You are on page 1of 4

Combined Mechanics-Materials Based Optimization of Superplastic Forming of

Magnesium AZ31 Alloy


M.K. Khraisheh1, F.K. Abu-Farha1, M.A. Nazzal1, K.J. Weinmann2 (1)
1
Center for Manufacturing, Department of Mechanical Engineering
University of Kentucky, Lexington, KY, USA
2
Department of Mechanical Engineering, University of California, Berkeley, CA, USA
Abstract
A new optimization approach for superplastic forming of Mg AZ31 alloy is presented and experimentally
validated. The proposed new optimization approach is based on a multiscale failure criterion that takes into
account both geometrical necking and microstructural evolution, yielding a variable strain rate forming path
instead of the commonly used constant strain rate approach. Uniaxial tensile tests and free bulge forming
experiments, in conjunction with finite element analysis, are used to evaluate the proposed optimization
approach. Significant reduction in forming time is achieved when following the proposed optimization
approach, without compromising the uniformity of deformation.
Keywords:
Sheet Metal, Magnesium, Superplastic Forming

1 INTRODUCTION
The increasing demand for lighter and therefore more
fuel-efficient vehicles has been the prime motivation for
the growing interest in magnesium and its alloys. Yet this
lightest constructional metal on earth exhibits inferior
tensile ductility at low temperatures, mainly because of its
hexagonal close-packed crystal structure. Warm forming
of magnesium between 220 and 350 C has been seen as
a relief for the problem by many investigators [1-3].
Despite these and other studies, more investigations and
data on the formability of magnesium alloys under wide
range of forming conditions are needed to advance the
utilization of magnesium alloys.
Superplastic Forming (SPF) is an innovative process that
stretches the boundaries of ductility in conventional
forming operations, offering a great potential for
successful sheet metal forming of hard-to-form materials
such as magnesium alloys. Though SPF has been
successfully used to form titanium and aluminum parts for
aerospace applications, its widespread industrial use is
still limited. One of the most critical issues that currently
hamper the widespread use of SPF is the low production
rate, due to the low deformation rates and limited
predictive capabilities of deformation and failure during
forming. Blow forming is considered the most common
practice employed in forming superplastic materials,
where the sheet is formed onto the die using pressurized
gas. The selection of the forming pressure profile is very
critical, as it ultimately determines the integrity of the
formed part and the production time. Forming pressure
profiles used in the industry are usually based on trialand-error practices, and employ low safe pressure to
prevent premature failure. The challenge is to develop
optimum forming pressure profiles that can reduce the
forming time and maintain the integrity of the formed part.
The main scheme usually employed to optimize the
process is by maintaining the strain rate during forming
within the optimum range, where maximum ductility is
achieved. This may prevent thinning and premature
failure, but would prolong the forming time since the
optimum strain rate is usually low. Some investigators
have reported recently that using variable strain rate
schemes may reduce the forming time, and yet maintain

Annals of the CIRP Vol. 55/1/2006

the integrity of the formed component [4,5]. Others


reported increased thickness strain at failure by applying
pulsating strain rate scheme [6]. Generally, these studies
are based on limited experimental observations and/or
simple models that cannot be generalized to optimize the
SPF of the various superplastic materials. To develop
accurate optimum forming paths, deformation behavior of
superplastic materials must be accurately described, and
a failure criterion that takes different failure mechanisms
into account must be used.
Most recently, a new multiscale failure criterion for
superplastic deformation, that takes both geometrical and
microstructural features into account, has been developed
[7]. In this work, the stability criterion is extended to
develop optimum forming loading paths for the AZ31
magnesium alloy. First, a constitutive model that accounts
for grain growth and cavitation is developed and
calibrated to capture the superplastic behavior of the
alloy. The model is then combined with the stability
criterion to generate optimum variable strain rate loading
paths. Uniaxial tensile tests and free bulge forming
experiments
were conducted to evaluate the
effectiveness of the proposed optimization technique. For
the bulge forming experiments, the multiscale stability
criterion is modified to account for the biaxial state of
stress, and the forming pressure profiles were generated
using finite element (FE) analysis.
2

MODEL DEVELOPMENT

2.1 Constitutive Model


Based on the continuum theory of viscoplasticity,
superplastic deformation is modeled using a constitutive
relation that accounts for microstructural features,
including grain growth and cavitation. A simplified form of
the model is given by [8,9]:
1

k m
& = p

d (1 fa )

(1)

2.2 Grain Growth and Cavitation Models


In order to account for the change in microstructure
during deformation, evolution equations for the average
grain size and the area fraction of voids are used. A
simple linear grain growth model similar to the one used
by Wilkinson et al [10] is used here:
d = d + C

Similar approach is used to calibrate the grain growth


equation (figure not shown here).

