Sheet Metal Forming Processes

Dorel Banabic

Sheet Metal Forming

Processes

Constitutive Modelling and Numerical

Simulation

123
Prof. Dr. Ing. Dorel Banabic
Technical University of Cluj-Napoca
Research Centre on Sheet Metal
Forming – CERTETA
27 Memorandumului
400114 Cluj Napoca
Romania
[email protected]

ISBN 978-3-540-88112-4 e-ISBN 978-3-540-88113-1

DOI 10.1007/978-3-540-88113-1
Springer Heidelberg Dordrecht London New York

Library of Congress Control Number: 2010927076

© Springer-Verlag Berlin Heidelberg 2010

This work is subject to copyright. All rights are reserved, whether the whole or part of the material is
concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting,
reproduction on microfilm or in any other way, and storage in data banks. Duplication of this publication
or parts thereof is permitted only under the provisions of the German Copyright Law of September 9,
1965, in its current version, and permission for use must always be obtained from Springer. Violations
are liable to prosecution under the German Copyright Law.
The use of general descriptive names, registered names, trademarks, etc. in this publication does not
imply, even in the absence of a specific statement, that such names are exempt from the relevant protective
laws and regulations and therefore free for general use.

Cover design: Frido Steinen

Printed on acid-free paper

Springer is part of Springer Science+Business Media (www.springer.com)

Preface

The concept of virtual manufacturing has been developed in order to increase the
industrial performances, being one of the most efficient ways of reducing the man-
ufacturing times and improving the quality of the products. Numerical simulation
of metal forming processes, as a component of the virtual manufacturing process,
has a very important contribution to the reduction of the lead time. The finite
element method is currently the most widely used numerical procedure for sim-
ulating sheet metal forming processes. The accuracy of the simulation programs
used in industry is influenced by the constitutive models and the forming limit
curves models incorporated in their structure. From the above discussion, we can
distinguish a very strong connection between virtual manufacturing as a general
concept, finite element method as a numerical analysis instrument and constitutive
laws, as well as forming limit curves as a specificity of the sheet metal forming
processes. Consequently, the material modeling is strategic when models of reality
have to be built.
The book gives a synthetic presentation of the research performed in the
field of sheet metal forming simulation during more than 20 years by the
members of three international teams: the Research Centre on Sheet Metal
Forming—CERTETA (Technical University of Cluj-Napoca, Romania); AutoForm
Company from Zürich, Switzerland and VOLVO automotive company from
Sweden.
The first chapter presents an overview of different Finite Element (FE) formula-
tions used for sheet metal forming simulation, now and in the past. The objective
of this chapter is to give a general understanding of the advantages and disadvan-
tages of the various methods in use. The first section is dedicated to some of the
necessary ingredients of the fundamentals of continuum mechanics for large defor-
mation problems. These are needed for a better understanding of the forthcoming
FE-formulations.
A more extended chapter is devoted to the presentation of the phenomenologi-
cal yield criteria. Due to the fact that this chapter is only a synthetic overview of
the yield criteria, the reader interested in some particular formulation should also
read the original paper listed in the reference section. We have tried to use the sym-
bols adopted by the authors, especially in the mathematical relationships defining

v
vi Preface

the yield stresses and the coefficients of plastic anisotropy. This decision has been
made in order to facilitate the reading of the original papers. Of course, under these
circumstances, the coherency of the notations cannot be preserved. As one may see
in the list of symbols, several identifiers have different meanings. The reader should
take this aspect into account. This chapter gives a more detailed presentation of the
yield criteria implemented in the commercial programs used for the finite element
simulation (emphasizing the formulations proposed by the CERTETA team—BBC
models—implemented in the AutoForm commercial code) or the yield criteria hav-
ing a major impact on the research progress. To improve the springback prediction
a novel approach to model the Bauschinger effect has been developed and imple-
mented in the commercial code AutoForm. Consequently, an extended section of
this chapter has been dedicated to the modeling of the Bauschinger effect, especially
in the AutoForm model.
The sheet metal formability is discussed in a separate chapter. After present-
ing the methods used for the formability assessment, the discussion focuses on the
Forming Limit Curves (FLC). Experimental methods used for limit strains determi-
nation and the main factors influencing the FLC are presented in detail. A section is
dedicated to the use of Forming Limit Diagrams in industrial practice. Theoretical
predictions of the FLCs are presented in an extended section. In this context, the
authors emphasize their contributions to the mathematical modeling of FLCs. A
special section has been devoted to present an original implicit formulation of the
Hutchinson–Neale model, developed by the authors of this chapter, used for cal-
culating the FLCs of thin sheet metals. The commercial programs (emphasizing
the FORM CERT program) and the semi-empirical models for FLC prediction are
presented in the last sections of the chapter.
The aspects related to the numerical simulation of the sheet metal forming pro-
cesses are discussed in the last chapter of the book. The role of simulation in process
planning, part feasibility and quality, process validation and robustness are presented
based on the AutoForm solutions. The performances of the material models are
proved by the numerical simulation of various sheet metal forming processes: bulge
and stretch forming, deep-drawing and forming of the complex parts. A section has
been devoted to the robust design of sheet metal forming processes. Springback is
the major quality concern in the stamping field. Consequently, two sections of this
chapter are focused on the springback analysis and Computer Aided Springback
Compensation (CASP).
The authors wish to express their gratitude to Dr. Waldemar Kubli, founder and
CEO, Dr. Mike Selig, CTO and Markus Thomma, CMD of AutoForm Company,
for their support of the book project. They have created favorable conditions for the
AutoForm team in order to make this book possible. The authors also wish to thank
Dr. Alan Leacock from University of Ulster (UK) for his help in proofing the English
of the manuscript. Prof. Banabic wishes to express his thanks to his former PhD
students Dr. L. Paraianu, Dr. P. Jurco, Dr. M. Vos, Dr. G. Cosovici and his current
PhD students G. Dragos and I. Bichis for their help in preparing and editing this
book.
Preface vii

The book will be of interest to both the research and industrial communities. It is
useful for the students, doctoral fellows, researchers and engineers who are mainly
interested in the material modeling and numerical simulation of sheet metal forming
processes.

Cluj-Napoca, Romania Dorel Banabic

December 2009
Contents

1 FE-Models of the Sheet Metal Forming Processes . . . . . . . . . . . 1

1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.2 Fundamentals of Continuum Mechanics . . . . . . . . . . . . . . 3
1.2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.2.2 Strain Measures . . . . . . . . . . . . . . . . . . . . . . . 4
1.2.3 Stress Measures . . . . . . . . . . . . . . . . . . . . . . . 8
1.3 Material Models . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.4 FE-Equations for Small Deformations . . . . . . . . . . . . . . . 11
1.5 FE-Equations for Finite Deformations . . . . . . . . . . . . . . . 13
1.6 The ‘Flow Approach’—Eulerian FE-Formulations
for Rigid-Plastic Sheet Metal Analysis . . . . . . . . . . . . . . . 16
1.7 The Dynamic, Explicit Method . . . . . . . . . . . . . . . . . . . 18
1.8 A Historical Review of Sheet Forming Simulation . . . . . . . . . 21
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
2 Plastic Behaviour of Sheet Metal . . . . . . . . . . . . . . . . . . . . 27
2.1 Anisotropy of Sheet Metals . . . . . . . . . . . . . . . . . . . . . 30
2.1.1 Uniaxial Anisotropy Coefficients . . . . . . . . . . . . . . 30
2.1.2 Biaxial Anisotropy Coefficient . . . . . . . . . . . . . . . 36
2.2 Yield Criteria for Isotropic Materials . . . . . . . . . . . . . . . . 39
2.2.1 Tresca Yield Criterion . . . . . . . . . . . . . . . . . . . . 41
2.2.2 Huber–Mises–Hencky Yield Criterion . . . . . . . . . . . 42
2.2.3 Drucker Yield Criterion . . . . . . . . . . . . . . . . . . . 43
2.2.4 Hershey Yield Criterion . . . . . . . . . . . . . . . . . . . 44
2.3 Classical Yield Criteria for Anisotropic Materials . . . . . . . . . 45
2.3.1 Hill’s Familly Yield Criteria . . . . . . . . . . . . . . . . 45
2.3.2 Yield Function Based on Crystal Plasticity
(Hershey’s Familly) . . . . . . . . . . . . . . . . . . . . . 61
2.3.3 Yield Criteria Expressed in Polar Coordinates . . . . . . . 74
2.3.4 Other Yield Criteria . . . . . . . . . . . . . . . . . . . . . 75
2.4 Advanced Anisotropic Yield Criteria . . . . . . . . . . . . . . . . 76
2.4.1 Barlat Yield Criteria . . . . . . . . . . . . . . . . . . . . . 77
2.4.2 Banabic–Balan–Comsa (BBC) Yield Criteria . . . . . . . 81

ix
x Contents

2.4.3 Cazacu–Barlat Yield Criteria . . . . . . . . . . . . . . . . 84

2.4.4 Vegter Yield Criterion . . . . . . . . . . . . . . . . . . . . 87
2.4.5 Polynomial Yield Criteria . . . . . . . . . . . . . . . . . . 88
2.5 BBC 2005 Yield Criterion . . . . . . . . . . . . . . . . . . . . . 91
2.5.1 Equation of the Yield Surface . . . . . . . . . . . . . . . . 91
2.5.2 Flow Rule Associated to the Yield Surface . . . . . . . . . 92
2.5.3 BBC 2005 Equivalent Stress . . . . . . . . . . . . . . . . 92
2.5.4 Identification Procedure . . . . . . . . . . . . . . . . . . . 94
2.5.5 Particular Formulations of the BBC 2005 Yield Criterion . 105
2.6 BBC 2008 Yield Criterion . . . . . . . . . . . . . . . . . . . . . 106
2.6.1 Equation of the Yield Surface . . . . . . . . . . . . . . . . 107
2.6.2 BBC 2008 Equivalent Stress . . . . . . . . . . . . . . . . 108
2.6.3 Identification Procedure . . . . . . . . . . . . . . . . . . . 109
2.7 Recommendations on the Choice of the Yield Criterion . . . . . . 113
2.7.1 Comparison of the Yield Criteria . . . . . . . . . . . . . . 113
2.7.2 Evaluating the Performances of the Yield Criteria . . . . . 116
2.7.3 Mechanical Parameters Used by the Identification
Procedure of the Yield Criteria . . . . . . . . . . . . . . . 118
2.7.4 Implementation of the Yield Criteria in Numerical
Simulation Programmes . . . . . . . . . . . . . . . . . . . 118
2.7.5 Overview of the Anisotropic Yield Criteria Developing . . 120
2.7.6 Perspectives . . . . . . . . . . . . . . . . . . . . . . . . . 120
2.8 Modeling of the Bauschinger Effect . . . . . . . . . . . . . . . . 121
2.8.1 Reversal Loading in Sheet Metal Forming Processes . . . . 121
2.8.2 Experimental Observations . . . . . . . . . . . . . . . . . 122
2.8.3 Physical Nature of the Bauschinger Effect . . . . . . . . . 124
2.8.4 Phenomenological Modelling . . . . . . . . . . . . . . . . 125
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
3 Formability of Sheet Metals . . . . . . . . . . . . . . . . . . . . . . . 141
3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142
3.2 Evaluation of the Sheet Metal Formability . . . . . . . . . . . . . 147
3.2.1 Methods Based on Simulating Tests . . . . . . . . . . . . 147
3.2.2 Limit Dome Height Method . . . . . . . . . . . . . . . . . 151
3.3 Forming Limit Diagram . . . . . . . . . . . . . . . . . . . . . . . 152
3.3.1 Definition: History . . . . . . . . . . . . . . . . . . . . . 152
3.3.2 Experimental Determination of the FLD . . . . . . . . . . 156
3.3.3 Methods of Determining the Limit Strains . . . . . . . . . 162
3.3.4 Factors Influencing the FLC . . . . . . . . . . . . . . . . 165
3.3.5 Use of Forming Limit Diagrams in Industrial Practice . . . 175
3.4 Theoretical Predictions of the Forming Limit Curves . . . . . . . 179
3.4.1 Swift’s Model . . . . . . . . . . . . . . . . . . . . . . . . 180
3.4.2 Hill’s Model . . . . . . . . . . . . . . . . . . . . . . . . . 182
3.4.3 Marciniak–Kuckzynski (M–K) and
Hutchinson–Neale (H–N) Models . . . . . . . . . . . . . 182
Contents xi

3.4.4 Implicit Formulation of the M–K and H–N Models . . . . 185

3.4.5 Linear Perturbation Theory . . . . . . . . . . . . . . . . . 194
3.4.6 Modified Maximum Force Criterion (MMFC) . . . . . . . 195
3.5 Commercial Programs for FLC Prediction . . . . . . . . . . . . . 197
3.5.1 FORM-CERT Program . . . . . . . . . . . . . . . . . . . 198
3.6 Semi-empirical Models . . . . . . . . . . . . . . . . . . . . . . . 203
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 204
4 Numerical Simulation of the Sheet Metal Forming Processes . . . . 213
4.1 AutoForm Solutions . . . . . . . . . . . . . . . . . . . . . . . . . 213
4.1.1 The Role of Simulation in Process Planning . . . . . . . . 213
4.1.2 Material Data in Digital Process Planning . . . . . . . . . 215
4.1.3 Feasibility (Part Feasibility) . . . . . . . . . . . . . . . . . 218
4.1.4 Manufacturability (Process Validation) . . . . . . . . . . . 225
4.1.5 Capability (Robustness) . . . . . . . . . . . . . . . . . . . 230
4.1.6 Simulation Result ‘Quality’ . . . . . . . . . . . . . . . . . 236
4.1.7 Comprehensive Digital Process Planning . . . . . . . . . . 236
4.2 Simulation of the Elementary Forming Processes . . . . . . . . . 238
4.2.1 Simulation of the Bulge Forming Process . . . . . . . . . 238
4.2.2 Simulation of Stretch Forming of Spherical Cup . . . . . . 241
4.2.3 Simulation of Cross Die . . . . . . . . . . . . . . . . . . . 244
4.3 Simulation of the Industrial Parts Forming Processes . . . . . . . 250
4.3.1 Simulation of an Outer Trunklid . . . . . . . . . . . . . . 251
4.3.2 Simulation of a Sill Reinforcement for Volvo C30 . . . . . 254
4.4 Robust Design of Sheet Metal Forming Processes . . . . . . . . . 255
4.4.1 Variability of the Material Parameters . . . . . . . . . . . 256
4.4.2 AutoForm-Sigma . . . . . . . . . . . . . . . . . . . . . . 257
4.4.3 Robust Design: Case Studies . . . . . . . . . . . . . . . . 258
4.4.4 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . 267
4.5 The Springback Analysis . . . . . . . . . . . . . . . . . . . . . . 267
4.5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 267
4.5.2 Example Description . . . . . . . . . . . . . . . . . . . . 268
4.5.3 The Influences on the Accuracy of Springback Simulation . 269
4.5.4 The Optimized Numerical Parameters of
Springback Simulation: Final Validation Settings . . . . . 277
4.5.5 The Simulation of Numisheet 2005 Benchmark #1:
Decklid Inner Panel . . . . . . . . . . . . . . . . . . . . . 277
4.5.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . 281
4.6 Computer Aided Springback Compensation . . . . . . . . . . . . 282
4.6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 282
4.6.2 The Basic Methodologies of Computer-Aided
Springback Compensation . . . . . . . . . . . . . . . . . 283
4.6.3 The Influences of the Quality of Computer Aided
Springback Compensation . . . . . . . . . . . . . . . . . 284
xii Contents

4.6.4 The Recommended Work Flow of Computer-Aided

Springback Compensation . . . . . . . . . . . . . . . . . 285
4.6.5 The Springback Compensation of Numisheet 2005
Benchmark #1 . . . . . . . . . . . . . . . . . . . . . . . . 287
4.6.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . 293
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 294
Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 297
List of the Authors

Prof. Dorel Banabic

Professor at the Technical University of Cluj-Napoca
Director of the Research Centre in Sheet Metal Forming – CERTETA
27 Memorandumului, 400114 Cluj Napoca, Romania
e-mail: [email protected]
URL: www.certeta.utcluj.ro

Dr. Bart Carleer

AutoForm Engineering Deutschland GmbH
Emil-Figge-Str. 76-80, 44227 Dortmund, Germany
e-mail: [email protected]
URL: www.autoform.com

Dr. Dan-Sorin Comsa

Reader at the Technical University of Cluj Napoca
15 C. Daicoviciu, 400020 Cluj Napoca, Romania
e-mail: [email protected]
URL: www.certeta.utcluj.ro

Eric Kam
AutoForm Engineering USA, Inc.
560 Kirts Blvd, Suite 113, Troy, Michigan 48084-4141, USA
e-mail: [email protected]
URL: www.autoform.com

Dr. Andriy Krasovskyy

Formerly AutoForm Development GmbH
Technoparkstrasse 1, CH-8005 Zurich, Switzerland
URL: www.autoform.com

xiii
xiv List of the Authors

Prof. Kjell Mattiasson

Chalmers University of Technology
SE-412 96 Goteborg, Sweden
e-mail: mailto:[email protected]
URL: www.chalmers.se

Volvo Cars Safety Centre

Dept. 91432/PV 22, SE-405 31 Goteborg, Sweden
e-mail: [email protected]
URL: www.volvocars.com

Dr. Matthias Sester

AutoForm Development GmbH
Technoparkstrasse 1, CH-8005 Zurich, Switzerland
e-mail: [email protected]
URL: www.autoform.com

Mats Sigvant PhD

Technical Expert, Sheet Metal Forming Simulation
Stamping CAE, Volvo Car Corporation
Dept. 81153/26HK3, Olofstrom, Sweden
e-mail: [email protected]
URL: www.volvocars.com

Xiaojing Zhang PhD

AutoForm Engineering Deutschland GmbH
Emil-Figge-Str. 76-80, 44227 Dortmund, Germany
e-mail: [email protected]
URL: www.autoform.com
Contributions of the Authors

Dorel Banabic Co-ordination of the book

Sections 2.1–2.4, 2.7; Chapter 3
Bart Carleer Section 4.4
Dan-Sorin Comsa Sections 2.5, 2.6, 3.4.4
Eric Kam Section 4.1
Andriy Krasovskyy Section 2.8
Kjell Mattiasson Chapter 1
Matthias Sester Sections 4.2.1, 4.2.3
Mats Sigvant Sections 4.2.2, 4.3
Xiaojing Zhang Sections 4.5, 4.6

xv
Chapter 1
FE-Models of the Sheet Metal Forming
Processes

List of Special Symbols

[A] matrix with constants describing the anisotropy

of the material
[B] strain matrix
[BL ] linear strain matrix
[BNL ] nonlinear strain matrix
c speed of sound
d rate of deformation tensor
dp plastic rate of deformation
D elastic constitutive tensor
[D] matrix of the elastic constitutive tensor
[DT ] matrix form of the constitutive tensor
ds lengths of the vector dx
dS lengths of the vector dX
dv volume element in the current configuration
dV volume element in the reference configuration
E Young’s modulus
E 1 , E2 principal values of Lagrangian strains
E Green’s strain tensor
Ė Green strain rate tensor
f yield function
f load
{f ext } external force vector
{f int } internal force vector
F deformation gradient tensor
[K] linear stiffness matrix
[KG ] geometric stiffness matrix
[KM ] material stiffness matrix
[KS ] initial stress stiffness matrix
[KT ] tangent matrix
L lenght
L velocity gradient tensor

D. Banabic, Sheet Metal Forming Processes, DOI 10.1007/978-3-540-88113-1_1, 1


C Springer-Verlag Berlin Heidelberg 2010
2 1 FE-Models of the Sheet Metal Forming Processes

[M] consistent mass matrix

n̂ unit vector
[ N(x) ] matrix with base functions
S surface area
S second Piola-Kirchhhoff (PK2) stress
t time
t traction or stress vector
{ũ} vector with nodal displacements
u̇ velocity
ü acceleration
V volume
W spin tensor
x position vector
X position of the particle in the reference configuration
ε̄ effective strain
ε̄p effective plastic strain
p
ε̄˙ effective plastic strain rate
ξ fraction of critical damping of the highest eigen-mode.
Λ1 , Λ2 principal stretch values
λ̇ scalar in the associated flow rule
ν Poisson’s ratio
ρ mass densities in the current configuration
ρ0 mass densities in the reference configuration
{} residual vector
σ Cauchy stress
σ̂ Jaumann or co-rotational rate
σ∇ objective stress rate tensor
σ̄ effective stress
τ Kirchhoff stress
ωmax maximum eigen-frequency

1.1 Introduction
In the current section an overview of different Finite Element (FE) formulations
used for sheet metal forming simulation, now and in the past, will be given. The
theories of FE-simulation of large deformation problems will be briefly touched
upon herein, but for thorough presentations the reader is referred to the many exist-
ing text books on the subject, e.g. Belytschko et al. [1], Zienkiewicz and Taylor [2]
and Crisfield [3]. The object is rather to give a general understanding of the advan-
tages and disadvantages of the various methods in use. Review articles on sheet
forming simulation and comparative studies of different FE-procedures are found in
e.g. Honecker and Mattiasson [4], Oñate and Agelet de Saracibar [5], Oñate et al.
[6], Mattiasson [7], Kawka et al. [8], Wenner [9], Wang et al. [10], Mattiasson [11],
1.2 Fundamentals of Continuum Mechanics 3

Makinouchi [12] and Wenner [13]. State-of-the-art articles on the utilization of sheet
forming simulation today, and outlooks against the future are given in Banabic and
Tekkaya [14] and Roll and Weigand [15].
Finite Element Methods (FEM) have been developed and used for sheet form-
ing simulations since the 1970s, when the continuum mechanics foundations for
problems involving large displacements and large strains became well established.
FE-procedures for sheet forming analysis can be classified into two main groups
depending on if they are based on an elastic-plastic or a rigid-plastic material model.
Large strain formulations may be based on either an Eulerian or a Lagrangian
description of motion, leading to two basically different FE-procedures with nodal
velocities and nodal incremental displacements, respectively, as primary unknowns.

1.2 Fundamentals of Continuum Mechanics

1.2.1 Introduction
In [16] and [17] an extended presentation of the fundamentals of continuum
mechanics for large deformation problems and the theory of phenomenological
plasticity are given. Here, some of the necessary ingredients for the forthcoming
FE-formulations will be briefly presented.
When discussing the kinematics of continua it is important to make clear the
meanings of the terms point and particle. The word point will be used to designate
a certain location in space, while the word particle will denote a small part of a
material continuum. There are basically two ways of describing the motion of such
a continuum.
In the material or Lagrangian descrition the independent variables are the parti-
cle P and the time t. The motion may then be expressed by an equation of the form

x = x(P, t ) (1.1)

where x is the position vector of the particle P at time t. It is common to

write this equation in terms of the position X of the particle P in the reference
configuration, i.e.

x = x( X, t ) (1.2)

In the spatial or Eulerian description attention is fixed to a given region in space

(a point) instead of a certain particle of a continuum. Independent variables are the
present time t and the present position x of the particle that occupied the point X at
time t=0. The motion is thus expressed by

X = X( x, t ) (1.3)
4 1 FE-Models of the Sheet Metal Forming Processes

If Eqs. (1.2) and (1.3) represent one-to-one mappings with continuous partial
derivatives, the two mappings are unique inverses of each other.
The Eulerian description is the description best suited for fluid mechanics prob-
lems, as it focuses attention on a certain region in space, which enables us to
observe the flow in a point in a channel or a wind tunnel. The Lagrangian for-
mulation has traditionally been used in solid and structural mechanics problems, in
which there normally exist a natural reference configuration with known stresses and
deformations. In a Finite Element context, the primary variables in the Lagrangian
formulation are displacements, while in the Eulerian formulation they are velocities.
In some metal forming processes, especially bulk forming processes, the metal
flow resembles that of a fluid, and the problems have consequently been solved by
means of the Eulerian formulation. But even a problem like sheet metal forming has
been analyzed by means of this approach as will be discussed in one of the coming
sections.
In the following we will focus on the Lagrangian approach. To generalize a
formulation for small deformations to large deformations adds a great deal of com-
plexity. There are a number of different ways to formulate and solve the problem.
Here we will only present a couple of possible alternatives. As mentioned in the
introduction of this chapter, the reader is referred to one or more of the previously
mentioned textbooks to get a more complete coverage of the subject.
The main challenge when formulating the basic equations in the Lagrangian for-
mulation is to do it in such a way that they become independent of rigid-body
rotations. This means that a rigid body rotation should not give rise to additional
strains and stresses. We say that the formulation must be objective or frame invari-
ant, which also implies that the strain and stress measures being a part of the
formulation also must satisfy the objectivity requirement. There are a number of
such strain and stress measures appearing in the literature, but we will here limit our
discussion to only a few of them.

1.2.2 Strain Measures

We will first introduce a tensor called the deformation gradient tensor F, defined by

∂x ∂ xi
F = ; FiJ = . (1.4)
∂X ∂ XJ
This tensor is not objective, but plays an important role in the derivation of the above
mentioned strain and stress tensors. The tensor relates a line segment in current and
reference configurations, respectively, as

dx = F · dX ; d xi = FiJ d XJ (1.5)

To emphasize the distinction between the two sets of coordinates, we will use capital
letters for indices on tensor components referred to the reference configuration and
1.2 Fundamentals of Continuum Mechanics 5

lower case letters on those referred to the current configuration. The deformation
gradient tensor is said to be a two-point tensor, since its components are referred to
both the reference and the current configuration.
Let dS and ds denote the lengths of the vectors dX and dx, respectively. The
squares of these lengths may be written as

d S2 = |d X |2 = d X · d X = d XI d XI
(1.6)
d s2 = |d x|2 = d x · d x = d xi d xi

Using Eq. (1.5), we can rewrite ds2 as

d s2 = ( F · d X ) · ( F · d X) = d X · ( FT · F) · d X
(1.7)
d s2 = (FkI dXI ) (FkJ dXJ ) = FkI FkJ dXI dXJ

The difference ds2 –dS2 for two neighboring particles of a continuum is used as a
measure of deformation. This difference can be written

d s2 − d S2 = d X · ( FT · F) · d X − d X · d X = d X · ( FT · F − I) · d X =
= 2 dX · E · dX
d s2 − d S2 = FkI FkJ dXI dXJ − dXI dXI = ( FkI FkJ − δIJ ) dXI dXJ =
= 2 EIJ dXI dXJ
(1.8)
From Eq. (1.8) the definition of Green’s strain tensor is found to be

1 T 1
E = (F · F − I) ; EIJ = (FkI FkJ − δIJ ) (1.9)
2 2
In a rigid body rotation the difference ds2 –dS2 is constant. This can only be
accomplished if also the tensor E is constant, which proves that the tensor is
invariant under a rigid body rotation and, thus, objective.
An especially useful form of the strain tensor is obtained when it is expressed in
displacement gradients. Define the displacement u from the following relation:

x = X + u; xI = XI + uI (1.10)

Introducing this expression into the equation for F in Eq. (1.4), we get

∂x ∂u ∂ xi ∂ ui
F = = I + ; FiJ = = δiJ + (1.11)
∂X ∂X ∂ XJ ∂ XJ
Green’s strain tensor can then be rewritten as
  
∂u T ∂u T
  
1 ∂u ∂u
E = + + ·
2 ∂X ∂X ∂X ∂X
  (1.12)
1 ∂ uI ∂ uJ ∂ uK ∂ uK
EIJ = + + ·
2 ∂ XJ ∂ XI ∂ XJ ∂ XI
6 1 FE-Models of the Sheet Metal Forming Processes

Now, consider two neighboring particles with instantaneous positions x and

x+dx, respectively. The difference in velocity between these two points point is

∂v ∂ vi
dv = dx ; d vi = d xj (1.13)
∂x ∂ xj

The gradient in the above equation is called the velocity gradient tensor and is, thus,
defined by

∂v ∂ vi
L = ; Lij = (1.14)
∂x ∂ xj

The velocity gradient can be decomposed into a symmetric and a skew-

symmetric part according to

1 1
L = ( L + LT ) + ( L − LT );
2 2
   (1.15)
1 ∂ vi ∂ vj 1 ∂ vi ∂ vj
Lij = + + −
2 ∂ xj ∂ xi 2 ∂ xj ∂ xi

The symmetric part d

 
1 1 ∂ vi ∂ vj
d = ( L + LT ) ; dij = + (1.16)
2 2 ∂ xj ∂ xi

is called rate of deformation tensor, while the skew-symmetric part W

 
1 1 ∂ vi ∂ vj
W = ( L − LT ) ; Wij = − (1.17)
2 2 ∂ xj ∂ xi

is called the spin tensor.

The physical interpretation of these tensors is now quite obvious. Rewriting
Eq. (1.13), we get

dv = (d + W) · dx ; d vi = ( dij + Wij ) d xj (1.18)

If all the components dij are equal to zero at a certain point P, the instantaneous
motion in the neighbourhood of P is a rigid body motion. This is a consequence
of the skew-symmetry of the tensor W. We then realize that, if all components Wij
are zero, we have a pure deformation without any rigid body rotation. The rate
of deformation tensor is, thus, an objective tensor. It should also be observed that
for a small deformation problem the rate of deformation tensor is simply the time
derivative of the strain tensor, i.e. dij = ε̇ij .
We shall now find the relationship between the rate of deformation tensor and the
time derivative of Green’s strain tensor. From the definition of Green’s strain tensor
in Eq. (1.8) we get
1.2 Fundamentals of Continuum Mechanics 7

d 2
( s ) = 2 dX · Ė · dX = 2 ĖIJ dXI dXJ (1.19)
dt
The left hand side of the above equation can also be written

d
dt ( s2 ) = 2 dx · ddt (dx) = 2 dx · dv = 2 dx · ∂x
∂v
· dx =
(1.20)
= dx · ( d + W) · dx = dxk ( dkm + Wkm ) dxm

Due to the skew-symmetry of the spin tensor W, i.e. Wij = −Wji , it is easy to
show that

dx · W · dx = dxk Wkm dxm = 0 (1.21)

Eq. (1.20) can then be rewritten as

d 2
( s ) = dx · d · dx = dxk dkm dxm (1.22)
dt
Introducing the definition of the deformation gradient tensor F according to
Eq. (1.5), we can rewrite the above equation as

d 2
( s ) = ( dX · FT ) · d · ( F · dX ) (1.23)
dt
Comparing Eqs. (1.19) and (1.23), we find the following relation between Green’s
strain rate tensor and the rate of deformation tensor:

Ė = FT · d · F ; ĖIJ = dkm FkI FmJ (1.24)

From Eq. (1.24) we note that also the components of the tensor Ė vanish when the
neighborhood of the particle considered moves like a rigid body. That is, the Green
strain rate tensor is objective.
It is much more convenient to express the Green strain rate in velocity gradients
referred to the reference coordinates. Taking the time derivative of Eq. (1.12) we get

  T  T  T 
1 ∂ u̇ ∂ u̇ ∂ u̇ ∂u ∂u ∂ u̇
Ė = + + · + ·
2 ∂X ∂X ∂X ∂X ∂X ∂X
  (1.25)
1 ∂ u̇I ∂ u̇J ∂ u̇K ∂ uK ∂ uK ∂ u̇K
ĖIJ = + + +
2 ∂ XJ ∂ XI ∂ XJ ∂ XI ∂ XJ ∂ XI

Finally, it should be mentioned (without proof) that the determinant of the defor-
mation gradient tensor expresses the relation between a volume element in the
current and the reference configuration, respectively, so that

dv = det ( F) dV = J dV (1.26)
8 1 FE-Models of the Sheet Metal Forming Processes

where dv is the volume element in the current configuration and, dV is the same
element in the reference configuration.

1.2.3 Stress Measures

There are a number of different stress measures appearing in the continuum mechan-
ics literature. Here we will only consider three of them: The Cauchy stress σ, the
Kirchhoff stress τ and the Second Piola-Kirchhhoff (PK2) stress S, together with
rates of these tensors.
The Cauchy stress tensor σ is also known as the true stress tensor and measures
the force per unit area in the current configuration. The traction or stress vector t at
a boundary point is related to the Cauchy stress tensor through the relation

t = n̂ · σ ; ti = n̂j σji (1.27)

where n̂ is an outward unit vector at the boundary point.

In a Lagrangian formulation equilibrium equations are often written in terms of
variables referred to the reference configuration. To this purpose a variety of so
called pseudo stress measures have been defined, of which two will be discussed
here. The first one is the Kirchhoff stress tensor τ, which frequently has been used
in constitutive equations in finite strain plasticity. It is related to the Cauchy stress
tensor by

ρ0 ρ0
τ = σ = Jσ ; τij = σij = J σij (1.28)
ρ ρ

where ρ0 and ρ are the mass densities in the reference and current configurations,
respectively, and J is the determinant of the deformation gradient tensor. It should
be noted that for an incompressible material the Cauchy and Kirchhoff stress tensors
are equal.
The Second Piola-Kirchhof f stress tensor S is defined to be energy conjugate to
the Green strain rate tensor. The work rate can then be written

1
  
Ẇ = σ : d dv = S : Ė dV = S : Ė dv (1.29)
J
v V v

Introducing the relation between the rate of deformation tensor and Green’s strain
rate according to Eq. (1.24), we get

J σ : d = S : Ė = S : ( FT · d · F ) (1.30)

After some tensor manipulations the following relation is obtained

−1 −1
S = J F−1 · σ · F−T ; SIJ = J FIk σkm FJm (1.31)
1.3 Material Models 9

It can be shown that the Cauchy stress tensor σ is symmetric and objective.
As a consequence of Eq. (1.31) it can be understood that also the PK2 tensor S
is symmetric and objective.
The constitutive equations for certain material types like hypoelastic and elastic-
plastic ones are formulated in rate form. The rate of deformation tensor d is a
suitable deformation measure in such material laws. The problem is to choose a
proper stress rate measure. The material time derivative of the Cauchy stress ten-
sor, σ̇ , can be shown not to be objective. Hence, the material time derivative of the
Cauchy stress tensor cannot serve as a proper stress rate measure. Of course this
defect is valid for the Kirchhoff stress rate τ̇ as well.
There do, however, exist a number of different objective rates of the Cauchy
stress tensor. Here we will only mention one of them: The Jaumann or co-rotational
rate, σ ∇ , defined by

σ ∇ = σ̇ + σ · W − W · σ ; σij∇ = σ̇ij + σik Wkj − Wik σkj (1.32)

1.3 Material Models

In the following we will assume that the yield condition can be expressed in the
form

f = σ̄ − H(ε̄ p ) = 0 (1.33)

This relation implies that yielding occurs when the effective stress σ̄ , which is
a scalar function of the state of stress, reaches a critical value H, which in turn is
a function of the effective plastic strain ε̄p . The function H(ε̄p ) is usually obtained
from a uniaxial stress-plastic strain curve.
The rate of deformation tensor d, defined in Eq. (1.16), is used as a strain rate
measure in elastic-plastic constitutive equations. We will in the following assume
that the rate of deformation tensor can be additively divided into an elastic and a
plastic part:

d = de + dp (1.34)

The normality condition states that the plastic rate of deformation dp is outward
normal to the yield surface f = 0. This is expressed as

∂f
dp = λ̇ (1.35)
∂σ
where λ̇ is a scalar function that depends on the current state of stress and strain.
The relationship (1.35) is called an associated flow rule.
During plastic loading the stress point remains on the yield surface. This implies
that ḟ = 0, which is known as the consistency condition. In the present case
we find
10 1 FE-Models of the Sheet Metal Forming Processes

∂f p ∂f p
ḟ = : σ ∇ − H ′ ε̄˙ = σ ∇ − H ′ ε̄˙ = 0 (1.36)
∂σ ∂ σkm km

where H ′ = d H/d ε̄p is the slope of the uniaxial stress-plastic strain curve, and σ ∇
is an objective stress rate tensor.
p
The effective plastic strain rate ε̄˙ is defined by the rate of plastic work equation

p
Ẇ p = σ̄ ε̄˙ = σ : dp (1.37)

Combining Eqs. (1.35) and (1.37), we note that

p
λ̇ = ε̄˙ (1.38)

Typical for an elastic-plastic law is that there exists a relation between rates
(or increments) of stress and strain. This relation can be written

σ ∇ = D : de = D : ( d − dp ) (1.39)

where D is the elastic constitutive tensor.

For simplicity, we will in the following gather the Cartesian components of the
tensors σ and dp in two vectors (column matrices) {σ} and {dp }. For the case of
plane stress these vectors will have the following appearances

T
{σ } = σx σy τxy
 (1.40)
dp = dxp dxp 2 dxy
p T

For a quadratic yield condition, e.g. the von Mises or the Hill’48 condition, the
effective stress can in matrix form be expressed as
 1/2
{σ }T [ A ] { σ }

σ̄ = (1.41)

where [A] is a matrix with constants describing the anisotropy of the material.
The gradient to the yield surface can then be expressed as
 
∂f 1
= [A] {σ } (1.42)
∂σ σ̄

and according to the normality condition the components of the plastic rate of
deformation tensor are given by

ε̄˙ p
dp = [A] {σ } (1.43)
σ̄
1.4 FE-Equations for Small Deformations 11

To give an example of the appearance of the matrix [A], let us consider a transver-
sally anisotropic material obeying the Hill’48 yield condition. Eq. (1.43) can then
be expressed in matrix form as

p
⎡ ⎤
1 − 1 +R R 0
⎡ ⎤
dx ⎡ ⎤
p
p
ε̄˙ ⎢ ⎥ σx
⎢ dy ⎦ =
⎥ ⎢ − 1 +R R 1 0 ⎥ ⎣ σy ⎦ (1.44)
σ̄ ⎣
⎣
p
⎦
2dxy 1 + 2R τxy
0 0 2 1+R

Inverting Eq. (1.43), we get

σ̄
{σ } = [A ] −1 dp (1.45)
˙ε̄ p

Note that this equation expresses total stress in terms of rate of plastic strain. Note
also that it is only for quadratic yield conditions that the normality condition can be
inverted to this form.
If the total strain rates in Eq. (1.45) are replaced by plastic strain rates, i.e. the
elastic part of the rate of deformation tensor is ignored, this equation will form the
basis of the rigid-plastic theory. A couple of the earlier FE formulations for sheet
forming simulation were based on this form of the constitutive equations. In matrix
form this equation takes the form
⎡ ⎤ ⎡ ⎤
σx 1+R R 0 ⎡ d ⎤
σ̄ 1 + R ⎢ x
⎣ R 1 + R 0 ⎦ ⎣ dy ⎦
⎢σ ⎥
⎣ y⎦ = p (1.46)
⎥
τxy ε̄˙ 1 + 2R 1 2dxy
0 0 2

Using the work Eq. (1.37), we can easily derive the following expression for the
effective strain rate
  1/2
p
ε̄˙ = {d }T [ A ] −1 { d } (1.47)

1.4 FE-Equations for Small Deformations

The dynamic equilibrium conditions of a body are readily expressed by means
of the principle of virtual velocities. In terms of variables referred to the current
configuration the principle is stated
   
σ : δd dV + ρ ü : δ u̇ d V = t : δ u̇ d S + f : δ u̇ d V (1.48)
V V S V
12 1 FE-Models of the Sheet Metal Forming Processes

where ü is the acceleration, u̇ is the velocity, t is the surface traction, and f is the
body load. Integration is performed over current volume V and surface area S.
Assume now for a moment that we are dealing with a small deformation problem.
FE-approximations can be introduced as

{ u(x)} = [ N(x) ] { ũ }
{ u̇(x)} = [ N(x) ] { ũ˙ } (1.49)
{ δ u̇(x)} = [ N(x) ] { δ ũ˙ }

where { ũ } is a vector with nodal displacements and [ N(x) ] is a matrix with base
functions.
Introducing the above FE-approximations into the equation for the strain rate,
Eq. (1.16), we obtain the expressions for the strain rate and virtual strain rate,
respectively, according to

{ ε̇ } = { d } = [ B ] { ũ˙ }
(1.50)
{ δ ε̇ } = { δ d } = [ B ] { δ ũ˙ }

where the matrix [B] is known as the strain matrix.

The discretized equilibrium equations will now take the form
 
f int ¨ =
+ [ M ] {ũ} f ext (1.51)

where [M] is the consistent mass matrix, defined by


[M] = ρ [ N ] T [ N ] dV (1.52)
V

The external force vector is given by

 
f ext = [ N ] T { b } dV + [ N ] T { t } dS (1.53)
V S

and, finally, the internal force or stress force vector is defined by

  
f int = [ B ] T { σ } dV (1.54)
V

In the case of linear elasticity the constitutive relation can be written in matrix
form as

{ σ } = [ D ] { ε} (1.55)
1.5 FE-Equations for Finite Deformations 13

where [D] is the matrix form of the elastic constitutive tensor. The internal force
vector can then be rewritten as
⎛ ⎞
  
int T
f = ⎝ [ B ] [ D ][ B ] dV ⎠ { ũ } = [ K ] { ũ } (1.56)
V

where [K] is the linear stiffness matrix.

1.5 FE-Equations for Finite Deformations

The virtual velocity equation (1.48) is a correct representation of the dynamic
equilibrium of a body, even if the deformations are of finite magnitudes, on the
assumption that all quantities are measured in the current configuration. In a
Lagrangian formulation it is, however, much more convenient to express equilib-
rium in variables referred to the reference configuration. The transformation of
Eq. (1.48) can be shown to give
   
S : δ Ė d V0 + ρ0 ü : δ u̇ d V0 = t0 : δ u̇ d S0 + f 0 : δ u̇ d V0
V0 V0 S0 V0
(1.57)

In Eq. (1.57) an index ‘0’ has been assigned to variables measured in the reference
configuration. The variables t0 and f 0 are pseudo forces per unit area and volume,
respectively, in the reference configuration. They are defined by

t0 d S0 = t d S ; f 0 d V0 = f d V (1.58)

In accordance with the small deformation formulation in Eq. (1.50), the expres-
sions for the strain rate and virtual strain rate in matrix notation now become


Ė = B̂ { ũ˙ }
 (1.59)
δ Ė = B̂ { δ ũ˙ }

The strain matrix B̂ can formally be divided in two parts according to

B̂ = [BL ] + [BNL ] (1.60)

Here [BL ] is the ordinary linear strain matrix, equivalent with the matrix [B] in
the small deformation theory, while [BNL ] is a nonlinear matrix, which arises as a
consequence of the displacement dependent terms in the expression for the Green
strain rate Ė according to Eq. (1.25).
14 1 FE-Models of the Sheet Metal Forming Processes

In accordance with the small deformation theory the discretized dynamic equi-
librium equations can be expressed as
 
f int ¨ =
+ [ M ] { ũ} f ext (1.61)

with the consistent mass matrix


[M] = ρ0 [ N ] T [ N ] dV0 (1.62)
V0

the external force vector

 
f ext = [ N ] T { b0 } dV0 + [ N ] T { t0 } dS0 (1.63)
V0 S0

and the internal force vector

   T
f int = B̂ { S } dV0 (1.64)
V0

We will henceforth assume that the problem is quasi-static, i.e. we neglect the
inertia term in the equilibrium equations in Eq. (1.61). The simplified equations can
then be written
 T
{ } = B̂ { S } dV0 − f ext = 0 (1.65)
V0

where {} is a residual vector, whose elements should be zero when the equilibrium
equations are satisfied.
In order to solve the resulting set of nonlinear equations, the Newton-Raphson
iterative solution procedure, or related techniques, is commonly used. This requires
a linearization of Eq. (1.65) around the last obtained solution. Taking the time
derivative of the equilibrium equation in Eq. (1.65), we get
  T 
˙ˆ T

˙
 = B̂ Ṡ + B {S} dV0 − ḟ ext = 0 (1.66)
V0

Assuming that the constitutive equations can be transformed to a form such as

Ṡ = [ DT ] Ė (1.67)

where [DT ] is a matrix form of the constitutive tensor. It can then be shown that
Eq. (1.66) can be rewritten as
1.5 FE-Equations for Finite Deformations 15
 
˙
 = [KT ] ũ˙ − ḟ ext = 0 (1.68)

where the tangent matrix [KT ] formally can be divided into three matrices:

[KT ] = [KM ] + [KG ] + [KS ] (1.69)

The first of these matrices, [KM ], is called the material stiffness matrix and is
defined by

[KM ] = { BL }T [DT ] {BL } dV0 (1.70)
V0

This matrix is recognized as the ordinary small deformation tangent stiffness matrix.
The second one, [KG ], is known as the geometric stiffness matrix, and is a conse-
quence of the nonlinear terms in the strain-displacement relation. Finally, the third
matrix, [KS ], is called the initial stress stiffness matrix, and is a function of the stress
state in the current configuration.
Various FE-approaches, based on a Lagrangian description, can be con-
structed depending on the choice of reference configuration. The most well-known
Lagrangian formulations are the Total (TL) and Updated Lagrangian (UL) formu-
lations, respectively. In the TL-formulation the initial, stress free configuration is
taken as reference configuration, while in the UL-formulation the last calculated
configuration is taken as reference state. These two formulations are described
below.
The TL-formulation follows basically the one described above. The reference
coordinates should here be interpreted as the initial ones, and the displacements
are the total ones. Integrations are, furthermore, performed over initial volume and
surface area, respectively. The Green strain tensor components at time t+∆t can
either be calculated from the total displacements at time t+∆t, or by adding the
strain increment during the time increment ∆t to the strains at time t. This is justified
by the fact that all tensors, even incremental ones, are referred to the same reference
configuration.
One of the primary objects of the simulations is to determine the strain distribu-
tion in the blank. Principal logarithmic strains in the plane of the sheet are given by

ε1 = ln 1 ; ε2 = ln 2 (1.71)

where Λ1 and Λ2 are principal stretch values. Stretch is a measure of extensional

strain of a differential line element, defined by Λ=ds/dS. The principal stretch val-
ues can be shown to be related to the in-plane principal values of Lagrangian strain,
E1 and E2 , by
16 1 FE-Models of the Sheet Metal Forming Processes
! !
1 = 2 E1 + 1 ; 2 = 2 E2 + 1 (1.72)

where
"
1 1 2 ;
E1, 2 = (Ex + Ey ) ± (Ex − Ey ) + Exy (1.73)
2 4

The UL FE-formulation follows largely the same pattern as outlined for the TL-
approach above. The main differences are as follow. All coordinates entering the
formulation should be the current ones. Since the displacements are measured from
the reference (current) configuration, all terms involving displacements in the gen-
eral formulation will vanish. This implies that the matrices [BNL ] and [KG ] of the
general formulation do not enter the UL-formulation. Furthermore, integrals are
carried out over current volume and area.
The calculation of total strain is much more complicated in the UL-formulation
than in the TL-formulation, since the Lagrangian strain increments in each time step
is referred to different configurations. This implies that they cannot be added to total
strains without complicated transformations.
For three-dimensional shell and membrane elements it is usually necessary to
use a local coordinate system for each element, which is redefined (updated) in
each step. This implies that a new transformation matrix has to be established in
each step, and that a number of transformations of displacement and load vectors
between local and global systems have to be performed.

