Journal of Materials Processing Technology 128 (2002) 302–312

Forming limit diagrams for kinematically hardened

voided sheet metals
A.R. Ragab*, Ch. Saleh, N.N. Zaafarani
Department of Mechanical Design and Production Engineering, Faculty of Engineering, Cairo University, Giza, 12613 Cairo, Egypt
Received 3 July 2001; received in revised form 3 July 2002; accepted 4 July 2002

Abstract

A model to predict forming limit diagrams for kinematically hardened voided sheet metals is established. The model is based on von Mises
yield criterion for a void-free matrix material, Gurson–Tvergaard yield criterion for voided solids, and a kinematic-hardening rule due to
Prager–Ziegler. A complete formulation of plasticity constitutive laws is given, e.g. yield function, flow rule and void growth law.
A computer program is developed using the above model and containing all parameters involved in the process of biaxial stretching of sheet
metals. The model which considers only rigid plastic behavior has been then applied to investigate the influence of various material
parameters on the prediction of limit strains. These include strain-hardening exponent ‘‘n’’, strain-rate sensitivity ‘‘m’’, average ratio of
normal anisotropy ‘‘
r ’’, initial void volume fraction ‘‘Cv0’’, and the initial sheet imperfection ‘‘f0’’.
Comparisons among theoretical predictions and experiments existing in the literature show that reasonable agreements are realized. This
has been done by using realistic material data for various metallic alloys of steel, aluminum, brass, and copper.
# 2002 Elsevier Science B.V. All rights reserved.

Keywords: Sheet metal forming; Forming limit diagram; Plastic instability; Kinematic hardening; Void growth

1. Introduction to the direction of maximum principal stress, as shown in

Fig. 1, represents this region. During stretching plane strain
In stamping operations, where the sheet metal is subjected condition is approached within the groove, and hence a
to biaxial stretching, the occurrence of non-uniform strains localized neck may form. The hypothesis of M–K model was
within a small region of the sheet results in the formation of elucidated by Sowerby and Duncan [2] for loading condi-
local necking leading to fracture. Limit strains are defined to tions ranging from plane strain to equibiaxial stretching of
be the strains that lie in a region near enough to the localized anisotropic sheet metals.
neck. The forming limit diagram (FLD) shows these limit Though M–K concept provides a powerful and fruitful
strains for different strain paths. Several factors affect the model, there are some discrepancies that still exist. It is
FLD, e.g. strain-hardening exponent ‘‘n’’, average normal found that in order to have acceptable agreement among
anisotropic ratio ‘‘
r ’’ and strain-rate sensitivity factor ‘‘m’’. predicted FLDs with experiments, unrealistic high values of
The sheet–tool contact conditions, such as degree and initial inhomogeneity (i.e. thickness defect) should be often
type of lubrication, punch speed, die geometry, and sheet assumed [1,3]. Another major inconsistency with experi-
thickness have also major effects. ments has been observed with predicted FLDs for sheets
Marciniak and Kuczynski [1] suggested a model (M–K with normal anisotropy
r < 1, e.g. soft aluminum and soft
model) based on von Mises constitutive law, in which 70/30 brass [4]. This led to considering constitutive laws
they assumed that a region of reduced load-carrying capacity other than von Mises—such as non-quadratic yield func-
resulting from either a microstructural or dimensional inho- tions—to calculate FLDs closer to experiments [5–7].
mogeneity. A thinning zone or simply a groove perpendicular An approach to predict FLD using M–K model has been
to combine the variation of void concentration within the
sheet metal together with the conventional thickness inho-
Abbreviations: FLD, forming limit diagram; I-H, isotropic-hardening;
K-H, kinematic-hardening
mogeneity [8–10] when the area supporting the applied load
*
Corresponding author. Fax: þ202-5703620. is considered. Void growth with strain included in the
E-mail address: a.r.ragab@link.net (A.R. Ragab). analysis was either evaluated from experiment as [8,9] or

0924-0136/02/$ – see front matter # 2002 Elsevier Science B.V. All rights reserved.
PII: S 0 9 2 4 - 0 1 3 6 ( 0 2 ) 0 0 4 7 0 - 3
 A.R. Ragab et al. / Journal of Materials Processing Technology 128 (2002) 302–312 303

