Académique Documents
Professionnel Documents
Culture Documents
Springback Simulation
Numisheet 2008 September 1 - 5, 2008 – Interlaken, Switzerland
ABSTRACT: As a way to resolve the recent environmental and high energy cost issues, automotive
companies are exerting major efforts to reduce the weight of vehicles by replacing conventional steel parts
with light weight alloys and/or with advanced high strength steels (AHSS) such as dual-phase (DP),
twinning induced plasticity (TWIP) and transformation induced plasticity (TRIP) steels. The main objective
of this work is to experimentally and numerically evaluate the macro-performance of the automotive TWIP
sheet in conjunction with formability and spring-back. In order to characterize the mechanical properties, the
simple tension and tension-compression tests were performed for anisotropic properties as well as the
Bauschinger, transient and permanent softening behaviors during reverse loading. Formability verification
was performed for the simple tension and hemispherical dome stretching tests, while the 2D draw bending
test was carried out for springback verification. For numerical simulations, the anisotropic yield function
Yld2000-2d was utilized along with the combined isotropic-kinematic hardening law based on the modified
Chaboche model.
467
Numisheet 2008 September 1 - 5, 2008 – Interlaken, Switzerland
hemispherical dome tests as well as the 2D-draw Table 1: Yield stress and R-value
bending test were simulated utilizing the ABAQUS Dir. YS (MPa) R-value
Standard/Explicit codes along with their user RD 457 0.816
subroutines. 45D 465 1.188
TD 481 1.339
2 MATERIAL
CHARACTERIZATION 2.2 SIMPLE COMPRESSION TEST
Simple compression test was conducted with an
2.1 SIMPLE TENSION TEST
anti-buckling device. The result is compared with
Simple tension tests were carried out based on the tensile curve in Fig.3. Due to buckling, the
ASTM E8M with specimens prepared by the measured compressive strain range is limited but
milling process. Tensile speed was 0.05mm/s with the data shows that the compression curve
specimen thickness, 1.4mm. As shown in Fig.1, the converges to the tensile curve rapidly, confirming
maximum elongation is larger than 60% and the almost the same hardening behavior in the tensile
ultimate tensile strength (UTS) is close to and compressive hardening:
1,000Mpa, while the initial yielding is as low as
that of DP600 steel with around 400Mpa. The
earlier failure of the specimen prepared by water- 800
jet cutting (therefore, its surface is rougher than
that prepared with milling) confirms the ductility 600
listed in Table 1.
0
600
2.3 STRAIN RATE SENSITIVITY
400
Strain rate sensitivity was measured using the
200
Milling
simple tension test at four tensile speeds: 0.05, 0.5,
0
Water-jet 5, 50 mm/s. With 50mm gauge length, these
DP steel
correspond to (approximate) strain rates 0.001,
-0.2 0.0 0.2 0.4 0.6 0.8
0.01, 0.1 and 1.0/s, respectively. Based on the
Eng. Strain
measured hardening curves shown in Fig. 4, the
Figure 1: Simple tension curves of TWIP steel strain rate sensitivity (m value) was calculated, as
shown in Fig. 5, from
1800 ln(V i / V 0 )
m (1)
1600
Exp.
ln(Hi / H0 )
Effective Stress (MPa)
1400 Fitting
1200 where Hi and H0 (with 0.001/s) denote strain rate
1000 and the reference strain rate, respectively. Unlike
800
V tot 2362 (H 0.078) 0.68 other materials, the TWIP steel has the m value
600 varying from positive to negative as strain
400 increases, while its average m value is negative
200 with -0.007.
0
0.0 0.1 0.2 0.3 0.4 0.5 2.4 NOTCH TEST
Effective Plastic Strain In order to investigate the sensitivity of ductility to
Figure 2: Fitting curves for hardening surface condition, the V-shape notch test was
carried out based on ASTM E338-03 [3]. Here, the
V-shape notch is considered as an isolated
exaggerated surface defect in the rough surface that
468
Numisheet 2008 September 1 - 5, 2008 – Interlaken, Switzerland
800
transformations; i.e.,
600
1
400 1
) ½M
200
0.05mm/s f M
® ¾ V (2)
0.5mm/s
5mm/s ¯2¿
50mm/s
0
where V is the effective stress and
-0.1 0.0 0.1 0.2 0.3 0.4 0.5 0.6
Eng. Strain M M M
) SIc SIIc 2 SIIcc SIcc 2 SIcc SIIcc (3)
Figure 4: Hardening curves with different
strain rates
In Eq. (3), Skc and Skcc ( k I , II ) are the principal
0.02 values of the modified deviatoric stress tensor s
0.05 vs 0.5 mm/s ( s c or s cc ), in which
0.05 vs 5 mm/s
0.01 0.05 vs 50 mm/s
Average
s c Cc s Cc T ı Lc ı ,
0.00 (4)
m value
Table 2: Yield stress and ultimate tensile stress 2.5.2 Formulation of yield functions
YS (MPa) UTS (MPa) Based on the simple tension test results
TWIP 440 960 summarized in Table 1, anisotropic coefficients
DP600 358 620 were calculated for the yield functions: Yld2000-
2d, Hill’s 1948 [4] (using threee R-values) and
Table 3: Results of V-shape notch test Mises. As for Yld2000-2d, yielding under the
Max. stress (Mpa) Max. stress/UTS balanced biaxial condition was assumed to be the
TWIP DP TWIP DP same as that of the simple tension yielding in the
602.6 554.5 0.62 0.92 rolling direction. Also, the assumption, L12cc L21
cc ,
was imposed as well [1]. The anisotropy
coefficients of Yld2000-2d was obtained by the
469
Numisheet 2008 September 1 - 5, 2008 – Interlaken, Switzerland
1.06
h2 83 223 e (13)
1.04
1.02 whose integration leads
1.00
H H
D1 (H ) ³ h1 d H ³ 49269 (H 0.1) d H
0.2
0.98 (14)
0 20 40 60 80 100 0 0
H H
Angle from RD (Degree)
³ ³
11.99 H
D 2 (H ) h2 d H 83 223 e dH (15)
Figure 7: Normalized yield stress variation 0 0
1.4 1000
Yld2000 800
1.3
Hill'48 600
Mises
True Stress (MPa)
1.2
Exp. 400
R-value
1.1 200
1.0 0
-200
0.9
-400
0.8 -600
0.7 -800
0 20 40 60 80 100 -0.02 0.00 0.02 0.04 0.06 0.08 0.10 0.12 0.14 0.16 0.18
470
Numisheet 2008 September 1 - 5, 2008 – Interlaken, Switzerland
3 VERIFICATION 500
3.1 FORMABILITY Experiment
FEM
400
3.1.1 Simple tension test Failure with FLD
Failure without FLD
Punch load (kN)
471
Numisheet 2008 September 1 - 5, 2008 – Interlaken, Switzerland
Exp.
FEM (without softening)
6 REFERENCES
800 FEM (with softening)
600
[1] F. Barlat, J. C. Brem, J. W. Yoon, K. Chung,
R. E. Dic, S-H. Choi. F. Pourbograt, E. Chu
400
and D. J. Lege: Plane Stress Yield Function
True Stress (MPa)
iso-kine
30 L. Wenner and K. Chung: Numerical sheet
iso
20 forming simulation of friction stir welded
10 Exp.
TWB automotive sheets. GM Final Report,
FEM (without softening) 2006
0
FEM (with softening)
-10
-20 0 20 40 60 80 100 120
X Coord.
Figure 15: Shapes after springback in the 2D draw
bending test
472
Numisheet 2008 September 1 - 5, 2008 – Interlaken, Switzerland
____________________
*Corresponding author: 3-10-19 Minami-semba, Chuo-ku, Osaka, 542-0081, Japan, phone +81 6 6243 5001, fax +81 6
6243 4870, ma.ninshu@jri-sol.co.jp
473
Numisheet 2008 September 1 - 5, 2008 – Interlaken, Switzerland
Young㵭s modulus
correctly re-produced by Y-U model programmed
in LS-DYNA. Fig.3 shows an example of the
stress-strain cycles for the comparison between
experiment and Y-U model, and a good accuracy
is achieved.
equations.
D E D* (3a)
2
E m( bH p E H p ) (3b)
3
B R Y a p (a) S-rail part [7]
D * C[ (V D )H p D * H ] (3c)
Y D*
The expanding rate of boundary surface F is
defined by Eq.(3d)
pad S-rail
R m( Rsat R)H p (3d )
Where, Y , C , B, m, b, Rsat are material parameters
punch
holder
of Y-U model. Besides these 6 parameters, another
parameter h is used to express hardening (b) Stamping tools
stagnation [2].
474
Numisheet 2008 September 1 - 5, 2008 – Interlaken, Switzerland
475
Numisheet 2008 September 1 - 5, 2008 – Interlaken, Switzerland
ᵕᵎ
ᵣᶖᶎᵌ Vˆ x V x dz ,Vˆ y V y dz ,Wˆxy
ᵷᶍᶑᶆᶇᶂᵿᵋᵳᶃᶋᶍᶐᶇᴾᶋᶍᶂᶃᶊ
ᵔᵎ
ᵓᵎ
ᵏᵎ
d3 d6 equation (4a) some values such as 0, 1 or -1,
ᵎ
springback reasons can be easily analyzed by
ᵏ ᵐ ᵑ
ᵮᶍᶑᶇᶒᶇᶍᶌᴾᶌᶓᶋᶀᶃᶐ
ᵒ ᵓ ᵔ simulation system JSTAMP. It is called as
springback research function in JSTAMP.
Figure 8: Draw-in displacement at 6 points
measured and computed
5.2 SprINGBaCK DUe tO BeNDING
4.3 SprINGBaCK reSULtS StreSS aND MeMBraNe StreSS
After the draw-in results are verified, springback is Figs.10 (a) and (b) shows springback displacement
computed by releasing drawing stresses. The when bending stress V M and membrane stress V T
profiles of section-I before and after springback are are released separately. Springback due to bending
shown in Fig.9 by lines ̆and ---, respectively. stress V M has a larger deformation at left side as
The experimental result is also plotted in this figure shown in Fig.10 (a). However, the springback
by line ̆. Springback deformation, especially the
displacement due to membrane stress V T is larger
deformation on the wall of the right side, is exactly
predicted by Y-U model. at the right side as shown in Fig.10 (b).
12
11
10
9.0
8.0
7.0
I ICompensated part 6.0
5.0
4.0
3.0
2.0
1.0
0.0
476
Numisheet 2008 September 1 - 5, 2008 – Interlaken, Switzerland
477
Numisheet 2008 September 1 - 5, 2008 – Interlaken, Switzerland
9 reFereNCeS
[1] Singuang Xu, Kunmin Zhao, Terry Lanker,
Jimmy Zhang and C.T.Wang: springback
Prediction, Compensation and Correlation for
478
Numisheet 2008 September 1 - 5, 2008 – Interlaken, Switzerland
ABSTRACT: As has been frequently reported the Bauschinger effect is an important phenomenon which
should be considered in the simulation of sheet metal forming processes, in particular for springback
prediction. The majority of material models in commercial codes is restricted to describe this effect by
classical kinematic hardening. The physical nature of this phenomenon, namely dislocation dynamics, leads
to more complex behavior including early re-plastification and work hardening stagnation during reverse
loading. This paper investigates reverse loading phenomena for several typical sheet metal alloys used in the
automotive industry, and presents a new approach to model such effects. In addition to an accurate modeling
of the observed behavior, special emphasis is placed on the applicability of the model: the required material
parameters should be as few as possible, should be independent of each other, should have a well defined
physical meaning and should be universal for a certain material group, so that expensive material tests to
measure the parameters can be reduced to a minimum. Finally, springback simulations are presented and
compared to experimental data.
479
Numisheet 2008 September 1 - 5, 2008 – Interlaken, Switzerland
From physical point of view dislocation dynamics early re-plastification and transient softening is the
is responsible for plasticity and hardening of metals immediate motion of less stable dislocation struc-
during forward and reverse loading. The classical tures, such as pile-ups. Other factors which contri-
phenomenological assumption of the existence of bute to such material behavior on the macroscopic
an elastic domain with linear stress-strain response level could be change of the crystallographic
is only a simplification. In fact, during unloading texture during plastic deformation, stress induced
and reverse loading a superposition of elastic phase transformation or porosity evolution [2].
lattice deformation and reorganization of dis-
location structures takes place. Already at the very 3 CONCEPT OF THE MODEL
beginning of unloading, motion of less stable
dislocation structures such as pile-ups occur. This A novel approach to model reverse loading effects
leads macroscopically to early re-plastification and has been developed and implemented in the
transient softening [2][4]. Finally the work harde- commercial code AutoForm. To account for the
ning stagnation is attributed to the dissolution of industrial environment in which this code is typi-
dislocation cell-block boundaries [4]. cally used, special emphasis is placed on the
For an accurate stress prediction in the forming applicability of the model: it should be able to
simulation it is important to model not only tran- accurately describe the aforementioned phenome-
sient softening and hardening stagnation but also na, should be compatible to any of the imple-
early re-plastification, so that strain release during mented yield functions and hardening laws, should
unloading and springback can be accurately pre- not significantly increase computation time, and
dicted. should require as few as possible additional
The aim of the present work was to develop a material parameters. These should be independent
phenomenological model for reverse (and non- of each other, should have a well defined physical
proportional) loading of metal sheets, primarily to meaning and should be universal for a certain
improve springback simulation. Experimental re- material group, so that expensive material tests can
sults of tension-compression tests and hat-profiles be reduced to a minimum.
are taken from the EFB/AIF project [5], performed The detailed constitutive equations of the model
at the Fraunhofer Institute of Mechanics of are undisclosed research; in the following, the
Materials (IWM) at Freiburg, Germany. In the basic concepts of the model are presented.
paper at hand, results are presented for four steels
H220, DP600, TRIP700 and CPW800. 3.1 EARLY RE-PLASTIFICATION AND
TRANSIENT SOFTENING
2 EXPERIMENTAL RESULTS The main idea of the model is to use the same
developing equation to describe the two partial
Fig. 7 shows measured tension and tension-com- effects early re-plastification and transient softe-
pression curves, with the reverse loading curves in ning resulting in a smooth stress function for the
mirrored stress-strain representation. The afore- entire unloading and reverse loading path, inclu-
mentioned three partial effects during load reversal ding the area which is treated as elastic in con-
have been observed for each material. Fig. 5 ventional models. The function consists of a linear
presents according tangent moduli for the reverse and a non-linear part representing an initial tangent
dσ r modulus after the load reversal and its further
loading. The tangent modulus is defined as
dε r
(see Fig. 1) and is computed from the raw experi-
mental data by means of numerical differentiation.
