See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/264422770

Finite element implementation of an orthotropic plasticity model for sheet

metal forming simulations

Article  in  Latin American Journal of Solids and Structures · June 2007

CITATIONS READS

8 1,090

1 author:

Luciano Pessanha Moreira

Universidade Federal Fluminense
90 PUBLICATIONS   236 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Simulação Estocástica da Propagação de Trincas View project

Limit strains in metallic sheets: experimental characterization and modeling View project

All content following this page was uploaded by Luciano Pessanha Moreira on 19 January 2015.

The user has requested enhancement of the downloaded file.

www.lajss.org
Latin American Journal of Solids and Structures 4 (2007) 149–176

Finite element implementation of an orthotropic plasticity model for

sheet metal forming simulations

Luciano P. Moreira1,∗ and Gérard Ferron2

1
Programa de Pós-graduação em Engenharia Metalúrgica, Universidade Federal Fluminense,
Volta Redonda, RJ – Brazil
2
Laboratoire de Physique et Mécanique des Matériaux, Université Paul-Verlaine Metz – France

Abstract
In this work, the implementation procedures for the user Fortran subroutines of the
commercial finite element code ABAQUS are described for the particular case of linear
elasticity and a plasticity model proposed within the context of the plastic flow theory along
with the isotropic work-hardening assumption. The general 3D case is firstly presented
and detailed for the implicit and explicit integration techniques available in the ABAQUS
code. Afterwards, the equations for the particular plane-stress state are outlined. The
corresponding procedures are applied to a plane-stress orthotropic plasticity model. Finally,
the implemented user subroutines are validated by means of numerical simulations of sheet
metal forming experiments, namely, the hemispherical punch stretching and cup-drawing.
The comparisons of the predicted either radial strains or earing formation with the measured
data obtained for an IF steel and a tinplate food-can shows the pertinent conditioning of the
user subroutines as well as the importance of an accurate description of the range of stresses
states involved in sheet metal forming processes.

Keywords: orthotropic plasticity, sheet metal forming, finite element modeling

1 Introduction

Sheet metal forming is an important manufacturing process widely used to produce complex
stamped parts starting from flat blank sheets in many sectors such as the automotive, food
and beverage cans industries. Due to the global economic context, these sectors are urged to
be extremely competitive by decreasing production costs and increasing the process efficiency.
The numerical simulation combined with the sheet metal forming skill represents one of the
technological innovations available to achieve these requirements since this may drastically de-
crease the cost of the time consuming traditional tryout steps. In the last 20 years, considerable
efforts have been made to improve the finite element method (FEM) based upon incremental
or one-step techniques for solving nonlinear problems arising from material behavior, geometry

∗
Corresp. author email: [email protected] Received 29 Nov 06; In revised form 7 Mar 07
150 Luciano P. Moreira and Gérard Ferron

and friction. Moreover, user-friendly graphical interfaces together with the increasing computer
capacity also promoted the numerical simulation to analyze the sheet metal forming process at
an industrial scale. In view of these advances, the questions arising from the accuracy or the lim-
itations related to the type of material description adopted to model the sheet behavior became
particularly important. From this standpoint, a phenomenological approach is generally favored
owing to its straightforward implementation in finite element codes and low computational costs
in comparison to physically motivated models based upon the polycrystalline plasticity theo-
ries [1, 4].
The present work aims at presenting the procedures used together with the Fortran user
material subroutines available in the commercial ABAQUS finite element code. Firstly, the
elastic-plastic constitutive equations are outlined for the linear isotropic Hooke’s law and the
associated plastic flow theory under the isotropic work-hardening assumption. Afterwards, the
phenomenological plasticity description proposed by Ferron et al. [8], herefater Ferron’s model,
is briefly recalled in order to explain the methods used to determine the coefficients defining
the initial material anisotropy. The equations needed for the finite element implementation
of Ferron’s model are then presented for the plane-stress and the 3D generalized stress states.
Finally, the numerical predictions obtained with Ferron’s model are compared to the measured
data determined from the hemispherical punch stretching and the cup-drawing experiments.
The discussions are focused on the influence of the yield surface shape in the stress range of
interest upon the strain distributions and the final geometry predicted for these sheet metal
forming experiments.

2 Elastic-plastic constitutive equations

First of all, the elastic behavior is described by the Hooke’s linear elasticity law defined by:

σ̇ = Ce : ²̇e (1)
where Ce is the 4th order elasticity tensor. By considering the small elastic strains assumption,
the elastic-plastic behavior can then be described by the additive decomposition of the total
Cauchy strain-rate tensor into an elastic part and a plastic part, that is,

²̇ = ²̇e + ²̇p (2)

in which the plastic strain-rate tensor is determined from the associated plastic flow rule or the
normality law defined by:

∂f
²̇p = λ̇ (3)
∂σ
where σ is the Cauchy stress tensor and λ̇ is the plastic multiplier.

Latin American Journal of Solids and Structures 4 (2007)

 Finite element implementation of plasticity model for sheet metal forming simulations 151

In Eq. (3), f is the yield function which can be defined under the isotropic work-hardening
assumption by:
f = F (σ) − σ̄ (4)
where F (σ) is a first degree homogeneous stress function and σ̄ is a scalar measure of the effective
stress. The plastic multiplier λ̇ in Eq. (3) is obtained from the elastic-plastic loading condition,
f = 0, by applying first the Euler’s identity to the yield function, Eq. (4),
∂f
F (σ) = σ̄ = σ (5)
∂σ
and then by calculating the associated plastic work per unit volume from the equivalent plastic
work principle, i.e.,
σ : ²̇p = σ̄ ²̄˙ p (6)
together with the normality law, Eq. (3), results that the plastic multiplier λ̇ is equal to the
effective plastic strain-rate ²̄˙p conjugated of the effective stress σ̄.
The elastic-plastic behavior description is completed by defining the work-hardening evo-
lution with respect to the plastic strain which relates the effective stress-rate to the effective
plastic strain-rate as:
σ̄˙ = H(σ̄)²̄˙ p (7)
where H(σ̄) = dσ̄/d²̄p is the work-hardening rate.

3 General integration method

The ABAQUS FE code provides at the beginning of the time step, for each element integration
point, the stress tensor components, σij , the increments of the total strain tensor, ∆²ij , and all
the internal or user defined state variables. Bearing in mind deformation-driven problems, as in
displacement based and mixed finite element formulations, the stress state at the end of the time
step in then obtained from the strain history by means of an integration of the rate constitutive
equations in an incremental procedure The general integration procedure is based upon the
elastic predictor-plastic corrector method commonly known as the return mapping algorithm.
Thus, in order to abridge the notation, let us denote ξ t as the current value of all variables at
the beginning of the time step, ξ T rial as the quantities referring to the elastic prediction and
ξ t+∆t as the corrected values determined at the end of the time step.
The trial stress components are computed from an elastic prediction by making use of the
Hooke’s law, Eq. (1), with the total strain increments ∆²kl , that is,
T rial t e
σij = σij + Cijkl ∆²kl (8)
e
where for the isotropy case the elasticity tensor Cijkl is defined by:
e
Cijkl = λδij δkl + µ (δik δjl + δil δjk ) (9)

Latin American Journal of Solids and Structures 4 (2007)

152 Luciano P. Moreira and Gérard Ferron

where λ and µ are the Lamè’s coefficients. On the other hand, elastic-plastic loading may occur
during the current time step if the yield condition is violated by the trial state of stress, that is
to say, if the following condition is verified:
T rial
F (σij ) − σ̄ ≥ 0 (10)

Then, the new state of stress is obtained from the plastic correction of the trial state of stress
according to the implicit and explicit integration techniques available in the ABAQUS finite
element code. In the following, the procedures used together with the Fortran user subroutines
UMAT (ABAQUS/Standard) and VUMAT (ABAQUS/Explicit) are presented in details for the
general 3D and the plane-stress cases.

