You are on page 1of 18

International Journal of Plasticity, Vol. 14, Nos 4-5, pp.

301 318, 1998


Pergamon © 1998 ElsevierScience Ltd
Printed in Great Britain. All rights reserved
0749-6419/98 $19.00+0.00
PlI: S0749-6419(97)00065-X

EXPERIMENTAL ANALYSIS OF ALUMINUM YIELD SURFACE


FOR BINARY AI-Mg ALLOY SHEET SAMPLES

Y. Maeda, 1. M. Yanagawa, 2 F. Barlat, 3 K. Chung, 4 Y. Hayashida, l S. Hattori, l


K. Matsui, 5 J. C. Brem, 3 D. J. Lege, 3 S. J. Murtha 3 and T. Ishikawa 6

IProcess T¢:hnotogy Research Laboratory Kobe Steel, Ltd., 5-5, Takatsukadai l-chome, Nishi-ku, Kobe,
Hyogo 651-22, Japan
2Material Research Laboratory Kobe Steel, Ltd., 5-5, Takatsukadai l-chome, Nishi-ku, Kobe,
Hyogo 651-22, Japan
3Alcoa Technical Center, Aluminum Company of America, Alcoa Center, PA 15069, U.S.A.
4Department of Fiber and Polymer Science, College of Engineering, Seoul National University, Kwanak-Ku,
Seoul, Korea
5Aluminum-Copper Division Kobe Steel, Ltd., 5-5 Takatsukadai 1-chrome, Nishi-ku, Kobe,
Hyogo 651-22, Japan
6Department of Materials Processing Engineering, Nagoya University, Furo-cho, Chikusa-ku,
Nagoya 464-01, Japan

(Received in final revisedform 23 September 1997)

Abstract--In this work, the yield surfaces of binary aluminum-magnesium alloy sheet samples
were measured using biaxial compression tests. Sheet samples of a given material were stacked and
bonded together with epoxy and cubic compression specimens were machined out of the laminate.
The yielding behavior was assumed to be independent of the hydrostatic pressure. In the analysis
of the biaxial compression tests, the effects of friction and of the elasticity of the die were accoun-
ted for. These effects were studied with the aid of finite element method (FEM) simulations of the
test which proved to be useful in avoiding systematic errors. The yield surfaces of three binary alloy
sheet samples containing 5 wt% Mg but with different crystallographic textures were analyzed. The
different textures resulted from processing under different thermomechanical conditions. The experi-
mental yield surfaces were compared to predictions made with the Taylor-Bishop and Hill (TBH)
model and with a phenomenological yield function. The experimental and polycrystal yield surfaces
were found to be in fair agreement. The yield function was found to be a suitable description of the
plastic behavio for only two of the materials studied. © 1998 Elsevier Science Ltd. All fights reserved

I. INTRODUCTION

Recently, aluminum alloy sheets have been applied to autobody applications. However,
aluminum alloy sheet exhibits less formability than steel. This difference arises from the
material properties in the plastic deformation range. To help in understanding this dif-
ference, the yield surface provides important information. In particular, finite element
method (FEM) simulations (Hayashida et al., 1995; Logan, 1995) can be applied to
simulate sheet forming processes and the yield source is an important input. The forming

*Corresponding author

301
302 Y. Maeda et al.

limit diagram (FLD) is also an important indicator of the formability and can be calcu-
lated using a theoretical model (Marciniak and Kuczinski, 1967; Jianshe et al., 1989) using
strain hardening equation and the yield function as inputs. For these reasons, researchers
in forming processes are interested in yield surfaces. However, it is difficult to realize
biaxial stress conditions that can permit the accurate measurements of stresses and strains
for determination of the yield surface. Some researchers (Tozawa, 1978; Tozawa and
Nakamura, 1992) tried to measure the yield surface and establish the associated yield
function formulation. Recently, direct calculations of the forming process (Mathur, 1995;
Zhou et al., 1995) using texture information have been developed without measurement of
the yield surface. In this paper, we discuss a method of yield surface measurement and
compare the experimental results with the predictions obtained from a polycrystal model
and a phenomenological yield function obtained on binary AI-Mg sheet samples.

