Professional Documents
Culture Documents
Modeling Evolution of Anisotropy and Hardening For Sheet Metals
Modeling Evolution of Anisotropy and Hardening For Sheet Metals
DISSERTATION
By
*****
Modeling the material behavior accurately in large strain and large rotation is
essential to sheet metal forming simulation and springback prediction, especially when
of the texture and the modification of dislocation structures. It is viewed that the
The plastic spin theory is introduced to define the rotation of texture and the
hardening model proposed by Chun et al. (2002) is modified to model the transient
is proposed to account for the rotation as well as the expansion (isotropic) and translation
(kinematic) of the yield loci in terms of the plastic spin theory and the anisotropic
when the rotation of anisotropy axes is not considered. The effects of reorientation of the
ii
backstress in alloys containing hard precipitates appear to be the dominant factor for the
effects of multi-path loadings: crossing and Bauschinger effects. The proposed model
reproduces the hardening in an arbitrary orientation as well for a wide range of strain.
Only two material parameters are added to describe the evolution of anisotropy
and related backstress, which can be identified from tensile tests in three orientations and
curve fittings. Three initial R-values can be obtained from the tensile tests. The other six
parameters for defining isotropic and combined kinematic hardening are identified by the
three-point-bend test and the inverse method developed by Zhao and Lee (2000),
modified by Chun et al. for the ANK model. Relatively straightforward material
The RIK model is applied to sheet metal forming and springback predictions.
Cylindrical cup and square cup drawings are simulated and compared with available
experimental data. The RIK model shows good trends in earing and draw-ins, compared
with other models. The draw bead test simulation shows a significant difference in the
springback shape when compared with other hardening models. The RIK model corrects
the unexpected twisting mode which is computed for the specimen cut in an arbitrary
orientation away from the RD when using conventional isotropic and combined
prediction of springback.
iii
Dedicated to my parents and family
iv
ACKNOWLEDGMENTS
I would like to express my heartfelt thanks to my adviser, Professor J.K. Lee, for
co-advisers: Professor R.H. Wagoner shared his expertise in the sheet metal forming
research. Professor M.E. Walter led the experimental observations and taught me how to
I appreciate Dr. C.-S. Han for sharing his ideas and mentoring this research and
I would also like to thank all of the members of ComPro Lab: Dr. B.K. Chun, J.
Choi, H.S. Kim and Hyungjun Kim for their interactive and fruitful discussions and help
acknowledged:
Finally but not at the least, I would like to express my sincere gratitude to my
parents for cheering me, and my wife and children for being with me throughout the long
v
VITA
PUBLICATIONS
Research Publication:
1. Choi, Y., Han, C.-S., Lee, J.K., and Wagoner, R.H., Effect of Anisotropy Axes
Rotation on Spring Back, in Numisheet 2002, Jeju Island, Korea, p. 325-330.
2. Han, C.-S., Choi, Y., Lee, J.K., and Wagoner, R.H., Modeling of Anisotropic Work
Hardening, in Numisheet 2002, Jeju Island, Korea, p. 103-108.
vi
3. Han, C.-S., Choi, Y., Lee, J.K., and Wagoner, R.H., A Fe Formulation for Elasto-
Plastic Materials with Planar Anisotropic Yield Functions and Plastic Spin,
International Journal of Solids and Structures, 39, 2002, p. 5153-5141
4. Choi. Y., Han, C.-S., Lee, J.K., Wagoner, R.H., Modeling Multi-Axial Elasto-Plastic
Deformation of Planar Anisotropic Material, Part I: Theory, to be submitted to
International Journal of Plasticity.
5. Choi. Y., Han, C.-S., Lee, J.K., Wagoner, R.H., Modeling Multi-Axial Elasto-Plastic
Deformation of Planar Anisotropic Material, Part II: Applications, to be submitted to
International Journal of Plasticity
6. Choi, Y., Walter, M.E., Lee, J.K., and Han, C.-S., Observations of Anisotropy
Evolution and Identification of the Plastic Spin Parameters, to be submitted to
International Journal of Solids and Structures
FIELDS OF STUDY
vii
TABLE OF CONTENTS
Page
Abstract ............................................................................................................................... ii
Acknowledgments............................................................................................................... v
Vita .................................................................................................................................... vi
List of tables....................................................................................................................... xi
Chapters:
1. Introduction..................................................................................................................... 1
3.1 Introduction....................................................................................................... 30
3.2 Material formulation ......................................................................................... 33
3.2.1 General model for isotropic, kinematic and rotational hardening ............ 33
3.2.2 Rotation of the symmetry axes of anisotropy (Rotational hardening)...... 36
3.2.3 Armstrong-Frederick type backstress components................................... 37
3.2.4 Nonlinear kinematic hardening with permanent softening....................... 38
viii
3.2.5 Correction of flow stress with kinematic hardening associated with the
anisotropy axes rotation ............................................................................................ 39
3.3 Application on plane stress problem for sheet metal forming. ......................... 41
3.4 Numerical investigation .................................................................................... 43
3.4.1 Material properties:................................................................................... 43
3.5 Conclusions....................................................................................................... 46
4.1 Introduction....................................................................................................... 60
4.2 Definition of R-value ........................................................................................ 63
4.3 Experimental set up........................................................................................... 64
4.4 Hardening models ............................................................................................. 67
4.5 Experimental results and material parameters .................................................. 68
4.5.1 DDQ (Experiment with DIC method)....................................................... 69
4.5.2 DDQ (NUMISHEET 2002) ...................................................................... 70
4.5.3 DQ............................................................................................................. 70
4.5.4 HSS ........................................................................................................... 71
4.6 Discussion ......................................................................................................... 71
4.7 Concluding remarks .......................................................................................... 72
5.1 Introduction....................................................................................................... 94
5.2 RIK hardening model........................................................................................ 95
5.3 Material properties ............................................................................................ 97
5.4 Applications ...................................................................................................... 98
5.4.1 Cylindrical cup drawing............................................................................ 98
5.4.2 Square cup drawing................................................................................. 100
5.4.3 Pulling the strip through a draw bead toolset and springback ................ 101
5.4.4 Hydroforming ......................................................................................... 102
5.5 Concluding remakrs ........................................................................................ 104
ix
7. Conclusion .................................................................................................................. 139
BIBLIOGRAPHY........................................................................................................... 141
Appendices:
x
LIST OF TABLES
Table Page
Table 2.1: Comparison of continuum setting in the intermediate configuration and the
( ε e ) ) is used. .... 20
current configuration. Small elastic strain assumption ( Ve = I +O
Table 4.1: Material properties of RIK hardening model for DDQ, DQ and HSS. .......... 74
Table 5.1: Material properties for RIK hardening model. ............................................. 105
Table 5.2: Clamping force and drawing force for different orientation and different
hardening for draw bead test................................................................................... 105
xi
LIST OF FIGURES
Figure Page
Figure 1.1: Macro-mechanical test: a) A big specimen is cut from a rolled sheet with an
orientation Ψ to the RD and is pre-stretched by ε*, b) smaller specimens are cut
from the pre-stretched specimen in every 15° from the RD and are stretched to
measure yield stress .................................................................................................... 5
Figure 1.2: Definition of angles used in this research. The symmetry axes ( eiφ ), initial
orientation to the RD ( ψ ), angle of the symmetry axes rotation ( η ), and the angle
between the symmetry axes and the straining directioin ( ϑ ). Symmetry axes of
anisotropy rotates toward pre-stretched direction (ε*) from the RD and TD
whichever closer. ........................................................................................................ 6
Figure 1.3: Rotation of the symmetry axes as the pre-strain increases: measured yield
stress (0.2% offset, ●) and curve fitting to Hill’s 48 yield function [6]. ................... 7
Figure 1.4: Rotation of the symmetry axes observed by using pole figures for 45° initial
orientation. The symmetry axis is rotating gradually from TD counterclockwise to
45° as strain increases [5]. .......................................................................................... 7
Figure 2.1: The configurations and relationship between the configurations: reference
( B0 ), intermediate ( B, B ), and current ( B ) configurations. In infinite number of
intermediate configuration, a specific intermediate configuration is chosen by
leaving only stretch in the elastic deformation and all relative rotations are contained
in the plastic deformation, F = Ve R*U p . .................................................................. 23
Figure 2.2: Rotation angle of the symmetry axes using analytical solution and FE
analysis: (a) Ψ=30°, (b) Ψ=45°, (c) Ψ=60°............................................................... 24
Figure 2.3: Rotation angle of the symmetry axes using modified plastic spin scalar (a = -
350): (a) Ψ=30°, (b) Ψ=45°, (c) Ψ=60...................................................................... 26
Figure 2.4: Flow stresses for the RD, the TD, and 45°. The 45° flow stress approaches
the flow stress of the TD when the rotation of symmetry axes is used with an
isotropic hardening.................................................................................................... 28
Figure 3.2: Multi-path loading response of conventional hardening models: (a) isotropic
hardening, (b) anisotropic nonlinear kinematic hardening model. Both models
cannot reproduce the ‘crossing’ effect while the Bauschinger effect is caputred in (b).
................................................................................................................................... 51
Figure 3.3: Rotational evolution of yield function: The yield function is expanded to the
stress state C when only the isotropic hardening is considered. The yield surface is
rotate by using rotational hardening and the iterated stress state is B. The yield
surface is moved toward C with the amount of backstress by the kinematic
hardening................................................................................................................... 52
Figure 3.4: Flow stress in the RD, 45°, and the TD orientation with a) ANK, b) RIK.
ANK predicts higher flow stress for 45° while RIK correlates the measurement data.
................................................................................................................................... 53
Figure 3.5: Flow curves for the 45° specimen: isotropic+kinematic hardening model
(ANK) produces rather higher stress ( ( R 45 )avg , ( R 45 )init ), rotation+isotropic
xiii
hardening model produces lower stress, while the RIK follows the measured data
rather closely............................................................................................................. 54
Figure 3.6: Rotation angle (degree) of the symmetry axes for 45° oriented specimen. .. 55
Figure 3.7: Stress offset tests for loading-unloading-reloading in different path: (a)
loading condition – give different loading path to the pre-stretched specimen, (b)
stress offset with respect to the plastic spin parameter. The stress is normalized with
respect to the offset of 180º ( ∆τϕ / ∆τ180o )................................................................. 56
Figure 3.8: Flow stress response when using the rotational hardening with isotropic
hardening scheme. Stress jump (crossing) is shown but Bauschinger effect is not
captured..................................................................................................................... 57
Figure 3.9: Flow stress result of the multi-path loading when the RIK hardening model is
used. Both ‘cross’ and ‘Bauschinger’ effects are captured...................................... 57
Figure 3.10: Multi-path loading responses show the flow stress crossing after various
pre-strains; more pre-strain yields higher jump in stress. ........................................ 58
Figure 4.2: Transverse strain versus longitudinal strain of DDQ. The slope of each curve
is used to compute the R-values. (Numisheet 2002 [61])......................................... 76
Figure 4.3: Evolution of R-values: R0 and R90 are not evolving compared to the R45. . 77
Figure 4.5: FEM validation of specimen dimensions to avoid clamping effect: a) strain
measurement locations, b) computed strains. The strains are closer enought to show
that the measured area is safe from the clamping effect........................................... 79
Figure 4.6: Schematic diagram of the experimental set up using DIC(Digital Image
Correlation). .............................................................................................................. 80
Figure 4.7: The random pattern sprayed on the specimen and captured by CCD camera: a)
before and b) after deformation. The small rectangles highlight a 100 x 100 pixels
subset of the image that has been deformed. ............................................................ 81
xiv
Figure 4.8: a) Comparison of stress-strain measurement by DIC(symbols) and strain
gauge (lines) for a small strain, b) comparison of strain measurement by DIC
(symbols) and NUMISHEET 2002 data (lines) for large strain. .............................. 82
Figure 4.10: Simulation results of averaged R value for 45° orientation comparing with
the experimental results: a) Transverse strain versus longitudinal strain, b) flow
stresses. Prediction of 45° orientation is underestimated. ....................................... 84
Figure 4.11: Sensitivity of the plastic spin parameter: a) R45 evolution, b) flow stresses of
45° orientation according to the plastic spin parameter. The gap between the
measurement and computed flow stress is the amount of backstress evolution. ...... 85
Figure 4.12: Simulation with the RIK hardening model and comparison with
experimental results: a) Transverse-strain versus longitudinal strain, b) flow stresses.
Prediction for 45° orientation captures the experiment well. ................................... 86
Figure 4.13: DDQ- data provided by NUMISHEET 2002. Computed by RIK hardening
model comparing with experimental results: a) Transverse strain versus longitudinal
strain, b) flow stresses, c) R45 evolution. .................................................................. 87
Figure 4.15: HSS – experiment. Computed by RIK hardening model comparing with
experimental results: a) Transverse strain versus longitudinal strain, b) Flow stresses.
................................................................................................................................... 91
Figure 4.16: Linear relationship between the plastic spin parameter (a) and the planar
anisotropy (∆R). This relationship can apply for mild steel which has B.C.C. crystal
structure..................................................................................................................... 92
Figure 5.1: Cylindrical cup drawing: a) Dimension of the tool set. (R1: punch
radius=50mm, R2: die radius=51.25mm, R3: punch fillet=9.5mm, R4: die fillet
radius=7.0mm, R0: blank radius=105mm<DDQ>), b) initial mesh....................... 106
xv
Figure 5.3: Cylindrical cup drawing: Residual von Mises stress distribution with respect
to (a) isotropic hardening, (b) RIK hardening ........................................................ 108
Figure 5.4: Rotation angle (in degree) of the symmetry axes of anisotropy................... 109
Figure 5.5: Square cup drawing - Forming tool for the deep drawing of a square cup (a)
and the FE model (b)............................................................................................... 110
Figure 5.6: Square cup drawing – Show the difference of a) Major and minor strain
distribution along the diagonal direction, b) thickness distribution along the diagonal
direction with respect to hardening models.. .......................................................... 111
Figure 5.7: Square cup drawing – Punch force-travel shows different history according to
the hardening models. ............................................................................................. 112
Figure 5.8: Square cup drawing – Draw-in with respect to hardening models compared to
the averaged experimental results of NUMISHEET ’93. ....................................... 113
Figure 5.10: Stress history during the process for the 30° oriented specimens .............. 115
Figure 5.11: Springback shape of the 30° oriented specimen: (a) rear view, (b) side view
................................................................................................................................. 116
Figure 5.12: Von Mises stress distribution before springback: a) isotropic, b) ANK, c)
RIK hardening model. RIK shows symmetric distribution of stress while the other
two show anti-symmetric distributions which causes unexpected twisting model
after springback....................................................................................................... 117
Figure 5.13: Hydroforming – Dimensions of the tool and initial tube and loading
conditions of the pressure and displacements at both ends of the tube. ................. 118
Figure 5.16: Rotation angle of the symmetry axes of anisotropy. Rotational evolution of
the anisotropy is concentrated at the highest stress area......................................... 121
Figure 5.17: Reaction force at both edges when feed the material axially toward center in
the rate of displacement shown in Figure 5.13b. .................................................... 122
xvi
Figure 6.1: Stress flow measurements of each direction for Al-3%Cu precipitate
hardened material. Hardening of 45° shows more anisotropy than the RD and TD.
................................................................................................................................. 132
Figure 6.2: Neither conventional isotropic hardening (a) nor kinematic hardening (b)
with Barlat YLD96 anisotropic yield function does trace the measurements. ....... 133
Figure 6.3: Flow stresses of 45° for different backstress formulations. Those
formulations cannot trace the trend of anisotropic hardening shown in the
measurement. .......................................................................................................... 134
Figure 6.4: Flow stresses using the fourth-order tensor for precipitate inclusion without
rotation (a) and with rotation (b). The anisotropic hardening for 45° is captured by
rotation of backstress. ............................................................................................. 135
Figure 6.5: Earing pattern of cylindrical cup according to the hardening models: a)
Isotropic hardening only, b) Isotorpic + kinematic hardening, c) Isotropic +
Anisotropic kinematic hardeningm, d) Isotropic + Anisotropic kinematic hardening
with rotation. ........................................................................................................... 136
Figure 6.6: Residual stress distributions with respect to the hardening model: (a)
isotropic hardening, (b) anisotropic hardening with the fourth-order tensor, but no
rotation, (c) anisotropic hardening with the fourth-order tensor and rotation. ....... 137
xvii
CHAPTER 1
INTRODUCTION
This research is focused on describing the anisotropic material behavior where the
texture development is significant. There are many sources for developing texture, such
preferred orientation which can be defined from the orientation distribution function
(ODF) [2, 3]. A preferred orientation is developed in certain crystallographic planes that
The preferred direction can be related to the symmetry axes of anisotropy and the
reorientation of the symmetry axes have been determined experimentally by both micro-
been performed with the texture analysis using the X-ray pole figures to see the rotation
of the symmetry axes directly, while macro-mechanical experiments have been done by
fitting the yield stress of the pre-strained specimen to the symmetric anisotropy function.
