Sheet Metal Forming Using

Y. Park Polymer Composite Rapid

Graduate Research Assistant

J. S. Colton
Prototype Tooling
Professor
Fellow ASME To meet the growing demand for rapid, low-cost die fabrication technology in the sheet
metal forming industry, easy-to-machine, polyurethane-based, composite board stock is
The George W. Woodruff School of Mechanical widely used as a rapid tooling material. However, the failure mechanisms of the rapid
Engineering, prototyped tools are not clearly understood, thus making the prediction of tool life diffi-
Georgia Institute of Technology, cult. As a fundamental step for effective tool life estimation, the microstructure and the
Atlanta, GA 30332-0405 mechanical properties of the polymer composite tooling material were characterized. A
finite element model of 90° V-die bending process was developed, and the effects of
process parameters on stress distribution in punch and die were investigated through
simulation. The simulation results were verified through experiments using instrumented,
laboratory-scale punch and die sets. 关DOI: 10.1115/1.1543971兴

1 Introduction based on the knowledge of underlying process mechanics and

Sheet metal forming is a major fabrication process in many
sectors of industry. Throughout the years, technological advances
have allowed the production of extremely complex parts. How-
ever, the evolution of die design and ultimate tryout has been a
slow, cautious process based on the trial-and-error and the expe-
riences of skilled workers 关1兴.
As today’s ever-competitive business environment demands re-
ductions in product development time and cost, the need for faster
turn-around times and more efficient means of producing proto-
type and short-run tooling has increased. As a result, rapid tooling
technologies have made inroads into conventional die fabrication
methods with the aim of reducing the lead time and investment
costs of tooling development.
One category of rapid tooling technology involves the applica-
tion of advanced polymers and composite materials to fabricate
sheet metal forming dies. One such material is aluminum
trihydrate共ATH兲-filled polyurethane, which offers easy handling,
fast lead times and low costs while maintaining high-quality stan-
dards. Compared with conventional die materials, durable poly-
urethane board stock allows high machining rates, which bring
about reductions in tooling time and overall production costs 关2兴.
Despite their advantages in terms of lead time and cost, poly-
mer composite dies for sheet metal forming application have sev-
eral drawbacks. Due to their lack of strength as compared to con-
ventional die materials, the use of polymer composite dies is often
limited to prototype or short-run production. Also, because the
mechanisms by which they fail are not fully understood, dies are
still designed on the basis of experience and intuition, often em-
ploying the design rules established for metal dies. These factors
often lead to an inappropriate selection of geometry and dimen-
sions, which may result in the premature fatigue failure of the
dies.
Unlike polymer composite dies, a number of studies have been
performed to analyze the fatigue failure of the die and to achieve
better tool design in the case of metal dies—both computationally
and experimentally. Most of them involve computer-aided tech-
niques with the common goal of reducing development time and
cost by replacing full-scale process trials with simulations. Nu-
merical simulation and modeling of metal forming processes,
validated by experimental results, provide a powerful tool for op-
timizing process parameters 关3兴.
Knoerr et al. 关4兴 and Geiger et al. 关5兴 introduced computer-
aided fatigue analysis concepts for cold forging with the aim of
estimating tool life and optimizing tool geometry. The analyses
provided an estimate of the cycles to crack initiation and a method
to reduce the effective stress at the tool surface. Falk et al. 关6兴
conducted a comparative study to assess the applicability of dif-
ferent fatigue failure concepts for a closed, cold forging die.
As for sheet metal forming, the formability and failure of sheet
metals have been of great interest. Yet, not much attention has
been paid to the failure and life prediction of the dies. A study has
been performed by Jensen et al. 关7兴 to obtain the distribution of
tool wear on the die profile in deep drawing. However, the fatigue
life of the die must be considered simultaneously, since it is im-
portant to design the tool in such a way that the service life of the
tool ends by wear rather than by catastrophic fatigue failure 关4兴.
A different approach is necessary to predict the fatigue life of
the polymer composite tool and to optimize tool geometry to en-
hance tool life. The stress state in the dies must be determined
accurately, and the fatigue mechanisms in brittle polymers must
be incorporated in the fatigue life prediction method. Ratner and
Potapova 关8兴 showed that the maximum principal stresses are the
criterion for brittle failure in multicycle fatigue. Suresh 关9兴 docu-
mented that cyclic deformation and the subcritical advance of fa-
tigue fracture in many polymers are dictated by the nucleation,
growth, and breakdown of crazes, and that the criterion for the
nucleation of a craze involves maximum and minimum principal
stresses. Ritchie et al. 关10兴 reported that the mechanisms of fa-
tigue damage in brittle materials are inherently different from
those in metals, and that much future effort must be devoted to the
topic.
As a fundamental step for predicting the fatigue life of the
polymer composite tool in sheet metal forming, this paper pre-
sents a parameter study to determine the dominant process param-
eters that govern tool life. A 90° V-die bending process was mod-
eled using a finite element method, and the finite element model
was verified experimentally. In parallel, material testing was per-
formed to characterize the composition and mechanical properties
of the tooling material.

