International Journal of Mechanical Sciences 84 (2014) 147–157

Contents lists available at ScienceDirect

International Journal of Mechanical Sciences

journal homepage: www.elsevier.com/locate/ijmecsci

An analytical solution for the estimation of the drawing force

in three dimensional plate drawing processes
Andrea Panteghini n
DICATAM, University of Brescia, Via Branze, 43, 25123 Brescia, Italy

art ic l e i nf o a b s t r a c t

Article history: An analytical solution for the estimation of the drawing force requested to perform a three dimensional
Received 7 June 2012 drawing of a rectangular plate is presented.
Received in revised form It has been developed using the limit analysis technique, on the basis of a three-dimensional velocity
25 March 2014
field, under the main assumptions of constant friction and perfect plasticity.
Accepted 17 April 2014
Available online 28 April 2014
To overcome the limitations given by these hypothesis, some extensions of the proposed upper
bound model are presented, accounting for the material strain hardening and for the Coulomb friction
Keywords: between the plate and the die, thus avoiding the need to calibrate physically meaningless coefficients.
Drawing process The effectiveness of the proposed results is demonstrated by comparison with both experimental
Plate drawing
results and new numerical solutions.
Limit analysis
& 2014 Elsevier Ltd. All rights reserved.
Drawing force

1. Introduction Moreover, if the numerical modelling involves not only the

The optimization of the plate drawing process has been slightly
improved by the recent theoretical and numerical developments
in mechanics: in the industrial practice, the determination of an
appropriate working condition during the drawing process and the
design of the optimum die shape and dimensions are still based on
empirical knowledges, and on trial-and-error approaches. This is
mainly due to the difficulties to develop simple techniques
allowing the optimization of the die geometry as a function of
the other process parameters (plate geometry, friction coefficient,
area reduction, etc.), in order to minimize the production costs.
One of the main tools today available to study metal forming
processes is the Finite Element Method (FEM). This numerical
procedure models rather accurately the process mechanics as a
function of the process parameters but, unfortunately, each
different combination of possible die geometry, area reduction,
or friction coefficient requires a single numerical analysis.
The optimization of a real industrial process implies parametric
studies that involve many different values of the parameters (for
example, different die geometries), and, using this method, several
different models should be developed [8]. Moreover, the numerical
simulation of metal forming processes requires in general a
complex three-dimensional non-linear analysis, and a large
amount of time must be spent to obtain a solution for each single
set of parameters.
n
Tel.: þ 39 030 3711218; fax: þ 39 030 3711312.
E-mail address: andrea.panteghini@ing.unibs.it
computation of the drawing stress as a function of the die
geometry, but also other complex features, such as, for instance,
the minimization of the residual stresses, or the prediction of
defects into the final piece, one should adopt rather complicate
constitutive models. For example, during the drawing process, the
material is subjected to non-monotonic loadings, and the assump-
tion of a kinematic strain hardening law is fundamental to the
correct evaluation of the residual stress profiles in the final
product [9]. Also the prediction of defects into the final piece
requires a material modelling richer than that furnished by the
von Mises isotropic plasticity, in order to simulate accurately the
material response at very high hydrostatic pressures [11].
Obviously, a more complex constitutive modelling of the material
requires higher computational costs.
For these reasons, even if the numerical analyses are probably
the most powerful tool available to simulate metal forming
processes, analytical design procedures, which allow (at least) an
initial design of the process are very important. Moreover, it
should be noted that the most adopted design procedures of metal
forming processes in the engineering practice are still based on the
limit analysis technique. This analytical method has been exten-
sively employed to study the material flow through conical
converging dies for rods and wires (see, for instance, [2,3,5]).
Recently, the upper bound (UBET) solutions reported in these
works, based on the assumption of perfect plasticity, have been
extended to consider the strain hardening of the material during
the drawing process [7,10]. Unfortunately, for dies with more
complex geometry, only a few theoretical approaches to drawing
or extrusion have been published (for instance, see [12] or [17]).

