Vous êtes sur la page 1sur 12

C.

Hari Manoj Simha


Department of Mechanical Engineering,
University of Waterloo,
Waterloo, ON, Canada, N2L 3G1

Prediction of Necking in Tubular


Hydroforming Using an Extended
Stress-Based Forming Limit
Curve

Javad Gholipour
Institute for Aerospace Research,
National Research Council,
Aerospace Manufacturing Technology Center,
5145 Decelles Avenue,
Campus of the University of Montreal,
Montreal, PQ, Canada H3T 2B2

Alexander Bardelcik
Michael J. Worswick1
e-mail: worswick@lagavulin.uwaterloo.ca
Department of Mechanical Engineering,
University of Waterloo,
Waterloo, ON, Canada, N2L 3G1

This paper presents an extended stress-based forming limit curve (XSFLC) that can be
used to predict the onset of necking in sheet metal loaded under non-proportional load
paths, as well as under three-dimensional stress states. The conventional strain-based
FLC is transformed into the stress-based FLC advanced by Stoughton (1999, Int. J.
Mech. Sci., 42, pp. 127). This, in turn, is converted into the XSFLC, which is characterized by the two invariants, mean stress and equivalent stress. Assuming that the stress
states at the onset of necking under plane stress loading are equivalent to those under
three-dimensional loading, the XSFLC is used in conjunction with finite element computations to predict the onset of necking during tubular hydroforming. Hydroforming of
straight and pre-bent tubes of EN-AW 5018 aluminum alloy and DP 600 steel are considered. Experiments carried out with these geometries and alloys are described and
modeled using finite element computations. These computations, in conjunction with the
XSFLC, allow quantitative predictions of necking pressures; and these predictions are
found to agree to within 10% of the experimentally obtained necking pressures. The
computations also provide a prediction of final failure location with remarkable accuracy. In some cases, the predictions using the XSFLC show some discrepancies when
compared with the experimental results, and this paper addresses potential causes for
these discrepancies. Potential improvements to the framework of the XSFLC are also
discussed. DOI: 10.1115/1.2400269

Introduction

The strain-based forming limit curve FLC, introduced by


Keeler and Backofen 1 and Goodwin 2, is widely used to
predict the onset of necking in sheet metal forming. These curves
are determined through plane stress experiments that subject sheet
samples to load paths that are nominally proportional in principal
strain space and strain paths that range from uniaxial to biaxial
stress. The strain state at the onset of necking in the sample,
characterized by the in-plane true minor and major principal
strains 2 , 1 for a given load path, is recorded and a forming
limit curve in strain space is determined. For linear load paths,
this curve represents the limit of formability for the as-received
sheet.
However, for load paths that are non-proportional in strain
space, the formability curves shift with respect to the FLC of the
as-received sheet and may change shape. A non-proportional load
path is one in which the principal loading directions vary. For
example, Ghosh and Laukonis 3 prestrained cold-rolled
aluminum-killed steel to several levels of prestrain under equi-biaxial tension. They measured the forming limit of the prestrained
samples and found that the forming limit curve shifted downward
with increasing prestrain and also changed shape. Shifting and a
change in shape of the formability limit curve was also observed
for samples that were subjected to uniaxial prestrain. Graf and
Hosford 4 measured the FLCs of Al 2008-T4 samples subjected
to prestrain under uniaxial, biaxial, and plane strain loading. They
found the FLC of the prestrained sheet to be different from that
of the as-received FLC. It is pertinent to mention that the pre1
Corresponding author.
Contributed by the Materials Division of ASME for publication in the JOURNAL OF
ENGINEERING MATERIALS AND TECHNOLOGY. Manuscript received September 17, 2005;
final manuscript received August 9, 2006. Review conducted by Somnath Ghosh.

36 / Vol. 129, JANUARY 2007

strain imposed in both these reports were tensile. The reason for
this comment will be clear when the hydroforming of bent tubes is
considered in a later section.
In light of the above, the use of the FLC is restricted to cases
in which the sheet necks due to load paths that are linear in strain
space. If the load path deviates from linearity, necking along the
new path cannot be predicted without a knowledge of the FLC
due to the prestrain imposed during the linear portion of the load
path. An alternative approach to describe the formability limit was
advanced by Stoughton 5 who developed a stress-based formability limit curve-FLC. By assuming a stress-strain response
and an appropriate yield function, he transformed the FLC into
true principal stress 2 , 1 space. The resulting limit curve in
stress space is somewhat sensitive to the assumed stress-strain
curve and the assumed yield function. However, the most noteworthy demonstration in Stoughtons paper 5 is that when the
FLCs of the prestrained sheet are transformed into stress space
after accounting for the prestrains, the resulting forming limit
curve is nearly coincident with the stress-based limit curve for the
as-received sheet. In other words, to within the limits of experimental uncertainty and within the framework of the constitutive
assumptions, there exists a single curve in principal stress space
that represents the formability limit of the sheet. It has been argued that the various FLCs transformed into nearly coincidental
curves in stress space because of the insensitivity of the stressstrain relation at large strains. Stoughton 6 addressed this by
showing that when the FLC was shifted by an order of magnitude, the FLC shifted by 5 MPa. In passing, it should be noted
that a similar procedure to transform FLCs to stress space was
proposed by Embury and LeRoy 7. However, they did not demonstrate that the stress-based limit curve for the as-received and
the prestrained sheet were nearly the same.
Since the FLC is measured through plane stress loading, the
FLC represents the formability limit for plane stress load paths.

Copyright 2007 by ASME

Transactions of the ASME

Downloaded 12 Oct 2007 to 129.97.69.217. Redistribution subject to ASME license or copyright, see http://www.asme.org/terms/Terms_Use.cfm

Table 1 Experimental details hydroforming. SCF Square cross-section, corner fill. OCF
Outside corner fill. ICF inside corner fill. Dtube diameter. R center-line bend radius. L
tube length.
Experiment
Straight tube SCF
Pre-bent tube OCF
Pre-bent tube ICF
Pre-bent tube SCF
Pre-bent tube OCF

Material

Geometry

Hydroforming
lubricant

Tube-die friction
coefficient

EN-AW 5018
EN-AW 5018
EN-AW 5018
DP600
DP600

L 453 mm
R/D 2.5
R/D 2.5
R/D 2
R/D 2.5

AL070
Hydrodraw 711
Hydrodraw 711
Hydrodraw 625
Hydrodraw 625

0.05
0.15
0.15
0.03
0.03

The FLC appears to be particularly attractive to predict necking


in sheet metal with or without prestrains and non-proportional
load paths under plane stress conditions. This paper considers the
application of stress-based forming limit approaches to the case of
seam-welded tubes, which are fabricated from sheet, and where
the fabrication process stores plastic strains in the part. In the
stress-based framework, the FLC of the tube is obtained by
transforming FLC of the as-received sheet with the stress-strain
curve into principal stress space.
Necking and/or failure in the hydroforming of tubes can occur
at locations where in addition to the in-plane stresses a throughthickness compressive component acts. For a review of hydroforming technology see Ko and Altan 8. During hydroforming,
when the tube is initially expanding due to the internal pressure,
the stress state in the thin-walled tube is indeed approximately
plane stress. However, after the tube wall contacts the die, the
plane stress approximation is no longer valid. For example, when
a straight tube of circular cross section is being formed into a
square cross section, when the expanding tube comes into contact
with the die, a through-thickness compressive stress due to the
internal pressure acts in addition to the in-plane stresses. Necking
under these conditions, usually originates at the location where the
tube wall becomes a tangent to the die wall.
The through-thickness compressive component of stress requires an additional consideration, which is a potential increase in
the formability and its influence on the FLC and the FLC. Gotoh et al. 9 have presented an analytical expression that predicts
an increase in the plane strain forming limit in strain space due to
the presence of through-thickness compressive stresses. Smith
et al. 10, arguing that the model of Gotoh and co-workers cannot
be used for rate sensitive materials, developed an alternative expression that predicts an increase in formability due to a compressive 3. However, they assume that the formability curve in stress
space is not affected by a compressive 3, where 3 acts in the
through-thickness direction. This assumption is adopted in the
current work.
The foregoing are continuum approaches to predicting the onset
of necking. That is, when the variables such as stress and strain
reach critical levels, a neck is predicted. Alternatively, damagemechanics based approaches have also been applied to metal
forming 11. For hydroforming, in particular, the article by Cherouat et al. 12, uses a fully coupled damage mechanics model,
which is used in finite element computations, to predict failure
during hydroforming. In this approach, damage is treated as a
scalar variable, and used to model failure. The dissertation by
Baradari 13, from our group, uses the Gurson-TvergaardNeedleman 14,15 constitutive model to model failure of aluminum alloy tube during hydroforming. In this work, it is assumed
that void growth, nucleation and coalescence are responsible for
failure. In this approach, the model predictions of the void volume
fraction at failure compare favorably to metallographic observations. Though the results of these efforts using damage mechanics
based approaches are promising, they involve either fairly complex modeling schemes or an involved calibration effort to obtain
material damage parameters.
The main objective of the current work is to develop a conJournal of Engineering Materials and Technology

