Acta Materialia 52 (2004) 3001–3007

www.actamat-journals.com

Using incremental forming to calibrate a void nucleation model

for automotive aluminum sheet alloys
W.B. Lievers a, A.K. Pilkey b,*
, D.J. Lloyd c

a
Centre for Automotive Materials and Manufacturing, 945 Princess Street, Kingston, Ont., Canada K7L 5L9
b
Department of Mechanical and Materials Engineering, Queen’s University, Nicol Hall, Kingston, Ont., Canada K7L 3N6
c
Alcan International Limited, Kingston Research and Development Centre, 945 Princess Street, Kingston, Ont., Canada K7L 5L9

Received 5 December 2003; received in revised form 18 February 2004; accepted 3 March 2004
Available online 12 April 2004

Abstract

A novel method for the quantification of void nucleation rates in sheet material is presented. An incremental sheet forming
process is employed to create large regions of homogeneous deformation, such that material density changes can then be used to
quantify the evolution of void volume fraction with applied strain. This technique is employed to calibrate the void nucleation
behaviour of three automotive aluminum sheet alloys (AA5182, AA5754 and AA6111) for incorporation into finite element method
models which employ the Gurson–Tvergaard–Needleman (GTN) constitutive softening equations.
Ó 2004 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved.

Keywords: Automotive aluminum; Nucleation; Gurson; Void damage; Incremental forming

1. Introduction ods [5–8] or Archimedes’ principle [9–13]. The funda-

mental drawback of these tests is that the necked or
In the vast majority of sheet metal forming opera- notched regions contain large gradients of stress and
tions, formability is limited by a local necking instabil- strain through the specimen cross-section. This makes
ity. Once a neck is initiated, the high hydrostatic stresses determination of the stress or strain state at any given
that develop within the neck lead to rapid void nucle- material point extremely difficult, complicating attempts
ation, growth and coalescence. However, in certain to quantify nucleation stresses or strains. Furthermore,
forming operations, such as bending and stretch flang- the exponential dependence of void growth on triaxiality
ing, necking is suppressed and the formability is con- [14] means that the accelerated growth in these regions
trolled by the evolution of void damage and shear band may overwhelm the otherwise subtle effects of void
instability [1,2]. As a result, a quantitative understand- nucleation.
ing of the void nucleation behaviour of the material is One of the most popular and convenient ways to
critical to obtain accurate predictions of these damage- model the nucleation, growth and coalescence of voids
controlled forming operations. in a continuum finite element method (FEM) formula-
A great deal of experimental work has been per- tion is with the Gurson–Tvergaard–Needleman (GTN)
formed to quantify void nucleation and growth. In constitutive softening equations. In the GTN model,
general, these studies involved the use of either smooth nucleation is most commonly seeded using the normal
or notched tensile specimens [3–6] and quantification of distribution scheme originally proposed by Chu and
the void volume fraction based on either optical meth- Needleman [15]. While a number of numerical studies
[1,15–17] have shown the importance of void nucleation
in determining forming limits, very little work has been
*
Corresponding author. Tel.: +1-613-533-6303; fax: +1-613-533- done to rigorously calibrate these model parameters for
6489/6610. real materials. A notable exception is the work of
E-mail address: pilkey@me.queensu.ca (A.K. Pilkey). Guillemer-Neel et al. [4], yet it too suffers from the

1359-6454/$30.00 Ó 2004 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved.
doi:10.1016/j.actamat.2004.03.002
3002 W.B. Lievers et al. / Acta Materialia 52 (2004) 3001–3007

problems associated with stress and strain gradients in

notched specimens.
The present work outlines a novel method for quan-
tifying void nucleation, based on an incremental sheet
forming process. The principal advantage of this method
over smooth and notched tensile specimens is that large
regions of homogeneous deformation are created.
A simple FEM model allows the GTN nucleation pa-
rameters to be calibrated from experimentally-measured
changes in material density.
Fig. 2. Geometry of an incremental forming specimen having a de-
formation angle of h ¼ 60°.
2. Experimental methods

