Int..I. Mech. Sci. Vol.24, No. 8. pp.

479-493, 1982
Printed in Great Britain.

0020-7403/82/080479-15503.00/0
Pergamon Press Ltd.

F I N I T E E L E M E N T A N A L Y S I S OF M E T A L F O R M I N G
P R O C E S S E S W I T H A R B I T R A R I L Y S H A P E D DIES
S. I. OH
Battelle Columbus Laboratories, 505 King Avenue, Columbus, OH 43201, U.S.A.
(Received 23 May 1981; in revised form 21 September 1981)

Summary--A method of discretizing the die boundary conditions is considered for the
analysis of metal forming processes by the rigid viscoplastic finite element method. The
method is natural to the finite element method and general enough to treat the boundary
conditions of arbitrarily shaped dies in a unified way. Solutions of the spike forging process
are obtained by using the method. The deformation mechanics of the spike forging process are
discussed and the solutions are compared with experiments.
NOTATION
v~ velocity component
iij strain rate tensor component
effective strain rate
effective strain
cr~j stress tensor component
~0' deviatoric stress component
effective stress
1 prescribed boundary traction
U prescribed boundary velocity
f frictional tration
q~ variational functional
E work function
Hi distribution function for node i
ui node velocity
vo die velocity
n die surface unit normal
s die surface unit tangent
K penalty constant
/z viscosity
x material coordinates
m friction factor
k local shear yield
Ro initial workpiece radius
H0 initial workpiece height
VD die velocity (IVo[)
INTRODUCTION

A major objective of mathematical modeling of metal forming processes is to provide

necessary information for proper design and control of these processes. Therefore,
the method of analysis must be capable of determining the effects of various
parameters on metal flow characteristics. Furthermore, the computation efficiency, as
well as solution accuracy, is an important consideration for the method to be useful in
analyzing metalworking problems. With this viewpoint, the rigid plastic (and rigid
viscoplastic) finite element method has been most successful in analyzing a wide class
of metal forming problems.
The rigid plastic finite element method (FEM), which was developed by Lee and
Kobayashi [ 1], has been applied to the analysis of various problem s such as solid cylinder
upsetting, ring compression, extrusion, and sheet bending [1-7]. The method has been
also successfully applied to the prediction of defect formation in upsetting[8] and
extrusion [9], and the measurement of die-workpiece friction [ 10]. It has been extended to
rigid viscoplastic material by Oh et al. [11] and applied to upsetting and ring compression
in the hot working range.
The major advantage of the finite element method is its ability to generalize. That
479

480

S.I. O H

is, the method can be applied to a wide class of boundary value problems without
restrictions of workpiece geometry. This is achieved by the proper discretization
procedure used in the finite element method. Metal forming processes, in general, are
characterized by loading with dies of various shapes. It seems that a method of
discretizing the die boundary conditions, which is comparable to that used in the finite
element method, is necessary in order to fully utilize the advantage of the finite
element method in metal forming analysis.
In the present paper a method is proposed to establish the theoretical foundations
of the rigid viscoplastic finite element analysis of metal-forming precesses with
arbitrarily shaped dies. Then the solutions of the spike forging process are obtained
by using the proposed

method.

RIGID VISCOPLASTIC FINITE ELEMENT FORMULATION

The rigid viscoplastic material is an idealization of an actual one, by neglecting the eleastic response.
The material shows the dependence of flow stress on strain rate in addition to the total strain and
temperature. Also it can sustain a finite load without deformation. The rigid viscoplastic material which was
introduced for the analytical convenience simplifies the solution process with less demanding computational
procedure. Moreover, it seems that the idelaization offers excellent solution accuracies due to the negligible
effects of elastic response at large strain in the actual material. Because of this reason, the rigid viscoplastic
material is especially suitable for metal forming analysis at elevated temperature. In this section a finite
element formulation of rigid viscoplastic material is described. The discussions are limited to the twodimensional isothermal deformations.
Consider a body of volume V at a generic moment with the traction F prescribed on a portion of the
surface, Sv, and the velocity U on Sv. Let Sc be the remainder of the surface where the frictional stress f
acts. The deformation of the body V is characterized by the following field equations.
(1) Equilibrium conditions, neglecting the body force;
wq.j- 0

(I)

where ~r~jis a stress component and "," denotes the differentiation.

