IIE Transactions (1998) 30, 913±921

Blank design optimization on deep drawing of square shells

HAE CHANG GEA and RAJESH RAMAMURTHY

Department of Mechanical and Aerospace Engineering, Rutgers, The State University of New Jersey, Piscataway, NJ 08855. USA
E-mail: [email protected]

Received November 1997 and accepted February 1998

A numerical scheme for improving the drawability in the deep drawing of square shells by blank design optimization is presented.
The numerical scheme is formulated as an optimization problem whose objective is to maximize the drawability, subject to the
constraint that fracture failure and draw-in failure do not occur. Appropriate blank design parameters are used as the design
variables of the formulation. To enable the numerical scheme to work models predicting the onset of fracture failure and draw-in
failure are required. Numerical models simulating the onset of these failures under the given process conditions are thus discussed.
Optimal designs for three cases are then presented. Finally, by considering both the drawability and the non-uniformity of the ®nal
¯ange pro®le it is shown that the circular pro®le can be considered to be the optimal blank shape for square cup drawing.

1. Introduction non-circular cup drawing processes to determine the

shape of the blank that can be formed into a cup of
Maximizing drawability in deep drawing operations di- uniform height. For this purpose, they ®rst analyze the
rectly reduces the manufacturing cost and time associated deformation of the ¯ange region of the cup using ®nite
with these operations. Furthermore, an a priori know- element analysis. The outcome of this analysis is then
ledge of the limits of drawability can also be useful in used to determine the optimal blank geometry. Their
avoiding the development of infeasible designs early on calculations use the assumption that the sheet metal is
thereby decreasing the total product design cycle time. isotropic. Kim and Kobayashi [4] have proposed a geo-
Deep drawing is essentially a metal ¯ow process and metrical scheme to determine the contour shape of the
the limits of sheet metal drawability are limited by the blank for rectangular cup drawing. They achieve this by
onset of process failures, such as fracture failure and calculating the ¯ow lines of the material points in the
draw-in failure. Consequently, drawability can be in- blank during drawing. Toshihiko et al. [5] have used yet
creased by delaying the onset of process failures. Towards another innovative scheme for this purpose. They start
this end, process parameters which in¯uence the drawing with a cup having the desired geometry properties and
operation need to be properly selected. Blank shape is one then based upon a slip line ®eld theory they step back-
of the most important process parameters that strongly wards to obtain the starting blank shape. Gloeckl and
in¯uences drawability i.e., by altering the blank design Lange [6] developed a numerical scheme using plane-
considerable improvement in the drawability may be strain slip line ®eld theory to predict the blank shape for
obtained. general stampings. The material behavior is assumed to
Blank design has in general been a particularly dicult be perfectly plastic. By conducting experiments and also
problem to handle. Trial and error methods [1] or em- by giving qualitative arguments, Hsu and Lee [7] have
pirical rules [2] have been used for blank design in in- suggested a simple blank pro®le that could help reduce
dustry. The experimental cut and trial technique of the amount of ®nal blank pro®le at the end of a square
®nding the best blank design is not only time consuming cup drawing operation. Recently, Majlessi and Lee [8,9]
but is also at times confusing because variations in the conducted experimental and numerical studies on various
mechanical properties of individual blanks can give dif- aspects of square cupping including the issue of corner
ferent results for similar blanks of the same material. cut-o€s in blanks. They observed that the presence of the
Many numerical analyses have been proposed to pre- corners contributed to increased diagonal straining and
dict the optimum blank shape. For general cupping op- that the removal of the material in the corner helped in
erations, most de®ne the optimal blank shape as that not only reducing the strains but also in improving the
blank pro®le which can be formed into a cup with either a drawability.
uniform ®nal ¯ange pro®le or a uniform rim height. Iseki In this paper, a numerical scheme for blank design
and Murota [3] have proposed the numerical modeling of optimization in the context of square shell drawing is

0740-817X Ó 1998 ``IIE''