One can clearly see the non-linear evolution of the
tangent modulus from the beginning of reverse
loading. Moreover, its initial value at the beginning
of reversal ( σ r = 0 ) is lower than the typical
value of Young’s modulus for steel.
According to the investigations in [1][2][3][6] the
tangent modulus at unloading even measured at
small reversal strains (e.g. vibrometrically) reduces
with increasing pre-strain. In the present paper, the
term elastic modulus, which is a physical material
constant, is not used, on purpose, because obvi-
ously a combination of elastic and plastic effects
takes place. As mentioned before a plausible
explanation for the reduction of initial tangent
modulus at the start of unloading as well as for Figure 2: Model concept
480
Numisheet 2008 September 1 - 5, 2008 – Interlaken, Switzerland
reduction. The model is formulated for the general reversal stress curve σ r (see Fig. 4). Typical
plane-stress condition, and is presented here sche- values range between 0.002 for materials with a
matically for the uniaxial case: small Bauschinger effect and 0.02 for a large
εr = εrl + εrn (1) effect.
2
where ε r is the reverse strain and ε rl and ε rn are 1.8
its linear and non-linear components (see Fig. 2). 1.6
K
[-]
The linear part, predominantly describing early re- 1.4
Normalized stress
1.2
modulus El : 1
σ r 0.8
In [1] and [6] it has been shown that the initial 0.2
El = E0 (1 − γ (1 − e − χp )) (3)
200
where E0 is the tangent modulus at p = 0 (equi-
valent to the Young’s modulus). γ is a material
[GPa]
parameter representing the maximal amount of 150
1.8 0
0 0.5 1 1.5 2
1.6
σr
Normalized stress [-]
[-]
σh
1.4
1.2
0.6
With this concept, the evolution of the tangent
0.4
modulus for the investigated materials can be well
0.2
described (see Fig. 5).
0 To incorporate this concept into the framework of
0 0.005 0.01 0.015 0.02 0.025 0.03
Strain ε r [-]
classical elasto-plasticity we abandon non-linearity
within the yield surface of the size a and use an
200
average secant modulus E s (see Fig. 2) instead:
a
Es = (5)
[GPa]
150
a
+ ε rn (a,σ h ( p), K )
El
Tangent modulus
100
Consequently, the tangent modulus for the elastic
part will be smaller than E 0 even if γ = 0 . This
50
is an important difference compared to other mo-
dels treating the load reversal by independent
0 elastic and plastic characteristics: The elastic
0 0.5 1 σr 1.5 2
Normalized stress
σh
[-] modulus here does not independently develop with
pre-strain as it does in other models. This reflects
Figure 3: Influence of the parameter El on the the aforementioned observation that dislocation
evolution of reversal stress and tangent modulus effects start at the very beginning of unloading and
cause a smooth stress curve over the entire un-
The non-linear part which is mostly responsible for
loading and reverse loading path. Consequently,
the transient softening is defined as:
the whole area of early re-plastification and transi-
εrn = εrn (σ r , σ h ( p ), K ) (4) ent softening including its evolution with pre-strain
where σ h ( p ) is the (isotropic) hardening stress can be described very accurately by just three
and K is a material parameter representing a physically well-defined material parameters ( γ , χ
typical strain distance affecting the steepness of the and K ).
481
Numisheet 2008 September 1 - 5, 2008 – Interlaken, Switzerland
100
experiment (small pre-strain) result, the stagnation including the stress reduction
50
experiment (large pre-strain) can be accurately described by just one additional
model (small pre-strain) material parameter.
model (large pre-strain)
0
In order to model delayed work hardening, p is
0 0.5 1 σ r 1.5 2 replaced by a new hardening parameter pd which
Normalized reversal stress [-]
σh behaves as follows: pd is identical with the
200 equivalent plastic strain p during proportional
(b) forward deformation and develops slower than p
[GPa]
100
0 0.5 1 1.5 2
σr 2
[-]
200
0
(d)
[GPa]
150 -1
-2
ξ =1
Tangent modulus
100 ξ = 0.5 -3
0
model (large pre-strain)
Figure 6: Influence of material parameter ξ on the
0 0.5 1 σr 1.5 2 work hardening during reverse loading
Normalized reversal stress [-]
σh
482
Numisheet 2008 September 1 - 5, 2008 – Interlaken, Switzerland
3 500
[-]
2.5 (a)
(a) 400
Stress [MPa]
300
1.5
experiment experiment
model 200
1 AutoForm
Yoshida-Uemori
0.5 100
0
0 0.1 0.2 0.3 0.4 0.5 0
Effective strain ε [-] 0 0.05 0.1 0.15 0.2 0.25 0.3 0.35
Effective strain ε [-]
2.5
[-]
1000
2
(b)
(b)
Effective normalized stress
800
1.5
Stress [MPa]
600
1 experiment
model
experiment
400
0.5 AutoForm
Yoshida-Uemori
200
0
0 0.05 0.1 0.15 0.2 0.25 0.3
Effective strain ε [-]
0
0 0.05 0.1 0.15 0.2 0.25
2.5 Effective strain ε [-]
[-]
1
to describe the reverse loading plasticity effects.
experiment
model
For the elasto-plastic effects during early re-plasti-
0.5 fication, it requires two material parameters γ and
0
χ , similar to the Yoshida-Uemori model.
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 Hardening curves were used in tabular form based
Effective strain ε [-]
on measured data and its extrapolation by Hockett-
1.5
Sherby hardening law.
Table 1: Material parameters
(d)
[-]
1.25
K, ξ, γ , χ, E0 ,
Effective normalized stress
1
Material [-] [-] [-] [-]
0.75 [GPa]
0.5
experiment
model
H220 0.004 0.62 0.245 30. 200.
DP600 0.011 1. 0.13 30. 200.
0.25
TRIP700 0.009 0.29 0.175 30. 200.
0 CPW800 0.005 0.52 0.1 30. 200.
0 0.02 0.04 0.06 0.08 0.1
Effective strain ε [-] SPCC 0.003 0.44 0.248 31. 206.
Figure 7: Measured and calculated tension- SPFC 0.013 0.83 0.117 61. 206.
compression curves.
(a) H220; (b) DP600; (c) TRIP700; (d) CPW800 4.2 SPRINGBACK SIMULATION
Using the presented material model fitted to
the data for the average Young’s modulus provided
tension-compression tests hat profiles with a sheet
in [1] for the range 0.75σ 0 ≤ σ ≤ 0.95σ 0 . Fig.
thickness of 1.5 mm and with two different die
8 shows also the curves obtained from Yoshida-
radii (5 mm and 2 mm) were simulated (for the
Uemori model taken from [7]. experimental set-up see [5]). The simulations were
Also here, one can see a very good agreement of run with AutoForm using shell elements with 11
the present model with the experiments. The model integration points over the thickness and a uniform
delivers similar results as the Yoshida-Uemori mesh with an element size of 1.3 mm for the large
model [7]. However, note that the present model die radius and with an element size of 0.6 mm for
uses standard hardening curves even in tabular the small die radius. These settings were
form, widely available in industry and which does determined with help of mesh convergence studies.
not need to be adapted, and only requires two A friction coefficient of 0.12 was used for all
additional material parameters K and ξ in order materials.
483
Numisheet 2008 September 1 - 5, 2008 – Interlaken, Switzerland
6 REFERENCES
[1] F. Yoshida, T. Uemori, K. Fujiwara: Elastic-
plastic behavior of steel sheets under in-plane
(d) cyclic tension-compression at large strain.
Int. J. Plasticity 18: 633-659, 2002.
[2] A. Krasovskyy: Verbesserte Vorhersage der
Rückfederung bei der Blechumformung durch
weiterentwickelte Werkstoffmodelle. Thesis
Fakultät für Maschinenbau der Universität
Karlsruhe, Germany, 2005.
[3] R.M. Cleveland, A.K. Ghosh: Inelastic effects
Figure 9: Measured and predicted springback for on springback in metals. Int. J. Plasticity 18:
hat-profile 769-785, 2002.
(a) H220; (b) DP600; (c) TRIP700; (d) CPW800 [4] B. Peeters, S.R. Kalidindi, C. Teodosiu, P.
In all cases the new model leads to an essential Van Houtte, E. Aernoudt: A theoretical
improvement of springback simulation results, in investigation of the influence of dislocation
particular for the wall curl prediction. For the small sheets on evolution of yield surfaces in single-
die radius, isotropic hardening always leads to curl- phase B.C.C. polycrystals. J. Mech. Phys.
in; the curl-out which is observed in reality can Solids, 50: 783-807, 2002.
only be achieved with kinematic hardening. [5] Beurteilung der Leistungsfähigkeit von
Materialmodellen zur
Blechumformsimulation. EFB/AIF-project
5 CONCLUSIONS
(AiF 13530 BG) report Nr. 244, 2005.
For an accurate springback simulation of sheet [6] S. Thibaud, N. Boudeau, J.C. Gelin: On the
metal forming processes it is important to model influence of the Young modulus evolution on
the material behavior during reverse loading the dynamic behaviour and springback of a
accurately. This involves not only the transient sheet metal forming component. In:
softening effect, but also work hardening stag- NUMISHEET, Vol.1: 149-154, 2002.
nation and early re-plastification effects in order to [7] F. Yoshida, T. Uemori: A model of large-
obtain an appropriate stress state during the strain cyclic plasticity describing the
forming simulation and to model unloading with Bauschinger effect and work hardening
the proper stress-strain response. A new material stagnation. Int. J. Plasticity 18: 661-686,
model has been presented (and implemented in the 2002.
code AutoForm) with the following advantages: [8] N. Ohno: A constitutive model of cyclic
• It can accurately describe tension-compression plasticity with a nonhardening strain region.
curves of different material classes, including Journal of Applied Mechanics 49: 721-727,
early re-plastification, transient softening, 1982.
work hardening stagnation and reduction of
tangent (“elastic” secant) modulus with pre-
strain.
484
Numisheet 2008 September 1 - 5, 2008 – Interlaken, Switzerland
ABSTRACT: The main purpose of this work is to predict the springback of Numisheet’05 Benchmark#3
using different material models. This Benchmark was carried out with different sheet materials (AKDQ-
HDG, HSLA-HDG, DP600-HDG and AA-6022-T43) and with drawbead penetrations that varied from 25%
to 100%. Simulations were performed using Hill's 1948 anisotropic yield function and two types of
hardening models: isotropic hardening and combined isotropic-nonlinear kinematic hardening. The work
hardening behaviour of the aluminium was described with the Voce model and that of the steel with
Hollomon’s power law. Kinematic hardening was modelled using the Armstrong-Fredrick nonlinear
kinematic hardening model which considers cyclic deformation phenomena such as the Bauschinger effect
and yield stress saturation. Comparisons between simulation results and experimental data show that none of
these models can individually predict the springback of sheet metal parts.
____________________
* Corresponding author: 224a, Essex Hall, Mechanical Eng. Dept., University of Windsor, 401 Sunset Ave., Windsor,
Ontario, Canada. Tel: 1-519-2533000-Ext 4854. Email: ghaei@uwindsor.ca
485
Numisheet 2008 September 1 - 5, 2008 – Interlaken, Switzerland
al [1]. After careful study of element type (shell, was used to develop a user-defined subroutine
solid), mesh size and integration scheme, the material model based on Hill and NKH for both
present authors found that the following conditions ABAQUS-Explicit (VUMAT) and ABAQUS-
yielded the best results: Standard (UMAT). Cyclic tension-compression
• Solid plane-strain elements (denoted by tests were carried out on the specimens made by
CPE4R in ABAQUS) DP600, AKDQ and HSLA. The stress-strain curve
• A mesh consisting of 4 elements through the presented in reference [18] was used to evaluate
sheet thickness and 1000 elements along the the mechanical properties of AA-6022. The least-
length (only one half of the channel section squares method was used to fit curves to the
was modeled due to symmetry) experimental data and Table 1 shows the
• An explicit-implicit scheme for forming and mechanical properties and hardening parameters
springback stages, respectively used in the simulation. It should be noted that the
transverse direction of the sheet was taken as the
For processes in which the sheet metal is subjected reference direction for defining the hardening
to relatively linear loading paths (such as constants (C, , Q, b) as the part was drawn in this
hemispherical punch stretching and cup-drawing direction during the forming stage.
tests), the simple assumption of isotropic hardening
may provide good springback predictions.
However, the results of simulations for processes Upper Die
Drawbead
in which sheet metal undergoes cyclic loading
(such as alternative tension-compression or
bending-unbending stresses in bending-drawing
tests or deep drawing tests including drawbeads),
this assumption should be relaxed to account for Binder Punch
the kinematic hardening effects associated with
stress reversal. Especially for springback analysis,
modeling of the Bauschinger effect and cyclic
hardening characteristics of materials is of vital
importance (it is well-known that high strength
steel and aluminium sheets exhibit more Figure 1: Sketch of the A/SP channel draw die with
springback than ordinary mild steel sheets) [14]. In location of drawbead inserts (side-view)[11]
an experimental study, Gau and Kinzel investigated
the influence of the Bauschinger effect on
springback predictions [9].