1.6 The ‘Flow Approach’—Eulerian FE-Formulations

for Rigid-Plastic Sheet Metal Analysis

The rigid-plastic constitutive relations in Sect. 1.3 have the form of the constitutive
relations for a non-Newtonian viscous fluid. In steady-state metal forming problems,
such as extrusion and rolling, the velocities at a given point in space remain constant
in time. The material behavior in this type of problems is similar to that of a fluid,
and an Eulerian FE-approach is a natural choice. The FE mesh in such problems is
fixed (Eulerian).
The Eulerian formulation has, however, also been used for the solution of rigid-
plastic, transient problems, such as stretch forming and deep-drawing of metal
sheets, although the material behavior in such problems bears small resemblance
with a fluid flow. In such transient problems the control volume is identified with the
sheet geometry in each deforming step. The element mesh has, thus, to be ‘updated’
in each step (Lagrangian).
The use of the flow approach in sheet metal forming problems has been advocated
particularly by Prof. O.C. Zienkiewicz and co-workers in Swansea, and by Prof. E.
Oñate in Barcelona [5, 6, 18–21]. A review of the flow approach in application to
various steady-state and transient forming problems is given in Zienkiewicz [22].
1.6 ‘Flow Approach’—Eulerian FE-Formulations for Rigid-Plastic Sheet Metal Analysis 17

In Sect. 1.3 the constitutive relations of a rigid-plasti c model were derived by

neglecting the elastic part of an elastic-plastic model (Eqs. 1.45, 1.46, 1.47). Some
writers have derived the constitutive relations for a rigid-plastic material starting
from the general form of a viscoplastic material as suggested by Perzyna [23]. If
the time dependent effects in the viscoplastic model are neglected, the equations of
the rigid-plastic model are recovered. It is interesting to note the analogy between
the equations of rigid-plastic or viscoplastic flow, and those of small strain, linear
elasticity. It is easily shown that completely analogous stress-strain relations are
obtained, if the velocity and rate of deformation variables of the flow equations are
interchanged by displacement and strain of the linear elasticity equations. Take for
instance Hooke’s generalized law in the case of plane stress and replace Poisson’s
ratio ν by R/(1+ R). Hooke’s law then takes the form
⎡ ⎤
⎡
εx
⎤ 1 − 1 +R R 0 ⎡ ⎤
1 ⎢ ⎥ σx
⎣ εy ⎦ = ⎢ − 1 +R R 1 0 ⎥ ⎣ σy ⎦ (1.74)
2εxy E ⎣
1 + 2R
⎦
τxy
0 0 2 1+R

The analogy between these equations and the corresponding plastic flow equa-
tions in Eq. (1.44) is immediately seen. It is noted that the modulus of elasticity
E in the elasticity relations plays the role of the ‘viscosity’ σ̄ /ε̄˙ in the flow relations.
The above discussed analogy makes it possible to use a standard FE-program for
linear elastic analysis in large strain viscoplastic analysis with only minor modifi-
cations of the program. Basically, the same FE-equations outlined in Sect. 1.4 for
linear elasticity are also applicable in the current flow approach, but with nodal dis-
placements replaced by nodal velocities, and the constitutive relationship modified
as described above. In the solution process a steady-state flow situation is assumed at
every deformation level. Due to the nonlinear ‘viscosity’, σ̄ /ε̄˙ , an iterative solution
scheme has to be employed at every step to ensure equilibrium.
The FE-equations are thus established in a standard fashion. We assume that
a local Catresian coordinate system is defined for each element and is updated in
each step. Components referred to these local axes are in the following marked by
a super-scribed star. Briefly, the major steps of the discretization process are the
following:
Velocity assumptions:
∗
u̇∗ = N∗ { ũ˙ } (1.75)

Rate of deformation-velocity relations:

∗ 1 ∗
dαβ = ( u̇ + u̇∗α )
2 α,β (1.76)
∗
d ∗
= N ∗
{ ũ˙ }
18 1 FE-Models of the Sheet Metal Forming Processes

Local equilibrium equations:

∗
 
K∗ ũ˙ = {f∗ }
T
K∗ B∗ [ C ] B∗ dV
#
= (1.77)
V
σ̄ −1
[C] = ε̄˙
[G]

Transformations between local and global system:

∗
   
{ f ∗ } = [ A ] f ext ; ũ˙ = [A] ũ˙ ; [ A ]T [ A ] = [ I ]
(1.78)
[ K ] = [ A ] T K∗ [ A ]

Global equilibrium equations:

 
[K ] ũ˙ = f ext (1.79)

When convergence is achieved, the geometry is updated by ũ˙ t. The effective
strain at time t+∆t is obtained as t+t ε̄ = t ε̄ + ε̄˙ t. A new value of the effective
stress is obtained from the hardening curve as t+t σ̄ = H(t+t ε̄).
In rigid or nearly rigid zones of the material the value of ε̄˙ tends to zero and, thus,
the ‘viscosity’ σ̄ /ε̄˙ tends to infinity. To avoid numerical difficulties due to this fact,
the use of a large but finite cut off value of the ‘viscosity’ is recommended in the
references above. Such a cut off value makes it possible to compute stresses even in
zones, where the stress state is below the yield stress.
Osakada et al. [24] have pointed out the importance of satisfying the equilib-
rium conditions at the end of the increment, at time t+∆t, in certain large strain
transient problems, such as sheet metal forming problems. Such a procedure incor-
porates the effects of shape change and work hardening during the incremental step,
and yields for certain problems more accurate results than the simple extrapolation
scheme previously described. In the method proposed in [24] the nodal velocities
are assumed constant during the time increment ∆t. In the equilibrium relations the
matrix [B] and the integration domains are functions of the unknown nodal coordi-
nates at the end of the increment. Furthermore, the effective stress σ̄ is a function of
the unknown effective strain ε̄
Concerning the strain calculation in the flow approach, the same comments can
be made as for the UL-formulation in Sect. 1.5.

1.7 The Dynamic, Explicit Method

In a mathematical sense there are basically two methods for solving the discretized,
dynamic equilibrium equations outlined in Eqs. (1.57), (1.58), (1.59), (1.60), (1.61),
1.7 The Dynamic, Explicit Method 19

(1.62), (1.63) and (1.64). These are the implicit and explicit methods, respectively.
The implicit method is the most common one. The problem can be formulated in the
following way: From a known solution at time step n, we like to compute the solu-
tion at time step n+1. In the implicit method the time derivatives of displacements
at time step n+1 is part of the solution. This necessitates the solution of a system of
nonlinear equations in every time step. The implicit time integration methods can be
proved to be unconditionally stable (at least for linear problems). One of the most
well-known integration schemes of this kind is the Newmark beta scheme.
Explicit time integration methods are mainly used for transient dynamic prob-
lems of very short duration, such as for structures submitted to blast and impact
loads. Such problems do normally have durations of the order of 1–100 ms. In an
explicit method the solution at time step n +1 is only based on known quantities
at time step n. This implies that no system of equations has to be solved in every
time step. Explicit time integration methods can be proved to be only conditionally
stable, i.e. there exist a critical time step size, which cannot be exceeded in order to
get a stable solution. The critical time step size is usually of the order of a fraction
of a microsecond. The most common explicit time integration method is the Central
difference method, whose basic equations are the following:
 
1 1 n+1
n+
ũ˙ =
2 ( { ũ } − n { ũ } ) (1.80)
t
  1 n+ 1   1
  1
n
ũ¨ = ( 2 ũ˙ − n − 2 ũ˙ ) = 2 ( n + 1 { ũ } − 2 n { ũ } + n−1
{ ũ } )
t t
(1.81)
Typical
 for the Central difference method is that displacements { ũ } and accel-
erations ũ¨ are evaluated at whole time steps n − 1 t, n t, n + 1 t, and so on, while
  1
velocities ũ˙ are evaluated at the midpoints of the time intervals, i.e. at n − 2 t
1
and n + 2 t.
To advance one step in the explicit calculation scheme, we start from known  
1 1
velocities n − 2 ũ˙ and displacements n { ũ }, and would like to calculate n + 2 ũ˙
and n + 1 { ũ }. The dynamic equilibrium equations at time n t can be written
 
[ M ] n { ũ¨ } + [ C ] n { ũ˙ } = n
f ext − n
f int (1.82)

Here a damping term has been included for the sake of completeness. Since the
velocity is not known at time n t, the following approximation is introduced:
1
[ C ] n { ũ˙ } ≈ [ C ] n− 2 { ũ˙ } (1.83)

In order to fully employ the benefits of the explicit method, the mass matrix will
in the following be assumed to be a diagonal matrix, thus avoiding a time consuming
inversion of the matrix. The explicit calculations are now performed according to the
following scheme:
20 1 FE-Models of the Sheet Metal Forming Processes
 1
 
n
{ ũ¨ } = [ M ]−1 − [C] n− 2 { ũ˙ } + n
f ext − n
f int (1.84)
1 1
n+ 2 { ũ˙ } = n− 2 { ũ˙ } + n
{ ũ¨ } t (1.85)
1
n+1
{ ũ } = n { ũ } + n+ 2 { ũ˙ } t (1.86)

In the above equations no time step index has been assigned to the time increment
∆t. It should, however, be understood that ∆t has to be continuously updated as the
analyzed structure deforms.
For a damped system the critical time step can be shown to be (Belytschko
et al. [1])

2
$
tcr = ( 1 + ξ2 − ξ ) (1.87)
ωmax

where ωmax is the maximum eigen-frequency of the system, and ξ is the fraction of
critical damping of the highest eigen-mode. From Eq. (1.87) we see that the critical
time step has its biggest value for ξ =0, i.e. for an undamped system. It can also be
noted that a critically damped system, i.e. for ξ =1, has a critical time step, which is
about 60% of that of an undamped system.
The critical time step according to Eq. (1.87) is only valid for linear systems. In
practice, therefore, a smaller time step should be used. Belytschko et al. [1] make
the following recommendation:

t = α tcr ; 0.8 ≤ α ≤ 0.98 (1.88)

The maximum eigen-frequency of a discretized structure is determined by the

smallest element in the FE-mesh. Let c denote the speed of sound in the structure.
Then, according to the so-called Courant condition

l
t ≤ (1.89)
c
where l is a characteristic length of the smallest element. The interpretation of the
above equation is that the time step should be so small that information does not
propagate across more than one element during a single time step. This observa-
tion is only true for constant strain elements. The speed of sound for two common
structural elements are
% %
E E
Bar : c = ; Shell : c = ; (1.90)
ρ ρ (1 − ν 2 )

The special characteristics of the dynamic, explicit method can now be summa-
rized as follow:
1.8 A Historical Review of Sheet Forming Simulation 21

• In explicit codes the Central difference time integration method is usually

employed.
• The explicit method is conditionally stable, i.e. there exists a critical time step ,
which must not be exceeded in order to get a stable solution.
• The critical time step is approximately equal to the time for a compression or a
bending wave to travel through the smallest element in the mesh.
• The number of time steps in a typical explicit analysis is tens of thousands.
• A typical critical step size is of the order of a fraction of a microsecond.
• When a diagonal mass matrix, e.g. a lumped matrix, is used, no system of
equations has to be solved, and the method is truly explicit.
• For a given FE-mesh the computing time is directly proportional to the duration
of the studied event.
• An explicit method is very robust. The frequently occurring convergence prob-
lems in implicit simulations are completely avoided in explicit analyses, since no
iterations are performed on a global level.
• The finite elements used in explicit codes are usually simple, linear displacement
based elements. They are often under-integrated, which is why hourglass control
is an important issue.
• Since the time steps are very small, the incremental deformations during a step
are also very small. This implies that the complex equations of the Lagrangian
description of motion can be ignored, and simplified, linear geometric relations
can be used instead. Also the contact conditions can be strongly simplified.
• Since no material tangent stiffness matrices have to be established, the implemen-
tation of new material laws is considerable simplified (compared to an implicit
method).
• Since no matrices are stored during an explicit simulation, the data storage
requirements are much smaller than for an implicit method.
• The explicit method is well adapted for parallelization.

1.8 A Historical Review of Sheet Forming Simulation

The possibility to perform simulations of sheet forming processes was for a long
time an unattainable desire in the sheet forming industry. The process engineers
liked to be able to reveal any possible forming defects at an early stage, and to min-
imize the need for expensive modifications of the tools in a trial-and-error process.
From a theoretical perspective the forming problem was, however, considered to
be a very complicated problem in the early days of FEM in the 1960s and early
1970s. Modelling sheet metal forming problems requires accurate characterisation
of effects like nonlinear material behavior, large deformations, and complicated
contact conditions between the tools and the blank. The simulation of a real, indus-
trial forming operation with the computing power of the time considered to be an
extremely computationally demanding task.
Not until 1970 the first theoretically correct FE-formulation of the large
deformation problem was presented by Hibbitt et al. [25]. These equations were
22 1 FE-Models of the Sheet Metal Forming Processes

written in a TL-frame (see Sect. 1.5). Later in 1975 McMeeking and Rice [26] and
Bathe et al. [27] presented correct UL-formulations of the problem. The attempts to
solve forming problems at that period of time were usually based on plane strain or
axi-symmetric formulations, and from a theoretically point of view they were often
not fully consistent.
The first theoretically correct, 3D formulation of the sheet forming problem
was presented by Wang and Budiansky [28] in 1978. The presented method was
a TL-formulation and involved triangular, constant strain membrane elements. The
material model was an elastic-plastic one. The solution was advanced with a simple
forward Euler, incremental scheme. Methods based on such a solution scheme are
sometimes denoted static-explicit methods.
The following decade saw a high activity in the field. The development took
place along several different paths. Methods based on 2D as well as 3D formulations
were developed. Depending on the choice of description of motion, type of consti-
tutive relations, and solution procedures, the methods were described as the solid
approach, the static-implicit approach, the static-explicit approach, the rigid-plastic
approach and the flow approach.
The solid approach and the static-implicit approach are two different names
of the same procedure. The basic equations of this method are the same as those
presented in Sect. 1.5, with the difference that the problem is considered as a qua-
sistatic one, and the inertia term is therefore neglected. This method is based on a
Lagrangian description of motion and an elastic-plastic constitutive relation. The
solution is advanced incrementally, and in every increment the equilibrium equa-
tions are solved iteratively with some Newton-Raphson-like procedure. The typical
number of incremental steps in such a simulation is of the order of hundreds..
The previous approach is implicit in the sense that an iterative procedure is
employed in each step in order to fulfill the static equilibrium conditions. However,
some authors have used an alternative technique, called the static-explicit approach,
in which no iterations at all are performed. The updating of the geometry is just
based on the tangent moduli in the previous step. This implies that equilibrium
is never satisfied. In order to reduce the errors involved, very small steps have
to be taken. Several thousand steps are common for an ordinary simulation. The
advantage of this approach is that it is quite robust, since there are no iterative pro-
cesses that have to converge. Even instability phenomena like wrinkling have been
simulated by means of this procedure.
The flow approach was previously described in Sect. 1.6. It uses a kind of
Updated Eulerian formulation and a rigid-plastic material law. Nodal velocities are
primary unknowns. The geometry is fixed in each time step, while the equilibrium is
iteratively solved for. The geometry is then updated based on the calculated veloc-
ities. One of the main advantages of the flow approach is, thus, that the governing
equations get a very simple appearance. An obvious disadvantage is of course that
no phenomena related to elasticity, such as springback, can be simulated.
The rigid-plastic approach is based on the same rigid-plastic constitutive rela-
tions as the flow approach (the two methods are sometimes mixed-up). However,
some writers have preferred to rewrite these relations in terms of increments
1.8 A Historical Review of Sheet Forming Simulation 23

of strain. This leads naturally to a Langrangian FE-formulation with nodal dis-

placements as primary unknowns. The disadvantages of such an approach are of
course the same as those of the flow approach. However, the present formulation
do also lack the simplicity of the flow formulation, since the kinematic rela-
tions in a Lagrangian formulation are much more complicated than those in an
Eulerian one.
During this period of time (the 1980s) the flow and rigid-plastic approaches were
more popular than the static-implicit one, mainly because they were more stable,
and considerably larger time-steps could be employed. In spite of a lot of promising
development, the practical application of sheet forming simulation was, however,
for many years hampered by too unstable numerical procedures and excessive com-
puting times, even for very small problems. As an example of state-of-the art at that
time, Tang et al. [29] published results from practical applications of a code, devel-
oped at Ford, to the simulation of the stamping of real 3D automotive parts. This
code was based on 3D shell theory and a static-implicit approach. Models with up
to 400 higher order shell elements were analyzed, and the reported computing time
was about 20 h on a super-computer.
In 1989 Honecker and Mattiasson [4] presented results from a study, in which
the dynamic-explicit approach was evaluated in application to sheet metal stamping.
The results from this study were very promising. Problems with up to 10,000 shell
elements could be solved within 1.5 h on a super computer. Also the robustness of
this approach was found to be widely superior to that of any other method.
Since sheet forming processes usually have much longer durations than what the
dynamic, explicit method normally is intended for, the prerequisites for the problem
have to be modified in some way, in order to fully utilize the benefits of the method.
This can be done in two different ways. Either the virtual velocity in the simula-
tion is substantially increased in comparison to the physical one, in order to reduce
the virtual process time, or the density of the sheet metal is increased, in order to
increase the critical time step size, and thereby reducing the total number of time
steps. Both these methods lead to an increase of the virtual inertia forces. It is, thus,
of utmost importance to have a strict control of these forces so that they have only
a marginal influence on the solution. Experience has shown that the physical veloc-
ity can be increased by a factor 10–50 without inertia forces having any influence.
Some guidelines for the choice of virtual quantities are given in e.g. Chung et al.
[30]. A comprehensive overview of the use of the dynamic, explicit method in sheet
metal forming simulations is given in Nielsen [31].
From the beginning of the 1990s there was an explosive increase of the practical
utilization of sheet forming simulations within the industry, and from the middle of
this decade most companies within the automotive industry were performing sheet
stamping simulations on a regular basis. Dynamic, explicit codes were dominating
the software market. General purpose codes like LS-DYNA and ABAQUS/Explicit,
and specialized codes such as PAM-STAMP and OPTRIS are examples of codes
in use. In Japan also a couple of codes based on the static-explicit approach found
some industrial usage; see Kawka et al. [8] and Makinouchi [12]. A few static-
implicit codes like INDEED in the German automotive industry and MTLFORM at
24 1 FE-Models of the Sheet Metal Forming Processes

Ford Motor Company were still in use. The static-implicit method was at that time,
and is still today, frequently used in academic circles.
The highly specialized code for stamping simulations, AutoForm emerged from
a research project at ETH in Zurich in the early 1990s. The code is based on the
static-implicit approach, but uses some innovative algorithms to enhance stability
and computational efficiency. Characteristic for the code in its original form was
the use of bending-enhanced membrane elements, and an iterative linear equation
solver. In course of time the code has developed and includes now also conventional
shell elements. For stamping applications the code is competitive with, or even supe-
rior to the dynamic, explicit codes with regard to efficiency and robustness. The code
will be treated more in detail in Sect. 4.1 of this book.
Today (2009), AutoForm is probably the most commonly used code in the indus-
try for sheet stamping simulations. Beside this code the software market is still
dominated by various dynamic, explicit codes like LS-DYNA, ABAQUS/Explicit,
PAM-STAMP 2G and STAMPACK. The use of other types of codes is now only
marginal.

References
1. Belytschko T, Liu WK, Moran B (2000) Nonlinear finite elements for continua and structures.
John Wiley & Sons, Chichester
2. Zienkiewicz OC, Taylor RL (2000) The finite element method, 5th edn, Vol 2. Solid
mechanics. Butterworth-Heinemann, Oxford
3. Crisfield MA (1991) Non-linear finite element analysis of solids and structures, Vol 1. John
Wiley & Sons, Hoboken, NJ
4. Honecker A, Mattiasson K (1989) Finite element procedures for 3D sheet forming simulation.
In: Thompson EG, Wood RD, Zienkiewicz OC, Samuelsson A (eds) NUMIFORM’89, AA
Balkema, Fort Collins
5. Oñate E, Agelet de Saracibar C (1992) Alternatives for finite element analysis of sheet metal
forming problems. In: Chenot JL, Wood RD, Zienkiewicz OC (eds) NUMIFORM’92, AA
Balkema, Sophia Antipolis
6. Oñate E, Garcia Garino C et al. (1993) NUMISTAMP: A research project for assessment
of finite element models for stamping processes. In: Makinouchi A, Nakamachi E, Oñate E,
Wagoner RH (eds) NUMISHEET’93, Isehara, Japan
7. Mattiasson K (1996) Computational strategies for sheet forming problems. ECCOMAS’96,
John Wiley & Sons, Paris
8. Kawka M, Makinouchi A, Wang SP, Nakamachi E (1997) Advances and trends in sheet metal
forming simulation in Japanese automotive industry. SAE Technical Paper 970432
9. Wenner M (1997) State-of-the-art of mathematical modeling of sheet metal forming of
automotive body panels. SAE Technical Paper 970431
10. Wang SP, Choudhry S, Wertheimer TB (1998) Comparison between the static implicit and
dynamic explicit methods for FEM simulation of sheet forming processes. In: Huétink J,
Baaijens FPT (eds) NUMIFORM’98, AA Balkema, Twente
11. Mattiasson K (2000) On Finite Element simulation of sheet metal forming processes in
industry. ECCOMAS 2000, Barcelona
12. Makinouchi A (2001) Recent developments in sheet metal forming simulation. In: Mori K
(ed) NUMIFORM 2001, AA Balkema, Toyohashi, Japan
13. Wenner M (2005) Overview – Simulation of sheet metal forming. In: Smith LM, Pourboghrat
F, Yoon JW, Stoughton TB (eds) NUMISHEET 2005, Detroit, AIP 0-7354-0265-5
References 25

14. Banabic D, Tekkaya E (2006) Forming simulation – Numerical simulation and material mod-
els of aluminium sheet forming. In: Hirsch J (ed) Virtual fabrication of aluminum products.
Microstructural modeling in industrial aluminum production. Wiley-VCH Verlag, Weinheim,
275–302
15. Roll K, Weigand K (2009) Tendencies and new requirements in the simulation of sheet metal
forming processes. Computer Methods in Materials Science 9:12–24
16. Malvern LE (1969) Introduction to the mechanics of a continuous medium. Prentice-Hall,
Englewood Cliffs, NJ
17. Gurtin ME (1981) An introduction to continuum mechanics. Academic Press, London
18. Zienkiewicz OC, Jain PC, Oñate E (1978) Flow of solids during forming and extrusion: Some
aspects of numerical solutions. International Journal of Solids and Structures 14:15–38
19. Baynham JMW, Zienkiewicz OC (1982) Developments in the finite element analysis of thin
sheet drawing and direct redrawing processes, using a rigid/plastic approach. In: Pittman JFT
et al. (eds) Numerical methods in industrial forming processes. Pineridge Press, Swansea
20. Oñate E, Zienkiewicz OC (1983) A viscous shell formulation for the analysis of thin sheet
metal forming. International Journal of Mechanical Sciences 25:305–335
21. Wood RD, Mattiasson K, Honnor ME, Zienkiewicz OC (1985) Viscous flow and solid
mechanics approaches to the analysis of thin sheet forming. In: Wang N-M, Tang SC (ed)
Proceedings of the Symposium on computer modeling of the sheet forming process – Theory,
verification and applications, Ann Arbor. The Metallurgical Society, Warrendale, PA
22. Zienkiewicz OC (1984) Flow formulation for the numerical solution of forming processes.
In: Pittman JFT et al. (eds) Numerical analysis of forming processes. Wiley-InterScience,
New York, NY
23. Perzyna P (1966) Fundamental problems in viscoplasticity. Recent advances in applied
mechanics. Academic Press, New York, NY
24. Osakada K, Nakano J, Mori K (1982) Finite element method for rigid-plastic analysis of metal
forming – Formulation for finite deformation. International Journal of Mechanical Sciences
24:459–468
25. Belytschko T, liu wk, Moran B (2000) Nonlinear Finite Elements for continua and Structures.
John Wiley & Sons Ltd, Chichester
26. Hibbitt HD, Marcal PV, Rice JR (1970) A finite element formulation for problems of large
strain and large displacement. International Journal of Solids and Structures 6:1069–1086
27. McMeeking RM, Rice JR (1975) Finite element formulations for problems of large elastic-
plastic deformations. International Journal of Solids and Structures 11:601–616
28. Bathe KJ, Ramm E, Wilson EL (1975) Finite element formulations for large deformation
dynamic analysis. International Journal for Numerical Methods in Engineering 9:353–386
29. Wang N-M, Budiansky B (1978) Analysis of sheet metal stamping by finite element method.
Journal of Applied Mechanics, Transaction ASME 45:73–82
30. Tang SC, Ilankamban R, Ling P (1988) A finite element modeling of the stretch-draw forming
process. SAE Paper 880527
31. Chung WJ, Cho JW, Belytschko T (1998) On the dynamic effects of explicit FEM in sheet
metal forming analysis. Engineering Computations 15:750–776
32. Nielsen KB (2000) Sheet metal forming simulation using explicit finite element methods, PhD
thesis, Third edition. Department of Production, Aalborg University, Denmark
Chapter 2
Plastic Behaviour of Sheet Metal

List of Special Symbols

a exponent in the Hershey and Hosford yield criteria

a, b coefficients in the Hill 1990 yield criterion
a, b, c, f, g, h material parameters in the Barlat 1991 yield criterion
a, b, c, h, p coefficients in the Barlat 1989 yield criterion
a, b, m, n, p, q parameters describing the planar anisotropy of the material
in the Ferron yield criterion
a, b, c, d, e, f, g coefficients in the BBC 2000 yield criterion
a1. . . a4 coefficients in the Cazacu–Barlat yield criteria
a1. . . a25 coefficients in the Soare yield criteria
A0 , . . ., A9 coefficients in the Gotoh yield criterion
b1 . . .b11 coefficients in the Cazacu–Barlat yield criteria
c weighting coefficient in the Karafillis–Boyce yield
criterion
c, p, q coefficients in the Hill 1993 yield criterion
c1 , c2 , c3 material coefficients describing the material anisotropy in
the Barlat 1994 yield criterion
c12 , c13 . . .c66 coefficients in the linear transformation in Barlat 2000
yield criterion
C elasticity tensor
C′ , C′′ linear transformations in Barlat 2000 yield criterion
CD material constant in the Drucker yield criterion
D strain-rate tensor
E elastic modulus
f, F, ϕ yield function
f, g, h, a, b, c coefficients in the Hill 1979 yield criterion
F, G, H, L, M, N coefficients in the Hill 1948 yield criterion
F1 , F2 functions in the expresion of the uniaxial yield stress
(Barlat 1989 yield criterion)
Fb function used to define the biaxial yield stress and the
biaxial anisotropy coefficient

D. Banabic, Sheet Metal Forming Processes, DOI 10.1007/978-3-540-88113-1_2, 27


C Springer-Verlag Berlin Heidelberg 2010
28 2 Plastic Behaviour of Sheet Metal

Fθ function used to define the uniaxial yield stress and the

anisotropy coefficient
g(α) function used to define the Budiansky yield criterion
g(θ , α) function used to define the Ferron yield criterion
h scalar parameter which defines the plastic deformation
accumulated in the material
hij anisotropy coefficients in the von Mises 1928 yield
criterion
I2 , I3 second and third invariants of the stress tensor
J2 , J3 second and third invariants of the stress tensor
k exponent in the the Karafillis–Boyce and BBC yield
criteria
k1 , k2 invariants of the stress tensor
l final gage length
l0 initial gage length
L linear transformation tensor in the Karafillis–Boyce yield
criterion
L, M, N function in the Comsa yield criterion
M integer exponent used by the yield criteria
M, N, P, Q, R, S, T coefficients in the BBC yield criteria
m, n exponents used by the yield criteria
p exponent in the generalized Drucker yield criterion
p accumulated equivalent plastic strain
p1 . . .p8 coefficients in the Comsa yield criterion
R material parameter in the Lin–Ding yield criterion
R isotropic hardening variable
R, S, T shear yield stresses in the principal anisotropic directions
(Hill 1948)
r, R normal anisotropy coefficient
rb biaxial anisotropy coefficient
rθ anisotropy coefficient associated to the direction θ
r0 , r45 , r90 anisotropy coefficients at 0, 45 and 90◦ from the rolling
direction
s exponent in the Lin–Ding yield criterion
s deviatoric stress tensor in Barlat 2000 yield criterion
S IPE stress tensor used by the Karafillis–Boyce yield
criterion
S stress deviatoric tensor
S1 , S2 , S3 principal deviatoric stresses
S11 , S22 , S33 , components of the IPE stress tensor used by the Karafillis–
S12 , S23 , S31 Boyce yield criterion
t0, t initial and final thickness of the specimen
t1 , t2 functions in the expresion of the uniaxial anisotropy coef-
ficient (Barlat 1989 yield criterion)
T transformation matrix in Barlat 2000 yield criterion
2 Plastic Behaviour of Sheet Metal 29

w final width of the specimen

w0 initial width of the specimen
Wf energy of distortion
Wp elastic potential energy
Wv volumetric change energy
X linear transformation stress tensor in Barlat 2000 yield
criterion
X, Y, Z tensile yield stresses in principal anisotropic directions
(Hill 1948)
Y yield stress
Yθ uniaxial yield stress in a sample inclined by θ with respect
to the rolling direction
Yb theoretical biaxial yield stress
α angle between principal stress σ1 and rolling direction
α, β, γ coefficients in the Wang yield criterion
α = σ 2 /σ 1 ratio of the principal stresses
α back-stress tensor
α1, α2, α3 coefficients in the Barlat 1994 yield criterion
α 1 , ... α 8 coefficients in the Barlat 2000 yield criterion
α 1 , α 2 , γ 1 ,γ 2 ,γ 3 , C parameters defining the anisotropy of the material in the
Karafillis-Boyce yield criterion
αx, αy, αz coefficients in the Barlat 1994 yield criterion
β, ϕ, δ, γ accuracy index of the yield criteria
β 1, β 2, β 3 auxiliary coefficients used to define the linear transforma-
tion tensor in the Karafillis–Boyce yield criterion
∆r variation of anisotropy coefficients
εe equivalent (effective) strain
ε1 , ε2 , ε3 principal (logarithmic) strains
ε, εe , εp tensors of total, elastic and plastic strain respectively
 plastic potential
φ invariant homogeneous function
Γ, Ψ, Λ function in the BBC yield criteria angle between the
specimen longitudinal axis and the rolling direction
λ parameter of the Bézier function used in Vegter’s yield
criterion
λ plastic multiplier in the flow rule
μ Poisson’s ratio
σ stress tensor
σ 0 , σ 45 , σ 90 uniaxial yield stress at 0, 45 and 90◦ from the rolling
direction
σ0 initial yield stress
σ 1, σ 2, σ 3 principal stresses
σb equibiaxial yield stress
σe equivalent (effective) stress
σk hardening stress
30 2 Plastic Behaviour of Sheet Metal

σu uniaxial yield stress

σ 11 , σ 22 , σ 33 , components of the actual stress tensor
σ 12 , σ 23 , σ 31
τ shear yield stress

In the Sect. 2.8 the summation convention over repeated indices is used. Let A
denote second-order tensor and B a forth-order tensor. One can define the double
contracted tensor√product as A : A = Aij Aij and (B : A)ij = Bijkl Akl . The norm
&
of A is A = A : A and its direction is n = A A . The time derivative is
&
Ȧ = dA dt.

2.1 Anisotropy of Sheet Metals

2.1.1 Uniaxial Anisotropy Coefficients

Due to their crystallographic structure and the characteristics of the rolling process,
sheet metals generally exhibit a significant anisotropy of mechanical properties. In
fact, the rolling process induces a particular anisotropy characterised by the sym-
metry of the mechanical properties with respect to three orthogonal planes. Such a
mechanical behaviour is called orthotropy. The intersection lines of the symmetry
planes are the orthotropy axes. In the case of the rolled sheet metals, their orienta-
tion is as follows (see Fig. 2.1): rolling direction (RD); transverse direction (TD);
normal direction (ND).
The variation of their plastic behavior with direction is assessed by a quantity
called Lankford parameter or anisotropy coefficient [1]. This coefficient is deter-
mined by uniaxial tensile tests on sheet specimens in the form of a strip. The
anisotropy coefficient r is defined by

ε22
r= (2.1)
ε33

Fig. 2.1 Orthotropy axes

of the rolled sheet metals:
LD—longitudinal direction;
TD—transversal direction;
ND—normal direction
2.1 Anisotropy of Sheet Metals 31

Fig. 2.2 Geometry of the specimen: (a) before and (b) after deformation

where, ε22 , ε33 are the strains in the width and thickness directions, respectively.
In the case of an isotropic material, the coefficient is one and the width and thick-
ness strains have the same value. If the coefficient is greater than one, the width
strains will be dominant (the ‘thinning resistance’ is more pronounced). On the other
hand, for the materials having a coefficient less than one, the thickness strains will
dominate.
Using the notations from Fig. 2.2, Eq. (2.1) can be written in the form

ln ww0
r= (2.2)
ln tt0

where w0 and w are the initial and final width, while t0 and t are the initial and final
thickness of the specimen, respectively.
As the thickness of the specimen is very small compared to its width (usually by
at least one order), the relative errors of measurement of the two strains will be quite
different. Therefore the above relationships are replaced by one implying quantities
having the same order of magnitude: length and width of the specimen. Taking into
account the condition of volume constancy

ε11 + ε22 + ε33 = 0 (2.3)

32 2 Plastic Behaviour of Sheet Metal

the following form of Eq. (2.1) is obtained

ε22
r=− (2.4)
ε11 + ε22

and Eq. (2.2) becomes

− ln ww0
r= (2.5)
ln ll0 + ln ww0

where l0 and l are the initial and final gage length. The length l0 is specified by
standards, see [2]. Equation (2.5) can be rearranged as follows:

w
ln
w0
r= . (2.6)
l0 · w0
ln
l·w

This relationship is used in practice for evaluating the anisotropy coefficient.

Experiments show that r depends on the in-plane direction. If the tensile spec-
imen is cut having its longitudinal axis inclined with the angle θ to the rolling
direction, the coefficient rθ is obtained (see Fig. 2.3). The subscript specifies the
angle between the axis of the specimen and the rolling direction.

Fig. 2.3 Tensile specimen prelevated at the angle θ (measured from the rolling direction)
2.1 Anisotropy of Sheet Metals 33

Fig. 2.4 Uniaxial yield stress TD

corresponding to a direction
σax = σθ = Yθ
in the plane of the sheet metal

θ RD

Another important element characterising the performances of a yield criterion

is the capability to predict the variations of the uniaxial yield stress and coefficient
of plastic anisotropy in the plane of the sheet metal.
In order to assess this capability, we shall establish the relationships defining the
dependence of the parameters mentioned above on the angle θ measured from the
rolling direction (Fig. 2.4).
Let us denote by Yθ the uniaxial yield stress corresponding to the direction
inclined at the angle θ with respect to the rolling direction. In the case of a uniaxial
load, the components of the stress tensor can be expressed as follows:

σ11 = Yθ cos2 θ ;
σ22 = Yθ sin2 θ ; (2.7)
σ21 = σ12 = Yθ sin θ cos θ

By replacing Eq. (2.7) in the relationship defining the equivalent stress and taking
into account its homogeneity, we obtain:

σ = Yθ · Fθ , (2.8)

where Fθ is a function depending on the angle θ . Of course, Fθ is defined according

to the specific formulation of the equivalent stress. If we combine Eq. (2.8) with the
coherence condition,

(σ , Y) := σ − Y(h) = 0, (2.9)

where Φ(σ ,Y) is the yield function associated to the yield criterion, Y—yield stress,
h—scalar parameter defining the plastic strain accumulated by the material, we get:

Y(h)
Yθ = . (2.10)
Fθ
34 2 Plastic Behaviour of Sheet Metal

Equation (2.10) defines the uniaxial yield stress corresponding to the planar
direction identified by the angle θ . If the reference yield stress is selected to be
the one corresponding to the rolling direction (Y(h) = Y0 ), we obtain the following
relationship:

Y0
Yθ = . (2.11)
Fθ
In this case, the yield stress corresponding to some planar direction will depend
only on the yield stress associated to the rolling direction and the function Fθ (which
is related to the yield criterion adopted in the model). The determination of the
function Fθ will be presented in the next subchapters, for each type of yield criterion.
In a similar way, we can establish the relationship defining the variation of the
coefficient of plastic anisotropy in the plane of the sheet metal. Let us consider the
specimen inclined at the angle θ with respect to the rolling direction (Fig. 2.4).
According to Eq. (2.1), the instantaneous coefficient of plastic anisotropy rθ is
defined as the ratio of the plastic strain rates associated to the width (inclined at
the angle θ +90◦ with respect to the rolling direction), ε̇θ +90 , and thickness, ε̇33 :

ε̇θ +90
rθ = (2.12)
ε̇33
Taking into account the incompressibility restraint (see Eq. 2.3), as well as the
expressions of the strain rate components along the principal directions,

ε̇θ = ε̇11 cos2 θ + ε̇22 sin2 θ + ε̇12 sin θ cos θ ,

(2.13)
ε̇33 = −(ε̇11 − ε̇22 ).

we obtain the relationship defining the coefficient of plastic anisotropy associated to

the direction θ :

ε̇11 cos2 θ + ε̇22 sin2 θ + ε̇12 sin θ cos θ

rθ = − 1. (2.14)
ε̇11 + ε̇22
Equation (2.14) can be rewritten in terms of the stress components if the
associated flow rule is used:
∂σ ∂σ ∂σ
σ11 + σ22 + σ12 ∂σ
∂σ11 ∂σ22 12
rθ = − 1. (2.15)
∂σ ∂σ
σ11 + σ22
∂σ11 ∂σ22
By coupling Eq. (2.15) with Euler’s identity, we obtain:
σ
rθ =   − 1. (2.16)
∂σ ∂σ
Yθ +
∂σ11 ∂σ22
2.1 Anisotropy of Sheet Metals 35

Finally, after replacing Eq. (2.8) in the last relationship, we get:

Fθ
rθ = − 1. (2.17)
∂σ ∂σ
+
∂σ11 ∂σ22

This formula defines the coefficient of plastic anisotropy as a dependence of the

specimen inclination. In order to make use of it, we need the expression of the equiv-
alent stress and the function Fθ , both of them being specific to the yield criterion
adopted in the plasticity model. The planar distribution of the coefficient of plastic
anisotropy will be determined in the next subchapters, for different yield criteria.
The average of the r-values obtained for different directions in the plane of the
sheet metal represents the so-called coefficient of normal anisotropy rn . Having
determined the values of r at specimens cut along three directions in the plane of
the sheet metal (0◦ , 45◦ , 90◦ , respectively), the coefficient of normal anisotropy is
determined by:

r0 + 2r45 + r90
rn = . (2.18)
4

A measure of the variation of normal anisotropy with the angle to the rolling
direction is given by the quantity:

r0 + r90 − 2r45
r = , (2.19)
2

known as planar anisotropy.

This quantity is related to the earring amplitude of the deep-drawn cups. More
precisely, if the value of the anisotropy coefficient is the same along all the direc-
tions in the plane of the sheet metal, the earring phenomenon will not be observed.
In order to get a more intuitive image of the planar anisotropy of sheet metals, a
polar coordinate representation is frequently used. The diagram in Fig. 2.5 [3] gives
such a representation for an aluminium alloy (AA5182-0). The polar diagrams are
preferred due to their capability to offer direct information on the tendency of the
sheet metal to form ears. One may notice the symmetry of the curve with respect
to the coordinate axes. This fact is a proof of the orthotropic behaviour of the sheet
metal.
By convention the r-values usually are determined at 20% elongation for the
purpose of comparison. Modern tensile testing machines perform instantaneous
measurement of the quantities required for evaluating the anisotropy coefficient and
calculate it during the test. Values of r at 20% elongation as well as its variation with
strain can be determined.
Figure 2.6 [4] shows the planar variation of the anisotropy coefficient for DDQ
mild steel at different straining degrees. One may notice a significant variation of
the r45 coefficient with respect to the straining degree.
36 2 Plastic Behaviour of Sheet Metal

Fig. 2.5 Variation of the

anisotropy coefficient in the
plane of the sheet metal [3]

Fig. 2.6 Variation of the

anisotropy coefficients with
respect to the straining degree
[4]

2.1.2 Biaxial Anisotropy Coefficient

The experimental research has proved that the yield surfaces are not symmetric in
the biaxial region [5, 6]. This fact is also a consequence of the plastic anisotropy.
In order to give a quantitative description of such a behaviour, the so-called coef-
ficient of biaxial anisotropy has been defined independently by Barlat et al. [5]
and Poehlandt et al. [7]. Barlat and his co-workers [5] have proposed the use of
2.1 Anisotropy of Sheet Metals 37

Fig. 2.7 Strains measured on

the specimen during the disk
compression test [5]

a compression test for the experimental determination of this mechanical parameter.

A set of circular specimens are subjected to a normal pressure. Due to the plas-
tic anisotropy, the discs become elliptic during the compression. The phenomenon
can be observed in Fig. 2.7 [5], showing the results of a compression test for an
AA6111-T4 aluminium alloy. By measuring the major and minor axes of the elliptic
specimen, the corresponding principal strains can be evaluated.
As in the case of the uniaxial tension, the ratio of the principal strains will define
the coefficient of biaxial anisotropy:

ε22
rb = (2.20)
ε11

If the material is isotropic, the coefficient will be one. The more pronounced is
the anisotropy, the farther is the coefficient from unity. This parameter is a direct
measure of the slope of the yield locus at the balanced biaxial stress state. Pöhlandt
et al. [7] have proposed to use the biaxial tensile testing machine to determine the
coefficient.
Figure 2.8 shows the method used for the determination of the principal strains
on a biaxial testing machine. This experimental procedure is limited by the fact that
the straining degree is rather small (less than 5%).
38 2 Plastic Behaviour of Sheet Metal

Fig. 2.8 Strains measured on

the cross specimen during the
biaxial tensile test [7]

As for the uniaxial coefficient of plastic anisotropy, we shall define a general

relationship defining the biaxial coefficient. The relationship will be usable for any
yield criterion.
In the case of biaxial tension along the rolling and transverse directions, Eqs.
(2.7) will get the following particular forms:

σ11 = σ22 = Yb ;
(2.21)
σ21 = σ12 = 0.

where, Yb is the theoretical biaxial yield stress.

Consequently, Eq. (2.8) becomes:

σ b = Yb · Fb , (2.22)

Here, σ b stands for the experimental biaxial yield stress, while Fb is a constant
quantity depending on the yield criterion adopted in the plasticity model.
Equation (2.22) provides the theoretical biaxial yield stress as a dependence of
the the experimental biaxial yield stress and the parameter Fb :

σb
Yb = . (2.23)
Fb

In a very close analogy with the case of the uniaxial coefficient of plastic
anisotropy (see Eqs. 2.12, 2.13, 2.14, 2.15, 2.16, and 2.17), one may deduce the
relationship defining the coefficient of biaxial plastic anisotropy:
2.2 Yield Criteria for Isotropic Materials 39

Fb
rb = ∂σ
− 1. (2.24)
∂σ11

This relationship involves only the parameter Fb and the expression of the equiv-
alent stress, both of them being specific to the yield criterion adopted in the plasticity
model. The determination of the biaxial coefficient of anisotropy will be presented
in the next subchapters, for different yield criteria.