The macro-mechanical experiment requires approximately 400 tensile tests for a material
[4].
The motivating experiments have been done by Boehler [5] and Kim and Yin [4].
1
shown in Figure 1.1(a) is prepared. The specimens are pre-strained up to 6%, and then
small specimens are cut at every 15° to the RD from the pre-strained specimen as shown
in Figure 1.1(b). They are stretched to measure the yield stresses and then fit the data of
the yield stresses to the symmetric yield function such as Hill’s ’48 [6]. The processes
for pre-strains of 6%, 14% and 36% are repeated. The angles used in here are defined
with the pre-straining direction and the orthogonal anisotropy axes in Figure 1.2. The
measured yield stresses are then fit to the curve shown in Figure 1.3, and the symmetry
axes of the curves can be set as shown with the blue lines in Figure 1.3. From the fitted
curve, one can observe that the symmetry axes are rotating toward the initial orientation
of 45°.
The micro-mechanical pole figures are shown in Figure 1.4. Interpreting the
figures, the symmetry axes originally located in a vertical direction rotate gradually, and
reach 45° at around 36% of strain. The rotation of the symmetry axes of anisotropy is
orientation and development of dislocation structures are known as the key features of
anisotropy evolution [29], as shown in Figure 1.5. These are different in scale but they
influence each other through slip processes. The preferred orientation is decided by the
slip process, and the slip process is defined by active slip system which lies closer to the
preferred orientation. The slip process influence the resistance against dislocation, again
the dislocation decides the slip process. In order to mimic the micro structural behavior
in phenomenological plasticity, the two key features should be modeled as in Figure 1.6.
assuming that the anisotropy axes are laid on the orthogonal axes of the yield function.
(to expand the yield surface), kinematic hardening (to move the yield surface), and
rotational hardening (to rotate the yield surface) as in Figure 1.7. The isotropic and
kinematic hardening descriptions have been developed by many researchers and can be
found in the literature. The rotational hardening is named for the reorientation of the
symmetry axes of anisotropy as assuming that the anisotropy axes are the basis of the
yield function, and the symmetry axes are rotating according to the reorientation of the
preferred direction.
The continuum setting for finite strain plasticity in intermediate configuration and
2. The setting is used basically in this research. The reorientation of the symmetry axes
using the plastic spin theory is also explained. The spin theory is combined with the
model, suggested by Chun et al. [7, 8], is combined with the rotational hardening theory.
The resulting RIK hardening model is introduced in Chapter 3. The relation of the
symmetry axes rotation with the loading path change is discussed. The capability of the
Experimental approach for the anisotropy evolution and the identification of the
material properties are introduced in Chapter 4. Mild steel (DDQ, DQ) and high strength
steel (HSS) have been tested. Mild steel has a high rate of rotation, while HSS does not.
Questions about defining ‘R-values’ are raised and a method to determine the rotation of
the anisotropy using simple tension tests are discussed. A linear relationship between the
3
plastic spin parameter and the difference of plane anisotropy along orientation, ∆R, is
found. A method of identifying the plastic spin parameter from the R-values is proposed
for the materials which have the B.C.C. crystal structure such as mild steel.
Numerical applications in sheet metal forming and spring back are compared with
are used. The RIK hardening model provided more reliable results than other
the RIK model in forming processes. The draw bead simulations in an arbitrary
properly by using existing models [9]. The preferred orientation of the precipitate
precipitate hardened material is decided by the precipitate. This can be treated in the
same manner as the rotation of the symmetry axes. With the hint, the anisotropic work
hardening for the aluminum alloy is investigated in Chapter 6. The material properties in
Barlat and Liu [10] are used, including Barlat 96 anisotropic yield function [11] and 3D
solid elements. The rotation of the symmetry axes is not limited to in-plane. As a
numerical application, a cylindrical cup drawing is demonstrated for the effect of the
4
(a) (b)
Figure 1.1: Macro-mechanical test: a) A big specimen is cut from a rolled sheet with an
orientation Ψ to the RD and is pre-stretched by ε*, b) smaller specimens are cut from the
pre-stretched specimen in every 15° from the RD and are stretched to measure yield
stress
5
Figure 1.2: Definition of angles used in this research. The symmetry axes ( eiφ ), initial
orientation to the RD ( ψ ), angle of the symmetry axes rotation ( η), and the angle
between the symmetry axes and the straining directioin ( ϑ ). Symmetry axes of
anisotropy rotates toward pre-stretched direction (ε*) from the RD and TD whichever
closer.
6
Figure 1.3: Rotation of the symmetry axes as the pre-strain increases: measured yield
stress (0.2% offset, ●) and curve fitting to Hill’s 48 yield function [6].
{1,0,0} Ψ= 45°
Figure 1.4: Rotation of the symmetry axes observed by using pole figures for 45° initial
orientation. The symmetry axis is rotating gradually from TD counterclockwise to 45° as
strain increases [5].
7
Figure 1.5: Treatment of anisotropy evolution in polycrystalline plasticity. Texture
orientation and substructures are considered to be the key features of anisotropy which
are related through slip processes. [29]
8
Figure 1.7: Schematic diagrams of hardening models: a) Isotropic hardening – expand
the yield surface uniformly, b) Kinematic hardening – move the yield surface to the
loading direction, c) Rotational hardening – rotate the yield surface with the rotation of
the symmetry axes of anisotropy, d) RIK hardening – combine all three hardening
definitions; the yield surface can grow, move, and rotate.
9
CHAPTER 2
2.1 Introduction
anisotropy axes are embedded in the material and thus rotate only if the material itself is
rotated. Induced deformations without material rotation do not change the direction of
the anisotropy axes, i.e., the deformation gradient is equal to the right or left stretch
tensor F = U = V , where the anisotropy axes are laid on symmetry planes of the yield
surface. Such a behavior can be verified experimentally in the rolling and transversal
directions of sheet metals. However, in uni-axial tension tests in other directions, e.g.,
30° to the RD, rotations of the anisotropy axes have been observed by several authors
using mechanical experiments and texture analysis, as discussed in Chapter 1. For mild
steels, rotations of up to 45° at 10% strain have been observed in these experiments,
related to the rotation of symmetry planes of the polycrystalline texture evolving with the
plastic deformation [5]. To model such a behavior, material formulations with a plastic
spin have been suggested by several investigators [12-14]. The attempt to relate the
material model with corotational shell elements is proposed by Han et al. [15], using an
isotropic hardening. The rotation rate of the anisotropy axes is supposed to be dependent
on materials because some aluminum sheets investigated in Bunge and Nielsen [3] and
Truong Qui and Lippmann [16] have much lower rotation rates of 5° at 20 percent strain.
10
The goal of this chapter is to summarize the continuum setting of the reorientation
finite strain. Also, the effect of the reorientation of the anisotropic axes in the plane
stress condition and detailed implicit algorithmic treatment for the rotation is discussed.
This model assumes that the anisotropy incurred in rolling processes is much larger than
the anisotropy due to subsequent strains by further forming. This assumption is used to
maintain the shape of the yield function that is defined as initial so the yield function can
only be rotated and expanded. The equations in the intermediate configuration are then
pushed forward into the current configuration using a small elastic deformation
assumption. It is generally accepted that elastic strains are negligible compare to plastic
rotations, finite strains, and assuming plane stress through the thickness direction.
The reorientations of the anisotropy axes with isotropic hardening are studied by
comparing the numerical results with the experimental data presented by Kim and Yin [4]
The material formulation for finite strain is generally done in the stress-free
where Q is an orthogonal tensor to represent the rigid body rotation. B0 and B are the
11
and B does the specific intermediate configuration which is used in this research. The
contains only small stretch ( Ve ) and all relative rotations are put in one rotation R * ,
deformation gradient
F = Fe Fp = Ve Fp . (2.2)
where the elastic and plastic parts in the current and the intermediate configuration are
defined by l e = Ve Ve−1 and L p = Fp Fp−1 , respectively. The plastic part in the intermediate
configuration can be split into the symmetric and the antisymmetric parts:
L p = Dp + Wp . (2.4)
(
Wp = ( L p ) = R *R*−1 ) + (R U U
* p
−1
p R *−1 ) . (2.5)
A A A
Wp = Θ + Ω p , (2.6)
12
where Θ is the constitutive spin for defining R * which is considered to be the anisotropy
Ω p = µ φ ( PDp − Dp P ) , (2.7)
where P = Ce S denotes the Mandel stress tensor defined by the elastic Cauchy-Green
tensor Ce = FeT Fe and the second Piola-Kirchhoff tensor S , defined in the intermediate
configuration.
The symmetric part of L p is formulated using the associated flow rule and
consistency relation
∂φ
Dp = γ . (2.8)
∂P
The assumption of a small elastic strain is used to transform those formulae to the
in Table 2.1.
ω p = µ φ ( τd p − d p τ ) , (2.9)
where τ is Kirchhoff stress. The anisotropy axes remain unchanged, i.e., eiφ = eiφ , for
small elastic strains as shown in Figure 2.1. The constitutive spin θ and the anisotropy
cφ
µφ = , (2.10)
φ
13
where cφ is material parameter and φ the yield flow stress. However, the definition is
insufficient to model different rates of rotation during deformation and different initial
orientations [14, 15]. The analytical solutions of Dafalias [14] and the FE model using
(2.10) are shown in Figure 2.2. They are not correlated with the experiments using a
single material parameter. Observing the experimental curve of the anisotropy axes
rotation, the rate of rotation is initially higher and it decreases gradually as the anisotropy
axes are rotated toward the straining direction. With this hint, a modified material
a
µφ = tan ( ϑ ) , (2.11)
φ
where a is material parameter and ϑ ∈ [0 o ,45 o ] describes the angle between the
shell formulations applying the plane stress assumption and assuming that the transversal
shear strains are small. The deformation tensor F is treated in a specific way for shell
formulations since the transversal shear components of F are selectively integrated and
do not enter the elasto-plastic material description (see Han et al. [15]).
s 2 ∂φ 2
γ= , where β = and s = dp , (2.12)
β 3 ∂τ 3
which is used for the description of the isotropic hardening and evolution of the yield
stress
14
iso
τ y = c iso (s o + s p ) n . (2.13)
.
The Oldroyd rates for stresses are given with (.) = (.)− l T (.) − (.)l which is identical to the
∇
∂
Lie-derivative Lv (.) = F (.) F T [19] and can be written in incremental form as
∂t
n +1
τ = F n τ F T + n +1∆τ , (2.14)
where the left superscript denotes the loading step: n is for current and n+1 is for next
loading step.
∆τ = Γ e (∆ε − ∆ε p ) , (2.15)
where ∆ε can be identified with the incremental Almansi strain tensor and Γ e is the
elasticity tensor.
The constitutive equations are performed with the plane stress condition in the
trial stresses, the plastic strain increments are assumed to be zero, yielding
trial
τ = U n τ U T + Γ e ∆ε . The Hill’s anisotropy yield function is applied and defined by
with the orthogonal axes as K αβγδeαφ ⊗ eβφ ⊗ eφγ ⊗ eδφ . It is convenient to convert the
15
symmetric tensor K into a matrix form P so that the plastic strain can be related to the
ε11
p
P11 P12 P13 τ11
p
ε 22 = P21 P22 P23 τ22 . (2.16)
p P33 τ12
2ε12 P31 P32
ε11p
K1111 K1112 K1121 K1122 τ11
p
ε12 K1211 K1212 K1221 K1222 τ12
p = (2.17)
ε 21 K 2111 K 2112 K 2121 K 2122 τ21
ε 22 K 2211 K 2222 τ22
p
K 2212 K 2221
By comparing (2.16) and (2.17), the matrix P can be described with the components of
1 −β12 0
0
P = −β12 β22 0 (2.20)
0 0 β66
16
where the three parameters (β12, β22, and β66) are defined in terms of the R-values:
R0 R (1 + R 90 ) ( R + R 90 )(1 + 2R 45 ) .
β12 = , β22 = 0 , β66 = 0 (2.21)
1+ R0 R 90 (1 + R 0 ) R 90 (1 + R 90 )
1 0 0 −β12
0 β66 / 4 β66 / 4 0
0
K mnpq = (2.22)
0 β66 / 4 β66 / 4 0
−β12 0 0 β22
Using such a vector notation, the algorithmic procedure from load step n to n+1 is
The trial orientation for the anisotropic yield function is computed with
trial φ
e = R n eφα . The directional evolution of eφα in plastic deformation is described as
α
n +1 φ
e = R*φ n eφα . The incremental orthogonal transformation R *φ is corresponding to the
α
R *φ = R Tωp R . (2.23)
Therefore the anisotropy axes direction in next step is directly related to the rotation
eφα i +1 = R ωT p trial φ
e .α (2.24)
( )
∆ω φp i +1 = µφ τ i +1∆ε p − ∆ε p τ i +1 with
17
cos ∆ωφp12i +1 sin ∆ωφp12i +1
R ωp = φ i +1 . (2.25)
− sin ∆ωp12 cos ∆ωφp12i +1
Updating the fourth-order anisotropic tensor K i +1 with the evolution of the symmetry
+1
K iαβγδ = ∑∑∑∑
m =1,2 n =1,2 p =1,2 q =1,2
0
( i +1
)( i +1
)( i +1
K mnpq eαφ ⋅ 0e m eβφ ⋅ 0e n eφγ ⋅ 0e p eφδ ⋅ 0eq , )( i +1
) (2.26)
where the superscript 0 means the initial loading step and initial direction of the
anisotropy axes. Therefore, e0m is (1,0) and (0,1) when e10 is along the RD. The matrix
i +1
K1111 K1122 2K1112
P = K 2211
i +1
K 2222 2K 2212 (2.27)
2K1211 2K1222 4K1212
stretch tests investigated by Dafalias [14] and Kim and Yin [4] are considered.
Rectangular samples are cut from a larger sheet in 30°, 45°, 60° angles from the rolling
direction (RD) and stretched along in these directions. An analytic solution for rigid-
plastic material, incorporating plastic spin and Hill’s yield function without hardening is
presented in Dafalias [14]. The material parameters used for the simulation with the
proposed algorithm are taken from the pre-stretched mild steel described in Kim and Yin
[4], and are given in Table 2.2. The results of this approach for the tensile stretch test in
the directions of 30°, 45° and 60° angles are presented in Figure 2.3 using a single
18
In Figure 2.4, the flow stresses obtained with the plastic spin are illustrated. The
flow stresses of 0°, 90° and 45° angles are illustrated as results of the FE analysis. The
flow stress of the 45° angle approaches the flow stress of the 90° angle as the symmetry
axes of anisotropy rotates. The phenomenon incurred by adopting the plastic spin will be
19
Intermediate Current
Configuration Configuration
L = Le + L p ≅ L p l = Fe Fe−1 + Fe Fp Fp−1Fe−1
= l e + Fe L p Fe−1
= l e + ˆl p
D = ( L )S d = ( l )S
= ( Lp ) = ( Dp + Wp )
S S
( )
= ( l e )S + ˆl p
S
= Dp = ( l e )S + ( Fe Dp Fe−1 ) + ( Fe Wp Fe−1 )
S S
−1
= d e + Ve Dp V e
= d e + dˆ p ≅ d e + d p
W = ( L p ) = ( Fp Fp−1 )
A A
w = ( l )A = ( l e )A + ˆl p ( ) A
= ( R*R −1
* A ) + (R U U
* p
−1
p R )
−1
* A (
= 0 + Ve Dp V ) + (V W V )
e
−1
e p e
−1
A A
= Θp + Ωp (
= VeΘ p Ve−1 ) + (V Ω V )e p e
−1
= θp + ωp
E = Ee + E p e = ee + ep
1 1
=
1 T
2
( 1
) (
Fe Fe − 1 + 1 − Fp− T Fp−1
2
) =
2
( ) (
1 − Ve−1Ve−1 + Ve−1Ve−1 − F − T F −1
2
)
Table 2.1: Comparison of continuum setting in the intermediate configuration and the
( ε e ) ) is used.
current configuration. Small elastic strain assumption ( Ve = I +O
20
Young’s modulus E = 206 GPa
Poisson’s ratio ν = 0.3
Initial yield stress τ0 = 107.06 MPa
Hill’s yield function β12 = 0.5837, β22 = 1.0092, β66 = 2.3550
Isotropic hardening ciso = 544 MPa, niso = 0.25
21
n
Input : F n +1 , τ n , s np , e φα
Compute: ∆ε , trial
τ , trial e φα , trial Φ = Φ ( trial τ, trial e φα , τ yn ) , trial
P
if Φ < tol then
n +1
τ n +1 = trial τ , s np +1 = s np , e φα = trial φ
e , Γ ep = Γ e
α
if Φ ≥ tol then
τ 0 = trial τ, s 0p = s pn , Φ 0 = trial Φ, P 0 = trial P
∆γ 0 = 0
do i = 0, imax
compute
• a = 2P τ , rτ = τ −
i i i
( trial
τ − ∆γ i Γ e a )
• (
Q = 1 + 2∆γ i Γ e P i )
Φ − aT Q −1rτ
• ∆∆γ =
aT Q −1Γ ea + 4H ' τ iT P i τ ( )
with H ′ = n iso c iso (s 0 + s p )
iso
n −1
• s ip+1 = s ip + ∆s p
with ∆s p = 2∆γ i τ iT P i τ i
• τ i +1 = τ i + ∆τ
with ∆τ = −Q −1 (rτ + ∆∆γΓ e a )
• ∆γ i +1 = ∆γ i + ∆∆γ
i +1
• eφα = R ωT p trial φ
e
α
)
n +1 i +1
τ n + 1 = τ i + 1, s np + 1 = s ip+ 1 , e φα = e φα
exit to Output
end if
end do
end if
n +1
Output : Γ ep , τ n +1 , s np +1 , e φα
22
Figure 2.1: The configurations and relationship between the configurations: reference
( B0 ), intermediate ( B, B ), and current ( B ) configurations. A specific intermediate
configuration is chosen by leaving only Ve in the elastic deformation and all relative
rotations are put in R*, as F = Ve R*U p .