3.1 Implicit procedure

The integration procedure used in the ABAQUS/Standard code is based upon the implicit
backward Euler scheme where it is firstly necessary to obtain the plastic corrections associated
to the trial stress state and then to define the elastic-plastic tangent operator.

3.1.1 Plastic corrections

The solution for the plastic corrections is made with the help of the Newton’s method. By
denoting c(ξ) the correction associated to the variable ξ, the Hooke’s law with the additive
decomposition of the total strain, Eqs. (1-2), provide a relationship between the corrections of
the stress components and the plastic strain increments as:
e
c(σij ) = −Cijkl c(∆²pkl ) (11)

Also, the effective stress-strain measures, Eq. (7), must be verified leading to:

∂H(σ̄)
c(σ̄) = H(σ̄) c(∆²̄p ) + ∆²̄p c(σ̄) (12)
∂ σ̄
or
c(σ̄) = Ĥ(σ̄) c(∆²̄p ) (13)
where
H(σ̄)
Ĥ(σ̄) = (14)
∂H(σ̄) p
1− ∆²̄
∂ σ̄
Then, the elastic-plastic loading condition, f = 0 in Eq. (4), is not satisfied until the solution
is reached within a given accuracy, that is,

F,ij c(σij ) − c(σ̄) = −f (15)

Latin American Journal of Solids and Structures 4 (2007)

 Finite element implementation of plasticity model for sheet metal forming simulations 153

where F,ij = ∂f (σij )/∂σij = ∂F (σij )/∂σij since it is assumed hereafter that the work-hardening
is a function only of the effective plastic strain.
Moreover, the associated plastic flow rule, see Eq.(3), gives the corrections of the plastic
strain increments:

c(∆²pij ) − F,ij c(∆²̄p ) − ∆²̄p F,ijkl c(σkl ) = F,ij ∆²̄p − ∆²pij (16)

where F,ijkl = ∂ 2 F (σij )/∂σij ∂σkl . Finally, by combining Eqs. (11-15), the following set of
Newton’s equations are obtained for the corrections upon the plastic strain components:
h ³ ´i ³ ´
c(∆²pmn ) δim δjn + Cklmn
e
F,ijkl ∆²̄p + F,ij F,kl /Ĥ(σ̄) = F,ij ∆²̄p + f /Ĥ(σ̄) − ∆²pij (17)

Once the plastic corrections are determined, the stress state at the end of the time step is
computed iteratively by correcting the trial stress state, Eq. (8), until the elastic-plastic loading
condition (f = 0) is satisfied within a given accuracy. Thus, the plastic corrections for the stress
components are: ¯
e
σij |N = σij |N −1 − Cijkl c(∆²pkl )¯N (18)
where N is the current Newton’s iteration.
The work-hardening evolution, Eq. (7), is then updated from the effective plastic strain
which is stored as a state variable defined at the end of Newton’s iterations as:

²̄p |N = ²̄p |N −1 + ∆²̄p |N (19)

where the effective plastic strain increment ∆²̄p is determined by:

∆²̄p |N = ∆²̄p |N −1 + c(∆²̄p )|N (20)

In Eq. (20), the correction upon the effective plastic strain increment is calculated by
the Newton’s equation obtained from the equivalent plastic work associated to the (N − 1)th
iteration: ¯
σ ∆²p
− σ̄∆²̄p¯
ij ij ¯
c(∆²̄p ) = ¯ ¯ (21)
σ̄ + H(σ)∆²̄ ¯ p
N −1

3.1.2 Elastic-plastic tangent operator

The elastic-plastic tangent operator is defined by consistent linearization of the response function
resulting from the integration algorithm as proposed by Simo and Taylor [18]. The consistent
stiffness matrix is determined by taking variations with respect to the all variables. Firstly, the
generalized Hooke’s law, Eq. (1), together with the total strain additive decomposition, Eq. (2),
provide the variations between the total strain and stress components:
e
∂²ij = Sijkl ∂σij + ∂²pij (22)

Latin American Journal of Solids and Structures 4 (2007)

154 Luciano P. Moreira and Gérard Ferron

where
e ν 1
Sijkl =− δij δkl + (δik δkl + δil δjl ) (23)
E 4µ
is the elastic compliances tensor in which E is the Young modulus and ν is the Poisson’s ratio.
Next, the variations in the associated plastic flow rule, Eq. (3), the elastic-plastic loading
condition, Eq. (4), and the effective stress-strain relationship, Eq. (7), give respectively:

∂²pij = ∂²̄p F,ij +∆²̄p F,ijkl ∂σkl (24)

F,ij ∂σij − ∂ σ̄ = 0 (25)

∂ σ̄ = H(σ̄)∂²̄p (26)
By eliminating ∂ σ̄ and ∂²̄p in Eqs. (25-26) and replacing the resulting expression in Eq.
(24):
F,ij F,kl
∂²pij = ∂σkl + ∆²̄p F,ijkl ∂σkl (27)
H(σ̄)
and then returning in Eq. (22), the following system of equations is obtained:
£ ¤ ep
∂²ij = (F,ij F,kl ) H(σ̄) + ∆²̄p F,ijkl +Sijkl
e
∂σkl = Sijkl ∂σkl (28)
ep
where the term in the bracktes defines the elastic-plastic compliances tensor,that is, Sijkl . There-
fore, the consistent elastic-plastic tangent operator at the end of the time step is given by:
h i−1
t+∆t ep
Jijkl = (∂σij /∂²kl )t+∆t = Sijkl (29)

3.2 Explicit procedure

In the ABAQUS/Explicit code the global equilibrium equations are solved by means of an explicit
integration technique based upon the central difference finite scheme, wherein the integration
of the constitutive equations is simply obtained from the plastic strain increments once the
elastic-plastic loading condition is verified. Thus, the stress state at the end of the time step
is determined by means of the plastic strain increments from the correction of the trial stress
state, Eq. (8), that is:
t+∆t
σij T rial
= σij − 2µ∆²pij (30)
It should be noted that the correction in Eq. (30) is colinear with the plastic strain increment
and, hence, the method should be referred to as the normal return algorithm. For the particular
case of the von Mises yield criterion represented in the deviatoric stress plane, the algorithm is
known as the radial return [13]. Therefore, since the plastic strain increments are known from
the associated plastic flow rule, Eq. (3), the only quantity which needs to be defined so as to
determine the values of all variables at the end of the time step is the effective plastic strain
increment ∆²̄p . An appropriate procedure to define the effective plastic strain increment consists

Latin American Journal of Solids and Structures 4 (2007)