Contributed by the Materials Division for publication in the JOURNAL OF ENGI- 2 Material Characterization
NEERING MATERIALS AND TECHNOLOGY. Manuscript received by the Materials
Division October 15, 2001; revision received July 2, 2002. Associate Editor: L. 2.1 Composition. This paper deals with a polymer compos-
Catherine Brinson. ite tooling material specially developed for metal forming appli-

Journal of Engineering Materials and Technology JULY 2003, Vol. 125 Õ 247
Copyright © 2003 by ASME
 Fig. 2 Load-displacement curves in fracture toughness tests

Ren Shape 5166 is that its compressive strength 共86 MPa; manu-
facturer’s data兲 is significantly higher than its tensile strength 共33
MPa; measured data兲. This qualifies Ren Shape 5166 as an effec-
tive tooling material since the dies tend to be subject to high
compressive loads in metal forming.
Plane strain fracture toughness tests were performed in accor-
dance with ASTM D 5045 to characterize the toughness of the
tooling material. Single-edge-notch bending 共SENB兲 method was
employed. The linearity of the load-displacement curves in Fig. 2
validates the assumption of linear elastic behavior of the cracked
specimens. Three replicate tests yield the critical stress intensity
factor K Ic of 2.20 MPa冑m with a standard deviation of
0.02 MPa冑m, which falls within the typical range for polymers
that are useful as engineering materials.
Subsequent fractographic analysis was performed on the frac-
Fig. 1 Mechanical behavior of Ren Shape 5166: „a… stress- ture surfaces of the SENB specimens to identify the mode of
strain curves in tensile tests; and „b… load-deflection curves in failure. On a macroscopic level, precise matches of the fracture
flexural tests. surfaces indicate the absence of significant plastic deformation.
The fractographs taken using an optical stereomicroscope 共Fig. 3兲
cations known as Ren Shape 5166 by Vantico Inc. The material reveal some key features of the fracture surfaces. The fracture
has a thermosetting polyurethane base, filled with aluminum tri- mechanism is dominated by debonding between the ATH particles
hydrate共ATH兲. The spherical ATH is randomly dispersed in the and the polyurethane matrix. A vast distribution of voids, as well
matrix to impart material isotropy and adhesion. More impor- as the sites of debonding over the surfaces, shows that the crack
tantly, ATH serves to increase the overall compressive strength initiated at the razor notch of the SENB specimen and propagated
and to improve tribological characteristics. through voids and along filler-matrix interfaces. Although no
Pyrolysis was performed according to ASTM D 2584 to obtain gross plastic yielding is evident, signs of localized plastic defor-
the weight percent of the filler. Five samples were heated at 565°C mation can be observed. The brittle fracture along the edges of the
for two hours to allow the complete ignition loss of the polyure- fracture surface is accompanied by plastic flow processes, where
thane matrix. The content of ATH is calculated to be 68.7% by the strain energy absorption is high relative to that in the bulk
weight with a standard deviation of 0.2%. region.
2.2 Mechanical Properties and Fracture Behavior. Fun- 2.3 Tribological Properties. Lubrication and friction con-
damental mechanical tests were performed to identify the tensile ditions in sheet metal forming are important in lowering forces,
and flexural properties of the material according to ASTM D 638 increasing drawability, and reducing tool wear. In conventional
and ASTM D 790, respectively. The corresponding stress-strain deep drawing, lubrication is minimized as the friction between the
and load-deflection curves are plotted in Fig. 1, and the numerical punch and the blank tends to improve drawability. However, the
results are summarized in Table 1. The linearity of the stress-strain type and the amount of lubrication applied require experience and
response indicates that the material is brittle and that little plastic reference to engineering data, as excessive friction may result in
deformation takes place before failure. One key characteristic of eventual necking and tearing of the sheet metal.
The polymer composite rapid tooling material is advantageous
because it eliminates the need for lubrication. As the material
Table 1 Mechanical properties of Ren Shape 5166
undergoes machining, the polyurethane matrix encapsulates the
Property Tested Manufacturer ATH filler particles. The heat generated during machining creates
polyurethane coating over the machined surface and reduces fric-
Elastic modulus 7.2 GPa 7.2 GPa tion. As reported by Tadmor and Gogos 关11兴, in polymer-metal
Yield strength 共0.2% offset兲 32 MPa 32 MPa
Ultimate tensile strength 33 MPa 34 MPa dry friction, both adhesion and ploughing contribute to friction,
Flexural modulus 3-point 6.8 GPa 6.7 GPa adhesion usually being the dominant factor. The soft nature of
4-point 6.2 GPa N/A polymers as compared to metals makes the former susceptible to
Flexural strength 3-point 62 MPa 55 MPa being ploughed. If the adhesive force between the metal and the
4-point 54 MPa N/A
Compressive modulus N/A 5.8 GPa polymer is greater than the cohesive polymer force, sliding occurs
Compressive strength 共0.2% offset兲 N/A 86 MPa at a plane within the polymer, resulting in the kinematic friction
coefficient of greater than 0.2. The sliding speed has only moder-