In the present study, Hill’s quadratic yield function
(1948) was used as the yield criterion and two
basic hardening models were employed to define
the evolution of yield surface: isotropic hardening
and a combined isotropic-nonlinear kinematic
model (NKH) proposed by Fredrick and Armstrong Figure 2: A channel formed in the A/SP channel
[15]. In this model, the evolution for the location draw die after springback
and size of the yield surface is defined respectively
as:
3 RESULTS
2
dα = Cdε p − γαdp
3 (1) A series of 16 simulations were carried out: 8 with
isotropic hardening and 8 with nonlinear kinematic
dr = b.(Q − r ).dp
(2) hardening (NKH). Simulation and experimental
where denotes the backstress, C and are results are compared in this section. It is observed,
material constants associated with nonlinear both experimentally and numerically, that the
kinematic hardening and Q and b are material springback decreases when the drawbead
constants associated with isotropic hardening. The penetration depth increases. That is, as the plastic
forming stage was first analysed using ABAQUS- deformation in the drawbeads increases,
Explicit and then the results were imported to springback reduces. In the deeper drawbead
ABAQUS-Standard to simulate the springback penetrations, the restraining force is higher
stage and obtain the final configuration of the part such that the sheet metal is stretched to a greater
at equilibrium. Unfortunately, ABAQUS-Standard extent of tension in the sidewall region after
still does not support the combined use of Hill and passing over the die shoulder. The higher tensile
NKH models in its library of material models. stresses help to decrease the residual stress gradient
Therefore, the so-called return mapping algorithm through the sheet thickness.
486
Numisheet 2008 September 1 - 5, 2008 – Interlaken, Switzerland
In order to measure the curvature in the channel is shown in Figures 5 and 6. Neither the isotropic
sidewalls (i.e. after springback), 3 or 5 channel or the NKH models are capable of predicting
sections for each prestrain condition were scanned springback accurately. Although it would be
using IRDI’s Virtek LaserQC™ 2D laser scanner. expected that the isotropic hardening should
Each channel section was placed on its edge on the overestimate the springback, it was found to
glass surface of the LaserQC™ very slowly in underestimate it. This may be due to the fact that a
order to avoid applying any constraints to the possible decrease in elastic modulus during
channel and thereby distorting its natural shape. unloading was not accounted for in this work.
The glass surface was also sufficiently slippery that Some researchers have recently indicated that the
the part would find its natural equilibrium. Channel elastic unloading modulus decreases with plastic
sections were positioned in such a way that the deformation [7, 16, 17]. The effect of this
laser could scan the edge of the RHS sidewall in phenomenon is taken into account only for DP in
contact with the glass without being obstructed. the next section.
The scanning accuracy of the LaserQC™ is up to AKDQ-25%-Iso AKDQ-25%-NKH Exp 25%
0.05 mm. The interested reader is referred to ref. 250
penetration where the range of cyclic strain is Figure 3: Profile of the sidewall for AKDQ at 25%
smaller than for 100% penetration. NKH divides penetration
the work hardening of the material into expansion
and translation of the yield surface in stress space. Figures 7 and 8 show profiles of the HSLA channel
However, the isotropic hardening assumes only an sidewall at 25 and 100% penetrations. Again, the
expansion of the yield surface due to global work isotropic hardening assumption underestimates the
hardening of the material. Consequently, it does springback. Neither the isotropic nor the NKH
not take the Bauschinger effect into account and is models are capable of predicting springback
expected to overestimate springback. accurately, although the general shape of sidewall
A comparison between the experimental profiles profile is correctly predicted with NKH. The
with those obtained by simulation at different decrease in elastic modulus may be required to
drawbead penetrations for DP600 channel sections obtain a better fit with the experimental profile.
However, due to a lack of experimental data this
phenomenon was not considered here.
487
Numisheet 2008 September 1 - 5, 2008 – Interlaken, Switzerland
0 200
0 50 100 150 200 250 300 350
X, mm 150
Y, mm
penetration 50
150
Figure 7: Profile of the sidewall for HSLA at 25%
drawbead penetration
Y, mm
100
HSLA-100%-Iso HSLA-100%-NKH Exp-HSLA 100%
250
50
200
0
0 50 100 150 200 250 300 350
150
X, mm
Y, mm
penetration
50
488
Numisheet 2008 September 1 - 5, 2008 – Interlaken, Switzerland
Figures 12 and 13 show the profile of channel DP-75%-Iso DP-75%-NKH Exp-DP 75%
Y, mm
the curve accurately at 25%. However, the NKH is
100
more capable of predicting springback compared to
isotropic hardening. This model has predicted the
50
radius of sidewall curl more accurately at 25%
penetration and the sidewall angle more precisely 0
75% penetration. 0 50 100 150 200 250 300 350
X, mm
AL-100%-NKH AL-100%-Iso Exp-Al 100%
250 Figure 13: Profile of the sidewall for DP at 75%
200
penetration considering decrease of unloading
elastic modulus
150
Y, mm
100 4 CONCLUSIONS
50 Numisheet’05 Benchmark#3 was simulated using
0
nonlinear FEA code ABAQUS. The channel draw
0 50 100 150 200 250 300 350 die tests were conducted with four different sheet
X, mm
materials (3 grades of steel and an aluminium
Figure 10: Profile of the sidewall for AL at 100% alloy). Two different drawbead penetrations were
drawbead penetration also applied in these simulations for each material.
Simulations were carried out using two hardening
models (isotropic and combined isotropic-
Section 1 Section 2 Section 3 Section 4
0.26
nonlinear kinematic). For both cases, Hill’s 1948
0.24 anisotropic yield function was used. A comparison
0.22
0.20
between predicted and experimental sidewall
Equivalent plastic strain
200
5 ACKNOWLEDGEMENT
Professor R. H. Wagoner and Dr. P. Raghupathy
150 from Ohio State University are gratefully
Y, mm
489
Numisheet 2008 September 1 - 5, 2008 – Interlaken, Switzerland
[2] Song N., Qian D., Cao J., Liu W. K., Li Proceedings IDDRG’06, edited by Santos,
S., Effective Models for Prediction of A.D., and Barata da Rocha, A., INEGI,
Springback In Flanging, ASME Journal Porto, 2006, pp.559-566.
of Engineering Materials and Technology [13] Shi M.F., Huang M., Specification for
123:456–461,2001. Benchmark Materials, Proceedings of
[3] Gan W., Wagoner R.H., Die design Numisheet 2005, edited by J. Cao, M. F.
method for sheet springback, International Shi, T. B. Stoughton, C.-T. Wang, and L.
Journal of Mechanical Sciences 46:1097– Zhang, American Institute of Physics,
1113,2004. CP778 Volume B, pp. 1173-1178, 2005.
[4] Carden W.D., Geng L.M., Matlock D.K., [14] Moreira L.P., Ferron G., Influence of the
Wagoner R.H., Measurement of plasticity model in sheet metal forming
springback, International Journal of simulations, Journal of Materials
Mechanical Sciences 44:79–101,2002. Processing Technology, 156:1596–1603,
[5] Chung K., Lee M., Kim D., Kim C., 2004.
Wenner M.L., Barlat F., Springback [15] Armstrong P.J, Frederick C.O, A
evaluation of automotive sheets based on mathematical representation of the
isotropic-kinematic hardening laws and multiaxial Bauschinger effect, Technical
non-quadratic anisotropic yield functions Report RD/B/N 731, Central Electricity
Part I: theory and formulation, Generating Board, 1966.
International Journal of Plasticity 21:861– [16] Levy B.S., Van Tyne C.J., Moon Y.H.,
882,2005. Mikalsen C., The Effective Unloading
[6] Choi Y., Han C., Lee J.K., Wagoner R.H., Modulus for Automotive Sheet Steels, In:
Modeling multi-axial deformation of Proceeding of SAE 2006 World Congress
planar anisotropic elasto-plastic & Exhibition, April 2006, Detroit, MI,
materials, part I: Theory, International USA.
Journal of Plasticity 22:1745–1764,2006. [17] Yang M., Akiyama Y., Sasaki T.,
[7] Yoshida F., Uemori T., A model of large- Evaluation of change in material
strain cyclic plasticity describing the properties due to plastic deformation,
Bauschinger effect and workhardening Journal of Materials Processing
stagnation, International Journal of Technology 151:232–236,2004.
Plasticity 18:661–686,2002. [18] Zhao K. and Lee J.K., Generation of
[8] Chun B.K., Kim H.Y., Lee J.K., Modeling Cyclic Stress-Strain Curves for Sheet
the Bauschinger effect for sheet metals, Metals, ASME Journal of Engineering
part I: theory, International Journal of Materials and Technology, 123, 2001,
Plasticity 18:571–595,2002. pp.391-397
[9] Gau J., Kinzel G.L., A new model for
springback prediction in which the
Bauschinger effect is considered,
International Journal of Mechanical
Sciences 43:1813–1832,2001.
[10] Stoughton T.B., Green D.E., Iadicola M.,
Specification for BM3: Two-stage
Channel/Cup Draw, In: Proceedings of
Numisheet 2005, edited by J. Cao, M. F.
Shi, T. B. Stoughton, C.-T. Wang, and L.
Zhang, American Institute of Physics,
CP778 Volume B, pp. 1157-1172, 2005.
[11] Green D.E., Description of Numisheet
2005 Benchmark #3 Stage-1: Channel
Draw with 75% Drawbead Penetration,
In: Proceedings of Numisheet 2005,
edited by J. Cao, M. F. Shi, T. B.
Stoughton, C.-T. Wang, and L. Zhang,
American Institute of Physics, CP778
Volume B, pp. 894-904.
[12] Green D.E., Stoughton T.B., Gnaeupel-
Herold T., Iadicola M.A., Foecke T.,
Influence of drawbeads in deep drawing
of plane-strain channel sections, In:
490
Numisheet 2008 September 1 - 5, 2008 – Interlaken, Switzerland
aBStraCt: The origin of springback lies in the elastic recovery of materials after forming operations
and this phenomenon is still a serious problem in sheet metal forming. Springback modifies the final shape of
the part and increases manufacturing costs. This study deals with experimental and numerical evaluation of
springback in an AA5754-O aluminum alloy performed with a split-ring test at several temperatures. A ring is
cut from the wall of a drawn cup. Residual stresses induced in the cup during manufacturing are then allowed
to relax by the ring splitting and create the opening of the ring. Experimental results are compared to predicted
springback simulations using the finite element code Abaqus. The whole deep drawing process of a semi-blank
is simulated and numerical splitting of the ring is performed. 3D solid elements with reduced integration are
used for the calculations. Several material models are analyzed, all model using isotropic and kinematic
hardening and any of the following plasticity criteria : Von Mises, Hill’48 and Barlat’91 yield criterion which
has been implemented in Abaqus using a UMAT subroutine. Main observed data are force-displacement
curves during forming, cup thickness according to material orientations and ring gap after splitting. Influence
of temperature during forming on springback is also studied.
KeYWOrDS: sheet metal forming, springback prediction, material modeling, aluminum alloy, influence
of temperature.
491
Numisheet 2008 September 1 - 5, 2008 – Interlaken, Switzerland
100
Due to the Portevin Le Chatelier (PLC) effect which
decreases measurement accuracy, the elastic modu-
0 lus is measured as the average value during loading
and unloading. However, the elastic modulus varia-
−100
tion has not yet been applied to the model proposed
below.
−200
−0.4 −0.3 −0.2 −0.1 0 0.1 0.2 0.3 0.4
2.1.1 Isotropic and kinematic hardening
Strain The isotropic hardening part of the model depends
(b)
300 on the viscoplastic deformation εvp and its evolution
is given by the Hocket-Sherby equation :
250 UT − 00 exp
UT − 00 sim
MS − 00 exp
200 R = B − (B − A) exp − C(εvp )n (2)
� �
MS − 00 sim
CS − 10 exp
150 CS − 10 sim
CS − 20 exp
CS − 20 sim where (B − A) is the isotropic hardening range, and
Stress (MPa)
100 CS − 30 exp
CS − 30 sim C is the saturation speed. A = σy is the yield stress.
50
The non-linear kinematic hardening is described by
0 the Armstrong-Frederick law with a linear Prager
−50 type contribution :
−100
3�
Cp α + Hp εvp where α̇ = ε̇vp − γ ṗα
�
−150 X =
2
−200 (3)
−0.4 −0.3 −0.2 −0.1 0 0.1 0.2 0.3 0.4
Strain Cp represents the kinematic hardening range, γ is
the saturation speed and Hp is the linear kinematic
Figure 1: Stress-strain database of AA5754-O : hardening modulus. ṗ is defined by :
tensile (UT), monotonic (MS) and cyclic shear (CS)
tests. (a) Experimental, (b) Identification for Barlat 91
�
2 vp vp
yield criterion ṗ = ε̇ : ε̇ (4)
3
The identification of material parameters was per- 2.1.2 Yield criterion
formed from the tensile and simple shear tests using Three plasticity yield criteria are considered : one
SiDoLo software [6]. isotropic of Von Mises [7] and two anisotropic crite-
The elastic modulus variation has also been obtained ria of Hill 48 and Barlat 91.
by performing cyclic loading-unloading tensile tests Young modulus and Poisson’s ratio are fixed to
(see Fig.2). An equation expressing the modulus vs. E0 = 74620 M P a and ν = 0.33 respectively.
492
Numisheet 2008 September 1 - 5, 2008 – Interlaken, Switzerland
Identified parameters of all behavior laws and crite- Rings are cut 15 mm from the bottom of the cup us-
rion parameters are presented in table 1. Simulated ing machine tool. The rings are 20 mm high and
curves show good agreement on anisotropy for ten- had an internal diameter of 100 mm before split-
sile and shear test simulations and the Bauschinger ting. Experimental springback data consist of rings
effect is well described by the Barlat 91 yield crite- gap measurement along straight lines connecting the
rion (see Fig.1 (b)). two ends of the split rings. 8 springback tests have
been performed. Opening gap after splitting is about
2.2 eXperIMeNtaL DeVICe 64 mm-wide (± 3 mm) at room temperature.
The experimental device has been developed in or-
der to perform reverse re-drawing of cylinder cups
[10]. However, only the device designed for the first
stage is used in this study (see Fig.3). Tool geometry
is given in table 2. Blanks are 170 mm in diameter
and 1 mm in thickness.