= 1.6
20

fa = fa exp( )

(3)

Experiment
Model (Eqn 3)
0
0

(4)

where d A& is the variation in the area increment rate and


dA is the variation in cross-section area. Assuming that
the stress is a function of strain and strain rate only, Hart
derived a stability criterion for a uniaxial loading case.
Incorporating the modified constitutive equation along
with the microstructural evolution equations into the
framework of Harts analysis, a new stability criterion
accounting for both geometrical instabilities and
microstructural aspects is developed [9]:

&

&

f &

+ a

&
d ,fa

d ,

d& &
&
2

=1

,fa

(5)

where is a parameter defined in terms of the bixiality


ratio ( = 1 for uniaxial loading). The first term represents
strain rate sensitivity, the second represents the effect of
cavitation, and the last represents strain hardening due to
grain coarsening.

2.4 Model Calibration


Recently, the deformation of the AZ31 magnesium alloy
was studied in detail under simple tension, within a wide
range of temperatures and strain rates [13]. The optimum
superplastic forming temperature was found to be within
400-450 C; 400 C is used in this work to calibrate the
proposed optimization model. Grain growth and cavitation
models were calibrated by microstructural examinations of
the deformed samples of interrupted tests and the
parameters in equations (2) and (3) were determined by
fitting the models to the experimental data. Figure 1
shows the evolution of cavitation during deformation. A

25

1.5

True Strain

10-3 s-1

Experiment
Model (Eqn 1)

5x10-4 s-1

20

True Stress (MPa)

(dA& / dA)P 0

0.5

Figure 1: Area fraction of voids vs. plastic strain.


The strain rate sensitivity index m was evaluated at
various strain rates using strain rate jump tests [13]. The
remaining parameters were obtained as functions of
strain and strain rate by fitting the model to the
experimental stress-strain curves as shown in Figure 2.
The figure shows that the calibrated model provides an
excellent fit to the experimental data at various strain
rates.

where fa0 is the initial area fraction of voids, and is a


void growth material parameter that depends on the strain
rate sensitivity index m, and the ratio between the mean
stress m and the effective stress .
2.3 Stability Criterion
The amount of stable and uniform deformation is limited
by the onset of localized necking and cavitation. The
condition for stable deformation as defined by Hart [12] is
given by:

= 1.15

10

(2)

where is the effective strain, d0 is the initial average


grain size, and C is a material parameter.
Cavitation model used here accounts for the growth of
pre-existing cavities; nucleation of new voids is not
considered. Because of the large plastic deformation
associated with SPF, void growth mechanism is believed
to be plasticity-controlled [11] where the area fraction of
voids is exponentially related to the effective plastic strain:

= 1.75

30

fa (%)

where & is the effective strain rate, is the effective


flow stress, m is the strain rate sensitivity index, d is the
average grain diameter, fa is the area fraction of voids, p
is the grain growth exponent, and k is a material
parameter.

2x10-4 s-1
15

1x10-4 s-1

5x10-5 s-1

10
5
0
0

0.25

0.5

0.75

True Strain

1.25

Figure 2: Stress strain curves, model vs. experiment.

2.5 Optimum Forming Path


After calibrating the model, and for a given strain rate,
equation (5), along with equations (1) through (3) were
solved numerically to yield the limiting strain at which the
onset of unstable deformation is expected to occur for
that specific strain rate. This process is repeated for other
strain rates and the resulting optimum forming path is
generated as shown in Figure 3. It is important to note
that the optimum variable strain rate profile shown in
Figure 3 is not arbitrary; it is based on actual deformation
behavior accounting for both geometrical necking and
microstructural evolution including grain growth and
cavitation. Initially, the material can be deformed at a fast
rate, and as deformation continues, strain rate is shifted
to a lower value to prevent premature failure from necking
and cavitation. The novelty of this optimization approach
is that it considers both failure modes (more details can
be found in [9]).
3

RESULTS AND DISCUSSION

The proposed optimization approach was tested and


evaluated through uniaxial tests and free bulge forming
experiments. Samples formed under constant strain rates
(fast and slow) are compared with samples formed
according to the optimum variable strain rate profile.

1.E-02

2x10-3 s-1

Stepped

True Strain Rate (1/s)

Continuous
(Eqn 5)

1.E-03

5x10-5 s-1

Optimum

1.E-04

0.1

True Strain

10

Figure 3: Optimum forming path for uniaxial loading case.

