You are on page 1of 13

Journal of Materials Processing Technology 128 (2002) 178±190

Wrinkling predictions in the deep-drawing process


of anisotropic metal sheets
JoaÄo Pedro de MagalhaÄes Correia, GeÂrard Ferron*
Laboratoire de Physique et MeÂcanique des MateÂriaux (LPMM), ISGMP, Ile du Saulcy, 57045 Metz Cedex 01, France
Received 9 August 2001; received in revised form 3 January 2002; accepted 26 June 2002

Abstract

The onset of wrinkling in sheet metals is ®rst analysed as an elastic±plastic bifurcation for thin and shallow shells with compound
curvatures. Plastic yielding is described using a criterion recently proposed for transversely anisotropic materials. A local analysis is
developed, which allows us to de®ne wrinkling limit curves depending on material properties and local geometry. Finite element (FE)
simulations of the conical cup test are also performed using the Abaqus/Explicit code. The FE results relating to the initiation of wrinkling in
the wall are compared with the predictions of the bifurcation model. In addition to the intrinsic effect of anisotropy on wrinkling tendencies, it
is shown that the attainment of critical wrinkling conditions is signi®cantly affected by the in¯uence of anisotropy on the stress state and sheet
curvature developed in the wall prior to bifurcation.
# 2002 Elsevier Science B.V. All rights reserved.

Keywords: Wrinkling; Bifurcation; Sheet metal forming; Anisotropy; Finite element simulations

1. Introduction ®eld where contact conditions and continuity conditions are


disregarded, which allows the derivation of critical stresses
Wrinkling is, together with necking, tearing and spring- that only depend on material properties and local geometry.
back, one of the most deleterious defects arising in sheet Wrinkling limit diagrams, de®ned in principal stress axes,
metal forming operations. Predictive calculations are thus can be obtained with this procedure [10]. These diagrams
highly desirable to reduce as much as possible lengthy and are reminiscent of the well-known forming limit diagrams
costly trial and error procedures aiming at removing the associated with the onset of necking, with the only differ-
formation of wrinkles, as well as to prevent blank destruc- ence that these latter are generally presented in principal
tion and tool damage in the manufacturing process. strain axes [11±13].
The formation of wrinkles can be predicted by analytical Nordlund and HaÈggblad [14] de®ned a wrinkle indicator
treatments using Hill's [1] bifurcation criterion, specialised based on the principle of virtual displacements, and imple-
by Hutchinson [2] to the case of thin and shallow plates and mented it in the explicit ®nite element (FE) code Dyna3d.
shells. One of the ®rst treatments performed in the context of Generally, a bifurcation algorithm is introduced in commer-
elastic±plastic stamped sheets is the study by Triantafyllidis cial FE codes. For instance, a modi®ed Riks method is
and Needleman [3], who analysed the onset of ¯ange wrink- implemented in the Abaqus code [15]. Such implementa-
ling in the Swift cup test by assuming an admissible bifurca- tions have been also performed in speci®c FE codes to
tion displacement ®eld in the region between the die and the analyse the in¯uence of material anisotropy on wrinkling
blankholder. Several further works [4±7] were devoted to the tendencies [16].
determination of critical wrinkling stresses, for isotropic The effect of plastic anisotropy is actually quite important
shells obeying the J2 ¯ow, or deformation theory of plas- in sheet metal forming processes. In particular, the shape of
ticity and for transversely isotropic shells following Hill's the yield surface strongly affects the strain distributions
[8] quadratic criterion or Hill's [9] non-quadratic criterion. obtained during stamping [17,18], or the onset of necking
All these studies assume a local bifurcation displacement [19]. The aim of this paper consists of investigating the
in¯uence of plastic anisotropy on the onset of wrinkling.
*
Corresponding author. Tel.: ‡33-3-87-31-53-60;
The analysis is made by using the plane-stress yield surface
fax: ‡33-3-87-31-53-66. proposed by Ferron et al. [20]. This description of yielding
E-mail address: ferron@lpmm.univ-metz.fr (G. Ferron). is referred to as Ferron's criterion in the following.

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 4 8 - X
J.P. de MagalhaÄes Correia, G. Ferron / Journal of Materials Processing Technology 128 (2002) 178±190 179