 Finite element implementation of plasticity model for sheet metal forming simulations 155

of calculating the increment of the plastic work associated to the time step increment. From
Eq. (6), this can be stated as:
t+∆t
σij ∆²pij = σ̄ t+∆t ∆²̄p (31)
and then, by using Eqs. (3) and (7), ∆²pij = ∆²̄p F,Tijrial and σ̄ t+∆t = σ̄ t + H(σ̄)t ∆²̄p , with Eq.
(30):
(σij F,ij )T rial − σ̄ t
∆²̄p = (32)
H(σ̄)t + 2µ (F,ij F,ij )T rial

3.3 Plane-stress case

For the plane-stress state, such as in the case of shell or membranes elements formulations,
the strain component normal to the element in-plane local directions (x,y), that is, εzz , is not
defined from the kinematics but by the restriction σzz = 0. As a result, the elastic prediction
must be carried from the Hooke’s law, Eq. (1), written in the plane-stress state together with
the in-plane total strain increments (∆²xx ,∆²yy ,∆γxy ), that is,

T rial T rial E
σxx = σxx + [∆²xx + ν∆²yy ] (33)
(1 − ν 2 )

T rial T rial E
σyy = σyy + [ν∆²xx + ∆²yy ] (34)
(1 − ν 2 )
T rial T rial
σxy = σxy + µ∆γxy (35)
where it should be noted that in the ABAQUS code the shear strains are defined as γij = 2εij (i 6=
j). Likewise, the effective plastic strain increment and the corresponding plastic corrections must
be redefined from the plane-stress state according to the integration procedure.
The increment of the total strain in the normal direction is defined from the additive strain
decomposition, Eq. (2), that is, ∆²zz = ∆²ezz + ∆²pzz , where the elastic part is obtained from
Hooke’s law, Eq. (1), by:
ν £¡ t+∆t ¢ ¡ t+∆t ¢¤
∆²ezz = − t
σxx − σxx t
+ σyy − σyy (36)
E
whereas the plastic part is calculated from the plastic incompressibility condition together with
the normality law, Eq. (3), as:
¡ ¢
∆²pzz = − ∆²pxx + ∆²pyy = −(F,xx +F,yy )t+∆t ∆²̄p (37)

It should be noted that in all equations defined here above, the terms F,ij (i 6= j) are
calculated from the expressions of the F function, see Eq. (4), taking into account the symmetry
of the Cauchy stress tensor, for instance, F,ij = ∂F/∂σij + ∂F/∂σji = 2F,ij ∀ (i 6= j).

Latin American Journal of Solids and Structures 4 (2007)

156 Luciano P. Moreira and Gérard Ferron

3.4 Finite-strain formulation

In the ABAQUS code, the description of the elastic-plastic behavior in finite strains is given by
the additive decomposition of the total strain-rate tensor, (1) D = De + Dp , the hypoelastic law,
(2) σ = Ce : De = Ce : (D − Dp ) and the associated plastic flow rule, (3) Dp = ²̄˙p F,ij ei ⊗ ej .
These constitutive equations are expressed in a corotational frame defined by an orthonormal
basis ei (i = 1, 2, 3) [17]. The strains components are then obtained by integration of the strain
rate components defined in this corotational frame which rotates according to the spin rate [9].
In the ABAQUS input keywords file (jobname.inp), the corotational frame is activated with
the ∗ORIEN T AT ION option [17]. Thus, the implementation of the constitutive equations
for any orthotropic plasticity model together with the ABAQUS user material subroutines is
performed directly since the rotation of the symmetry material axes are taken into account by
the corotational frame.

4 Ferron’s plasticity model

Phenomenological anisotropic yield criteria are usually defined in the axes of material orthotropic
symmetry (x, y, z). In a rolled metal sheet, these axes are parallel to the rolling direction
(x ≡ RD), the in-plane transverse (y ≡ T D) and the normal (z ≡ N D) directions respectively.
For the particular plane-stress case, the yield function defined by Eq. (4) can be written as:

f = F (σxx , σyy , σxy ) − σ̄ (38)

The yield function in Eq. (38) can be rewritten using the principal stress components (σ1 , σ2 )
and the orientation between the principal stress axes(1, 2) and the in-plane orthotropy directions
(x, y) , i.e., α ≡ (x, 1) ≡ (y, 2) as:

f = φ (σ1 , σ2 , α) − σ̄ (39)

Likewise, by introducing a change of variables given by x1 = (σ1 + σ2 ) /2 and x2 = (σ1 − σ2 ) /2,

where x1 and x2 define the centre and the radius of the Mohr’s stress circle respectively, the
plane-stress yield function can be cast as:

f = ψ (x1 , x2 , α) − σ̄ (40)

Also, the variables x1 and x2 can be described in polar coordinates by:

x1 = G (θ, α, σ̄) cos θ (41)

x2 = G (θ, α, σ̄) sin θ (42)

where G (θ, α, σ̄) is the polar radius defining a point on the yield locus whereas the polar angle
θ defines the current plane-stress state. According to the isotropic work-hardening assumption

Latin American Journal of Solids and Structures 4 (2007)

 Finite element implementation of plasticity model for sheet metal forming simulations 157

it is possible to redefine the polar radius as:

G (θ, α, σ̄) = σ̄g (θ, α) (43)
where g (θ, α) is radius normalized by the effective stress σ̄. Figure 1 illustrates this description
in the principal stress space as a function of the orientation angle α where the yield loci are
normalized by the effective stress σ̄ which, in turn, is chosen as the equibiaxial yield stress, that
is, σb = σ1 = σ2 .

Figure 1: Yield loci of Ferron’s plasticity model plotted in the principal stress space (σ1 , σ2 )
normalized by the equibiaxial yield stress σb as a function of the orientation angle α between the
principal stress axes and the in-plane orthotropy directions.

The function g (θ, α), fullfilling a certain number of consistence conditions vis-à-vis of the
orthotropic symmetry, is defined as an extension of a normal anisotropy yield function g (θ) [8]:
(1 − k)m/6 g (θ, α)−m = F (θ)m/6 − 2a sin θ cos2n−1 θ cos 2α + b sin2p θ cos2q 2α (44)
where the exponents m, n, p and q are positive integers whereas a and b are dimensionless param-
eters describing the initial in-plane sheet material anisotropy. The proposed normal anisotropy
yield function g (θ) is an extension of the Drucker’s [7] isotropic yield criterion and is given by:
¡ ¢3 ¡ ¢2
(1 − k) g (θ)−6 = F (θ) = cos2 θ + A sin2 θ − k cos2 θ cos2 θ − B sin2 θ (45)
where A and B are positive constants and a positive k-value accounts for a flattening of the
yield surface near to the plane-strain tension or compression and pure shear stress states, as
observed in the yield loci obtained from polycrystalline models [2].