2.2 Yield Criteria for Isotropic Materials

In this section the focus is on the conditions under which a material passes from the
elastic to the plastic state and on the associated flow rules. In order to describe the
plastic behavior of a material in a general stress state, three elements are needed:

a) a yield criterion expressing a relationship between the stress components at the

moment when plastic ‘yielding’ occurs
b) an associated flow rule expressing the relationship between the components of
the strain-rate and stress
c) a hardening rule describing the evolution of the initial yield stress during the
forming process.

The transition from the elastic to the plastic state occurs when the stress reaches
the yield point of the material. The yield point in uniaxial tension is established
using the stress-strain curve of the material whereby a convention is necessary in
order to define it, or by temperature measurement.
In case of a multiaxial stress state it is more difficult to define a criterion for the
transition from the elastic to the plastic state. A relationship between the principal
stresses is needed specifying the conditions under which plastic flow occurs. Such
a relationship is usually defined in the form of an implicit function (known as the
‘yield function’):

F (σ1 , σ2 , σ3 , Y) = 0 (2.25)

where σ 1 , σ 2 , σ 3 are the principal stresses and Y is the yield stress obtained from a
simple test (tension, compression or shearing).
Equation (2.25) can be interpreted as the mathematical description of a surface
in the three dimensional space of the principal stresses usually called the ‘yield
surface’. It must be closed, smooth and convex. For incompressible materials it is a
cylinder the cross section of which depends on the material (only for the von Mises
criterion—see below—it is a circular cylinder as shown in Fig. 2.9).
All the points located in the inside of the surface (F < 0) are related to an elas-
tic state of the material. The points belonging to the surface (F = 0) are related
to a plastic state. The points located outside the surface (F > 0) have no physical
meaning.
40 2 Plastic Behaviour of Sheet Metal

Fig. 2.9 von Mises yield

surface in the stress space

In the case of plane stress (e.g. σ3 = 0) the yield surface reduces to a curve in
the plane of the principal stresses σ 1 and σ2 .
The expression of the yield function is established on the basis of some pheno-
menological considerations concerning the transition from the elastic to the plastic
state.
The most widely used yield criteria for isotropic materials have been proposed
by Tresca (the ‘maximum shear stress criterion’) and Huber–von Mises (the ‘strain
energy criterion’) [8].
Basically the yield function may be defined in two different ways [9]: either
by assuming that plastic yield begins when some physical quantity (energy, stress,
etc.) attains a critical value or by approximating experimental data by an analytical
function.
The latter class of yield functions are not obtained from a calculus based on the
crystallographic structure of the material; they are purely phenomenological func-
tions. The advantages of using such phenomenological yield functions instead of
those based on the crystallographic texture are [10]:

• they usually have a simpler mathematical form

• they are easy to understand and manipulate (‘user-friendly’)
• they may be easily implemented in finite element codes
• they may be generalized to describe the anisotropic behavior of the materials
• they may also be easily adapted to describe the behavior of fcc instead of bcc
materials (for example, by simply changing the exponent of the Hershey family
yield criteria).

The main disadvantage of the phenomenological approach is the poor accuracy

under multi-axial and non-proportional loading conditions.
2.2 Yield Criteria for Isotropic Materials 41

2.2.1 Tresca Yield Criterion

The oldest yield criterion was proposed by Tresca in 1864 [11] on the basis of his
observation that plastic strains appear by crystallographic gliding under acting shear
stresses.
According to this criterion the material passes from an elastic to a plastic state
when the maximum shear stress τ max reaches a critical value.
In the general case, the criterion may be written as follows:

max {|σ1 − σ2 | , |σ2 − σ3 | , |σ3 − σ1 |} = σ0 (2.26)

where σ 1 , σ 2 and σ 3 are principal stresses.

Under plane stress condition (σ 3 = 0), Eq. (2.26) becomes

σ1 − σ2 = σ0 = 2 K; σ1 > σ 2 (2.27)

σ2 − σ1 = σ0 = 2 K; σ1 < σ 2 (2.28)

Equations (2.27) and (2.28) can be unified:

σ1 − σ2 = ±σ0 = ±2 K; (2.29)

Equation (2.29) represents a polygon in the plane of the principal stresses σ 1 and
σ 2 and a hexagonal prism in the space, see Fig. 2.10.
By squaring Eq. (2.29) it is obtained

(σ1 − σ2 )2 = σ02 (2.30)

Fig. 2.10 Tresca yield

surface in the stress space
42 2 Plastic Behaviour of Sheet Metal

In the case when the stress components σ 11 and σ 22 do not coincide with the
principal stresses, the latter takes the following form:

(σ11 − σ22 )2 + 4σ12

2
= σ02 (2.31)

2.2.2 Huber–Mises–Hencky Yield Criterion

This criterion was proposed independently by Huber [12] and von Mises [13] and
further developed by Hencky [14]. It is based on the observation that a hydrostatic
pressure cannot cause plastic yielding of the material. Thus, the conclusion that only
the elastic energy of distortion influences the transition from an elastic to a plastic
state comes naturally. This idea has been proposed first by Maxwell in a letter to
Thomson [15] but it was not published at the time. For simplicity, in the following
text the criterion will often be referred to as the Mises criterion.
This criterion can be formulated as follows: the material passes from an elastic
to a plastic state when the elastic energy of distortion reaches a critical value that is
independent of the type of the stress state.
The elastic potential energy, Wp , has two components: the volumetric change
energy Wv and the energy of distortion, Wf .

Wp = W v + W f (2.32)

Thus Wf can be written in the form

W f = W p − Wv (2.33)

After replacing the expressions of the elastic potential energy and energy of
distortion in Eq. (2.33), it is obtained

1+μ 
Wf = · (σ1 − σ2 )2 + (σ2 − σ3 )2 + (σ3 − σ1 )2 (2.34)
6E

In the case of uniaxial tension (σ 2 = σ 3 = 0) yielding of the material occurs if

σ 1 = σ 0 , where σ 0 is the uniaxial yield stress in direction 1. Thus, the critical value
of the elastic energy of distortion at which the material passes from an elastic to the
plastic state is:

1+μ 2
Wf = 2σ0 (2.35)
6E
Then the Mises criterion may be written in the form:

1+μ 2 1+μ 
2σ0 = (σ1 − σ2 )2 + (σ2 − σ3 )2 + (σ3 − σ1 )2 , (2.36)
6E 6E
2.2 Yield Criteria for Isotropic Materials 43

Fig. 2.11 Graphical σ1

representation of the Tresca σ0
and von Mises yield criteria Von Mises

Tresca

–σ0

σ0 σ2

–σ0

or

(σ1 − σ2 )2 + (σ2 − σ3 )2 + (σ3 − σ1 )2 = 2σ02 (2.37)

For plane stress (σ3 = 0) Eq. (2.37) becomes

(σ1 − σ2 )2 + σ12 + σ22 = 2σ02 (2.38)

rewritten in the form

σ12 + σ22 − σ1 · σ2 = σ02 (2.39)

Another equivalent of the Mises criterion is

 2  2
σ1 − σ2 σ1 + σ2
3 + =4 (2.40)
σ0 σ0

Equations (2.39) and (2.40) represent an ellipse in the plane of the principal
stresses σ1 –σ2 which is circumscribed to the polygon given by Tresca criterion,
see Fig. 2.11.

2.2.3 Drucker Yield Criterion

In order to represent the experimental data located between the Tresca and Mises
yield surfaces, Drucker [16] proposed the following criterion:

J23 − CD J32 = F (2.41)

where J2 and J3 are the second and third invariants of the stress tensor, respectively,
and CD is a constant.
44 2 Plastic Behaviour of Sheet Metal

Equation (2.41) may be generalized in the form

3p 2p
J 2 − C D J3 = F (2.42)

p being an integer.

2.2.4 Hershey Yield Criterion

Based on the Norton [17] and Bailey [18] laws used for non-linear creep, Hershey
introduced a non-quadratic formulation of the yield criterion [19]

(σ1 − σ2 )a + (σ2 − σ3 )a + (σ3 − σ1 )a = 2Y a (2.43)

Here Y is the uniaxial yield stress and a is an exponent determined based on the
crystallographic structure of the material. For a = 2, Eq. (2.43) reduces to the Mises
yield condition, whereas for a = 1 and in the limit case a → ∞ it leads to the
Tresca yield condition. For 2 < a < 4, the corresponding surface lies outside the
Mises circular cylinder, whereas for 1 < a < 2 and for a > 4, it lies between Mises
and Tresca. The Hershey formulation has been used later (1972) by Hosford [20].
The Hershey’s formulation has been generalized by Karafillis and Boyce [21] in
the following form

 = (1 − c) 1 − c2 , (2.44)
where

1 = |S1 − S2 |2 k + |S2 − S3 |2 k + |S3 − S1 |2 k = 2σe2 k (2.45)

and
22 k + 2 2 k
2 = |S1 |2 k + |S2 |2 k + |S3 |2 k = σe (2.46)
32 k
Here S1 , S2 , and S3 are the principal deviatoric stresses, c is a weighting coef-
ficient, and 2k is an exponent having the same significance as the exponent a in
Hosford’s criterion.
For k = 1 the Eqs. (2.45) and (2.46) take the form given by von Mises, however,
for k = ∞ Eq. (2.45) becomes the Tresca function and (2.46) gives an upper limit
of the yield surface.
The value of the coefficient c is in the range [0, 1]. It determines the weight of
the functions F1 and F2 in the yield function F. As a consequence, there are two
parameters k and c that may be used in order to ‘adjust’ the shape of the yield locus
whereas the other criteria use only one parameter (exponent a or m) for this purpose.
Therefore the new criterion is very flexible.
More examples of isotropic yield functions are reviewed by Życzkowski [9] and
Yu [22].
2.3 Classical Yield Criteria for Anisotropic Materials 45

2.3 Classical Yield Criteria for Anisotropic Materials

The first yield criterion for anisotropic materials was proposed by von Mises in
the form of a quadratic function [23]. Though it was initially used for describ-
ing the plastic behavior of an anisotropic single crystal, later it was also used for
polycrystals.
This criterion is given by

 = h11 σ11 2 + h σ 2 + h σ 2 + h σ 2 + h σ 2 + h σ 2 + 2h σ σ +
22 22 33 33 44 12 55 23 66 31 12 11 22
2h13 σ11 σ33 + 2h14 σ11 σ12 + 2h15 σ11 σ23 + 2h16 σ11 σ31 + 2h23 σ22 σ33 +
2h24 σ22 σ12 + 2h25 σ22 σ23 + 2h26 σ22 σ31 + 2h34 σ33 σ12 + 2h35 σ33 σ23 +
2h36 σ33 σ31 + 2h45 σ12 σ12 + 2h46 σ12 σ31 + 2h56 σ23 σ31
(2.47)
where hij (i, j = 1, 2, . . ., 6) are coefficients of anisotropy which can be identified
by mechanical tests. Equation (2.47) gives a quadratic function containing products
implying both normal and shear stresses.
Olszak [24] gave a generalization of this function for non-homogeneous
anisotropic materials. In the case of an orthotropic material, it can be reduced to
a quadratic function having only six terms and coefficients of anisotropy. This is the
same as the function proposed by Hill in 1948 [25].

2.3.1 Hill’s Familly Yield Criteria

2.3.1.1 Hill 1948 Yield Criterion

In 1948 Hill [25] proposed an anisotropic yield criterion as a generalization of the
Huber-Mises-Hencky criterion. The material is supposed to have an anisotropy with
three orthogonal symmetry planes.
The yield criterion is expressed by a quadratic function of the following type:

2f σij ≡ F (σ22 − σ33 )2 + G (σ33 − σ11 )2 + H (σ11 − σ22 )2

 
2 + 2 Mσ 2 + 2 Nσ 2 = 1, (2.48)
+2Lσ23 31 12

Here f is the yield function; F, G, H, L, M and N are constants specific to the

anisotropy state of the material, and x, y, z are the principal anisotropic axes.
In the case of sheet metals, axis 1 is usually parallel to the rolling direction, 2 is
parallel to the transverse direction and 3 is collinear with the normal direction.
If the tensile yield stresses in the principal anisotropy directions are denoted by
X, Y and Z it can easily be shown that

1 1 1
= G + H; = H + F; =F+G (2.49)
X2 Y2 Z2
From this equation, by some simple mathematical calculations the coefficients F,
G and H are obtained as functions of the uniaxial yield stresses:
46 2 Plastic Behaviour of Sheet Metal

1 1 1 1 1 1 1 1 1
2F = + 2 − 2; 2G = + 2 − 2; 2H = + 2 − 2 . (2.50)
Y2 Z X Z2 X Y X2 Y Z
If R, S and T are the shear yield stresses associated to the same directions, then

1 1 1
2L = ,2M = 2,2N = 2. (2.51)
R2 S T
Only one of the parameters F, G, H can be negative. This situation rarely occurs
in practice (it would cause great differences between the stresses); F > G if and only
if X > Y, etc. L, M and N are always positive.
As a consequence, in order to give a complete description of the anisotropy of
the material, six independent yield stresses (X, Y, Z, R, S and T) have to be known as
well as the orientation of the principal anisotropy axes.
The yield criterion may be interpreted as a surface in a six-dimensional space of
the stress components. The points located at the interior of the surface represent the
elastic states of the material, while points belonging to the surface correspond to the
plastic state.
For plane stress (σ33 = σ31 = σ23 = 0; σ11 = 0; σ22 = 0; σ12 = 0), the yield
criterion becomes

2 2 2
 
2f σij ≡ (G + H) σ11 − 2Hσ11 σ22 + (H + F) σ22 + 2 Nσ12 = 1. (2.52)

After introducing the yield stress X, Y, Z and T, Eq. (2.52) may be rewritten as
 
1 2 1 1 1 1 2 1 2
σ − + 2− 2 σ11 σ22 + σ22 + 2 σ12 = 1, (2.53)
X 2 11 X 2 Y Z Y 2 T

When the principal directions of the stress tensor coincide with the principal
anisotropic axes, the Hill 1948 yield criterion has the form
 
1 2 1 1 1 1 2
σ − + 2− 2 σ1 σ2 + σ = 1, (2.54)
X2 1 X2 Y Z Y2 2

where σ 1 , σ 2 are the non-zero principal stresses.

When simulating sheet metal forming processes, the anisotropy coefficients
are denoted (r0 , r45 , r90 ) and the yield stresses in the directions of the principal
anisotropic axes are denoted as follows: X = σ 0 , Y = σ 90 . The relations between
the anisotropy coefficients and the coefficients F, G, H,. . . may be easily obtained
from the flow rule associated to the yield function:

H H N 1
r0 = ; r90 = ; r45 = − . (2.55)
G F F+G 2
It can be shown that the following relation between the yield stresses and the
anisotropy coefficients applies:
2.3 Classical Yield Criteria for Anisotropic Materials 47
%
σ0 r0 (1 + r90 )
= (2.56)
σ90 r90 (1 + r0 )

This equation implies that from r0 > r90 it follows σ 0 > σ 90 and the reciprocal,
however, some materials do not satisfy this condition.
The last of the three Eqs. (2.55) leads to
 
1
N = (F + G) r45 + (2.57)
2

or, according to Eq. (2.57),

1 2r45 + 1 1 r0 + r90
2N = = 2 (2r45 + 1) . (2.58)
Z2 2 σ0 r90 (1 + r0 )

Finally it is obtained
' (
1 1 1 1 r0 + r90 1 2
σ2 − + 2 − 2 σ11 σ22 + 2 σ22 +
σ02 11 2
σ0 σ90 σ0 r90 (1 + r0 ) r90 (2.59)
1 r0 + r90 2 =1
+ 2 (2r45 + 1) σ12
σ0 r90 (1 + r0 )

As σ0 and σ90 are not independent, but related by (2.57), Eq. (2.59) may also be
written as

2 2r0 r0 (1 + r90 ) 2 r0 + r90 2

σ11 − σ11 σ22 + σ + (2r45 + 1) σ12 = σ02 (2.60)
1 + r0 r90 (1 + r0 ) 22 r90 (1 + r0 )
In case that the principal directions of the stress tensor are coincident with the
anisotropic axes (σ 11 = σ 1 , σ 22 = σ 2 , σ 12 = 0), the Hill 1948 yield criterion can
be written as a dependence of the principal stress in the form

2r0 r0 (1 + r90 ) 2
σ12 − σ1 σ2 + σ = σ02 (2.61)
1 + r0 r90 (1 + r0 ) 2
or, taking into account Eq. (2.56)

2r0 r0 (1 + r90 ) 2 r0 (1 + r90 ) 2

σ12 − σ1 σ2 + σ2 = σ . (2.62)
1 + r0 r90 (1 + r0 ) r90 (1 + r0 ) 90
From Eqs. (2.61) and (2.62) it follows that in order to define the yield under
plane stress condition, three mechanical parameters, namely the coefficients r0 and
r90 and one of the uniaxial yield stresses σ 0 and σ 90 are needed.
Equations (2.61) or (2.62) geometrically represent families of ellipses depending
on the parameters r0 and r90 . The influence of these parameters upon the yield loci
is demonstrated in Figs. 2.12 and 2.13.
48 2 Plastic Behaviour of Sheet Metal

Fig. 2.12 Influence of the r0 1.5

anisotropy coefficient upon
the shape of the yield locus
defined by the Hill 1948 1.0
criterion r0

0.5

σ2/Y
0.0

–0.5

–1.0

–1.5
–1.5 –1.0 –0.5 0.0 0.5 1.0 1.5
σ1/Y

Fig. 2.13 Influence of the r90 1.5

anisotropy coefficient upon
the shape of the yield locus
defined by the Hill 1948 1.0
criterion
r90
0.5
σ2/Y

0.0

–0.5

–1.0

–1.5
–1.5 –1.0 –0.5 0.0 0.5 1.0 1.5
σ1/Y

In case of a material exhibiting only normal anisotropy (r0 = r90 = r) Eq. (2.56)
imposes that σ 0 = σ 90 and Eqs. (2.61) and (2.62) take the same form:

2r
σ12 − σ1 σ2 + σ22 = σu2 (2.63)
1+r
2.3 Classical Yield Criteria for Anisotropic Materials 49

Fig 2.14 Influence of the 1.5

normal anisotropic coefficient
on the shape of the yield r
locus defined by the Hill 1948 1.0
criterion

0.5

σ2/Y
0.0

–0.5

–1.0

–1.5
–1.5 –1.0 –0.5 0.0 0.5 1.0 1.5
σ1/Y

where σu is the uniaxial yield stress.

The influence of the parameters r and σ u upon the shape of the yield locus is
shown in Figs. 2.14 and 2.15.
It can be seen that if r < 1, the yield locus predicted by the Hill 1948 criterion
is located inside the one given by von Mises; if r > 1 the Hill 1948 yield locus
is outside the von Mises yield locus. Woodthrope and Pearce [26] and Pearce [27]
noticed that some materials (in particular aluminum alloys) have the yield locus
outside the von Mises surface though their r-coefficient was less than one.
This behavior cannot be properly described by the Hill 1948 yield criterion and
materials exhibiting it are called ‘anomalous’.
Indeed, by rewriting Eq. (2.63) for the case of the equibiaxial tension (σ 1 = σ 2 =
σ b ) one obtains

"
1+r
σb = σu , (2.64)
2

where σ b is the equibiaxial yield stress.

It follows that if r > 1, then σ b > σ u and if r < 1 then σ b < σ u . This problem
will be considered again below.
It can be seen from Fig. 2.15 that if the uniaxial yield stress increases, the yield
surface expands uniformly. This is called isotropic hardening.
Equation (2.63) may be rewritten in the form
50 2 Plastic Behaviour of Sheet Metal

Fig 2.15 Influence of the 1.5

uniaxial yield stress on the
shape of the yield locus
defined by the Hill 1948 1.0
criterion
σu
0.5

σ2/Y
0.0

–0.5

–1.0

–1.5
–1.5 –1.0 –0.5 0.0 0.5 1.0 1.5
σ1/Y

 2  2
σ1 − σ2 σ1 + σ2
(1 + 2r) + = 2 (1 + r) (2.65)
σu σu

In case of an isotropic material (r = 1), Eqs. (2.63) and (2.65) reduce to the von
Mises relationships (2.38) and (2.39), respectively.
In the above equations, the yielding condition is expressed by relations between
components of the stress tensor. This defines the shape of the yield surface. Its exten-
sion in the space of the stress components is given by the equivalent or effective
stress σ e . This is the stress associated to a simple mechanical test that causes the
transition of the material from an elastic state to a plastic state. Yield criteria are
frequently expressed using this parameter.
If we take into account Eqs. (2.52) and (2.7), the equivalent stress can be
expressed as

1
σ = Yθ [G cos2 θ + F sin2 θ + H(cos2 θ − sin2 θ ) + 2 N sin2 θ cos2 θ ] 2 (2.66)

Consequently, Fθ will be defined by the relationship

1
Fθ = [G cos2 θ + F sin2 θ + H(cos2 θ − sin2 θ ) + 2 N sin2 θ cos2 θ ] 2 (2.67)

In the case of the Hill’48 yield criterion, the uniaxial yield stress corresponding
to a direction inclined at the angle θ with respect to the rolling direction is
2.3 Classical Yield Criteria for Anisotropic Materials 51

Y(h)
Yθ = 1
(2.68)
2
[G cos2 θ + F sin θ + H(cos2 θ − sin2 θ ) + 2 N sin2 θ cos2 θ ] 2

If the yield parameter Y(h) is set equal to the uniaxial yield stress σu , the uniaxial
yield stress predicted by this criterion is

σu
Y0 = √ . (2.69)
G+H

The expression of the uniaxial anisotropy predicted by the Hill’48 yield criterion
is obtained by replacing Eq. (2.67) in Eq. (2.17):

G cos4 θ + F sin4 θ + H cos2 2θ + 12 N sin2 2θ

rθ = − 1. (2.70)
G cos2 θ + F sin2 θ

Equations (2.69) and (2.70) are used for predicting the uniaxial yield stress and
coefficient of plastic anisotropy, in the case when the parameters F, G, H and N
of the Hill 1948 yield criterion are related to the experimental yield stress σu and
the experimental coefficients of plastic anisotropy r0 , r45 and r90 . The identifi-
cation of the yield criterion can be also performed by using three experimental
values of the yield stress and one experimental value of the coefficient of plastic
anisotropy.

2.3.1.2 Comments on the Hill’48 Yield Criterion

When describing the anisotropy of metals, the Hill 1948 yield criterion has the
advantage that its basic assumptions are easy to understand. The parameters
included in the yield functions have a direct physical meaning. This explains its wide
use in practice. In addition, the model has a simple formulation for the 3D case. The
criterion needs a small number of mechanical parameters for determining the yield
function. Under plane stress conditions, four parameters are sufficient, namely, r0 ,
r45 , r90 and σ 0 or r0 , r45 , r90 and σ 90 because the parameters r0 , r90 , σ 0 and σ 90
are related by Eq. (2.56). In practice, the values of the anisotropy coefficients and
an average of the uniaxial yield stress (σ u ) are used.
Besides its advantages, the criterion also has some drawbacks:

1. It cannot represent the ‘anomalous’ behavior observed by Woodthrope and

Pearce [26]: r < 1 and σ b > σ u (or the reciprocal) because the criterion predicts
(see Eq. 2.64)

"
1+r
σ b = σu (2.71)
2
52 2 Plastic Behaviour of Sheet Metal

2. It cannot represent the ‘second order anomalous’ behaviour (this new concept
has been introduced in the literature by the author in the paper [28]): rr900 > 1
and σσ900 < 1 (or vice-versa) because the criterion predicts

%
σ0 r0 (1 + r90 )
= (2.72)
σ90 r90 (1 + r0 )

3. It can only be applied to materials forming four ‘ears’ in axisymmetric deep-

drawing processes whereas in practice also different numbers of ears are
observed.
4. The curves equivalent stress σe vs. equivalent strain εe for different materials
depend on the loading path, although they should be unique and intrinsic for a
given material.
5. In uniaxial tensile tests, the dependence of the yield stress on direction is poorly
predicted by the theory, although the variation of the r-coefficient is properly
determinate. There is also the possibility to perform the identification of the Hill
1948 model by using only yield stresses and not coefficients of plastic anisotropy.
Of course, in such a case, the accuracy of the predictions related to the variability
of the anisotropy coefficients is poorer. The model cannot use the same num-
ber of yield stresses and anisotropy coefficients in the identification procedure.
This characteristic is an important drawback, especially in the case of mate-
rials exhibiting significant variations of the mechanical parameters mentioned
above [28].

2.3.1.3 Hill 1979 Yield Criterion

As mentioned above the quadratic yield criteria cannot describe the plastic behavior
of some materials such as aluminum alloys. According to Woodthrope and Pearce
[26, 27], these metals, though having anisotropy coefficients less than unity, have an
experimental yield surface located outside the surface predicted by the von Mises
yield criterion (σ b > σ u ). Since the Hill 1948 yield criterion cannot describe such
a situation, the above authors improperly called this behavior ‘anomalous’ (see also
the comments to Eq. 2.64).
Starting from this experimental observation, in the 1970s several researchers con-
cluded independently from each other that only non-quadratic functions are suitable.
In the next sections various criteria of this kind are presented in order to describe
the so-called ‘anomalous’ behavior of some materials.
Whereas the quadratic criterion by Hill can be applied both to sheet metal and to
round bars, the non-quadratic yield criteria described below generally can only be
applied to sheet metal.
This function was expressed in its general form as well as in four special cases
[29]. If the directions of the principal stresses coincide with the axes of anisotropy
the criterion is written as
2.3 Classical Yield Criteria for Anisotropic Materials 53

f |σ2 − σ3 |m + g |σ3 − σ1 |m + h |σ1 − σ2 |m + a |2σ1 − σ2 − σ3 |m +

(2.73)
+b |2σ2 − σ1 − σ3 |m + c |σ3 − σ1 − σ2 |m = σem

Here f, g, h, a, b and c are anisotropy coefficients; the exponent m can be

calculated from the non-linear relationship obtained from Eq. (2.73) written for
equibiaxial tension (σ 1 = σ 2 = σ b ; σ 3 = 0) [29]:
m '  m−1  (
2 − 2 · (a − c)

σb 1
= (1 + r) · 1 + , (2.74)
σm 2 a + 2m−1 · c + f

This equation can be solved numerically.

If m = 2, there are seven parameters taking different values (according to the
convexity conditions), ensuring an improved flexibility as compared to the Hill 1948
yield criterion.
For plane stress (σ3 = 0) Eq. (2.73) can be written in four particular forms given
by Hill [29]. The most frequently used formulation is the so-called Case 4, according
to Hill’s paper [29]. As a consequence, we shall describe only this model.
Case 4 is defined by the following constraint acting on the parameters: a = b =
f = g = 0. Equation (2.73) reads now

c |σ1 + σ2 |m + h |σ1 − σ2 |m = σem (2.75)

Lian, Zhou and Baudelet [30] proved that the four forms of the Hill 1979 yield
criterion can be expressed as functions of only two coefficients depending on the
parameters r and m. The coefficients in Eq. (2.75) is given by

r 1+2·r
c= , h= (2.76)
2 (1 + r) 2 (1 + r)

These particular forms of the Hill’79 are based on the assumption of planar
isotropy whereby the axes 1 and 2 can be arbitrarily oriented in the plane of the
sheet metal and the terms associated to shear stress are not necessary.
However, it is possible to generalize the Hill 1979 criterion for taking into
account planar anisotropy [31–33].
For planar isotropy, the most widely used expression of the Hill 1979 yield cri-
terion is in the form (2.75). This expression shall be considered more thoroughly. It
can be rewritten in the form

|σ1 + σ2 |m + (1 + 2 · r) |σ1 − σ2 |m = 2 (1 + r) σum (2.77)

The convexity condition requires that m be greater than unity. In this particular
case Eq. (2.74) has the form
 m
σb 1+r
= (2.78)
σu 2m−1
54 2 Plastic Behaviour of Sheet Metal

By solving Eq. (2.78), usually non-integer values of m are obtained.

For the aluminum alloy DIN-AlMgSi1 having σ u = 138 MPa, σ b = 156 MPa,
r = 0.50 it follows m = 1.358 [34]. For m = 2 the criterion is reduced to Eq. (2.63)
associated to the Hill 1948 yield criterion.
In 1977 Bassani published a paper [35] where he proved independently of Hill
that a family of yield functions depending on four parameters may be constructed
in order to approximate a relatively broad category of transversally isotropic yield
surfaces predicted by the Bishop–Hill theory [36].
He proposed the yield function:

) σ1 + σ2 )n ) σ1 − σ2 )m
) ) ) )
f =)
) ) +)
) ) −1 (2.79)
2σ )
b 2τ )

where τ is the yield stress in pure shearing, while n and m are two constants greater
than unity.
One may notice that this function is a generalised expression of Case 4 proposed
by Hill in 1979 (for n = m, Eq. (2.79) reduces to Eq. (2.75)). The difference consists
in the way of defining the coefficients.
Bassani concluded that the proposed family of yield functions, for arbitrarily
chosen values of m or n approximates the yield surfaces predicted by the Bishop-Hill
theory.
Extensions of the Hill 1979 model for stress states with a planar shearing
component have been proposed by Chu [32], Zhou [33], [37] and Monteillet
[38].

Advantages of the Hill 1979 yield criterion are [30]:

• it can describe the Woodthrope-Pearce ‘anomalous’ behavior of materials

• it has a relatively simple form
• it leads to of an analytical expression of the associated flow rule and
equivalent strain.

Disadvantages are:

• it cannot describe the behavior of the materials having rr900 = 1 and

σ0
σ90  = 1
• due to the non-integer value of the exponent m, it requires numerical
procedures even for the solution of quite simple cases
• although it allows the description of the ‘anomalous’ behavior, the
predicted yield surfaces are sometimes far from the experimental surfaces
predicted by the Bishop-Hill theory [36].

2.3.1.4 Hill 1990 Yield Criterion

As mentioned above, the yield criterion proposed by Hill in 1979 can only be used
when the directions of the principal stresses are coincident with the orthotropic axes.
2.3 Classical Yield Criteria for Anisotropic Materials 55

This causes severe limitations of application. The generalisation of the criterion by

expressing it in a general coordinate system was realised by Hill in 1990 [39]:

2 )m/2 + )σ 2 + σ 2 + 2σ 2 )(m/2)−1 ·
ϕ = |σ11 + σ22 |m + σbm τ m )(σ11 − σ22 )2 +4σ12
 & ) ) ) )
11 22 12

2 + b (σ − σ )2 = (2σ )m
 2 
· −2a σ11 − σ22 11 22 b
(2.80)
Here σ b is the yield stress in equibiaxial tension, τ is the yield stress in pure
shear deformation (σ 1 = –σ 2 ), a and b are material constants.
The m exponent is obtain by solving the following equation

 m
2 · σb
= 2 · (1 + r45 ) (2.81)
σ45

The value of the m exponent is

ln [2 (r45 + 1)]
m= (2.82)
2σb
ln
σ45

The constants a and b are determined from the equations

1 )) 2σb m 2σb m ))
)    )
a= − );
4 ) σ90 σ0
 m  m   m (2.83)
1 2σb 2σb 2σb
b= + − .
2 σ0 σ90 σ45

The ratio σ b /τ may be also expressed as a function of coefficient r45 (see

Eq. 2.57):

 σ m
b
= 1 + 2 · r45 (2.84)
τ

It can be shown that the parameters a and b can also be determined as functions
of the anisotropy coefficients r0 , r45 and r90 :
 & 
(r0 − r90 ) 1 − (m − 2) 2 · r45
a= ;
(r0 + r90 ) − (m − 2) · r0 · r90
(2.85)
m · [2 · r0 · r90 − r45 · (r0 + r90 )]
b= .
(r0 + r90 ) − (m − 2) · r0 · r90
56 2 Plastic Behaviour of Sheet Metal

F−G F + G + 4H − 2 N
a= ; b= (2.86)
F+G F+G
or as functions of the coefficients in Eqs. (2.50) and (2.51).
By using the methodology described in Sect. 2.1.1, the equivalent stress associ-
ated to the Hill 1990 yield criterion can be expressed as follows:

1  σ m
b 1
σ = Yθ [1 + − 2a cos 2θ + b cos2 2θ ] m (2.87)
2 τ
The corresponding function Fθ is

1  σ m
b 1
Fθ = [1 + − 2a cos 2θ + b cos2 2θ ] m (2.88)
2 τ
In the case of Hill 1990 model, the yield parameter Y(h) is set equal to the biaxial
yield stress (Y(h)=Yb ). In order to preserve the original notations, we shall use the
symbol σ b for the biaxial yield stress. According to Eq. (2.10), the uniaxial yield
stress is defined as follows:

4σb
Yθ =  σ m 1
(2.89)
b
[1 + − 2a cos 2θ + b cos2 2θ ] m
τ
The planar distribution of the uniaxial anisotropy coefficient is predicted by the
formula
 σ m
b 2
−1+b cos2 2θ
rθ = τ m −1 (2.90)
m−2
2 − 2a cos 2θ + b cos2 2θ
m
Equations (2.89) and (2.90) allow the determination of the planar distribution of
the uniaxial yield stress and coefficient of plastic anisotropy.
A detailed discussion of the computational methodology of the parameters in
Hill’s 1990 yield function is presented in [40] and [41]. By comparing the yield loci
computed with Eq. (2.80) using the coefficients a and b evaluated on the basis of the
yield stresses (Eq. 2.83) as well as on the basis of the anisotropy coefficients (Eq.
2.85), with the yield loci given by the Taylor theory, Lin and Ding [40] concluded
that the identification procedure using stresses ensured a better approximation.
Using Mohr’s circle, the yield function (2.80) can be expressed in principal
stresses:
 m
σb
|σ1 − σ2 |m + )σ12 + σ22 ) /
)(m 2)−1
ϕ = |σ1 + σ2 |m +
)
m
·
τ
(2.91)
· −2a σ1 − σ22 + b (σ1 − σ2 )2 cos 2θ cos 2θ = (2σb )m
 2 
2.3 Classical Yield Criteria for Anisotropic Materials 57

Here θ is the angle between the direction of σ 1 and the rolling direction. If a = b
or θ = π /4, Eqs. (2.80) and (2.91) reduce to Case 4◦ of the Hill 1979 yield criterion
(Eq. 2.75).
The flow rule associated to this yield criterion is presented in [41]. The Hill 1990
yield criterion preserves all the advantages of the Hill 1979 criterion. In addition,
it includes all the planar components of the stress tensor (σ 11 , σ 22 and σ 12 ). As a
consequence it is possible to evaluate the distribution of the uniaxial yield stress and
anisotropy coefficient in the plane of the sheet metal. From Eqs. (2.81) to (2.83) it
can be seen that five parameters are needed for defining the yield function: σ 0 , σ 45 ,
σ 90 , σ b and r45 .
If Eqs. (2.84) and (2.85) are used for calculating the coefficients a and b, the
mechanical parameters necessary for defining the yield function are σ 45 , σ b , r0 , r45
and r90 and again three uniaxial tensile tests and a biaxial tensile test are required.
In their analysis of Hill’s 1990 yield criterion, Lin and Ding [40] proposed a more
general expression:
)m− 1
ϕ = |σ1 + σ2 |m + (1 + 2R) |σ1 − σ2 |m + )σ12 + σ22 ) 2 ·
)
(2.92)
s |σ s s m
· {−2a (|σ1 | − |
2 )+b 1 |σ − σ 2 | cos 2θ} cos 2θ = (2σb )

where R and s are material parameters. For θ = π /4 and s = 2 Eq. (2.92) is simpli-
◦
fied to Eq. (2.75) i.e. Hill’s 1979 criterion (Case 4 ). The new characteristic of this
criterion is that the additional terms are not quadratic. The exponent s is calculated
from the equation
  m 
mr0 2 (r90 − R) + r90 2σ b
σ90 − 2 (1 + R)
s=−  m  (2.93)
(1 + R) (r0 − r90 ) − 12 (r0 + r90 + 2r0 r90 ) 2σ b
σ90 − 2 (1 + R)

where R is determined from

 m 
1 2σb
R= −1 . (2.94)
2 σ45

If the identification of the parameters a and b is made using stresses then

 m  m 
1 2σb 2σb
a= − ;
4 σ90 σ0
 m  m  (2.95)
1 2σb 2σb
b= + − 2 (1 + R) .
2 σ0 σ90
where by s is calculated from

m [(b − 2a) r0 + 2 (r0 − r45 )]

s= (2.96)
(1 + r0 ) b
58 2 Plastic Behaviour of Sheet Metal

The above equations show that the determination of the proposed yield function
demands six parameters (σ 0 , σ 90 , σ b , r0 , r45 and r90 ) in the first identification pro-
cedure, or σ 45 , σ b , r0 , r45 and r90 in the second one. In order to establish these
parameters, three uniaxial tensile tests and a biaxial tensile test have to be carried
out. The paper by Lin and Ding [40] also gives a comparison of the yield surfaces
computed with coefficients a, b, s and R identified in the two manners described
above with the yield surfaces predicted by the Taylor theory. The identification of
the coefficients a and b could be made on the basis of the anisotropy coefficients
[40]. In this case a better approximation of the yield surfaces predicted by the Taylor
theory would be obtained (contrary to the situation occurring when the original Hill
1990 yield criterion is used). This is due to the non-quadratic terms added to the
yield function.
Leacock [42] extended the formulation proposed by Hill in 1990 [39] by defining
the following yield criterion:
)(m/2)−2  2
|σ1 + σ2 |m + Am |σ1 − σ2 |m + )σ12 + σ22 ) σ1 − σ22
) 

H(σ12 + σ22 ) + I(σ12 − σ22 ) cos 2θ + (σ1 − σ2 )2 (2.97)

J(σ12 + σ22 ) + K(σ12 − σ22 ) cos 2θ cos 2θ ] = (2σb )m .

The evaluation of the constants A, H, I, J, K and m is based on 6 mechanical

parameters, namely σ 0 , σ 45 , σ 90 , σ b , r0 , r45 and r90.
Equation (2.97) can be reduced to the formulation proposed by Hill in 1990
(Eq. 2.80).
The uniaxial yield stress is defined by the relationship
 −1/m
σθ = 2σb 1 + Am + (H + I cos 2θ) + (J + K cos 2θ) cos2 2θ . (2.98)

while the coefficient of uniaxial anisotropy is expressed as

Am + 0.5(H + I(cos 2θ ) cos 2θ + (m + 2)/(2m) /(J + K cos 2θ ) cos2 2θ

rθ = −1/2.
2 1 + 0.5(H + I(cos 2θ ) cos 2θ + (m − 2)/(2m) /(J + K cos 2θ ) cos2 2θ
(2.99)
Due to the use of seven mechanical parameters in the identification procedure,
Leacock’s model gives better predictions than Hill 1990 and Lin 1996 models both
for the yield surfaces and the planar distributions of the uniaxial yield stresses and
coefficient of plastic anisotropy.
Advantages of the Hill 1990 yield criterion are the following:

• it allows to describe both the ‘first order anomalous behavior’ (r < 1, σ b < σ u )
and the ‘second order anomalous behavior’ (r0 < r90 , σ 0 > σ 90 and vice-versa)
• it is able to describe very well the variation of the anisotropy coefficient and of
the uniaxial yield stress in the plane of the sheet
2.3 Classical Yield Criteria for Anisotropic Materials 59

• it has a great flexibility due to the high number of the mechanical parameters
incorporated
• Leacock’s modification [42] is shown to explicitly deal with the second order
anomalous behaviour.
However, there are also some disadvantages:
• the formulation is not user-friendly
• due to the trigonometric functions incorporated in their formulation, both Lin
and Ding [40] and Leacock [42] extensions of the Hill 1990 model need larger
CPU times when used in the numerical simulation of sheet metal forming
processes.

2.3.1.5 Hill 1993 Yield Criterion

In 1993 Hill [43] improved the model of plastic behavior of textured sheet metals,
namely those observed when complex loads are applied along the planar orthotropic
axes.
There are some situations that cannot be dealt with completely by any of the
criteria described above. A comprehensive analysis of these problems is given in
[43] where Hill showed that for all the criteria described above the condition σ 0 =
σ 90 enforces r0 = r90 and the reciprocal.
However, it has been observed experimentally that some materials, especially
aluminum alloys and brass have almost equal yield stresses but different anisotropy
coefficients in rolling and transverse direction. This applies to the results reported
in [44] for a brass 70-30 sheet (σ 0 = 126 MPa, σ 90 = 125 MPa, r0 = 1.51, r90 =
0.37) and in [34] for AlMgSi1 sheet (σ 0 = 143 MPa, σ 90 =133 MPa, r0 = 0.48,
r90 = 0.61).
For this phenomenon the term ‘anomalous behavior of second order’ should be
used [28], in order to distinguish it from the ‘anomalous behavior’ of the aluminum
alloy described in [26, 27] whereby r < 1 though the equibiaxial yield stress is
higher than the uniaxial one: σ b /σ u > 1.
For this reason Hill [43] proposed a new yield criterion that should preserve the
generality of the one proposed in 1979, that is, the capability of modelling both the
‘anomalous behavior’ and the ‘anomalous behavior of second order’.
These constraints are satisfied by the following polynomial function, valid for
stress states in the first quadrant (biaxial tension) which is most relevant for thin
sheet metals:

σ12 σ22
 
cσ1 σ2 (pσ1 + qσ2 ) σ1 σ2
− σ1 σ2 + 2 + (p + q) − = σu2 (2.100)
σ02 σ0 σ90 σ90 σb σ0 σ90

where

c 1 1 1
= 2+ 2 − 2 (2.101)
σ0 σ90 σ0 σ90 σb
60 2 Plastic Behaviour of Sheet Metal

while p and q are calculated with the normality condition of the strain rate tensor to
the yield surface applied to function (2.100) at the intersection with the coordinate
axes:
 
2r0 (σb − σ90 ) 2r90 σb c 1
p= − + ; (2.102)
(1 + r0 ) σ02 2
(1 + r90 ) σ90 σ0 1 1 1
+ −
σ0 σ90 σb
 
2r90 (σb − σ90 ) 2r0 σb c 1
q= − + ; (2.103)
2
(1 + r90 ) σ90 2
(1 + r0 ) σ0 σ90 1 1 1
+ −
σ0 σ90 σb

From Eqs. (2.100) to (2.103) it follows that in order to define the yield function,
five mechanical parameters are required (r0 , r90 , σ 0 , σ 90 and σ b ). These parameters
can be determined by two uniaxial tensile tests and an equibiaxial tensile test.
By expressing the modulus of the principal stress σ 1 and σ 2 from the third order
term of Eq. (2.100) the criterion can be extrapolated to the other quadrants of the
plane (σ 1 , σ 2 ). This leads to some discontinuity of the yield locus which, however,
can be tolerated if the discontinuity errors are within the limits of the experimental
errors.
Equation (2.100) show that the proposed yield function is nonhomogeneous with
respect to σ 1 and σ 2 . Consequently it is not possible to get an explicit expression of
the strain increment from the normality condition.
Advantages of the Hill 1993 yield criterion are the following:

• it allows to describe both the ‘first order anomalous behavior’ (r < 1, σ b < σ u )
and the ‘second order anomalous behavior’ (r0 < r90 , σ 0 > σ 90 and vice versa)
• it has a relatively simple and user-friendly expression; it has a great flexibility
due to the five mechanical parameters incorporated.

However, there are also some disadvantages:

• the yield function is non-homogenous with respect to σ 1 and σ 2 and hence does
not allow to obtain explicit expressions of the strain increments
• it can be used only if the directions of the principal stresses are coincident with
the orthotropic axes
• it does not allow to describe the variation of the anisotropy coefficient and of the
uniaxial yield stress in the plane of the sheet
• the yield surface predicted by this function is far from that obtained from
polycrystal theories (Taylor or Bishop-Hill).

Due to the disadvantages mentioned above, the applicability of this yield criterion
is limited.
2.3 Classical Yield Criteria for Anisotropic Materials 61

2.3.2 Yield Function Based on Crystal Plasticity

(Hershey’s Familly)
Besides the family of Hill yield criteria presented in the previous subchapter, we
have to mention another class of models based on the isotropic formulation pro-
posed by Hershey [19]. Hosford ‘rediscovered’ Hershey’s model in 1972 [20] and
used it for the development of an anisotropic yield criterion. Barlat et al. [5] and
Banabic et al. [51] as well as some other researchers proposed further extensions of
the model.

2.3.2.1 Hosford Yield Criterion

Independently of Hill, Hosford proposed a yield criterion in the form [31]

F |σ22 − σ33 |a + G |σ33 − σ11 |a + H |σ11 − σ22 |a = σ a (2.104)

This is a generalization of his own isotropic criterion [20].

One may consider this criterion as a particular expression (a = b = c = 0 and
f = g) of Case 4◦ of Hill’s 1979 yield criterion (see Eq. 2.75).
The essential difference between the approaches by Hosford and Hill consists in
the different way of determining the exponent a. Hosford related a to the crystallo-
graphic structure of the material [45–48]. He concluded that the best approximation
was given by a = 6 for BCC materials and a = 8 for FCC materials, see also [45].
For plane stress Eq. (2.104) reduces to

r90 |σ11 |a + r0 |σ22 |a + r0 r90 |σ11 − σ22 |a = r90 (r0 + 1) σ0a (2.105)

The main advantage of the Hosford 1979 yield criterion is that by fitting the
value of the exponent a it ensures a good approximation of the yield locus computed
through the Bishop-Hill theory [36] as well as from experimental data.
An important drawback of this criterion is caused by the lack of shear stress: it
cannot predict the variation of the coefficient r with direction (planar anisotropy).