914 Chang Gea and Ramamurthy

proposed. First, we present numerical techniques for 2.2. Fracture failure

predicting the onset of fracture failure and draw-in fail-
Fracture failure in drawing operations has received con-
ure. Next, we discuss our optimization formulation for
siderable attention in the literature because it has long
maximizing drawability in square shell drawing opera-
been recognized as the most undesirable form of failure
tions. Numerical examples from the formulation are then
[11±13]. It has now been established that the limits of
presented and discussed. The overall optimum blank
sheet metal formability depend on many factors such as
shape for square shell drawing is ®nally deduced.
the material properties and the deformation geometry.
Subsequent extensive experimental investigations have
led to the introduction of the Forming Limit Diagram
2. Numerical scheme (FLD) [12]. The FLD can be determined either analyti-
cally [11,13] or experimentally. They are now available
The goals of our numerical scheme are to maximize the for a number of forming materials and have become
drawability and to determine the optimal starting blank widely accepted as indicating the limits of sheet metal
design for square cupping operations. Drawability is de- formability.
®ned as the maximum possible punch penetration. At the A typical FLD is shown in Fig. 1. The abscissa of the
maximum possible punch penetration, process failures set plot is the minor strain (second principal strain), e2 , while
in making further drawing infeasible. Delaying these the ordinate represents the major strain (®rst principal
failure mechanisms by a proper choice of the process strain), e1 . In this plot, point A represents a deformation
parameters to improve drawability is thus the central state that is considered safe while point B represents an
theme of our numerical scheme. unacceptable deformation state. The existence of a
For the working of our numerical scheme, several is- necking zone between the failure and safe zones is nota-
sues need to be addressed. First, we need a means of ble. Deformation states lying in the necking zone are
numerically simulating the square shell drawing process. considered unacceptable.
PAM-STAMPä [10], a commercially available ®nite el- Now consider the point A with e1;A ˆ 0:5 and
ement analysis package has been used for this purpose. e2;A ˆ ÿ0:1 in Fig. 1. Suppose, it represents the strain
We brie¯y discuss issues pertinent to the drawing process state at a speci®c point on the blank. With a knowledge of
simulation in the following section. The modeling of the the FLD of the material i.e., the function ecr …e2 †, the
onset of fracture failure and draw-in failure is a basic ``critical'' major strain (major strain at failure) at the
necessity for our numerical scheme. Thus, we consider the point A, ecr;A , can be computed. In this case it represents
issues related to the modeling of these failures in Sections the point C on the plot with e1;C ˆ 0:6. Note that the
2.2 and 2.3. Then, we discuss our numerical scheme, critical major strain ecr;A ˆ e1;C is higher than e1;A , the
which is an optimization formulation, in Section 2.4. To major strain corresponding to the deformation state at
deduce the overall optimum blank shape for square shell the point A. Thus, the deformation state at the point A is
drawing operations we shall also require the notions of acceptable. By the same arguments, it can be seen that the
``out-of-squareness'' for the ¯ange pro®le and ``out-of- deformation state corresponding to the point B is unac-
circularity'' for the blank pro®le. We discuss these no-
tions in Section 2.5.

2.1. Deep drawing simulation

The square shell drawing process is simulated using
PAM-STAMPä, a commercial available ®nite element
analysis package. Towards this end, the punch, die,
blankholder, and the sheet are discretized using ®nite ele-
ments. Thus, blanks of any shape and/or size can be
considered by appropriately discretizing them into ele-
ments. Other process variables such as tool pro®le radii,
material properties and blank holding pressure can be
provided as inputs for the simulation. Several parameters
corresponding to a speci®ed punch penetration can be
obtained as outputs from the simulation. Of interest to us
is the strain undergone by the sheet and the displacement
of the sheet. Using these data it is possible to predict the Fig. 1. Forming Limit Diagram of the blank material. e1 is the
onset of fracture failure and draw-in failure as discussed major strain and e2 is the minor strain. Deformations at A are
in the following sections. acceptable whilst those at B are unacceptable.
Blank design optimization on deep drawing of square shells 915