A blank-holder adjusted on the punch diameter and Figure 4: Experimental drawn cup, cut ring and
allowing the force control is maintained by 8 screws springback after splitting
with Belleville washers stacked in both parallel and
series. Force washers are used to control the blank- It is interesting to note that the shape of the ring af-
holder pressure at the beginning and during all the ter splitting is slightly conical due to the variation of
drawing stage. thickness and the non-symmetric stress distribution
Experiments were carried out on a classical tensile in the wall of the ring.
test machine of 500 kN maximum load capacity.
Blank-holder force is imposed at 24 kN and draw-
ing speed is 0.5 mm.s−1 . Blanks are lubricated on
493
Numisheet 2008 September 1 - 5, 2008 – Interlaken, Switzerland
494
Numisheet 2008 September 1 - 5, 2008 – Interlaken, Switzerland
100 300
25°C
100°C
150°C
50 250 200°C
0 200
Stress (MPa)
Stress (MPa)
−100 100
−150 50
−200 0
0 10 20 30 40 50 60 70 80 90 0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18
Angle from Rolling Direction (°) Strain
Figure 7: Hoop and axial stress evolution in the mid Figure 8: Tensile stress-strain curves of AA5754-O
section of the cup wall for several positions from RD for several temperatures at 2.10−3 s−1
- Barlat 91 yield criteria
perfomed in a heating furnace, for a temperature
Table 3: Springback values (in mm) for several ma-
range from 25◦ C to 200◦C. Tools dimensions are
terial models
the same as those given in table 2, except for the
blank-holder opening which is 104.5 mm in diame-
Experimental 64 ± 3 mm ter due to thermal dilatation of tools. Temperature
Hardening isotropic mixed is the same for all tools and is maintained constant
Von Mises 85 74 during all the forming process. Force-displacement
Hill 48 61.5 49 curves are proposed on Fig.9.
Barlat 91 95 85
60
495
Numisheet 2008 September 1 - 5, 2008 – Interlaken, Switzerland
6 aCKNOWLeDGeMeNt
496
497
498
ρ
θ
∆
∆ ∆ ∆
499
500
501
Numisheet 2008 September 1 - 5, 2008 – Interlaken, Switzerland
ABSTRACT: Nowadays, springback prediction for mild steels is an established technique which is
successfully applied in the automotive industry. The rising importance of light weight constructions in car
body design lead to the development of advanced high strength steels such as dual phase, TRIP or TWIP
steels. The application of these steels in automotive products leads to the demand of predicting springback of
such steel grades with the same accuracy as of mild deep drawing grades, which still is a challenging task.
This article deals with the current approaches and proceedings in the field of springback prediction of
advanced high strength steels. The springback is investigated using a principle deep drawing part. The part is
drawn, the final geometry is digitalized. The deep drawing process is then modeled in the dynamic explicit
part of the FEM-code LS-DYNA, following an implicit springback analysis. The predicted part geometry is
compared to the experimentally gathered geometry and the differences are evaluated. Special attention is
spent here to the influence of different hardening models and the used yield loci. Investigated are
conventional isotropic hardening, isotropic kinematic hardening and anisotropic or directional hardening.
These hardening models are each combined with yield loci models of different complexity, starting with the
simple von Mises yield locus leading to complex models like actual Barlat yield loci.
503
Numisheet 2008 September 1 - 5, 2008 – Interlaken, Switzerland
but some effects, especially in modern high A more complex model for the Bauschinger effect
strength steels, are not covered by this hardening is provided by the user material model MF GenYld
model. These steels, especially those with of the company MATFEM. It’s so called Isokin
microstructural changes during forming, can model is very similar to the Chaboche model and
exhibit a severe anisotropy in hardening depending defines the back stress through the following
on the stress state. To take this effect into account, equation [8]:
the shape of the yield surface has to be depending
on the amount of strain and the nature of the
applied stress. The consequences of anisotropic Xx =
C1
γ1
(
⋅ 1− e
−γ 1⋅ε eq
) + Cγ ⋅ (1 − e
2 −γ 2 ⋅ε eq
) (1)
hardening on the strain distribution of deep drawn 2
parts were evaluated earlier by the authors [6]. The
Bauschinger effect is another hardening Xx is the experimental back stress in uniaxial
phenomenon, which is not covered by a simple reversal and C1, C2, γ1 and γ2 are material
isotropic hardening model. The Bauschinger effect parameters which have to be determined. As shown
occurs when the direction of an applied load in figure 1, this model allows a more accurate
changes and is modeled by a transition of the yield description of the material behavior in the case of a
locus during forming which is usually called load reversal. A drawback are the long calculation
kinematic hardening. times which are not primarily due to the hardening
Since the aim of this paper is to evaluate model, but rather because this is a user material
commercial models that are available in model and though not as fast as a native LS-DYNA
commercial FEM codes and so applicable in material model.
industrial use, only yield loci and hardening
models which are available in LS-DYNA are 2 EXPERIMENTAL AND
focused. In LS-DYNA, different kinematic NUMERICAL SETUP
hardening laws are implemented but they are
usually combined with a v. Mises yield locus what 2.1 EXPERIMENTAL
makes them inadequate for the simulation of For this article, three different steel grades were
anisotropic sheet metals. There are two practical chosen. As a reference, a low alloyed H340LAD, a
options which overcome this limitation. The dual phase steel DP600 and to represent materials
implementation of the Barlat 2000 yield locus with strain induced phase transitions a TRIP
features a very simple model that accounts for assisted high strength steel TRIP700.
kinematic hardening. A parameter β is created, To evaluate the springback behavior
which defines the drop of the current yield stress in experimentally deep drawing experiments were
the case of a load reversal (see figure 1). performed using a cruciform punch as it was used
in [9] before. The shape of the blank was chosen in
GenYld_Yld2k DYNA_Yld2k, BETA 0.1 a way that the final part is not a closed cross but
0.8 forms a u-shaped profile (see figure 2).
0.7
effective stress / GPa
0.6
0.5
0.4
0.3
0.2
tension compression
0.1
load reversal
0
0 0.05 0.1 0.15 0.2 0.25 0.3
effective plastic strain / -
Figure 2: drawn part and location of the section cut
Figure 1: comparison of different kinematic
for springback determination
hardening models
This shape was chosen to reduce the stiffness of the
The main advantage of this model is it’s simplicity
final part and increase the springback. A cruciform
since it comes along with only one parameter and
punch instead of straight u-shaped profile was used
provides short calculation times. The parameter β
to get more complex stress states within the final
is very similar to the back stress component in the
part. The produced specimens have been
Chaboche model [7]. But in contrast to the
digitalized without further cutting operations. In
Chaboche model, the parameter β is constant and this way, the experimental achieved final geometry
not a function of the plastic strain. This is a severe was available in a digital form as STL-data and so
simplification of the real material behavior. This could be treated in the post processing in the same
simple model also can’t deal with effects like early way than the numerical results.
replastification or work hardening stagnation.
504
Numisheet 2008 September 1 - 5, 2008 – Interlaken, Switzerland
total deviation / mm
minimize the influence of numerical effects to the 0.6
springback results and to maximize the 0.5
comparability between the results. 0.4
At the time the simulations were done, LS-DYNA 0.3
did not support free springback without nodal 0.2
constraints, so a set of boundary conditions under 0.1
utilization of symmetry conditions was chosen 0
which matches a free springback as close as 0 20 40 60 80 100 120 140 160
treated length / mm
possible (figure 3).
Figure 4: comparison of different yield loci in
combination with isotropic hardening for H340LAD
1
Yld2k BETA0 Yld2k BETA 0.055
0.9 Yld2k BETA 0.075 Yld2k BETA 0.095
0.8
0.7
total deviation / mm
0.6
0.5
0.4
0.3
0.2
0.1
0
0 20 40 60 80 100 120 140 160
0.8
0.6
von Mises Hill 48 Barlat 89 Barlat 2000
isotropic x x x x 0.5
anisotropic 0.4
x x
(MF GenYld) 0.3
simple kinematic 0.2
x
(MAT 133)
0.1
Isokin
x x 0
(MF GenYld)
0 20 40 60 80 100 120 140 160
treated length / mm
505
Numisheet 2008 September 1 - 5, 2008 – Interlaken, Switzerland
model gives the biggest deviation from the sophisticated yield loci leads to a substantial
measurement in combination with the Hill 48 yield improvement of the predicted springback. The best
locus, while the results for this yield surface in result is gained using the classical combination of a
combination with simple isotropic hardening are Hill 48 yield surface and an isotropic hardening
comparable to that of the Yld2k-Isokin rule and again the simple kinematic model
combination. The effect of the simple kinematic implemented in MAT 133 has no significant effect.
hardening model of Material type 133 is very small
and does not lead to an improvement of the result. 3.3 TRIP700
1.4 1.8
MISES Hill48 iso Yld89 iso Yld2k iso
1.6
total deviation / mm
1.2
1.4
1
1.2
total deviation / mm
0.8 1
0.6 0.8
0.6
0.4
0.4
0.2 0 20 40 60 80 100 120 140 160
treated length / mm
0
0 20 40 60 80 100 120 140 160 Figure 10: comparison of different yield loci in
treated length / mm
combination with isotropic hardening for TRIP700
Figure 7: comparison of different yield loci in
combination with isotropic hardening for DP600 1.8
Yld2k BETA 0.00 Yld2k BETA 0.10
Yld2k BETA 0.13 Yld2k BETA 0.016
1.6
1.2
Yld2k BETA 0.00 Yld2k BETA 0.095
Yld2k BETA 0.116 Ylad2k BETA 0.138 1.4
total deviation / mm
1
1.2
0.8
total deviation / mm
0.6 0.8
0.4 0.6
0.4
0.2
0 20 40 60 80 100 120 140 160
treated length / mm
0
0 20 40 60 80 100 120 140 160 Figure 11: effect of parameter β of LS-DYNA
treated length / mm
Material 133 for TRIP700
Figure 8: effect of parameter β of LS-DYNA
Material 133 for DP600 2
Hill48 iso Hill48 aniso Hill48 iso-kin
1.8 Yld2k iso Yld2k iso-kin
2
Hill48 iso Hill48 aniso Hill48 iso-kin 1.6
1.8 Yld2k iso Yld2k iso-kin
1.4
total deviation / mm
1.6
1.2
1.4
total deviation / mm
1
1.2
0.8
1
0.8 0.6
0.6 0.4
0.4 0.2
0 20 40 60 80 100 120 140 160
0.2
treated length / mm
0
0 20 40 60 80 100 120 140 160 Figure 12: comparison of isotropic hardening and
treated length / mm
the Isokin model combined with Hill48 and Barlat
Figure 9: comparison of isotropic hardening and 2000 yield loci for TRIP700
the Isokin model combined with Hill48 and Barlat
2000 yield loci for DP600 The results for the TRIP700 steel grade are in
principal very similar to the ones from the
For the DP600 steel, neither the use of the H340LAD, but a higher overall level of deviation
kinematic hardening models, nor of the between measurement and simulation. Again the
506
Numisheet 2008 September 1 - 5, 2008 – Interlaken, Switzerland
Isokin hardening model combined with the Barlat more competitive when using fully austenitic
2000 yield surface gives the best results and again stainless steels or TWIP-steels, since these types of
the combination of the Hill 48 yield locus with the materials are showing a strong dependency of their
kinematic model is not accompanied by an hardening behavior from the current stress state.
improvement of the result while this yield function The final conclusion is that against the background
in combination with isotropic hardening leads to of the here presented results there is no general
satisfying results. recommendation for the use of kinematic
hardening models. The user has to decide based on
4 CONCLUSIONS the used material, the geometry of the formed part
and the requirements regarding the accuracy of the
As expected the accuracy of the spring back springback prediction whether the cost of the
prediction decreases with increasing strength of the kinematic hardening model is justified or not. The
material and the complexity of the microstructure. employment of a Hill 48 yield locus and an
It points out, that for none of the investigated steel isotropic hardening rule leads to results which
grades the use of kinematic hardening model can satisfy the needs of most industrial applications.
be recommended in general. The first conclusion is
that to take into account the kinematic hardening
5 REFERENCES
one has to use a model of adequate complexity.
The simple dropdown of the current yield stress by [1] Hill, R.: A theory of the yielding and plastic
the parameter β, which is done by the LS-DYNA flow of anisotropic metals. Proc. Roy. Soc.
Material 133 is useless to counterproductive. The A193, 281-297, 1948.
investigations showed that in this case it is better to [2] Barlat, F., Lian, J.: Plastic behavior and
use a simple yield locus with a simple isotropic stretchability of sheet metals. Part I: A yield
hardening, since this combination leads to better function for orthotropic sheets under plane
results than an inadequate description of the stress conditions. International Journal of
kinematic hardening. In contrast to this, the Plasticity 5, 51-66, 1989.
application of the Isokin model available in MF [3] Barlat, F., Brem, J. C., Yoon, J. W., Chung,
GenYld combined with the very flexible Barlat K., Dick, R. E., Lege, D. J., Pourboghrat, F.,
2000 yield function generates a real improvement Choi, S.-H., Chu E.: Plane stress yield
of spring back prediction, expect for the DP600 function for aluminium alloy sheets – part 1:
steel. But at this point the user has to deliberate theory. International Journal of Plasticity, 19,
about costs and benefits. The prediction of the final 1297-1319, 2003.
geometry of the H340LAD part is even with the [4] Banabic, D., Aretz, H., Comsa, D. S.,
simple isotropic Hill 48 model quite good and in Paraianu, L.: An improved analytical
industrial applications the use of more cost description of orthotropy in metallic sheets.
intensive model is not justifiable. The situation International Journal of Plasticity 21, 493-
changes when one uses TRIP steels. For these steel 512, 2005.
grades the predicted shape of the final part is not as [5] von Mises, R.: Mechanik der plastischen
close to the reality as it should be, so there is a real Formänderung von Kristallen. Z. angew.
need for improved modeling. The consideration of Math. Mech. 8, 161, 1928
the kinematic hardening is here an adequate way to [6] Faust, A., Roll, K., Kessler, L.: Deep Drawing
get this improvement. Another approach would be Simulation Of High And Ultrahigh Strength
to take into account the phase transition during Steels Under Consideration Of Anisotropic
forming in these kinds of steels. The current Hardening. In: NUMIFORM 2007, June
modeling ignores these phase transitions 2007, Porto.
completely which leads to wrong stress predictions [7] Lemaitre, J., Chaboche, J.-L.: Mechanics of
and thus incorrect springback calculations. But solid materials. Camebridge University Press,
current approaches to this topic (e.g. by Haensel 1990.