2.1. Incremental forming 2.4 mm2 grid. The specimen edges were clamped into a
fixture such that the base of the frustum measured
Incremental forming (IF) is a rapid prototyping 200 mm
200 mm.
process, whereby a forming tool is used to trace a pat- After forming, 50 mm diameter circular sections were
tern into a sheet blank at increasing depths [18,19]. cut from a longitudinal and transverse side of each
A schematic of the process is given in Fig. 1. It should be specimen (i.e. from two adjacent faces as indicated in
noted that the apparatus employed for the current study Fig. 2), and acetone was used to remove the gridding.
did not use a back-up tool to support the deforming The circles were then hand cut into smaller pieces, using
sheet. Hence, it is referred to as unsupported IF. While a piercing saw, for insertion into the pycnometer sample
much slower than a traditional stamping operation (i.e. cup. The material around the circumference of the sec-
the geometries produced herein required roughly 30–45 tions was excluded to reduce the influence of the circular
minutes to form), IF allows very complex convex parts cutting process on the measurements.
to be created. A further advantage is that the incre- The masses and volumes of the deformed and unde-
mental nature of the deformation distributes strain very formed materials were measured using a balance (Ac-
evenly throughout the material, allowing very large culab AL-104) and a 1 cm3 helium pycnometer
strains to be achieved. (Micromeritics Accupyc 1330), respectively. Due to the
The geometry chosen for the present study is a frus- minute differences in density, multiple measurements
tum of a quadrangular pyramid [20], as shown in Fig. 2. were obtained for each specimen. A minimum of five
The advantage of such a geometry is that it provides mass measurements and five sets of volume measure-
regions of plane strain at the center of each side in both ments were taken, with each set consisting of five indi-
the transverse and rolling directions. The sample blanks vidual measurements. Ten sets of measurements were
were 240 mm
240 mm sheet specimens gridded with a taken of the undeformed material. To further improve
the accuracy of the density measurements, a calibration
volume was measured following each sample and a de-
viation of more than 0.0005 cm3 caused the preceding
values to be rejected.

2.2. Materials

Three different automotive aluminum alloys were

investigated in this work: AA5182-O, AA5754-O and
AA6111-T4P. Uniaxial tensile tests were performed and
the plastic portion of the tensile curve was fit to a Voce
[21] curve of the form:

 h
b i
rM ¼ rs  rs  ry exp  a ePM ; ð1Þ

where rM is the matrix flow stress and ePM is the effective

plastic strain. For all cases, Young’s modulus (E) and
Poisson’s ratio (m) were taken to be 70 GPa and 0.3,
respectively. The results are summarized in Table 1.
Fig. 1. Schematic of the unsupported incremental forming test Sections of each alloy were mounted, polished and
apparatus. analyzed using optical microscopy. Arrays of overlap-
 W.B. Lievers et al. / Acta Materialia 52 (2004) 3001–3007 3003

Table 1
Material descriptions for the three alloys obtained from tensile testing and the fit parameters for the Voce curve given in Eq. (1)
Alloy t (mm) R ry (MPa) rs (MPa) a b
AA5182-O 0.93 0.88 118 379 7.35 0.890
AA5754-O 1.02 0.76 93 298 7.48 0.844
AA6111-T4P 0.93 0.70 164 445 5.42 0.837