Latin American Journal of Solids and Structures 4 (2007)

158 Luciano P. Moreira and Gérard Ferron

The parameters A, B and k are determined from the experimental values of (σb /σ45 ), (σb /τ0 )
and R45 where τ0 is the shear yield stress allong the rolling direction whereas σ45 and R45 are the
uniaxial tensile yield stress and the Lankford plastic strain-ratio at 450 with respect to the rolling
direction respectivelly. The recommended values for the exponents are m = 2, n = 1 or 2, p = 1
or 2 and q = 1 or 2, as discussed elsewhere [8]. Then, the parameters a and b can be determined
from the R-values obtained at 00 and 900 from the rolling direction, hereafter, the method R, or
from the uniaxial tensile yield or flow stresses (σ0 , σ45 , σ90 ), the method σ, as detailed in [14].
Moreover, Hill’s quadratic orthotropic yield criterion [10] can be obtained as a particular case
of the Ferron’s description by setting in Eq. (44) k = 0, m = 2 and n = p = q = 1.
For finite element implementation purposes, it is more useful to write the Ferron’s model
with the help of Eq. (37), which for the elastic-plastic loading condition can be rewritten as:

ψ (x̄1 , x̄2 , α) = 1 (46)

where x̄1 = x1 /σ̄ and x̄2 = x2 /σ̄. Then, the effective stress for the Ferron’s model is achieved
by expressing Eq. (46) together with the Eqs. (41-45):
" ¡ # 1/m
 x2 + Ax2 ¢3 − kx2 ¡x2 − Bx2 ¢2 m/6 
1 2 1 1 2
ψ (x1 , x2 , α) = − C cos 2α + D cos2q 2α (47)
 1−k 

2a x2 x2n−1
1 b x2p
2
where C = ¡ ¢ and D = ¡ ¢ .
(1 − k)m/6 x21 + x22 n−m/2 (1 − k)m/6 x21 + x22 p−m/2
On the other hand, the partial derivatives of the function F (σ), see Eq. (4), are obtained
from the consistency condition of the yield function cast in the form of Eq. (40), that is, df = 0:
∂ψ ∂ψ ∂ψ
df = dx1 + dx2 + dα − dσ̄ = 0 (48)
∂x1 ∂x2 ∂α
by first calculating dx1 , dx2 and dα from Eqs. (49-51) as a function of dσxx , dσyy and dσxy ,
σ1 + σ2 σxx + σyy
x1 = = (49)
2 2
s 
µ ¶2
σ1 − σ2  σxx − σyy 2 
x2 = = + σxy sign (σxy ) (50)
2 2

2σxy
tan 2α = (51)
(σxx − σyy )
and then returning to Eq. (48), one obtains:
· ¸
1 ∂ψ ∂ψ sin 2α ∂ψ
F,xx = + cos 2α − (52)
2 ∂x1 ∂x2 2x2 ∂α

Latin American Journal of Solids and Structures 4 (2007)

 Finite element implementation of plasticity model for sheet metal forming simulations 159

· ¸
1 ∂ψ ∂ψ sin 2α ∂ψ
F,yy = − cos 2α + (53)
2 ∂x1 ∂x2 2x2 ∂α
· ¸
1 ∂ψ cos 2α ∂ψ
F,xy = 2 sin 2α + (54)
2 ∂x2 x2 ∂α
where sin 2α = σxy /2 and cos 2α = (σxx − σyy )/(2x2 ). Likewise, the normal component is
obtained from the plastic incompressibility condition together with the associated plastic flow
rule, Eq.(3), that is, F,zz = −(F,xx +F,yy ). Once the first derivatives of the function F (σ) are
known as a function of the variables x1 , x2 and α, the definition of the partial derivatives in the
form F,ijkl , see Eq. (16), is straightforward.
The extension to the 3D general case is done assuming firstly that the plastic flow is in-
dependent of the hydrostatic pressure. Then, the effect of the normal stress component σzz
is accounted for by replacing σxx and σyy in the plane-stress description by (σxx − σzz ) and
(σyy − σzz ) respectively. Next, the effects of the through-thickness shear stress components
(σxz , σyz ) are considered by means of quadratic terms. Finally, the proposed 3D version of the
Ferron’s model is given by:
q
f = Φ (x1 , x2 , α, σxz , σyz ) − σ̄ = ψ 2 (x1 , x2 , α) + βσxz
2 + γσ 2 − σ̄
yz (55)

where β = γ = 3 in the same way as for the von Mises isotropic yield criterion and ψ (x1 , x2 , α) is
the Ferron’s yield function for the plane-stress state defined by Eq. (45). The partial derivatives
of the 3D Ferron’s yield function, required for the calculation of the plastic strain increments
corrections and the elastic-plastic tangent operator, are obtained in the same way as for the
plane-stress case.

5 Numerical simulations

In all the numerical simulations discussed hereafter, the effective plastic stress-strain relationship
is defined by the Swift-Krupkwoski power law:

σ̄ = K (²0 + ²̄p )N (56)

where K, ²0 and N are the strength coefficient, the pre-strain and the strain-hardening exponent
respectively. In all cases, the elastic constants are taken equal to E = 200, 000 MPa and ν = 0.29.
Moreover, the contact between the blank and the tooling is described by the Coulomb friction
law. The simulations were performed with a workstation HP-UX 9000/785 J5600 model with
1.5 Gb RAM. For consistency, all the predictions related to the Hill’s quadratic yield criterion
discussed herefater have been obtained as a particular case of the Ferron’s model unless othewise
stated.

Latin American Journal of Solids and Structures 4 (2007)

160 Luciano P. Moreira and Gérard Ferron

Figure 2: Hemispherical punch stretching experiment for metal sheets: (a) schematic view with
standard dimensions and (b) characteristical regions of a deformed blank.

5.1 Hemispherical punch stretching

Figure 2(a) shows schematically the geometry adopted in the simulations of the hemispher-
ical punch stretching experiment, which is the same as in the experimental and numerical
work conducted by Knibloe and Wagoner [12]. Owing to the axial symmetry and assum-
ing a normal anisotropy behavior, the analysis is described by an axisymmetric model. The
simulations were performed with the implicit code ABAQUS/Standard together with the user
material subroutine UMAT. The tooling is described by analytical rigid surfaces whereas the
sheet blank by 50 axisymmetric shell elements with 9 thickness integration points, SAX1, as
defined in the ABAQUS terminology. On the other hand, the contact is modeled by the option
∗CON T ACT P AIR, IN T ERACT ION = ROU GH wherein the friction behavior is described
by the Coulomb’s law. The blank is fully clamped at the edge (r = 59.2 mm) to assure the
strectching boundary conditions replacing, in this way, the need for modeling either the lock-
bead or the blank-holder.
Figure 2(b) shows the deformed blank obtained for an isotropic material defined with Drucker’s
yield criterion. From this figure it is possible to distinguish three typical regions of the hemi-
spherical punch strecthing. In the first one, between points P and A, the blank is in contact
with the punch. Next, there is a unsupported region located between points A and B. Finally,
the blank envelops the die shoulder between the points B and C, which corresponds to the
clamped edge. It should be observed that the range of stress states in the hemispherical punch
stretching varies from the equibiaxial stretching at the pole (r = 0) to the plane-strain tension
at the clamped blank edge. Therefore, the ratio between the major principal stress component

Latin American Journal of Solids and Structures 4 (2007)

 Finite element implementation of plasticity model for sheet metal forming simulations 161

in plane-strain tension and the equibiaxial tension yield stresses, namely, P = σP S1 /σb , seems to
be the appropriate material parameter controlling the strain distributions in the hemispherical
punch stretching experiment. A numerical study is firstly presented in order to analyze the
effects of the rheological and the tribological parameters upon the numerical predictions of the
radial and the hoop strains obtained from the simulations of the hemispherical punch stretching
experiment. Then, the experimental data obtained by Knibloe and Wagoner [12] is compared
with the numerical predictions determined with Ferron’s model.