tinuum mechanics based approach to predict necking under threedimensional stress states and apply the method to predict necking
during tubular hydroforming. To this end, an extended stressbased FLC XSFLC is proposed. With suitable assumptions,
Stoughtons FLC is transformed into the XSFLC; the XSFLC
can predict necking under a three-dimensional state of stress. The
XSFLC is applied to two alloys: EN-AW 5018 aluminum alloy
and dual phase DP600 steel. Finite element simulations using the
explicit dynamic finite element code LS-DYNA, in conjunction
with the XSFLC, are used to predict necking during hydroforming
of straight and pre-bent tubes. For the case of pre-bent tube hydroforming, additional assumptions required to use the XSFLC
are presented. The computations are used to estimate the internal
hydroforming pressure at the onset of necking and these values
are compared with the experimental results. The paper concludes
with an appraisal of the stress-based method.

Hydroforming Experiments

This section presents a brief description of the hydroforming of


straight and pre-bent tubes and discusses the conditions under
which a neck initiates. Two materials are considered in this work.
The first is EN-AW 5018, an Al-Mg-Mn alloy, and the second is a
dual-phase steel, DP600. Table 1 summarizes details of the hydroforming experiments using these materials, including the tube material, geometry, and hydroforming lubricant. The table also presents values for the tube-die friction coefficients that were
determined from twist-compression tests. At least, two hydroforming trials were performed for each case. Figure 1 shows sectional views of the tubes in the hydroforming dies, and the die
cross sections. The EN-AW 5018 aluminum tubes had a wall
thickness of 2 mm, whereas the DP600 wall thickness was
1.85 mm. All tubes had an initial outside diameter of 76.2 mm.
The EN-AW 5018 tubes were annealed after tubing, and the
DP600 tubes were not.

Fig. 1 Cross section of dies used in the hydroforming


experiments

JANUARY 2007, Vol. 129 / 37

Downloaded 12 Oct 2007 to 129.97.69.217. Redistribution subject to ASME license or copyright, see http://www.asme.org/terms/Terms_Use.cfm

Table 2 Comparison of experimental and computational necking pressures and cross-section


geometry

Experiment
Straight tube SCF
Prebent tube OCF
Prebent tube ICF
Prebent tube SCF
Prebent tube OCF

Material
EN-AW 5018
EN-AW 5018
EN-AW 5018
DP600
DP600

Necking pressure
MPa
Experiment Computation
31.8
20.8
21.6
44. 5*
96.5

32.2
22.4
27.8
48.5
105.3

Cross-section Geometrya
mm
Experiment Computation
2Ro + s
2Ro + s
2Ro + s
2Ro + s

91.6
79
78.5
86.8

91.5
80.2
82.7
87.4

a
Ro-radius of tube before hydroforming. See Fig. 1 for definition of s. This cross section is at 45 in the bend angle.
necking pressure was not measured in this case; the reported value is 95% of the burst pressure.

In the straight tube corner fill experiments, EN-AW 5018 aluminum alloy tubes of 453 mm length were constrained in a die of
square cross section with rounded corners radius 3.17 mm. A
solid film lubricant, ALO70, was applied to the tubes. Pressurized
fluid was admitted into the tube through the end plugs. The end
plugs were loaded with sufficient force to provide a seal for the
pressurized fluid, and there was no end feeding applied to the tube
in these experiments. The fluid pressure and the forces on the end
seals were recorded, and the internal pressure was increased until
the tubes burst. Subsequently, experiments were interrupted to
prevent bursting and tubes with an incipient neck were recovered.
More details about the experiments can be found in the article by
Gholipour et al. 16. The values of the corner fill expansion and
the internal pressure at the onset of the neck are presented in Table
2.
EN-AW 5018 tubes were pre-bent and then hydroformed in the
inside corner fill and outside corner fill dies Fig. 1. To carry out
hydroforming of bent tubes, the tubes were bent in an instrumented rotary draw bender. Mandrels were used to prevent ovalization, and bend torque and forces on the tooling were monitored
during the bending process. After bending, the tubes were
trimmed to size for hydroforming. Bend geometries of the
EN-AW 5018 tubes are presented in Table 1. As in the case of
straight tube hydroforming, the process variables, which included
internal pressure, end plug forces, and expansion of the tube were
monitored. The burst pressure and the pressure to cause incipient
necks were determined for bent tube hydroforming. Necking pressures and the final cross-section geometry at necking at 45 deg
in the bend angle are presented in Tables 1 and 2, respectively.
More details of the bending and hydroforming can be found in the
articles by Dwyer et al. 17 and Oliveira et al. 18.
DP600 tubes were pre-bent and hydroformed in the die with a
square cross section Fig. 1. The methods employed were similar
to the bending of EN-AW 5018 tubes. Tables 1 and 2 present
details of the DP600 experiments, and more details about similar
experiments can be found in article 19.
Failure of the tubes during hydroforming showed a marked feature. Though the failure location depended on the tube geometry
and material, failure was found to occur at locations where the
expanding tube came into contact with the die. Figure 2 illustrates
the conditions at which a neck originates in the tube. As the tube

Fig. 2 Schematic of conditions under which a neck originates


in tubular hydroforming

38 / Vol. 129, JANUARY 2007

The

expands and fills the die, a greater portion of the tube comes into
contact with the die wall. A segment of the tube is shown in the
figure. One portion of the segment is in contact with the die wall,
which sets up a through-thickness compressive component of
stress, and the outer surface of this portion is consequently subjected to a tube-die friction force that retards the motion of this
portion. The compressive component acts in addition to the tensile
hoop stress and, possibly, an axial component of stress if end
feeding is applied. The portion of the segment not in contact with
the die is undergoing free expansion and is only acted on by a
tensile hoop stress and the internal fluid pressure. This portion is
approximately under plane stress loading and plastic deformation
is higher in this portion. Thinning and expansion in the plane
stress portion can be reduced by material flow from the threedimensional portion. However, when the magnitude of the friction
force on the three-dimensional portion is large enough, which occurs when a sufficient portion of the tube is in contact with the die
wall, material forming into the plane stress region is restricted.
Under these conditions, a neck and eventual failure originates at
the interface of the three dimensional portion and the plane stress
portion. Thus, the critical conditions for neck formation are characterized by a three-dimensional stress state and the friction force
due to the tube-die interaction. The computational modeling, described in a later section, highlights these two conditions. Since
the FLC and FLC describe the formability limit for plane stress
loading, a formability limit curve in three-dimensions is developed from the FLC under the assumption that stress state at the
onset of necking under in-plane plane stress loading are equivalent to those at the onset of necking under three-dimensional
loading.