ceptable. More generally, the deformation state at any

point `i' in the sheet is unacceptable if the major strain e1;i
at that point is higher than the critical major strain ecr;i
permissible at that point.
Based upon the above discussion, we can now explain
the technique used to detect fracture failure for the given
process conditions and a speci®ed punch penetration.
First, we simulate the drawing process for the given
process conditions and the speci®ed punch penetration.
Then, we determine the strains at each point on the sheet
and check whether they are acceptable or are unaccept-
able using the FLD of the sheet material. If the strains are
found to be unacceptable at some point on the sheet we
detect the onset of fracture failure.

2.3. Draw-in failure

Fig. 3. The geometry of the square pan obtained by the deep
In many circumstances it may be necessary to provide for drawing simulation.
an adequate width for the ®nal ¯ange in the ®nished box.
The ®nal ¯ange pro®le is de®ned as the resulting pro®le of
the ¯ange upon completion of drawing. For example, it
may be necessary to place rivets on the ¯ange of the box another view of the entire square box. Recall that for the
during an assembly operation. Such a situation requires purposes of the ®nite element simulation of the drawing
us to ensure that a minimum ¯ange width is available. In process, the sheet is discretized into elements. Thus the
other words, if the minimum width of the ®nal ¯ange is displacement of the sheet due to the penetration of the
found to be smaller than the required minimum ¯ange punch is given by the displacements of the nodes which
width, draw-in failure is deemed to have occurred. Thus, comprise the ®nite element model of the sheet. To com-
by computing the ®nal ¯ange width, we can detect the pute the minimum ®nal ¯ange width it suces to obtain
occurrence of draw-in failure for the given process con- the displacements of the nodes which comprise the ¯ange
ditions and the speci®ed punch penetration. The magni- rim (shown in the ®gure) from the ®nite element analysis.
tude of the minimum ®nal ¯ange width can be obtained Furthermore, owing to the symmetry it suces to con-
through a simulation of the drawing process for the given sider only one-eighth of the square shell for these calcu-
process conditions and the speci®ed punch penetration as lations.
follows. Now consider a node on the ¯ange rim, say at C, to the
Figure 2(a) shows a one-eighth section of the geometry left of the reference line AB in Fig. 2(a). Line AB passes
of the square box from the top view, while Fig. 3, shows through the point of tangency B of the corner circle and
the straight line which constitutes an edge of the square
shell. The ¯ange inner boundary remains ®xed through-
out the drawing operation and is thus used as a reference
edge to measure the ®nal ¯ange width at the node C. The
width of the ¯ange, wi , at point C is obtained by com-
puting the di€erence between the y-coordinate of point C
from the y-coordinate of the ¯ange inner boundary as
shown. In this manner the ®nal ¯ange width at each node
of the ¯ange rim lying to the left of the reference line AB
is computed. Next, consider a node to the right of line
AB, say at D in the ®gure. The width of the ®nal ¯ange,
wi , at node D is obtained by taking the di€erence of AD
from r as shown in the ®gure with r being the distance of
the ¯ange inner boundary from the reference point A,
while AD gives the distance of the node D from the ref-
erence point A. This calculation is repeated for each of
Fig. 2. (a) Notation for the calculation of the minimum ®nal the nodes on the ¯ange rim lying to the right of line AB.
¯ange width and the out-of-squareness of the ®nal ¯ange pro- Thus in this manner the ®nal ¯ange width corresponding
®le; (b) notation for the calculation of the out-of-circularity of to each node on the ¯ange rim is obtained. The minimum
the initial design. among these ¯ange widths computed gives the minimum
916 Chang Gea and Ramamurthy

®nal ¯ange width for the given process conditions and the
speci®ed punch penetration. If the value of the minimum
®nal ¯ange width, shown as wnom is smaller than the re-
quired minimum ¯ange width w then draw-in failure is
deemed to have occurred.