Table 2 The evolution of the void volume fraction, f , is as-

Particle measures obtained from optical microscopic evaluation in- sumed to be the result of both the growth of existing
cluding: the number of particles per mm2 , NP ; the particle area frac-
tion, AP ; and, the average particle size in lm2 , SP
voids and the nucleation of new voids as given by
Alloy NP AP SP f_ ¼ f_growth þ f_nucleation : ð3Þ
AA5182-O L 6453 0.0111 1.72 The growth of voids is a function of the plastic strain
P 7149 0.0134 1.87 rate, such that:
T 6725 0.0114 1.70
AA5754-O L 24525 0.0192 0.78 f_growth ¼ ð1  f Þ_ePkk ð4Þ
P 20106 0.0209 1.04 and, based on the work of Chu and Needleman [15],
T 25952 0.0191 0.74
AA6111-T4P L 3169 0.0097 3.06
void nucleation is assumed to follow a plastic strain-
P 2281 0.0092 4.03 controlled normal distribution specified by
T 3155 0.0092 2.92
f_nucleation ¼ A_ePM ; ð5Þ
where
ping metallographic images were acquired from each  2 !
fn 1 ePM  en
material in the longitudinal, planar and transverse sheet A ¼ pffiffiffiffiffiffi exp  ; ð6Þ
directions. These grayscale images were converted to sn 2p 2 sn
binary images using the minimum cross-entropy algo- and fn is the volume fraction of void-nucleating parti-
rithm of Li and Lee [22] combined with the cost function cles, en is the average void-nucleating strain, sn is the
proposed by Yen et al. [23], as described elsewhere [24]. standard deviation of the void-nucleating strain, and ePM
Each family of images was then ‘‘stitched’’ into a large is the effective plastic strain. It should be noted that the
compound image using custom software. In each case, parameter en is an average value in a normal distribution
the square pixel size was 0.136 lm and the total area and should not be confused with the onset of nucleation.
analyzed averaged roughly 1.5 mm2 . The particle area
fraction, the number of particles per unit area, and the 3.2. Finite element method model
average particle size were then calculated, as shown in
Table 2. While it is recognized that clustering of particles In order to calibrate the GTN nucleation parameters,
can also influence damage rates in the different sheet a single-element single-step model was employed as il-
directions [25,26], measures of particle clustering were lustrated in Fig. 3. This model is representative of the
not obtained in the current study. total deformation experienced by the central region of
the frustum side. Thus, the major true strain can be
calculated as
3. Numerical modelling
e ¼ lnðsec hÞ; ð7Þ
3.1. Gurson–Tvergaard–Needleman equations where h is the angle of deformation. The calculation of
strain using a single-step nodal projection approach has
The Gurson [27] constitutive softening equations, as been confirmed experimentally by Kim and Yang [30].
modified by Tvergaard and Needleman, have been de- Employing this single-element single-step model and
scribed in great detail in the literature [28]. Accordingly, the GTN user-material subroutine (umat) developed by
only a brief overview of the material model will be the authors [31] for LS-DYNA, several settings for the
presented here. nucleation parameters given in Eq. (6) were simulated.
The isotropic hardening yield function is defined as The parameters fn , en and sn were investigated at in-
  crements of 0.00005, 0.05 and 0.01, respectively. The
r2 q2 rkk 2
U ¼ 2e þ 2q1 f cosh  1  ð q1 f Þ ¼ 0; ð2Þ only constraint on the nucleation parameters was im-
rM 2 rM
qffiffiffiffiffiffiffiffiffiffiffi posed upon the standard deviation of strain, such that
where re ¼ 32 sij sij , sij ¼ rij  13 rkk dij , rij is the Cauchy sn 6 13 en .
stress tensor, rM is the matrix flow stress, and q1 and q2 The porosity values, f , as a function of major true
are the fitting parameters introduced by Tvergaard [29]. strain, were extracted from the FEM simulation results
3004 W.B. Lievers et al. / Acta Materialia 52 (2004) 3001–3007

Fig. 4. Agreement of measured true strains (emeas ) with those obtained

Fig. 3. Illustration of the single-step single-element FEM model. from the theoretical calculation (etheo ¼ lnðsec hÞ). Error bars indicate
one standard deviation, based on 25 measurements.

for each permutation of the three nucleation parameters.

To permit comparison with the experimental data, the
porosities were converted to densities, q, using:
q ¼ q0 ð1  f Þ; ð8Þ
where q0 is the experimentally-measured density of the
undeformed material. A least-squares assessment of the
error between predicted curves and experimental data
was used to identify the most appropriate nucleation
parameters.

4. Results

A series of frustum geometries with varying defor-

mation angles, h, were incrementally formed using the
three aluminum sheet alloys (i.e. AA5182, AA5754,
Fig. 5. Evolution of density as a function of deformation for AA5182
AA6111). Strain measurements were obtained from the with the best-fit curve for fn ¼ 0:00055, en ¼ 0:65, sn ¼ 0:08, q1 ¼ 1:5
deformed specimens using a hand-held digital strain and q2 ¼ 1. Error bars indicate one standard deviation.
measurement system.
In all cases, plane strain was achieved on the faces of
the frustum, the average minor strain being less than 5%
of the major strain. Fig. 4 shows the average major true
strain measured from each specimen plotted against the
theoretical true strain predicted based on the nodal
projection assumption made in the FEM model. The
measured values all lie very close to the theoretical line
given by Eq. (7), confirming that a projection strain
model is valid, as concluded by Kim and Yang [30].
In the vast majority of cases, very little measurable
difference in density was observed between the material
deformed in the rolling versus transverse sheet direc-
tions. This is not surprising given the planar isotropy of
the particle measures shown in Table 2. Due to this
isotropy, the nucleation parameter fitting was performed
on the combined sets of longitudinal and transverse data
points for each material. The best-fit density curves are Fig. 6. Evolution of density as a function of deformation for AA5754
shown in Figs. 5–7 for q1 ¼ 1:5 and q2 ¼ 1 in Eq. (2). with the best-fit curve for fn ¼ 0:00035, en ¼ 0:65, sn ¼ 0:03, q1 ¼ 1:5
Because the parameter en indicates the average nucle- and q2 ¼ 1. Error bars indicate one standard deviation.
 W.B. Lievers et al. / Acta Materialia 52 (2004) 3001–3007 3005