23
40
ο
Ψ = 30
35 φ
c = -300
30
ο
η
25 φ
c = -100
Rotation angle 20 φ
c = -200
15
Exp. (Kim & Yin)
10
FEM
5 Analytical Solution
(Dafalias)
0
0 2 4 6 8 10
Strain (%)
(a)
0
ο
Ψ =45
-10
φ
c = -100
ο
η
-20
Rotation angle
φ
c = -200
φ
-30 c = -300
-40
-50
0 2 4 6 8 10
Strain (%)
(b)
(continued)
Figure 2.2: Rotation angle of the symmetry axes using analytical solution and FE
analysis: (a) Ψ=30°, (b) Ψ=45°, (c) Ψ=60°.
24
(Figure 2.2 continued)
0
ο
Ψ = 60
-5
-10
φ
ο
c = -100
η
-15
Rotation angle
φ
-20 c = -200
-25
-30 φ
c = -300
-35
-40
0 2 4 6 8 10
Strain (%)
(c)
25
40
ο
Ψ = 30
35
30 FEM
ο
η
25
Rotation angle
20
15
10 EXP
(Kim & Yin)
5
0
0 2 4 6 8 10
Strain (%)
(a)
0
ο
Ψ = 45
-10
ο
η
-20
Rotation angle
-30 FEM
-40
EXP
(Kim & Yin)
-50
0 2 4 6 8 10
Strain (%)
(b)
(continued)
Figure 2.3: Rotation angle of the symmetry axes using modified plastic spin scalar (a = -
350): (a) Ψ=30°, (b) Ψ=45°, (c) Ψ=60
26
(Figure 2.3 continued)
0
ο
Ψ = 60
-5
-10
ο
η
-15
Rotation angle
-20 FEM
-25
-30
EXP
(Kim & Yin)
-35
-40
0 2 4 6 8 10
Strain (%)
(c)
27
400
o
0
o
300 45
Stress (MPa)
o
90
200
100
0
0 5 10 15 20
Strain (%)
Figure 2.4: Flow stresses for the RD, the TD, and 45°. The 45° flow stress approaches
the flow stress of the TD when the rotation of symmetry axes is used with an isotropic
hardening.
28
CHAPTER 3
The behaviors of the material points under multi-axial and multi-path loadings are
not modeled accurately by phenomenological models. With the help of the advances in
loadings has been being unveiled in polycrystalline plasticity. However, the formulation
problems, even though it has accuracy in describing the material behavior. Motivated by
behaviors to a phenomenological model for the computation. The proposed RIK model
allows the yield function to grow (isotropic hardening), move (kinematic hardening), and
The RIK model has been formulated based on the Armstrong-Frederick type
kinematic hardening model and on the rotation of the symmetry axes of anisotropy with
the plastic spin theory. The model is accounting for the different backstress evolutions
for loading and reversal loading with respect to the difference of the loading direction for
29
the multi-axial loading and correcting the kinematic hardening with respect to the
reorientation of the symmetry axes. The ability of the model is demonstrated by the
‘Bauschinger’ effects. The effects of the RIK model in deformation and springback are
3.1 Introduction.
The goal of this article is to describe the material behavior under multi-axial/-path
loadings. Tensile loading in the transverse direction (TD) after pre-straining in the
rolling direction (RD) was done by Doucet and Wagoner in plane stress condition [20,
21]. Simple shear test after various amounts of pre-strain was investigated by Rauch [22]
and shear test with various orientation angles to the RD, after pre-strain in the RD was
done by Thuillier and Rauch [23]. Hiwatashi et al. [24, 25] suggested a non-proportional
biaxial loading experiment and the material model for the loading. Takahasi et al. [26]
Teodosiu and Hu [27] suggested three simple tests: (case a), a simple shear test with the
shearing direction (SD) parallel to RD and the shear plane normal parallel to the TD;
(case b), tension-shear test, pre-tensile strain in the RD, followed by simple shear with
the SD also parallel to the RD; and (case c), a reverse-shear test, an amount of simple
shear in the RD, and then simple shear in the opposite direction. The schematic diagrams
of the tests are shown in Figure 3.1 with the corresponding flow stress results. The
experiments show two major effects of multi-path loading, ‘crossing’ and ‘Bauschinger’
30
abrupt change of the strain path causes the ‘crossing’ of the flow stress, and it gradually
approaches to the monotonic shear curve. The ‘Bauschinger’ effect and the permanent
However, these approaches remain impractical for use with large-scale boundary value
problems such as those encountered in metal forming or other structural problems. The
principal obstacle is the inability to represent the essential aspects of texture-based results
literature. Hiwatashi et al. [24] approached this strain-path change problem with the
With the investigation of Peeters et al. [28, 29], the micro-structural developments
are mainly dislocation structures and reorientation of the preferred direction. The
kinematic hardening, while the reorientation of the preferred direction can be considered
to be the rotation of the symmetry axes, which is related with the plastic spin [5, 15, 31-
non-linear kinematic hardening models [34] have been used to capture the nonlinear
hardening behavior and smooth transition from elastic to plastic deformation relatively
well. However, the reversal loading of the models approaches the monotonic hardening
curve and does not show the permanent softening characteristics. Chun et al. [7]
evolution of the bounding surface to reproduce the permanent offset, and also suggested
31
an advanced form of isotropic hardening to represent the uniaxial tensile experiment with
the conventional isotropic, kinematic hardening and Chun’s ANK hardening model. The
results of isotropic hardening and the ANK hardening are shown in Figure 3.2. The
‘Bauschinger’ effect and permanent offset are captured by the ANK model, but the
‘crossing’ has not been captured with those hardening models. The capability of
reproducing the characteristics with respect to the hardening models is compared in Table
3.1. The existing models cannot reproduce the ‘crossing’ effect, even though the latest
ANK model can account for the directional effect between loading and reloading and can
capture the permanent offset. It is because the models do not have the ability to account
formulated with the rotation of the symmetry axes of anisotropy and the formulation has
been introduced by the authors using the plastic spin theory [15]. The rotation of the
symmetry axes have been investigated experimentally by Boehler [5] and Kim and Yin
[4] for mild steel, and by Bunge [3] and Truong Qui [35] for aluminum alloy. Macro-
mechanical modeling of the plastic spin is done by several authors, e.g., Dafalias [14] and
Kuroda [13]. The reorientation of the anisotropy axes affects the yield function, and the
rotated yield function makes different evolutions in flow stress. Therefore, the anisotropy
evolution due to the rotation of the yield function can be called ‘rotational hardening’
can handle the rotation, translation, and expansion of the yield function simultaneously as
32
motivated by the micro-mechanical behavior of the material. This research will also
define the relation of the backstress and the rotational hardening in correcting the flow
The results of the new RIK hardening model will be compared qualitatively with
the experimental results of Peeters [28]. The comparison will show that the suggested
framework.
The model is derived in general form, and then formulated in planar anisotropic
and a time-independent plane stress problem specifically for the sheet metal forming
process. Demonstrating the flow curves in the RD, TD and 45° all follow the
effect after various pre-strain also demonstrate the effect of pre-straining. The
formulation and numerical algorithms are done by the return mapping method and are
material subroutines.
where the elastic and plastic parts in the intermediate and current configuration are
elastic strain assumption, the symmetric part of l can be decomposed into an elastic and
( l )S = d = de + d p . (3.2)
part and eφα is the direction vector of the symmetry axes of the anisotropy which is
anisotropic yield function such as Hill’s [6] and Barlat’s [11], etc. The backstress y in
y = ∑qj . (3.4)
j
Those stress-like-internal variables are related to their work conjugated strain like
variables a j , s as
q j = −c ja j , σ y = σ y ( s ) (3.5)
where c j is a positive definite anisotropy tensor and it can be tensor, vector, and scalar
Following Dafalias [14, 33], the anti-symmetric part can be decomposed into the
34
( l )A = w = θ + ω p , (3.6)
and the orientation of the anisotropic yield function is presumed to be defined by the
anisotropy axes eφα . The anisotropy axes eφα are related to the constitutive spin θ by a co-
rotational rate
o
eφα = eφα − θeφα = 0 . (3.7)
The anisotropy axes rotation does not evolve internal stress or strains, but the flow stress
can change the orientation of the yield function. Therefore, it can be called “rotational
hardening” [36].
The constitutive relation for plasticity is defined by the associated flow rule as
∂φ
dp = γ (3.8)
∂τ
s
and the consistent parameter γ can be described as γ = where s = 2
3 d p and
β
∂φ
β= 2
3 ∂τ .
motivated by considerations on smaller scales. The anisotropy axes may affect the other
variables and the other state variables may affect the anisotropy axes.
τ = f τ ( d e , d p , eαφ )
q = f q ( d p , s, eφα )
(3.9)
σ y = f σy ( d e , d p , s, eφα )
eφα = f e ( τ, q, d p , σ y )
35
The relationship between the isotropic hardening and the kinematic hardening can
with isotropic hardening was suggested by Chaboche [38, 39] and modified by Chun et al.
[7]. The relationship between the rotational hardening and the isotropic hardening was
investigated by Dafalias [14, 33], [13] and Han et al. [15]. Loret [40], Dafalias [17, 41,
42], Van der Gissen [43] suggest the plastic spin formulation for kinematic hardening.
Kuroda [13, 18, 44] and Dafalias [14] define a simple form of the plastic spin by
defining the plastic spin with the non-coaxiality of the stress and the plastic strain
increment direction. The relationship between the isotropic hardening, the kinematic
Dafalias [14], Kuroda [13, 18, 44] and Zbib [45] define the plastic spin in the
current configuration as
ω p = µφ ( τd p − d p τ ) or (3.10)
ω p = µ φ {( τ − y ) d p − d p ( τ − y )} . (3.11)
From the observation of the experimental results of Kim and Yin [4], the rate of the
rotation is initially faster and then decreases as the symmetry axes rotate toward the
function of the angle between the anisotropy axes direction and the straining direction [15,
46],
a
µφ = µφ ( ϑ) = tan ( ϑ ) , (3.12)
φ
36
φ
−1 e α ⋅ n π
ϑ = min cos φ
, ϑ∈ 0, (3.13)
α=1,2,3
eα n 4
The vector n in (3.13) is the eigenvector of d p . For the plane stress setting such as sheet
Detailed formulation and the numerical algorithm are presented in Chapter 2 with
the isotropic hardening model and compared with Kim and Yin’s experimental results [4].
Choi et al. [46] showed the effect of the anisotropy axes rotation on springback.
From an analogy to the plastic strain rate, the evolution of the strain-like internal
a ∆j = sb ja j − d p (3.14)
Following the derivative of the stress-like internal variable q j following Haupt and
q∇j = −c ja ∆j (3.15)
q∇j = c jd p − sb jq j (3.16)
The evolution of the backstress in (3.16) would have variation according to the
material concerned. The variable c j in (3.16) can be a constant, scalar function for multi-
axial loading, and a tensor which describes the accommodation tensor of the material for
elastic inclusion of the precipitate [48] and also for texture development.
37
Frederick type hardening from the intermediate configuration can be found in the
In order to capture the Bauschinger effect with permanent offset in the planar
backstress component has been introduced by Chun et al. [7] and Geng and Wagoner
[51]. The model proposed by Chun et al. has two backstress components and the first
The evolution equations of the first component are described in the Armstrong-
Frederick type nonlinear hardening, and the second component could be a simpler linear
form for springback prediction is suggested when a cyclic loading is expected during the
The parameter φ denotes the angle between the loading and reloading,
d p ⋅ d*p
ϕ = cos −1 , (3.18)
d p d*p
where, d p is for the current loading, while d*p is for the previous loading step.
The second backstress component is evolving with respect to the loading path
history. Once the loading path is in the reverse direction, the second backstress evolves
38
in a different scheme. The evolution of the second backstress is then considered as a
function of the angle between the previous and current loading paths. The angle can be
considered to be the difference between the directions of the symmetry axes and the
loading direction. Because the symmetry axes of anisotropy is rotating toward the
loading direction, the symmetry axes direction of the previous loading path is the
reference to measure the directional change of the current loading path. The evolution of
3.2.5 Correction of flow stress with kinematic hardening associated with the anisotropy
axes rotation
material is under plastic loading, each grain has micro-mechanical structures, such as slip
planes in its own direction. The directions are represented as the preferred direction of
the aggregates or anisotropy axes which can be found statistically by the orientation
distribution function (ODF). The mechanism of deciding the direction of slip planes of
assumed to be maintained as long as the external loading direction is not changed. When
the loading path is changed, new slip planes of each grain develop in a different direction
according to the new loading direction. The mechanism is not well known and the
Now consider the interpretation of Peeters et al. [28,29]. Once the slip planes are
developed in a direction according to the previous loading path, the existing slip planes
act as obstacles to the newly developing dislocation according to the current loading.
The dislocation density would increase at the ‘old’ slip planes and the resistance would
39
increase. As the loading is continuous in the new direction, the micro-band occurs as the
resistance is over the theoretical shear stress of the ‘old’ slip planes in the region and they
cut the ‘old’ slip planes. More micro-band occurs as the loading continues, and gradually
the broken structures of the old slip planes are aligned to the new direction.
This procedure looks as if the slip planes are rotated according to the new loading
path. The resistance due to the pile up at the old slip planes decreases as the obstacles
rotate toward a new loading direction. The direction of each grain defines the preferred
behavior of the aggregate can be formulated with the rotation of the symmetry axes of the
anisotropy.
If the loading is in the RD, then the initial direction of the symmetry axes are
aligned with the loading path. Hence the angle ϑ is zero. Therefore, the second
backstress evolution in (3.17) will be c2. When the loading path is changed to the
opposite direction, it is reversal loading, but the angle between the loading direction and
the symmetry axes direction will still be zero. Therefore, the second backstress evolution
is zero. This is identical to the ANK hardening model as long as the loading and
reloading are in the RD. In contrast to the ANK model, the RIK model can account for
rotation makes flow stress in any orientation approaching to the RD or the TD flow stress
rapidly and following the RD or the TD flow stress afterwards as shown in Figure 2.4.
However, the flow curves in the experiment do not approach the RD or the TD.
Therefore, it is necessary to correct the flow stress as long as the rotational hardening is
used. The mechanism of the hardening models with the rotational effect of the yield
surface is illustrated in Figure 3.3. Yield function will grow from A to C when only the
40
isotropic hardening is used. If the rotational hardening is added, the yield surface will
rotate and grow to a stable condition. And then the stress that satisfies the yield function
is iterated from A to B, not C. Once the yield surface is rotated and the direction of the
symmetry axes and the direction of loading are aligned, the yield surface is not rotated
any further. The hardening model in this article does not change the yield surface shape,
but only changes the size, location, and direction. This is the explanation for why all of
the flow curves are approaching the RD or the TD after the anisotropy axes rotate to the
loading direction, and it could be corrected with kinematic hardening. The distance of
the two stress states B and C would be recovered by the associated backstress, α , and it
The formulations are now applied to the plane stress problem with shell element
to apply the theory to sheet metal forming process. The quadratic yield function is
Φ = φ − σ 2y , (3.19)
where the components are related to the R-values as [52] and [7]
R0 R (1 + R 90 ) ( R + R 90 )(1 + 2R 45 ) .