2.4. Optimization formulation

We formulate our numerical scheme as an optimization
problem whose objective is to maximize the drawability
while satisfying fracture and draw-in failure constraints.
Thus, it can be given as:
Maximize: Drawability,
Subject to:
e1;i  ecr …e2;i †;
w  wnom ;
The constraint functions of the above formulation
represent the criteria for fracture and draw-in failure
respectively. The formulation can be further simpli®ed
and the calculation of the sensitivities of the constraint
functions can be avoided if we incorporate the fracture
and draw-in failure constraints into the objective func-
tion. The ¯ow diagram of this simpli®ed formulation is Fig. 5. Symmetry planes in a square shell. Only one-eighth of
shown in Fig. 4. A gradient-based method, sequential the blank is used in the drawing simulation by applying ap-
propriate displacement boundary conditions on the nodes lying
quadratic programming (SQP) available in IDESIGN
on each of the two planes of symmetry.
[14], has been used to solve this optimization problem.
The evaluation of the gradients of the objective function
is done using the ®nite di€erence technique because the propriate displacement boundary conditions applied on
dependence of the drawing failures upon the process the nodes lying on the planes of symmetry, only one-
parameters (design variables) is implicit and therefore the eighth of the tools and blank need to be modeled as
analytical expressions for its evaluation are not available. shown in Fig. 5. Refer to Fig. 2. Boundary conditions are
The ®nite element analysis of the drawing process is imposed on the sheet nodes lying on the planes of sym-
quite computationally intensive, the computational load metry in such a manner so that they are permitted to
depending directly upon the number of elements used in translate only on their respective symmetry planes and
the discretization of the blank. To reduce the computa- are permitted to rotate only about an axis parallel to a
tional load, we take advantage of the symmetry existing normal to these symmetry planes. Speci®cally, nodes ly-
in square shells as shown in Figs. 3 and 5. With the ap- ing on the edge OE of the sheet are permitted to translate
only on the yz plane and have rotations only about a
normal to this plane. Similarly nodes lying on the edge
OD of the sheet are permitted to translate only on the
plane containing OD and are permitted to rotate about
an axis perpendicular to this plane.
The stamping tools are assumed to be rigid while the
behavior of the sheet is characterized as elasto-plastic.
The stress±strain relationship of the sheet is shown in
Fig. 6 and the FLD of the sheet material is shown in
Fig. 1. The other material properties of the sheet are as
follows: the sheet thickness is 0.7 mm; its Poisson ratio, v,
is 0.33; its Young's modulus is 210 MPa; its yield stress is
139 MPa; and the coecient of friction in the tool/blank
interface is 0.15. The dimension of the punch is
158.4 mm ´ 158.4 mm and the punch pro®le radius is
Fig. 4. The ¯ow diagram of the optimization formulation for 9.2 mm; the dimension of the die is 160.0 mm ´ 160.0 mm
maximizing drawability. Failure draw depth is evaluated using and the die pro®le radius is 10.0 mm. The blankholding
the failure criteria for draw-in failure and fracture failure. pressure is set at 1.92 MPa. Finally as the failure depth
Blank design optimization on deep drawing of square shells 917

Fig. 6. The stress±strain curve of the material used for the

drawing simulation.

for the given set of process conditions is not known a

priori, the sheet is drawn in each case to a large depth
which has been set at 68 mm in the numerical simulation.