3.1 Uniaxial Tensile Tests


Uniaxial tensile samples were cut from a 3.22 mm thick
AZ31 commercial magnesium sheet with a 19x6.35 mm
gauge section. A 5582 INSTRON load frame equipped
with a heating chamber, enabling testing at temperatures
up to 610 C, was used. The testing machine cannot
perform a variable strain rate test following the smooth
continuous optimum curve shown in Figure 3. Therefore,
the continuous curve is approximated by multi-steps of
-3 -1
constant strain rates starting at 3x10 s , and ending at
-5 -1
5x10 s as shown in Figure 3. Following the variable
strain rate optimum profile, samples were stretched at
400 C to strain values of 250%, 300% and 350%
elongation. Additional samples were also stretched to the
same strain values at selected constant true strain rates
-3 -1
-5 -1
of 2x10 s (fast forming) and 5x10 s (slow forming)
for comparison. The results of these tests are
summarized in Table 1.

Strain
Rate
(s-1)

Thinning Factor ( tmin / tave )


Forming Time [min]
250%
Elong.

300%
Elong.

350%
Elong.

(0.339)
[10.44]

(Fail)

(Fail)

5x10

(0.856)
[417.6]

(0.805)
[462.1]

(0.751)
[501.4]

Optimum

(0.738)
[84.2]

(0.687)
[125]

(0.624)
[160.8]

-3

2x10

-5

Table 1: Thinning factor and forming time in simple


tension, for the different strains and strain rates.
For the sample that was stretched to 250% elongation at
2x10-3 s-1 constant strain rate, forming time was 10.44
minutes and the thinning factor (the ratio between the
minimum and average thickness) was 0.339, indicating a
highly localized necking (see Figure 4). To improve
deformation uniformity, lower strain rates must be used.
-5 -1
For a constant strain rate of 5x10 s , the uniformity is
significantly improved to a 0.856 thinning factor; however,
forming time was increased drastically to 417.6 minutes.
This trade-off presents the challenge we face. The
sample that was formed to 250% elongation according to
the proposed optimum profile clearly shows the
effectiveness of the optimization scheme. Forming time
was reduced from 417.6 minutes to only 84.2 minutes
while maintaining almost the same deformation
uniformity, embodied by the 0.738 thinning factor.
Samples deformed to 250% are all shown in Figure 4.

Figure 4: Samples strained to 250% elongation.


Similar results are achieved for the samples stretched to
-3 300% and 350% elongation. Samples strained at 2x10 s
1
could not even reach these limits, as they failed at about
250% due to severe localized necking. Samples formed
according to the optimum profiles exhibit uniform
deformation with significant reduction in forming time, as
shown in the table. This important result is also illustrated
in Figure 5, where thickness distributions of the samples
tested at different forming schemes are compared. Not
only the forming time is significantly reduced, but the
uniformity of deformation is also maintained. These
represent clear experimental evidence supporting the
proposed multiscale-based optimization approach.
0.75
0.6
0.45

t / t0

1.E-05
0.01

2E-3
5E-5
Opt

0.3

0.15
0
-0.15
0

20

40

L/L0 (%)

60

80

100

Figure 5: Thickness distribution along tensile samples


strained to 250% elongation.

3.2 Superplastic Free Bulge Forming


Bulge forming of 1 mm thick AZ31 sheets at 400 C were
carried out onto a 62.5 mm open circular die using
pressurized Argon. This geometry was chosen because
the state of stress at the pole of the dome is known to be
balanced biaxial. Therefore, the constitutive model, the
evolution equations, and the stability criterion were
extended to account for biaxial stress state by using the
effective values of stress, strain and strain rate [9,14].
The model and the evolution equations were then
implemented into the commercial FE code ABAQUS
through
user-defined
subroutines.
Different
FE
simulations were conducted to generate forming pressure
profiles: constant strain rate at the pole and variable strain
rate according to the optimum profile determined from the
-3 -1
stability criterion. Constant strain rates of 1x10 s (fast)
and 5x10-5 s-1 (slow) were used to evaluate the results of
the optimum profile. In order to have a meaningful
comparison, different samples were formed to the same
average circumferential strain (an arc line passing
through the pole of the dome) using the different pressure
profiles generated from the FE analysis. The target
circumferential strain was set at 0.524, the maximum
deformation achieved for the fast strain rate. For each
test, the achieved dome height, forming time, and the
thickness distribution were measured. The results of
these tests are summarized in Table 2.

-5

5x10 s

-1

Strain Rate (s )

Parameter

1x10-3

5x10-5

Opt.