This criterion is ®rst introduced in a bifurcation analysis for where g…y; a† represents the adimensional length of the
a doubly curved shell, which allows us to de®ne wrinkling radius to a point on the yield locus normalised by s . This
limit curves that only depend on material properties and polar-coordinate description of the yield surface is sketched
local geometry (sheet curvature). in Fig. 1.
Based on previous works [17,18] in which Ferron's As a ®rst step, a yield function g(y) for transversely
criterion was introduced in the Abaqus FE code, the analy- isotropic behaviour is de®ned. This function is chosen as
tical predictions obtained with the bifurcation analysis have an extension of Drucker [21] isotropic criterion, given by
been compared with the results from numerical simulations. 6
This comparison is made for the conical cup test, which …1 k†g…y† ˆ …cos2 y ‡ A sin2 y†3
represents a good check of wrinkling predictions. k cos2 y…cos2 y B sin2 y†2 (7)
where A and B are positive constants (A ˆ 3 and B ˆ 9 for
2. Wrinkling analysis Drucker's criterion). For transverse anisotropy, an expres-
sion of g…y; a† that satis®es a number of consistency con-
2.1. Constitutive equations ditions for orthotropic plasticity can be obtained as an
extension of g(y), de®ned by
The constitutive equations for elastic±plastic materials are g…y; a† m
ˆ g…y† m
2a sin y cos2n 1 y cos 2a
handled in a classical way to express the tangent stiffness
tensor L relating the stress increments s_ ij to the total strain ‡ b sin2p y cos2q 2a (8)
increments e_ kl , i.e. where m, n, p and q are positive integers and a and b are
s_ ij ˆ Lijkl e_ kl (1) dimensionless parameters characterising transverse aniso-
tropy. Denoting by r the coef®cient of normal anisotropy,
The general procedure for calculating the tangent stiffness
Hill's quadratic criterion for transverse isotropy is retrieved
components is presented in Appendix A, for any yield
with k ˆ 0 and A ˆ 1 ‡ 2r in Eq. (7). Hill's quadratic
surface used in association with the assumption of isotropic
criterion for transverse anisotropy is retrieved with k ˆ 0,
hardening. The derivation of the tangent stiffness compo-
A ˆ 1 ‡ 2r45 , m ˆ 2 and n ˆ p ˆ q ˆ 1 in Eq. (8).
nents is made for both ¯ow and deformation theory, and it is
As discussed in more detail in [17,20], this criterion can
particularised to plane-stress conditions. Finally, the calcu-
provide signi®cant improvements in the description of
lations are developed using Ferron's yield criterion, which
yielding. In particular, a positive k-value in Eq. (7) leads
applies for transversely anisotropic materials submitted to
to a ¯attening of the yield surface near pure shear and plane-
plane-stress loading. This description of yielding is brie¯y
strain tension or compression. Also, for transversely aniso-
recalled now.
tropic materials, a small orientation dependence of yield
The stress state is speci®ed by the principal stress com-
stresses s(a) can be obtained along with a sharp orientation
ponents (s1, s2) and by the angle a de®ning the orientation of
dependence of the anisotropy coef®cient r(a) by using
the principal stress directions (s1, s2) with respect to the
exponents n and p larger than 1 in Eq. (8). Finally, the
principal in-plane orthotropic axes (x1, x2). The yield func-
¯exibility of the criterion often permits a better account of
tion f is thus de®ned by
yield surfaces determined by experiments or by micro±
f ˆ F…s1 ; s2 ; a† 
s (2) macro computations. For steels, and when only uniaxial
where s is the effective stress. Using the change of variables tension tests are available, a good approximation of material
de®ned by x1 ˆ …s1 ‡ s2 †=2 and x2 ˆ …s1 s2 †=2, the behaviour is often obtained by setting: k ˆ 0:2, B ˆ 3A,
yield function can be re-expressed as m ˆ n ˆ p ˆ 2 and q ˆ 1. The material parameters A, a and
b are then determined so as to ®t the experimental values of
f ˆ c…x1 ; x2 ; a† 
s (3)
r0, r45 and r90.
Finally the yield surface, f ˆ 0, is described in parametric The description of elastic±plastic behaviour is completed
form by by the speci®cation of the hardening law between effective
 and effective plastic strain ep . A Ludwik form will be
stress s
†;
x1 ˆ x1 …y; a; s †
x2 ˆ x2 …y; a; s (4)
used, which is de®ned by
The parameter y in (4) is taken as the polar angle associated  ˆ s0 ‡ Kep
s
n
(9)
with the polar-coordinate description de®ned by
where s0, K and n are material constants.
† cos y;
x1 ˆ G…y; a; s † sin y
x2 ˆ G…y; a; s (5)
† represents the length of the radius to a point
where G…y; a; s 2.2. Bifurcation analysis
on the yield locus in (x1, x2) axes. Assuming an isotropic
expansion of the yield surface, we have A sheet element is characterised at the current stage of
loading by the principal radii of curvatures R1 and R2, by
† ˆ s
G…y; a; s g…y; a† (6) the thickness t and by the stress state applied to the element.
180 J.P. de MagalhaÄes Correia, G. Ferron / Journal of Materials Processing Technology 128 (2002) 178±190

Fig. 1. Sketch of the yield surface in principal stress space (s1, s2). The yield loci are parameterised by the angle a between the principal in-plane orthotropic
axes (x1, x2) and the principal stress directions (s1, s2).

The different axes and orientations are de®ned in Fig. 2. The


direction normal to the sheet is x3. The angle between
principal in-plane orthotropic axes (x1, x2) and principal
curvature axes (k1, k2) is Z. The principal stress directions
(s1, s2) are inclined at an angle a from the orthotropic axes.
Finally, wrinkling is assumed to occur along one direction,
w1 , inclined at the angle d from x1.
As in several previous studies [4±7] the Donnell±Mush-
tari±Vlasov shallow shell theory is assumed, which limits
the analysis to values of the characteristic wrinkle wave-
length that are large compared to the sheet thickness t,
though small compared to the sheet radii of curvature
(R1, R2). The bifurcation functional derived by Hutchinson
[2] can then be used to determine the critical stress state for
wrinkling. The functional expresses as
Z  
t3
F…u_ i ; u_ 3 † ˆ Lijkl K_ ij K_ kl ‡ tLijkl E_ ij E_ kl ‡ Nij u_ 3;i u_ 3;j dS
s 12
(10)

where …u_ i ; u_ 3 † ˆ …u_ 1 ; u_ 2 ; u_ 3 † de®nes the incremental bifurca-


tion displacement ®eld. The incremental stretching strains
E_ ij and bending strains K_ ij that occur at the instant of
wrinkling in addition to the incremental strains associated
with the fundamental solution are de®ned by

E_ ij ˆ 12 …u_ i;j ‡ u_ j;i † ‡ bij u_ 3 (11)