(2) Strain rate-velocity relation;

e~j = ~ (v~j + vj,~)

(2)

where ~q and v~ are a strain rate component and a velocity component, respectively.
(3) Constitutive relation;
2iT

o,~j= 7_. e~J

(3)

3e

where tTi~' is the deviatoric stress component, and # and ~ are defined by X/(3/2~rU%r0') and X/(2/3~i:~0)
respectively. The flow stress #, in general, is a function of total strain, strain rate and temperature.
(4) Boundary condition;

a,jni

= Fj o n SF

vi = Ui on Sv

(4)

JLJ = given, sign(/,) = - sign(Av,) on Sc

where n~ is a component of unit normal to the surface, and Av, is slipping velocity.
The field equations (I)-(4) can be put into variational principle as

8*=8[f:<'*)dV+L~K'"dV--fsc{ f Ldv.}dS-fsF~V~d$]=O
I

"*

v~

(5)

where the work function E(~) can be expressed as[12]

E(~) = ~0~ ~ d~.

(6)

Here K is a large positive constant which penalizes the dilatational strain, and "*" denotes the restriction of
the velocity fields to the trial space.

Finite element analysis of metal forming processes with arbitrarily shaped dies

481

The discretization of this functional follows the standard procedure and can be found elsewhere[13].
The distribution function H, which is assigned to each node, is introduced such that

U(x) = ~. H~(x)U~
(7)

V(x) = ~,/-/~(x) Vi
where U(x) and V(x) are the velocity components in x~ and x2 direction, respectively, U~ and V~ are those
at nodal points, and the summations are done over nodal points. In the present investigation 9-node
isoparametric elements are used.
Substituting equations (6) and (7) into equation (5), the variational functional ~ becomes nonlinear
algebraic equations with unknown u,'s which are velocity components of the nodal points. The stationary
point of the functional can be obtained by solving the simultaneous equations
a~P (ul . . . . . U2M) = 0,
Oui

i = 1. . . . . 2M

(8)

where u = ( u l , u2. . . . . u2M-,,u2u)=(Ul, V1. . . . . UM, VM), and M is the number of nodal points. The
solution of the highly nonlinear algebraic simultaneous equations are obtained iteratively by the NewtonRaphson method. The linearized form of equation (8) near the assumed velocity field u0 becomes

00]

+ r 02~ ]

Auj = O.

a-~J.=,~ Laujau~Ju=mo

(9)

Solving equation (9) with respect to Auj, the assumed velocity field is updated by ui + aAu~, where a ~< 1.
The iterative process continues until the fraction of the correction term, I[Aul[/llul[and the calculated nodal
point force error norm reach a preassigned value.
In certain cases of rigid viscoplastic deformation, a portion of the workpiece remains rigid. It can be
proved that the rigid zone is determined uniquely by the field equations if the work function E(~) is convex.
The variational function tb in equation (5) becomes non-analytic at the solution when the rigid zones are
involved and the Newton-Raphson method cannot be used for the solution procedure. The difficulty has
been overcome by assuming the material to be linear if ~ becomes smaller than a small positive number ~0[11].
The rigid viscoplastic material in this approach is considered as a limiting case when Cobecomes infinitely small.
In the finite element method, it is well known that the velocity field given by equation (7) is not flexible
enough to provide adequate solutions for incompressible or nearly incompressible deformation. Various
suggestions have been made in order to overcome this difficulty in analyzing incompressible linear elastic
material[14, 15] and the elastic-plastic material in fully plastic range[16]. Numerical tests show that the
reduced integration for volume strain component gives best results for the present investigation. That is, a
3 x 3 integration formula is used for distortional energy rate, while a 2 x 2 integration is used for volumetric
strain rate with quadratic elements. In relation to the incompressibility, the strain rates and stresses were
evaluated at the reduced integration points for best results.

BOUNDARY CONDITIONS AND ARBITRARILY SHAPED DIES

The implementation of a regular boundary condition on SF or Su is straightforward and can be found
elsewbere[13]. In the practical analysis of metal forming processes by the finite element method, a
particular attention must be paid to die boundary conditions. The frictional stress, in general, changes its
direction at the "neutral point", but the location of this point is not known a priori. The "neutral point"
problem has been considered by various investigators[3,4, 17] in their analyses of the ring compression
test.
The shapes of dies used in metal forming processes change considerably from process to process. In the
finite element analysis of metal forming processes the individual implementation of the die boundary
conditions for a particular shaped die requires a substantial amount of programming effort. Therefore it is
desirable to use a technique which can be applied without the restrictions of die geometries. Thus, the
method becomes a practical and economical tool for the metal forming analysis. Such a technique, which is
natural to FEM, is described in the present paper. This technique allows the systematic implementation of
frictional boundary conditions in a unified way for arbitrarily shaped dies. It also provides the capability of
predicting the neutral point location at the arbitrarily shaped die-workpiece interface. In the present
investigation the constant shear friction law is employed which is a common assumption in metal forming
analysis. The die boundary conditions along curved die-workpiece interfaces have been considered in the
framework of FEM by several investigators [7, 18-21]. Their main emphases were on the effects of rotation
on the boundary conditions in relation with the definition of the traction.
Consider, at a generic moment, a curved die surface which is in contact with workpiece as shown in Fig.
l(a). The die boundary condition at the interface can be given by
V. = v D . n
f, = _ m k - ~Avs
,~ =