2.3.2.2 Barlat 1989 Yield Criterion

Barlat and Richmond [49] proposed a more general form of Hosford’s criterion for
isotropic materials [20] by expressing it in an x, y, z coordinate system, not necessar-
ily coincident with the principal directions (the so-called ‘tricomponent plane stress
yield surface’):

f = |k1 + k2 |M + |k1 − k2 |M + |2k2 |M = 2σeM (2.106)

Here k1 and k2 are invariants of the stress tensor while M is an integer exponent
having the same significance as the exponent a used by Hosford; k1 and k2 are
obtained from
62 2 Plastic Behaviour of Sheet Metal

σ11 + σ22
k1 =
2
% (2.107)
σ11 − σ22 2

k2 = 2
+ σ12
2

For σ 12 = 0 it follows σ 11 → σ 1 and σ 22 → σ 2 and Eq. (2.106) is reduced to

the isotropic Hosford criterion which now includes the shear stresses (incorporated
in the invariant k2 ).
The yield function given by Eq. (2.106) was also extended to the case of normal
anisotropy in the form:

f = a |k1 + k2 |M + b |k1 − k2 |M + c |2k2 |M = 2σeM (2.108)

where a, b and c depend on the anisotropy coefficients while k1 and k2 are calculated
from the equation

2
a=b=2−c= (2.109)
1+r
In 1989, Barlat and Lian [50] published a generalisation of Eq. (2.106) for
materials exhibiting planar anisotropy by introducing the following yield function:

f = a |k1 + k2 |M + a |k1 − k2 |M + c |2k2 |M = 2σeM (2.110)

The coefficients k1 and k2 are given by

 2 1/ 2
σ11 + hσ22 σ11 − hσ22
k1 = ; k2 = + p2 σ12
2
(2.111)
2 2

and a, c, h and p are material parameters identified by

σe M σe M
' ( ' (
2 −2 1+
τ σ
a = 2 − c = ' s2 (M ' 90 (M
σe σe
1+ − 1+
σ90 σ90 (2.112)
σe
h=
σ90
 1
σe 2 M
p= M
τs1 2a + 2 c

Here τ s1 and τ s2 are yield stresses for two different types of shear tests: σ 12 =
τ s1 for σ 11 = σ 22 = 0 and σ 12 = 0 for σ 22 = −σ 11 = τ s2 .
Using another identification procedure based on the coefficients r0 and r90 it is
obtained
2.3 Classical Yield Criteria for Anisotropic Materials 63

r90
"
r0
a=2−c=2−2 · ;
1 + r0 1 + r90
" (2.113)
r0 1 + r90
h= · .
1 + r0 r90

The coefficient p has to be calculated by a numerical procedure or by using

Eq. (2.112) instead. For σ 12 = 0, Eqs. (2.110) and (2.111) are practically reduced
to the Hosford yield criterion in principal stresses (except of the constant h, see
Eq. 2.105).
In order to establish the expression of the uniaxial yield stress, Eq. (2.7) will be
replaced in Eqs. (2.110) and (2.111). We get the relationship

Y0
Yθ = 1
, (2.114)
[a(F1 + F2 )M + a(F1 − F2 )M + (1 − a)(2F2 )M ] M

where

⎡' (2 ⎤
h sin2 θ + cos2 θ h sin2
θ − cos2θ
F1 = ; F2 = ⎣ + p2 sin2 θ cos2 θ ⎦ .
2 2
(2.115)
The function Fθ is obtained from Eq. (2.114):

1
Fθ = [a(F1 + F2 )M + a(F1 − F2 )M + (1 − a)(2F2 )M ] M (2.116)

The yield parameter Y(h) in Eq. (2.114) has been set equal to the uniaxial yield
stress corresponding to the rolling direction (Y(h) = Y0 ).
By replacing in Eq. (2.17) the Fθ expression given by Eq. (2.116) and perform-
ing some computations, we get the relationship defining the coefficient of plastic
anisotropy:

1
[a(F1 + F2 )M + a(F1 − F2 )M + (1 − a)(2F2 )M ] M
rθ = M−1
−1
a(K1 + K2 ) (t1 + t2 ) + a(K1 − K2 )M−1 (t1 − t2 ) + 2(a − 1)(2K2 )M−1 t2
(2.117)
where

 
h sin2 θ + cos2 θ h sin2 θ − cos2 θ 2 2 2
K1 = Yθ ; K2 = + p sin θ cos θ .
2 2
(2.118)
and t1 and t2 are
64 2 Plastic Behaviour of Sheet Metal

Fig. 2.16 Influence of r0 1.5

anisotropy coefficient on the
Barlat 1989 yield locus
1.0
r0

0.5

σy
0.0

–0.5

–1.0

–1.5
–1.5 –1.0 –0.5 0.0 0.5 1.0 1.5
σx

a(K1 + K2 )M−1 + a(K1 − K2 )M−1

t1 = ;
σ M−1
(2.119)
a(K1 + K2 )M−1 − a(K1 − K2 )M−1 + 2(a − 1)(2K2 )M−1
t2 = .
σ M−1

where σ represents the equivalent stress corresponding to the Barlat 1989 yield
criterion.
Equations (2.114) and (2.117) allow the calculation of the uniaxial yield stress
and the coefficient of plastic anisotropy corresponding to different directions in the
plane of the sheet metal.
Figures 2.16 and 2.17 shows that the anisotropy coefficient r0 and r90 act in a
different manner on the yield locus. In case of the ‘tricomponent plane stress’ the
influence of the exponent M extends to the region of biaxial tension (Fig. 2.18). The
figures demonstrates that the yield criterion by Barlat 1989 has a great flexibility.
This is due to the large number of parameters (four material parameters and M
chosen in accordance with the crystallographic structure of the material).
The advantages of the Barlat 1989 yield criterion are:

• the reduced number of the mechanical parameters (four parameters) used in the
identification
• relatively easy identification (except for the coefficient p)
• a relativ good prediction of the yield locus for aluminium alloys without high
anisotropy;
• by correctly choosing the exponent M a very good correlation with the yield locus
predicted by the Bishop-Hill theory is obtained.
2.3 Classical Yield Criteria for Anisotropic Materials 65

Fig. 2.17 Influence of r90 1.5

anisotropy coefficient on the
Barlat 1989 yield locus
1.0
r90
0.5

σy
0.0

–0.5

–1.0

–1.5
–1.5 –1.0 –0.5 0.0 0.5 1.0 1.5
σx

Fig. 2.18 Influence of 1.5

parameter M on the Barlat
1989 yield locus
1.0
M

0.5
σy

0.0

–0.5

–1.0

–1.5
–1.5 –1.0 –0.5 0.0 0.5 1.0 1.5
σx

The most important disadvantages are as follows:

• the coefficients of the yield function have not a direct and intuitive physical
significance.
• the evaluation of the parameter p can be performed only numerically, by solving
a non-linear equation.
66 2 Plastic Behaviour of Sheet Metal

• the model does not give accurate predictions of the biaxial yield stress, especially
in the case of aluminium alloys exhibiting a pronounced anisotropy
• the model cannot capture simultaneously the planar variation of the uniaxial yield
stress and uniaxial coefficient of plastic anisotropy
• the model does not give accurate predictions of the biaxial coefficient of plastic
anisotropy in case of highly-anisotropic materials.

Despite its limitations, the Barlat 1989 yield criterion is still frequently used in
the numerical simulation of sheet metal forming processes.
In 1991, Barlat proposed a 3D extension of his yield criterion [10] (see the
next section). Banabic et al. [51–53] also proposed extensions of the Barlat 1989
yield criterion (with seven and eight coefficients), aiming to remove the disadvan-
tages mentioned above (a detailed description of these formulations can be found in
Sect. 2.4.2).

2.3.2.3 Barlat 1991 Yield Criterion

Barlat [10] proposed a general six-component yield criterion (σ 11 , σ 22 , σ 33 , σ 12 ,
σ 23 , σ 31 ) that could be adopted with no restrictions to any stress state. With this
aim in view, the isotropic Hershey criterion [19] is rewritten in a form containing
the deviator principal stresses S1 , S2 and S3 :

 = |S1 − S2 |m + |S2 − S3 |m + |S3 − S1 |m = 2 · σem , (2.120)

After a complex number transformation and the Bishop-Hill notation (see [10]),

A = σ22 −σ33 , B = σ33 −σ11 , C = σ11 −σ22 , F = σ12 , G = σ31 , H = σ12 , (2.121)

Barlat obtained the following expression of the isotropic Hershey yield criterion:

2θ + π ))m )) 2θ + 3π ))m
)  ) )  )
m)
 = (3I2 ) )2 cos
2)
) + )2 cos ) +
6 6
(2.122)
2θ + 5π ))m
)  ) 
) m
+ )−2 cos
)
) = 2σe ,
6
where
' (
I3
θ = arccos 3/2
, (2.123)
I2

where I2 and I3 are the second and third invariant of the stress determinant,
respectively:

F 2 + G2 + H 2 (A − C)2 + (C − B)2 + (B − A)2

I2 = + , (2.124)
3 54
2.3 Classical Yield Criteria for Anisotropic Materials 67

(C − B) · (A − C) · (B − A)
I3 = + F · G · H−
54
(2.125)
(C − B) · F 2 + (A − C) · G2 + (B − A) · H 2
−
6

The yield function defined above was generalised to the anisotropic case by
using weighting coefficients (a, b, c, f, g, h), multiplying the stress components
given by Eq. (2.121). After this modification, the expressions of the yield func-
tion φ (Eq. 2.122) and angle θ (Eq. 2.123) remain the same. The expression of the
invariants I2 and I3 become:

(fF)2 + (gG)2 + (hH)2 (aA − cC)2 + (cC − bB)2 + (bB − aA)2

I2 = + (2.126)
3 54
(cC − bB) · (aA − cC) · (bB − aA)
I3 = + (fF) · (gG) · (hH) −
54
(2.127)
(cC − bB) · (fF)2 + (aA − cC) · (gG)2 + (bB − aA) · (hH)2
−
6

The weight factors a, b, c, f, g, h describe the anisotropy of the material. They

are identified by three uniaxial tensile tests in the directions of the orthotropic axes
(a, b, c) and three pure shearing tests (f, g, h).
For the case of plane stress, the number of coefficients diminishes to four (a, b,
c and h). They may be established by three uniaxial tensile tests (at 0◦ , 45◦ and 90◦
with the rolling directions) and a uniaxial compressive test through the thickness.
Alternatively, the identification of the coefficients may be performed using the
uniaxial yield stresses (σ 0 , σ 45 , σ 90 ) and the equibiaxial yield stress (σ b ). By intro-
ducing these values in the expression of the yield function given by Eq. (2.120),
one gets a set of four non-linear equations that can be solved numerically. Another
identification method makes use of the anisotropy coefficients r0 , r45 and r90 . In this
case, the value of one coefficient is assumed as being known. The rest of three is
obtained by numerically solving a set of non-linear equations.
The exponent m is above unity and is chosen in the same manner as in the case
of Hershey’s yield criterion. For a = b = c = f = g = h = 1 the criterion given by
Eq. (2.122) reduces to the isotropic Hershey expression (Eq. 2.120).
The methodology used for obtaining the associative flow rules is rather difficult.
It is fully detailed in [41] and [54].
More recently, Choi et al. [54] used the transformation proposed by Karafillis and
Boyce [21] for his criterion. As will be shown below, the Barlat 1991 yield criterion
may be considered as a special case of the Karafillis-Boyce yield criterion (c = 0).
The Barlat 1991 criterion has the advantage of being general and flexible (like the
Hill 1948 criterion it has six stress components); it predicts yield surfaces in good
agreement with those calculated from polycristal theories (Taylor and Bishop-Hill);
68 2 Plastic Behaviour of Sheet Metal

it is easy to implement in finite-element codes; it gives a realistic estimation of the

distribution of the uniaxial yield stress and coefficient r in the plane of the sheet.
The main disadvantage of the yield criterion is the complicated flow rule which
is not user-friendly.
A similar methodology was used by Lian and Chen [55] for generalising the Hill
1979 criterion in the case of the three-dimensional stress state. They established the
following six-component expression of the Hill 1979 criterion:

m/2
2m J2 f |sin (ϕ/3)|m + g |cos (ϕ/3 + π/6)|m + h |cos (ϕ/3 − π/6)|m +

+ 3m/2 a |cos (ϕ/3)|m + b |cos (ϕ/3 − π/3)|m + c |cos (ϕ/3 + π/3)|m = σem
(2.128)
where ϕ depends on the second and third invariant of the deviatoric stress tensor.
The a, b, c, f, g and h are the material coefficients, like in the Barlat model.
The methodology proposed by Lian and Chen is very general and can be applied
to any yield criterion written by using principal stresses in order to obtain a six-
component expression.

2.3.2.4 Yield Criteria by Barlat 1994 and 1996

Experimental studies showed that aluminum alloys have a plastic behavior that is
very difficult to model with the criteria above. At the beginning of 1990 several
researchers have focused their interest on this problem.
In [56] Barlat and co-workers proposed a more general expression of the yield
function introduced by himself in 1991. The generalization consisted in giving
weight factors to the terms of Eq. (2.120):
)m )m
 = αx )Sy − Sz ) + αy |Sz − Sx |m + αz )Sx − Sy ) = 2σem
) )
(2.129)

The weight factors α x , α y and α z are related to the anisotropy of the materials; Sx ,
Sy , and Sz are normal component of the stress tensor modified with the linear trans-
formation like in the Karafillis-Boyce proposal [21]; m has not the same significance
as in case of the Barlat 1991 criterion.
Assuming the shear stresses to be zero, the linear-transformation operator from
Karafillis-Boyce proposal (see [21]) becomes
⎡ c2 + c3 c3 c2 ⎤
− −
⎢ 3 3 3 ⎥
⎢ c3 c3 + c1 c1 ⎥
L=⎢ − ⎢ − ⎥ (2.130)
3 3 3 ⎥
c2 c2 + c3 c1 + c2
⎣ ⎦
−
3 3 3
where c1 , c2 and c3 are material coefficients describing the anisotropy of the
material.
The generalization of this function in order to include the shear stresses
(six-component formulation) as obtained by rewriting Eq. (2.130) in the principal
stresses:
2.3 Classical Yield Criteria for Anisotropic Materials 69

 = α1 |S2 − S2 |m + α2 |S3 − S1 |m + α3 |S1 − S2 |m = 2σem (2.131)

where the coefficients α 1 , α 2 and α 3 are computed from α x , α y and α z by the

transformation:

αk = αx p21 k + αy p22 k + αz p23 k , (2.132)

The coefficients p1 k , p2 k , p3 k are components of a rotation matrix relating the

anisotropy axes to the principal directions of the stress tensor S.
In the case of plane stress six independent coefficients must be determined. As a
consequence, four mechanical tests have to be made (three uniaxial tensile tests for
establishing σ 0 , σ 45 , σ 90 , r0 and r90 and one biaxial tensile test for establishing σ b ).
The yield surfaces predicted by this criterion are good approximations of those
calculated by the Bishop-Hill theory and of experimental data. However, finite-
element simulations based on this criterion [57] revealed some inaccuracies in
predicting blank earing in deep-drawing.
In order to improve the performance of his criterion, Barlat and co-workers [58]
proposed a generalisation of the 1994 criterion whereby α x , α y and α z are not con-
stant anymore. They depend on the angles β1 , β2 , β3 between the principal directions
of the stress tensor and the anisotropic axes:

αx = αx0 cos2 2β1 + αx1 cos2 2β1 ;

αy = αy0 cos2 2β2 + αy1 cos2 2β2 ; (2.133)
αz = αz0 cos2 2β3 + αz1 cos2 2β3 .

where

αi0 = αi for βi = 0;
(2.134)
αi1 = αi for βi = π/2.

and

y · 1 if |S1 | ≥ |S3 | or
*
cos2 2β1 =
y · 3 if |S1 | < |S3 | ;
z · 1 if |S1 | ≥ |S3 | or
*
cos2 2β2 = (2.135)
z · 3 if |S1 | < |S3 | ;
*
x · 1 if |S1 | ≥ |S3 | or
cos2 2β3 =
x · 3 if |S1 | < |S3 | .

For plane stress, after applying the linear transformation introduced by Karafillis
and Boyce (see [21]) the components of the IPE (isotropic plastic equivalent) stress
tensor are written
70 2 Plastic Behaviour of Sheet Metal

c3 (σ11 − σ22 ) + c2 σ11

S11 = ;
3
c1 σ22 + c3 (σ11 − σ22 )
S22 = ; (2.136)
3
−c2 σ11 − c1 σ22
S33 = ;
3
S12 = c6 σ12 .

where c1 , c2 , c3 and c6 are material coefficients defining the anisotropy. The

principal deviatoric stresses are computed as follows:
%
S11 + S22 2
 
S11 + S22 2 ,
S1,2 = ± + S12 (2.137)
2 2
S3 = −S1 − S2 = −S11 − S22 .

In this case the coefficients α 1 , α 2 and α 3 are

α1 = αx cos2 2θ + αx cos2 2θ ;
α2 = αy cos2 2θ + αy cos2 2θ ; (2.138)
α3 = αz0 cos2 2θ + αz1 cos2 2θ.

whereby αz0 = 1 and

 
−1 S11 − S22
θ = tan . (2.139)
S12

The orthotropic axes 1, 2 and 3 are oriented along the rolling, transverse and
normal direction, respectively. The methodology used for establishing the yield
function is the same as that used by Karafillis and Boyce (see [21]).
Equations (2.136) and (2.138) show that in order to determine the yield function
eight parameters are necessary: c1 , c2 , c3 , c6 , α x , α y and α 0 and the exponent set in
accordance with the crystallographic structure of the material. The great number of
parameters ensures a good flexibility of the criterion but implies a large number of
mechanical tests.
Simulations of deep-drawing of cylindrical cups using the new criterion [59]
revealed a very good agreement of the predicted earing with experimental data. The
computed yield surfaces are also in good agreement with those predicted by the
Bishop-Hill theory and experiments. A very good agreement between theory and
experiment has also been found for the distribution of the uniaxial yield stresses and
anisotropy coefficients in the plane of sheet.
The most important disadvantages of these models are as follows:

• the convexity of the yield functions is not guaranteed

• the derivatives of the equivalent stress are difficult to obtain analytically and,
consequently, the usability of the models in the numerical simulation codes is
reduced.
2.3 Classical Yield Criteria for Anisotropic Materials 71

• the application for full stress states leads to numerical problems due to the
complexity of the yield functions
• the CPU time is considerably larger than in the case of simpler models. This is a
major drawback in the numerical simulation of sheet metal forming processes.

2.3.2.5 Karafillis–Boyce Yield Criterion

As mentioned in Sect. 2.2.4 Karafillis and Boyce [21] proposed a very general yield
criterion. Its originality is given by the expression of the yield function (a weighted
combination between the von Mises and Tresca expression) as well as the use of a
linear transformation in order to pass from the isotropic to the anisotropic case.
The proposed isotropic yield function is presented in the Sect. 2.2.4. In order to
adopt the yield criterion for anisotropic materials, Karafillis and Boyce proposed an
original method based on the linear transformation:

S̃ = L̃ · σ̃ , (2.140)

where S̃ is a deviatoric stress tensor associated to an ‘isotropic plastic equivalent’

(IPE) state, σ̃ is the actual anisotropic stress tensor; L̃ is a linear operator depending
on the material which is defined as follows:
⎡ ⎤
1 β1 β2 0 0 0
⎢ β1 α1 β3 0 0 0 ⎥
⎢ ⎥
⎢ β2 β3 α2 0 0 0 ⎥
L̃ = C ⎢
⎢ 0 0 0 γ1 0 0 ⎥
⎥ (2.141)
⎢ ⎥
⎣ 0 0 0 0 γ2 0 ⎦
0 0 0 0 0 γ3

where

α2 − α1 − 1
β1 = ,
2
α1 − α2 − 1
β2 = , (2.142)
2
1 − α1 − α2
β3 = ,
2
and α 1 , α 2 , γ 1 , γ 2 , γ 3 and C are parameters defining the anisotropy of the metallic
material.
Thus the Karafillis-Boyce yield function is defined be eight coefficients (α 1 , α 2 ,
γ 1 , γ 2 , γ 3 , c, C and k).
In case of an isotropic material the parameters have the values

2 3
c= , α1 = α2 = 1, γ1 = γ2 = γ3 = (2.143)
3 2
72 2 Plastic Behaviour of Sheet Metal

For loading with negligible non-planar shear stresses (σ 32 = σ 13 ≈ 0) it is

obtained γ 2 = γ 3 = 3/2.
Equation (2.140) may be written in the form
⎡ i ⎤
S11
⎡ ⎤ ⎡ a ⎤
1 β1 β2 0 0 0 σ11
⎢ Si ⎥ ⎢ β1 α1 β3 0 0 0⎥ ⎢σa ⎥
⎢ 22 ⎥ ⎢ 22
⎢ Si ⎥ ⎢ a ⎥
⎥ ⎢ ⎥
⎢ 33
⎢ β2 β3 α2 0 0 0⎥ ⎥ · ⎢ σ33
⎢ Si ⎥ = C · ⎢ 0 ⎢ a ⎥ (2.144)
⎥ ⎢ ⎥
⎢ 23 ⎥ 0 0 γ1 0 0⎥ ⎥ ⎢ σ23a ⎦
⎢ ⎥
⎣ Si ⎦ ⎣0 0 0 0 γ2 0 ⎦ ⎣ σ31
31
i 0 0 0 0 0 γ3 a
S12 σ12

The ‘i’ and ‘a’ superscripts specify the ‘isotropic’ and ‘anisotropic’ state,
respectively. In case of a plane-stress state, Eq. (2.144) becomes:
⎡ i
⎤ ⎡ α2 −α1 −1
⎤ ⎡ a ⎤
S11 1 0 2 σ11
⎣ Si ⎦ = C · ⎣ α2 −α1 −1 a
22 2 0 ⎦ · ⎣ σ22 ⎦
α1 (2.145)
i a
S12 0 0 γ3 σ 12
and only six parameters needed for defining the yield surface (α 1 , α 2 , γ 3 , c, C
and k), one more than for the Barlat 1991 criterion. Therefore the Karafillis-Boyce
criterion is more flexible than Barlat 1991.
a
The linear transformation of an anisotropic stress state S̃ to an equivalent iso-
i
tropic one, S̃ , has been called by Karafillis and Boyce as ‘Isotropic Plasticity
Equivalent’ (IPE). A similar transformation, although not in the same form, was
used by Barlat in 1991 [10] in order to change the isotropic Hosford criterion into a
six-component anisotropic one.
The methodology used to establish the Karafillis-Boyce yield function for plane
stress is as follows:
1◦ Let σ 11 , σ 22 and σ 12 be the planar components of the anisotropic stress tensor
2◦ By using the linear transformation (2.145) the components of the IPE deviatoric
stress tensor are obtained:
 
S11 = C σ11 + α2 −α2 1 −1 σ22 ;
 
S22 = C α2 −α1 −1 σ11 + α1 σ22 ; (2.146)
2
S12 = γ3 · σ12 .

3◦ The principal deviatoric stresses are calculated:

" 2
S11 +S22 S11 −S22 2 ;
S1 = 2 + 2 + S12
"
S11 +S22 S11 −S22
2
2 ;
(2.147)
S2 = 2 − 2 + S12
S3 = − (S1 + S2 ) .
2.3 Classical Yield Criteria for Anisotropic Materials 73

4◦ Using Eqs. (2.146) the principal deviatoric stress tensor σ is obtained. By

inserting into Eq. (2.44), an expression of the yield function in terms of the
components of the anisotropic stress tensor and the coefficients α 1 , α 2 , γ 3 , c, C
and k is obtained:

 =  (σ11 , σ22 , σ12 , c, C, α1 , α2 , γ3 , k, σe ) (2.148)

This methodology is used in order to establish the yield function for the Barlat
1994 and 1996 yield criteria (presented above).
Karafillis and Boyce applied the inverse transformation

a i
D̃ = L̃ · D̃ (2.149)

in order to determinate the associated flow rules in the anisotropic state as functions
a
to the isotropic ones. Here D̃ is the anisotropic strain-rate tensor (for the begin-
i
ning of plastic yielding) while D̃ is the same tensor associated to the IPE material.
i
D̃ may be calculated from the associated flow rule assuming the yield function
Φ (Eq. 2.44):

i ∂Φ
D̃ = λ (2.150)
∂S
The transformation (2.145) is used for identifying the anisotropy coefficients r0 ,
r45 and r90 . The numerical procedure used for the inverse determination of these
coefficients in the operator L (Eq. 2.141) is presented in [21] as a flowchart.
Bron and Besson [60] have proposed a very general model that extends both
Barlat 1991 [10] and Karafillis–Boyce [21] formulations.
The equivalent stress of the new yield function is defined in the folowing form:

' K (1/a
+
k k a
σ = α (σ ) (2.151)
k=1

where the K functions σ k are convex, positive and homogeneous of degree 1 and α k
are positive coefficients (the sum of which is 1).
In their original paper [60], Bron and Besson use only two functions (K = 2). In
this case, the general formulation of the criterion (2.151) reduces to
k
&
σ k = (ψ k )1 b (2.152)

1 )) 1 )b1 ) )b1 ) ) b1 
1 1) ) 1 1) ) 1 1)
 = )S2 − S3 ) + )S3 − S1 ) + )S1 − S2 ) (2.153)
2
2
3b
) ) 2 ) ) 2 ) ) 2 
) 2 )b ) )b ) )b
2 = 2 )S1 ) + )S22 ) + )S32 ) (2.154)
2b + 2
74 2 Plastic Behaviour of Sheet Metal

where Sk i=1–3 are the principal values of the stress deviator Sk . The stress deviator
is determined by a linear transformation Lk defined in the paper [60].
In the formulation proposed by Bron and Besson, a total number of 16 parameters
are involved. Due to this fact, the model is very flexible. The convexity of the yield
function has been proved.
A similar model with the Karafillis and Boyce one has been developed for
anisotropic modelling of the polimeric foams by Wang and Pan [61]. An extra
parameter is used to model the different yield behaviours under tension and
compression.
The yield surfaces predicted by Karafillis-Boyce criterion are in very good agree-
ment with experimental data as well as with the predictions of the Bishop-Hill theory
[21]. The same agreement is also obtained when comparing the variation of the
uniaxial yield stress and anisotropy coefficients in the plane of the sheet with exper-
imental data [21]. Another advantage of the criterion is that it uses only uniaxial
tensile tests for identifying the material parameters. From a mathematical point of
view the method proposed by Karafillis and Boyce is both elegant and rigorous.
A disadvantage of the criterion is that the identification procedure of the ten-
sor operator is complex and requires a numerical solution. But this is not a major
difficulty when implementing the yield criterion into an FE code.

2.3.3 Yield Criteria Expressed in Polar Coordinates

2.3.3.1 Budiansky Yield Criterion
Budiansky proposed a general yield criterion [62] in a form that seems to be
attractive for applications, especially when using it together with appropriate pla-
nar constitutive equations. Planar isotropy allows that any two-dimensional yield
criterion is expressed as a function of the stress points (σ2 + σ1 ) and (σ2 − σ1 ).
Budiansky’s criterion can be written in the form

σ1 + σ2
x= = g(α) cos α;
2σb (2.155)
σ2 − σ1
y≡ = g (α) sin α.
2σs

where g (α) > 0 is the radial coordinate of a point located on the yield surface, α
is the associated polar angle, σs is the yield stress in pure shear, and σb is the yield
stress in equibiaxial tension.
The problem that arises is to establish the function g(α). By using the ratios
X = σb /σu and Y = σb /σs as non-dimensional parameters characteristic of the
material Eq. (2.155) can be rewritten in the form:

σ2 + σ1
= Xg (α) cos α
2σu (2.156)
σ2 − σ1 Xg (α) sin α
=
2σu Y
2.3 Classical Yield Criteria for Anisotropic Materials 75

Hence,
 
−1 Y (σ2 − σ1 )
α = tan (2.157)
σ2 + σ1

The associated flow rule is fully detailed in [62].

Ferron [63] proposed a yield criterion more general that by Budiansky [62] since
it allows to include the case of planar anisotropy due to the dependence of g on two
parameters θ and α. The function g(θ , α) must satisfy the symmetry and convexity
conditions imposed to the function g(α) in Budiansky’s criterion and is determined
starting from the isotropic criterion by Drucker [16].
Ferron [63] yield criterion is defined as follows:

F (θ )m/6 + 2a sin θ cos2n−1 θ cos 2α+

 
m/6 −m
(1 − k) g (θ, α) = (2.158)
+6 sin2p θ cos2q 2α

Here a, m, n, p and q are material parameters. The identification of these

parameters is made by uniaxial tensile and shearing tests.
Among the parameters in Ferron’s yield function five are independent. This
ensures a great flexibility of the criterion.
The obtained yield surfaces are in good agreement with experiments. The yield
criterion can also model accurately the distribution of the uniaxial yield stress and
coefficient r in the plane of the sheet [63].

2.3.4 Other Yield Criteria

In addition to the criteria described in the previous sections, several other non-
quadratic yield criteria have been developed. With respect to their restrained use
they are only described briefly.

2.3.4.1 Gotoh Yield Criterion

In order to overcome the disadvantages of the Hill 1948 criterion, Gotoh proposed
in 1977 [64] a polynomial yield function of fourth degree (instead of the quadratic
one) for orthotropic rolled sheet metals by writing the yield function in the form
given by Hill in 1950 [65].

j
+
i 2k
f = Aijk σ11 σ22 σ12 (2.159)
i,j,k

where i, j, 2 k ≤ 4, x, y are the orthotropic axes, and Aijk are constant coefficients.
The conditions of orthotropy and wrinkling of the blank in axisymetric deep-
drawing necessitate to write for the function f
76 2 Plastic Behaviour of Sheet Metal

f = A0 (σ11 + σ22 )2 + A1 σ11

4 + A σ3 σ + A σ2 σ2 + A σ σ3 + A σ4 +
 2 11 22 3 11 22 4 11 22 5 22
2 + A σ σ + A σ2 σ2 + A σ4
A6 σ11 7 11 22 8 22 12 9 12
(2.160)

The first term may be considered as a function of the mean normal pressure and
thus, assuming an incompressible material, A0 is obtained. The condition to avoid
wrinkling leads to the equations [64]

cos α sin αF (c) = 0

c = cos2 α (2.161)
F (c) = 4Ac3 + 3Bc2 + 2Cc + D

A. . . D are coefficients depending on A1 ...A9 :

A = (A1 + A3 + A5 + A7 + A9 ) − (A2 + A4 + A6 + A8 )
B = (A2 + 3A4 + A6 + 3A8 ) − 2 (A3 + 2A5 + A7 + A9 )
(2.162)
C = (A3 + 6A5 + A7 + A9 ) − 3 (A4 + A8 )
D = A4 + A8 − 4A5

Since the identification of the coefficients A1 –A9 is made by means of an equi-

biaxial tensile test, this criterion can model the ‘anomalous behavior’ observed by
Woodthrope and Pearce. This criterion is presented in detail in [64].
The large number of coefficients will generally allow a high flexibility of the
yield criterion. The model captures the planar variation of both the uniaxial yield
stress and uniaxial coefficient of plastic anisotropy. The advantages previously men-
tioned above explain why the model is frequently used, especially by the Japanese
researchers.
A disadvantage of the criterion is its complicated form and the large number of
mechanical tests needed for identifying the coefficients.

2.4 Advanced Anisotropic Yield Criteria

During the last years, the competition in the automotive and aeronautical indus-
try has become more intense. This fact has lead to the development of new steel
alloys (Bake Hardenable, Dual Phase, Complex Phase, Transformation Induced
Plasticity-TRIP, Martensitic Steels, Hot-stamping boron-alloyed steels), aluminium
alloys having better performances and increased interest on the use of magnesium
and superplastic alloys. Since 2000, the modelling of the anisotropic behaviour of
these materials has encouraged the research activities focused on the development
of yield criteria. Several new models have been proposed during the last years.
These models allow a very good description of the anisotropic behaviour both of
steel alloys (BCC crystallographic structure), aluminium alloys (FCC structure) and
magnesium alloys (HCP structure). The new yield criteria incorporate a large num-
ber of coefficients (usually, at least 8 coefficients). Due to this fact, they are able to
2.4 Advanced Anisotropic Yield Criteria 77

give an accurate description of the yield surface and follow closely the planar varia-
tions of the uniaxial yield stress and the coefficient of plastic anisotropy. Even more,
some of the recently developed models can also capture the non-symmetric response
in tension/compression specific to the HCP alloys. Due to the significant impact of
these advanced yield criteria, they will be described in a separate subchapter entitled
‘Advanced Anisotropic Yield Criteria’.

2.4.1 Barlat Yield Criteria

In order to remove the disadvantages of the Barlat 1994 and Barlat 1996 yield crite-
ria (see Sect. 2.3.2), but aiming to preserve their flexibility, Barlat proposed in 2000
[5, 66] a new model particularized for plane stress (2D).
Let us consider a liniear transformation defined as follows:

X=C·s (2.163)

where s is the deviatoric stress tensor and X the linearly transformed stress tensor.
This gives 9 independent coefficients for the general case and 7 for plane stress.
However, applied to plane stress conditions, only one coefficient is available to
account for σ 45 and r45 . As pointed out in Barlat et al. [5] additional coefficients in
the context of linear transformations can be obtained by using two transformations
associated to two different isotropic yield functions, respectively.
As a consequence, Barlat et al. [5] proposed a yield function expressed by the
relationship

 = ′ + ′′ = 2σ a , (2.164)

where

′ = |S1 - S2 |a (2.165)

′′ = |2S2 + S1 |a + |2S2 + S2 |a (2.166)

S1 , and S2 are the principal deviatoric stresses and ‘a’ is an exponent determined
based on the crystallographic structure of the material.
By applying a linear transformation to each of the isotropic functions defined by
Eqs. (2.165) and (2.166), we obtain the yield function

 = ′ X ′ + ′′ X ′′ = 2σ a
   
(2.167)

where σ is the effective stress, ‘a’ is a material coefficient and

)a
′ = )X1′ + X2′ )
)
(2.168)
)a ) )a
′′ = )2X2′′ + X1′′ ) + )2X1′′ + X2′′ )
)
(2.169)
78 2 Plastic Behaviour of Sheet Metal

and

X′ = C′ .s = C′ .T.σ = L′ .σ
(2.170)
X′′ = C′′ .s = C′′ .T.σ = L′′ .σ

T is a matrix that transforms the Cauchy stress tensor σ to its deviator s:

⎡ & & ⎤
2 3 −1 3 0
& &
T = ⎣ −1 3 2 3 0⎦ (2.171)
0 0 1

′ ′
C and C being the linear transformations.
In the reference frame associated with the material symmetry,

′
⎡ ⎤ ⎡ ′ ′
⎤⎡ ⎤
X11 C11 C12 0 s11
⎣ X ′ ⎦ = ⎣ C′ C′ 0 ⎦ ⎣ s22 ⎦ (2.172)
22 21 22
′
X12 0 0 C66′ s12

and

′′
⎡ ⎤ ⎡ ′′ ′′ ⎤⎡ ⎤
X11 C11 C12 0 s11
⎣ X ′′ ⎦ = ⎣ C′′ C′′ 0 ⎦ ⎣ s22 ⎦ (2.173)
22 21 22
′′
X12 0 0 C66′′ s12

Because ′ depends on X1′ − X2′ , only three coefficients remain independent in

C′ (see more details in [5]). In C′′ are five independent coefficients. Totally, in both
transformations, are 8 independent coefficients.
The principal values X1 and X2 of there X′′ and X′′ are the follows:
 
1
$
X1 = 2 2
X11 + X22 + (X11 − X22 ) + 4X12 , (2.174)
2
 
1
$
X2 = 2
X11 + X22 − (X11 − X22 )2 + 4X12 . (2.175)
2

The coefficients of L′ and L′′ are expressed as follows

⎡ ′
⎤ ⎡ & ⎤
L11 2 &3 0 0
⎢ L′ ⎥ ⎢ −1 3
⎡ ⎤
⎢ 12 0& 0⎥⎥ α1
⎢ L′ ⎥ = ⎢ 0
⎥ ⎢
⎢ 21 −1 3 0⎥⎥ α2
⎣ ⎦ (2.176)
⎣ L′ ⎦ ⎣ 0
⎥ ⎢ &
22 2 3 0 ⎦ α7
′
L66 0 0 1
2.4 Advanced Anisotropic Yield Criteria 79

′′
⎡ ⎤ ⎡ ⎤⎡ ⎤
L11 −2 2 8 −2 0 α3
⎢ L′′ ⎥ ⎢ 1 −4 −4 4 0 ⎥ ⎢ α4 ⎥
⎢ 12
⎢ L′′ ⎥ = 1 ⎢ 4 −4 −4 1 0 ⎥ ⎢ α5 ⎥
⎥ ⎢ ⎥⎢ ⎥
⎢ 21 (2.177)
⎣ L′′ ⎦ 9 ⎣ −2 8 2 −2 0 ⎦ ⎣ α6 ⎦
⎥ ⎢ ⎥⎢ ⎥
22
′′
L66 0 0 0 0 1 α8

Due to the fact that 8 coefficients are incorporated in the linear transformations,
we need 8 material characteristics for evaluating them. The uniaxial tension test
along the rolling, diagonal and transversal directions, together with the biaxial ten-
sion test can provide only 7 characteristics (3 uniaxial yield stresses, 3 coefficients
of uniaxial anisotropy and the biaxial yield stress). Barlat adopted the coefficient of
biaxial anisotropy rb as the eighth characteristic in the identification procedure. The
experimental procedure used for the determination of this mechanical parameter is
described in Sect. 2.1.2.
By using the same methodology as the one described above, Aretz and Barlat
[67] and Barlat et al. [68] proposed a 3D yield criterion called Barlat 2004-18p:
)a ) )a ) )a ) )a ) )a
 = )s)′1 − s′′1 ) ) + )s)′1 − s′′2 ) ) + )s)′1 − s′′3 ) ) + )s)′2 − s′′1 ) ) + )s′2 − s′′2 ) +
)
a a a a (2.178)
+ )s′2 − s′′3 ) + )s′3 − s′′1 ) + )s′3 − s′′2 ) + )s′3 − s′′3 ) = 4σ a ,

where, σ represent the uniaxial yield stress (any other yield stress may be use as
reference yield stress) and a is an exponent determined based on the crystallographic
structure of the material.
The associated linear transformation on the stress deviator is defined:
⎡ ⎤
0 −c12 −c13 0 0 0
⎢ −c21 0 −c23 0 0 0 ⎥
⎢ ⎥
⎢ −c31 −c32 0 0 0 0 ⎥
C=⎢ 0⎥ (2.179)
⎢ 0
⎢ 0 0 c44 0 0 ⎥ ⎥
⎣ 0 0 0 0 c55 0 ⎦
0 0 0 0 0 c66

and C′ and C′′ are obtained by adding prime and double prime symbols.
Each transformation provides 9 coefficients and totally both transformations give
18 coefficients. In order to determine all this coefficients an the minimization of the
error function method is used (see [51]). If only one linear transformation is assumed
the Barlat 2004-18p formulation reduce to Barlat 1991 yield criterion.
The uniaxial yield stresses and anisotropy coefficients in seven directions in
the plane of the sheets (0, 15, 30, 45, 60, 75 and 90 degree to the rolling
direction), the biaxial yield stress, the biaxial anisotropy coefficient and four addi-
tional data characterizing out-of-plane properties (two tensile and two simple shear
yield stresses) are used in the identification of all the coeficients. For determi-
nation of the out-of-plane parameters the crystal plasticity models are nedded
(see [68]).
80 2 Plastic Behaviour of Sheet Metal

If we adopt a yield function defined by the relationship

′′ )a ′′ )a ′′ )a
) ′ )a ) ′ )a ) ′ )a ) ′′ )a ) ′′ )a
 =)s′1 −
) ) )′ ) )′ )
) ′′s)2a + sa2 − s3 + s3 − s1 − { s1 + s2 + s3 } + s1 + s2
) ) ) ) ) ) ) ) ) ) ) )
+ s3 = 4σ ,
) )
(2.180)

the number of coefficients incorporated in the linear transformations will reduce to

13:

−1 −c′13
⎡ ⎤
0 0 0 0
⎢ −c′ 0 −c′23 0 0 0 ⎥
⎢ 21 ⎥
⎢ −1 −1 0 0 0 0 ⎥
C′ = ⎢ 0⎥ (2.181)
⎢ 0 0 0 c′44 0 0 ⎥
c′55
⎢ ⎥
⎣ 0 0 0 0 0 ⎦
0 0 0 0 0 c′66

0 −c′′12 −c′′13
⎡ ⎤
0 0 0
⎢ −c′′ 0 −c′′ 0 0 0 ⎥
⎢ 21 23 ⎥
′′
⎢ −1 −1 0 0 0 0 ⎥
C = (2.182)
c′′44
⎢ ⎥
⎢ 0 0 0 0 0 ⎥
c′′55
⎢ ⎥
⎣ 0 0 0 0 0 ⎦
0 0 0 0 0 c′′66

For the plane stress case the number of the coefficients are reduced from 13 to 9.
The yield function has been tested for different aluminium alloys exhibiting a
pronounced anisotropy. The model has proved its capability to provide an accurate
prediction of the planar variations of the uniaxial yield stress and coefficient of
plastic anisotropy.
The implementation of the Barlat 2004-18p model in finite-element codes [69]
proved its capability to predict the occurrence of six and eight ears in the pro-
cess of cup drawing. Barlat 2004-18p is one of the phenomenological model being
able to capture more than 4 ears. This is the most important advantage of the
yield criterion. Of course, it is possible to develop models incorporating more and
more linear transformations and thus having a larger number of coefficients. The
practical difficulty related to the use of such yield criteria consists in the experi-
mental determination of the mechanical parameters needed for the evaluation of the
coefficients.
The disadvantages of the models presented above are:

• due to the complexity of the formulation, they are not user-friendly

• they need crystal plasticity models for the evaluation of some parameters.

The Barlat 2004-18p is implemented in the LS Dyna commercial code.

2.4 Advanced Anisotropic Yield Criteria 81

2.4.2 Banabic–Balan–Comsa (BBC) Yield Criteria

In 2000 the members of the CERTETA1 team started a research programme hav-
ing as principal objective the development of a model able to provide an accurate
description of the yield surfaces predicted by texture computations. The new formu-
lation was developed on the basis of the isotropic formulation proposed by Hershey.
By adding weight coefficients to that model, the researchers succeeded in develop-
ing a flexible yield criterion. The last version incorporates a number of 8 coefficients
and, consequently, its identification procedure uses 8 mechanical parameters (3 uni-
axial yield stresses, 3 uniaxial coefficients of anisotropy, the biaxial yield stress and
the biaxial coefficient of plastic anisotropy).
The first formulation of the yield criterion was proposed by Banabic et al. [51] in
the form
 1
2k
σ = a (bŴ + c)2 k + a (bŴ − c)2 k + (1 − a) (2c)2 k (2.183)

where a, b, c, and k are material parameters, while Ŵ and  are functions of the
second and third invariants of a transformed stress tensor s′ = Lσ, where L is a 4th
order tensor. In this formulation anisotropy is described by means of the tensor L,
which satisfies: (i) the symmetry conditions Lijkl = Ljikl = Ljilk = Lklij (i, j, k,
l =1. . .3), (ii) the requirement of invariance with respect to the symmetry group of
the material, and (iii) the three conditions L1 k + L2 k + L3 k = 0 (for k = 1,
2, and 3), which ensures that s′ is traceless (see Karafillis–Boyce [21]). Hence, in
the reference system associated with the directions of orthotropy, the tensor L has 6
non-zero components for 3D conditions and 4 components for plane stress state.
Let define (1, 2, 3), the reference frame associated with orthotropy. For a rolled
sheet, 1, 2, and 3 represent the rolling direction, the long transverse direction, and
the short transverse direction, respectively. In the reference system (1, 2, 3):

s′11 = dσ11 + eσ22

s′22 = eσ11 + f σ22
s′33 = − (d + e) σ11 − (e + f ) σ22 (2.184)
s′12 = gσ12
s′13 = s′23 = 0

where d, e, f, and g are the four independent components of the tensor L.

The expressions of Ŵ and  in terms of the stress components are

Ŵ = Mσ
$ 11 + Nσ22 (2.185)
 = (Pσ11 + Qσ22 )2 + Rσ12
2

1Research Centre in Sheet Metal Forming Technology belong the Technical University of Cluj
Napoca, Romania (http://www.certeta.utcluj.ro).
82 2 Plastic Behaviour of Sheet Metal

where

M =d+e
N =e+f
d−e
P= (2.186)
2
e−f
Q=
2
R = g2

(for more details see Banabic et al. [52]).