[10]) have some limitations. On the one hand, the [8] Gese, H., Oberhofer, G.: MF GenYld +
parameter identification of these models is very CrachFEM Theory Manual, MATFEM, 2007
complex, on the other hand they do not take into [9] Roll, K., Faust, A., Kessler, L.:
account the stress state dependency of the phase Tiefziehsimulation hochfester Stähle unter
transitions. Berücksichtigung anisotroper Verfestigung.
For all the investigated materials, the calculations In: EFB-Kolloquium Blechverarveitung 2007,
with the anisotropic hardening model don’t March 2007, Fellbach.
produce improved results. So from the point of [10] Haensel, A.: Nicht isothermes
springback prediction the utilization of this model Werkstoffmodell für die FE-Simulation von
makes no sense. It has to be stated that the steel Blechwerkstoffen mit metastabilen
grades considered in this article does not exhibit a austenitischen CrNi-Stählen. Fortschr.-Ber.
severe anisotropic hardening. It is supposable that VDI Reihe 2 Nr. 491, 1998.
the results of the anisotropic hardening model are
507
Numisheet 2008 September 1 - 5, 2008 – Interlaken, Switzerland
aBStraCt: The objective of this study is to analyze the accuracy of springback prediction by finite
element analysis (FEA). Predicting springback by FEA requires the correct choice of model asumptions and
the settings of input parameters such as mesh density, friction, and material properties. In this paper, the
springback prediction for the hat shape drawing test (similar to Numisheet´93 Benchmark) is analyzed. The
material behaviour is defined in tensile tests and cyclic shear tests. The observed Bauschinger effect is
considered by a combined isotropic-kinematic hardening model. The influence of numerical input
parameters is studied and optimized by the Response-Surface-Methodolgy. Using the defined parameter
settings, springback experiments were reproduced numerically. The investigations show that springback
prediction is more accurate if a combined isotropic-kinematic hardening is taken into account. Further, it is
shown that a springback prediction is very sensitive to numerical input parameters. Small changes in
numerical input parameters can result in big differences in the predicted springback. But a numerical
reasonable parameter setting which allows efficient and correct springback predictions for all experiments in
a general way could not be found.
509
Numisheet 2008 September 1 - 5, 2008 – Interlaken, Switzerland
2 eXperIMeNtaL reSULtS
Evaluating the quality of a springback prediction
requires experimental data to compare simulation Figure 2: Experimental springback results
and experiment. For these investigations, hat-shape
drawing was selected, because springback is higher In order to carry out a FE-simulation, it is
in comparison to a cup drawing due to the lower necessary to characterize the material behavior. In
part stiffness. Further, the springback prediction by sheet metal forming, material characterization
FEA is difficult, because the dominating stress usually comprises uni-axial tensile tests in different
state is caused by a stretch-bending. The details of rolling directions for the determination of elasto-
the process setup are displayed in Figure 1. plastic material properties and of plane anisotropy.
To describe the material behavior after a load
reversal, a conventional tensile test is not suitable.
Therefore, cyclic shear tests were conducted.
Details about the test setup can be found in [8,9].
The observed stress-strain curves for the material
in sheet thickness 0.3 mm are displayed
in Figure 3. It has to be mentioned that the
material is a special grade (non-standard). It is a
zinc coated and microalloyed low carbon steel.
510
Numisheet 2008 September 1 - 5, 2008 – Interlaken, Switzerland
the displacement of the yield surface in the stress observed punch forces (see Figure 4).
space during plastic deformation. The backstress Figure 4 (left) illustrates the development of the
tensor is defined by the kinematic saturation stress experimentally observed punch force during the
Xsat and the kinematic hardening rate CX. Further, punch stroke. In a numerical study, the friction
the hardening model comprises an isotropic part, coefficient was varied for different blankholder
which describes an expanding of the yield surface. forces. Figure 4 (right) shows the numerically
This is defined by the isotropic saturation stress observed average punch force. The numerical
Rsat and the isotropic hardening rate CR. The punch force fits to the experimental punch force for
material parameters were determined by using a a friction coefficient of µ = 0.04. The excellent
best fit over the experimental data. sliding is caused by the sheet metal and the
Therefore, equation (1) is used for the flowstress kf lubricant. The sheet metal is zinc coated and the
before load reversal and equation (2) for the lubricant is wax (Gleitmo 2345V).
flowstress after load reversal:
kf0 V0 R X0
k f 1 V 0 R X1
511
Numisheet 2008 September 1 - 5, 2008 – Interlaken, Switzerland
0.015
U in mm
4 CONCLUSIONS
Figure 6: Deviation of simulations to experiments
Considering the stress states occurring in sheet
Poor springback predictions were made for small metal forming processes requires a realistic
punch strokes in the flange and for small material model describing true hardening effects. If
blankholder forces in the side wall. The side wall a material exhibits Bauschinger effect, a combined
curvature ȡ is underpredicted for small blankholder isotropic-kinematic hardening model improves the
forces and the flange angle ș2 is overpredicted for simulation result compared to pure isotropic
small punch strokes. For the Blankholder force, an hardening. The considered material model is not
opposing effect can be observed. For small able to fit exactly to the experimentally observed
512
Numisheet 2008 September 1 - 5, 2008 – Interlaken, Switzerland
cyclic stress-strain curve. This may be a reason for NUMISHEET Conference, 2001, Jeju Island,
the poor springback predictions for small Korea, pp. 85-90.
blankholder forces. If the blankholder force is [2] Olivera, M.; Alves, J.; Chaparro, B.; Menezes,
small, the bending stresses dominate the stress L.: Study on the influence of work-hardening
distribution. So springback is larger and the modelling in springback prediction.
simulation is more challenging and more sensitive International Journal of Plasticity, Vol. 23,
to the Bauschinger effect. No. 3, pp. 516-543, 2006, ISSN 0749-6419.
The springback prediction and respectively the [3] Roll, K.; Lembke, K.; Wiegand, K.:
predicted stress state are very sensitive to Possibilities and Strategies for Simulations
numerical parameters. But a numerical reasonable and Compensations for Springback.
and efficient parameter setting could not be found, Proceedings of the Numisheet 2005, Detroit,
which is compatible with all experimentally USA, Melville, New York, pp. 295 – 302.
observed springback values. The blank element ISBN 0-7354-0265-5
size and the number of elements about tool radii [4] Lee, S.W.; Yang, D.Y.: An assessment of
have strong effects on the simulation result. For numerical parameters influencing springback
these parameters, it is possible to define robust in explicit finite element analysis of sheet
parameter settings, which means: Slight changes in metal forming process. Journal of Materials
the parameter settings have only little effect on the Processing Technology, Vol. 80-81, 1998, pp.
simulation result. For the punch velocity and the 60-67, ISSN 0924-0136.
number of integration points, such a parameter [5] Xu, W.L.; Ma, C.H.; Li, C.H.; Feng, W.J.:
setting could not be found. Such a parameter Sensitive factors in springback simululation
setting exists probably in a parameter range with for sheet metal forming. Journal of Materials
more integration points and slower punch Processing Technology, Vol. 151, 2004, pp.
velocities. But this would increase the calculation 217-222, ISSN 0924-0136.
time. So this is not an alternative solution from a [6] Wagoner, R.H.; Li, M.: Simulation of
practical point of view. Further, the obtained springback: Through-thickness integration.
results show that this would not improve the International Journal of Plasticity. Vol. 23,
results. But from a theoretical point of view, more 2007, pp. 345-360, ISSN 0749-6419.
integration points and slower punch velocities [7] Gösling, M.; Kracker, H.; Brosius, A.;
should improve the results. Gather, U.; Tekkaya, A.E.: Study of the
This seems to be a fundamental problem of an influence of input parameters on a springback
explicit forming simulation with shell elements. prediction by FEA. Proceedings of the
For predicting the stress state in an accurate way, IDDRG 2007 Conference, Györ, Hungary.
such a modelling strategy requires a lot of [8] Schikorra, M.; Brosius, A.; Kleiner, M.:
experience. Small changes in numerical input Determination of Anisotopic Hardening Sheet
parameters can result in big differences in the Metals by Shear Tests. Proceedings of the
predicted stress state and, consequently, in the Numisheet 2005, Detroit, USA, Melville,
predicted springback. New York, pp. 389 – 394.
Further investigations are planed to analyze the [9] Govindarajan, R.; Schikorra, M.; Brosius, A.;
prediction accuracy using an implicit forming Kleiner, M.: Investigation of Anistotropic
simulation with solid elements. The aim of these Hardening Behavior using Cyclic Shear Tests
studies is to clarify if different time integration and Orthogonal Tests. Proceedings of the
scheme and different element types improve the ESAFORM 2006 Conference, Glasgow UK,
results and are less sensitive to numerical input pp. 319– 322, 2006, ISBN 83-89541-68-8.
parameters. [10] Armstrong, P.J.; Frederick, C.O.: A
Mathematical Representation of the
5 aCKNOWLeDGeMeNt Multiaxial Bauschinger Effect. Report
RD/B/N 731, Central Electricity, Generating
The cooperative research program “3D-Surface Board Berkeley Nuclear Laboratories, CA,
Engineering of Tools for Sheet Metal Forming – 1966.
Manufacturing, Modelling, Machining (SFB708)” [11] Lemaitre, J.; Chaboche, J.-L.: Mechanics of
is kindly supported by the German Research Solid Materials. Cambridge University Press,
Foundation (Deutsche Forschungsgemeinschaft Cambridge, 1985.
DFG), the central public funding organisation for [12] Krasovskyy, A.; Schmitt W.; Riedel, H.:
academic research in Germany. Material Characterisation for Reliable and
Efficient Springback Prediction in Sheet
6 reFereNCeS Metal Forming. Steel Research int. 77 (2006)
[1] Duffett, G.; Wyler, R.; De la Cruz, C.: No. 9-10, pp 747. – 753, ISSN 0177-4832.
Material Hardening Model Sensitivity in [13] Myers, R; Montgomery, D.: Response Surface
Springback Predictions. Proceedings of Methodology, Wiley, New York, 1995.
513
515
516
517
518
519
520
Numisheet 2008 September 1 - 5, 2008 – Interlaken, Switzerland
aBStraCt: The purpose of the present paper is to present the results obtained during the
experimentation of a new methodology for the springback reduction in sheet metal forming operations. The
proposed innovative approach involves the following steps: 1) cold forming a sheet metal part to a specified
shape, 2) locally heating the part up to a warm temperature, 3) waiting until the part cools down, 4)
removing the loads and tools. The temperature increase is only localized, thanks to a concentrated source
that affects exclusively the component areas subjected to greater stress concentration or to greater bending
deformation. This operation allows a localized stress relieving that reduces the elastic recovery of the
deformed piece, once removed from the dies. The finite element method has been used to model the thermo-
mechanical deformation of the workpiece and it has helped in understanding the effect of the post-
deformation heating on the workpiece and in improving the proposed methodology, which could be of great
interest for industrial applications.
KeYWOrDS: Springback, warm forming, stretch bending, draw bending, numerical simulation
____________________
* Corresponding author: via Di Biasio 43, 03043 Cassino (FR), Italy, +3907762993719, m.strano@unicas.it
521
Numisheet 2008 September 1 - 5, 2008 – Interlaken, Switzerland
2 eXperIMeNtaL SetUp
Two different kinds of mechanical tests, simple
bending (Figure 1) and stretch bending (Figure 2)
conducted with and without post-deformation
heating will be described as follows. They have
been used in order to develop and verify the
proposed methodology, together with a third kind
of test (draw bending), whose description will be
omitted. All tests have been performed on
rectangular thin sheet metal blanks made by
different materials and with different wall
thickness values. The same equipment and tooling,
Figure 1: Simple bending test
shown in Figure 3, can be used for both tests. The
blankholder force BHF clamps with a maximum
force of 6 tons, which is able to inhibit any draw-in
towards the punch. The punch is dome-shaped, the
central part having a larger curvature radius (100
mm) than the outer portion (50 mm). The punch is
equipped with internal holes and cavities for the
insertion, when needed, of heating devices and
sensors. In simple bending tests, the rectangular
sample is clamped only on one side and its length
is half the length of the specimens used in stretch
drawing, which are clamped on both sides. The
state of stress and strain can be obviously changed
by controlling the velocity and total stroke s of the
punch. However, in the present study the velocity Figure 2: Stretch bending test
profile of the punch has been kept constant to a
trapezoidal shape with a maximum value of 1
mm/s. The combinations of materials and shapes of
the specimens used in the study are reported in
Table 1 and Table 2.
simple bending
stretch bending
522
Numisheet 2008 September 1 - 5, 2008 – Interlaken, Switzerland
Al 1050-0,
wall thickness 0.6 mm
Dc
heating
Figure 4: Angle D measured after bending tests Figure 5: Final shapes after bending tests
As mentioned in the Introduction, tests with post-
deformation heating have been conducted as well.
'D%
Since the largest stress and strain values can be
registered at the bottom fibres of the sample, close
to the radius of the blankholder (as pointed by the
arrow in Figure 4), during tests with warm heating
after forming, local heating is performed only in
this area, by means of a small Bunsen burner
generally used for mini-welding tasks, able to heat
the strips up to 350°C, depending on the material
and the heating time. The temperature has been
monitored with a type k thermocouple placed on
the opposite face of the specimen, i.e. between the
sample and the blankholder radius. An infrared
thermocamera has been placed over the strips in
order to register the temperature map on the top
free surface. The test procedure is as follows.
