Journal of Materials Processing Technology 190 (2007) 300304

Effect of normal anisotropy on springback

, A. Haldar
Research and Development Division, Tata Steel Ltd., Jamshedpur 831001, India
Received 7 April 2006; received in revised form 23 October 2006; accepted 20 February 2007
Abstract
Share of high and advanced high strength steels in automobile is increasing, however, such steels generally have poor formability and high
amount of springback. One of the focus areas of research in high strength automotive steel is to increase the normal anisotropy to get better
formability. Effect of strength and process parameters on springback has been studied by many researchers but that of anisotropy has not been
studied by many. In the present work the effect of anisotropy on springback is predicted using nite element analysis for the benchmark problem of
Numisheet-2005 [2005 Numisheet Benchmark 2, Springback prediction of a cross member, Proceedings of the 6th International Conference and
Workshop on Numerical Simulation of 3D Sheet Metal Forming Processes, Detroit, USA, August 1519, 2005]. An analytical model is developed
to cross check the trends predicted from the nite element analysis. The effective stress has not been treated as a constant and the radial stress is
considered in the present model. Both the models (FE and analytical) predict that higher anisotropy, in general, gives higher springback. Finite
element analysis of the problem shows that springback is minimum for an isotropic material.
2007 Elsevier B.V. All rights reserved.
Keywords: Springback; Normal anisotropy; FEA; High strength steels
1. Introduction
High and advanced high strength steels are nding wide
acceptance in the automotive industry. However, one of the
major problems in stamping automotive parts with high strength
steel sheets is the increased levels of springback. In springback,
there are two issues:
Control of springback during forming and
Prediction of springback so that it is compensated during dies
design stage.
Various researchers developed a variety of manufacturing
techniques to control the springback. Sunseri et al. [1] showed
that a variable binder force history during forming operation can
reduce the springback amount while maintaining a relatively low
maximum strain. In their study an initial low binder force fol-
lowed by a higher binder force was used. Rufni and Cao [2]
and Cao et al. [3] proposed neural network based models to min-
imize the springback in a channel forming process. It was shown
that even for large variations in friction condition and material