5.1.1 Numerical study

The blank has an initial thickness of 0.89 mm with the plastic behavior defined in uniaxial
tension by K = 500 MPa, ²0 = 0.002 and N = 0.2 whereas the material anisotropy is described
by the yield loci plotted in Fig. 3 as a function of the normal anisotropy parameter R.

(a) (b)

Figure 3: Yield loci for different values of the normal anisotropy coefficient R determined with:
(a) Hill’s quadratic yield criterion (k = 0) and (b) Ferron’s model (k = 0.2).

On the other hand, Figs. 4 and 5 shows the corresponding numerical predictions of the radial
strains determined for a punch height P H = 32.5 mm using a Coulomb’s friction coefficient
µ = 0.10 between the blank and the tooling. These strain distributions were obtained at the
shell mid-surface with respect to the original distance from the pole. In a qualitative way, both
yield loci descriptions adopted in the simulations provide the same trends, namely, as the R-
value increases the radial strains at the pole decreases while higher stretching is obtained close
to the blank clamped edge. These effects are better explained as a function of the material
parameter P = σP S1 /σb . From Fig. 3 one can observe that the material parameter P decreases

Latin American Journal of Solids and Structures 4 (2007)

162 Luciano P. Moreira and Gérard Ferron

with either the R-value or the k parameter in Eq. (45), i.e., from 0 for Hill’s quadratic yield
criterion to 0.2 with Ferron’s model which, in turn, provides a flattenning of the yield locus near
to the plane-strain tension stress state. Consequently, the radial strains at the blank edge are
favored, wherein the stress state is plane-strain tension, in detriment to the equibiaxial stretching
at the pole. Moreover, the peak strain is located at larger distances from the pole, as shown
in Fig. 5(a) for the Ferron’s model, since the radial displacements are smaller for decreasing
values of the parameter P . On the other hand, the hoop strains distributions are consistent with
the equibiaxial stretching at the pole and the plane-strain tension at the clamped edge. Also,
the small radial displacements resulting from the higher values of the parameters R and or k,
explains the decreasing of the hoop strains for small P -values.

(a) (b)

Figure 4: Numerical predictions of the radial (²rr ) and the hoop (²θθ ) strains distributions
obtained for the hemispherical punch stretching with Hill’s quadratic yield criterion as a function
of the normal anisotropy coefficient R.

Figure 6 shows the influence of the Coulomb’s friction coefficient µ upon the strains distribu-
tions determined assuming an isotropic material described by the Drucker’s yield criterion with
k = 0.2 and B = 3A in Eq. (45), corresponding to P = 1.112. In the contact region between the
blank and the punch, increasing the friction coefficient reduces the blank slip and thus either the
radial or the hoop strains in the regions near to the pole. For a given punch height, the radial
peak strain starts prematurely and is shifted away from the pole as the µ-value is increased.
In the unsupported region, the radial strains increase with the friction coefficient whereas the
hoop strains are slightly affetcted. Conversely, the plane-strain tension observed in the regions
situated along the die arc and near to the clamped independs of the friction behavior.

Latin American Journal of Solids and Structures 4 (2007)

Finite element implementation of plasticity model for sheet metal forming simulations 163

Figure 5: Numerical predictions of the radial (²rr ) and the hoop (²θθ ) strains distributions
obtained for the hemispherical punch stretching with Ferron’s model as a function of the normal
anisotropy coefficient R.

Figure 6: Numerical predictions of the radial (²rr ) and the hoop (²θθ ) strains distributions
obtained for the hemispherical punch stretching with Drucker’s yield criterion as a function of
the Coulomb’s friction coefficient µ.

Latin American Journal of Solids and Structures 4 (2007)

164 Luciano P. Moreira and Gérard Ferron

At this point, it is interesting to resume the role of either the rheological or tribological
parameters upon the strains predictions of the hemispherical punch stretching experiment. From
the sheet metal formability standpoint, the preponderant material parameters upon the limiting
strains are the hardening exponent N , the strain-rate sensitivity M and the normal anisotropy
coefficient or the Lankford value R. For a given tooling geometry, the limiting strains are also
influenced by the temperature, the initial blank thickness and the tribology condition between
the blank and the tooling surfaces. From the parametric study conducted by Burford et al. [6] for
the full-dome hemispherical punch strecthing test, the parameters controlling the critical dome
height at the pole can be classified in a decreasing order of importance as µ, N , R and M . On
the other hand, the numerical study performed in the present work points out, for a given work
and strain-rate hardening descriptions, that the strains resulting from the hemispherical punch
strecthing and, as a consequence, the critical dome height at the pole as a function of either
the material anisotropy or the friction behavior, may be better explained from the stress-ratio
P = σP S1 /σb in preference to the traditional R-value. Actually, the effects of higher µ-values
are, in a qualitative way, the same as those caused by decreasing P -values. This coupling effect
between the rheological and tribological parameters is better illustrated in Fig. 7 where the
strain-paths, determined at an original distance from the pole equal to 27.2 mm, are plotted
as a function of the parameters P and µ. The peak-strain leading to the strain localization,
characterized by an abrupt change in the strain-path towards the plane-strain state, can be
ascribed to the effects of either the material or the tribological parameters P and µ respectively.

(a) (b)

Figure 7: Strain-paths (²rr , ²θθ ) determined for an element located at 27.2 mm from the original
distance of the pole as a function of (a) the rheological and (b) the tribological parameters.

Latin American Journal of Solids and Structures 4 (2007)

 Finite element implementation of plasticity model for sheet metal forming simulations 165

5.1.2 Comparisons with experimental data

In the work of Knibloe and Wagoner [12], an aluminum killed steel with an initial thickness
of 1.04 mm is tested under dry conditions. The material behavior of this steel is described in
uniaxial tension by K = 598 MPa, ²0 = 0.0023 and N = 0.23. Knibloe and Wagoner [12] have
also performed FEM simulations with the Hill’s quadratic and non-quadratic [11] yield criteria,
the later defined as:

|σ1 + σ2 |m + (1 + 2R) |σ1 − σ2 |m = (1 + R) σ̄ m (57)

where m is a positive exponent which can be obtained to fit the uniaxial and plane-strain
tension data [15]. In the present numerical simulations, the parameters A, B and k defining
the normal plastic anisotropy in Ferron’s model, see Eq. (45), have been identified so as to
respect the ratio of the flow stresses determined by Eq. (57) with m = 2.38 between the major
principal plane-strain tension, σP S1 , and the uniaxial tension, σu , together with the experimental
normal anisotropy coefficient R = 1.48. Table 1 compares the stresses-ratios determined for the
aluminum killed steel with Ferron’s model and Hill’s quadratic and non-quadratic yield criteria
where one can observe that the smallest P = σP S1 /σb value is obtained by Ferron’s model with
A = 3.65, B = 12.23 and k = 0.148. The corresponding strains distributions obtained from
the numerical simulation of the hemispherical punch stretching together with Ferron’s model
are compared in Fig. 8 to the experimental data measured by Knibloe and Wagoner [12]. The
numerical predictions determined by Ferron’s model with a friction coefficient µ = 0.20 are in
good agreement with the experimental values whereas Knibloe and Wagoner [12] have adopted
µ = 0.268 and 0.639 for the Hill’s quadratic and non-quadratic yield criteria respectively. Clearly,
this unreasonable µ-value is due to the high value of the parameter P = σP S1 /σb predicted by
the Hill’s non-quadratic yield criterion, as indicated in Table 1.