Temperature Tmax [°C]
1. Rectangular strips are cut and clamped on one
short side with maximum BHF. Figure 6: Final shapes after bending tests
2. Punch is moved up to 40-60 mm with nearly
constant speed (trapezoidal profile) and sheets cases (except for steel, which is evidently not
are bent at room temperature. affected by the Tmax reached), the residual angle
3. Punch is stopped and samples are heated up to increases with increasing maximum temperature
a known maximum temperature Tmax, for a and seems to increase with decreasing thickness t0.
know heating time, with local heating applied
where pointed by the arrow in Figure 4. 3.1 NUMerICaL SIMULatIONS
4. Heater is removed and punch holds until the An FEM simulation model has been developed in
specimen is cooled to about 30°C. order to describe and explain the mechanisms of
5. Punch moves down, the strip is unclamped and the proposed process. A plain strain 2D model has
the final bending angle Dw is measured. been used with 4 through-the-thickness brick
Each experimental condition has been replicated elements (4 nodes) and coupled thermo-mechanical
twice. Tests clearly show that the treatment analysis. Given the small amount of plastic strain
significantly reduces the angle of springback. As observed during the experiments, the Green-
an example, in Figure 5, the final shapes and angles Lagrange strain tensor and the 2nd Piola- Kirchhoff
of two aluminium specimens are shown, one has stress tensor have been used. Implicit Lagrangian
been formed and unloaded at room temperature, integration has been performed. The material is
the other has undergone the mentioned procedure elastic-plastic with isotropic Johnson-Cook [10]
up to 290°C. In the same figure, both samples are hardening:
also shown by a side view, when they are still m
e · § T 18 · (2)
clamped by the blankholder. After this test, the s §
©
n
A Be ¨1 C ln
0.01 ¸ ¨¨1 T
¹©
¸
¸
melting 18 ¹
percentage increase of the final angle obtained
thanks to the proposed method can be measured: Some material properties, namely the elastic E
D w Dc modulus and Poisson’s coefficient and the thermal
'D % 100 (1) expansion coefficient have been modelled as
Dc
523
Numisheet 2008 September 1 - 5, 2008 – Interlaken, Switzerland
dependent on the temperature T. The data used for 'D%=15 %, which is not very different from the
material AZ31, t0=1.6 mm are reported in Table 3. experimental value (19 %, see also Figure 6).
This example demonstrates that the FEM
Table 3: material data used for AZ31 underestimates the effect of the proposed method,
but it can be used to study the evolution of stress,
A B C n m Tmelting
170 412.16 0.0223 0.223 0.06943 618
strain and forces during the process, i.e. to explain
the thermo mechanics of the phenomenon. In fact,
T [C°] E [MPa] poisson thermal exp [mm/mm*C°] the evolution of the process could be followed with
the aid of Figure 8, where the stress-strain history
0 49500 0.30 0.00002400
of the node corresponding to the point indicated by
20 49000 0.31 0.00002415 the arrow in Figure 4 is shown.
100 46000 0.32 0.00002505
200 43000 0.34 0.00002580 120
300 40000 0.36 0.00002655
400 37000 0.38 0.00002715 100
Stress [MPa]
The temperature history recorded by the thermo- cold forming
60
couple and mapped by the thermo-camera has been heating
imposed to the corresponding nodes of the meshed 40 cooling
blank, hence no heat source has been modelled and springback
no heat exchange has been modelled between the 20
workpiece and the environment nor the tools.
Therefore, the temperature of the remaining nodes 0
is determined by the code only thanks to 0.00 0.01 0.02 0.03 0.04 0.05 0.06
conduction through the workpiece. In Figure 7, the Equivalent total strain
map of the equivalent total strain in the bending
Figure 8: Final shapes after bending tests
area of the sample can be seen at the end of the
cold bending phase, with punch stopped. The plot is divided into four sections: 1) “cold
forming”: while the punch moves and the sheet is
kept up at room temperature, stress and strain both
increase; 2) “heating”: while the punch is hold in
the upper position and the sheet is warmed up, the
stress goes rapidly down and the strain remains
nearly constant; 3) “cooling”: while the punch still
holds the point is compressed due to thermal
shrinking and stresses goes up again, nearly to the
previous level; 4) “springback”: the punch moves
down with the sheet nearly at room temperature;
the level of stress increases as the sheet springs
back, due to the formation of residual stresses.
524
Numisheet 2008 September 1 - 5, 2008 – Interlaken, Switzerland
1. Rectangular strips are cut, painted and As a further example, the plot printed in Figure 11
clamped on both sides with maximum BHF. can be considered, where the 'h% is plotted vs.
2. Punch is moved up to a maximum dome height Tmax for different total punch stroke s. The effect is
h1 (10 to 30 mm, depending on the sheet more relevant with decreasing level of
material and thickness) with nearly constant deformation, i.e. with decreasing s-value.
speed and sheets are bend-stretched at room
temperature. 10%
' h%
3. Punch is stopped and samples are heated up to
8%
a known maximum temperature Tmax, for a
know heating time, with local heating applied 6% s= 12 mm
where pointed by the arrow in Figure 9. s= 14 mm
4% s= 15 mm
4. IR Heater is switched off and the punch holds
until the specimen is cooled to about 30°C. 2%
5. Punch moves down, and the residual dome 0%
height h2 is measured; strips are unclamped. 50 70 90 110 130 150
Tests clearly show that the post-deformation T max (°C)
heating is able to reduce the amount of springback,
which can be measured as (h1-h2). This reduction Figure 11: Springback reduction 'h% for stretch
can be expressed by the following equation (3): bending tests on Al 2024-T3 samples (t0=0.3 mm)
~160 °C
23 °C
525
Numisheet 2008 September 1 - 5, 2008 – Interlaken, Switzerland
Table 4: material data used for Aluminium alloy Numerical simulation run with a fully implicit
integration, a small strain formulation and solid
T (°C) 20 500
E [MPa] 69000 60000 (brick) elements is able to provide a thermo
Poisson 0.33 0.36 mechanical description of the phenomena involved
Thermal exp. 0.0000240 0.0000274 in the process, which is effectively summarised by
[mm/mm*C°] Figure 8.
Yield Stress (MPa) 110 48
On the contrary, mixed explicit/implicit simulation
Plastic hardening moduli 0.20 0.20
with shell elements did not provide satisfactory
results.
6 aCKNOWLeDGeMeNt
The authors wish to acknowledge the Italian
Ministry of research (MUR) for funding the present
research under the “PRIN ‘05” program titled
SMART.
7 reFereNCeS
[1] Moon Y.H., Kang S.S., Cho J.R., Kim T.G.:
Effect of tool temperature on the reduction of
the springback of aluminum sheets. Journal of
Materials Processing Technology, 132:365–
368, 2003.
[2] Yoshida T. et al.: Shape Control Techniques
for High Strength Steel in Sheet Metal
Forming. Nippon Steel Technical Report, 88:
27-32, 2003.
[3] Keum Y.T., Han B.Y.: Springback of FCC
Figure 13: Effective Plastic Strain distribution after sheet in warm forming. Journal of Ceramic
cold stretch bending for Aluminium alloy
Processing Research, 3(3):59-165, 2002.
[4] Mori K., Maki S., Tanaka Y.: Warm and Hot
Without post-deformation heating Stamping of Ultra High Tensile Strength Steel
Sheets Using Resistance Heating. CIRP
With post-deformation heating
Annals - Manufacturing Technology, 54(1):
209-212, 2005.
Figure 14: Comparison between cold and warm [5] Yanagimoto J., Oyamada K., Nakagawa T.:
springback for Aluminium alloy Springback of High-Strength Steel after Hot
and Warm Sheet Formings. CIRP Annals -
Figure 13 shows the effective plastic strain of the
Manufacturing Technology, 54(1):213-216,
cold formed blank at the end of the die stroke.
2005.
Finally, Figure 14 shows a comparison of the
[6] Watanabe H., Mukai T., Sugioka M.,
formed part after springback in cold conditions and
Ishikawa K., Elastic and damping properties
with warming after deformation. The figure clearly
from room temperature to 673 K in an AZ31
shows that the dome height of the warmed part is
magnesium alloy. Scripta Materialia, 51: 291-
smaller, therefore 'h% is negative, and this goes 295, 2004.
against the experimental evidence. Several other [7] Perrin G.: Theoretical determination of the
different approaches have been used for modelling temperature dependence of elastic properties
the material properties. However the results of in cubic polycristals. J. Phys.Chem Solids,
simulations were still unsatisfactory, since the 58(6):1019-1025, 1997.
simulated value of 'h% has always been negative [8] Koster W: Die Temperaturabhängigkeit des
or insignificant. Elastizitätsmoduls reiner Metalle. Z Metallk.
39:1-9, 1948.
5 CONCLUSIONS [9] Geiger M., Vollertsen F.: The mechanisms of
A new method for the reduction of springback has laser forming. 43rd Annals of CIRP: 301-304,
been presented, based on warm heating of the sheet 1993.
metal blanks after forming and before springback. [10] Johnson G. R., Cook W. H.: A constitutive
The experimental evidence has shown the method model and data for metals subjected to large
is effective for simple bending, stretch bending and strains, high strain rates and high
draw bending (draw bending tests and simulations temperatures. In: 7th International Symposium
have not been described in this paper). on Ballistics, 514-546, 1983.
526
Numisheet 2008 September 1 - 5, 2008 – Interlaken, Switzerland
aBStraCt: In this paper a FE simulation of springback in reverse cup drawing process, taken from
NUMISHEET’99 benchmark is carried out, utilizing commercial code, ABAQUS. The forming process is
simulated in ABAQUS/Explicit while the springback simulation is performed in ABAQUS/Standard. The
utilized method is probing a relationship between the springback and other important factors such as initial
blank thickness die gap and type of material, i.e. Aluminum alloys and HSS. It will be shown that increasing
the initial blank thickness and decreasing the die gap leads to springback reduction in the cup. Considering
the required maximum punch load for different values of initial blank thickness and die gap reveals that
these parameters should be controlled appropriately. Moreover, comparing the obtained results shows that
the sheets produced from HSS exhibit smaller amount of springback after unloading, however, they require
much larger punch load during the forming process.
527
Numisheet 2008 September 1 - 5, 2008 – Interlaken, Switzerland
calculation accuracy will have an effect on the dimension of the tools are shown in Figure 1. One
amount of springback. It may be indicated that the quarter of the blank is modeled with a total of 900
element size in FE analysis and the material’s shell elements (S4R) with the symmetry boundary
hardening model have greater effects on the stress conditions along the X and Y axis (Figure 2).
calculation. The material’s hardening model, viz Twenty five integration points through the
the material’s stress–strain relationship, expresses thickness are used in the modeling. Mass densities
the basic properties of the material during plastic used for dynamic explicit code are 2.7 gr/cm3 for
deformation. It is important to correctly select and the aluminum alloys and 7.8 gr/cm3 for the high
reasonably pre-digest the stress–strain curve to strength steel. The major parameters assumed in
enhance the accuracy of the springback simulation the process are:
of bending with FE codes [7, 8].
In this paper the springback occurred in the reverse
cup drawing process, taken from the
NUMISHEET’99 benchmark, is numerically
studied. Influence of initial sheet thickness, die gap
and material type on springback is investigated by
means of the FE software, ABAQUS.
528
Numisheet 2008 September 1 - 5, 2008 – Interlaken, Switzerland
Blank diameter: 170mm displays that the springback amount for 1.15mm
Coefficient of friction: 0.168 and 1.17mm initial thicknesses are considerably
The reverse cup drawing process is composed of close. This fact demonstrates the sensitivity of
two stages; forward draw with total punch stroke of springback to value of initial blank thickness in
55 (mm) and reverse draw with total punch stroke sheet metal forming. As mentioned earlier, the
of 85 (mm). Figure 3 shows the complete deformed required maximum punch load is one of factors
shape of the cup in three conditions for both that may limit the value of the initial blank
numerical and experimental results [13]. When the thickness. The results of required maximum punch
anisotropy is considered, the earring predicted by load for the two stages are presented in Figure 6.
the ABAQUS can be observed clearly in the The results verify clearly that using larger blank
experimental specimen. Instead of applying blank thickness in the process leads to requiring higher
holder force in the model, fixed gap was amount of maximum punch load. Investigating the
considered between the blankholder and the die. In results reveals that increasing the blank thickness
simulations, the punch velocity was increased to 15 of 1.05mm by 11% up to 1.17mm, causes 33%
m/s without mass scaling that resulted in very enlargement in the maximum punch load.
small oscillation in the kinetic energy which is
acceptable for a quasi static process. In order to a)
measure the springback of the cup two parameters 39.6 BS
AS
are considered, i.e. the mean edge radius and the 39.4
bottom angel of the cup (theta), as shown in Figure
38.8
38.6
38.4
38.2
38
1.05 (mm) 1.15 (mm) 1.17 (mm)
Initial sheet thickness (mm)
b)
0.9 BS
AS
0.8
0.7
0.6
Theta (degree)
0.3
0.1
Initial thickness is one of the parameters that
clearly affects on springback in sheet metal
0
1.05 (mm) 1.15 (mm) 1.17 (mm)
forming which in fact may be utilized to control it. Initial sheet thickness (mm)
On the other hand, increasing the initial sheet Figure 5: Influence of the initial sheet thickness on
thickness causes increasing of required punch load springback: a) mean radius of the cup edge, b)
and weight of the blank that are undesirable things theta
in design parameters. Therefore, finding the
Thickness=1.05 (mm)
optimum value for initial sheet thickness is of vital 90
Thickness=1.15 (mm)
importance for the purpose of using suitable values 80 Thickness=1.17 (mm)
thickness apparently is a way of reducing Figure 6: Required maximum punch load for
springback. Paying more attention to the figure, different values of initial sheet thickness
529
Numisheet 2008 September 1 - 5, 2008 – Interlaken, Switzerland
39.4
Mean radius of cup edge (mm)
90 Gap 1.3(mm)
Gap 1.22(mm)
39.2
80 NoGap
39 70
Maximum punch load (kN)
60
38.8
50
38.6
40
38.4
30
38.2 20
Gap 1.3 (mm) Gap 1.22 (mm) NoGap
Die gap (mm) 10
b) 0
BS Stage1 Stage2
0.8
AS
0.7
Figure 8: Required maximum punch load for
different values of die gap
0.6
a)
Theta (degree)
0.5
39.6 BS
0.4 AS
0.3 39.4
Mean radius of cup edge (mm)
0.2 39.2
0.1 39
0
38.8
Gap 1.3 (mm) Gap 1.22 (mm) No Gap
Die gap (mm)
38.6
530
Numisheet 2008 September 1 - 5, 2008 – Interlaken, Switzerland
0.6
0.5 1998.