Nomenclature

A cross-sectional area of sheet

c material characteristic
Cv0 initial value of void volume fraction
dQ scalar multiplier
Ep plastic tangent modulus of the material
f0 initial thickness inhomogeneity
F yield function
K strength coefficient
m, n strain-rate sensitivity and strain-hardening
Fig. 1. Schematic drawing for a biaxially stretched sheet metal having a
exponent, respectively
zone with void inhomogeneity.
qT adjusting material parameter in
Gurson–Tvergaard yield function

r average ratio of normal anisotropy
Constitutive equations were developed for several yield
t current sheet thickness
criteria, namely Gurson [12], Gurson–Tvergaard [15], Gur-
V total volume of porous solid
son–Richmond [16], Shima–Oyane [17] and Green [18].
X reduced deviatoric stress ratio
They have extended the latter to a non-quadratic form in
Y0 initial yield stress of the material
order to deal with materials possessing normal anisotropy.
For sheet metals having
r < 1, an obvious but not ideal
Greek letters
improvement in FLD predictions compared to experiments
a coordinates of yield surface center
has been attained considering reasonable values for the
(back stress)
initial void volume fraction and thickness imperfection.
a0 back stress in deviatoric stress space
The consistency has not been the same for all loading paths
e strain
and higher discrepancies were found close to plane strain
s, s0 macroscopic stress and macroscopic deviatoric
conditions [14].
stress, respectively
All the above-mentioned efforts have considered an iso-
s

macroscopic effective stress
tropically hardened (I-H) matrix material obeying von Mises
sm, sm macroscopic mean stress and reduced
yield function. Another behavior based on kinematic-hard-
macroscopic mean stress, respectively
ening (K-H) as suggested by Prager [19] and Ziegler [20]
may be also considered to describe the plastic behavior of
Subscripts
sheet metals. This behavior postulates that the yield surface
a homogeneous zone
maintains its initial size while its center translates in a
b inhomogeneity zone
certain prescribed direction as a result of plastic deforma-
i component index; i ¼ 1; 2; 3
tion. The K-H model has been already employed to predict
M matrix material
FLD by Tvergaard [21] followed by Lu and Lee [22]. They
v voids
introduced the modified Prager–Ziegler model [20] in order
0 initial value
to investigate local necking for an elastic–plastic material.
Their analysis used M–K model and a void-free material.
by using Rice and Tracey theoretical model [10,11]. Padwal Mear and Hutchinson [23] extended the work of Tvergaard
et al. [8] used an inhomogeneity index related to both void [21] by assuming a K-H model for voided material in order
volume fraction levels and thickness inhomogeneity coupled to investigate the effect of curvature of the yield surface on
with a non-linear form of Rice and Tracey void growth law. shear instability. Their constitutive model uses Gurson’s
The effect of density change on limit strains using Hill’s criterion [12] for void growth and suggests a yield func-
criterion for anisotropic materials was investigated by Kim tion to account for mixed isotropic–kinematic-hardening
and Kim [9]. Melander [10] used an expression due to behavior.
Gurson with graphically fitted parameters to include void The objective of the present work is to establish a K-H
growth in M–K analysis. model to predict FLD for voided sheet metals. The model
In order to predict FLD, Needleman and Triantafyllidis is based on three basic yield criteria namely: von Mises
[13] derived a complete model based on Gurson’s yield criterion for a void-free matrix material, Gurson–Tvergaard
criterion for voided biaxially stretched sheet metals. They criterion for voided-solids and kinematic-hardening rule
calculated limit strains due to either a given initial void due to Prager–Ziegler. The model which considers the rigid
defect or conventional thickness inhomogeneity. Ragab and plastic deformation is applied to investigate the influence
Saleh [14] have presented FLD for biaxially stretched aniso- of various material parameters on FLD and then hold
tropic sheets using different yield functions for voided-solids. comparisons with experiments for several sheet metals.
304 A.R. Ragab et al. / Journal of Materials Processing Technology 128 (2002) 302–312

2. Theoretical This condition requires that by the end of a loading incre-

ment ‘‘dsi’’, the final loading point must still lie on the
A yield function describing the behavior of anisotropic translated yield surface whose current coordinates are
voided-solid that hardens kinematically is suggested to take ðai þ dai Þ. Satisfying this condition for expression (1) gives
the form
( ð@F=@si Þ dsi þ ð@F=@Cv Þ dCv
1 dQ ¼ (4)
F¼ ½
r ½ðs01  a01 Þ  ðs02  a02 Þ 2 þ ½ðs02  a02 Þ ðsi  ai Þð@F=@si Þ
1 þ
r
) where the summation convention is applied for the recurring
letter suffix. Performing the lengthy algebra, expression (4)
 ðs03  a03 Þ 2 þ ½ðs03  a03 Þ  ðs01  a01 Þ 2
becomes
 