2 mk tan_l ( _ ~ )

(10)

482

S.I. OH

Die

DieS,rfoce/I ~loeu~::
y

"x

>

i--

(b]

Ca)

FIG. 1. Schematic diagram of a curved die and a workpiece.

where n is unit normal as shown in the figure, and the subscripts n and s denote the normal and tangential
direction to the interface, respectively. The expression of f, in equation (10) has been used for the smooth
transition of frictional stress near the neutral point for the flat die surface [4]. Here Av, is slipping velocity,
m friction factor, k local flow stress in shear, and Uo is a very small positive number compared to Av,.
The implementation of equation (10) for a curved die is not straightforward since the die surface does
not coincide with the workpiece surface defined in FEM approximation. These discrepancies are shown in
Fig. l(b). Because of the discrepancy, the necessary quantities in equation (10) are not well defined except
at the contact node.
In order to overcome the difficulties the boundary condition normal to the surface is enforced at the
contacting node and is given by
V n = v D n~i )

where n~i)is the unit normal to the die surface, rather than element surface, at node i. It is assumed that the
slipping velocity Av, can be approximated by

Avs = ~ HiAv.i = .~, H,-(v~i-vo' s.)).

(ll)

Here, Hi is the FEM shape function on the surface, v~i the tangential velocity of the ith node, and s(0 the
unit tangent to the die surface at the contact node L
By substituting equation (11) into fs expression, the integrand of the die boundary term in equation (5)
can be evaluated for the two-dimensional case as
r

t
f~' fs dr, = f ' - Zq -7r mkHj tan-' |(v,i

VD ' so,)H, ]
/~0

dvo.

(12)

The integration of the die boundary term in equation (5) is carried out by Gaussian quadrature considering
the element surface as a one-dimensional isoparametric element. A higher order integration is necessary in
order that the term can reflect the accurate behavior of fs near the neutral point where the integrand has
large higher order contributions than that of the shape function. Numerical tests show that a five-point
integration formula is sufficient for this purpose with a quadratic element.
In a case when a portion of the element boundary is not in contact with the die, the slipping velocities at
the free nodes are not defined in equation (12). This difficulty can be overcome easily. The tangential die
velocity vD - %) is set to zero for the free node and the small positive number ue is replaced by a very large
number for the free boundary position. This procedure ensures the consistency of equation (12) on the
contact portion and makes the rest traction free.
The present formulation of the die boundary conditions becomes the same as that of Chen and
Kobayashi [4] in the case of the flat die surface with vD s = 0, except that the present investigation employs
quadratic elements and surface integration is done numerically. It can be easily shown that the error
introduced by the present approach is of the same order of magnitude as that introduced in the FEM
calculation. Therefore, the present formulation does not effect the accuracies of the FEM analysis.
INITIAL GUESS
The Newton-Raphson method, used in solving the nonlinear equation (8), is known to be a fastconverging solution process. A drawback of the Newton-Raphson method is that it requires an initial guess

Finite element analysis of metal forming processes with arbitrarily shaped dies

483

for the velocity field as it can be seen in equation (9). Even though the initial guess velocity field does not
affect the solution itself, it should be close enough to the solution to achieve the convergence in the
iteration process. The initial guess, for the first step solution in general, can be provided by a simple upper
bound solution. It seems, however, that the systematic means of providing the initial guess is desirable for
the effective application of the rigid viscoplastic FEM to complicated metal forming processes. In the
present investigation the initial guess was obtained by treating the workpiece as linear viscous material with
the viscosity as a function of the material point.
Consider the constitutive equation of linear viscous material
~ ' = 2/.*~i~

(13)

where/., is viscosity. Comparing this with equation (3) it can be seen that the solution of rigid viscoplastic
material becomes identical to that of the fictitious linear viscous material with the viscosity
_.
= ~(*)

_ I___~
-

(14)

3~

where ~ is the flow stress of the material.