0.4 [3] Oliveira M.C., Alves J.L., Chaparro B.M.,
0.3 Menezes L.F.: Study on the Influence of
0.2 Work-Hardening Modeling in Springback
0.1 Prediction. International Journal of Plasticity,
0 23:516-543, 2007.
Al60616-T4 Al6022-T4 HSS
Material [4] Lee S.-W., Kim Y.-T.: A study on the
springback in the sheet metal flange drawing.
Figure 9: Influence of the material on springback:
Journal of Materials Processing Technology,
a) mean radius of the cup edge, b) theta
187:89-93, 2007.
140 Al6016-T4 [5] Esat V., Darendeliler H., Gokler M.I.: Finite
Al6022-T4
HSS
Element Analysis of Springback in Bending of
120
Aluminum Sheets. Journal of Materials and
Design, 23:223-229, 2002.
Maximum punch load (kN)
100
531
Numisheet 2008 September 1 - 5, 2008 – Interlaken, Switzerland
aBStraCt: The aim of this paper is to study the springback that occurs in L-bending process after
unloading, by means of ABAQUS, a finite element code. The material’s anisotropy is considered during
simulations. The forming process is simulated in ABAQUS/Explicit while the springback simulation is
preformed in ABAQUS/Standard. The influence of die radius and die clearance on springback for an
aluminum alloy, AA6111-T4, has been investigated considering final amount equivalent plastic strain
achieved in blank and a relationship between them has been explored. The obtained results show that the
higher amount of equivalent plastic strain causes the smaller amount of springback at the end of process.
The relation of die radius and die clearance with amount of required maximum punch load is investigated as
well. The effect of three different hardening models utilized in the FE simulation on springback prediction is
studied. Finally, the springback of various materials are compared to each other.
533
Numisheet 2008 September 1 - 5, 2008 – Interlaken, Switzerland
over bending, stretching, arc bottoming, pinching of sheet thickness, and are mainly caused by the
die, spanking and movement (double bend) unbalanced through-thickness stresses of the sheet
techniques used in ‘U’ channel bending. Math and once it is taken out of stamping tooling. With the
Grizelj [8] reported springback and residual progress of FE methods along with the
stresses of bent plates, designed for assembling computational hardware and software technologies,
spherical tanks made of steel, using elastic–plastic the explicit and implicit incremental formulations
incremental FE calculations and experimental have been developed for the process modeling and
validation. Lei et al. [9] analyzed the free bending analysis. The explicit dynamic and static
and square cup deep drawing to predict the incremental methods have found widespread use in
springback, stress distribution, etc. for stainless the modeling and analysis of 3-D sheet metal
steel using finite element method (FEM). forming due to its ability of better contact handling
Springback is caused by the release of internal and relatively low computational cost when
stress during the unloading phase in sheet metal compared to the implicit static incremental method.
forming, so factors affecting the stress calculation In the forming analysis phase, an initially flat sheet
accuracy will affect the springback calculation. It is is placed between the stamping die elements
indicated that the finite element dimensions and the usually involving the die, punch and blankholder.
material’s hardening model have greater effects on It is common, in sheet metal forming analysis, to
the stress calculation. The material’s hardening include only the surface of the tooling in the FE
model, viz the material’s stress–strain relationship, model, rather than the complete geometry, as rigid
expresses the basic properties of the material geometric entities.
during plastic deformation. It is important to The L-bending process as shown in Figure 1 is a
correctly select and reasonably pre-digest the case studied in this paper for three materials:
stress–strain curve to enhance the accuracy of the AA5754-O, AA6111-T4 and DP-Steel. The
springback simulation of bending with FE codes materials basic properties are summarized in Table
[10,11]. 1. To increase the computational efficiency, the
In this paper the finite element simulation of the simulation of the L-bending process is modeled in
springback in L-bending process using FE code, the finite element program ABAQUS\Explicit,
ABAQUS, is studied. The influence of die radius, while the springback analysis is simulated in
die clearance and material on springback is ABAQUS\Standard as it would take a long time to
investigated by considering the required maximum obtain a quasi-static solution of springback analysis
punch load and the achieved amount of equivalent in ABAQUS\Explicit. Half of the blank is modeled
plastic strain in sheet. Also, three different with a total of 300 shell elements (S4R) and 9
hardening models are utilized in the simulations in integration points through the thickness. For
order to study their effects on springback definition of contact in ABAQUS/Explicit, the
prediction. general contact algorithm was utilized. The Hill48
anisotropic yield function is utilized to consider the
2 Fe SIMULatION material anisotropy. Mass densities used for
dynamic explicit code are 2.7 gr/cm3 for the
In this part, the computer simulation of the aluminum alloy and 7.8 gr/cm3 for the high
stamping process is conducted in two major steps. strength steel. The initial dimension of the sheet
Firstly, a forming analysis is conducted, including was 127mm (length) u 25.4mm (width) u
the blank and tooling, in order to determine the 1.016mm (thickness) with the 70mm total punch
sheet metal deformation during the stamping stroke. The contact between tools and the sheet
process and, secondly, the sheet metal springback blank is simulated as a frictionless choice in the FE
deformations following the removal of the code, while lubricant is used in experimental
stamping tooling are computed using the forming procedure. The punch velocity was speeded up to 3
stress distribution and the deformed geometry m/s in the dynamic explicit code. The change in
along with thickness distribution. There are some parameter ș after unloading is considered as
fundamental differences in the characteristics of springback. Simulations are preformed for three
both computation phases. The forming process is different die radiuses (Ri), die clearances (d) and
controlled by the time-dependent interactions of hardening models.
the blank and stamping tooling through a frictional
contact-interface, and results in gross shape
3 reSULtS aND DISCUSSION
changes of the sheet metal. Consequently, the
computational modeling of the forming process 3.1 eFFeCt OF DIe raDIUS
necessitates an incremental formulation due to the Three different values for die radius, say 12.7mm,
geometrically non-linear kinematics of sheet metal 9.525mm and 4.7625mm are considered in our
deformation involving large displacements, large simulations while other parameters are assumed to
rotations and finite plastic strains. On the other be constant. The variation of springback with
hand, the springback deformations of a typical changes in die radius for the aluminum alloy,
stamping part are comparatively small, on the order
534
Numisheet 2008 September 1 - 5, 2008 – Interlaken, Switzerland
88
Thetal (degree)
84
AA5754- AA6111- DP- 82
O T4 Steel 80
Young’s 78
Modulus 73.25 75.25 205.35 76
(GPa) 74
Poisson’s
0.33 0.33 0.3 72
ratio 12.7mm 9.525mm 4.7625mm
Die radius (mm)
Yield
strength 102.4 149.1 358.7 Figure 2: Effect of die radius on springback
(MPa)
Ultimate 350
tensile 300
234.2 279.3 570.9
strength 250
(MPa)
150
100
50
0
12.7mm 9.525mm 4.7625mm
Die radius (mm)
0.12
Figure 1: Schematic view of tools Die radius
12.7mm
0.1
9.525mm
springback occurrence (AF). Therefore, the greater 4.7625mm
Equivalent plastic strain
punch load for the three values of die radius. The Figure 4: Distribution of equivalent plastic strain
obtained results show that using smaller die radius along Path1 for different values of die radius
requires higher value of punch load. The reason of
this fact, springback reduction by decreasing the
die radius, may be attributed to the equivalent 3.2 eFFeCt OF DIe CLearaNCe
plastic (characteristic) strain achieved in the sheet In order to study the influence of die clearance on
for the different values of die radius through the springback in the L-bending process, three
process. In Figure 4 distribution of the equivalent different values are chosen which are 1.55mm,
plastic strain along Path1, located along the sheet 1.35mm and 1.1mm. The obtained results by the
length on the front side and top layer of the sheet, FE simulation for AA6111-T4 are shown in Figure
is shown. As it can be observed, decreasing the 5, while the die radius is assumed to be 12.7mm.
amount of die radius causes the level of the Referring to the figure, springback increases while
equivalent plastic strains to ascend. Consequently, the die clearance is enlarging, although some
the plastic strains, especially axial plastic strain, restrictions take place during this enlargement. One
increases. It is noticeable that the area under the of the most important limitations that occur is the
equivalent plastic curve for the smaller die radius is maximum required punch load, similar to the die
larger which justifies the larger required maximum radius as shown in Figure 6. It is observed that the
punch load for it.
535
Numisheet 2008 September 1 - 5, 2008 – Interlaken, Switzerland
0.015
84
0.01
82
80 0.005
78 0
0 5 10 15 20 25 30 35 40 45
76
True distance (mm)
74
72
Figure 7: Distribution of equivalent plastic strain
1.55mm 1.35mm 1.1mm along Path1 for different values of die clearance
Die clearance (mm)
116
utilizing the isotropic model in FE simulating still
are reliable.
114
112 BF
90
AF
110 88
86
108
1.55mm 1.35mm 1.1mm 84
Theta (degree)
of die clearance 78
76
72
On of the remarkable factors in the simulation of ISO-KIN ISO KIN
536
Numisheet 2008 September 1 - 5, 2008 – Interlaken, Switzerland
reveals that the higher level of the equivalent more springback. Besides, the superiority of
plastic strain prediction leads to the higher amount AA5754-O may be probed in the amount of
of maximum punch load anticipated by the maximum required punch load, as shown in Figure
hardening model. 12. It may be observed clearly that the bending of
AA5754-O requires less maximum punch force
118 than the other two materials which means less
116
amount of energy is needed for performing of the
process.
114
Punch load (N)
112 90 BF
AF
88
110
86
108
Theta (degree)
84
106
82
104
ISO-KIN ISO KIN 80
Hardening model
78
ISO
0.03 KIN 350
0.025 300
0.02
250
Punch load (N)
0.015
200
0.01
150
0.005
0 100
0 5 10 15 20 25 30 35 40 45
True distance (mm) 50
537
Numisheet 2008 September 1 - 5, 2008 – Interlaken, Switzerland
has the lowest one. This relation can also be found [6] Papeleux L., Ponthot J.P.: Finite element
in the springback angles for the three materials simulation of springback in sheet metal
(Figure 11). forming. Journal of Materials Processing
Technology, 125-126:785-791, 2002.
4 CONCLUSIONS [7] Chou I.N., Hung C: Finite element analysis
and optimization on springback reduction.
A numerical study of springback phenomenon in International Journal of Machine Tools and
the L-bending process utilizing the FE code, Manufacturing, 39:517-536, 1999.
ABAQUS was provided in this paper. Effects of [8] Math M., Grizelj B.: Finite element approach
different factors such as die radius, die clearance in the plate bending process, Journal of
and material type on springback have been Materials Processing Technology, 125-
investigated. Also, the influence of different 126:778-784, 2002.
hardening models on predicting springback was [9] Lei L.P., Hwang S.M., Kang B.S.: Finite
presented. element analysis and design in stainless steel
Decreasing the die radius causes the reduction of sheet forming and its experimental
the springback, because the process let the blank comparison, Journal of Materials Processing
achieve higher level of equivalent plastic strain Technology, 110:70-77, 2001.
after the forming stage. On the other hand, the [10] Zhang Z.T., Lee D.: Development of a new
required maximum punch load increased for the model for plane strain bending and
higher values of die radius. Similar event occurred springback analysis. Journal of Material
for the die clearance. The investigation was Engineering Performance, 4:291-300, 1995.
extended for different types of materials where [11] Zhang Z.T., Hu S.J.: Stress and residual
AA5754-O lead to smaller springback by reaching stress distribution in plane strain bending.
to the higher amount of equivalent plastic strain. International Journal of Mechanical Sciences,
Comparing the results of springback for the 40:533-543, 1998.
different materials revealed that AA5754-O [12] Gau J.T., Kinzel G.L.: An experimental
required less maximum punch load through the investigation of the influence of the
process. After all, the influence of hardening Bauschinger effect on springback predictions.
models on springback prediction was investigated. Journal of Materials Processing Technology,
It was found that results were so close to each 108:369-375, 2001.
others because the elements of the blank did not [13] Prager W.: Recent developments in the
undergo reversal loading. mathematical theory of plasticity.
International Journal of Applied Mechanics,
5 reFe reNCeS 78-493, 1956.
[1] Lingbeek R., Huetink J., Ohnimus S., Petzoldt [14] Ziegler H.: A modification on Pragr's
M., Weiher J.: The development of a finite hardening rule. Quarte. of Applied
elements based springback compensation tool Mathematic, 17-55 1959.
for sheet Metal products. Journal of Materials [15] Armstrong P.J., Frederick: A mathematical
Processing Technology, 169:115-125, 2005. representation of the multiaxial Bauschinger
[2] Panthi S.K., Ramarishnan N., Pathak K.K., effect. G.E.G.B. Report RD/d/B/N, 1966.
Chouhan J.S.: An analysis of springback in [16] Lemaitre J., Chaboche J.-L.: Mechanics of
sheet metal bending using finite element Solid Materials, Cambridge University Press,
method(FEM). Journal of Materials 1990.
Processing Technology, 186:120-124, 2007.
[3] 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, 141:109-116, 2003.
[4] 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, 123:209-211, 2002.
[5] Bui Q.V., Papeleux L., Ponthot J.P.:
Numerical simulation of springback using
enhanced assumed strain elements. Journal of
Materials Processing Technology,
153í154:314-318, 2004.
538
Numisheet 2008 September 1 - 5, 2008 – Interlaken, Switzerland
t. de Souza1, B. F. rolfe2*
1
Centre for Material and Fibre Innovation, Geelong Technology Precinct. Deakin
University. Waurn Ponds. 3217. Victoria, Australia.
