You are on page 1of 7

Metal working

Calculation of metal forming processes by means of FEM


Andreas Greve, Folker Haase, Lutz KeBler and Robert Schilling

The simulation of metal forming processes with the finite element method (FEM) is becoming more and more significant for process ana-
lysis and optimising. With regard to the quality of the calculations care must be taken to select a suitable FE model. Possibilities for FEM
applications will be shown using the examples of die bending sheets and profile bending. A practical extension of this application is the
support of the FE systems user by access to a material database.

Berechnung von Umformprozessen mit der FEM. Die Simulation von Umformvorgangen mit der Finite-Elemente-Methode (FEM) wird fur
die ProzeBanalyse und -optimierung zunehmend bedeutungsvoller. Ein geeignetes FE-Modell isl mit Hinblick auf eine qualitativ hochwertige
Berechnung sorgfaltig auszuwahlen. Am Beispiel des Gesenkbiegens von Blechen und des Profilbiegens werden die Einsatzmoglichkeiten
der FEM aufgezeigt. Eine UnterstOtzung des Anwenders von FE-Systemen durch Zugriff auf eine Materialdatenbank bildet dabei eine
sinnvolle Erganzung.

The market for the metal-working industry today is char-


acterised by increasing varieties of parts, decreasing batch
• Increasing dynamic of market
sizes and increasing quality requirements. To remain com- - Shorter live of products
petitive companies are forced to ensure that their production -Increasing varieties of parts
complies with the increased demands made for productiv- - Decreasing batch sizes
• Increasing quality requirements
ity, flexibility and safety [1]. • Higher price-cutting
These demands made on production lead to the process-
oriented development focal points shown in figure 1 and
bring about the increased use of modem computer sys-
tems and computer-aided technologies. In the field of metal
forming production processes the focal points at present of
applications of these technologies are mainly to be found Productivity
in fields which involve coping with general mechanical en-
gineering construction and production planning tasks [2].
Computer-aided simulation of forming processes, in partic-
ular in the case of precision forming, is gaining in signifi-
Flexible Flexible Optimized Safety Process
cance as an aid to process optimising. tools manufacture produktion process control
Figure 2 shows the integration of process simulation of sequence design
in a computer-aided metal forming production process. To geometry

avoid extensive series of experiments the simulation of Figure 1. Market trends determine the demands made on produc-
forming processes is used as a numerical experiment [3] tion
and supplies starting points for selective experiment plan-
ning, so that experiments, which are often bound up with method of analysis which has multiple uses. Where these
considerable expense, can be reduced to a minimum. As numerical processes are used the question is often raised
among other things a reliable determination of material as to the quality of the calculations which can only be an-
data through experiments has a considerable influence on swered under the significant aspect of model influences.
the quality of the calculation, process simulation will never For an optimum generation of the FE model, however, not
be able to replace experiments. The aim of simulation is only the aspect of accuracy is important but also the time
rather to clarify as early as possible, i.e. in the design and required for the calculations which is governed by the sim-
construction phase, questions concerning: ulation times and by the amount of pre- and post-processing
the feasibility of the metal forming process; involved.
changes in the characteristics of the workpiece and pos-
sible optimising of the process, for example to minimise Use of process analysis in forming by bending
residual stresses; technology
current physical processes to extend understanding
Simulating sheet bending with the FEM. Using the
through parameter variations [3].
example of die bending the possibilities for process analy-
Methods of numerical analysis of metal forming processes ses will be introduced in the following pages. Moreover, the
can now look back on twenty years' development. At a influence of the FE model on the calculated results will be
very early stage these were used in the fields of automo- made clear. For some years the commercially available FE
tive engineering and aerospace technology. However, they system Marc has been used in the Department of Forming
are only now starting to be used by small and medium- Technology to simulate metal forming processes. In con-
sized businesses [3]. The finite element method (FEM) is a trast to elementary methods of calculating, the FEM allows
a wide-ranged analysis of the bending process including the
Dr.-Ing. Andreas Greve; Dip/.-lng. Folker Haase; Dipl.-Ing. Lutz KejJler;
Dr.-lng. Robert Schilling, Department of Forming Technology, Dortmund die-stamping process. The FE model required for the sim-
University, Germany. ulation was based on the following assumptions:

284 steel research 65 (1994) No. 7


Metal working

Workpiece Equivalent plastic


Product design strain (v. Mises)
Material : X 5 Cr Ni 18 9
Development CAD Sheet width: b = 40 mm Do -o. 01
Construction MARC element type 11 [:SJ 0, 01 - 0, 09
Drafting (Plane strain) Ill§§§ 0, 09-0, 17
El§ ~ 111111110, 17-0, 25
Product data Corrections
'8111, ll!II • 0, 25-0, 50

Process development
I I
Development CAD
Construction
Drafting
. Experiment
,
:I I
Experimental arrangement
. Measure
En g ineerin g -
dat a
181!
Corrections Parameter Recommendations
'D'

s:31

Process simulation
I I
Model generation
Calculation
Drafting
' '
I Job engineering I
Choice of operating material CAP
NC - programming
&NC - data = 7, 4 mm
Construction of «, = goo
Tool Fst. max= 100 kN
operating material
Rigid
Control system CAM
Checking Friction
(by Roll/Lange/ coefficient : µ= 0, 15
Operational data collecting
Herrmann/Wilhelm)
Figure 2. Position of process simulation in a computer-aided metal Figure 3. FE model and equivalent plastic strain distribution during
forming technology [3] die bending

- plane strain in the sheet; gridded specimens made of special stainless steel, which
rigid tools; were bent in stages, the stress distribution in the sheet was
consideration of the friction between workpiece and tool established by determining the material flow. Figure 4 il-
with the help of Coulomb's law; lustrates the good degree of consistency between the visio-
sheet free of residual stress before forming; plastically and the numerically determined circumferential
linear elastic behaviour of the sheet material and non- stress distribution in the axis of symmetry of the bending
linear, non-compressible behaviour in the area of plastic part under load in the tensile area. On the other hand, devi-
deformations; ations can be seen in the compression area of the bending
isotropic flow behaviour according to v. Mises flow cri- part. This can also be observed in the radial stress curves
terion, associated flow law and isotropic compacting be-
haviour of the material.
The work hardening curve of the sheet material, austenitic I
I
I Bending angle: & 1 = 60°
steel X5CrNi 18-9, was determined in a computer-aided I FE - calculation:
flat tensile test before the calculations. Figure 3 shows the 3, 2 f - - 1 - - - - + + - - - - h - - / - - , f - - - - - Clxx
E
FE structure and the distribution of the equivalent plastic E - - - Cfyy

strain on air bending and on stamping. Typical for bending c Visioplasticity:


·rl
are the v-shaped isolines not only of the equivalent plastic .c: 2. 4
strain but also of the equivalent stress in the middle area .c

of the sheet, which is not shown here [4]. Maximum strain D.
QJ
or stress values occur in the upper side of the sheet near TI
1, 6
the symmetrical section. In the case of sheets which are .µ
Cll
relatively thick in comparison to the width of the lower QJ
.c
die, the influence of the force initiation of the punch and en
o. 8
the die edge on the deformation distribution can be seen. \
\
For example, with a punch displacement of 4 mm the sheet
already displays a local plasticised zone at the die edge. 0 ~--~----'---~~--~----'-->---~
Verifying calculations, e.g. with other methods, or ex- -1200 -BOO -400 0 400 800 1200
perimental examinations are important for the determina- Stress Oxx • Oyy in N/mm 2
tion of the quality of the simulation. For this reason, as a Figure 4. Distribution of normal stress in circumferential and radial
supplement to the simulation calculations visioplastic ex- directions in the symmetry section of the bending part with a bending
aminations were carried out [5]. With the help of special, angle of 60 °

steel research 65 (1994) No. 7 285


Metal working

and according to Rothstein [6] is to be regarded as caused


evice or bending without lateral force
by the process. With the visioplastic method, starting with
the outer fibres, the radial stress was calculated first of all
and then the circumferential stress. This means that errors
(which can scarcely be avoided in the experiment) can add
up from the outside to the inside.
After the external force has been removed residual Bending sample
stresses caused by forming remain in the sheet which can without
also be calculated with the FEM - as has been shown in [4;
7; 8]. Suitable modelling of the material behaviour, in order
to take into account the Bauschinger effect, for example,
is of particular importance for the quality of this kind of
residual stress calculations.