β12 = , β22 = 0 , β66 = 0 (3.21)
1+ R0 R 90 (1 + R 0 ) R 90 (1 + R 0 )
41
In order to avoid unwanted inconsistency for unidirectional loading with the
s
q1∇ = c1 ( τ − y ) − sb1q1
β
(3.22)
s
q∇2 = kc 2 ( τ − y )
β
The yield function is completed by defining the size of yield function q in (3.3),
which is modified from the Chaboche [39] to follow the simple tension test curve by
c1
(
σ y = σo + K 1 − e − Ns − ) b1
( )
1 − e − b1s − c 2s . (3.23)
t +∆t
eiφ = t +∆t R φ t eiφ . (3.24)
In the case of a purely elastic increment, the plastic spin is identical to the zero
tensor K would rotate along the basis, which is rotated with the symmetry axes of
The formulation and numerical algorithms are completed with a return mapping
interface material subroutines. The constitutive equations used in this paper are
The developed model is applied for several loading cases to qualitatively show
that it can reproduce the multi-axial/-path loading responses. The model is formulated
and the algorithm is developed using the return mapping method. A detailed algorithmic
approach of the model is described in the appendix. In order to see the effect of the
model for controlled loading, uniaxial stretch and multi-axial loading problems are
The material concerned in this paper is mild steel. The material properties for
DDQ steel are determined from the series of experiments to identify the anisotropy and
plastic spin parameters. Using the RIK theory, the parameters can be limited to a for
plastic spin, c1, c2, and bl for kinematic hardening, and κ for defining the scale of the
Chapter 4. Zhao et al. [53, 54] and Chun et al. [7] introduced the procedures to identify
the kinematic hardening parameters by an inverse method. The material parameters used
Uniaxial tension test: The uniaxial tension test is done to identify the material parameters
for the initial anisotropy and for isotropic hardening. The computed flow stresses along
43
the RD, TD and 45° are compared with measured data in Figure 3.4. The lines represent
the computed results and the symbols for measured data. In contrast to the RD and TD,
the stress evolves higher than that of the experiment in 45° orientation when the latest
kinematic hardening model (ANK) is used, as shown in Figure 3.4(a). Although it is not
shown in this diagram, the isotropic hardening model gives similar results. However, by
using the rotational hardening scheme the flow stress in 45° shows a good correlation
with measured data. The RD and TD curves are not affected with the rotational
hardening, because those directions lie in the symmetry direction of the anisotropy.
because the anisotropy is evolving. The R-values may be different with respect to the
longitudinal straining of the specimen, so defining the R-value measured at low strain as
the initial R-value. In contrast to that, the R-value up to 25% strain is defined as the
averaged R-value. In order to see the effect of the hardening models and anisotropy, it is
necessary to compute the 45° oriented tensile experiment with various combinations of
The resulting flow stresses of the two R-values are rather higher than the
experimental results when use the ANK model. If the rotational hardening scheme is
applied, the R-value is changing according to the evolution of the anisotropy, and the
flow curve is decreased. Without the correction of the backstress associated with the
rotation of the symmetry axes, the flow curve is converging to the RD or TD, and in this
case it is converged to the TD (the TD flow curve is omitted here to compare with 45°
clearly.) The difference between the experimental curve and the computed curve by
using rotational scheme only is the amount of the backstress evolution, which is
associated with the rotational hardening. Therefore, the hardening curve is fitted with the
Offset test: The purpose of this test is to see the effect of anisotropy evolution in a
different loading direction, and see the offset stress as we choose the plastic spin
parameter a. The offset amount of stress due to the Bauschinger effect is measured from
the numerical simulation for a single element. The tests are done in three steps. To make
equivalent plastic strain, and then release from the loading. The second loading is in a
different path which has a certain angle from the first loading. Measure the equivalent
stress at 13% equivalent plastic strain for comparison. In this multi-step loading
condition, the stress in the reloading stage is dependent on the different evolution of the
second backstress component. The loading sequence is illustrated in Figure 3.7(a), and
the response of the offset for the reloading stage with respect to the loading angle is
shown in Figure 3.7(b). The stress offsets are normalized with respect to the offset at
180° (full reversal loading). The stress offset is responses to the rate of rotation of the
symmetry axes.
Cross and Bauschinger effect: The loading conditions shown in Figure 3.1were used to
demonstrate the ability of the RIK hardening model in capturing the response of the
multi-path loading. It is already discussed in the introduction that neither the isotropic
hardening model nor the isotropic + kinematic hardening model can reproduce the key
The same model is run with the rotational hardening scheme with and without the
45
kinematic hardening model. With these results, the anisotropy axes rotation is in charge
of the “cross” effect, and the kinematic hardening is of the “Bauschinger” effect.
The tension shear tests with different pre-strains are simulated for 3%, 7%, and
10% strain. The flow stress curves are illustrated in Figure 3.10. At every pre-strain, the
curve crosses and approaches the simple shear curve. As larger pre-strain, the amount of
stress “cross” gets bigger. Dislocation structures are developed more in the direction of
the dislocation structure which is previously developed due to the rolling process, and
then the effect of resistance to the dislocation structures in a different directional loading
gets bigger [29]. More texture develops in the tensile direction as the tensile strain
3.5 Conclusions
By combining the rotational hardening with the isotropic hardening and with the
framework.
- capture the flow stress in 45° in correlating with the experimental results,
and
The results show that the suggested model can reproduce the “cross effect” and
the “Bauschinger effect” clearly. The anisotropy evolution with the anisotropy axes
rotation due to the plastic spin is responsible for the “cross effect.” Even if the rotational
hardening can change the axes of orientation, it cannot change the yield surface shape
46
itself. The yield surface shape in this model can grow, move, and rotate. The model
could be better if we considered the yield surface shape evolution by relating the yield
47
Model Cross Bauschinger Permanent Offset
Isotropic No No No
Kinematic No Yes No
ANK No Yes Yes
∂φ
dp = γ
∂τ
s s
y = ∑ q j ; q1 = c1 ( τ − y ) − sb1q1 ; q 2 = kc2 ( τ − y )
j β β
c
( ) ( )
σ y = σo + K 1 − e − Ns − 1 1 − e − b1s − c 2s
b1
(
ωp = µφ τ d p − d p τ )
eφα = θeφα
48
Young’s Modulus, E 210 Gpa
Poisson’s Ratio, ν 0.3
Initial yield stress, τ 0 152.0 Mpa
Isotropic Hardening parameters; K, N 235.81 Mpa, 9.23
Initial anisotropy; R0,R45,R90 2.64, 1.47, 2.17
Backstress parameter; c1,b1,c2,κ 3.0GPa, 300.0, 70.0, 1.8
Plastic spin parameter, µ φ -57.0
49
σ12
∆c
∆b
εp
50
300
250
100
50
Reverse-shear
0
0 0.04 0.08 0.12 0.16 0.2
Eqv. plastic strain
(a)
300
200 Tension-shear
150
100
50
Reverse-shear
0
0 0.05 0.1 0.15 0.2
Eqv. plastic strain
(b)
Figure 3.2: Multi-path loading response of conventional hardening models: (a) isotropic
hardening, (b) anisotropic nonlinear kinematic hardening model. Both models cannot
reproduce the ‘crossing’ effect while the Bauschinger effect is caputred in (b).
51
moved yield surface
α
backstress
Figure 3.3: Rotational evolution of yield function: The yield function is expanded to the
stress state C when only the isotropic hardening is considered. The yield surface is rotate
by using rotational hardening and the iterated stress state is B. The yield surface is
moved toward C with the amount of backstress by the kinematic hardening.
52
500
400
EXP ANK
200 RD
o
45
TD
100
0 0.05 0.1 0.15 0.2 0.25
Eqv. Plastic strain
(a)
500
400
True stress (MPa)
300
EXP RIK
200 RD
o
45
TD
100
0 0.05 0.1 0.15 0.2 0.25
Eqv. Plastic strain
(b)
Figure 3.4: Flow stress in the RD, 45°, and the TD orientation with a) ANK, b) RIK.
ANK predicts higher flow stress for 45° while RIK correlates the measurement data.
53
500
(R )
45 init
(R )
400 45 av g
True stress (MPa)
300 RIK
Rotation + Isotropic
200
EXP
100
0 0.05 0.1 0.15 0.2 0.25
Eqv. Plastic strain
Figure 3.5: Flow curves for the 45° specimen: isotropic+kinematic hardening model
(ANK) produces rather higher stress ( ( R 45 )avg , ( R 45 )init ), rotation+isotropic hardening
model produces lower stress, while the RIK follows the measured data rather closely.
54
0
-10
Rotation angle (η)
-20
-30
-40
-50
0 0.05 0.1 0.15 0.2 0.25
Eqv. Plastic strain
Figure 3.6: Rotation angle (degree) of the symmetry axes for 45° oriented specimen.
55
(a)
1.2
1
a=-100
0.8
Normalized stress
0.6
0.4
a=-350
0.2
-0.2
0 60 120 180 240 300 360
Loading angle
(b)
Figure 3.7: Stress offset tests for loading-unloading-reloading in different path: (a)
loading condition – give different loading path to the pre-stretched specimen, (b) stress
offset with respect to the plastic spin parameter. The stress is normalized with respect to
the offset of 180º ( ∆τϕ / ∆τ180o )
56
300
250
150
Simple shear
100
50
0
0 0.05 0.1 0.15 0.2
Accum. Plastic strain
Figure 3.8: Flow stress response when using the rotational hardening with isotropic
hardening scheme. Stress jump (crossing) is shown but Bauschinger effect is not captured.
300
250
Shear stress (MPa) |σ 12 |
Tension-shear
Simple shear
200
150
Reverse-shear
100
50
0
0 0.05 0.1 0.15 0.2
Accum. Plastic strain
Figure 3.9: Flow stress result of the multi-path loading when the RIK hardening model is
used. Both ‘cross’ and ‘Bauschinger’ effects are captured.
57
300
250
Shear stess (MPa)
200
150
100 3%
7%
10%
50
0
0 0.05 0.1 0.15 0.2 0.25
Accm. Eqv. Plastic strani
Figure 3.10: Multi-path loading responses show the flow stress crossing after various
pre-strains; more pre-strain yields higher jump in stress.
58
CHAPTER 4
Specimen orientation is specified with respect to the rolling direction (RD). The
R-value is defined as the ratio of the transverse strain versus thickness strain at a certain
elongation, and the R-values are computed using the slope of the transverse strain versus
longitudinal strain curves. The R-values are measured from the experiment, but because
of the evolution of anisotropy, the hardening behavior may not be predictable for
orientations that are not in the RD or TD. With the rotational evolution of the anisotropy,
the slope of the strain-strain curve changes continuously. Anisotropy evolution for large
strains can be observed in simple uniaxial tension tests when measuring the transverse
and longitudinal strain continuously with the Digital Image Correlation (DIC) method.
model is investigated. With the RIK hardening model and its ability to represent the
rotational evolution of the anisotropy, the hardening behavior is predictable for non-RD
59
and non-TD directions. Methods to identify the plastic spin parameter and kinematic
4.1 Introduction
material anisotropy is important for making accurate predictions. Cold rolled sheet
metals have initial anisotropy which is incurred during the rolling process. The
anisotropy causes different flow stress with respect to the orientation angle that is
measured from the rolling direction (RD). The different flow stresses have significant
affects on the forming process (e.g., different earing of edges, punch force-displacements,
functions such as those of Hill [6, 55] and Barlat [11, 56] have been developed. Hill’s
quadratic yield function [6] is known to be appropriate for BCC materials such as mild
steel, while Barlat’s yield function [11] is good for FCC materials such as aluminum
alloys. However, these yield functions do not account for the evolution of material
The evolution of the anisotropy of materials has been observed by both micro-
development which changes the preferred orientation of a grain aggregate. Using micro-
[57], Asaro [31], Beaudoin [58], Kocks [59] and Nakamachi [60]. On the other hand,
Boehler [5], Kim and Yin [4], Kuroda [13], Dafalias [14] and Han et al. [15] have
observed and modeled anisotropy evolution using macro-mechanical approaches that rely
on phenomenological descriptions.
60
For sheet metals, R-values are provided to represent the planar anisotropy and to
compare the anisotropy for different materials [2]. The R-value is defined as the ratio of
transverse strain to thickness strain at a certain longitudinal strain. Usually R-values are
provided with the longitudinal strain at which the transverse strains are measured because
the measured R-values are different with respect to the longitudinal strain.
Even if the R-values are used to define the anisotropic relationship between a
certain direction and the RD and the experimental hardening curve for the RD is used
with Hill’s 48 anisotropic yield function, the computed hardening curve may not properly
model the experiment results for orientations not in the RD or TD. The computed
hardening curves in various directions are compared with the experimental results in
Figure 4.1. The curves of the RD and TD correlate quite well with the experimental
results. However, the computed 45° orientation curve overestimates the experimentally
observed result. The computed curves were generated by ABAQUS standard using the
Hill’s quadratic yield function and isotropic hardening model. The experimental results
are provided from the Numisheet 2002 benchmark problem set [61]. The material was
yield criterion has major drawbacks in reproducing flow stress curves that depend on the
loading path. In other words, the dependence of the yield stress on direction is poorly
predicted by the theory [2]. These shortcomings are usually interpreted to result from the
tensile tests for RD, TD and 45° specimens were repeated on different materials. The
digital image correlation (DIC) method was used to measure the strain continuously up to
large strains.
61
The discussion will focus on the hardening model which can properly describe the
flow stresses. The R-values are effective for characterizing the RD and TD, however, it
is not capable of predictive modeling for the 45° orientation. It is assumed that the
unexpected higher estimation of the stress evolution in 45° in Figure 4.1 is caused by the
evolution of the anisotropy and that the R values can be used to determine the ‘initial’
A simple method to measure the rotation of the symmetry axes is suggested with
the assumption that the anisotropy evolution is dominant to the rotation of the symmetry
measurement of R-values for specimens cut in 45° to the RD. According to the theory,
the symmetry axes do not rotate for the deformation along the RD or TD. To measure the
strain continuously up to large strains, the Digital Image Correlation (DIC) method is
used. The DIC method was developed by Sutton et al. [62] and improved by Vendroux
and Knauss [63], although it lacks accuracy at small strains, it can be used at large strains
and is simple to implement. Although Rao et al. [64] measured large strains using digital
images of grid marks on uniaxial sheet metal specimens, the grid deformation was not
used to determine R-values. The authors are not aware of any other investigations that
62
This paper presents a DIC-based experimental method for determining anisotropy
evolution and demonstrates how the material parameters for the RIK hardening model
The definition of the R-value for a planar anisotropic material is the ratio of
εw
R= . (4.1)
εt
Because the specimen thickness is generally very small, the thickness strain ε t for sheet
metal is very difficult to measure accurately. Therefore, the constant volume condition in
εw εw
R= =− , (4.2)
εt ( εl + ε w )
which is widely used to measure R-values. Then the R-values are defined by a linear
fitting of the slope of the ε w − εl (transverse strain versus longitudinal strain) curve in
Figure 4.2.
The R-value from the ε w − ε l for the RD is R90, for 45° is R45, and for the TD is
R0, respectively. The R-values are shown in Figure 4.3. The similar characteristics of
the R-values had been measured on a cube of copper with a rolling texture as shown in
As assuming as the anisotropy axes are not rotating when the loading is along the
RD and TD, the R0 and R90 are computed by linear fit through whole strain. Anisotropy
63
evolution is significant in 45°, so the strain curve requires non-linear function to be fitted.
ε w = {a ln ( εl ) + b} εl . (4.3)
with an Instron 8500 digital controller. The controller was operated in displacement
control mode. Load was measured from a 25 MT load cell and was subsequently
converted to true stress. For some experiments, strain gages were used to assure the
accuracy of the strain measurement by the DIC. The strain gages were connected to a
Measurements Group 2100 strain gage conditioner. The strain conditioner output and the
load signal from the Instron controller were recorded by with National Instruments SCXI
Specimens used in the experiments were cut from sheet metal with an angle to the
RD direction of 0°, 45°, and 90°. Specimens were 1 inch wide and 0.039 inch (1mm)
thick. Since the field of view of the CCD camera is fixed and only the lower grip of the
load frame moves, the region of interest has a significant downward displacement. The
size of the specimen must be carefully chosen to ensure that enough of the originally
undeformed image remains inside the camera’s field of view for all subsequent
deformation. For this reason, 1.5 inch gage length specimens were used. FEM
simulations were performed to confirm that the region where strains were measured were
far enough away from grips. The results of these FEM simulations are shown in Figure
4.5. By comparing strain-strain curves of four points, along the specimen centerline
64
(shown in Figure 4.5a), no significant differences were observed in the strains values
Large strains were measured using digital image correlation (DIC). DIC is a
technique that compares digital images of a specimen surface before and after
deformation to deduce its two-dimensional surface displacement field and strains [63].