Table 1: Stress-ratios for the aluminun killed steel analyzed by Knibloe and Wagoner [12].

Yield criterion σb /τ σb /σu σP S1 /σb σP S2 /σb σP S1 /σu σP S2 /σu

Hill’s quadratic 1.190 1.113 1.119 0.668 1.246 0.744
Hill’s non-quadratic 1.783 0.980 1.200 0.553 1.176 0.542
Ferron’s model 1.962 1.070 1.099 0.642 1.176 0.687

5.2 Cup-drawing experiments

In this section the earing profile determined for an IF steel is firstly compared to the numerical
predictions obtained from Hill’s quadratic yield criterion and Ferron’s model by using either
shell and linear solid elements to mesh the blank. Then, the earing formation obtained for a

Latin American Journal of Solids and Structures 4 (2007)

166 Luciano P. Moreira and Gérard Ferron

Figure 8: Comparison between the numerical predictions determined with Ferron’s model and the
experimental strains distributions obtained from the hemispherical punch stretching by Knibloe
and Wagoner [12].

tinplate steel is discussed from the standpoint of the material parameters describing the effects
of the in-plane anisotropy commonly observed in the deep-drawing of circular blank sheets.

5.2.1 Cup-drawing of an IF steel

The geometry and the blank mesh used in the simulations of the cup-drawing experiment per-
formed on an IF steel with an initial thickness of 0.7 mm are depicted in Fig. 9. Due to the
material orthotropy and the axial symmetry, only one quarter of the cup-drawing geometry is
taken into account. The cup-drawing tooling mesh is described by analytical rigid surfaces and
4-node bilinear discrete rigid elements (R3D4) in ABAQUS/Standard and ABAQUS/Explicit
codes respectively. The blank is meshed by either 756 quadrangular shell with 5 integration
points through the thickness or 1,512 two-layer 3D linear solid elements. Only reduced element
integration rule have been tested with both shell (S4R) and solid (C3D8R) elements formula-
tions. In all cases analyzed herefater, the blank-holder force is taken equal to 42.5 kN whereas
the Coulomb friction coefficient has been taken equal to µ = 0.15 between the punch and the
blank and µ = 0.05 elsewhere.
Table 2 presents the material parameters determined for the IF steel from two uniaxial tensile
tests conducted on specimens taken along the parallel (00 ), diagonal (450 ) and transverse (900 )
angular orientations with respect to the sheet rolling direction (RD). In this table, the yield σy

Latin American Journal of Solids and Structures 4 (2007)

 Finite element implementation of plasticity model for sheet metal forming simulations 167

(a) (b)

Figure 9: Numerical simulation of the cup-drawing experiment: (a) tooling dimensions and (b)
finite element mesh of the blank.

and the flow σf stresses have been determined at 0.2 and 20 % plastic strain levels respectively
whereas the plastic anisotropy coefficients R have been defined at 18% of plastic strain. In this
case, the choice is made upon the flow stress definition which is more pertinent than the initial
yield stress to model the plastic behavior at larger strains when the assumption of isotropic
work-hardening is assumed. This IF steel presents a four-ear profile located at 00 and 900 with
respect to the RD with mean cup heights of 45.99 ± 0.11mm at the peaks and 43.97 ± 0.05 mm
at the troughs situated near to the diagonal angular orientation [5].

Table 2: Material parameters for the IF steel used in the cup-drawing experiment.

Angle (Degrees) σy (MPa) K(MPa) ²0 N R(18%) σf (MPa)

0 155.0 510 0.0048 0.223 1.884 358.0
45 161.5 523 0.0036 0.209 1.527 375.4
90 153.0 510 0.0038 0.216 2.268 361.4

Provided that only uniaxial tensile data is available, the parameters of Ferron’s model were
determined by adopting a positive k-value (0.2) and imposing the relationship B = 3A. Af-
terwards, the parameter A is obtained from the experimental R45 value. The parameters a
and b in Eq. (44) describing the initial material anisotropy, are then determinated by applying
the methods R and σ introduced in Section 4. In order to assure a good compromise between

Latin American Journal of Solids and Structures 4 (2007)

168 Luciano P. Moreira and Gérard Ferron

the experimental data and the theoretical predictions for the angular evolutions of the plastic
anisotropy coefficient, R(α) , and the uniaxial tensile flow stress, σ(α) , different values of the
exponents (n, p) can be considered. In fact, the combinations of interest are (n = 1, p = 2),
(n = 2, p = 1) and (n = 2, p = 2), in addition to the pair (n = 1, p = 1) which provides the cor-
relation between R and σ corresponding to the Hill’s quadratic yield criterion.
The effects of the method adopted to determine the parameters a and b are illustrated
in Fig. 10 as a function the exponents (n, p) together with the experimental values of the
normalized uniaxial tensile flow stresses (σ45 /σ0 , σ90 /σ0 ) and the plastic anisotropy coefficients
(R0 , R45 , R90 ). It is interesting to note that neither of (n, p) pairs provides a good agreement
between both R(α) and σ(α) angular evolutions. Nevertheless, for a given plastic anisotropy
angular evolution, the angular dependence of the uniaxial tensile flow stress is clearly decreased
when the exponents (n, p) are equal to 2.

Figure 10: Comparison between the predictions determined with Ferron’s model and the exper-
imental data obtained for an IF steel : (a) orientation dependence of the normalised uniaxial
tensile flow stress and (b) orientation dependence of the anisotropy coefficient.

Tables 3 compares the number of time increments and the CPU time by adopting reduced
integration shell elements to mesh the blank obtained with Ferron’s model (method R) and
the keyword ∗P OT EN T IAL, which is used to define the Hill’s quadratic yield criterion in
the ABAQUS code. Although the CPU times are larger with use of the user subroutines, the
implementation of Ferron’s model shows a pertinent conditioning since it leads to almost the
same number of increments in comparison to the use of the option ∗P OT EN T IAL for either
the implicit or the explicit integration techniques available in the ABAQUS finite element code.
Fig. 11 shows the predicted earing profile percentage determined for an IF steel with Hill’s
quadratic yield criterion obtained either as a particular case of Ferron’s model and by means

Latin American Journal of Solids and Structures 4 (2007)

 Finite element implementation of plasticity model for sheet metal forming simulations 169

of the keyword ∗P OT EN T IAL. For comparisons purposes, the earing profile percentage is
defined as:

EP% = (hα − hmin ) /hmin (58)

where hα and hmin correspond to the cup height at an angular orientation α with respect to
the rolling direction and the minimum cup heigth respectively. The results obtained with the
user subroutines UMAT (ABAQUS/Standard) and VUMAT (ABAQUS/Explicit) are in good
agreement with the earing profile determined with the option ∗P OT EN T IAL as well as with
experimental trend. Nonetheless, the implicit and explicit integration techniques forecasted
quite different ears amplitudes. Actually, this is due to the contact treatement used together
with the shell elements in the ABAQUS/Standard wherein the sheet thickness variation is not
taken into account when the tooling is represented by analytical rigid surfaces.