3 Extended Stress-Based Forming Limit Curve (XSFLC)


The FLCs adopted for the EN-AW 5018 aluminum alloy and
the DP600 steel are shown in Fig. 3a. No conventional strainbased FLC was available for the aluminum; instead, the left side
of the curve was obtained from free-expansion experiments carried out on EN-AW 5018 tubes 20, whereas the right side of the
curve is approximated by that of AA5754 aluminum alloy sheet.
These two alloys are similar in composition, but differ somewhat
in the magnesium content and stress-strain curves, EN-AW 5018
having higher strength and a higher magnesium content 3.5%
versus 3.0%. The plane strain intercept in this approximate FLC
is equal to that obtained from free expansion experiments 20.
Since the tubes were annealed, the plastic strains produced during
tubing need not be accounted for. For the DP600 steel, the approximation due to Keeler-Brazier is used. This involves assuming the shape of the FLC as the one given by Keeler and Brazier
21, and determining the plane strain intercept from the hardening exponent, n, and sheet thickness, t. A value of n equal to 0.115
was obtained from the tensile stress-strain curve.
Figure 3 also shows the stress-strain response of the two alloys.
These were determined by using tensile test samples obtained
Transactions of the ASME

Downloaded 12 Oct 2007 to 129.97.69.217. Redistribution subject to ASME license or copyright, see http://www.asme.org/terms/Terms_Use.cfm

Fig. 3 a Strain-based FLCs for EN-AW 5018 aluminum alloy and DP600 steel,
b stress strain curves for the alloys, c stress-based FLCs for the alloys
obtained assuming isotropic hardening, and d XSFLCs for the two alloys. The
mean stress is assumed to be positive in tension.

from the tubes and conform to the ASTM E8M standards. Note
that the EN-AW 5018 and DP600 tensile-test samples fail at
20% elongation and that the rest of each curve is an extrapolation. These curves, which are relationships between the longitudinal stress and longitudinal strain, are equivalent to a functional
relationship between the equivalent stress, eq defined below,
and effective plastic strain, p =2 / 3ijij. Here, the Einstein
summation convention is used.
Two assumptions are introduced to transform the FLC into the
XSFLC. They are as follows:
ASSUMPTION 1. The alloys are assumed to be described by the J2
flow theory with isotropic hardening. Hardening is described by
the functional relationship y = y p, assuming proportionality
of the load path.
ASSUMPTION 2. The stress invariants, namely the equivalent
stress and mean stress, that characterize the formability limit under plane stress loading are representative of the formability limit
under three-dimensional stress states.
Using equations developed by Stoughton 5 and Assumption 1,
the FLC is transformed into the FLC and these curves are
shown in Fig. 3c. The FLC for DP600 lies above that of the
aluminum alloy since the dual phase steel has higher strength and
is more formable than the aluminum alloy. From Assumption 2,
the equivalent stress and mean stress in the neck are described
through the invariants, equivalent stress, eq, and mean stress,
hyd, given by
eq = 21 + 22 12

and

hyd =

1 + 2
3

where the i is the principal stress the FLC. The variables eq


and hyd are the variables that constitute the XSFLC.
The XSFLCs for the two alloys are shown in Fig. 3d. Mean
stress is assumed to be positive in tension. The left edge of the
XSFLC corresponds to the left edge of the FLC and is the formability limit under a uniaxial stress load path. The right edge corresponds to the formability limit under biaxial loading, and the dip
in the XSFLC corresponds to the plane strain limit.
Journal of Engineering Materials and Technology

To apply the XSFLC to predict the onset of necking in sheet


metal forming, the evolution of eq and hyd in the part being
formed are required. When the load path described by these variables, at any location in the part, intersects the XSFLC, the onset
of necking is predicted. Since the XSFLC is described in terms of
stress invariants, it can be used to predict the onset of necking in
parts being formed under plane stress loading as well as parts
formed under three-dimensional loading. Here, its utility in the
prediction of necking in tubular hydroforming is investigated.

Computational Details

4.1 Constitutive Model. The experiments described above


were modelled using the explicit dynamic finite element code LSDYNA. The computations were geared toward using the XSFLC
to predict the onset of necking in tubular hydroforming. To this
end, a user-subroutine was programmed to model the tube material. The tube material was modeled as an elastic-plastic material
using J2 flow theory with isotropic hardening. In the context of
small elastic strains, the increment of the strain tensor, d, is given
by the additive decomposition of the elastic and plastic strain
tensors as d = de + d p. Bold-faced letters are used to denote tensors and subscripts e and p denote elastic and plastic components,
respectively. The plastic strain is computed through an associated
flow rule as
d p = d p

where is the yield function, d p the increment of the effective


plastic strain, and the stress tensor. The yield function is
given by

= 3J2 y p =

3
sijsij y p = 0
2

where y p is the hardening curve obtained through the uniaxial tensile test and s is the deviatoric stress.
Algorithmic implementation of the stress update for the mateJANUARY 2007, Vol. 129 / 39

Downloaded 12 Oct 2007 to 129.97.69.217. Redistribution subject to ASME license or copyright, see http://www.asme.org/terms/Terms_Use.cfm

rial model utilized the radial return scheme of Wilkins 22. A


brief description of this scheme is provided in the Appendix; a
thorough treatment can be found in the book by Belytschko et al.
23. The XSFLC serves as an input to the subroutine. In addition
to implementing the stress update for the element, the subroutine
also tracks whether the load path described by the variables eq
and hyd has crossed the XSFLC. A formability variable for a
given hyd is defined as follows:

eq
XSFLC

where eq is the equivalent stress, and XSFLC is the equivalent


stress from the XSFLC, for the given hyd. This variable determines the proximity of the load path hyd , eq to the XSFLC.
When is unity the load path has intersected the XSFLC. An
alternative definition of as a binary variable = 0, safe, and
= 1, failed, is sometimes useful to highlight the potential failure
location in the mesh. Contour plots of can be used to determine
locations in the mesh where the element load paths have crossed
the XSFLC.
4.2 Material Parameters. For the EN-AW 5018 aluminum
alloy, the shear modulus and bulk modulus K were taken to be
26 GPa and 68 GPa, respectively. For the DP600 steel,
= 80 GPa and K = 164 GPa. The uniaxial tensile stress strain
curves presented in Fig. 3b were provided in tabular form to the
constitutive subroutine. In the standard ASTM uniaxial tensile test
for tube samples, the total strain at uniform elongation is less than
about 20%, after which the test results are not usable due to the
onset of necking instability. In the straight tube hydroforming
problem studied in this work, the effective plastic strain at the
onset of necking is about 20%; thus, the flow stress relationship
obtained from the ASTM uniaxial tensile test can be used to
model the flow stress of the tube.
The pre-bent tubes, however, present an additional challenge.
Since the effective plastic strain at the end of bending can reach
30%, an extrapolation of the uniaxial stress strain curve is
needed to model the tube material. Such an extrapolation will
introduce uncertainities into the flow stress versus effective plastic
strain curve. It is worthwhile to note that the increment of plastic
strain during the hydroforming of pre-bent tubes, for the materials
considered in this work, is about 1 2%.
To circumvent such an extrapolation, Koc et al. 24 proposed a
hydraulic bulge test to evaluate the flow stress of tubular materials
under biaxial strain conditions that approximate hydroforming.
They presented flow stress-strain data for tubes of SS304 stainless
steel, 6260-T4 aluminum alloy, and 1008 low carbon steel. However, Ko and Altan 8 point out that the stress strain curves
obtained from the bulge test agree with those obtained from the
ASTM uniaxial test for materials such as copper, brass, and aluminum, with the exception of titanium.
On the other hand, Levy et al. 25 used hydraulic burst tests on
AKDQ and HSLA steel tubes and showed that the tensile properties measured from uni-axial ASTM tests on tube samples could
be used to adequately model the tube flow stress. Indeed, they also
proposed a way to use the uniaxial stress strain curve obtained
from the sheet, after accounting for the tube-making strains, to
predict the flow stress of the tube.
Thus, there is no consensus in the literature as to how to characterize the flow stress of tubular material. Papers by Gholipour
et al. 16, Dwyer et al. 17, and Oliveira et al. 18 present
results of bending computations wherein the tubes were modeled
by using an extrapolation of the ASTM tensile test curves. The
computational results were in excellent agreement with the strains
effective strain during bending is 30% and loads measured
during the bending process. Thus, extrapolations of the uniaxial
stress-strain curves obtained from the ASTM tensile tests were
adopted in the current work due the absence of hydraulic bulge
data.
40 / Vol. 129, JANUARY 2007