K_ ij ˆ u_ 3;ij (12)
Fig. 2. Axes and angles defining sheet geometry, loading and wrinkling.
J.P. de MagalhaÄes Correia, G. Ferron / Journal of Materials Processing Technology 128 (2002) 178±190 181

where bij denote the curvature components. Stress resultants tension. However, wrinkling was never attained in the simu-
Nij are given by lations in the biaxial stretching range. The results are there-
fore limited to the quadrant de®ned by (s1  0; s2  0).
Nij ˆ tsij (13) The effects of sheet geometry and loading orientation are
®rst examined. The plastic behaviour is assumed here to be
Finally, the incremental wrinkling displacement ®eld is
described by the J2 ¯ow theory. For isotropic or transversely
assumed to be de®ned in …w1 ; w2 ; x3 † axes by
isotropic materials, the in-plane orthotropic axes are unde-
    ®ned, and the orientation of loading is fully speci®ed by the
l l
u_ 1 ˆ Bt sin w1 ; u_ 2 ˆ 0; u_ 3 ˆ At cos w1 angle …a Z† between the principal curvature axes and
l l
principal stress axes. The wrinkling limit curves obtained
(14) for different geometries and orientations are shown in Fig. 3.
p The normalised characteristic wavelength, L/R2, is drawn in
where l is an adimensional wavenumber, and with l ˆ Rt,
where the radius R takes the value R1 or R2. The constants Fig. 4 for Case c (R1 ˆ 110 mm, R2 ˆ 213 mm). The fol-
A and B represent the relative displacement amplitudes of lowing observations can be made, with some exceptions
the wrinkling mode. when the stress is not codirectional with the principal
The calculations relating to the function F (Eq. (10)) are curvature axes:
given in Appendix B. This function can be expressed in the (1) The shape of the wrinkling limit curve is roughly the
matrix form as same in all cases, except for Case b with a Z ˆ 45 ,
where the wrinkling condition is found only for stress
F ˆ 12 St3 u‰MŠut (15)
states close to uniaxial compression.
where S is the sheet surface over which wrinkling spans and (2) The stress level at which wrinkling occurs increases
u ˆ …A; B† is the displacement amplitude vector. The wrink- as the radius of curvature normal to the compressive
ling condition, F ˆ 0, is obtained by setting the determinant direction decreases. In contrast, the wrinkling limit
M equal to 0, and by minimising M with respect to the curve is independent of the radius along the compres-
wavenumber l. Thus, wrinkling is obtained when sive direction, as can be seen for instance by comparing
@ det‰MŠ (Case a, a Z ˆ 0 ) and (Case b, a Z ˆ 0 ), (Case a,
det‰MŠ ˆ 0; ˆ0 (16) a Z ˆ 90 ) and (Case c, a Z ˆ 0 ), or (Case b,
@l
a Z ˆ 90 ) and (Case c, a Z ˆ 90 ).
The critical wrinkling stress and the wavenumber l are (3) The wrinkling direction is along the compressive
derived from (16). The most favourable orientation for direction for a Z ˆ 0 and 908. For a Z ˆ 45,
wrinkling is ®nally obtained by determining numerically the most favourable orientation for wrinkling depends
the lowest value of |s1| as a function of d. The wavelength of on the stress state and on the curvature radii. The
the wrinkles is given by disorientation between the wrinkling direction and the
  compressive direction is larger for widely different
l
L ˆ 2p (17) curvature radii.
l
(4) The normalised characteristic wavelength decreases as
the compressive stress |s2| increases, i.e., as the loading
varies from uniaxial tension in direction s1 to uniaxial
3. Wrinkling predictions
compression in direction s2. For a given stress ratio (i.e.
for a given value of angle y on the yield surface, see
The wrinkling limit curves obtained from the analysis
Fig. 1) the wavelength is smaller when the radius of
described in Section 2.2 have been determined by assuming
curvature along the direction normal to the compressive
that the elastic behaviour is characterised by the values
stress is smaller.
of Young's modulus, E ˆ 210,000 MPa, and of Poisson's
ratio, n ˆ 0:3. The effective plastic stress±strain law is given The wrinkling limit curves determined in Case a with the
by (9) with s0 ˆ 100 MPa, K ˆ 400 MPa and n ˆ 0:3. The J2 deformation theory are drawn in Fig. 5. As is amply
thickness of the sheet is equal to 0.7 mm. Three (R1, R2) documented in the literature, the critical stresses obtained
pairs are considered. Case a (R1 ˆ 10 mm, R2 ˆ 100 mm) with this theory are substantially lower than those deter-
and Case b (R1 ˆ 10 mm, R2 ˆ 290 mm) can be associa- mined with ¯ow theory (Fig. 3). The discrepancy is less
ted with the geometry along a punch radius and a die radius, important for a Z ˆ 90. In that case the critical stresses
respectively, and Case c (R1 ˆ 110 mm, R2 ˆ 213 mm) with are the lower, owing to the fact that the radius of curvature
the unsupported region between a punch and a die. The along the direction normal to the compressive stress is
above values are inferred from the geometry developed in the larger. Despite the better ®t of experiments obtained
the conical cup test. with deformation theory in some bifurcation problems [2]
The condition of wrinkling was searched a priori in the the analytical predictions and the numerical simulations
stress range between uniaxial compression and equibiaxial (Section 4) will be carried on with ¯ow theory.
182 J.P. de MagalhaÄes Correia, G. Ferron / Journal of Materials Processing Technology 128 (2002) 178±190

Fig. 3. Wrinkling limit curves obtained with the J2 flow theory for three geometrical conditions: (a) R1 ˆ 10 mm, R2 ˆ 100 mm; (b) R1 ˆ 10 mm,
R2 ˆ 290 mm; (c) R1 ˆ 110 mm, R2 ˆ 213 mm, and for three orientations of loading defined by a Z ˆ 08, 458 and 908.