Simulation of profile bending with the FEM. Today,


along with process analyses of forming sheet bending, sim-
ulation of profile bending with the FEM is also of interest.
However, other FE models are important for profile bend-
ing (figure 8). Profile bending is often used for the further
processing of extruded aluminium profiles. Because of the
complicated design possibilities for extruded profiles it is
often not possible to make a simple theoretical prediction
of the cross-section deformations which can be expected or
of the process limits.
Trials of bending without lateral force, in which the
bending process is reduced to a pure moment bending, have
proved themselves in experiments. Because of the slight F==~=-=--=-=--=-=--=-=--=-=--=-=--=-~
influence of machine parameters, this bending process is Rigid
easily reproducible, and is specially suitable for fundamen-
tal examinations, for example, to determine characteristic
values, which can then be transferred to other bending pro- Figure 5. Experimental appliance and FE model for profile bend-
cesses [9; 10]. Numerical studies with the FEM can be used ing
as a supplement to these experiments.
The sketch in figure S shows the appliance used for Material: Al Mg Si O, 7
bending without lateral force with the rectangular hollow
profile used for the experiments. The appliance can be
placed on a press-frame. On a downward movement of the
upper press bar, the forces required for forming the profile
are transmitted through a lever arm to the profile clamp-
ing position. Both ends of the lever, at which the force is
supplied are fitted with roller bearings, so that they can be
moved freely in a horizontal direction. This type of bear-
ing has the effect that no force components are created in
a horizontal direction, and also that there is a bending load 60
free of lateral force in the area of the profile to be formed.
Because of the double symmetry of the bending process
the FE model was able to be reduced to a quarter of the
bending trials. In order to take the boundary conditions into
consideration, the nodes in the cutting planes were fixed so
that they could only be displaced within these planes. The
profile was filled with a plastic core in the clamping area,
so that only slight deformations were possible here. This Rigid
behaviour was taken into account in the calculation by a Figure 6. FE model for the simulation of bending without lateral
rigid final area of the model. The bending moment was force
applied by forces at the rigid face of the profile.
rectangular hollow profile used in the experiment measured
Figure 6 shows the FE structure used for the simulation 60 mm x 40 mm x 4 mm and was made of AlMgSi0,7.
of profile bending. This model consists of 1624 elements The strains and cross-sectional deformation occurring
and 2494 nodes. The hatched section at the front end of during forming are shown in figure 7 for an outer external
the profile marks the elements for which rigid material be- bending radius of 29.3 mm, which corresponds to a cur-
haviour was assumed, in order to simulate the hindering of vature of 3.4 m-1. The rigid clamping end is at the front
deformation in the clamping area. Comparative calculations end of the profile shown. Because of the material behaviour
have shown that an even finer FE structure does not lead selected for this area, there is no strain or cross-sectional
to an improvement in the result of the calculations [10]. The deformation in this area. Outside the area of influence of

286 steel research 65 (1994) No. 7


Metal working

1,
Beam model I'-...
I '-...
I '-...
I -...
I I
/' ........... I
I I
I '--... I
/ ................. I
I
I
I
I I I
I I I
I I I
I I I
I I
I I ',, I
I
I I ,.J
I
I I
I I
............ I
'--...
........ ,
I

Shell model

Workpiece: Profile 60x40x4


Material : Al Mg Si 0, 7
Curvature: 3, 41 1/m

Equivalent plastic
strain (v. Mi.ses)
Solid model
0 - o. 020
0, 020 - o. 039
0, 039 - o. 059
0, 059 o. 078
0, 078 - o. 097
0, 097 - o. 117