4.3 Finite Element Mesh Details. All simulations were carried out with eight-noded solids elements with reduced integration. For the straight-tube hydroforming, due to symmetry, oneeighth of the tube was modeled with five elements through the
thickness and a total of 45,000 elements. To contrast these computations with those in which the through-thickness compressive
stress is zero, shell element computations were also carried out
using the shell element developed by Belytschko et al. 26 with
seven through-thickness integration points. The loading conditions
for bending and hydroforming that were measured in the experiments were used as inputs to the simulations. The bend tooling,
the mandrels, wiper die, and bend die, were modeled as rigid
bodies. A similar approach was used for the modeling of the hydroforming dies and tooling. The computational time and the
loading history were adjusted so that the computations were performed in a reasonable amount of time and also minimized inertial
effects. Further details of such bending and hydroforming computations can be found in the article by Bardelcik and Worswick
27.
For bent-tube hydroforming, one-half of the tube and tooling
were meshed in the bending and hydroforming computations. In
these models, the tube was modeled using eight-noded solid elements with five elements through the thickness and a total of
21,600 elements. These computations were performed in a sequence of steps: bending, springback, die-close, and hydroforming. At the end of the bending computations, springback calculations were performed. The level of hardening described by p in
the elements at the end of bending and the elemental stresses at
the end of springback were transferred with the mesh to the hydroforming computation.
4.4 Tube-Die Friction. Friction in tube hydroforming is a
complex phenomenon depending on the interacting materials, lubricant, the interfacial pressure, and sliding distance. A variety of
techniques exist to characterize the coefficient of friction between
the tube and die. In tubular hydroforming, three distinct friction
zones, namely, the guiding, transition and expansion zone have
been identified 28. However, in computational practice, a constant value of the coefficient of friction COF is often adopted in
the contact algorithm to model tube-die friction for all the zones.
Ko 29 used a combined experimental-numerical procedure
and reported COF values of 0.080.09 and 0.120.14 for two wet
lubes. In addition, values of 0.040.05 for a dry lube and 0.19
0.22 for a paste lube were reported 29 specifies the compositions of these lubricants. Ngaile et al. 30 used a limiting dome
height test to estimate an overall COF in the expansion zone for
four lubricants given as Lube A 0.125, Lube B 0.15, Lube C
0.15, and Lube D 0.075. In addition, they used the pear-shaped
test to estimate the COF in the expansion zone for the same lubricants as 0.074, 0.22, 0.20, and 0.10. Vollertsen and Plancak
31 have used a tube-upsetting test to estimate COF to be between 0.01 and 0.1, for three different lubricants.
The methods described above and the estimates of COF these
methods yield show that there is not a clear consensus as to how
this quantity can be characterized. In the current work, the twist
compression test 32 was used to estimate a COF between the
tube and the die; however, it is recognized that this value may not
be valid for all regions of the hydroforming die. The values for
tube-die friction presented in Table 1 are considered reasonable
when compared with the estimates from other methods. These
values were used in the penalty function-based contact algorithms
33 available in LS-DYNA to model tube-die friction.

5 Computational Results: Application of XSFLC to


Tubular Hydroforming
The application of the XSFLC to predict the onset of necking
during tubular hydroforming is presented here. A straight tube
corner fill expansion operation is examined first; this case represents a relatively monotonic strain path. Next, the XSFLC is apTransactions of the ASME

Downloaded 12 Oct 2007 to 129.97.69.217. Redistribution subject to ASME license or copyright, see http://www.asme.org/terms/Terms_Use.cfm

Fig. 4 Contour plots of straight tube hydroforming EN-AW


5018 aluminum using solid elements. Formability variable
Eq. 2 indicates whether the load path hyd , eq in the element has crossed the XSFLC. a Shows one-quarter of the
tube when the internal pressure is 32.2 MPa. Several elements
on the inside of the tube have crossed the XSFLC. b Close up
of the tube center where the locations under a threedimensional state of stress are labeled as 3D. At these locations, the load paths in all of the elements through the thickness of the tube have crossed the XSFLC = 1. The regions
designated as free expansion are under plane stress loading
approximately. The regions labeled die contact stay in contact with the die during the entire process.

plied to the hydroforming of pre-bent tubes for which the strain


path changes between the pre-bend and hydroforming operations.
5.1 Application
of
XSFLC
to
Straight
Tube
Hydroforming. Figure 4 shows results of the simulations of hydroforming of an EN-AW 5018 straight tube. The contour plots
are the results from solid element calculations and plot formability
variable as defined in Eq. 2. A value of unity indicates the
crossing of the XSFLC by the load path hyd , eq. Note that
these computations were carried out by modeling one-eighth of
the tube. The contour plots were reflected about the xz plane. The
locations designated as 3D in Fig. 4b are under a threedimensional state of stress, and the locations designated as die
contact remain in contact with the die from the start of hydroforming. The portions of the tube not in contact with the die are under
approximately plane stress loading and are designated as the free
expansion region.
In Fig. 4a, when the internal pressure is 32.2 MPa, the initially circular tube has expanded and the variable indicates that
many elements are at or above 95% of their formability in stress
space. At this state, many elements located on the inside surface
of the tube have crossed the XSFLC = 1. In particular, = 1 at
Journal of Engineering Materials and Technology

the locations on the inside of the tube designated as 3D. In Fig.


4b, when the internal pressure is 34.1 MPa, at the locations labeled as 3D, the load paths in all of the elements through the
thickness of the tube have crossed the XSFLC. This is the only
location in the mesh where this feature is observed; consequently,
these are the locations where the tube will eventually fail. Thus,
necking is predicted to originate at the locations labeled as 3D
when the internal pressure is 32.2 MPa. It must be emphasized
that the mesh is not designed to model an actual neck. The equivalent stress and mean stress in the load path are compared with the
values from the XSFLC curve. The XSFLC, in the present example, can be used to predict the internal pressure 32.2 MPa at
which a neck will start to form in the tube. Table 2 compares the
necking pressure and cross-section geometry obtained through the
computation with those measured in the experiment and it can be
seen that they are in excellent agreement. Furthermore qualitative
predictions of failure location can be obtained from the contour
plot in Fig. 4b; the location designated as 3D is where the tube
will burst. Since this is the only location where the load paths in
all of the elements through the thickness of the tube cross the
XSFLC.
Figures 5a and 5b present load paths from solid elements,
obtained through the explicit time integration scheme, located in
the 3D, free expansion, and die contact regions. These paths are
nonsmooth, which is a characteristic of the explicit dynamic time
integration scheme used in the finite element computation see the
Appendix. Load paths from elements in the free expansion and
die contact regions do not cross the XSFLC. The load path from
the element at location 3D, on the other hand, does cross the
XSFLC. Note that the path changes slope while crossing the XSFLC; this is the point at which the mesh comes into contact with
the die. Until this point, the stresses are predominantly tensile.
Once the tube contacts the die the through-thickness pressure acts
as a compressive component of stress and serves to reduce the
mean stress; consequently, the curve acquires a negative slope.
Figures 5c and 5d present load paths from solid elements,
obtained through the implicit time integration scheme, located at
the 3D, the free expansion and the die contact regions. In this
case, the built-in implicit solver within LS-DYNA was used.
These results show that there are negligible difference between the
results obtained using explicit integration and those obtained using implicit integration. The implicit results are smoother than
those obtained using explicit integration, since both the constitutive update and the time integration use implicit integration. In
addition, in this computation, the onset of necking is predicted at
an internal pressure of 32 MPa, which is identical to the value
predicted using explicit time integration.
To investigate the effect of neglecting the through-thickness
component of compressive stress, a computation wherein the tube
was meshed with plane stress shell elements was carried out. This
computation shows a rather different result when compared with
the one using the solid elements. Results of the computation using
plane stress shell elements can be seen in Fig. 6, where load paths,
hyd , eq, are plotted with respect to the XSFLC. This computation is similar to the solid element one with regard to loading
and boundary conditions. The only difference is in the type of
elements that constitute the mesh of the tube. It must be noted that
the shell elements in this mesh do not admit a three-dimensional
state of stress, and in this mesh, the locations labeled 3D are not
under a three-dimensional state of stress; the notation has been
retained to compare load paths from equivalent locations in the
shell element mesh with those from the solid element computation. It can be seen in Fig. 6 that the load path from element in the
free expansion region crosses the XSFLC, whereas, the load path
from the element in the 3D region does not. Since these shells are
plane stress elements, load paths in principal stress space 2 , 1
can be plotted with respect to the FLC Fig. 7. Again, the load
path from an element in the free expansion region crosses the
FLC, whereas the load path from the element in the 3D region
JANUARY 2007, Vol. 129 / 41