In many practical situations where deep-drawing pre- orientation of loading is thus appropriately de®ned by
dominates, the compressive stress s2 is along the hoop a Z ˆ 0. This coincidence of principal directions of stress
direction, and the tensile stress s1 along the radial direction. and curvature is exactly (respectively, approximately) satis-
For an axisymmetrical process where R1 and R2 denote ®ed for transversely isotropic (respectively, transversely
the meridional and circumferential radii, respectively, the anisotropic) materials. Therefore, the effect of material
anisotropy is examined now for a Z ˆ 0.
The yield surfaces of transversely isotropic materials are
drawn in Fig. 6 for different values of the normal anisotropy

Fig. 4. Wrinkle wavelengths obtained with the J2 flow theory for R1 ˆ


110 mm, R2 ˆ 213 mm, and for three orientations of loading. EBT:
equibiaxial tension; UT: uniaxial tension; PS: pure shear; UC: uniaxial Fig. 5. Wrinkling limit curves obtained with the J2 deformation theory for
compression. R1 ˆ 10 mm, R2 ˆ 100 mm, and for three orientations of loading.
J.P. de MagalhaÄes Correia, G. Ferron / Journal of Materials Processing Technology 128 (2002) 178±190 183

Fig. 6. Normalised yield surfaces for transversely isotropic materials defined with Ferron's model (with k ˆ 0:2 and B ˆ 3A in Eq. (7)).

coef®cient r using Ferron's model (with k ˆ 0:2 and B ˆ 3A and that the yield stresses in the deep-drawing range are
in Eq. (7)). The wrinkling limit curves obtained in Case c are higher for smaller r-values (Fig. 6). In the stress range of
plotted in Fig. 7. The limit stresses are higher for smaller interest, a material having a low r-value is thus harder, and
r-values. It should be noted that the stress±strain curve is more resistant to wrinkling, than a material with a high
chosen to be the same for all materials in uniaxial tension, r-value. In comparison with the predictions obtained with

Fig. 7. Wrinkling limit curves obtained with flow theory for R1 ˆ 110 mm, R2 ˆ 213 mm, a Z ˆ 0, for the transversely isotropic materials defined in
Fig. 6.
184 J.P. de MagalhaÄes Correia, G. Ferron / Journal of Materials Processing Technology 128 (2002) 178±190

Fig. 8. Normalised yield surfaces for a transversely anisotropic material with r0 ˆ 2, r45 ˆ 1:5, r90 ˆ 3. The material parameters are defined by k ˆ 0:2,
B ˆ 3A, m ˆ n ˆ p ˆ 2, q ˆ 1, A ˆ 3:69, a ˆ 0:12 and b ˆ 0:72 in Eq. (8).

Hill's quadratic criterion (not shown here) this effect of r is curvature- and stress-axes. An illustration of the results is
slightly weaker with Ferron's model, owing to the fact that given here for a material with a minimum of the anisotropy
the yield surface is less dependent on r in that case. coef®cient r(a) for 0  a  p=2 (r0 ˆ 2, r45 ˆ 1:5, r90 ˆ 3).
For transversely anisotropic materials, the predictions are The material parameters of the yield function are determined
still presented for a Z ˆ 0, but by specifying now the from Eq. (8), with k ˆ 0:2, B ˆ 3A, m ˆ n ˆ p ˆ 2 and
angle a ˆ Z between in-plane orthotropic axes and principal q ˆ 1. The other material parameters (A, a and b) are then
determined to ®t the prescribed r-values. The yield surfaces
are drawn in Fig. 8. The wrinkling limit curves for
a ˆ Z ˆ 08, 458 and 908 (Fig. 9) show that the limit stresses
are higher when the loading is applied at 458 from the
orthotropic axes, i.e., along the directions of smaller r-values.
These directions correspond to higher yield stresses (Fig. 8),
and the material is therefore, stiffer when it is loaded at 458.

4. Comparison with FE simulations of


the conical cup test

The dimensions of the tooling and of the blank are


assumed to be de®ned by the following values: punch radius,
50 mm; punch pro®le radius, 6 mm; die opening radius,
100 mm; die pro®le radius, 10 mm; initial blank radius,
150 mm; initial sheet thickness, 0.7 mm. These dimensions
are those used in the work by Ellagoune and Boivin [22]. For
the purpose of comparison, 11 wrinkles were observed by
these authors around the wall in experiments on mild steel.
Numerical simulations have been performed with the
Fig. 9. Wrinkling limit curves obtained with flow theory for R1 ˆ 110 mm,
Abaqus/Explicit FE code, using quadrangular shell elements
R2 ˆ 213 mm, a ˆ Z, for the transversely anisotropic material defined in with ®ve integration points through the thickness, denoted
Fig. 8. S4R. Two FE meshes have been used (Fig. 10).
J.P. de MagalhaÄes Correia, G. Ferron / Journal of Materials Processing Technology 128 (2002) 178±190 185