2
Department of Engineering and Information Technology. Deakin University. Waurn
Ponds. 3217. Victoria, Australia.
aBStraCt: The sheet forming industry is plagued by inherent variations in its many input variables,
making quality control and improvements a major hurdle. This is particularly poignant for Advanced High
Strength Steels (AHSS), which exhibit a large degree of property variability. Current FE-based simulation
packages are successful at predicting the manufacturability of a particular sheet metal components, however,
due to their numerical deterministic nature are inherently unable to predict the performance of a real-life
production process. Though they are now beginning to incorporate the stochastic nature of production in
their codes. This work investigates the accuracy and precision of a current stochastic simulation package,
AutoForm Sigma v4.1, by developing an experimental data set where all main sources of variation are
captured through precise measurements and standard tensile tests. Using a Dual Phase 600Mpa grade steel a
series of semi-cylindrical channels are formed at two Blank Holder Pressure levels where the response
metric is the variation in springback determined by the flange angle. The process is replicated in AutoForm
Sigma and an assessment of accuracy and precision of the predictions are performed. Results indicate a very
good correspondence to the experimental trials, with mean springback response predicted to within 1° of the
flange angle and the interquartile spread of results to within 0.22°.
539
Numisheet 2008 September 1 - 5, 2008 – Interlaken, Switzerland
540
Numisheet 2008 September 1 - 5, 2008 – Interlaken, Switzerland
UTS (MPa)
650
grey and the mean experimental flow curve in red. Yield Stress (MPa)
A relative large dispersion of flow behaviour is
Figure 4: Correlation between Yield Stress and
observed, primarily caused by shift in the yield Ultimate Tensile Stress for DP600.
stress and not so much variation in the strain
hardening properties. Table 1: Variation of material parameters extracted
A power law approximation (Equation 1) was used from intrinsic tensile tests and blank geometry
to characterise the material behaviour of each test measurements
specimen. In order to accurately represent the true
Material Mean Std. Min Max
scattering behaviour of the DP600, correlation
Parameter Dev
between the material parameters was performed to
ıy (0.2%) 489.46 12.55 467.69 535.01
establish if any dependence occurred. It was found
UTS 632.99 14.37 605.43 674.60
that the Yield Stress and the Ultimate Tensile
Stress were linearly dependent upon one another as t0 (BHP-21MPa) 1.958 0.017 1.919 1.979
shown in Figure 4, with the strength co-efficient, t0 (BHP-57MPa) 1.963 0.013 1.914 1.982
K, related by Equation 2. The strain hardening, n, w0 (BHP-21MPa) 19.55 0.234 18.94 19.94
remained constant at 0.143. w0 (BHP-57MPa) 19.55 0.155 19.20 19.80
V KH n (1)
2.3 prOCeSS SetUp
Vy An Erichsen (Model 145) sheet metal formability
K (2)
H 0n tester has been used for the forming process.
Relatively slow punch speeds of approximately
Quantitative characterisation of the spread in
3.4mm/s have been used to allow an adequate
material parameters was performed by determining
sampling rate of data for tool forces and punch
the standard deviation, which assumes a normal
stroke displacement. To limit frictional variation
distribution, and minimum and maximum limits,
effects no lubricant was used and both the tools and
which are represented in Table 1. Normal
blank were cleaned with an acetone covered cloth
probability plots were performed to ensure a
before each stamping.
normal distribution fit is acceptable.
Blank holder force was monitored and precisely
Blank thickness and width variation was also
controlled to within 1% of the desired value due to
measured for each specimen before forming and is
the resolution of the load cell.
recorded in Table 1 for each of the BHP operating
windows explored.
DP600 Material Variation
3 SprINGBaCK MeaSUreMeNt
900
Average Experimental Max AF Sigma Achieving an accurate, precise and reliable method
Material
800 Mean
Variation to determine the amount of springback was crucial
Min
700 Approximation
to ensuring that the subtle changes in geometry
True Stress (MPa)
600
from stamping to stamping were captured. A
500 simple yet accurate method to characterise the final
400 part shape has been developed which involves the
300 following process.
200 Firstly the channel edges are marked with a white
100 paint marker to ensure a good level of contrast
0 when each channel is scanned. The channels are
then scanned on a flat-bed at a resolution of 300dpi
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35
True Strain
using a black and white filter. The image (*.tif) is
Figure 3: Material variation of the DP600 steel, then inverted to provide a white background
indicating the variation approximation used within leaving a solid outline of the channel profile. A
the AutoForm FEM simulations.
MATLAB script has been developed to process the
image, utilising chain coding which converts the
541
Numisheet 2008 September 1 - 5, 2008 – Interlaken, Switzerland
image file into a x-y data set. To allow reference to 5. The selected experimental data set represents the
the tooling geometry when analysing the amount of average punch force response, whilst the nominal
springback, the bottom edge of the channel is simulations were chosen, with all noise parameters
traced. From here the 2D profile is rotated, to set at their mean values. This process also allowed
ensure the flange angles are balanced, providing an the selection of a suitable friction co-efficient
axis of symmetry. A linear regression fit of flange which accurately captured the forming forces.
region allows the angle from horizontal to be Table 2: Description of Numerical Parameters used
determined. The flange is a common mating face, in the FEM simulation.
and its dimensional accuracy is critical at the
assembly stage, hence is a useful metric. Numerical Parameter Value
Tools
3.1 MeaSUreMeNt UNCertaINtY Tool Stiffness 100 N/mm3
To establish the uncertainty in the measurement Tool Columns (Binder) Tool Centre
technique two processes where employed. The first Material
analysed the accuracy and determined if any bias Flow Curve Ludwik
error was prevelant. This involved generating a Yielding Criterion Hill (Isotropic)
CAD profile of a semi-cylindrical channel with a Bi-axial stress factor 1
specifed amount of springback i.e. flange angle of Lubrication – all tools 0.135
5° and performing the springback charaterisation Element Type Elastic-Plastic
measurement process. This was repeated ten times. Shell
The average measured flange angle was 4.96° Meshing
indicating a very small negative bias 0.04° (0.09%) Accuracy User Defined
concluding excellent accuracy in the measurement Radius Penetration 0.04 mm
process. Max Element Angle 22.5°
To assess the precision of the process, the spread of Initial Number of Elements 2500
the results from the ten tests were investigated, Max Refinement Level 0
with a total range of flange angle measurements Layers 11
equalling 0.09º. Hence, the measurement technique Time Steps
exhibits excellent resolution with the ability to Max Displacement 0.2
accurately and reliably measure the final flange Comparison of Punch Forces
angle of the component. 50
Average Experimental
45 BHP = 57 MPa
AutoForm Prediction
40
35
30
BHP = 21 MPa
The FEM simulation was performed using 25
10
parameters used are described in Table 2 for all Punch Stroke (mm)
542
Numisheet 2008 September 1 - 5, 2008 – Interlaken, Switzerland
Normalised Frequency
Min = 8.35
Mean = 9.14
statistical descriptors determined as shown in Max = 9.55
StDev = 0.37
Cumulative (%)
Range = 1.20
0.15 Median = 9.14 60%
Figure 6 and Figure 7. A normalised frequency is
Range (%) =
IQR = 0.55
Min = 8.14
used to allow direct comparison between different 0.1
Max = 9.83
Range = 1.69 40%
sample sizes. Cumulative density plots are also Range (%) =
18.49
Assessing the accuracy of the AutoForm’s ability Springback Distribution (BHP = 57 MPa)
to predict springback, involves an analysis of the 0.25 100%
Mean = 5.43
StDev = 0.26 StDev = 0.39
of accuracy, a difference of only 0.28° (approx 3%) Median = 4.16
Cumulative (%)
Median = 5.45
0.15 IQR = 0.24 60%
IQR = 0.46
when compared to the experimental mean. The Min = 3.69
Max = 5.45
Min = 4.32
Max = 6.17
Hence, the two metrics provide a good overview to 1 0.34 0.55 0.24 0.46
543
Numisheet 2008 September 1 - 5, 2008 – Interlaken, Switzerland
544
Numisheet 2008 September 1 - 5, 2008 – Interlaken, Switzerland
C.Barthel1∗, B.Svendsen1 ,
1
Department of Mechanical Engineering, D-44227 Dortmund, Germany
ABSTRACT: The work deals with a draw-bending process where a strip is being pulled from right to left
over a roller under tension, and with deep drawing of a cup. After the description of the material model used in
the simulations the identification of the hardening parameters for isotropic, kinematic and combined hardening
is discussed. Finally the simulation results for the draw bending are presented and compared to experimental
results. The deep drawing simulation is work in process and will be presented at the conference.
545
Numisheet 2008 September 1 - 5, 2008 – Interlaken, Switzerland
DP = λ NP
NP = ∂M−X φ
Here, cr and rsat are the saturation rate and the satu-
ration value of r respectively. Kinematic hardening
is modeled by the Armstrong-Frederick form
Ẋ = cx (xsat NP − X) λ
3 PARAMTER IDENTIFICATION
Setting rsat , cr , xsat and cx to zero results in ideal
plasticity in which case the initial yield stress is
σY0 . If xsat and cx are chosen equal to zero, pure
isotropic hardening results. On the other hand, set-
ting rsat and cr to zero results in purely kinematic
hardening.
The total amount of hardening is given by the ten-
sion test. Each valid parameter set has to fulfill this
condition. To this end a Matlab-routine was pro- Figure 2: Simulation of the tension-compression ex-
grammed which iteratively evaluates the “isotropic periment
parameters” depending on the given “kinematic pa-
rameters”. For the uniaxial tension case the stress Figure 1 shows the test result and the simulation re-
can be expressed as a function of the strain by in- sults for the uniaxial tension test. The re-calculation
tegrating the evolution equations belonging to the
546
Numisheet 2008 September 1 - 5, 2008 – Interlaken, Switzerland
of the tension test succeeds with all three investi- 4 DRAW-BENDING PROCESS
gated parameter sets.
The finite element model for the draw-bending test
In Figure 2 the results for a tension compression test
is shown in Figure 3. This process is simulated here
can be seen. The parameters in the simulations are
in 5 steps by analogy with the experimental proce-
the same as in Figure 1. In the case of unloading
dure. In the first step, the sheet metal strip is pre-
and reloading in the opposite direction the behavior
bent to 90 degrees using the bending tool. Secondly,
is symmetric only in the isotropic case as one would
the bending tool is removed. Thirdly, the strip is
expect. For the combined hardening case the simu-
clamped into the machine at both ends and pulled
lation result is closer to the experiments compared
tight over a rubber roller from both ends, putting it
with the two other hardening types. Table 1 to Ta-
under tension. The actual test begins with the strip
ble 2 show the identified parameters for the different
being pulled from right to left over the roller under
hardening models.
tension, where it experiences bending-unbending.
Table 1: Isotropic hardening parameters Since the roller moves with the strip, there is little or
no friction between them. In the last step, the strip
Parameter Value is unclamped and the resulting springback process is
E 200450 MPa simulated (Figure 4 and Figure 5).
ν 0.3
σY0 340.61 MPa
rsat 276.57 MPa
cr 12.26
Parameter Value
E 200450 MPa
ν 0.3
σY0 340.61 MPa
xsat 276.57 MPa
cx 12.26
Figure 3: FE-Model for draw-bending
Parameter Value
E 200450 MPa
ν 0.3
σY0 340.61 MPa
xsat 100.00 MPa
cx 25
rsat 225.23 MPa
cr 5.96
547
Numisheet 2008 September 1 - 5, 2008 – Interlaken, Switzerland
6 ACKNOWLEDGEMENT
Financial support for this work was provided by the
German National Science Foundation (DFG) under
the contract SV 8/9-1 in the priority program 1204
and is gratefully acknowledged.
REFERENCES
[1] J. Wang, V. Levkovitch, F. Reusch,
B. Svendsen, J. Huéting, and M. Van Reel. On
the modeling of hardening in metals during
non-proportional loading. In International
Journal of Plasticity 24, pages 1039–1070,
2008.
[2] V. Levkovitch and B. Svendsen. Accurate
hardening modeling as basis for the realistic
simulation of sheet forming processes with
Figure 5: Comparison of experimental and FE sim- complex strain-path changes. In Proceedings of
ulation results for springback in metal strips subject the 9th international Conference on Numerical
to draw-bending (r=15mm) Methods in Industrial Forming Processes,
pages 1331–1336, 2007, Porto.
tional shell elements. For efficiency, the symmetry [3] J. Wang, V. Levkovitch, F. Reusch, and
of the strip (Figure 3) has been exploited in the sim- B. Svendsen. On the modeling and simulation
ulation. The minimal friction taking place between of induced anisotropy in polycrystalline metals
the roller and strip has been modelled in ABAQUS with application to springback. In Archive of
via the definition of a ”Contact Property” of ”Fric- Applied Mechanics 74, pages 890–899, 2005.
tion” type. A comparison of the simulation and ex-
perimental results for springback of the strip using
rollers of diameter 10 mm and 15 mm are shown in
Figure 4 and Figure 5. The agreement between these
is very good in the case of the combined hardening
modeling. For comparison, simulation results based
on models for purely isotropic and purely kinematic
hardening are also shown for the 10 mm case in Fig-
ure 4. As shown, the purely isotropic model results
in an overestimate, and the purely kinematic model
in an underestimate, of the amount of springback. In
particular, note that the purely isotropic model un-
derestimates the amount of inelastic deformation. In
order to satisfy the boundary conditions, then, the
amount of elastic deformation is overestimated, re-
sulting in too much springback. Similar results have
been obtained for other types of steels, e.g., DP 600
[3].
5 CONCLUSIONS
Using the tension-compression test results, models
for purely isotropic, purely kinematic, and com-
bined hardening have been identified for the new
steel LH800. The identified model was validated
with the help of the finite-element simulation of
draw-bending. The deep drawing process is work in
progress which will be presented at the conference.
This holds for the experiments and the simulations.
In particular spring-back and distortional hardening
548