1 INTRODUCTION

Metal forming processes involve large plastic strains

and severe strain path changes. Large plastic strains
lead in many metals to strong flow anisotropy. This
induced anisotropic behavior manifests itself in the
case of a strain path change by very different stress-
strain responses depending on the type of the strain
path change [1]. To describe two-stage strain path
changes, Schmitt and al. introduced in [2] the scalar
parameter

2 1
N N = (1)

Here, N
1
and N
2
are the strain-rate direction before
and after the strain-path change, respectively. If
equals -1, we have a load reversal, and if equals 0,
the strain-path change is referred to as orthogonal.
While many metals exhibit a drop of the yield stress
(Bauschinger effect) after a load reversal, some
metals show an increase of the yield stress after an
orthogonal strain path change (so-called cross
hardening effect, Figure 1). The reason for this
induced flow anisotropy is the development of
persistent dislocation structures during large
deformations [3]. These consist of walls of high
dislocation density separating low dislocation
density areas. The one side of each wall contains
excess dislocations of the same sign, and the other
side such dislocations of the opposite sign. After a
load reversal, plastic deformation takes place due to
the slip on the same slip systems but in opposite
direction. Excess dislocations, since they repel each
other, facilitate this slip, resulting in the Bauschinger
effect. After an orthogonal strain path change, new
slip systems are activated and the existing
dislocation walls act as obstacles, resulting in the
cross hardening effect.
To model the Bauschinger effect, the concept of
kinematic hardening has been successfully used for
years. However, the presence of kinematic
hardening results in a drop of the yield stress after an
orthogonal strain path change, which contradicts
tests results on materials exhibiting the cross
hardening effect. Accordingly, the combined
isotropic-kinematic hardening ansatz is insufficient
to describe the constitutive behavior of materials,
exhibiting both the Bauschinger and the cross
hardening effects, and has to be extended. Another
effect, not accounted for in the classical elasto-
plasticity, is the difference between the tensile and
compressive strengths (strength differential effect),
ABSTRACT: Sheet metal forming involves large strains and severe strain-path changes. Large plastic strains
lead in many metals to the development of persistent dislocation structures resulting in strong flow
anisotropy. While many metals exhibit a drop of the yield stress (Bauschinger effect) after a load reversal,
some metals show an increase of the yield stress after an orthogonal strain-path change (so-called cross
hardening). In this work we present a phenomenological material model whose structure is motivated by
polycrystalline modeling that takes into account the evolution of polarized dislocation structures on the grain
level the main cause of the induced flow anisotropy on the macroscopic level. The model considers besides
the movement of the yield surface and its proportional expansion, as it is the case in conventional plasticity,
also the changes of the yield surface shape (distortional hardening) and accounts for the pressure dependence
of the flow stress. All these additional attributes turn out to be essential to model the stress-strain response of
dual phase high strength steels subjected to non-proportional loading.
Key words: induced flow anisotropy, distortional hardening, cross hardening, strain-path changes, pressure
dependent plasticity, sheet forming
Modeling of sheet metal forming processes taking into account
distortional hardening
C. Barthel
1
, V. Levkovitch
1
, B. Svendsen
1

1
Department of Mechanical Engineering -D-44227 Dortmund, Germany
URL: www.mech.mb.uni-dortmund.de e-mail: c.barthel@mech.mb.uni-dortmund.de
exhibited e.g. by some steel materials [4].

Fig1. Orthogonal strain-path changes in FeP06 [3].

In this work we analyze the mechanical response of
a dual phase high strength sheet steel in one- and
two-stage loading processes. It turns out that the
material exhibits the strength differential effect and
that the deformation induced flow anisotropy is to
complex to be modeled by the combined hardening
ansatz. To describe the constitutive response of the
investigated material, we extend the isotropic-
kinematic-distortional hardening model, presented
by two of the authors at the ESAFORM 2006 [5], by
incorporating the first invariant of the stress tensor
into the yield condition.
2 MODEL DESCRIPTION
The current model, presented in [5] at the
ESAFORM06, is based on the standard elasto-
plastic framework with the yield function. Besides a
shift of the yield surface and its proportional
expansion as in the case of conventional plasticity,
the model also accounts for the changes of the yield
surface shape (distortional hardening). To be able to
describe the pressure dependent yielding behavior,
exhibited e.g. by some steel materials, we
incorporate the trace of the stress tensor into the
yield function [4]

Y tr a
e
+ = ) (T (2)

where Y is the yield stress, whose evolution is
determined by Voce isotropic hardening, a is a
material parameter, governing the plastic flow
pressure sensitivity, and

H M ) ( + =
Hill e
(3)
represents the equivalent stress measure. Here,

X T = :
(4)

is the difference between the deviatoric part of the
Cauchy stress tensor T and the back stress tensor X,
M
Hill
is the orthotropic fourth-order Hill tensor and
H the fourth-order tensor, introduced to represent
distortion of the yield surface. The plastic strain rate
is given by the associative flow rule

T
D
p

&
(5)

Note, that taking the derivative of the yield function
with respect to the deviatoric stress part insures the
plastic incompressibility of the material, which is a
good approximation also for steel materials
exhibiting the strength differential effect [4].
Ignoring any texture effects for simplicity, the
J aumann rate of the Cauchy stress tensor is given by
the isotropic hypo-elastic relation

( ) ( )I D D D D T
p p
+ = tr 2
o
(6)

Further, the evolution of the kinematic hardening is
given by the Armstrong-Frederick relation

( )
&
o
X D X =
p Sat X
X C (7)

The fourth-order tensor H, representing distortional
hardening, is described by the following evolution
equation

&
&
o
] ) ( [
) )( (
L
H N N I
N N H

+ =
dev Sat L
D Sat D
L C
H D C
(8)

Here, I
dev
is the deviatoric part of the fourth-order
identity tensor, N the direction of the plastic strain-
rate, H
D
the projection of H onto the direction of the
plastic strain-rate

) ( N N H =
D
H
(9)

and H
L
is the part of H orthogonal to N

) ( N N H H
L
=
D
H (10)

In addition, D
Sat
and C
D
represent the saturation
value and saturation rate parameters, respectively,
for H
D
, while L
Sat
and C
L
, respectively, represent
these parameters for the latent part H
L
.
To understand the behavior of the proposed model,
we consider an initially annealed material state with
vanishing H and assume L
Sat
< 0 as well as for
simplicity D
Sat
=0. If the material is subjected to
proportional loading in the direction N
1
, the
directional part H
D
does not evolve since D
Sat
=0.
Consequently, only H
L
evolves, saturating to the
value L
Sat
(I
dev
-N
1
N
1
), which is orthogonal to N
1
.
Such an evolution of H does not influence the yield
strength of the material in the loading direction N
1
,
while the strength in all directions, being orthogonal
to that of the loading direction, increases. After an
orthogonal strain-path change with re-loading
direction N
2
, the directional part as the projection of
H onto N
2
takes on a negative value and saturates
toward zero during re-loading due to D
Sat
=0. This
results in shrinkage of the yield surface in the
direction of re-loading. On the other hand, the yield
surface expands in direction N
1
since N
1
is now
orthogonal to the current loading direction. This
behavior corresponds qualitatively to that obtained
by Peeters et al. in [6], using the microstructural
model.
With the presented model it is possible to describe
both the Bauschinger and the cross hardening
effects, simultaneously. The model was
implemented into commercial FE code ABAQUS
via UMAT interface.

3 MODEL APPLICATION
To demonstrate the capability of the presented
model to describe complex pressure dependent
hardening behavior under non-proportional loading,
we use it to simulate the constitutive behavior of a
dual phase high strength sheet steel. Due to
commercial dependency, the experimental results
can be given only in scaled format.
The investigated material exhibits a difference
between the tensile and compressive yield strength.
Tension and compression experiments in rolling
(RD) and transverse (TD) directions after 10% pre-
tension in RD show that the material exhibits both
the Bauschinger and the cross hardening effects.
We first try to simulate the experimental data using
the combination of Voce isotropic and Armstrong-
Frederick kinematic hardening (the classical
combined hardening ansatz). The hardening
parameters were fitted to the tensile test in RD and
to the compression test in RD after 10% pre-tension
in RD. Both tests (monotonic and reversed) are
modeled adequately by the combined hardening
ansatz. However, if we use the combined hardening
model with parameters extracted from these both
tests to model the RD monotonic compression test
on as-received material and the TD compression test
after 10% pre-tension in RD, the predictions are not
satisfactory. We conclude, that such a behavior as
exhibited by the investigated dual phase steel cannot
be described by a classical pressure independent
combined hardening ansatz.
In contrast, applying the model, described in the
previous section, it was possible to simulate with a
good accuracy the whole set of experimental data
with only one set of material parameters (Fig. 2 and
3).


Figure 2. Comparison of experimental and
numerical results. Above: uniaxial tension in RD.
Below: uniaxial compression in RD.

Accordingly, the mixture of isotropic, kinematic and
distortional hardening in combination with the yield
function, depending on the trace of the stress tensor,
seems to be an adequate constitutive ansatz to model
the pressure dependent hardening behavior of dual
phase high strength steels. Due to commercial
dependency, it is not possible to give the material
parameters.


Figure 3. Comparison of experimental and numerical results.
Above: uniaxial compression in RD after 10% tensile pre-strain
in RD. Below: uniaxial tension and compression in TD,
respectively, after 10% tensile pre-strain in RD

As an application example for a structural spring
back simulation we consider the strip drawing test
proposed as a spring back benchmark test at the
NUMISHEET93 conference. Figure 4 shows
experimental and numerical results for the released
strips have being drawn under the following process
characteristics: the ratio of the sheet thickness to the
punch radius and to the die radius were 0.2 and 0.14,
respectively. The sheet strip used in this test was not
pre-deformed. The application of the proposed
model yields a good agreement with the experiment.
Similarly good agreements were also reported in
different works where only the combined hardening
were used. The superiority of the proposed model
over the combined hardening approach due to its
more realistic description of the induced anisotropy
is expected to become evident in the drawing test on
strips, pre-deformed in different directions before
drawing.
An other model capable to model the cross
hardening effect was published in [7].

Figure 4. Comparison of experimental and numerical results:
released strips after drawing (the half of the geometry is
displaced).
Acknowledgments
The authors would like to thank ThyssenKrupp Steel
Company for providing the experimental data. Partial financial
support provided by the German National Science Foundation
(DFG) is gratefully acknowledged.
REFERENCES
1. E.F. Rauch and S. Thuillier, Rheological behaviour
of mild steel under monotonic loading conditions and cross-
loading, Mater. Sci. Engng. A164 (1993), pp. 255-259.
2. J .H. Schmitt, E. Aernoudt and B. Baudelet, Yield loci
for polycrystalline metals without texture, Mater. Sci. Eng. 75
(1985), pp. 13-20.
3. S. Bouvier, J .L. Alves, M.C. Oliveira and L.F.
Menezes, Modelling of anisotropic work-hardening behaviour
of metallic materials subjected to strain path changes Comp.
Mater. Sci. 32 (2005), pp. 301-315.
4. W.A. Sptitzig, R.J . Sober and O. Richmond, The
effect of hydrostatic pressure on the deformation behavior of
maraging and HY-80 steels and its implication for plasticity
theory Metal. Trans. A 7A, (1976), pp. 1703-1710.
5. V. Levkovitch, B. Svendsen., Wang,
Micromechanically motivated phenomenological modeling of
induced flow anisotropy and its application to metal forming
processes with complex strain path changes, The 9
th

International ESAFORM Conference on Material Forming,
Glasgow, UK, April 26-28, 2006.
6. B. Peeters, S.R. Kalidindi, C. Teodosiu, P. Van Houtte
and E. Aernoudt, A theoretical investigation of the influence
of dislocation sheets on evolution of yield surfaces in single-
phase b.c.c. Polycrystals, Journal of the Mechanics and
Physics of Solids 50 (2002), pp. 783-807.
7. J . Wang, V. Levkovitch, F. Reusch, B. Svendsen, J .
Hutink, M. van Riel, On the modeling of hardening in metals
during non-proportional loading, International Journal of
Plasticity 2007, online doi:10.1016/j.ijplas.2007.08.009
1 INTRODUCTION
In our laboratory, an extensive work has been
developed since ten years, in order to describe the
ductile damage modelling in bulk and sheet metal
forming by using advanced constitutive modelling
[1-8]. Based on the thermodynamics of irreversible
processes with state variables, the advanced
approach aims to describe the coupling between the
main thermo-mechanical fields and the ductile
damage. These models have been implemented in
ABAQUS using the available user subroutines
(Umat, Vumat, Uel and Vuel). An adaptive meshing
and remeshing procedure with geometrical (local
curvature at the contact points between the tools and
the part) and physical error estimates (stress, plastic
strain, damage) which kills the fully damaged
elements has been developed based on the work by
Rassineux et al [9, 10].
In the present work, this approach is shortly
discussed from both theoretical and numerical points
of view. Applications are made to the simulation of
a hole blanking and a hole expansion processes
using a high-strength steel material provided by
Arcelor Mittal company. A comparison between the
cases of the hole expansion of an ideal and
previously blanked holes is made.
2 ABOUT THE COUPLED CONSTITUTIVE
EQUATIONS
The constitutive equations are formulated on an
appropriated Eulerean intermediate configuration
having the same Lagrangian orientation as the initial
undeformed configuration according to the RFF
method (see [6-8] among many others). Using this
rotated objective formulation a complete set of
constitutive equations can be obtained. In this paper
a non associative and anisotropic plastic formulation
accounting for the nonlinear isotropic and kinematic
hardening fully coupled with the isotropic damage is
ABSTRACT: In this work, an anisotropic elastoplastic finite element model strongly coupled with ductile
damage is applied to simulate some metal forming tests used by Arcelor Research. The F.E. code is linked to
a 2D adaptive remeshing procedure. First, the anisotropic elastoplastic model with non linear kinematic and
isotropic hardening strongly coupled with ductile damage is presented. This model is written in finite plastic
deformation through the so called rotated frame formulation using a non associative plasticity assumption
with state variables. The adaptive analysis including the 2D mesh adaptation together with adaptive loading
sequences and fully damaged elements deletion is described. This 2D adaptive procedure is applied to some
metal forming tests with various high properties steel materials as the hole blanking and expansion. Two
different cases are performed: (i) an initial hole is expanded starting from the zero stress at virgin state; (ii) the
hole is first formed by blanking operation followed by the expansion process taking into account the residual
fields (stress, strain, damage).

Key words: Finite anisotropic plasticity, ductile damage, numerical simulation, adaptive mesh.
F.E. elastoplastic damage model with 2D adaptive remeshing procedure
for fracture prediction in metal forming simulation
H. Badreddine
1
, C. Labergre
1
, K. Saanouni
1
, W. Rajhi
1,2
, A. Rassineux
3
, D. Kircher
4

1
UTT de Troyes, ICD/Lasmis 12 rue Marie Curie BP, 10010 Troyes, France.
URL: www.utt.fr e-mail: houssem.badreddine@utt.fr ; carl.labergre@utt.fr; khemais.saanouni@utt.fr

2
ESST de Tunis LMMP, rue Taha HusseinB.P. 56, 1008 Tunis, Tunisia

3
University of technology of Compiegne, Laboratoire Roberval, UMR 6253 du CNRS , centre de Recherches
de Royallieu, BP 20529, 60205 Compiegne e-mail: alain.rassineux@utc.com

4
Arcelor Innovation R&D Voie Romaine, 57283 Maizires-ls-Metz, France

considered. This model introduces several material
parameters which should be identified given in the
following:
is the fourth order elastic properties tensor of
the non damaged material.

y
is the initial yield stress in simple tension.
C is the kinematic hardening modulus and Q is
the linear isotropic hardening modulus.
a and b characterize respectively the kinematic
and isotropic hardening non linearity.
, S , s and
0
Y characterize the ductile damage
evolution.
H is fourth order plastic anisotropic tensor
characterized by six material constants
F, G, H, L, M and N .
For the sake of shortness the constitutive are not
given. The reader is invited to refer to [6-8] for more
details about the formulation of the overall fully
coupled constitutive equations.
3 NUMERICAL ASPECTS
The model developed above has been implemented
into ABAQUS/Explicit FE software for metal
forming simulation thanks to the user subroutine
Vumat (ABAQUS Theory Manual). The dynamic
explicit global resolution schema is developed in
detail in (ABAQUS Theory Manual) considering
the contact with friction of Coulomb type
characterized by the friction parameter . The
computation of the stress tensor on the rotated
(Lagrangien) configuration is required in order to
evaluate the internal stress vector at each integration
point inside each finite element for the end of each
time increment. This is achieved by integrating all
the constitutive equations of the model presented
above including the ductile damage. The classical
incremental and iterative elastic predictor plastic
elastic prediction plastic correction method [11-13]
is used together with the reduction of the number of
differential equations to be solved. This procedure is
fully described in ([1-8]) and therefore is not
presented here.
4 APPLICATIONS
4.1 Identification of the material parameters
Arcelor Research provided several experimental
results of uniaxial tension tests until final fracture
with sheet specimen cut on 0, 45 and 90
orientations with respect to the rolling direction. By
using this experimental database, the quasi-isotropy
of the material has been shown. Accordingly, the
values of the different parameters obtained for the
studied material are the following:
E= 195000 MPa, =0.3,
y
=405 MPa, Q=5500
MPa, b=10, C=38000 MPa, a=290, F=G=H=0.5,
L=M=N=1.5, S=45 MPa, s=1 , =2,
0
y =0MPa
and the characteristic mesh length h
min
=0.1 mm.
4.2 Cutting process simulation
The scheme of the cutting process is given in figure
1.

Fig.1. Scheme of the cutting process
The problem is supposed as axis-symmetric. The
simulation is made using the isotropic version of the
model presented above and with the adaptive
remeshing procedure with CAX4R axis-symmetric
elements from the ABAQUS element library. The
parameters governing the adaptive analysis are:
max
h = 0.8mm,
p
max
h =0.2mm,
p
min
h = 0.07 and
d
min
h =0.03mm.
Some steps of the cutting process are given in figure
2. The path (i.e. the location of the deleted fully
damaged elements) followed by the macroscopic
crack is clearly shown. We can also observe that the
mesh is refined in the vicinity of the contact points
between the tools and the sheet due to the small
radius of the tools. Also note that the mesh size is
coarsened after the formation. The sheet is
completely cut after a punch displacement about
0.63 mm, which represents about 25% of the sheet
thickness.
4.3 Hole expansion simulation
The scheme of the hole expansion process is given
in figure 3. The problem is supposed to be axis-
symmetric and the tools are taken as rigid bodies.
The computation is performed without adaptive
remeshing with CAX4R axis-symmetric elements
from ABAQUS element library. Three cases are
considered and compared:
Case 1: Hole expansion starting from a perfect
hole. In that case no adaptive remeshing is used.
Case 2: Hole expansion using pre-blanked hole
and the expansion is made in the blanking
direction.
Case 3: Hole expansion using pre-blanked hole
and the expansion is made in the inverse direction
of the blanking direction.
For the last two cases, all the mechanical residual
fields (stress, plastic strain, damage,) due to
previous blanking operation are transferred as well
as the final geometry of the blanked hole (see figure
4).


(a) U=0.07mm (b) U=0.31mm

(c) U=0.50mm (a) U=0.63mm
Fig.2. Damage maps obtained for different punch
displacement of the cutting process.

Fig.3. Scheme and geometry of the hole expansion process.


Fig.4. Transferred geometry and residual fields of the pre-
blanked hole.
Some steps of the simulation of the case 1 are shown
in figure 5. In this figure we can observe that for a
5.62mm punch displacement, some elements located
at the contact area between the punch and the sheet
in the vicinity of the hole are damaged due to the
contact pressure. A macroscopic crack is initiated in
the external side of the hole for a punch
displacement of about 24.75mm (see figure 5.b).
Some steps of the simulation of cases 2 and 3 are
given in figures 6 and 7 respectively. We can
observe that in both cases, a damaged zone develops
at the area of contact with the tool for different tool
displacements u=3.2 mm for case 2 and u=5.1mm
for case 3. When comparing case 1 (Figure 5) and
case 2 (Figure 6) the damage distributions are
clearly different indicating the important role of the
residual fields generated by the previous blanking
operation.


(a) U=5.62mm

(b) U=24.75mm
Fig.5. Damage maps obtained for different values of the punch
displacement (process with perfect hole).


(a) U=1.5mm

(b) U=3.2mm
Fig.6. Damage maps obtained for different values of the punch
displacement (process with pre-blanked hole made in the
cutting direction).

(a) U=1.5mm

(b) U=5.1mm
Fig.7. Damage maps obtained for different values of the punch
displacement (process with pre-blanked hole made in the
inverse cutting direction)
5 CONCLUSION
In this paper, an elastoplastic model strongly
coupled with ductile damage has been briefly
presented. This mechanical model have been both
implemented into ABAQUS F.E code and used in
connection with a 2D adaptive meshing and
remeshing procedure. This adaptive numerical
methodology is used to simulate blanking and
expansion of holes inside a quasi-isotropic sheet.
The results are very encouraging and show the
ability of the proposed modelling to predict the
rupture in sheet metal forming. This adaptive
analysis should be extended to 3D analysis and
compared to experimental results in future research.
REFERENCES
1. Saanouni K., Hammi Y., Numerical simulation of damage in
metal forming processes , in Continuous Damage and
Fracture , Editor A. Benallal, Elsevier, ISBN. 2-84299-247-4,
pp :353-363, 2000
2. Saanouni, K., Forster C. and Ben Hatira F., On the Anelastic
Flow with Damage , Int. J. Dam. Mech., 3:140-169, 1994
3. Saanouni K. and Chaboche J.L., Computational Damage
Mechanics. Application to Metal Forming, Chapter 7 of the
Volume 3 : Numerical and Computational methods (Editors:
R. de Borst, H. A. Mang), in Comprehencive Structural
Integrity, Edited by I. Milne, R.O. Ritchie and B. Karihaloo,
ISBN: 0-08-043749-4, 2003, Elsevier, Oxford.
4. Lemaitre, J., and Chaboche J.L., Mcanique des Matriaux
Solides, Dunod, Paris, 1985.
5. Khelifa M., Badreddine H., Belamri N., Gahbich M. A.,
Saanouni K., Cherouat A., Dogui A., Effect of anisotropic
plastic flow on the ductile damage evolution in hydrobulging
test of thin sheet metal, Int. Journal of Forming Processes, Vol.
8, N:2, pp : 271-289, 2005
6. M.,Khelifa, K., Saanouni, H., Badreddine, M.-A., Gahbiche,
and A., Dogui, "Plasticit anisotrope couple un
endommagement ductile isotrope: Appliction au gonflement
hydraulique de tles minces", Revue Europenne des Elments
Finis, Vol15., N7-8, pp. 891-908, 2006.
7. Badreddine H., Saanouni K., Dogui A., Gahbich M.A.,
Elastoplasticit anisotrope non normale en grandes
dformations avec endommagement. Application la mise en
forme de tles minces. Revue Europenne de Mcanique
Numrique, Volume 16, N6 et 7,2007, pp :913-940.
8. H., Badreddine, "Elastoplasticit anisotrope endommageable
en transformations finies : Aspects thoriques, numriques et
applications". Thse de Doctorat, 2006, Universit de
Technologie de Troyes (France) Ecole Nationale
dIngnieurs de Monastir (Tunisie).
9. A. Rassineux, An automatic mesh generator for planar
domains, StruCome (1991), p. 519-531.
10. C. Labergere, A. Rassineux,, K. Saanouni, Endommagement et
procd de mise en forme. Apport du maillage adaptatif, 8
th

Colloque National en Calcul des Structures, Giens, France, 21-
25 Mai, (2007), CD.
11. Simo J.C., Hughes T.J.R., Computational inelasticity, Springer,
New York, 1998
12. J.C. Simo, and M., Ortiz, A Unified Approach to Finite
Element Deformation Elastoplastic Analysis Based on the Use
of Hyperelastic Equations , Comp. Meth. Appl. Meth. Engng.,
vol 49, 1985, pp, 221-245.
13. T. J. R., Hughes, J., Winget, Finite rotation effects in
numerical integration of rate constitutive equations arising in
large-deformation analysis , Int. J. Nume. Meth. Engng, 15,
1980: 1862-1867.
14. D. Freed, and K. P., Walker, Exponential integration
algorithm for first-order ODEs with application to
viscoplasticity. ASME Summer Conf. On Mechanics and
Materials Recent Advances on Damage Mechanics and
Plasticity, Tempe, 1992.

1 INTRODUCTION
Most sheet steels commonly used for automotive
and other manufacturing applications have good
mechanical properties such as ductility, resistance
and Youngs modulus. These materials can be bent
under severe conditions and be subjected to high
loading without fracture or cracking. In a pioneers
work [1], it has been established that the final
mechanical characteristics of sheet metal
components do not depend only on the initial
material properties but also on the geometry used for
producing sheet components. Davies et al. [2]
conducted one of the few experimental studies
available in the literature on straight-edge flanging
(wiping die bending). They studied the influence of
the die radius, the punch-die gap, the kind of steel
and the yield stress on sheet springback.
The optimisation of forming processes aimed at the
production of net-shape components and high
resistant products is nowadays of fundamental
importance. As a reliable methodology, design of
experiments associated with response surface
approximation is retained in several cases for the
analysis and optimisation of sheet-metal forming
problems [3, 4]. Todoroki et al. [5] have described a
new experimental method to optimise stacking
sequence by applying a response surfaces method to
composite cylinder products. Similarly, Chou et al.
[6] have used the response surface technique for
analysing springback as a function of both material
properties and tooling parameters.
In this paper, the bending and unbending operations
of sheet metal using a mechanical press are
investigated. The study is based on experiments and
the simulation of wiping-die bending. A special
specimen geometry was designed, having a central
oblong hole and is representative of the real
geometry of automotive safety parts. The
mechanical behaviour of this specimen is also
similar in unbending situations. Results for the
maximum unbending load were plotted by applying
a response surface technique.
2 DESIGN PROCEDURE
2.1 Experimental study of unbending
The shock resistance of the bend specimens has been
characterised by quasi-static unbending tests on
ABSTRACT: Anchorage parts for automobile safety belts and other safety parts must resist shock loadings
without breaking. They are typically made from High Strength Low Alloy sheet metal and fabricated by
blanking and bending operations. The study of their behaviour during fabrication and their resulting
mechanical properties has been conducted experimentally and numerically. The experimental results were
used to validate the numerical simulation. The resulting material damage is taken into account by a user
subroutine in the Abaqus Standard Finite Element code. Damage is one of the objective functions intervening
in the forming process of safety parts as well as in the prediction of unbending and shock loads. This study is
based on the use of the design of experiments technique and the approximated representation by response
surfaces. For unbending operations representative of dynamic shock loading conditions, the objective
functions represent the maximum unbending load and maximum damage. The parameters that represent the
die radius R
d
and the sheet/die clearance C are optimised in order to obtain the most resistant safety part
possible.
Key words: Safety parts, Bending, Unbending, F.E simulation, Damage, Optimisation, Response Surfaces.
Optimisation of the bending process of High Strength Low Alloy sheet
metal: numerical and experimental approach
R .Bahloul
1
, Ph. Dal Santo
2
, A. Potiron
2

1
ENIM - Dpartement de Gnie Mcanique, LGM, Avenue Ibn Eljazzar, 5019 Monastir, Tunisie
URL: www.enim.rnu.tn/ e-mail: bahloul_riadh@yahoo.fr

2
ENSAM Angers - 2 Boulevard du Ronceray, 49035 Angers, France
URL: www.angers.ensam.fr/ e-mail: philippe.dalsanto@angers.ensam.fr; alain.potiron@angers.ensam.fr

specimens with holes. A series of experiments was
carried out with a tensile testing machine (ZWICK-
50KN). Figure 1 shows the experimental setup, with
universal cardan joints. The tests consist of
unbending then tensile loading of specimens at low
and constant velocity. A data acquisition system was
used to record force-displacement data.






Fig. 1. Tensile testing machine
In order to determine the influence of the process
parameters on the mechanical behaviour of the bent
specimen, tests were conducted on specimens
fabricated with different die radii and clearances.
The specimens were made from 4mm thick steel
sheets H.S.L.A S500MC. The values of the
maximum unbending loads are summarised in the
following table:

Table1.Experimental results of maximum unbending load (kN)
R
d
(mm)
C (mm)
-0.6 -0.4 -0.2 0 0.2 0.4 0.6
1 1.6 0.75 2.5 2.25 1.75 3.1 2.55
2 9.6 17.3 15.25 21.5 23.9 24.25 24.875
4 24.3 27.85 27.75 27.25 30.05 29.875 29.5
6 28.25 29.5 29 28.95 30.375 31.625 31.95
3 FINITE ELEMENT APPROACH
3.1 Lemaitres damage model
In this paper, the Lemaitre [7] material behaviour
law coupled with damage has been used to simulate
the process. The algorithms implemented in the
finite element code for integration of the non linear
constitutive equations are the so-called radial return
algorithms, and they are used to solve the equations
in an incremental form. They are based upon the
notion of an elastic predictor-plastic corrector where
a purely elastic trial state is followed by a plastic
corrector phase [8, 9]. The Von Mises yield function
coupled with damage is given by:
] D)[ (1 f
0 el eq
+ =
(1)
where D is the damage variable,
el
is the yield
stress,
eq
is the equivalent stress and
0
is the non
linear hardening law given by:
n
eq 0
K = (2)
According to Lematre and Chaboche, the increment
of the damage variable D is defined by:
eq
2
eq
H
D R
c
) 2 - 3(1 ) (1
3
2 D
D
(
(

|
|

\
|

+ +

=
(3)
This model uses 4 material constants to define the
damage: represents the Poissons ratio,
D
is a
threshold strain at which damage is initiated,
R
is
the logarithmic strain at fracture and
eq
represents
the increment of the equivalent logarithmic plastic
strain. Identification of this model consists of
determining the four coefficients
c
D ,
R
,
D
and
by mean of one-dimensional tensile tests [10].
3.2 Numerical simulation of the unbending
operation
3.2.a Geometry
A 3-Dimensional model of the specimen with the
oblong hole is used to simulate the process. Bending
and unbending are consecutively modelled, taking
material damage into account.







Fig. 2. Three-dimensional numerical simulation of bending-
unbending tests
The tooling (punch and die) were modelled as rigid
bodies while the workpiece was considered to be
deformable with an elastic-plastic behaviour. In
these simulations the finite element mesh consists of
linear cubic elements (C3D8), which have been
showed to be sufficient to describe the bending and
the unbending process. The material behaviour on
the part/tool interfaces is governed by the Coulombs
friction law with a value of 0.15 for the coefficient
of friction. Figure 2 shows the different steps of the
numerical simulation by the ABAQUS/Standard
code with the use of the UMAT subroutine
discussed above.
Displacements were applied in the 45 direction
(that of unbending) at all nodes situated on the mid-
thickness level in the lower half of the specimen.

Upper cardan
joint
Lower cardan
joint
Bent part

Mobile
support
Fixed
support
Hardening effect
F
b) Sequence of
unbending
a) Bending end.
Springback
Fracture zone
c) Unbending end and
tensile test
F
The forces on these nodes are thus noted and their
sum provides the value of the unbending force.
3.2.b Evolution of damage and unbending load
This study is aimed at improving the understanding
of the physical mechanisms of damage and fracture.
Therefore, as already discussed, the Lemaitre
damage model coupled with an elasto-plastic
behaviour law has been used. The state of damage in
the specimens at the end of the unbending operation
has been investigated. At a certain time in the
operation, both the interior and exterior zones are
subjected to uniaxial tension, which correspond to
final stage of unbending. Figure 3 highlights this at
the end of the unbending step.







Fig. 3. Damage fields in tensile and compressive zones during
the unbending test
The damaged zones are located in the middle of
specimen, on the inside and outside of bend. Figure
3 shows that the interior zone is damaged all along
the width of specimen with more damage near the
hole where the damage accumulates during bending
and unbending.
In the same way, we chose to characterise the shock
resistance of pre-bent components. Unbending loads
determined via simulation have been compared with
the experimental results. Figure 4 shows the
variation of unbending force with respect to
displacement of the mobile support for the two
approaches.











Fig. 4. Evolution of the unbending force versus the
displacement
At the beginning of the unbending operation, the
straightening of the specimen corresponds to a large
displacement of the mobile support until the
specimen is essentially straight. During this first
stage the force increases slowly. In the second stage,
a rapid increase in the unbending force is noted. This
is because the specimen is simply loaded in tension
until it final ruptures.
The results provided by the two approaches are
coherent since the trends shown by both curves are
very similar. Also, the value of the relative error for
the maximum unbending force does not exceed
3.26%.
3.2.c Construction of the experiment matrix
This study is aimed at determining the influence of
two principal geometrical process parameters (the
die radius R
d
and the clearance C between the sheet
and the tooling) on the maximum load reached
during the unbending operation. They vary in the
intervals R
d
[1, 6] mm and C [-0.6, 0.6] mm
respectively with a uniform increment of 0.2 mm.
The optimisation is based on a design of experiment
obtained by numerical modelling in view to define
approximations of objective functions. The results of
simulations are reported in table 2.

Table2. Numerical results of maximum unbending load (KN)
given in form of experiment matrix
R
d
(mm)
C (mm)
-0.6 -0.4 -0.2 0 0.2 0.4 0.6
1 2.2 2.575 2.95 3.375 4.26 4.67 4.96
2 10.75 13.92 17.1 23.65 26.21 27.13 27.98
4 25.458 28.5 29.1 29.43 30.32 31.25 31.72
6 29.37 30.71 30.98 31.266 32.68 33.09 33.679
4 RESULTS AND ANALYSIS
Figures 5.a and 5.b illustrate the global evolutions of
the maximum relative unbending force determined
by experimental tests compared to the numerical
predictions. They are given by response surfaces
obtained by a cubic approximation of the objective
functions with respect to two process parameters R
d

and C previously described. Good agreement can be
seen between the simulated and experimental trends
of the two response surfaces. The variation of
unbending force is more sensitive to the die radius
than to the clearance. The first observation concerns
the material strengths. It can be seen that the
unbending force increases with the increasing of die
radius and relative clearance.


0
4
8
12
16
20
24
28
32
0 5 10 15 20 25 30
Displacement (mm)
U
n
b
e
n
d
i
n
g

l
o
a
d

(
K
N
)
Experiment_Rd4_C0
Numerical_Rd4_C0
Internal zone
of the bent part
External zone
of the bent part





















Fig. 5. Evolution of the maximum relative unbending force
versus relative punch-sheet clearance and die radius.
a) Experimental approach b) Results of Finite Element analysis

The second observation concerns the influence of
negative clearance on the behaviour of parts in
unbending. As noticed in figures 5.a

and 5.b, they
result in a decrease of strength. The unbending force
reaches its maximum value for bigger clearances for
all values of the die radius.
5 CONCLUSIONS
Straightening of specimens after forming by bending
was studied experimentally by the means of an
unbending test and numerically in order to predict
their mechanical behaviour. The distribution of
damage in concentration zones situated in the middle
of the specimens has been observed. The FE results
using the Lemaitre damage model (for which
damage increases only when the stress tensile) show
most damaged zones. They are located in the central
part at the inside of the bend. Both the numerical
model and experimental tests allowed the prediction
of the evolutions of the static unbending force with
respect to displacement of the mobile support. Good
coherence between the unbending loads predicted by
the Finite Element analysis and the experimental
data is achieved. This confirms the reliability of the
numerical model. It was shown that the process
parameters like die radius and punch-sheet clearance
have a significant influence on the mechanical
behaviour of safety parts.
ACKNOWLEDGEMENTS
The authors are grateful to Deville s.a. Company for its
technical support.
REFERENCES
1. L. Kurt, Handbook of Metal Forming, First edition,
McGraw-Hill Book Company, U.S.A (1985).
2. Davies RJ., Liu YC., Control of springback in flanging
operation, Journal of applied Metalworking, 3, (1984)
142-147
3. Bahloul R., Ben-Elechi S., Potiron A., Optimisation of
springback predicted by experimental and numerical
approach by using response surface methodology,
Journal of Materials Processing Technology, 173,
(2006) 101-110
4. Bahloul R., Mkaddem A., Dal Santo Ph., Potiron A.,
Sheet metal bending optimisation using response
surface method, numerical simulation and design of
experiments, Int. Journal of Mechanical Sciences, 48,
(2006) 991-1003
5. Todoroki A. and Ishikawa T., Design of experiments for
stacking sequence optimizations with genetic algorithm
using surface approximation, Composites Structures,
64, (2004) 349-357
6. Chou IN. and Hung C., Finite element analysis and
optimization of springback reduction, Int. Journal of
Machine Tools and Manufacture, 39, (1999) 517-536
7. Lemaitre J., A continuous damage mechanics model for
ductile fracture, Journal of Engineering Materials and
Technology, 107, (1985) 83-89
8. M.A. Criesfield, Non linear finite element analysis of
solids and structures, Vol.1, Wiley (1991).
9. Marques J.M.M.C., Stress computation in
elastoplasticity, Engineering computations, 1, (1984)
42-51
10. Mkaddem A., Exprimentation et simulation du pliage
de tles H.L.E. Prvision du comportement en service
des pices plies, Thse de doctorat, ENSAM Angers
16 dcembre 2003
Experimental
(a)
Rd/t
C/t
(b)
Numerical
Rd/t
C/t
1 INTRODUCTION
The technologies of sheet metal forming as
stamping, deep drawing, stretching and incremental
forming are relevant and complex production
processes in the automotive, aeronautic and kitchen
device industries. The main metals utilized in these
industries are steel, aluminium and titanium alloys.
Formability of sheet metals is the important and
complex issue related to the optimization and quality
control of the final product.
Historically, formability has been assessed by simple
testing as the Erichsen test. However, formability of
sheet metals is a complex attribute that involves
different variables as the process parameters and the
material properties. In addition, sheet metals defects
or inhomogeneities as thickness variations, porosity,
roughness and variations in the plastic properties
influence the limiting strains of sheet metal forming.
Later, the concept of Forming Limit Curves - FLC
has been developed to assess sheet metal formability
[1]. The Forming Limit Diagram - FLD displays the
principal in-plane true strains,
1
and
2
, attained by
the sheet metal at critical points during testing
methods or production process, i.e. the FLC. Two
types of curves can be plotted: local necking or
fractures strains.
Experimental and theoretical predictions of the local
necking and fracture strains have been investigated
by academic researchers and industry professionals.
Various mathematical models have been proposed to
predict the limiting curves of sheet metals for deep
drawing, stretching and constant or variable strain
path processes [2,3,4,5,6]. These theoretical models
considered the material plastic properties and
thickness imperfections only to predict the limiting
strains, but do not taking into account the nominal
value of thickness size.
2 LOCAL NECKING MODELLING
The stretch forming or deep drawing of sheet metal
are considered failed when terminated by fracture or
local necking. Within the biaxial stretching region of
the FLD, experimental investigations have shown
that rupture is generally preceded by local necking
or by shear process [5]. The process of strain
localization in sheet metal forming have been
investigated by the author [6], using the concept of
strain gradient development. The mathematical
model assumes that the process of neck initiation
and growth is a continuous process of strain
localization due to initial variations in thickness of
the sheet metal. This initial variation in thickness is
characterized by the parameter which is the initial
normalized gradient in the transversal area or defect.
ABSTRACT: A new approach on the influence of sheet thickness and material plastic properties on the limit
strains in thin sheet metal forming is developed, using the strain gradient model that predict the local necking
onset from initial thickness imperfections. As a result of the analysis, the definition of a roughness concept
parameter is presented: the initial roughness profile inclination to thickness ratio parameter. Also, the critical
normalized strain gradient at the onset of local necking can be calculated from the initial assumed roughness
profile. A brief review of stretch forming of sheet metal, the diagrams and the limiting strain curves for local
necking, FLC, and the limit strain theories are presented. The main characteristics of the sheet metal forming
processes are also identified and are based in the press shop practice. The limit strains for different thickness
sizes are obtained using the present model and a numerical code developed by the author. Present theoretical
model produced reasonable predictions about the influence of thickness size on the FLC.
Key words: Stamping, Deep drawing, Forming Limit Curve, Defects, Thickness.
Influence of thickness size in sheet metal forming
J.D. Bressan


Department of Mechanical Engineering - Centre for Technological Sciences - Santa Catarina State
University - Campus Universitrio - 89223-100 Joinville, Brazil.
URL: www.joinville.udesc.br e-mail: dem2jdb@joinville.udesc.br
The local necking or limit strain occurs when the
strain gradient attains a critical value
crit
= 20.
In the present approach, the influence of thickness
size h
o
on the limit strain is analysed through the
initial parameter and the adopted initial roughness
profile as seen in Fig.1. Thus, in the initial element
of sheet metal, the local imperfections in thickness
h
o
(x) can be related to the roughness profile by,

= = = tg
h
1
dx
dh
h
1
dx
dA
A
1
o
o
o
o
o
(1)

where A
o
= initial transversal area, h
o
= h
o
(x) = initial
thickness size, tg = roughness profile inclination,
x = coordinate axis perpendicular to the local neck.