The convexity of the yield surface described by (2.183) is ensured if a ∈ [0, 1]
and k is a strictly positive integer number.
The yield stress in uniaxial tension along an axis at orientation θ to the rolling
direction x, the equibiaxial yield stress and the coefficient of plastic anisotropy asso-
ciated to a direction inclined at an angle θ ∈ [0, 90◦ ] with the rolling direction are
presented in the paper [70]. Further details related to the evaluation of the coeffi-
cients as well as to the experimental validation of the BBC2000 yield criterion can
be found in the paper [70].
The shape of the yield surface is defined by the material parameters a, b, c, d, e,
f, g, and k. Among these parameters, k has a distinct status. More precisely, its value
is set in accordance with the crystallographic structure of the material [45]: k = 3
for BCC alloys, and k = 4 for FCC alloys. The other 7 parameters are determined
such that the model reproduces as well as possible the following experimental char-
acteristics of the orthotropic sheet metal: σ0 exp , σ90 exp , σ45 exp , σb exp , r0 exp , r90 exp
and r45 exp . It is possible to obtain their values by solving a set of seven non-linear
equations.
An improvement of this criterion was proposed by Banabic et al. [53] in order
to account for an additional mechanical parameter, namely, the biaxial anisotropy
coefficient. The new formulation is given by:
 1
2k
σ = a (Ŵ + )2 k + a (Ŵ − )2 k + (1 − a) (2)2 k (2.187)

where k ∈ N ≥ 1 and 0 ≤ a ≤ 1are material parameters, while Ŵ,  and  are

functions depending on the planar components of the stress tensor:

Ŵ = σ"11 + 2M σ 22
 2
 = Nσ 11 2- P σ 22 + Q2 σ12 σ21 (2.188)
" 2
 = R σ 11 2- S σ 22 + T2 σ12 σ21

where the quantities M, N, P, Q, R, S and T are also material parameters.

More details converning the uniaxial yield stress and anisotropy coeffi-
cients variations are presented in the paper [53]. Further details related to the
evaluation of the quantities mentioned above as well as to the convergence of
2.4 Advanced Anisotropic Yield Criteria 83

Table 2.1 Different strategies to identify the coefficients in the BBC2003 yield function
Mechanical
parameters BBC2003-8 BBC2003-7 BBC2003-6 BBC2003-5 BBC2003-4 BBC2003-2

σ0
σ45
σ90
σb
r0
r45
r90
rb

the identification procedure of the BBC2003 yield criterion can be found in the
paper [71].
The other eight parameters are determined such that the model reproduces the
experimental characteristics of the orthotropic sheet metal as well as possible,
namely, σ 0 , σ 45 , σ 90 , σ b , r0 , r45 , r90 and rb . It is possible to obtain the value of
these parameters by solving a set of eight non-linear equations. However, this set
of equations has multiple solutions. A more effective strategy of identification is to
impose the minimization of the following error function:
 p
2  p
2  p
2  p
2
r0 r45 r90 rb
F (a, M, N, P, Q, R, S, T) = r0 −1 + r45 −1 + r90 −1 + −1 rb
 p
2  p
2  p
2  p 2
σ0 σ45 σ90 σ
+ σ0 −1 + σ45 −1 + σ90 − 1 + σb − 1
b
(2.189)
where the superscript (.) pdenotes the values predicted by the constitutive equation.
For the numerical minimization, the downhill simplex method proposed by
Nelder and Mead [72] has been adopted because it does not need the evaluation
of the gradients. The identification procedure can also use a reduced number of
mechanical parameters (2, 4, 5, 6 or 7), as shown in Table 2.1. The particular set
of mechanical parameters used by each identification strategy is specified in the
table. The author have also developed identification procedures based on uniaxial
and plane-strain experimental data [71].
A version of the BBC 2003 yield criterion has been improved by Aretz [73].
The BBC2003 yield criterion is reducible both to Hill 1948 and Barlat 1989
formulations (see more details in Sect. 2.3).
Barlat et al. [74] showed that the BBC 2003 and Barlat 2000 are the same. But
one should notice that the development procedures adopted by the authors where
different: the BBC models emerged in a classical manner by adding coefficients to
Hershey’s formulation, while Barlat 2000 used two linear transformations.
The most important advantages of these models are:

• the yield functions have simple expressions

• they are able to describe accurately the yield surface and also give good
predictions of the planar distribution of the uniaxial yield stress and uniaxial
coefficient of plastic anisotropy
84 2 Plastic Behaviour of Sheet Metal

• the predicted shape of the yield surface closely follows the results of the texture
models
• the CPU time needed for the simulation of complex sheet metal forming
processes is not considerably increased
• the models can be used also in the cases when less then 8 mechanical parameters
are available (e.g., 2, 4, 5, 6 or 7 parameters)
• the models are reducible to classical formulations such as Hill 1948 or Barlat
1989.

The following disadvantages can be mentioned:

• the formulation of the yield criterion is not very user-friendly

• the coefficients of the yield function do not have a direct physical meaning
• the development of 3D extensions is difficult.

A modified version of this criterion (BBC 2005) has been implemented in the
AutoForm 4.1 commercial Finite Element program (issued May 2007).

2.4.3 Cazacu–Barlat Yield Criteria

To introduce orthotropy in the expression of an isotropic criterion, Cazacu and

Barlat [75] proposed an alternative method based of the theory of the representation
of tensor functions. They developed a method of generalizations of the invariants
of the stress deviator J2 and J3 . Based on this method, an anisotropic yield crite-
rion is obtained by substituting the expression of the stress deviator invariants in the
isotropic criterion by their respective anisotropic forms.
The generalized forms of the invariants, J30 and J20 , respectively, are:

1 3 1 3 1 3
J3o = (b1 + b2 ) σ11 + (b3 +b4 ) σ22 + [2 (b1 + b4 )−b2 − b3 ] σ33
27 27 27
1 2 1 2
− (b1 σ22 + b2 σ33 ) σ11 − (b3 σ33 + b4 σ11 ) σ22
9 9
1 2
− [(b1 − b2 + b4 ) σ11 + (b1 − b3 + b4 ) σ22 ] σ33
9
2 σ2
+ (b1 + b4 ) σ11 σ22 σ33 − xz [2 b9 σ22 − b8 σ33 − (2b9 − b8 ) σ11 ]
9 3
σ122
− [2 b10 σ33 − b5 σ22 − (2b10 − b5 ) σ11 ]
3
σ2
− 23 [(b6 + b7 ) σ11 − b6 σ22 − b7 σ33 ] + 2 b11 σ12 σ13 σ23 .
3
(2.190)
2.4 Advanced Anisotropic Yield Criteria 85

where the coefficients bk ( k = 1. . .11) describe the anisotropy and they reduce to
unity for isotropic conditions.

a1 a2 a3
J2o = (σ11 − σ22 )2 + (σ22 − σ33 )2 + (σ11 − σ33 )2 + a4 σ12 2
6 6 6
2 2
+ a5 σ13 + a6 σ23
(2.191)
where the coefficients ak ( k = 1. . .6) describe the anisotropy and they reduce to
unity in the isotropic case. Note that J20 is Hill’s [25] quadratic yield function.
In Cazacu and Barlat [75], this approach was used to extend Drucker’s [16]
isotropic yield criterion to an orthotropic one. For this case the expression of the
proposed orthotropic criterion is:
 3  2
f O = J2o − c J3o = k2 . (2.192)

where c is a constant,
 6
2 Y
k = 18 (2.193)
3

and Y is the uniaxial limit stress.

For the in-plane case the yield function may be writen in the form:
3
f2O ≡ 1 2 − a1 σ σ
(a1 + a3 ) σ11 1 2 2
6 3 11 22 + 6 (a1 + a2 ) σ22 + a4 σ12
⎧ 1 3 + 1 3
⎫2
⎨ 27 (b1 + b2 ) σ11 27 (b3 + b4 ) σ22 ⎬ (2.194)
− c − 91 (b1 σ11 + b4 σ22 ) σ11 σ22 = k2 .
⎩ 1 2
− 3 σ12 [(b5 − 2 b10 ) σ11 − b5 σ22 ]
⎭

where a1 −a4 and b1 −b5 and b10 are coefficients describing the anisotropy, c is a
constant and k is expressed by Eq. (2.193).
As one may see, the yield function incorporates 10 anisotropy coefficients and
an extra constant c. The 10 anisotropy coefficients and the value of c can be deter-
mined from the measured uniaxial yield stresses σθ and strain ratios rθ in 5 different
orientations and σ b , the value of the equibiaxial tensile stress. In the 3D case, the
model incorporates 18 coefficients.
The yield stress in uniaxial tension along an axis at orientation θ to the rolling
direction is predicted by:
⎧ 3 ⎫−1/6
1 4 θ +(a − a /3) cos2 θ sin2 θ+ 1 (a + a )sin4 θ ⎪
⎪
⎪ (a 1 + a3 ) cos 4 1 1 2 ⎪
⎨ 6 6
⎪
⎪ ⎪
⎪
2
1 ⎡ 1 ⎬
6 1 6
⎤
σθ = k 3 27(b 1 +b 2 ) cos θ + 27 (b 3 +b 4 ) sin θ
− c ⎣ 1 (b1 + 3b5 −6 b10 ) cos2 θ+
⎪ 
⎦ ⎪
sin2 θ cos2 θ ⎪
⎪ ⎪
−
⎪ ⎪
9 + (b −3 b )sin2 θ
⎪
⎩ ⎭
4 5
(2.195)
86 2 Plastic Behaviour of Sheet Metal

and the biaxial yield stress by:

 3  2 − 16
1 a2 + a3 −2 b1 + b2 + b3 − 2b4
σb = k 3 − c (2.196)
6 27

Yielding under pure shear parallel to the orthotropic axes occurs when σxy is
equal to
1 1
τ = k 3 (a4 )− 2 (2.197)

In order to predict the distribution of the anisotropy coefficient rθ , the function

f20 defined by Eq. (2.194) should be replaced in the relationship

∂ fo ∂ fo ∂ fo
sin2 θ ∂σx − sin 2 θ ∂ σxy + cos2 θ ∂ σy
rθ = − ∂ fo ∂ fo
(2.198)
∂σx + ∂σy

Cazacu and Barlat [76] also applied the representation theorems for transverse
isotropy and cubic symmetries. The general expressions of the invariants of the
stress deviators in these conditions are presented in detail in [75]. The method is
applied for the extension of Drucker’s isotropic yield criterion to transverse isotropy
and cubic symmetries.
Aiming to develop models of the asymmetrical tension/compression behaviour
specific to the alloys having a Hexagonal Closed Packed-HCP structure, Cazacu and
Barlat have successfully used the representation theory of tensor functions. They
have proposed an isotropic yield function in the form [77]:

f = (J2 )3/2 − cJ3 = τY3 (2.199)

where τ Y is the yield stress in pure shear and c a constant. This constant can be
expressed in the terms of the uniaxial yield stresses in tension σ T and compression
σ C , respectively, as follow:
√
3 3(σT3 − σC3 )
c= . (2.200)
2(σT2 + σC2 )

Anisotropy was introduced in the formulation using the same method presented
above.
For plane stress conditions, the yield locus is:
 
1 2 3/2 c
2
σ1 − σ1 σ2 + σ2 − [2σ13 + σ23 − 3(σ1 + σ2 )σ1 σ2 ] = τY3 , (2.201)
3 27

where σ 1 and σ 2 are the principal stresses.

2.4 Advanced Anisotropic Yield Criteria 87

The expressions of the anisotropic yield function and of the uniaxial yield
stresses in tension and compression along an axis at orientation θ to the rolling
direction direction are presented in [77]. The predictions of the biaxial yield stresses
corresponding to the tension and compression, as well as the planar distribution of
the anisotropy coefficient are also presented.
The yield function defined by Eq. (2.190) is a third-order expression. The exper-
imental researches [78] have shown that for some HCP alloys (e.g., titanium based
alloys) the yield surface is better described by fourth order functions. As a con-
sequence, in order to describe such a behaviour, Cazacu et al. [79] proposed an
isotropic yield function for which the degree of homogeneity a is not fixed:
) )a ) )a ) )a
 = )|S1 | − kS1 ) + )|S2 | − kS2 ) + )|S3 | − kS3 ) , (2.202)

where, S1 , S2 , S3 are the principal values of the stress deviator, a—an positive integer
and k—the strength differential parameter.
In order to extend the isotropic criterion defined by Eq. (2.202) to an anisotropic
formulation, the principal values of the deviatoric stress (S1 , S2 , S3 ) are replaced by
the principal values of the transformed tensor (Σ 1 , Σ 2 , Σ 3 ), obtained after applying
a linear transformation. In this way, the new anisotropic yield criterion (CPB05) can
be written as
) )a ) )a ) )a
 = )|1 | − k1 ) + )|2 | − k2 ) + )|3 | − k3 ) . (2.203)

The paper [79] gives a detailed presentation of the relationships used to predict
the uniaxial yield stresses and the coefficients of plastic anisotropy both for tension
and compression states. Additional linear transformations can be incorporated into
the CPB 2005 criterion for an improved representation of the anisotropy.
The most important advantage of this yield criterion consists in its capability to
provide an accurate description of the tension/compression behaviour specific to the
magnesium and titanium alloys.

2.4.4 Vegter Yield Criterion

Using points of the yield locus which had been determined directly by experiments,
Vegter [80, 81] obtained the yield locus in the first quadrant by applying a Bezier
interpolation. The Vegter criterion requires the determination of three parameters
for each reference point (two principal stresses σ 1 and σ 2 and the strain vector ρ =
dε 2 /dε1 ). In order to describe planar anisotropy Vegter’s criterion needs as many as
17 parameters.
The analytical expression of the criterion is
   r  h  r
σ1 σ1 σ1 σ1
= (1 − λ)2 + 2λ (1 − λ) + λ2 (2.204)
σ2 σ2 i
σ2 i
σ2 i+1
88 2 Plastic Behaviour of Sheet Metal

for σ e and angle ϕ where

 r m cos
' j
(r
σ1 + a1
= j cos (2jϕ) (2.205)
σ2 i+1 a2
j=0 i

is a trigonometric expansion associated to the reference point;

m
+ cos
R (ϕ) = bj cos (2jϕ) (2.206)
j=0

is cosine interpolation of the function R(ϕ); ϕ is the angle between the principal
directions and the orthotropic axes; λ is a parameter of the Bézier function; r is a
superscript denoting the reference point; h is a superscript denoting the breaking
' (
j r
a1
point; j are parameters of the trigonometric interpolation to be determined
a2 i
at the reference points; bj are parameters of the trigonometric interpolation of the
R-function.
The most important advantage of the criterion is the flexibility ensured by the
large number of parameters. Disadvantages are the unfriendly form of the yield
function making it improper for analytical computation; the large number of exper-
iments required (uniaxial tension, biaxial tension, plane strain and pure shearing)
and the necessity of mathematical abilities of the user.
The Vegter’s model has been implemented in the PAMSTAMP FE commercial
program.
Mollica and Srinivasa [82] proposed a method for generating the yield locus
similar to the one presented above. A simple way to obtain a closed convex surface
is to consider the intersection of a sufficient number of elementary convex surfaces.
Each elementary surface is defined by an equation having the form fi = 0. In order
to avoid the sharp corners and edges a special regularization procedure is proposed.
The method is illustrated in [82] for Hill 1948 [25] and Hosford [31] criteria.

2.4.5 Polynomial Yield Criteria

Hill [65] proposed in 1950 a general formulation of a plane-stress anisotropic yield
criterion having the polynomial expression defined by Eq. (2.159). Gotoh [64] suc-
ceeded to apply that idea in the 1970s by developing a polynomial yield function of
fourth degree (see Sect. 2.3.4).
During the last years, a new family of polynomial yield criteria has been created
on the basis of Hill’s idea.
2.4 Advanced Anisotropic Yield Criteria 89

2.4.5.1 Hu Yield Criteria

Hu [83] proposed a yield criterion in the form

1 4 4R0 1 1 1 4R90
f (σij ) = σ − σ3 σ +[ − − + 4 +
σ04 11 (1+R0 )σ04 11 22 σb4 σ04 4
σ90 (1+R90 )σ90

4R0 4R90 3 1 4
+ ]σ 2 σ 2 − 4 σ11 σ22 + 4 σ22 +
(1+R0 )σ04 11 22 (1+R90 )σ90 σ90

16R45
+[ (1+R16 )σ 4 − 2
σb4
2
](σ11 2 − σ σ )σ 2 + [
+ σ22 11 22 12
1
σb4
+ 4
4 ]σ12
(1+R45 )σ45
= 1.
45 45
(2.207)
The planar distributions of the uniaxial yield stress and of the anisotropy coef-
ficient are presented in [83]. Hu also succeeded to develop a 3D extension of his
criterion [84]. A quadratic formulation of the yield function has been presented by
Hu in [85].

2.4.5.2 Wang Yield Criterion

Wang (2005, Constitutive Modeling of Orthotropic Plasticity in Sheet Metals, pri-
vate communication) proposed a 2D polynomial formulation of the yield function
as:

m+1 β m+1 β m+1 γ m+1

σ11 m
− Fσ11 σ22 − G σ11 σ22 + σ22 + σ12 = σYm+1 (2.208)
α α α
where m is an odd integer 3, 5 or 7 (depending on the crystallographic structure), F,
G, β/α and γ /α are anisotropic coefficients and σ Y is the uniaxial yield stress in the
rolling direction.

2.4.5.3 Comsa Yield Criterion

Comsa also developed in his PhD thesis [86] a polynomial criterion of sixth order:

  1/ 6
p1 (L + M)(cL + M)(L + cM)+
= − Y, (2.209)
+p2 (L + N)(cL + N)(L + cN)

where,

L = (p3 σ11 − p4 σ22 )2 + σ12 σ21 ,

M = (p5 σ11 − p6 σ22 )2 + σ12 σ21 , (2.210)
N = (p7 σ11 − p8 σ22 )2 .
√
and c = (2+ 3)2 , p1 –p8 (p1 −p8 > 0) are the material parameters and Y is the yield
parameter. As one may see in the above relationship, the yield function incorporates
90 2 Plastic Behaviour of Sheet Metal

8 coefficients. These coefficients can be evaluated by using three uniaxial tensile

tests and a biaxial tensile experiment.
The convexity of the yield surface is proved in [86]. The relationships used
to evaluate the uniaxial/biaxial yield stress and the uniaxial/biaxial coefficients of
plastic anisotropy are also presented.
The predictions of the model have been tested by comparison with experimental
data for several types of materials (steel and aluminium alloys). The yield criterion
has been also implemented by the author in the LS-DYNA programme and used for
the numerical simulation of various sheet metal forming processes (deep-drawing
and bending) [87].

2.4.5.4 Soare Yield Criteria

Soare [88] proposed three yield criteria expressed by polynomial functions of 4th,
6th and 8th order, respectively (Poly 4, 6 and 8).
The yield function is defined as

f (σ11 , σ22 , σ12 ) := [Pn (σ11 , σ22 , σ12 )]1/n , (2.211)

where, Pn is the polynomial function and n—the order of the polynomial function.
The form of the orthotropic fourth order polynomial (Poly 4) is:

4 + a σ3 σ + a σ2 σ2 + a σ σ3 + a σ4 +
P4 = a1 σ11 2 11 22 3 11 22 4 11 22 5 22
2 + a σ σ + a σ 2 )σ 2 + a σ 4
(2.212)
+(a6 σ11 7 11 22 8 22 12 9 12

where a1 −a9 are coefficents describing the anisotropy.

The formulation proposed by Soare for the fourth order yield function differs
from Gotoh’s model [64] only in one aspect: elimination of the first term in Eq.
(2.160). Soare has paid a special attention to the convexity analysis. He has proposed
an original identification strategy which is able to remove the large overall error
affecting Gotoh’s procedure.
The 6th and 8th order polynomial functions developed by Soare are

6 5 4 σ2 + a σ3 σ3 + a σ2 σ4 + a σ σ5 + a σ6 +
P6 = a1 σ11 + a2 σ11 σ22 + a3 σ11 22 4 11 22 5 11 22 6 11 22 7 22
4 3
+(a8 σ11 + a9 σ11 σ22 + a10 σ112 σ 2 + a σ σ 3 + a σ 4 )σ 2 +
22 11 11 22 12 22 12
+(a13 σ112 + a σ σ + a σ 2 )σ 4 + a σ 2 ,
14 11 22 15 22 12 16 12
(2.213)
respectively,

8
P8 = a1 σ11 7 6 2 5 3 4 σ4 + a σ3 σ5 + a σ2 σ6 +
+a2 σ11 σ22 +a3 σ11 σ22 +a4 σ11 σ22 + a5 σ11 22 6 11 22 7 11 22
7 8 6 5
+a8 σ11 σ22 +a9 σ22 +(a10 σ11 +a11 σ11 σ22 +a12 σ11 4 σ 2 +a σ 3 σ 3 +a σ 2 σ 4 +
22 13 11 22 14 11 22
+a15 σ11 σ225 + a σ 6 )σ 2 + (a σ 4 + a σ 3 σ + a σ 2 σ 2 + a σ σ 3 +
16 22 12 17 11 18 11 22 19 11 22 20 11 22
4 )σ 4 + (a σ 2 + a σ σ + a σ 2 )σ 6 + a σ 2 .
+a21 σ22 12 22 11 23 11 22 24 22 12 25 12
(2.214)
2.5 BBC 2005 Yield Criterion 91

As one may notice, Poly 6 and 8 have 16 and 25 coefficients. The procedure
used for evaluating them is based on the minimisation of an error-function. Due
to the large number of coefficients, Poly 6 and 8 allow a better description of the
plastic behaviour, even in the case of materials exhibiting a pronounced variation
of the anisotropy characteristics. By implementing them in finite-element codes, the
author has proved the ability of the new models to capture the occurrence of 6 or 8
ears in the deep-drawing process of cylindrical cups. The strength-differential effect
into the yield surface has been also introduced in these formulations [89].
The most important advantages of these yield criteria are as follows:

• Simplicity of the formulation

• Direct formulation (use of the actual stress components)
• Flexibility ensured by the large number of coefficients (Poly 6 and 8)
• Easy extension to 3D stress states (see [88]).

The are also some disadvantages that should be mentioned:

• Not all the formulations are convex. Due to this fact, the variation range of some
coefficients must be bounded
• The identification procedure is quite complex, especially for the Poly 6 and 8
models.

A quadratic yield model to describe the ortotropic behaviour of the sheet metals
has been proposed by Oller et al. [90]. It deals with the case in which the yield stress
in simple tension is different from the one in compression.

2.5 BBC 2005 Yield Criterion

2.5.1 Equation of the Yield Surface

The sheet metal is assumed to behave as a plastically orthotropic membrane under
plane stress conditions. By making this assumption, we can use the following
description of the yield surface:
   
 σαβ , Y := σ σαβ − Y = 0 (2.215)
 
where σ σαβ > 0 is the BBC 2005 equivalent stress (see Sect. 2.5.3), Y > 0
is a yield parameter, and σαβ = σβα (α, β = 1, 2) are planar components of the
stress tensor expressed in an orthonormal basis superimposed to the axes of plas-
tic orthotropy: (1) rolling direction (RD), (2) transverse direction (TD), (3) normal
direction (ND). The other components are subjected to the restriction

σ3i = σi3 = 0, (i = 1, 2, 3) (2.216)

92 2 Plastic Behaviour of Sheet Metal

arising from the plane stress hypothesis. Whenever not clearly specified, we shall
use the following convention: Greek indices take the values 1 and 2, while the Latin
ones take the values 1, 2 and 3.
The BBC 2005 yield criterion does not enforce some special constraints on the
choice of the yield parameter (Y). In fact, any quantity representing a yield stress
can act as Y. For example, Y may be the uniaxial yield stress Yθ associated to a
direction defined by the angle θ measured from RD, an average of several uniaxial
yield stresses, or the biaxial yield stress Yb associated to RD and TD.

2.5.2 Flow Rule Associated to the Yield Surface

The flow rule associated to the yield surface described by Eq. (2.215) is

σ3i = σi3 = 0, (i = 1, 2, 3) (2.216)

p ∂
ε̇αβ = λ̇ , α, β = 1, 2 (2.217)
∂σαβ
p p
where ε̇αβ = ε̇αβ (α, β = 1, 2) are planar components of the plastic strain-rate
tensor (expressed in the same basis as the corresponding components of the stress
tensor), and λ̇ ≥ 0 is a scalar multiplier (its significance is not essential for our dis-
cussion). The out of plane components of the plastic strain-rate tensor are subjected
to the restrictions
p p
ε̇3α = ε̇α3 = 0, α = 1, 2 (2.218)
p p p
ε̇33 = − ε̇11 − ε̇22 (2.219)

arising from the plane stress hypothesis and the isochoric character of the plastic
deformation.
When using Eq. (2.217) we need the partial derivatives of the function  with
respect to the planar components of the stress tensor. Equation (2.215) allows us to
calculate them as partial derivatives of the equivalent stress:

∂ ∂σ
= , α, β = 1, 2 (2.220)
∂σαβ ∂σαβ

2.5.3 BBC 2005 Equivalent Stress

The equivalent stress used in Eq. (2.215) is defined by the following formula:
 1
2k
σ = a ( + Ŵ)2 k + a ( − Ŵ)2 k + b ( + )2 k + b ( − )2 k (2.221)
2.5 BBC 2005 Yield Criterion 93

where k ∈ N ≥ 1 and a, b > 0 are material parameters, while Ŵ, and  are

functions depending on the planar components of the stress tensor:

Ŵ = Lσ
! 11 + Mσ22
 = !(Nσ11 − Pσ22 )2 + σ12 σ21 (2.222)
 = (Qσ11 − Rσ22 )2 + σ12 σ21

The coefficients L, M, N, P, Q, and R involved in Eqs. (2.222) are also material

parameters.
Despite the fact that Eqs. (2.221) and (2.222) do not enforce any constraint on the
sign of the coefficients L, M, N, P, Q, and R, the numerical tests performed by the
authors have shown that positive values of these parameters lead to better predictions
of the BBC 2005 yield criterion.
The conditions k ∈ N ≥ 1 and a, b > 0 ensure the convexity of the yield surface
defined by Eqs. (2.215), (2.221) and (2.222). The parameters L, M, N, P, Q, and R
are not subjected to any constraint from this point of view.
Nine material parameters are involved in the expression of the BBC 2005 equiv-
alent stress: k, a, b, L, M, N, P, Q, and R (see Eqs. 2.221 and 2.222). The integer
exponent k has a special status, due to the fact that its value is fixed from the very
beginning in accordance with the crystallographic structure of the material: k = 3
for BCC materials, k = 4 for FCC materials.
The identification procedure calculates the other parameters (a, b, L, M, N, P, Q,
and R) by forcing the constitutive equations associated to the BBC 2005 yield
criterion to reproduce the following experimental data:

• The uniaxial yield stresses associated to the directions defined by 0◦ , 45◦ and 90◦
angles measured from RD (denoted as Y0 , Y45 and Y90 )
• The coefficients of uniaxial plastic anisotropy associated to the directions defined
by 0◦ , 45◦ and 90◦ angles measured from RD (denoted as r0 , r45 and r90 )
• The biaxial yield stress associated to RD and TD (denoted as Yb )
• The coefficient of biaxial plastic anisotropy associated to RD and TD (denoted
as rb ).

There are 8 constraints acting on 8 material parameters. The identification proce-

dure has enough data to generate a set of equations having a, b, L, M, N, P, Q, and R
as unknowns. The structure of this set of equations, as well as the solution strategy
will be presented in Sect. 2.5.4.
When using the flow rule given by Eq. (2.217), we need the partial derivatives of
the function φ with respect to the planar components of the stress tensor. Equations
(2.220), (2.221) and (2.222) lead to the following formula:

∂ ∂σ ∂Ŵ ∂σ ∂ ∂σ ∂
= + + , α, β = 1, 2 (2.223)
∂σαβ ∂Ŵ ∂σαβ ∂ ∂σαβ ∂ ∂σαβ
94 2 Plastic Behaviour of Sheet Metal

where
∂σ a
= 2 k−1 ( + Ŵ)2 k−1 − ( − Ŵ)2 k−1
∂Ŵ σ̄
∂σ 1
= 2 k−1 a ( + Ŵ)2 k−1 +( − Ŵ)2 k−1 +b (+ )2 k−1+( − )2 k−1
∂ σ̄
∂σ b
= 2 k−1 ( + )2 k−1 − ( − )2 k−1
∂ σ̄
(2.224)
and

∂Ŵ ∂Ŵ ∂Ŵ ∂Ŵ
= L, = M, = 0, = 0,
∂σ11 ∂σ22 ∂σ12 ∂σ21

∂ N (Nσ11 − Pσ22 ) ∂ P (Nσ11 − Pσ22 )

= , = − ,
∂σ11  ∂σ22 

∂ σ21 ∂ σ12
= , = , (2.225)
∂σ12 2 ∂σ21 2

∂ Q (Qσ11 − Rσ22 ) ∂ R (Qσ11 − Rσ22 )

= , = − ,
∂σ11  ∂σ22 

∂ σ21 ∂ σ12
= , = ,
∂σ12 2 ∂σ21 2
Equations (2.221), (2.222), (2.223), (2.224) and (2.225) allow us to express
the flow rule given by Eq. (2.217) as a dependency of the stress components
σαβ (α, β = 1, 2).

2.5.4 Identification Procedure

As mentioned in Sect. 2.5.3, the parameters a, b, L, M, N, P, Q, and R are obtained by
constraining the constitutive equations associated to the BBC 2005 yield criterion
to reproduce the following experimental data: Y0 , Y45 , Y90 , r0 , r45 , r90 , Yb , and
rb . In fact, the identification procedure will solve the following set of 8 equations
considering a, b, L, M, N, P, Q, and R as unknowns:

Ỹ0 = Y0 , Ỹ45 = Y45 , Ỹ90 = Y90

r̃0 = r0 , r̃45 = r45 , r̃90 = r90 (2.226)
Ỹb = Yb , r̃b = rb

where:

• Ỹ0 , Ỹ45 and Ỹ90 are the theoretical yield stresses corresponding to pure tension
along the directions defined by 0◦ , 45◦ and 90◦ angles measured from RD
• r̃0 , r̃45 and r̃90 are the theoretical coefficients of uniaxial plastic anisotropy
associated to the directions mentioned above
2.5 BBC 2005 Yield Criterion 95

• Ỹb is the theoretical yield stress corresponding to biaxial tension along RD and
TD
• r̃b is the theoretical coefficient of biaxial plastic anisotropy associated to RD
and TD.

It is obvious that the identification procedure needs formulas for evaluating Ỹ0 ,
Ỹ45 , Ỹ90 , r̃0 , r̃45 , r̃90 , Ỹb , and r̃b . These formulas will be presented below.

2.5.4.1 Theoretical Yield Stress in Pure Tension

Let Ỹθ be the theoretical yield stress corresponding to pure tension along a direction
defined by the angle θ measured from RD. The planar components of the stress
tensor are

σ11 = Ỹθ cos2 θ , σ22 = Ỹθ sin2 θ , σ12 = σ21 = Ỹθ sin θ cos θ (2.227)

The quantities Ŵ,  and defined by Eqs. (2.222) become

Ŵ = Ỹθ Ŵθ ,  = Ỹθ θ ,  = Ỹθ θ (2.228)

where

L cos2 θ + M sin2 θ
Ŵθ = $
2
N cos2 θ − P sin2 θ + sin2 θ cos2 θ

θ = (2.229)
$ 2
θ = Q cos2 θ − R sin2 θ + sin2 θ cos2 θ

Equations (2.221) and (2.228) lead to the following expression of the equivalent
stress when pure tension is applied along the θ direction:

σ̄ |θ = Ỹθ · F (θ ) (2.230)

where
1
2k
F (θ ) = a (θ + Ŵθ )2 k + a (θ − Ŵθ )2 k + b (θ + θ )2 k + b (θ − θ )2 k
(2.231)

σ̄ |θ given by Eq. (2.230) should be replaced in Eq. (2.215). We thus obtain the
desired formula of the theoretical yield stress Ỹθ :

Y
Ỹθ = (2.232)
F (θ )

Ỹ0 , Ỹ45 and Ỹ90 can be calculated from Eqs. (2.232) and (2.231) using θ =0◦ , 45◦
and 90◦ , respectively.
96 2 Plastic Behaviour of Sheet Metal

2.5.4.2 Theoretical Coefficient of Uniaxial Plastic Anisotropy

The theoretical coefficient of uniaxial plastic anisotropy associated to a direction
inclined at the angle θ measured from RD is defined as follows:
p
ε̇θ +90◦
r̃θ = p (2.233)
ε̇DN
p
where: ε̇θ +90◦ is the plastic strain-rate component associated to the direction defined
p
by the angle θ + 90◦ , and ε̇DN is the component of the same tensor associated to
ND. After using the condition of plastic incompressibility
p p p
ε̇θ + ε̇θ +90◦ + ε̇DN = 0 (2.234)

Equation (2.233) becomes

p
ε̇θ
r̃θ = − p − 1 (2.235)
ε̇DN
p
The symbol ε̇θ denotes the plastic strain-rate component associated to the θ direc-
p p
tion. ε̇θ and ε̇DN may be rewritten using the components of the plastic strain-rate
tensor expressed in the orthotropy basis:
p p p p p 
ε̇θ = ε̇11 cos2 θ + ε̇22 sin2 θ +

ε̇12 + ε̇21 sin θ cos θ
p p p p
(2.236)
ε̇DN = ε̇33 = − ε̇11 − ε̇22
p p
We can replace now ε̇θ and ε̇DN given by Eqs. (2.236) into Eq. (2.235):
p p p p 
ε̇11 cos2 θ + ε̇22 sin2 θ +

ε̇12 + ε̇21 sin θ cos θ
r̃θ = p p − 1 (2.237)
ε̇11 + ε̇22

The right-hand side of Eq. (2.237) should be expressed in terms of the planar
stress components. This transformation is achieved using the flow rule (see Eqs.
2.217 and 2.220), as well as Eqs. (2.227) (they are valid because r̃θ is defined for a
uniaxial stress state):
 )
σαβ ∂σ∂ σ̄αβ )
)
1 θ
r̃θ =  ) − 1 (2.238)
Ỹθ ∂ σ̄ + ∂ σ̄ ))
∂σ11 ∂σ22 θ

The notation (·)|θ means that the expression enclosed by parentheses should be
calculated for pure tension along the θ direction. The summation rule for tensor
components has been used in Eq. (2.238).
The equivalent stress defined by Eqs. (2.221) and (2.222) is a homogeneous func-
tion of the stress components σαβ (α, β = 1, 2), its degree of homogeneity being
one. Thus we can use Euler’s theorem:
2.5 BBC 2005 Yield Criterion 97

∂ σ̄
σ̄ = σαβ (2.239)
∂σαβ

Equations (2.238), (2.239) and (2.230) lead to the following formula for r̃θ :

F (θ )
r̃θ =  ) − 1 (2.240)
∂ σ̄ ∂ σ̄
+
)
∂σ11 ∂σ22 )
θ
 )
∂ σ̄ ∂ σ̄ )
We shall express now ∂σ 11
+ ∂σ 22
) as a dependency of the θ angle. We start
θ
by rewriting Eq. (2.223) both for α = β = 1 and α = β = 2, assuming a uniaxial
stress state along the θ direction. We have two relationships that can be added, thus
obtaining
 ) )  )
∂ σ̄ ∂ σ̄ ∂ σ̄ ) ∂Ŵ ∂Ŵ
+ ) = + ) +
) )
∂σ11 ∂σ22 θ ∂Ŵ θ ∂σ11 ∂σ22 θ
(2.241)
)  ∂
∂ σ̄ )
)
∂ )
) 
∂ σ̄ ) ∂ ∂
)
+ ∂ + ∂σ22 ) + +
)
θ ∂σ11 θ ∂ θ ∂σ11 ∂σ22 )
θ

∂ σ̄ )
)
Equations (2.224), (2.228) and (2.230) allows us to express the derivatives ∂Ŵ θ ,
∂ σ̄ ) ∂ σ̄ )
) )
∂ θ and ∂ θ as functions of the θ angle:

∂σ ) a
(θ + Ŵθ )2 k−1 − (θ − Ŵθ )2 k−1
)
=
∂Ŵ θ [F(θ)]2 k−1

1
∂σ )
a (θ + Ŵθ )2 k−1 + (θ − Ŵθ )2 k−1 + b (θ + θ )2 k−1
)
=
∂ θ [F(θ)]2 k−1
+ (θ − θ )2 k−1

∂σ ) b
(θ + θ )2 k−1 − (θ − θ )2 k−1
)
=
∂ θ [F(θ)]2 k−1
(2.242)

where Ŵθ , θ and θ are defined by Eqs. (2.229). The other derivatives appearing
in the right-hand side of Eq. (2.241) can be also expressed as functions of the θ
angle (see Eqs. 2.225, 2.227, and 2.228):
 )
∂Ŵ ∂Ŵ
+ ) = L + M,
)
∂σ11 ∂σ22 θ

(N − P) N cos2 θ − P sin2 θ
 )  
∂ ∂ )
∂σ11 + ∂σ 22
) = θ , (2.243)
θ

(Q − R) Q cos2 θ − R sin2 θ
 )  
∂ ∂
+ ) =
)
∂σ11 ∂σ22 θ θ

After replacing the quantities given by Eqs. (2.242) and (2.243) into Eq.
(2.241)
 and making
) some rearrangements, we get the following relationship for
∂ σ̄ ∂ σ̄ )
∂σ11 + ∂σ22 ) : θ
98 2 Plastic Behaviour of Sheet Metal
 )
∂ σ̄ ∂ σ̄ )) G (θ )
+ = (2.244)
∂σ11 ∂σ22 θ ) [F (θ)]2 k−1
where
 
(N−P) N cos2 θ −P sin2 θ
 
G (θ ) = a θ + L + M (θ + Ŵθ )2 k−1 +

 
(N−P) N cos2 θ −P sin2 θ
 
a θ − L − M (θ − Ŵθ )2 k−1 +
  (2.245)
(N−P) N cos2 θ −P sin2 θ (Q−R) Q cos2 θ −R sin2 θ
  
2 k−1
b θ + θ (θ + θ ) +
 
(N−P) N cos2 θ − P sin2 θ (Q −R) Q cos2 θ −R sin2 θ
  
b θ − θ (θ − θ )2 k−1

We can now combine Eqs. (2.240) and (2.244) to obtain a formula for evaluating
the coefficient of uniaxial plastic anisotropy:

[F (θ )]2 k
r̃θ = −1 (2.246)
G (θ )

r̃0 , r̃45 and r̃90 can be calculated from Eqs. (2.246), (2.245) and (2.231) using
θ = 0◦ , 45◦ and 90◦ , respectively.

2.5.4.3 Theoretical Yield Stress in Biaxial Tension Along RD and TD

Let Ỹb be the theoretical yield stress corresponding to biaxial tension along RD and
TD. The planar components of the stress tensor are

σ11 = Ỹb , σ22 = Ỹb , σ12 = σ21 = 0 (2.247)

The quantities Ŵ,  and  defined by Eqs. (2.222) become

Ŵ = Ỹb Ŵb ,  = Ỹb b ,  = Ỹb b (2.248)

where

Ŵb = L + M ,
!
b = (N − P)2 = |N − P| , (2.249)
!
b = (Q − R)2 = |Q − R|

Equations (2.221) and (2.248) lead to the following expression of the equivalent
stress when biaxial tension is applied along RD and TD:
2.5 BBC 2005 Yield Criterion 99

σ̄ |b = Ỹb · Fb (2.250)

where
1
2k
Fb = a (b + Ŵb )2 k + a (b − Ŵb )2 k + b (b + b )2 k + b (b − b )2 k
(2.251)

σ̄ |b given by Eq. (2.250) should be replaced in Eq. (2.215). We thus obtain the
desired formula of the theoretical yield stress Ỹb :

Y
Ỹb = (2.252)
Fb

Ỹb can be calculated from Eqs. (2.252) and (2.251).

2.5.4.4 Theoretical Coefficient of Biaxial Plastic Anisotropy

The theoretical coefficient of biaxial plastic anisotropy associated to RD and TD is
defined as follows:
p
ε̇TD
r̃b = p (2.253)
ε̇RD
p p
where ε̇RD and ε̇TD are the components of the plastic strain-rate tensor correspond-
ing to RD and TD, respectively. The choice of the orthonormal basis allows us to
write the equalities

p p p p
ε̇RD = ε̇11 , ε̇TD = ε̇22 (2.254)
p p
We can replace now ε̇RD and ε̇TD given by Eqs. (2.254) into Eq. (2.253):

p p p
ε̇22 ε̇11 + ε̇22
r̃b = p = p − 1 (2.255)
ε̇11 ε̇11

The right-hand side of Eq. (2.255) should be expressed in terms of the planar
stress components. This transformation is achieved using the flow rule (see Eqs.
2.217 and 2.220), as well as Eqs. (2.247) (they are valid because r̃b is defined for a
biaxial stress state):
 )
∂ σ̄ ))
σαβ
1 ∂σαβ )b
r̃b =  ) − 1 (2.256)
Ỹb ∂ σ̄ ))
∂σ11 )b
100 2 Plastic Behaviour of Sheet Metal

The notation (·)|b means that the expression enclosed by parentheses should be
calculated for biaxial tension along RD and TD. Eqs. (2.256), (2.239) and (2.250)
lead to the following formula for r̃b :

Fb
r̃b =  ) − 1 (2.257)
∂ σ̄ ))
∂σ11 )b
 )
∂ σ̄ )
We shall find now the expression of the denominator ∂σ 11
) . We start by
b
rewriting Eq. (2.223) for α = β = 1, assuming a biaxial stress state along RD
and TD:
 ) ) ) ) ) ) )
∂ σ̄ )) ∂ σ̄ )) ∂Ŵ )) ∂ σ̄ )) ∂ )) ∂ σ̄ )) ∂ ))
= + + (2.258)
∂σ11 )b ∂Ŵ )b ∂σ11 )b ∂ )b ∂σ11 )b ∂ )b ∂σ11 )b

Equations (2.224), (2.248) and (2.250) allows us to express the derivatives ∂∂Ŵσ̄ )b ,
)
∂ σ̄ ) ∂ σ̄ )
) )
∂ b and ∂ b :
)
∂σ )) a
= 2 k−1 (b + Ŵb )2 k−1 − (b − Ŵb )2 k−1
∂Ŵ )b Fb
)
∂σ )) 1
= 2 k−1 a (b + Ŵb )2 k−1 + (b − Ŵb )2 k−1 + b (b + b )2 k−1
∂ )b Fb
+ (b − b )2 k−1
)
∂σ )) b
= 2 k−1 (b + b )2 k−1 − (b − b )2 k−1
∂ )b Fb
(2.259)

where Ŵb , b and b are defined by Eqs. (2.249). The other derivatives appearing in
the right-hand side of Eq. (2.258) can be also expressed from Eqs. (2.225), (2.247)
and (2.248):
) ) )
∂Ŵ )) ∂ )) N (N − P) ∂ )) Q (Q − R)
= L, = , = (2.260)
∂σ11 )b ∂σ11 )b b ∂σ11 )b b

After replacing the quantities givenby Eqs.) (2.259) and (2.260) into Eq. (2.258),
∂ σ̄ )
we get the following relationship for ∂σ 11
) (see also Eqs. 2.249):
b
 )
∂ σ̄ )) Gb
= 2 k−1 (2.261)
∂σ11 b )
Fb
2.5 BBC 2005 Yield Criterion 101

where
 
N(N−P) N(N−P)
Gb = a b +L (b + Ŵb )2 k−1 + a b −L (b − Ŵb )2 k−1 +
 
N(N− P) Q(Q− R) N(N− P) Q(Q− R)
b b + b (b + b )2 k−1 + b b − b (b − b )2 k−1
(2.262)

We can combine Eqs. (2.257) and (2.261) to obtain a formula for evaluating the
coefficient of biaxial plastic anisotropy:

Fb2 k
r̃b = − 1 (2.263)
Gb

Now we have all the quantities needed to construct the identification conditions
(see Eqs. 2.226). At first, we shall refer to Eqs. (2.226.1), (2.226.2), (2.226.3) and
(2.226.7) (the constraints associated to the yield stresses). Equations (2.232) and
(2.252) allow us to rewrite Eqs. (2.226.1), (2.226.2), (2.226.3) and (2.226.7) in a
more convenient form:

[F (0◦ )]2 k = y20 k , [F (45◦ )]2 k = y245k ,

(2.264)
[F (90◦ )]2 k = y290k , Fb2 k = y2b k .

where

Y Y Y Y
y0 = , y45 = , y90 = , yb = (2.265)
Y0 Y45 Y90 Yb

are normalized values of the experimental yield stresses.