Downloaded 12 Oct 2007 to 129.97.69.217. Redistribution subject to ASME license or copyright, see http://www.asme.org/terms/Terms_Use.cfm

Fig. 5 Load paths from solid elements computations of straight tube hydroforming computations EN-AW 5018 aluminum. These results were obtained
using explicit time integration. Load path from an element at location 3D. This
load path crosses the XSFLC and while doing so the path changes slope point
when the mesh comes into contact with the die. c Load paths from elements
in the die contact and at the free expansion regions. These paths do not cross
the XSFLC. The load paths are from elements that are located in the middle
layer. b , d Results obtained using implicit time integration.

does not. That is, the shell element computations indicate that
failure will occur in the free expansion region. The feature of the
load path acquiring a negative slope when the mesh comes into
contact with the die is also not seen in the shell element results.
These results highlight the effect of the through-thickness stress
and show that incorrect predictions can be obtained if this component of stress is neglected. Furthermore, the internal pressure at
the onset of necking was predicted to be 27 MPa, which is much
lower than the experimentally observed value of 32 MPa. It is
worthwhile to note that the FLC, in conjunction with shell elements, can be used to predict necking in tubes undergoing free
expansion 34. In the case of the free expansion of tubes, since
there is no die, the plane stress approximation is reasonably valid
through the entire process, and shell elements can be utilized.
5.2 Application of XSFLC to Hydroforming of Pre-bent
Tubes. The simulations in which the XSFLC was applied to predict the onset of necking during the hydroforming of pre-bent
tubes indicated that additional assumptions were required to obtain realistic predictions. For example, in pre-bent tube OCF hydroforming, the method predicts the onset of necking at the inside
of the bend as well as at the outside of the bend. This prediction is
counter to the experimental observation in which the tube failed at
neither of these locations. The reason for the incorrect prediction
is that the XSFLC, as presented, does not predict the formability
of the material when the direction and state of loading change
between the bending and hydroforming operations.
During bending primary loading, the material at the outside of
the bend is plastically deformed under tension along the longitudinal axis of the tube. In the inside of the bend, though the plastic
deformation is also along the longitudinal axis, the plastic strain is
compressive. During hydroforming secondary loading, the loading is approximately perpendicular to the direction of the primary
loading and is tensile on the inside of the bend as well as on the
42 / Vol. 129, JANUARY 2007

Fig. 6 Load paths from Beytschko-Lin-Tsay plane stress shell


element computations of straight tube hydroforming EN-AW
5018 aluminum alloy. a Load path hyd , eq from element at
location 3D plotted with the XSFLC. b Load path hyd , eq
from an element in the free expansion region plotted with the
XSFLC. The load path in the free expansion region crosses the
XSFLC, whereas, the path in the 3D region does not.

Transactions of the ASME

Downloaded 12 Oct 2007 to 129.97.69.217. Redistribution subject to ASME license or copyright, see http://www.asme.org/terms/Terms_Use.cfm

Fig. 7 Load paths from Beytschko-Lin-Tsay plane stress shell


element computations of straight tube hydroforming EN-AW
5018 aluminum alloy. a Load path 2 , 1 from element at
location 3D plotted with the FLC. b Load path 2 , 1 from
an element in the free expansion region plotted with the FLC.
The load path in the free expansion region crosses the XSFLC,
whereas, the path in the 3D region does not.

outside. The load path change between bending and hydroforming


in the inside of the bend is designated as compression-tension, and
the path change in the outside of the bend is designated as
tension-tension.
It will be argued below that the XSFLC, when strictly applied
as presented in previous sections, cannot account for the formability limit during secondary loading. However, with some additional
assumptions the XSFLC can be applied successfully. Before introducing these assumptions, it is useful to define a variable, *, as

* =

sgnhyd p dt

where p is the rate of effective plastic strain, and sgn is the sign
function. It can be seen that if the material yields under a compressive state of stress * will be negative since hyd is negative
for compressive stress states. For yielding under tensile stress
states, * will be positive. From a computational standpoint, this
variable identifies whether or not the plastic strain in the element
accumulated due to yielding in compression or tension. It must be
emphasized that this is a nonphysical variable since effective
plastic strain cannot be negative and is only used to track the
stress state that is the source of the plastic deformation, especially
during bending. Depending on the sign of *, the tension-tension
load path can be distinguished from the compression-tension path.
5.2.1 Tension-Tension. The constitutive response of material
that has been subjected to tensile prestrains, of the magnitudes
encountered in the outside of the bend, and then unloaded can be
classified into two types. Zandrahimi et al. 35 present a succinct
discussion of the two. Figure 8a presents a schematic that shows
the two types of response, and these are compared with the response to monotonic single loading of the as-received material.
Journal of Engineering Materials and Technology

Fig. 8 Schematic illustrating Assumption 3. a Material response under monotonic single loading solid line. Dashed
lines illustrate Type 1 and Type 2 responses. These curves are
shifted by the prestrain values. b Effect of Assumption 3.
Load paths oa and ob have been drawn as straight lines for the
purpose of illustration, in reality they are not.

The secondary deformation is along an axis different from that of


the primary loading, as is the case during bending followed by
hydroforming. Upon secondary loading, in the Type 1 response,
the material displays a reduced yield stress but a higher hardening
rate. For a Type 2 response, on the other hand, the material shows
an increased yield stress but a lower hardening rate. The Graf and
Hosford 5 stress-strain data for as-received and prestrained
sheets of Al 2008-T4 display Type 1 response. By way of contrast,
Lloyd and Sang 36 present data for an aluminum AA3003-0
alloy that displays Type 2 response. Other representative examples
of Type 1 and Type 2 responses for steels and other alloys can be
found in the articles by Laukonis and Ghosh 37 and Zandrahimi
et al. 35. A marked feature of both these responses is that if the
secondary deformation imposes sufficient plastic strain, the response of the material tends to that obtained under single loading.
However, during the secondary loading, there exists a transient
regime when the plastic strains accrued during the secondary
loading are small, and when isotropic hardening fails to adequately model the constitutive response. Consequently, the formability during the secondary loading will not be correctly modeled
by isotropic hardening. In addition, anisotropic strength effects
may be playing a role since there is a change in loading between
bending and hydroforming. Therefore, under Assumption 1, the
XSFLC will predict a premature neck during hydroforming in the
region of the bend that has tensile prestrains. The introduction of
kinematic hardening and anisotropy of strength will be one way to
obtain a better formability estimate during the secondary loading.
It is assumed that during secondary loading the transient regime
is in operation and causes the material to have a higher formability than predicted by the XSFLC. For Type 1 response, since the
material displays a higher hardening rate, the formability is higher
than that derived using the monotonic stress-strain curve obtained
JANUARY 2007, Vol. 129 / 43