Fig.1 Initial thickness profile and roughness model of sheet
metal related to equation (1).


Fig.2 Element of sheet metal under biaxial stretching, showing
a local neck and the definition of the strain gradient .

Present approach analyses a thin sheet metal with
strain hardening and strain rate hardening behavior
which constitutive equation for flow stress is,

( )
M n
o
k + =
&
(2)

where k = strength coefficient, = equivalent true
strain,
o
=prestrain, n = strain hardening coefficient,

&
= equivalent strain rate, M =strain rate sensitivity
coefficient.
Also, the anisotropic yield criterion proposed by Hill
[7] which accommodate R-value less than 1, is used,

] ) R 2 1 [(
) R 1 ( 2
1 m
2 1
m
2 1
m
+ + +
+
= (3)

where R = normal anisotropy, m = parameter of
anisotropy (m = 1.14+0.86R) [5],
1
and
2
are the
sheet in-plane principal stresses, see Fig.2.
The governing equation for the local necking
formation and growth [6] from the initial thickness
imperfection in sheet metal forming processes is,

( )

=


o
n
z ) 1 ( M
1
M
(4)

where x / = = strain gradient in the local neck,
2 1
/ = = strain path and z = subtangent which
is defined as,

m
m
m m
m
m m
m
R
R
z
1
) 1 /(
) 1 /( 1
) 1 /(
/ 1
1
) 2 1 (
1
) 1 ( 2
)] 1 ( 2 [

+ +
+

+
+
=

(5)

Equation (4) can be solved analytically or
numerically to describe the development of local
strain gradient during sheet metal forming processes.
Varying the strain path , the limiting strain curve or
FLC can be calculated when the strain gradient
attains a critical value
crit
= 20 or when / = (/)

= constant. Introducing equation (1) in (/)

,

(/)

crit
o
tg
h

= (6)

3 RESULTS AND DISCUSSIONS
The roughness inclination parameter tg can be
evaluated from roughness measurements for short
wave or long wave profiles. From experimental
results, this parameter is less than tg 10
o
= 0.176 ,
thus, the critical normalized strain gradient is,

(/)

o
h 4 . 113 (7)

Therefore, the critical normalized strain gradient at
the instant of local necking inception (/)

increases linearly with the sheet thickness h
o
,
consequently, the FLC curve also moves upwards in
the FLD. Thicker sheet metal must generate higher
limit strains than thinner ones, but the correlation of
limit strains with the sheet thickness is not linear.
h
o
(x)
x
tg
For a material exhibiting M = 0 and
o
= 0, the
critical normalized strain gradient at the instant of
local necking is,

(/)

1
n
/ z ) 1 (
*
1

+
= (8)

where
*
1
is the limit strain in the principal direction 1.
Introducing equation (7) into equation (8), the limit
true strain
*
1
or the FLC curve can be evaluated as,

*
1
n
h 4 . 113
/ z ) 1 (
1
n
o

+
+
= (9)

Thus, the thickness size h
o
has neglecting effect on
the limit strains for strain rate insensitive materials,
but the n-value may vary slightly with thickness.
Alternatively, for material strain rate sensitive or
exhibiting M-value, the thickness size h
o
has an
important role on the limit strains or the FLC curve.
The theoretical effect of h
o
on the limit true strain
curve for steel sheets of 0.5 , 1.0 , 1.5 , 2.0 , 2.5 and
3.0 mm can be evaluated in Fig.3. The FLC is
plotted in the positive region or the biaxial
stretching region of the FLD. The steel is assumed
isotropic, R = 1 and m = 2, and to have the
equivalent flow stress of ( )
012 . 0 22 . 0
05 . 0 k + =
&
MPa ,
i.e, n = 0.22 and M = 0.012.
According to equation (1), the initial thickness
imperfection parameter decrease from 0.352 to
0.059 as the thickness size h
o
increases from 0.5
mm to 3.0 mm, hence, the limiting true strains
increases about 20%. Therefore, thinner sheet metal
will have lower limit strain or FLC, i.e., lower
resistance to local necking. However, thicker sheet
metal tends to a maximum theoretical limit of true
strain or major true strain of about 0.34 for this
adopted material exhibiting work hardening
coefficient n = 0.22 and strain rate sensitivity
coefficient M = 0.012. Although M-value is very
small, it is sufficient to delay the onset of local
necking, increasing the major true strain
*
1
from
0.22 to 0.31 for a steel sheet of thickness of 1 mm.
4 CONCLUSIONS
From the prediction of limit strains curves in the
biaxial stretching region of FLC for thin steel sheets
with thickness h
o
varying from 0.5 mm to 3.0 mm,
0
0,1
0,2
0,3
0,4
0 0,1 0,2 0,3 0,4
Minor true strain
2
M
a
j
o
r

t
r
u
e

s
t
r
a
i
n

1
ho= 3.0 mm
ho= 2.5 mm
ho= 2.0 mm
ho= 1.5 mm
ho= 1.0 mm
ho = 0.5 mm
Fig.2 Prediction of the influence of thickness size h
o
on the
limit strains or FLC for isotropic steel sheet with flow stress
( )
012 . 0 22 . 0
05 . 0 k + =
&
, according to present model.

the following conclusions can be drawn:
- present theoretical model of strain gradient
development produced reasonable predictions about
the influence of thickness size on the FLC,
- the critical normalized strain gradient at the onset
of local necking can be assumed (/)

o
h 4 . 113
- thicker steel sheets have higher limit strains of up
to 20%, thus, have higher resistance to local necking
- steel sheets exhibiting the equivalent flow stress of
( )
012 . 0 22 . 0
05 . 0 k + =
&
have a maximum major true
strain of 0.34 in the stretching region of the FLD.
ACKNOWLEDGEMENTS
The author would like to gratefully acknowledge the financial
support received from CNPq of Brazil and the University of
Santa Catarina State/Brazil.
REFERENCES
1. S.P. Keeler, Sheet Metal Industries, (1968) 633.
2. R. Hill, J. Mech. Phys. Solids, 1 (1952) 19.
3. Z. Marciniak and K. Kuczynski, Int. J. Mech. Sci., 9
(1967) 609.
4. S. Storen and J.R. Rice, J. Mech. Phys. Solids, 23 (1980)
421.
5. J.D. Bressan and J.A. Williams, Int. J. Mech. Sci., 25
(1983) 155.
6. J.D. Bressan and J.A. Williams, J. Mech. Working Tech.,
11 (1985) 291.
7. R. Hill, Math. Proc. Cam. Phil. Soc., 85 (1979) 179.
R = 1.0
m = 2

o
= 0.05
n = 0.22
M= 0.012
1 INTRODUCTION
In the modern manufacturing world wide scenario,
the requirement of customized production, cost
reduction, life-cycle shortening is still growing.
Scientists have to find answers to these urgent
requests and new technologies seem to be a
sustainable approach to reach such aims in all
sectors of activities including sheet metal forming.
Thanks to a basic concept different than
conventional metal forming processes, the
introduction of incremental forming may represent a
progress in the manufacturing processes evolution
[1]. In Single Point Incremental Forming, the final
geometry is generated by the movement of a simple
punch controlled by a CNC milling machine which
deformed a clamped blank. This concept allows to
avoid the use of traditional die differently than
conventional stamping processes.
Regarding the concept of such a technology, a set of
advantages provided by SPIF can be listed.
Incremental forming technology allows to reduce
set-up costs significantly [2-3] and presents a very
high flexibility. Furthermore, due to the favourable
stress state induced by the punch during the local
deformation, the material formability is higher in
comparison with conventional stamping operation
[4].
As SPIF is an emerging process, it remains to be
improved to make an industrially suitable process. In
this sense, it is very important to increase the
knowledge of such a technology through both
experimental and numerical investigations.
Therefore, this paper is based on an accurate FE
analysis of the process used to emphasize the
necessity to control all process parameters in order
to make an industrially suitable technology. The
forming strategy is taken as example to underline its
influence on the formed part.
2 MODELLING AND NUMERICAL ANALYSIS
2.1 Finite Element Model
In recent years, Finite Element Analysis has been
considered to be an effective tool for simulating
ABSTRACT: As it is well known, the design of a mechanical component requires some decisions about
tolerances and the product has to be manufactured with a careful definition of the process set up. Although
standard sheet metal forming processes are strongly controlled, new processes like Single Point Incremental
Sheet Forming remain to be improved. In SPIF, the final geometry is generated by the envelopment of all
positions assumed by a simple forming tool which deforms a clamped blank. No dies are required differently
than any conventional sheet metal forming processes. Although ISF concept allows to increase the flexibility
and to reduce set up costs, such a process has a negative effect on the shape accuracy by initiating undesired
rigid movement, elastic springback and sheet thinning. This paper emphasizes the necessity to control all
process parameters to improve final shape accuracy. To attend to this aim, a finite element analysis is
performed in order to study the influence of forming strategy on the opening or the closing of rings taken
from a truncated cone manufactured by SPIF. The results obtained allow to have a better knowledge of
springback effect on parts manufactured by SPIF with the aim to improve their accuracy.
Key words: Single Point Incremental Sheet Forming, FEM analysis, Forming strategy, Elastic springback,
Sheet thickness.
Finite element analysis and experimental investigations for improving
precision in single point incremental sheet forming process
S. Dejardin
1
, S. Thibaud
2
, J.C. Gelin
1

1
ENSMM 26 chemin de lEpitaphe, 25000. Besanon. FRANCE
e-mail: steeve.dejardin@ens2m.fr; jean-claude.gelin@ens2m.fr

2
FEMTO-ST Institue, Applied Mechanics Department- 24 chemin de lEpitaphe, 25000. Besanon. FRANCE
e-mail: sebastien.thibaud@univ-fcomte.fr
such an emerging metal forming process. As
incremental forming is a progressive sheet metal
forming process characterized by large
displacements and localized strains, an explicit
solution scheme was adopted, resulting in the choice
of LS-Dyna as the FEM simulation code. In a next
section, elastic springback simulation of formed part
is running by adopting an implicit scheme.
According to previous experimental investigations
[5], the investigated shape to perform simulation is a
truncated cone. Main data are summarized in Table
1.

Table1. Dimension of the investigated shape
Major base 140mm
Minor base 40mm
Depth 50mm
Wall inclination 45

Due to the three-dimensional tool path, a fully three-
dimensional analysis is required. As a consequence,
shell elements with 4 nodes and 6 degrees of
freedom per node and five integration points along
the thickness were used. These are reduced
integration elements (one point in the plane).
Furthermore, an adaptive mesh refinement, which
allows four levels of refinement, was performed in
order to reduce element size when the distortion
level reached a maximum value. These ingredients
allow a proper modelling of the progressive
deformation of the sheet by increasing the number of
nodes in contact with the tool surface. The sheet was
initially meshed with 3600 elements.
The 10 mm diameter tool head is considered as a
rigid body and the corresponding boundary
conditions are related to the path that it should
follow during the process.
Sheet metal behaviour over the yield stress has been
accounted by means of a Swift type hardening law:
( )
n
p
k + = .

Parameters were defined through results of sheet
bulging test. Material used in this study is an
aluminium alloy 1050.
2.2 Results and discussion
2.2.a. Analysis framework

In order to validate the numerical model, the final
shape was measured offline and compared with
numerical ones. A set of several series of
measurements taking into account the transverse
sections highlighted in Figure 1 was performed. As
results demonstrate that the part is properly formed
and maintains its symmetry, next sections will focus
on the zone AB where the tool has its vertical
displacement.














Fig. 1. Definition of profiles used in measurement.
The tool path is made up of a series of rotational
movement around the vertical axis of the milling
machine generated transverse to the long axis of the
cone. Once the rotation is done, the tool moves in
horizontal direction and penetrates the blank in the
vertical direction to follow the next contour (Figure
2).











Fig. 2. Forming strategy used in experimental measurement
Tool path A.
From this modelling, a numerical study was
performed to observe the potential relationship
between the tool path and geometric defects due to
elastic springback. Two strategies in simulating the
SPIF were deployed. In the first one (strategy A) the
followed tool paths in both the simulation and the
experiment are equal, whereas in the second strategy
(strategy B), the simulated
tool path alternates in clockwise and
counterclockwise direction.

2.2.b. Strain history and distribution
profile 2
B A
profile 1
X
Y
profile 3
profile 4
C
step 1
step 3
contour
step 2
In order to study the influence of tool path on the
strain history, the strain path was analyzed for the
elements indicated in Figure 4. During simulation,
the selected elements were consecutively affected by
the tool movement. Element 4191 was located in the
bending zone, close to the major base of the cone.
Elements 4215, 4239 and 4263 correspond
respectively to a depth of 15mm, 30mm and 45mm.
These elements are located along the section AB of
the profile 1 (Figure 3).









Fig. 3. Location of elements along profile 1 in strain analysis
As it has already been underlined in previous studies
[6], figure 4 shows increments which characterized
the strain history of elements in ISF.
Regarding the influence of the forming strategy on
strain history, it can be noticed that the effective
plastic strain value obtained with tool path B is
lower than the one obtained with strategy A. Indeed,
the relative variation between the two strategies goes
from 13% for element 4191 to 19% for element
4239.























Fig. 4. Strain history along profile 1.
So, from observations made on global aspect of
experimental parts manufactured by ISF using
strategy B, this first analysis emphasizes the
influence of the forming strategy on the strain
distribution and represents a first remark which has
to be taken account to improve the quality of the
formed part.

2.2.c. Thickness measurements

As incremental sheet forming is mainly
characterized by stretching deformation mode of the
sheet metal added to the lack of any dies, a
significant sheet thinning determines accurate limits
of the process.
The numerical analysis has demonstrated that the
used of strategy B allows to reduce the local sheet
thinning of the zone AB by up to 8% (Figure 5).














Fig. 5. Influence of tool path on thickness history.
To conclude on this section, the analysis of thickness
distribution and strain history emphasizes the
necessity to control all process parameters to
improve final shape accuracy by showing directly
the influence of the forming strategy on the formed
part.
2.3 Elastic springback analysis
Problems of dimensional accuracy on parts formed
by ISF due to elastic springback effects are well
known. The next section deals with an original way
to analyze this mechanical effect by focusing on the
influence of tool path localized in the wall of the
cone [7].
The elastic springback analysis is obtained from
rings taken from the formed part as it is shown in
Figure 6.
0.000
0.200
0.400
0.600
0.800
1.000
1.200
1.400
0.00 0.02 0.04 0.06 0.08 0.10 0.12
simulation time (s)
e
f
f
e
c
t
i
v
e

p
l
a
s
t
i
c

s
t
r
a
i
n
0.55
0.60
0.65
0.70
0.75
0.80
0.85
0.90
0.95
1.00
1.05
0.000 0.020 0.040 0.060 0.080 0.100 0.120
simulation time (s)
t
h
i
c
k
n
e
s
s

(
m
m
)
Element 4215 strategy B
Element 4215 strategy A
4191
4215
4239
4263

Fig. 6. Location of rings taken from the formed part at different
depth.
Each ring is cut along an axis. Nodes which are
localized on the opposite axis are completely
constrained to allow the prediction of elastic
springback and to avoid rigid movement in the
prediction phase.
Elastic springback is characterized by closing of
each ring with a bigger gap in the horizontal plan of
rings -20/-22, -24/-26, -28/-30 corresponding at the
middle of the wall.
Figure 7 shows that nodes which are initially joined
deviate from each other in X direction with a bigger
opening in the case of strategy A than in strategy B,
contrary to in Y direction, a bigger closing is
obtained with strategy B.






















Fig. 7. Tool path influence on elastic springback of rings
localised in the middle of the wall of the formed part.
3 CONCLUSIONS
As it is well known, incremental sheet forming
process is a very promising manufacturing process
which still requires further optimizations. Numerous
studies have demonstrated the effect of process
parameters like advancing speed, forming force, tool
depth step in the characteristics of the formed parts.
This paper added knowledge on the influence of the
forming strategy on the formed parts not only on a
geometrical point of view, with an original elastic
springback analysis, but also on the evolution of
strain and thickness distribution, showing the
possibility to improve the quality of the final parts
by an accurate control of all process parameters. In
the one hand, such an improvement is directly linked
with the use of optimised tool trajectories. In the
other hand, a better knowledge of experimental
evolution of the characteristics of formed parts
during the process is necessary to go farther with the
aim to make an industrially suitable technology.
REFERENCES
1. F. Micari, G. Ambrogio, L. Filice, Shape and
dimensional accuracy in single point incremental
forming: State of the art and future trends, Int. J. Mater.
Process. Technol. (2007).
2. G. Ambrogio, I. Consatntino, L. De Napoli, L. Filice, M.
Muzzupappa, Influence of some relevant process
parameters on the dimensional accuracy in incremental
forming : a numerical and experimental investigation,
Int. J. Mater. Process. Technol. 153C/154C (2004) 501-
507.
3. E. Ceretti, C. Giardini, A. Attanasio, Experimental and
simulative results in sheet incremental forming on CNC
machines, Int. J. Mater. Process. Technol. 152 (2004)
176-184.
4. T.J. Kim, D.Y. Yang, Improvement of formability for
the incremental sheet metal forming process, Int. J.
Mech. Sci. 42 (2001) 1271-1286.
5. S. Dejardin, S. Thibaud, J.C. Gelin, Experimental and
numerical investigations in single point incremental
sheet forming, in: Proceedings of the 9
th
International
Conference on Numerical Methods in Industrial Forming
processes (2007), 889-894.
6. Q. Qin, E.S. Masuku, A.N. Bramley, A.R. Mileham,
G.W. Owen, Incremental sheet forming simulation and
accuracy, in: Proceedings of the 8
th
International
Conference on Technology of Plasticity (2005).
7. S. Thibaud, Contributions for modelling TRiP steels in
sheet metal forming Simulations and influences of
manufacturing processes on in-use properties, PhD
Thesis (2004).

-12
-16
-44
Depth of the
formed part
(mm)
-20
-24
-28
-32
-36
-40
2mm
ring -20/-22 strategy A
ring -20/-22 strategy B
ring -24/-26 strategy A
ring -24/-26 strategy B
ring -28/-30 strategy A
ring -28/-30 strategy B
-8
-6
-4
-2
0
2
4
6
8
40 42 44 46 48 50 52 54
first horizontal axis X (mm)
s
e
c
o
n
d

h
o
r
i
z
o
n
t
a
l

a
x
i
s

Y

(
m
m
)
Finite Element Analysis of Deep drawing and hole flanging processing
of an oil filter cover
T. Elbitar, A. Gemeal

Central Metallurgical R&D Institute (cmrdi) ZIP code 11421, P.O. 87 Helwan, Cairo, Egypt.
URL: www.cmrdi.sci.eg
email:elbitart@cmrdi.sci.eg
1 INTRODUCTION
This search deals with a practical industrial
problem. It investigates problems of ripping of the
inner flanged edge of a filter cover (figure 1),
produced in Egyptian industry. The outer edge
flange is made by deep drawing process while the
inner edge is made by hole flanging.

Fig. 1. Filter cover of passenger car
The industrial production encounters a serious
problem as shown in figure 1. Cracks occur during
the hole flanging process. Despite advances in
FEA, the prediction of material failure that may
result from the gradual internal deterioration
associated with high straining remains, to a great
extent, relegated to post-simulation analysis.
Typically, simulation results are post-processed
and the location and likelihood of failure are
assessed by empirical methods (of questionable
accuracy). One of the possible tools to solve this
problem is continuum damage mechanics (CDM).

The current production process plan of the filter
cover is described table 1. Unfortunately this plan
gives permanent failure in the inner hole flanging
because metal can not withstand associated tensile
stresses.

Table1. Current Production Process planning
N. Process
1 Inner and outer blanking and piercing
2 Inner hole flanging
3 Deep drawing of outer rim
2 FEA
The objective of FE simulation is to check the
experimental observation which indicates that
unavoidable cracks always initiate in the tip of
inner flange. FEA was carried out by commercial
MSC.SuperForm 2005 using the following
conditions: explicit finite element analyses, rigid
plastic, axisymmetric model, the elements of blank
are 4-node axisymmetric quadrilateral, 4 points
Gaussian integration. The finite element mesh of
the cover is composed of 2 layers through the
thickness and 33 elements along the diameter that
is a total of 66 elements. The behavior of cover
material was fed to FE code by 5 points on the
experimentally determined flow curve of the cover
material. The tools were defined as rigid body.
Table 2 shows the geometrical conditions of FEA
and the industrial production while properties of
used material are shown in table 5. FEA simulated
, gemeal2000@cmrdi.sci.eg; gemeal2000@yahoo.com
ABSTRACT: Trial and error is a tedious way of working which has been implemented for long time in the
sheet metal forming processing. Finite Element Analysis (FEA) is one of the most advanced solutions that
can eliminate time and money consumption. The present investigated part is an oil filter cover and contains
outer flange processed by deep drawing and inner flange processed by hole flanging. Inner flange ripping
usually happens on the rim. Both flanging techniques have been analyzed by FEA. Failure modelling
according to three damage theories was applied to detect reasons of inner flange ripping. Metal simply
cannot withstand elongation required for hole flanging. FEA proposed deep drawing technique of the inner
flange followed by hole piercing and finally deep drawing is used for outer flange.
Key words: Deep drawing, hole flanging, CDM, FEA, ductile fracture
Table2. Analytical and experimental conditions
Material stainless and heat resisting
steel, Material number
1.4005, Symbol 2380,
Standard 142380 of
thickness 0.5
Blank diameter 100 mm
Height of outer flange 5 mm
Height of inner flange 5 mm
Punch diameter 26.5 mm
Punchdie clearance 1 mm
Tool geometry tapered punch
Blankholder None
Coulomb Friction
coefficient
0.1

- initially without damage modeling- propagation of
processes numbers 2 and 3 together in figure 3. The
equivalent plastic strain distribution is observed for
four different increments. Table 3 shows the
corresponding values of punch displacement and the
maximum value of equivalent plastic strain that
occurs in the cover. This maximum value of strains
and therefore the stresses appear in the inner flange
tip. This result is in good agreement with the
experimental observations.. Analysis based on
Continuum Damage Mechanics (CDM) was carried
out to check the necessary conditions for crack
initiation.

Table3. Analytical and experimental conditions
Punch displacement
(mm)
Max. total equivalent
plastic strain
0 0
4.464 0.1034
5.245 0.182
12.33 0.2476
14.98 0.3362

(a)

(b)

(c)

(d)

(e)
Fig. 3. Distribution of the equivalent plastic strain at four
different punch penetrations, a) 0 mm, b) 4.464 mm, c) 5.245
mm, d) 12.33 mm and e) 14.98 mm
3 DAMAGE MODELING
Numerous attempts have been made to relate the
fracture strains of metals to macroscopic variables
associated with the material, the process, or both.
Three damage models of Cockroft and Latham,
Oyane and Lemaitre were used in this research to
verify the experimental results of failure of the
processes [1]. Cockroft and Latham observed that
ductile fracture occurs in the region of largest tensile
stress [2],
T
p
d d

Where
T
= maximum principal tension stress, =
effective stress, p = effective plastic strain and d =
damage value which sets the maximum damage
indicator so that d d
max
. Oyane model [3] took
hydrostatic pressure effect into account
( )
m
p
d B d

= +

Where B = material constant represents hydrostatic
pressure effect and
m
= mean stress. Lemaitre
model [4] is designed to capture the evolution of
ductile damage under large plastic deformation.
When damage model of Cockroft-Latham was used,
it was noticed that: when Crack threshold = 0.05,
damage threshold = 0.06 (calculated from the tensile
test results), the initiation of crack appears in inner
fiber of tip of hole flanging side. For Oyane model,
for hydrostatic constant = 0.05, it also gave close
results to Cockroft-Latham model. Lemaitre
Damage model [5] was introduced and the constants
were as shown in table 5. Where, critical uniaxial
damage represents the damage when material fails in
tension. This is used to evaluate the critical damage.

Table5. Material parameters related to Lematre model
Name of constant Value
Critical uniaxial damage 0.06
Corrected ultimate stress (tensile test) 300 MPa
Damage resistance parameter 0.2
Uniform elongation 29.4 %
Fracture closure parameters 0.2

Corrected ultimate stress (tensile test) is the
corrected stress at failure. Damage resistance
parameter is a parameter representing resistance of
material to damage. Uniform elongation is
elongation under tension, Fracture closure Parameter
is an experimentally determined value between 0
and 1, that indicates how much the voids close
during loading (0 indicates no closure). The results
are also close to the other 2 models as shown in
figure 4. All FE conditions are the same mentioned
in section 2 except that each damage model was
introduced once at a time.


Fig. 4. Crack initiation on the outer side of the inner flange, by
applying Cockroft-Latham, Oyane or Lemaitre criteria for the
corresponding mentioned values, exaggerated
4 EXPERIMENTAL WORK
The material is a stainless steel with heat resisting.
The cover thickness is 0.5 mm. The detailed
chemical composition is given in table 4.

Table4. Chemical composition of the cover material
Element C Si P S Mn
% 0.0891 0.0143 0.01 0.0123< 0.269

Tensile specimens were cut from the sheet metal
used at angles of 0, 45, 90 degrees to rolling
direction and tested on a tensile test machine
(Shimatzu with 5 Tons capacity) at a constant speed
of 5 mm/min. The identified mechanical properties
are presented in table 5.

Table5. Mechanical properties of the cover material
Mechanical property Value
Elongation 29.4%
Yield stress 170 MPa
Ultimate stress 300 MPa
n-value 0.703
r-value 6.2

A solution for the problem (fracture occurrence) is
to change material and use extra deep drawing
(EDDQ) steel or, at least, DDQ steel. That is
because EDDQ steel has higher formability than
filter cover stainless steel. Comparing tables 6 and 5
shows that they have higher elongation. Comparing
chemical composition between tables 6 and 4 shows
that filter cover material is lower in P, S and much
lower in Si. But due to unavailability of those types
of steels in local market, alternative solution was
studied by processing i.e. changing the production
process from hole flanging to deep drawing.

Table6. Properties of DQ, DDQ, EDDQ {6}
Chemical Properties Mechanical Pr.
P S
Typ
e
DIN C
%
Si
% Max. %
Mn
%

u
Elong
.%
DD
Q
130
3
0.1 0.1 0.04 0.04
0.2-
0.5
2
6
28-
38
34
ED
DQ
140
3
0.08 0.08 0.04 0.04
0.5-
0.45
2
4
28-
38
36
5 SUGGESTED PROCESS PLAN
The suggested production plan is that the inner
flange is made by deep drawing followed by bottom
piercing and after that the outer flange is made (table
6). There must be a hole in punch head to avoid the
previously made inner flange.

Table6. Suggested production process plan
N. Process
1 Deep drawing of inner hole flange figure 5
2 Bottom piercing
3 Deep drawing of outer rim

Fig. 5. Distribution of the equivalent plastic strain for the
suggested deep drawing, punch displacement=5.05 mm, for
punch nose radius 7 mm and die corner radius 5 mm.
Figure 5 shows the equivalent plastic strain for the
mentioned conditions. For a flange depth equivalent
to that made by hole flanging i.e. 5 mm the total
equivalent plastic strain = 2.134. Guaranteeing the
production of a crack free 5 mm flange height, thus
guarantees a well oil leakage preventing filter cover
All FE conditions are the same as mentioned in
section 2. Applying damage criteria on deep drawing
did not give crack initiation as in hole flanging case.
6 DISCUSSIONS
1- In hole flanging to get flange height of 5 mm the
total equivalent plastic strain = 0.3362. It is
obviously higher than its value if made by deep
drawing which is 2.134.
2- On the other hand, hole flanging gives right
angle flange while in deep drawing gives conical
flange. However, angle of the cone depends on
clearance between punch and die and punch
displacement. If greater angle is required, then
lower punch-die clearance is designed or more
punch displacement can be applied. But in fact,
this is not required in filter cover function
because 5 mm flange is enough for oil sealing;
7 CONCLUSIONS
1- Despite that the highest equivalent plastic strain
level occurs in both sides of the part (the deep
drawing side and hole flanging side) but it is
higher in the tip of the hole flanging side;
2- The three damage models as well as
experimentations allies each other in predicting
unavoidable crack initiation in hole flange side;
3- Changing the process of producing the inner
flange from hole flanging to deep drawing will
produce this successfully. In spite of that deep
drawing ratio in the inner flange is higher than
the outer, but finite element analysis proves it is
still successful;
REFERENCES
1. S. Thipprakmas, M. Jin and M. Murakawa, Study on
flanged shapes in fine blanked - hole flanging process
(FB-hole flanging process) using finite element method
(FEM), Journal of Materials Processing Technology
192193 (2007) 128133.
2. M.G. Cockroft and D.J. Latham, Ductility and the
Workability of Metals, Inst. Met. 96 (1968) 33.
3. M. Oyane, T. Sato, K. Okimoto and S. Shima, Criteria
for ductile fracture and their applications, J. Mech. Work
Techno. 4 (1980) 65.
4. J. Lemaitre, A continuous damage mechanics model for
ductile fracture, J. Eng. Mater. Technol. 107 (1985) 83
89.
5. J. Lemaitre and J.L. Chaboche, Mechanics of Solid
Materials, Cambridge University Press, (1990).
6. Egyptian Iron & Steel Co., Products and specifications,
Cairo (1992)
1 INTRODUCTION
The object of non-linear solid and structural
mechanics is the modeling and the computation of
structures with strong non-linearities, both geometric
(finite transformations, evolving unilateral contact,
friction, etc.) and physical (plasticity, hardening,
damage, temperature, etc.). The aim is to simulate
numerically the behavior of a mechanical object
subjected to various mechanical loadings, in order to
improve its endurance, or even to optimize its
manufacturing process. The reliability and the
performance of such a simulation are based on
different types of tools: theoretical (constitutive
relations representing the physical phenomena),
numerical (algorithms to integrate the ODEs,
schemes to solve non-linear systems PDEs, etc.) and
geometric (representation of the object shape, finite
element discretization or meshing, remeshing and
adaptive meshing during the simulation).
Theoretical aspects (plasticity with damage
models), and related numerical aspects have been
widely developed for many years, and more or less
performing tools have been proposed [1].
Concerning geometric aspects, the 2D or 3D object
representation, as well as the initial finite element
discretization, have also given rise to many
development efforts. As for adaptive remeshing,
which is necessary for strongly non-linear problems,
the interest is really high nowadays and the proposed
solutions are not quite satisfactory [2, 3]. Let us
mention that the main difficulty lies in the fact that,
in large deformations, the damaged solid geometry
is variable and cannot be defined in an explicit way.
We are interested in the problem of remeshing a
mechanical structure subjected to large plastic
deformations, including damage. Contributions to
the adaptive strategy using adaptive meshing and a
posteriori error estimation in large elasto-plasticity
with damage have been developed [4]. These
methods need to entirely remeshing the piece. This
paper presents a simpler and rapid method only
based on refinement and coarsening techniques. This
method avoids entirely remeshing the piece. This
paper gives the necessary steps to remesh the
ABSTRACT: This paper presents an adaptive remeshing scheme to solve problems, where large plastic
deformations with ductile damage are possible, in metal forming processes. During simulations of metal
forming processes, severe mesh distortion occurs and it is then necessary to remesh the part to carry out the
finite element analysis. The proposed remeshing method is based on refinement and coarsening techniques
and avoids entirely remeshing the piece. Its advantages are the simplicity and the rapidity. This paper presents
the necessary steps to remesh the damage structure: refinement of the elements in the boundary of the domain,
adaptive refinement and coarsening of elements according to physical and geometrical criteria. In addition,
the macroscopic crack propagation is modelled by removing the fully damaged elements. Numerical
examples show the efficiency of the proposed method.

Key words: adaptive remeshing, refinement and coarsening techniques, large deformation, damage
Adaptive remeshing for sheet metal forming in large plastic deformations
with damage
L. Giraud-Moreau
1
, H. Borouchaki
1
, A. Cherouat
1

1
Charles Delaunay Institute University of Technology of Troyes BP 2060 10010 Troyes cedex - France
e-mail:Laurence.moreau@utt.fr;houman.borouchaki@utt.fr; abel.cherouat@utt.fr
mechanical structure subjected to large elasto-plastic
deformations with damage during simulation of
sheet metal forming processes by using the proposed
adaptive remeshing method. These steps are divided
into two main categories: the refinement of the
elements in the boundary of the domain, the adaptive
refinement and coarsening of the part elements. The
remeshing is governed by a mesh element size map
representing the conformity with the underlying
geometry of the deformed domain and the
improvement of the accuracy of mechanical fields
(stress, plastic strain, damage). In addition, the
macroscopic cracks propagations are modelled by
removing the fully damaged elements.
2 GENERAL REMESHING SCHEME
The simulation of the forming process is based on an
iterative process. At first, a coarse initial mesh of the
part is generated with triangular or quadrilateral
elements. At each iteration, a finite element
computation is realized in order to simulate
numerically the forming process for a small
displacement step of forming tools. Then, remeshing
is applied after each deformation increment, if
necessary, according to the following scheme:

definition of the new geometry after deformation
by eliminating totally damaged elements,
definition of a physical size map based on the
adaptation of the mesh element size with respect
to one of the mechanical fields,
definition of a geometrical size map based on the
geometric curvature of the boundary,
adaptive remeshing of the domain based on
refinement and coarsening techniques with
respect to the physical and geometrical size map.
2.1 Definition of the new geometry
Totally damaged elements are eliminated from the
mesh. A new geometry is then defined. In the model
of applied mechanics, macroscopic craks are
deduced from the suppression of totally damaged
elements. To approach the physical reality at best, a
minimal size is imposed to the damaged elements.
2.2 Definition of a physical size map
A physical size map is defined by calculating a
physical size (h
D
) for each element of the part. This
physical size is defined with respect of one of the
mechanical field. In this paper, the ductile damage
has been chosen. This mechanical field is quantified
by a real value between 0 and 1. With a totally
damaged element (value 1) is associated a minimal
element size and, with a 0-damaged element, a
maximal element size. A critical value Dc (for
example: 0.8) has been defined from which the
minimal size must be reached. For the other
elements, a linear size variation can be used. For a
given element, if the ratio between the average size
of its edges ( h ) and its physical size (h
D
) is greater
than a given threshold, the element must be refined.
During the step of remeshing, the refinement is
repeated as long as the physical size is not reached.
2.3 Definition of a geometrical size map
The geometrical size map indicates if a boundary
element must be refined or not. The geometric
curvature is estimated at each boundary vertex of the
domain. If this curvature has been modified during
the deformation of the computational domain, all
elements sharing this boundary vertex must be
refined. The geometrical size map only contains the
list of elements which must be refined one time
because of the curvature variation.
2.4 Adaptive remeshing based on refinement and
coarsening techniques
The adaptive remeshing technique consists in
improving the mesh by coarsening and refinement
methods in order to conform to the geometry and the
mechanical fields of the current part surface during
deformation. Two consecutive steps are executed:
a coarsening step during which the mesh is
coarsened with respect to the physical size map,
a refinement step during which the mesh is
refined according to the geometrical size map
and then to the physical size map.