Table 3: Number of time increments and CPU time obtained in the numerical simulations of the
cup-drawing experiment of an IF steel performed with shell reduced integration elements.

ABAQUS/Standard ABAQUS/Explicit
Yield criterion Increments CPU time (s) Increments CPU time (s)
*POTENTIAL 1,373 10,897 18,207 502
Ferron’s model 1,317 16,418 18,607 859

Fig. 12 compares the earing profile measured for an IF steel with the predictions determined
from the numerical simulations performed with the user subroutine VUMAT using reduced
integration shell elements. These results are plotted as a function of the method to identify
the anisotropy material parameters a and b in Ferron’s model. Firstly, the experimental ear
height, i.e., (h0,90 − h45 ) ∼
= 2 mm, is overestimated and underestimated when the method R is
used together with Hill’s quadratic yield criterion and Ferron’s model respectively. Likewise, it
is worthwhile to note that the orientation dependence of the uniaxial tensile flow stress, σ(α) ,
is overpredicted by Hill’s quadratic yield criterion and underpredicted by Ferron’s model, as
shown in Fig. 10(a). Conversely, a better prediction is achieved when the material parameters
describing the in-plane anisotropy are adjusted to the experimental σ(α) values, in this case by
Ferron’s model together with n = 2 and p = 2.
The finite element earing predictions determined with linear solid elements using reduced
integration rule are shown in Fig. 13. For the case of the user subroutine UMAT, one can observe
a good agreement with the results obtained from the simulations realized with ABAQUS/Explicit
using reduced integration shell elements. On the other hand, though the predictions given by
VUMAT are qualitatively in agreement with the material descriptions adopted, the ears heights
at the angular orientations 00 and 900 are overestimated. This can be atributed to the excitation
of hourglass modes which may produce mesh distortions in reduced integration elements.

Latin American Journal of Solids and Structures 4 (2007)

170 Luciano P. Moreira and Gérard Ferron

Figure 11: Comparison between the predicted earing profile for an IF steel determined with
reduced integration shell elements and Hill’s quadratic yield criterion as a function of the finite
element integration technique.

Figure 12: Comparison of the experimental and predicted earing profile for an IF steel determined
with reduced integration shell elements as a function of the identification method of the anisotropy
parameters in Ferron’s model.

Latin American Journal of Solids and Structures 4 (2007)

 Finite element implementation of plasticity model for sheet metal forming simulations 171

Figure 13: Comparison of the experimental and predicted earing profile for an IF steel deter-
mined with linear solid reduced integration elements as a function of the identification method
of the anisotropy parameters in Ferron’s model: (a) ABAQUS/Standard with UMAT and (b)
ABAQUS/Explicit with VUMAT.

5.2.2 Earing formation on a tinplate steel food-can

At this stage, is interesting to discuss the different explanations proposed in the literature
concerning the earing formation in the cup-drawing experiment of metal sheets. According
to the first one, the ears and respectively the troughs are formed at the angular orientations
α corresponding to the maximum and the minimum values of the plastic anisotropy coefficient
R [5,10]. This idea is based upon the higher and or lower radial strains obtained from orthoradial
compression in the cup flange, namely, at α + π/2, for larger and lower R−values respectively.
Another explanation, introduced by Panchanadeeswaran et al. [16], is based on the fact that
the stress states encountered at the edge of the flange and the flange cup varies between an
uniaxial orthoradial compression to a pure shear. In this case, the ear formation is related to
the shearing effects appearing when the compressive orthoradial or hoop stress is not coaxial
with the material in-plane orthotropy directions. This off-axis anisotropy effect can be quantified
by the parameter Γ(α) = ²̇p12 /²̇p11 proposed by Barlat and Richmond [3], which is defined with
Ferron’s model as: ¯
∂g (θ, α) /∂α ¯¯
Γ(α) = − (59)
2g (θ, α) ¯θ=π/4
√
Moreover, since the uniaxial yield stress at the angular orientation α is σ(α) = σb 2 g (π/4, α),

Latin American Journal of Solids and Structures 4 (2007)

172 Luciano P. Moreira and Gérard Ferron

see Fig. 1, the parameter Γ(α) for a fixed effective stress σ̄ is given by:

∂σ(α) /∂α
Γ(α) = − (60)
2σ(α)

where it should be noted that this relationship holds true for any anisotropic yield criterion
under the assumption of an isotropic work-hardening.
The two explanations lead to the same qualitative conclusions that the ears develops at 00
and 900 for materials with a minimum of R(α) and a maximum of σ(α) for 0 < α < π/2 and
near to 450 for materials with a minimum of σ(α) or a maximum of R(α) . However, as pointed
out from the numerical predictions of the cup-drawing for an IF steel, the explanation for earing
formation based upon the orientation dependence of the yield and/or flow stress determined in
the uniaxial tensile test, σ(α) , seems to be more quantitatively convincing than the one based
on the orientation dependence of the anisotropy coefficient, R(α) .
The idea of earing formation based upon the parameter Γ(α) has been firstly advanced in
the work of Panchanadeeswaran et al. [16] and further discussed by Barlat et al. [10]. Assuming
either a plane-stress state or a plane-strain hoop compression at the cup flange, Barlat et al. [10]
demostrate that the normalised angular variations of the radial velocity of the rim cup, vr , and
the uniaxial compression yield stress, σ(α) , are both linked to the parameter Γ(α) by:

1 dvr 1 dσ(α)
= = 2Γ(α) (61)
vr dα σ(α) dα

Panchanadeeswaran et al. [16] derived a simplified model to assess the normalised radius of
the cup flange which is given as:
· ¸½ · Z α ¸¾
r(α) r
=1+ − 1 1 − exp 2 Γ(α+π/2) dα (62)
rα0 rα0 0

where it should be noted that the parameter Γ is evaluated at angular orientation α = α + π/2,
i.e., under uniaxial compression whereas r and α0 are the initial blank radius and the reference
ear orientation angle respectively. This model can be solved together with Eq. (61) so as to
obtain a normalised amplitude of the cup flange profile as:
¯
∆r(α) r(α) − r ∆σ(α) ¯
= = ¯ (63)
∆rα0 rα0 − r σ(α) ¯α=α+π/2

In the following, the experimental flange profile of the deep-drawing of a tinplate steel food-
can [5] is compared to the predictions determined with the help of the simplified model proposed
by Panchanadeeswaran et al. [16], as defined by Eq. (63). The food-can is deep-drawn from a
tinplate steel with an initial blank radius of 74 mm and 0.24 mm thickness and tooling dimensions
given by: punch radius, 41.675 mm, punch nose radius, 1.25 mm, die gap, 42.05 mm, die profile
radius, 2 mm. The blank-holder force and the punch displacement are taken equal as 30 kN and

Latin American Journal of Solids and Structures 4 (2007)

 Finite element implementation of plasticity model for sheet metal forming simulations 173

39 mm respectively. The mechanical properties of the tinplate steel determined from uniaxial
tensile tests are given in Table 4. Thus, as for the IF steel, the value k = 0.2 and the relationship
B = 3A have been imposed in Eq. (45) so as to obtain the parameter A from the experimental
value of R45 . For this material, the orientation dependencies of both R(α) and σ(α) are quite
well described by using the method R together with the exponents n = p = 2, as shown in
Fig. 14. Figure 15 shows the predicted tinplate steel food-can determined with Hill’s quadratic
yield criterion, where a 3D geometry showing the complete flange profile has been adopted for
illustration purposes. Figure 16(a) compares the averaged experimental flange profile of the
tinplate steel food-can and the calculated results given by the simplified model, see Eq. (63),
with Hill’s yield criterion and Ferron’s model. A good agreement is achieved with Ferron’s model
since this description provides a small angular variation of either the uniaxial tensile yield stress
, σ(α) , and hence of the parameter Γ(α) , as shown in Fig. 16(b).