248 Õ Vol. 125, JULY 2003 Transactions of the ASME

 Fig. 5 Stages in V-die bending †12‡

Fig. 3 Stereomicroscopic fractographs of a fracture surface:
„a… at 10X magnification; and „b… at 30X magnification.
3 Process Simulation
3.1 Finite Element Model. Bending is one of the most
commonly practiced metal forming operations. A 90° V-die bend-
ing process is illustrated in Fig. 5 关12兴. During V-die bending, the
punch slides down, coming first into contact with the unsupported
sheet metal. By progressing farther down, the punch forces the
material to follow along, until finally bottoming on the ‘‘V’’ shape
of the die. It is evident from the figure that the loading condition
experienced by the sheet metal resembles three-point beam bend-
ing. Conversely, although it may not be as obvious, the die itself is
subject to a similar condition with the contact points against the
sheet metal as beam supports, the significance of which will be
addressed later.
A two-dimensional ‘‘base’’ model was constructed using the
commercial package ABAQUS/CAE as shown in Fig. 6. The
‘‘base’’ model will later be used as a reference model in the pa-
rameter study. Due to symmetry, only half of the model was taken

ate effect on the coefficient of friction. An increase in speed tends

to increase the friction coefficient. A typical polymer has a kine-
matic friction coefficient of approximately 0.3 when the speed
reaches 10 cm/s.
An experimental setup was designed to measure static friction
coefficient between the surfaces of metal sheet and Ren Shape
5166 at room temperature. The coefficients were measured at
seven different normal loads, and five replications were made for
each load. The polymer sliding blocks were polished to the degree
equivalent to an actual die surface, and they were allowed to slide
on a 3003-H14 aluminum sheet surface. All contact surfaces were
treated with isopropyl alcohol prior to each test to remove con-
tamination. As Fig. 4 suggests, static friction coefficient decreases
with increasing normal load and becomes constant at high normal
loads, with a value of approximately 0.25.

Fig. 6 Finite element model with boundary conditions and

Fig. 4 Friction coefficient versus normal load mesh

Journal of Engineering Materials and Technology JULY 2003, Vol. 125 Õ 249
Table 2 Process conditions for finite element analysis and
experiment

A B
FE Model
Process Parameter ‘‘Base’’ FE Model for Validation
Bend radius R b 5 mm 5 mm
Die shoulder radius R s 5 mm 5 mm
Die opening W 30 mm 25 mm
Die width W d 100 mm 100 mm
Die height H 45 mm 45 mm
Die thickness B 1 mm 8.5 mm
共into the paper兲
Sheet material 1100-O Al 3003-H14 Al
Sheet thickness T 1 mm 0.8 mm
Punch travel distance d p 12.5 mm 9.5 mm
Punch speed v p 120 mm/min 120 mm/min
Friction No friction ␮⫽0.2 Fig. 7 Maximum principal stress field in the die „magnified at
Miscellaneous Plane strain Plane stress
the bend regions…: „a… Ren Shape 5166; and „b… Steel.

The punch was not considered because it is subject to lower

into account. The model consisted of three parts: punch, die and
workpiece. The punch and the die were modeled as isotropic,
elastically deformable bodies with small elastic strains 共typically
less than 5%兲 and the mechanical properties of Ren Shape 5166.
Elastic modulus E of 7.2 GPa 共from tensile tests兲 and Poisson’s
ratio ␯ of 0.34 共typical for a thermoset 关13兴兲 were used. 1100-O
aluminum was selected as the sheet material, and it was modeled
as elastic and strain-hardening to realistically account for plastic
deformation. It should be noted that plane strain condition was
assumed throughout the analysis. Process conditions are summa-
rized in Column A of Table 2.
Figure 6 also illustrates the boundary conditions imposed on the
model. The displacements of all three components along the plane
of symmetry were constrained in the horizontal direction. The
bottom of the die was constrained only in the vertical direction,
allowing the die to deform in the horizontal direction to the right
of the plane of symmetry. A uniform displacement was applied at
the top surface of the punch to simulate the downward stroke of
the bending process. The punch was allowed to travel so as to
bottom exactly at the end of the stroke.
Sheet metal forming involves large deformation, and therefore
it can be treated as a nonlinear static problem. Due to the nonlin-
ear nature of the process, instability in the solution may result.
Accordingly, an iterative method was used to find the optimum
mesh density. The minimum mesh size was determined by refin-
ing the mesh until solution convergence was reached and localized
instability was eliminated. Different mesh sizes were employed
for different regions of the model; a smaller mesh was applied in
the regions where the stress gradient was expected to be large.
Since the location of crack initiation corresponds to the location of
the highest maximum tensile principal stress, the magnitude of the
maximum tensile principal stress in the die was employed as a
measure for convergence.
The minimum mesh size along the ‘‘V’’ profile in the die was
varied from 0.2 mm to 0.5 mm by increments of 0.1 mm. The
maximum principal stresses when the punch was bottomed were
compared, and the size was selected where further refinement no
longer resulted in significant improvement in the accuracy of the
solution. When the minimum mesh size of 0.4 mm or higher was
used, instability in solution was manifested by chatter in the stress
values along the ‘‘V’’ profile. As the minimum mesh size was
reduced to 0.3 mm, chatter was no longer present; further reduc-
tion to 0.2 mm did not change the plot significantly. Therefore, 0.3
mm was chosen as the optimum value for the minimum mesh size
in the current model. The model contained 1078, 2256, and 1114
elements and 1000, 1959, and 980 nodes in the punch, die and
sheet, respectively.
3.2 Model Validation. As a preliminary step to validate the
finite element analysis results, stress fields in the die were studied.
stresses compared to the die in V-die bending. Figure 7 compares
the maximum principal stress fields in Ren Shape 5166 and steel
dies when the punch is bottomed on the die. Typical mechanical
properties of E ⫽ 200 GPa and ␯ ⫽ 0.3 were used for steel. In
both cases, large stress gradients are observed in the ‘‘V’’ region,
whereas stresses are uniformly distributed in the bulk region. It
should be noted from the magnified views in the bend regions that
the steel die exhibits a greater degree of stress concentration due
to its higher stiffness. However, the polymer composite die is
more prone to crack initiation because its stress level is signifi-
cantly higher relative to its tensile strength.
To further validate the finite element model, the simulation re-
sults were compared with experimental results and the literature.
A new set of process conditions was selected to accommodate
machining limitations, such as the capacity of the load frame and
load cell, and material availability. 共column B of Table 2兲 A finite
element analysis was performed accordingly to obtain the bending
force and the stress-strain response in the die. Plane stress condi-
tion 共i.e., small thickness兲 was assumed in order to allow the
comparison with experimental results using strain gages, as will
be discussed later in the section. The kinetic friction coefficient at
punch-sheet-die interfaces was chosen to be 0.2, based on the
previous friction tests 共Fig. 4兲.
Figure 8 illustrates the laboratory-designed experimental setup
for sheet metal bending. An Instron Universal Testing System
共Model 4466兲 was used to apply the bending load, and Instron
Series IX Software was used for data acquisition. The die set was
CNC-machined from Ren Shape 5166, and 3003-H14 aluminum
was selected as the sheet material. The operation was programmed
Fig. 8 V-die bending experimental setup