The refinement technique consists in subdividing
mesh elements. An element is refined if it is a
boundary element which needs to be refined
(element which belongs to the list of the geometrical
size map) or if its size is greater than its physical
size (physical size map). There is only one element
subdivision which allows preserving the element
shape quality: the uniform subdivision into four new
elements. A triangular or a quadrilateral element
which needs to be refined is subdivided into four
elements. For this subdivision, a node is added is the
middle of each edge of the element. Boundary
elements which belong to the geometrical size map
are first refined. The refinement is then applied
according to the physical size map. In this case, the
refinement procedure is repeated as long as the
physical size map is not reached. After each
refinement procedure (geometrical criterion or
physical criterion), an iterative refinement to restore
mesh conformity is necessary. Indeed, after applying
the subdivision according to the geometrical or
physical criteria, adjacent elements to subdivided
elements must be modified. A procedure of
subdivision has been proposed for the adjacent
elements in order to stop the propagation of the
homothetic subdivision. The details of the
refinement procedure can be found in [5,6].

Fig. 1. Flowchart of the used shell Script
The coarsening technique is the reciprocal
operation of the refinement procedure. It can only be
applied to a set of four elements, obtained during a
homothetic element refinement. A set of four
elements can only be coarsened if the physical size
specified in the physical size map for these four
elements is at least two times greater than their
actual average element size.

For the simulations of metal forming processes, a
special procedure has been developed in order to
execute Abaqus step by step (see figure 1). At each
load increment, and after the convergence has been
reached, the over all elements are tested in order to
detect the fully damaged elements (elements where
the damage variable has reach its critical value in all
Gauss points). If so, the fully damaged element is
killed from the structure. Then remeshing is applied.
Mechanical fields are simply induced from the old
mesh to the new mesh.
3 APPLICATION
3.1 2D forming process
In this example, hollow cylinder is side-pressed
between two rigid planar tools. The cylinder is made
on aluminium alloy and has 58 mm external
diameter and 36 mm internal diameter. Figure 2
show the meshes adapted to the damage fields of the
aluminium hollow cylinder corresponding to
different displacements of the moving tool. One can
observe that a macroscopic crack initiates similar to
the experimental result.




Fig. 2. Cylinder side-pressed between rigid planar tools


Input ABAQUS data
for new loads and new boundary
diti
Coarse Initial
mesh of the piece
ABAQUS
Stand./Explicit
If damage
zone is
required
Automatic Re-meshing
(RAFF2D)
(geometrical curvature+ physical
Store new
mesh
and new state
variables
Analysis
complete
d
NO
Kill
elements
YES
Transfer of state
variables to new mesh
NO
STOP
YES
START

3.2 3D deep-drawing process
3D sheet metal process will be presented in order to
test the capability of the Gurson model to predict the
ductile damage occurrence during the cross-die
deep-drawing process. The punch-die clearance is
very small compared to the blank diameter.
Isoparametric sheet finite elements with reduce
integration (S3R and S4R of ABAQUS/EXPLICIT)
with adaptive size are used. The tools (punch, die
and blank-holder) are discretizated by rigid surfaces.
The sheet metal is defined by material coefficients
summarized in Table 1. It is worth noting that the
used material exhibits a non linear hardening and the
damage induced softening, the maximum stress is
about
eq
894MPa = for
p
18.5% while the
ductility (plastic strain at fracture) is about 29.5%.
See the results for the cross-die deep-drawing (figure
3). We can note that, the initial blank sheet is
computed using an initial coarse mesh (elements),
the mesh is again refined uniformly and the adaptive
mesh refinement procedure is activated where
elements are created automatically in regions of
large curvature to even more accurately represent the
complex material flow (large stretching) around the
punch and die radii.

Table1. Mechanical properties of the used material
Initial thickness 1.0 mm
Elasticity modulus E = 200.0 GPa and = 0.3
Flow Stress(MPa)
( )
n
y p e p
K = +
n = 0.42
e
= 410.6 MPa K = 968.92 MPa
Damage parameters
1 2 3
q 1.5 q 1. q 2.25 = = =
f c
f 0.75 f 0.05 = =

n n n
0.25 S 0.15 f 0.03 = = =



Fig. 3. Part shapes at different punch displacement
The second example concerns the 3D sheet forming
process of a front door panel proposed at
Numisheet'2002. Meshes adapted to the part
curvature corresponding to different punch
displacements (10, 130 mm) are shown in figure 4.

Fig. 4. Part shapes at different punch displacement
4 CONCLUSIONS
A simple remeshing scheme based on refinement
and coarsening techniques for problems in large
elastoplastic deformations, taking into account the
ductile damage, has been proposed. This adaptive
remeshing strategy has been applied to solve the
elastoplasticity problem in large deformation with
damage using the finite element code
ABAQUS/Explicit. Identification of contacts and
introduction of a projection procedure on the tools
elements could improve the remeshing method. The
extension in three dimensions for massive structures
is an important point to study.
REFERENCES
1. J. Lemaitre, A course on damage Mechanics, Springer
Verlag, (1992).
2. P. Coorevits, J.P. Dumeau and J.P. Pelle, Analyses
elements finis adaptatives pour les structures
tridimensionnelles en lasticit, In: Revue europenne
des elements finis 5 (2001), 3, 341-373.
3. H. Borouchaki, D. Chapelle, P.L. George, P. Laug and
J.P. Frey, Estimateurs derreur gomtriques et
adaptation de maillage, in: Maillage et adaptation, trait
Mcanique et Ingnierie des Matriaux, Mthodes
numriques, Herms, (2001), 279-310.
4. A. Cherouat, K. Saanouni, H. Borouchaki and P. Laug,
Virtual metal forming with damage occurrence using
adaptive remeshing, In: International journal of
forming processes, (2005), vol 8, n2-3, 311-289.
5. L. Giraud-Moreau, H. Borouchaki and A. Cherouat,
Adaptive remeshing method based on refinement and
coarsening techniques, In: Proc. 10th ESAFORM
Conference on Material Forming, , Espagne, (2007).
6. L. Giraud-Moreau, H. Borouchaki and A. Cherouat,
Sheet metal forming using adaptive remeshing, In:
Proc. of NUMIFORM04, Etats Unis, (2004).

1 INTRODUCTION
In recent years, the demand for three-dimensionally
bent steel and aluminum profiles as important
structural and design elements in traffic systems as
well as in civil engineering has increased strongly.
3D-bent profiles provide the design engineer with
new degrees of freedom and allow the construction
of lightweight structures with more advantages
regarding e.g. space saving and aerodynamics [1].
In the field of tube and profile bending there are well
known procedures offering a high potential for
three-dimensionally bending of semi-finished
products. The problem is that most of these
procedures are specialized and optimised for tube
bending; involving profiles with circular cross-
section [2, 3]. There is no well working procedure
available now in industry offering a high flexibility
to bend profiles with arbitrary cross-sections and
materials three-dimensionally. When analyzing
several procedures suitable for 3D-shaping of
profiles like stretch bending or curved profile
extrusion it was found out that these procedures
show obvious disadvantages or restrictions which
justify new considerations for the realization of a
new procedure variant for 3D-bending of profiles [4,
5]. Procedures using roll bending systems to deflect
the profile towards the third axis to produce arbitrary
space curves seem to be very suitable. Roll systems
are highly flexible concerning adjustment and
degrees of freedom and are capable of protecting the
profile surface due to the low friction between rolls
and profile. Other guiding systems like sliding
guides or ceramic guides are expensive or need
lubrication, otherwise they lead to a high surface
damage. Furthermore, the superposition of stresses
is easier when using rolls, giving the opportunity to
superpose a torsion moment which prevents twisting
of asymmetrical profiles during bending. For these
reasons it was necessary to carry out pre-
investigations in order to develop a new bending
procedure for 3D-bending of profiles. It seemed to
be reasonable to find and systematically investigate
new opportunities for the purposeful influencing of
the forming zone in the sense of stress superposition.
ABSTRACT: The paper shows a new method for three-dimensional bending especially of profiles. The
bending process only works in one plane according to the machine geometry. To leave the two-dimensional
bending plane, the profile cross-section is turned by an overlaid torsional moment. By this process very long
symmetrical and asymmetrical profiles can be bent three-dimensionally without surface damage and the
unwanted torsion of asymmetrical profiles can be prevented by means of a similar compensation moment.
The paper focuses on the bending method which is realized in the new experimental setup. The device
consists of a hybrid machine design of hydraulic and electrical drives, both numerically controlled. The new
device realizes a new roll-bending method that uses 6 transportation rolls and also a roll-based bending head.
The bending moment is applied in the profile by a transverse force driven by the bending head and a servo-
hydraulic cylinder. The turning of the cross-section during the process is realized by a special machine design,
making it possible to turn the guiding rolls and the bending rolls of the machine synchronistically.
Key words: 3D-Bending of Profiles, Roll-Bending, Superposition of Stresses, Torsion Compensation
Three-Dimensional Bending of Profiles with Stress Superposition
M. Hermes
1
, S. Chatti
1
, A. Weinrich
1
, A. E. Tekkaya
1

1
Institute of Forming Technology and Lightweight Construction (IUL) Technische Universitt Dortmund,
Baroper Str. 301, 44227 Dortmund, Germany
URL: www.iul.uni-dortmund.de e-mail:
matthias.hermes@iul.uni-dortmund.de;
sami.chatti@iul.uni-dortmund.de;
erman.tekkaya@udo.edu
2 PRE-INVESTIGATIONS
To pre-investigate the potential of a stress
superposition for 3D-bending of profiles, the
universal and flexible three-roll-bending process was
chosen. The kinematic 3D-bending of the profile
running through the rolls can be carried out by
means of a deflecting device influencing the forming
process between the three rolls. The profile
curvature, which is normally created by three-roll-
bending, is probably also irrelevant so that the first
process only serves the creation of the plastic zone
which is used for the easier second 3D-shaping
process. Irrespective of these aspects, the profile
curvature is produced kinematically. Figure 1 shows
the tool arrangement of the combined procedure
which has been realized at the IUL.
Movable
deflecting tool
Three-roll arrangement
Motion direction of the profile
z
y
x
d
d =0

Fig. 1. Superposed three-roll-bending with subsequent profile
deflection [4, 5]
The experimental set-up for this new way of profile
bending consists of two essential parts: a
conventional CNC-three-roll-bending-machine for
the bending step in the first plane and a special
device that makes it possible to deflect the profiles
towards the third axis directly after the first bending
process. The force for the bending in this axis is
given by a position controlled hydraulic cylinder.
For the first experimental phase the only
automatically moving axis is the hydraulic cylinder
(z-axis). Figure 2 shows a photo of the machine
during an early bending experiment. The guidance
of the profile after leaving the three-roll-system
works with a window fixed on the hydraulic cylinder
axis and furnished with PTFE plackets to reduce the
friction in the contact zone [4, 5]. The objective of
the procedure pre-investigations was to prove the
existence of certain effects which have the potential
for improvements in the field of three-dimensional
bending of profiles. It was important to highlight
some tendencies for solutions of classical bending
problems which are not solved or only solved with
extremely efforts by means of many other
procedures.
Three-roll-bending-machine
Hydraulic axis
Deflecting device

Fig. 2. Bending device for 3D-Profiles [4, 5]
First experiments carried out at the IUL included
investigations to achieve 3D-bent symmetrical
profiles (20x20x2 mm made of S 235 (EN 10 025)).
For this purpose, both the 3-roll-bending machine
(d-axis in xy-plane) and the hydraulic device (z-axis)
have been adjusted and synchronized in order to
produce the desired profile contour (figure 3). The
first profile bending in the xy-plane step has been
superposed with the second one, which is defined by
the position of the hydraulic device and the leading
window in the x, y, and z-axis.
z
x
y

Fig. 3. 3D-bending example of a profile by superposition of
bending stresses [4, 5]
To use and to make the advantage of the previous
plastic deformation in the three-roll-bending zone
for an easier bending in the orthogonal plane
evident, different roll adjustment values (d =0 mm,
13 mm, and 16 mm) have been selected at the
middle roll and a superposed force of the hydraulic
cylinder with the same adjustment value for the z-
axis of 40 mm has been used. It was found out that
the larger the middle roll adjustment causing a
higher previous plastic deformation of the profiles
is, the larger the profile curvature (smaller profile
radius) in the third plane will be (figure 4).
d = 0 mm
d = 13 mm
d = 16 mm
x
y
z
=
=
=
const.
const.
const.

Fig. 4. Influence of previous plastification on bending radii
A reason for the curvature increase is the decrease of
the profile springback due to the stress
superposition. In addition to the decrease of the
springback a force decrease in z-axis was measured
in the deflecting tool, which is related to a smaller
cross-sectional deformation [4, 5].
The concept of a machine set-up for 3D-profile
bending based on a conventional 3-roll bending
machine works only for relatively large profile radii.
The bending of smaller radii is problematic. The
transportation of the profile through the process is
no longer possible because the friction between roll
and profile is insufficient. To solve these problems
in connection with the second demand to develop a
concept for the 3D-bending of profiles, a systematic
proceeding is required. To create new procedure
variants a proceeding for design systematics,
recommended in [6, 7], has been chosen. It is often
used to design new products, but it is not common in
production technology. In [7], a similar method is
used, based on a morphological system, for the
production process of ring rolling.
3 A NEW CONCEPT GENERATED BY
SYSTEMATIC ENGINEERING DESIGN
The first step in the design analysis according to the
VDI 2221 is to reduce the technical problem
influencing the main function and then to make a
subdivision into main functions and auxiliary
functions [6]. The systematic for 3D-bending of
profiles is shown in figure 5. This simple system can
be subdivided into three main functions. The first is
the transportation of the profiles over their
longitudinal axis. The second is the plastification of
the material to allow a forming process. The third is
an essential function which defines the bending
geometry of the workpiece. Furthermore, the
realization of a suitable bending system requires a
guiding system for the profile to reduce the cross-
section deformation.
It is reasonable to limit the subdivision into further
subfunctions to a special degree of abstraction. The
following search for solutions for each subfunction
has been made on the basis of the only demand of a
roll-based system. The recombination of this work
produces a morphological matrix of solutions for the
problem. After an evaluation the best variant needs
to be chosen. A useful recombination for a process,
designed especially for open and closed non-circular
profile cross-sections, is shown in figure 6.
Systemboundary
Main function
bending of profiles with
free defineable contours
Material
(Profiles)
(Geometry)
Information
Energy
(3D-curved profile)
(3D-contour saved
in the product)
Transportation of the
profile over the
longitudinal axis
Energy transformation
Guiding of the profile
section
Generation of plasticity
and a forming zone
Prevention of cross
section deformation
Definition of the
curvature
Energy transformation Energy transformation
Guiding of the profile
section
Material
Information
Energy

Fig. 5. Design systematics and analysis
Unlike the system analysis in figure 5, the function
definition of the bending contour has been
subdivided into two subfunctions according to
dimensional aspects. The 2D-bending process has
been realized by a plane bending system and by the
positioning of the tangential position of the roll pair
at the end of the process. This solution realizes the
full function for bending 2D-curves. On this basis it
is possible to easily create a 3D-contour system by
turning the whole system over the longitudinal axis
of the profile, but without changing the position of
the roll system in the other axes. By this movement
the bending plane of the profile cross-section
changes during the process and the profile obtains a
3D-bending curve. Hence, it is a 2D-bending with a
superposed 3D-bending process.
Plastification of forming
zone
Transport, push over
profile longitudinal axis
Definition of 2D-contour
Definition of a 3D-contour
Support and guide
the profile
Guide and support the
profile to avoid cross-section
deformation
F
F
F, s
influence
through
friction lining
Additional
counter roll
Adjustment of third roll
1 1,1
1,2
2,1
3,2
1
2
2
3
3

Twisting all
rolls
X2
X1
X3

4-roll guidance
Several mandrel concepts
possible
Tangential adjusment
of third roll pair
1,1
1,2
3,1
z

R
F

Fig. 6. Combination of possible solutions
According to the method of VDI 2221 [6], the
different solutions have to be recombined to obtain a
new system for 3D-bending of profiles. The result of
the recombination is shown in figure 7. The new
system has three pairs of rolls and a roll-based
guiding system that defines the bending curve in a
horizontal plane. This bending axis is realized by
one horizontally mounted machine axis. The profile
leaves the device during bending in any case in the
horizontal plane. The turning axis for the definition
of the 3D-curve is realized by a torsion bearing and
a compensation axis positioned in the bending head.
By this function the bending plane can be changed
and a 3D-shape is produced. Furthermore, the
adjustment of a difference between the angle of the
roll pairs and the angle of the compensational axis
enables a superposition of torsion moments with the
bending process. This can be used for the prevention
of the twisting of asymmetrical cross-sections.
Bending roll
Compensation
axis of the
bending head
Profile transportation axis
Torsion bearing

Fig. 7. Principle of the new 3D-bending process
These new aspects are not easy to investigate by
experimental equipment based on a conventional 3-
roll-bending-system. Thus, a new experimental set-
up has been planned and designed at the IUL.
4 DESIGN CHARACTERISTICS OF THE NEW
DEVICE
In Figure 8, the system and the additional drives are
shown as a CAD-plot. The main 6-roll unit for the
transportation of the profile over the longitudinal
axis is pivot-mounted in a lunette-based bearing that
allows a high stiffness. The drive system is a hybrid
construction combining a servo hydraulic and
electrical drives. The twisting system is realized by
an electrical servo engine with a synchronous belt
drive with a maximum of 2000 Nm. The profile
transport is realized by an electrical six-roll-driven
system with a chain gear. The four-roll-based
bending head can be adapted to several profile types.
It is mounted on a hydraulic axis of 60 kN. By
means of this set-up it is possible to realize a
maximum bending moment of approx. 21000 Nm.
This provides the potential to bend typical industrial
profiles.
Workpiece
4-roll bending
head
6-roll unit
Lunette-based
bearing
Transportation drive
3000 Nm
Bending axis
60 kN

Fig. 8. New machine set-up (front view)
5 CONCLUSIONS AND OUTLOOK
A new machine set-up for the 3D-bending of
profiles has been developed (kindly supported by the
German Research Foundation (DFG)). The device
will be available at the IUL in the beginning of
2008. Detailed experimental as well as theoretical
investigations of the process using semi-analytical
and FE-process simulations are planned.
REFERENCES
1. S. Chatti, Production of Profiles for Lightweight
Structures, Habilitation thesis, University of Franche
Comt, Books on Demand GmbH, Germany, 2006.
2. R. Neugebauer, W.-G. Drossel, U. Lorenz, N. Luetz,
Hexabend - A new Concept for 3D-free-form Bending of
Tubes and Profiles to preform Hydroforming Parts and
Endform Space-frame-components, In: Proceedings of
the 7 ICTP, Yokohama, 28.-31.10.2002, Advanced
Technology of Plasticity (2002), Vol. 2, pp. 1465 -1470.
3. M. Murata, T. Kuboti, K. Takahashi, Characteristics of
Tube Bending by MOS Bending Machine, In: Proc. of
the 2nd Int. Conf. on New Forming Technology, 20. -
21.9.2007, Bremen, Germany, (2007), pp.135-144.
4. S. Chatti, M. Hermes, M. Kleiner, Three-Dimensional-
Bending of Profiles by Stress Superposition, In: Proc.
8th ESAFORM Conference on Material Forming, 27.-
29.4.2005, Cluj-Napoca, Romania, (2005), pp. 245248.
5. S. Chatti, M. Hermes, M. Kleiner, Three-Dimensional-
Bending of Profiles by Stress Superposition, In:
Advanced Methods in Material Forming, Springer
Verlag, (2006), pp. 101-118.
6. G. Pahl, W. Beitz, J . Feldhusen, K.-H. Grote,
Konstruktionslehre, Springer, Germany, 2004.
7. J .M. Allwood, A structured search for novel manufac-
turing processes leading to a periodic table of ring
rolling machines, In: ASME J. of Mechanical Design
May 2007, Volume 129, Issue 5, pp. 502-511.

1 INTRODUCTION
Laser forming is a process of gradually adding
plastic strain to a metal component to generate
desired shape. Laser forming can be used for
forming straight bends in high strength metal such as
titanium instead of hot brake forming. The laser
forming process involves scanning a focused or
partially defocused laser beam over the surface of a
workpiece to cause localized heating along the bend
line. The sharp thermal gradients in the material
cause the sheet to bend either toward or away from
the laser source. The resulting deformation of the
material, which is, bending toward the laser beam, is
permanent. By repeating the laser forming process,
either with over lapping or parallel scans, bend of
desired angle and radius can be obtained. Some of
the earliest works on laser forming of sheet metal
into Two-dimensional shape are attributed to Namba
[1, 2] in 1985. The laser forming process was first
modelled by Vollertsen, Geiger, and Li using both of
the FDM and FEM [3]. Vollertsen has suggested a
semi empirical model to predict bending angle as a
function of material and laser parameters [4, 5].
Kyrsanidi has developed a numerical model of the
laser forming process for steel by using a coupled
transient thermal-structural finite element analysis
[6]. Edwardson presents an investigation into the 2D
and 3D laser forming of metallic component [7].
2 EXPERIMENTAL EQUIPMENT AND SETUP
A pulsed Nd:Yag laser, Model IQL-10,with
maximum mean laser power of 400 W was used for
the experiments. Square shape pulse is the standard
output of this laser. The available ranges for the
laser parameters were 1-1000 Hz for pulse
frequency, 0.2-20 ms for pulse duration, and 0-40 J
for pulse energy. The experiments were conducted
with frequency 20 Hz and workpiece velocities from
2 to 4 mm/s. Two major factors were important for
selecting the 20 Hz frequency; firstly the required
overlapping of alternative laser pulses, regarding
process travel speed and absolute irradiated energy
per unit length of the workpiece and secondly
technical limitations of laser source that confines our
choices about each combinations of laser pulse
energy, pulse duration and frequency for each value
of average output power. The focusing optical
system was composed of three lenses with 75mm
ABSTRACT: The laser forming process can be most useful in the automation of sheet metal forming. Three-
dimensional transient temperature and stress fields complicate the simulation of the process. The aim of this
experimental study is to identify the response related to deformation and characterize the effects of process
parameters such as laser power, beam diameter, scans velocity and pulse duration, in terms of bending angle
for a square sheet part. Extensive experimentation, including a design of experiments, is performed to address
the above-mentioned aims. From these experiments it has been determined that laser forming using Nd:YAG
laser is a flexible manufacturing process for steel sheet bending.
Key words: Laser Forming, Laser Bending, DOE, Taguchi Method.
An experimental study of sheet metal bending by pulsed Nd:YAG laser
with DOE method
M. Hoseinpour Gollo
1
, H. Moslemi Naeini
1
, G.H. Liaghat
1
, M. J. Torkamany
2
, S. Jelvani
2
,
V. Panahizade
1

1
Dept. of Mechanical Eng., Faculty of Engineering, Tarbiat Modares University P.O.Box 14115/143,
Tehran, Iran
URL: www.Modares.ac.ir e-mail:hoseinpour@modares.ac.ir;moslemi @ modares.ac.ir

2
Paya Partov Laser Research Center, P.O.Box 14665-576, Tehran, Iran
URL: www.payapartov.com e-mail: mjtorkamny@ yahoo.com;sjelvani@yahoo.com
focal length and 250m minimum spot size. For
each combination of pulse energy and duration, the
laser beam was defocused to different extents to
obtain various spot diameters and power densities on
the workpiece surface. A 5000 W-Lp Ophir power
meter and LA3000 W-Lp joule meter were used to
measure average power and pulse energy. Pure
argon gas with a coaxial nozzle, and flow rate 5-10
liter/min was used as shielding gas. Coaxial
shielding supports the safety of optical elements
when operating in an industrial environment.
The shorter wave length Nd:YAG laser
light(1.06m) is more effective on heating sheet
metal because more energy is absorbed by metal
surface. The investigation presented is the laser
bending of 1mm thick mild steel St12 (AISI1010), a
cold rolled low carbon steel sheet. The size of the
samples is 100x100 mm. The samples were cleaned
using ethyl alcohol. The bending of the samples was
measured using a coordinate measuring machine
(CMM) at 3 to 5 locations along the scanning path
and their average was calculated. Material data is
given in table (1).

Fig. 1 Experimental setup for 2D laser forming
Table 1- Mechanical and Thermal properties ofAISI1010 [8]
Parameter Unit Value
Density [kg/m3] 7870
Youngs Modulus [GPa] 205
Tensile Strength [MPa] 365
Yield Strength [MPa] 305
Thermal conductivity [W/mK] 49.8
Coefficient of Thermal Expansion
at 20C,250C,500C
[10-6/K] 12.2, 13.5,
14.2
Specific Heat Capacity at
100C,300C,450C,700C
[J/kg.K] 448, 536,
649, 825
3 DESIGN OF EXPERIMENT
3.1 Taguchi Experiments
A set of experiments was performed to determine
the bending angle of sheet components formed by
laser. The effects of laser power, beam diameter,
scan velocity and pulse duration on the bending
angle are investigated experimentally.
To limit the experimental costs a Taguchi
experimental design is used. A L-9 Taguchi array
with four factors (power, beam diameter, scan
velocity and pulse duration) and three levels for each
factor is given in tables (2) and (3). Because
replication is used, the total number of experiments
is (3x9=27). The objective of these experiments is to
determine bending angle as function of the process
parameter mentioned above.
Table2- Factors and their corresponding levels
Factors Level 1 Level 2 Level 3
P -Laser-Power (W) 200 230 260
S -Beam Diameter (mm) 2 2.5 3
V -Scan Velocity(mm/s) 2 3 4
D -Pulse Duration(ms) 7 9 11

Table3-Orthogonal array or Taguchi design
Exp. P S V D P
(W)
S
(mm)
V
(mm/s)
D
(ms)
Bending
Angle()
1 1 1 1 1 200 2 2 7 3.285667
2 1 2 2 2 200 2.5 3 9 3.712
3 1 3 3 3 200 3 4 11 3.642667
4 2 1 2 3 230 2 3 11 4.226
5 2 2 3 1 230 2.5 4 7 3.730333
6 2 3 1 2 230 3 2 9 5.464667
7 3 1 3 2 260 2 4 9 2.179333
8 3 2 1 3 260 2.5 2 11 4.634667
9 3 3 2 1 260 3 3 7 4.351333
A Taguchi design of experiments has the advantage
of allowing the effect of each process variable
(called Main Effect) as well as any suspected
interactions between them (called Interaction
effect) to be statistically evaluated. In this case, a
popular statistical technique called ANOVA or
Analysis of Variance has been used. The software
MINITAB is used to perform the ANOVA.
3.2 Data analysis
The ANOVA indicates that the bending angle is
influenced by beam diameter, pulse duration, scan
velocity and laser power. There does not appear to
have any significant interactions between the beam
diameter and other process parameters, such as scan
velocity and pulse duration [9].
3.2.a Effect of factors on bending angle
Main effects plot of factors can be used to draw a
draft conclusion about effects of factors. These plots
are shown in Fig.2 for the sake of quick reference.
Fig.2a shows that laser power factor has a significant
effect on bending angle. It can also be seen from this
figure that the effect of this factor is directly
proportional to bending angle. In addition, it can be
stated that by increasing the laser power factor,
bending angle increases significantly.
3
3.5
4
4.5
5
5.5
190 210 230 250 270
Power (Watt)
B
e
n
d
i
n
g

A
n
g
l
e

(
D
e
g
r
e
e
)


Fig. 2a Effect of laser power on bending angle
3.6
3.8
4
4.2
4.4
4.6
1.9 2.4 2.9
Beam Diameter(mm)
B
e
n
d
i
n
g

A
n
g
l
e

(
D
e
g
r
e
e
)


Fig. 2b Effect of beam diameter on bending angle
3.6
4.1
4.6
1.8 2.8 3.8
Scan Velocity (mm/s)
B
e
n
d
i
n
g

A
n
g
l
e

(
D
e
g
r
e
e
)


Fig. 2c Effect of scan velocity on bending angle
3.6
4.1
4.6
6 8 10 12
Pulse Duration(ms)
B
e
n
d
i
n
g

A
n
g
l
e

(
D
e
g
r
e
e
)


Fig. 2d Effect of pulse duration on bending angle
Fig.2b illustrates the effect of Beam diameter on
bending angle. It can be seen that the increasing rate
of the bending angle is reduced by increasing beam
diameter. The relationship between the scan velocity
and the bending angle is shown in Fig.2c.It is clear
that the bending angle is decreased by increasing
scan velocity. The relationship also appears to be
quasi-linear. Eventually Fig.2d shows the effect of
pulse duration on bending angle. The bending angle
increases with increasing of the pulse duration.
3.2.b Effect of laser power and beam diameter on
bending angle
Fig. 3 shows the interaction between laser power and
beam diameter. It can be seen that the maximum
bending angle is obtained by higher laser powers at
beam diameter of 2.5 mm approximately.
Beam Diameter(mm)
L
a
s
e
r

P
o
w
e
r
(
W
)
3.0 2.8 2.6 2.4 2.2 2.0
260
250
240
230
220
210
200
Bending
4.0 - 4.5
4.5 - 5.0
5.0
Angle(Degree)
- 5.5
> 5.5
< 3.5
3.5 - 4.0

Fig. 3 Contour Plot of Laser Power and Beam Diameter on the
Bending Angle
3.2.c Effect of laser power and scan velocity on
bending angle
Fig. 4 indicates the effects of laser power and scan
velocity on the bending angle. The results show that
using lower scan velocity and higher laser power
leads to higher bending angle.

Fig. 4 3-D effect plot of Laser Power and Scan velocity on the
Bending Angle
3.2.d Effect of laser power and pulse duration on
bending angle
Fig. 5 shows the interaction between laser power and
pulse duration. The maximum bending angle is
obtained using the highest values of laser power and
pulse duration.
Pulse Duration
L
a
s
e
r

P
o
w
e
r
(
W
)
11 10 9 8 7
260
250
240
230
220
210
200
Bending
4.0 - 4.5
4.5 - 5.0
5.0
Angle(Degree)
- 5.5
> 5.5
< 3.5
3.5 - 4.0
Fig. 5 Contour Plot of Laser Power and Pulse Duration on the
Bending Angle
3.3 Regression analysis
Regression analysis is performed to find out the
relationship between factors and bending angle.
Accordingly, a first order polynomial best predicts
the observation. The regression equation in terms of
factors (table 2) is obtained and presented as below:

Bending Angle-(Degree) = - 3.10 + 0.0257P(W) +
0.489 S(mm)- 0.389 V(mm/s) + 0.161 D(ms)
(1)
3
3.5
4
4.5
5
5.5
6
3 3.5 4 4.5 5 5.5 6 6.5
Bending Angle (Degree)
B
e
n
d
i
n
g

A
n
g
l
e

(
D
e
g
r
e
e
)
Calculated from Regression
Experimental
Power 200-260 (W)
Beam Diameter 2-3 (mm)
Scan Velocity 2-4(mm/s)
Pulse Duration 7-11(ms)

Fig. 6 Comparison between experimental results and calculated
data from regression analysis
The R
2
value for equation (1) is 97.7%, indicating
that this model can be used with sufficient accuracy.
Fig.6 shows the differences between experimental
results and calculated data using regression equation
for bending angle. It demonstrates that the
regression equation gives the bending angle for
specific conditions with good accuracy.
4 CONCLUSION
In this paper laser bending of sheet metals has been
studied experimentally. Influences of main process
parameters including laser power, beam diameter,
scan velocity and pulse duration on bending angle
were investigated.
Bending angle is most strongly affected by beam
diameter, followed by pulse duration, scan
velocity and laser power.
Increasing laser power, beam diameter, pulse
duration and decreasing scan velocity increased
the bending angle.
A formula is obtained using a regression analysis
to predict bending angle.
REFERENCES
1. Y. Namba, laser forming in space, Proc. of Int. Conf.
onLasers, Las Vegas, NV, USA, (1985) 403-407.
2. Y. Namba, laser forming of Metals and alloys, Proc. of
LAMP, Osaka, Japan, (1987) 601-606.
3. F. Vollertsen, M. Geiger and W. Li, FDM and FEM
Simulation of laser forming: A Comparative Study, Proc.
of 4
th
Int. Conf. on Technology of plasticity, (1993) 1793-
1798.
4. F. Vollertsen, An Analytical Model for Laser Bending,
Lasers in Eng., 2 (1994) 261-276.
5. F. Vollertsen, Mechanism and Models for Laser
Forming, Proc. of Laser assisted Net shape Eng.,
Meisenbach Bamberg, Germany, 1(1994)345-359.
6. A.K. Kyrsanidi, T.B. Kermanidis and S.G. Pantelkis,
Numerical and Experimental Investigation of the Laser
forming process, Journal of Materials Processing
Technology, 87 (1999) 281-290.
7. S. Edwardson A Study into the 2D and 3D Laser
Forming of Metallic Components, PhD thesis, Laser
Group Department of Engineering the University of
Liverpool, UK (2004).
8. ASM Metals Handbook, ASM International, 10
th
Ed.
(1990).
9. R.L. Mason, R.F. Gunt and J.L. Hess, Statistical design
and analysis of experiment, John Wiley & Sons Inc.,
Hoboken, New Jersey, USA (2003).

1 INTRODUCTION
Springback is generally defined as the additional
deformation of sheet metal parts after the loading is
removed. In recent years, the high strength steels
and aluminum alloys are increasingly used for sheet
metal parts in the automotive industry to reduce
mass. But, because of these materials higher ratios
of yield strength to elastic modulus, precise
prediction and control of springback become
essential [1]. The precision in dimension is a major
concern in sheet metal bending process because of
the considerable elastic recovery during unloading
leading to springback. The elastic recovery is
influenced by a combination of various process
parameters such as tool shape and dimension,
contact friction condition, material properties,
thickness and so on.
Lee and Kim [2] focused on the evaluation of
springback occurring in the sheet metal flange
drawing by controlling some process factors like
punch corner radius (PR), die corner radius (DR),
and blank holding force. Esat et al. [3] carried out
springback analysis of different aluminum sheets
with different thicknesses and explored the relation
between the amount of springback and total

characteristic strain and also characteristic stress.
They concluded that the material with higher yield
strength and smaller characteristic strain has higher
amount of springback than the material with lower
yield strength and higher characteristic strain. Liu et
al. [4] proposed a method to control the forming
process of a U-shaped part by means of a reasonable
blankholder force curve.
In recent years, the rapid development of computer
technologies enables numerical simulation of sheet
metal forming operations and finite element codes to
be used in an industrial environment. The
springback prediction in sheet metal forming
processes using FEA has been studied by many
researchers in the past. Cho et al. [5] carried out a
numerical investigation on springback
characteristics in plane strain U bending process by
thermo-elastoplastic FEA. Li et al. [6] mainly dealt
with material hardening to analyze V bending and
showed that the material-hardening model directly
affects on springback simulation accuracy.
The goal of this paper is to explore a possible
relation between the amount of springback after
unloading and the characteristic strain distribution in
the sheet at the final stage in a sheet draw bending
process through the finite element code, ABAQUS
ABSTRACT: In this paper the relation between characteristic (equivalent plastic) strain and springback,
taking a benchmark of NUMISHEET'93 2-D draw bending and using the commercial FEM code, ABAQUS,
has been investigated. The simulations are preformed for three test materials: AA5754-O, AA6111-T4 and
DP-Steel. The effect of process parameters such as blankholder force, friction and blank thickness, on
springback has been studied. The obtained results for springback have been compared with some
experimental data. It is found that higher amount of characteristic strain in the sheet causes less springback
after unloading.
Key words: Characteristic strain, Springback, Draw bending, FEM
On the relation of equivalent plastic strain and springback in sheet draw
bending
M. Kadkhodayan, I. Zafarparandeh


Department of Mechanical Engineering, Ferdowsi University of Mashhad 91775-1111, Mashhad, Iran
E-mail: kadkhoda@um.ac.ir; i.zafarparandeh@yahoo.com

[7]. For this purpose, the influence of various factors
such as initial sheet thickness, blankholder force,
friction and different hardening models on
springback is studied. The obtained results are
verified by some experimental data reported in
literature.
2 FINITE ELEMENT MODELING
The 2D draw-bending problem in NUMISHEET'93
as shown in figure 1 is a case studied in this paper
for three materials: AA5754-O, AA6111-T4 and
DP-Steel [8]. The materials basic properties are
summarized in table 1. To obtain numerical
efficiency, the simulation of the U-bending process
is modeled in the finite element program
ABAQUS\Explicit, while the springback analysis is
simulated in ABAQUS\Standard as it would take a
long time to obtain a quasi-static solution of
springback analysis in ABAQUS\Explicit. Half of
the blank is modeled with a total of 300 shell
elements (S4R) and 9 integration points through the
thickness, with the symmetry boundary condition
along the Y axis. Mass densities used for the
dynamic explicit code are 2.7 gr/cm
3
for aluminum
alloy and 7.8 gr/cm
3
for high strength steel.

Table1. Basic material properties for the three test materials
AA5754-
O
AA6111-
T4
DP-
Steel
Thickness (mm) 1.5 1.25 1.2
Youngs Modulus (GPa) 73.25 75.25 205.35
Poissons ratio 0.33 0.33 0.3
Yield strength (MPa) 102.4 149.1 358.7
Ultimate tensile strength
(MPa)
234.2 279.3 570.9

Fig. 1. The 2D draw bending.
The initial dimension of the sheet was 300mm
(length) 35mm (width) with the 70mm total punch
stroke for three test materials. The punch velocity
was speed up to 10 m/s in the dynamic explicit code.
The springback parameters
1
and
2
studied by this
benchmark are shown in figure 2.

Fig. 2. The springback parameters (at the punch and die
corners).
3 RESULTS AND DISCUSSIONS
3.1 Initial sheet thickness
The value of initial thickness of the sheet clearly
affects on the amount of springback for DP-Steel as
it can be seen in table 2. It may be observed that
increasing the thickness causes reduction of
springback parameters, which means increasing of

2
and decreasing of
1
. In figure 3 the distribution
of characteristic strain for three different thicknesses
along a sample path, Path1 which locates along the
sheet length on the front side and top layer of the
sheet, is shown. A glance on the results of table 2
and figure 3 demonstrates that the sheet with larger
initial thickness will have higher level of
characteristic strain in the final stage of process and
exhibits smaller springback after unloading. This
idea can easily be extended to the other paths along
the sheet length.
3.2 Blankholder force
In order to investigate the effect of blankholder force
on springback two different blankholder forces, say
high and low, are considered while other factors
such as initial sheet thickness and friction are
assumed to be constant as they are presented in table
3. As it may be seen from the table, the increasing of
blankholder force, decreases the
1
and increases the

2
which leads to a reduced springback at the end of
process.
Table2. Influence of sheet thickness on the springback angels
Sheet thickness (mm)
Experiment[8] Simulation
1.2 1.5 1.25 1.2

1
() 108.26 98.79 102.21 104.22

2
() 81.79 84.54 81.9 81
0
0.05
0.1
0.15
0.2
0.25
0 20 40 60 80 100 120 140
Distance from the plane of symmetry (mm)
C
h
a
r
a
c
t
e
r
i
s
t
i
c

s
t
r
a
i
n
Thickness=1.2 (mm)
Thickness=1.25 (mm)
Thickness=1.5 (mm)
BHF=2.5 (kN)
Friction=0.1

Fig. 3. Distribution of the characteristic strain along the Path1
for three thicknesses.