http://dx.doi.org/10.1016/j.ijmecsci.2014.04.012
0020-7403/& 2014 Elsevier Ltd. All rights reserved.
148 A. Panteghini / International Journal of Mechanical Sciences 84 (2014) 147–157
This is mainly due to the difficulties to develop analytical
expressions in a closed form to compute the drawing force taking
into account the three-dimensional die shapes.
Rubio et al. [14] have recently published an analytical solution
for two-dimensional (2D) plate drawing process assuming a
perfect plastic material under plane strain condition. Their solu-
tion, which has been extended in [13] to consider the material
strain hardening, is relatively simple. Unfortunately, since it has
been developed for a 2D problem, it cannot be employed directly
to optimize an industrial 3D plate drawing process.
In order to obtain estimations of the drawing force for three
dimensional die shapes, the limit analysis techniques have been
extended using the SERR (Spatial Elementary Rigid Region) tech-
nique (see, for instance, [15,16]). This mixed numerical–analytical
technique is based on the subdivision of the deformation zone into
tetrahedral blocks. The velocity of each block is then determined
numerically, by imposing the compatibility conditions between
the block interfaces. Then, this method does not furnish an
expression for the drawing force in a closed form.
In this paper, a new analytical solution for the drawing force in
three dimensional plate drawing processes is presented.
The proposed analytical model, based on the limit analysis
technique, is obtained from a new three dimensional velocity field,
assuming constant friction between die and plate, and perfect
plasticity. The new solution is also extended considering Coulomb
friction and accounting for the material strain hardening.
The proposed solutions is finally validated by comparison with
new FEM solutions.
2. The velocity-based analytical solution
The upper bound solution here presented allows the estimation
of the drawing stress szd necessary to perform a reduction of a
rectangular plate, with initial dimensions equal to 2bi  2hi to the
final dimensions 2bf  2hf .
Let assume that the initial rectangular shapes of the die and of
the workpiece are similar, i.e., that the ratio between the width
and the height of the die at the entrance is equal to that of the
plate. If this condition is not respected, the plate becomes initially
in contact only with two faces of the die (top-bottom or left-right),
and the reduction is initially limited to one dimension (height, or
width), until all the lateral surface of the plate is in contact with
the die. It should be noted that the condition of different dimen-
sional ratios between plate and die is not common in the industrial
practice. In fact, the nonhomogeneous deformation of the plate
faces initially free causes high friction stresses, that provoke a
quick wearing of the die surfaces and a worst quality of the final
product (possible surface defects).
Under this assumption, the die geometry can be fully described
as a function of the plate initial and final dimensions, and of a
single inlet reduction angle (for example, as a function of height
reduction angle α).
The problem here studied has two symmetry planes, and then,
only a quarter of the plate is considered.
Initially, the material is assumed to be rigid perfectly plastic,
and the shear along the contact interfaces between die and plate is
modelled assuming a constant friction. This model will be
extended in the next sections, considering a Coulomb friction
model to describe the interactions between die and plate, and
accounting for the strain hardening of the material.
2.1. The assumed three-dimensional velocity field
An upper bound analytical approach is based on a postulated
velocity field, that must respect the boundary conditions, i.e., for
this problem, the plate symmetry planes, and the constraints given
by the contact with the die surfaces.
Here we assume that the plate in the plastic zone can be
subdivided into eight blocks, in which the material is subjected to
rigid body motion.
The blocks subdivision here adopted is reported in Fig. 1. This
particular geometry of the blocks assures two main features, both
of them related to the directions of the velocities of the blocks
themselves. In particular, the directions of both the absolute
velocity of each block and that of the relative velocity between
two neighboring blocks are a priori known.
The first feature derives from the external boundary condi-
tions: the plate has two symmetry planes (yz and xz in Fig. 2a), and
it is constrained by the two planes π1 and π2 (Fig. 2b), on which it
is in contact with the die surfaces. Hence, the direction of the
absolute velocity of each rigid block can be determined a priori, as
a function of the interactions of the block itself with these planes
(Fig. 3a).
The second feature results from the compatibility condition
between each couple of neighboring blocks: the only admissible
relative motion between two neighboring blocks is a slip along the
face that they share (the discontinuity surface). This condition
implies that the relative velocity between each couple of contig-
uous blocks must lie on the plane of their discontinuity surface
(Fig. 3b). In order to respect this geometric constraint, two
different features have been adopted for the block geometry. First,
the block subdivision has been obtained in such a way that the
absolute velocities of each couple of neighboring blocks are
coplanar. This obviously implies that also their relative velocity
lies on the plane of their absolute velocities (the plane of the
velocities in Fig. 3b). Second, the discontinuity surface (i.e., their
shared face) is constructed to be always normal to this velocity
plane. In this way, for each couple of rigid blocks, the direction of
their relative velocity is initially known, since it is given by the
intersection between these two planes.
The geometry of each block can be described as a function of
the die geometry, and of a distance η from the plate cross-section
at the die exit and the point B, placed on the drawing axis. The die
geometry is initially known as a function of the plate initial and
final dimensions, and of the reduction angle α on the plane xz,
whilst η will be determined in order to optimize the upper bound,
i.e., to minimize the computed drawing stress. Both the die and the
blocks geometry can be easily determined from Fig. 4: all the
Fig. 1. Assumed blocks subdivision for the plastic zone.
 A. Panteghini / International Journal of Mechanical Sciences 84 (2014) 147–157
Fig. 2. Symmetry planes and boundary conditions for the rigid blocks.
Fig. 3. Interaction between two generic blocks Si and Sj , the symmetry plane, and
the die plane. The figure shows also that the discontinuity surface between Si and Sj
is always normal to the plane of the velocities of Si and Sj .
geometric quantities necessary in the presented solution are
reported in Appendix.
At the exit of the die (block S2), the material is in motion, with a
prescribed constant absolute velocity equal to v2 . The modulus of
the absolute velocity of the material at the entrance of the plastic
zone v1 (i.e., the modulus of the velocity of the block S1) can be
determined using the mass conservation equation, which implies
bf hf
jv1 j ¼ jv2 j ð1Þ
bi hi
With reference to Figs. 1 and 5, one can observe that the
absolute velocities v1 , v2 , v3a and v3b of the blocks S1, S2, S3a and
S3b respectively are coplanar with the plane xz. Moreover, because
of the symmetry planes yz and xz, v1 and v2 are parallel to the z
axis, while, because of the constraint given by the plane π2, v3a and
v3b must be directed as the segment EF. The discontinuity surfaces
sBDE and sBJI between S1 and S3a, and S2 and S3b are normal to the
plane of the velocities xz. A two-dimensional representation of
these two blocks in the plane xz is reported in Fig. 4a. Since v1 and
v2 are known, and all the directions of the unknown velocities are
given, the moduli of v3a and of the relative velocity Δv1  3a
between S1 and S3a can be computed by the trivial solution of
the triangle in Fig. 6a, that yields
jv3a j ¼ jv1 jsecðα þ θÞ cos θ ð2Þ
jΔv1  3a j ¼ jv1 jsecðα þ θÞ sin α ð3Þ
where θ is the incident angle of sBDE on the plane xz with respect
to the plate cross-section at the die entrance (Fig. 4a).
Analogously, solving the triangle of Fig. 6b. One obtains
jv3b j ¼ jv2 jsecðα  φÞ cos φ ð4Þ
149
jΔv2  3b j ¼ jv2 jsecðα  φÞ sin α ð5Þ
where φ is the incident angle of sBJI on the plane xz with respect to
the plate cross-section at the die exit (Fig. 4a).
It is important to observe that, in general, jv3a j ajv3b j. However,
the blocks S3a and S3b are not neighboring blocks, since they do not
share any face (Fig. 1), and they do not slip one on the other, but
they both slide on S5a. Hence, their different velocities do not
imply a violation of the compatibility conditions. This is a simple
extension to the three dimensional case of what results in other
two dimensional velocity fields presented in the literature (see, for