2.5. Flange pro®le

Fig. 7. Final and nominal ¯ange pro®le of the optimum for the
Inappropriate choice of the initial blank design could ®rst one parameter design. The out-of-squareness of the ®nal
result in a very non-uniform ®nal ¯ange pro®le. In many ¯ange pro®le for this case is higher than that for the two pa-
instances it may also be necessary to obtain as uniform a rameter design.
®nal ¯ange pro®le as possible at the end of the drawing
operation. The consequence of choosing poor blank de-
signs are thus not limited to only obtaining low permis-
sible draw depths but also to obtaining non-uniform ®nal
¯ange pro®les. An overall optimum initial blank shape
for square shell drawing would be one which has high
drawability and results in a ®nal ¯ange pro®le having a
high degree of uniformity.
The uniformity of the ®nal ¯ange pro®le is measured
by the magnitude of the ``out-of-squareness'' of the rim of
the ¯ange. Intuitively, a ®nal ¯ange pro®le having a low
out-of-squareness corresponds to a very uniform ¯ange
pro®le while one having a high out-of-squareness corre-
sponds to a non-uniform ¯ange pro®le.
For the purpose of de®ning out-of-squareness, we ®rst
de®ne the ``nominal blank pro®le'' as the largest square
pro®le with rounded corners that can be inscribed within
the ®nal ¯ange pro®le. A one-eighth section of the
nominal blank pro®le is shown in Fig. 2(a). More ex-
amples of the nominal blank pro®les are also shown in
Figs. 7±9. The nominal blank pro®le is the pro®le that is
desired in the ideal case and the extent of deviation from
the nominal blank pro®le therefore gives a good measure
of the out-of-squareness of the ®nal ¯ange pro®le. The
out-of-squareness of the ®nal ¯ange pro®le is measured Fig. 8. Final and nominal ¯ange pro®le of the optimal blank
using the expression: design for the second one parameter design. Even though the
s drawability for this design is comparable with the drawability
Pn 2 for the two parameter design, the ®nal ¯ange pro®le for this
iˆ0 …wi ÿ wnom † case has a higher out-of-squareness as compared to the two
rˆ ; …1†
n parameter design.
918 Chang Gea and Ramamurthy

compare the out-of-circularity of the initial/starting blank

pro®les has a radius wnom ˆ 150 mm for all the three
design cases. The ®nal ¯ange pro®les of these designs are
also compared.

3.1. Example 1
For this example, the shape of the blank is octagonal as is
shown in Fig. 10. The central vertical line has a ®xed
length of 150 mm and the design variable x1 is chosen as
the horizontal distance of the vertex of the octagon from
the central vertical line. To avoid unrealistic designs, x1 is
bounded within the interval [94.50 mm, 129.50 mm]. The
initial value of x1 is arbitrarily selected as 113.21 mm. The
drawability for this design is found to be 23 mm. Using
the optimization formulation an improved value for the
design variable x1 ˆ 106:47 mm is obtained after six it-
erations. The drawability for this blank design is found to
be 37 mm. Thus, considerable improvement in drawabi-
lity is noted from the initial to the ®nal design. The initial
Fig. 9. The ®nal and nominal ¯ange pro®le of the optimal de- and the optimal designs are shown in Fig. 10. The itera-
sign for the two variable case. The ®nal ¯ange pro®le is seen to tion history of the maximum draw depth (drawability) is
have a high degree of squareness. shown in Fig. 11.
The reference circle with a radius of 150 mm is used to
compare the initial and ®nal designs. It is found that the
where wi is the width of the ®nal ¯ange corresponding to design variable x1 continually decreases in magnitude.
the ith node on the ¯ange rim, wnom is the width of the Note that the reference circle circumscribes the optimal
nominal ¯ange pro®le from the ¯ange inner boundary, blank design. Consider a reduction in the value of the
and n is the number of nodes along the rim of the ¯ange. design variable from its optimum value. This would result
Note from the discussion in the previous section that in a reduction of the blank material causing early draw-in
wnom is also the minimum ®nal ¯ange width.
For reasons that will become apparent in the following
section we shall also require to de®ne the notation of
``out-of-circularity'' of the initial/starting blank pro®le.
The out-of-circularity of the initial blank pro®le is qual-
itatively a measure of the deviation of the initial blank
pro®le from a circular shape. Figure 2(b) shows one-eighth
of a typical initial blank pro®le. The out-of-circularity of
the initial blank pro®le is computed using Eq. (1) where
wi is the distance of the node i on the blank rim from the
center O, wnom is the radius of the reference circle, and n is
the number of nodes on the blank rim ABC.