Dome Height (mm)

35.3

34.0

34.2

Forming Time (min)

17.3

500

67.5

Parts Status

Fail

Ok

Ok

Arc Length (mm)

105.5

106

105.5

Circumferential Strain

0.524

0.528

0.524

tMin (mm)

0.21

0.41

0.35

tAve (mm)

0.525

0.525

0.526

Thinning Factor

0.4

0.781

0.684

tMin / tMax

0.256

0.503

0.463

Table 2: Results of the bulge forming experiments


The results clearly show that the proposed optimization
scheme significantly reduces the forming time without
compromising the uniformity of the formed sheets. The
-3 -1
sheet formed at the fast strain rate of 1x10 s took only
17.3 minutes; however, it failed as a result of severe
thinning. On the other hand, the sheet formed at the slow
-5
-1
strain rate of 5x10 s had a uniform thickness
distribution; yet, it took 500 minutes. The sheet formed
according to the optimum profile took only 67.5 minutes (a
significant reduction from 500 minutes!) without sacrificing
the deformation uniformity. This result is also illustrated
by the thickness distributions of the three formed sheets
as shown in Figure 6. The profiles of the three sheets are
shown in Figure 7 along with the profile generated using
FE simulations. It is also important to note that the
thickness distribution predicted by the FE analysis agrees
to a large extent with the experimental results as
highlighted in Figure 6.
0.9
1E-03
5E-5
Opt-Exp
Opt-FE

Thickness (mm)

0.8
0.7
0.6
0.5
0.4

50

0.3
100

0.2
0

10

20

30

40

50

Distance along the Dome x (%)

Figure 6: Thickness distribution of sheets bulged at


various strain rates

4 SUMMARY
A new optimization scheme for superplastic forming that
takes into account different failure modes was presented.
Detailed uniaxial and biaxial experiments on the AZ31 Mg
alloy clearly indicate the effectiveness of the proposed
optimization scheme in reducing the forming time (e.g.,
from 500 to 67.5 minutes) without significantly sacrificing
the uniformity of the formed sheet. Moreover, the
experimental results support the predictions of the finite
element simulations.

-1

Opt.

-3

-1

10 s

Optimum

Figure 7: Deformed sheets vs. FE simulation

5 ACKNOWLEDGMENTS
The support of the US National Science Foundation,
CAREER Award # DMI-0238712, is acknowledged.
6 REFERENCES
[1] Doege, E., Drder, K., 2001, Sheet Metal Forming of
Magnesium Wrought Alloys Formability and
Process
Technology,
Journal
of
Materials
Processing Technology, 115/1:14-19.
[2] Doege, E., Drder, K., 1997, Processing of
Magnesium Sheet Metals by Deep Drawing and
Stretch Forming, Materiaux & Techniques, 7-8:1923.
[3] Siegert, K., Jger, S., Vulcan, M., 2003, Pneumatic
Bulging of Magnesium AZ31 Sheet Metals at
Elevated Temperatures, Annals of the CIRP,
52:241-244.
[4] Khraisheh, M., Zbib, H., 1999, Optimum Forming
Loading Paths for Pb-Sn Superplastic Sheet
Materials, ASME JEMT, 121:341-345.
[5] Ding, X., Zbib, H., Hamilton, C., Bayoumi, A., 1995,
On the Optimization of Superplastic Blow-Forming
Processes, ASM JEMP, 4/4:474-485.
[6] Banabic, D., Vulcan, M., Siegert, K., 2005, Bulge
Testing under Constant and Variable Strain Rates of
Superplastic Aluminium Alloys, Annals of the CIRP,
54:205-208.
[7] Thuramalla, N., Deshmukh, P., Khraisheh, M., 2004,
Multi-Scale Analysis of Failure during Superplastic
Deformation, Mater. Sci. Forum, 447-448:105-110.
[8] Khraisheh, M., Zbib, H., Hamilton, C., Bayoumi, A.,
1997, Constitutive Modeling of Superplastic
Deformation. Part 1: Theory and Experiments, Int. J.
of Plasticity, 13:143-164.
[9] Thuramalla, N., Khraisheh, M., 2004, Multiscale
Based Optimization of Superplastic Forming,
Transactions of NAMRI/SME, 32: 637-643.
[10] Wilkinson, D., Caceres, C., 1984, Large Strain
Behavior of a Suprerplastic Copper Alloy
Deformation, Acta Metallurgica, 32:415-422.
[11] Pilling, J., Ridley, N., 1989, Superplasticity in
Crystalline solids, The Institute of Metals, London.
[12] Hart, E., 1967, Theory of Tensile Test, Acta
Metallurgica, 15:351-355.
[13] Abu-Farha, F., Khraisheh, M., 2006, Mechanical
Characteristics of Superplastic Deformation of AZ31
Magnesium Alloy, ASM JEMP, In Press.
[14] Abu-Farha, F., Khraisheh, M., 2005, Modeling of
Anisotropic Deformation in Superplastic Sheet Metal
Stretching, ASME JEMT, 127:159-164.

You might also like