Table3. Influence of the blankholder force on the springback
angels
Blankholder force (kN), Thickness=1.2(mm)

Experiment[8] Simulation

2.5 25 2.5 25

1
()
108.26 100.27 104.22 102.53

2
()
81.79 82.48 81 83.2
0
0.05
0.1
0.15
0.2
0.25
0 20 40 60 80 100 120 140
Distance from the plane of symmetry (mm)
C
h
a
r
a
c
t
e
r
i
s
t
i
c

s
t
r
a
i
n
BHF=2.5kN
BHF=25kN
Sheet thickness= 1.2 (mm)
Friction=0.1

Fig.4. Distribution of the characteristic strain along Path1 for
different blankholder forces.
The final level of characteristic strain evaluated
along the Path1 for two different values of
blankholder forces is displayed in figure 4. It is
found that the higher blankholder force which
generates itself a smaller amount of springback,
causes the sheet to achieve a larger characteristic
strain.
3.3 Friction
The friction between the blank and tools is one of
the factors that can be taken into account to control
the springback. In order to study the influence of
friction on springback, two friction coefficients are
selected in our simulations, i.e. 0.1 and 0.16 and the
Coulombs model is used to simulate the contact
friction. The corresponding obtained results are
summarized in table 4.

Table4. Influence of the friction coefficient on the springback
angels
Friction coefficient, Thickness=1.2(mm)
Experiment[8] Simulation
0.1 0.1 0.16

1
() 108.26 104.22 103.86

2
() 81.79 81 81.54
0
0.05
0.1
0.15
0.2
0.25
0 20 40 60 80 100 120 140
Distance from the plane of symmetry (mm)
C
h
a
r
a
c
t
e
r
i
s
t
i
c

s
t
r
a
i
n
Friction=0.1
Friction=0.16
Sheet thickness= 1.2 (mm)
BHF=2.5 (kN)

Fig. 5. Distribution of the characteristic strain along Path1 for
two friction coefficients.
Figure 5 shows the distribution of characteristic
strain along Path1 for different friction coefficients.
The results show that the increasing of friction
coefficient, reduces the springback, and at the same
time increases the final amount of characteristic
strain in the sheet.

Table5. Influence of the hardening models on springback
angels
Hardening model
Experiment[8] Simulation
ISO-KIN ISO KIN

1
() 100.9 102.53 99.94 102.29

2
() 82.31 84 83.2 87.66

3.4 Hardening models
Three hardening models are used in this study i.e.,
isotropic hardening model (ISO), kinematic
hardening model (KIN) and combined hardening
model (ISO-KIN) based on the Lemaitre and
Chaboche [9] work. The results of springback angels
predicted by the hardening models for DP-Steel are
presented in table 5. The isotropic hardening has
almost predicted less springback than other
hardening models. Figure 6 displays the distribution
of characteristic strain along Path1 at the final stage
of process. It is found that the isotropic hardening
which has predicted the higher amount of
characteristic strain expects less possible springback.
When it is important to consider the Bauschinger
effect in sheet draw bending, like for Aluminum
alloys' where there are several parts in the sheet that
undergo reverse loading, the results of the combined
hardening model will be more reliable.

0
0.05
0.1
0.15
0.2
0.25
0 20 40 60 80 100 120 140
Distance from the plane of symmetry (mm)
C
h
a
r
a
c
t
e
r
i
s
t
i
c

s
t
r
a
i
n
ISO-KIN
ISO
KIN
Sheet thickness= 1.2 (mm)
BHF=25 (kN)
Friction=0.1

Fig. 6. Distribution of the characteristic strain along Path1 for
three hardening models.
3.5 Material
The springback of angels and distribution of the
characteristic strain for the three test materials are
presented in table 6 and figure 7, respectively. The
material which has subjected to the higher
characteristic strain shows less springback.

Table6. Springback of angels for three test materials
Material
AA5754-O AA6111-T4 DP-Steel

1
() 97.66 98.42 102.53

2
() 86.68 84.5 83.2
4 CONCLUSIONS
In this paper the relation between the amount of
achieved characteristic strain in the sheet and the
springback of angels after unloading was studied. In
order to study this relation, influence of different
factors on springback such as initial sheet thickness
and friction was investigated. It was found that
increasing of friction, blankholder force or initial
sheet thickness could reduce the springback.
Generally, the results confirm that higher amount of
characteristic strain in the sheet leads to less
springback.
0
0.05
0.1
0.15
0.2
0.25
0.3
0 20 40 60 80 100 120 140
Distance from the plane of symmetry (mm)
C
h
a
r
a
c
t
e
r
i
s
t
i
c

s
t
r
a
i
n
DP-Steel
AA6111-T4
AA5754-O
BHF=2.5 (kN)
Friction=0.1

Fig. 7. Distribution of the characteristic strain along Path1 for
three test materials.
REFERENCES
1. Firat M., U-channel forming analysis with an emphasis
on springback deformation, Journal of Materials and
Design, 28, (2007) 147-154.
2. Lee S.-W., Kim Y.-T., A study on the springback in the
sheet metal flange drawing, Journal of Materials
Processing Technology, 187, (2007) 89-93.
3. Esat V., Darendeliler H., Gokler M.I., Finite element
analysis of springback in bending of aluminium sheets,
Journal of Materials and Design, 23, (2002) 223-229.
4. Liu G., Lin Z., Xu W., Bao Y., Variable blankholder
force in U-shaped part forming for eliminating
springback error, Journal of Materials Processing
Technology, 120, (2002) 259-264.
5. Cho J.R., Moon S.J., Moon Y.H., Kang S.S., Finite
element investigation on springback characteristics in
sheet metal U-bending process, Journal of Materials
Processing Technology, 141, (2003) 109-116.
6. Li X., Yang Y., Wang Y., Bao J., Li S., Effect of
material-hardening mode on the springback simulation
accuracy of V-free bending, Journal of Materials
Processing Technology, 123, (2002) 209-211.
7. ABAQUS Users Manual Version6.5, Hibbit, Karlsson
& Sorensen, Inc., (2003).
8. Lee M.G., Kim D., Kim C., Wenner M.L., Chung K.,
Springback evaluation of automotive sheets based on
isotropic-kinematic hardening laws and non-quadratic
anisotropic yield functions Part : applications, Int.
Journal of Plasticity, 21, (2005) 915-953.
9. Lemaitre J., Chaboche J.-L., Mechanics of solid
materials, Cambridge University Press, London, (1990)
161-241.
1 INTRODUCTION
In a typical sheet metal forming process, a lot of
material points undergo cyclic plastification, for
example during passing over a tool radius or through
a drawbead. Since the hardening process for metals
differs between forward and reverse loading
(Bauschinger effect), the stress state in the simu-
lation will depend directly on the ability of the
material model to describe this phenomenon.
Moreover, modelling the proper stress-strain res-
ponse for unloading is essential for accurate spring-
back prediction. In the last years numerous research
groups investigated the Bauschinger effect for sheet
metal forming and developed according models.
However, its industrial application is not wide-
spread, primarily because of the quite expensive
procedure to identify the many material parameters
usually required by such models. Several techniques
to perform tension-compression tests have been
developed during last years and the material
behaviour at reverse loading has been investigated
for some typical deep drawing alloys [1][2][3][4].
Based thereon we can dismember the Bauschinger
effect into three partial effects (Fig. 1): early re-
plastification (I), transient softening (II) and harde-
ning stagnation leading to reduced yielding (III).
The classical phenomenological assumption of the
existence of a linear-elastic domain is only a
simplification. In fact, during load reversal a super-
position of elastic lattice deformation and reorgan-
ization of dislocation structures takes place. Already
at the very beginning of unloading, motion of less
stable dislocation structures such as pile-ups occur.
This leads to early re-plastification and transient
softening [2][5]. Finally the workhardening stag-
nation is attributed to the dissolution of dislocation
cell-block boundaries [5].
The aim of the present work was to develop a
phenomenological model for reverse (and non-
proportional) loading of metal sheets, mainly to
improve springback simulation. Experimental results
of tension-compression tests and hat profiles are
taken from the EFB/AIF-project [6] performed at the
Fraunhofer Institute of Mechanics of Materials
(IWM) at Freiburg, Germany. In this paper, results
are presented for three high strength steels DP600,
TRIP700 and CPW800.
2 EXPERIMENTAL RESULTS
For the studied materials tension and tension-
compression tests for two different amounts of pre-
strain have been performed. The three mentioned
partial effects during load reversal have been obser-
ved for each material (Fig. 2). Remarkable is that the
initial value of tangent modulus at the strain reversal
is lower than the typical value of Youngs modulus
for steel (Fig. 3). According to the investigations in
[1][2][3][7] the tangent modulus at unloading even
measured at small reversal strains (e.g. vibrometri-
cally) reduces with increasing pre-strain. Here the
term elastic or Youngs modulus, which is a physical
material constant, is not used, on purpose, because
obviously a combination of elastic and plastic effects
takes place. A plausible explanation for the reduc-
ABSTRACT: This paper deals with the Bauschinger effect in sheet metal forming simulation. The concept of
a new material model is presented able to accurately handle the phenomena of early re-plastification, transient
softening and workhardening stagnation at load reversal. Besides the standard data used for simulations based
on isotropic hardening, the model requires only four additional, physically comprehensible material
parameters. The model is verified by means of tension-compression tests and springback simulations for hat
profiles for three typical high-strength steel alloys.
Key words: Bauschinger effect, Kinematic hardening, Sheet metal forming, Springback
Modeling of reverse loading effects including workhardening stagnation
and early re-plastification



W. Kubli, A. Krasovskyy, M. Sester



AutoForm Engineering GmbH, Technoparkstrasse 1, 8005 Zurich, Switzerland
URL: www.autoform.com mail: andriy.krasovskyy@autoform.ch
tion of tangent modulus at the start of unloading as
well as for early re-plastification and transient
softening is the immediate motion of less stable
dislocation structures, such as pile-ups. Other factors
which contribute to such material behaviour could
be change of the crystallographic texture during
plastic deformation, stress induced phase transfor-
mation or porosity evolution [2].
3 CONCEPT OF THE MATERIAL MODEL
A novel approach to model such reverse loading
effects has been implemented in the commercial
code AutoForm. To account for the industrial envi-
ronment in which this code is typically used, special
emphasis is placed on the applicability of the model:
it should be able to accurately describe the afore-
mentioned phenomena, should be compatible to any
of the implemented yield functions and hardening
laws, should not significantly increase computation
time, and should require as few as possible
additional material parameters. These should be in-
dependent of each other, should have a well defined
physical meaning and should be universal for a cer-
tain material group, so that expensive material tests
can be reduced to a minimum.
The detailed constitutive equations of the model are
undisclosed research; in the following, the basic
concepts of the model are presented.
3.1 Early re-plastification and transient softening
The precise description of the unloading and early
reverse loading phase is essential for accurate
springback prediction. The main idea of the model is
to use the same developing equation to describe the
two partial effects early re-plastification and tran-
sient softening resulting in a smooth stress function
for the entire unloading and reverse loading path,
including the area which is treated as elastic in
conventional models. The function consists of two
parts representing an initial tangent modulus after
the load reversal and its further reduction. The
model formulated for the general case is presented
here schematically for the uniaxial case:
rn rl r
& & & + = (1)
where
r
is the reverse strain and
rl
and
rn
are its
linear and non-linear parts (Fig. 1). The linear part
predominantly describing early re-plastification is
defined by the initial tangent modulus
l
E :

Fig. 1. Schematic of the model concept.
l
r
rl
E

&
& = (2)
In [1] and [7] it has been shown that the initial tang-
ent modulus
l
E typically reduces exponentially with
accumulated equivalent plastic strain p :
)) 1 ( 1 (
0
p
l
e E E

= (3)
where
0
E is the tangent modulus at 0 = p (equiva-
lent to the Youngs modulus). is a material
parameter representing the amount of reduction of
0
E , typically ranging between 0.1 and 0.2. is the
saturation constant with values between 20 and 50.
The non-linear part which is mostly responsible for
the transient softening is defined as:
) ), ( , ( K p
h r rn rn
& & & = (4)
where ) ( p
h
is the (isotropic) hardening stress and
K is a material parameter representing a typical
strain distance affecting the steepness of the reversal
stress curve
r
. Typical values range between 0.002
for materials with a small Bauschinger effect and
0.02 for a large effect.
Note the difference to other models treating the load
reversal by independent elastic and plastic charac-
teristics: If, in order to incorporate this concept into
the framework of classical elasto-plasticity, one
abandons the non-linearity within the yield surface
of the size a and uses an average secant modulus
s
E instead (Fig. 1), the tangent modulus for the
elastic part will be smaller than
0
E even if 0 = .
I.e. the elastic modulus does not independently deve-
lop with pre-strain as it does in other models. This
reflects the aforementioned observation that disloca-
tion effects start at the very beginning of unloading
and cause a smooth stress curve over the entire un-
loading and reverse loading path. Consequently, the
whole area of early re-plastification and transient

0
0.5
1
1.5
2
2.5
0 0.05 0.1 0.15 0.2 0.25 0.3
Effective strain [-]
E
f
f
e
c
t
i
v
e

n
o
r
m
a
l
i
z
e
d

s
t
r
e
s
s








[
-
]

experiment
model

0
0.5
1
1.5
2
2.5
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4
Effective strain [-]
E
f
f
e
c
t
i
v
e

n
o
r
m
a
l
i
z
e
d

s
t
r
e
s
s







[
-
]

experiment
model

0
0.25
0.5
0.75
1
1.25
1.5
0 0.02 0.04 0.06 0.08 0.1
Effective strain [-]
E
f
f
e
c
t
i
v
e

n
o
r
m
a
l
i
z
e
d

s
t
r
e
s
s









[
-
]

experiment
model

Fig. 2. Measured and calculated tension-compression curves.
(a) DP600; (b) TRIP700; (c) CPW800.
softening can be described very accurately by just
three physically well-defined material parameters
( , and K ).
3.2 Workhardening stagnation
For each of the materials studied in [6] workharde-
ning stagnation has been observed, presumably
caused by the dissolution of dislocation cell-block
boundaries. During stagnation, plastic flow is partly
achieved by reversal of dislocation structures and
not by generating new dislocations, so that work-
hardening is delayed. This results in a stress reduc-
tion at the end of the stagnation phase compared to
the stress achieved under isotropic hardening at the
same accumulated strain. In [4], this is called per-
manent softening and modelled as separate effect,
with its own parameters. Here, by contrast, the stress
reduction is treated as pure consequence of the
workhardening delay. As a result, the stagnation in-
cluding the stress reduction can be accurately descri-
bed by just one additional material parameter.
0
50
100
150
200
0 0.5 1 1.5 2
Normalized reversal stress [-]
T
a
n
g
e
n
t

m
o
d
u
l
u
s









[
G
P
a
]
experiment (small pre-strain)
experiment (large pre-strain)
model (small pre-strain)
model (large pre-strain)
h
r

0
50
100
150
200
0 0.5 1 1.5 2
Normalized reversal stress [-]
T
a
n
g
e
n
t

m
o
d
u
l
u
s









[
G
P
a
]
experiment (small pre-strain)
experiment (large pre-strain)
model (small pre-strain)
model (large pre-strain)
h
r

0
50
100
150
200
0 0.5 1 1.5 2
Normalized reversal stress [-]
T
a
n
g
e
n
t

m
o
d
u
l
u
s









[
G
P
a
]
experiment (small pre-strain)
experiment (large pre-strain)
model (small pre-strain)
model (large pre-strain)
h
r

Fig. 3. Measured and calculated tangent moduli after strain
reversal at different stress points. (a) DP600; (b) TRIP700; (c)
CPW800.
In order to model delayed workhardening, instead of
p , a new hardening parameter
d
p is introduced.
d
p
is identical with the equivalent plastic strain p
during proportional forward deformation. During
reverse or non-proportional deformation,
d
p deve-
lops slower than p . In order to determine if a defor-
mation is reverse, a storage surface in strain space is
introduced, similar to the concept described in [8]
and also used in modified form in [4]. A material
parameter controls the size of the storage surface
and thus the amount of stagnation; characterizes
the amount of forward strain which can be reversed
without workhardening taking place, and ranges
between 0 and 1.
4 MODEL EVALUATION
4.1 Tension-compression tests
With the model described above one can reproduce
the tension-compression curves to a high degree of
a)
b)
c)
a)
b)
c)
accuracy. On the mirrored stress-strain represen-
tation the workhardening stagnation is clearly visible
(Fig. 2). Also the initial value of the tangent
modulus and its evolution during load reversal can
be described well (Fig. 3).
4.2 Springback simulation
Using the presented material model fitted to tension-
compression tests hat profiles with a sheet thickness
of 1.5 mm and with two different die radii (5 mm
and 2 mm) have been simulated. The simulations
were run with AutoForm using shell elements with
11 integration points over the thickness and a
friction coefficient of 0.12. A uniform mesh with an
element size of 1.3 mm for the large die radius and
with an element size of 0.6 mm for the small die
radius was used.
In all cases the new model leads to an essential
improvement of springback prediction, in particular
for the wall curl prediction. For the small die radius,
isotropic hardening always leads to curl-in; the
observed curl-out can only be achieved with kine-
matic hardening.
5 CONCLUSIONS
A new material model for sheet metal forming has
been presented with the following advantages:
It can accurately describe tension-compression
curves of different material classes, including
early re-plastification, transient softening, work-
hardening stagnation and reduction of tangent
(elastic secant) modulus with pre-strain.
Only four additional, physically comprehensible
material parameters are required, which can be
determined from tension-compression tests.
They seem to be constant per material class
(needs to be further investigated); as
consequence, material tests may be avoided once
the according knowledge is built up.
The model can be used with any of the
implemented yield criteria and hardening curve
descriptions even in tabular form. No changes to
the material data used for simulations with
isotropic hardening, widely available in industry,
are necessary.
As shown for hat profiles, the new model im-
proves springback prediction considerably.
The increase of computational cost is below 5%
for most applications.




Fig. 4. Measured and calculated tension-compression curves.
(a) DP600; (b) TRIP700; (c) CPW800.
REFERENCES
1. F. Yoshida, T. Uemori, K. Fujiwara: Elastic-plastic behavior
of steel sheets under in-plane cyclic tension-compression at
large strain. Int. J. Plasticity 18 (2002) 633-659.
2. A. Krasovskyy: Verbesserte Vorhersage der Rckfederung
bei der Blechumformung durch weiterentwickelte
Werkstoffmodelle. Thesis Fakultt fr Maschinenbau der
Universitt Karlsruhe, Germany (2005).
3. R.M. Cleveland, A.K. Ghosh: Inelastic effects on springback
in metals. Int. J. Plasticity 18 (2002) 769-785.
4. F. Yoshida, T. Uemori: A model of large-strain cyclic
plasticity describing the Bauschinger effect and
workhardening stagnation. Int. J. Plasticity 18 (2002) 661-
686.
5. B. Peeters, S.R. Kalidindi, C. Teodosiu, P. Van Houtte, E.
Aernoudt: A theoretical investigation of the influence of
dislocation sheets on evolution of yield surfaces in single-
phase B.C.C. polycrystals. J. Mech. Phys. Solids, 50, (2002)
783-807.
6. Beurteilung der Leistungsfhigkeit von Materialmodellen zur
Blechumformsimulation. EFB/AIF-project (AiF 13530 BG)
report Nr. 244 (2005).
7. S. Thibaud, N. Boudeau, J.C. Gelin: On the influence of the
Young modulus evolution on the dynamic behaviour and
springback of a sheet metal forming component. Proceedings
of NUMISHEET 2002, Jeju Island, Korea, Vol.1 (2002) 149-
154.
8. N. Ohno: A constitutive model of cyclic plasticity with a
nonhardening strain region. Journal of Applied Mechanics 49
(1982) 721-727.
a)
b)
c)
1 INTRODUCTION
Today the food industry is locking to develop
attractive market products with a simple and useful
packaging. Particularly, for the sterilized food cans
the so called easy-opening lids are developed as an
attractive commercial product. This consists to open
the can without using utensils thanks to the effect of
a score line along the circumference of the lid. This
score line acts as local stress concentrator and
facilitates the crack initiation and its propagation
along the circumference. An optimal easy-opening
lid for food cans, consists to obtain a complete
opening using a minimum applied force and
avoiding any inadvertent opening. Many parameters
influence the performance of the easy-opening as:
geometrical parameters (residual thickness after
indentation, angle of indent, geometries of the panel
...) and the mechanical properties of the steel
(ductility, hardening, ) [1, 2].

This work aims to propose FEM based numerical
methodology able to optimize the easy-opening
lids using advanced constitutive equations
accounting for non linear isotropic and kinematic
hardenings strongly coupled with ductile isotropic
damage. The coupling between the ductile damage
and the elastoplastic constitutive equations is
formulated in the framework of the thermodynamics
of irreversible processes together with the
Continuum Damage Mechanics (CDM) theory. The
associated numerical aspects are discussed and
implemented into ABAQUS/Explicit using the
Vumat users subroutine.
2 THERMO MECHANICAL DAMAGE
CONSTITUTIVE EQUATION
The fully coupled elasto-plastic behaviour is
modelled in the framework of the thermodynamics
of irreversible processes with state variables ([3-4])
assuming the small elastic strain hypothesis with
large plastic strain. According to the first gradient
formulation, the external state variables: (, ) for
total strain tensor and the Cauchy stress tensor. The
internal state variables and their conjugate forces
are : (
e
, ) for small elastic strain tensor and the
Cauchy stress tensor; (, X) for the back-strain and
back-stress deviator tensors that describe the
kinematic hardening (i.e. translation of the yield
surface centre); (r, R) equivalent plastic driving
strain and stress representing the isotropic hardening
(i.e. variation of the yielding surface size) and (D, Y)
for isotropic ductile damage and its conjugate force,
which is also known as a damage strain energy
release rate.
The fully coupled constitutive equation formulated
in the rotated configuration according to the
objectivity requirement using the above defined state
variables ([3, 4]):
ABSTRACT: This paper presents the modelling and numerical simulation of the easy opening process
(indentation, perforation and tearing) of food lids. The objective is the virtual prediction of the tearing load by
using 3D finite element analysis accounting for the mixed isotropic and kinematic hardening together with the
ductile damage effect. The different materials parameters are identified using experimental tensile tests
conducted until the final fracture. The overall process is numerically simulated and the results compared to
the experimental measurement of the load-displacement curves.

Key words: Ductile damage, FEA, indentation process, easy-opening can lid, plasticity
Numerical simulation of easy opening lids for food cans using fully
coupled advanced constitutives equations with ductile damage
C. Labergere
1
, C. Dubois
1
, K. Saanouni
1
, O. Beigneux
2
, J.J. Li
2

1
University of technology of Troyes, ICD/LASMIS FRE CNRS 2848, 12 rue Marie Curie BP2060 10010
Troyes
e-mail: carl.labergere@utt.fr; christophe.dubois@utt.fr; khemais.saanouni@utt.fr

2
ArcelorMittal Maizires, voie Romaine BP 30320, 57283 Maizires les Metz
e-mail: olivier.beigneux@arcelormittal.com; jing-jing.li@arcelormittal.com;
The state relations:
( ) ( )
2
e
1
g D : = (1)
( ) ( )
2
1
2
X g D C
3
= (2)
( ) ( )
2
2
R g D Qr = (3)
e r a
Y Y Y Y = + + (4)

( )
( ) ( )
( ) ( )
2
e e e
1
e 1
e e
e
3 4
tr
dg (D)
3
Y g D
dD
2 dev : dev
+
+


=




(5)
( )
( )
1
1
2
:
3
=
a
dg D
Y g D C
dD
(6)
( )
( )
2 2
2
=
r
dg D
Y g D Qr
dD
(7)
Evolution Equation :
p
D n =

,
1
3 1 s X
n
2 g (D) X

(8)
( ) n a =

(9)
( )
2
1
r br
g D

=

(10)
s
0
Y Y
D
S (1 D)

(11)
In these equations is the fourth order symmetric
elastic properties tensor
e e
2 1 1 1 = + ; C is the
kinematic hardening modulus and Q is the isotropic
hardening modulus; a and b characterize the non
linearity of the kinematic and isotropic hardening
respectively; Y
0
(Threshold), S, s and characterize
the ductile damage evolution. Finally the Macaulay
brackets Z are used to define the positive part of
Z.
The deviatoric second order tensor n is the
outward normal to the isotropic Mises yield surface f
with damage effect defined by:
( ) ( )
1 2
0

= =
y
X
R
f
g D g D
(12)
in which
y
is the limit yield stress in uniaxial
tension i.e. the initial size of the yield surface in the
stress space. The function g
1
(D) and g
2
(D) are
positives and decreasing function of damage
variable D representing the effect of the ductile
damage on the mechanical behavior. Various forms
can be taken for these damage effect functions as:
( )
i
g D 1 D

= (13)
In this work, the case with =1 is used. Finally, the
friction between the tools and the sheet is taken as
the classical Coulomb model with the friction
parameter =0.1.
3 IDENTIFACTION METHODOLOGY
The identification of the materials parameters is
based on experimental results of tensile tests
conducted until the final fracture. The gage length of
the specimen is discretized with hexahedral trilinear
elements (C3D4R from Abaqus element library)
with a constant size of 0.2 mm. The dimensions of
the specimen gage length are 80x20x0,17 mm.
The identification procedure gives the following
values of the material parameters: E=210 GPa,
=0.35,
y
=285 MPa, Q=704 MPa, b=1.3, C=2000
MPa, a=80, S=80 MPa, s=1.3, B=10, Y
0
=0
As shown in figure 1 the damage zone localizes
inside along one shear band giving the final fracture
of the specimen.



Fig. 1. Damage distribution after 30 mm displacement
The global force-displacement curve predicted by
the model is shown in figure 2 compared to the
experimental data.

Fig. 2. Comparison between the numerical and experimental
global force-displacement curves.
4 THE SCORE LINE FORMING
The goal of the indentation is to weaken the zone of
drilling and opening in order to facilitate the crack
propagation during the opining. This operation
consists to form a score line along the circumference
of the lid (see figure 5). The sheet is placed between
the indent and the cylindrical anvil tools; the
displacement of the indent deforms the sheet and the
score line with a certain residual thickness. In certain
case, a large displacement of the indent leads to a
traversing cracks.
The figure 3 shows some micrography of score line
with different indent tools without any traversing
crack.

Fig. 3. Micrographs of different indentation
A 2D adaptive meshing is developed to enhance the
prediction of the numerical model [5]. In this case,
the configuration of plane strain is choosen. The
description of this test is shown in figure 4. The
initial thickness of the flange is equal to 0.17 mm.
Fig. 4. Geometry of the score line forming
The distribution of the damage for different residual
thickness h is summarized in figure 5. One can
verify that no traversing cracks develop along the
thickness of the flange.

h=0.16mm h=0.125mm

h=0.08mm
Fig. 5. Damage distribution at different indent tool
displacement values (h is the residual thickness)
The same operation is realized in the case of a 3D
real geometry of a lid. To save the CPU time only
five elements are put along the thickness in the score
line zone. The result is described in figure 7.



Fig. 6. 3D forming of the score line

5 NUMERICAL SIMULATION OF EASY
OPENING LIDS
The first step consists to perforate the lid with a
special form of a ring tool. This operation generates
generally the maximum of opening force. The ring
tool is supposed as rigid body and only the
symmetric part of the sheet is considered (see figure
7)

(a)
(b)
Fig. 7. The perforating operation of the sheet
An experimental procedure to measure the opening
force versus the displacement during the easy-
opening operation is performed by Arcelor Mittal
Company. The experimental facility is shown in
figure 8.
indent
anvil
flange
Edges of the lid

Fig. 8. Experimental apparatus for the opening test
In figure 9 are summarized different steps of the
opening process. The comparison between the
predicted and the experimental force-displacement
curves is shown in figure 10. Clearly the predicted
force is quite different from the experiment mainly
at the beginning of the tearing operation. Also some
oscillations of the force are obtained due to some
numerical instability caused by a bad choice of the
mass scaling factor. This aspect should be enhanced
in the future.


Step 1

Step 2

Step 3
Step 4
Fig. 9. The tearing operation of a lid for different steps

Fig. 10. Comparison between experimental and numerical
tearing forces
6 CONCLUSION
An advanced elastoplastic model accounting
for mixed non linear hardening fully coupled with
ductile damage has been shown helpful to predict
the opening force of a lid. A 2D adaptive remeshing
methodology has been used to optimize the easy-
opening lids of cans by minimizing the opening
force and avoiding the formation of any macroscopic
crack during the score line simulation by
indentation. However, some open aspects are still
under progress and will be addressed in prospect as
developing a 3D adaptive remeshing facility.
REFERENCES
1. R.H.J. Peerlings, J. Mediavilla, R.A.B. Engelen, M.G.D.
Geers, Towards a micomechanics-based modeling of
damage development during the forming of food can lids,
Journal of Enginneering Fracture Mechanics, (2007)
2. A. Monsalve, A. Artigas, D. Celentano, J.L. Basoalto, C.
Alvarez, Study and modelling of the opening and tearing-
off process of tinplate lids, Journal of Enginneering Failure
Analysis, vol 13, pp 210-225, (2006).
3. K. Saanouni., J.L. Chaboche, Computational Damage
Mechanics. Application to Metal Forming , Chapter 7 of
the Volume 3: Numerical and Computational methods
(Editors: R. de Borst, H. A. Mang), in Comprehencive
Structural Integrity, Edited by I. Milne, R.O. Ritchie and
B. Karihaloo, ISBN: 0-08-043749-4, (2003)
4. K. Saanouni, Y. Hammi, Numerical simulation of damage
in metal forming process, in Continuous Damage and
Fracture, Editor A. Benallal, Elsevier, ISBN 2-84299-247-
4, (2000), pp 353-363
5. C. Labergere, A. Rassineux,, K. Saanouni,
Endommagement et procd de mise en forme. Apport du
maillage adaptatif, 8
th
Colloque National en Calcul des
Structures, Giens, France, 21-25 Mai, (2007), CD.

1 INTRODUCTION
Aluminium-Foam-Sandwich (AFS), consisting of a
porous aluminium core with aluminium cover
sheets, combines a high stiffness to weight ratio with
good acoustic and thermal damping properties.
However, so far AFS is rarely used in industrial
applications because of its limited formability and
the problem of joining AFS and conventional sheets.
In order to expand the range of applications for this
material with regard to a future utilisation in vehicle
manufacturing, a process chain for the production
and processing of tailored blanks out of AFS in
combination with conventional aluminium sheets
has been developed at the Chair of Manufacturing
Technology within the DFG-Collaborative Research
Center 396. The sequence starts with an initial
cutting step of the unfoamed precursor material,
followed by the joining of the different sheets by
friction stir welding. Afterwards the AFS tailored
blank is formed, then foamed and finally trimmed
[1].
Recently a change from AFS material with cover
layers out of normalised aluminium alloys to AFS
material with cover layers of hardenable alloys of
the 6000 series has taken place. This material allows
a targeted enhancement of the hardness of the cover
layers and therefore of the stability of the whole
sandwich structure by adapted heat treatment.
In order to expand the shape forming possibilities of
AFS, studies of the failure mechanisms and the
formability at elevated temperatures were
performed. Besides, the foaming behaviour of
formed AFS and AFS-tailored blanks was
investigated.
2 FORMING CHARACTERISATION OF AFS
2.1 Failure Mechanism of AFS
Although AFS is a very promising material for light
weight construction, there are only few cases of
applications in commercial products. The main
reason for this missing breakthrough is the poor
formability of the AFS material, due to the porous
core layer consisting of a sintered compacted
mixture of aluminium alloy and foaming agent [2].
At room temperature only very low degrees of
deformation can be achieved with the unfoamed
ABSTRACT: Aluminium-Foam-Sandwich (AFS) is, because of its high stiffness and good damping
properties, a material with great potential for light-weight applications. However, the forming properties of
unfoamed AFS precursor material are very poor at room temperature. The formation of cracks at the TiH
2

grains of the foaming agent leads to a refractory failure of the material. Yet by the change to elevated forming
temperatures the core material becomes more and more ductile and the achievable degree of deformation can
be significantly improved. The forming operation shows a great influence on the subsequent foaming process
so that the compliance of certain construction guidelines is necessary.
Key words: Aluminium-Foam-Sandwich (AFS), Tailored Blanks, Sheet Forming.
Forming of AFS-Tailored Blanks at Elevated Temperatures
K.-H. Leitz
1
, A. Otto
1

1
Chair of Manufacturing Technology University of Erlangen-Nuremberg, Paul-Gordan-Strae 3, D-91052
Erlangen
URL: www.lft.uni-erlangen.de e-mail: leitz@lft.uni-erlangen.de; A.Otto@lft.uni-erlangen.de
AFS precursor material due to a refractory failure of
the core layer. A microscopic analysis of formed
AFS material shows that the hard particles of the
foaming agent TiH
2
constrain a gliding of the grains
and are causal for small microcracks (see figure
1 a)). These cracks lead to a refractory breakage of
the core layer (see figure 1 b)), a necking of the
cover layers and finally to the early failure of the
material.


Fig. 1. Microcracks (a)) caused by the TiH2 grains of the
foaming agent lead to a refractory breakage of the core layer
(b))

2.2 Formability of AFS at elevated temperatures
Because of the limited formability of AFS-material
at room temperature studies of the forming
behaviour at elevated temperatures in the range of
25 to 400 C were performed. Higher forming
temperatures are disadvantageous, because of a
beginning decomposition of the foaming agent.
In order to characterise the formability of AFS the
forming-limit diagram (FLD) was determined for
different temperatures according to ISO 12004 on a
heatable test bench for the determination of forming-
limit diagrams [3]. Different probe geometries were
used in order to cover strains from biaxial strech-
forming to uniaxial draw.


Fig. 2. Stamp force-pile height curves (a)) and achievable pile
heights (b)) for different temperatures
In figure 2 the force-pile height curves for solid
specimen and formed samples with maximum
achievable pile heights for different temperatures
can be seen. With increasing temperatures the
necessary stamp force decreases, while the
achievable pile height increases significantly.



Fig. 3. Forming-limit diagram of AFS for different
temperatures

The full forming-limit diagram of AFS at different
temperatures can be seen in figure 3. The achievable
deformation degree strongly increases with rising
temperature. The bending on the right hand side of
the FLD in the area of biaxial strech-forming is due
to the refractory failure of the core layer [4].
These investigations show that by the change to
elevated forming temperatures the formability of
AFS can be notably improved. Although a refractory
failure of AFS can not be prevented, it can at least
be significantly delayed to higher deformation
degrees.


Fig. 4. Deep drawing ratio of AFS for different temperatures
In order to cover the left hand side of the FLD and
evaluate the deep drawability of AFS the maximum
drawing ratio
max
was determined for different
temperatures after the method of Schmidt-
Kapfenberg. Whereas at room temperature AFS is
not deep drawable at all, the maximum drawing ratio
increases with rising temperatures to values of

max
=2.3 at 400 C (see figure 4). These studies
show that by the change to elevated forming
temperatures the formability of AFS can be
increased to a level comparable to conventional
blank sheets, so that the production of more complex
formed parts becomes possible. Besides, the process
becomes more and more robust with rising forming
temperature.
3 FOAMING PROPERTIES OF FORMED AFS
So far investigations on the foamability of AFS-
tailored blanks mainly concentrated on the influence
of the weld seam [5]. However the foaming
behaviour of AFS is also significantly influenced by
a preceding forming process, especially at extended
degrees of deformation. In figure 5 two probes of
AFS material that were firstly formed at elevated
temperatures and afterwards foamed by heating it up
to 585 C can be seen. The middle pore diameter in
the formed region is about 30 % higher than in the
non-formed one and the deformation leads to
connection failures between the cover layer and the
core. The so called U-problem leads to a higher
pressure of the foaming agent in the formed region.
The microcracks mentioned in 1.1, in combination
with this high pressure, act as seeds for big pores.


Fig. 5. Foaming of formed AFS

Besides, the U-problem leads to a lateral shift
between cover and core layer during the foam
expansion. This explains the numerous occurrence
of long, big pores on the border between cover and
core layer. The higher the degree of deformation, the
worse is the foam quality in the formed region of the
sandwich. Whereas the sample formed at 300 C
with a pole height of 25 mm still shows an
acceptable foam quality (see figure 5 a)), the sample
formed at 400 C with a pole height of 30 mm shows
very large pores and buckling of the cover layer (see
figure 5 b)). A comparison of the unformed border
regions with the formed center shows that the foam
quality is mainly influenced by the precedent
forming process. The heat treatment during the
forming process in contrast shows no notable effect.
4 PART FABRICATION OUT OF AFS-
TAILORED BLANKS
The so acquired knowledge about the extended
formability of AFS at elevated temperatures was
used in order to produce more complex parts out of
AFS-tailored blanks. AFS-tailored blanks allow an
adapted use of AFS material. Regions in which high
stiffness is required can be made of AFS, those that
require less stiffness or need to be attached to the
rest of the construction, can be made of cheaper
conventional aluminium sheets.


Fig. 6. U-formed parts out of different AFS-tailored blank-
geometries

Figure 6 shows U-formed parts produced by forming
at 300 C out of different AFS-tailored blank
geometries. Both the geometry of the part and the
geometry of the tailored blank show a significant
influence on the final part.
Besides, these investigations showed that for the
production of more complex geometries out of AFS-
tailored blanks certain construction guidelines have
to be considered. Firstly the friction stir weld seam
should not be located parallel inside the flange.
Otherwise the bend-strech load weakens the seam so
strong that it can indeed stand the forming operation
but collapses during the foaming process (figure
7 a)). Secondly fully embedded parts of AFS are
critical because of the high pressures occurring
during the foaming process can lead to a breakage of
the weld seam and big pores. This can be avoided by
inserting degassing possibilities (figure 7 b)). An
alternative solution might be the foaming inside a
mould.