instance, [14]): with reference to Fig. 7, one can observe that the
velocity of the point B is different if it is considered belonging to
the block 1 or to the block 2. However, the two-dimensional
velocity field is consistent, since both the blocks 1 and 2 slide on
the block 3.
The velocities v4a and v4b of the blocks S4a and S4b can be
determined using the same considerations adopted to compute v3a
and v3b . In this case, their directions are coplanar with the plane yz
and they must be parallel to the segment CG. A planar representa-
tion in plane of the velocities yz is reported in Fig. 4b. It result
jv4a j ¼ jv1 jsecðβ þ ξÞ cos ξ ð6Þ
jv4b j ¼ jv2 jsecðβ  ωÞ cos ω ð7Þ
where β is the reduction angle in the yz plane, ξ and ω are the
incident angles of the discontinuity surfaces sBCD and sBHI on the
plane yz with respect to the plate cross section at the die entrance
and exit respectively (Fig. 4b).
The relative velocities Δv1  4a and Δv2  4b between the blocks
S1 and S4a, and S2 and S4b lie on the planes BCD and BHI
respectively and are parallel to the segments BC and BH respec-
tively (Fig. 4b). As done for Δv1  3a and Δv2  3b they can be
computed as
jΔv1  4a j ¼ jv1 jsecðβ þ ξÞ sin β ð8Þ
and
jΔv2  4b j ¼ jv2 jsecðβ  ωÞ sin β ð9Þ
The blocks S5a, S3a and S3b are contiguous, and their absolute
velocities v5a , v3a and v3b have directions that lie on the plane π2
(Fig. 8). In particular, S3a and S3b are in contact only with the top
surface of the die (i.e. their velocity directions belong to the plane
π2), while S5b interacts also with its lateral face through the
segment DI (i.e. its velocity direction must lie also on the plane
π1). Hence, while v3a and v3b are parallel to the segment EF, v5a
must be directed as the segment DI. Both the discontinuity
surfaces sBDF and sBIF between S5 and S3a, and S5 and S3b
respectively, are normal to the plane of the velocities π2. A planar
representation of the velocities in the plane π2 is reported in
Fig. 4c.
The directions of the relative velocities Δv3a  5a and Δv3b  5a
are given, since they lie on π2 and also on the plane BDF and BIF
respectively: they must be parallel as the segment DF and IF
respectively. Since v3a and v3b are known, and the directions of v5a ,
150 A. Panteghini / International Journal of Mechanical Sciences 84 (2014) 147–157
Fig. 4. 2D representation of the blocks in their relative planes of the velocities.
Fig. 5. Interactions between the symmetry planes, the die planes, and the blocks
S1 , S2 , S3a and S3b .
Δv3a  5a and Δv3b  5a are given, jv5a j and jΔv3a  5a j can be
computed by the trivial solution of the triangle of Fig. 9a, which
gives
jv5a j ¼ jv3a jsecðγ þ χ Þ cos χ
jΔv3a  5a j ¼ jv3a jsecðγ þ χ Þ sin γ
where γ is the reduction angle on the plane π2.
The modulus of Δv3b  5a can be obtained solving the triangle of
Fig. 9b, which yields
jΔv3b  5a j ¼ jv3b jsecðγ  κ Þ sin γ :
With the same considerations, it is possible to compute also the
modulus of the absolute velocity jv5b j of the S5b block, and the
relative velocities jΔv4a  5b j and jΔv3b  5b j between S5b  S4a and
S5b S4b respectively: v4a , v4b , v5b , Δv4a  5b and Δv3b  5b have
directions that lie on the plane π1. A planar representation of the
block velocities in this plane is reported in Fig. 4d. The disconti-
nuity surfaces sBDG and SBGI are normal to the plane of the
velocities π1. With reference to Fig. 4d, it results
jv5b j ¼ jv4a jsecðψ þ δÞ cos ψ
jΔv4a  5b j ¼ jv4a jsecðδ þ ψ Þ sin δ ð14Þ
jΔv4b  5b j ¼ jv4b jsecðδ  ζ Þ sin δ ð15Þ
where δ is the reduction angle in the π1 plane, ψ and ζ are the
incident angles of sBDG and sBGI on π1 (Fig. 4d).
Finally, S5a and S5b are adjacent, and they are in motion with
absolute velocities v5a and v5b respectively. Moreover, v5a and v5b
are parallel, and their direction lies on the discontinuity surface
sDBI. Then, S5a and S5b slide along this surface, with a relative
velocity that can be easily computed as
jΔv4a  5b j ¼ jjv5a j  jv5b jj ð16Þ
2.2. The upper bound solution under constant friction hypothesis
ð10Þ An upper bound solution is based on the estimation of the
powers spent for ideal plastic deformation, for sliding along the
discontinuity surfaces, and for friction. The solution is then
ð11Þ
obtained equating these internal powers to the external power,
which, in this case, is a function of the drawing stress. Such
estimation is obtained postulating a kinematically admissible
velocity field in which, in general, the internal stresses violate
the equilibrium, and then, only an upper bound of the exact
ð12Þ
solution can be obtained. The velocity field described in the
previous section is based on the subdivision of the plastic zone
into rigid blocks, i.e., the plastic deformation of each single block is
nil. For this reason, the power spent for the ideal plastic deforma-
tion Wp is nil.
Assuming a Tresca rigid-perfectly plastic material, with yield
stress equal to s0, the power spent for the slip along the
discontinuity surfaces can be written as
nd Z s  
0
Wd ¼ ∑ Δvk dΣ k ð17Þ
ð13Þ k¼1 k s 2
 A. Panteghini / International Journal of Mechanical Sciences 84 (2014) 147–157
Fig. 6. Geometric relationship between the velocities of the blocks S1 and S3a (a) and S2 and S3b (b).
Fig. 7. Geometry and design parameters for a literature plane strain upper bound
solution [14].
Fig. 8. Interactions between the die planes, and the blocks S3a , S3b , and S5a .
where nd is the number of the discontinuity surfaces, sk is the k-th
discontinuity surface between the two generic blocks i and j, Δvk
is the k-th relative velocity between the two generic blocks i and j,
and Σk is the area of sk. One obtains
s0     
W d ¼ ðΣ 1  3a Δv1  3a  þ Σ 1  4a Δv1  4a  þ Σ 2  4b Δv2  4b j
2
þ Σ 2  3b jΔv2  3b j þ Σ 3a  5a jΔv3a  5a j þ Σ 3b  5a jΔv3b  5a j
þ Σ 4a  5b jΔv4a  5b jþ Σ 4b  5b jΔv4b  5b j þ Σ 5a  5b jΔv5a  5b jÞ
All the area measures Σk of the discontinuity surfaces sk can be
easily determined from Fig. 4, and their values are reported in
Appendix.
The intensity of the shear due to the friction between the plate
and the die is assumed to be constant in each k-th die surface πk
and equal to
mπ k s0
τπ k ¼
2 the Coulomb friction coefficient μ.
where mπ k is the shear factor along the k-th surface πk.
The power dissipated by friction Wf can be estimated as
nd nf Z
Wf ¼ ∑ ∑ τπj jvk jdΣ k;πj
j¼1k¼1 sk;π
j
where nd is the number of surfaces of the die, nf is the number of
surfaces of the blocks in contact with the die, vk is the absolute
velocity of the k-th block, and sk;π j is the surface of the k-th block
in contact with the j-th die surface πj, of area equal to Σ k;π j . With
reference to Figs. 1 and 2, we assume two different shear factors,
151
equal to mπ 2 and mπ 1 for the top (plane π2) and lateral (plane π1)
surfaces of the die respectively. Eq. (20) can be expanded as
s0      
Wf ¼ ½mπ 2 ðΣ 5a;π 2 v5a  þ Σ 3a;π 2 v3a  þ Σ 3b;π 2 v3b Þ
2
þmπ 1 ðΣ 5b;π 1 jv5b j þ Σ 4a;π 1 jv4a j þ Σ 4b;π 1 jv4b jÞ ð21Þ
The measures of the contact surfaces area Σ k;π j are reported in
Appendix.
The external power is equal to
W e ¼ jv2 jhf bf szd ð22Þ
The average drawing stress szd can be finally obtained from the
following equation:
We ¼ Wd þWf ð23Þ
which gives
Wd þWf
szd ¼ ð24Þ
hf bf jv2 j
Note that jv2 j is the prescribed velocity of S2, and it is a dummy
variable, since all moduli of the blocks velocities are scaled by jv2 j.
As described above, fixed initial and final geometry of the plate,
all the geometric quantities involving this upper bound solution
can be expressed as a function of the reduction angle α in the
plane xz and of a distance η between a point on the plate axis and
the plate cross section at the die exit (Fig. 4). Hence, for a single set
of process parameters, this last equation furnishes a class of
possible values of the drawing stress szd as a function of η. In
order to estimate an optimized upper bound solution for the
drawing stress szd, Eq. (23) can be minimized with respect to η.
Thus, an optimized upper bound solution for the drawing stress
can be computed as
s zd ¼ minfszd ; ηg ð25Þ
ð18Þ
2.3. Extension to the Coulomb friction hypothesis
In this section, it is described a three-dimensional lower bound
analytical solution, that furnishes an estimation of an average
pressure on each surface contact between plate and die.
Once each average pressure is known, one can evaluate the
shear factors mπ 1 and mπ 2 as a function of the die geometry and of
ð19Þ
The three-dimensional geometry here considered is reported in
Figs. 10 and 11. The origin of the reference coordinate system is
placed on the intersection point between the plate axis and the
ð20Þ symmetry axis of the top surface of the die (Fig. 11). It is assumed
that this is a principal coordinate system for the stress field, i.e. sx,
sy and sz are principal stresses.
In this reference coordinate system, the measure of plate height
along the die axis can be expressed as a function of the z
coordinate as follows:
h ¼ z tan α ð26Þ
152 A. Panteghini / International Journal of Mechanical Sciences 84 (2014) 147–157