3. Results and discussions

In this section, three blank design optimization problems

for square shells are studied. The objective is to maximize
the drawing depth while the constraints are no fracture
and draw-in failures. The minimum ®nal ¯ange width in
the draw-in failure criterion is chosen as 15 mm. The ®rst
design uses a single variable to de®ne an octagonal blank.
The second design uses a single parameter to de®ne a
square blank with each of its corners cut-o€ symmetri- Fig. 10. Initial and optimum designs for the ®rst one parameter
cally. The ®nal design uses two parameters to de®ne the design. The optimal value of x1 is such that the vertex of the
initial shape of the blank. The reference circle used to octagon lies on the reference circle of radius 150 mm.
Blank design optimization on deep drawing of square shells 919

Fig. 11. Iteration history for the ®rst one parameter design.
Considerable improvement in the drawability, obtained
through the optimization process, may be noted.

failure. Thereby, the drawability decreases upon de-

creasing the value of the design variable. On the other Fig. 12. The initial and optimal ¯ange pro®le for the second
hand if the design variable value is increased beyond the one parameter design. The corner cut-o€ increased to about
optimum value early fracture sets in. This occurs pri- 30% at the optimum.
marily because the increase in blank material provides for
a better clamping condition under the blank holder. This
means that for the same holding pressure increased tensile
stresses are generated in the sheet resulting in early frac-
ture failure. These conclusions in general are true irre-
spective of the magnitude of reduction or increase in the
value of the design variable. Since neither an increase nor
a decrease in the value of the design variable results in an
increase in the value of the objective function (drawabi-
lity) it is felt that the minimum value obtained for the
objective function is in the vicinity of its global minimum
for this formulation. Figure 7 shows the ®nal ¯ange
pro®le corresponding to the optimal blank design.

3.2. Example 2
For this example, we consider a square blank with its
corner cut-o€. The design variable is the horizontal dis-
tance of the center of the cut-o€ from the center line as is Fig. 13. Iteration history for the second one parameter design.
shown in Fig. 12. The initial value of the design variable, In this case considerable improvement in the drawability from
x1 , is chosen as 113.45 mm and its lower and upper the initial design to the optimal design may also be noted.
bounds are set as 90 and 135 mm respectively. The
drawability corresponding to the starting value of the p
design variable was found to be 25.2 mm. After eight starting blank design, since 105:41 mm  2 ˆ 149 mm
iterations, the value of the design variable was obtained as is shown in Fig. 12. Arguing along similar lines as for
as x1 ˆ 105:41 and the drawability at this value of the the previous example it would again be reasonable to
design variable was found to be 54.7 mm. The initial and conclude that the optimum obtained in this case is also
optimal designs are shown in Fig. 12. The design iteration close to the global optimum. The ®nal ¯ange pro®le
history is given in Fig. 13. The reference circle with a corresponding to the optimal blank design is shown in
radius of 150 mm can be inscribed within the optimal Fig. 8.
920 Chang Gea and Ramamurthy