Figure 7. Equivalent plastic strain on bending a rectangular hollow


profile

Figure 8. Basic sketches of the FE model tor bending without


the clamping position, in other words in the rear area of lateral force
the part of the profile shown here, there is an even strain
distribution in a longitudinal direction of the profile. With gous to the solid model, i.e. rigid elements were planned
increasing distance from the bending axis the calculated for the clamping area here as well, and the double sym-
strain values increase so that maximum strain occurs at metry was used. At the symmetrical planes, a restriction of
the inner and outer bending radius. In addition the section' the rotatory degrees of freedom was planned along with the
displays an even curvature, which is characteristic for the restriction of the translatory degree of freedom vertical to
process of bending without lateral force. The strain distri- the cutting plane. This is necessary to guarantee a vertical
bution in the plane of symmetry section, in other words in transfer of the elements into the symmetrical plane.
the area of the middle of the profile, is made clear in the The third method of describing the process of bending
section shown below. A displacement of the layer free of without lateral force is based on the use of beam elements.
equivalent strain can be seen here. This layer has moved These elements are described by two or three nodes which
around 5 mm from the original symmetrical axis in the di- describe the position of the shear centre in the global co-
rection of the inner bending radius. As a consequence of ordinates system. The cross-section is divided into individ-
this displacement the maximum equivalent strains at the ual webs and described in local coordinates, at which the
outer radius of the profile are slightly greater than at the shear centre is the origin. The integration points lie at the
inner radius. The bending process causes a deformation of web end points as well as being distributed evenly along
the profile cross-section. The webs parallel to the bending the web lengths in a frequency determined by the user. In
axis are curved inwards and those vertical to the bending the case of simulations using beam elements the whole of
axis are curved outwards. The calculated deformations were the cross-section was always modelled. Three-dimensional
also observed in the experiment. non-linear beam elements were used.
A calculating period of 46 h on a workstation was re- A comparison of the momentum/curvature ratios, fig-
quired to simulate the bending process. Along with solid ure 9, shows an extremely good correspondence of the
elements, shell and beam elements were used in the simu- curves calculated with shells and solid models. In the case
lation of the bending process, to check whether the calcu- of the beam models a slightly greater increase in the re-
lating period can be reduced using other types of elements quired bending momentum with increasing curvature can
and how the element class used affected the results of the be seen than with the shell or the solid model. The rea-
calculation, figure 8 [12]. son for this is that the cross-section of the beam element
Shell elements are defined by the position of their middle does not alter on bending. In the case of the shell and solid
plane and by the shell thickness. An isoparametric bilinear models cross-section deformation has the effect that the
thick shell element with seven layers above the shell thick- distance between the outer and the inner radius becomes
ness was used for the calculations. The model was analo- smaller, which causes a reduction in the flexural strength.

steel research 65 (1994) No. 7 287


Metal working

Al Mg Si 0, 7 (T4)
- - - Beam model
E hollow profile: 60x40x4
- - - - - Shell model _ _ _+ - - - - - f - - - - - - - - - <
~ 2. 5 - - Solid model
E 50 q = 3, 15 1/m
;y_
----- ----- ·rl
c

J __--~-:::;:::;---,__J_:::;;;:~:=:"::1::=::::r===J
.
o L__
·: 2.
::;.:
+-'
-~- w
.......
·rl

~ 1. 5
~
f-----------<-----+-- 0 30
E (_
0 0.
E
~

g1 1. 0 0 20
·rl +-'
TI .c - - - Beam model
c Material :AlMgSi0.7 (T4) 01
- - - - - Shell model
~ 0, 5 10
·rl
Rectangular hollow prof i l : 60x40x4 - + - - - - - - <
w - - Solid model
I

0 ~--~>-----~-----+---~------' 0
0 2 3 4 5 -180 -120 -60 0 60 120 180
Curvature q in 1/m Normal stress oxx in N/mm 2
Figure 9. Influence of the selected element type on the momen- Figure 10. Normal stress distribution in circumferential direction
tum/curvature ratio