Downloaded 12 Oct 2007 to 129.97.69.217. Redistribution subject to ASME license or copyright, see http://www.asme.org/terms/Terms_Use.cfm

form single loading. Type 2 response, on the other hand, will have
a higher yield strength and consequently a higher formability. This
higher formability will operate during the transient regime and
will only hold when the increment of plastic strain imposed by the
secondary deformation process is small. The following assumption is made to account for the higher formability due to Type 1
and Type 2 material responses.
ASSUMPTION 3. The formability of a material element that has
an tensile effective tensile pre-strain, p, is assumed to be
maxy p , XSFLC.
Note that Assumption 3 is equivalent to assuming that onset of
necking, during secondary loading, is not possible until the material yields during the secondary loading. Yoshida et al. 38 have
presented data for an aluminum alloy that supports this assumption. This assumption is required because isotropic hardening, as
argued above, underpredicts the formability and a kinematic hardening model may better account for the increased formability during the transient regime. The incorporation of kinematic hardening
in the stress-based FLC framework will be addressed later. Assumption 3 is a catch-all and augments the XSFLC irrespective of
whether Type 1 or Type 2 response is in operation. Figure 8b
presents a scenario where the assumption plays a role. Path oa is
approximately uniaxial, as is the case during bending, and when it
crosses o the material yields and starts to harden. Load path oa is
terminated at a, the end of bending, and the hardening corresponds to an equivalent stress of oa. Note that the path terminates before the material necks or intersects the XSFLC, and the
material is unloaded. The material is then loaded along the planestrain path ob. This path roughly corresponds to the hydroforming
process. This load path first intersects the XSFLC, however, the
material element will not neck under Assumption 3, but necks
when it intersects its yield surface oa. At the end of the bending
computation, if * 0, and depending on the level of hardening,
Assumption 3 is invoked during the hydroforming computation.
5.2.2 Compression-Tension. The formability of material elements that underwent yielding due to compression, as in the inside
of the bend, is augmented according to the following assumption.
ASSUMPTION 4. The formability of a material element that has a
compressive effective plastic pre-strain, p, is assumed to be
maxy2 p , XSFLC.
At the beginning of the hydroforming computation, this assumption is applied to elements in which * 0. Unlike Assumption 3, there can be no appeal to experimental evidence as support
for Assumption 4. However, microstructural arguments can be invoked. For example, if the necking under tensile loading is governed by growth and nucleation of voids, compressive prestrain
will suppress these mechanisms and delay the onset of necking
under a secondary tensile loading. A consideration of the GursonTvergaard-Needleman 14,15 framework for void growth will
support this claim. A material element that underwent yielding
under compression during primary deformation will have stored in
it an effective plastic strain of magnitude p1 that will be equal to
*. When the secondary tensile deformation process results in an
increment of effective plastic strain of p1, the magnitude of *
will be zero and the total effective plastic strain will be 2 p1. The
factor 2 arises from the assumption, that the increment p1 under
tensile loading will annihilate the effects the compressive prestrain of p1.The factor 2 is, at this stage, purely arbitrary.
Assumptions 3 and 4 do not affect the results of the straight
tube computations. These assumptions were added to the userdefined subroutine and applied to the problem of bent-tube hydroforming. At the end of the bending and springback computations,
in addition to the effective plastic strain and the stress components, the elemental values of * are transferred with the mesh to
the hydroforming calculation. Depending on the sign of *, either
Assumption 3 or 4 is invoked at the beginning of the hydroforming computation.
44 / Vol. 129, JANUARY 2007

5.2.3 Results for Pre-Bent Hydroformed Tubes. Figure 9 presents the load paths in bending and hydroforming from the computations for the pre-bent OCF EN-AW 5018 aluminum alloy. The
columns correspond to element locations in the outside and inside
of the bend, and the failure location in the tube. In Fig. 9, the rows
a, b, and c are plots in principal strain, principal stress and
invariant space. The FLCs in each space are also shown. The
succeeding discussion treats each plot as an element of a nineelement matrix. The strain paths are continuous; however, the
stress paths are not. Partly due to the complexity of the forming
processes under consideration and partly due to the nature of the
explicit time integration, the stress-based paths are complex and
ragged. In the interest of clarity, salient inflection points have been
chosen and connected by straight lines. All hydroforming load
paths are terminated when the internal pressure has reached the
condition for the predicted onset of the neck. For the plots in
strain and stress space, since these are results of solid element
computations, the third component has been ignored. First, consider the plots in strain space. The largest amount of plastic strain
is accumulated during the bending process; whereas, the increment of plastic strain is small prior to failure during hydroforming. This relative magnitude of plastic strain emphasizes the major
impact of the pre-bend on the formability during hydroforming.
These plots suggest that the strain paths are linear during bending
and hydroforming. However, note that the slope of the path in
bending is different from that during hydroforming. As pointed
out in the Introduction, the path during hydroforming has to be
compared with the FLC corresponding to the level of prestrain
imposed during bending to predict the onset of necking during
hydroforming. As plotted in Fig. 9, the FLC of the as-received
sheet provides no useful information.
The FLCs plots in Fig. 9 provide no useful information either.
For example, in Fig. 9 Outside-b the bend path is roughly
uniaxial along the major principal stress axis and the path actually
crosses the FLC, indicating necking during bending, which is an
incorrect prediction. The hydroforming path in this figure suggests
uniaxial loading from the springback state S and intersects the
FLC, indicating failure. Again, this is not where the pre-bent
OCF aluminum alloy tube fails. Figure 8 Inside-b, on the other
hand, indicates that the loading is completely compressive, as expected, and failure in the inside of the bend is indicated incorrectly, when the hydroforming path intersects the FLC. Figure 9
Fail-b, also incorrectly predicts that necking will occur during
bending.
A more complex picture is revealed by the row of plots in
invariant space the XSFLC plots. The bending paths show complex changes in mean stress, as when the mesh comes into contact
with the mandrel labeled M. At this point, a through-thickness
compressive component is imposed on the tube, which decreases
the mean stress that is predominantly tensile. This leads to a
change in slope of the load path Fig. 9c. Consider the load
paths in the outside of the bend, Fig. 9 Outside-c. After two
slope changes due to contact with the mandrel, the highest hardening level equivalent stress of 324 MPa is attained during
bending. Subsequently, during hydroforming, the load path undergoes a slope change when the tube undergoes a draw-in due to the
motion of the end seals and proceeds to intersect the XSFLC.
However, the formability has been augmented in accordance with
Assumption 3 and is shown by the dashed horizontal line. Since
the load path during hydroforming does not intersect this upper
limit, a neck does not originate in this location.
The role of Assumption 4 can be seen in the load path from the
element in the inside of the bend Fig. 9 Inside-c. The bending
path results in a hardening of 319 MPa, which corresponds to
an effective plastic strain of 24%. In accordance with Assumption
4, the formability limit is set to an equivalent stress that corresponds to a plastic strain of 48%, which corresponds to an equivalent stress of 359 MPa and is indicated by the dashed horizontal
Transactions of the ASME

Downloaded 12 Oct 2007 to 129.97.69.217. Redistribution subject to ASME license or copyright, see http://www.asme.org/terms/Terms_Use.cfm

Fig. 9 Strain paths and stress paths for bending and hydroforming of pre-bent OCF EN-AW 5018 aluminum
alloy tube. Plots in each column correspond to element locations at the inside, outside, and the failure location,
respectively see mesh outline plot at bottom left. Plots in rows correspond to strain, stress and XSFLC space.
The linear paths shown are simplifications to those computed. For the purpose of discussion, each plot is
treated as an element in a matrix. Note that the equivalent stress axes in the XSFLC plots Outside-c and Fail-c
do not start at zero. Also note that the augmented formability, as per Assumptions 3 and 4, is shown as a
dashed horizontal line in plots Outside-c and Inside-c, respectively.

line in the figure. Though the load path during hydroforming


crosses the XSFLC, it does not intersect the dashed line and consequently, a neck does not originate in the element.
In the element at the failure location, Fig. 9 Fail-c, the bend
path results in a hardening level of 316 MPa as indicated by the
dashed horizontal line. During hydroforming, the load path
crosses this line and then undergoes a slope change due to contact
with the die and finally intersects the XSFLC leading to failure.
This feature of a slope change in the load path when the mesh
comes into contact with the die was also pointed out in the straight
tube computation.
The preceding discussion and the accompanying plots highlight
the complexity of the forming operations under discussion. Two
points deserve to be emphasized. First, plots in strain space are
misleading. If the third component of stress is neglected, then the
complexity of the forming process is hidden. Second, it can be be
seen that both Assumptions 3 and 4 introduce a stress path dependence into the stress-based framework for describing formability.
Table 2 presents quantitative results from the hydroforming
computations. The necking pressures and final cross section obtained in the experiments are compared with the experimental
values. In the context of the bent tube OCF experiments, AssumpJournal of Engineering Materials and Technology

tion 4 plays a role and prevents the load paths in the elements in
the inside of the bend from crossing the XSFLC, and it is Assumption 3 that controls the final failure locations. For all of the
bent tube OCF cases, the predicted necking pressures are only
10% higher than the experimentally obtained value and the final
cross section obtained in the computations are within 1% of the
experimentally measured final cross section. Therefore, this assumption appears to be reasonable. For the bent tube ICF predictions, Assumption 4 plays a key role, since the failure location is
in the region that yielded under compression during bending. The
predicted necking pressure is 28% higher than the experimental
value in the computation for the EN-AW 5018 bent tube ICF.
Although Assumption 4 played a satisfactory role in the bent tube
OCF case, it results in an over-estimate of the formability of the
inside of the tube in the bent tube ICF. This suggests that the
factor of two used in Assumption 4 is too high. When more data is
available, a better estimate of this factor may be possible.