The rigid viscoplastic solution can be obtained by searching for the fictitious material which satisfies the
equation (14). under an equilibrium state. This can be done iteratively by updating/~ --/~(~) using equation
(14) until/~(~) does not change. This method of solution had been used by Lyness et al. [22], in their finite
element analysis of non-Newtonian fluid. Comparing with the Newton-Raphson method, this method
replaces the initial guess velocity field by the initial viscosity distribution. The iteration scheme with this
method converges fast during the early stage of iteration and it is insensitive to the initial viscosity distribution.
The convergence of this method is slower, especially near the solution, than that of the Newton-Raphson
method.
In the present formulation, advantages of both methods were utilized to obtain the best solution
efficiencies. That is, the Newton-Raphson method was used for the solution process, while initial guesses
were generated by the method described above. Numerical tests show that a "good" initial guess can be
obtained within 2--4 iterations using a uniform initial viscosity distribution throughout the workpiece. The
generation of initial guess by the method is only necessary for the first step solution. For the subsequent
steps, the previous step solutions are used as initial guesses.
SPIKE FORGING ANALYSIS
In order to validate the present formulation, solutions of the spike forging process were obtained. In
spike forging, a cylindrical billet is forged in an impression die containing a central cavity. The deformation
characteristics of the spike forging are such that the portion of the material near the outside diameter flows
radially, while the portion near the center of the top surface is extruded forming a spike. It has been
reported [24, 25] that the material flow, which is characterized by spike height variation, depends on the
interface friction as well as the geometries of dies and billets used.
The deformation mechanics of spike forging processes have been studied by several investigators [2327]. Jain et at. [24] analyzed extrusion forging using a flat top die with a sharp cornered hole by the upper
bound method. Their analysis successfully explained the three distinctive modes of deformation found in
experiments. Those modes of deformation consist of the stages where the spike height is reduced with
increasing deformation, and the spike height remains constant until it finally increases very rapidly.
Newnham and Rowe [25] also considered the flow behavior of spike forging process using a flat die with a
sharp cornered hole. They analyzed the process using the slip line theory and explained the deformation
mechanics. Price and Alexander[26] analyzed the spike forging process by flat platen die with a central
orifice. In their analysis only the sticking die boundary condition was considered. Aitan et al. [27], analyzed
spike forging process with tapered top and bottom dies by the modular upper bound method.
Computational conditions and procedure
Analysis of isothermal spike forging was performed for two different frictions. The friction factors used
in the analyses are m = 0.3 and m = 0.6. The first value, m = 0.3, seems to represent an approximate value
of the friction factor in actual hot forging operations under lubricated conditions [28, 29]. The second value,
m = 0.6, is also a representative number in hot forging without lubrication[30]. The undeformed circular
billet has the dimension of 57.2 mm (2.25 in.) in height and 50.8 mm (2.0 in.) in diameter with chopped corner
to secure the stable initial position. The material used in the analysis is a +/3 microstructure Ti-6242-0.1S~
at 954 C (1750 F)[31]. The stress-strain rate relation is shown in Fig. 2. The shape and the dimension of the
spike forging dies are shown in Fig. 3. The velocity of the upper die used for the analysis was 25.4 mm/sec
(l.0in./sec), and the bottom die is stationary. The non-steady state deformation of spike forging was
analyzed in a step-by-step manner by treating it quasilinearly during each incremental deformation. The
incremental displacement of upper die for each step is chosen to be 0.02 times of the original workpiece
height.
The measure of convergence represented by IIAull/llull was chosen to be 0.00001.
Results and discussions
Fig. 4 shows the undeformed FEM grid used for analysis and the calculated grid distortions at die
displacements of 0.2, 0.4 and 0.58 Ho. For better comparison the grid distortions of both frictional cases are
superimposed on each other. As it can be seen from the figure, the general deformation behavior of spike
forging for the two different cases is similar. The solution shows that at an early stage, the deformation is
concentrated near the right upper corner of the workpiece where it touches the top die, while the rest of the
workpiece undergoes virtually no deformation. Because of this concentration, folding of the top surface

484

S.I. 0 8
250

35

300

25.o

~~-

200

200

150

in
I00 ~
I0.0

--

50
.5.0--

0.

0.0

8.0

4.0

12.0
Strain Rate,

16.0

20.0

sec -i

FIG. 2. Stress-strain rate relation of a +//microstructure Ti-6242-0.1Si at 954 C (1750 F).

Top Die

2 3.i mm

9.Smm ~1

25.4mm

i=.[

~ o

Y/'JJJJJ///////.>"
Bottom

Die

FIG. 3. Schematic drawing of spike forging dies.