3sm AþBþC
þ 2qT Cv Y02
cosh  Y02 ð1 þ q2T Cv2 Þ ¼ 0 (1) dQ ¼ (5)
2Y0 D
where s0 denotes the deviatoric stress and a0 the deviatoric where
back stress.
Note that in this expression the mean stress sm is defined 2
A¼ fð
r ½ðs01  a01 Þ  ðs02  a02 Þ
in terms of the reduced stress components as 1 þ
r
sm ¼ 13 ½ðs1  a1 Þ þ ðs2  a2 Þ þ ðs3  a3 Þ (2) þ ½ðs01  a01 Þ  ðs03  a03 Þ Þ ds1
þ ð
r ½ðs02  a02 Þ  ðs01  a01 Þ þ ½ðs02  a02 Þ
Obviously expression (1) reduces to Gurson–Tvergaard
yield function [15] for ai ¼ 0, where i ¼ 1, 2 and 3, and  ðs03  a03 Þ Þ ds2 þ ½3ðs03  a03 Þ ds3 g
considering isotropy, i.e.
r ¼ 1. Moreover, Hill’s yield func-  
3sm
tion [24] for planar isotropy is also restored if the effect of B ¼ qT Cv Y0 sinh dsi
2Y0
voids is omitted, i.e. Cv ¼ 0. This yield function can be     
shown to be convex in the stress space [25,26]. It also 3sm
C ¼ 2qT Y02 cosh  qT Cv dCv
resembles that suggested previously by Mear and Hutch- 2Y0
inson [23] and reduces to that proposed by Ragab and Saleh    
3sm
for I-H [14,26]. A graphical representation of the yield D ¼ 2Y02 ð1 þ q2T C 2v Þ  2qT Cv cosh
function (1) is shown in Fig. 2. 2Y0
 
The most commonly used rule to describe the incremental 3s
þ 3qT Cv sm Y0 sinh m
translation of coordinates of the center of the yield surface, 2Y0
i.e. ‘‘dai’’ postulates that it translates according to Prager–
Again for a void-free isotropic solid, expression (4) reduces
Ziegler rule [20], namely
to the known definition of dQ for pure K-H condition [27],
dai ¼ dQðsi  ai Þ; i ¼ 1; 2; 3 (3) namely
where dQ is a positive scalar multiplier that is determined
by obeying the consistency condition, namely ðdF ¼ 0Þ. dQ ¼ 32 Y02 ½ðs0i  a0i Þ dsi (6)

The flow rule associated with the yield function (1) is

derived by assuming that normality of the incremental
plastic strain vector to the yield surface at the loading point
holds [23,25,26]. This simply states that
@F
dei ¼ dl (7)
@si
Combining expression (7) with Prager’s suggestion [19] that
dai ¼ c dei (8)
and making use of Eq. (4) results in
dQ si  ai
dl ¼ (9)
c @F=@si

where ‘‘c’’ is a positive scalar quantity to be determined

later. The final expression for ‘‘dl’’ is thus found to be

Fig. 2. Successive yield loci for equibiaxial stretching within the dQ D

dl ¼ (10)
inhomogeneity zone according to K-H model. c N
 A.R. Ragab et al. / Journal of Materials Processing Technology 128 (2002) 302–312 305

where D is given before and N is expressed by 3. Calculation procedure

 2
2 The limit strains considering only rigid plastic deforma-
N¼ fð2
r 2 þ 2
r Þ½ðs01  a01 Þ  ðs02  a02 Þ 2
1 þ
r tion are computed using the above constitutive relations for
þ ½ðs02  a02 Þ  ðs03  a03 Þ 2 þ ½ðs03  a03 Þ  ðs01  a01 Þ 2 conditions ranging from equibiaxial tension to near plane
  strain loading. First it is assumed that the sheet metal is
0 0 2 2 2 3sm
þ ½9ðs3  a3 Þ g þ 3ðqT Cv Y0 Þ sinh either free from thickness imperfection or having a very
2Y0
small thickness imperfection ‘‘f0’’. The other major source
The associated flow rule is thus explicitly expressed as of inhomogeneity is due to the presence of an initial
" void volume fraction ‘‘Cv0’’ in the zone ‘‘b’’ higher than
2 that in the rest of the sheet metal. For simplicity zone ‘‘a’’
de1 ¼ dl f
r ½ðs01  a01 Þ  ðs02  a02 Þ
1 þ
r is assumed to consist of a void-free material as shown in
  # Fig. 1.
0 0 0 0 3sm At initial yielding (corresponding to an initial strain of
þ ½ðs1  a1 Þ  ðs3  a3 Þ g þ qT Cv Y0 sinh
2Y0 0.002) the coordinates of yield surface (ai) are equal to zero,
(11a) so for a constant loading stress ratios ‘‘X’’ the initial con-
" ditions can be calculated. This stress ratio is assumed to be
2 the same as the reduced deviatoric stress ratio and hence
de2 ¼ dl f
r ½ðs02  a02 Þ  ðs01  a01 Þ represents also the incremental strain ratio (de2a/de1a) in the
1 þ
r
  # uniform zone ‘‘a’’. The stress ratio remains constant during
3sm the entire deformation process, however, it continuously
þ ½ðs02  a02 Þ  ðs03  a03 Þ g þ qT Cv Y0 sinh
2Y0 changes with deformation in zone ‘‘b’’ due to the presence
of inhomogeneity. The solution depends on calculating the
(11b) states of stress and strain inside and outside the inhomo-
    geneity due to an incremental change of strain. So an
2 0 0 3sm
de3 ¼ dl f3ðs3  a3 Þg þ qT Cv Y0 sinh appropriate constant major strain increment ‘‘de1a’’ is
1 þ
r 2Y0
assumed. Then using Eqs. (3), (5), (10), (11a)–(11c) and
(11c)
(15) the stress and strain states in zone ‘‘a’’ are completely
Now it remains to define the incremental change of void defined.
volume fraction, which is simply obtained, from the volu- Following M–K analysis [1] and due to the large size of
metric strain as the sheet in the minor direction compared to its thickness,
the minor strain increments of zones ‘‘a’’ and ‘‘b’’ are
dV dCv
dev ¼ ¼ ¼ de1 þ de2 þ de3 (12) assumed to be equal, i.e.:
V 1  Cv
From the flow rule this becomes de2a ¼ de2b (16)
 