The images required for the DIC method were acquired by a Pulnix TMC-9701 digital
CCD camera with Nikon AF micro Nikkor 105mm lens and National Instruments PCI-
1424 frame grabber. TIFF images that were 768 by 472 pixels were written to a PC at a
continuous rate of 4 frames/sec. Because of the large amount of PC bus activity and the
large numbers of image files, separate computers are used for DIC image storage and
load and strain gage data acquisition. To minimize intensity change during the
experiment, directed light source (KL1500, SCHOTT) was used to illuminate the
The DIC method works best when the starting image contains a random pattern
that can carrys the specimen deformation exactly. To create such a random pattern on the
current specimens, spray paint was used. First white paint provided a bright background
and then black spray paint gave a high contrast speckle pattern. The black paint was
applied by spraying parallel to the specimen surface and letting paint droplets fall
randomly on the specimen. Examples of the random pattern before and after deformation
The DIC program was set to measure incremental transverse and longitudinal
strain between subsequent images. The DIC program was first used in an interactive
mode to search good correlation parameters. For random patterns such as the one shown
in Figure 4.7, the best subset size was 100 x 100 pixels which corresponded to 4.2 x
4.2mm. The displacement was measured at 25 equally spaced points in the subset. The
65
maximum search range was set to 40 pixels in the longitudinal and transversal directions,
and the tolerance for optimization is set to 1.0x10-7. Subsequently, using the same
correlation parameters, a batch mode was used to process all the data for a particular
experiment. Since the subset used for correlation at the beginning of the loading has
undergone significant translation by the end the deformation, it was necessary to devise
an algorithm to track the approximate position of each subset used for correlation. Total
strain was obtained by adding the incremental strains. In order to reduce measurement
error, the correlation was performed on all 25 points in the subset, and then strains values
there were beyond one standard deviation were eliminated. The average strain was then
determined from the remaining strain values. The original DIC program was written by
Vendroux and Knauss in FORTRAN and C [63]. The modifications to track the overall
displacement of the subsets and to statistically treat the strain measurements were done
by the authors. The load is measured from the load cell, those strain data and the load
comparison with a strain gage for lower strains and with the data provided by
NUMISHEET 2002 for higher strains, and the results are shown in Figure 4.8. The strain
measurement with DIC has poor resolution at lower strains but shows quite good
correlation with the NUMISHEET 2002 data at large strain. Usually it is difficult to
measure large strains with a strain gauge, but DIC can measure 25% to 30% strain with
relative ease. The limitation at strains greater than 30% is related to the loss of adherence
66
4.4 Hardening models
The hardening model used to fit the experimental curve in this paper is the RIK
following Chun et al. [7] which is modified from the Chaboche model [39] in order to
better reproduce the tension test stress-strain curve. The yield function is defined as
follows:
c1
( )
σ y = σ0 + K 1 − e Ns −
b1
( )
1 − e b1s − c 2s , (4.4)
where σ0 is the initial yield stress, K, N are the fitting parameters of the hardening curve
measured in the RD, and c1, c2, b1 are the parameters for kinematic hardening, and s is the
c1
q1∇ = ( τ − y ) s − b1sq1 , (4.5)
β
c
q∇2 = k 2 ( τ − y ) s , and (4.6)
β
1 + κϑ for initial loading
k ( ϕ, ϑ) = (4.7)
κϑ for reversal loading
where ϑ is defined the angle of the direction of the symmetry axes and direction of
The rotational hardening is defined by the rotation of the yield function φ whose
basis functions eiφ are assumed to lay on the symmetry axes of anisotropy. The rotation
of the symmetry axes is defined with the constitutive spin θφ which is defined from
67
θφ = w p − ω φp , and (4.8)
φ φ φ
e =θ e
i i (4.9)
The plastic spin in (4.8) is defined as follows and is due to the non-coaxiality of the stress
ωφp = µφ ( τε p − ε p τ ) . (4.10)
The plastic spin parameter µφ in (4.10) is defined through the angle ϑ [15, 46],
a
µφ = tan ( ϑ ) , (4.11)
σy
and a is the only parameter to be defined for the plastic spin. The parameter can be
identified with the results of the transverse strain vs. longitudinal strain curve.
(4.6) are identified using the inverse method which optimize the parameters from the
results of the three point bend experiments performed by Chun et al. [7]. If the rotational
hardening is ignored, a and κ are zero, then the RIK hardening model returns to ANK
model. In addition, the hardening model turns to Chaboche model if c2 becomes zero,
and the model becomes an isotropic hardening model if c1,b1 are set to zero.
The experimental results of the materials are transverse strain versus longitudinal
strain curves to measure the anisotropy values (R-values) and flow stress curves in the
RD, TD and 45° directions. In order to show the dependence on material parameters, the
procedure to define parameters will be explained with DDQ steel data provided from
both the current experiments and NUMISHEET 2002. The experimental results of others
68
are shown with theoretically fitted curves and the parameters of the materials will be
summarized.
The sheet metal of DDQ used in the experiments is the same material used in
NUMISHEET 2002 benchmark problems [61]. The measured transverse strain versus
longitudinal strain curves for the RD, TD and 45° are shown in Figure 4.9(a), and the
corresponding flow curves are shown in Figure 4.9(b). Also shown in Figure 4.9 are
simulation results of initial anisotropy without rotational hardening. When the rotational
evolution of the anisotropy is not considered, the strain-strain relation for the 45°
orientation is higher than the experimental curve. The estimated flow curve for the 45°
orientation differs from the experimental results after approximately 5% strain. The ratio
of stress evolution between 0% and 5% strain correlates very well with the experiment.
For larger deformations, one can actually get better results when using the averaged R45
value which is computed from the linear fit of the strain-strain from 0% to 20% strain.
The comparison is in Figure 4.10. However, even though the stress level is much closer
to the experimental results than the estimation by using initial anisotropy values in Figure
4.9(b), the rate of stress evolution does not correlated with the experimental results for
because the anisotropy evolution in 45° orientation is not accounted in the conventional
hardening models. Following the assumption, the anisotropy evolution can be seen for
only 45° orientation, so consider R45 related to the flow stress. To model the R45
evolution, as shown in Figure 4.11(a), only the rotational hardening scheme can change
the evolution rate according to the deformation by using plastic spin theory. The
69
corresponding flow stresses for the 45° orientation then evolved with an abrupt change in
stress as the symmetry axes are rotate due to the plastic spin. The evolution occurs
because the basis of the yield function is laid on the symmetry axes of anisotropy. The
results for the flow stress are shown in Figure 4.11(b). The RD and TD curves are
omitted because there is no rotation in those two directions and the predictions in these
directions are identical to those without rotational hardening. The plastic parameter a = -
155 is selected as the computed R45 curve that best correlates to the experimental curve.
The difference in flow stress curves between the experimental and computed
results is then corrected by kinematic hardening (4.7). As shown in Figure 4.12(a) and
(b), kinematic hardening has corrected the flow stress curves but did not change the slope
of the strain-strain curves. The parameters for this modeling are summarized in Table 4.1.
The flow stress curves are shown in Figure 4.1 and the transverse strain versus
longitudinal strain curves are shown in Figure 4.2. As was shown in Figure 4.1, the
hardening evolution for the 45° orientation is overestimated with average R value. The
simulation results with RIK hardening are shown in Figure 4.13 and are compared with
the experimental results. For plastic spin parameter, R45 evolving curve is shown in
4.5.3 DQ
The drawing quality (DQ) mild steel was also tested. Following the same
procedure that was described earlier for DDQ, the final fitted curves are shown in Figure
4.14.
70
4.5.4 HSS
High strength steel (HSS) was fitted to the RIK hardening model. In contrast to
mild steels, HSS does not show much rotation. Experimental results are compared with
As summarized in Table 4.1, when comparing R-values for all directions, only for
HSS the value of R45 is higher than other values. Even with high plastic spin factors, for
high R45 values, the anisotropy axes are not rotating toward the straining direction for the
45° orientation. In addition to that, the flow curves are accurately modeled with initial R
values chosen from the transverse strain versus longitudinal strain curve. Therefore, it
may be concluded that the rotational hardening in HSS is very small. The anisotropy of
this material is not effectively changed by the strains and deformation in these tests.
Therefore, it may be assumed that the texture of this material is not severely affected by
the deformation of these experiments. To verify this claim, the symmetry axes would
(ODF).
4.6 Discussion
The R-values measured in tensile experiments are not useful for predicting the rotational
R 0 + 2R 45 + R 90 R − 2R 45 + R 90
R= , ∆R = 0 , (4.12)
4 2
are used to judge the formability ( R ) and the earing pattern ( ∆R ) in circular cup
drawing. Formability increases as R increases. For mild steel with the rotational
hardening, the yield function shape evolution in σ1 , σ 2 space looks as if the R increases
71
with isotropic expansion of the yield locus. Therefore, it could be said that models with
rotational hardening can describe the formability better than non-rotational hardening
models.
The plastic spin in sheet metals defines the planar rotation of the symmetry axes
must be a relationship between the parameters of the plastic spin and the planar
anisotropy ∆R. The plastic spin and planar anisotropy values for the tested materials are
shown in the Table 4.1. As shown in Figure 4.16, the plastic spin is linearly related to
anisotropy is required to verify these results. Nonetheless, in Figure 4.16 the comparison
between the plastic spin parameter in Han et al. [15] using the measured data of the
rotation by Kim and Yin [4] shows excellent correlation. Therefore, the unknown plastic
spin parameter can be identified from the R-values of mild steel. However, more
experiments are required to verify this methodology for other alloys and crystal structures.
For two mild steels and high strength steel sheet metal, experimental results of
conventional flow stress curves and transverse versus longitudinal strain curves for
various orientations to the RD have been provided. Using the assumption that anisotropy
axes rotation is the dominant reason for the evolution of anisotropy, the material
parameters for plastic spin and for correcting kinematic hardening can be identified from
the simple tensile tests. With the help of rotational evolution of the anisotropy, the RIK
model can capture the material plastic hardening behavior with Hill’s quadratic yield
criteria.
72
Using the DIC method, it is relatively easy to measure large strains. The DIC
method was used to generate plots of transverse strain versus longitudinal strain for
simple tension experiments. The evolution of anisotropy is then correlated with the slope
of the transverse strain versus longitudinal strain curve. Only the plastic spin parameter
can change the slope of the computed transverse strain versus longitudinal strain curve.
The flow stress results are correctable with the kinematic hardening parameter. The
parameters for the RIK model are defined by three-point-bend-test [53, 65] and tensile
tests for three directions: RD, TD, and 45°. For mild steel, a linear relationship between
the planar anisotropy ∆R and the plastic spin parameter a has been found.
73
Material Elastic Anisotropy Isotropic Kinematic Rotational
R0 = 2.137 σ0 =152.22MPa c1 =3.0GPa
DDQ E=210GPa a = -155
R45 = 0.935 K =222.01MPa b1 =300
(Exp.) ν=0.3 κ = 50.0
R90 = 1.508 N =-7.87 c2 = 70.0MPa
R0 = 2.722 σ0 = 152.0MPa c1 = 3.0GPa
DDQ E=210GPa a = -57
R45 = 1.474 K = 235.81MPa b1 = 300
(NS2002) ν=0.3 κ = 1.8
R90 = 2.169 N = 9.23 c2 = 70.0MPa
R0 = 1.60 σ0 = 198.0MPa c1 = 3.3GPa
E=180GPa a = -100
DQ R45 = 1.010 K = 242.47MPa b1 = 220
ν=0.3 κ = 1.5
R90 = 1.46 N = 9.95 c2 = 103MPa
R0 = 0.832 σ0 = 332.22MPa c1 = 3.0GPa
E=210GPa a=0
HSS R45 = 1.185 K = 430.55MPa b1 = 150
ν=0.3 κ=0
R90 = 0.560 N = 1.7 c2 = 80MPa
Table 4.1: Material properties of RIK hardening model for DDQ, DQ and HSS.
74
400
350
300
True stress (MPa)
250
200
EXP. FEM
RD
150 45
o
TD
100
0 0.05 0.1 0.15 0.2 0.25
Eqv. Plastic strain
75
0
-0.05
o
45
Transverse strain
-0.1
-0.15 TD
RD
-0.2
0 0.05 0.1 0.15 0.2 0.25
Longitudinal strain
Figure 4.2: Transverse strain versus longitudinal strain of DDQ. The slope of each curve
is used to compute the R-values. (Numisheet 2002 [61])
76
3
2.5 R
0
2 R
90
1.5 R
R
45
0.5
0
0 0.05 0.1 0.15 0.2 0.25
Longitudinal strain
Figure 4.3: Evolution of R-values: R0 and R90 are not evolving compared to the R45.
77
Figure 4.4: R-value measurements for rolled copper in compression experiments.
Particularly R45 shows evolution.[59]
78
1 2 3 4
(a)
0.2
1
2
0.15
Transverse strain
3
0.1 4
0.05
0
0 0.05 0.1 0.15 0.2 0.25 0.3
Longitudinal strain
(b)
Figure 4.5: FEM validation of specimen dimensions to avoid clamping effect: a) strain
measurement locations, b) computed strains. The strains are closer enought to show that
the measured area is safe from the clamping effect.
79
LIGHT
SOURCE
MTS Controller
CCD
CAMERA
STRAIN GAUGE
CONDITIONER PC
STORE STRAIN
AND LOAD
PC
STORE IMAGE DATA A/D
CONVERTER
MTS
Figure 4.6: Schematic diagram of the experimental set up using DIC(Digital Image
Correlation).
80
(a)
(b)
Figure 4.7: The random pattern sprayed on the specimen and captured by CCD camera:
a) before and b) after deformation. The small rectangles highlight a 100 x 100 pixels
subset of the image that has been deformed.
81
300
250
200 Longitudinal
100
Transverse
50 Comparison in
small strain
0
-0.01 -0.005 0 0.005 0.01 0.015
Strain
(a)
-0.05
Transverse strain
-0.1
-0.15
RD (DIC)
-0.2
RD (NS 2002)
-0.25
0 0.05 0.1 0.15 0.2 0.25 0.3
Longitudinal strain
(b)
82
0
-0.05
Transverse strain
-0.1
EXP. FEM
-0.15
RD
o
45
TD
-0.2
0 0.05 0.1 0.15 0.2 0.25
Longitudinal strain
(a)
500
400
True stress (MPa)
300
EXP. FEM
200 RD
o
45
TD
100
0 0.05 0.1 0.15 0.2 0.25
Eqv. Plastic strain
(b)
83
0
-0.05
Transverse strain
-0.1
EXP. FEM
-0.15
RD
o
45
TD
-0.2
0 0.05 0.1 0.15 0.2 0.25
Longitudinal strain
(a)
500
400
True stress (MPa)
300
EXP. FEM
200
RD
o
45
TD
100
0 0.05 0.1 0.15 0.2 0.25
Eqv. Plastic strain
(b)
Figure 4.10: Simulation results of averaged R value for 45° orientation comparing with
the experimental results: a) Transverse strain versus longitudinal strain, b) flow stresses.
Prediction of 45° orientation is underestimated.
84
2
DDQ(DIC)
1.5
a = -250
a = -155
a = -50
45
1
R
a=0
EXP
0.5
0
0 0.05 0.1 0.15 0.2 0.25
Longitudinal strain
(a)
500
a=0
400 EXP
True stress (MPa)
a = -50
300
200
a = -155
a = -250
100
DDQ(DIC)
0
0 0.05 0.1 0.15 0.2 0.25
Eqv. plastic strain
(b)
Figure 4.11: Sensitivity of the plastic spin parameter: a) R45 evolution, b) flow stresses of
45° orientation according to the plastic spin parameter. The gap between the
measurement and computed flow stress is the amount of backstress evolution.
85
0
-0.05
Transverse strain
-0.1
EXP. FEM
-0.15
RD
o
45
TD
-0.2
0 0.05 0.1 0.15 0.2 0.25
Longitudinal strain
(a)
500
400
True stress (MPa)
300
EXP. RIK
200
RD
o
45
TD
100
0 0.05 0.1 0.15 0.2 0.25
Eqv. Plastic strain
(b)
Figure 4.12: Simulation with the RIK hardening model and comparison with
experimental results: a) Transverse-strain versus longitudinal strain, b) flow stresses.
Prediction for 45° orientation captures the experiment well.
86
0
-0.05
Transverse strain
-0.1
EXP. FEM
-0.15 RD
o
45
TD
-0.2
0 0.05 0.1 0.15 0.2 0.25
Longitudinal strain
(a)
500
400
True stress (MPa)
300
EXP. RIK
200 RD
o
45
TD
100
0 0.05 0.1 0.15 0.2 0.25
Eqv. Plastic strain
(b)
(continued)
Figure 4.13: DDQ- data provided by NUMISHEET 2002. Computed by RIK hardening
model comparing with experimental results: a) Transverse strain versus longitudinal
strain, b) flow stresses, c) R45 evolution.