Fig. 9. Geometric relationship between the velocities of the blocks S3a and S5a (a) and S3b and S5a (b).

the z direction gives

sz ðh dbþ b dhÞ þ bh dsz þ ½qx bð tan α þ μÞ þ qy hð tan β þ μÞ dz ¼ 0
ð31Þ
The equilibrium along the x and y directions gives
sx ¼ qx ðμ sin α  cos αÞ ð32Þ
and
sy ¼ qy ðμ sin β  cos βÞ ð33Þ

In order to impose the Tresca yield criterion, one must determine

which are the maximum and the minimum principal stresses. Let
observe that, for the drawing process, sz is always positive (i.e. it is
a traction stress), while sx and sy are negative. Hence, sz is always
the maximum principal stress.
Fig. 10. Assumed three-dimensional stress field for the lower bound solution. However, at this point, it is unknown which of sx and sy is the
minimum principal stress. One could assume, for instance, a
parametric stress field, in which sx is the minimum principal
stress whilst sy is the intermediate one, expressed as a linear
combination of sx and sz, with the condition that sy r 0, i.e.:
sy ¼ ksx þ ð1  kÞsz ð34Þ

where 0 r sz =ðsz  sx Þ rk r1. Once a solution is obtained, one

can maximize the drawing stress as a function of k in order to
optimize the lower bound. The other possible stress field could be
obtained assuming sy as minimum principal stress and expressing
sx as a linear combination of sy and sz. Obviously, if sx ¼ sy , the
two stress fields are coincident. It will be shown that this is the
condition that maximize the lower the bounds.
Let assume, for instance, that sx is the minimum stress:
Fig. 11. Planar representation of the assumed three-dimensional stress field for the observing that sz is positive and sx is negative, one can write
lower bound solution. the Tresca yield criterion as follows:
sz  sx ¼ s0 ð35Þ
Hence, the z coordinate at the die entrance (z ¼ zi ) and at the die
exit (z ¼ z) are respectively equal to Solving this equation for sx and substituting the result into Eq.
(32), gives
zi ¼ hi cot α ð27Þ
s0  sz
and qx ¼ ð36Þ
cos α  μ sin α
z ¼ hf cot α ð28Þ Analogously, substituting sx obtained from Eq. (35) into Eq. (34),
In general, the intersection point between the plate axis and the and placing the result into Eq. (33) yields
symmetry axis of the lateral surface of the die is not coincident ks0  sz
qy ¼ ð37Þ
with the origin of the assumed reference coordinate system. This cos β  μ sin β
condition occurs only when the die is a truncated right pyramid.
Calling z0 the z coordinate of this intersection point, with reference Finally, substituting these two last expressions into Eq. (31), one
to Fig. 11, the width of the plate can be expressed as obtains the following linear differential equation:
   
b ¼ ðz þ z0 Þ tan β ð29Þ dsz 1 1 a1 b1
¼  sz þ þ ðs0  sz Þ þ ðks0  sz Þ ð38Þ
dz z þ z0 z z z þz0
where
  where
hi  hf
z0 ¼ bi  hi cot α ð30Þ tan α þ μ
bi  bf a1 ¼ ð39Þ
ð cos α  μ sin αÞ tan α
The postulated stress field is reported in Figs. 10 and 11.
With reference to these figures, and omitting the dependence tan β þ μ
b1 ¼ ð40Þ
of the quantities with respect to z, the equilibrium equation along ð cos β  μ sin βÞ tan β
 A. Panteghini / International Journal of Mechanical Sciences 84 (2014) 147–157 153