Finite element analysis of metal forming processes with arbitrarily shaped dies

485

.2-

"- u~
o

-c" I

.2

u~

o.o

l.o
RADIUS/R o

2'.o

ko

(a)

2'.o

(b)

:.o
RADIUS/R o

-r"

u~

o-.:,
d-

d"

0.0

l.O

RADIUS/Ro
(c)

2.0

0.0

l'.O

2'.0

RADIUS/Ro
(d)

FIG. 4. Calculated FEM grid distortions at die displacements of (a) 0.0 Ho, (b) 0.2 Ho, (c)
0.4 Ho and (d) 0.58 Ho. Solid lines are for m = 0.6 and dotted lines for m = 0.3. Units are
multiples of the undeformed radius.
starts to take place during the early stage. As the deformation progresses, the deforming zone spreads
throughout the workpiece, forming the side surface bulge. Then folding from the side surface to the top and
the bottom dies takes place. Throughout the spike forging process the material near the center of the upper
surface undergoes virtually no defromation as can be seen from the grid distortions. It can also be seen
from the figure that the spike heights are not affected by the friction during the early stages of deformation.
The figure shows, however, that higher spike formed with higher friction and that the difference in spike heights
with different frictions is large at the die displacement of 0.58 Ho.
The predictions are compared with the experimentally measured spike cross sections. The experimental
measurements used for comparison were obtained by Altan et at. [27], who carried out spike forging tests
with lead at room temperature. These data were selected for an initial evaluation of the present FEM
because they were readily available. Lead being a strain-rate dependent material, it is expected to simulate
the flow of the other strain-rate dependent materials, at least approximately. This point is discussed later
again in connection with Fig. 10.
The predictions with m = 0.3 are compared with lubricated case at the die stroke of 0.58 Ho in Fig. 5. In

486

S.I. OH

-I

Theory at 58 %
x~.,erime nt oi 59%

l-=I
L9

0.0

1,0

2,0

RADIUSIR o

FIG. 5. Predicted cross sections compared with experiment at die displacement of 0.58 Ho.
Prediction: Ti-6242-0.1Si, 954 C (1750 F), m = 0.3. Experiment: Lead, Room Temperature,
Lubricated.
Fig. 6, the predicted cross sections are superimposed with those of the dry friction experiments at die
displacements of 0.3, 0.44 and 0.58 Ho. These figures show that the predictions are in excellent agreement
with experiments, even though the effects of the different material behaviors of lead and titanium at their
hot working range on the spike height are not known precisely. This point should be further investigated.
The details of the solutions are examined in terms of loads, velocity fields, neutral point locations,
changes in spike height, and effective strain distributions. The calculated load displacement relations are
shown in Fig. 7 for both frictions. The results reveal that the load increases slowly during early stages of
deformation and then it increases rapidly as the deformation progresses. Similar trends were found in spike
forging experiments with lead and aluminum[27]. The load values for both frictions are almost identical at
the early stages and the load is higher for the higher friction.

~ / ~ ~ "

Undeformed
- ----

Theory ot 50 %
Experiment at 50 %

(a)

(hi
- -

(C

Theory

ot

44%

- - - ~ ~

Theory at 5 8 %
Experiment at 5 9 %

(d)

FIG. 6. Predicted cross sections compared with experiments at die displacements of (a)
0.0Ho, (b) 0.3 Ho, (c) 0.44Ho and (d) 0.58 Ho. Prediction: Ti-6242-0.]Si, 954 C (1750 F),
m = 0.6 Experiment: Lead, Room Temperature, Dry Friction.

Finite element analysis of metal forming processes with arbitrarily shaped dies

487

300
--~--

rn : 0 . 3

--o--

rn=0.6

1.25

I.OO

200

0.75

0
0
0
x

I00

x/
x~X ~

x/x/

0.50

x~X/x~

x_x~ x~x
- 0.25

0
O.O

I
0.5

I
1.0
Displocemenl/R

5oo

FIG. 7. Calculated load-displacement relations when m = 0.3 and m = 0.6.