87
(Figure 4.13 continued)
1.5
1
45
R
Exp.
0.5
FEA
0
0 0.05 0.1 0.15 0.2 0.25 0.3
Longitudinal strain
88
0
Transverse strain
-0.05
-0.1
EXP. FEM
RD
o
45
TD
-0.15
0 0.05 0.1 0.15 0.2 0.25
Longitudinal strain
(a)
500
450
True stress (MPa)
400
350
300
EXP. RIK
RD
250 o
45
TD
200
0 0.05 0.1 0.15 0.2 0.25
Eqv. Plastic strain
(b)
(continued)
Figure 4.14: DQ – experiment. Computed by RIK hardening model comparing with
experimental results: a) Transverse strain versus longitudinal strain, b) Flow stresses, c)
R45 evolution.
89
(Figure 4.14 continued)
2
DQ (DIC)
1.5
1
45
R
0.5
Exp.
RIK
0
0 0.05 0.1 0.15 0.2 0.25
Longitudinal strain
90
0
-0.03
Transvers strain
-0.06
-0.09
EXP. FEM
RD
-0.12 o
45
TD
-0.15
0 0.05 0.1 0.15 0.2 0.25
Longitudinal strain
(a)
600
500
True stress (MPa)
400
EXP. RIK
300 RD
o
45
TD
200
0 0.05 0.1 0.15 0.2 0.25
Eqv. Plastic strain
(b)
Figure 4.15: HSS – experiment. Computed by RIK hardening model comparing with
experimental results: a) Transverse strain versus longitudinal strain, b) Flow stresses.
91
0
DDQ (NS)
-100
DQ (DIC)
DDQ (DIC)
-200
a
-300
Figure 4.16: Linear relationship between the plastic spin parameter (a) and the planar
anisotropy (∆R). This relationship can apply for mild steel which has B.C.C. crystal
structure.
92
CHAPTER 5
The model combines isotropic, kinematic, and rotational hardening, which is motivated
texture. Armstrong-Frederick type backstress components are used for the kinematic
hardening description. The plastic spin theory in literature is used to describe the rotation
of the symmetry axes of anisotropy, which is considered to be rotated along the texture
preferred direction.
The purpose of this chapter is to show the significance of the RIK hardening
model in the sheet metal forming process as well as springback prediction. With
rotational hardening and related kinematic hardening, the flow stress in each orientation
from the RD can be reproduced with a limited number of material parameters. Several
benchmark problems are considered to demonstrate the effect of the RIK hardening
model in comparison with other hardening models and experimental results. While the
deep drawing examples show the effect of the RIK model in the forming process,
simulation of pulling a sheet metal strip from a draw bead tool is used to demonstrate the
93
effect of the RIK model in springback. A typical form of hydroforming is explored using
5.1 INTRODUCTION
experiments, knowledge about the origin of the evolution of anisotropy has greatly
structures and texture [24, 29, 66-68]. The development of dislocation structures has
been considered to be related to isotropic and kinematic hardening, while the texture
development is related to the rotation of the symmetry axes of anisotropy [24, 36, 66].
Isotropic and kinematic hardening models have been developed so well that they are able
to capture the permanent offset of cyclic loading for some materials [7, 69, 70]. However,
the reorientation of the texture has not been incorporated into a phenomenological
rotational evolution of anisotropy directions [4, 5, 14, 15, 67]. The RIK hardening model
applied problems. The material behavior under those loadings has been described in
polycrystalline plasticity with complex material properties which are defined by many
experiments [25, 28, 29, 30, 71]. The polycrystalline approach requires massive
computation, so it has been hardly used in practical forming problems even though it may
yield a better accuracy than conventional phenomenological approaches. The RIK model
94
is capable of capturing multi-axial or multi-path loading by translating the key features of
the material behavior with a proper Bauschinger effect and rotational hardening, as
shown in [13].
In order to assess the capability of the RIK model in the application of industrial
problems, the results of some conventional sheet metal forming processes and spring
The suggested RIK hardening model is able to describe the growth of the yield
surface (isotropic hardening), its transition (kinematic hardening), and its rotation
(anisotropy axes rotation, via plastic spin). The yield function of the RIK hardening
( ) (
Φ τ, y, q;eiφ = φ ( τ − y ) ; eiφ + q , ) (5.1)
where q describes the isotropic hardening, y the backstress, and eiφ the orientation of the
( )
yield surface. Such a yield function φ τ, y; eiφ can represent Hill’s or Barlat’s yield
function for mild steels or aluminum alloys, respectively. Hill’s quadratic yield function
Φ = φ − σ 2y , (5.2)
R0 R (1 + R 90 ) ( R + R 90 )(1 + 2R 45 )
β12 = , β22 = 0 , β66 = 0 (5.4)
1+ R0 R 90 (1 + R 0 ) R 90 (1 + R 0 )
and the orientation of the tensor K φ is redefined in the deformation process by the basis
of eiφ which is defined with the plastic spin and the material spin. The reorientation can
be determined via
φ
K ijkl = ∑ K 0αβγδ ( eiφ ⋅ eα0 )( eφj ⋅ eβ0 )( eφk ⋅ e0γ )( eφl ⋅ eδ0 ) , (5.5)
and the constitutive spin is defined with material spin and plastic spin as
θ = w − ωp (5.7)
For the plastic spin the definition of Dafalias [14] and Kuroda [13, 18, 44] has been
modified as
a
ωp = tan ( ϑ ) ( τε p − ε p τ ) . (5.8)
φ
The isotropic hardening part of the yield function is described, as in Chun et al.
[7],
c1
(
σ y = σo + K 1 − e − Ns − ) b1
( )
1 − e − b1s − c 2s , (5.9)
96
with the equivalent plastic strain s, the initial yield stress σo and hardening coefficients K
and N, and other material parameters which are related to the kinematic hardening is
described below.
y = ∑qj , (5.10)
j
s
q1∇ = c1 ( τ − y ) − b1sq1 and (5.11)
β
s
q∇2 = kc 2 ( τ − y ) . (5.12)
β
The k in equation (5.12) is dependent on the loading direction and the angle of the
difference between the anisotropy axes direction and the plastic straining direction ( ϑ ),
1 + κϑ initial loading
k= (5.13)
κϑ reversal loading
Three material parameters are used in the RIK hardening model for the isotropic
hardening and three for kinematic hardening. In addition, the R values in three directions
are required to define the anisotropy yield function of Hill’s [6] for plane stress
conditions. The identification of these material parameters is clearly a non trivial problem.
observation and discussions in Chapter 4. The RIK hardening model can describe the
stress flow in an arbitrary angle to the RD, while a simple application of Hill’s quadratic
yield criterion is in general not capable to describe the stress evolution in the arbitrary
97
direction [2]. The advantage of the RIK hardening model is its capability to account for
the evolution of the anisotropy. The kinematic hardening parameters c1,c2 and b1 can be
identified from the three-point-bend test and an inverse method with a genetic algorithm
presented in Zhao and Lee [53, 54, 65] and modified by Chun et al. [7, 8].
If the rotational parameters (a, κ) are zeros, then the RIK hardening model
Chun et al. [7]. The computational investigations for the deep drawing and springback
processes were performed with respect to three hardening models: the isotropic, the ANK,
and the RIK hardening models. The materials used in this presentation were mild steel
DDQ, DQ and the material of NUMISHEET ’93. The material properties are
summarized in Table 1.
5.4 APPLICATIONS
In order to compare the simulation results with the experimental data, the
NUMISHEET 2002 benchmark problem [61] of a cylindrical cup drawing has been
considered to assess the behavior of the RIK hardening model on the sheet metal forming
process. The problem is commonly considered for an assessment of the planar anisotropy
via earing pattern along the outside of the flange after the forming process. The punch
force-displacement characteristics and the earing pattern are of specific interest in this
problem. The dimensions of the tool set are shown in Figure 5.1(a). The undeformed
mesh is shown in Figure 5.1(b) in which 900 reduced integrated Kirchhoff-type shell
elements are used, where the transversal shear strains are not explicitly contained in the
the flange. The considered material of the blank is a DDQ steel. Two friction
98
coefficients have been used in the simulations. A lower friction coefficient of 0.049 as
coefficient of 0.12.
3. The punch forces are shown in Figure 5.2a for the isotropic, kinematic, and the RIK
hardening models. The differences in the punch forces are within the range of the
patterns of the hardening models show many more differences, as can be seen in Figure
5.2b. The dots are the averaged values of the NUMISHEET benchmark experimental
The earing pattern of the isotropic hardening and the ANK models are almost
identical and more draw-in is determined than that of the RIK hardening model. Also,
the differences of the earing pattern between the RD and TD are small in isotropic and
kinematic hardening in contrast to the big difference in the RIK model. In relation with
the experimental results, the RIK hardening model yields a better agreement than the
other two hardening models. The difference of the amount of the punch force and the
earing pattern are considered to be due to the friction coefficient which often have a
strong influence in the structural response of such problems. The punch force and the
earing patterns are closer to the results of the experiments. It is interesting to note that for
the friction effect on the earing pattern, the friction is more effective in the region of the
sheet of 0° to 45° to RD than in the region between 45° to 90° to RD. Therefore, a
smaller force and more draw-in are determined in the simulation than the experiment due
99
Without the rotational hardening, the initial anisotropy is attached to the material
points and remains basically unchanged in this problem, while the orientation of the
anisotropy evolves with the deformation with rotational hardening. The stress
distribution plots are shown in Figure 5.3, showing the differences of the isotropic
Severe differences in stress distribution at the diagonal direction and the rounded
corner between the side wall and flange can be found. It may be worth noting that there
are quite a few differences in stress distribution, although the punch force-displacement
are essentially the same. This can be explained by the rotation evolution of the symmetry
axes of anisotropy. The rotation angles of the anisotropy axes are shown in Figure 5.4,
and maximal rotations of about 45° are reached in the region of 43.8° and -40.8° to RD.
Another deep drawing example – the square cup of the NUMISHEET ’93
benchmark – is considered. The diagram of the tools is shown in Figure 5.5a and the FE
discretization is illustrated in Figure 5.5b. The tools were modeled with analytical rigid
surfaces and the blank sheet metal is descritized with 2500 reduced integrated shell
elements. A blank holding force 19.6kN and the friction coefficient of 0.15 are applied.
25 integration points in the thickness direction have been used. In order to see the effect
of multi-axial loading, major and minor strains and thickness distribution along the
diagonal directions are compared with respect to the hardening models in Figure 5.6. The
diagonal direction is chosen because a severe multi-axial loading is expected due to the
shape and complex material flow around the corner area. Significant differences of the
punch force-displacement are shown in Figure 5.7. The punch force prediction using the
RIK hardening model is 20% lower than the force with the isotropic hardening model.
100
As a result, the draw-in of the RD, TD, and diagonal direction are compared with the
5.8.
5.4.3 Pulling the strip through a draw bead toolset and springback
geometry of the draw bead tool is shown in Figure 5.9. A strip is clamped between the
tools where a gap of twice the thickness between the dies, dg. The clamped strip is pulled
out for 165 mm while the tool remains fixed. In the last stage of this problem, the tools
The dimension of the specimen is 325mm in length, 25.5 mm in width and 0.9mm
in thickness. The elements are descritized 100 in length and 5 in width, respectively.
The width of the specimen is chosen that narrow in order to minimize the curvature
effects and to be as close to the plane stress condition as possible, to assess the twisting
The results are shown in comparison to the isotropic hardening, the ANK
hardening, and the RIK hardening. Comparing the shapes of the 30° oriented specimens
after springback, differences in twisting modes as well as the heights can be seen. The
anisotropy in the RD and the TD. The differences in the modes after springback are due
to the stress evolutions during the process. The ANK and RIK hardening models show
lower values in springback height, because the models have a permanent softening
scheme. The material experiences cyclic loading by bending and unbending while
101
passing through the draw bead tool. The stress histories of a material point are shown in
The shapes are different depending upon the hardening rules. The effect of the
hardening rules in the twisting mode is illustrated in Figure 5.11 for a 30° oriented
specimen.
The reason for developing different twisting modes is the stress distribution just
before springback. As shown in Figure 5.12, the stress distributions in the isotropic or
the ANK hardening model are anti-symmetric in both edges, while the RIK hardening
model develops symmetric stress distribution. The different stresses in both sides make
The clamping force and drawing force for each case are summarized in Table 5.2.
The forces for the initial clamping are increase when the rotational hardening scheme is
added. In contrast to the clamping force, the drawing force requires less force as the
rotational hardening scheme is used. With the numerical results, we can conclude the
rotational hardening in the RIK model has a key role in reproducing both the twisting
5.4.4 Hydroforming
discussed in the literature [72-74]. In the deformation process of the T-tube expansion,
material points in the expansion zone experience multi-axial loading situations because
material points have compression and tension in several directions as they move along
the tool surface. There are hoop and radial stresses due to the pressure inside the tube
and compression from both ends to feed the material. There have been many
investigations to optimize the pressure and the material feeding at both ends. Therefore,
102
the effect of the hardening models is discussed in this paper for the isotropic, the ANK,
and the RIK hardening models. The material used in this simulation is mild steel (DDQ).
The dimension of die set and loading conditions are shown in Figure 5.13. The tube is
modeled with 3500 shell elements. In order to see the difference of the expansion zone,
the top portion of the die is set to open. The pressure applied inside the tube is calculated
to slightly over the yield stress to make the tube expand. The movement of both ends to
feed the material at the center to make the expansion easier is chosen to prevent excessive
The deformation height is not effective with the hardening model. The maximum
difference at the center of the dome is only 0.5% (45.54mm for isotropic hardening and
45.78mm for RIK hardening). However, the stress distribution shows much difference
with respect to the hardening models as shown in Figure 5.14. The highest stressed
region is shown at the connected area (marked with a circle) for isotropic and RIK
hardening. But the equivalent plastic strain is much larger in the ANK model, and the
region of the RIK model is smaller than other models (Figure 5.15). It is unique for the
RIK model that the plastic strain at the top of the dome is distributed in an elliptical shape
along the longitudinal axes (marked with an arrow). The rotational evolution of the
symmetry axes are shown in Figure 5.16 and the maximal rotation is 55° in accumulating
the rotation in loading and unloading. It should be noted that the region of severe change
forces at both edges, while feeding the material toward the center of the tube in a certain
103
5.5 CONCLUDING REMARKS
Three kinds of the forming process and one kind of the springback process are
investigated numerically. The rotational hardening scheme makes the material behavior
in hardening differently, and then the different behavior results in the springback. The
accuracy of the RIK hardening model in the forming process is shown at the draw-in of
the cylindrical cup. Only the RIK model can follow the hardening history around the
diagonal direction.
square cup drawing problem. The corner of the cup is the area. The stress and strain
distributions in the area are different so much that the contour of the RIK model does not
resemble the other two hardening models. It is more useful to predict formability of the
hydroforming process, because the material points on the tube have experienced multi-
Those flow stress effects are also revealed in the springback. The springback of a
strip which is oriented a certain angle to the RD initially has a different shape with
respect to the amount of the symmetry axes rotation. The main effect is at the curvature
of the strip in the springback, and also the twisting mode which can be corrected only
104
Material Elastic Anisotropy Isotropic Kinematic Rotational
R0 = 2.137 σ0 =152.22MPa c1 =3.0GPa
DDQ E=210GPa a = -155
R45 = 0.935 K =222.01MPa b1 =300
(Exp.) ν=0.3 κ = 50.0
R90 =1.508 N =-7.87 c2 = 70.0MPa
R0 = 1.491 σ0 = 198.0MPa c1 = 3.3GPa
E=180GPa a = -100
DQ R45 = 1.01 K = 62.47MPa b1 = 220
ν=0.3 κ = 1.5
R90 = 1.180 N = 8.95 c2 = 103MPa
R0 = 1.790 σ0 = 154.0MPa c1 = 3.0GPa
E=206GPa a = -215
NS ‘93 R45 = 1.410 K = 253.1MPa b1 = 210
ν =0.3 κ = 5.0
R90 = 2.270 N = 9.78 c2 = 110MPa
Table 5.2: Clamping force and drawing force for different orientation and different
hardening for draw bead test.
105
(a)
(b)
Figure 5.1: Cylindrical cup drawing: a) Dimension of the tool set. (R1: punch
radius=50mm, R2: die radius=51.25mm, R3: punch fillet=9.5mm, R4: die fillet
radius=7.0mm, R0: blank radius=105mm<DDQ>), b) initial mesh
106
120
Experiment
100
20
0
0 10 20 30 40
Punch displacement (mm)
(a)
96
94
Distance to the center (mm)
RIK
High Friction ANK
92
Experiment
90
RIK
88
86
Isotropic
84
0 10 20 30 40 50 60 o 70 80 90
Angle to the RD ( )
(b)
107
(a)
(b)
Figure 5.3: Cylindrical cup drawing: Residual von Mises stress distribution with respect
to (a) isotropic hardening, (b) RIK hardening
108
Figure 5.4: Rotation angle (in degree) of the symmetry axes of anisotropy.