  
Let observe that in the plastic zone it results z Z0, z þ z0 Z 0 and ða1 þ 3b1  2Þz0 zi f ða1  b1 Þ f 5 ða1  b1 Þ
 þ 7  z0 zi z2i ð49Þ
sz Z 0, with the initial condition that, at the die entrance (z ¼ zi ), a1 þ b1  1 a1  1 a1 þ b1  1
sz ¼ 0. The parameters a1 and b1 are always positive terms.1 Z
Moreover, from Eqs. (34) and (35), considering that in the drawing zi qy h s0 sec β tan α
Fy ¼ dz ¼
process must be sx r0 and sy r 0, it results that ks0 Z sz . Hence, z cos β ð cos β  μ sin βÞða1 þ b1  2Þ
(
the right-hand side of the differential equation (38) is the sum of ða1  b1 Þz0 ½ðf 4 1Þz þ ð1  f 5 Þzi 
three negative terms, and sz will be a monotonic decreasing  þ z 2  z2i
a1 þ b1  1
function of z. Since sz is positive in the plastic zone, the maximum " #
value will be at the die exit, i.e. for z ¼ z. ða1  b1 Þf 1 ðz0 þ zi Þz2i 
þ a1 þ b1  f
Since z o zi , and sz ðzi Þ ¼ 0, the maximum value of sz ðz; kÞ will a1  1 ð1 þa1 Þz0 9
be when the term ðks0  sz Þ is maximum. This condition is  a1 þ 1   #)
z z þz0 b1
satisfied when k ¼1, that implies sx ¼ sy . Under this condition, f 8 ð50Þ
Eq. (38) can be rewritten as zi z0 þ zi
    
d sz 1  a1 1  b1 a1 b1 where
¼  sz þ þ s0 þ ð41Þ  
dz z z þ z0 z z þ z0 z
f 2 ¼ 2 F 1 1; 1 þ a1 þ b1 ; 1 þa1 ;  ð51Þ
This differential equation can be integrated in a closed form, under z0
the condition that sz ðzi Þ ¼ 0. The solution is2  
( " # z
 a1  1   f 3 ¼ 2 F 1 1; 1 þ a1 þ b1 ; 1 þa1 ;  i ð52Þ
s0 z z þ z0 b1  1 z0
sz ¼ ða1 þ b1 Þ 1 
a1 þ b1  2 zi zi þ z0  
  "  a1  1  b1  1 #)
z
a1  b1 z z þ z0 f 4 ¼ pFq 1; 1; 1  a1  b1 ; f2; 1  a1 g;  ð53Þ
þ f0 f1 ð43Þ z0
1  a1 zi zi þz0
 

z
where f 5 ¼ pFq 1; 1; 1  a1  b1 ; f2; 1  a1 g;  i ð54Þ
z0
 
z  
f 0 ¼ 2 F 1 1; 2 a1  b1 ; 2  a1 ;  ð44Þ
z
z0 f 6 ¼ pFq 1; 1; 2  a1  b1 ; f2; 2  a1 g;  ð55Þ
z0
 
zi  
f 1 ¼ 2 F 1 1; 2 a1  b1 ; 2  a1 ;  ð45Þ
z
z0 f 7 ¼ pFq 1; 1; 2  a1  b1 ; f2; 2  a1 g;  i ð56Þ
z0
3
2 F 1 is the hypergeometric function, defined as  
1 ðrÞk ðsÞk k z
f 8 ¼ 2 F 1 1; 1 þ a1 þ b1 ; 2 þa1 ;  ð57Þ
2 F 1 ðr; s; t; zÞ ¼ ∑ z ð46Þ z0
k ¼ 0 ðtÞk k!
 
Here ðrÞk is the (rising) Pochhammer symbol [1]. The drawing z
f 9 ¼ 2 F 1 1; 1 þ a1 þ b1 ; 2 þa1 ;  i ; ð58Þ
stress szf can be computed evaluating sz at z ¼ z. Once sz is z0
known, in order to estimate an average friction stress, one can
pFq is the generalized hypergeometric function, defined as
compute the total normal force on each contact surface between