II. YIELD SURFACE EXPERIMENTS

II.1. Experimental conditions

The yield surfaces of AI-5%Mg alloys were measured using a biaxial compression test.
The biaxial compression test machine was constructed by combining horizontal and ver-
tical compression test machines. The testpiece blocks were machined from many sheet
samples stacked and glued together with 30/zm epoxy, each block being a cube of side
12 mm (Fig. 1). Silicone oil was used as the lubricant between the sample and the die. The
blocks were compressed using two hydraulic cylinders. A constant velocity was imposed
on the vertical cylinder, and the displacement and the force generated by the die head were

12mm
Fig. 1. Size of test piece.
Experimental analysis of aluminum yield surface 303

measured. "lqaevelocity of the horizontal cylinder was controlled using a micro-computer,


so as to maintain a constant force ratio between the two loading axes. The displacements
and forces d~eveloped along both axes were automatically measured and accumulated on a
floppy disk as binary data every 30ms. These data were converted into true stress and
strain continuously, every 150 ms. The stress ratios of each sample were measured along
16 directions, as shown schematically in Fig. 2 and under three conditions involving shear
strains in the RD and TD reference frame (Fig. 3). For these three conditions, the com-
pression directions were at 45 ° and 35° from the rolling direction.

11.2. Experi!mental analysis method for yield surface determination


11.2.1. Effect of Elasticity. Doing each compression test, the stresses and strains were
calculated, taking into account the geometrical changes associated with the specimens.
The yield surface is defined as the set of stresses at the yield points. To construct a

TD

Fig. 2. Stress direction on ~r plane.


304 Y. Maedaet al.

Fig. 3. Shearedsample.

yield surface from the raw data, first, the plastic deformation must be separated from
the total deformation. However, real materials do not exhibit a sharp transition
between their elastic and plastic states. Moreover, in the experiment, the compression
machine itself behaves as an elastic spring. This make it difficult to define the yield
stresses and separate the elastic and plastic strains. Another difficulty arises from the
non-uniform deformation of the compression specimen due to imperfections in the
parallelism of the dies. Therefore, as a result, the initial slope of the stress-strain curve
is very low.
In order to confirm both effects (elasticity of the testing machine and lack of paralle-
lism), experimental loading-unloading sequences were carried out. Figure 4 shows the
results of these tests with the corresponding stress conditions. In general, the initial slope
of the stress-strain curve was about 20 GPa and significantly lower than the expected
value, about 70 GPa for aluminum alloys. The slopes after unloading and reloading were
observed to increase with the amount of plastic strain but to remain below the expected
value. FEM simulations (ABAQUS/Explicit) were performed in order to understand this
phenomenon. First, the stress-strain curves obtained with rigid dies and with elastic dies
were compared, as shown in Fig. 5. T h e effect of the elasticity of the dies appeared
during both loading and unloading, but was not particularly significant. Then a small
gap was introduced at the contact surface, which resulted in a lower force during initial
loading. Non-uniform contact leads to local elastic deformation, and the total force
remains small, until uniform plastic deformation occurs in the specimen. From these
results, the experimental loading-unloading test was better understood and led to the
following conclusions:
Experimental analysis of aluminum yield surface 305

200
24.2GPa . ...-"°'""

¢n
~1°°11/"
;-"
(
iiiiiii
~
29.4 GPa
j
Ij

0 5 10
Total Strain /

#.
300

200
3/ .......

a.
31.4

oo. ,
GPa

100'

5 10
Total Strain / %
Fig. 4. Examples of stress-strain curves on load-unload tests.

. The value of the elastic slope during loading should be ignored.


2. The effect of the elasticity of the dies is accounted for by using the experimental
unloading modulus. Esys (the system elastic modulus after yielding) can be used to
calculate the die elastic modulus Ed through the equation

l/Esys : l/Emat Jr- 1lEd (1)

Here, Esys :is the system elastic slope displayed by an experimental unloading test and Emat
is the material modulus, which is a known constant (68.6 GPa).
306 Y. Maeda et al.