109
(a)
(b)
Figure 5.5: Square cup drawing - Forming tool for the deep drawing of a square cup (a)
and the FE model (b).
110
0.3
0.2 ε RIK
1
ISO
0.1
Strain
0
-0.1
-0.2 ε ISO
2
RIK
-0.3
0 20 40 60 80 100
Distance from center (mm)
(a)
0.9
0.85
Thickness (mm)
0.8
RIK
0.75
ISO
0.7
0.65
0 20 40 60 80 100
Distance from center (mm)
(b)
Figure 5.6: Square cup drawing – Show the difference of a) Major and minor strain
distribution along the diagonal direction, b) thickness distribution along the diagonal
direction with respect to hardening models..
111
80
ISO
ANK
60
Punch force (kN)
40
RIK
20
0
0 10 20 30 40
Punch travel (mm)
Figure 5.7: Square cup drawing – Punch force-travel shows different history according to
the hardening models.
112
30
25
20
Draw-in (mm
ISO
15 RIK
EXP
10
0
DX DD DY
Figure 5.8: Square cup drawing – Draw-in with respect to hardening models compared to
the averaged experimental results of NUMISHEET ’93.
113
FC
99
16.2
dg R=6.5
FD
R=6.5
unit : mm
114
3
Isotropic
2
RIK
Normalized stress
ANK
0
-1
Clamp Draw
-2
0 0.5 1 1.5 2
Loading history
Figure 5.10: Stress history during the process for the 30° oriented specimens
115
140
Isotropic
120
ANK
100 RIK
Z-coordinate (mm)
80
60
40
20
0
-40 -20 0 20 40 60
Y-coordinate (mm)
(a)
200
150 Isotropic
ANK
Z-coordinate (mm)
100
50 RIK
-50
-100
0 50 100 150 200 250 300
X-coordinate (mm)
(b)
Figure 5.11: Springback shape of the 30° oriented specimen: (a) rear view, (b) side view
116
(a)
(b)
(c)
Figure 5.12: Von Mises stress distribution before springback: a) isotropic, b) ANK, c)
RIK hardening model. RIK shows symmetric distribution of stress while the other two
show anti-symmetric distributions which causes unexpected twisting model after
springback.
117
(a)
1
Pressure (Max: 20Pa)
0.8
Rate of loading
0.6
0.4
Axial feeding
0.2
(Max: 8mm)
0
0 0.2 0.4 0.6 0.8 1
Loading history
(b)
Figure 5.13: Hydroforming – Dimensions of the tool and initial tube and loading
conditions of the pressure and displacements at both ends of the tube.
118
(a)
(b)
(c)
(b)
(c)
120
Figure 5.16: Rotation angle of the symmetry axes of anisotropy. Rotational evolution of
the anisotropy is concentrated at the highest stress area.
121
32
Axial Force (kN)
28
RIK
24
ISO
20 ANK
16
0 2 4 6 8
Axial feed (mm)
Figure 5.17: Reaction force at both edges when feed the material axially toward center in
the rate of displacement shown in Figure 5.13b.
122
CHAPTER 6
in plastic loading. If dislocations move with the same degree of resistance in every
direction, then the material shows isotropic characteristics. If dislocations move with a
different degree of resistance in accordance with directions, then the material shows
considered to be grain boundaries, precipitate, and dislocation walls, etc. In the sources
The rotation of texture and relating backstress evolution has been discussed in
earlier chapters for mild steels which have a BCC crystal structure. In contrast to the
BCC material, aluminum alloys are known for the small rotation in texture. However,
some aluminum alloys show strong anisotropy and anisotropy evolution in the orientation
away from the RD due to the non-shearable precipitate inclusion in the material matrix.
Usually a higher-order anisotropic yield criterion such as Barlat ’96 [11] is used for an
Therefore, the model cannot predict anisotropic hardening for an arbitrary angle to the
123
fourth-order tensor that Barlat and Liu [10] proposed for the precipitate hardened material,
e.g., Al-Cu3%.
6.1 Introduction
models are used instead, because of their benefit in the cost and ease in implementing in a
finite element approach. However, for a certain problem, like springback analysis, these
models are not accurate, because they do not account for the anisotropy hardening in an
compression were performed by Barlat et al. [9] in Figure 6.1, which shows the flow
hardening using conventional isotropic hardening and kinematic hardening are shown in
Figure 6.2. Neither isotropic hardening nor kinematic hardening can predict the
anisotropic hardening for 45° because they do not account for the anisotropy evolution.
In order to account for such anisotropic hardening, Barlat and Liu [10], Barlat et
considerations using the ellipsoidal precipitate inclusion model introduced by Mura [75].
From the research of Barlat and Liu [10], anisotropy evolution for the precipitate
hardened material is dominant to the rotation of backstress. From the rotation of yield
function using plastic spin theory for the texture spin [14, 15], it is hinted that the
backstress also can be rotated as the precipitate rotates along the metal matrix
124
reorientation using a plastic spin similar to the one used in the rotation of the yield
function.
Because the process is for the finite strain, the material formulation is based on
summarized in Chapter 1.
The assumptions used in the formulation are: (i) matrix and precipitates have
approximately the same elastic characteristics, (ii) precipitates deform only elastically
Φ = f ( τ − q ) − 2 τya (6.1)
where
a a a
f = α1 s 2 − s3 + α 2 s3 − s1 + α 3 s1 − s 2 (6.2)
with a=8 for fcc materials. The details of the coefficient αk can be found in Barlat et al.
[11].
The isotropic parts to define the yield function size are defined by
(
τy = τ0 + K 1 − e
− N εp
) (6.3)
ε ep = ( I − Λ ) ε p , (6.4)
125
where I and Λ are identity and a tensor for precipitate inclusion. The tensor is defined
by Mura [75] as
1
Λ = ∑
f ( p)
f (p) Λ(p) , (6.5)
where f(p) is the volume fraction of a precipitate (p), and Λ ( p ) is Eshelby tensor for a
precipitate (p).
The applied stress can be decomposed to the matrix part and precipitate part as
{ }
τ = (1 − f ) τ m + f τ p , (6.6)
where f is the volume fraction of the precipitate, {τ p } is the average over all precipitates,
and τ m is for matrix. The alloy containing non-shearable precipitates shows a strong
τ p = Γ e (1 − Λ ) ε p , (6.7)
where Γ e is the elasticity tensor. For the part of precipitate inclusion in (6.6), the
backstress for the precipitate inclusion is modeled by Barlat and Liu [10] as
q = fΓ e {I − Λ } ε p h ( ε ) , (6.8)
where ε is the effective strain in the matrix. With the hint of (6.8), backstress possible
hardening are
126
q∇ = cAd p − sbq
q∇ = cAd p − sbAq , (6.9)
q∇ = cAd p − sbA −1q
The initial value of tensor A in matrix form M has been provided in Barlat and Liu [10]
as
With the plastic deformation, the texture are changed which in turn affects the
orientation of the precipitates. In addition, it has been observed that the precipitates
themselves can also rotate relative to the crystal orientation [76, 77]. Although
precipitates are often assumed to be embedded in the habit planes of the lattice,
precipitates enable rotations of the precipitates relative to the crystal orientation. The
o
nia = nia − θnia = 0 , (6.12)
where the constitutive spin θ can be defined with material spin and the plastic spin as we
127
θ = w − ωp . (6.13)
From the heuristic definitions of the plastic spin, the following definition is effective to
a
ωp =
φ
{( τ − q ) dp − dp ( τ − q )} . (6.14)
cases. In contrast to the plane stress problem, the spin is not limited within planar
the three-dimensional spin in Cartesian coordinates, it is assumed that spins along each
axis are small enough to ignore the higher-order term. Spins along each axis can be
1 0 0 1 0 −dθ y 1 dθz 0
R x = 0 1 dθ x , R y = 0 1 0 , R z = −dθz 1 0 , (6.15)
0 −dθ x 1 dθ y 0 1 0 0 1
1 dθz −dθ y
R = R x R y R z = −dθz 1 dθx (6.16)
dθ y −dθz 1
decide which form of the backstress evolution is the most efficient to follow the
measured anisotropic hardening. Flow stresses of 45° oriented specimen are examined
128
with and without the backstress rotation in Figure 6.3 to see the influence of backstress
reorientation in the formulation. When the fourth-order tensor is used in the non-linear
term, the stress flow shows an unexpected trend as the backstress rotates. By observing
the characteristics of each backstress model on the stress flow, the equation (6.9)a shows
The measurements are Al-3%Cu of Barlat et al. [9]. As expected, the isotropic
hardening model does not follow the experimental measurement as shown in Figure 6.2.
Even if the higher-order yield function is used, it is insufficient to predict the anisotropic
hardening in 45° orientation. By adding kinematic hardening in (6.9), it does not capture
orientation. The two models fit the hardening in the RD and TD fairly well.
evolution, the model gives under-estimate of the hardening for 45° as shown in Figure
6.4a. The backstress in 45° evolves very little. The model using the fourth-order tensor
and its reorientation with a plastic spin shows anisotropic evolution in the hardening and
gives a fairly good correlation to the experiment, as shown in Figure 6.4b. The
simulation in Figure 5.1 is used. In contrast to the model used in Chapter 5, solid 3d
elements are used in the model. Two layers of elements are used in the thickness
direction. The material properties regarding hardening models are from the paper of
129
Barlat et al. [9] and summarized in Table 6.2. Comparisons between the hardening
The results of earing patterns of each model are shown in Figure 6.5. The models
which do not account for the kinematic hardening develop a wide-banded earing pattern
while the models which account for the anisotropy for kinematic hardening develop a
smaller difference in the earing pattern. The difference of using the rotational scheme on
the kinematic hardening is not severe, as long as the anisotropy is accounted for the
kinematic hardening. The residual stress distributions are shown in Figure 6.6 with
The thickness distributions along the 45 degree line are compared in Figure 6.5.
As in the earing pattern, the difference between the hardening models is dependent on the
evolution anisotropy. It should be noted that the comparison is done only by numerical
analysis.
6.7 Conclusion
Even though there are insufficient validations of the model by comparing proper
experiments, the hardening curves show the effect of reorientation of the backstress to
describe the anisotropic hardening behavior of the precipitate hardened materials. This
could mean that in the precipitate hardened materials, anisotropy evolution is dominant to
the reorientation of the backstress, not to the reorientation of yield function. With the
130
c1, c2, c3, c4, c5, c6 1.144, 1.076, 1.066, 1.126, 1.023, 1.075
a 8.0
Ciso Cm
Hardening Model biso b α M
(MPa) (GPa)
B. Isotropic + Kinematic
110 15 3.0 80 0 -
hardening
C. Isotropic + Anisotropic
110 15 0.08 80 0 Defined
Kinematic hardening
D. Isotropic + Anisotropic
110 15 0.08 80 60 Defined
Rotational Kinematic hardening
* Properties for Elasticity and Barlat YLD96 are common.
131
350
Al-3%Cu
300
250
Stress (MPa)
200
150
100 RD
o
45
50
TD
0
0 0.05 0.1 0.15 0.2 0.25 0.3
Compressive strain
Figure 6.1: Stress flow measurements of each direction for Al-3%Cu precipitate
hardened material. Hardening of 45° shows more anisotropy than the RD and TD.
132
300
250
200
Stress (MPa)
150
100
Exp.
50 Stress (FEM)
0
0 0.05 0.1 0.15 0.2 0.25 0.3
Compressive strain
(a)
300
250
200
Stress (MPa)
150
100
Exp.
50 Stress (FEM)
0
0 0.05 0.1 0.15 0.2 0.25 0.3
Compressive strain
(b)
Figure 6.2: Neither conventional isotropic hardening (a) nor kinematic hardening (b)
with Barlat YLD96 anisotropic yield function does trace the measurements.
133
350
300
No spin
250 Spin
Stress (MPa)
200
150
100
-1
50 q=c M-sb M
m m
0
0 0.05 0.1 0.15 0.2 0.25 0.3
Compressive strain
(a)
350
Spin
300
No spin
250
Stress (MPa)
200
150
100
50 q=c M-sb M
m m
0
0 0.05 0.1 0.15 0.2 0.25 0.3
Compressive strain
(b)
Figure 6.3: Flow stresses of 45° for different backstress formulations. Those
formulations cannot trace the trend of anisotropic hardening shown in the measurement.
134
300
250
200
Stress (MPa)
150
Exp.
Stress (FEM)
100
Backstress (FEM)
50
0
0 0.05 0.1 0.15 0.2 0.25 0.3
Compressive strain
(a)
300
250
200
Stress (MPa)
150
Exp.
Stress (FEM)
100
Backstress (FEM)
50
0
0 0.05 0.1 0.15 0.2 0.25 0.3
Compressive strain
(b)
Figure 6.4: Flow stresses using the fourth-order tensor for precipitate inclusion without
rotation (a) and with rotation (b). The anisotropic hardening for 45° is captured by
rotation of backstress.
135
95
(A)
92.5
(B)
90 (D)
(C)
87.5
85
0 10 20 30 40 50 60 70 80 90
Figure 6.5: Earing pattern of cylindrical cup according to the hardening models: a)
Isotropic hardening only, b) Isotorpic + kinematic hardening, c) Isotropic + Anisotropic
kinematic hardeningm, d) Isotropic + Anisotropic kinematic hardening with rotation.
136
(a)
(b)
(c)
Figure 6.6: Residual stress distributions with respect to the hardening model: (a)
isotropic hardening, (b) anisotropic hardening with the fourth-order tensor, but no
rotation, (c) anisotropic hardening with the fourth-order tensor and rotation.
137
1.1
1.05
Thickness (mm)
1
(C) (D)
0.95
0.9
0.85 (B)
0.8
(A)
0.75
0 35 70 105
Distance from center, undeformed (mm)
138
CHAPTER 7
CONCLUSION
arbitrary orientation other than the RD and the TD. The number of material properties to
express the rotational hardening is two. A simple method to identify the required
parameters is introduced.
process of the cylindrical cup and square cup of the NUMISHEET benchmark problems.
An isotropic hardening and the ANK non-linear kinematic hardening model are used to
compare the results with that of the RIK hardening model and the averaged experimental
results. The RIK model shows more accurate results in the prediction of earing or draw-
For the springback prediction, draw bead simulation is done by all the three
the height and twisting mode. With the rotational hardening scheme, the stress
distributions after the drawing process are quite a lot different than the hardening without
139
Anisotropy of materials is evolving during deformation. For BCC mild steel, the
evolution is rotational and the effect is dominant to the yield function. For FCC materials,
such as aluminum alloy, the evolution appears more dominant to the backstress evolution.
Further research is required to assess the rotation of the symmetry axes using the
texture analysis. Also in order to generalize the model more experimental investigation is
needed for materials other than mild steel. It is necessary to model the yield function
evolution using the texture analysis directly into the yield function to predict material
140
BIBLIOGRAPHY
2. Banadic, D., Bunge, H.-J., Pohlandt, K., and Tekkaya, A.E., Formability of
Metallic Materials, Springer, (2000).
4. Kim, K.H. and Yin, J.J., Evolution of Anisotropy under Plane Stress, Journal of
Mechanics and Physics of Solids, 45, p.841-851, (1997).
6. Hill, R., A Theory of the Yielding and Plastic Flow of Anisotropic Metals, in Roy.
Soc. London A, p.281-297, (1948).
7. Chun, B.K., Jinn, J.T., and Lee, J.K., Modeling of the Bauschinger Effect for
Sheet Metals, Part I: Theory., International Journal of Plasticity, 18, p.571-595,
(2002).
8. Chun, B.K., Kim, H.Y., and Lee, J.K., Modeling the Bauschinger Effect for Sheet
Metals, Part II: Applications., International Journal of Plasticity, 18, p.597-616,
(2002).
9. Barlat, F., Liu, J., and C., B.J., Plastic Anisotropy in Precipitation-Strengthened
Aluminum Alloys Deformed in Uniaxial and Plane Strain Deformation,
Modelling Simul. Mater. Sci. Eng., p.435-443, (2000).
10. Barlat, F. and Liu, J., Precipitate-Induced Anisotropy in Binary Al-Cu Alloys,
Materials Science and Engineering, A257, p.47, (1998).
141
11. Barlat, F., Maedat, Y., Chung, K., Yanagawa, M., Brem, J.C., Hayashida, Y.,
Lege, D.J., Matsui, K., Murtha, S.J., Hattori, S., Becker, R.C., and Makosey, S.,
Yield Function Development for Aluminum Alloy Sheets., Journal of Mechanic
and Physics of Solids, (1997).