Fig. 9. Porosity predictions from the single-element model using dif-

Fig. 7. Evolution of density as a function of deformation for AA6111
ferent pairs of q1 and q2 values. The nucleation parameters are
with the best-fit curve for fn ¼ 0:00125, en ¼ 0:5, sn ¼ 0:16, q1 ¼ 1:5
fn ¼ 0:00125, en ¼ 0:5, and sn ¼ 0:16.
and q2 ¼ 1. Error bars indicate one standard deviation.

ation strain based on the normal distribution model of dently by both Becker et al. [32] and Worswick [33]. The
Chu and Needleman [15], its value is much larger than best-fit nucleation terms for all three of these combina-
the strain level at which nucleation begins. tions are presented in Table 3.
An advantage of the strain-based nucleation formu- Sections of the deformed material were also mounted
lation is that it is independent of the material flow stress. perpendicular to the direction of deformation and pol-
As demonstrated in Fig. 8, no difference is observed for ished for metallographic examination. The micrographs
the three sets of Voce parameters presented in Table 1 in Fig. 10 show representative regions of damage for the
despite their different stress-strain behaviours. This in- three aluminum sheet alloys. For all alloys, void nucle-
dependence is important because the IF deformation ation appears to be a result of particle cracking as op-
results in strains far beyond those obtained using a posed to particle-matrix decohesion. The specimens
uniaxial tensile test. were also examined for evidence of void coalescence,
The FEM predictions are very sensitive, however, to which would confound the numerical modelling as-
the selection of the q1 and q2 fitting parameters as seen sumption of only nucleation and growth; none was
in Fig. 9. In the literature, the values of q1 and q2 are observed.
generally taken to be 1.5 and 1, respectively, as sug- Accurate quantification of void volume fraction from
gested by Tvergaard [29]. Other common combinations digital images is a non-trivial task. For example, Gui-
include values of q1 ¼ 1 and q2 ¼ 1, which reduces glionda and Poole [34], while investigating damage in an
Eq. (2) to the original Gurson yield surface, or the val- Al–Si alloy, observed that optical microscopy yielded
ues of q1 ¼ 1:25 and q2 ¼ 0:95 determined indepen- lower porosity measurements than those obtained by
density measurements and they attributed these differ-
ences to the difficulties of metallographic preparation.
Accordingly, the current study did not attempt to
quantify void damage levels from the metallographic
images.

5. Discussion

A method for the quantification of void nucleation in

sheet material has been presented based on an IF tech-
nique. The principal advantage of an IF approach is that
large regions of homogeneous deformation can be easily
obtained, avoiding the high triaxialities and large gra-
dients found in typical tensile and notched tensile
methods. Furthermore, the fact that IF techniques can
Fig. 8. Porosity predictions from the single-element model using the be carried out with commercially available equipment
Voce curve parameters for each alloy given in Table 1. The GTN makes it extremely useful in either academic or indus-
parameters are fn ¼ 0:00125, en ¼ 0:5, sn ¼ 0:16, q1 ¼ 1:5 and q2 ¼ 1. trial settings. While the porosity data obtained in the
3006 W.B. Lievers et al. / Acta Materialia 52 (2004) 3001–3007

Table 3
Best-fit nucleation parameters for three pairs of qi values
Alloy q1 ¼ 1:5, q2 ¼ 1 q1 ¼ 1, q2 ¼ 1 q1 ¼ 1:25, q2 ¼ 0:95
fn en sn fn en sn fn en sn
AA5182 0.00055 0.65 0.08 0.00080 0.70 0.11 0.00075 0.70 0.10
AA5754 0.00035 0.65 0.03 0.00055 0.75 0.12 0.00045 0.70 0.10
AA6111 0.00125 0.50 0.16 0.00170 0.55 0.18 0.00165 0.55 0.18

Fig. 10. Micrographs of the three aluminum alloys showing regions of representative damage, in which the arrows indicate voids formed through
particle cracking. Major strain applied in horizontal image direction; through-thickness sheet direction is vertical.