Loading slope 47.4GPa


Unloading slope 46.2GPa
Elastic flat die
400 Loading slope 68.9GPa ,,~"
\,fl
t~
Unloadingslope 78.7GPa
Rigid flat .die ,-../,4 Z-'
0
2
0

0
° ~ 200 !
E ,-/,K, . / Elastic die with

/
0
non-uniform
contact condition
Loading slope 2.9GPa
Unloadingslope 40.5GPa
O9
w-
w g/ I I I I I i n

0 0.1 0.2
Strain from displacement of die
Fig. 5. Simulated stress-strain curve for several conditions of die.

. The material elastic modulus exhibits considerable elastic anisotropy because of the
laminated structure of the specimen. Thus, it can be assumed that Em,t,RD and
Emat,TD are both equal to 68.6 GPa and that Emat,ND is 50.2 GPa. (RD, T D and N D
represent the rolling, transverse and normal directions, respectively).

11.2.2. Effect of friction. The next correction accounts for the effect of friction between
the die surfaces and the specimen. In this compression test, friction is an important factor
that affects the measurement of the stress. To estimate this effect for uniaxial and biaxial
compression tests, numerical Calculations were carried out using ABAQUS/Explicit (Figs 6
and 7). In these figures, the stress-strain curves were calculated for different friction
coefficients from the forces and displacements on the dies and compared to the true stress-
strain curves given as input to the simulations. The agreement was good in the cases where the
friction coefficient/z was equal to zero. With increasing/z, the discrepancy between the true
and calculated strain-stress curves increased. The difference was largest in the biaxial case.
The "excess stress" is defined here as the stress with/z = 0.0 subtracted from the cal-
culated stress with/z # 0.0. The excess stress ratio is defined as the excess stress divided by
the stress with/z = 0.0. In Figs 8 and 9, the excess stress ratio is shown for the uniaxial
and biaxial tests. In the uniaxial case, if/z is less than 0.1, the friction effect can be ignored
and the error is less than 5%. However, in the biaxial case, the error can be more than
10%, if the influence of friction is not taken into account. From Fig. 9, it is evident that
the excess stress ratio is approximately equal to the friction coefficient for the biaxial
Experimental analysis of aluminum yield surface 307

5OO

-- " ..< //'d' •

o 25o

/ / _ _ _ .:o.1
f g=0.2

IX=0.3

True S-S

i I i I i I , I
0 0.1 0.2
Strain
Fig. 6. Effect of friction on uniaxial test.

compression test. However, it is difficult to determine the friction coefficient in these


experiments directly.
To detet~rnine the friction coefficient, we compared uniaxial compression test results for
the thickness direction with biaxial compression test results for deformation along the
rolling and transverse directions. In the former case, the stress state is equivalent to biaxial
tension, while in the latter case it is biaxial compression. If symmetry between compres-
sion and biaxial tension is assumed (no strength differential effect), both stresses should be
the same. However, if friction plays a role, these two stresses are different. Therefore, to
correct for the friction effect, a simple model which can be used under any stress condition
and for any friction coefficient/z was developed. In the model, the measured forces Fml,
Fro2 and the effective forces Fel, Fe2 satisfy the following equations:

Fml = Eel + / £ F m 2
Fro2 = Fe2 + ~ F m l - (2)
If the biaxial tension and compression stresses are the same, the friction coefficient /z
which satisfies this condition is/z = 0.1.
308 Y. Maeda et al.

~l.t=0,3
f
. . . . p=0.2 f
500 ~J
-- -- -- p=O. l 7
t~ f
. . . . . . . p=O,05
f I.~ ~ J
~g=O.O ./" t.1"
//C~/
CO 250

0 0.05 0.1
Strain
Fig. 7. Effect o f friction on biaxial test.

II.2.3. Definition of yielding The yield stresses are difficult to determine accurately,
because for real materials there is no sharp transitiori from the elastic to the plastic
state. Moreover, the slope of the stress-strain curve is very steep near the yield condi-
tion. Therefore, to determine the yield stress, it is necessary to distinguish between the
plasticity stage and the elastic stage. By means of isotropic elastic deformation for plane
stress, trx + try and ex + ey satisfy eqn (3).