Corresponding author.
E-mail address: [email protected] (R.K. Verma).
properties the springback is quite low. Cao et al. [3] showed that
a stepped binder force can reduce the springback signicantly.
Two critical values, high binder force and percent punch of total
punch displacement (the outputs from the neural network) are
used for the process control. In their study, the material was
assumed to be isotropic, elasto-plastic following the von-Mises
yield criterion. Hardening was assumed to be isotropic.
Though control of springback is important, the prediction of
it is desirable. Both analytical and numerical models have been
developed for the prediction.
Zhang and Lin [4] proposed an analytical solution for spring-
back in components stamped by a rigid punch and an elastic
die whereas, Morestin et al. [5] proposed a model in which
the PrandtlReuss plasticity equations associated with a non-
linear kinematic hardening model was solved. The calculation
takes into account the change in Youngs modulus with plastic
strain. Liu [6] proposed a simple model for predicting the spring-
back and bendability. He considered the normal anisotropy and
strain-hardening exponent in his model, however, the effective
stress is considered to be constant through out the bending. His
prediction showed that with increase in normal anisotropy the
springback increases but did not give any reason. Wang et al. [7]
established a mathematical solution for plane-strain bending of
sheet and plate. Using their model they predicted that bending
moment and, therefore, the springback increases with increase
0924-0136/$ see front matter 2007 Elsevier B.V. All rights reserved.
doi:10.1016/j.jmatprotec.2007.02.033
R.K. Verma, A. Haldar / Journal of Materials Processing Technology 190 (2007) 300304 301
in strength, strain-hardening and normal anisotropy. They used
Hills (1979) non-quadratic yield criteria but did not consider
the radial stress and, therefore, their model may not be suitable
for thick sheets or in severe bending condition. Pourboghrat
and Chu [8] used a momentcurvature relationship derived for
sheets undergoing plane-strain stretching, bending and unbend-
ing deformations. They used membrane nite element method
with bending/unbending corrections to calculate springback.
The use of membrane element in the solution reduced CPUtime
considerably.
Though the analytical models are simple and easy to use for
parametric studies, these in general do not represent the real
condition. Numerical methods like FEM are frequently used to
solve the real life problems. Samual [9] provided an FE pro-
gram for predicting springback and sidewall curl whereas Lee
and Yang [10] evaluated quantitatively the numerical factors
inuencing the springback prediction. Effects of contact damp-
ing parameter, penalty parameter, blank element size, number
of corner elements and punch velocity were evaluated using the
Taguchi method and found that the number of corner elements
and the blank element size are the main factors inuencing the
springback prediction. It was concluded by them that kinematic
hardening should be included for more realistic simulations.
Gomes et al. [11] investigated the springback in simple U-
shape fromNumisheet-93 using three different material models;
Barlats yield criteria, Hills transverse anisotropic model and
von-Mises yield criteria. They found discrepancies between
the results obtained from the different material models. Ragai
et al. [12] studied the effect of sheet anisotropy on springback
in draw bending of stainless steel 410 numerical methods and
experiments.
Though many techniques were proposed, its prediction is still
a challenging issue. Many researchers have reported that poor
representation of Bauschinger effect is one of the main reasons
for poor springback prediction accuracy. In a recent study [13]
it was shown that the stress paths of a material point moving
over a radius can be quite different when using an isotropic or a
kinematic hardening model. The stress magnitudes are similar at
the beginning and the end of the stamping but the stress state is
actually quite different. Gau and Kinzel [14] proposed a bending
experiment to study the effect of Bauschinger effect on spring-
back in aluminium and concluded that the total strain method is
not sufcient to model springback when sheet undergoes cyclic
deformation. In another work Gau and Kinzel [15] proposed
a new hardening model in which they assumed that actuated
surface both translates and expands during deformation. All sur-
faces within the actuated surface have a rigid body translation as
described in Mroz method. Kim et al. [16] did the measurement
of anisotropy, Bauschinger effect and transient behaviour of
automotive dual phase steels. For the anisotropy measurement,
non-quadratic anisotropic yield function Yld2000-2d has been
utilized and its material parameters have been obtained using
the uni-axial tension tests as well as the hydraulic bulge test.
To measure the hardening behaviour including the Bauschinger
and transient behaviour, they proposed a new tension and com-
pression test. It was concluded that the Chaboche model well
represented the Bauschinger effect and the transient behaviour.
Most of the FE simulations used for sheet metal forming
prediction assume that springback is purely elastic and linear.
However, Cleveland and Ghosh [17] showed that the unload-
ing is not linear. It was demonstrated that the inelastic strain
released from the formed state can be a major source of addi-
tional strain recovery, the magnitude of which depends on the
forming stress state. It has been shown by them that 1020%
error in the estimation of springback is possible if the anelastic
effects during loading and unloading, is ignored. Such effects
arise from mobile dislocations which can move in response to
the internal repulsive forces.
Many works have so far been done on prediction and control
of the springback. However, there are a few literatures deal-
ing with the effects of material properties other than strength.
Material scientists are working towards increasing the normal
anisotropy by having a favourable texture for better formabil-
ity, and therefore, it is required to know the effect of normal
anisotropy on springback. In the present work the effect of nor-
mal anisotropy on the amount of springback is evaluated for
the benchmark problem of Numisheet-2005 [18]. An analyti-
cal model for plane-strain bending is also developed and used
for assessing the effect of various parameters, including normal
anisotropy, on springback. In the present analytical model radial
stress has been considered and the effective stress is not treated
as a constant value (unlike [6,7]). Also this model provides a
closed form solution.
2. Analysis
In the present model the material is assumed to be rigid plas-
tic and strain-hardening with normal anisotropy. Bauschinger
effect is neglected and Hollomans equation is used for mod-
elling the hardening. The neutral axis is assumed to coincide
with the geometrical mid plane of the sheet.
Fig. 1 shows the stresses acting on a small element in the
deforming sheet.
The force equilibriumequation for the small element making
an angle d at the centre is,
d
r
dr
=

r
r
(1)
Using the Hillss yield criteria and associated ow rule the
expressions for equivalent stress and strain can be written as
Fig. 1. State of stress on a small material element in pure bending.
302 R.K. Verma, A. Haldar / Journal of Materials Processing Technology 190 (2007) 300304
below:
d
eq
=
1 +

R

1 +2

R
d
r
= Cd
r
, C =
1 +

R

1 +2

R
(2)

r
= C
eq
(3)
where
eq
= K
n
eq
and for plane-strain condition d
r
=d

.
Using Eqs. (2) and (3) in (1) and integrating fromneutral line
to the outer most bre (2t is the sheet thickness),
_
d
r
=
_
t
r
C
eq
dr
r
=
_
t
r
KC
n+1
_
ln
_
R +r
R
__
n
dr
r
(4)
Using logarithmic series expansion and neglecting higher order
terms (as the ratio r/R is small), the above expression can be
written as
_
d
r
= KC
n+1
_
t
r
_
r
R

r
2
2R
2
_
n
dr
r
(5)
Using binomial series expansion in Eq. (5) and neglecting higher
order terms, the above expression can be written as
_
d
r
=KC
n+1
_
t
r
_
r
R
_
n
_
1
nr
2R
+
n(n 1)r
2
8R
2
_
dr
r
(6)
Integrating the above expression and using the limit
r
=0 at
r =t and
r
=
r
at r =r. Sheet thickness is 2t.