Fig. 7. Construction guidelines for AFS-tailored blanks

Furthermore, the potential of forming AFS at
elevated temperatures was shown by the production
of a box shaped part out of AFS by hydroforming at
250 C (see figure 8). Despite an obvious U-problem
the box shaped part could be foamed. However,
degassing possibilities were necessary in order to
avoid big pores and achieve a homogenous foam
structure in the whole part.


Fig. 8. Box-formed part out of AFS produced by hydroforming
at 250 C without (a)) and with (b)) degassing possibilities
5 CONCLUSIONS
It could be shown that by the change to elevated
forming temperatures between 300 and 400 C the
formability of AFS can be significantly improved
and it becomes possible to produce complex parts of
AFS and AFS-tailored blanks. The forming
operation shows a significant influence on the
subsequent foaming process. By the consideration of
certain construction guidelines it is possible to
produce complex highly adapted parts out of AFS-
tailored blanks. The fundamentals for an application
of AFS in automotive light weight structures are laid
and the technology ready for transfer to an industrial
application. A further shortening of the process
chain could be achieved by a combined forming and
foaming process, where in a first step the AFS
material is formed in a heatable forming tool, then
the tool opens in order to make way for the foam
expansion. The construction of a suitable heatable
forming tool that can resist the high foaming
temperatures poses a challenge for future system
developments.
ACKNOWLEDGEMENTS
The authors would like to gratefully acknowledge the funding
of this project by the German Research Foundation (DFG)
within the DFG-Collaborative Research Center 396 and the
support received from the Erlangen Graduate School in
Advanced Optical Technologies.
REFERENCES
1. M. Schmidt, A. Otto, F. Albert, S. Drfler, A. Grimm, C.
Kgeler: Fehlertolerante Fgeprozesse im automobilen
Umfeld. In: M. Geiger, R.F. Singer, Robuste, verkrzte
Prozessketten fr flchige Leichtbauteile. Tagungsband
zum Industriekolloquium des SFB 396 (2006), Bamberg,
Germany: Meisenbach Verlag, p.121-132.
2. H.-W. Seeliger: Aluminium Foam Sandwich (AFS) Ready
for Market Introduction. In: Advanced Engineering
Materials 2004, 6, No. 6, Wiley-VCH, Weinheim,
Germany, p. 448-451.
3. International Standard ISO 12004 Metallic materials
guidelines for the determination of forming-limit
diagrams, 1997.
4. Y. Brchet: Microsturctures, Mechanical Properties and
Processes Computer Simulation and Modelling.
EUROMAT 99 Volume 3, Wiley-VCH, Weinheim,
Germany, p. 161-165.
5. S. Drfler; A. Otto: Investigations on foaming AFS-
tailored blanks. In: Cellular Metals and Polymers, Trans
Tech Publications, Zrich, 2005. p. 189-194.

Time-dependent Springback
H. Lim
1
, M. G. Lee
2
, J. H. Sung
1
, R. H. Wagoner
1

1
The Ohio State University, Department of Materials Science and Engineering,
2041 College Rd., Columbus, OH 43210 USA
1 INTRODUCTION
Springback, a result of bending and unbending
combined with stretching for formed sheet-metal
parts, is the elastically driven change of a part shape
after forming and unloading. Prediction and
compensation of springback are important to achieve
precise final part shape to avoid assembly problems.
In previous work, aluminum alloys were observed to
change shape for long periods after draw-bend tests
[1, 2]. Several autobody steels were reported to have
no such time-dependent behavior following similar
forming and unloading [2].
Wang et al. [1] suggested two possible underlying
mechanisms for the time-dependent springback in
aluminum alloys; room temperature creep and
anelasticity. Experimental results showed that creep
plays a dominant role in long term time-dependent
springback in aluminum alloys.
Similar experiments were recently performed using
traditional and advanced high strength steels
(AHSSs). Some AHSSs exhibit time-dependent
springback behavior, an effect not reported
previously for ferrous alloys. Draw restraining
forces and radius to thickness (R/t) ratios were
varied in the experiments. After forming and
unloading, angular changes were measured for 6
months. These results were compared with
simulations using a simple finite element model
based on residual stress- driven creep model.
2 EXPERIMENTAL PROCEDURES
2.1 Materials
In order to compare the time-dependent springback
of steels, three conventional steels (AKDQ, DQSK,
HSLA steels) and four AHSSs, (DP600, DP800,
DP980 and TRIP 780) were considered. Mechanical
properties of tested steels are listed in Table 1.

Table 1: Mechanical properties for tested steels

Materials AKDQ DQSK HSLA DP600 DP800 DP980 TRIP780
Y.S (MPa) 190 168 398 425 537 679 505
UTS(MPa) 312 294 459 672 807 988 846
E
u
(%) 26.3 24.1 16.6 16.5 11.5 9.4 14.1
2.2 Draw-bend test
Each sheet material was sheared to a length of
635mm parallel to the rolling direction and a width
of 50.8mm. With one end clamped to the left grip,
the strip was hand-formed around the radius to 90
and then the other side was clamped. The left
hydraulic actuator was programmed to maintain a
constant back force at a fraction of the materials
yield strength, while the right actuator pulls a
distance of 127mm at a constant speed of 25.4
mm/sec. Tool radii varying from 9.5 to 38.1mm
were used to assess the effect of R/t ratio on
springback. In this work, the tool, or roller, was set
to rotate at the same speed as the specimen was
ABSTRACT: Draw-bend tests performed some years ago on four aluminum alloys (2008-T4, 5182-O, 6022-
T4, and 6111-T4) revealed that specimens can continue to change shape for long periods, up to 15 months,
following forming and unloading [1]. Contemporaneous tests of autobody steels (DQSK, AKDQ and HSLA
steels) tested under identical conditions showed no such time-dependent springback over a 7 year period [2].
Current, preliminary results for a few advanced high strength steels (DP600, DP800, DP980 and TRIP780)
revealed time-dependent springback at room temperature; the sign of the springback reversing for certain
combinations of process conditions. Time-dependent behavior of four advanced high strength steel was
measured and creep-simulated for various test conditions. Comparisons show qualitative agreement, but the
simulations over-predict the magnitude of the effect

Key words: Springback, Time-dependent, AHSS, Draw-bend test, Anelasticity, Creep, Residual stress
URL: http://www.mse.eng.ohio-state.edu e-mail:{lim,sungj,wagoner}@matsceng.ohio-state.edu

2
Korean Institute of Machinery and Materials,
66 Sangnam-dong, Changwon-city, Kyeongnam, 641-010, South Korea
URL: http://www.kimm.re.kr e-mail: mang92@ kims.re.kr

pulled to minimize friction.


Fig. 1: Schematic drawings of draw bend test [3]

Upon starting the draw-bend test, the material
underwent tensile loading, bending, and unbending
at constant speed over the cylinder. At the end of the
test, the strip was taken out of the grips immediately
and the profile of the sample was traced onto the
paper to measure the initial springback angle. Time-
dependent angle changes were then measured at
various time intervals up to 6 months and later were
digitalized to calculate precise angles.
3 RESULTS
3.1 Static (time-independent) draw-bend tests
Initial springback angles were measured within 30
seconds of unloading the sample after forming.
Initial springback angles of AHSSs at different
normalized back forces are shown in Fig. 2. Both
conventional steels and AHSSs showed a decline of
springback angle with increasing back force and tool
radius, consistent with previous works [4, 5]. AHSSs
with higher yield stress showed larger initial
springback angles compared to conventional steels.

0
20
40
60
80
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
DP800 (R/t=4.5)
DP980 (R/t=4.5)
TRIP780 (R/t=4.1)
S
p
r
i
n
g
b
a
c
k

a
n
g
l
e
s

(
d
e
g
r
e
e
)
Normalized Back Force
HSLA (R/t=4.1)
AKDQ (R/t=5.9)
DQSK (R/t=4.2)
DP600 (R/t=4.5)

Fig. 2: Initial springback angles for tested steels

3.2 Time-dependent springback
Measurement of time-dependent springback has
been carried out for both conventional steels and
AHSSs at various back forces and R/t ratios. In
agreement with previous work by Wang et al. [1],
the tested conventional steels, AKDQ, DQSK, and
HSLA steels, did not show any time-dependent
behavior for 6 months after forming. However, all
tested AHSSs showed angle changes after forming.
Fig. 3 shows profiles of deformed samples measured
at various times after forming. At long times near
saturation, the angle changes are 7 for Al 6022-T4
and 2.6 for DP600. Fig. 3 (c), (d) and (e) show
maximum angle changes after 2 weeks for three
other AHSSs with thicknesses near 1.4mm.
Materials with higher yield strength showed larger
variation in the nearly saturated time-dependent
springback angle, approximately proportional to the
time-independent springback angle.


Fig. 3: Time-dependent springback of Al 6022-T4 and AHSSs
(a) Al602-T4 (b) DP600 (c) TRIP780 (d) DP800 (e) DP980

40
45
50
55
60
65
70
75
10 100 1000 10
4
10
5
10
6
10
7
DP600
R/t=4.8
S
p
r
i
n
g
b
a
c
k

a
n
g
l
e

(
d
e
g
r
e
e
)
Time (s)
F
b
=0.20YS
F
b
=0.40YS
0.38
0.36
0.40
F
b
=0.60YS

Fig. 4: Time-dependent springback angles of DP600

Fig. 4 shows springback angles of DP600 as a
function of log (time). The springback angle change,
, is linear with log (time) having a slope m
before it gradually saturates and the angle change
becomes negligible. In some test conditions,
especially with large back force, the direction of
springback angle reversed and then saturates at a
new smaller value. The initial linear response can be
represented as follows [1];

( )
0 0 1 0 0
( ) ( ) log / ( 1 ) m = + = + = s

where is the initial springback ( ),
is the angle change at (s) and m is the slope.
m values of aluminum alloys and AHSSs are listed
in Table 2.
0

1
30s =
( )

Table 2. m values for different materials

Materials m
HSLA, AKDQ, DQSK 0 [1, 2]
Al 5182-H18 1.07 ~1.58 [1]
Al 6111-T4 0.74 ~ 1.09 [1]
Al 6022-T4 1.14 ~ 1.59 [1]
Al2008-T4 0.57 ~ 0.99 [1]
DP 600 0.15 ~ 0.66
DP 800 0.46 ~ 0.68
DP 980 0.64 ~ 0.94
TRIP 780 0.43 ~ 0.54

The average m value of AHSSs is approximately one
half the value for aluminium alloys as shown in
Table 2. Saturation occurred at approximately 10
7
s
(3.5 months) for aluminum alloys, 510
6
s (1
month) for AHSSs.
3.3 Anelasticity and room temperature creep
In order to understand the basis of time-dependent
behavior in AHSS, two mechanisms were
investigated [1]: anelastic deformation and residual
stress driven creep. Anelastic strains were measured
after unloading from 1) uniaxial tension and 2)
compression and then tension, for up to an hour.

0
0.03
0.06
0.09
0.12
0.15
0.18
0 1250 2500 3750 5000

0
=0.05
0
=632MPa

0
=0.10
0
=667MPa

0
=0.15
0
=674MPa

0
=0.20
0
=670MPa
DP600
N
o
r
m
a
l
i
z
e
d

a
n
e
l
a
s
t
i
c

s
t
r
a
i
n
Time (s)

Fig. 5: Normalized anelastic strain after uniaxial tension for
DP600

Similar to aluminum alloys, anelastic strain for
AHSSs saturated within an hour, much shorter than
the saturation time for time-dependent springback.
The second mechanism considered, residual stress
driven creep, was measured by applying a constant
load and recording the creep strain digitally for 2
hours. Measured creep properties were fitted using
simple steady-state power law [6] as shown in Fig. 6


0
0.002
0.004
0.006
0.008
0.01
0.012
0.014
0 500 1000 1500 2000
= 470 MPa
= 538 MPa
= 605 MPa
DP600
Time (s)
C
r
e
e
p

S
t
r
a
i
n
Steady state creep
=7.53x10
-28
()
7.59

Fig. 6: Room temperature creep tests for AHSSs
3.4 Simulation of springback
A simple finite element model was constructed using
ABAQUS/Standard to simulate time-dependent
springback based on residual stress driven creep. A
shell element (ABAQUS element type S4R) with 51
through-thickness integration points, Von Mises
yield and isotropic hardening were employed. The
simulation process consists of three consecutive
stages: (1) time-independent elastic-plastic loading,
(2) time-independent elastic-plastic initial unloading,
and (3) creep of the unloaded specimen driven by
internal residual stress. Creep properties were
implemented in a form of steady state creep power
law. In order to improve the accuracy, the friction
coefficient (Fig. 7) between the material and the tool
were determined by comparing the measured and
simulated pulling forces. Fig. 7 compares measured
and simulated initial springback angles (t=30s).
Predicted initial springback angles showed good
agreement at F
b
<0.5 with deviation less then 10%.

10
20
30
40
50
60
70
0 0.2 0.4 0.6 0.8 1 1.2
Normalized Back Force
S
p
r
i
n
g
b
a
c
k

A
n
g
l
e

(
d
e
g
r
e
e
)
DP 600
R/t = 4.8
u=0.03
TRIP 780
R/t = 4.1
u=0.05
Measured
Simulated

Fig. 7: Static (time-independent) springback angles: simulated
and experimental results

The internal residual stress through the thickness of
the sheet after each simulation step is shown in Fig.
8. At the end of forming and unloading, the
maximum tensile residual stress is reduced by 70%
after 1.810
7
s (~7 months).

-600
-400
-200
0
200
400
600
800
-0.4 -0.2 0 0.2 0.4
S
t
r
e
s
s

(
M
P
a
)
Through thickness coordinate (mm)
Loaded
Unloaded
(t=0)
After creep
(t=1.8x10
7
s)
DP600
R/t=4.8
Fb=0.5
Steady state law


Fig.8: Simulated through thickness stress at each stages

0
1
2
3
4
10 100 1000 10
4
10
5
10
6
10
7
Exp1
Exp2
Simulated

(
d
e
g
r
e
e
)
Time (s)
DP600
R/t=4.8
F
b
=0.5

Fig. 9: Time-dependent springback angles of DP600: simulated
and experimental results

Fig. 9 shows simulated and measured time-
dependent springback angles for DP600. Results
show good qualitative agreement but the predicted
angle changes overestimate experimental results by
approximately a factor of two. Previous work on
aluminum alloys showed opposite results; simulated
results were approximately two times smaller than
the measured values [1]. Quantitative deviations
between simulated and measured time-dependent
springback angles may be attributed to approximate
material law implemented and the complex loading
states that would affect the creep behavior.

Table 3: Times to reach fractions of saturation strains ( ) or
springback angles ( ) for DP600.

Time (s) 0.5 0.8 0.9

Draw-bend measured 110
3
1.510
4
~10
5
Draw-bend, creep model 510
3
6.510
4
3.510
5
Anelasticity 210
2
1.210
3
210
3

Table 3 shows kinetics of measured time-dependent
springback, residual stress-driven creep model and
anelasticity. Times to reach fractions of saturation
strains or springback angles are compared.
Saturation here is defined by a zero slope of the
variable with respect to time. The kinetics of
anelasticity is 1-2 orders of magnitude faster than
measured and simulated springback. Therefore,
anelastic deformation contributes only to the short-
term response of the time-dependent springback,
consistent with previous work with aluminum alloys
[1].
4. CONCLUSIONS
Time-dependent springback was observed in AHSSs
for some combinations of sheet tension and R/t. In
general, increases with increasing back force
and started to drop when the front force exceeds
yield stress.
Room temperature creep and anelasticity were tested
as possible origins of the observed time-dependent
behavior. Anelasticity becomes negligible 1-2 hours
after unloading, making this mechanism unlikely to
dominate time-dependent springback, which occurs
over a period of several months.
The residual stress driven creep simulation showed
good qualitative agreement with experiment but
simulated results overestimated the time-dependent
springback angle.
REFERENCES
1. Wang, J.F., Wagoner, R. H., Carden, W. D., Matlock, D.
K., Barlat, F., Creep and anelasticity in the springback
of aluminum. Int. J. Plast., 2004. 20: p. 2209-2232.
2. Wagoner, R., Carden, W., Carden, W., Matlock, D.
Springback after drawing and bending of metal sheets.
in IPMM97 Intelligent Processing and Manufacturing
of Materials, vol 1. 1997. University of Wollongong:
Intelligent Systems Applications.
3. Li, K., Carden, W., Wagoner, R., Simulation of
springback. Int. J. Mech. Sci., 2002. 44(1): p. 103-122.
4. Takahashi, S., Kuwabara, T., Ito, K. Springback analysis
of sheet metal subjected to bendingunbending under
tension Part 2 (experimental verification). in
Advanced Technology of Plasticity Proceedings of the
5th ICTP, vol. 2. 1996. The Ohio State University,
Columbus,OH.
5. Carden, W.D., Geng, L. M., Matlock, D. K., Wagoner,
R. H., Measurement of springback. Int. J. Mech. Sci.,
2002. 44: p. 79-101.
6. Grafalo, F., Fundamentals of Creep and Creep Rupture
in Metals. 1966, New York: The MacMillan Company.
7. Li, K.P., Wagoner, R. H. Simulation of deep drawing
with various elements. in Proceedings of
NUMESHEET'99. 1999. University of Franche-
Compte, Besancon, France.


Tool And Blank Interaction In The Cross-Die Forming Process
R.A. Lingbeek
1,2
, T. Meinders
3
, A. Rietman
2
1
Netherlands Institute for Metals Research - Mekelweg 2 P.O. Box 5008 2600GA Delft
URL: www.nimr.nl e-mail: roald.lingbeek@inpro.de
2
INPRO GmbH - Hallerstrae 1, D-10587 Berlin
URL: www.inpro.de
3
University of Twente, Faculty of Engineering Technology - Postbus 217 7500 AE Enschede
URL: www.utwente.nl
ABSTRACT: The deformation of the press and the forming tools during a deep drawing process is small.
However, it has a signicant inuence on the formed product, since the draw-in is affected signicantly by this
deformation. This effect is demonstrated for the cross-die forming process. The process was simulated using the
commercial code ABAQUS, comparing different models for the forming tools and blank. The simulated process
behaves quite differently when rigid or deformable tools are applied. In the latter case, so-called tool-spacers
absorb a signicant part of the blankholder load, resulting in a stronger draw-in of the blank. In all cases, the
results depended heavily on the blank element type and on numerical settings for the contact algorithm. These
should be treated with great care when accurate results are required.
KEYWORDS: Die-design, Tool deformation, Sheet metal forming, Finite Element Method, Contact, Friction
1 INTRODUCTION
The deep drawing process is an extremely sensitive
procedure. Even phenomena that are hard to measure
may inuence the blank ow and therefore the prod-
uct quality. The deformation of the press and tools
during forming is such a phenomenon. The defor-
mations are small, but the inuence on the contact
pressure distribution is very large. In Finite Element
(FE) simulations, the tools are generally modeled as
rigid bodies, and simplications are applied in the
blank-tool contact calculations. Due to errors in the
calculated pressure distribution, the blank draw-in is
not always predicted accurately. As a solution, [4]
shows an industrial strategy where the draw-in on the
real press is made to t the simulation results. The re-
quired tool reworkings are time-consuming, accord-
ing to [3] approximately 350-500 hours are spent on
the average forming tool, and they need to be carried
out by experienced die technicians. The problems
are aggravated by the increased use of high-strength
steels.
The accurate prediction of the contact pressure dis-
tribution on the blank can help to reduce the amount
of tool reworking and it is the main focus of this pa-
per. The inuence of the following three items will be
investigated
Tool deformations
Blank thickness changes
Contact parameters
2 THE CROSS-DIE BENCHMARK
The interaction between blank and tools is complex
in any industrial deep drawing process. In order to
provide a comprehensible overview, it is useful to as-
sess the problem with a simpler forming process. The
focus in this paper will be on a benchmark process
called the cross-die, shown in Figure 1. It is used
industrially as a material test [1] and provides insight
in the formability of a steel grade: The idea is to in-
crease the blank-size in a series of forming tests until
fracture occurs. The maximum allowable blank-size
is dened as the cross-die benchmark value.
During the experiments, the process revealed a high
sensitivity to tool deformation. In the prototype press,
1
the tools are supported by a set of pins. Depending
on congurations of these pins different benchmark
results were found [1]. In order to reduce this sensitiv-
ity, small squares called spacers were placed around
the blank. These spacers are made from the same
sheet-material as the blank. The experimenters in-
tended to make the gap between blankholder and die
more even, because due to tool deformations, the
gap-width had become nonuniform. Unfortunately,
the spacers made the problem worse. In regular form-
ing processes, the forming tools are supported by a
larger surface, however, problems due to tool deec-
tion occur too.
Figure 1: The cross-die process
Table 1: Settings for the ABAQUS simulations
Tool model Rigid Deformable
Type Static implicit
Blank 4-node red. integ. shell
8-node solid-shell
Tools Rigid-body elements Solid elements
Contact Penalty
# of elements 33124 187298
A FE analysis is a good way to analyze the process
and to show the inuence of tool deformation and
the use of spacers. Due to limitations of the forming
simulation software, the tool deformation was cal-
culated in a separate structural FE simulation in [1].
However, a full calculation with deformable tools is
possible with a general purpose FE code. ABAQUS
has been used here to perform both a regular simu-
lation using rigid tool models, and a simulation with
deformable tools. Table 1 shows the settings of both
simulations.
3 TOOL DEFORMATION
The forming process is divided into two phases,
blankholder loading and forming. The contact pres-
sure from the tools onto the blank denes the amount
of friction and therefore the amount of draw-in. The
contact pressure distributions for deformable and
rigid calculations are compared after the blankholder
closing phase in Figure 2. Note that there is no pres-
sure in the middle area of the blank, as forming has
not yet started. In this process the blankholder area,
the part of the blank where it is clamped between die
and blankholder, is completely at. Therefore, a ho-
mogeneous pressure distribution was expected.
Figure 2: Pressure distribution after blankholder clos-
ing for deformable (left) and rigid tools (right)
This is the case for the calculation with rigid tools.
When the tools are allowed to deform, even the slight-
est deection (in the order of magnitude of 0.01mm)
of the tools results in a localization of the pressure
eld to the edge of the blank. The reason for this is
made clear schematically in Figure 3 (left). The de-
formation of the die after the completed forming stage
is visualized in Figure 3 (right). The deformation was
multiplied by 5000 for visualization purposes. Note
that the spacers also cause deformation in the tools;
they carry a part of the blankholder load.
2
Due to the in-plane compression the blank thick-
ens considerably during draw-in. The contact pres-
sure maximizes at the thickest spots, lifting up the
blankholder slightly thereby relieving the spacers.
These thickening spots can be observed on a photo-
graph of an experimental blank (Figure 4) as shiny
spots. In these areas the blank was polished due to
the high friction.
Figure 3: Deformation of tools (schematically) and FE
result (x5000)
Figure 4: Shiny spots on the blank reveal high-pressure
zones (picture courtesy of Corus RD&T)
The ABAQUS calculation shows the same pressure
spots (see Figure 5). In the left picture, rigid tools
were used. Because the blankholder is rigid, it is lifted
up entirely, almost completely relieving the spacers.
However, when the tools are allowed to deform, they
do take a considerable amount of the blankholder
force away from the blank. In the right picture, this
can be seen clearly: There is a high pressure on the
spacers, and the size of the high pressure spots is
reduced. Because of the reduction in blankholder
pressure on the blank, the draw-in is larger, as Figure
7 shows. Due to the larger draw-in, the calculation
with deformable tools predicts a higher tendency for
blank-wrinkling, whereas the calculation with rigid
tools predicts a higher risk for rupture.
Figure 5: Pressure distribution for rigid (left) and de-
formable tools (right)
4 BLANK THICKNESS CHANGES AND CON-
TACT MODELING
When comparing the high pressure spot predicted by
ABAQUS and the results from the experiment, the
calculation with rigid tools appears to be closer to re-
ality. The reason for this is not that the modeling of
the deformable tools is wrong, experiments conrm
that the spacers do carry a part of the blankholder load
and allow larger blank draw-in. Instead, it is likely
that the thickness distribution is predicted wrongly by
the ABAQUS solid-shell elements. The use of these
elements is required because the thickness change of
regular shells is not taken into consideration during
contact calculations.
3
As a comparison, a simulation was carried out with
regular shell elements using the FE codes PAM-
STAMP and DiekA. Each of these simulations pre-
dicts much more thinning in the vertical walls and
thickening on the blankholder area of the blank. The
results are shown in Figure 6. At the high-pressure
spots the thickening now amounts 0.2mm instead of
0.02mm. Therefore, the thickening is now an order of
magnitude larger than the tool deection.
Figure 6: Thickness for solid (left) and regular shells
(right)
5 CONTACT MODELING
Penalty contact is used most frequently in forming
simulations. Even when the settings are numerical
parameters, they inuence the physical outcome of
the simulation heavily. Two ABAQUS calculations
were carried out, one with default contact stiffness
and one where the contact stiffness was multiplied
with a factor of 0.1. Regular shells and elastic tools
were used. For the regular contact settings, the blank
only sticks at the corners and slips at the other loca-
tions. In the case of the softer contact settings, the
pressure is much more uniformly distributed around
the edge of the blank. It is still high enough to prevent
slip at the blank edge so draw-in is almost reduced
to zero, an erroneous result. However, the contact
settings are generally adjusted by the simulant to en-
sure convergence rather than to reect reality. Softer
penalty-factors generally reduce numerical problems
and also the calculation time.
6 CONCLUSION
The cross-die benchmark is specically sensitive to
tool deformations, as shown in experiments and sim-
ulations. The simulation is able to reproduce the in-
creased blank draw-in, caused by the spacers. Figure
7 shows that the difference is not negligible. There-
fore, taking tool deformations into account increases
the simulation accuracy. In this case the prediction
of rupture risk was improved, which is essential for a
material benchmark.
Figure 7: Blank contour for rigid (solid line) and de-
formable tools (dashed line)
It has proven to be even more important to predict
blank thickness changes correctly. Also, the FE re-
sults have been found to depend heavily on the pa-
rameters of the contact algorithm. Because of these
ndings the authors believe, more research into con-
tact and friction modeling could make a signicant
improvement in the accuracy of forming simulations.
ACKNOWLEDGEMENT
This work was carried out under projectnumber MC1.03166,
in the framework of the Strategic Research Programme of the
Netherlands Institute for Metals Research (www.nimr.nl). Corus
RD&T is kindly thanked for supporting the project.
REFERENCES
[1] E.H. Atzema, C.H.L.J. ten Horn, and H. Vegter. Inuence of
tooling layout on sheet forming process analysis. In P. Neit-
taanm aki et. al. (eds.) Jyv askyl a, editor, Proceedings EC-
COMAS, 2004.
[2] J. Cao et.al. Benchmark study report. In L. Smith et al.,
editor, Proceedings 6th NUMISHEET conference, Part B,
2005.
[3] H. Hayashi. Elastic deformation of tools in stamping of
large-scale autobody panels. In M. Tisza et al., editor, pro-
ceedings IDDRG, 2007.
[4] C.T. Wang, D. Zhang, and N. Goan. Draw-in map - a
roadmap for simulation-guided die tryout and stamping pro-
cess control. In L. Smith et al., editor, Proceedings 6th NU-
MISHEET conference, pages 6669, 2005.
4
1 INTRODUCTION
All methods conventionally used for bending tubular
sections (rotary draw bending, compression bending,
roll bending, etc. [1]) share some common features
present several differences in terms of productivity,
flexibility, cost and parts quality. Selection of the
right bending process for tubing depends on the
required quality of bend and production rate, as well
as some geometrical and technological parameters
(tube outer diameter, wall thickness, minimum bend
radius and material characteristics). Rotary draw-
bending is the most versatile, cost-effective and
precise method used to bend thin or light-wall tubes.
This method basically requires three items: a centre
bend die, a pressure die and a clamping die (left part
of Fig. 1). The workpiece is secured to the bending
form by the clamping die and, as the bending die
rotates, it draws the workpiece against the pressure
die. In the worst cases, the die set can be equipped
with an internal mandrel, in order to prevent wall
collapse, with a stationary insert, the wiper die, that
provides additional support just behind the tangent
point of the bend and works, in conjunction with the
mandrel, to prevent wrinkling on the inside of the
bend and with a pushing device, the booster, that
forces the trailing end of the workpiece, as it draws
into the bending dies [2].
Although it is an established cold forming process,
rotary draw bending of metal tubes is evolving into a
precision metalworking process that requires high
quality assurance, advanced knowledge and
experience. This is mainly due to the growing
number of applications of cold formed metal tubes:
in the automotive and aerospace fields, in air
conditioning and exhaust systems, for fluid lines and
even furnishing and design. Therefore, the main
challenge facing the tubing manufacturing industry
is to fulfil the customers demand on complex parts
and tight tolerances. This often requires difficult
investigations into the best process parameters and
leads to onset and growth of defects and failures in
tubular parts, such as undesired deformation,
inaccuracy of bend angles and geometry, wall-
thinning, flattening, wrinkling, etc.
The setup and correct dimensions of the tooling
generally depends on process parameters such as the
tube material, the wall factor WF (ratio of the tube
outside diameter OD to the wall thickness t), the
ABSTRACT: A new computational methodology has been developed that enables to rapidly obtain feasible,
close to optimal displacement curves for the control of some of the key tools, namely the booster and the
pressure die. Considering the main geometrical data (thickness, external diameter, bending radius and bending
angle) and the strain distribution in a draw-bending process, the algorithm generates the right correlation
between the booster and the pressure die velocity curves. The development of the proposed method has been
carried out with an explicit FEM commercial code and has been verified comparing the numerical output and
the experimental results. In this paper, the effects of the variation of booster and pressure die displacement
curves will be also highlighted.
Key words: Tube bending, Rotary draw bending, Booster, Pressure die, Severe bends, FEM.
A new method for feasibility study and determination of the loading
curves in the rotary draw-bending process
A. Mentella
1
, M. Strano
1
, R. Gemignani
2

1
Dip. Ingegneria Industriale Universit di Cassino 43, via G. Di Biasio - 03043 Cassino (FR) - Italy
URL: www.dii.unicas.it e-mail: alessia.mentella@unicas.it;
m.strano@unicas.it

2
BLM s.p.a. 30, via Selvaregina 22030 Cant (CO) - Italy
URL: www.blm.it e-mail: roberto.gemignani@blm.it;

tube
pressure die
bend die
clamping
die
wiper die
RM RM mandrel mandrel
booster booster
v
S
v
B

P
R
E
S
S
U
R
E

D
I
E
BOOST
BLOCK
a. Pressure die only
(a
1
stationary, a
2
follower, a
3
boosted)
b. Pressure die
(boosted) with
connected booster block
c. Pressure die (c
1
follower, c
2
boosted) and
independent axial booster
P
R
E
S
S
U
R
E

D
I
E
P
R
E
S
S
U
R
E

D
I
E
BOOSTER
P
R
E
S
S
U
R
E

D
I
E
BOOST
BLOCK
P
R
E
S
S
U
R
E

D
I
E
BOOST
BLOCK
a. Pressure die only
(a
1
stationary, a
2
follower, a
3
boosted)
b. Pressure die
(boosted) with
connected booster block
c. Pressure die (c
1
follower, c
2
boosted) and
independent axial booster
P
R
E
S
S
U
R
E

D
I
E
P
R
E
S
S
U
R
E

D
I
E
P
R
E
S
S
U
R
E

D
I
E
BOOSTER
P
R
E
S
S
U
R
E

D
I
E
BOOSTER

Fig. 1. Scheme of rotary draw bending (left) and different configurations for axial assist (right)
bending factor BF (ratio of the centreline bending
radius RM to the diameter OD), the difficulty factor
FD=OD
2
/(tRM). Once identified the specific
tooling, the loading curves also depend on the
process data, on the tooling conditions and on the
machine setup. Several alternative methods are
traditionally used which combine the pressure die
assist and the axial boost. In the right part of Fig. 1,
six different approaches (a
1
, a
2
, a
3
, b, c
1
, c
2
) are
presented with three possible tooling configurations
(labelled as a, b and c). When the pressure die
is boosted (a
3
, b, c
2
), it is generally controlled by
displacement. The c configurations, with
independent booster, provide the greatest process
flexibility and performance. The traditional way of
controlling the independent booster is to apply a
constant axial load. Very limited studies or
procedures are available in the scientific literature
providing criteria for selecting the axial assist of the
rotary draw bending process [3]. Therefore, in order
to aid the design of the rotary draw bending process,
a simple computational methodology has been
developed, that enables to rapidly obtain feasible,
close to optimal velocity curves for the most critical
cases, when configuration c
2
must be adopted. The
proposed method is based on displacement control
of both booster and pressure die. In fact, the
proposed design procedure is FEM based and a
stronger correlation between experimental and
numerical results is observed when position control
is preferred over force control.
2 DETERMINATION OF THE DISPLACEMENT
OF BOOSTER AND PRESSURE DIE
As already mentioned in the previous section, the
coordinated control of both the pressure die and the
independent booster is very important for a
successful bending operation. A procedure has been
developed and is hereby presented, aimed at
searching for the optimal displacement curves of
these two tools. First of all, define v
M
as the
tangential velocity of the bend die, given by:
RM v
M
= (1)
where is the angular velocity of the bend die, v
S

and v
B
are, respectively, the tangential velocity of
the pressure die and of the booster. As a reasonable
assumption, these two variables should be strictly
correlated to v
M
, which can also be interpreted as the
axial velocity of the tube at the section immediately
before the bending region (see left side of Fig. 1), if
no axial assist is provided. Two factors
S
and
B
can
be used for tuning a proportional law between v
M

and the tools velocity:
B M B
S M S
v v
v v

=
=
(2)
The maximum principal true strain at the tube
extrados (axial strain
1
), if assuming volume
constancy, isotropic material behaviour and no shift
of the neutral axis, can be roughly calculated as:

+
=
RM
d
t OD
RM

0
1
2
ln (3)
where is the bend angle. Mainly due to the axial
assist of the pressure die and the booster, the actual
length of the external fibre is reduced by an amount
d
0
. In pure bending, the component d
0
is null. It can
be normalised as:
RM
d

0
(4)
Under the assumptions listed above, the axial strain

1
is approximately double the thickness strain th.
th = 2
1
(5)
can now be correlated to the desired value of wall
thinning th. Considering Eq. (4) and (5), Eq. (3) can
be rewritten as:
( )
max
2 exp
2
1 th
RM
t OD

+ = (6)
The total axial stroke d of the tube at the extrados,
immediately before the bending region, can be
written as:
0
d RM RM d + = = (7)
The term RM is mainly due to the bend die and the
term d
0
is mainly due to the assist tools. A correction
factor can now be calculated as:
+ =

+ =

+
= 1 1
0 0
RM
d
RM
d RM

(8)
The factor can be interpreted as the amount of
extra axial feed which must be provided by the assist
tools, if the target maximum thinning th
max
must be
reached. By substituting Eq. (8) into (6) one gets:
( )
max
2 exp
2
2 th
RM
t OD

+ = (9)
Finally, can be used to find the values of
S
and
B
.
In other words, the axial assist effect should be
distributed between the booster and the pressure die.
Since the pressure die acts mostly on the extrados,
while the booster acts on the whole tube section,
S

should be greater than
B
. Another reasonable
condition is that the average value of
S
and
B

should be greater than the calculated , since
obviously not all the displacement of the booster and
of the pressure die can be transformed into
displacement of the tube at the section immediately
before the bending region. Following the previous
considerations and observing the results of FEM
simulations, the following constraints have been set:


+
>
1 . 1 2
S B
B S
(10)
By running several FEM analyses with different
combinations of
S
and
B
, the following rules have
been identified as optimal:
( )
( ) 1 , 92 . 0 max
1 , 25 . 1 max


=
=
B
S
(11)
In the following Section, the FEM model will be
presented along with its experimental evaluation. In
Section 4, the proposed method is applied to 3
different cases, producing final maximum thinning
values which are very close to the targets.
3 FINITE ELEMENT MODELING
3.1 Description of the FEM model
All simulations have been carried out with the
commercial explicit code LS-DYNA; pre-processing
have been carried out with a special purpose macro,
which made possible the automation of the model
generation, in terms of tools setup, contact
conditions, kinematics constraints and material
properties. Post-processing has been carried out with
Ls-Prepost. Due to the model geometrical symmetry,
the overall analysis has been led only considering
half tube. The model is shown in Fig. 2. The
mechanical properties of the material are shown in
Table 1. The constitutive equation used to describe
the material elastic-plastic behavior is:
( ) ( )
n
p e p
K + =
(12)
where K is the strength coefficient,
e
the offset
elastic strain, n the work-hardening exponent and
p

is the effective plastic strain. In order to better
describe the tube bending deformation, a four-node
thin shell (Belytschko-Tsay) element has been used,
with five integration points across the thickness,
while the tools have been set as rigid bodies and
modeled with four-node shells. Surface to surface
contact and the classical Coulomb friction model
have been chosen to represent the interfaces friction
conditions [4]. Table 2 shows the friction
coefficients in each contact; the friction coefficient
value for the clamps has been chosen equal to 1.99,
in order to effectively represent the locking action.
Tube
Bend die
Mandrel body
Booster
Pressure
die
Mandrel balls

Fig. 2. Illustration of FEM model for CNC bending process.
Table1. Mechanical properties of tube
Material AISI 304
Ultimate tensile strength
f
788 MPa
Extensibility A% 53
Poissons ratio 0.28
Initial yield stress
s
205 MPa
Hardening exponent n
Strenght coefficient K
Youngs modulus E
0.224
954 MPa
196.5 GPa

Table2. Friction coefficients in various contact interfaces
Contact interface Static f. c. Dynamic f. c.
Tube/Pressure and bend dies

0.57 0.35
Tube/Wiper die 0.30 0.15
Tube/Clamp die 1.99 1.99
Tube/Mandrel
Ball/Mandrel (spherical joint)
0.075
0.055
0.055
0.055

Table3. Tube geometrical data and process parameters
Outside diameter, OD

35mm
Initial Thickness, t 0.8mm
Mean radius of the bend, RM 40mm
Bend angle, 90

Table4. Tube geometrical data, process parameters and outputs.
Process data Case 1 Case 2 Case 3
Outside diameter, OD

35mm 85mm 76mm
Initial Thickness, t 0.8mm 2mm 1.5mm
Mean radius of the bend, RM 35mm 85mm 114mm
Bending angle,
Difficulty Ratio FD=(OD
2
)/(tRM)
90
43.75
90
42.5
90
33.8
Booster coefficient,
B

Pressure die coefficient,
S

1.05
1.42
1.05
1.42
1.0
1.25
Target maximum thinning, th
MAX
0.15 0.15 0.14
Output maximum thinning, th
MAX
0.15 0.14 0.14
3.2 Model validation by experiments
The described elastic-plastic FE model has been
verified by using a stainless steel (AISI 304) tube,
whose geometrical characteristics, together with the
main process parameters are shown in Table 3.
As shown in Fig. 3 all the points representative of
the strain fall below the critical FLC and the tube is
safe both in the experiment and simulation.
Moreover, as shown in Fig. 4, theres a good relation
between the thickness measured at the outside of the
deformed tube and the thickness obtained from the
simulation. The shape of the simulated bending
moment vs. time curve is very similar to the
experimental moment.
4 APPLICATION OF THE METHOD
The procedure described in Section 2 has been
applied to 3 different bending operations using the
FEM model described in Section 3 and the
conditions of Tab. 1 and 2. Specific conditions can
be found in Tab. 4. These cases have been selected
with significantly different values of process
parameters. Cases 1 and 2 are particularly critical, as
signaled by the large value of the difficulty ratio FD.
In all of these application examples the calculated
factors
S
and
B
yielded a final value of maximum
thinning very close to the target.
5 CONCLUSIONS
The amount of axial displacement of the assist tools
(booster and pressure die), expressed through the
parameters
S
and
B
, is particularly important in
critical bending conditions. In fact, experiments and
simulations show that the maximum thinning
decreases as
S
and
B
increase. However, these
values cannot be indefinitely increased, since
wrinkling may occur, especially as
B
increases.
The study proposed a method for determining
S
and

B
. The method has been evaluated by successfully
applying it to three different critical bending
operations.