1 ðr 1 Þk …ðr p Þk k
the plate and the die. It results4
p F q ½ r 1 …r p ; s1 …sq ; z ¼ ∑ z ð59Þ
Z zi k ¼ 0 ðs1 Þk …ðsq Þk k!
qx b s0 secα tan β
Fx ¼ dz ¼ Assuming that the average friction stress τ f along the die surfaces
z cos α ð cos α  μ sin αÞða1 þ b1  2Þ
( " # can be expressed respectively as τ π 1 ¼ mπ 1 s0 =2 and
ða1  b1 Þf 1 a1 þ b1 z a1 ðz þ z0 Þb1 þ 1
  f 2 a 1  f z ðz
3 i 0 þ z i Þ 2 τ π2 ¼ mπ 2 s0 =2, the shear coefficients can be estimated as
ða1  1Þa1 z0 a1 z 0 z 1 ðz0 þ zi Þb1  1
i
4μ F y cos β
    mπ 1 ¼ ð60Þ
f ðb1  a1 Þ f 4 ðb1  a1 Þ ða1 þ3b1  2Þz0 ðhi þ hf Þl0 s0
þ 6  zz0 þ z z þ
a1  1 a1 þ b1  1 a1 þ b1  1
4μ F x cos α
mπ 2 ¼ ð61Þ
pffiffiffiffiffiffiffiffiffiffiffiffiffi
ðbi þ bf Þl0 s0
1
Note that p is positive if 0 r α r 2arctan½ 1 þ μ2  μ and b1 is positive if
a1ffiffiffiffiffiffiffiffiffiffiffiffiffi
0 r β r 2arctan½ 1þ μ  μ. However, considering that α 5 π =2, β 5 π =2, and
2 These last expressions give an estimation of the values of mπ 1
μ 51 they are always positive for the commonly adopted drawing parameters. and mπ 2 as a function of the reduction and of the Coulomb friction
coefficient μ. These estimations can be adopted in Eq. (21) to
2
If z0 ¼ 0, the die geometry degenerates in a truncated pyramid. In this case,
the stress distribution can be derived either as a limit case of the general solution,
compute the power dissipated by friction, and then the drawing
or, in a simplest way, integrating Eq. (41) under the assumption z0 ¼ 0. It results
"  a1 þ b1  2 #
stress, by means of the upper bound method (Eq. (25)), as
s ða þ b Þ z described in the precedent section.
sz ¼ 0 1 1 1  ð42Þ
ða1 þ b1  2Þ zi
3
For an easier implementation and utilization of the proposed analytical 2.4. A simplified empiric way to taking into account the material
model in practical engineering, a Taylor series expansion of all the hypergeometric strain hardening
functions necessary in the lower bound solution is reported in Appendix.
4
If z0 ¼ 0 the total normal force on each die surface is respectively equal to:
"  a1 þ b1  2 #
The upper bound solution presented in Section 2.2 furnishes an
s0 z 2 sec α tan β z estimation of the drawing stress during the process under the
Fx ¼ 1 ð47Þ
ða1 þ b1  2Þð cos α  μ sin αÞ zi
assumption of a perfect plastic material. It is well-known that
"  a1 þ b1  2 # metals exhibit strain hardening after the first yield, and this
s0 z 2 sec β tan α z
Fy ¼ 1 ð48Þ strengthening, that is not considered under the assumption of
ða1 þ b1  2Þð cos β  μ sin β Þ zi
perfect plasticity, strongly influences the drawing stress. In this
154 A. Panteghini / International Journal of Mechanical Sciences 84 (2014) 147–157
section, we present an extension allowing the consideration of this
effect, at least in an empiric and simplified way. A similar approach
has been adopted in [13], for a two dimensional problem.
It should be noted that the limit analysis technique in drawing
processes can be applied also in the presence of hardening, since a
steady-state process is considered in which, at each spatial point,
the mechanical quantities are constant in time and, therefore, the
material could be considered as perfectly plastic with an “infinite
degree of heterogeneity” (yield stress varying in space, but con-
stant with time) [10].
Let assume that the material flow stress sðεÞ can be expressed as
sðεÞ ¼ s0 þC εn
where s0 is the yield stress, C and n are material parameters.
The material yield stress is assumed to be respectively equal to
s0 in the cross section at the die entrance, and sðεf Þ in the cross
section at each die exit where
hi bi
εf ¼ ln
hf bf
is an estimation of the average total plastic at the end of the
process. With reference to Fig. 1, one can observe that in S1 the
material is virgin. For this reason, the yield stress on the dis-
continuity surfaces between S1 and S3a and between S1 and S4a is
assumed to be equal to s0. In S2, the material is outside the plastic
zone: at this point, the plate is completely deformed, and its
plastic deformation is equal to εf. The yield stress on the velocity
discontinuity surfaces between the block S2 and the blocks S3b and
S4b can be estimated as sf ¼ sðεf Þ. In all the other surfaces
(discontinuity surfaces, but also contact surfaces between plate
and die), the material is in an intermediate state of deformation,
between 0 and εf. One can simply assume that, in all these
surfaces, the yield stress is equal to
s ¼ 12 ðs0 þ sf Þ
With these assumptions, Eqs. (18) and (21) can be rewritten as
^ d ¼ s0 ðΣ 1  3a jΔv1  3a j þ Σ 1  4a jΔv1  4a jÞ þ sf ðΣ 2  4b jΔv2  4b j
W determined as a function of both the reduction and of the die inlet
2 2
s angles α and β, in order to have a central part of the plate in a
þ Σ 2  3b jΔv2  3b jÞ þ ðΣ 3a  5a jΔv3a  5a j þ Σ 3b  5a jΔv3b  5a j
2
þ Σ 4a  5b jΔv4a  5b j þ Σ 4b  5b jΔv4b  5b þ Σ 5a  5b jΔv5a  5b jÞ
and
^ f ¼ s ½mπ ðΣ 5a;π v5a  þ Σ 3a;π v3a  þ Σ 3b;π v3b Þ
     
W 2
2 2 2 2
þ mπ 1 ðΣ 5b;π 1 jv5b jþ Σ 4a;π 1 jv4a j þ Σ 4b;π 1 jv4b jÞ
Then, Eq. (24) simply becomes
^ þW
W ^
s^ zd ¼ d f
hf bf jv2 j
An optimized bound can be finally obtained minimizing s^ zd with
respect to η:
s^ zd ¼ minfs^ zd ; ηg
3. Validation of the new model
The obvious way to verify the analytical model described in the
precedent sections would be to compare its results with experi-
mental results covering a suitable range of the parameters of
interest. Unfortunately, to the best of our knowledge, experimental
values of the drawing force for a wide range of drawing para-
meters are unavailable in literature for plate drawing processes.
Therefore, one is forced to resort to numerical simulations. A
rather refined parametric finite element model has been devel-
oped, whose results will be used as a reference for the assessment
of the accuracy of the proposed analytical estimations.
Three different plate geometries have been studied. The geo-
metric dimensions of the initial and final cross sections of the
considered dies are reported in Table 1.
For each die, different values of the inlet angles α have been
considered. For each value of α, the reduction angle β has been
determined assuming that the plate must be initially in contact at
the same time with all the die surfaces.
The plates investigated are in C45 steel. The flow stress curve,
ð62Þ
adopted for both the numerical and the analytical models, has
been taken from [4] and it is reported in Fig. 12. With reference to
Eq. (62), the following material parameters have been adopted:
s0 ¼ 345:0 MPa, C ¼667.6 MPa, and n ¼ 0.30.
In the numerical model, the material mechanical behavior has
been assumed elastoplastic. The von Mises yield function, with
ð63Þ non-linear isotropic hardening law has been adopted as constitu-
tive model. It should be noted that this model is in general too
poor to simulate defects, or other features like the residual stress
profiles, which are influenced also by the unloading of the work-
piece at the end of the process, but allows the correct computation
of the drawing force [9,11].
The Young modulus and the Poisson coefficient have been
assumed respectively equal to E ¼200,000 MPa and ν ¼ 0:3.
The model is three-dimensional, and, to reduce locking phe-
nomena, it makes use of 8-noded brick with linear shape functions
and selective reduced integration for both stresses and stiffness
matrix. Because of the geometry of the problem, only a quarter of
the complete geometry of the process has been modelled (Fig. 13).
The die has been considered rigid. A Coulomb contact with
friction coefficient μ ¼ 0:035 has been assumed. Such a value has
been found to provide the best fit with the only experimental
ð64Þ
result available: for the geometry 1, the drawing force has been
measured experimentally in 93 kN, for α ¼ 31.
The length Ltot of the considered portion of the plate has been
steady-state condition.
ð65Þ Table 1
Geometric dimensions of the considered dies.
Geometry no. Die entrance Die exit
2bi (mm) 2hi (mm) 2bf (mm) 2hf (mm)
ð66Þ
1 71.5 6.7 69.88 5.89
2 46.5 9.0 45.06 7.89
3 33.5 20.0 31.91 17.88
ð67Þ
ð68Þ
Fig. 12. Flow stress curve of C45 steel adopted in the numerical model.
 A. Panteghini / International Journal of Mechanical Sciences 84 (2014) 147–157 155

180

160

140

120

F [kN]
100

80 FEM solution
Experimental measure
Coulomb friction, μ=0.07
60 Constant friction, m=0.12

0 2 4 6 8 10 12
2α [°]

Fig. 14. Comparison between experimental, numerical and analytical results for
geometry 1.