(ex + ey) = (1 -- v)(trx q- Cry)/E. (3)

From eqn (3), yielding is defined as a critical slope (trx + try) versus (ex + ey). If the
material is isotropic elastic, this slope is constant in the elastic region at any stress ratio.
However, the experimental slopes change continuously and rapidly from the elastic to the
plastic region for the reasons given above. In order to define the yielding point, the con-
dition (trx + try)/(ex + ey) 5 GPa (elastic cut-off parameter) was used. The yield strain eel
and the yield stress tryld were determined from the measured strain and stress at the yield
condition, respectively.

11.2.4. Post yielding behavior. These yield stresses included uncertainty, because of the
artificial elastic cut-off parameter. To obtain an accurate yield surface shape, the post
yielding behavio was analyzed. From these yield strains and stresses, the plastic strain was
defined by the following equations:
Experimental analysis o f aluminum yield surface 309

/- /1 =0.3
f
/
0,1 /
O
. I
/
U i
i f / ~=0.2
I/ f
¢,o
(1) .i' [
-I.-
! /
i
oo 0 . 0 5 /
// ................
~/ .._...- ~ =0.1
4

I r !

r,// ~ =0.05

I I I I I I I I I I
0 250 500
True stress M P a
Fig. 8. Excess stress ratios in the uniaxial tests.

eplx = eobx - - eelx -- [trobx -- Oryldx]/Ematxx -- [O'oby -- O'yldy]/Ematyx


(4)
gply : Eoby --/~ely -- [O'oby - - O'yldy]/Ematxy - - [orobx -- O'yldx]/Ematxy.

are the plastic strains, eobx,ythe measured strains, eelx,y the yield strains, trobx,y
H e r e , eplx,y
the measured stresses and O'yldx,ythe yield stresses in each direction. The gmatxx,xy,yx,yy a r e
the materi~Ll elastic moduli in the different material directions.
The plastic work Wp defined by numerical integration of the stress with respect to the
plastic strain can be calculated.

Wp=JCTobx'dEplx"l"IOroby'dEply (5)

For different values of the plastic work Wp}, the compression stresses are converted into
tension anti compression stresses along RD and TD, assuming that a hydrostatic pressure
does not irLfluence yielding. An example of the resulting yield surface is given in Fig. 10.

11.2.5. Accuracy of the measurements. In this procedure, several variables were intro-
duced, for example the friction coefficient/z and the elastic cut-off parameter whose values
are somewhat ambiguous. These ambiguities can cause systematic errors. For example, if
310 Y. M a e d a e t a l .

0.5
f~

J
~f
~f /1 = 0 . 3
~f
f
B f
j f l
/ f
f
0 B I
/ f,~"
U / f
f l ,~ = 0 . 2
0") 0.25 !
f l I f

.4--- - /ii II
/1 =o.1
-/// ,"
11
r .... "-'"

----,u =0.05

i s J I i t i i
0 250 5OO
T r u e stress M P a
F i g . 9. E x c e s s stress r a t i o s in t h e b i a x i a l tests.

too large a value of/z is assumed, then the yield stress in uniaxial compression becomes
larger than the yield stress in biaxial compression at the same amount of plastic work. If
the elastic cut-off parameter is chosen incorrectly, calculation of the plastic work can be
affected depending on whether the loading condition is uniaxial compression or biaxial
compression.
Comparing the stress-strain curves between the uniaxial compression tests along the
thickness and the biaxial compression tests along the TD and RD in Fig. 11, these are in
agreement with each other. When the yield surface is plotted, no systematic large differ-
ence between compression and tension along any direction can be observed which sug-
gests that the selected elastic cut-off parameter and the friction coefficient are reasonable.
From Fig. 11, it was suggested that the effect of the epoxy is negligible after these collec-
tions.
Another possible systematic error is the plastic work dependence on the strain. At small
amounts of plastic work, the yield stresses are very sensitive to the elastic cut-off parameter.
However, at large amounts of plastic work, the laminated specimens may be destroyed,
leading to large systematic errors. In particular, the TD and RD uniaxial compression
conditions are the most sensitive to buckling and specimen destruction. For these reasons,
the data obtained at intermediate amounts of deformation are the most reliable. Finally,
Experimentalanalysisof aluminumyieldsurface 31!