Figs. 8 and 9 show the local velocity distributions at the die displacements of 0.02, 0.2, 0.4 and 0.58 Ho
for m = 0.3 and m = 0.6, respectively. The velocities shown in the figures are relative quantities with
respect to the upper die. It can be seen from the figures that the deformations at the initial stages of both
frictional cases are concentrated near the right upper corner of the workpiece, while the rest of the portion
is rigid. The relative velocities of spike tip are 0.188 Vv and 0.204 Vv at the die displacements of 0.2 and
0.4Ho, respectively when m =0.3, and they are 0.193 VD and 0.238 Vv when m =0.6. Here, VD is the
downward velocity of the upper die. These results show that the differences in spike tip velocities with
different frictions are relatively small at early stages. The figures show, however, that the flow pattern
changes significantly at a die displacement of 0.58 Ho. At this stage the relative velocity of spike tip is about
1.25 Vo when m = 0.6 while it is 0.465 VD when m = 0.3.
As it can be se~n from the relative velocity patterns, a part of the workpiece flows into the central cavity
while the rest of the material flows radially. Because of these two different flow directions, a neutral point
forms on the upper die workpiece interface where the relative movement between die and material becomes
zero. Figs. 8 and 9 show the locations of neutral points. In the case of low friction the neutral point does not
form at die displacements of 0.02 and 0.2 Ho, suggesting that the material flows outward at early stages of
deformation. As the deformation progresses the neutral point appears near the fillet of the upper die.
Comparing Figs 8 and 9 it can be seen that the neutral point forms earlier for higher friction and that its
location is farther away from the center axis. In spite of the differences in the locations of the neutral points with
different frictions, the relative movements are small along the workpiece-upper die interface for both frictions
up to 0.4 Ho die displacement.
Fig. I0 shows the change of spike height as a function of die displacements. It can be seen from the
figure that the top surface of the spike moves downward for both friction cases during early stages of
deformation. The downward speed slows down and then the spike top surface moves up near the end of the
process for both cases. The spike tip reaches its minimum earlier for the higher friction case. The figure
shows that the minimum height of 34.8 mm (1.37 in.) is reached at die displacement of 31.5 mm (1.24 in.) in
case of m = 0.6, while that of 31.0 mm (1.22 in.) is reached at die displacement of 35.6 mm (1.40 in.) when
m = 0.3. The figure also reveals that the effect of friction on the spike height is minimal during the early
stage of deformation. The variations of spike height shown in Fig. 10 are in general agreement with earlier
experimental observations [24, 25].
In Fig. 10 the calculated spike height variations are compared with experimental data which were
obtained from isothermally forged lead and aluminum specimens with and without lubricant. The lead was
forged at room temperature and the aluminum at 31YC. It is interesting to note that "hot" aluminum and
"cold" lead behave similarly during isothermal forging. Fig. 10 shows that the spike heights measured from
both lead and aluminum specimens without libricant match the prediction with m = 0.3. The predictions of
the friction factor, m, may require further study because of the uncertainties in spike height variations for
the different materials involved. However, if it is assumed that the spike height variation is relatively
insensitive to the differences in material properties involved, it can be said that the friction factor is approx.
0.6 in the case of dry friction, and it is about 0.3 in the case of lubricated forging of lead and aluminum.
MS Vol. 24, No. 8--C

488

S.I. OH

vl

ill

1.0

all

II/

I"-T-'r-

1lttlll

..;.

(.D

,,,,~IZ

La.J

IIIIIII
o

~.Polllll

IIIIII1

l l l l l l l

11IIIII
e;"

lllltll

o"

1111111

lllllll
IIIIIII

1111111
0.0

1.0

~.0

0.0

RADIUS/Ro
(a)

01J

1~0

RADIUS/R 0
(b)

1.0

o a.
F"--"
"r-

Q
i

0"

/ //~

tl/////p

e;-

'~ /

l' ; I I I I//~.~

t. , t. .z , - -zj't- _ ' ~

IIII
o.o

Lo
RADIUS/R o
(c)

0.0

1.0

RADIUSIRo
(d)

FIG. 8. Relative velocity distributions at die displacements of (a) 0.02 Ho, (b) 0.2 Ho, (c)
0.4 Ho and (d) 0.58 Ho when m = 0.3. A denotes the location of neutral point.

;I.O

Finite element analysis of metal forming processes with arbitrarily shaped dies

1.0

lll/~

l--'r"

~D

I
I
I
I
I
I

,4'

d"

0.0

o ,_,

I~

I t # / / /t

.~ l l l / l l l

11[tltl/

: tTtlltl

1.0
RADIUSIR o

1 1 1 1 /

l~.O

i'.o

o.o

RADIUS/Ro

(a)

ILl
"r

489

(b)

..

0
i

<~<'

.,r

"I

/ i//L;

llillfl,

0.0

:
1.0

RADIUS/Ro

(c)

2.0

J o.

o"

tt
;.o

O.O

RADIUS/R o

(d)

FIG. 9. Relative velocity distributions at die displacements of (a) 0.02 Ho, (b) 0.2 Ho, (c)
0.4 Ho and (d) 0.58 Ho when m = 0.6. ~ denotes the location of neutral point.

i.o

490

S.I. OH
3.0
- .....

Calculated
Calculated

m = 03
m = 0.6

Experiment
Lead W i t h no L u b r i c a n t
o Lead With Fel- Pro- C300
AI W i t h n o L u b r i c a n t
t, AI W i t h F e I - P r o - C 3 0 0

Z.

t.O

oo

i
0.5

I
1.0

1.5

Displacement IRo

F1G. 10. Variation of spike heights when m = 0.3 and m = 0.6.