3sm This condition is used to calculate ‘‘dlb’’ and thus from the
dCv ¼ 3 dl qT Cv Y0 ð1  Cv Þ sinh (13)
2Y0 flow rule, the strain state in zone ‘‘b’’ is totally defined. The
Combining Eqs. (11c) and (13) for equibiaxial stretching incremental void volume fraction is calculated from Eq. (13)
and integrating yields while Eq. (5) is used to determine ‘‘dQ’’, hence the incre-
mental translation of the yield center ‘‘dai’’ is found from
2 sm Eq. (3).
e3 ¼ 
1 þ
r qT Y0 sinhð3sm =2Y0 Þ As the matrix material supports the load, then the equili-
   
Cv ðCv0  1Þ 1 Cv0  1 brium condition along the first direction can be written as
 ln þ ln (14)
Cv0 ðCv  1Þ 3 Cv  1
s1Ma AMa ¼ s1Mb AMb (17)
The parameter ‘‘c’’ in ‘‘dl’’ of the flow rule is determined as
where AMa and AMb are the cross-sectional areas of the
usual from the specialization of the above expressions to
matrix material in zones ‘‘a’’ and ‘‘b’’, respectively. Rewrit-
simple uniaxial tension for which the flow curve is given by
ing Eq. (17) may be rewritten in the form
s1 ¼ Ken1
s1Ma AMb tb ð1  Cvb Þ
In such a case ¼ ¼
s1Mb AMa ta
2 ds1 2 p
c¼ or E (15) ¼ f0 ð1  Cvb Þ expðe3b  e3a Þ (18)
3 de1 3
where Ep is the tangent modulus of the flow curve at the Note that s1a ¼ s1Ma and Aa ¼ AMa since zone ‘‘a’’ is
prescribed stress. assumed to be a voided free zone. The yield function
306 A.R. Ragab et al. / Journal of Materials Processing Technology 128 (2002) 302–312

representing the matrix material is a K-H von Mises one,

i.e.:
2
F ¼ 12 f½ðs1Mb  a1Mb Þ  ðs2Mb  a2Mb Þ
2
þ ½ðs2Mb  a2Mb Þ  ðs3Mb  a3Mb Þ
2
þ ½ðs3Mb  a3Mb Þ  ðs1Mb  a1Mb Þ g  Y02 ¼ 0 (19)
Considering only the growth of homogeneously distributed
voids allows adopting the assumption of having two con-
centric yield surfaces for the matrix and the aggregate
materials, i.e.:
aiMb ¼ aib (20a)
and accordingly
daiMb ¼ daib (20b)
And hence from the translation rule of Ziegler [20]—
Eq. (3)—it is concluded that
s2b  a2b s2Mb  a2Mb Fig. 3. Comparison between FLDs predicted according to the present K-H
¼ (21) formulation and Tvergaard’s model [21].
s1b  a1b s1Mb  a1Mb
Using assumptions (20), relation (21) and yield functions (1)
and (19) the state of stress inside zone ‘‘b’’ is thus com- initial inhomogeneity in the form of voids. Such a case is
pletely defined. presented firstly by Needleman and Triantafyllidis [13] who
The instability condition is taken to be such that considered elastic–plastic behavior and secondly by Ragab
and Saleh [14]. In these two analyses a unique effective
de1a
!0 (22) stress–effective strain flow curve representing the matrix
de1b material was used, namely in the form s
m

M ¼ K
enM
e_ M . Fig. 4
at which calculations are stopped when it is satisfied. At the illustrates a favorable qualitative comparison for an initially
instant of instability, the calculated major and minor strains high Cv0 ¼ 0:01 as assigned in Ref. [13].
for zone ‘‘a’’—e1a and e2a, respectively—are representing Based on a yield function similar to (1) Mear and
the limit strains for the given strain ratio ‘‘X’’. At this instant Hutchinson [23] developed constitutive relations for isotro-
zone ‘‘b’’ suffers from localized straining under condition pic material. They postulated that for proportional stressing,
very close to plane strain. This defines a point on the FLD
associated with the assumed stress—or strain ratio—‘‘X’’.