J
Sal iple with s ~ear i n ~
RD ,TD frame
0
o

o
Z
-1 O
t

-1 0 1
Normalized Stress Ox/O
Fig. 10. Example of analyzed yield surface.

200.
19o; .........'."'

~60.
SO1701
.; f •".~":": "
tSO; ..~?
140.
130.
~ 120.
110. ~ ' a x i a l compression test
~ 90 ~ a x i ~ . compression test
80 / ~niaxial compression test
.......... Uniaxial compression test
70
60 .......... Uniaxial compression test
,o
40
30
20
10 Bi~ial eoml~esiontest Umaxi
0
o " l' " ~ ' ~ ' I ' ~ " d " i ' ~
Plastic strain (%)

Fig. 11. Comparison of the stress-strain curves between the biaxial compression and the uniaixial compression
tests.
312 Y. Maeda et al.

statistical error was limited, because each yield surface point was measured three times
under the same conditions. Comparing these triplicate tests, it was found that the largest
and average statistical errors were about 10 and 5 MPa, respectively.

11.3. ot - 0 plot analysis


The yield surface is the most popular description for material plastic anisotropy. In the
classical flow theory of plasticity, the associated plastic flow rule theory indicates that the
direction of incremental plastic strain is normal to the yield surface. For this reason, not
only stresses but also strains were measured at the same time in the biaxial compression
tests. A new description for material anisotropy based on the concept of the associated
flow rule, the a - 0 plot is introduced. Figure 12 shows three different angles that char-
acterize the material anisotropy. 0 is the angle between the loading direction and uniaxial
loading along the rolling direction. 0e is the angle between the strain increment and uni-
axial loading along the rolling direction. The angle ot is defined as follows:

a = 10 - 0,1. (6)

From the experimental data, 0, 0e and ot can be easily determined, except for uniaxial
compression, because, for these states, only the strain in the compression direction was
measured.

11.4. Comparison between theoretical predictions and experimental results


The Taylor-Bishop and Hill (TBH) model (Bishop and Hill, 1951) is useful to explain
the plastic deformation of polycrystalline materials. Aluminum alloys have a FCC crystal
structure and can be adequately studied using the TBH model. First, the (111) and (200)
pole figures were measured using X-ray techniques. From the pole figures, orientation
distribution functions (ODFs) were computed. This information was used as an input for
the TBH model. The experimental of yield surfaces and ot - 0 plots results were compared
with the TBH model predictions (Fig. 13). The materials had the same thickness, 1.0 mm,
but different amounts of cold reduction before solution heat treatment. As a result, they

1.0'
/

d-
0.5"
u~
¢1}

CO

0,0"
0.0 0.5 1.0 E
Stress GRD
Fig. 12. Schematic figure o f a and 0.
Experimental analysis of aluminum yield surface 313
(a) (b) I
'Ofl I l I l I u I l I u I I I I I I I l l I I I
CR 50 % ;SHT 10eec/530C Grain size 70p m
------ TBH Model
• , . . . . I . . . . , "
• MeNUCmelX

;
/ Xe

,/ l
=,
o.4 ~ i •
,, f ii
7 o~:
i i/ ~l X [ U
z LI X/ .~
~'----e ....I"/ V :!!
-1
, . . . .
0 . . . .
i
1 "
o ,I . . . . . . . .t. . . . . . . +," I

-150 0 150
Normalized Stress o RD Stress direction O / degrees
Material A Material A

0"8t'11 U I u I I I U U I I I U I U U U U U I I
CR 80% ;SHT 15sec/530C Grain size 70p. m
• I . . . . I . . . . , •

/ 0.6 I

/
| /
0.4

-! I

! /
¢D
It
0.2
LI'