process and material symmetries, the FE results obtained


with one-quarter of the structure were compared with those
for the full blank, and also with those obtained by introdu-
cing a ``material defect'' in the form of a slight transverse
anisotropy, de®ned by r0 ˆ 0:99, r45 ˆ 0:98 and r90 ˆ 1:05.
For a given mesh, no signi®cant differences were observed
between these various simulations. Likewise, the initiation
and growth of wrinkles was not modi®ed by re®ning the
mesh (for instance, by using 1255 elements instead of 655
with a mesh of type I). On the other hand, the number of
wrinkles is systematically different with the two meshes.
With mesh I, the number of wrinkles (for the full blank) is
equal to 8 and next to 16. For mesh II it remains equal to 12
from the onset of wrinkling.
The development of a bulge on the wall is con®rmed by
numerical simulations, as can be seen in Fig. 11, where the
sheet pro®le is drawn for H ˆ 32:8 mm. At this stage the
pro®le is still independent of the angular position around the
blank, and of the mesh employed. The curvature radii in the
region of the bulge are then equal to R1  110 mm and
R2  213 mm.
The results obtained with mesh I were analysed in more
detail, by recording stresses, strains and Eulerian coordi-
nates for the nodes initially situated at 104.5 mm from the
pole. The circumferential distribution of hoop stresses s2
for increasing punch strokes, Fig. 12, shows a fairly com-
plicated evolution, with eight waves in a ®rst stage, and next
16 waves, in agreement with the 16 wrinkles observed for
H ˆ 52 mm.
For the purpose of comparison with the bifurcation model,
it must be emphasised that the stress values at each point of
the structure should be compared at each time with the
Fig. 10. FE meshes used for numerical simulations of the conical cup test. wrinkling limit curve, which itself depends on the instanta-
neous local geometry (local curvature) at the point of
interest. Fig. 12 tends to indicate that a ®rst wrinkling mode
Again, the elastic constants are equal to E ˆ 210,000 MPa with eight waves initiates for H  20 mm. The amplitude of
and n ˆ 0:3, and the effective stress±strain law is assumed to wrinkling around the wall is still very shallow at this stage.
be described by (9), with s0 ˆ 100 MPa, K ˆ 400 MPa and For H  20 mm the nodes of interest are situated just below
n ˆ 0:3. The friction coef®cient is equal to m ˆ 0:15 for the the die radius, and outward bulging is already present at this
contacts of the blank with the punch and the die, and equal to early stage. The curvature radii at the location of interest are
0 with the blankholder. The blankholder force is equal to equal to R1  200 mm and R2  460 mm. The evolutions of
10 kN. A total punch stroke of 52 mm is imposed. stress states (s1, s2) for node 608 at 22.58 and node 1108 at
458 are drawn in Fig. 13, along with the wrinkling limit
4.1. Wrinkling initiation and growth curve corresponding to R1 ˆ 200 mm and R2 ˆ 460 mm.
The bifurcation point obtained from the comparison of the
During the deep-drawing of a conical cup, wrinkles can stress paths of the two nodes is fairly close to the wrinkling
form in the wall, for a punch stroke that depends on the limit curve. The wavelength of the wrinkling mode mea-
geometry, blankholder force and material properties. For sured on the wall is equal to L  81 mm. The wavelength
analysing the appearance of wrinkles in the wall, it is can be also tentatively estimated from the bifurcation
important to recall that the prebifurcation behaviour is model, taking into account the stress ratio (s2/s1) and the
characterised by the development of an outward bulging values of R1 and R2 obtained in FE simulations at the
on the upper part of the wall. This bulge formation, which is bifurcation point. The wavelength predicted in this way is
the result of compressive hoop stresses, was observed by equal to 69 mm.
Havranek [23]. As the punch goes down into the die cavity, the bulge
Wrinkling is ®rst analysed for a material obeying the J2 develops on the upper part of the wall, and the 8-wave
¯ow theory. As a precaution against the consequences of wrinkling mode vanishes, to the bene®t of a 16-wave mode,
186 J.P. de MagalhaÄes Correia, G. Ferron / Journal of Materials Processing Technology 128 (2002) 178±190

Fig. 11. Sheet profile obtained with the J2 flow theory for a punch stroke of 32.8 mm.

which ampli®es dramatically in the region of the bulge. Note for R1 ˆ 110 mm and R2 ˆ 213 mm. However, the wave-
that the bulge is still almost independent of the angular length of the wrinkling mode measured on the wall is equal
position around the cup, and thus practically unaffected by to L  38 mm, and the value estimated from the bifurcation
the ®rst wrinkling mode when the 16-wave mode emerges. analysis is equal to 72 mm.
The evolutions of (s1, s2), but now for node 308 at 98 and Finally, the two values of the characteristic wavelength
node 608 at 22.58, Fig. 14, show that this mode is detected estimated from the bifurcation analysis for the two ``criti-
for H  32:8 mm, viz. for R1  110 mm and R2  213 mm. cal'' H-values are close to each other (L ˆ 69 and 72 mm),
Again, this second bifurcation point obtained in FE simula- and they are compatible with the development of 10
tions is fairly close to the wrinkling limit curve determined wrinkles around the blank. As noted by Kawka et al. [24]

Fig. 12. Circumferential distribution of hoop stresses s2 obtained with mesh I at the initial distance of 104.5 mm from the poleÐJ2 flow theory.
J.P. de MagalhaÄes Correia, G. Ferron / Journal of Materials Processing Technology 128 (2002) 178±190 187

pre-existing ®elds associated with the 8-wave mode. Such a


drop from 8 to 16 wrinkles was already observed by Kim
et al. [16] in numerical simulations of a spherical cup deep-
drawing test using a similar mesh. In contrast, the 12-wave
mode that initiates with mesh II continuously ampli®es
throughout the drawing process. The 12-wave mode may
be closer to the ``exact'' result.
In summary, the critical stress states for wrinkling
obtained with the bifurcation analysis are in good accor-
dance with the FE results. The characteristic wavelength
(or the number of wrinkles) obtained with the bifurcation
analysis agrees fairly well with the FE results obtained with
mesh II.