Fig. 3. Forming limit diagram.

Fig. 4. Thickness distribution at outside of the bend.

REFERENCES
1. R.J. Kervick, R.K. Springborn, Cold bending and
Forming Tube and Other Sections, American Society of
Tool and Manufacturing Engineers Michigan (1966).
2. G. Miller, Tube forming processes: a comprehensive
guide, SME, Society of Manufacturing Engineers
Dearborn, Michigan (2003).
3. M. J. Worswick, A. Bardelcik, Numerical Investigation
Into the Effects of Bending Boost and Hydroforming
End-Feed on the Hydroformability of Dp600 Tube, SAE
2005 World Congress & Exhibition, April 2005, Detroit,
USA, SAE Paper No. 2005-01-0094.
4. J.O. Hallquist, D.W. Stillman, T.L. Lin, LS-DYNA
Keyword Users Manual, Livermore Software
Technology Corporation Livermore (1997).
1 INTRODUCTION
All process of 2D drawing of a rectangular metal
sheet can be divided in two types. The first one is
the non stationary drawing when at the beginning of
the process the sheet is in tools and some parts of
tools move during this operation to shape metal
sheet. In this case different parts of sheet undergone
different loading histories. The examples of non
stationary drawing are draw-bend test [1], V-
bending [2].
Another type of 2D drawing is the stationary
drawing (or strip-drawing). In this case the tools are
fixed while the sheet moves in tools. Each part of
sheet undergone successively the same loading
history. The most famous process in this family is
the draw-bead test [3].
The behaviour of sheet metal in strip-drawing is of
significant interest to engineers in stamping industry.
Indeed, during this operation the sheet is stretched,
bent, unbent several times, so the sheet after this
operation has a changed micrographic texture and a
significant level of residual stress. Therefore, in the
modelling of this process the particular cyclic
behaviour of the material of sheet is required.
This work proposes a semi-analytical model to
describe the process of strip-drawing. This model is
a generalization of the approach of Inkin et al. [4]
which was based on the analytical study of Sanchez
and Weinman [5]. It allows the determination of the
profile of metal sheet in tools, stress and strain fields
and forces acting on tools during drawing.
2 MODEL
2.1 General description
Sheet metal in tools can be divided in three types of
elements: unsupported arcs, punctual and distributed
contacts. It must be remarked here that for instance
only plane and cylindrical types of contact surfaces
are used in the presented model.
For the mechanical behaviour description of the
sheet material some basis equations were used:
- Hill48 yield surface with planar anisotropy
ABSTRACT: A semi-analytical model which describes the process of strip-drawing is proposed. In this
model the strip in tools is divided in three types of zones: unsupported arcs, zones of line contact and zones of
distributed contact. In each zone the using of the forces and the moments equilibrium with the relation
between internal forces, moments and components of stress allows the determination of strain state and
subsequently stress state in each point of sheet. For the correct modelling of hardening effect the step by step
incremental plasticity was used and for the correct modelling of bending/reverse bending cyclic effect the
mixed isotropic-kinematic hardening laws were implemented in the program code. At the output the program
provides the deformed profile of sheet, stress and strain states of any point of the strip, thickness variation,
internal forces and moment of each sheet element and the forces acting on each surface of the tools in contact
with the strip.
Key words: analytical model, strip-drawing, cyclic plasticity, draw-bead
Analytical model for strip-drawing
R.A. Nazarov
1,2
, Z. Ayadi
1
, S.A. Nikulin
2
, M. Nivoit
1

1
NANCY UNIVERSITY, LSGS (EEIGM) 6 rue Bastien Lepage 54010 Nancy, France
URL: http://www.mines.inpl-nancy.fr/wwwlsgs/ e-mail:Roman.Nazarov@eeigm.inpl-nancy.fr;
Zoubir.Ayadi@eeigm.inpl-nancy.fr; Michel.Nivoit@eeigm.inpl-nancy.fr

2
MISIS 4 avenue Leninskij, 119991Moscow, Russia
URL: http://www.misis.ru/ e-mail: nikulin@misis.ru;

( ) ( ) ( )
r
r
+
+ +
=
1
2
3 2
2
3 1
2
2 1

(1)
where
4
2
90 45 0
r r r
r
+ +
= (2)
where r
0
, r
45
, r
90
are coefficients of anisotropy in
different directions (0, 45 and 90 to the rolling
direction)
- Lemaitre-Chaboche mixed hardening law
( ) 0
0 , ,
= = R X f
j i j i
(3)
where X
i,j
is a tensor of displacement of the centre of
yield surface and R is a parameter of isotropic
hardening.
The corresponding evolution laws are
( ) ( )
r sat y
C R R R + = exp 1 (4)
d X Cd dX
j i j i j i , , ,
= (5)
where R
y
, R
sat
, C
r
, C and are material constants.
- The flow law is

d
d
df
d
j i
j i
,
,
= (6)
where d is the effective strain increment.
2.2 Description of each zone
2.2.a. Unsupported arc
Stress and strain states in each section of the metal
sheet can be described by two parameters: z
0
is the
position of the neutral fibre with respect to middle
fibre and k is the curvature of the middle fibre (fig.
1), so
1
=
1
(z,z
0
,k) (a detailed description for this is
given in [6]).

Fig. 1. Unsupported arc with a selected element
Therefore there is a direct relation between the pairs
N-M and z
0
-k, given by a system of equations:
( )

=
2 /
2 /
0 1
, ,
h
h
dz k z z N (7)
( )

=
2 /
2 /
0 1
, ,
h
h
zdz k z z M (8)
Knowing the internal forces (N,M) allows us to
determine the parameters z
0
and k and to find later
the internal forces in the next section (at a small
distance ds) by means of equilibrium laws:
0 = +
t
p kT
ds
dN
(9)
0 = + +
n
p
ds
dT
Nk (10)
0 = + +
z
m T
ds
dM
(11)
where p
n
and p
t
are the superficial densities of
cutting and normal external forces, m
z
is the
superficial density of external bending moment. In
the case of the unsupported arc the equations (9)-
(11) are greatly simplified by the fact that all
external densities of forces and moment are nil
(p
t
=p
n
=m
z
=0). So, for the knowing forces at the
beginning of unsupported arc the stress and strain
fields can be found in each following point in a such
way. Unsupported arc is considered as an element
with no degrees of freedom.

2.2.b. Punctual contact
At the point of punctual contact only cutting force T
c

is acting on the sheet what also creates a normal
force of contact N
c
owing to the friction between
sheet and die (Coulomb friction law is used, N
c
=T
c
,
where is a friction coefficient).

Fig. 2. Punctual contact
So the internal forces just after the punctual contact
are
c in
T T T + = (12)
c in
T N N = (13)
in
M M = (14)
So, one degree of freedom representing by T
c
is
adjusted to the sheet.

2.2.c. Distributed contact
A distributed contact is simulated by an application
of 2 punctual cutting forces T
c
and T
c
and a
M
in

F
in

T
in

N
in

M=M
in

N
c

N
F
F
c

T
c

T
1/k
c

in

z
0

F
in

F
F+dF
T
in

T
T+dT
N
in

N
N+dN
M
in

M
M+dM
1/k
ds
T
f

N
f

F
f

M
f

superficial density of cutting force p
t
. All these
forces create normal forces (N
c
and N
c
) and density
of normal force (p
n
).
Value of T
c
can be found using the following
reflections. Just after an application of the first
cutting force T
c
the internal forces of the sheet are
defined by equations (12)-(14) (T
c
is used instead of
T
c
). The curvature of sheet is now equal to the
curvature of cylindrical contact. So, the value of z
0

can be simply found from the equation (8) with the
known values of M and k. After that, using equations
(7) and (13) the first cutting force of a distributed
contact can be found by:
( )

2 /
2 /
0 1
, ,
1
'
h
h
in c
dz k z z N T

(15)
Density of normal force can be found by use of the
rope formulae:

+
=
c
c
c
c in n
h k
k
k N p
2 / / 1
/ 1
exp (16)
where
c
denotes the length of the distributed contact
and k
c
is the curvature of the contact surface.

Fig. 3. Distributed contact
So the internal forces at the end of distributed
contact will be the function only on two parameters

c
and T
c
. So, two degrees of freedom (
c
and T
c
)
are adjusted to metal sheet.

2.3 Description of a complete algorithm
Each drawing begins with a punctual contact
(unsupported arcs before drawing are straight lines,
we dont consider the effect of prestraining and the
effect of gravity) and a distributed contact is not
allowed at the beginning as the moment that bends
the sheet along the die cant be created by one
contact.
There is only one independent parameter that
describes the position of the punctual contact: if this
is a contact with a plane surface then this is one of
the coordinates (for example x
in
for planes except for
planes parallel to an y-axis) all others parameters
can be found from the equation of plane surface
(y
in
=x
in
tan
in
+y
0
, where y
0
is the known constant); if
this is a contact with a cylindrical surface then the
simplest parameter that can determine the position of
the contact is a contact angle (
in
) all others
parameters can be found from the equation of
cylindrical surface (x
in
=R
0
cos
in
+x
0
, y
in
=R
0
sin
in
+y
0,
where x
0
, y
0
, R
0
are the known constants defining the
cylindrical surface).
Fig. 4. Step-bead configuration
There are no forces before the drawing so
T
in
=N
in
=M
in
=0. As a result the configuration of sheet
after this first punctual contact depends on two
parameters x
in
(it can be also y
in
or
in
) and T
c
, so
there are two degrees of freedom q
0
and q
p
.
A punctual contact of the sheet is followed normally
by an unsupported arc. The position of the beginning
of the arc (x
a
in
, y
a
in
,
a
in
) is defined by the preceding
punctual contact: x
a
in
=x
p
f
, y
a
in
=y
p
f
,
a
in
=
p
f
.
Moreover, the forces at the issue of a punctual
contact (T
p
f
, N
p
f
, M
p
f
) determined completely the
forces at the beginning of the unsupported arc:
T
a
in
=T
p
f
, N
a
in
=N
p
f
, M
a
in
=M
p
f
. As a result the position
and the internal forces along the unsupported arc are
completely defined by the position and the forces at
the beginning (in our case this is fixed by the
preceding punctual contact). At the end the
unsupported arc must be in contact with a die that
delete one degree of freedom.
Consider now a distributed contact. The position of
the beginning of the contact (x
d
in
, y
d
in
,
d
in
) is given
by the end of the unsupported arc: x
d
in
=x
a
f
, y
d
in
=y
a
f
,

d
in
=
a
f
, as well as the forces: T
d
in
=T
a
f
, N
d
in
=N
a
f
,
M
d
in
=M
a
f
. As a result the issue of distributed contact
is completely defined by two parameters T
c
,
c
. So
there are another two degrees of freedom that are
adjusted to sheet by a distributed contact (q
d
=2).
2.4 Example of application (step bead)
For the configuration of step-bead that is presented
in the fig.4 there are 2 degrees of freedom:
F
in

T
in

N
in

M
in

M
N
c

N
F
F
c

T
c

T 1/k
c

in

c

F
c

T
c

N
c

p
n

p
t

xin=0.019
c=5

2
in

3
in
( ) ( ) 2 1 1
0
= + + =
d d p p
n q n q q q (17)
where q
0
=1 represents the uncertainty of the position
of initial contact, q
p
=1 is the number of degree of
freedom adjusted by a punctual contact (T
c
), n
p
is the
number of punctual contacts (in our case n
p
=1), q
d
=2
is the number of degree of freedom adjusted by a
distributed contact (T
c
,
c
) and n
p
is the number of
distributed contacts (n
d
=1 in our case). The position
of the first contact (x
in
) and the angle of the
distributed contact (
c
) were chosen as two degrees
of freedom. There are some geometrical restrictions
which are imposed on these parameters. They are
presented on the plane (x
in
-
c
) as the area of possible
values of these parameters (fig.5).

Fig. 5. Mechanical energy as a function of x
in
and
c

Different level of accumulated mechanical energy
corresponds to each couple of these parameters from
the described above area. This level can be obtained
by the following formulae:
( ) ( ) ds dz d s z b W
s h
h

=
0
2 /
2 /
, (18)
where s is the total curvilinear length of sheet in
tools, b is the width of sheet.
The level of mechanical energy for step-bead from
the fig. 4 and for material parameters from the table
1 was calculated on the area of possible values of x
in

and
c
. These results are presented on the fig. 5.

Table1. Material parameters for step-bead drawing
Parameters values
r
0
0.72
r
45
1.38
r
90
1.03
R
y
265.4 MPa
R
sat
334.8 MPa
C
r
7.57
C 89.1 MPa
12.3

There is a global minimum of mechanical energy
which corresponds to the most probable profile of
the metal sheet. This configuration is presented on
the fig. 4.
3 CONCLUSIONS
A semi-analytical model for stationary strip-drawing
process is proposed. Strain and stress fields in each
section of the strip are completely defined by the
curvature of middle fibre and the position of neutral
fibre. However the mechanical state in each next
section is defined from equilibrium laws. The final
profile of the sheet is governed by degrees of
freedom. The optimal values for parameters which
represent degrees of freedom are found from the
principle of minimisation of mechanical energy.
REFERENCES
1. Z. Ayadi, R. Nazarov, M. Nivoit, I. Inkin, J.-P.
Bettembourg, Analytical Model and Experimental
Validation of Springback Prediction in U-shape
Stamping Test, 12
th
Int. Symposium on Plasticity and Its
Current Applications, Hallifax, (2006).
2. D. Fei, P. Hodgson, Experimental and Numerical Studies
of Springback in Air V-bending Process for Colled Rolled
TRIP steels. Nuclear Engineering and Design 236 (2006)
1847-51.
3. H.D. Nine, New Draw Bead Concepts for Sheet Metal
Forming. J. Applied Metalworking 2 (1982) 185-192.
4. I. Inkin, M. Nivoit, Z. Ayadi, J.-P. Bettembourg, A.
Hildenbrand, L.M. Kaputkina, Application of Analytical
Drawbead Model for Stable and Metastable Steels,
Proceedings of the Int. Conf ESAFORM 7, Trondheim,
(2004) 191-194.
5. L.R. Sanchez, K.J. Weinmann, An Analytical and
Experimental Study of the Flow of Sheet Metal Between
Circular Drawbeads. ASME J. of Engineering for Industry
118 (1996) 45-54.
6. I. Inkin, Etude Exprimentale et Modlisation des
Processus de Dformation des Tles en Acier Stable et
Mtastable, PhD thesis, Nancy, France (2004).

1 INTRODUCTION
Sheet metal bending is one of the most widely-used
methods in sheet forming operations to produce
frames, channels, and other non-symmetrical sheet
metal parts [1]. After bending, some elastic spring
back occurs which causes undesirable effects in final
product. For solving this problem, as in stretch
forming, tension is applied simultaneously with
bending. Therefore sheet bending without or under
tension have been the subject of many researches
which mostly focused on simple cases of material
models.
One of the first mathematical descriptions of plastic
sheet bending was published by Ludwik, a century
ago [2]. The theory of plane strain pure bending for
rigid-perfectly plastic materials was formulated by
Hill [3]. Dadras and Majlessi [4] studied bending of
rigid-work hardening materials in a cylindrical pure
bending. Duncan and Bird [5] presented a model for
sheet stretch forming. Hosford and Caddell [6]
presented a simple derivation for bending, and then
developed further by superimposed tension.
Calculation of strains and stresses when a rigid
plastic sheet is bent and stretched under plane strain
conditions was carried out by Pourbograt and Chu
[7]. Lazim [1] analyzed the draw-bending of work-
hardening materials. Marciniak et al [8] described
mechanics of sheet metal forming. They presented
also the moment-tension curve for elastic-perfectly
plastic sheet.
In this paper, a model for bending under tension of
elasto-plastic sheet materials with work-hardening
characteristic has been proposed and effect of
tension force on moment has been investigated.
2 THEORY
Figure 1 shows a unit width of sheet in which a
cylindrical bent region with radius of curvature is
flanked by flat sheet. A moment M, and a tension T
are applied at the middle surface of the sheet
Since the width of sheet is much larger than the
thickness, t, it can be assumed that plane strain
condition, i.e.
3
=0. For simplicity, cylindrical
bending is assumed and bauschinger effect, strain
rate and friction are neglected. At present, the
behavior of an element through the cross section will
ABSTRACT: Most sheet metal operations involve some kind of bending. To avoid unwanted spring back
after bending, which causes undesirable effects in final product, as in stretch forming, tension is applied
simultaneously with bending. Since the tension plays major role and strongly affects the bending moment,
analysis of sheet metal bending under tension is important. In this study, an analytical model has been
developed for sheet metal subjected to plane strain bending under tension. The model has been used to
describe the effect of tension force on the bending moment in a linear elastic-work hardening sheet material.
Then, the bending moment versus tension has been plotted. Finally influence of punch radius and material
parameter on the shape of moment-tension curve will be discussed.
Key words: Sheet Metal Bending, Bending Moment, Tension, Work-Hardening Material
Bending of Work Hardening Sheet Metals subjected to Tension
M.H. Parsa, S. Nasher Al Ahkami
School of Metallurgy and Materials Engineering, University College of Engineering, University of Tehran-
P.O.Box 111155/456, Tehran, Iran
e-mail: mhparsa@ut.ac.ir; saeid.nasher@gmail.com
be taken to model the entire specimen.
Fig. 1. Coordinate system for analysis of bending (a). A unit
length of a sheet bent along a line (b). Transverse section of
curved sheet under simultaneous moment and tension (c) [8].
2.1 Strain-stress relationship
If a sheet is bent and simultaneously subjected to
tension as showing in figure 2(a), the strain
distribution shown in figure 2(b). It can be assumed
that the tangential strain
1
is linear sum of bending
strain (
b
c =y / ) and tension strain (
T
c ), i.e.
(1)
T
y
c

c + =
1
Where y takes values between -0.5t and 0.5t.
T
c denotes strain coming from superimposed tension
force which depends on the material behavior and
value of tension force. The position of the neutral
axis, y
0
, depends on the tension force or the tensile
strain
T
c , which is expressed as equation (2) [9].
(2)
T
y c .
0
=
Fig. 2. Equilibrium diagram for a section through a unit width
of sheet (a), strain distribution through the thickness (b) and
stress distribution of elastic-work hardening sheet (c).
An Elastic-plastic sheet material with work-
hardening characteristic has been assumed. Principle
stress component
1
can be calculated using equation
(3) with taking into account plane strain condition
and elasto-plastic behavior.
(3)

> ' +
s '
=
y
n p
y
y
e
if K
if E
o o c o
o o c
o
) .(
.
1
1
1
Where E and K are the modulus of elasticity and
strength coefficient in plane strain, respectively, and
are given by equations (4) and (5).
(4)
2
1 v
= '
E
E
(5)
2
1
) 3 4 .(
+
= '
n
K K
Using equation (3), the following elastic strain and
the plastic one are obtained.
(6)
E
y
T e
'
+ =
o

c
1
(7)

>
'
+
'
+
s
'
+
=
y T
n T T
y T
T
p
if
K E
y
if
E
y
o o
o o

o o
o

c
1
1
) (
Where tensile stress resulted from superimposed
tension force is t T
T
= o .
In a sheet bending under tension, with an increasing
tensile force, T, the neutral plane shifts towards the
inside of the bend and in many operations, this
tension is sufficient to move the neutral plane
completely out of the sheet so that the entire cross
section yields in tension. The strain and stress
distribution for such a case are sketched in figure 3.
Fig. 3. Distribution of strain (a) and stress (b) in an elastic,
work-hardening sheet bent to a gentle curvature and stretched.
2.2 Bending moment
The bending moment at each cross-section can be
calculated by integrating principal stresses
1
in
equation (3) over the current thickness of each
element as follows
(8)
ydy M
t
t
)

=
2
2
1
o
Since the stress-strain relation varies with
T
o , the
bending moment M, can be divided M
1
and M
2
.
The bending moment in the
y T
o o s range is
(9)
p e
M M M
1 1 1
+ =
According to equation (8), these moments are given
by equations (10) and (11).
(10)
dy y
E
y
E M
t
m
t
T e
)

'
+ ' =
2
2
1
) (
o

and
(11)
dy y
E
y
K M
t
t
m
n T
y
p
)
'
+ ' + =
2
2
1
) ) ( (
o

o
The elastic-plastic interface is located at a distance
2 mt from the middle surface as shown in figure 3,
where 1 1 < < m . Parameter m is found by inserting
equating (6) in equation (3) at elastic-plastic
transition point with 2 mt y = .
(12)
y T
T n
E E
K
t
m o o
o
s
'

'
'
=

hen w ) ) ((
2
1
1
The bending moment in the
y T
o o > range is given
by equation (13).
(13)
p e
M M M
2 2 2
+ =
Where
e
M
2
and
p
M
2
are the elastic and plastic
portions respectively. These moments are given by
(14)
dy y
K E
y
E M
t
m
t
n T T e
)

'
+
'
+ ' =
2
2
1
2
) ) ( (
o o

and
(15)
dy y
K E
y
K M
t
t
m
n
n T T
y
p
)
'
+
'
+ ' + =
2
2
1
2
) ) ) ( ( (
o o

o
Where m parameter for this case, as obtained before,
is calculated by equating (7) to strain
p
1
c in equation
(3) at the yield point with 2 mt y = .
(16)
y T
n
T T
n
K E E
K
t
m o o
o o
>
'

'

'
'
=

when ) ) ( ) ((
2
1
1
1
For the better perception of influence of T on M, the
plot of bending moment versus applied tension has
received significant importance. Utilizing equations
(9) and (13), it becomes possible to plot the
moment-tension diagram for a specified sheet
material.
3 MATERIALS
To study the influence of material properties on
moment-tension diagram of the proposed model, two
materials; plane carbon steel, st-14, and aluminum
alloy, AA5754 has been selected. The mechanical
properties and sheet thicknesses of the two materials
are shown in table 1.
Table1. Mechanical properties and thickness of sheet materials
used in this investigation [10, 11].
Material
Youngs
modulus
E (GPa)
Yield
strength

y
(MPa)
K
(MPa)
n
Thickness
t (mm)
St-14 200 220 625 0.27 0.3 1
AA5057 71 136 577 0.359 0.34 1
4 RESULT AND DISCUSSION
Using material properties listed in the table 1, the
computed M-T curves for the steel and aluminum
alloy sheets are plotted in figures 4 and 5
respectively. The effects of curvature variations
from 0.05m to 0.002m are also shown in the
mentioned figures.
Fig.4. Moment-tension plot for St-14 steel sheet characterised
in Table 1.
Fig.5. Moment-tension plot for AA5754 sheet characterised in
Table 1.
The resulted M-T curves can be divided into three
sections. In the first section, when the moment M >0
and T =0, for a given curvature, with increasing T,
the moment will be constant until the elastic limit is
reached. Also in this section, the neutral axis
coincides with the center line.
Applying tension results in a shift in the position of
the neutral axis, y
0
, as expressed in equation (2), and
leads to enlarging the zone which is subject to
tensile strains and stresses (figure 1). By increasing
the tension above the elastic limit, bent sheet entered
into the section two of M-T curves in figures 4 and 5
where, y
0
grows, M decreases, and the portion of the
cross section deformed elastically shifts towards the
inner surface. In the other word, at section two of M-
T curves, as shown in the figures, for a given
curvature an increase in tension significantly reduces
bending moment during bending process until the
neutral plane completely go out of cross section.
At the third section of M-T curves, by increasing
tension, the moment rises up suddenly, and a
minimum is created. In this region, the neutral axis
completely exited from sheet section. The growth of
bending moment at large strains can be attributed to
increasing resistance of sheet to bending under high
stretching. In addition, the sheet will be work
hardened with increasing strains, so in the higher
stretches, larger loads required to bend sheet, i.e.
increasing tension lead to increasing moment until
ultimate tensile strain reached. Naturally with
declining in the rate of work-hardening at high
strains, the slop of M-T curves decreases.
As observed in the figures 4 and 5, the pattern of
bending moment variations with tension is greatly
affected by the bending radius and material
properties. In the M-T curves before minimum, at a
constant tension, with increasing radius of curvature
, the moment decreases. For the small curvature
radiuses, according to equation (1), strain and
associated stress is large, therefore a larger load is
required to bend sheet in comparison to the large
curvature radiuses. In addition, according to
equation (2) for small radius of curvatures, exit of
the neutral axis from sheet section requires high
tensions T. Also, with increase in radius of
curvature, the curves and minimum shift left to
smaller tensions since lower tension loads need to
exit the neutral axis from sheet section.
At the third section in M-T curves, after the
minimum point, at constant applied tension,
increasing of curvature radius increases the bending
moment. It is due to the fact that with increasing
curvature radius, the angle between sheet direction
and horizontal line reduced. Therefore, for
compensating the growing required normal force
component, the moment should be enlarged
accordingly.
5 CONCLUSIONS
In this paper a theoretical model has been presented
for calculating the Stress-strain relationships and
bending moment in the bending under tension of
work-hardening sheet metals. The effects of radius
of curvature and material parameters have been also
evaluated. It can be seen that Tension strongly
affects the bending moment. In addition, the pattern
of bending moment variations with tension is greatly
affected by the bending radius and material
properties. Obtained results also show that the
moment-tension curve can be separated to three
portions: constant moment, moment decreasing and
eventually ascending moment. In the other hand, by
increasing radius of curvature, the curves and
minimum shift left to smaller tensions. Besides, in
the ascending moment region, an increase in radius
of curvature increases bending moment.
REFERENCES
1. D. Lazim, Springback in draw-bending on aerospace
alloys, M.S. Thesis, McGill University, Canada (2003).
2. B. Heller, M. Kleiner, Semi-analytical process modeling
and simulation of air bending, J. Strain Anal. Eng. Des.,
41(1), (2006) 57-80.
3. R. Hill, The mathematical theory of plasticity, Oxford
University Press, London, UK (1950).
4. P. Dadras, S.A. Majlessi, Plastic bending of work
hardening materials, Trans. ASME (1982) 224-230.
5. J.L. Duncan, J.E. Bird, Die forming approximations for
aluminum sheet, Sheet Metal Industries (1978) 10-15.
6. W.F. Hosford, R.M. Caddell, Metal forming, Mechanics
and metallurgy, Prentice-Hall Inc., USA (1983).
7. F. Pourboghrat, E. Chu, Springback in plane strain stretch-
draw sheet forming, Int. J. Mech. Sci. 36(3), (1995) 327-
341.
8. Z. Marciniak, J.L. Duncan, S.J. Hu, Mechanics of Sheet
Metal Forming, Butterworth-Heinemann, UK (2002).
9. R.H. Wagoner, M. Li, Advances in springback, In: CP778
(A), Numisheet'05, eds, L.M. Smith, F. Pourboghrat, J.W.
Yoon, T.B. Stoughton, Detroit (2005) 209-305.
10. M.H. Parsa, P. Pournia, Optimization of initial blank
shape predicted based on inverse finite element method,
Finite Elem. Anal. Des. 43 (2007) 218233.
11. M. Samuel, Experimental and numerical prediction of
springback and side wall curl in U-bendings of anisotropic
sheet metals, J. Mater. Process. Technol. 105 (2000) 382-
393.
1 INTRODUCTION
Advanced high strength steels (AHSS) are
engineered to have both high strength and enhanced
formability characteristics. Yet, the formed
components undergo springback due to elastic
recovery after removal of the forming tools. This
severely affects the dimensional accuracy of the
part. Many techniques have been evolved to deal
with springback. A simple approach is to design
forming tools that compensate for springback [1]. As
the springback increase, its variability will also
increase that result in the die correction estimation
difficulties, especially when the part has complex
geometry [2]. An effective approach to reduce
springback is by attaining adequate strain levels in
the part. This can be achieved by utilizing draw
beads to induce large strain and reduce elastic
recovery process in the part [3]. Dual phase steels
have excellent ductility and work-hardening rate
resulting in a superior strength-ductility balance.
Higher strength levels and improved formability are
achieved in the parts due to increased values of
work-hardening exponent of the dual-phase steels.
This in turn increases problems such as springback,
higher loads on the forming tools, etc. This study
focuses on the influence of drawing restraining force
on the springback characteristics of the dual-phase
steel using draw-bend test. Draw-bend tests imitate
typical forming procedure involving tensile loading,
bending, and unbending as the specimen is drawn
over the roller [4]. As the sheet tension is increased,
through a hydraulic actuator in the draw-bend tester,
springback reduces at different rates for different
materials [5].
2 EXPERIMENTAL PROGRAM
2.1 Material
Table 1. Material Properties
Property DP600 DP800 DP980
Youngs Modulus (GPa) 200 210 210
Poisson Ratio 0.3 0.3 0.3
Yield Stress (MPa) 362 512 578
UTS (MPa) 651 807 979
Thickness (mm) 1.44 1.44 1.44
Three grades of dual phase steel, namely DP600,
ABSTRACT: Draw beads are used in sheet metal forming processes to regulate material flow and achieve
higher quality in the parts. The draw beads impose draw restraining force on the blank which eliminates
defects such as wrinkling, tearing and reduces springback. In a draw-bend test, the draw restraining force is
imitated by the back force applied through a hydraulic ram. As the back force increases, large strain is
induced which reduces springback tendency. This study evaluates the influence of back force on the
springback behavior in advanced high strength steel (AHSS). Three grades of dual-phase steel, namely
DP600, DP800, DP980, are used with normalized back forces ranging from 0.5 to 1.1. The springback
decreases as the back force increases. Anticlastic curvature as well as curl radius has significant influence on
the springback behavior of dual phase steels.
Key words: Forming, AHSS, Springback, Draw-bend test, Anticlastic curvature
Influence of draw restraining force on the springback in advanced high
strength steels
R. Padmanabhan
1
, Jihyun Sung
2
, H. Lim
2
, M.C. Oliveira
1
, L.F. Menezes
1
, R.H. Wagoner
2
1
CEMUC, Department of Mechanical Engineering
University of Coimbra, Polo II e-mail: {padmanabhan, marta.oliveira, luis.menezes}@dem.uc.pt
Coimbra 3030 788, Portugal
2
Department of Materials Science and Engineering
The Ohio State University, 177 Watts Hall
2041 College Road, Columbus
OH 43210-1179, USA e-mail: {sungj, lim, wagoner}@matsceng.ohio-state.edu
DP800, DP980, were used in this study. Some
material properties are listed in table 1.
2.2 Draw-bend test specimen
The draw-bend test specimens were cut along the
rolling direction for 710 mm length and 25.4 mm
width. In addition, 50.1 mm wide specimens made
from DP600 steel were used for comparison.
2.3 Draw-bend test procedure and Parameters
The draw-bend tester is capable of imposing longer
drawn distance and various front and back forces on
the specimen [6, 7].
Fig. 1. Schematic of the stages of the draw-bend test and
unloaded specimen geometry [8]
The specimen was mounted between the front and
back actuators and a pretension (back force) was
applied using the back actuator. The back force was
varied between 50% yield strength of the material to
110 % yield strength at regular intervals of 10%.
Immediately after the application of back force, the
specimen is drawn to a distance of 127 mm at a rate
of 25.4 mm per second. At the end of drawing, the
specimen was released from the tester and its profile
was traced on paper. The time between release from
the tester and tracing was close to 30 seconds. The
anticlastic curvature (R
a
) and curl radius (r) were
measured using a Kreon (KZ50) LASER device.
They were measured on each specimen in region 3,
at a position 76 mm from the location of the
boundary between region 1 and region 2 (fig. 1). The
R/t ratios under analysis are close to 4.6 and 7.9 for
the 6.4 and 11.1 mm roller radii, respectively.
Minimum friction (free rolling) was used in all
experiments.
3 RESULTS AND DISCUSSION
3.1 Springback Angle
Figure 2 shows the springback angle () for
different materials at different back forces, ranging
from 50% yield strength to 110% yield strength for a
roller radius of 6.4 mm. As the back forces increase,
the springback angle reduces. The rate of reduction
in springback angle increases with increasing
material strength. The springback angle at high back
forces is controlled by anticlastic curvature [9].
Higher back forces accentuate anticlastic curvature,
which in turn increases the moment of inertia of the
specimen cross section, thus reducing springback. At
low back forces (Fb), the springback angles for
increasing material strength are higher. Beyond Fb =
0.8 this trend reverses, and the springback angles for
increasing material strength become smaller.
Fig. 2. Springback angle obtained using R = 6.4 mm
To determine the influence of specimen width on
springback, specimens made of DP600 steel were
tested in two widths: 50.1 mm and 25.4 mm.
Springback angle results for these specimens are
also presented in figure 2. At low back forces, both
specimens produced same springback while back
forces beyond 80% yield strength produced
considerable deviations in springback angle. Due to
high contact forces, back forces more than 90% of
yield strength cannot be tested for wider specimens.
Figure 3 shows the springback angle for various
back forces using a roller radius of 11.1 mm. As the
back forces increase, the springback angle reduces
with marginal difference in the rate of reduction.
A
B
C
Fig. 3. Springback angle obtained using R = 11.1 mm
The springback angle obtained using 11.1 mm roller
radius is less than that obtained by using a roller
radius of 6.4 mm. Over the range of back forces
DP980 produced more springback compared to
DP800 and DP600 respectively.
3.2 Anticlastic curvature
Fig. 4. Anticlastic radius of curvature using R = 6.4 mm
Anticlastic curvature develops due to differential
lateral contraction through thickness of the sheet
under the principal bending action. It depends on the
magnitude of the back force and the roller radius. As
the back force increases in the sheet, the springback
angle can drop drastically due to the appearance of
anticlastic curvature. Figure 4 shows the increase in
anticlastic radius of curvature as the back forces
increase, using a roller radius of 6.4 mm. As the
relative back force increases beyond 0.7, significant
increase in the anticlastic curvature is observed,
especially in DP600 steel.
The trend is similar for DP600 steel using 11.1 mm
roller radius, as shown in Fig. 5. In DP800
specimens, the anticlastic radius of curvature
remained almost the same for all back forces and it
reduced marginally in DP980 specimens. In higher
grade steels, especially DP980, the anticlastic radius
of curvature reduces due to the development of a
tertiary curvature between region 2 and region 3 in
figure 1. This behavior is further explained at the
end of section 3.3.
Fig. 5. Anticlastic radius of curvature using R = 11.1 mm
3.3 Curl radius
Fig. 6. Curl radius using R = 6.4 mm
Fig. 7. Curl radius using R = 11.1 mm
The curl radius (r) at the region that has undergone
bending and unbending, over an arc length of 127
mm, is related to the springback angle. As shown in
figure 6, increasing the back force increases the curl
radius. The trend is opposite to that observed for
springback angle. The curl radius in DP980
increased sharply as the back force increases, while
the increase is moderate in other steels. Figure 7
shows the increase in curl radius using 11.1 mm
roller radius. The increase is similar in the three
grades of steel. In DP980, as the back force
increased beyond 100% yield strength, a tertiary
curvature developed between region 2 and region 3.
Higher back forces, yield strength and beyond,
causes prestrain in the specimen before the draw-
bend procedure. This increased the strength in the
specimen and consequently resulted in a complex
geometry after the draw-bend procedure. Figure 8
shows the profile of the draw-bend portion of the
specimen as two sections, AB and BC (indicated in
figure 1). As shown in Fig. 8, curvature 1 and
curvature 3 have their centers on one side, and
curvature 2 has its center on the opposite side.
Normalized back forces beyond 1.0 result in a
complex stress state in the specimen and hence an
intermediate curvature is developed.
Fig. 8. Unloaded profile of the draw-bend region
4 CONCLUSIONS
Three grades of dual-phase steel specimens were
subjected to different back forces to determine their
springback behavior using two different roller radii.
As the back force increase, the springback angle can
drop quickly with increasing material strength. This
behavior is pronounced at low R/t ratios. Draw
restraining force has significant influence on the
springback; higher forces result in less springback
and consequently produce dimensional stability in
the formed part. Anticlastic radius of curvature and
curl radius have direct implication on the springback
angle in the draw-bend specimen. A tertiary
curvature was observed in DP980 specimens at
higher back forces.
ACKNOWLEDGEMENTS
The authors acknowledge the Ohio State University for
providing material and facilities to conduct the experiments for
this study. The authors from Portuguese Institution are grateful
to the Portuguese Foundation for Science and Technology
(FCT) for the financial support through the program POCI
2010.
REFERENCES
1. Wei Gan, R.H. Wagoner, Die design method for sheet
springback, International Journal of Mechanical
Sciences 46 (2004) 1097-1113.
2. M.C. Oliveira, J.L. Alves, B.M. Chaparro, L.F. Menezes,
Study on the influence of work-hardening modeling in
springback prediction, International Journal of Plasticity
23 (2007) 516543.
3. J-H. Song, H. Huh, S-H. Kim, A simulation-based
design parameter study in the stamping process of an
automotive member, Journal of Materials Processing
Technology 189 (2007) 450-458.
4. W.D. Carden, L.M. Geng, D.K. Matlock, R.H. Wagoner,
Measurement of Springback, International Journal of
Mechanical Sciences 44 (2002) 79-101.
5. K.P. Li, W.P. Carden, R.H. Wagoner, Simulation of
Springback, International Journal of Mechanical
Sciences 44 (2002) 103-122.
6. G.J. Wenzloff, T.A. Hylton, D.K. Matlock, A new
procedure for the bending under tension friction test.
Journal of Material Engineering and Performance, 1(5)
(1992) 609613.
7. D.W. Vallance, D.K. Matlock, Application of the
bending-under-tension friction test to coated sheet steels,
Journal of Material Engineering and Performance, 1(5)
( 1992) 685693.
8. J.F. Wang, R.H. Wagoner, D.K. Matlock, F. Barlat,
Anticlastic curvature in draw-bend springback,
International Journal of Solids and Structures 42 (2005)
1287-1307.
9. L. Geng, R.H. Wagoner, Role of plastic anisotropy and
its evolution on springback, International Journal of
Mechanical Sciences 44 (2002) 123-148.
1 INTRODUCTION
In the last decades important efforts were made to
increase the flexibility in metal forming processes,
including the stage of product development. Some
flexibility systems for plastic deformation
technology can be identified today. Among them,
multipoint deformation using reconfigurable dies is
an advanced manufacturing technique for obtaining
three-dimensional sheet-metal parts (Fig. 1). The
concept of multipoint deformation using
reconfigurable dies means that the working surface
of the die is made up of individual active punches,
called pins. In general each pin has a square cross
section, a hemispherical end and can be
independently and moved in vertical direction.
At international level a series of reserches in this
field could be remarked. Thus, Walczyk and Hardt
[1, 2] developed a series of dies for multipoint
deformation. The pins could be positioned both
equally distanced and densely packed. To obtain a
surface, the heights of pins is made by numerical
controll. Papazian [3] developed a system for strech
forming. To avoid the dimpling phenomenon a
rubber sheet has been inserted between the pins and
the blank. The method is used in the aeronautical
industry.
Fig. 1. Multipoint forming die
Cai and Li [4, 5] studied the theoretical aspects of
the process, made the numerical simulation of the
process using a special sofware, and also, studied the
influence of the material restraint during the process
of multipoint deformation. They developed two new
methods of multipoint deformation, the secvential
deformation of large parts and the deformation along
ABSTRACT: Forming with reconfigurable multipoint dies is a new flexible manufacturing forming
technology which uses discrete pins to materialize a continuous 3-D surface for deformation. The desired
tooling surface shape is obtained by adjusting the pins heights and could have a fixed or an active
configuration. In this paper, finite element simulation of the sheet metal forming was performed to investigate
the influence of the pins network type on the deformation process in reconfigurable multipoint forming
(RMF) with fixed configuration. The deformation process is evaluated in terms of thickness, stresses, forces
and springback variations. The results demonstrate that the final decision in choosing a network type depends
mainly on the design reasons.
Key words: Reconfigurable die, Multipoint forming, Sheet metal forming
Numerical Simulations in Reconfigurable Multipoint Forming
V. Paunoiu
1
, P. Cekan
2
, E. Gavan
3
, D. Nicoara
1
1
University of Galati, Department of Manufacturing, 800008 Galati, Romania
URL: www.ugal.ro e-mail: viorel.paunoiu@ugal.ro;
2
PC Engineering Solutions s.r.o., Centrum 5/9 (OC Manin), 01701 Povazska Bystrica, Slovakia
URL: www.pcsolutions.sk e-mail: cekan@pcsolutions.sk;
3
University of Galati, Department of Shipbuilding, 800008 Galati, Romania
URL: www.ugal.ro e-mail: eugen.gavan@ugal.ro
a forming path. The ideea of finding an optimum
forming path in RMF was also developed in Europe
by Boers [6]. The Plastic Deformation Team from
the Department of Manufacturing Technologies -
Dunarea de Jos University of Galati, Paunoiu [7-9],
started the researches in this field from 2004. The
team performed the numerical simulation of the
process, springback compensation and designed an
equipment, manually operated for this technology.
In this paper numerical results regarding the
deformation with reconfigurable multipoint forming
die are presented.
2 GENERAL CLASSIFICATION
Depending on the type of application a classification
of reconfigurable multipoint die forming is
presented in table 1.
Table1. Process classification and characteristics
Type of
process
Type of
configuration
Main
characteristics
Fixed The geometry configuration
is fixed from the begining to
the end of the process
For stretch
forming
Active The geometry configuration
is changeable in process
according to a path-forming
Fixed As for stretch forming
Active As for stretch forming
For bending
with or
without blank-
holder
Sectional The forming is made step
by step, the deformation
region, effective region,
transition region and
underformed region runs
up, respectively, to the final
part geometry
Fixed As for stretch forming
Active As for stretch forming
For deep-
drawing
with or
without blank-
holder
Sectional As for bending
3 SIMULATION MODELS
Today, the characteristics of a product or a process
are analyzed using advanced computational methods
such as FEM. One of the powerful computer
simulation programs based on FEM is Dynaform-
PC.
In order to simulate the RMF process using
Dynaform-PC, the final part geometry to be obtained
is a simple curved plate with an interior radius of 95
mm, a width of 120 mm (maximum depth is 21.345
mm) and a length of 130 mm.
In the simulation two pins networks which
materialize the punch and the die were used. The
first one is a tool with networks of circular pins, with
a radius of each pin end of 5 mm (Fig. 2). The
second one is a tool with networks of square pins,
with a radius of each pin end of 7.07 mm according
to the 10 mm square section (Fig. 3).
Fig. 2. FEM model with circular pins
The FE mesh consists of 4-node Belytschko-Tsay
shell elements, with five integration points through
the thickness of the sheet [10]. The Belytschko-Lin-
Tsay shell element are based on a combined co-
rotational and velocity-strain formulation. The co-
rotational portion of the formulation avoids the
complexities of nonlinear mechanics by embedding
a coordinate system in the element. The choice of
velocity strain, or rate of deformation, in the
formulation facilitates the constitutive evaluation,
since the conjugate stress is the more familiar
Cauchy stress. [10]
Fig. 3. FEM model with square pins
The material used in experiments was mild steel,
with 1 mm thickness. The yielding of the material
was modeled using a power law:
n
K = (1)
in which: K is the material characteristic; n
hardening exponent. In simulation the n-value = 0.22
and K = 648 MPa. The R-values were set to: R
00