r
= C
n+1
_
1
R
_
n
K
_
r
n
t
n
n

n(r
n+1
t
n+1
)
2(n +1)R
+
n(n 1)(r
n+2
t
n+2
)
8(n +2)R
2
_
(7)
The moment can be calculated by integrating the following
equation:
M = 2w
_
t
0

r dr
= 2w
_
t
0
_

r
+C
n+1
K
_
ln
_
1 +
r
R
__
n
_
r dr (8)
I
1
=
_
t
0

r
r dr, I
2
=
_
t
0
C
n+1
K
_
ln
_
1 +
r
R
__
n
r dr
The above expressions can be simplied (using binomial and
logarithmic series expansions as above) as below,
I
1
+I
2
= C
n+1
K
t
n+2
2R
n
_
1
n +2

n
2(n +3)
t
R
+
n(n 1)
8(n +4)
t
2
R
2
_
(9)
As proposed in [6], the elastic strain recovery after unloading
causes the springback. On assuming that unloading moment M
having the same magnitude but opposite sign to that of applied
bending moment M. The unloading moment M is given as [6],
M =
2wEt
3
3(1
2
)
_
1
R

1
R

_
(10)
Table 1
Material properties used for the benchmark problem
Yield stress (MPa) 403.8
Stress constant (K) (MPa) 1040.4
Strain-hardening exponent (n) 0.159
Average normal anisotropy 0.62.0
Coefcient of friction 0.12
Here R is the radius before unloading and R* is the radius after
unloading. Equating Eqs. (8) and (10) the change in curvature
can be obtained as below:
_
1
R

1
R

_
=
3(1
2
)
Et
3
[I
1
+I
2
] (11)
where I
1
and I
2
are dened as above.
As R =R** the springback ratio / can be expressed by
Rx(1/R1/R*).
3. Results and discussions
Finite element simulation for the benchmark problem of
Numisheet-2005 [18] is carried out using the commercial soft-
ware PamStamp. Reasonfor selectingthis problemis tosimulate
a condition having a more realistic and severe strain path. The
default numerical parameters were used in the simulation. The
Hills material model withnormal anisotropyandplaner isotropy
is used in this simulation. The hardening is assumed to be
isotropic. The material properties used for this analysis is given
in Table 1. Fig. 2 shows the comparison of the present prediction
with the experimental values [18]. The springback is plotted for
Section 4 of the benchmark [18]. It can be seen that there is a
good agreement between the prediction and the experiment.
After this validation the same set-up (for simulations) is used
to predict the effect of normal anisotropy on springback. The
springback is taken at the location of maximum value. Fig. 3
shows the effect of normal anisotropy on springback. It is to be
notedthat withincrease inanisotropythe springbackamount rst
decreases and then increases. The analysis shows that spring-
back amount is the minimum for isotropic case (average normal
anisotropy equal to 1).
Fig. 2. Comparison between nite element prediction and the experiment for
the benchmark problem of Numisheet-2005 [18].
R.K. Verma, A. Haldar / Journal of Materials Processing Technology 190 (2007) 300304 303
Fig. 3. Effect of normal anisotropy on springback for the benchmark problem
of Numisheet-2005 [18].
To conrm the effect of normal anisotropy on the amount
of springback the above mathematical model was used. Fig. 4
shows the effect of normal anisotropy on springback for differ-
ent bend radii. This is consistent with the prediction of Liu [6];
however, it is observed that springback is monotonously increas-
ing with normal anisotropy and is not the minimumfor isotropic
material. The reason for the disagreement could be, either the
strain paths are not same and/or analytical model may not be
good enough for springback prediction.
Parametric studies are carried out to see the effect of
strain-hardening exponent, strength coefcient, bend radius and
sheet thickness. The effect of bend radius for different strain-
hardening exponents (on absolute change in curvature) is shown
in Fig. 5. It shows, with increase in bend radius springback
decreases. This happens because higher bend radius means
less severe bending and less plastic deformation. However, if
a graph is plotted between the percent changes in curvature
versus bend radius then the trend is reversed. The reason for
this reversed trend is the fact that for smaller total strain the
ratio of elastic to plastic strain is higher. Fig. 6 shows the effect
of sheet thickness whereas Fig. 7 shows the effect of strain-
hardening exponent on springback for different bend radii. With
increase in thickness and strain-hardening exponent springback
Fig. 4. Effect of normal anisotropy on springback. Using analytical model
( =662.0
0.25
, YS=197 MPa, t =1.0 mm).
Fig. 5. Effect of bend radius on springback. Using analytical model
( =662.0
0.25
, YS=197 MPa, t =1.0 mm,

R=1.5).
Fig. 6. Effect of sheet thickness on springback. Using analytical model
( =662.0
0.25
, YS=197 MPa,