I-'1
~aI ~x

~
• /""
I l da\

. . . a. X
~I ['
z \ ,,.~
k I
L ,~ ~/ .~ ~4
/
1% - ~ ":
ol i I ~ , , , , ,'~, * , a , , ~',, , , , , , n
-1 0 1 -150 0 150
Normalized Stress o RD Stress direction e / degrees
Material B Material B
0.8 i i i i i i i i i i l i i i i i i I i i i i

CR 17°/. :SHT30sec/530C Grain size 1501zm


• , . . . . u . . . . , "

~l a / / I
/ ~' ~ o.~

i ,/ ~ ~ ]1 I I "¢"
\ L,,. -'4 ~ ~ !
z I/ i
,,f ~ |
-t .... ~ ..... ,,,,,V, Lu,,w,,,,,,,n
1 -1.5( 0 150
Normalized Stress o RD ;tress direction 8 / degrees
Material C Material C

Fig. 13. (a) Measured yield surfaces (O) compared with polycrystal model predictions for various materials (.)
Material A: cold reduction 50%, heat treatment 530°C for 10 s, grain size 70/zm; material B: cold reduction 80%,
heat treatmerLt 530°C for 15 s, grain size 70/zm; and material C: cold reduction 17%, heat treatment 530°C for 30 s,
g r a i n size 1 5 0 / z m . (b) M e a s u r e d tz - 0 p l o t s ( O ) c o m p a r e d w i t h p o l y c r y s t a l m o d e l p r e d i c t i o n s f o r v a r i o u s m a t e r i a l s .

had different textures and different anisotropies. F o r the yield surfaces, the measurements
and polycrystal results are in g o o d agreement. F o r the ot - 0 plots, in general, the mea-
surements and polycrystal results are also in g o o d agreement. So, it appears that the
standard T B H model explains the yield surface shape and the relationship between the
stress and the strain increments.
314 Y. Maeda et al.

II.5. Empirical yield surface function

The goal of this project was to define a constitutive equation for the material studied
and to conduct FEM simulations of forming processes using this constitutive equation. If
the yield function has an analytical description, it is easy to apply to a complex forming
process with short computation time. In this work, the six-component yield function
(Barlat et al., 1991) was used to describe the AI-Mg alloy sheet, because this function was
developed mainly for aluminum alloy sheet. This function is defined by means of the fol-
lowing equations:

A _~-
ayy - fzz, B = fzz - fxx, C = fxx - fyy, F = fyz, G - azx, H = O'xy
I2= {(fF) 2 + (gG)2 + (hH)2}/3 + {(aA - cC) 2 + (cC - bB) 2 + (bB - aA)2}/54

h = (cC - bB)(aA - cC)(bB - a A ) / 5 4 + f g h F G H - {(cC - bB)0CF)2}


+ (aA - cC)(gG) 2 + (bB - aA)(hH)2}/6
0p= arccos( /2/ ~ V2)
2fm -----(3/2)m/2{[2 cos((20p + zr)/6)] m + [2 cos((20p - 3rr)/6)] m + [-2 cos((20p + 5zr)/6)]m}.

Here, a, b, c,f, g and h are anisotropy parameters. From these equations, the yield surface
and the a - 0 plot can be defined as:

O'm ~-- Fsurface(fxx,fyy, fizz, O'xy,fyz, fzx, a, b, c,f, g, h, m)


(8)
0 = F~-o(fxx, fyy, a, b, c,f, g, h, m)

In order to determine the yield function coefficients, the least squares method was used
with the error function ) 2.

X2 =
~-"~{O'm F surfaceI,Qyob
- xx, O,~r~)}2 ..[_ fl ~-'~{0ob - F~-o(fxx,
ob ob 2
%y)} (9)

Here, fl is a weight function (constant) introduced to normalize the different measure-


ments. In this case, fl is 10000, because the error is roughly 0.1 in 0 and 10MPa in f. The
superscript " o b " indicates measured values. However, not all of the measured data was
used, because of the large errors described previously. The data ignored were associated
with uniaxial compression along the RD and TD directions, because the laminated
structure was destroyed in these cases. To minimize X2, eqn (10) must be satisfied.