In a similar way, Eqs. (2.244), (2.263) and (2.264) lead to the following expres-
sions of Eqs. (2.226.4), (2.226.5), (2.226.6) and (2.226.8) (the constraints associated
to the coefficients of plastic anisotropy):

1 1
G (0◦ ) = y2 k , G (45◦ ) = y2 k ,
r0 + 1 0 r45 + 1 45
(2.266)
1 1
G (90◦ ) = y2 k , Gb = y2 k .
r90 + 1 90 rb + 1 b

Finally, we use Eqs. (2.231), (2.229), (2.251), (2.249), (2.245) and (2.262) to
put into evidence the unknown material parameters a, b, L, M, N, P, Q, and R in the
left-hand side of Eqs. (2.264) and (2.266):
102 2 Plastic Behaviour of Sheet Metal

a (N + L)2 k + a (N − L)2 k + b (N + Q)2 k + b (N − Q)2 k = y20 k

! 2 k ! 2 k
a (N − P)2 + 1 + L + M + a (N − P)2 + 1 − L − M +
! ! 2 k
b (N − P)2 + 1 + (Q − R)2 + 1 +
! ! 2 k
b (N − P)2 + 1 − (Q − R)2 + 1 = (2 y45 )2 k

a (P + M)2 k + a (P − M)2 k + b (P + R)2 k + b (P − R)2 k = y290k

a (N − P + L + M)2 k + a (N − P − L − M)2 k +
b (N − P + Q − R)2 k + b (N − P − Q + R)2 k = y2b k
a (P − M) (N + L)2 k−1 + a (P + M) (N − L)2 k−1 +
r0
b (P + R) (N + Q)2 k−1 + b (P − R) (N − Q)2 k−1 = 2k
r0 + 1 y0
!  ! 2 k−1
a (Q − R)2 + 1 (N − P)2 + 1 + L + M +

! 2 k−1 
2
(N − P) + 1 − L − M +
 ! ! 2 k
b (N − P)2 + 1 + (Q − R)2 + 1 −

! ! 2 k 
2 2
(N − P) + 1 − (Q − R) + 1 =
&
!
2
!
2 r45 + 1 2
(N − P) + 1 (Q − R) + 1 (2 y45 )2 k
r45 + 1
a (N − L) (P + M)2 k−1 + a (N + L) (P − M)2 k−1 +
r90
b (N + Q) (P + R)2 k−1 + b (N − Q) (P − R)2 k−1 = 2k
r90 + 1 y90

a (N + L) (N − P + L + M)2 k−1 + a (N − L) (N − P − L − M)2 k−1 +

b (N + Q) (N − P + Q − R)2 k−1 +
1
b (N − Q) (N − P − Q + R)2 k−1 = 2k
rb + 1 yb
(2.267)

Equations (2.267) form together a set of 8 non-linear equations. The identifica-

tion procedure uses Newton’s method to obtain its numerical solution.
From now on we shall manipulate Eqs. (2.267) in a generic form:

fi (a, b, L, M, N, P, Q, R) = 0, i = 1, 2, . . . , 8 (2.268)
2.5 BBC 2005 Yield Criterion 103

where

1. f1 = a (N + L)2 k + a (N − L)2 k + b (N + Q)2 k + b (N − Q)2 k − y20 k

! 2 k ! 2 k
2. f2 = a (N − P)2 + 1 + L + M + a (N − P)2 + 1 − L − M +
! ! 2 k
b (N − P)2 + 1 + (Q − R)2 + 1 +
! ! 2 k
b (N − P)2 + 1 − (Q − R)2 + 1 − (2 y45 )2 k

3. f3 = a (P + M)2 k + a (P − M)2 k + b (P + R)2 k + b (P − R)2 k − y290k

4. f4 = a (N − P + L + M)2 k + a (N − P − L − M)2 k +
b (N − P + Q − R)2 k + b (N − P − Q + R)2 k − y2b k
5. f5 = a (P − M) (N + L)2 k−1 + a (P + M) (N − L)2 k−1 +
b (P + R) (N + Q)2 k−1 + b (P − R) (N − Q)2 k−1 − r0r+0 1 y20 k
!  ! 2 k−1
6. f6 = a (Q − R)2 + 1 (N − P)2 + 1 + L + M +

! 2 k−1 
2
(N − P) + 1 − L − M +
 ! ! 2 k
b (N − P)2 + 1 + (Q − R)2 + 1 −

! ! 2 k 
2 2
(N − P) + 1 − (Q − R) + 1 −

+ 1/ 2
! !
(N − P)2 + 1 (Q − R)2 + 1 r45
r45 + 1 (2 y45 )
2k

7. f7 = a (N − L) (P + M)2 k−1 + a (N + L) (P − M)2 k−1 +

r90
b (N + Q) (P + R)2 k−1 + b (N − Q) (P − R)2 k−1 − 2k
r90 + 1 y90

8. f8 = a (N + L) (N − P + L + M)2 k−1 + a (N − L) (N − P − L − M)2 k−1 +

b (N + Q) (N − P + Q − R)2 k−1 +
1
b (N − Q) (N − P − Q + R)2 k−1 − 2k
rb + 1 yb
(2.269)

As mentioned in Sect. 2.5.3, the parameters involved in the expression of the

equivalent stress are subjected to constraints: a > 0, b > 0, L > 0, M > 0, N >
0, P > 0, Q > 0, and R > 0. Aiming to ensure a natural treatment of these restric-
tions, the identification procedure replaces the parameters a, b, L, M, N, P, Q, and
R with the following substitutes:
104 2 Plastic Behaviour of Sheet Metal

 ′ 2  2  2  2
a = a , b = b ′ , L = L′ , M = M ′ ,
 2  2  2  2
N = N ′ , P = P′ , Q = Q′ , R = R′ , (2.270)
a′ , b′ , L′ , M′, N′, P′ , Q′ , R′ ∈R

In this way, f1 , f2 , . . . , f8 will become functions of the variables

a′ , b′ , L′ , M ′ , N ′ , P′ , Q′ , and R′ . As a consequence, Eqs. (2.268) should be
rewritten in the form

fi a a′ , b b′ , L L′ , M M ′ , N N ′ , P P′ , Q Q′ , R R′
               
= 0,
i = 1, 2, . . . , 8
(2.271)

The identification procedure will solve Eqs. (2.271) considering

a′ , b′ , L′ , M ′ , N ′ , P′ , Q′ , and R′ as unknowns. After finding the numerical
solution, the values of the actual parameters a, b, L, M, N, P, Q, and R can be
obtained from Eqs. (2.270).
Let a′k , b′k , Lk′ , Mk′ , Nk′ , P′k , Q′k , and R′k be the approximations of the numerical
solution corresponding to the k-th Newton iteration. Eqs. (2.271) are linearised in
the vicinity of this approximation using a truncated Taylor expansion:
) ) ) )
∂fi )) ∂a ))  ′  ∂fi )) ∂b ))  ′ 
a k + bk +
∂a )k ∂a′ )k ∂b )k ∂b′ )k
) ) ) )
∂fi )) ∂L ))  ′  ∂fi )) ∂M )) 
Mk′ +

′
L k + ′
∂L k ∂L k
) ) ∂M k ∂M k
) )
) ) ) )
∂fi )) ∂N ))  ′ ∂fi )) ∂P ))  ′ 
(2.272)

N k + Pk +
∂N )k ∂N ′ )k ∂P )k ∂P′ )k
) ) ) )
∂fi )) ∂Q ))  ′
 ∂fi )) ∂R ))  ′ 
Qk + Rk = − fi |k ,
∂Q )k ∂Q′ )k ∂R )k ∂R′ )k
i = 1, 2, . . . , 8

The symbol ·|k means that the associated expression should be evaluated
considering
a′ = a′k , b′ = b′k , L′ = Lk′ , M ′ = Mk′ , N ′ = Nk′ , P′ = P′k , Q′ = Q′k , and
R = R′k .
′

The unknowns of the linearised set (see Eqs. 2.271 and 2.270) are the cor-
rections a′k , b′k , Lk′ ,Mk′ , Nk′ , P′k , Q′k , and R′k . After adding them to
a′k , b′k , Lk′ , Mk′ , Nk′ , P′k , Q′k , and R′k , respectively, we obtain a new approximation
of the numerical solution that should be used in the next iteration:
2.5 BBC 2005 Yield Criterion 105

a′k+1 = a′k + a′k

b′k+1 = b′k + b′k
′
Lk+1 = Lk′ + Lk′
Mk+1 = Mk′ + Mk′
′
′ (2.273)
Nk+1 = Nk′ + Nk′
Pk+1 = P′k + P′k
′

Q′k+1 = Q′k + Q′k

R′k+1 = R′k + R′k

Two convergence criteria are used to stop the iterations:

%
(a′k )2 + (b′k )2 + (Lk′ )2 + (Mk′ )2 + (Nk′ )2 + (P′k )2 + (Q′k )2 + (R′k )2
 2  2  2  2  2  2  2  2 < 10−7
a′k+1 ′
+ b′k+1 + Lk+1 ′
+ Mk+1 ′
+ Nk+1 + P′k+1 + Q′k+1 + R′k+1
%
8  2
< 10−5
3
fi |k
i=1
(2.274)

The convergence of the Newton iterations is strongly influenced by the initial

guess.

2.5.5 Particular Formulations of the BBC 2005 Yield Criterion

We can reduce BBC 2005 to Hill 1948 yield criterion if we choose the material
parameters as follows:

Y = Y0
k = 1
"
1 + r90
1+ r0
r0 −1    
r90 1 r0 1
a = " 1+ r45 +
1 + r0 1 + r90 1 + r0 r90 2
+1
r0 r90
a (2.275)
b= "
1 + r0 1 + r90
−1
r0 r90
1
L = N = Q = √
2 %a + b
r0 1 + r90
1 1 + r0 r90
M = P = R =
2 a+b
106 2 Plastic Behaviour of Sheet Metal

In this case, the identification procedure needs only r0 , r45 and r90 as input data.
The yield criterion proposed by Barlat and Lian in 1989 can be also obtained by
enforcing the following constraints on the material parameters:

Y = Y0 , k = 3 or 4, L = N = Q, M = P = R (2.276)

As above, the identification procedure needs only r0 , r45 and r90 as input data.
Another situation of practical interest is the so-called normal anisotropy (r0 =
r45 =r90 =r, Y0 =Y45 =Y90 =Y). In this case, BBC 2005 also reduces to the Hill 1948
or Barlat 1989 yield criteria (depending on the value of the exponent k):

k = 1 (Hill 1948), k = 3 or 4 (Barlat 1989),

1 r 1 (2.277)
a = 1+r , b = 1+r , L = N = Q = M = P = R= 2

There are many situations when the coefficient of biaxial plastic anisotropy (rb )
is not available as input data. The most convenient strategy for handling such cases
consists in replacing Eq. (2.267.8) with the following constraint:

N = P (2.278)

In this way, the number of identification equations is preserved. As a conse-

quence, the solution procedure discussed in Sect. 2.5.4 will remain usable. The
numerical tests performed for several sorts of sheet metals have shown that the
above constraint leads to well-shaped yield loci (i.e. the predicted value of rb is
not far from 1).
The situations when the biaxial yield stress (Yb ) and the coefficient of biaxial
plastic anisotropy (rb ) are not available can be handled by replacing Eqs. (2.267.4)
and (2.267.8) with the following constraints:

L + M = 2 N, N = P (2.279)

This strategy has the same advantages as those mentioned above.

The situations when the uniaxial yield stresses Y45 and Y90 , as well as the coef-
ficient of biaxial plastic anisotropy rb are not available can be handled by replacing
Eqs. (2.267.2), (2.267.3) and (2.267.8) with the following constraints:

M = R, L + M = 2 N, N = P (2.280)

2.6 BBC 2008 Yield Criterion

In order to enhance the flexibility of the BBC 2005 yield criterion, a new version of
this model has been developed [91]. The model is expressed as a finite series that can
be expanded to retain more or less terms, depending on the volume of experimental
2.6 BBC 2008 Yield Criterion 107

data. Different identification strategies (using 8, 16, 24, etc. input values) could be
used in order to determine the coefficients of the yield function.

2.6.1 Equation of the Yield Surface

The sheet metal is assumed to behave as a plastically orthotropic membrane under
plane-stress conditions. We use the following description of the yield surface:

σ̄ (σαβ ) − Y = 0 (2.281)

where σ̄ (σαβ ) ≥ 0 is the equivalent stress defined in Sect. 2.6.2, Y > 0 is the yield
parameter, and σαβ = σβα (α, β = 1, 2) are planar components of the stress tensor
expressed in an orthonormal basis superimposed to the axes of plastic orthotropy:
(1) rolling direction (RD), (2) transverse direction (TD), (3) normal direction (ND).
The other components are subjected to the constraint

σ3i = σi3 = 0, i = 1, 2, 3 (2.282)

arising from the plane-stress hypothesis. Whenever not specified, the following con-
vention will be adopted: Latin subscripts take the values 1, 2 and 3, while the Greek
ones take only the values 1 and 2.
The equivalent stress defined in Sect. 2.6.2 does not enforce constraints on the
choice of the parameter Y. In fact, any quantity representing a yield stress can act
as Y. For example, Y may be the uniaxial yield stress Yθ associated to a planar
direction defined by the angle θ measured from RD, an average of several uniaxial
yield stresses, or the biaxial yield stress corresponding to the tension along RD
and TD.
The flow rule associated to the yield surface described by Eq. (2.281) is

(p) ∂ σ̄
ε̇αβ = λ̇ (2.283)
∂σαβ

(p) (p)
where ε̇αβ = ε̇βα are planar components of the plastic strain-rate tensor (expressed
in the same basis as the corresponding components of the stress tensor), and λ̇ ≥ 0
is a scalar multiplier (its significance is not essential for our discussion). The out of
plane components of the plastic strain-rate are subjected to the constraints

(p) (p) (p) (p) (p)

ε̇3α = ε̇α3 = 0, ε̇33 = −ε̇11 − ε̇22 (2.284)

arising from the plane-stress hypothesis and the isochoric character of the plastic
deformation [29].
108 2 Plastic Behaviour of Sheet Metal

2.6.2 BBC 2008 Equivalent Stress

The equivalent stress used in Eq. (2.281) is defined as follows:
s   
σ̄ 2 k 2k 2k
wi−1 L(i) + M (i) + L(i) − M (i)
3
w−1 = +
 i=1 
2k 2k
ws−i M (i) + N (i) + M (i) − N (i)

k, s ∈ N∗ w = 3 2 / > 1
 & 1 s

(i) (i)
L(i) = ℓ1 σ11 + ℓ2 σ22
" (2.285)
2 2
(i) (i) (i)
M (i) = m1 σ11 − m2 σ22 + m3 (σ12 + σ21 )

" 2 2
(i) (i) (i)
N (i) = n1 σ11 − n2 σ22 + n3 (σ12 + σ21 )

(i) (i) (i) (i) (i) (i) (i) (i)

ℓ1 , ℓ2 , m1 , m2 , m3 , n1 , n2 , n3 ∈ R.

(i) (i) (i) (i) (i) (i) (i) (i)

The quantities denoted k, ℓ1 , ℓ2 , m1 , m2 , m3 , n1 , n2 , n3 (i = 1, . . . , s) are
material parameters. One may prove that k ∈ N∗ is a sufficient condition for the
convexity of the yield surface defined by Eqs. (2.281) and (2.285). From this point
of view, there is no constraint acting on the admissible values of the other material
parameters.
It is easily noticeable that Eq. (2.285) reduces to the isotropic formulation
proposed by Barlat and Richmond [49] if
(i) (i) (i) (i) (i) (i) (i) (i) &
ℓ1 = ℓ2 = m1 = m2 = m3 = n1 = n2 = n3 = 1 2, i = 1, . . . , s
(2.286)
Under these circumstances, the exponent k may be chosen as in Barlat and
Richmond’s model, i.e. according to the crystallographic structure of the sheet
metal: k = 3 for BCC materials (2 k = 6), and k = 4 for FCC materials (2 k = 8).
The other parameters involved in Eq. (2.285) result from an identification
procedure (see 2.6.3). Their number (np ) is defined by the summation limit s :

np = 8s (2.287)

Let ne be the number of experimental values describing the plastic anisotropy.

The summation limit should be chosen according to the following constraint:

np = 8s ≤ ne (2.288)
2.6 BBC 2008 Yield Criterion 109

i.e.

s ∈ N∗
&
s ≤ ne 8, (2.289)

Apparently, Eq. (2.285) is usable only when ne ≥ 8. In fact, it also works with
less experimental values. When such a situation occurs, the summation limit should
be s = 1, and the ne < 8 identification constraints arisen from experiments should
be accompanied by at least 8 − ne artificial conditions involving the material param-
(1) (1)
eters. For example, if ne = 6, we may enforce the equalities m1 = n1 and
(1) (1)
m2 = n2 .

2.6.3 Identification Procedure

Due to the expandable structure of the yield criterion, many identification strategies
can be devised. We shall restrict our discussion to a procedure that uses only nor-
malized yield stresses and r-coefficients obtained from uniaxial and biaxial tensile
tests.
Let Yθ be the yield stress predicted by the yield criterion in the case of a uniaxial
traction along the direction defined by the angle θ measured from RD. The planar
components of the stress tensor are in this case

σ11 |θ = Yθ cos2 θ , σ22 |θ = Yθ sin2 θ ,

(2.290)
σ12 |θ = σ21 |θ = Yθ sin θ cos θ

After replacing them in Eq. (2.285), we get the associated equivalent stress

σ̄ |θ = Yθ Fθ (2.291)

where Fθ is defined by the relationships

s
  
Fθ2 k (i) 2 k (i) 2 k
 
3 i−1 (i) (i)
w−1 = w Lθ + Mθ + Lθ − Mθ +
i=1
 
(i) 2 k (i) 2 k
 
(i) (i)
ws−i Mθ + Nθ + Mθ − Nθ

Lθ(i) = ℓ(i) 2 (i) 2

1 cos θ + ℓ2 sin θ (2.292)
" 2 2
(i) (i) (i) (i)
Mθ = m1 cos2 θ − m2 sin2 θ + m3 sin 2θ

" 2 2
(i) (i) (i) (i)
Nθ = n1 cos2 θ − n2 sin2 θ + n3 sin 2θ
110 2 Plastic Behaviour of Sheet Metal

Equations (2.281) and (2.291) lead to the following expression of the normalized
uniaxial yield stress:

Yθ 1
yθ = = (2.293)
Y Fθ
The r-coefficient corresponding to the uniaxial traction along a direction inclined
at the angle θ measured from RD is defined by the formula

(p)
ε̇θ +90◦
rθ = (p)
(2.294)
ε̇ND
(p)
where ε̇θ +90◦ is the plastic strain-rate component associated to the θ + 90◦ planar
(p)
direction, and ε̇ND is the through-thickness component of the same tensor. After
some simple mathematical manipulations, Eq. (2.294) becomes

Fθ
rθ = −1 (2.295)
Gθ
where Gθ is defined by the relationships

s

Fθ2 k−1 Gθ (i) 2 k−1
 
(i) (i) (i)
wi−1 L̂θ + M̂θ
3
w−1 = Lθ + Mθ +
i=1

(i) 2 k−1 (i) 2 k−1

   
(i) (i) (i) (i) (i) (i)
wi−1 L̂θ − M̂θ Lθ − Mθ + ws−i M̂θ + N̂θ Mθ + N θ +

(i) 2 k−1
 
(i) (i) (i)
ws−i M̂θ − N̂θ Mθ − Nθ

L̂θ(i) = ℓ(i) (i)

1 + ℓ2
 4
(i) (i) (i) (i) (i) (i)
M̂θ = m1 − m2 m1 cos2 θ − m2 sin2 θ Mθ
 4
N̂θ(i) = n(i) (i)
1 − n2
(i) (i)
n1 cos2 θ − n2 sin2 θ
(i)
Nθ

(2.296)
together with Eq. (2.292).
Let us denote by Yb the yield stress predicted in the case of a biaxial traction
along RD and TD. The corresponding planar components of the stress tensor are

σ11 |b = Yb , σ22 |b = Yb , σ12 |b = σ21 |b = 0 (2.297)

After replacing them in Eq. (2.285), we get the associated equivalent stress
2.6 BBC 2008 Yield Criterion 111

σ̄ |b = Yb Fb (2.298)

where Fb is defined by the relationships

s
  
Fb2 k (i) 2 k (i) 2 k
 
(i) (i)
wi−1
3
w−1 = Lb + Mb + Lb − M b +
i=1
 
(i) 2 k (i) 2 k (2.299)
 
(i) (i)
ws−i Mb + Nb + Mb − N b

(i) (i) (i) (i) (i) (i) (i) (i) (i)

Lb = ℓ1 + ℓ2 , Mb = m1 − m2 , Nb = n1 − n2

Equations (2.281) and (2.298) lead to the following expression of the normalized
biaxial yield stress:

Yb 1
yb = = (2.300)
Y Fb
The r-coefficient corresponding to the biaxial traction along RD and TD is
defined by the formula

(p)
ε̇TD
rb = (p)
(2.301)
ε̇RD
(p) (p)
where ε̇RD and ε̇TD are the plastic strain-rate components associated to the rolling
and transverse directions, respectively. After some simple mathematical manipula-
tions, Eq. (2.301) becomes

Fb
rb = −1 (2.302)
Gb
where Gb is defined by the relationships

s

Fb2 k−1 Gb (i) 2 k−1
 
(i) (i) (i)
wi−1 L̂b + M̂b
3
w−1 = Lb + M b +
i=1

(i) 2 k−1 (i) 2 k−1

   
(i) (i) (i) (i) (i) (i)
wi−1 L̂b − M̂b Lb − Mb + ws−i M̂b + N̂b Mb + N b +

(i) 2 k−1
 
(i) (i) (i)
ws−i M̂b − N̂b Mb − Nb

(i) (i) (i) (i) (i) (i)

L̂b = ℓ1 , M̂b = m1 , N̂b = n1
(2.303)
together with Eq. (2.299).
112 2 Plastic Behaviour of Sheet Metal

An identification procedure that strictly enforces a large number of experimen-

tal constraints on the yield criterion would be inefficient in practical applications.
The failure probability of such a strategy increases when the external restrictions
become stronger. Taking into account this aspect, the authors have developed an
identification procedure based on the minimization of the following error-function:

) 
(i) (i) (i) (i) (i) (i) (i) (i) )
E ℓ1 , ℓ2 , m1 , m2 , m3 , n1 , n2 , n3 ) i = 1, . . . , s =

 2
(exp)
yθ (exp)
2  (exp)
yb
2
(exp)
2 (2.304)
j
3 3
yθj −1 + rθj − rθj + yb −1 + rb − rb
θj θj

where θj represents an individual element from a finite set of angles defining the ori-
entation of the specimens used in the uniaxial tensile tests. One may notice that Eq.
(2.304) describes a square-distance between the experimental and predicted values
of the anisotropy characteristics.
Two versions of the BBC 2008 yield criterion have been evaluated from the point
of view of their performances (see [91]). They include 8 and 16 material coefficients,
respectively, and correspond to the smallest values of the summation limit (s = 1
and s = 2). The identification of the BBC 2008 (16 parameters) model has been
(exp) (exp) (exp) (exp)
performed using the following mechanical parameters: y0◦ , y15◦ , y30◦ , y45◦ ,
(exp) (exp) (exp) (exp) (exp) (exp) (exp) (exp) (exp) (exp) (exp) (exp)
y60◦ , y75◦ , y90◦ , yb , r0◦ , r15◦ , r30◦ , r45◦ , r60◦ , r75◦ , r90◦ and rb . In
the case of BBC 2008 (8 parameters), the input data has been restricted to the values
(exp) (exp) (exp) (exp) (exp) (exp) (exp) (exp)
y0◦ , y45◦ , y90◦ , yb , r0◦ , r45◦ , r90◦ and rb .
The predictions of the BBC 2008 model with 16 parameters are superior to those
given by the 8-parameters version. The improvement is noticeable especially in the
case of the r-coefficients. This capability of the 16-parameter version is relevant for
the accurate prediction of the thickness when simulating sheet metal forming pro-
cesses. For the materials exhibiting a distribution of the anisotropy characteristics
that would lead to the occurrence of 8 ears in a cylindrical deep-drawing process
[69] the planar distribution of the r-coefficient predicted by the BBC 2008 yield
criterion with 8 parameters is very inaccurate (see [91]). This model would not be
able to predict the occurrence of more than 4 ears at the top edge of a cup deep-
drawn from a circular blank. In contrast, the variation of the r-coefficient described
by BBC 2008 with 16 parameters closely follows the reference data. In conclu-
sion, this model would predict the occurrence of 8 ears as reported by Yoon et al.
[69].
As compared with other formulations described in the literature, the new model
does not use linear transformations of the stress tensor. Due to this fact, its com-
putational efficiency should be superior in the simulation of sheet metal forming
processes.
2.7 Recommendations on the Choice of the Yield Criterion 113

2.7 Recommendations on the Choice of the Yield Criterion

As concerns the modelling of the anisotropic plasticity, the present state-of-the-art is
somewhat confusing since most of the above-described yield criteria are still being
used. The most important factors that must be taken into account when choosing the
yield criterion are as follows:

• Accuracy of the prediction both of the yield locus and the uniaxial yield stress
and uniaxial coefficient of plastic anisotropy
• Computational efficiency and ease of implementation in numerical simulation
codes
• Flexibility of the yield criterion
• Degree of generality
• Number of mechanical parameters needed by the identification procedure
• Robustness of the identification procedure
• Experimental difficulties caused by the determination of the mechanical parame-
ters involved in the identification procedure
• User-friendliness of the yield criterion
• Acceptance of the yield criterion in the scientific/industrial community.

We shall now analyse some of these factors.

2.7.1 Comparison of the Yield Criteria

At present, the most frequently used yield criteria are Hill 1948, Hill 1990 and
Barlat 1989. Due to this fact, they have been chosen as reference models for com-
parison with one of the advanced yield criteria, namely BBC 2000. The comparison
is focused on the following performance aspects: prediction of the yield locus geom-
etry and description of the planar distribution both for uniaxial yield stress and
uniaxial coefficient of plastic anisotropy. The aluminium alloy AA3103-0 has been
chosen as a test material.
Figure 2.19 shows the normalised yield loci predicted by Hill 1948, Hill 1990,
Barlat 1989 and BBC 2000 yield criteria together with experimental data [6, 70].
One may notice that the best predictions are provided by the BBC 2000 and Hill
1990 models, while the performances of the Hill 1948 and Barlat 1989 models are
unsatisfactory, especially in the biaxial tension region.
The planar distribution of the uniaxial yield stresses predicted by the mod-
els is shown in Fig. 2.20. As one may notice, BBC 2000 and Hill 1990 have
better performances. The predictions of the Hill 1948 and Barlat 1989 are in
poorer agreement with the experimental data, due to the fact that only the uniax-
ial yield stress corresponding to the rolling direction is used by their identification
procedure.
114 2 Plastic Behaviour of Sheet Metal

Fig. 2.19 Yield loci 1.4

predicted by the Hill 1948,
Hill 1990, Barlat 1989 and 1.2
BBC 2000 criteria versus
experimental one for the
AA3103-0 aluminium alloy 1.0

0.8

σ2 /Y
0.6

0.4
Hill 1948
Hill 1990
0.2 Barlat 1989
BBC 2000
Experiments
0.0
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4
σ1/ Y

Fig. 2.20 Uniaxial yield 1.15

stresses predicted by the Hill Hill 1948
Hill 1990
1948, Hill 1990, Barlat 1989 Barlat 1989
and BBC 2000 yield criteria 1.10 BBC 2000
versus experimental one for Experiments
the AA3103-0 aluminium
alloy 1.05
σ/Y

1.00

0.95
0 15 30 45 60 75 90
θ [°]

As concerns the prediction of the uniaxial coefficient of plastic anisotropy

(Fig. 2.21), the BBC 2000, Hill 1948 and Barlat 1989 have the best performances.
Due to the fact that only the coefficient corresponding to the diagonal direction is
used in its identification procedure, Hill 1990 cannot provide good predictions.
We can conclude that the best quality of the predictions will be ensured by the
yield criteria having an identification procedure based both on uniaxial and biaxial
tension experimental data. As concerns the experimental data obtained by uniaxial
tension tests, the identification should use at least the yield stresses and the coeffi-
cients of plastic anisotropy corresponding to three planar directions (0, 45 and 90◦ ).
Under such circumstances, the yield criterion will have to fit a minimum number of
seven mechanical parameters.
2.7 Recommendations on the Choice of the Yield Criterion 115

Fig. 2.21 Anisotropy 0.7

coefficients predicted by the
Hill 1948, Hill 1990, Barlat
1989 and BBC 2000 yield 0.6
criteria versus experimental
one for the AA3103-0
aluminium alloy

r [–]
0.5

Hill 1948
0.4 Hill 1990
Barlat 1989
BBC 2000
Experiments
0.3
0 15 30 45 60 75 90
θ [°]

Fig. 2.22 Yield loci for 1.50

3103-0 alloy sheet predicted 3103-O
with the Yld2000-2d, CB
2001 and BBC 2000 yield 1.25
functions
1.00
σyy / σ

0.75

0.50

Yld2000-2d
0.25 CB2001
BBC2000

0.00
0.00 0.25 0.50 0.75 1.00 1.25 1.50
σxx / σ

The yield criteria using at least seven mechanical parameters in the identifica-
tion procedure provide almost the same predictions of the anisotropic behaviour for
usual materials. This fact can be noticed by comparing three of the most recently
developed models, namely Barlat 2000 (Yld2000-2d) [5], Cazacu–Barlat 2001
(CB 2001) [75] and BBC 2000 [51]. The AA3103-0 aluminium alloy has been used
in order to evaluate the accuracy of the predictions (see [92, 93]). Figure 2.22 shows
the yield locus for this material measured in the first quadrant using biaxial ten-
sile testing of cruciform specimens [92] and predicted with the Yld2000-2d [5],
CB 2001 [75] and BBC 2000 [51] yield functions.
Figure 2.23 shows the experimental normalized yield stress and r value as a func-
tion of the tensile direction for a 3103-0 aluminium alloy sheet sample. The three
116 2 Plastic Behaviour of Sheet Metal

Fig. 2.23 Tensile anisotropy 1.15 0.65

for 3103-0 alloy sheet Yld2000-2d
CB2001
BBC2000
1.10

Normalized stress •
stress 0.6

r value ∇
1.05

0.55
r value
1.00

3103-O
0.95 0.5
0 20 40 60 80
Angle from rolling

models are able to capture tensile anisotropy and the yield locus shape of this alloy
very well.
The yield criteria that use a larger number of mechanical parameters in the iden-
tification (13 or even more—Barlat 2004, Soare 2007, BBC 2008 etc.) are able to
provide highly accurate descriptions of the anisotropic behaviour. It is especially
notable their capability to capture the occurrence of six or eight ears in the case of
deep-drawing of cylindrical cups (see [69, 89]).

2.7.2 Evaluating the Performances of the Yield Criteria

As emphasized in the previous section, the performances of the yield criteria must
be evaluated by thorough comparisons with experimental data. These comparisons
should not be limited to the analysis of the yield locus shape. They should also
envisage the planar distributions of the uniaxial yield stress and uniaxial coefficient
of plastic anisotropy. In order to have a comprehensive evaluation tool, the members
of the CERTETA research team have developed a global accuracy index defined as
follows [70, 71, 94]:

β = ϕ + δ + γ [%] (2.305)

where: φ is an accuracy index associated to the prediction of the yield locus shape
in the plane of the principal stresses; δ is the accuracy index associated to the pre-
diction of the planar distribution of the uniaxial yield stress; γ is the accuracy index
associated to the prediction of the planar distribution of the uniaxial coefficient of
plastic anisotropy.
2.7 Recommendations on the Choice of the Yield Criterion 117

Fig. 2.24 Projection Qi of an σ2

experimental point Pi onto
the yield locus predicted by a
yield criterion

Pi
Qi

O σ1

ϕ is computed by using the formula

%
n
d2 (Pi , Qi )
3
i=1
ϕ= 100 [%] (2.306)
Y
where Y is the reference yield stress, d2 (Pi , Qi ) is the squared distance from an
experimental point Pi and its projection Qi onto the yield locus predicted by the
yield criterion under testing (see Fig. 2.24), and n is the total number of the available
experimental points.
δ is computed by using the formula
5
6 n ' exp (
6+ σθi − σθti
δ=7 exp 100 [%] (2.307)
i=1
σθ i

exp
where; σθi is the experimental uniaxial yield stress corresponding to the direction
defined by the angle θ i (measured from the rolling direction); σθti is the predicted
uniaxial yield stress associated to the same direction; n is the total number of
experimental points.
γ is computed by using the formula
5
6 n ' exp (
6+ rθi − rθt i
γ =7 exp 100 [%] (2.308)
i=1
rθi

exp
where: rθi is the experimental anisotropy coefficient corresponding to the direction
defined by the angle θ i (measured from the rolling direction); rθt i is the predicted
anisotropy coefficient corresponding to the same direction; n is the total number of
experimental points.
The practical use of the global accuracy index is exemplified by comparing
four yield criteria: Hill 1948, Hill 1990, Barlat 1989 and BBC 2000. The iden-
tification procedure of the BBC 2000 model is based on the minimisation of an
118 2 Plastic Behaviour of Sheet Metal

Table 2.2 Quality index for different yield criteria

Material Quality index Hill 1948 Hill 1990 Barlat 1989 BBC 2000

AA3103-0 ϕ [%] 42.64 18.8 30.64 13.41

δ [%] 11.65 0 11.92 2.64
γ [%] 0 36.05 0 1.60
β [%] 54.29 56.03 42.56 17.65
AA5182-0 ϕ [%] 30.33 20.26 21.14 12.95
δ [%] 8.58 0 8.54 1.96
γ [%] 0 48.51 0 0.39
β [%] 38.91 68.77 29.68 15.3
AA6111-T4 ϕ [%] 12.84 0 7.47 2.57
δ [%] 16.76 1.95 12.47 6.69
γ [%] 8.69 54.20 6.98 8.38
β [%] 38.29 56.15 26.62 17.64

error-function. The comparison has been performed in the case of three aluminium
alloys: AA3103-0, AA5182-0 and AA6111-T4. The values of the individual accu-
racy indices are listed in Table 2.2. The best overall performance corresponds to the
lowest value of the global index β. Table 2.2 shows that BBC 2000 has better per-
formances as compared to the other yield criteria (the corresponding overall index
is three times smaller than in the case of the Hill 1948 yield criterion).

2.7.3 Mechanical Parameters Used by the Identification Procedure

of the Yield Criteria

Table 2.3 shows the mechanical parameters needed for the identification of several
yield criteria. On the basis of this list, we can estimate the amount of experimen-
tal tests and the costs required for identification of various yield criteria. The main
question of interest is whether a biaxial yield stress and biaxial anisotropy coeffi-
cient have to be determined since this requires a special apparatus: either for cross
tensile tests, hydraulic bulge tests or for disk compression, respectively.
Table 2.3 makes reference only to plane-stress models (2D). The following nota-
tions have been used in the table: 3D signifies the fact that the model is extendable
to spatial stress states; A1 shows that the yield criterion is able to describe ‘the
first order anomalous behaviour’ (see [26]); A2 shows that the yield criterion is
able to describe ‘the second order anomalous behaviour’ (see [28]). The yield cri-
teria belonging to Hershey family use an exponent chosen in accordance with the
crystallographic structure of the material.

2.7.4 Implementation of the Yield Criteria in Numerical

Simulation Programmes
The main criteria for selecting the yield function for implementation in Finite
Element code are the prediction precision of the anisotropic behaviour and the CPU
2.7 Recommendations on the Choice of the Yield Criterion 119

Table 2.3 The mechanical parameters needed for the identification of several yield criteria

Author, year σ 0 σ 30 σ 45 σ 75 σ 90 σ b r0 r30 r45 r75 r90 rb 3D A1 A2

Hill’s family
Hill 1948 x x x x x
Hill 1979 x x x x x
Hill 1990 x x x x x x
Hill 1993 x x x x x x x
Lin, Ding 1996 x x x x x x x x
Hu 2005 x x x x x x x x
Leacock 2006 x x x x x x x
Hershey’s family
Hosford 1979 x x x x x
Barlat 1989 x x x x
Barlat 1991 x x x x x x
Karafillis Boyce 1993 x x x x x x x x x
Barlat 1997 x x x x x x x x x x
BBC 2000 x x x x x x x x x x
Barlat 2000 x x x x x x x x x
Bron, Besson 2003 x x x x x x x x x x x
Barlat 2004 x x x x x x x x x x x
BBC 2005 x x x x x x x x x x x
Drucker’s family
Cazacu–Barlat 2001 x x x x x x x x x x x x x x
Cazacu–Barlat 2003 x x x x x x x x x x x x x x
C-P – B 2006 x x x x x x x x x x x x x x
Polinomial criteria
Comsa 2006 x x x x x x x x x x x
Soare 2007 (Poly 4) x x x x x x x x x x x x

time efficiency of the program. Table 2.4 presents the main commercial FE software
and the anisotropic yield criteria implemented in them. The Barlat 2000, Vegter
and BBC 2005 models have been implemented by various users in the material
subroutines of ABAQUS and LS-DYNA.

Table 2.4 Various yield criteria implemented in some commercial codes

Software Hill 1948 Hill 1990 Barlat 1989 Barlat 2000 Vegter BBC 2005

ABAQUS X X X
AUTOFORM X X X X
LS-DYNA X X X X
OPTRIS X X
PAM STAMP X X X
STAM PACK X X
120 2 Plastic Behaviour of Sheet Metal

2.7.5 Overview of the Anisotropic Yield Criteria Developing

The yield criterion proposed by Hill in 1948 [25] is still very used in industrial
applications. Its general acceptance is due to the mathematical simplicity and the
direct significance of the coefficients. The identification procedure of this yield cri-
terion uses only four mechanical parameters. As a consequence, it cannot accurately
describe the planar distribution of both the uniaxial yield stress and the uniaxial
coefficient of plastic anisotropy. In addition, it cannot capture the biaxial yield stress.
The discussion in the previous sections of the chapter has shown the stages
followed by the development of the anisotropic yield criteria:

Methods used for transforming isotropic formulations into anisotropic ones:

• inclusion of new coefficients into isotropic models: Hill 1948 [25], Hill
1979 [29], Hosford 1979 [31]
• use of linear transformations: Barlat 1991 [10], Karafillis-Boyce [21]
• representation functions: Cazacu–Barlat 2001 [75]
• geometrical methods: Vegter [80], Mollica [82]
• extension of the yield criteria outside the orthotropy axes by using
coordinate transformations: Barlat 1989 [50], Hill 1990 [39].

Identification strategies:

• analytical solutions: Hill 1948 [25], Hill 1990 [39]

• numerical procedures:
– Newton–Raphson: Barlat 1989 [50], Barlat 2000 [5], BBC 2005 [53],
BBC 2008 [91] etc.
– minimization of an error-function: BBC 2000 [51], BBC 2008 [91]
– genetic algorithms: Chaparro et al. [95].

Further information referring to the most recent anisotropic yield criteria can be
found in the synthesis papers/monographs [96–101].

2.7.6 Perspectives
As it can be seen from the previous sub-chapters, advanced yield criteria allow
accurate prediction of the anisotropic behaviour of materials. On the one hand, it is
possible to simultaneously describe both the uniaxial yield stress variation and the
anisotropic coefficient in the sheet. On the other hand, it is also possible to model
both „first and second order anisotropic behaviour anomalies’. As well as this, the
yield criteria have been extended to 3D. The asymmetry of the yield loci can be
accurately predicted, thus allowing modeling the strength differential effect specific
to materials with hexagonal close packed structure.
2.8 Modeling of the Bauschinger Effect 121

In the future the research in this field of study will be oriented towards devel-
oping new models which include special properties (superplastic materials, shape
memory materials etc.). By including the evolution of the coefficients in yield func-
tions it will be possible to predict the yield loci for nonlinear loading. Stochastic
modeling will be used for a more robust prediction of the yield loci (taking into
account the variability of the mechanical parameters). Coupling of the phenomeno-
logical models with the ones based on crystal plasticity will allow better simulation
of the parameters evolution in technological processes (these include temperature,
strain rate, strain path, structural evolution).
Therefore, the virtual process chain will be described more accurately, allowing
it to be used in real fabrication processes.

2.8 Modeling of the Bauschinger Effect

2.8.1 Reversal Loading in Sheet Metal Forming Processes

In a typical sheet metal forming process a considerable amount of material under-
goes non-proportional loading. Such as during passing over a tool radius or through
a draw bead material is subjected to the multiple bending-unbending, which means
reverse tension-compression loading over the sheet thickness. Figure 2.25 shows
the stress distribution for these two cases, calculated with the commercial FE-code
ABAQUS [102].
It is well known that the mechanical response of metals depends not only on
the current state but also on the previous deformation history. One of the phe-
nomena related to the change of mechanical properties during a non-proportional
loading is the Bauschinger effect. It is usually associated with conditions where
the yield strength of a metal decreases when the direction of strain is changed.
Figure 2.26a illustrates the Bauschinger effect for the uniaxial tension-compression

Fig. 2.25 Reversal tension-compression loading (red for the tension stress) during passing over a
tool radius (left) or through a draw bead. Arrows show the direction of drawing
122 2 Plastic Behaviour of Sheet Metal

Fig. 2.26 Schematic of the Bauschinger effect (a) and its influence on the stress evolution during
passing over a die radius (b)

test. First material hardens in tension to the stress σh and then loaded in compres-
sion. Plastic deformation occurs before negative yield strength (dashed line presents
material behavior without the Bauschinger effect). Similar reversal stress-strain his-
tory occurs when material passes over a die radius (Fig. 2.26b). During the first
bending there is a tension on the outer side of a sheet and a compression inside.
During the second unbending the stress state over the sheets thickness reverses and
is influenced directly by the Bauschinger effect.
Especially for a springback prediction an accurate description of the material
behavior during reversal loading is essential: on the one hand it is important to
know the exact stress distribution at the end of the forming process before unloading
starts, on the other hand it is necessary to model unloading with the proper stress-
strain response [103]. Since the magnitude of a springback depends on the yield
strength and the Young’s modulus, more proper material modeling including the
Bauschinger effect has become necessary particularly for the new materials like
high strength steels and aluminium alloys.

2.8.2 Experimental Observations

The experimental procedure in order to get cyclic tension-compression curves for
sheet metals is complex as sheet metal tends to buckle during compression loading.
This makes the measurement delicate, especially for large strains. Several tech-
niques to perform such tests have been developed recently [103–105]. Based on
these investigations one can separate the Bauschinger effect into three partial effects:
early re-plastification, transient softening and workhardening stagnation leading to
2.8 Modeling of the Bauschinger Effect 123

Fig. 2.27 Uniaxial tension and tension-compression curves with different pre-strains for DP600
steel

reduced yielding. Figure 2.27 shows tension and two tension-compression curves
at different pre-strains of 8 and 14% for the high strength steel DP600. A special
representation of the compression curves as the positive effective stress over the
accumulated true strain permits the observation of the workhardening stagnation,
which is typical for many materials and looks like a delay in the hardening for a
certain amount of strain.
Noticeable there is a non-linear character on the stress-strain curve directly after
the load reversal, so that the tangential modulus is lower than the elastic modulus.
One call this phenomena can early re-yielding or early re-plastification [104–106].

Fig. 2.28 The Bauschinger effect during reversal shear loading and the cross effect during
orthogonal loading for mild steel DC06 [107]
124 2 Plastic Behaviour of Sheet Metal

Another phenomenon related to the Bauschinger effect is the so-called reduc-

tion of the Young’s modulus depending on the pre-strain. In [104, 105] it has been
shown that the initial modulus typically reduces exponentially with the accumulated
pre-strain. In [107] not only the sign of the loading but also its direction has been
verified. During the orthogonal loading (uniaxial tension followed by simple shear)
the increase of the yield stress has been observed (see Fig. 2.28). This phenomena
is the so-colled cross effect.

2.8.3 Physical Nature of the Bauschinger Effect

Before dealing with phenomenological modeling of the Bauschinger effect, it is
advisable to get a good understanding of the physical origin of it, what may lead to
more refined plasticity models and may ultimately improve the simulation results.
As for the metal plasticity at room temperature in general, the main source for the
Bauschinger effect is a dislocation structure.
As deformation occurs, dislocations move, activating slip on the energeti-
cally favorable slip systems, and the dislocation density increases gradually.
Dislocations overlap, accumulate at obstacles producing dislocation tangles and
pile-ups (Fig. 2.29). This increases the resistance to further dislocation motions and
causes a hardening of the metal.
The Bauschinger effect can be generally ascribed to long-range effects, such as
internal stresses due to dislocation interactions, dislocation pile-ups at grain bound-
aries or Orowan loops around strong precipitates, and to short-range effects, such as
directionality of mobile dislocations in their resistance to motion or annihilation of
the dislocations during the reverse loading.

Fig. 2.29 Edge dislocation (a) and dislocation pile-ups on the grain boundaries (b)
2.8 Modeling of the Bauschinger Effect 125

The primary driving force of the Bauschinger effect can be explained by the
motion of the less stable dislocation structures such as pile-ups. Pile-up occurs
as a cluster of dislocations is unable to move past the barrier. As accumulated
dislocations generate microscopic back-stresses, they will assist the movement of
dislocations in the reverse direction and the yield strength becomes lower. This
occurs directly after the change of load direction or during unloading and takes
place simultaneously with elastic deformation. With this microscopic mechanism
one can explain such macroscopic phenomena as the transient softening, the early
re-plastification and the reduction of the Young’s modulus.
Another mechanism is, when the strain direction is reversed, dislocations of the
opposite sign can be produced from the same source that produced the slip-causing
dislocations in the initial direction. Dislocations with opposite signs can attract and
annihilate each other. Since strain hardening is related to an increased dislocation
density, reducing the number of dislocations reduces strength.
The workhardening stagnation can be explained by the partial disintegration of
the performed dislocation cell structures and the subsequent resumption of work-
hardening to the formation of new dislocation structures [105, 107].
The so-called cross effect during orthogonal loading is referred to the fact that
the dislocation structures which developed during pre-loading in a given direction
act as obstacles to slip on systems activated in the orthogonal direction after the
change of loading direction [107].
Other mechanisms beside the crystallographic slip can also macroscopically con-
tribute to the Bauschinger effect. Twinning is crucial particularly for the metals with
hexagonal close-packed lattice such as magnesium or zircon. During the cold form-
ing of the magnesium alloys the twinning under compression can occur, which leads
to the essential reduction of the yield strength. Other factors which contribute to
such material behavior on the macroscopic level could be a change of the crystallo-
graphic texture during plastic deformation, stress induced phase transformation or
porosity evolution.