3.3. Example 3
In this example, a polygon with sixteen sides is used to
model the blank. Due to the symmetry conditions, we
only need two design variables to completely de®ne the
blank shape as is shown in Fig. 14. These design variables
are chosen as the horizontal distance of the two vertices
from the centerline as is shown in Fig. 14. The starting
value of these variables x1 and x2 are 56.3 and 100.0 mm
respectively. x1 and x2 are restricted to lie within the
closed intervals [49.0 mm, 60.0 mm] and [90.0 mm,
135.0 mm], respectively. The punch penetration corre-
sponding to the starting values of the two design variables
is 32.9 mm. The drawability corresponding to the optimal
design variable is 52.4 mm. The optimal design was ob-
tained in six iterations with the optimal values for x1 and
x2 being 58.827 and 106.387 mm respectively. The initial
and ®nal designs are shown in Fig. 14 and the iteration Fig. 15. Iteration history for the two variable design.
history is shown in Fig. 15. For reasons similar to those
given in example 1, it would be reasonable to conclude
that the optimum obtained after six iterations is indeed the circles on the plot. The out-of-circularity of each of
near the global optimum. Figure 9 shows the ®nal ¯ange the initial designs is indicated beside the curve. For ex-
pro®le corresponding to the optimal blank design. ample, for design #2 the out-of-circularity is 5.08 mm and
for the square blank the out-of-circularity is 27.38 mm.
At the beginning, the out-of-squareness of the square
3.4. Comparison of ®nal ¯ange pro®les blank is the least (its value is di€erent from zero since the
Figure 16 shows the out-of-squareness of the ®nal ¯ange nominal ¯ange pro®le has rounded corners). However, as
pro®le as a function of the draw depth for the three op- the punch moves into the die cavity the out-of-squareness
timal starting blank designs evaluated in examples 1, 2, of the ¯ange pro®le continually increases and fracture
and 3. In addition, two reference starting blank shapes, failure is initiated before drawing depth 20.0 mm. This is
one circular blank (radius 150 mm) and the other a to be expected since the material movement near the
square blank (300 ´ 300 mm) have been included for center of the edges is greater than those at the corners.
comparison. All other process conditions were held con- See Fig. 5. We can conclude that the performance of an
stant while generating these plots. Each curve terminates initially square design is poor not only from the point of
upon the initiation of failure, which has been indicated by view of drawability but also because of the continually
increasing out-of-squareness of the ®nal ¯ange pro®le. On
the other hand, although the circular pro®le has a higher

Fig. 16. The variation in the out-of-squareness with the punch

Fig. 14. The initial and optimal designs for the two parameter depth for the three optimal designs, the circle and the square
design. The out-of-circularity of the optimum design may be blank pro®le. Circles at the end of the curves indicate failure
noted to be small. initiation.
Blank design optimization on deep drawing of square shells 921