4. Comparison with other formulations

It may be instructive to compare, at first, the predictions of

FLD using the present formulation with others existing in
the literature. Three main comparisons may be held. The first
is concerned with a kinematically hardening solid without
voids but with an initial thickness inhomogeneity. Tvergaard
[21] has presented the formulation for this case with a rather
arbitrarily selected effective stress s
¼ Y0 þ
a where s

and
a are defined similarly, e.g. s
2 ¼ ð3=2Þs0ij s0ij and
a2 ¼ ð3=2Þa0ij a0ij . By adapting the present formulation to


include these definitions, comparison between predicted
FLDs using Tvergaard’s, the present formulation and that
of M–K model is illustrated in Fig. 3. Note that for M–K
model a higher inhomogeneity (f0 ¼ 0:995 instead of 0.999)
is assigned to bring calculated limit strains within the order
of magnitude of the other formulations. Fig. 4. Comparison between FLDs predicted according to the present
The second comparison is devoted to the results obtained formulation for I-H and Needleman’s [13] and Ragab and Saleh [14]
by the present formulation reduced to an I-H model with an models.
 A.R. Ragab et al. / Journal of Materials Processing Technology 128 (2002) 302–312 307

the rate-constitutive relations has been brought into coin- Fig. 5 and compared with those predicted by the present
cidence with Gurson’s relation by choosing the H and Q (in model for different ‘‘Cv0’’, indicating that both formulations
the following relations) such that give almost coincident results.
si  ai si
¼
Y0 s

M
5. Effect of material parameters
Their flow rule is derived according to
  The matrix material in the present formulation is assumed
1 @F @F
dei ¼  dsi (23) to obey the law s1M ¼ KM en1M e_ m 1M in uniaxial tension. The
H @si @si
effect of the various material parameters namely, ‘‘n’’, ‘‘m’’,
where ‘‘H’’ is a function of the current state of deformation. ‘‘qT’’ and ‘‘Cv0’’ on limit strains are shown in Figs. 6–9. The
For pure K-H it takes the form general trends of the shown FLDs are as expected where the
(  2 limit strains increase with increasing ‘‘n’’, ‘‘m’’ and
 1 Ep @F decreasing ‘‘Cv0’’ within the selected range of variation
H ¼ ðsi  ai Þ
1  Cv Y02 @si for these parameters. Worth noting that for 0:3 < n < 0:6,
 ) the FLD changes its trend as often observed for higher
@F @F @F @F values of strain-hardening exponents [29]. The FLDs
 ð1  Cv Þ þ þ (24)
@Cv @s1 @s2 @s3 obtained by K-H formulation tend to be more sensitive
to the value of ‘‘n’’—as shown in Fig. 6—in the region
The translation rule of the yield surface is suggested as
  close to plane strain conditions than near equibiaxial con-
@F ditions. For both K-H and I-H predictions, the major limit
dai ¼ dQ ðsi  ai Þ dsi (25) strains tend to take a value equal to ‘‘n’’ as suggested by
@si
Swift diffuse instability [30]. As expected higher values
where ‘‘dQ’’ is another function of the current state. For of limit strains are obtained for higher values of ‘‘m’’ as
pure K-H it takes the form shown in Fig. 7.

f1 þ ð1=H  Þð1  Cv Þð@F=@Cv Þ½ð@F=@s1 Þ þ ð@F=@s2 Þ þ ð@F=@s3 Þ g

dQ ¼ (26)
½ð@F=@si Þðsi  ai Þ

The detailed unabridged form of ‘‘H’’ and ‘‘dQ’’ for To indicate the effect of the initial void volume fraction
combined isotropic–kinematic-hardening solid are given ‘‘Cv0’’ it is assumed that the major source of inhomogeneity
in [28]. is the presence of voids in a certain region where the rest of
Repeating the computational procedure discussed earlier the sheet consists of homogeneous material, i.e. f0 ¼ 1.
for determining limit strains, FLDs based on Mear–Hutch- FLDs illustrated in Fig. 8 show that an increase in initial
inson’s above formulation are found. These are shown in void volume fraction causes reduction of limit strain. Higher