2.8.4 Phenomenological Modelling

The Bauschinger effect is usually being predicted by using the anisotropic (also
known as distortional) hardening, which describes in general a transformation of the
yield surface during the plastic deformation. However, as the experimental obser-
vations show, the distortion of the yield surface is very complex and includes its
translation and change of its size and shape [108]. To describe such a behavior phe-
nomenologicaly a large amount of experiments is needed, however with reasonable
effort one can describe the translation of the yield surface (kinematic hardening) and
the change of its size (isotropic hardening, Fig. 2.30).
It is necessary to distinguish between the aspects: the word isotropic can be
related to both yield criteria and to the evolution of yield surface, while kinematic
is applied only to the evolution. A position of the yield surface will be described
with the back-stress tensor α. It is obvious, that as soon as plastic pre-strain occurs
126 2 Plastic Behaviour of Sheet Metal

Fig. 2.30 Kinematic

hardening—translation of the
yield surface with the same
shape and size

and the back-stress tensor has a value, the kinematic hardening leads to the plastic
anisotropy even if isotropic yield criterion is used.
Before starting with particular models describing kinematic hardening, let us first
refer to the classical framework usually being used for elastoplastic modeling with
isotropic hardening.
Assuming small elastic and large plastic deformations for metals, the rate of
deformation can be decomposed into elastic and plastic parts as

ε̇ = ε̇e + ε̇ p (2.309)

The stress-strain response in elastic range is assumed to be linear

σ̇ = C: ε̇e (2.310)

where C is the elasticity tensor.

The plastic flow occurs when the yield criterion is valid:

f = φ(σ ) − R − σ0 = 0 (2.311)

where σ 0 is the initial yield stress, R is the scalar function, which introduces hard-
ening and φ is the homogeneous function, which indicates the form of the yield
surface. For simplicity, let us assume the von Mises isotropic yield function derived
from the second invariant of the stress tensor
"
3
φ = J2 (σ ) = S (2.312)
2
where S is stress deviator tensor.
2.8 Modeling of the Bauschinger Effect 127

The consistency condition during the plastic flow is

∂f ∂f
ḟ = : σ̇ + Ṙ = 0 (2.313)
∂σ ∂R
The relation between plastic strains and stresses is defined by the flow rule:

∂
ε̇p = λ̇ (2.314)
∂σ
where λ is the Lagrange multiplier and φ is the plastic potential, which defines
the direction of plastic flow. From (2.314) follows the normality of ε̇p to the yield
surface for the associated flow rule ( = f ).
The equivalent plastic strain rate is defined as

∂
ṗ = −λ̇ (2.315)
∂R
For the associated flow rule and von Mises plasticity one can now derive the
plastic multiplier
"
28 8
λ̇ = ṗ = 8ε̇p 8 (2.316)
3
Finally substituting (2.315) into (2.310) one can write

∂
σ̇ = C : (ε̇ − λ̇ ) (2.317)
∂σ
Additionally, the hypothesis of plastic incompressibility (independence on the
hydrostatic stress) will be assumed:

∂
=0 (2.318)
∂(tr[σ ])

2.8.4.1 Prager’s Model

The hardening in this formulation corresponds to the pure translation of the yield
surface in stress space and will be described by the back-stress tensor α. The yield
condition in this case is
"
3
f = S − α − σ0 = 0 (2.319)
2
In the theory of Prager, it is assumed that during loading the back-stress develops
in the same direction as the plastic strain increment [108]:
128 2 Plastic Behaviour of Sheet Metal

α̇ = cε̇p (2.320)

where c is a material constant. This model is the simplest case of the pure kinematic
hardening. However, it can not accurately describe material behavior using the linear
hardening.

2.8.4.2 Model of Armstrong and Frederick

To get a better description of a hardening and to make a smoother transition from
elastic to inelastic behaviour, Armstrong and Frederick [109] suggested to enhance
the evolution equation (2.320) to one of the Bailey-Orowan type:
 
2
α̇ = c aε̇p − α ṗ (2.321)
3

where c and a are the material parameters. The additional recovery term with the
accumulated plastic strain p describes a kind of memory effect and leads to an
exponential evolution character of the back-stress, which corresponds much better
to experimental observations.

2.8.4.3 Chaboche’s Model

The previous models described the hardening only by the translation of the yield
surface. With further improvement, Chaboche presented a mixed hardening rule as
a superposition of nonlinear isotropic and kinematic hardening [108, 110].
Then the yield criterion is
"
3
f = S − α − R − σ0 = 0 (2.322)
2
with the back-stress
2
α̇ = c( aε̇p − α ṗ) (2.323)
3
and the isotropic part of the hardening

Ṙ = b(Q − R)ṗ (2.324)

where c, a, b and Q are the material constants.

As it has been mentioned in [110] one non-linear term of the back-stress is not
sufficient for the description of large-strain hardening. To improve it one can use
several back-stresses simultaneously

M
+
α= α (m) (2.325)
m=1
2.8 Modeling of the Bauschinger Effect 129

2
α̇ (m) = c(m) ( a(m) ε̇p − α (m) ṗ) (2.326)
3
including one linear

2
α̇ (M) = H ε̇p (2.327)
3
The model of Chaboche with many back-stress components is able to describe
the transient softening quite accurately. However, the phenomena of the early re-
plastification and the workhardening stagnation are still uncovered.

2.8.4.4 Yoshida–Uemori Model

The model of Yoshida and Uemori [105, 111] is a further improvement of the
Chaboche’s model which additionally describes the early re-plastification and the
workhardening stagnation.
This model assumes two surfaces: the yield surface f develops within the
bounding surface F. Based on the wide experimental investigations Yoshida and
co-workers apply the pure kinematic hardening for the yield surface since the early
re-plastification occurs at a very early stage of load reversal. At the same time the
bounding surface has mixed isotropic-kinematic hardening:
"
3
f = S − α − σ0 = 0 (2.328)
2
"
3
F= S−β −R−B=0 (2.329)
2
where β denotes the center of the bounding surface, and B and R are its initial size
and isotropic hardening.
The flow rule is defined as
∂f
ε̇p = λ̇ (2.330)
∂σ
One can write the back-stress as the sum

α = α∗ + β (2.331)

where α ∗ is the relative motion of the yield surface with respect to the bounding
surface
" "
2 ᾱ∗
α̇ ∗ = ca(np − n∗ )ṗ (2.332)
3 a
in terms of the non-dimensional measures
130 2 Plastic Behaviour of Sheet Metal

S−α α
np = , n∗ = (2.333)
S−α α
with
" "
28 8 3
ṗ = 8ε̇p 8 , ᾱ∗ = α∗ , a = B + R − σ0 (2.334)
3 2
Kinematic hardening for the bounding surface is
'" (
2
β̇ = m bnp − β ṗ (2.335)
3

and its isotropic hardening

Ṙ = m(Rsat − R)ṗ (2.336)

To describe the workhardening stagnation Yoshida and Uemori proposed to mod-

ify the evolution of the bounding surface by providing a new surface g in the stress
space
"
3
g= S−q −r =0 (2.337)
2
where q and r denote the center and the size of the surface.
It will be assumed that the center of the bounding surface β exists either on or
inside of the surface g and the isotropic hardening of the bounding surface R takes
place when β stays on the surface g, namely Ṙ > 0 when
"
3 ∂g
g= β − q − r = 0 and : β̇ > 0 (2.338)
2 ∂β

and Ṙ = 0 otherwise (see Fig. 2.31).

The kinematic motion of the surface g is defined as

Fig. 2.31 Schematic

illustration of the surface g in
stress space
2.8 Modeling of the Bauschinger Effect 131

q̇ = μ(β − q) (2.339)

with

Ŵ − ṙ 3(β − q) : β̇
μ= andŴ = (2.340)
r 2r

the evolution of r is

ṙ = hŴ when Ṙ > 0
(2.341)
ṙ = 0 when Ṙ = 0

where h(0 ≤ h ≤ 1) is a material parameter.

After analysing the ability of the above model to describe experimental results
it was found that the early re-plastification phenomenon can not be covered well
enough. The proposed solution is to use the variable Young’s modulus as a function
of the accumulated plastic strain p:

E = E0 − (E0 − Ea ) 1 − e−ξ p
 
(2.342)

where E0 and Ea are the values of Young’s modulus for virgin and infinitely large
pre-strained material, respectively, and ξ is a material parameter. For the definition
of the Young’s modulus Yoshida suggests an average value in the stress range 0 ≤
σ ≤ 0.95σr , where σr indicates the stress reversal point.

2.8.4.5 AutoForm-Model
To improve the springback prediction a novel approach to model the Bauschinger
effect has been developed and implemented in the commercial code AutoForm
[112]. The main idea of the model is to use the same evolution equation for the
entire unloading and reverse loading path, including the area, which is treated as an
elastic in conventional models. As the model is part of an undisclosed research, the
principal of it will be presented here for the uniaxial tension-compression case using
fictive values of the reversal stress σr and the reversal strain εr with the coordinate
center in the point of load reversal
 2
σr σr
εr = + K arctan h2 (2.343)
El (p) 2σh (p)

with σh (p) as hardening stress used to describe tension curve (see Fig. 2.32) and
El as the initial Young’s modulus at the moment of load reversal. Analog to the
Yoshida-Uemori model El is a function of the accumulated plastic strain p:

El = E 1 − γ 1 − e−χp
  
(2.344)
132 2 Plastic Behaviour of Sheet Metal

Fig. 2.32 Schematic of the

tension-compression curve
with reversal stress and strain

where γ and χ are two material parameters and E is the elastic modulus.
One can then write the definition for tension and compression parts of the curve

σh (p) for tension
σ = (2.345)
σh (p) − σr for compression

On the contrary to the previous models, here the evolution equations for the hard-
ening stress and for the back-stress are much more decoupled. The definition of
the hardening during proportional loading is independent from the definition of the
back-stress and hardening curve can be defined even in tabular form.
As in the model of Yoshida-Uemori, the pure kinematic hardening is assumed
here. However, the model in AutoForm uses the non-linear stress-strain response
within the yield surface and can describe the early re-plastification very accurately.
In order to model the workhardening stagnation, the accumulated equivalent plas-
tic strain p is replaced by a new hardening parameter pd , which behaves as follows:
pd is identical with p during proportional deformation and develops slower than p
during reverse or non-proportional deformation. To determine if a deformation is
reverse or non-proportional, a storage surface g is introduced (Fig. 2.33). During
the initial proportional loading (time t), the progression of plastic strain extends the
storage surface and the strain-like variable q lies on the surface g. During the further
reverse or non-proportional loading (time t + t), the tensor q traverses the storage
surface; during the time it is inside, the variable pd develops slower than p. This
behavior is controlled by a material parameter ξ (0 ≤ ξ ≤ 1), which characterizes
the fraction of forward strain that can be reversed with the delay of work hardening.

The set of equations is:

ṗd = f · ṗ (2.346)
2.8 Modeling of the Bauschinger Effect 133

Fig. 2.33 Schematic

illustration of the storage
surface g

f = sign(f̃ )f̃ 2 (2.347)

3r
f̃ = 1 − (2.348)
4g
In the beginning is g = 0 and growths proportionally to ṗd :

ġ = ξ ṗd (2.349)

r in Eq. (2.348) is a scalar value

r = φ (g − q) (2.350)

where q is a tensor representing the position within the surface g

 −1
∂r
q̇ = ġ + ṗ (2.351)
∂g

and
ġ
ġ = σ (2.352)
ϕ(σ )

Considering Eqs. (2.350) and (2.351) and since q can not exceed the surface g,
one derive the rate of r as

ṙ = −ṗ when r > 0

(2.353)
ṙ = 0 when r = 0

Finally, assuming all the discussed above phenomena, the model is able to
describe accurately experimental tension-compression curves. Figure 2.34a shows
the uniaxial tension and tension-compression data for the high-strength steel DP600
134 2 Plastic Behaviour of Sheet Metal

Fig. 2.34 Uniaxial tension and tension-compression curves (a) and evolution of the tangential
modulus (b) during compression for DP600 steel calculated with AutoForm-model together with
experimental results

Table 2.5 Kinematic

hardening parameters for K ξ (–) γ (–) χ El (GPa)
DP600 steel
0.011 1 0.13 30 200
References 135

for two tests with different pre-strains of 8 and 14%. The second picture on the fig-
ure presents the evolution of tangent modulus starting from the reversal point, which
allows to see the non-linear character of the stress-strain response directly after the
load reversal.
Table 2.5 presents the model parameters for the DP600 steel. For the description
of tensile hardening curve tabulated data from the tension test has been used.

References
1. Lankford WI, Snyder SC, Bauscher JA (1950) New criteria for predicting the press
performance of deep-drawing sheets. Transaction ASM 42:1196–1232
2. Wech PI, Radtke L, Bunge HJ (1983) Comparison of plastic anisotropy parameters. Sheet
Metal Industries 60:594–597
3. Siegert K, TALAT Lecture 3705 (http://www.eaa.net/eaa/education/TALAT/index.htm)
4. Choi Y, Walter ME, Lee JK, Han CS (2006) Observations of anisotropy evolution and iden-
tification of plastic spin parameters by uniaxial tensile tests. International Journal of Solids
and Structures 1:303–325
5. Barlat F, Brem JC, Yoon JW, Chung K, Dick RE, Choi SH, Pourboghrat F, Chu E, Lege DJ
(2003) Plane stress yield function for aluminium alloy sheets – Part 1: Theory. International
Journal of Plasticity 19:297–319
6. Banabic D, Wagner S (2002) Anisotropic behaviour of aluminium alloy sheets. Aluminium
78:926–930
7. Pöhlandt K, Banabic D, Lange K (2002) Equi-biaxial anisotropy coefficient used to describe
the plastic behavior of sheet metal. Proceedings of the ESAFORM Conference, Krakow,
723–727
8. Findley WN, Michno MJ (1976) A historical perspective of yield surface investigations for
Metals. International Journal of Non-Linear Mechanics 11:59–80
9. Zyczkowski M (1981) Combined loadings in the theory of plasticity. Polish Scientific
Publishers, Warsaw
10. Barlat F, Lege DJ, Brem JC (1991) A six-component yield function for anisotropic materials.
International Journal of Plasticity 7:693–712
11. Tresca H (1864) On the yield of solids at high pressures. Comptes Rendus Academie des
Sciences 59:754 (in French)
12. Huber MT (1904) Przyczynek do podstaw wytorymalosci. Czasopismo Techniczne 22:
34–81
13. Mises R (1913) Mechanics of solids in plastic state. Göttinger Nachrichten Mathematical
Physics 4:582–592 (in German)
14. Hencky H (1924) On the theory of plastic deformations. Zeitschrift für Angewandte
Mathematik und Mechanik 4:323–334 (in German)
15. Timoshenko SP (1953) History of strenght of materials. McGraw-Hill, New York, NY
16. Drucker DC (1949) Relations of experiments to mathematical theories of plasticity. Journal
of Applied Mechanics 16:349–357
17. Norton FH (1929) The creep of steel at high temperatures. McGraw-Hill, New York, NY
18. Bailey RW (1929) Creep of steel under simple and compound stresses and the use of high
initial temperature in steam power plants. Transmission in Tokyo Section Meeting World
Power Conference, Konai-kai Publishing, Tokyo
19. Hershey AV (1954) The plasticity of an isotropic aggregate of anisotropic face centred cubic
crystals. Journal of Applied Mechanics 21:241–249
20. Hosford WF (1972) A generalised isotropic yield criterion. Journal of Applied Mechanics
39:607–609
136 2 Plastic Behaviour of Sheet Metal

21. Karafillis AP, Boyce MC (1993) A general anisotropic yield criterion using bounds and a
transformation weighting tensor. Journal of the Mechanics and Physics of Solids 41:1859–
1886
22. Yu MH (2002) Advances in strength theories for materials under complex stress state in the
20th century. Applied Mechanics Reviews 55:198–218
23. von Mises RV (1928) Mechanics of plastic deformation of crystals. Zeitschrift für
Angewandte Mathematik und Mechanik 8:161–185 (in German)
24. Olszak W, Urbanowski W (1956) The orthotropy and the non-homogeneity in the theory of
plasticity. Polska Archiwum Mechaniki Stosowanej 8:85–110
25. Hill R (1948) A theory of the yielding and plastic flow of anisotropic metals. Proceedings of
the Royal Society London A 193:281–297
26. Woodthrope J, Pearce R (1970) The anomalous behaviour of aluminium sheet under
balanced biaxial tension. International Journal of Mechanical Sciences 12:341–347
27. Pearce R (1968) Some aspects of anisotropic plasticity in sheet metals. International Journal
of Mechanical Sciences 10:995–1001
28. Banabic D, Müller W, Pöhlandt K (1998) Determination of yield loci from cross tensile
tests assuming various kinds of yield criteria. Sheet Metal Forming Beyond 2000, Brussels,
343–349
29. Hill R (1979) Theoretical plasticity of textured aggregates. Mathematical Proceedings of the
Cambridge Philosophical Society 85:179–191
30. Lian J, Zhou D, Baudelet B (1989) Application of Hill’s new theory to sheet metal forming
– I. Hill’s 1979 criterion and its application to predicting sheet forming limits. International
Journal of Mechanical Sciences 31:237–244
31. Hosford WF (1979) On yield loci of anisotropic cubic metals. In: Proceedings of the 7th
North American Metalworking Conference (NMRC), SME, Dearborn, MI, 191–197
32. Chu E (1995) Generalization of Hill’s 1979 anisotropic yield criteria. Journal of Materials
Processing Technology 50:207–215
33. Zhou W (1990) A new non-quadratic orthotropic yield criterion. International Journal of
Mechanical Sciences 32:513–520
34. Müller W (1996) Characterization of sheet metal under multiaxial load. Berichte aus dem
Institut für Umformtechnik, Universität Stuttgart, Nr. 123, Springer, Berlin (in German)
35. Bassani JL (1977) Yield characterisation of metals with transversally isotropic plastic
properties. International Journal of Mechanical Sciences 19:651–654
36. Bishop JFW, Hill R (1951) A theory of the plastic distortion of polycrystalline agregates
under combined stress. Philosophical Magazine 42:414–427
37. Zhou W (1994) A new orthotropic yield function describable anomalous behaviour of
materials. Transactions of Nonferrous Metals Society of China 4:431–449
38. Montheillet F, Jonas JJ, Benferrah M (1991) Development of anisotropy during the cold
rolling of aluminium sheet. International Journal of Mechanical Sciences 33:197–209
39. Hill R (1990) Constitutive modelling of orthotropic plasticity in sheet metals. Journal of the
Mechanics and Physics of Solids 38:405–417
40. Lin SB, Ding JL (1996) A modified form of Hill’s orientation-dependent yield criterion for
orthotropic sheet metals. Journal of the Mechanics and Physics of Solids 44:1739–1764
41. Chung K, Shah K (1992) Finite element simulation of sheet metal forming for planar
anisotropic metals. International Journal of Plasticity 8:453–476
42. Leacock AG (2006) A mathematical description of orthotropy in sheet metals. Journal of the
Mechanics and Physics of Solids 54:425–444
43. Hill R (1993) A user-friendly theory of orthotropic plasticity in sheet metals. International
Journal of Mechanical Sciences 15:19–25
44. Stout MG, Hecker SS (1983) Role of geometry in plastic instability and fracture of tubes
sheet. Mechanics of Materials 2:23–31
45. Logan R, Hosford WF (1980) Upper-bound anisotropic yield locus calculations assuming
(111) – Pencil glide. International Journal of Mechanical Sciences 22:419–430
References 137

46. Hosford WF (1985) Comments on anisotropic yield criteria, International Journal of

Mechanical Sciences 27:423–427
47. Hosford WF (1993) The mechanics of crystals and textured polycrystals. Oxford University
Press, Oxford
48. Hosford WF (1996) On the crystallographic basis of yield criteria. Texture and
Microstructures 26–27:479–493
49. Barlat F, Richmond O (1987) Prediction of tricomponent plane stress yield surfaces and asso-
ciated flow and failure behaviour of strongly textured FCC polycrystalline sheets. Materials
Science and Engineering 91:15–29
50. Barlat F, Lian J (1989) Plastic behaviour and stretchability of sheet metals (Part I): A
yield function for orthotropic sheet under plane stress conditions. International Journal of
Plasticity 5:51–56
51. Banabic D, Balan T, Comsa DS (2000) A new yield criterion for orthotropic sheet metals
under plane-stress conditions. Proceedings of the 7th Conference ‘TPR2000’, Cluj Napoca,
Romania, 217–224
52. Banabic D, Kuwabara T, Balan T, Comsa DS, Julean D (2003) Non-quadratic yield cri-
terion for orthotropic sheet metals under plane-stress conditions. International Journal of
Mechanical Sciences 45:797–811
53. Banabic D, Aretz H, Comsa DS, Paraianu L (2005) An improved analytical description of
orthotropy in metallic sheets. International Journal of Plasticity 21:493–512
54. Choi SH et al. (1999) Prediction of yield surfaces of textured sheet metals. Metallurgical and
Materials Transactions 30A:377–379
55. Lian J, Chen J (1991) Isotropic polycrystal yield surfaces of BCC and FCC met-
als: Crystallographic and continuum mechanics approaches. Acta Metallurgica 39:2285–
2294
56. Barlat F et al. (1997) Yielding description for solution strengthened aluminium alloys.
International Journal of Plasticity 13:185–401
57. Hayashida Y et al. (1995) FEM analysis of punch stretching and cup drawing tests for
aluminium alloys using a planar anisotropic yield function. In: Shen SF, Dawson PR
(eds) Simulation of materials processing-theory, methods and applications. AA Balkema,
Rotterdam, 712–722
58. Barlat F et al. (1997) Yield function development for aluminium alloy sheets. Journal of the
Mechanics and Physics of Solids 45:1727–1763
59. Yoon JW et al. (1999) A general elasto-plastic finite element formulation based on incre-
mental deformation theory for planar anisotropy and its application to sheet metal forming.
International Journal of Plasticity 15:35–67
60. Bron F, Besson J (2004) A yield function for anisotropic materials. Application to aluminium
alloys. International Journal of Plasticity 20:937–963
61. Wang D-A, Pan J (2006) A non-quadratic yield function for polimeric foams. International
Journal of Plasticity 22:434–458
62. Budiansky B (1984) Anisotropic plasticity of plane-isotropic sheets. In: Dvorak GJ, Shield
RT (eds) Mechanics of material behavior. Elsevier, Amsterdam, 15–29
63. Tourki Z et al. (1994) Orthotropic plasticity in metal sheets. Journal of Materials Processing
Technology 45:453–458
64. Gotoh M (1977) A theory of plastic anisotropy based on a yield function of fourth order.
International Journal of Mechanical Sciences 19:505–520
65. Hill R (1950) The mathematical theory of plasticity. Oxford University Press, Oxford
66. Barlat F et al. (2000) Constitutive modeling for aluminium sheet forming simulations.
In: Khan AS, Zhang H, Yuan Y (eds) Plastic and viscoplastic response of materials and
metal forming. Proceedings of the 8th International Symposium in Plasticity and its Current
Applications, Whistley, Canada, Neat Press, Fulton, MD, 591–593
67. Aretz H, Barlat F (2004) General orthotropic yield function based on linear stress deviator
transformations. In: Ghosh S, Castro GC, Lee JK (eds) Materials processing and design:
138 2 Plastic Behaviour of Sheet Metal

Modelling, simulation and applications. Proceedings of the NUMIFORM 2004 Conference,

Columbus, OH, 147–151
68. Barlat F et al. (2005) Linear transformation-based anisotropic yield functions. International
Journal of Plasticity 21:1009–1039
69. Yoon JW et al. (2006) Prediction of six or eight ears in a drawn cup based on a new
anisotropic yield function. International Journal of Plasticity 22:174–193
70. Cosovici GA (2006) Implementation of the new yield criteria in the FE programs for sheet
metal forming simulation. PhD Thesis, Cluj Napoca, Romania (in Romanian)
71. Paraianu L (2006) Modelling of the FLC using the large deformation theory. PhD Thesis,
Cluj Napoca, Romania (in Romanian)
72. Press WH et al. (1992) Numerical recipes in C. The art of scientific computing. Cambridge
University Press, Cambridge
73. Aretz H (2005) A non-quadratic plane stress yield function for orthotropic sheet metals.
Journal of Materials Processing Technology 168:1–9
74. Barlat F, Yoon JW, Cazacu O (2006) On linear transformations of stress tensors for the
description of plastic anisotropy. International Journal of Plasticity 22:876–896
75. Cazacu O, Barlat F (2001) Generalization of Drucker’s yield criterion in orthotropy.
Mathematics and Mechanics of Solids 6:613–630
76. Cazacu O, Barlat F (2003) Application of representation theory to describe yielding
of anisotropic aluminium alloys. International Journal of Engineering Science 41:1367–
1385
77. Cazacu O, Barlat F (2004) A criterion for description of anisotropy and yield differential
effects in pressure-insensitive metals. International Journal of Plasticity 20:2027–2045
78. Liu C, Huang Y, Stout MG (1997) On the asymmetric yield surface of plastically orthotropic
materials: A phenomenological study. Acta Materialia 45:2397–2406
79. Cazacu O, Plunkett B, Barlat F (2006) Orthotropic yield criterion for hexagonal close packed
metals. International Journal of Plasticity 22:1171–1194
80. Vegter H, Drent P, Huetink J (1995) A planar isotropic yield criterion based on material
testing at multi-axial stress state. In: Shen SF, Dawson PR (eds) Simulation of materials
processing-theory, methods and applications. AA Balkema, Rotterdam, 345–350
81. Vegter D, van den Boogaard AH (2006) A plane stress yield function for anisotropic sheet
material by interpolation of biaxial stress states. International Journal of Plasticity 22:
557–580
82. Mollica F, Srinivasa AR (2002) A general framework for generating convex yield surfaces
for anisotropic metals. Acta Mechanica 154:61–84
83. Hu W (2003) Characterized behaviors and corresponding yield criterion of anisotropic sheet
metals. Materials Science and Engineering A345:139–144
84. Hu W (2005) An orthotropic yield criterion in a 3-D general stress state. International Journal
of Plasticity 21:1771–1796
85. Hu W (2007) A novel quadratic yield model to describe the feature of multi-yield-surface of
rolled sheet metals. International Journal of Plasticity 23:2004–2028
86. Comsa DS (2006) Numerical simulation of the sheet metal forming prodesses using a new
yield criterion. PhD Thesis, Tehnical University of Cluj Napoca (in Romanian)
87. Comsa DS, Banabic D (2007) Numerical simulation of sheet metal forming processes using
a new yield criterion. Key Engineering Materials 344:833–840
88. Soare S (2007) On the use of homogeneous polynomials to develop anisotropic yield
functions with applications to sheet metal forming. PhD Thesis, University of Florida
89. Soare S, Yoon JW, Cazacu O (2007) On using homogeneous polynomials to design
anisotropic yield functions with tension/compression symmetry/asymmetry. In: Cesar de
Sa JMA, Santos AD (eds) Materials processing and design: Modeling, simulation and
applications. Proceedings of the NUMIFORM 2007 Conference, Porto, 607–612
90. Oller S, Car E, Lubliner J (2003) Definition of a general implicit orthotropic yield criterion.
Computer Methods in Applied Mechanics and Engineering 192:895–912
References 139

91. Comsa DS, Banabic D (2008) Plane-stress yield criterion for highly-anisotropic sheet
metals. In: Hora P (ed) Proceedings of the 7th International Conference and Workshop
on Numerical Simulation of 3D Sheet Metal Forming Processes, NUMISHEET 2008,
Interlaken, Switzerland, 43–48
92. Banabic D, Cazacu O, Barlat F, Comsa DS, Wagner S, Siegert K (2002) Recent anisotropic
yield criteria for sheet metals. Proceedings of the Romanian Academy 3:91–99
93. Barlat F, Banabic D, Cazacu O (2002) Anisotropy in sheet metals. In: Yang D-Y, Oh SI,
Huh H, Kim YH (eds) Design inovation through virtual manufacturing. Proceedings of the
NUMISHEET 2002 Conference, Jeju Island, Korea, 515–524
94. Paraianu L, Banabic D (2006) Predictive accuracy of different yield criteria. Proceedings of
the SISOM Conference, Bucharest, 465–574
95. Chaparro BM, Alves JL, Menezes LF, Fernandes JV (2007) Optimization of the phe-
nomenological constitutive models parameters using genetic algorithms. In: Banabic D (ed)
Advanced methods in material forming, Springer, Heidelberg, 35–54
96. Barlat F, Cazacu O, Zyczkowski M, Banabic D, Yoon J-W (2004) Yield surface plasticity
and anisotropy. In: Raabe D, Chen L-Q, Barlat F, Roters F (eds) Continuum scale simulation
of engineering materials fundamentals-microstructures-process applications. Wiley-VCH,
Weinheim, 145–185
97. Banabic D, Tekkaya EA (2006) Forming Simulation. In: Hirsch J (ed) Virtual fabrication of
aluminum alloys: Microstructural modeling in industrial aluminum production. Wiley-VCH,
Weinheim, 275–302
98. Yoon JW, Barlat F (2006) Modeling and simulation of the forming of aluminium sheet
alloys. In: Semiatin SL (ed) ASM handbook, Vol 14B, Metalworking: Sheet forming. ASM
International, Materials Park, OH, 792–826
99. Barlat F (2007) Constitutive modeling for metals. In: Banabic D (ed) Advanced methods in
material forming, Springer, Heidelberg, 1–18
100. Banabic D, Barlat F, Cazacu O, Kuwabara T (2007) Anisotropy and formability. In:
Chinesta F, Cueto E (eds) Advances in material forming-ESAFORM 10 years on. Springer,
Heidelberg, 143–173
101. Barlat F (2007) Constitutive descriptions for metal forming simulations, In: Cesar de
Sa JMA, Santos AD (eds) Materials processing and design: Modeling, simulation and
applications. Proceedings of the NUMIFORM 2007 Conference, Porto, 3–22
102. ABAQUS (2004), Analysis user’s manual, Version 6.4. Hibbit, Karlsson & Sorensen Inc.,
Providence, RI
103. Krasovskyy A (2005) Verbesserte Vorhersage der Rückfederung bei der Blechumformung
durch weiterentwickelte Werkstoffmodelle. PhD Thesis, University of Karlsruhe (in
German)
104. Yoshida F, Urabe M, Toropov VV (1998) Identification of material parameters in constitu-
tive model for sheet metals from cyclic bending test. International Journal of Mechanical
Sciences 40:237–249
105. Yoshida F, Uemori T, Fujiwara K (2002) Elastic-plastic behavior of steel sheets under
in-plane cyclic tension-compression at large strain. International Journal of Plasticity 18:
633–659
106. Cleveland RM, Ghosh AK (2002) Inelastic effects on springback in metals. International
Journal of Plasticity 18:769–785
107. Peeters B, Kalidindi SR, Teodosiu C, Van Houtte P, Aernoudt E (2002) A theoretical inves-
tigation of the influence of dislocation sheets on evolution of yield surfaces in single-phase
B.C.C. polycrystals. Journal of the Mechanics and Physics of Solids 50:783–807
108. Lemaitre J, Chaboche JL (1990) Mechanics of solid materials. Cambridge University Press,
Cambridge
109. Armstrong PJ, Frederick CO (1996) A mathematical representation of the multiaxial
Bauschinger effect. GEGB Report RD/B/N731. Berkeley Nuclear Laboratories
140 2 Plastic Behaviour of Sheet Metal

110. Chaboche JL, Rousselier G (1983) On the plastic and viscoplastic constitutive equa-
tions – Part 1: Rules developed with internal variable concept. Journal of Pressure Vessel
Technology 105:153–158
111. Yoshida F, Uemori T (2002) A model of large-strain cyclic plasticity describing the
Bauschinger effect and workhardening stagnation. International Journal of Plasticity
18:661–686
112. Kubli W, Krasovskyy A, Sester M (2008) Advanced modeling of reverse loading effects
for sheet metal forming processes. In: Hora P (ed) Proceedings of the 7th International
Conference and Workshop on Numerical Simulation of 3D Sheet Metal Forming Processes,
NUMISHEET 2008, Interlaken, Switzerland, 479–484
Chapter 3
Formability of Sheet Metals

List of Special Symbols

a, b, c material constants in the Cayssials model

D blank diameter
d cup diameter
dm mean cup diameter
E0 coefficient in the eMMFC model
eu ultimate strain
F formability index; Force
f, f0 , f1 yield function; non-homogeneity coefficients in MK and HN
models
g11,22,12 partial derivatives
k material constant in the Hollomon and Swift strain-hardening
laws
m strain-rate sensitivity index
n hardening coefficient
p exponent in the eMMFC model
r normal anisotropy (Lankford) coefficient
s thickness
s0 initial thickness
t time parameter
T11,22, 12, 21 tensors
Y yield function
α loading ratio
ß drawing ratio
β strain rate ratio
(t, ε, σ ) increment of time, strain or stress
ε logarithmic strain
ε̄ effective strain
ε̇ effective strain rate
ε̇11,22 principal strain rates
ε̇12, 21 angular strain rates

D. Banabic, Sheet Metal Forming Processes, DOI 10.1007/978-3-540-88113-1_3, 141


C Springer-Verlag Berlin Heidelberg 2010
142 3 Formability of Sheet Metals

ε0 pre-deformation in the Swift strain-hardening

law
ε 1,2 major and minor strains
ε3 principal strain in sheet thickness direction
ε ∗ 10 limit strain corresponding to the plane strain
ε ∗ 11, 22 major and minor limit strains
ζ principal stress ratio
μp,d friction coefficients
ϕ angle of the geometrical non-homogeneity
ρ ratio of incremental strains or principal strain rates
σ max maximum principal stress
σ e , σ̄ effective stress
σ stress
σ 1,2 principal stresses

3.1 Introduction
The formability is the capability of sheet metal to undergo plastic deformation to a
given shape without defects. The defects have to be considered separately for the
fundamental sheet metal forming procedures of deep-drawing and stretching. The
difference between these types of stamping procedures is based on the mechanics of
the forming process. For deep-drawing, the usual defects of the produced parts are
presented in Fig. 3.1 [1].
Some of these defects are caused by the forming tools (types 5, 9, 10, 14), by
the friction regime (types 4, 13) or by the mechanical and metallurgical properties
of the material as well as by geometrical parameters (types 1, 2, 3, 6, 7, 8, 11,
12). Only the defects of type 3, 6, 8 are related to stretching processes, the others
are specific to deep-drawing. Concerning the defect type 3 (e.g. in hemispherical
punch stretching), the tear is oriented along the circumference and located near the
pole. Tearing is usually preceded by strain localization (necking) which causes a
reduction of the part’s strength, worsens its appearance and is a reason for rejecting
it. Necking, tearing, wrinkling, modification of the roughness or a poor appearance
are factors that generally define a limit to the deformation by stretching (see also
[2–8]).
Necking is a limiting criterion not only for stretching but also for other processes
leading to similar strain states in the plastic zone (Fig. 3.2).
Figure 3.3 demonstrates that formability is a complex characteristic.
As an example, the influence of various parameters on the formability in deep
drawing is presented in Fig. 3.4.
The formability, expressed by the drawing ratio ß = D/d (D being the blank
diameter and d the cup diameter, depends on the strain limiting criterion as well as
on a process parameter (blank holding pressure p).
3.1 Introduction 143

Fig. 3.1 Defects in deep-drawing: 1—flange wrinkling; 2—wall wrinkling; 3—part wrinkling;
4—ring prints; 5—traces; 6—orange skin; 7—Lüders strips; 8—bottom fracture; 9—corner
fracture; 10, 11, 12—folding; 13, 14—corner folding [1]
144 3 Formability of Sheet Metals

Fig. 3.2 Examples of processes where deformation is limited by necking [7]

Marciniak [4–8] developed the idea of Cockcroft and Latham [9] by introducing
formability indices depending on the type of the simulating test as well as the strain
bounding criterion.
Of course, these indices depend also on the mechanical parameters of the
material:

Fkl = F(n, m, r, eu , f ) (3.1)

where n is the hardening coefficient, m the strain rate sensitivity, r the anisotropy
coefficient, eu the ultimate strain, and f the non-homogeneity coefficient.
The increment of the formability index may be expressed as

∂F ∂F ∂F ∂F ∂F
dF = dn + dm + dr + dεu + df (3.2)
∂n ∂m ∂r ∂εu ∂f

Each derivative in this equation can be determined theoretically or experimen-

tally for a given sheet forming process [5]. Figure 3.5 shows the factors influencing
the formability index for some important sheet forming processes. In this figure, μp
and μd are the friction coefficients at the interface between blank and punch and
blank and die, respectively.
3.1 Introduction 145

Fig. 3.3 Parameters influencing sheet metal formability

146 3 Formability of Sheet Metals

Fig. 3.4 Influence of various parameters on formability in deep-drawing [6]

Fig. 3.5 Factors influencing the formability index for the most important sheet metal forming
processes [7]
3.2 Evaluation of the Sheet Metal Formability 147

3.2 Evaluation of the Sheet Metal Formability

At the end of the nineteenth century, due to the development of the sheet forming
technology, sheet metal formability became a research topic. A detailed presen-
tation of this research can be found in [10–13]. Some of the first researchers
interested in this field were Bessemer and Parkes, Adamson, Considere [14] and
Erichsen [15].
Since then, various method for evaluating sheet metal formability have been
developed. One may subdivide them into four classes:

• simulating tests
• methods based on mechanical tests
• method of the limiting dome height
• methods based on forming limit diagrams.

3.2.1 Methods Based on Simulating Tests

Sheet metal forming processes can be characterized by two basic types of deforma-
tion patterns: drawing and bending [16]. Concerning drawing one has to distinguish
stretching (ε1 > 0; ε2 > 0 ) and deep drawing (ε1 > 0; ε2 < 0), see Fig. 3.6.
Various formability tests specific to each of these deformation patterns have been
developed. In the case of drawing processes there are simulating tests for stretching,
deep-drawing and combined drawing. A detailed presentation of the simulating tests
is given in [10, 13, 17–32].

Fig. 3.6 Deformation patterns specific to sheet metal forming processes

148 3 Formability of Sheet Metals

3.2.1.1 Typical Punch-Stretching Methods

The first test for evaluating the formability of sheet metals was proposed by Erichsen
[15]. The test consists in stretching a sheet specimen by means of a hemispherical
punch until the occurrence of fracture, see Fig. 3.7.
The depth of the punch indentation in the specimen expressed in millimetres is
the so-called Erichsen index (IE). This is the most commonly used parameter for
expressing the formability of sheet metals. Various researchers have analyzed the
accuracy, the limits of applicability as well as the factors influencing the test results
[12, 25, 33–37]. Olsen [38] introduced a test similar to that proposed by Erichsen
but with a different size of the tools.
The index established using these methods has a low accuracy due to the small
size of the tools, the impossibility of ensuring a sound fixing of the specimen and
poor lubrication.
In order to diminish these errors Hecker [39] proposed a formability test based
on stretching a punch having a greater diameter than that in the Erichsen and Olsen
tests. In this case the blank is also held by draw-beads whereby there is a dry fric-
tion regime between punch and specimen. Like for the Erichsen and Olsen test the
formability index is expressed by the depth of the punch indentation at the beginning
of fracture.
Since most of the defects in deep-drawing of complex parts are located near the
region on plane stretching (plane strain), Ghosh [40] proposed a modification of the
Hecker test in the order of simulating this state of strain. For this purpose he used
Hecker’s device with strip specimens of various widths. This test, called the limit
dome height test (LDH), will be described in Sect. 3.3.2.

Fig. 3.7 Scheme of Erichsen

test
3.2 Evaluation of the Sheet Metal Formability 149

Jovignot [41] proposed a testing method based on the hydraulic bulge test. Siebel
and Pomp [42] developed a test consisting of deep-drawing a circular specimen
with a central circular hole until the occurrence of fracture. The formability index is
expressed as the diameter of the hole at the initiation of fracture.

3.2.1.2 Typical Deep-Drawing Methods

Sachs [43] proposed a test based on tensioning a specimen having the shape of a
wedge. This simulates the straining of a longitudinal strip as a part of a cup during
a deep-drawing process.
The research initiated by Sachs was continued by Swift [44] and Guyot [45]
who proposed different shapes of the specimens. Swift also established a linear
relationship between the maximum traction load and the drawing ratio.
The advantage of wedge drawing tests is the simple shape of the specimen and the
suitability of general-purpose machines for carrying out the tests. The outstanding
disadvantage is the fact that the mechanics of straining is rather different from that
of real deep-drawing processes since neither bending nor the effect of friction are
simulated in the test.
Chung and Swift [46] published the first coherent and complete mathematical
model of the deep-drawing process which also was confirmed by experiments [47,
48]. Swift proposed a deep-drawing test method of his own. This test consists of
deep-drawing cylindrical parts having different diameters (Fig. 3.8) and determining
the limit drawing ratio LDR, i.e. the ratio of the maximum diameter of the workpiece
that is drawn without tearing and the diameter of the punch:

Dmax
LDR = (3.3)
d

Fig. 3.8 Swift’s cup-drawing

test
150 3 Formability of Sheet Metals

Sometimes a percent expression of the LDR is used:

Dmax − d
100[%] (3.4)
Dmax

Swift’s method has been widely used and is considered as a standard test by
the International Deep-Drawing Research Group (IDDRG) [16, 29]. However, the
efforts of Swift and co-workers in order to find a unique formability index which can
be obtained by a single test have been unsuccessful. Further research was carried
out by Willis [49]. Siebel and co-workers analyzed the factors such as lubrication,
drawing speed influencing the limit drawing ratio [13]. Fukui [50] proposed a deep-
drawing test using a conical die. The advantage of this method is that the ‘diameter
ratio’ D/D0 (D = upper diameter of the part at fracture) as a measure of formability
may be established by a single test.
The results of Japanese research in this field are summarized in [51, 52].
From all the above-mentioned experiments the one proposed by Swift is the most
accurate one, giving the most reliable results.

3.2.1.3 Combined Deep-Drawing Methods

Real sheet metal forming processes usually are a combination of the basic processes
of deep-drawing and stretching. For simulating this, combined tests have been devel-
oped. Sachs [43, 53] developed a method based on the use of a conical punch with a
flat bottom. After tearing the test is continued until the occurrence of fracture at the
edge of the hole obtained previously.
Petrasch [27] proposed a method of evaluating the formability from the defor-
mation energy. At first the specimen is drawn by impact whereby the deformation
energy is obtained as a measure of deep-drawability. After this the upper edge of the
cup is enlarged until the occurrence of fracture. The ratio of the maximum diam-
eter and the diameter of the cylindrical section of the part is a measure of stretch
formability.
Swift stated that deep-drawing with a hemispherical punch gives more compre-
hensive information about the suitability for combined deep-drawing [48]. In this
test deformation takes place by stretching at the centre of the part and by deep-
drawing in the region of the blank. Fukui obtained a similar conclusion by using a
hemispherical punch and a conical die. The formability is expressed as the ratio of
the bottom diameter of the cone (corresponding to the occurrence of fracture) and
of the diameter of the workpiece.
Eisenkolb [54, 55] proposed a test consisting of forming a cup on the wall of
another cup that had already been produced by deep-drawing. By this a residual
formability is determined.
Engelhardt [56] introduced the so-called ‘deep-drawing potential’ as a measure
of safety against failure by bottom cracks. This method has obtained some practical
3.2 Evaluation of the Sheet Metal Formability 151

importance. Compared to the cupping test by Swift [48] this test is easier to carry
out since only one specimen is needed.

3.2.2 Limit Dome Height Method

This method combines advantages of simulating tests and of the forming limit dia-
gram. Based on observations by Drewes [57], Ghosh [40] proposed to represent
the heights of the parts as functions of the minimum strains occurring in rectangular
specimens (of Nakazima type) stretched on a hemispherical punch until fracture. By
drawing a curve through the experimental points obtained with specimens of differ-
ent width, a diagram like shown in Fig. 3.9 is obtained [11]. Later on, the method
has been modified by English researchers under the name of Strip Stretch Test and
by American researchers, named Limiting Dome Height test.
The height of the part corresponding to plane strain is a formability index denoted
by LDH0 . This is the minimum compared to the heights obtained for other states of
strain. The width of the specimen corresponding to plane strain is a characteristic of
the material. In spite of its advantages the method has been little used in industry,
due to the large dispersion of the LDH0 values and the large amount of experimental
work.
More recently, Wagoner [58] introduced the OSUFT method in order to over-
come the applicability limits of the LDH method. For this purpose the shape of the
tools was optimized by FE simulations aiming to achieve as close as possible a state
of plane strain. The results showed a reduced dispersion and good agreement with
industrial practice. Also the amount of experimental work is reduced.