250 Õ Vol. 125, JULY 2003 Transactions of the ASME

 Fig. 9 Comparison of bending forces

such that the punch is stopped when the bending force reaches
100 N. Load and punch travel distance were recorded as the sheet
metal was bent. Fig. 10 Strain gage locations and strain measurement
Experimental and simulated bending forces are plotted as the directions
functions of bend angle in Fig. 9, which illustrates a good agree-
ment between them. As explained by Kalpakjian 关13兴, the bending
force is a function of punch travel. It increases from zero to a
maximum and may decrease as the bend is completed. The force
4 Parameter Study
then increases sharply as the punch bottoms. The stress state in the die is the most important factor that
Bending forces can be estimated by assuming that the process is governs fatigue failure. Especially in the case of mechanical fa-
a case of simple beam bending. Then the maximum bending force tigue failure, the principal stresses must be taken into account.
can be expressed as 关13兴: The primary purpose of the parameter study is to find out the
effects of parameter variation on the stress state in the dies. From
共 UTS兲 LT 2 the outcome, dominant parameters can be determined, which can
P max ⫽ k (1) be used to optimize die geometry during the design phase and to
W
aid in setting process parameters.
where 4.1 Fatigue Failure Criteria. It was found from the pre-
k ⫽ die opening factor 共1.2⬃1.33兲 liminary finite element simulation that the deformation of the die
takes place mainly in the elastic regime. Therefore, in sheet metal
UTS ⫽ ultimate tensile strength of sheet material
forming, the fatigue mode of the die can be described as high
L ⫽ length of bend cycle fatigue since the stress amplitude is typically below the
T ⫽ sheet thickness yield stress of the material. Moreover, if it is assumed that no
cracks pre-exist, the fatigue life is controlled by the number of
W ⫽ die opening. cycles to crack initiation, and the contribution of crack propaga-
Equation 共1兲 yields the estimated maximum bending force of 41.6 tion can be neglected.
N when k⫽1.2 is used. 共k varies from 1.2 for W⫽16T to 1.33 for In this study, the stress-life (S-N) approach was investigated,
W⫽8T 关14兴.兲 The simulation and experimental values show good which is typically used in high cycle fatigue involving constant
agreement with the estimated value. amplitude loading and negligible plastic strains 关15兴. One such
While the comparison of bending force profiles validates the approach is the local stress-based fatigue approach, based on the
finite element analysis results in a global sense, it also is necessary assumption that a crack develops at a point that experiences the
to check if the results are locally satisfied. Three 45° rectangular greatest tensile cyclic loading. One example of such points occurs
stacked rosette strain gages were bonded to the die surface at at the bend region of the die as shown in Fig. 11. The sharp peak
representative points 共Figs. 8 and 10兲, and corresponding strains evident in the simulated data corresponds to the instance when the
were measured in the directions as indicated in Fig. 10. Two gages punch has traveled its maximum stroke. The stress increases
with 0.38 mm gage lengths measured the strains in the bend and sharply to its maximum value as the punch bottoms and drops to
die shoulder regions, where strain gradients were large; one gage zero as the punch retracts. This suggests that the die is subject to
with a 3.05 mm gage length measured the strains in the bulk zero-to-tension cyclic loading with the conditions listed in Eq. 共2兲.
region of the die. A larger gage was selected for the latter because
the strain gradients are relatively small in the bulk region, and
therefore averaging the strains over the gage length provides a
Table 3 Comparison of strains at various locations in the die
more accurate value corresponding to the strain in the coarse bulk
element in the finite element model. Strain 关␮-strain兴
The punch was allowed to travel until the bending force
reached 300 N at the instant of bottoming. The elastic strains in Location Orientation Notation Simulated Experimental
the selected directions at the gage locations were collected at this Die bend 0° (␧ 1 ) 0 354 395
instant and compared with simulated values. As shown in Table 3, 90° (␧ 1 ) 90 ⫺976 ⫺955
the two sets of data indicate that the finite element model and Die shoulder 45° (␧ 2 ) 45 22 25
90° (␧ 2 ) 90 ⫺40 ⫺39
analysis yield accurate results. As the gage locations encompass Die center 45° (␧ 3 ) 45 ⫺42 ⫺38
both the regions of high and low strain gradients, the validity of 共bulk兲 90° (␧ 3 ) 90 ⫺71 ⫺83
the finite element model can be extended to the entire model.