1.87; R
45
1.27; R
90
2.17. The Coulomb friction
law was used with a friction coefficient of 0.125.
The punch speed was 100 mm/second.
The blank is a rectangular plate with the dimensions
of 130x130x1 mm and the mesh consists of 900
finite elements.
The tooling was modelled as rigid surfaces. In both
cases the geometrical model of die-punch tool was
composed from two working networks with 100 pins
for each network, 10 rows on x-direction and 10
rows on y-direction. The pins are disposed face to
face, both on x-direction and y-direction. The mesh
consists of 103488 numbers of finite elements in the
first deformation case and 258804 in the second
case. No rubber interpolator was used and no
blankholder.
4 NUMERICAL RESULTS AND DISCUSSIONS
In all cases the measurements were made along the
curved profile in the middle of the surface part,
starting from the right margin to the left margin, in
the nodes of the FE deformed mesh, using arbitrary
units (a.u.).
Figure 4 presents the thickness variation of the
simple curved deformed part. As one can see, the
thickness variations are very small in both cases.
From qualitative point of view, in both cases of
deformation, thickness is more reduced in some
regions of contacts between the pins and the blank.
This phenomenon is called dimpling and affects the
part quality.
999,75
999,8
999,85
999,9
999,95
1000
1000,05
0 5 10 15 20 25 30 35
Distance, [a.u.]
T
h
i
c
k
n
e
s
s
*
1
0
0
0
,

[
m
m
]
Circular Pins
Square Pins
Fig. 4. Thickness variation for circular and square pins
network in the middle of the part
The values of stresses are presented in figures 5-7.
The simulations show small values of stresses, and
prove the character of plane stresses during this
RMF process.
-15
-10
-5
0
5
10
15
20
25
30
0 5 10 15 20 25 30 35
Distance, [a.u.]
S
i
g
m
a

X
,

[
M
P
a
]
Circular Pins
Square Pins
Fig. 5. Sigma X variation for circular and square pins
network in the middle of the part
-10
0
10
20
30
40
50
0 5 10 15 20 25 30 35
Distance, [a.u.]
S
i
g
m
a

Y
,

[
M
P
a
]
Circular Pins
Square Pins
Fig. 6. Sigma Y variation for circular and square pins
network in the middle of the part
-20
-15
-10
-5
0
5
10
15
20
25
30
0 5 10 15 20 25 30 35
Distance, [a.u.]
S
i
g
m
a

Z
,

[
M
P
a
]
Circular Pins
Square pins
Fig. 7. Sigma Z variation for circular and square pins
network in the middle of the part
The values of Von-Mises stresses showed that the
material was entirely plastically deformed. The
stresses state is not uniform along the part surface.
However, the stresses were more uniform in the
second case of deformation. In the first case the
maximum stress was 207 MPa while in the second
case was 212.5 MPa.
The maximum deformation force, according to
simulation program, is given by [11]:
[ ]
[ ]
8896
N
t
F
F = (2)
where F is the maximum force which results from
the simulation.
In the first case using circular pins, the maximum
force obtained was 0.13 tf. In the second case using
square pins the force was 0.16 tf.
In the deformation case with circular pins the
maximum value of springback was 2.907 mm. In the
deformation case with square pins the maximum
value of springback was 3.098 mm. and the
distributions of springback is presented in figure 8.
Fig. 8. The distributions of springback in deformation
with square pins network
5 CONCLUSIONS
The numerical simulation shows that using the RMF
process it possible to obtain a sound part. It is
important to know the correct positions of pins
centers in respect with the part geometry. After the
deformation, even the small thickness variation
shows the presence of dimpling phenomenon. This
phenomenon is more present in the case of circular
pins. In the second case, because the pins radius is
higher the dimpling phenomenon is less present. We
can conclude that the process of deformation using
circular pins is almost identical with the one using
square pins since the level of stresses, the level of
the total force and the level of springback are
approximately the same. Thus the final decision in
choosing one of the two types of network depends
mainly on the design reasons.
AKNOWLEDGEMENTS
The work reported here was an activity in the frame of
Romanian Grants: CEEX-P-CD no. 130/2006 Materials and
technologies for plane and spatial profiling and CNCSIS Grant
A686/2007 A new theory of nanostructuring as a recursive
chaotic process and designing of a new method of
nanostructuring by controlled multidirectional deformation.
The authors appreciate the continuing encouragement and
support of this program.
REFERENCES
1. Walczyk, D.F. and Hardt, D.E., A Comparison of Rapid
Fabrication Methods for Sheet Metal Forming Dies,
ASME Journal of Manufacturing Science and
Engineering, 1999, 121(1), 214-224.
2. Walczyk, D.F. and Hardt, D.E., Design and Analysis of
Reconfigurable Discrete Dies for Sheet Metal Forming,
Journal of Manufacturing Systems, 1998, 17(6), 436-
454.
3. Nardiello, J., Christ, R., and Papazian, J.M., Block Set
Form Die Assembly, USA Patent 6,053,026, April 2000.
4. Cai, Z.Y., Li, M.Z., Multi-point forming of three-
dimensional sheet metal and the control of the forming
process, Int. J. Pressure Vessels Piping 79 4 (2002),
289296.
5. Cai, Z., Li, M., Optimum path forming technique for
sheet metal and its realization in multi-point forming,
Journal of Materials Processing Technology, Vol. 110,
136-141, 2001.
6. Boers, S.H.A., Schreurs, P.J.G., Geers, M.G.D., Path-
Dependent Plasticity and 3D Discrete Forming, VIII
International Conference on Computational Plasticity,
COMPLAS VIII, CIMNE, Barcelona, 2005.
7. Paunoiu, V., Nicoara, D., Design an experimental
reconfigurable die for sheet metal forming, The Annals
of Dunrea de Jos University of Galai, fasc. V, 2006,
81-85
8. Paunoiu, V., Oancea, N., Nicoara, D., Simulation of
Plates Deformation Using Discrete Surfaces, Materials
Processing and Design: Simulation and Application,
NUMIFORM, 2004, OHIO State University, American
Institute of Physics 0-7354-0188-8, 1007-1010.
9. Gavan, E., Paunoiu, V., Dimache, A., Comparative
Study for Single-Curved Plates Forming with
Continuous and Reconfigurable Die-Punch Assembly,
The Annals of Dunrea de Jos University of Galai, fasc.
V, 2004, 81-85.
10. Belytschko, T., Liu, W.K., Moran, B., Finite Elements
for Nonlinear Continua and Structures, Wiley, New
York (1996).
11. eta/DYNAFORM, Application manual, 2006
1 INTRODUCTION
The global competition requires that manufacturing
industry besides the skill and the experience
accumulated in the shop practice should
increasingly utilise proven techniques of Computer
Aided Engineering including numerical simulation
and modelling for rapid and cost effective process
design and tool manufacturing. The application of
various methods of Computer Aided Engineering
has become one of the most important topics in
manufacturing industries and particularly in the
automotive industry [1].
The application of various CAE techniques
practically should cover the full product develop-
ment cycle from the conceptual product design
through the process planning and die design up to
the manufacturing phase of the production. CAE
techniques are widely used in sheet metal forming,
for example to predict the formability, to determine
the type and sequences of manufacturing processes
and their parameters, to design forming tools, etc.
[2]. The importance of the application of CAE tools
becoming more and more important as the
manufactured parts are becoming ever increasingly
complex since the sophisticated Computer Aided
Design packages provide practically unlimited
possibilities for stylists to realise their creative
fantasy. The need for the widespread application of
CAE techniques is driven by the demand of global
competitiveness, thus a robust and streamlined
Process and Die Design Engineering (PDDE)
becomes more and more crucial [3].
In this paper, the integration of various CAE
techniques as Knowledge and Simulation Based
Systems (KSBS) will be described through the
example of sheet metal forming practice. The
forming simulation in sheet metal forming
technology [4] and its industrial applications have
greatly impacted the automotive sheet metal product
design, die developments, die construction and
tryout, and production stamping in the past decades.
In todays die and stamping industry, the simulation
for virtual validations of die developments before
production trials is a critical business for lead-time
reduction, cost reduction and quality improvements
[5]. The global competitions drive higher quality
requirements, lower cost, and shorter lead-time. All
these new trends create new challenges for stamping
simulation and to production applications.
ABSTRACT: During the recent 10-15 years, Computer Aided Process Planning and Die Design evolved as
one of the most important engineering tools in sheet metal forming, particularly in the automotive industry.
This emerging role is strongly emphasized by the rapid development of Finite Element Modelling, as well. In
this paper, an integrated process simulation and die design system developed at the University of Miskolc,
Department of Mechanical Engineering will be introduced. The proposed integrated solution has great
practical importance to improve the global competitiveness of sheet metal forming in the very important
segment of industry. The concept to be described in this paper may have specific value both for process
planning and die design engineers.
Key words: Process planning, Die-design, Integrated FEM simulations
Integrated Process Simulation and Die-Design in Sheet Metal Forming
M. Tisza
1
, Zs. Lukcs
2
, G. Gl
3

1
University of Miskolc H-3515 Miskolc-Egyetemvros, Hungary
URL: www.met.uni-miskolc.hu e-mail: tisza.miklos@uni-miskolc.hu

2
MTA ME Technological Research Group Miskolc H-3515 Miskolc-Egyetemvros, Hungary
URL: www.met.uni-miskolc.hu e-mail: lzsolt@kugli.met.uni-miskolc.hu

3
UM DME Metal Forming Division H-3515 Miskolc-Egyetemvros, Hungary
URL: www.met.uni-miskolc.hu e-mail: metgalga@ uni-miskolc.hu
2 SHORT HISTORICAL REVIEW
Sheet metal forming is one of the most widely
applied manufacturing processes in manufacturing
industry. Parts made from sheet metal can provide,
with appropriate design, a high strength to weight
ratio. They are increasingly used from small
electrical components through the automobile
industry up to large aircraft structures for various
purposes. Despite the increasing number of
applications of sheet metal parts, surprisingly little
quantitative design information is available in the
technical literature. Most companies use internal
guidelines for part design, based on experience with
the geometries and materials used in that specific
company [6]. While such design guidelines are
extremely useful and practical, they do not
necessarily consider in detail the fundamental
reasons for selecting a given design. Thus, when a
new part, a new material, or a new process is
introduced the entire set of experience-based design
guidelines must be re-evaluated and modified.
Therefore, it is necessary to develop generic design
methods based on metal forming analysis and on
systematic experimental investigation. This tendency
can be clearly observed in the development of
various knowledge-based systems for designing
sheet metal parts and for process planning of
forming processes [7]. As in many other metal
forming applications, process planning and design of
dies for sheet forming can benefit from a combined
application of knowledge based systems and process
modelling. Recently, many companies are applying
CAD/CAM techniques and knowledge-based expert
systems to improve and partially automate die
design and manufacturing function [8].
Several program packages were elaborated for metal
forming processes at the University of Miskolc at
the Department of Mechanical Engineering. Among
them, a general system for the process planning of
sheet forming processes performed in progressive
dies should be mentioned [9]. In this system, the
process planning and the die-design functions are
integrated into a knowledge-based expert system. It
has a modular structure with well defined tasks of
each module and providing streamlined data and
information flow between the various modules.
It consists of a geometric module for creating,
exporting and importing the object geometry, a
blank module for determining the optimum shape,
size, and nesting of blanks, a technological design
module for designing the process sequence based on
empirical rules and technological parameters, a tool
design module for designing the tools and selecting a
tool of standard size, and an NC/CNC post
processor module for preparing programs for
NC/CNC manufacturing of tool elements.
The forming simulation in sheet metal forming
technology and its industrial applications have
greatly impacted the automotive sheet metal product
design, die developments, die construction and
tryout, and production stamping in the past decades
[10]. It led to significant progresses not only in
fundamental understanding of sheet metal formabili-
ty, forming mechanics, numerical methods, but also
to the fruitful industrial applications in a wide range
of industrial production [11].
The automotive die and stamping industry benefit
most from the stamping simulations. The technology
advancement speeds up the historical transition in
automotive die development and stamping from a
tryout-based workshop practice to a science-based,
technology-driven engineering solution. The
applications and benefits may be summarized as
follows [12]:
Stamping simulation is used as a Design for
Manufacturability (DFM) tool to assess and
validate the product styling surface designs to
ensure a formable sheet product design;
It may be used as a die engineering tool in
stamping die developments.
It may be used as a tryout tool to shorten
production die tryout and thus to significantly
reduce die cost and lead-time.
It may be used as a problem solving tool for
production troubleshooting to reproduce
manufacturing problems, and to provide
solutions for process control improvements.
It may be used as a simulation-based
manufacturing guide to use the simulation output
to drive consistency among die engineering, die
construction, and production stamping.
3 INTEGRATED PROCESS PLANNING AND
DIE-DESIGN IN SHEET METAL FORMING
Due to the global competition and this is
particularly valid for the automotive industry there
is an overall demand to improve the efficiency in
both the process planning and in the die design
phase, as well as to reduce the time and product
development costs and to shorten the lead times. It
requires the efficient use of simulation techniques
from the earliest stage of product development, to
give feedback from each step to make the necessary
corrections and improvement when it takes the least
cost. This principle is illustrated in the schematic
flow chart of simulation based process planning and
die design as shown in Fig. 1.
With this approach, stamping defects may be
minimized and even eliminated before the real die
construction stage. If any correction or redesign is
needed, it can be done immediately, with a very
short feedback time, thus it leads to a much
smoother die try-out if necessary at all and to
significantly shorter lead times with less
development costs.
However, even with this approach, there are some
further shortfalls in the die design process, since
most of the simulation programs do not provide die
construction in sufficient details, which can be easily
used in most of the CAD systems to complete the
die design task. This shortage may be overcome by
integrating the CAD and FEM systems through a
special interface module, which can provide a
smooth, continuous and reliable data exchange
between the two important parts of design process.
This solution will be described through the example
of automotive sheet metal components using the
Unigraphics NX (version 4.0) as the CAD system,
and the AutoForm 4.05 as the FEM package,
however, the principles applied here, can be adopted
to other programs as well [13].
The CAD model of the component is created by the
product design engineer in the UG-NX CAD system
as a solid model. However, FEM systems dedicated
for sheet metal forming usually require surface
models. Therefore, before exporting the part model a
surface model should be created. This function is
well-supported in most CAD systems. Depending on
the simulation requirements, even we can decide
which surface (top, middle or bottom) will be
exported into the surface model.
In most cases, process planning engineers would like
to know right at the beginning whether the component
can be manufactured with the planned formability
operations. Therefore, after importing the surface
model of the component with the AutoForm input
generator, first a fast feasibility study should be done.
The AutoForm has an extremely well suited module
for this purpose: in the so-called One-Step simulation
module, this formability analysis can be done even if
we do not have any or just very few information on the
forming tools. Using this One-Step simulation
procedure, a quick decision can be made if any
modification of the part is required.
Besides the formability validation in this very early
stage of product development, further important possi-
bilities are also offered in this module including the
analysis of slight part modifications, studying alterna-
tive material types and grade, or various thicknesses,
material cost estimation and optimization, etc.
Even if the One-Step simulation resulted in good
formability, the final decision on the whole process
realization can be made only after performing a
detailed incremental modeling particularly concerning
the critical forming steps. For this detailed simulation
we need already very detailed knowledge on the tools
and process parameters. Formerly, most die surfaces
were created within general purpose CAD systems,
which is a time consuming procedure and by this way
it is nearly impossible to integrate geometry
modifications into an automated simulation and design
optimization procedure. However, linking the
AutoForm with the Unigraphics CAD system provides
an associative linking. This integration not only greatly
improves the speed and efficiency of process planning
and die design but also significantly improves the data
consistency. With this integration, the active surfaces
of the forming tools can be derived from the imported
surface model of the component utilizing the many
useful possibilities offered by the AF Die-Designer
module to create the binder and addendum surfaces, as
well as the so-called reference surface, which can be
Product
design
OneStep
Simulation
Feasibility
study
Process
planning &
Tool
design
Incremental
Forming
Simulation
Die
manufacturing
Good Good
Prototyping
& Tool
try-out
Good
Production
Product
redesign
Failed
Process
or die
redesign
Failed
Die
correction
Failed

Fig. 1. Workflow in simulation based process planning and die design
used to quickly derive the punch and die surfaces.
These surfaces are created using a parameterization
methodology, which provides an order of magnitude
faster die face generation compared to conventional
CAD systems. The applied parameterization makes
also possible to combine surface generation with
optimization algorithms to determine optimum die
surfaces for various process parameters, too. The main
objective of this optimization is to improve the quality
of the product and the reliability (robustness) of the
forming process. In this optimization, the most
important criteria are as follows: 1) crack criteria (i.e.
no cracks should occur); 2) thinning criteria (i.e.
thinning should not exceed a definite value); 3)
wrinkling criteria (i.e. no wrinkles can occur); 4)
stretching criteria (i.e. a minimum overall stretching
should be achieved). These criteria are mainly
evaluated on the basis of the forming limit diagram of
the sheet material. Defining the set of design variables,
the effect of them on the forming process and on the
optimum die surfaces can be studied in an integrated
simulation environment. By this integrated solution,
the influence and sensitivity of process and design
parameters can be easily analysed that leads to
improved process know-how, robust and more
optimum forming processes and dies, together with
shorter development times, better product quality, and
less rejects.
4 CONCLUSIONS
The application of various methods of computer
aided engineering has a vital and central role in the
recent developments in sheet metal forming
concerning the whole product development cycle.
The application of various methods and techniques
of CAE activities resulted in significant
developments: the formerly trial-and-error based
workshop practice has been continuously
transformed into a science-based and technology
driven engineering solution.
In this paper, an integrated approach for the
application of knowledge based systems and finite
element simulation is introduced. Applying this
knowledge and simulation based concept for the
whole product development cycle from the
conceptual design through the process planning and
die design as an integrated CAE tool provides
nificant advantages both in the design and in the
manufacturing phase. Sheet metal forming
simulation results today are already reliable and
accurate enough that even tryout tools and the time
consuming tryout processes may be eliminated or at
least significantly reduced. Thus, the integrated
solution described in this paper results in
significantly shorter lead times, better product
quality and as a consequence more cost-effective
design and production.
ACKNOWLEDGEMENTS
This research work was financially supported by the
Hungarian Academy of Sciences within the National
Science Foundation Grant (OTKA NI-61724), which is
gratefully acknowledged.
REFERENCES
1. C. T. Wang: Evolutions of Advanced Stamping CAE,
NumiSheet2005 Conf., 19-24. September 2005. pp. 78-82.
2. S.K. Sitaraman, T. Altan: A Knowledge Based System for
Process Sequence Design in Sheet Metal Forming, J . of Mat.
Proc. Techn. (1991) pp. 247-271.
3. M. Tisza: Numerical Modelling and Knowledge Based
Systems in Metal Forming, ICTP Conf., Nuremberg, 19-24.
Sept. 1999. in Advanced Technology of Plasticity, 1999. v. 1.
pp. 145-154.
4. A. E. Tekkaya: State of the art of Simulation in Sheet Metal
Forming, J . Mat. Proc. Techn., v.103. (2000) pp. 14-22.
5. C. T. Wang: Advanced Stamping Simulation Technology
State of Business and Industrial Prospect, Numisheet Conf.,
Besancon, 13-17. Sept. 1999. pp. 250-256.
6. T. Altan, V. Vasquez: New Concepts in Die Design Physical
and Computer Modelling, J . of Mat. Proc. Techn., v. 98. (2000)
pp. 212-223.
7. M. Tisza: Numerical Modelling and Simulation in Sheet Metal
Forming, J ournal of Materials Processing Technology, v.151.
(2004) No. 1-3. pp. 58-62.
8. A. Tang et al: CAE Based Die Face Engineering Development,
NumiSheet2005 Conf., 19-24. Sept. 2005. pp. 50-60.
9. M. Tisza: Expert System for Sheet Metal Forming, J . of Mat.
Proc. Techn. (1995) pp. 423-432.
10. M. Tisza: Numerical Modelling and Simulation: Academic and
Industrial Perspectives, Materials Science Forum, v. 473-474.
(2005) pp. 407-414.
11. A. Andersson: Comparison of Sheet Metal Forming Simulation
and Try-out Tools in Design of Forming Tools, J . of
Engineering Design, v. 15. (2004) No. 6. pp. 551-561.
12. M. Tisza: Rapid Parametric Process Design using FEM,
Advanced Materials Research, v. 6-8. (2005) pp. 235-240.
13. Zs. Lukcs, M. Tisza: Multi-step forming simulation,
ICIT2005, Bled, Slovenia, 12-14. April 2005. pp. 331-
336.

1 INTRODUCTION
A problem that has attracted some attention in the
recent past years is optimum blank shape design for
the sheet metal forming. Optimum blank shape has
many evident advantages. The optimum blank not
only improves formability and product quality but
also reduces material cost, number of trials in the
try-out stage and product development period.
Moreover, the optimum blank shape leads to the
prevention of tearing, the uniform thickness
distribution and the reduction of the press load
during drawing. However, it is not easy to find
optimum blank shapes because of the complexity of
material behavior especially in the actual stamping
dies described with 3D CAD data.
Designing of blank shape have been widely studied
by many researchers. Hazek and Lange [1] and
Karima [2] used the slip-line method to design the
initial blank. Chung and Richmond [3, 4] proposed a
direct design method and its theoretical basis to get
an initial blank shape of the sheet metal component.
These studies did not consider the real forming
conditions, such as blank holding forces, friction and
tool geometry. Barlet et al. [5] and Lee et al. [6]
proposed an inverse design approach using a
mathematical technique to obtain blank shapes,
considering contact conditions between the tool and
metal sheet. Analytical methods using the FE
analysis code were also developed [7, 8]. However,
expensive computational time is required in order to
obtain a precise blank shape.
In this paper, an effective algorithm called boundary
projection method based on finite element
simulation is presented for arriving at the optimum
blank shape. This technique approximately develops
the component shape onto a 2D plane. FE analysis is
used to simulate the forming process with real
processing parameters and a shape of the formed
part is obtained, which is compared with the shape
of the initial CAD design.
The FE forming simulation is carried out by using
the commercial FE code, Ls-Dyna. The material
properties, the friction between the material and the
tools and the blank holding force affect the
deformation of the material and thus affect the blank
dimension. The effects of these parameters on blank
shape can be rationalized via FE simulation. Two
examples are used to evaluate the proposed
algorithm. The results show that the optimum blank
shape can be obtained in a few iterations compared
to other methods.
ABSTRACT: The optimum blank shape is the minimization of the difference between the target contour of
the part and the outer contour of the deformed blank. The main objective of this paper is to introduce a new
blank design method based on iterative finite element (FE) sheet metal forming simulations. The algorithm is
based on the projection of the target contour on the deformed blank and modifying the blank shape
accordingly. The developed algorithm is applied a square cup drawing in order to confirm its validity.
Keywords: Optimum blank design-sheet metal forming-FE simulation
Optimum blank shape design in sheet metal forming by boundary
projection method
A. Vafaeesefat
School of Engineering, Department of Mechanical Engineering, Imam-Hussein University, Tehran, Iran.
URL: www.ihu.ac.ir e-mail: Abbas_v@yahoo.com
2 BOUNDARY PROJECTION METHOD
Figure 1 shows a detailed algorithm of the proposed
method. In the process of the blank design, an
interface program is developed to connecting the FE
analysis package, the blank design module, and the
re-meshing module. All parts including die, punch,
blank holder, and initial blank are first modeled. The
process parameters such as punch movement, blank
holder force, friction coefficient, and contact
conditions are determined. The initial blank is
deformed using FE analysis.
Fig. 1: Schematic diagram of boundary projection method
If the shape error becomes greater than specified
value, the boundary projection method is used to
minimize the shape error. The geometrical shape
error is defined as the difference between the
deformed blank contour and the target contour. The
target contour is generated from outer contour of the
product. In this stage, the procedure to define the
new blank shape is started from the projection of the
target contour to the deformed blank. Knowing the
elements that the boundary is projected on, the
position of the projected boundary is defined on the
initial blank. In this way, a new blank contour is
defined and the re-meshing module is used to define
a new blank shape model.
The new blank is deformed using the FE analysis
again and its boundary is compared with the target
shape. If any shape error is remaining, the whole
procedure is repeated until the error becomes smaller
than the given error bound.
In FE modeling, the blank is defined by triangular
elements and the target contour is defined by a set of
points. The first step to define the new blank contour
is to project each point of the target contour on the
deformed blank elements obtained by FE analysis.
To define the elements close to each point, a search
is done, and the element that the point has the
perpendicular projection on it is selected.
To transfer the position of this point on the same
element on the initial blank, the element is redefined
by a rectangle bilinear parametric surface. After
projection of the point on the selected element, the
parameters (u, v) of this point are first determined
(Figure 2). To allocate the location of this point on
the initial blank, the same parameters (u,v) are used
on the surface defined by the same elements on the
initial blank. Therefore, the point is mapped back
into the mesh system of the initial blank.
Fig. 2: Define the new blank boundary by point transformation
If the projection point falls out of deformed blank,
the closest element to this point (ie. E
2
in Figure 3)
is first determined. Then, the minimum distance d
between the projected point and this element is
calculated. Next, a point is generated with the same
distance from the defined element (E
2
) on the blank
boundary. In this way, a new blank boundary is
defined and a new blank mesh is accordingly
generated. The new blank shape is used for the next
iteration.
Fig. 3: Point transformation out of deformed blank
The geometrical shape error is defined as root mean
square of the shape difference between the target
shape and the deformed shape as:
E2
Q
d
Deformed blank Initial blank
Projected point
New blank
boundary
E1
E3
E2
Q
d
New boundary
point
E3
E4
E4
E4
Pi Pj
Pk
Q
u
v
Pi
Pj
Pk
Q
u
v
Deformed element Initial blank
Projected point
New boundary
point
New blank
boundary
El
El
Optimum
blank shape
Initial blank
shape
FE-analysis
Deformed blank
shape
Define shape error
Shape-
Error<
Define a new blank
contour by blank
projection method
END
Re-meshing the
blank
Yes
No
Blank
design
Re-meshing
module
FE analysis
package

N
i
d
N
Error
2
1
(1)
where
i
d is the distance between the target shape
and the deformed shape along the deformation paths,
and N is the number of nodal points along the
boundary of the blank.
3 EXAMPLES
To verify the validity of the boundary projection
method, the algorithm is applied to the process of
square cup deep drawing shown in Figure 4. In this
work, the desired blank shape for a square cup has
31 mm height and 5 mm width of flange. The sheet
metal is cold-rolled steel; sheet thickness, 1.0 mm;
friction coefficient, 0.125; press stroke speed, 2000
mm/s; blanking holding force, 200 kN; Youngs
modulus, 210 GPa; Poissons ratio, 0.3. When an
165 mm 165 mm square blank is used for an initial
blank, the results show that the shape error
significant (Figure 5). The outer contour of
deformed blank, the target contour, and the initial
blank boundary are shown in Figure 6. The first
modified blank shape can be calculated with the
result of the initial square blank. The analysis result
of this step is shown in Figure 7. The difference
between the deformed shape and the target contour
is remarkable. When the blank design process is
repeated three times the difference decreases and
converges to zero (Figure 8). Hence a square cup
with a uniform flange at its periphery can be made.
The final blank shape of square cup is shown in
Figure 9. The FE analysis of the final stage is shown
in Figure 10.
Fig 4: The square cup part
Fig 5: The FE analysis result of initial blank
Fig 6: The result of first step
Fig 7: The result of second step
Initial blank
Target contour
Deformed blank
Initial blank
Target contour
Deformed blank
Fig 8: The result of third step
Fig 9: The optimum blank shape
Fig 10: The FE analysis result of optimum blank
4 CONCLUSIONS
A new method of optimum blank design has been
proposed by using the boundary projection method.
The method was integrated in the finite element
modeling of sheet metal-forming process. Deep
drawings of an S rail and a square cup have been
treated as examples. It has been found out that with a
few iterations (three), the deformed contour shape
becomes almost coincident with the target shape.
The results show that the proposed method can reach
the desired blank shape within a few iterations. The
proposed method can be further applied to optimum
blank design of other practical sheet metal-forming
problems.
REFERENCES
1. V.V. Hazek, K. Lange, Use of slip line field
method in deep drawing of large irregular shaped
components, Proc. 7
th
NAMRC, 1979, pp. 65.
2. M. Karima, Blank development and tooling
design drawn parts using a modified slip line
field based approach, ASME Trans. J. Eng. Ind.
111 (1989), pp. 345.
3. Chung, K., Richmond, O., Ideal forming. I.
Homogeneous deformation with minimum
plastic work, Int. J. Mech. Sci. 34 (7), (1992),
pp. 575591.
4. Chung, K., Richmond, O., Ideal forming. II.
Sheet forming with optimum deformation, Int.
J. Mech. Sci. 34 (8), (1992), pp. 617633.
5. Barlet , O., Batoz, J.L., Guo, Y.Q., Mercier, F.,
Naceur, H., Knopf-Lenoir, C., Optimum design
of blank contour using the inverse approach and
a mathematical programming technique,
Numisheet96, (1996), pp. 178185.
6. Lee , C.H., Huh, H., Blank design and strain
prediction of automobile stamping parts by an
inverse finite element approach, J. Mater. Proc.
Tech., 63, (1997), pp. 645-650.
7. Toh, CH., Kobayashi, S., Deformation analysis
and blank design in square cup drawing, Int J
Mach Tool Des Res, 25(1), (1985), pp. 15-32.
8. Iseki, H, Murota, T., On the determination of
the optimum blank shape of nonaxisymmetric
drawn cup by the finite element method, Bull
JSME, 29(249), (1986), pp. 1033-40.
Initial blank
Target
Deformed
1 INTRODUCTION
Today the automotive industry is challenged by
increasing environmental regulations. One of the
main arguments for weight reduction of car body
assemblies is the decrease of fuel consumption. At
the same time, car body panels continue to be
subject to other complex demands such as
diversifying customer preferences and increasing
quality standards. In the past decades, the main
focus of efforts to meet these demands was put on
the use of lightweight or high strength materials, in
order to save weight by lower density or decreased
sheet thickness. When assessing the quality of sheet
metal car body panels in terms of their mechanical
durability, the denting resistance plays a key role.
Further quality attributes include the shape accuracy
and the sound radiation of the body panel. These
quality aspects can be controlled while work
hardening the sheet metal as part of the forming
process (1, 2). The strain dependency of basic
material parameters allows for a predefined
adjustment of the component properties, such as
denting resistance or denting stiffness. Thus, plane
stretch-forming combined with deep drawing allows
a controlled manipulation of the sheet metal part
properties (2). Current stretch-forming tool concepts
do not appear cost effective and flexible enough for
small series production to meet the high
diversification requirements of todays automotive
industry.
2 SHORT CYCLE STRETCH-FORMING (SCS)
2.1 General Function
The basic tool assembly includes two interlocking
sets of convex bead elements. The upper tool part is
fixed at the press ram. The lower tool part is located
on the press table. The constitutive function is based
on a bending and friction effect similar to
conventional draw beads.
CLAMPING SPECIMEN
UPPER
BEAD
SET
TENSILE LOAD
STRETCH-FORMING TENSILE FORCE
GENERATION
A
LOWER
BEAD
SET
Fig. 1. Basic tool assembly