Fig. 5. Comparison between FLDs predicted according to the present K-H

formulation and Mear–Hutchinson’s model [23]. Fig. 6. Effect of the strain-hardening exponent ‘‘n’’ on predicted FLD.
308 A.R. Ragab et al. / Journal of Materials Processing Technology 128 (2002) 302–312

Fig. 7. Effect of strain-rate sensitivity exponent ‘‘m’’ on predicted FLD. Fig. 9. Effect of the adjusting parameter ‘‘qT’’ on predicted FLD.

and probably unrealistic values of Cv0 is required to bring Note that in most of the previous investigations using the
limit strains calculated by I-H model to acceptable results, a K-H model the effect of normal anisotropy is omitted for
drawback which is not encountered by the present K-H simplicity. However, the translation of the yield surface
model. inherently implies an anisotropic effect. This is not generally
Fig. 9 indicates that the adjusting parameter ‘‘qT’’ in taken to be adequate when comparing with experiments
the modified Gurson–Tvergaard yield function has great since the sheet usually possesses an initial amount of normal
influence on limit strains. Higher values of qT, qT > 1, anisotropy which should be considered as discussed later.
result in accentuating void growth rate as expected from
Eq. (13). The maximum differences in limit strains are
noticed at equibiaxial tensile loading conditions and these 6. Comparison with experimental results
differences decrease when the loading approaches plane
strain conditions. It is essential to assign proper values to the material
parameters used in comparing theoretical predictions of
FLDs with experimental results. Void growth models are
evaluated by comparing the void growth characteristics with
experiments. This allows for adjusting the parameters,
incorporated in the modified Gurson–Tvergaard model,
namely the initial void volume fraction ‘‘Cv0’’ and the
parameter ‘‘qT’’. The rest of the material properties, e.g.
‘‘n’’, ‘‘m’’ and ‘‘
r ’’, are evaluated from flow curves of the
sheet metals.

6.1. Determination of initial void volume fraction

‘‘Cv0’’ and ‘‘qT’’

Parmar and Mellor [31] have experimentally determined

the relative density changes ‘‘Dr/r’’ occurring in some
commercial sheet metals as a result of plastic deformation as
typically shown in Fig. 10. A theoretical relation for plane
stress conditions between the thickness strain ‘‘e3’’ and the
current void volume fraction ‘‘Cv’’ is given by Eq. (14).
Fitting this equation to experimental points such as these of
Fig. 10, leads to the estimation of appropriate values for the
Fig. 8. Effect of the initial void volume fraction ‘‘Cv0’’ on predicted FLD. parameter ‘‘qT’’. It may be also suggested [4,9,14] that an
 A.R. Ragab et al. / Journal of Materials Processing Technology 128 (2002) 302–312 309

Fig. 10. Determination of the parameter ‘‘qT’’ for steel.

appropriate value of the initial void volume fraction is Fig. 12. Comparison between experimental and calculated FLD for dual-
estimated at the intersection of the fitted line with the y- phase steel according to I-H and K-H models.
axis, e.g. Cv0 ffi 104 in Fig. 10. The results of determining
‘‘Cv0’’ and ‘‘qT’’ for several sheet metals are summarized in
Table 1 together with other material parameters namely, Table 1
‘‘n’’, ‘‘m’’, and ‘‘
r ’’. Material parameters used in comparisons with experiments

Material n m
r Cv0 qT
6.2. Comparisons with experimental FLDs for various
4
sheet metals Rimming steel [4] 0.21 0.01 1.37 10 2
Dual phase steel [32] 0.15 0.01 1.04 104 2
LPG steel ‘‘B’’ [33] 0.22 0.01 0.85 104 2
Experimental FLDs for different materials found in the OFHC copper [14] 0.354 0 0.828 5  105 3.5
literature are compared with the predicted ones in Figs. 11–19. TPC copper [10] 0.265 0 0.88 104 4
They illustrate this comparison for various types of steels, Cu–Ni alloy [10] 0.23 0 0.8 5  105 3.5
copper, and aluminum alloys using I-H and K-H models. Brass [10] 0.22 0.01 0.85 5  105 3.5
Soft aluminum [4] 0.24 0.01 0.72 3  105 1.5
The major finding is the much lower predicted FLDs from Annealed Al [34] 0.23 0.008 0.56 5  105 1.8
K-H model than those obtained from I-H model. Generally

Fig. 11. Comparison between experimental and calculated FLD for Fig. 13. Comparison between experimental and calculated FLD for LPG
rimming steel according to I-H and K-H models. steel according to I-H and K-H models.
310 A.R. Ragab et al. / Journal of Materials Processing Technology 128 (2002) 302–312