current work was applied to the Chu and Needleman employing this technique should focus on materials
[15] void nucleation model within the GTN constitutive having larger and/or more clustered particles.
softening equations, it could easily be used to calibrate While very few published values exist for the GTN
any void damage based model. nucleation parameters of the three aluminum sheet al-
Despite the obvious advantages to the experimental loys examined herein, previous studies by Heinstein and
methods presented herein, there are a few potential Yang [36] and Brunet et al. [37] have investigated the
sources of error. Alterations to the sheet material may alloys AA5182 and AA5754, as summarized in Table 4.
have been made during pycnometer specimen prepara- These papers do not include results for AA6111, but a
tion, through the release of residual stresses in the ma- similar alloy (AA6016) is included for comparison pur-
terial during sectioning or damage due to cutting. While poses. Parameters were chosen to fit FEM predictions
every attempt was made to reduce these effects, there is with experimental load–displacement curves obtained
no way to quantify their influence apart from in situ from smooth tensile tests and no effort was made in ei-
measurements [35]. Although the accuracy of the helium ther study to experimentally quantify the void volume
pycnometer is dependent on the volume of material fraction. The values for volume fraction of nucleating
measured, no effort was made to ensure an equivalent particles, fn , in Tables 3 and 4 differ by orders of mag-
volume was tested for each specimen. The large strains nitude. In fact, the fn values of Table 4 are greater than
achieved in the IF process lead to considerable thinning the area fraction of particles, AP , obtained for these al-
of the sheet, which made extracting equivalent sample loys and presented in Table 2. Therefore, the authors
volumes more difficult. Finally, it was assumed that the feel that the current results should be viewed much more
initially undeformed material contained no porosity. reliably, despite differences in the method of deforma-
A certain amount of ‘‘unhealed’’ damage is expected tion and questions about the transferability of nucle-
from the sheet fabrication process (e.g. rolling), but it is ation parameters from one case to the other.
difficult to quantify. As a result, f0 ¼ 0 is considered a Nevertheless, more rigorous methods must be employed
reasonable assumption. to determine values of fn , rather than simply using the
The limitations of the method must also be recog- GTN model phenomenologically.
nized. A minimum level of void damage is needed to
obtain a statistically significant measure with the pyc- Table 4
nometer system employed herein. In two of the alloys Nucleation parameters obtained by [36] and [37]
(AA5182 and AA6111), this threshold is met, producing Alloy fn en sn Reference
significant changes in density as a function of applied AA5182 0.08 0.2 0.075 [36]
deformation. However, in the AA5754 sheet, the level of AA5182 0.035 0.270 0.1 [37]
damage is very small and a trend is not clearly estab- AA5754 0.034 0.320 0.1 [37]
lished. Since void nucleation rate is observed to increase AA6016 0.036 0.276 0.1 [37]
with increasing particle size (see Table 2), future work Values of q1 ¼ 1:5 and q2 ¼ 1 were used in all cases.
 W.B. Lievers et al. / Acta Materialia 52 (2004) 3001–3007 3007