In the finite element method the local information such as strains, strain rates, and stresses can be easily
obtained. The information can be used not only to investigate the effect of flow stress on the overall
deformation but also to predict defect formation in metal forming processes. In the present study the total
effective strain concentrations are shown in Fig. 11 for both friction cases at die displacements of 0.2, 0.4
and 0.58 Ha. As it can be expected from the previous discussions, the general patterns of strain
distributions are similar to each other, showing almost no deformation near the spike tip and the highest
strain concentration along the band joining the right upper corner of the specimen and middle of the height
near the center. It is, however, interesting to note that the strain gradient is higher for the higher friction
case, and that the deformation is more uniform in the case of lower friction. The difference in the strain
gradient is more evident near the bottom surface of the workpiece.

CONCLUSION

A method is established to discretize the die boundary conditions in the framework of rigid viscoplastic finite element method. The method is natural to FEM and
general enough to treat the arbitrarily shaped dies in a unified manner. Improvements
are made on the computational efficieneies of rigid viscoplastic FEM by introducing
an automated initial guess generation scheme. It seems that the discretization of die
boundary conditions, together with the initial guess generation scheme, is an essential
step toward the development of a general purpose FEM code for metal forming analysis.
The present work is now being expanded to simulate metal flow in relatively
complex two-dimensional metal forging applications. It is expected that this future
effort will, eventually, allow the design of preform shapes, involving massive forming
operations in industrial production processes.
Solutions of the spike forging process with two different frictional conditions are
obtained by the present method. The deformation mechanics of the spike forging
process and effects of friction on the process are discussed. The solutions are

Finite element analysis of metal forming processes with arbitrarily shaped dies

I
i

O"

~______
0.?.5.~
*

O.ZS--'~

O.Z~--~"

/
0

'% l
I

(a)

(d)

(b)

"

(c)

(e)

o8

O.~

(f)

FIG. 11. Effective strain distributions at die displacements of (a) 0.2 Ho, (b) 0.4 Ho and (c)
0.58 Ho when m = 0.3 and (d) 0.2 Ho, (e) 0.4 Ho and (f) 0.58 Ho when m = 0.6.

491

492
c o m p a r e d with e x p e r i m e n t s . T h e
a g r e e m e n t with e a c h other.

S.I. OH
comparison shows

that t h e y are in e x c e l l e n t

Acknowledgements--This paper is based on information developed in Processing Science Program sponsored by the Air Force Wright Aeronautical Laboratories, AFSC, under contract F33615-78-C-5025, with
Dr. Harold L. Gegel as program manager. The author would like to acknowledge gratefully the helpful
discussions with Professor Shiro Kobayashi of the University of California, Berkeley. The author also
would like to express his thanks to Drs. T. Altan and G. D. Lahoti at Battell's Columbus Laboratories for
their encouragement and suggestions during the course of this investigation.

