™77, Tkk — (A.11)

These results can also be obtained from equation (A. 10) using
Equations (A.3) can then be written as follows, when Tti > Tn the normal flow rule (as in equations (A.8) and (A.9)), provided
and Em > En t h a t it is recognized t h a t

3Ty
W = <r» \T„ v(El2 — En) + - TyyEyy = 0 (A.13)
BTa

Journal of Engineering Materials and Technology JANUARY 1977 / 15

Downloaded From: http://materialstechnology.asmedigitalcollection.asme.org/ on 09/13/2013 Terms of Use: http://asme.org/terms