Journal of Engineering Materials and Technology JULY 2003, Vol. 125 Õ 251

Fig. 11 Time history plot of maximum principal stress at bend
region
␴ min
R⫽ ⫽0 (2a)
␴ max
␴ max
␴m ⫽ ␴a ⫽ (2b)
2 maximum stresses of 50, 40, and 35 MPa each. The specimens
where R is the stress ratio, ␴ m is the mean stress, and ␴ a is the
stress amplitude. Once the stress amplitude is known, the con-
straint given by Eq. 共2b兲 reduces the number of S-N curves that
are necessary to determine fatigue life to one.
Another S-N model that is especially useful for glass-fiber-
reinforced polymer composites was provided by Adkins and
Kander 关16兴 as in Eq. 共3兲.
␴ max ⫽ ␴ u 共 1⫺m log N f 兲 (3)
where ␴ u ⫽ ultimate static stress of die material
m ⫽ slope of the S-N line on a log-linear plot
N f ⫽ number of cycles to failure.
Equation 共3兲 also assumes zero-to-tension cyclic loading, that is,
the conditions in Eqs. 共2a兲 and 共2b兲 apply, and ␴ max is the local
maximum tensile principal stress during each cycle. In their ex-
perimental work, Adkins and Kander showed that the glass-
reinforced composites follow a common S-N line with a slope, m,
of approximately 0.1. This simple rule allows fatigue life estima-
tion for preliminary design and material selection.
Time history plots of stresses at various points in the die indi-
cate that the points, in general, experience a range of stress levels,
from compression to tension, during each bending cycle. There-
fore, mean stresses may have significant effects on the fatigue
behavior of the die. In order to account for mean stress effects,
Smith, Watson, and Topper 关17兴 proposed the following equation
as a cumulative damage evaluation method. 共Eq. 共4兲兲
冑␴ max␴ a ⫽ ␴ ⬘f 共 2N f 兲 b (4)
where ␴ ⬘f and b are material constants. It should be noted that Eq.
共4兲 is applicable when stresses are approximately proportional to
strains and when the S-N curves for completely reversed cyclic
loading are available. It is also assumed that no fatigue damage
occurs when ␴ max ⬍ 0. The assumption that small laboratory
specimens can be used to simulate the behavior of material at
critical locations indicates that the equation should be applicable
to crack initiation and the early propagation stage in more com-
plex structures.
Figure 12 shows the S-N data obtained from zero-to-tension
and fully reversed fatigue tests for Ren Shape 5166. The tests
were performed in a four-point flexural test mode 共which tends to
be more conservative than a three-point mode兲 with the specimens
prepared according to ASTM D 790. The test specimens measured
252 Õ Vol. 125, JULY 2003
Fig. 12 S-N data for Ren Shape 5166
127 mm by 12.7 mm with a thickness of 3.2 mm, and the support
span-to-depth ratio was 16. The cyclic loading was strain-
controlled, and the loading frequency of 1 Hz was used to mini-
mize hysteretic heating.
For the zero-to-tension tests, three specimens were tested at the
tested below 35 MPa lasted more than 105 cycles, which can be
loosely interpreted as the endurance limit from the prototype and
short-run production viewpoints. The S-N data in Fig. 12 indicate
a linear relation between the maximum stress and the life, and it
can be expressed as Eq. 共5兲.
␴ max ⫽ 54共 0.97 ⫺ 0.080 log N f 兲 关 MPa兴 (5)
Equation 共5兲 resembles Adkins and Kander’s model 共Eq. 共3兲兲
when the ultimate flexural strength of ␴ u ⫽ 54 MPa 共Table 1兲 and
the slope of m ⫽ 0.1 are used. Therefore, it can be concluded that
the fatigue behavior of Ren Shape 5166 is similar to that of glass-
reinforced composites.
For the fully reversed tests, three specimens were tested at the
stress amplitudes of 45, 40, 35, and 32.5 MPa each. The S-N
relationship is also linear and can be expressed as Eq. 共6兲.
␴ a ⫽ 51.33 ⫺ 4.56 log N f 关 MPa兴 (6)
In both cases, it should be noted that the data become more
scattered as the stress level decreases.
4.2 Effects of Parameter Variations. Finite element analy-
sis was performed for 90° V-die bending, varying one parameter
at a time while holding all the others constant. The ‘‘base’’ model
was employed as the reference model 共Column A of Table 2兲. The
parameters under consideration include: 共1兲 geometric parameters,
such as bend radius R b , die shoulder radius R s , die opening W,
overall die size 共die width兲 W d , sheet thickness T, and die notch-
to-height ratio D/H 共Fig. 10兲, 共2兲 material parameters, such as
sheet material and friction at punch-sheet-die interfaces, and 共3兲
process parameters, such as punch travel d p and boundary con-
straint on the die imposed by the fixture 共Fig. 6兲. Two types of
output were investigated, the maximum tensile principal stress
␴ max and the SWT 共Smith-Watson-Topper兲 parameter ␴ SWT
⫽ 冑␴ max␴a, because they can be employed most effectively to
describe critical local stresses in relation to fatigue.
The range of each parameter was selected based on the guide-
line provided by die design handbooks 关12,18兴 and reflects the
most commonly used dimensions and working conditions in in-
dustry. Table 4 shows the sensitivity of the stress state to param-
eter variations. A plane strain condition was assumed in all cases.
For each simulation, the following steps were taken to find the
greatest values of ␴ max and ␴ SWT throughout each bending cycle:
共1兲 first, the magnitudes and directions of maximum and minimum
principal stresses were found at each time step at each element in
the finite element model, 共2兲 concurrently, the magnitudes of SWT
Transactions of the ASME
 Table 4 Effects of parameter variation