Although the present study has focused exclusively of Alcan International Limited (Alcan KRDC), the
on nucleation as a function of plastic strain, a great deal Natural Sciences and Engineering Research Council of
of evidence in the literature suggests that stress-con- Canada (NSERC) and the Centre for Automotive Ma-
trolled nucleation models [4,5,38] are more appropriate. terials and Manufacturing (CAMM). They also wish to
Indeed, the current results exhibit an increased void thank Dr. J. Jeswiet of Queen’s University for producing
nucleation rate with increasing yield strength, suggesting the IF specimens, and H. Buie (Queen’s) for preparing
a stress dependence. The decision was made to limit the and imaging the metallographic samples.
study to strain-controlled nucleation based on the ease
with which experimental strains could be obtained and
robustly compared with the FEM results. Furthermore, References
as discussed in Section 4, it is unclear how best to
quantify the flow stress behaviour of the material at very [1] Lievers WB, Pilkey AK, Lloyd DJ. Mater Sci Eng A 2003;361:312.
large strains. Although hydraulic bulge tests could [2] Chen ZT, Worswick MJ, Cinotti N, Pilkey AK, Lloyd DJ. Int J
Plast 2003;19:2099.
provide additional information, they would not produce [3] Beremin F. Metall Trans A 1981;12:723.
strains as large as 70% or 80%. The authors are currently [4] Guillemer-Neel C, Feaugas X, Clavel M. Metal Mater Trans A
investigating methods for estimating the flow stress, so 2000;31:3063.
that stress-controlled nucleation can be explored. Nev- [5] Kwon D, Asaro RJ. Metal Mater Trans A 1990;21:117.
ertheless, the excellent agreement obtained for all three [6] Agarwal H, Gokhale AM, Graham S, Horstemeyer MF. Mater
Sci Eng A 2003;341:35.
alloys suggests that a plastic-strain controlled condition [7] Cox T, Low J. J Metall Trans 1974;5:1457.
is appropriate for this mode of deformation. [8] Brownrigg A, Spitzig W, Richmond O, Teirlinck D, Embury J.
The normal distribution proposed by Chu and Nee- Acta Metall 1983;31:1141.
dleman [15] has received a great deal of acceptance in [9] Barlat F, Jalinier J. J Mater Sci 1985;20:3385.
the literature, yet a physical basis for this model has [10] Br
athe L. Scand J Metall 1978;7:199.
[11] Rosochowski A, Olejnik L. Int J Mech Sci 1988;30:51.
never been established. Given the effect of particle size [12] Spitzig WA. Acta Metall Mater 1990;38:1445.
and particle clustering on void nucleation, the correla- [13] Koenigsmann HJ, Starke Jr EA, Allaire PE. Acta Mater
tion between these parameters should be investigated in 1996;44:3069.
greater detail. [14] Rice JR, Tracy DM. J Mech Phys Solids 1969;17:201.
A final issue that remains to be addressed is whether [15] Chu C, Needleman A. J Eng Mater Technol 1980;102:249.
[16] Needleman A, Rice JR. In: Koistinen D, Wang NM, editors.
the void nucleation parameters determined using the IF Mechanics of sheet metal forming: material behaviour and
method are transferable to other forming operations. deformation analysis. New York: Plenum Press; 1978. p. 237–67.
Namely, the IF produces a through-thickness shearing [17] Saje M, Pan J, Needleman A. Int J Fract 1982;19:163.
form of deformation, whereas many forming operations [18] Shim MS, Park JJ. J Mater Process Technol 2001;113:654.
produce in-plane biaxial strains and/or bending. As a [19] Park JJ, Kim YH. J Mater Process Technol 2003;140:447.
[20] Iseki H. J Mater Process Technol 2001;111:150.
result, the authors are also investigating the application [21] Voce E. J Inst Met 1948;74:537.
of these parameters to the prediction of porosity devel- [22] Li CH, Lee CK. Pattern Recogn 1993;26:617.
opment in stretch flange forming operations [2]. [23] Yen JC, Chang FJ, Chang S. IEEE Trans Image Process
1995;4:370.
[24] Lievers WB, Pilkey AK. Mater Sci Eng A, accepted for publica-
6. Conclusions tion.
[25] Thomson CIA, Worswick MJ, Pilkey AK, Lloyd DJ, Burger G. J
Density measurements, obtained from a series of in- Mech Phys Solids 1999;47:1.
[26] Thomson CIA, Worswick MJ, Pilkey AK, Lloyd DJ. J Mech Phys
crementally formed specimens, have been used to cali- Solids 2003;51:127.
brate the nucleation terms for the GTN constitutive [27] Gurson AL. J Eng Mater Technol 1977;99:2.
softening equations. This novel sheet forming technique [28] Tvergaard V. Adv Appl Mech 1990;27:83.
creates large regions of homogeneous deformation, and [29] Tvergaard V. Int J Fract 1981;37:389.
avoids the large stress and strain gradients associated [30] Kim TJ, Yang DY. Int J Mech Sci 2000;42:1271.
[31] Lievers WB, Pilkey AK, Worswick MJ. Mech Mater 2003;35:661.
with smooth or notched tensile specimens. The plastic [32] Becker R, Needleman A, Richmond O, Tvergaard V. J Mech Phys
strain-controlled nucleation parameters for three auto- Solids 1988;36:317.
motive aluminum sheet alloys (AA5182, AA5754, and [33] Worswick MJ. Cavity growth and constitutive softering in ductile
AA6111) have been fit to the experimental data using a solids. Ph.D. Universtiy of Waterloo, 1988.
single-step single-element model. [34] Guiglionda G, Poole WJ. Mater Sci Eng A 2002;336:159.
[35] Babout L, Maire E, Buffiere J, Fougeres R. Acta Mater
2001;49:2055.
Acknowledgements [36] Heinstein MW, Yang HTY. Int J Numer Meth Eng 1992;33:317.
[37] Brunet M, Mguil S, Morestin F. J Mater Process Technol 1998;80-
81:40.
The authors wish to acknowledge the financial sup- [38] Fowler J, Worswick M, Pilkey A, Nahme H. Metal Mater Trans
port of the Kingston Research and Development Centre A 2000;31:831.