Y
X

Z
140

120

Fig. 13. Mesh adopted for the geometry 1, 2α ¼ 61.

100

F [kN]
80
Table 2
Mesh adopted in the FEM models for the geometry 1. Ltot is the considered plate
60
length, nx, ny and nz are the number of elements adopted to mesh the quarter part
of the plate in x, y, and z direction respectively, and dof is the total number of the 40 FEM solutions
Coulomb friction, μ=0.07
degrees of freedom of the mesh of the plate. Constant shear, m=0.12
20
2α (1) Ltot (mm) nx ny nz dof 0 5 10 15 20 25 30
2α [°]
3 140 8 20 280 159,327
4 105 8 20 210 119,637 Fig. 15. Comparison between numerical and analytical results for geometry 2.
5 105 8 20 210 119,637
6 70 8 20 140 79,947
7 70 8 20 140 79,947
8 70 8 20 140 79,947 In total, 20 different analyses have been carried out. The
10 70 8 20 140 79,947 simulations have been performed using the commercial finite
12 70 8 20 140 79,947 element code ABAQUS [6].
To validate the analytical model, a friction coefficient μ ¼ 0:07
has been adopted for the analytical model. Other results have been
Table 3 obtained also under the assumption of constant friction, assuming
Mesh adopted in the FEM models for the geometry 2.
m ¼ mπ 1 ¼ mπ 2 ¼ 0:12.
2α (1) Ltot (mm) nx ny nz dof In Fig. 14, it is reported a comparison between the drawing
force Fd computed by the FEM simulations, measured experimen-
3 200 12 12 133 67,938 tally, and obtained using the new analytical model, for several
6 150 12 12 200 101,907 values of the die inlet angle α, for the die geometry 1. Figs. 15 and
8 150 12 12 200 101,907
10 150 12 12 200 101,907
16 show a comparison between the total drawing force Fd
14 100 12 12 167 85,176 computed by the FEM simulations and estimated using the new
22 100 12 12 186 94,809 analytical model, for several values of the die inlet angle α, for the
die geometries 2 and 3 respectively.
These comparisons show that the drawing forces computed
Table 4 with the new analytical method is in good agreement with those
Mesh adopted in the FEM models for the geometry 3.
obtained using the FEM simulations for all the considered combi-
2α (1) Ltot (mm) nx ny nz dof
nations of drawing parameters. Moreover, the new model gives
results closer to those obtained numerically especially for the
3 240 12 12 160 81,627 values of the die inlet angles near to the optimum values (i.e. the
6 180 12 12 180 91,767 values that minimize the drawing force). The better results are
8 180 12 12 180 91,767
obtained for the geometries 2 and 3, i.e., for the plates in which the
10 180 12 12 180 91,767
12 120 12 12 240 122,187 ratio between the plate cross sectional dimensions are more
20 120 12 12 240 122,187 closer to 1.
The comparisons between the analytical and numerical results
reported above, obtained for different plate geometries and draw-
The adopted meshes and plate lengths are reported in Tables 2–4 ing conditions, demonstrate that the analytical method here
for the three studied geometries. proposed allows one to obtain in a relatively simple way an
The numerical analyses have been performed under displace- estimation of the drawing force as a function of the process
ment control, pulling the plate through the die, assuming arbitrary parameters and can be easily adopted as an engineering tool to
large strains and displacements. optimize real industrial plate drawing processes.
156 A. Panteghini / International Journal of Mechanical Sciences 84 (2014) 147–157

180
Angles of incidence of the discontinuity surfaces on the plane π1:
bi sin ðβ þ ξÞsec ξ bf sin ðω  βÞ sec ω
160 FEM solution ψ ¼ arctan ; ζ ¼ arctan
Constant shear, m=0.12 hi hf
Coulomb friction, μ=0.07
ðA:5Þ
140
F [kN]

Angles of incidence of the discontinuity surfaces on the plane π2:

120
hi sin ðα þ θÞsec θ hf sin ðφ  αÞsec φ
χ ¼ arctan ; κ ¼ arctan
bi bf
100
ðA:6Þ
80
0 5 10 15 20 25 Area of the discontinuity surfaces:
2α [°] Σ 1;3a ¼ 12 bi hi sec θ; Σ 2;3b ¼ 12 bf hf sec φ ðA:7Þ
Fig. 16. Comparison between numerical and analytical results for geometry 3.
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Σ 1;4a ¼ 12 hi ðl0  ηÞ2 þ b2i ; Σ 2;4b ¼ 12 hf bf þ η2
2
ðA:8Þ
4. Conclusions
Σ 3a;5a ¼ 12 bi ρ sec χ ; Σ 3b;5a ¼ 12 bf ρ sec κ ðA:9Þ
A new analytical solution to estimate the drawing stress in
three-dimensional plate drawing processes has been developed. Σ 4a;5b ¼ 12 hi λ sec ψ ; Σ 4b;5b ¼ 12 hf λ sec ζ ðA:10Þ
The new analytical model, which has been also extended to
account for strain hardening and Coulomb friction, has been Σ 5a;5b ¼ 14 ½2ðhi  hf Þ2 csc2 δðh2i þ h2f þ b2i sec2 ξ þ b2f sec2 ωÞ
validated comparing the drawing force computed analytically with
þðhi hf Þ4 csc4 δ  ðhi  hf þ bi sec2 ξ  bf sec2 ωÞ2 1=2
2 2 2 2
that computed by FEM simulations, considering different die ðA:11Þ
geometries and plate reductions. These comparisons show that
where
the new analytical model gives good results and could be adopted
in the engineering practice as a design procedure to optimize λ ¼ bf ð cos β þ sin β tan ωÞ ðA:12Þ
three-dimensional plate drawing processes.
Future work, partially in progress, may concern the improve- and
ment of the description of the strain hardening during the process. ρ ¼ hf ð cos α þ sin α tan φÞ ðA:13Þ

Area of the contact surfaces between the die and the blocks:
Acknowledgments Σ 3a;π 2 ¼ 12 b2i tan χ ; Σ 3b;π 2 ¼ 12 b2f tan κ ðA:14Þ