Prediction of Failure Location Using the XSFLC

The proposed method also provides qualitative predictions of


the final failure location during tubular hydroforming. The locaJANUARY 2007, Vol. 129 / 45

Downloaded 12 Oct 2007 to 129.97.69.217. Redistribution subject to ASME license or copyright, see http://www.asme.org/terms/Terms_Use.cfm

Fig. 10 Comparison of contour plots predicting failure locations left and photographs right from the experiments. The
white arrows indicate the failure location. The gray arrows indicate additional locations in the DP600 tubes where the XSFLC approach indicates failure. Though the variable = 1 in the
location indicated by the black arrow DP600 SCF, all the elements throught the thickness of the mesh have not crossed the
XSFLC. The binary definition of the formability variable, , is
used to plot these contours.

tion in the mesh where all of the elements through the thickness of
the mesh cross the XSFLC is identified as the location where the
tube will burst. Contour plots of from the hydroforming computations are shown in Fig. 10 along with the photographs of
tubes tested to failure. In these contours, is treated as a binary
variable for clarity, and it can be seen that the predicted failure
location are in excellent agreement with those seen in the experiments. Note that for the two DP600 tubes, the computations indicate additional failure locations as shown by the gray arrows.
Several approximations have been made to implement the
XSFLCthe stress-strain curves, the FLC, and Assumptions 3
and 4. These are all potential sources of the incorrect prediction. It
is not possible, though, to pinpoint the cause of the incorrect prediction at this stage of the research.

Summary

This paper presents an extended stress-based FLC that allows


the prediction of the onset of necking in sheet metal under threedimensional states of stress. A conventional plane stress FLC
was converted to a plane stress FLC, using the method proposed
by Stoughton 5, which was transformed into the XSFLC. The
key assumption behind the XSFLC is that the triaxial stress states
generated in the neck under plane stress loading can be described
by the invariants hyd and eq and can be used to describe the
stress states in necks that originate under three-dimensional states
of stress. Necking was shown to occur in tubular hydroforming
under three-dimensional stress states. Consequently, solid element
computations were used in conjunction with the XSFLC to predict
necking pressures and tube expansion. From the good agreement
between the computational predictions and the experiments for the
straight tube hydroforming and some of the bent tube cases, it can
be concluded that Assumption 2 is valid. Inasmuch as the quali46 / Vol. 129, JANUARY 2007

tative and quantitative predictions of the approach advanced in


our paper provide support for this assumption, independent validation of this assumption is warranted. A forthcoming publication
will present an approach that can be used to validate this assumption. Furthermore, computations carried out using plane stress
shell elements, in which the through-thickness stress is zero, were
shown to lead to erroneous predictions.
Isotropic hardening Assumption 1 appears to be sufficient to
describe the material response to predict the onset of necking
using the XSFLC. However, as already mentioned, in the case of
bent-tube hydroforming, the isotropic hardening assumption was
not sufficient to obtain reasonable predictions of the experimental
results. It was argued that because of the presence of prestrains at
the onset of hydroforming and a change in the loading direction
between bending and hydroforming, the isotropic hardening assumption under predicted the formability during the secondary
deformation due to hydroforming. Assumptions 3 and 4 were introduced to circumvent complications that would arise with the
physically more realistic kinematic hardening. However, these assumptions introduce a stress path dependence in the XSFLC
framework. The predictions for the case of bent-tube OCF hydroforming EN-AW 5016 and DP600 tubes are not in as good
agreement with the experimental results as in the case of straight
tube hydroforming. A potential improvement to the approach presented in this work is to incorporate kinematic hardening and
anisotropy of strength. Though it is not clear as to how kinematic
hardening can be incorporated into the stress-based formability
assessments. Consequently, the approach adopted in this work has
been to make the simplest set of assumptions to develop a tool
with predictive capability.

Acknowledgment
The authors wish to acknowledge the following agencies and
companies for their support: Auto21 Network Centres of Excellence, the Ontario R&D Challenge fund, General Motors of
Canada, Dofasco, Stelco, Nova Tube, DA Stuart, Eagle Precision
Technologies, Natural Sciences and Engineering Research Council of Canada, and the Canada Research Chair Directorate.

Appendix
In a finite element code that uses an explicit time integration
scheme, for stability, the time step is governed by the CourantFriedrichs-Lewy criterion 23, which is given by
t min

le
Ce

where le is the characteristic length of element e, and Ce is the


wave speed in the element. At the end of the computation for time
tn, for each element, a time step is computed and the least value
of these time steps is used as the time step for the entire domain.
The stress update algorithm is identical to the the built-in material model designated as MAT 24 in LS-DYNA 33. The userdeveloped subroutine that was used for the computations, in addition to carrying out the constitutive update, checks whether the
load path described by the equivalent stress and mean stress has
crossed the XSFLC. At the start of the time step tn+1, the rate of
the increment of the total strain tensor, n+1 is known. The task
of the constitutive subroutine is to update the stress tensor, n,
from time tn to the tensor n+1 at tn+1. The stress tensor n
has been corrected for the rotation of the element between tn and
tn+1 before entry into the subroutine 33. The Lam constants
and are used for the elastic update. Index notation for tensors
and the summation convention is used below.
1. Compute the trial elastic stresses through the co-rotational
Jaumann rate.
Transactions of the ASME

Downloaded 12 Oct 2007 to 129.97.69.217. Redistribution subject to ASME license or copyright, see http://www.asme.org/terms/Terms_Use.cfm

ijT = ijn + 2 ij + kkt


13

where T denotes trial state.


2. Compute the trial deviators, mean stress, and equivalent
stress.

14

1
pT = Tkk
3

15

1
sijT = ijT Tkkij,
3

and

Teq =

3 T T
s s
2 ij ij

3. Compute and check for yielding.


T

=
T

Teq

16

ypn

if 0 then material is elastic go to step 6, other wise


compute plastic strain increment.
4. Compute plastic strain increment.
d p =

Teq ypn
dy n
3 +
d p

17

18

19

where dy / d p is the slope of the hardening curve.


5. Scale the deviators back to the yield surface.
n
T y p

sijn+1 = sij

21

Teq

6. Update stresses, mean stress, and effective plastic strain.

ijn+1

sijn+1

+p

23
24

and
pn+1 = pn + d p
7. Check whether the stress state p
XSFLC curve.

22

pn+1 = pT

n+1

20

n+1
, eq

25

has crossed the

Note that the stress state at the end of the update, n+1, is on
n
the yield surface whose radius in deviatoric space is y p . That
is, the yield surface corresponding to the hardening at tn, and not
the yield surface at tn+1. This feature and inertial effects in the
time explicit integration could lead to some non-smoothness in the
stress computations.