4.2. Influence of material anisotropy

The effect of normal anisotropy is examined now. The


Fig. 13. Evolution of the stress state for the two nodes at the initial
simulations are made for the different materials de®ned in
distance of 104.5 mm from the pole, and at the angular positions: 22.58 and Fig. 6. Two main effects are identi®ed, which can modify the
458. The wrinkling limit curve (WLC) for R1 ˆ 200 mm and wrinkling behaviour:
R2 ˆ 460 mm is drawn for comparisonÐJ2 flow theory.
(1) The profile of the wall depends on r. For a given punch
stroke, the curvature in the region of the bulge (before
wrinkling) is more pronounced for large r-values.
the results from FE simulations should be handled with great
(2) The stress path for a material element in the wall also
caution in the present context. This is specially true for the
depends on r. For a given punch stroke, the stress state
number of wrinkles in processes combining axisymmetrical
in the region of the bulge is nearer to uniaxial tension
geometry and transverse isotropy of the material. For mesh I,
for larger r-values.
the initial number of wrinkles (eight wrinkles around the
blank) likely becomes too low to match the evolutions of Since the wrinkling limit curve and the characteristic
the bulge curvature as the punch penetrates further into the wavelength also depend on r, a comprehensive analysis of
die cavity, and it is enforced to drop to 16 as a result of the the in¯uence of anisotropy in a real forming process is quite

Fig. 14. Evolution of the stress state for the two nodes at the initial distance of 104.5 mm from the pole, and at the angular positions: 98 and 22.58. The WLC
for R1 ˆ 110 mm and R2 ˆ 213 mm is drawn for comparisonÐJ2 flow theory.
188 J.P. de MagalhaÄes Correia, G. Ferron / Journal of Materials Processing Technology 128 (2002) 178±190

Fig. 15. Circumferential distribution of hoop stresses s2 obtained with mesh I at the initial distance of 104.5 mm from the pole, for the transversely
anisotropic material defined in Fig. 8.

dif®cult. However, the general trends obtained from FE in limit stresses predicted for a given geometry as r increases
simulations can be summarised as follows: (Fig. 7). It follows that the larger punch stroke obtained with
a larger r at the onset of wrinkling should be consistently
(1) With mesh I, the 8-wave wrinkling mode is more
ascribed to the dependence of prebifurcation stress ®elds and
persistent for larger r-values. It is actually not observed
wall geometry on anisotropy.
with r ˆ 0:5, in which case 16 wrinkles are obtained
Wrinkling tendencies are investigated now for the trans-
throughout the process, and is still present for
versely anisotropic material de®ned in Fig. 8 (with r0 ˆ 2,
H ˆ 52 mm with r ˆ 4. With mesh II, the number of
r45 ˆ 1:5, r90 ˆ 3). In that case the in¯uence of the mesh is
wrinkles always remains equal to 12 for all materials.
erased by the strong in-plane anisotropy, and the results
(2) The punch stroke at the onset of wrinkling tends to
obtained with either mesh are quite similar. The circumfer-
increase as r increases.
ential distribution of hoop stresses s2 is drawn in Fig. 15 for
No clear tendency can be drawn from the ®rst observation increasing punch strokes. For H ˆ 7:25 mm, we can observe
concerning the number of waves. Again, the wavelength was that the circumferential stress distribution prior to bifurca-
estimated from the bifurcation analysis, using the values of tion is not uniform, owing to the transverse anisotropy of the
the stress ratio and of curvature radii obtained with the FE material. After a long period of hesitation, wrinkling clearly
simulations at the bifurcation point. The wavelength is found develops with 12 waves for H  45:6 mm. The bifurcation
to increase from L  70 mm for r ˆ 0:5, to L  90 mm model, where wrinkling is analysed as a function of the local
for r ˆ 4. These values correspond to a number of wrinkles orientation with respect to the orthotropic axes, is not suited
that decreases from 10 to 8 as r increases from 0.5 to 4. to the present situation, where three wrinkles develop inter-
This trend is observed with either Hill's or Ferron's yield actively over an angular sector of 908. However, an averaged
surface. value of 10 wrinkles can be predicted by the bifurcation
The dependence of stress path on anisotropy is quite model.
important. For r ˆ 0:5, the stress-state in the region of
the bulge for a punch stroke of 20±26 mm is fairly close
to pure shear. For r ˆ 4, it almost coincides with uniaxial 5. Conclusions
tension, before turning into the deep-drawing range for
larger punch strokes. Also, the bifurcation is delayed for A bifurcation analysis has been developed for predicting
large r-values, owing to the fact that the curvature of the wall the occurrence of wrinkling in metal sheets with isotropic
increases as r increases. These features of the response of the elasticity and transversely anisotropic plasticity. Based on
structure prior to bifurcation are much more important to the local conditions describing sheet geometry, loading and
de®ne the attainment of bifurcation than the slight decrease material anisotropy, wrinkling is predicted in the form of a
J.P. de MagalhaÄes Correia, G. Ferron / Journal of Materials Processing Technology 128 (2002) 178±190 189