Parameter Value ␴ max 关MPa兴 ␴ SWT 关MPa]

Bend radius R b 3 8.77 8.35
关mm兴 5 4.31 5.50
7 2.54 2.06
10 2.11 1.55
Die shoulder radius R s 3 4.39 5.89
关mm兴 5 4.31 5.50
7 4.19 4.99
10 4.15 4.55
Die opening W 20 3.89 4.70
关mm兴 30 4.31 5.50
40 5.72 7.86
50 6.13 8.13
Die width W d 80 4.70 5.71
关mm兴 90 4.47 5.28
100 4.31 5.50
110 4.19 5.25
Sheet thickness T 0.6 2.95 3.13
关mm兴 1.0 4.31 5.50
1.4 6.82 7.10
Die notch-to-height 0.50 4.31 5.50
ratio D/H 0.64 4.87 6.54
0.72 5.60 7.27
Friction coefficient ␮ 0 4.31 5.50
0.1 3.42 4.83
0.2 3.41 5.77
0.3 3.68 6.19
0.4 4.46 5.68
Sheet strength 125 4.31 5.50
共Yield stress⫹UTS兲 243.3 7.66 7.95
关MPa兴 285 9.46 8.63
400 12.28 13.67
Punch over-travel ⌬d p 0.1 19.10 13.51
关mm兴 0.15 26.11 18.47
0.2 32.03 22.65
0.25 35.67 25.23
0.3 41.42 29.29
0.35 46.26 34.58
Die constraint Unconstrained 4.31 5.50
Constrained 3.98 5.39

Fig. 13 Magnified view at bend region: „a… schematics of

punch-sheet-die interfaces; and „b… top view of ‘‘V’’ surface of
the die after 1000 stamping cycles.
parameters were calculated at corresponding instances and loca-
tions, 共3兲 then, all the values of ␴ max and ␴ SWT obtained from
steps 共1兲 and 共2兲 were compared to one another to determine the
maximum value of each. As the stress state tends to be most
severe along the ‘‘V’’ profile of the die, where a direct contact
between the sheet and die occurs, only the elements along the
profile from the bottom of the ‘‘V’’ to the die shoulder were taken
into account.
Before investigating the effects of parameter variations, it is
important to note how the physical constraints imposed by the
geometry of the die set result in different magnitudes and loca-
tions of maximum stresses. The magnified view of the bend region
in Fig. 13共a兲 illustrates the effects of die geometry at the final
stage of V-die bending. Although the punch, sheet, and die are
modeled to match perfectly after deformation, it can be noted that
the three entities do not conform completely at their contact sur-
faces. The gap at region A suggests that the die may be subject to
a loading condition similar to three-point beam bending, due to
the deformation of the parts and the stiffness of the sheet metal.
As a result, region B 共where the sheet metal comes into contact
with the die兲 acts as one of the beam supports and subjects the die
to compressive stresses. Therefore, regions A and B are the points
of interest, which are dominated by high tensile and compressive
stresses, respectively.
This phenomenon is illustrated in Fig. 13共b兲, which shows a top
view of the ‘‘V’’ surface of the die after 1000 hits. The dark
regions indicate the areas where the aluminum particles rubbed
off, and the light regions indicate the areas that have remained
intact from any contact with the aluminum sheet.
The sensitivity of the stress level to a parameter was evaluated
qualitatively by studying the change of the stress values, namely
␴ max and ␴ SWT , within the range of the parameter of interest.
Table 4 clearly shows that punch over-travel ⌬d p 共defined as the
amount of punch travel in addition to the distance for exact bot-
toming兲 has the most significant effects on both ␴ max and ␴ SWT .
Referring to the S-N data in Fig. 12, in all cases except when ⌬d p
is greater than 0.2 mm, the die will fail by wear. When ⌬d p is
0.25, 0.3, 0.35 mm, Eq. 共5兲 yields estimated fatigue lives of 7867,
367, 28 cycles, respectively. When ␴ SWT is used as the fatigue
failure criterion, the die is likely to fail by wear in all cases. Thus,
in terms of fatigue failure prediction, it can be said that ␴ max is the
more conservative criterion.
In addition to punch over-travel, bend radius, sheet thickness,
and strength of sheet material, in all three of which the strength is
represented by the sum of the yield and ultimate tensile strengths,
have significant effects on ␴ max and ␴ SWT . Sheet material has a
significant influence because even when the punch is bottomed,
the sheet, punch, and die surfaces are not in full contact 共Fig.
13共a兲兲, the configuration depending on the sheet strength. Die
opening also has an effect to some degree, but it is not as influ-
ential as other parameters. In addition, it can be noted that the
critical stress level can be reduced to a certain extent if the out-
ward expansion of the die is constrained, for example, by using an
insert-type die or by applying the fixture to the sides of the die.
The die notch-to-height ratio D/H has a moderate effect on die
stress. However, its influence becomes much more significant
when it is coupled with punch over-travel as is further discussed
in detail in the next section.
It should be pointed out that the greatest value of ␴ max occurs in
region A of Fig. 13共a兲, while that of ␴ SWT occurs in region B. This
is because ␴ max is by definition the maximum ‘‘tensile’’ principal