8x218a = 8x2/Sb = 8x218c = 8x2/Sf = 8x=lSg = 8x218h = 8x218fm = 0.0 (10)

II.6. Fitting results

The results of this analysis for the yield surfaces and a - 0 plots with the best fits
obtained using the phenomenological function are shown in Fig. 14. The best fit para-
meters are listed in Table 1. For material A, the phenomenological yield surface is in
better agreement with the experiments than with the TBH model. However, the ot - 0 plot
Experimental analysis of aluminum yield surface 315

(a) (b) 0.8 i i I i l i i i i i i i i i i i i i i i i I

• , . . . . . . . . i - i
i ~--- TSH Modll

0.6 ~ v u o ~

t 0.4
~J

¢b
o 0.2

,|,
-150 0 150
N o r m a l i z e d S t r e s s o RD Stress direction 0 / degree~
Material A Material A

0.8 ii i i i i i i i i i i i i i i i i i i i i I

• , . . . . i . . . . , •
---- TBHMoaa
_h--.--aL ~,
~ YLD FueClk~
0.6
o
/~// ,

~ 0.4
p |

o II 0.2
/
-1
• i . . . . J | 1
nJiJ|, ,, n
-1 0 1 -15O 0 150
Normalized Stress o RO Stress direction 8 / dgrees
Material B Material B

0.8 i i i i i i i i i i i i i i i i i i i i I

• i . . . . i . . . . J '
- - - - - - Tgtl I~da
YLO FunCUO~
0.6
o

'--~ 0.4

¢D

II 0.2
z

, i . . . . i |

-1 0 1
-150 0 150
Normalized Stress o RD Stress direction 0 / degrees
Material C Material C

Fig. 14. (a) Measured yield surfaces ( 0 ) and polycrystal model predictions for various materials. Material A:
cold reduction 50%, heat treatment 530°C for 10s, grain size 70/zm; material B: cold reduction 80%, heat
treatment 530°C for 15 s, grain size 70 p.m; and material C: cold reduction 17%, heat treatment 530°C for 30 s,
grain size 150/zm. Solid lines show the fitted phenomenological yield functions and dashed lines show the TBH
model. (b) Measured ct - 0 plots ( 0 ) compared with polycrystal model predictions for various materials.

Table 1. Best fit parameters for the phenomenological yield function

m a b c f g h

Material A 4.98 1.00 1.00 0.95 1.00 1.00 0.91


Material B 7.10 1.00 0.98 0.84 1.00 1.00 0.93
Material C 6.36 1.00 1.00 0.90 1.00 ! .00 1.00
316 Y. Maeda et al.

-- v ~ F=~-~ Material A
o.

o. .O

o. /"'Y O"~'""• • .... /"..Y ~",...

o.

> o, .... . . . . .
L
o.

o.

o.

Direction from rolling (°)

...... vto F ~ Material B


0.

o. •

o. O/ :, ........ / ~.-.
~o.
0/ 'o ./ ......
o.

o,

o.

0
c 30 ~10 90 120 150
Direction from rolling (o)

• ~m~a
...... YLD FlXclion Material C
......... TBH Model
1.8

1.e-"
......... •. "/ ~iiz
1,4i "\
1.2"~

1.0"~
0.8 -~

0.6 ~
f......,'
0,41

0.2'
0.0
180
Direction from rolling (°)

Fig. 15. Measured angular distribution of r-value compared with the predictions obtained from the TBH model
and the fitted empirical yield functions.
Experimental analysis of aluminum yield surface 317