Prof. Lorenzo Bardella and Prof. Rocco Lagioia are acknowl- Σ 4a;π 1 ¼ 12 h2i tan ψ ; Σ 4b;π 1 ¼ 12 h2f tan ζ ðA:15Þ
edged for the helpful discussions. Dr. Massimo Depaoli is acknowl-
edged for his guidance on the preparation of the figures. The
Σ 5b;π 1 ¼ 12 ðhi þ hf Þðhi tan ψ þ hf tan ζ Þ  Σ 4a;π1  Σ 4b;π 1 ðA:16Þ
commercial finite element code ABAQUS has been run under
academic license at the Department of Civil Engineering, Archi-
tecture, Land, Environment, and of Mathematics (DICATAM) of the
Σ 5a;π 2 ¼ 12 ðbi þ bf Þðbi tan χ þ bf tan κ Þ  Σ 3a;π2  Σ 3b;π2 ðA:17Þ
University of Brescia.

Appendix B. Taylor series for the hypergeometric functions

Appendix A. Geometric quantities necessaries in the velocity-
based solution The hypergeometric functions are not usually implemented in
standard programs for engineering. For this reason, in order to
Reduction angles: simplify the implementation of the proposed analytical model, in
bi  bf b b this section we furnish the Taylor expansions of the hypergeo-
β ¼ arctan ; γ ¼ arctanqffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
i f
; metric functions used for the stress-based solutions:
l0 2
ðhi  hf Þ2 þ l0  
x 1 ð  z Þ  i Γ ½p þ iΓ ½p 
0 1 2 i
h h f 0;1;2;3;8;9 ¼ 2 F 1 1; p1 ; p2 ;  ¼ ∑ x ðB:1Þ
δ ¼ arctanqffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
i f
ðA:1Þ z0 i¼0 Γ ½p1 Γ ½p2 þ i
2
ðbi  bf Þ2 þl0
  1 ð  z Þ  i Γ ½p þ iΓ ½p 


x 0 1 2 i
f 4;5;6;7 ¼ p F q 1; 1; p1 ; 2; p2 ;  ¼ ∑ x
i ¼ 0 ð1 þ iÞΓ ½p1 Γ ½p2 þi
where z0
l0 ¼ ðhi  hf Þcot α ðA:2Þ ðB:2Þ

Angles of incidence of the discontinuity surfaces on the planes xz: where Γ is the Euler Gamma function, defined as 5

Z 1
l η η Γ ½t ¼ xt  1 e  x dx ðB:3Þ
θ ¼ arctan 0 ; φ ¼ arctan ðA:3Þ
hi hf 0

Angles of incidence of the discontinuity surfaces on the planes yz:

5
The Euler Gamma function Γ can be directly computed using various
l η η
ξ ¼ arctan 0 ; ω ¼ arctan ðA:4Þ software applications, such as Fortran, Cþ þ , also Microsoft Excel and
bi bf OpenOffice Calc.
 A. Panteghini / International Journal of Mechanical Sciences 84 (2014) 147–157
References
[1] Abramowitz M, Stegun, editors. Handbook of mathematical functions with
formulas, graphs, and mathematical tables. 9th edition. New York: Dover;
1972.
[2] Avitzur B. Flow characteristics through conical converging dies. Trans ASME - J
Eng Ind 1964;88:410–20.
[3] Avitzur B, Hahn W, Isovici S. Limit analysis of flow through conical converging
dies. J Frankl Inst 1975;299:339–58.
[4] Eleö
d A, Baillet L, Berthier Y, Tö rkö
ly T. Deformability of the near surface
layer of the first body. In: Dowson D, Priest M, D. G, Lubrecht A, editors.
Tribological research and design for engineering systems proceedings of the
29th Leeds-Lyon symposium on tribology. Tribology series, vol. 41. Amsterdam,
The Netherlands: Elsevier; 2003. p. 123–32.
[5] Fueyo J, Avitzur B. Limit analysis of flow through conical converging dies. In:
Second canadian congress of applied mechanics, University of Waterloo Press;
Waterloo, Ontario, Canada, 1969. p. 13–4.
[6] Hibbitt D, Karlsson B, Sorensen P. Analysis user's manual version 6.7. USA:
Dassault Systèmes; 2007.
[7] Luis C, León J, Luri R. Comparison between finite element method and
analytical methods for studying wire drawing processes. J Mater Process
Technol 2005;164–165:1218–25.
[8] Panteghini A. Numerical simulations of multipass wire drawing processes [Ph.
D. thesis], Dept. Civil Enrgr., University of Brescia, Italy; 2009.
157
[9] Panteghini A, Genna F. Effects of the strain-hardening law in the numerical
simulation of wire drawing processes. Comput Mater Sci 2010;49:236–42.
[10] Panteghini A, Genna F. An engineering analytical approach to the design of
cold wire drawing processes for strain-hardening materials. Int J Mater Form
2010;3:279–89.
[11] Panteghini A, Genna F. Numerical integration of a pressure-dependent, non-
linear kinematic hardening constitutive model for large strain cyclic plasticity
of metals. Int J Numer Methods Eng 2012;89:1047–67.
[12] Prakash R, Khan QM. An analysis of plastic flow through polygonal converging
dies with generalized boundaries of the zone of plastic deformation. Int J
Mach Tool Des Res 1979;19:1–19.
[13] Rubio Alvir E, Mariín M, Domingo R, Sebastián Pérez M. Analysis of plate
drawing processes by the upper bound method using theoretical work-
hardening materials. Int J Adv Manuf Technol 2009;40:261–9.
[14] Rubio Alvir E, Sebastián Pérez M, Sanz Lobera A. Mechanical solutions for
drawing processes under plane strain conditions by the upper bound method.
J Mater Process Technol 2003;143–144:539–45.
[15] Sahoo S, Kar P, Singh K. A numerical application of the upper-bound technique
for round-to-hexagon extrusion through linearly converging dies. J Mater
Process Technol 1999;91:105–10.
[16] Sahoo S, Sahoo B, Patra L, Paltasingh U, Samantaray P. Three-dimensional
analysis of round-to-angle section extrusion through straight converging die.
Int J Adv Manuf Technol 2010;49:505–12.
[17] Yang DY, Lee CH. Analysis of three-dimensional extrusion of through curved
dies by conformal transformation. Int J Mech Sci 1978;20:541–52.