Journal of Engineering Materials and Technology JULY 2003, Vol. 125 Õ 253

Fig. 14 Failure modes of V-bending dies: „a… after 5000 stamp-
ing cycles „R b Ä5 mm, F maxÄ8 kN…; „b… after one stamping cycle
„R b Ä3 mm, F maxÄ10 kN…; and „c… after 507 stamping cycles
„R b Ä3 mm, F maxÄ8 kN….
stress while ␴ SWT incorporates both the maximum tensile and
compressive stresses 共in absolute values兲 by the use of stress am-
plitude ␴ a . Moreover, Table 4 indicates that in most cases ␴ SWT is
slightly higher than ␴ max . In V-die bending, a vast majority of the
die is subject to compressive loading, especially in region B of
Fig. 13共a兲, and large compressive stresses contribute to the rela-
tively larger values of ␴ SWT .
Another interpretation of the difference between the two stress
quantities can be made from the viewpoint of their functions as
indices for determining the fatigue life from S-N curves. As ex-
plained in the previous section, the methods for estimating fatigue
life using stress-based approach are two-fold: 共1兲 to compare the
greatest maximum tensile principal stress ␴ max to an S-N curve
obtained from zero-to-tension fatigue tests and 共2兲 to compare the
greatest SWT parameter ␴ SWT to an S-N curve obtained from
completely reversed fatigue tests, both performed at various stress
amplitudes. It can be expected intuitively that the S-N curve in the
former case will be lower than that in the latter case, since in-
creasing the mean stress in the tensile direction must be accom-
panied by a decrease in stress amplitude to maintain the same life.
Accordingly, the values of two stresses must differ in order to
yield the same fatigue life.
5 Experimental Results
V-bending experiments were performed to verify and compare
the two proposed fatigue failure criteria and to demonstrate the
effect of parameter variation. In order to accommodate the load
frame and load cell capacity, the conditions in Column B of Table
2 were applied except that the die opening of 30 mm, die height of
17 mm, die thickness of 10 mm, and sheet thickness of 1 mm
were used. The selected die height represented the D/H ratio of
0.7, which would be large enough to drive the fatigue failure in
the die under the given load limitation. Two different values of
bend radius were applied: 3 mm and 5 mm. In each case, the
punch travel distance was selected such that the maximum bend-
ing force was limited to 8 kN, which is 2 kN shy of the load frame
and load cell capacity. The corresponding finite element analyses
were performed to yield die stresses.
Figure 14共a兲 shows the ‘‘V’’ region of a 5 mm bend radius die
after 5000 stamping cycles. The crazing near the ‘‘V’’ corner be-
came visible after a few hits. A crack began to develop after
several thousand stamping cycles, and its location corresponded to
region A of Fig. 13共a兲. However, the die could hold another sev-
254 Õ Vol. 125, JULY 2003
eral thousand cycles up to 5000 hits. The experiment was stopped,
and it was concluded that the selected parameters fulfilled the life
requirement of a rapid tool.
Figures 14共b兲 and 14共c兲 illustrate two different modes of failure
when the bend radius of 3 mm was used. Figure 14共b兲 shows the
failure of the die after one cycle when the maximum bending
force was controlled to 10 kN. Again, a crazing zone was evident
at the ‘‘V’’ corner. However, the crack initiated at the bottom of
the die, which led to a catastrophic fracture. Due to excessive
punch over-travel, the compressive stress incurred by the nose of
the punch became dominant and led to crack initiation at the cen-
ter of the die bottom.
To avoid fracture at initial impact and to maintain the maximum
bending force at a prescribed level of 8 kN, the punch over-travel
was reduced to 1 mm. Figure 14共c兲 shows the die failure after 507
stamping cycles. The crack developed on the right side of the
crazing boundary on the ‘‘V’’ surface and immediately propagated
through to the bottom of the die. This presents a clear illustration
of the effect of bend radius on die failure. In the case of a larger
bend radius as shown in Fig. 14共a兲, the bulk of the die was able to
withstand fast crack propagation, as the die was less subject to
stress concentration.
The three cases presented in Fig. 14 suggest that the maximum
principal stress criterion is the most appropriate method in die
failure prediction. This is supported by the location of crack de-
velopment and the direction of crack propagation. The crack ini-
tiates exactly in region A of Fig. 13共a兲, and it propagates perpen-
dicular to the die surface. Also, as the punch over-travel increases
and the punch, sheet, and die are forced to come into full contact,
the ‘‘V’’ corner directly under the punch nose 共and not region B in
Fig. 13共a兲兲 becomes a more significant region. However, the SWT
parameter can help see the effect of mean and minimum principal
stresses when the effect of punch over-travel is not so dominant.
6 Conclusion
In this study, a computer-aided method to estimate the fatigue
life of a sheet metal forming die fabricated from ATH-filled poly-
urethane was presented. First, engineering data were obtained by