References
1 Keeler, S. P., and Backofen, W. A., 1963, Plastic Instability and Fracture in
Sheets Stretched Over Rigid Punches, ASM Trans. Q., 56, pp. 2548.
2 Goodwin, G. M., 1968, Application of Strain Analysis to Sheet Metal Forming in the Press Shop, SAE Paper, No. 680093.
3 Ghosh, A. K., and Laukonis, J. V., 1976, The Influence of Strain Path
Changes on the Formability of Sheet Steel, 9th Biennial Congress of The
International Deep Drawing Research Group, Sheet Metal Forming and Energy Conservation, ASM, Metals Park, OH.
4 Graf, A., and Hosford, W., 1993, Effect of Changing Strain Paths on Forming
Limit Diagrams of Al 2008-T4, Metall. Trans. A, 24A, pp. 25032512.
5 Stoughton, T. B., 1999, A General Forming Limit Criterion for Sheet Metal
Forming, Int. J. Mech. Sci., 42, pp. 127.
6 Stoughton, T. B., 2001, Stress-Based Forming Limits in Sheet-Metal Forming, ASME J. Eng. Mater. Technol., 1234, pp. 417422.
7 Embury, J. D., and LeRoy, G. H., 1977, Failure Maps Applied to Metal
Deformation Processes, Advances in Research on the Strength and Fracture
of Materials, Pergamon Press, London, pp. 1542.
8 Ko, M., and Altan, T., 2001, An Overall Review of the Tube Hydroforming
THF Technology, J. Mater. Process. Technol., 1083, pp. 384393.
9 Gotoh, M., Chung, C., and Iwata, N., 1995, Effect of Out-of-Plane Stress on
the Forming Limit Strain of Sheet Metals, JSME Int. J., Ser. A, 381, pp.
123132.
10 Smith, L. M., Averill, R. C., Lucas, J. P., Stoughton, T. B., and Matin, P. H.,
2003, Influence of Transverse Normal Stress on Sheet Metal Formability,
Int. J. Plast., 1910, p. 1567-1583.
11 Saanouni, K., Nesnas, K., and Hammi, Y., 2000, Damage Modeling in Metal
Forming Processes, Int. J. Damage Mech., 9, pp. 196240.
12 Cherouat, A., Saanouni, K., and Hammi, Y., 2002, Numerical Improvement

Journal of Engineering Materials and Technology

26
27

28
29
30

31
32
33
34
35
36
37
38

of Thin Tubes Hydroforming With Respect to Ductile Damage, Int. J. Mech.


Sci., 4412, pp. 24272446.
Baradari, G. J., 2006, Damage in Hydroforming of Pre-Bent Aluminum Alloy
Tubes, Ph.D. thesis, Department of Mechanical Engineering, University of
Waterloo.
Gurson, A. L., 1977, Continuum Theory and Ductile Rupture by Void Nucleation and Growth: Part I Yield Criteria and Flow Rules for Porous Ductile
Media, ASME J. Eng. Mater. Technol., 991, pp. 215.
Tvergaard, V., and Neeldeman, A., 1984, Analysis of the Cup-Cone Fracture
in a Round Tensile Bar, Acta Metall., 32, pp. 157169.
Gholipour, J., Worswick, M. J., and Oliveira, D., 2004, Application of Damage Models in Bending and HydroForming of Aluminum Alloy Tube, SAE
2004 World Congress, SAE Paper No. 2004-01-0835.
Dwyer, N., Worswick, M. J., Gholipour, J., Xia, C., and Khodayari, G., 2002,
Pre-Bending and Subsequent Hydroforming of Tube: Simulation and Experiment, NUMIFORM 2002, Proceedings of the 5th International Conference
and Workshop on Numerical Simulation of 3D Sheet Forming Processes, Jaejon Korea, pp. 447452.
Oliveira, D. A., Worswick, M. J., and Grantab, R., 2005, Effect of Lubricant
in Mandrel-Rotary Draw Tube Bending of Steel and Aluminum, Can. Metall.
Q., 44, pp. 7178.
Dyment, J., Worswick, M., Normani, F., Oliveira, D., and Khodayari, G., 2003,
Effect of Endfeed on Strains and Thickness During Bending and on the Subsequent Hydroformability of Steel Tubes, Proceedings IBEC2003, SAE Paper
No. 2003-01-2837.
Khodayari, G., 2001, Final Report USAMP Hydroforming of Aluminum
Tubes, Technical Report, Industrial Research and Development Institute.
Keeler, S. P., and Brazier, W. G., 1977, Relationship Between Laboratory
Material Characterization and Press Shop Formability, Proc. of Microalloy
75, Union Carbide, New York, pp. 447452.
Wilkins, M. L., 1964, Calculation of Elastic-Plastic Flow, Methods in Computational Physics, Vol. 3, Fundamental Methods in Hydrodynamics, Academic Press, New York.
Belytschko, T., Liu, W. K., and Moran, B., 2000, Nonlinear Finite Elements
for Continua and Structures, Wiley, New York.
Koc, M., Aue-u-lan, Y., and Altan, T., 2001, On the Characteristics of Tubular
Materials for Hydroforming: Experimentation and Analysis, Int. J. Mach.
Tools Manuf., 415, pp. 761772.
Levy, B. S., Tyne, C. J. V., and Stringfield, J. M., 2004, Characterizing Steel
Tube for Hydroforming Applications, J. Mater. Process. Technol., 1503, pp.
280289.
Belytschko, T., Lin, J., and Tsay, C. S., 1984, Explicit Algorithms for the
Nonlinear Dynamics of Shells, Comput. Methods Appl. Mech. Eng., 42, pp.
225251.
Bardelcik, A., and Worswick, M. J., 2005, Evaluation of Load Control EndFeed in Hydroforming of Pre-Bent DP600 Steel Tube Using the Extended
Stress-Based Forming Limit Curve XSFLC Failure Criterion, Numerical
Methods in Continuum Mechanics 2005.
Prier, M., and Schmoeckel, D., 1999, Tribology of Internal High Pressure
Forming, Proceedings of Internatioinal Conference on Hydroforming, Stuttgart, Germany, Stuttgart, Germany.
Ko, M., 2003, Tribological Issues in the Tube Hydroforming ProcessSelection of a Lubricant for Robust Process Conditions for an Automotive
Structural Frame Part, ASME J. Manuf. Sci. Eng., 1253, pp. 48492.
Ngaile, G., Jaeger, S., and Altan, T., 2004, Lubrication in Tube Hydroforming
THF: Part II. Performance Evaluation of Lubricants Using LDH Test and
Pear-Shaped Tube Expansion Test, J. Mater. Process. Technol., 1461, pp.
116123.
Vollertsen, F., and Plancak, M., 2002, On Possibilities for the Determination
of the Coefficient of Friction in Hydroforming of Tubes, J. Mater. Process.
Technol., 125126, pp. 412420.
Schey, J. A., and Nautiyal, C., 1990, Effects of Surface Roughness on Friction and Metal Transfer in Lubricated Sliding of Aluminum Alloys Against
Steel Surfaces, Wear, 146, pp. 3751.
Hallquist, J. O., 1998, LS-DYNA Theoretical Manual, Livermore Software
Technology Corp., Livermore, CA
Green, D. E., and Stoughton, T. B., 2004, Evaluating Hydroforming Severity
Using Stress-Based Forming Limits, Second Annual Hydroforming Conference.
Zandrahimi, M., Platias, S., Price, D., Barrett, D., Bate, P., Roberts, W., and
Wilson, D., 1989, Effects of Changes in Strain Path on Work Hardening in
Cubic Metals, Metall. Trans. A, 20A11, pp. 24712482.
Lloyd, D. J., and Sang, H., 1979, Influence of Strain Path on Subsequent
Mechanical PropertiesOrthogonal Tensile Paths, Metall. Trans. A,
10A11, pp. 17671772.
Laukonis, J. V., and Ghosh, A. K., 1978, Effect of Strain Path Changes on the
Formability of Sheet Metals, Metall. Trans. A, 9A12, pp. 18491856.
Yoshida, K., Kuwabara, T., and Kuroda, M., 2005, Forming Limit Stresses of
Sheet Metal Under Proportional and Combined Loadings, NUMISHEET
2005: Proceedings of the 6th International Conference and Workshop on Numerical Simulation of 3D Sheet Metal Forming Process, Smith, L. M., Pourboghrat, F., Yoon, J.-W., and Stoughton, T. B., eds., AIP, New York, Vol. 778.

JANUARY 2007, Vol. 129 / 47

Downloaded 12 Oct 2007 to 129.97.69.217. Redistribution subject to ASME license or copyright, see http://www.asme.org/terms/Terms_Use.cfm

Vous aimerez peut-être aussi