initial out-of-squareness, it continually decreases with References

increasing material utilization till failure is initiated. The
circular pro®le outperforms the square pro®le with re- [1] Cope, J.S. (1953) You can draw rectangular shells by following
gards to the degree of out-of-squareness of the ®nal ¯ange certain rules. American Machinist, 97, 156±159.
[2] Gopinathan, V. (1977) Optimum blank pro®le determination for
pro®le just beyond a moderate draw depth (20 mm). rectangular deep drawing. Indian Journal of Technology, 15, 330±333.
Figures 7±9 show the out-of-squareness of the ®nal [3] Iseki, H. and Murota, T. (1986) On the determination of the
¯ange pro®le of the optimal designs of examples 1, 2, and optimum blank shape of non-axisymmetric drawn cup by the ®-
3. The starting out-of-squareness of the optimum ¯ange nite element method. Bulletin of JSME, 29 (249), 1033±1040.
pro®le for the ®rst design case is about 16.5 mm. That is [4] Kim, N. and Kobayashi, S. (1986) Blank design in rectangular cup
drawing by an approximate method. International Journal of
lower than that of the circular blank (19.0 mm). However Machine Tool Design and Research, 26 (2), 125±135.
the rate of decline of the out-of-squareness of the circular [5] Toshihiko, M., Kawai, N., Marumo, Y. and Senga, M. (1987)
blank is much higher. The advantage of the ®rst design Square shell deep drawability of commercially pure aluminium
over the circular blank vanishes at a punch depth of about sheets. Nippon Kikai Gakkai Ronbunshu, 53 (487), 771±779, (in
37 mm when the blank corresponding to the ®rst design Japanese).
[6] Gloeckl, H. and Lange, K. (1983) Computer aided design of
case fails. This shows that the circular pro®le is better than blanks for deep drawn irregular shaped components. Manufac-
the ®rst design. It may be also observed that although the turing Engineering Transactions, 11, 243±251.
optimal design from example 2 has greater drawability [7] Hsu, T.C. and Lee, S.Y. (1976) A de®nition of drawability and its
than the optimal design from example 3, the optimal de- relation to deformation in the drawing of square cups. Journal of
sign from example 3 retains much lower out-of-squareness Testing and Evaluation, 4 (5), 340±346.
[8] Majlessi, S.A. and Lee, D. (1993) Deep drawing of square-shaped
as compared to the optimal design of example 2. Thus, sheet metal parts: Part I. ASME Journal of Engineering for In-
from Fig. 16 we note the following general trend: as the dustry, 115, 102±109.
out-of-circularity of the initial blank pro®le decreases, the [9] Majlessi, S.A. and Lee, D. (1993) Deep drawing of square-shaped
resulting out-of-squareness of the ®nal ¯ange pro®le de- sheet metal parts: Part II. ASME Journal of Engineering for In-
creases much quicker with punch depth and consequently dustry, 115, 110±117.
[10] PAM-STAMP user's manual (1994) Pam system international,
the circle shape can be considered to be an overall opti- ESI group software product Company.
mum blank shape for square shell drawing. A similar [11] Hill, R. (1958) A general theory of uniqueness and stability in
conclusion has been arrived at by El-Wakil et al. in an elastic-plastic solids. Journal of Mechanics and Physics of Solids, 6
experimental study of square box forming operations [15]. (3), 236±249.
[12] Keeler, S. P. (1965) Determination of forming limits in automo-
tive stampings. Sheet Metal Industry, 42, 683±691.
[13] Marciniak, Z. and Kuczynski, K. (1967) Limit strains in the
4. Conclusions processes of stretch forming of sheet metal. International Journal
of Mechanical Sciences, 9, 608±620.
This numerical study represents an initial step towards [14] IDESIGN user's manual (1989) Technical Report No. ODL-89-7,
the development of a rigorous computer aided tool for Optimal Design Laboratory, College of Engineering, The Uni-
versity of Iowa.
blank design optimization of deep drawn parts. The [15] El-Wakil, S. D., Kamal, M. N. E. M. and Darwish, A. H. (1980)
methodology is based upon an optimization formulation Mechanics of the square box forming operation of aluminium
whose goal is to maximize the drawability in the deep blanks. Sheet Metal Industries, 57, 679±689.
drawing process of square shells. Three blank design
cases are studied and compared. The optimal designs
obtained from the formulation has considerably im- Biographies
proved drawability as compared to the drawability of the
initial designs in each of the three cases. Note that the Hae Chang Gea obtained a Bachelor of Science Degree in Mechanical
values of the design variables corresponding to the opti- Engineering from National Taiwan University in 1982. He then entered
the University of Southern California (USC) and completed a Master
mal designs generally cannot be found by a trial and error of Science degree in Mechanical Engineering in 1985. He then worked
technique thereby highlighting the usefulness of our op- as a design engineer and senior system analyst at various engineering
timization formulation for this purpose. companies for 4 years. In 1993, he completed his Ph.D. at the Uni-
The shapes of the optimum blank pro®les were found versity of Michigan and joined the faculty in the Department of Me-
to be very nearly circular. By comparing the out-of- chanical and Aerospace Engineering at Rutgers, The State University
of New Jersey. His research interests are in design optimization, sto-
squareness of the ®nal ¯ange pro®les for each of the three chastic optimization and structural optimization. His research work
design cases along with the out-of-circularity of their has received considerable attention from federal agencies and indus-
optimal blank pro®les, we observe the following trend: tries. He has authored/co-authored over 30 refereed journal and con-
the lower the out-of-circularity of the starting blank ference articles and was awarded ASME Design Automation The Best
pro®les the smaller is the out-of-squareness of the ®nal Technical Paper in 1992 and again in 1997.
¯ange pro®le. i.e., more uniform is the ®nal ¯ange pro®le. Rajesh Ramamurthy was a graduate assistant at Rutgers University.
Thus, the circle shape can be considered to be an overall He is currently a graduate student at the University of Michigan, Ann
optimum blank shape for square shell drawing. Arbor.