characterizing the material in various perspectives: microstruc-
ture, mechanical and fatigue behavior, and tribological properties.
On the basis of material data, a finite element method was em-
ployed to obtain the stress distribution in the die. In the proposed
local stress-based fatigue approach, the stress levels at critical
regions serve as the indices for determining the cycles to fatigue
failure. Such a method can be applied to a general polymer com-
posite tooling material for metal forming, where the plasticity of
the die is negligible.
A finite element method was also used to investigate the effects
of process parameters on the fatigue life of the die. The parameter
study suggested that all parameters have an influence on the stress
response of die components with varying degrees. Based on the
comparisons of the highest maximum principal stresses and SWT
parameters in the die, it can be concluded that punch travel acts as
the dominant parameter that characterizes the fatigue behavior of
the die in sheet metal bending. Although not as dominant, bend
radius, sheet thickness, strength of sheet material, and die notch-
to-height ratio must be considered to obtain the desired die life.
The dominant process parameters can be used in the design stage
to effectively control the stress level in the die, and thus the re-
sistance to crack initiation. The experimental results suggested
that the maximum principal stress criterion is the better method
for die failure prediction.
Since formed sheet metal parts often include complex geom-
etry, the present study can be extended to investigate more general
sheet metal forming processes, such as deep drawing. The fatigue
data of the polymer composite die material and the parameter
study in this paper establish a foundation for providing reliable
tool life prediction and design optimization guidelines for ad-
vanced polymer tooling materials in metal forming.
Transactions of the ASME
Acknowledgments
The authors are grateful to Vantico Inc. for their technical sup-
port and supply of their products. Thanks are due to Professors
David McDowell, Christopher Lynch, Min Zhou, and Richard
Neu in the George W. Woodruff School of Mechanical Engineer-
ing at Georgia Institute of Technology for their helpful
discussions.
References
关1兴 Keeler, S. P., 1978, ‘‘Sheet Metal Stamping Technology—Need for Fundamen-
tal Understanding,’’ Mechanics of Sheet Metal Forming: Material Behavior
and Deformation Analysis, Koistinen and Wang, eds., Plenum Press, New
York, NY, pp. 3–18.
关2兴 Miller, W., 1999, ‘‘Producing Dies for Rapid Prototyping of Metal Formed
Parts,’’ The Fabricator, 29共4兲, Fabricators and Manufacturers Association In-
ternational 共FMA兲, pp. 44 – 47.
关3兴 Altan, T., and Vazquez, V., 1996, ‘‘Numerical Process Simulation for Tool and
Process Design in Bulk Metal Forming,’’ CIRP Ann., 45共2兲, pp. 599– 615.
关4兴 Knoerr, M., Lange, K., and Altan, T., 1994, ‘‘Fatigue Failure of Cold Forging
Tooling: Causes and Possible Solutions Through Fatigue Analysis,’’ J. Mater.
Process. Technol., 46, pp. 57–71.
关5兴 Geiger, M., Hansel, M., and Rebhan, T., 1992, ‘‘Improving the Fatigue Resis-
tance of Cold Forging Tools by FE Simulation and Computer Aided Die Shape
Optimization,’’ IMechE, Part B: Journal of Engineering Manufacture, 206, pp.
143–150.
关6兴 Falk, B., Engel, U., and Geiger, M., 1998, ‘‘Estimation of Tool Life in Bulk
Journal of Engineering Materials and Technology
Metal Forming Based on Different Failure Concepts,’’ J. Mater. Process. Tech-
nol., 80–81, pp. 602– 607.
关7兴 Jensen, M. R., Damborg, F. F., Nielsen, K. B., and Danckert, J., 1998, ‘‘Ap-
plying the Finite-Element Method for Determination of Tool Wear in Conven-
tional Deep-Drawing,’’ J. Mater. Process. Technol., 83, pp. 98 –105.
关8兴 Ratner, S. B., and Potapova, L. B., 1991, ‘‘Multicycle Fatigue Resistance of
Brittle Polymers,’’ Mech. Compos. Mater., 26共4兲, pp. 463– 467.
关9兴 Suresh, S., 1998, Fatigue of Materials, Cambridge University Press, Cam-
bridge, UK.
关10兴 Ritchie, R. O., Gilbert, J. M., and McNaney, J. M., 2000, ‘‘Mechanics and
Mechanisms of Fatigue Damage and Crack Growth in Advanced Materials,’’
Int. J. Solids Struct., 37共1兲, pp. 311–329.
关11兴 Tadmor, Z., and Gogos, C. G., 1979, Principles of Polymer Processing, John
Wiley and Sons, New York, NY.
关12兴 Suchy, I., 1998, Handbook of Die Design, McGraw-Hill, New York, NY.
关13兴 Kalpakjian, S., 1997, Manufacturing Processes for Engineering Materials, 3rd
ed., Addison-Wesley, Menlo Park, CA.
关14兴 American Society for Metals, 1985, Metals Handbook, ASM, Metals Park,
OH.
关15兴 Bannantine, J. A., Comer, J. J., and Handrock, J. L., 1990, Fundamentals of
Metal Fatigue Analysis, Prentice Hall, Englewood Cliffs, NJ.
关16兴 Adkins, D. W., and Kander, R. G., 1988, ‘‘Fatigue Performance of Glass Re-
inforced Thermoplastics,’’ Proceedings of the Fourth Annual Conference on
Advanced Composites, Dearborn, MI, pp. 437– 445.
关17兴 Smith, K. N., Watson, P., and Topper, T. H., 1970, ‘‘A Stress-Strain Function
for the Fatigue of Metals,’’ J. Mater., 5共4兲, pp. 767–778.
关18兴 American Society of Tool and Manufacturing Engineers, 1965, Die Design
Handbook, 2nd ed., McGraw-Hill, New York, NY.
JULY 2003, Vol. 125 Õ 255