R=1.5, R=5.0 mm).
Fig. 7. Effect of strain-hardening exponent on springback. Using analytical
model ( =662.0
0.25
, YS=197 MPa,

R=1.5, R=5.0 mm).
304 R.K. Verma, A. Haldar / Journal of Materials Processing Technology 190 (2007) 300304
decreases. It can be seen that for higher bend radius the effect is
more.
4. Conclusions
Researchers worldwide are trying to improve the anisotropy
by improving on the texture in the high strength steel sheets for
better formability. In the present work the effect of anisotropy on
springback amount is predicted for the benchmark problem of
Numisheet-2005 [18]. An analytical model is developed to cross
check the prediction from the nite element analysis. Both the
models predict that higher anisotropy is not good for springback.
Finite element analysis of the problem shows that springback is
the minimumfor an isotropic material, however, it does not agree
with the analytical model prediction for normal anisotropy less
than 1.
Acknowledgements
The authors gratefully acknowledge the permission fromTata
Steel Ltd. to publish the present work. They also thank Dr. N.
Gope for his constant support during this work.
References
[1] M. Sunseri, J. Cao, A.P. Karallis, M.C. Boyce, Accommodation of spring-
back in channel forming using active binder force control: numerical
simulations and experiments, J. Eng. Mater. Technol., Trans. ASME 118
(1996) 426435.
[2] R. Rufni, J. Cao, Using neural for springback minimization in a channel
forming process, J. Mater. Manuf. 107 (1998) 6573.
[3] J. Cao, B. Kinsey, S.A. Solla, Consistent and minimal springback using
stepped binder force trajectory and neural network control, J. Eng. Mater.
Tech., Trans. ASME 122 (1998) 6573.
[4] L.C. Zhang, Z. Lin, An analytical solution to springback of sheet metals
stamped by a rigid punch and an elastic die, J. Mater. Process. Technol. 63
(1997) 4954.
[5] F. Morestin, M. Boivin, C. Silva, Elasto plastic formulation using a kine-
matic hardening model for springback analysis in sheet metal forming, J.
Mater. Process. Technol. 56 (1996) 619630.
[6] D.K. Liu, A simplied approach for evaluating bendability and springback
in plastic bending of anisotropic sheet metals, J. Mater. Process. Technol.
66 (1997) 917.
[7] C. Wang, G. Kinzel, T. Altan, Mathematical modeling of plane-strain bend-
ing of sheet and plate, J. Mater. Process. Technol. 39 (1993) 279304.
[8] F. Pourboghrat, E. Chu, Prediction of springback and side-wall curl in 2-D
draw bending, J. Mater. Process. Technol. 50 (1995) 361374.
[9] M. Samual, Experimental and numerical prediction of springback and side
wall curl in U-bending of anisotropic sheet metals, J. Mater. Process. Tech-
nol. 105 (2000) 382393.
[10] S.W. Lee, D.Y. Yang, An inuence of numerical parameters inuencing
springback in explicit nite element analysis of sheet metal forming pro-
cess, J. Mater. Process. Technol. 80/81 (1998) 6067.
[11] C. Gomes, O. Onipede, M. Lovell, Investigation of springback in
high strength anisotropic steels, J. Mater. Process. Technol. 159 (2005)
9198.
[12] I. Ragai, D. Lazim, J.A. Nemes, Anisotropy and springback in draw-
bending of stainless steel 410: experimental and numerical study, J. Mater.
Process. Technol. 166 (2005) 116127.
[13] Gino A. Duffet, L. Neamtu, S. Oliveira, R. Weyler, The study
of various parameters in the numerical simulation of springback,
http://www.nafems.org (2004) 2829.
[14] J.T. Gau, G.L. Kinzel, An experimental investigation of the inuence of the
Bauschinger effect on springback prediction, J. Mater. Process. Technol.
108 (2001) 369375.
[15] J.T. Gau, G.L. Kinzel, A new model for springback prediction in which
the Bauschinger effect is considered, Int. J. Mech. Sci. 43 (2001) 1813
1832.
[16] D. Kim, M.G. Lee, C. Kim, M.L. Wenner, R.H. Wagoner, F. Barlat,
K.S. Chung, J.R. Youn, T.J. Kang, Measurement of anisotropy yielding,
Bauschinger and transient behaviour of automotive dual phase steel sheets,
Met. Mater. Int. 9 (2003) 561570.
[17] R.M. Cleveland, A.K. Ghosh, Inelastic effects on springback in metals, Int.
J. Plasticity 18 (2001) 769785.
[18] 2005 Numisheet Benchmark 2, Springback prediction of a cross member,
Proceedings of the 6th International Conference and Workshop on Numeri-
cal Simulation of 3DSheet Metal Forming Processes, Detroit, USA, August
1519, 2005.