REFERENCES
1. C. H. LEE and S. KOaAVASnl,New solutions to rigid-plastic deformation problems using a matrix
method. Trans. ASME, J. Engr. for Ind. 95, 865 (1973).
2. S. N. SHAH,C. H. LEE and S. KOEAYASm,Compression of tall, circular, solid cylinders between parallel
flat dies. Proc. Int. Conf. Prod. Engr., Tokyo, p. 295 (1974).
3. H. MATSUMOTO,S. I. OH and S. KORAYASHI,A note on the matrix method for rigid-plastic analysis of ring
compression. Proc. 18th MTDR Conf., London, p. 3 (Sept. 1977).
4. C. C. CHEN and S. KOBAYASHi,Rigid-plastic finite-element analysis of ring compression. Applications
of Numerical Method of Forming Processes, ASME, AMD, Vol. 28, p. 163 (1978).
5. C. C. CHEN, S. I. OH and S. KOBAYASHI,Ductile fracture in axisymmetric extrusion and drawing--Part
1. Deformation mechanics of extrusion and drawing. Trans. ASME, J. Engr. for Ind. 101, 23 (1979).
6. C. C. CHEN and S. KOBAVASHI,Rigid-plastic finite element analysis of plane-strain closed-die forging.
Process Modeling--Fundamentals and Applications to Metals, ASM, p. 167 (1980).
7. S. I. OH and S. KOS^YASNI,Finite element analysis of plane-strain sheet bending. Int. J. Mech. Sci. 22,
538 (1980).
8. S. I. On and S. KOBAYASHI,Workability of aluminum alloy 7075-T6 in upsetting and rolling. Trans.
ASME, J. Engr. Ind. 98, 800 (1976).
9. S. I. 08, C. C. CnEN and S. KOBAYASHI,Ductile fracture in axisymmetric extrusion and drawing--Part
2. Workability in extrusion and drawing. Trans. ASME, J. Engr. for Ind. 101, 36 (1979).
10. S. I. OH, S. KOBAVASI~Iand E. G. THOMSEN, Calculation of frictional stress distribution at the
die-workpiece interface for simple upsetting. Proc. 3rd North American Metalworking Research Conf,,
Pittsburgh, p. 159 (1975).
11. S. I. OH, N. REBELO and S. KOBAYASHI. Finite element formulation for the analysis of plastic
deformation of rate-sensitive materials for metal forming. Metal Forming Plasticity, IUTAM Symposium, Tutzing, Germany, p. 273 (1978).
12. R. HILL, New horizons in the mechanics of solids. J. Mech. Phys. Solids 5, 66 (1956).
13. O. C. ZmlqKIEWIEZ,The Finite Element Method, 3rd Ed. McGraw-Hill, New York (1977).
14. I. FRIED,Finite element analysis of incompressible material by residual energy balancing. Int. Z Solids
Structures 10, 993 (1974).
15. D. S. MALKUS,A finite element displacement model valid for any value of the compressibility. Int. J.
Solids Structures 12, 231 (1976).
16. J. C. NAGTeGAAL.D. M. PARKSand J. R. RICE, On numerically accurate finite element solutions in the
fully plastic range. Comput. Meth. Appl. Mech. Engng 4, 153 (1974).
17. P. HARTLeY,C. E. N. STUROESSand G. W. ROWE, Friction in finite-element analyses of metalforming
processes. Int. J. Mech. Sci. 21,301 (1979).
18. Y. YAMADAand H. HlP.AKAWA,Large deformation and instability analysis in metal forming process.
Applications of Numerical Method to Forming Processes, ASME, AMD, 28, 27 (1978).
19. A. S. WIl~, An incremental complete solution of the Stretch-forming and deep-drawing of a circular
blank using a hemispherical punch. Int. J. Mech. Sci. lg, 23 (1976).
20. S. KEy and R. D. KRmG, On application of the finite element method to metal forming process--1.
Comput. Meth. Appl. Mech. Engng 17, 597 (1978).
21. E. H. LEE, R. L. MALLEYrand W. H. YANO, Stress and deformation analysis of the metal extrusion
process. Comput. Meth. Appl. Mech. Engng 10, 339 (1977).
22. J. F. LVNeSS, D. R. J. OWEN and O. C. ZIENKIEVOCZ,Finite element analysis of the steady flow of
non-Newtonian fluids through parallel sided conduits. Int Syrup. on Finite Element Method in Flow
Problems, Swansea, p. 489 (1974).
23. H. KuDo, An upper-bound approach to plane strain forging and extrusion--I. Int. J. Mech. Sci. 1, 57
(1960).
24. s. c. JAnq, A. N. BRAMLE,C. H. LEE and S. KOBAYASHI,Theory and experiment in extrusion forging.
Proc. l lth Int. Conf. of MTDR, Birmingham, England, p. 1097 (1971).
25. J. A. NEWlqHAMand G. W. Rowe, An analysis of compound flow of metal in a simple extrusion/forging
process. Z Inst. Metals 101, 1 (1973).
26. J. W. H. PRICE and J. M. ALEXANDER,Specimen geometries predicted by computer model of high
deformation forging. Int. J. Mech. Sci. 21, 417 (1979).
27. T. ALTAN,A. H. CLAUREN,G. D. LAHOTI,A. R. ROSmqFIELDand T. L. SUBP,AMANIAN,Effect of hot
forging variables upon microstructure and properties of metals and alloys. Final Pep., AFOSR-76-2971.
28. J. R. DOUaLASand T. ALTAN,Characteristics of forging presses: determination and comparison. Proc.
13th Int. Conf. of MTDR, p. 535 (1972).
29. J. R. DOUOHLASand T. ALTAN,Flow stress determination for metals at forging rates and temperatures.
Trans. ASME, Z Engr. for Ind. 97, 66 (1975).

Finite element analysis of metal forming processes with arbitrarily shaped dies

493

30. G.D. LAHOTI,P. S. RAGHUPATHIand T. ALTAN,Application of the radial forging process to cold and warm
forging of cannon tubes. Final Rep., DAAA22-78-CO109 (1978).
31. S. L. SEML~TIN,G. D. LAHOTIand T. ALTAN,Determination and analysis of flow stress data for Ti-6242
at hot working temperatures. Process Modeling--Fundamentals and Applications to Metals, ASM, p.
387 (1980).