Fig. 14. Comparison between experimental and calculated FLD for OFHC Fig. 16. Comparison between experimental and calculated FLD for Cu–Ni
copper according to I-H and K-H models. Alloy according to I-H and K-H models.

this leads to better agreements with experiments. The larger approach numerically that value of ‘‘n’’. For I-H model these
difference lies when the loading is near the equibiaxial values seem to be underestimated.
region. An explanation lies in the fact that the curvature The K-H model incorporates ever-changing anisotropy of
of the K-H yield surface is higher than that of I-H model thus the material during plastic deformation due to the translation
yielding to a faster localization, i.e. movement towards plane of the yield surface. This may overcome the commonly
strain conditions within the groove. This made it possible to assumed constant ratio of normal anisotropy ‘‘
r ’’ in most
reduce or even omit the initial thickness inhomogeneity cases. Comparisons for coppers and aluminum for which
in K-H predictions. Hence in some cases such as those
r < 1, the predicted FLDs from K-H model agree with
presented in Figs. 11–13 the only considered source of experiments better than other FLDs pertaining to I-H model
defect becomes the presence of voids. In addition at high as seen from Figs. 15–19. This finding may exclude the need
values of strain-hardening exponent, when loading tends to for using non-quadratic I-H yield functions for these alloys
that of plane strain conditions, K-H limit strain predictions as previously suggested [5–7,14].

Fig. 15. Comparison between experimental and calculated FLD for TPC Fig. 17. Comparison between experimental and calculated FLD for brass
copper according to I-H and K-H models. according to I-H and K-H models.
 A.R. Ragab et al. / Journal of Materials Processing Technology 128 (2002) 302–312 311

 The K-H model with void growth gives FLDs in better

agreement both in shape and magnitude with experimen-
tal results especially for materials with high strain-hard-
ening exponent than those obtained from the I-H model.
When loading tends to that of plane strain conditions, K-H
model predicts major limit strain approaching numeri-
cally the value of the strain-hardening exponent ‘‘n’’. This
is not clearly seen with I-H model especially for high
values of ‘‘n’’.
 For most materials the initial thickness inhomogeneity
factor ‘‘f0’’, using the present formulation, tends to be
unity or at most 0.998, in order to agree with the experi-
mental limit strains. This represents a major advantage
over the I-H model for which it is not uncommon to
assume values as low as 0.95 for the thickness inhomo-
geneity, which represents a non-realistic value in rolled
sheet metals. It may thus be satisfactory to assume the
presence of voids to be the only source of defect in the
Fig. 18. Comparison between experimental and calculated FLD for soft sheet in order to reach reasonable agreements with experi-
aluminum according to I-H and K-H models. ments for materials obeying the K-H model. The values
taken for initial void volume fractions are realistic and
their values are assigned according to experimental evi-
dence for different sheet metals.
 The K-H model incorporates the evolution of the material
anisotropy during deformation due to the translation of
the yield surface. This may overcome the drawbacks of
the commonly assumed constant normal anisotropy ratio.
An advantage gained by applying K-H model is that no
need to use non-quadratic yield functions to achieve better
agreement among predicted FLDs with experiments for
normal anisotropy ratios,
r < 1.

References

[1] Z. Marciniak, K. Kuczynski, T. Pokora, Influence of the plastic

properties of a material on the forming limit diagram for sheet metal
in tension, Int. J. Mech. Sci. 15 (1973) 789–805.
[2] R. Sowerby, J.L. Duncan, Failure in sheet metal in biaxial tension,
Int. J. Mech. Sci. 13 (1971) 217–229.
[3] M. Azrin, W.A. Backofen, The deformation and failure of a biaxially
Fig. 19. Comparison between experimental and calculated FLD for stretched sheet, Metall. Trans. 1 (1970) 2857–2865.
annealed aluminum according to I-H and K-H models. [4] A.K. Tadros, P.B. Mellor, An experimental study of the in-plane
stretching of sheet metal, Int. J. Mech. Sci. 20 (1977) 121–134.
[5] A. Parmar, P.B. Mellor, Prediction of limit strains in sheet metal using
7. Conclusions a more general yield criterion, Int. J. Mech. Sci. 20 (1978) 385–391.
[6] A. Graf, W. Hosford, Calculations of forming limit diagrams, Metall.
Trans. A 21 (1990) 87–94.
The predicted FLDs using the present K-H model incor- [7] A.R. Ragab, A.T. Abbas, Assessement of work-hardening character-
porating void growth have led to some conclusions based on istics and limit strains of anisotropic aluminum sheets in biaxial
comparing the results with other models or with experi- stretching, J. Eng. Mater. Technol. 108 (1986) 250–256.
mental work. [8] S.B. Padwal, R.C. Chaturvedi, U.S. Rao, Influence of superimposed
hydrostatic tension on void growth in the neck of a metal sheet in
 The qualitative effects of material parameters (n, m,
r ) on biaxial stress field. Part II. Stress instability, J. Mater. Process.
the FLDs seem to be the same for both K-H and I-H Technol. 32 (1992) 99–107.
models. However much lower FLDs are predicted from [9] K.H. Kim, D.W. Kim, The effect of void growth on the limit strain of
steel sheets, Int. J. Mech. Sci. 25 (1983) 293–300.
the K-H model than those obtained from the I-H one. The [10] A. Melander, A new model of the forming limit diagram applied to
largest differences lie when the loading is at or near the experiments on four copper–brass alloys, Mater. Sci. Eng. 58 (1983)
equibiaxial region. 63–88.
312 A.R. Ragab et al. / Journal of Materials Processing Technology 128 (2002) 302–312