12. Kuroda, M., Interpretation of Behavior of Metals under Large Plastic Shear
Deformations: Comparison of Macroscopic Predictions to Physically Based
Predictions, Int. J. Plast., 15, p.1217-1236, (1999).
13. Kuroda, M., Interpretation of the Behavior of Metals under Large Plastic Shear
Deformations: A Macroscopic Approach, International Journal of Plasticity,
13(4), p.359-383, (1997).
15. Han, C.-S., Choi, Y., Lee, J.K., and Wagoner, R.H., A Fe Formulation for Elasto-
Plastic Materials with Planar Anisotropic Yield Functions and Plastic Spin,
International Journal of Solids and Structures, 39, p.5123-5141, (2002).
16. Truong Qui, K.H. and Lippmann, H., On the Impact of Local Rotation on the
Evolution of an Anisotropic Plastic Yield Condition, J. Mech. Phys. Solids, 49,
p.2577-2591, (2001).
17. Dafalias, Y.F., The Plastic Spin, Journal of Applied Mechanics, 52, p.865, (1985).
18. Kuroda, M., Plastic Spin Associated with Corner Theory of Plasticity, Int. J.
Plast., 11(5), p.547-570, (1995).
19. Simo, J.C. and Hughes, T.J.R., Computational Inelasticity, Springer, (1998).
20. Doucet, A.B. and Wagoner, R.H., Plane-Strain Work Hardening and Transient
Behavior of Interstitial-Free Steel, Metallurgical Transaction A, 18A, p.2139,
(1987).
21. Doucet, A.B. and Wagoner, R.H., Transient Tensile Behavior of Interstitial-Free
Steel and 70/30 Brass Following Plane-Strain Deformation, Metallurgical
Transaction A, 20A, p.1483, (1989).
22. Rauch, E.F., The Flow Law of Mild Steel under Monotonic or Complex Strain
Path, Solid State Phenomena, 23-24, p.317-334, (1992).
23. Thuillier, S. and Rauch, E.F., Development of Microbands in Mild Steel During
Cross Loading, Acta Metall. Mater., 42(6), p.1973-1983, (1994).
142
24. Hiwatashi, S., Van Bael, A., Van Houtte, P., and Teodosiu, C., Modelling of
Plastic Anisotropy Based on Texture and Dislocation Structure, Computational
Materials Science, 9, p.274-284, (1997).
25. Hiwatashi, S., Van Bael, A., Van Houtte, P., and Teodosiu, C., Prediction of
Forming Limit Strains under Strain-Path Changes: Application of an Anisotropic
Model Based on Texture and Dislocation Structure, Int. J. Plast., 14(7), p.647-669,
(1998).
26. Takahasi, H., Fujiwara, K., and Nakagawa, T., Multiple-Slip Work-Hardening
Model in Crystals with Application to Torsion-Tension Behavior of Aluminum
Tubes, International Journal of Plasticity, 14(6), p.489-509, (1998).
27. Teodosiu, C. and Hu, Z., Simulation of Materials Processing: Theory, Methods
and Applications, in Numiform 95, Balkema: Rotterdam, p.173, (1995).
28. Peeters, B., Seefeldt, M., Tedodosiu, C., Kalidindi, S. R., Van Houtte, P. and
Aernoudt, E., Work-Hardening/Softening Behavior of B.C.C. Polycrystals During
Changing Strain Paths: I. An Integrated Model Based on Substructure and Texture
Evolution, and Its Prediction of the Stress-Strain Behavior of an If Steel During
Two-Stage Strain Paths, Acta Materialaia, 49, p.1607-1629, (2001).
29. Peeters, B., Bacroix, B., Teodosiu, C., Van Houtte, P., and Aernoudt, E., Work-
Hardening/Softening Behavior of B.C.C. Polycrystals During Changing Strain
Paths: II. Tem Observations of Dislocation Sheets in an If Steel During Two-
Stage Strain Paths and Their Representation in Terms of Dislocations Densities,
Acta Materialia, 49, p.1621-1632, (2001).
30. Peeters, B., Kalidinidi, S.R., Teodosiu, C., Van Houtte, P., and Aernoudt, E.,
Modelling Microstructure-Based Anisotropy in Bcc Polycrystals During
Changing Strain Paths, Diffusion and Defect Data Pt. B: Solid State Phenomena,
87, p.163-168, (2002).
143
35. Truong Qui, K.H. and Lippmann, H., Plastic Spin and Evolution of an
Anisotropic Yield Condition, Int. J. Mech. Sci., 43, p.1969-1983, (2001).
36. Brocato, M., Tamagny, P., and Ehrlacher, A., Texture Evolution and Rotational
Hardening in Multiple Slip Plasticity: A Two Dimensional Study, Eur. J. Mech.
A/Solids, 20, p.345-365, (2001).
39. Chaboche, J.L., Constitutive Equations for Cyclic Plasticity and Cyclic
Viscoplasticity, Int. J. Plast., 5, p.247-302, (1989).
40. Loret, B., On the Effects of Plastic Rotation in the Finite Deformation of
Anisotropic Elastoplastic Materials, Mech. Mater., 2, p.287-304, (1983).
42. Dafalias, Y.F. and Rashid, M.M., The Effect of Plastic Spin on Anisotropic
Material Behavior, International Journal of Plasticity, 5, p.227-246, (1989).
43. Van der Giessen, E., Micromechanical and Thermodynamical Aspects of the
Plastic Spin, International Journal of Plasticity, 7, p.365-386, (1991).
44. Kuroda, M., Roles of Plastic Spin in Shear Banding, International Journal of
Plasticity, 12(5), p.671-698, (1996).
45. Zbib, H.M. and Aifantis, E.C., On the Concept of Relative and Plastic Spins and
Its Implications to Large Deformation Theories, Part I: Hypoelasticity and
Vertex-Type Plasticity., Acta Mechanica, 75, p.15, (1988).
46. Choi, Y., Han, C.-S., Lee, J.K., and Wagoner, R.H., Effect of Anisotropy Axes
Rotation on Spring Back, Numisheet 2002 proceedings, (2002).
47. Haupt, P. and Tsakmakis, C., On Kinematic Hardening and Large Plastic
Deformations, International Journal of Plasticity, 2, p.279-293, (1986).
48. Han, C.-S., Choi, Y., Lee, J.K., and Wagoner, R.H., Modeling of Anisotropic
Work Hardening, Numisheet 2002 proceedings, (2002).
144
49. Tsakmakis, C., Kinematic Hardening Rules in Finite Plasticity, Part I: A
Constitutive Approach, Continuum Mechanics and Thermodyanamics, 8(215-231),
(1996).
50. Han, C.-S., Chung, K., Wagoner, R.H., and Oh, S.-I., A Multiplicative Finite
Elasto-Plastic Formulation with Anisotropic Yield Functions, Int. J. Plast., 19,
p.197-211, (2003).
51. Geng, L. and Wagoner, R.H., Springback Analysis with a Modified Nonlinear
Hardening Model, in SAE, Detroit, p.SAE2000-01-0410, (2000).
52. Valliappan, S., Boonlaulohr, P., and Lee, I.K., Non-Linear Analysis for
Anisotropic Material, International Journal of Numerical Method of Engineering,
10, p.597-606, (1976).
54. Zhao, K. and Lee, J.K., Generation of Cyclic Stress-Strain Curves for Sheet
Metals, ASME MED, 11, p.667-674, (2000).
55. Hill, R., Constitutive Modelling of Orthotropic Plasticity in Sheet Metals, J. Mech.
Phys. Solids, 38, p.405-417, (1990).
56. Barlat, F., Lege, D.J., and Brem, J.C., A Six-Component Yield Function for
Anisotropic Materials, International Journal of Plasticity, 7, p.693-712, (1991).
57. Agnew, S.R. and Weertman, J.R., The Influence of Texture on the Elastic
Properties of Ultrafine-Grain Copper, Materials Science and Engineering, A242,
p.174-180, (1998).
58. Beaudoin, A.J., Dawson, P.R., Mathur, K.K., Kocks, U.F., and Korzekwa, D.A.,
Application of Polycrystal Plasticity to Sheet Forming, Computer Methods in
Applied Mechanics and Engineering, 117, p.49-70, (1994).
59. Kocks, U.F., Tome, C.N., and Wenk, H.-R., Texture and Anisotropy, Cambridge
University Press, (1998).
60. Nakamachi, E. and Doug, X., Study of Texture Effect on Sheet Failure in a Limit
Dome Height Test by Using Elastic/Crystalline Viscoplastic Finite Element
Analysis, Journal of Applied Mechanics, 64, p.519-524, (1997).
61. Yang, D.-Y., Oh, S.I., Huh, H., and Kim, Y.H., eds., Numisheet 2002 Proceedings
-Verification of Simulation with Experiment-. Proceedings of the 5th International
145
Conference and Workshop on Numerical Simulation of 3d Sheet Forming
Processes, ed. Series, Vol. 2: Jeju Island, (2002).
62. Sutton, M.A., Wolters, W.J., Peters, W.H., and Chao, Y.J., Determination of
Displacements Using an Improved Digital Correlation Method, Image Vision
Computing, 1(3), p.133-139, (1983).
64. Rao, K.P. and Mohan, E.V.R., A Vision-Integrated Tension Test for Use in Sheet-
Metal Formability Studies, Journal of Materials Processing Tech., 118, p.238-245,
(2001).
65. Zhao, K. and Lee, J.K., Generation of Cyclic Stress-Strain Curves for Sheet
Metals, Journal of Engineering Materials and Technology, 124, p.391-397,
(2001).
66. Kubin, L., Canova, G.R., Condat, M., Devinere, B., Pontikis, V., and Brechet, Y.,
Dislocation Microstructure and Plastic Flow: A 3d Simulation, Solid State
Phenom., 23-24, p.455, (1992).
67. Peeters, B., Kalidinidi, S.R., Teodosiu, C., and Aernoudt, E., A Theoretical
Investigation of the Influence of Dislocation Sheets on Evolution of Yield
Surfaces in Single-Phase B.C.C. Polycrystals, Journal of Mechanics and Physics
of Solids, 50, p.783-807, (2002).
68. Shizawa, K. and Zbib, H.M., Thermodynamical Theory of Plastic Spin and
Internal Stress with Dislocation Density Tensor, J. Eng. Mater. Technol. Trans.
ASME, 121(2), p.247-253, (1999).
70. Geng, L. and Wagoner, R.H., Role of Plastic Anisotropy and Its Evolution on
Springback, Int. J. Mech. Sci., 44(1), p.123-148, (2002).
71. Schmitt, J.H., Shen, E.L., and Raphanel, J.L., A Parameter for Measuring the
Magnitude of a Change of Strain Path: Validation and Comparison with
Experiments on Low Carbon Steel, Int. J. Plast., 10(5), p.535-551, (1994).
77. Hosford, W.F. and Zeisloft, R.H., The Anisotropy of Age-Hardened Al-4%Cu
Single Crystals During Plane-Strain Compression, Metall. Trans. A, 3, p.113-121,
(1972).
78. Oldroyd, J.G., On the Formulation of Rheological Equations of State, in Roy. Soc.
Lond., p.523, (1950).
147
APPENDICES
148
APPENDIX A
EXPLICIT FORMULATION
fi
∆γ i +1 = ∆γ i − (A.1)
( )
i
∂f
∂∆γ
i
∂f
where superscript i is iterative step. The derivation of is as follows.
∂∆γ
The yield function f is
2
f = φ− τy (A.2)
3
where φ and τy are defined as Hill’s 48 anisotropic yield function and the size of
2
φ= τ ⋅ Pτ
3 . (A.3)
τy = Ciso ( ε + ∆εp )
* Niso
149
Take a derivative from (A.2)
∂f ∂φ ∂τ 2 ∂τy ∂∆εp
= ⋅ − ⋅ (A.4)
∂∆γ ∂τ ∂∆γ 3 ∂∆εp ∂∆γ
where
∂φ
τ = τ 0 − Γ e ∆ε p = τ 0 − ∆γΓ e
, and (A.5)
∂τ
2 2 2
∆εp = ∆ε p = ∆γ τ ⋅ Pτ = ∆γ τ ⋅ Pτ = ∆γφ . (A.6)
3 3 3
From (A.5),
∂τ ∂φ
= −Γe + H.O.T. (A.7)
∂∆γ ∂τ
∂∆εp
= φ. (A.8)
∂∆γ
∂f ∂φ ∂φ 2 ∂τy
= − ⋅ Γe − φ. (A.9)
∂∆γ ∂τ ∂τ 3 ∂∆εp
Using the value computed at (A.9), the Newton-Raphson iteration (A.1) is used to find
150
APPENDIX B
From the step n to the step n+1, the strain increments ∆ε , and the variables
n
τ n , y n , q1n , q n2 , ε np , ε pn , e φα from the previous load step n are assumed to be given. For the
i
which correlates to the Oldroyd derivative for stress like variables (.) = (.)− L (.) − (.) LT ,
∇
setting [19, 78]. Γ e in (B.1) denotes the elasticity tensor. It may be helpful to formulate
the equation (B.1) relative to the referential coordinate system eiφ , because formulation is
for the plane stress set which is defined in the tangential coordinate system of the shell as,
2
y trial = ∑ q trial
j , qj
trial
= Fq nj F T
j=1
(B.3)
151
The trial isotropic hardening is set to
( )
2
Φ trial = φtrial − σ trial
y , (B.5)
( ) (
where φtrial = τ trial − y trial iK φtrial τ trial − y trial . )
Elastic loading
(
Φ trial = Φ τ trial , y trial , σ trial
y ; eα
φ trial
≤ TOL ) (B.6)
Plastic loading
In the case of plastic loading, Φ trial ≥ 0 the consistency of the yield function can
∆Φ = ∆φ − 2σ y ∆σ y (B.8)
∂φ
∆φ = ⋅ ∆σ, σ = τ − y . (B.9)
∂σ
where
152
∂φ
= 2K φσ = a (B.10)
∂σ
Residual method:
∆γ i +1 = ∆γ i + ∆ 2 γ i (B.11)
With the residual method, the increment of the consistent parameter is formulated as
∆ γ =
2 i (
Φ i − ai ⋅ Q −1rτi − ryi ) (B.12)
(a ⋅ Q
i −1
Γ ea + A + A ikin
i i
iso )
∂ai
where Qi = 1 + ∆γ i Γ ebi , bi = = 2K iφ . (B.13)
∂σ i
Where the residual values used in equation (B.12) are for the stress residual,
(
rτi = τ i − τ trial − ∆γΓ eai , ) (B.14)
(
rqi1 = q1i − q1tr + 2c1∆γ i σ i − 2∆γ i σiy b1q1i , ) (B.16)
(
rqi 2 = qi2 − q 2tr + 2kc 2 ∆γ i qi2 ) (B.17)
153
∂σ y
A iso = 4Hσ2y where H = and (B.18)
∂εp
A kin = 2a ⋅ {( d1c1 + d 2 kc 2 ) σ − d1σ y b1q1} (B.19)
Once the increment of consistent parameter is set with (B.12) then the increments
∆γ i +1 = ∆γ i + ∆2 γ i (B.20)
∆τ = −Q
i −1
(r
i
τ + ∆ γ Γ ea
2 i i
) (B.21)
(
∆q1i = −d1 rqi1 − 2c1∆ 2 γ iσ i + 2c1∆γ iQ −1rτi + 2∆ 2 γ i σiy b1q1i ) (B.22)
n
∆y i +1 = ∑ ∆qij+1 (B.24)
j=1
For the reorientation of anisotropy axis using the plastic spin is computed from
∆ω ip+1 = µ φ (τ∆ε p − ∆ε p τ ) .
i +1
(B.25)
The plastic spin is an anti-symmetric tensor, so the components in matrix form are
0 ∆ω12
∆ω ip+1 = . (B.26)
− ∆ω12 0
From the non-diagonal terms and the assumption of keeping orthogonal between
the two planar anisotropy axes the anisotropy direction evolution is driven.
154
(K )i +1
φ ijkl
= ∑ 0
K αβγδ ( )( )( )(
eii +1 ⋅ eα0 eij+1 ⋅ eβ0 eik+1 ⋅ e0γ eik+1 ⋅ eδ0 ) (B.28)
αβγδ=1,2
( )
2
Φ i +1 = φi +1 − σiy+1 (B.29)
If it returns greater than tolerance, repeat the iteration. If it returns less than or
equal to the tolerance, then transfer the last iterative values to next step values and quit
2
τ n +1 = τ i , y n +1 = ∑ q nj +1 , q1n +1 = q1i , q n2 +1 = q i2 ε pn +1 = ε pi , e φn +1 = e iφ (B.30)
j=1
Γ ep = M −
( Ma ) ⋅ ( Ma ) , (B.31)
a ⋅ Ma + A iso + A kin
where M = Q −1Γe .
155