The effects are slight up to the curvatures under observa- Support for model generation through access to
tion here. Beam elements are therefore extremely suitable material database
for use where the momentum/curvature ratios have to be
determined in a short computational period. In the case of Material data is often transferred to the corresponding
other analysis parameters the definition-determined limits simulation program by an input data record or interactively
of beam elements are encountered, which figure 10 shows at the start of calculating. For the FE user, this transfer
clearly. This figure shows the normal stress distribution in mechanism means a relatively long period of time spent
the direction of the circumference in the lateral web of the on research as well as a source of errors during data trans-
profile. In this case the outer profile curvature is 3.15 m- 1 fer, and can therefore no longer be regarded as practical.
and bending momentum is approximately 2200 Nm. In the For this reason, the concept of a central material data man-
case of the beam model, stress distribution runs symmet- agement system consisting of a material database with the
rically to the centre line of the profile. Compressive stress appropriate interface functions has been designed [14) and
occurs at the inner bending radius and tensile stress at the implemented within the scope of the joint project "Process
outer. In the case of the shell and solid models cross-section Simulation for Metal forming - PSU" [13). In addition,
deformation leads to a displacement of the stress free area the database was connected to the commercial FE system
in the direction of the inner bending edge. The deviations Marc.
between the shell and the solid models are mainly caused The demands the system made on the material database
here by the coarse discretion of the shell model in the area with regard to universal applicability, problem-free extend-
of the centre of the profile. The maximum stress calculated ability and short reaction times in areas critical to runtime
lie at the inner and outer bending radius. They are approx- led to the concept sketched in figure 13. The central com-
imately equal for all calculation models at approximately ponent is an SQL-based material database, the contents of
160 N/mm 2 . which are accessed via corresponding interface modules.
Figure 11 shows the outline of the deformed cross- Along with a data input system, which supports direct in-
section for the solid model as well as the profile defor- put of the material data generated in the experiment, the
mation of the web centre lines calculated with the help of material data management covers a research system for in-
the shell model. A comparison of the cross-sectional defor- teractive applications as well as an initialising system which
mations in the shell and solid models shows that the results calls the material data required for the simulation from the
of the calculations are consistent with one another. database before the calculation. The runtime system covers
Figure 12 shows some of the differences between beam, the area critical for the runtime and makes the processed
shell and solid elements. Where beam elements are used material data available during the simulation.
opportunities for analysis are restricted. It is of particu- The use of the material database in an FE bending simu-
lar disadvantage that the cross-sectional deformation and lation is characterised by the following procedure. Starting,
the changes in profile thickness are not taken into account. for example, from work hardening curves generated in flat
This is contrasted with the advantage of a short computa- tensile tests, the processed material data is written to the
tional period. With shell and solid elements, calculation material database, figure 14. In computer-aided flat tensile
results can be visualised better with the pre- and post- tests for gathering and processing material data a PC is
processor Mentat which was used here. In addition, the available for the intermediate storage of the raw data in the
cross-sectional deformation and in general the changes in form of an ASCII file. This file is then processed in accor-
thickness can also be calculated. However, the computa- dance with the necessary syntax for the data input system
tional period required for the analysis of the bending pro- and transferred to the database.
cess with shell or solid models is much greater than the The pre-processing system is divided into two fields (fig-
computational period required for a beam model. ure 14). On the one hand Mentat is used to enter the FE

288 steel research 65 (1994) No. 7


Metal working

Material : AlMgSiO, 7 (T4)

~
70 Rectangular Beam element She 11 e 1ement Solid element
hollow profil: 60X40X4
s

<> ©
Curvature q = 3, 15 1/m

60 _____ ,
r
e
/
- -- - - I Computable
I
E I Curvature for the Computable Computable
E neutral line
c 50 Cross -
·rl
sectional Nat
computable Computable Computable
deformation

,-,
"'
Q)
40 Change of Not Computable
thickness computable for some Computable
·rl
'I- shell elements
0
L
cross -
I
0. sectional Any Any Any
30 I form
'I-
I
0 I -- --- She 11 mode 1
I Computational

I
- - - Solid model time Short Long Long
.c I
OJ 20
·rl
I
Q)
I
I I
Figure 12. Possibilities for analysis in dependency on the element
I
I
classes used
10 I

0
------
~ Data acquisi
0 10 20 30 Experiment Data base

CID
Width of profile y

-~
in mm

...
Figure 11. Calculated cross-sectional deformation on the shell or
solid model

Preprocessin Preprocessing
Preprocessing Simulation
Research system With PATRAN
MDV anager B_esearch
Materials
:[nterfaces
I>
S.M
Input of:
Data types C>
Goods C>
Geometry,
Data recordsC> Mater1al-Nr. Boundary
ata record-ID-Nr. conditions etc.