[11] J.R. Rice, D.M. Tracey, On the ductile enlargement of void in triaxial [23] M.E. Mear, J.W. Hutchinson, Influence of yield surface curvature on
stress field, J. Mech. Phys. Solids 17 (1969) 201–217. flow localization in dilatant plasticity, Mech. Mater. 4 (1985) 395–407.
[12] A.L. Gurson, Continuum theory of ductile rupture by void nucleation [24] R. Hill, Theory of yielding and plastic flow of anisotropic metals,
and growth. Part I. Yield criteria and flow rules for porous ductile Proc. Roy. Soc. Lond. A 193 (1948) 198–281.
materials, ASME J. Eng. Mater. Technol. 99 (1977) 2–15. [25] A.C. Palmer, G. Maier, D.C. Drucker, Normality relations and
[13] A. Needleman, N. Triantafyllidis, Void growth and local necking in convexity of yield surfaces for unstable materials or structural
biaxially stretched sheets, ASME J. Eng. Mater. Technol. 100 (1978) elements, ASME J. Appl. Mech. 34 (1967) 464–470.
164–169. [26] A.R. Ragab, Ch.A.R. Saleh, Evaluation of constitutive models for
[14] A.R. Ragab, Ch.A.R. Saleh, Effect of void growth on predicting voided solids, Int. J. Plasticity 15 (1999) 1041–1065.
forming limit strains for planar isotropic sheet metals, Mech. Mater. [27] A.R. Ragab, S.E. Bayoumi, Engineering Solid Mechanics (Funda-
32 (2000) 71–84. mentals and Applications), CRC Press, Boca Raton, FL, 1999,
[15] V. Tvergaard, Influence of voids on shear band instabilities under pp. 697–715.
plane strain conditions, Int. J. Fract. 17 (1981) 389–407. [28] N.N. Zaafarani, Prediction of forming limit curves for kinematically
[16] O. Richmond, R.E. Smelser, Aloca Technical Center Memorandum, hardened voided sheet metals, M.Sc. Thesis, Faculty of Engineering,
1985. Cairo University, 2001.
[17] S. Shima, M. Oyane, Plasticity theory for porous metals, Int. J. [29] A. Melander, A. Thuvander, Influence of surface roughness and void
Mech. Sci. 18 (1976) 285–291. growth at inclusions on the forming limit diagram of brass, Scand. J.
[18] R.J. Green, A plasticity theory for porous solids, Int. J. Mech. Sci. 14 Metall. 12 (1983) 217–226.
(1972) 215–224. [30] H.W. Swift, Plastic instability under plane stress, J. Mech. Phys.
[19] W. Prager, A new method of analyzing stresses and strains in work- Solids 1 (1952) 1–18.
hardening plastic solids, J. Appl. Mech. 23 (1956) 493–496. [31] A. Parmar, P.B. Mellor, Growth of voids in biaxial stress fields, Int. J.
[20] H. Ziegler, A modification of Prager’s hardening rule, Quart. Appl. Mech. Sci. 22 (1980) 133–150.
Math. 17 (1959) 55–65. [32] A. Melander, E. Schedin, On the forming limit in sheet metal
[21] V. Tvergaard, Effect of kinematic hardening on localized necking in forming, Adv. Technol. Plasticity 1 (1984) 635–639.
biaxially stretched sheets, Int. J. Mech. Sci. 20 (1978) 651–658. [33] P.P. Date, K.A. Padmanabham, On the prediction of the forming limit
[22] Z.H. Lu, D. Lee, Prediction of history-dependent forming limits by diagram of sheet metals, Int. J. Mech. Sci. 34 (5) (1992) 363–374.
applying different hardening models, Int. J. Mech. Sci. 29 (1987) [34] A.R. Ragab, B. Baudelet, Forming limit curves: out-of-plane and in-
123–137. plane stretching, J. Mech. Work. Technol. 6 (1982) 267–276.