limit curve de®ning bifurcation in principal stress space, 4. The consistency condition for elastic±plastic loading,
along with the wavelength and orientation of the wrinkles. f_ ˆ 0, which reads
The practical relevance of this approach was checked by F;ij s_ ij _ ˆ 0
s (A.4)
comparison with FE predictions of wall wrinkling in the
conical cup test. Apart from the fact that the FE results where the abbreviated notation: F;ij ˆ @F=@sij is used.
should be considered with great care, specially concerning 5. The plastic strain increments e_ pij , which depend on the
the number of wrinkles for transversely isotropic materials, plasticity theory employed.
the critical stress levels and wrinkle wavelengths predicted
For ¯ow theory, the plastic strain increments are obtained
with the bifurcation analysis are in good accordance with the
from the normality rule
FE results.
p
The overall effect of plastic anisotropy on wrinkling in e_ pij ˆ e_ F;ij (A.5)
conical cup drawing comes from the combination of two p
where e_ is the work-conjugated effective plastic strain
factors: (1) the dependence of cup curvature and prebifurca-
rate. Introducing the tangent strain-hardening coef®cient:
tion stress ®elds in the wall on anisotropy and (2) the p
Ht ˆ d s=dep ˆ s_ =e_ , and using (A.1), (A.2), (A.4) and
intrinsic dependence of bifurcation on anisotropy. Although
(A.5), the strain increments e_ ij are determined as functions
the critical stress levels obtained with the bifurcation
of the stress increments s_ kl by
analysis for a given geometry slightly decrease as the  
anisotropy coef®cient r increases, the effect of anisotropy 1
e_ ij ˆ Sijkl ‡ F ;ij F;kl s_ kl (A.6)
on the stress state and the sheet curvature developed in Ht
the wall seems to be the most important factor. As a ®nal
result, wrinkling tends to initiate for larger punch strokes For deformation theory, the constitutive law is given in the
as r increases. integrated form
epij ˆ ep F;ij (A.7)

Appendix A where ep is the accumulated plastic strain. Differentiating


(A.7) under conditions of plastic loading and using (A.1),
The tangent stiffness tensor L is obtained from the set of (A.2) and (A.4), the strain increments are obtained in the
constitutive equations for elastic±plastic behaviour de®ned form
by  
1 1
e_ ij ˆ Sijkl ‡ F ;ij F;kl ‡ F;ijkl s_ kl (A.8)
1. The splitting of the total strain increments e_ ij into elastic Ht Hs
and plastic strains where Hs ˆ s =ep and F;ijkl ˆ @ 2 F=@sij @skl . The matrix S
has to be inverted to obtain the components of the stiffness
e_ ij ˆ e_ eij ‡ e_ pij (A.1)
matrix L. Eqs. (A.6) and (A.8) can be easily specialised to
2. The incremental form of Hooke's law for linear isotropic elasticity and plane-stress conditions.
elasticity, which relates the elastic strain increments
to the stress increments via the tensor of elastic com-
pliance S Appendix B

e_ eij ˆ Sijkl s_ kl (A.2) The calculations relating to the functional F (Eq. (10)) are
made in …w1 ; w2 ; x3 † axes. Then, all the useful quantities are
3. The yield function f for isotropic hardening defined by expressed in these axes, and are denoted by bdij, sdij , Nijd , Ldijkl .
f ˆ F…sij † 
s (A.3) Likewise, partial derivatives in (11) and (12) are taken with
respect to the coordinates …w1 ; w2 †, using the displacement
where s is a measure of the current effective stress and F ®eld de®ned by (14). The bifurcation functional F then
is a homogeneous function of degree one. becomes