exhibits some discrepancies with respect to the experimental plot near 0 = 15° and 75°.
For material B, there is some discrepancy between the experimental and phenomen-
ological yield surfaces from 0 = 45 ° to 90° and at 0 = 135° and 0 = 315 °. For the a - 0
plot, the phenomenological description does not match the experiment or the TBH pre-
dictions ne~Lr 0 = 0° and 0 = 90°. For material C, the results of the phenomenological
yield function are in good agreement with those of the experiments and of the TBH pre-
dictions for the yield surface and the o t - 0 plot. In general, there are differences in the
u - 0 plot l~etween the experimental results and the predictions from the phenomen-
ological yield function. As a result, the r-value, which is the width-to-thickness strain ratio
in a simple tension test, is not very well predicted with the phenomenological model.
In Fig. 15, the r-value anisotropy is shown for the three materials considered. For
material B, the r-value anisotropy from the TBH model is in good agreement with the
experiment~tl results. For the other materials, some discrepancies are observed between
the TBH models and the experiments. Regarding the predictions of the phenomenological
yield functions, the predictions are in better agreement with the experiments for material
C, but not in good agreement for the other two materials. This is consistent with the
trends of the ot - 0 plots.

III. CONCLUSIONS

The yield surfaces of three binary AI-Mg sheet samples were measured using biaxial
compression tests on specimens machined from sheet samples stacked and bonded toge-
ther. Triplicate tests performed for the same loading state showed that the variation in
stress was less than 10 MPa. For uniaxial compression in the sheet rolling and transverse
directions, the test resulted in delamination and buckling of the specimens. Analysis of the
raw data was taking the friction between the dies and the compression specimens into
account, as well as the elasticity of the testing machine. The data generated with the uni-
axial compression tests were ignored because of the experimental difficulties observed
during these particular tests. In general, the Taylor-Bishop and Hill polycrystal model of
the yield surface was in good agreement with the experimental results. However, the phe-
nomenological yield function model was suitable for only two out of the three materials
studied in this paper. To characterize the relation between the stress and strain increment
directions, the concept of the ot - 0 plot was introduced. This plot is a much more sensi-
tive index than the yield surface to characterize the agreement between the theoretical and
experimental results. It was found that when the phenomenological and polycrystal pre-
dictions of this plot do not agree with the experimental results, the respective r-value
predictions are not in good agreement with the experimental r-values.

REFERENCES
Barlat, F., Lege, D. J. and Brem, J. C. (1991) A six-component yield function for anisotropic materials. Int. J.
Plasticity 7, 693-712.
Bishop, J. W. F. and Hill, R. (1951) A theory of the plastic distortion of a polycrystalline aggregate under com-
bined stresses, and A theoretical derivation of the plastic properties of polycrystalline face-centered metals.
Philos. Mag. 42, 414-427,1298-1307.
Hayashida, Y., Maeda, Y., Matsui, K., Hashimoto, N., Hattori, S., Yanagawa, M., Chung, K., Barlat, F.,
Brem, J. C., Lege, D. J. and Murtha, S. J. (1995) FEM analysis of punch stretching and cup drawing test for
aluminum alloys using a planar anisotropy yield function. Simulation of Materials Processing 717-722.
318 Y. Maeda et al.

Jianshe, I., Dajun, Z. and Bernard, B. (1989) Application of Hill's new yield theory to sheet metal forming--Part
I. Hill's 1979 criterion and its application to predicting sheet forming limit. Int. J. Mech. Sci. 31(4), 237-247.
Logan, W. R. (1995) Finite-element analysis of earing using non-quadratic yield surfaces. Simulation of
Materials Processing, 755-760.
Marciniak, Z. and Kuczinski, K. (1967) Limit strains in the processes of stretch-forming sheet metal. Int.
J. Mechanical Science 9, 609--620.
Mathur, K. K. (1995) Parallel algorithms for large scale simulations in materials processing. Simulation of
Materials Processing 109-114.
Tozawa, Y. (1978) Plastic deformation behavior under the compression of combined stress. In Mechanics of
Sheet Metal Forming, eds D. P. Koistinen and N. M. Wang, pp. 81-110, Plenum Press, New York.
Tozawa, Y. and Nakamura, M. (1992) A biaxial compression testing method for thin sheets. Plasticity and
Processing 13, 538-541.
Zhou, Y., Jonas, J. J., Jain, M. and Macewen, A. R. (1995) Calculation of polycrystal yield surfaces and forming
limit diagrams of aluminum alloy sheets. Simulation o f Materials Processing, 369-374.

You might also like