Table 4: Material parameters of the tinplate steel food-can.

Angle (Degrees) σy (MPa) K(MPa) ²0 N R(18%)

0 352 668 0.023 0.167 0.670
45 347 656 0.021 0.164 1.017
90 349 665 0.017 0.159 0.858

Figure 14: Comparison between the experimental normalised uniaxial tensile yield stress of the
tinplate steel and the predictions determined with Hill’s quadratic yield criterion and Ferron’s
model.

Latin American Journal of Solids and Structures 4 (2007)

174 Luciano P. Moreira and Gérard Ferron

Figure 15: Predicted tinplate steel food can obtained with ABAQUS/Explicit using Hill’s
quadratic yield criterion and reduced integration shell elements.

(a) (b)

Figure 16: Earing formation in the deep-drawing of a tinplate steel food can: (a) comparison
between the measured normalised flange profile and the predictions obtained using the simplified
model proposed by Panchanadeeswaran et al. [3] and (b) angular evolution of the parameter Γ(α)
introduced by Barlat and Richmond [10].

Latin American Journal of Solids and Structures 4 (2007)

 Finite element implementation of plasticity model for sheet metal forming simulations 175

6 Conclusions

The implementation procedures used with the Fortran user material subroutines UMAT and
VUMAT, available in the commercial finite element code ABAQUS, were firstly detailed for
either the plane-stress and general 3D stress states in the case of an isotropic linear elastic
material under the assumption of an isotropic work-hardening. Afterwards, the parametric
description of orthotropic plasticity in metal sheets under plane-stress conditions proposed by
Ferron et al. [8] is firstly presented and then extended to the 3D general case. The numerical
predictions obtained from the simulations of typical sheet metal forming experiments, namely,
the hemispherical punch stretching and the cup-drawing, have shown the good consistency of the
implementation of Ferron’s model in either implicit and explicit integration techniques available
in the ABAQUS code. Also, it has been verified that the response of the plastically deformed
sheet closely depends upon the yield stress levels in the different regions of the blank. In all cases
analyzed in this work, the strength aspects, i.e., the yield surface shape, are much more important
than the kinematic aspects of plastic yielding as measured, for instance, by the anisotropy or
Lankford coefficient R. In the hemispherical punch stretching and cup-drawing experiments, the
simple assumption of isotropic work-hardening seems to be sufficient to achieve a good agreement
with the experimental data, provided a yield function giving an accurate description of the yield
or flow stress levels for the stress states of interest is used in the numerical simulations. This
observation holds true for either the radial strains distributions or the earing formation in the
hemispherical punch stretching and cup-drawing sheet metal forming experiments respectively.

Acknowledgements: LPM is grateful to the “Fundação de Amparo à Pesquisa do Estado do

Rio de Janeiro” (FAPERJ) for the research grant APQ1 No. E 26/170.566/2004.

References

[1] B. Bacroix and P. Gilormini. Finite-element simulations of earing in polycrytalline materials us-
ing a texture-adjusted strain-rate potential. Modelling and Simulation in Materials Science and
Engineering, 3:1–21, 1995.
[2] F. Barlat. Crystallographic texture, anisotropic yield surfaces and forming limits of sheet metals.
Materials Science and Engineering A, 91:55–72, 1987.
[3] F. Barlat and O. Richmond. Prediction of tricomponent plane stress yield surfaces and associated
flow and failure behavior of strongly textured f.c.c polycrystalline sheets. Materials Science and
Engineering A, 95:15–29, 1987.
[4] A.J. Beaudoin, P.R. Dawson, K.K. Mathur, U.F. Kocks, D.A., and Korzekwa. Application of
polycrystal plasticity to sheet forming. Computer Methods in Applied Mechanics and Engineering,
117:49–70, 1994.
[5] L. Bourne and R. Hill. On the correlation of the directional properties of rolled sheets in tension
and cupping tests. Philosophical Magazine, 41:671–681, 1950.

Latin American Journal of Solids and Structures 4 (2007)

 View publication stats

176 Luciano P. Moreira and Gérard Ferron

[6] D. A. Burford, K. Narasimhan, and R.H. Wagoner. A theoretical sensitivity analysis of full-dome
formability tests : parameter study for n, m, r and µ. Metallurgical Transactions, 22A:1775–1788,
1991.
[7] D.C. Drucker. Relation of experiments to mathematical theories of plasticity. Journal of Applied
Mechanics, Transactions of the ASME, 16:349–360, 1949.
[8] G. Ferron, R. Makkouk, and J. Morreale. A parametric description of orthotropic plasticity in metal
sheets. International Journal of Plasticity, 10:431–449, 1994.
[9] P. Gilormini and Ph. Roudier. Abaqus and finite strain. Technical report lmt n, Laboratoire de
Mécanique et Technologie, Université de Paris VI, France, 1993.
[10] R. Hill. A theory of the yielding and plastic flow of anisotropic metals. Proceedings of the Royal
Society of London, A193:281–297, 1948.
[11] R. Hill. Theoretical plasticity of textured aggregates. Mathematical Proceedings of the Cambridge
Philosophical Society, 85:179–191, 1979.
[12] J.R. Knibloe and R.H. Wagoner. Experimental investigation and finite element modeling of hemi-
spherically stretched steel sheet. Metallurgical Transactions, 10A:1509–1521, 1989.
[13] R.D. Krieg and S.W. Key. Implementation of a time independent plasticity theory into structural
computer programs. Constitutive Equations in Viscoplasticity: Computational and Engineering
Aspects, ASME, 20:125–137, 1976.
[14] L.P. Moreira. Étude numérique de l’influence du modle de plasticité sur le comportement plastique
des tôles minces lors de l’emboutissage. Thèse de doctorat, Universit de Mez, France, 2002.
[15] L.P. Moreira, G. Ferron, and G. Ferran. Experimental and numerical analysis of the cup drawing
test for orthotropic metal sheets. Journal of Materials Processing Technology, 108:78–86, 2000.
[16] S. Panchanadeeswaran, O. Richmond, W. G. Fricke, and L. A. Lalli. The effect of texture on earing in
cup drawing of rolled high purity aluminum sheets. In Eighth International Conference on Textures
of Materials ICOTOM, pages 1103–1110, 1988.
[17] Pawtucket. ABAQUS Theory Manual Version 6.6. ABAQUS Inc., Rhode Island, USA, 2006.
[18] J. C. Simo and R. L. Taylor. Consistent tangent operators for rate-independent elastoplasticity.
Computer Methods in Applied Mechanics and Engineering, 48:101–118, 1985.

Latin American Journal of Solids and Structures 4 (2007)