"  2  #Z  
3 l l l
F…u_ i ; u_ 3 † ˆ t B2 Ld11 d d d d d d
‡2AB…L11 b11 ‡ 2L16 b12 ‡ 2L12 b22 † cos2 w1 dS
l l S l
"  2 Z     Z  
l l t5 l 4 l
‡ A2 t3 sd11 sin2 w1 dS ‡ Ld11 cos2 w1 dS
l S l 12 l S l
Z   #
3 d d2 d d2 d d2 d d d d d d d d d 2 l
‡ t …L11 b11 ‡ L22 b22 ‡ 4L66 b12 ‡ 4b11 b22 L16 ‡ 4b22 b12 L26 ‡ 2b11 b22 L12 † cos w1 dS (B.1)
S l
190 J.P. de MagalhaÄes Correia, G. Ferron / Journal of Materials Processing Technology 128 (2002) 178±190
R R
Observing that S sin2 …lw1 =l† dS ˆ S=2; S cos2 …lw1 =l† dS ˆ [3] N. Triantafyllidis, A. Needleman, An analysis of wrinkling in the
S=2, the bifurcation functional can be expressed in the Swift cup test, J. Eng. Mater. Tech. 102 (1980) 241±248.
[4] J.W. Hutchinson, K.W. Neale, Wrinkling of curved thin sheet metal,
matrix form as: F ˆ 12 St3 u‰MŠut , where S is the sheet surface in: Proceedings of the International Symposium on Plastic Instability,
over which wrinkling spans and u ˆ …A; B† is the displace- Considere Memorial, Presses des Ponts et ChausseÂes, Paris, 1985,
ment amplitude vector. The matrix components Mij are pp. 71±78.
de®ned by [5] H. AmeÂziane-Hassani, K.W. Neale, On the analysis of sheet metal
wrinkling, Int. J. Mech. Sci. 33 (1991) 13±30.
 2 2  
d l t d l 4 d d2 2 [6] C.H. Wang, G. Kinzel, T. Altan, Wrinkling criterion for an
M11 ˆ s11 ‡ L11 ‡L11 b11 ‡ Ld22 bd22 ‡ 2Ld12 bd11 bd22 anisotropic shell with compound curvatures in sheet forming, Int. J.
l 12 l Mech. Sci. 36 (1994) 945±960.
2
‡ 4…Ld66 bd12 ‡ Ld16 bd11 bd12 ‡ Ld26 bd22 bd12 †; [7] K.W. Neale, Numerical analysis of sheet metal wrinkling, in:
 2 Thomson, et al. (Eds.), Proceedings of the Numiform'89, Balkema,
d l 1989, pp. 501±505.
M22 ˆ L11 ; [8] R. Hill, A theory of the yielding and plastic flow of anisotropic
l
  metals, Proc. Roy. Soc. Lond. A 193 (1948) 281±297.
l [9] R. Hill, Theoretical plasticity of textured aggregates, Math. Proc.
M12 ˆ M21 ˆ ‰Ld11 bd11 ‡ Ld12 bd22 ‡ 2Ld16 bd12 Š (B.2)
l Camb. Phil. Soc. 85 (1979) 179±191.
[10] K.W. Neale, P. Tugcu, A numerical analysis of wrinkle formation
The wrinkling condition is obtained by setting the determi- tendencies in sheet metals, Int. J. Num. Meth. Eng. 30 (1990) 1595±
nant M equal to 0, and by minimising M with respect to the 1608.
wavenumber l. The following values are obtained for the [11] Z. Marciniak, K. Kuczynski, Limit strains in the processes of stretch-
forming sheet metal, Int. J. Mech. Sci. 9 (1967) 609±620.
critical normal stress in the direction of wrinkling, sd11 , and
[12] S. StoÈren, J.R. Rice, Localized necking in thin sheets, J. Mech. Phys.
for the wavenumber l: Solids 23 (1975) 421±441.
  [13] G. Ferron, A. Molinari, Mechanical and physical aspects of sheet-
t2 Ld11 l 2 metal ductility, in: R.H. Wagoner, K.S. Chan, S.P. Keeler (Eds.),
sd11 ˆ (B.3)
6 l Forming Limit Diagrams: Concepts, Methods and Applications,
TMS, Warrendale, PA, 1989, pp. 111±151.
and [14] P. Nordlund, B. HaÈggblad, Prediction of wrinkle tendencies in
s explicit sheet metal-forming simulations, Int. J. Num. Meth. Eng. 40
2
4 12R d (1997) 4079±4095.
lˆ 2 ‰L11 A BŠ (B.4)
Ld11 [15] Abaqus/Explicit, Version 5.8 Manuals, 1998, Hibbit, Karlsson and
Sorensen Inc., Pawtucket, RI, USA.
with [16] J.-B. Kim, D.-Y. Yang, J.-W. Yoon, F. Barlat, The effect of plastic
anisotropy on compressive instability in sheet metal forming, Int. J.
2 2 2
A ˆ Ld11 bd11 ‡ Ld22 bd22 ‡ 2Ld12 bd11 bd22 ‡ 4…Ld66 bd12 ‡ Ld16 bd11 bd12 Plasticity 16 (2000) 649±676.
[17] L.P. Moreira, G. Ferron, G. Ferran, Experimental and numerical
‡ Ld26 bd22 bd12 †; analysis of the cup drawing test for orthotropic metal sheets, J. Mater.
Proc. Tech. 108 (2000) 78±86.
B ˆ …Ld11 bd11 ‡ Ld12 bd22 ‡ 2Ld16 bd12 †2 (B.5) [18] S. Fromentin, M. Martiny, G. Ferron, Z. Tourki, L.P. Moreira, G.
Ferran, Finite element simulations of sheet metal forming processes
The principal stress s1 along direction s1 is obtained from
for planar anisotropic materials, Int. J. Mech. Sci. 43 (2001) 1833±1852.
(B.3) as [19] R. Mesrar, S. Fromentin, R. Makkouk, M. Martiny, G. Ferron, Limits
t Ld11 to the ductility of metal sheets submitted to complex strain paths, Int.
s1 ˆ l2 (B.6) J. Plasticity 14 (1998) 391±411.
6R cos2 …d a† ‡ Rs sin2 …d a† [20] G. Ferron, R. Makkouk, J. Morreale, A parametric description
of orthotropic plasticity in metal sheets, Int. J. Plasticity 10 (1994)
where Rs ˆ s1 =s2 denotes the principal stress ratio. The 431±449.
principal stress s1 depends on the orientation of wrinkles, d, [21] D.C. Drucker, Relation of experiments to mathematical theories of
through its explicit dependence in (B.6), and via the orienta- plasticity, J. Appl. Mech. 16 (1949) 349±360.
tion dependencies of Ldij and bdij (Eqs. (B.4)±(B.6)). [22] S. Ellagoune, M. Boivin, CriteÁre d'apparition des plis en embou-
tissage, MeÂmoires et Etudes Scientifiques, Revue de MeÂtallurgie 85
(1988) 211±221.
[23] J. Havranek, The effect of mechanical properties of sheet steels on
References the wrinkling behaviour during deep drawing of conical shells, J.
Mech. Working Tech. 1 (1977) 115±129.
[1] R. Hill, A general theory of uniqueness and stability in elastic/plastic [24] M. Kawka, L. Olejnik, A. Rosochowski, H. Sunaga, A. Makinouchi,
solids, J. Mech. Phys. Solids 6 (1958) 236±249. Simulation of wrinkling in sheet metal forming, J. Mater. Proc. Tech.
[2] J.W. Hutchinson, Plastic buckling, Adv. Appl. Mech. 14 (1974) 67±144. 109 (2001) 283±289.

You might also like