ABSTRACT: Today stretch-forming is primarily used for the production of subconvex parts, such as aircraft
body panels. However, in the case of car body panels such as doors, roof or trunk lids, the application of
current stretch-forming technologies proves to be less efficient and more cost intensive. As part of current
research at the Institute for Metal Forming Technology in Stuttgart, a new stretch-forming technology called
SCS was developed which aims at improving these forming processes in order to meet economic
requirements. The SCS Technology allows for the application of highly efficient processes of combined plane
stretching and deep drawing using low-cost tools. The research and development was supplemented by newly
defined benchmarking methods. Several tools suitable for basic research work and one close-to-production
tool were developed and field-tested. The applied optimisation methods are described in this paper.
Key words: pre-defined work hardening, stretch-forming, car body panels, cost efficient tools
Benchmarking Methods for Short Cycle Stretch-Forming
D. Vlahovic
1
, M. Liewald
1

1
Institute for Metal Forming Technology
Universitt Stuttgart
Holzgartenstrae 17
70174 Stuttgart
Germany
www.ifu.uni-stuttgart.de e-mail:dejan.vlahovic@ifu.uni-stuttgart.de;
mathias.liewald@ifu.uni-stuttgart.de
When closing the tool, shifted bead segments of the
upper and lower tool section bend the fringe of the
blank alternately and pull it over the opposed
shoulder radii (Fig.1). Starting at the free end of the
blank, the tension stress increases stepwise along the
blank section due alternate bending and rebending,
increased by friction at upper R
Ui
and lower R
Li

shoulder radii (Fig.2). The last interfering upper
bead acts as a drawing ring; it pulls the sheet into the
bead geometry and stretches the fixed side of the
blank. Thereby the maximum tension results at the
equilibrium point A below the first acting shoulder
radius (Fig. 1). According to the proceeding of (3),
the above mentioned increasing tension stress along
the specimen section can be exemplary illustrated
over three process stages (k
fs1
, k
fs2
, k
fs3
) in Fig. 2.
T
e
n
s
i
l
e

S
t
r
e
s
s
RL1 RU1 RU2 RL2 RL3
kfs3
kfs2
kfs1
u
kfz3
kfz2
kfz1

Fig. 2. Increasing tension stress through the forming process
The process can be controlled by manipulating the
dimensions of the shoulder radii and the order in
which the bead elements interlock. Thus the load
equilibrium is being displaced depending on the
process state (Fig. 2). It is obvious that the process is
highly complex and transient, depending on the
provided set and geometry of the bead elements.
Hence the primary goal in development research on
this technology was a maximum reduction of the
possible geometry spectrum to an optimal tool
geometry and process sequence, in terms of cost
reduction and process efficiency.
2.2 Process Limits
During the forming process, the tensile stress along
the blank section increases continuously until the
ultimate stress (
u
=k
fz
) results in necking at point A.
Thus, the process is limited to a specific, material-
dependent flow stress k
fs
in the stretched region of
the blank. The tensile stress k
fz
at the equilibrium
point A can be determined by calculating the
increase of the flow stress k
fs
at the first shoulder
radius R
i
. A suitable theoretical solution for this
calculation has been proposed by Stoughton (4).
Primarily developed for the calculation of
restraining forces generated in draw beads, the
approach of Stoughton (4) considers all relevant
terms within the process examined here. The
primary shoulder radius, which the prestrained sheet
is pulled over, is the most important geometry
parameter in this case. The initial stress k
fs
in a
prestrained sheet pulled over the first shoulder radius
R
i
is increased by bending, friction and rebending
forces. An increase of the shoulder radius results in a
decrease of the restraining force up to a limit value
(5). In feasibility studies, the increase of the initial
stress by pulling a prestrained specimen over a
shoulder radius was calculated in dependency on the
shoulder radius dimension according to Stoughton
(4). Final dimensions of both shoulder radii in
addition on one bead element, are limited by the
bead width. An optimal dimension of shoulder
radius has to be determined. For further
examination, the primary shoulder radius above A
was specified to R=10mm. The determined flow
stress k
fz
at the equilibrium point A shown in Fig. 3
was calculated for mild steel DC04 with 0,75mm
sheet thickness and a primary shoulder radius of R=
10mm. Results suggest that with an effective strain
of 0,09 in the stretched region of the blank, the
stress value k
fz
at the equilibrium point exceeds
0
50
100
150
200
250
300
350
400
450
0 0,01 0,02 0,03 0,04 0,05 0,06 0,07 0,08 0,09 0,1 0,11
F
l
o
w

S
t
r
e
s
s

i
n

M
P
a
Effective Strain
g
in the Stretched Region
k
fs
k
fz
k
fs
k
fz
ultimate Stress
A Material: DC04
Initial Thickness: 0,75mm

Fig. 3. Theoretical process limit approximation
the ultimate stress
u
. It can be concluded that the
theoretically reachable effective strain in the
stretched region of the specimen averages 0,09.
3 PROCESS PERFORMANCE
As stated above, the examined stretch-forming
method features complex relations between the
process sequence and the bead set geometry.
General evaluation methods used in the development
of common sheet metal forming tool sets appear
insufficient in terms of detected process conditions.
In order to systematically identify the optimal
geometry parameters of the bead set, new bench-
marking methods had to be developed. In addition to
already presented systematic valuation methods in
(6,7), further significant methods were defined,
analysing the dependence of process efficiency and
process effectivity on stress and strain progression.
3.1 Geometry-related Performance
General demands to the SCS Technology include
minimal sheet cut and a minimal number of bead
elements featuring respective dimensions in order to
reduce production costs per unit and tool costs.
Feasibility studies suggested that two exemplary
SCS tool geometries can exhibit similar stress and
strain progressions, despite strongly differing in their
bead set geometries. Due to the SCS process limit
described in paragraph 2.2, it is not possible to
determine the most effective tool geometry by
means of their maximally reachable strain values
(7). The newly-developed benchmarking parameter
geometry-related performance allows for a more
precise statement. It describes the ratio of the work
performed by tension in the plane stretched blank
area W and the sum of overlapping bead set element
lengths E
Tn
(6).

= =
) ( ) (
0
t E
d k V
t E
W
W
Tn
f
Tn
V

(1)
The defined parameter has to be understood as
process performance over the press stroke in
dependence on the number of needed bead set
elements and the sum of their lengths, needed to
effect a specific forming work.
3.2 Geometry-related Efficiency
Supplementary to this geometry-related performance
benchmark, a further benchmarking value was
defined by proposing a geometry-related
efficiency. The efficiency of a tool geometry is
defined as the ratio of the minimally required tensile
stress
min
at the equilibrium point A, related to the
theoretically predictable maximum tensile stress
th

induced by the sum of bending and friction effects
within the bead set geometry. The maximal value of
the tensile stress
min
is equivalent to the ultimate
stress
u
.
th

min
=
(2)
Both tensile stresses
min
and
th
can be calculated
according to (4) and (7).
3.3 Improvement Results
The application of benchmarking methods described
within the present scope allowed for a development
of two highly optimised bead set geometries G2 and
G3 derived from the initially tested bead set
geometry G1, shown in Fig. 4.
G2 G1 G3

Fig. 4. Bead set geometry optimisation levels
Basically, two fundamental function principles can
be distinguished regarding the developed bead set
geometries. These are on the one hand the
cumulative tension strain induction by a multiple
bead set geometry (G1, G2), on the other hand the
singular tension strain induction (G3) with two
upper and lower beads, respectively.

Table1. Experimentally determined performance
Geometry Parameter G1 G2 G3
Number of upper bead elements 4 3 2
Number of lower bead elements 5 3 2
Total bead length 90mm 80mm 80mm
Required bead length for max. strain 50mm 50mm 40mm
Maximally reachable effective strain 0.08 0.09 0.095

As can be observed, the highest optimisation level
G3 offers excellent performance at minimal
geometry requirements. The G2 geometry offers
higher variability, but requires two more beads and
0
0,1
0,2
0,3
0,4
0,5
0,6
0,7
G1
G2
G3
G
e
o
m
e
t
r
y

-
R
e
l
a
t
e
d

E
f
f
i
c
i
e
n
c
y

Fig. 5. Bead set geometry optimisation levels
an additional 10mm of bead length to reach the same
performance. These advantages of the G3 geometry
can be quantified using the calculated geometry-
related efficiency shown in Fig. 5. Further
performance characteristics during the process
sequence can be expressed using the experimentally
determined geometry-related performance W shown
in Fig. 6.
0
500
1000
1500
2000
2500
3000
3500
0 10 20 30 40 50 60
G
e
o
m
e
t
r
y

-
R
e
l
a
t
e
d

P
e
r
f
o
r
m
a
n
c
e

W
Press Stroke in mm
G1
G2
G3

Fig. 6. Bead set geometry optimisation levels
4 CLOSE-TO-PRODUCTION TOOL CONCEPT
A final verification of the SCS Technology was the
implementation of a close-to-production tool. The
developed tool concept offers the option to combine
a stretch-forming function with a subsequent deep
drawing operation.

Fig. 7. Automobile door outer planking
The tool design also offers the option of stretching
the blank up to a predefined strain value. In field
tests, a scaled automobile door body panel was
realised using DC04 mild steel with a sheet
thickness of 0,75 mm. By changing the blank
geometry, defined strain values of 0%, 2%, 4%, 6%
and maximally 7% in the centre region of the door
panel were reached. In sum, it can be stated that the
SCS Technology offers new approaches to improve
component properties of sheet metal body panels.
5 CONCLUSIONS
In this paper the newly developed stretch-forming
technology SCS was presented and discussed in
terms of principal function and process limits. The
obviously complex and transient nature of the
examined technology requires systematic
approaches within the development cycle of the tool
geometry. Besides the already examined process
benchmarking parameters, new benchmarking
approaches were defined. These new benchmarking
parameters allow for an estimation of the process
performance depending on the geometry parameters
of a bead element set. Applying the defined
benchmarking methods, two most effective bead set
geometries were found. Based on results of this
optimization work, a close-to-production tool for
combined plane stretching and deep drawing was
developed and field-tested. The final result of the
conducted research work is an innovative, cost
effective tool concept.
REFERENCES
1. Taylor, B.: Stretch-Forming. ASM Handbook, ASM
International, 591-598.
2. Vlahovic, D.; Liewald, M.; Siegert, K.: Optimierung
von Ziehteileigenschaften durch gezielte Vorverfesti-
gung. Neuere Entwicklungen in der Blechumformung,
MAT INFO, Frankfurt, 478-510
3. Filzek, J.: Kombinierte Prfmethode fr das Reib-,
Verschlei- und Abriebverhalten beim Tief- und
Streckziehen. Berichte aus der Umformtechnik, Band
62, TU Darmstadt, Darmstadt 2004.
4. Stoughton, T. B.: Model of Drawbead Forces in Sheet
Metal Forming. Proceedings of the 15th Biennal
IDDRG Congress, 1988, 205-214.
5. Farr, M. T.: Zieh- und Stempelkantenradien beim
Tiefziehen. Beitrge zur Umformtechnik 31, DGM
Verlag, Frankfurt am Main 2002, .
6. Vlahovic, D.;Liewald, M.: Neue Technologieanstze
zum Recken von Blechen. WT-Online, Vol.97, H10,
Springer-VDI-Verlag, Dsseldorf 2007, .
7. Vlahovic, D.; Liewald, M.: Improvement of Car Body
Outer Skin Properties Using New Sheet Metal Forming
Technologies. 7. Int. Stuttgarter Symp. Automobil- und
Motorentechnik, Vieweg, Wiesbaden 2007, .
8. Vlahovic, D.; Liewald, M.: Neue Technologieanstze
zum Recken von Blechen Entwicklung kosten-
optimierter Werkzeugkonzepte. WT-Online, Springer-
VDI-Verlag, Heidelberg 2007, .

Adaptive Bending of Aluminium Extrusions Using an Automated
Closed-Loop Feedback Approach
T. Welo
1
, K. Stertr
1
, O.P. Svik
2

1
NTNU-Dept. of Engineering Design and Materials Rich. Birkelands v. 2B, N-7491 Trondheim, Norway
1 INTRODUCTION
1.1 Motivation and objective
European manufacturers are currently facing
increased competition from companies based in low
cost countries. Hence, future competitiveness is
strongly related to their capability in developing and
integrating new technology, followed by
commercialization into a stream of products that
provide additional customer value in terms of
reduced cost, improved quality as well as increased
features and functionality.
One strategy to meet this challenge is developing
more automated production technology, providing
reduced labour cost while improving product
quality. Adaptive processing is one of several
technologies that support the desire of creating
competitive advantages by offering improved
products in the market place. It is applicable to
numerous manufacturing processes that require
high-quality parts. Moreover, adaptive processing
may be executed at different levels of sophistication.
For example, conventional stretch bending may be
considered as a low-level adaptive process since
simultaneous stretching and bending are known to
reduce springback, hence improving the dimensional
accuracy caused by variability in mechanical
properties and geometric dimensions of incoming
parts. The next level of sophistication may be
associated with the method of manually adjusting
the settings of a tool or machine, using data from a
few test trials (of a new batch) and experience data
from previous production batches. The highest level
of sophistication in connection with adaptive
processing is to integrate an automated closed-loop
feedback scheme for instantaneous process control.
Hence, in-line measurements are utilized in order to
correct settings and process parameters while the
component is being processed [1].
Development of automated, closed-loop feedback
URL: www.ntnu.no e-mail: torgeir.welo@ntnu.no; kristian.satertro@simulix.com

3
Hydro Aluminium Structures, Design & Engineering - Building 5, Industriv. 1, N-2831 Raufoss, Norway
URL: www.hydro.com e-mail: odd.perry.sovik@hydro.com
ABSTRACT: A new rotary, compression bending set-up with automated closed-loop feedback control is
being developed. The overall goal is to improve the dimensional accuracy of formed shapes using elastic
springback compensation. In-process measurement data are transferred into an algorithm (steering model) for
prediction of springback and bend angle prior to unloading. Emphasis was placed on developing a physically-
based steering model. More than 150 bending tests of AA6060 extrusions were conducted to demonstrate the
capability of the technology. Prior to forming, the material was exposed to different heat treatments to
provoke a range of stress-strain behaviours, which are known to affect elastic springback. An optical
measurement procedure was established to determine key dimensions and their associated statistical
distributions. When using traditional compression bending, the results show that the variability in springback
of a selected reference point was in the range of 10 % of the nominal springback. Using the closed-loop
feedback system, the corresponding variability in springback was in the range of 3 %, representing a factor-
three improvement in terms of dimensional process capability (C
p
). It is concluded that the present technology
has a high industrial potential, in particular for volume components with tight dimensional requirements.
Key words: Adaptive, Bending, Closed-loop Feedback, Dimensional Accuracy, Springback, Aluminium, Extrusions
control, applied to rotary compression bending, is
the primary focus in this paper. The objective of the
work is to establish a method for in-process control
of profile bending, focusing on springback
compensation, along with the associated steering
model.
1.2 Closed-loop adaptive control strategies
There are multiple strategies for springback
compensation in a bending operation, see the
principles in Figure 1. One method (A) is (i) to
unload the part at an intermediate forming stage, (ii)
record springback characteristics, (iii) use the
measured data to estimated stop position using a
predetermined algorithm, (iv) reload the part to the
predicted stop position, and finally (v) unload the
part. A second strategy (B) is to (i) form the part to a
prescribed stop position that would normally result
in an under-bent part, (ii) unload the part and
measure the springback, (iii) utilize the data for
predicting a new stop position using a predetermined
algorithm, (iv) reload the part to a new stop position
closer to the nominal one, (v) repeat the procedure
until the part geometry meets the desired part
geometry. Both these strategies are mainly suitable
to small-batch production of customized products
since the loading-unloading scheme increase cycle
time.
A third strategy (C), which is more applicable to
high volume production, is to run the operation as a
one-hit, conventional process. Rather than
measuring springback directly, other more indirect
(underlying) parameters such as bending moment,
stretch and section dimensions need to be measured
instantaneously. A successful outcome, however,
presumes the existence of an accurate steering
model. Moreover, the measurement technology and


Bending angle,
B
e
n
d
i
n
g

m
o
m
e
n
t
,

M
p
Strategy A
Strategy B
Strategy C
Bending angle,
B
e
n
d
i
n
g

m
o
m
e
n
t
,

M
p
Strategy A
Strategy B
Strategy C

Fig. 1. Control strategies for adaptive bending using automated
closed-loop feedback.
equipment must be robust, accurate and reliable to
provide reliable input to the steering model.
2 EXPERIMENTS
2.1 Set-up
The lay-out of the (rotary) compression bender is
shown in Figure 2. The assembly consists of an
electric power unit that is connected to a gear box. A
torque transducer is placed between the exit of the
gear box and the entry shaft of the upper bending
arm. The rotation of the bending arm is measured
directly using a rotational transducer connected to
the gear. A drawback arrangement is mounted at the
underside of the bending arm to eliminate friction as
the profile slides towards the upper bending tool
during bending. The drawback is hinged locally at
the bending arm to ensure free rotation of the front
end of the profile. A device (not shown in the figure)
that is operated with air pressure clamps the rear end
of the profile, constraining rotation and translation in
the length direction. The lower tool has a constant
radius and is fixed. The tools contact surface is
made with a protruding ridge to make a local imprint
along the inner flange of the profile during bending.
During forming and unloading, both torque and
rotation are continuously recorded and fed into a
PC-operated control system, which automatically
calculates and updates the stop position using a pre-
determined steering model. The process is entirely
controlled by the control system, without any human
interference other than specifying desired bend angle
of profile, loading the part, click the go-button, and
removing the finished part. Due to the control
strategy adopted (C), the cycle time of the bending
machine is the same as for conventional
compression bending technology.
2.2 Calibration and test procedures
Since the torque ( ) ( M ) and rotation ( ) are
measured directly on the shaft that connects the gear
and the bending arm, the effects of gravity forces
( ) (
g
M ) and bearing friction ( ) , (
i
M

) have to
be reset to zero, hence
) ( ) , ( ) ( ) (
g i p
M M M M = (1)
where ) (
p
M is the bending resistance of the profile.
Readings obtained by running the machine without
profile for were used to calibrate the
o
90 0

Fig. 2. Outline of bending machine and tooling.
transducer such that only the contribution from the
profile would be measured in the tests. After
resetting the moment readings, an additional ten
tests were run without profile to determine
variability (from
i
). The results showed that the
recorded torques standard deviation was in the
range of 1.22.0 Nm within one cycle, and that the
mean value increased slightly from the first to the
last test (1.45.0 Nm), reflecting the overall
variations from friction and measurements.
The measurements procedure included clamping the
profile loosely to a fixture and measuring the angle
made up by the fixed and the free ends. A digital
protractor, Clinotronic-Wyler (resolution of 1/60),
was used to measure the final bend angle. The
repeatability of the procedure was checked by
performing a number of consecutive measurement
trials on the same profile.
A thin-walled, rectangular hollow AA 6060-T1
profile was used in the tests. In order to provoke
different material characteristics, the profiles were
aged to different temper conditions, including as
is(T1), and, respectively, 60 minutes and 120
minutes at 175 C, providing a 17 % range in yield
stress. The test overview is given in Table 1.


Table 1. Test overview.
Series no. Control model Material #of profiles
1 Manual As is
25
2 Manual 175C / 60min 25
3 Manual 175C / 120min
25
4 Adaptive As is
25
5 Adaptive 175C / 60min
19
6 Adaptive 175C / 120min
20


Fig. 3. Bent part clamped in fixture and ready for dimensional
measurements.
3 STEERING MODEL
Establishing a physical steering model for
springback compensation is a tedious matter, whose
details cannot be reported in detail within the scope
of this article. The procedure was based on beam
theory using a non-linear, closed-form moment-
curvature relationship as basis. The result may be
converted into a steering model on the form:

+ + +
=
)
~
(
~
) ( 1
)
~
(
~ 2 2
) (
) (
2
1
) (
~
0
2
0 0
0
3
0
0 2
1
0
0 0

M
EI
R
n f
M
L
D
L
R
n f
EI
L n f
n f
R
L
(2)
Here
~
is the die rotation at end of forming,
is measured torque, n is strain hardening parameter,
is initial bending stiffness, and and
)
~
(
~
M
0
EI
0

0
are
the parts desired bend angle and the difference
between the initial rotation of the die and the profile,
respectively. Some other geometrical parameters are
illustrated in Figure 4. If the measurements are
limited to the die rotation and torque, the above
equation may be simplified:
( ) )
~
(
~
1
)
~
(
~

~
3
1 0 0

M c
M c c

+ +
= (3)
4 RESULTS
A summary of the results obtained from more than
140 tests is given in Table 2. For the adaptive
process, the targeted angle ( ) was 80, whereas
the traditional process was run with a pre-specified
bend angle (
0

) of 85 without any attempts made to

hit the same nominal. The dimensional capabilities
can be evaluated by considering the process
capability index :
B
A

BA
v
AB

B
L
L
0
R
F

AB

C
D
2
D
clamp
tool con-
straints
B
A

BA
v
AB

B
L
L
0
R
F

AB

C
D
2
D
clamp
tool con-
straints

Fig. 4. Lay out of structural and kinematical model for segment
A-B prior to unloading, including key dimensions.
) ( 6

=
SD
LSL USL
C
p
(5)
In which USL and LSL are the upper and lower
specification limits, respectively, and SD() is the
standard deviation of the realized bend angle.
Assuming a tolerance band of 1.0, the adaptive
process shows a dimensional process capability that
is more than three times better than the traditional
process. If the bend angle is considered being a
standard dimensional feature (with C
p
>1.33) of a
specific part, the traditional process would require a
tolerance band of 3.26, whereas the adaptive
process would only need a tolerance band of 0.53
in order to provide good parts. This result clearly
demonstrates that adaptive processing has a high
industrial potential for improving part quality and
reducing quality cost.
The statistical distributions of the two processes are
shown in Figure 5. For illustration purposes, the
distribution for the two processes is moved to have
the same nominal bend angle (average). The
traditional process shows three clusters, one for each
heat treatment, with T1-profile results to the far right
in the figure. The steering model does merge the
results together, indicating that the main influential
parameters are utilized in the steering model.
One can obviously argue that the heat treatment
made to provoke different material characteristics

Table 2. Result summary.

Traditional proc. Adaptive proc.
Average angle, 80.75 79.82
Max angle,
max

81.32 80.05
Min angle,
min
Std. dev. ()

80.05 79.60
0.41 0.13
p
C
(0.5) 0.41 1.25
LSL USL (C
p
=1.33) 3.26 1.06
Unit: [] []
resulted in a somewhat larger spread in properties
than one would normally see in industrial practice
for, say, T1 material. This is a correct statement if
the manufacturer is capable of controlling the
casting and extrusion processes as well as the shelf
life of the material. Therefore, additional statistical
analyses were made between batchs 1 and 2, and
batches 2 and 3, for the two bending methods. The
results showed that the traditional bending
technology provided tolerance bands 2-4 times
wider than those of the adaptive technology at the
same process capability.
0
0,5
1
1,5
2
2,5
3
3,5
79,800 80,000 80,200 80,400 80,600 80,800 81,000 81,200 81,400
N
o
r
m
a
ld
is
t
r
i
b
u
t
i
o
n
,

d
e
n
s
i
t
y

f
u
n
c
t
i
o
n
Finishedprofile angle
Distribution
Manualprocess
Adaptiveprocess

Fig. 5. Statistical distributions of bend angle with the two
methods.
5 CONCLUSIONS
Based on the work presented herein, the following
conclusions can be drawn:
A new, adaptive bending technology with
closed-loop feedback has been developed
and validated using full scale experiments;
The adaptive bending method has proven to
dramatically improve the dimensional
process capability;
The technology has a great industrial
potential in terms of improved dimensional
quality and reduced manufacturing costs.
ACKNOWLEDGEMENTS
The financial support of The Norwegian Research Council
through the project ALUPART, as well as the financial and
technical support from Hydro Aluminium Structures A.S.,
Raufoss are gratefully acknowledged.
REFERENCES
1. H. Chu and K.A. Stelson, Modeling and Closed-Loop
Control of Stretch Bending of Aluminum Rect. Tubes, J .
Mnf. Sci. and Engn., Vol. 125, pp. 113-119, (2003).
1 INTRODUCTION
In the FP6 project DATAFORM [1], advanced panel
forming methods, i.e. multi-point forming (MPF)
technologies [2], are developed and applied mainly
in aeronautical panel manufacturing. Traditional
solid dies are replaced by discrete punch matrices
that may be digitally controlled using specific CAD
software, Fig.1. An important issue is to understand
how to predict and correct the springback (SB) of
panels when designing the CAD controlling code.
Generally, MPF technology provides new and
effective way to correct the forming deviations due
to SB, because the discrete punch matrix surface
may be easily adjusted according to measured or
estimated SB values. On the other hand, the use of
discrete punches, and also of interpolators between
the punch matrices and the sheet metal for avoiding
forming defaults (dimple, wrinkle, etc.), makes finite
element simulations complex and difficult.
Fortunately, as shown by Socrate & Boyce [3] in
their numerical research of sheet-metal forming
process on reconfigurable tools, it is possible to
simplify the numerical modelling by replacing
discrete punch matrices with equivalent continuous
surfaces. This simplification has been adopted in the
present numerical calculations.
In this work, investigation on SB was limited mainly
in aeronautical panel production by stretch-forming
processes. The objective is to develop simple
methods of predicting and preventing the SB effect
based on theoretical and numerical results, which
may be easily implemented in CAD controlling
software, being developed for the MPF prototypes.
2 SPRINGBACK PROBLEMS OF STRETCH
FORMING PROCESS
As many aeronautical panels have relatively small
curvature and large dimension, simple bend forming
or stamping process could not be effective and
convenient to form sheets to designed shapes. To
produce these panels, stretch forming process is
extensively adopted: besides necessary bend loads,
stretch loads are also applied in the sheet plane
(Fig.1). In this paper, the following problems will be
mainly dealt with:






Fig.1 Stretch forming process
1) If applied stretch strain is large enough, stretch
deformation dominates everywhere in the whole
across section of sheets. The SB due to non uniform
stress along thickness will be constrained by the
applied stretch strain. This will be highlighted in 3
by finite element calculations to propose an optimal
stretch strain for general stretch forming process.
2) Generally the SB values of sheets may be well
related to material flow strength
f
and stiffness
(Young module E) in the case of pure bending. It
ABSTRACT: This paper studies the springback behaviors of panel forming productions using multi-point
stretch forming technique. Various possible influencing factors on the springback effect were analyzed. Finite
element simulations were carried out by using a commercial finite element code SAMCEF/MECANO.
Special concern was placed on the prediction and prevention of the springback effects in the stretch forming
process for the production of aeronautical panels, where the springback becomes an important issue due to
high precision requirements, and the use of aluminum alloy materials of high strength/low elastic modulus.
Key words: Springback, Material stretch forming, Finite element simulation
Springback in stretch forming process of aeronautic panel production
by finite element simulation
A-M. Yan, I. Klappka


Open Engineering S.A. SAMTECH Group - 8 rue des Chasseurs Ardennais, 4031 Angleur, Belgium
URL: www.open-engineering.com e-mail:am.yan@open-engineering.com
Initial sheet under stretch forming unloading ->small SB
Punch matrix
will be evaluated by FE simulations in the present
stretch forming how the following simple relation
may be approximately applied:
/
f
SB E (1)
3) Although SB values in a usual bending forming
process depend generally on geometrical features of
sheets (thickness, curvature radius), this geometric
effect may be less important in the present stretch
forming process, which will also be investigated.
4) Since stretch forming process involves large
plastic deformation, it is logical to assume that the
SB values of panel forming process is loading-path
dependent. Optimal loading paths will be determined
by numerical simulations.
5) Shape error estimations and measurements of
panels are generally necessary to correct the SB by
adjusting the height of punches. In many practical
situations, only some typical S
b
values (e.g. at die
boundary) may be easily measured. So it is
interesting to investigate the SB distribution in
panel. The following equation is proposed to
describe SB distribution along the stretch direction
based on numerical simulations on simple panels [4]
( )
0.85
[ cos( / ) 1]
b
SB S x L = (2)
where S
b
is the measured SB value at the die
extremity; L is the half length of die, x is distance
from the symmetric centre of die.
All above problems will be detailed in Section 3 by
FE calculations on panels using SAMCEF code [5].
3 FE CALCULATIONS OF STRETCH
FORMING ON PANELS
3.1 Sheet material model
0
100
200
300
400
500
600
0 0,1 0,2 0,3 0,4 0,5 0,6 0,7 0,8 0,9 1
2024T3-like
2024O-like

Fig. 2 Stress-strain curves used for numerical calculations
Two nonlinear material models were used for FE
simulations, Fig. 2. Punch material is assumed rigid.
3.2 SB calculation
SB may be simply defined as back displacement of
sheet from forming limit state to the free state after
unloading. Fig. 3 gives an example of spherical shell
(taking symmetrical part for FE calculations). Due
to the nonlinearity of material and geometry,
stepping loads are applied. During this process, the
FE mesh is deformed according to the material
models. Then the calculated deformation is
converted to a measure of stress. The further
unloading accompanies elastic back of each area of
the sheet. The deformation interaction of all adjacent
areas induces the final SB. For the numerical
convenience, loading-unloading at sheet extremities
was replaced by a move-up then move-down of the
die with appropriate constraint conditions at the
extremities of the sheet. FE meshes were verified to
assure that the SB results are mesh-independent.

Fig. 3 Displacement history (mm) of a set of points at line OC
3.3 Effect of sheet thickness and curvature on SB
Springback-different thickness
0,0
0,2
0,4
0,6
0,8
1,0
1,2
1,4
1,6
1,8
2,0
1 2 3 4 5 6 7 8 9 10 11 12 13
Position
S
p
r
i
n
g
b
a
c
k
(
m
m
)
t=2mm, p_x=320N/mm, maxStrain=5%
t=3mm, p_x=480N/mm, maxStrain=5%

Fig. 4 SB calculation of cylindrical shell under stretch forming

Consider first a cylindrical shell subjected to
stretching load while warping the sheet. Then
unloading leads to the SB, which seems independent
of sheet thickness and curvature (Fig. 4). However,


Loading by
moving-up
the die
SB at a set of points
due to unloading &
relaxing fixation
C
A
O
B
in the case of spherical shell, the thickness and
curvature influences somewhat the SB: the thinner is
the thickness, the larger the SB. It is noticed that, in
comparison with usual bend forming process, this
geometric effect seems not very important especially
when the sheet has larger thickness.
Springback along CO and BA
0,0
0,5
1,0
1,5
2,0
2,5
3,0
3,5
1 2 3 4 5 6 7 8 9 10 11
stretching end <-- Position --> symetirc centre
S
p
r
i
n
g
b
a
c
k
(
m
m
)
t=1mm-CO
t=2mm-CO
t=3mm-CO
t=1mm-BA
t=2mm-BA
t=3mm-BA

Fig.5 SB calculation of spherical shells under stretch forming
3.4 Effect of loading paths on SB












Fig. 6 SB values depend on loading paths
Taking example of a cylindrical shell, three loading
paths are examined with two forming cases: small or
large bending deformation, to show their effects on
the SB values:
Path1: apply bending and stretching at same time
Path2: apply bending first and then stretching
Path3: apply stretching first, and then bending
In both forming cases, loading path 3 is the worst
because of large SB values, path 1 seems better, but
path 2 is the best. This conclusion is just from the
point of view of reducing the SB values.
3.5 Effect of material mechanical properties on SB
It is evaluated, in the case of stretch forming, if the
SB value may still be related by materials stiffness
(Young modulus) and typical flow strength (that is
the attained maximum stress in a uniform stretch
forming process, for example, in the case of
cylindrical shell, or a stress between the initial yield
stress and the attained maximum stress in the case of
general shells)
In Fig. 7, the SB values at the extremity of sheets are
related to the ratio of the yield stress and Young
modulus, showing a perfect linear relation in the
case of cylindrical sheets, while an approximate
linear relation is still useful in the case of spherical
sheet forming.

Fig. 7 Effect of materials properties on SB
3.6 Effect of applied stretching strain on SB
springback - stretching strain relation
44,9
9,7
5,3
4,55
2,34
1,74 1,42 1,22 1,08
11,249
6,93
3,586
2,90
2,149 1,776 1,431
0
5
10
15
20
25
30
35
40
45
50
0 1 2 3 4 5 6 7 8 9 10 11 12 13
stretching (plastic) strain (%)
s
p
r
i
n
g
b
a
c
k

v
a
l
u
e

a
t

o
r
i
g
i
n
a
l

p
o
i
n
t

(
m
m
)
2024O_like
2024T3_like

Fig.8 S
b
decreases with applied stretch strain

As discussed in 2, applying an appropriate stretch
strain may reduce the SB effect in the sheet forming
process. In this work, a series of stretch strain were
tested to show their effects on S
b
(reference to Fig7)
of a given cylindrical shell (R=1000mm). The results
are presented in Fig. 8 in the case of small bending
deformation with two materials of 2024 O-like
material effect on springback
0
0,5
1
1,5
2
2,5
3
3,5
0 0,5 1 1,5 2 2,5
Sigma/E(e-3)
S
p
r
i
n
g
b
a
c
k

v
a
l
u
e

a
t

s
y
m
m
e
t
r
i
c

c
e
n
t
e
r
cylindrical shell
spherical sheet
Springback-different loading paths
0,0
1,0
2,0
3,0
4,0
5,0
1 2 3 4 5 6 7 8 9 10 11 12 13
stretching end < Position > symmetric center
S
p
r
i
n
g
b
a
c
k
(
m
m
)
case1-path1
case1-path2
case2-path1
case2-path2
case1-path3
case 2-path3
Case 1: small bending
Case 2: large bending
before SB
after SB
S
b

Die
S
b

f
/E 10
3
Optimal stretch strain range
and 2024 T3-like. When the applied stretch strain
is small (say smaller than 1%), S
b
may be very
important. But it decreases rapidly with the
increased stretch strains. However we could not
apply too large stretch strain, because it does not
reduce much more S
b
but may cause over-thinning
and damage of shells. So we may propose an
optimal stretch strain in the range of 2-5%, suitable
to different materials. This result is in very good
agreement with our investigation in aeronautical
manufacturing industries [6].
3.7 SB compensation by adjusting punch matrices
-10,0
-9,0
-8,0
-7,0
-6,0
-5,0
-4,0
-3,0
-2,0
-1,0
0,0
0 1 2 3 4 5 6 7 8 9 10 11 12
symmetric centre < --Position -- > extrimity of die
C
o
r
r
e
c
t
i
o
n

v
a
l
u
e

(
m
m
)
ep=12%, S/S0=2,44
ep=9,2%, S/S0=2,22
ep=6,6%, S/S0=2
ep=4,4%, S/S0=1,78
ep=2,7%, S/S0=1,56
ep=1,55%, S/S0=1,33
ep=0,74%, S/S0=1,11
ep=0,3%, S/S0=1

Fig. 9 Compensation values of die for correcting SB
-3,0
-2,5
-2,0
-1,5
-1,0
-0,5
0,0
0 1 2 3 4 5 6 7 8 9
symmetric centre <-- Position --> extrimity of die
C
o
r
r
e
c
t
i
o
n

v
a
l
u
e

(
m
m
)
ep=0,75%, S/S0=1,11
A(cos(X/L)-1)
A[(cos(x/L)^0,85-1]

Fig. 10 SB distribution along the stretch direction is described
by a modified cosine function
To compensate the SB in a shell forming process,
we need to know how the SB distributes in the panel
if only S
b
in Fig. 7 may be measured. To be easy, we
take the symmetrical centre point of sheet as
reference to define a position of zero, so that the SB
values in different forming cases presented in Fig. 9
indicate just the necessary displacement of punches
for compensating the SB effect. The plate is
generally longer than the dimension of forming die
of punch matrix; of course, we are interested only in
the inside part where the SB varies in a nonlinear
way. We found that the SB distribution along the
stretch direction may be well described by a
modified cosine function (2), Fig. 10. This result
may be conveniently implemented in the CAD
software design.
4 CONCLUSIONS
The stretch strain has been numerically proven to
play a key role in reducing the SB in the stretch
forming process of panels. An optimal stretch strain
was found in 2-5% by a synthetic consideration.
Numerical simulations indicated that the SB effect is
loading-path dependent and may be linearly related
to material properties in the case of simple
geometry. However, the effect of sheet thickness and
curvature may be less significant. This is due to the
fact that the strain distribution through the sheet
thickness, which affects the SB of sheets, is
controlled mainly by the applied stretch strain. The
proposed SB distribution in panels along the
stretching direction may be easily implemented in
CAD code design for MPF prototypes.
ACKNOWLEDGEMENTS
Researches were performed in DATAFORM project funded by
the European Commission under the 6th Framework
Programme in the area of Aeronautics (Contract N 030877).
REFERENCES
1. DATAFORMDigitally Adjustable Tooling for
manufacturing of Aircraft panels using multi-point
FORMing methodology, FP6 Aeronautic and Space
Activity project, Annex I - Description of work, 2006
2. Jilin University (Dieless Forming Technology Center),
Technical Report for Multi-Point Forming, DATAFORM
project research report, Dec. 2006
3. S Socrate, MC Boyce, A finite element based die design
algorithm for sheet-metal forming on reconfigurable
tools. Journal of Engineering Materials and Technology,
Transactions of the ASME V.123, 2001, Pages 489-495.
4. A-M Yan, C-G Liu, Z-Y Cai, F-X Tan, I Klapka, S-H
Wang, G Sun, F Massabe V. Madhavan, L. DAlvise, S-
Z Su, X-W Wang, Deformation Mechanism of Multi-
Point Dieless Forming, DATAFORM project deliverable
D2.1, Oct. 2007
5. SAMCEF V12.1, Samtech Company, Belgium, 2007.
6. A-M Yan, F Massabe, A Farrell, S Su, I Klapka, C Liu,
User Need Analysis on Multi-Point Forming Techniques
Based on Interviews and Investigations to European
Aeronautical Industries, DATAFORM project report
D1.1. April, 2007.

Punch matrix
(die)