Computation
With
I
[K] t.{u}= t.{F}

Postprocessin g

With PATRAN
• Output of:
Displacements,
Strains, Stress.
~~::: Temperatures etc .
• 350-420

Figure 13. System overview of the material data management Figure 14. Course of a die bending simulation
(MDV) of the program system PSU

structure, the rigid tools and the boundary conditions, and individual PSU modules for polling. These calls occur in-
on the other hand making the material parameters available ternally and are invisible to the user. After the calculation,
for the material model necessitates research in the material the results are available in the archive for post-processing
database. The material found can be selected for the simu- (figure 14) and can be evaluated with Mentat via an inter-
lation with PSU or Marc and written to the archive, which face and outputted in the form of graphics for the analysis
also contains the data generated with Mentat. report.
At the start of the simulation the material data is written
in an initialising step to the running time system of the (A 00 914; received: 04. December 1993;
material management system, where it is available to the in completed form: 16. April 1994)

steel research 65 (1994) No. 7 289


Metal working

References [7] Schilling, R.: steel res. 64 (1993) No. 5, p. 255/61.


[8] Schilling, R.: Bander Bleche Rohre 34 (1993) No. 7, p. 29/38.
[9] Finckenstein, E. v.; Adelhof, A.; Haase, F.; Kleiner, M.; Schilling,
[I] Eversheim, W.; Konig, W.; Weck, M.; Pfeifer, T. [eds.:]: Produktions- R.: Blech Rohre Profile 40 (1993) No. 3, p. 215/20.
technik auf dem Weg zu integrierten Systemen. AWK'87 - Aach- [10] Finckenstein, E.v.; Adelhof, A.; Haase, F.; Kleiner, M.; Schilling, R.:
ener Werkzeugmaschinen Kolloquium, Aachen, 11.-12.6.1987, VDI- Flexibles Rundbiegen am Beispiel von stranggepreBten Aluminium-
Verlag, Dtisseldorf 1987. Profilen, [in:] Neuere Entwicklungen in der Massivumformung;
[2] Geiger, M.: CAD in der Umformtechnik - Stand der Entwicklungs- DGM-Symp., Feilbach, 8./9. June 1993; K. Siegert [ed.:], DGM-
tendenzen, [in:] UKD'85 - 2. Umformtechnisches Kolloquium Darm- Informationsgesellschaft, Oberursel 1993.
stadt, D. Schmoeckel [ed.:], Darmstadt, 6./7.3.1985, p. 2.5-1/25. [11] Haase, F.; Schilling, R.: Finite-Elemente Analyse des Biegeumfor-
[3] Lange, K.; Roll, K.; Wilhelm, M.; Herrmann, M.: ProzeBsimulation in mens von stranggepreBten Aluminiumprofilen, [in:] Umformtech-
der Umformtechnik. 4. Aachener Stahlkolloquium Umformtechnik, nik - ldeen, Konzepte und Entwicklungen; M. Kleiner, R. Schilling
Aachen, 30.6./1.7.1988, p. 5.2-1/15. [eds.:], Teubner-Verlag, Stuttgart 1992.
[4] Schilling, R.: Finite-Elemente-Analyse des Biegeumformens von [12] Averhage, M.; Schilling, R.; Haase, F.: Analyse des Biegeumformens
Blechen, Dortmund 1992, (Dr.-Ing. thesis); see also ProzeBsimulation von Aluminium-Profilen mit der Finiten-Elemente-Methode, internal
in der Umformtechnik, K. Lange [ed.:], No. 2, Springer-Verlag, report, Department of Forming Technology, Dortmund University,
Berlin Heidelberg, 1992. 1992.
[5] Rothstein, R.; Schilling, R.: Bander Bleche Rohre 31 (1990) No. 4, [13] AbschluBkolloquium des PSU-Projektes, Stuttgart, 31. March 1993;
p. 40/45. ProzeBsimulation in der Umformtechnik, K. Lange [ed.:], No. 4,
[6] Rothstein, R.: Einsatz der ProzeBsimulation zur Analyse und Weit- Springer-Verlag, Berlin Heidelberg 1993.
erentwicklung von Gesenkbiegeverfahren, Dortmund 1990 (Dr.-Ing. [14] Greve, A.: Wissens- und datenbankbasiertes Beratungssystem ftir die
thesis); see also Fortschritt-Ber. VDI, Reihe 2, No. 194, VDI-Verlag, FE-Simulation von Umformprozessen, Dortmund 1994 (Dr.-Ing. the-
Dtisseldorf 1990. sis).

290 steel research 65 (1994) No. 7

You might also like