Fig. 3.9 Limit dome height

curve
152 3 Formability of Sheet Metals

3.3 Forming Limit Diagram

3.3.1 Definition: History

The maximum values of the principal strains ε 1 and ε2 can be determined by mea-
suring the strains at failure (necking, fracture, wrinkling etc.) on sheet components
covered with grids of circles. The research in this field was pioneered by Keeler [59],
based on the observations of Gensamer [60] that instead of using global indices the
local deformations have to be considered (Fig. 3.10).
During forming the initial circles of the grid become ellipses. Keeler plotted the
major strains against the minor strains obtained from such ellipses at fracture of
parts after biaxial stretching (ε1 > 0; ε2 > 0 ) [59] (Fig. 3.11).
For numerous materials the critical area between the domains has been detected
both by means of laboratory tests and by forming of industrial components. These
measurements were conducted for various materials. The excellent correlation of the
results was a proof that the forming limits in sheet metal forming can be evaluated
very well by determining the Forming Limit Curve (FLC).
Later, Goodwin [62] plotted the curve for the tension/compression domain (ε1 >
0; ε2 < 0 ) by using different mechanical tests. In this case, transverse compression
allows for obtaining high values of tensile strains like in rolling or drawing.
The diagrams of Keeler (right side) and Goodwin (left side) are currently called
the Forming Limit Diagram (FLD), see Fig. 3.12. Connecting all of the points cor-
responding to limit strains leads to a Forming Limit Curve (FLC). The FLC splits

Fig. 3.10 The original Forming limit diagram defined by Gensamer [60]
3.3 Forming Limit Diagram 153

Fig. 3.11 Forming limit diagram defined by Keeler [61]

Fig. 3.12 Forming limit diagrams defined by Keeler and Goodwin [62]

the ‘fail’ (i.e. above the FLC) and ‘save’ (i.e. below the FLC) regions. The Forming
Limit Curve (FLC) is plotted on a Forming Limit Diagram (FLD). The intersection
of the limit curve with the vertical axis (which represents the plane strain deforma-
tion (ε2 = 0)) is an important point of the FLD and is noted FLD0 . The position of
this point depends mainly on the strain hardening coefficient and also on thickness
(see Sect. 3.3.4).
154 3 Formability of Sheet Metals

Fig. 3.13 Forming limit

diagrams for necking and for
fracture

Today, depending on the kind of limit strains that is measured different types of
FLD’s are determined: for necking and for fracture, see Fig. 3.13.
From subsequent experimental and theoretical research, even two more types of
FLDs have emerged: the wrinkling limit diagram by Havranek (Fig. 3.14) [63] and
the Stress Forming Limit Diagram (SFLD) by Arrieux (Fig. 3.15) [64]. The latter is
not sensitive to the strain path.
In order to extend the application of stress limit curves to a 3-D stress state (pres-
ence of through-thickness components of compressive stress), Simha et al. [65] has
introduced a new concept, namely Extended Stress-Based Limit Curve (XSFLC).
The XSFLC represents the equivalent stress and mean stress at the onset of necking
during in-plane loading. Figure 3.16 shows the three formulations of the Forming
Limit Curve concept, namely: strain-based FLC (εFLC), stress-based FLC (σFLC)
and Extended Stress-Based FLC (XSFLC), respectively. The Equivalent stress and
the Mean stress are obtained through the expressions
$
σeq = σ12 + σ22 − σ1 σ2 ; (3.5)
3.3 Forming Limit Diagram 155

Fig. 3.14 Forming limit

diagram for wrinkling [63]

Fig. 3.15 Stress forming

limit diagram defined by
Arrieux [64]

Fig. 3.16 Schematic of the strain-based forming limit curve (εFLC), the stress-based forming
limit curve (σ FLC) and the extended stress-based forming limit curve (XSFLC) [66]
156 3 Formability of Sheet Metals

Fig. 3.17 Forming limit

band (FLB) for two steel
grades [67]

σ1 + σ2
σmean = , (3.6)
3
where σeq is the equivalent stress, and σ mean the mean stress, which is assumed to
be positive in tension.
Figure 3.16 also presents the loading paths for the three cases: uniaxial stress,
plane strain and biaxial stress. A thorough analysis of the conditions for the use
of the XSFLC as a Formability Limit Curve under three-dimensional loading is
presented in [66].
Forming Limit Curves are valid for one particular material alloy, temper and
gauge combination. However material properties vary from batch to batch due to
variation in the production process. Therefor a single Forming Limit Curve cannot
be an exact description of the forming limit. Janssens et al. [67] have proposed a
more general concept, namely the Forming Limit Band (FLB) as a region covering
the entire dispersion of the Forming Limit Curves (Fig. 3.17).

3.3.2 Experimental Determination of the FLD

3.3.2.1 Experimental Tests
FLDs must cover as much as possible the strain domain which occurs in indus-
trial sheet metal forming processes. The diagrams are established by experiments
that provide pairs of values of the limit strains ε 1 and ε2 obtained for various load-
ing patterns (equibiaxial, biaxial, uniaxial etc.). In order to determine a FLD one
must generate load paths ranging from equibiaxial tension (ε1 = ε2 ) to pure shear
3.3 Forming Limit Diagram 157

Fig. 3.18 Rectangular

specimens with and without
notches

(ε1 = −ε2 ). In practice the state of simple tension (ε1 = −2ε2 for isotropic
materials) is never exceeded in the blank holder region.
It is necessary to deform the specimen along a linear strain path, i.e. the trajectory
followed by a point in the ε1 , ε2 -plane until reaching the forming limit must be a
straight line.

3.3.2.2 Uniaxial Tensile Test

This test allows determining a section of the FLD in the range ε2 < 0. For obtaining
different strain paths, specimens with different shapes and sizes with and without
notches are used, see Fig. 3.18 [68].
The advantages of the tensile test for determining FLDs are that the specimens
are easy to manufacture and a general purpose testing machine can be used; fur-
thermore the specimens remain plane during the test which facilitates measurement.
However, only the negative section of the forming limit diagram (ε2 < 0) can be
determined.

3.3.2.3 Hydraulic Bulge Test

This test was first proposed by Olsen [38]. For varying the strain path elliptical
dies of different shape can be used, see Fig. 3.19. The most important advantage of
the test is the absence of friction. The disadvantage is that only the section of the
forming limit diagram for positive values of ε2 can be obtained.

3.3.2.4 Punch Stretching Test

This test was first proposed and used by Keeler himself. It consists of stretching a
specimen (2) clamped between a blank-holder (3) and a die (4) using a spherical of
elliptical punch, see Fig. 3.20. The strain path is mainly varied by using specimens
of different width; is can also be varied by varying the punch radius and the lubri-
cant. Instead of using rectangular specimens Hasek applied specimens with circular
recesses.
158 3 Formability of Sheet Metals

Fig. 3.19 Scheme of

hydraulic bulge test with
elliptical dies

Fig. 3.20 Schematic layout

of the punch stretching test

3.3.2.5 Keeler Test

This test consists in the use of punches having different radii in order to vary the
stress state, see Fig. 3.21. Disadvantages of the test are the large amount of experi-
mental work; only the positive section of the forming limit diagram is obtained, and
the shape and position of the forming limit is influenced by the punch radii.
3.3 Forming Limit Diagram 159

Fig. 3.21 Punches used in the Keeler test

3.3.2.6 Hecker Test

In this test [69] only one type and size of punch and specimen is needed whereby
the friction regime is varied by varying the lubricant for obtaining different strain
paths. A disadvantage is that only the positive region of the forming limit diagram
is determined.

3.3.2.7 Marciniak Test

In deep-drawing with a flat bottom punch tearing of the part usually occurs at the
connection between the bottom and the cylindrical wall.
In order to produce the tearing at the planar bottom of the cup, Marciniak pro-
posed [4] to use a hollow punch and an intermediate part having a circular hole
placed between punch and workpiece, see Fig. 3.22.
To obtain different strain paths are used punches with different cross sections
(circular, elliptical, rectangular). The advantage of this test is that tearing appears at

Fig. 3.22 Schematic layout

of the device used in the
Marciniak test
160 3 Formability of Sheet Metals

Fig. 3.23 Shapes of the

specimens and the
intermediate part

the planar bottom of the part thus eliminating the errors of measurement caused by a
curvature. Disadvantages are the complex shapes of punch and die and the limitation
of the test to the positive domain of the forming limit diagram. In order to overcome
these drawbacks, the test can be modified by using specimens and intermediate parts
having different shapes, see Fig. 3.23. By varying the radius of the recesses the entire
domain of the FLD is obtained using only one ring punch.

3.3.2.8 Nakazima Test

The test [70, 71] consists of drawing rectangular specimens having different widths
using a hemispherical punch and a circular die. By varying the width of the specimen
and the lubricant one may obtain both the positive and the negative domain of the
FLD. The shape of the specimen recommended by ISO 12004 standard is presented
in the Fig. 3.24. The recommended length is longer than 25% of punch diameter.
For the punch diameter and the die diameter are recommended the values 100 and
105 mm, respectively.
Advantages of the test are the simplicity of the tools, the simple shape of the spec-
imens and the possibility of covering the entire domain of the FLDs. Disadvantages
are the possibility of wrinkling and errors of measurement caused by the curvature

a) b)
Fig. 3.24 (a) Shape of the specimens used in the Nakazima test [70]; (b) the photo of a set of
specimens for a complete FLD (courtesy by GOM company)
3.3 Forming Limit Diagram 161

Fig. 3.25 Shape of the

specimens used in the Hasek
test [73]

of the punch. This method is used actually as standard method by the ISO 12004
standard ‘Metallic materials. Determination of the forming limit curves’ [72].

3.3.2.9 Hasek Test

In order to avoid wrinkling of the specimens Hasek proposed [23] the use of circular
specimens with recesses of different radii, see Fig. 3.25. This requires an increased
amount of work for manufacturing the specimens.

3.3.2.10 Comparison of Different Tests

In addition to the above described methods there are some other tests but their use is
limited to a reduced number of strain states (Myauchi test using shearing, Marciniak

Fig. 3.26 FLDs established

using different testing
methods: 1—Hasek;
2—Nakazima; 3—uniaxial
tension; 4—Keeler;
5—hydraulic bulge test [73]
162 3 Formability of Sheet Metals

test using sheet torsion etc.). Hasek [73] published a systematic study of the influ-
ence of the testing method upon the obtained FLDs. The key results are summarized
in Fig. 3.26.
From the tests described above, the following ones are be recommended:
Marciniak test or hydraulic bulge test for eliminating friction; uniaxial test if
simplicity is sought for; Nakazima test for covering a great variety of strain
paths.

3.3.3 Methods of Determining the Limit Strains

Defining the limit strains of sheet metals has been a constant preoccupation of sev-
eral IDDRG workgroups. Almost all methods known until now have been developed
in these groups. The Benelux [74, 75], French [76, 77], Japan [22] and the German
[78, 79] IDDRG groups were very active in this area (see also [19, 80–82]).
Next we will briefly present the most well known methods developed over the
years.
Takashina and his co-workers [83] first proposed a simply method to determine
the limit strains (so-called ‘three circle method’). The method has been improved
by Veerman [84, 85]. Bragard [86] developed in 1972 a more precise method of
determining the limit strains based on interpolation (Fig. 3.27).
This method is later improved (1975) by D’Haeyer and Bragard [87] using the
name of ‘the double profile method’. In 1972 Hecker proposed a method [69, 88]
based on the determination of three types of ellipses around the fracture: frac-
tured, necked and acceptable (Fig. 3.28). The method consists in determining the
major and minor strains of the different types of ellipses in the neighborhood of the

Fig. 3.27 Bragard

(interpolation) method to
determine the limit strain
3.3 Forming Limit Diagram 163

Fig. 3.28 Hecker method to

determine the limit strain

fracture on the deformed piece and transposing them on FLD. The limit curve is
traced between the point corresponding to the ellipses affected by necking and the
acceptable ones (Fig. 3.28). The method has been used on a large scale because of
simplicity.
Kobayashi [89] defines the limit strain based on the accelerated increase of
the roughness in the necking area. The Zurich meeting in 1973 of the IDDRG
workgroup, following an analysis of several versions of limit strain determination,
recommends using an improved version of the Bragard method. This is known as
the ‘Zurich Nr.5 method’ [90].
Together with the development ‘of online’ video strain measuring methods, new
methods if determining the limit strains have been proposed in the last years.
A new criterion based on the evolution of the strain rate as a function of time
during the forming process has been proposed by the SOLLAC team [91]. The
method is based on the observation than the beginning of the necking is accom-
panied by a considerable increase of the strain rate. According with this method
164 3 Formability of Sheet Metals

Fig. 3.29 Strain-rate versus

time

the start necking point corresponds to the dramatically changing in the strain-rate
versus time variation (characteristic point). This point could be determined by the
intersection of the two straight lines corresponding to the first and the last sector
of the curve (Fig. 3.29). The strain—rate evolutions are automatically determined
by images analysis. The strain-rate method has been used recently by Volk [92]. He
used the idea to identify a regular grid for the optical measurement as a typical mesh
of a finite element method.
The Nakajima workgroup of the IDDRG has developed a new method [78], the
so-called ‘in-process measurement’ method (Fig. 3.30).
A guideline for the determination of FLC based on this method is presented in the
paper [93]. It is describe in detail the geometry and the number of the specimens, the
geometry of the tool, testing conditions (punch velocity, blank holding force lubri-
cant) as well as measurement and mathematical analysis of the deformed specimen.
The method is similar with the Bragard one. Using a video camera system, a film of
the forming process is made. Based on the film of the forming process, the devel-
opment of the strain distribution starting from the onset of necking and finally up

Fig. 3.30 IDDRG method to

determine the limit strain [78]
3.3 Forming Limit Diagram 165

to the fracture is analyzed. The method is a very robust one and gives a very good
repeatability of the results. Base of this achievements, the expert group of Nakajima
workgroup proposed a revision of the ISO 12004 standard ‘Metallic materials-sheet
and strip-Determination of the forming limit curves’ [72].
Based on the video camera measurement some systems have been developed
by the commercial company to determine automatically the FLC. CAMSYS com-
pany has developed the first automatically system (ASAME—Automated Strain
Analysis and Measurement Environment) used on the large scale, both in research
laboratory and industry [94]. The INSA Lyon developed a FLD determination sys-
tem (IcaForm) based on the spray of a random pattern of paint at the surface of
the sample to determine strain distribution [95]. An opto-mechanical device adapt-
able allows determining easily the FLD. An objective criterion to identify the start
of local necking automatically has been proposed recently [96]. The ‘Autogrid’ sys-
tem developed by Vialux company offer the possibility to determine the limit strains
automatically and independent of any operator. The methodology used to define the
limit strain is presented in details in the paper [97]. GOM Company has developed
for the FLC determination so-called ARAMIS system [98]. The methodology used
is according with the Nakajima workgroup recommendation. For an FLC are used
five different geometries, for each geometry three specimens and for each speci-
men three to five parallel sections. The FLC determination procedure can be done
automatically.

3.3.4 Factors Influencing the FLC

3.3.4.1 Sheet Thickness
Besides using sheet of higher quality, the most common solution for the success
of difficult sheet metal forming processes is to increase the sheet thickness. The
influence of the sheet thickness on the limit strains has been studied by Haberfield
[99], Romano [100], Hiam [101], Kleemola [102] etc.
The influence of sheet thickness on the FLD is characterized by the following
relationships:

• the FLD for necking depends on sheet thickness (t0 )(see Fig. 3.31 [103])
• as the thickness rises, the curve rises on the plot (ε1 ; ε2 )
• the influence is high for pure expansion and vanishes for pure compression
• the influence of the thickness on the FLD0 increases linearly (see Fig. 3.32 [104])
• along a linear strain path the rise of the FLD is proportional to the increase of
thickness but this influence vanishes above a critical value.

The engineer can decide if an unsuccessful forming process may be improved

by increasing the sheet thickness. This is especially important if the stress acting
during the forming process is tensile in both principal directions.
166 3 Formability of Sheet Metals

Fig. 3.31 Influence of the

thickness on the FLC [103]

Fig. 3.32 Influence of the thickness and strain hardening coefficient on the FLD0 [104]

3.3.4.2 Grid Size

If the circles of the initial grid have a diameter d, the measured strain is a mean
value averaged over a distance equal to d. Hence, in presence of a strain gradient, a
smaller grid size improves the resolution.
When determining the FLD for necking, the strain gradients are rather low; in
this case the effect of the grid size on the test results is only weak. However, when
determining the FLD for fracture, higher strain gradients occur [105].
3.3 Forming Limit Diagram 167

Grumbach and Sanz [105] studied the influence of the diameter d on the obtained
values of ε1 and ε 2 , using grids with radii in the range from 5 to 0.5 mm. The results
confirmed that the circle diameter of the grid has a strong influence on the strain
obtained in the direction of the strain gradient whereas the lateral strain may be
obtained without any gradient at all.
The data obtained for various circle diameters of the grid allow for an extrapola-
tion of the principal strains to a diameter of zero diameters. If the strain values for
this ideal grid are denoted by ε1∗ and ε2∗ one can write:

ε1∗ + ε2∗ + ε3∗ = 0 (3.7)

where ε3∗ is the strain in thickness direction. Assuming that there is a strain gradient
only in the direction of ε1 the strains in the two other directions are independent of
the grid diameter: ε2∗ = ε2 and ε3∗ = ε3 . Hence

ε1∗ = −(ε2 + ε3 ) (3.8)

As a consequence, the high dispersion of ε 1 one side and the other of the vertical
axis ε2 = 0 disappears. From this point of view several authors [26, 105] proposed
the use of an intrinsic FLD, representing the fracture FLD determined using a grid

Fig. 3.33 Intrinsic forming

limit diagram [105]
168 3 Formability of Sheet Metals

of infinitely small size. The plot of this intrinsic curve is very close to a straight line,
see Fig. 3.33 [105].

3.3.4.3 Strain Path

When the limit strains are sought, the path of deformation can be plotted, showing
the successive strains (ε1 ; ε2 ) imposed to the circle where fracture eventually occurs.
These paths always start at the origin of the two axes.
The curves in Fig. 3.34 were plotted using simple laboratory tests which are used
for determining limit diagrams. These tests were:

I ISO 50—simple tensile test using standard specimen

II TPE 1—tensile test using thinned specimen
III Swift 50—cylindrical cup  = 50 mm with flat bottom
IV hydraulic expansion through circular die (Jovignot).

The following conclusions have been formulated:

• the paths are practically linear as long as necking does not appear; in this case the
size of the grid circles has an influence on the measured values as shown above
• the slope of the path corresponding to the simple tensile test follows from the
relation

Fig. 3.34 Strain paths used in different tests [105]

3.3 Forming Limit Diagram 169

1+r
ε1 = − ε2 (3.9)
r

• the slope of the path corresponding to the hydraulic bulge test using a circular die
(Jovignot) is equal unity due to the equibiaxial tensile stress.

Kikuma [106] showed the influence of the strain path on the FLDs (Fig. 3.35).
For industrial sheet metal forming processes which require several passes (e.g. in
the automotive industry), Japanese authors have shown that, as long as the deforma-
tion pattern changes from one pass to another, the strain path is a broken line [52]. If
a tensile load path is followed by a compressive one the limit strains are smaller than
the ones corresponding to the FLD; on the other hand, a compressive load preceding
a tensile one improves the limit strain values. Yet, engineers from industry always
proceed first with a tensile or a compressive deformation, and only eventually apply
expansion.
The complex strain paths undergone by industrial parts during multi-pass
forming can be simulated in laboratory tests by applying to the same specimen
successively two different load paths, for example, a tensile test followed by deep-
drawing or biaxial expansion followed by a local tensile test. The analysis of the
specimens provides results similar to those for real industrial components:

• a path change has a strong influence on FLDs both for necking and fracture
• as long as the initial deformation implies ε2 < 0 (compression test, simple tensile
test) followed by biaxial expansion, the limit curve II is higher than the FLD I
(corresponding to simple, linear load paths, see Fig. 3.35)
• on the contrary, if ε2 > 0 during the first stage and compression or simple tensile
load follows later the limit curve III is below the FLD I for simple paths.

Fig. 3.35 Influence of strain

path on the forming limit
curve [106]
170 3 Formability of Sheet Metals

These results confirm the practical rule mentioned above. Actually, the residual
formability depends on the strain evolution. It is well known that the defects
induced by plastic deformation are a consequence of voids developing at the
matrix/inclusion interface; their growth and coalescence also depend on the strain
history.

3.3.4.4 Mechanical Properties

The numerous determinations of forming limit diagrams realized after Keeler and
Goodwin first published their results confirmed their conclusion that for materials
with relatively close mechanical properties the experimental FLDs are close as well.
Using the load path plots, it has been possible to determine the influence of the
mechanical properties not only on the position of the FLD, but also on the position
of the limit state corresponding to the part of interest. Woodthorpe [107] and Pearce
[108] have studied the influence of the mechanical parameters (strain hardening, n,
and anisotropy coefficients, r) on the FLC for the first time. Later, Haberfield [99],
Hiam [101] and Conrad [109] have extended Woodthorpe’s research. For example,
Conrad [109] studied the influence of the strain rate sensitivity exponent, m, on the
limit strains.
Since the strain paths for simple tensile tests are straight lines having the slope—
(1+r)/r in the ε1 , ε2 -plane the load paths will shift to the left by increasing r. Higher
values of r also allow for higher values of the final strain ε1 . This influence has been
outlined in [110], see Fig. 3.36.

Fig. 3.36 Influence of

anisotropy on strain path
[110]
3.3 Forming Limit Diagram 171

In the same manner, if n is increasing (at a given r), one may notice that the
uniform elongation and the ultimate elongation increase and the point designating
fracture moves to higher values of ε1 .
It has also been shown that in the range of expansion (ε1 > 0; ε2 > 0) with
the increase of n, the corresponding point of the FLD is shifted to the right i.e. to
higher values of equivalent strain. This may be explained by the delayed initiation of
necking whereby the strain path remains linear for a longer period. As a conclusion,
even if the FLDs for different materials are close to each other they are shifted in
accordance with the values of the mechanical properties (r and n).

3.3.4.5 Influence of the Punch Curvature

Ghosh and Hecker [111] showed that the choice of the experimental method used
for the FLC determination (in-plane versus out-of-plane) influences the position of
the limit curves. The influence of the punch curvature on the stretching limits has
been studied first by Charpentier [112]. He used punches having different curvature
radii (ranging from the infinite radius—in-plane stretching—to a relatively small
radius—15.85 mm) and noticed that the decrease of the radius was accompanied by
an increase of the formability (upwards motion of the FLC). Shi and Gerdeen [113]
performed a theoretical analysis of this influence using the Marciniak–Kuckzinsky
model (see Sect. 3.4.3) (Fig. 3.37).

3.3.4.6 Influence of the Temperature

The influence of the temperature on the limit strains was studied first by Lange [114]
and later by Ayres [115], Kumpulainen [116], Li and Ghosh [117], Abedrabbo [118],
van den Boogaard [119] etc. According to these researchers, the temperature has a
different influence on the formability of different metallic alloys. For example, the
formability of the AA 5754 alloy has a significant increase when the temperature
rises even with small amounts (from 250 to 350◦ C) (Fig. 3.38 [117]), while temper-
ature variations in the same range have a very little influence on the formability of
the AA 6111-T4 alloy (Fig. 3.39 [117]). The increase of the formability by raising
the temperature of the material is frequently used in the case of the sheet metals hav-
ing a poor formability at room temperature (some aluminum or magnesium alloys,
high-strength steels, etc.).

3.3.4.7 Influence of the Strain Rate

Drewes [120] followed later by Ayres [115] and Percy [121] have analyzed the
influence of the strain rate on the limit strain. In general, the increase of the strain
rate causes a downward displacement of the FLC, that is a diminishment of the
formability. Such an example is shown in Fig. 3.40 [121] and corresponds to the
SPCEN-SD steel. The difference between the strain rates is 104 . Similar results
were also obtained by Ayres [115]. On the other hand, more recently, Balanethiram
172 3 Formability of Sheet Metals

Fig. 3.37 Influence of punch curvature on the FLC [113]

and Daehn [122] have reported a significant increase of the formability when the
strain rate is also increased for an OFHC copper. Gerdooei and Dariani [123] have
explained this effect based on the Johnson–Cook law (see Fig. 3.41). The different
behavior of the metallic materials from this point of view is a consequence of
the different values of the strain-rate sensitivity index, as well as of the different
mechanical response when the strain rate is modified.

3.3.4.8 Influence of the Normal Pressure

The effect of the normal pressure on the formability of sheet metals is well know and
already used from long time ago in industry [124]. During some forming operations
(hydrostatic forming, incremental forming) the sheet is subjected to a significant
normal pressure. The increase in formability is usually explained by the closing of
3.3 Forming Limit Diagram 173

Fig. 3.38 Influence of the

temperature on the FLC for
the 5754 aluminium alloy
[117]

Fig. 3.39 Influence of the

temperature on the FLC for
the 6111-T4 aluminium
alloy [117]

the micro-voids in the sheet and the slowing down of the nucleation of new ones
due to the normal pressure exerted by the surrounding fluid [125]. But, an analysis
of sheet failure under normal pressure without assuming ductile damage has been
done in the last period. Such an analysis was performed by using Swift Hill models
by Gotoh [126], Smith [127] and Matin [128]. Recently, Banabic and Soare [129],
Wu et al. [130] and Alwood and Shouler [131] have analyzed the influence of the
normal pressure on the Forming Limit Curve using an enhanced Marciniak model.
174 3 Formability of Sheet Metals

1.0

0.9

0.8

0.7
CONVENTIONAL RATE
MAJOR STRAIN

0.6 ABOUT 10 mm/s

0.5

0.4
HIGH RATE

0.3

0.2

0.1 HIGH RATE

ABOUT 100 m/s
0
–0.4 –0.3 –0.2 –0.1 0 0.1 0.2 0.3 0.4 0.5
MINOR STRAIN

Fig. 3.40 Influence of the strain-rate on the FLC for SPCEN-SD steel [121]

0.8

0.6
Major Strain

0.4

0.2 Exp. at Quasi-Static Loading

(Balanethiram et al.,1994)
Theo. at Quasi-Static Loading

Exp.at Strain Rate = 428 (1/s)

(Balanethiram et al.,1994)
Theo.at Strain Rate = 428 (1/s)

0
–0.3 –0.2 –0.1 0 0.1 0.2 0.3 0.4
Minor Strain

Fig. 3.41 Influence of the strain-rate on the FLC for OFHC copper [123]
3.3 Forming Limit Diagram 175

Fig. 3.42 Forming limit curves for several values of the normal pressure for AA3104-H19
aluminium alloy [129]

The results presented in the last papers are very closed one to another one. In the
Fig. 3.42 is presented this influence based on the modified Marciniak model.

3.3.4.9 Other Parameters

Besides the influencing factors listed above some other parameters have an effect on
the FLCs, see [1, 3, 16, 18, 23, 31, 82, 81, 132–137].

3.3.5 Use of Forming Limit Diagrams in Industrial Practice

The most realistic and efficient way to estimate the technological possibilities of
producing a part and to avoid unsuccessful design is to apply the FLD method. This
enables to determine

• the range of safety for deep-drawing

• the critical zones where necking or fracture are most likely to occur
• the strain level
• the favorable working conditions (blank holding, lubrication etc.).

In practice this method is applied as follows. After the shape, dimensions and
material quality of a given part have been prescribed by the designer the form-
ing technology and the tools have to be designed. For this purpose the maximum
176 3 Formability of Sheet Metals

strains in the part must be known as well as the forming limit diagram of the
material.
By comparing the points corresponding to the maximum strains in the parts
with the FLD one can estimate whether fracture or necking could appear during
forming.
If no defect is to be expected one has information about how far from the limit
the material is deformed.
If the points defined by the maximum strain are beyond the limit curve, some
modifications have to be made of

• working conditions (blank-holding, lubrication)

• design of the part itself (fillet radii, angles etc.)
• material (quality, thickness).

The effect of these modifications is illustrated by Fig. 3.43. In the first two cases
the characteristic point is shifted either from A to A′ or A′′ or from B to B′ or B′′ .
In the third case the forming limit curve itself is shifted to the dashed line.
The FLD method also gives an estimation of the severity of deformation through
the so-called severity index [138]. This parameter is defined as shown in Fig. 3.44.
If the point of the maximum strain is below and far from the forming limit curve,
the severity index is small and the safety margin is so large that material is wasted
(point C in Fig. 3.43).
Therefore it is possible either to modify the forming process by increasing the
strains (move from C to C′ ) or to use a material having a lower forming limit
curve (yet still beyond the point of maximum strains). A good compromise between
reasonable safety margins and prevention of material waste appears to be in the
zone where the severity index is 7–8. The safety margin depends on the quality

Fig. 3.43 Examples of the

shifting of critical points
3.3 Forming Limit Diagram 177

Fig. 3.44 Definition of the severity indices [138]

of the material as well as on the process parameters. The choice of the mate-
rial not only influences the FLD but also the strain path and the deformation
front.
The determination of the maximum strains to compare with the FLD can be dif-
ficult for complex parts, since the deformations cannot be calculated by analytical
methods. In such cases two other methods are available for the estimation of the
maximum strains: experimental studies on small scale models or numerical simu-
lation. In the past experimental methods have been most widely used. However, in
the last two decades numerical simulation has become more and more the method
of choice. On the other hand, important progress has been achieved in numerical
solution methods, especially the finite element method.
The structure of an expert system for the analysis of sheet metal formability is
illustrated by Fig. 3.45. The steps of process design in sheet metal forming, based
on formability analysis, are presented in Fig. 3.46 [139].
178 3 Formability of Sheet Metals

Fig. 3.45 Structure of an expert system for the analysis of sheet metal formability [139]

Besides software packages of finite element analysis and CAD, computer aided
process design also requires databases with FLDs for various working conditions.
Experimental testing is a very valuable input for these databases, but it is also rather
expensive. The mathematical modeling based on theoretical developments described
in the Sect. 3.4 is also used for this purpose. A detailed presentation of the way the
FLD concept is used in FE simulation software is made in Sect. 4.3.
3.4 Theoretical Predictions of the Forming Limit Curves 179

Fig. 3.46 Flowchart of technological design process in sheet metal forming on the basis [139]

3.4 Theoretical Predictions of the Forming Limit Curves

Various theoretical models have been developing for the calculation of forming
limit curves (Fig. 3.47). The first ones were proposed by Swift [140] and Hill [141]
assuming homogeneous sheet metals (the so-called models of diffuse necking and
localized necking), respectively). The Swift model has been developed later by Hora
(so-called Modified Maximum Force Criterion-MMFC) [142]. Marciniak [143]
proposed a model taking into account that sheet metals are non-homogeneous from
both the geometrical and the structural point of view. Storen and Rice [144] have
180 3 Formability of Sheet Metals

Fig. 3.47 Theoretical models used in FLC calculation

been developed a model based on the bifurcation theory. Dudzinski and Molinari
[145] used the method of linear perturbations for analyzing the strain localization
and computing the limit strains.
Since the theoretical models are rather complex and need a profound knowl-
edge of continuum mechanics and mathematics while their results are not always in
agreement with experiments, some semi-empirical models have been developed in
recent years.
In the next sections the most commonly used models are presented briefly
with the focus on those based on the necking phenomenon (Swift and Hill), the
Marciniak–Kuczynski and MMFC model, respectively.

3.4.1 Swift’s Model

Considère [146] approached for the first time the problem of plastic instability in
uniaxial tension. In the case of ductile materials, two domains may be distinguished
in the region of plastic straining. In the first domain the hardening influence on the
traction force is stronger than the influence of the cross-section reduction. This is the
so-called ‘domain of stable plastic straining’, being characterized by the fact that an
increase of the traction force is needed in order to obtain an additional deformation
of the specimen. In the second domain material hardening cannot compensate the
decrease of the traction force due to the reduction of the specimen’s cross-section.
This is the so-called ‘domain of unstable plastic straining’, being characterized by a
decrease of the traction force, although the stress continues to increase.
The beginning of necking corresponds to the maximum of the traction force.
From the mathematical point of view, this condition can be written in the form
3.4 Theoretical Predictions of the Forming Limit Curves 181

dF = 0 (3.10)

By simple mathematical manipulations the following condition of plastic instability

is obtain:

dσ
=1+σ (3.11)
dε
Assuming a Ludwik–Hollomon strain-hardening law,

σ = kε̄n (3.12)

then the condition (3.11) becomes

ε̄ = n (3.13)

Hence, according to Considère’s criterion, a material obeying the Ludwick–

Hollomon hardening law starts to neck when the strain is equal to the hardening
coefficient.
Swift [140] used the Considère criterion in order to determine the limit strains
in biaxial tension. He analysed a sheet element loaded along two perpendicular
directions and applied the Considère criterion for each direction. Assuming a strain
hardening described by Eq. (3.12), he obtained the following expressions for the
limit strains:
 2  
∂f ∂f ∂f
σ1 ∂σ1 + σ2 ∂σ 2 ∂σ1
ε1∗ =  2  2 n (3.14)
∂f ∂f
σ1 ∂σ 1
+ σ2 ∂σ2

 2
  
∂f ∂f ∂f
σ2 ∂σ1+ σ1 ∂σ 1 ∂σ2
ε2∗ =  2  2 n (3.15)
∂f ∂f
σ1 ∂σ1 + σ2 ∂σ 2

where f is the yield function.

By using different yield functions, it is possible to evaluate the limit strains as
functions of the loading ratio α and the mathematical parameters of the material
(hardening coefficient n, anisotropy coefficient r, strain-rate sensitivity m, etc.). As
an example, if the Hill 1948 criterion is used (Eq. 2.63), the limit strains are as
follows:
 
2r
[1 + r (1 − α)] 1 − 1+r α + α2
ε1∗ = n (3.16)
1+4r+2r2
(1 + r) (1 + α) 1 − α + α2
(1+r)2
182 3 Formability of Sheet Metals
 
2r
[(1 + r) α − r] 1 − 1+r α + α2
ε2∗ = n (3.17)
1+4r+2r2
(1 + r) (1 + α) 1 − α + α2
(1+r)2

The expressions of the limit strains associated to some other yield criteria (such as
Hill 1979 and Hill 1993) are presented in [147].
By computing the values of ε 1 ∗ and ε 2 ∗ for different loading ratios α and
recording them in a rectangular coordinate system ε 1 , ε2 the necking limit curve
is obtain.

3.4.2 Hill’s Model

In the case of uniaxial tension, the localized necking develops along a direction,
which is inclined with respect to the loading direction. Hill [141] assumed that the
necking direction is coincident with the direction of zero-elongation and thus the
straining in the necking region is due only to the sheet thinning.
The method used for obtaining the limit strains in this case is presented in [10].
The expressions of these strains are as follows:

∂f
∂σ1
ε1∗ = ∂f ∂f
n (3.18)
∂σ1 + ∂σ2

∂f
∂σ2
ε2∗ = ∂f ∂f
n (3.19)
∂σ1 + ∂σ2

It can be seen that

ε1∗ + ε2∗ = n (3.20)

This is the equation of a line parallel with the second bisectrix of the rectangular
coordinate system ε1 , ε2 and intersecting the vertical axis at the point (0, n).
According to Eq. (3.20), the FLC computed on the basis of the Hill’s model does
not depend on the yield criterion, but only on the value of the hardening coefficient.

3.4.3 Marciniak–Kuckzynski (M–K) and Hutchinson–Neale

(H–N) Models
On the basis of the experimental investigations concerning the strain localization of
some specimens subjected to hydraulic bulging or punch stretching, Marciniak [148]
has concluded that the necking is usually initiated by a geometrical or structural
non-homogeneity of the material. This non-homogeneity may be associated to a
3.4 Theoretical Predictions of the Forming Limit Curves 183

variation of the sheet thickness (geometrical non-homogeneity) or some defects of

the lattice (structural non-homogeneity).
The analysis of the necking process has been performed assuming a geometrical
non-homogeneity in the form of a thickness variation. This variation is usually due
to some defects in the technological procedure used to obtain the sheet metal. The
thickness variation is generally gentle. However, the theoretical model assumes a
sudden variation in order to simplify the calculations (Fig. 3.48).
The theoretical model proposed by Marciniak assumes that the specimen has
two regions: region A having a uniform thickness s0 (A) , and region B having the
thickness s0 (B) (Fig. 3.48). The initial geometrical non-homogeneity of the speci-
men is described by the so-called ‘coefficient of geometrical non-homogeneity’, f,
expressed as the ratio of the thickness in the two regions:
(B)
s0
f = (A)
(3.21)
s0
The strain and stress states in the two regions are analysed with respect to the
principal strain ε1 (B) in region B and the principal strain ε 1 (A) in region A. When the
ratio ε1 (B) /ε 1 (A) becomes too high (infinite in theory, above 10 in practice), one may
consider that the deformation of the specimen is localized in region B (Fig. 3.49).
The shape and position of the curve ε 1 (B) (ε1 (A) ) depend on the value of the coef-
ficient f. If f = 1 (geometrically homogeneous sheet), the curve becomes coincident
with the first bisectrix. Thus this theory cannot model the strain localization for
geometrically homogeneous sheets.
The value of the principal strain ε 1 (A) in region A corresponding to non-
significant straining of this region as compared to region B (the straining being local-
ized in region B) represents the limit strain ε 1 (A)∗ (Fig. 3.49). This strain together

Fig. 3.48 Geometrical model

of the M–K theory
184 3 Formability of Sheet Metals

Fig. 3.49 The dependence ε 1 (B) /ε 1 (A)

with the second principal strain ε2 (A∗) in region A define a point of the forming limit
curve. By varying the strain ratios ρ = d ε2 (A) /d ε 1 (A) , different points on the FLC
are obtained. By scrolling the range 0 < ρ < 1, the FLC for biaxial tension (ε1 > 0,
ε 2 > 0) is obtained. In this range the orientation of the geometrical non-homogeneity
with respect to the principal directions is assumed to be the same during the entire
forming process.
The Marciniak model was further developed by Marciniak and Kuczynski [143]
and Marciniak, Kuczynski and Pokora [149], usually being briefly denominated the
M–K model.
The M–K model was extended to the negative range of the FLD’s (ε2 < 0) by
Hutchinson and Neale (H–N model) [150–152]. According with the original paper
of Hutchinson and Neale [151], the inclination of the non-homogeneity varies with
the main strains by a law having the form:

(A)
1 + dε1
tan(ϕ + dϕ) = (A)
tan ϕ (3.22)
1 + dε2

and the non-uniformity coefficient varies by a law having the form:

 
(B) (A)
f 1 = f 0 d ε3 −d ε3 , (3.23)

where, f1 and f0 are the current and initial non-uniformity coefficients, respectively.
The M–K and H–N models are thoroughly described in [153–155] together with
the numerical algorithms. The explicit algorithms are usually used to solve the M–K
and H–N models. Newton’s method is used to solve the non linear system of equa-
tions. Since Newton’s method usually has a non-convergence problem, different
methods are used (for example, backtracking algorithm [155]) to eliminate this
drawback. However, the use of such an algorithm significantly increases the com-
putation time. A synthetic presentation of the M–K and H–N models’ evolution is
given in [156, 157].
3.4 Theoretical Predictions of the Forming Limit Curves 185

In order to increase the robustness of the algorithms used to solve the M–K and
the H–N models, an implicit formulation of the models is proposed in [158]. This
formulation is presented in detail in the next section.

3.4.4 Implicit Formulation of the M–K and H–N Models

Both M–K and H–N models assume that the strain localization is caused by a thick-
ness imperfection represented as a groove in Fig. 3.50 [143, 150]. According to
this hypothesis, two regions of the sheet metal should be distinguished: A – non-
defective zone; B – groove. At different stages of the straining process (identified
by the time parameter t), the ratio
4
t
f = t s(B) t s(A) , 0 < tf < 1 (3.24)

is used to describe the amplitude of the imperfection (t s(A) and t s(B) denote the
current thickness of regions A and B, respectively – see Fig. 3.50).
Throughout this section, the sheet metal is considered to behave as an orthotropic
membrane under the plane-stress conditions
t σ = t σ = 0, i = 1, 2, 3,
i3 3i
t ε̇ t (3.25)
α3 = ε̇3α = 0, α = 1, 2.

The constraints written above are valid both for region A and region B. Eq. (3.25)
involves the components of the stress and strain-rate tensors expressed in the plastic
orthotropy frame (1 and 2 are the indices associated to the rolling and transverse

Fig. 3.50 Schematic view of

the thickness imperfection
assumed by the H–N model
186 3 Formability of Sheet Metals

directions, respectively – see Fig. 3.50, while 3 is the index corresponding to the
normal direction – not shown in Fig. 3.50).
We also assume that the sheet metal is subjected to loads which do not produce
tangential stresses and strains in the plastic orthotropy frame:

t
σ12 = t σ21 = 0, t
ε̇12 = t ε̇21 = 0. (3.26)

This constraint will be applied not only to the non-defective zone (as in the clas-
sical formulation of the Hutchinson–Neale model), but also to the groove. Under
such circumstances, the diagonal components of the stress and strain-rate tensors
automatically become eigenvalues. In order to emphasize their significance, the
following notations will be used:

tσ = 1, 2, 3) – principal stresses t σ1 = t σ11 , t σ2 = t σ22 , t σ3 = t σ33 = 0

 
i (i
t ε̇ (i = 1, 2, 3) – principal strain rates t ε̇ = t ε̇ , t ε̇ = t ε̇ , t ε̇ = t ε̇
i 1 11 2 22 3 33 .

The mechanical response of the sheet metal will be described by a rigid-plastic

model. The main ingredient of the constitutive model is the yield criterion:

t
σ t σ1 , t σ2 = t Y t ε .
   
(3.27)

Equation (3.27) involves the following quantities:

tσ = t σ t σ1 , t σ2 ≥ 0 – equivalent stress (homogeneous function of the first

 

degree)
t ε ≥ 0 – equivalent (plastic) strain
t Y = t Y t ε > 0 – yield parameter controlled by a strictly increasing hardening
 

law.

The non-zero components of the strain-rate tensor (considered fully plastic) are
defined by the flow rule

t ∂tσ
ε̇α = t ε̇ , α = 1, 2, (3.28)
∂ t σα

and the incompressibility constraint

t
ε̇3 = −t ε̇1 − t ε̇2 . (3.29)

In order to preserve the simplicity of the formulation, we assume that region A

evolves along linear strain paths defined as follows:

t (A) t (A) (A)

ε̇1 > 0, ε̇2 = ρ (A) · t ε̇1 , −1 < ρ (A) ≤ 1. (3.30)
3.4 Theoretical Predictions of the Forming Limit Curves 187

Each strain path investigated when calculating a forming limit curve will be iden-
tified by a constant value of the parameter ρ (A) . Eq. (3.30) automatically implies that
t ε̇ (A) has the status of a minor principal strain-rate.
2
As shown in Fig. 3.50, the orientation of the groove is described by the angular
parameter ϕ. We adopt the hypothesis 0◦ ≤ ϕ < 45◦ , thus considering that the
(A)
necking band is closer to the direction of the minor principal strain-rate t ε̇2 . In
order to find a formula for the calculation of the angular parameter ϕ, we define a
local frame associated to the groove. Its planar axes are identified by the indices 1′
and 2′ , being oriented as in Fig. 3.50. Let

t (A) (A) (A) (A)
ε̇2′ 2′ = t ε̇1 sin2 ϕ + t ε̇2 cos2 ϕ = t ε̇1 sin2 ϕ + ρ (A) cos2 ϕ (3.31)

be the strain-rate along the necking band. If −1 < ρ (A) ≤ 0, Eq. (3.31) could be
used to find a zero-extension direction. Indeed, by enforcing

t (A) (A)
ε̇2′ 2′ = t ε̇1 sin2 ϕ + ρ (A) cos2 ϕ = 0, −1 < ρ (A) ≤ 0, (3.32)

one obtains

tan2 ϕ + ρ (A) = 0, −1 < ρ (A) ≤ 0, (3.33)

i.e.
$
ϕ = arctan −ρ (A) , −1 < ρ (A) ≤ 0. (3.34)

Equation (3.34) defines the orientation of the necking band for the left branch of
the forming limit curve. In fact, this formula is similar to that found by Hill for the
same type of strain paths [141].
If 0 < ρ (A) ≤ 1, Eq. (3.31) does not allow the existence of zero-extension direc-
tions in the plane of the sheet metal. In such cases, as in the classical M–K model, we
assume that the necking band is oriented along the direction of the minor principal
(A)
strain-rate t ε̇2 :

ϕ = 0◦ , 0 < ρ (A) ≤ 1. (3.35)

Equations (3.34) and (3.35) can be unified in the general formula

$
ϕ = arctan max −ρ (A) , 0 , −1 < ρ (A) ≤ 1. (3.36)

It is easily noticeable that, for linear strain paths ρ (A) = const. , Eq. (3.36)
 

implies the constancy of the angular parameter ϕ.

For any load state having the property t σ1 > 0, the equivalent stress could be
expressed as follows:
188 3 Formability of Sheet Metals

t
σ = t σ1 · F t ζ , t
ζ = t σ2 t σ1 , t
  &
σ1 > 0. (3.37)

Equation (3.37) results from the t

t
& tfact that σ is a first-degree homogeneous func-
tion. The partial derivatives ∂ σ ∂ σα (α = 1, 2) are also homogeneous functions
but of zero-degree. As a consequence, they are expressible in the form

∂ tσ
= Gα t ζ , t
ζ = t σ2
  &t t
t
σ1 , σ1 > 0, α = 1, 2. (3.38)
∂ σα
The functions F and Gα (α = 1, 2) are related only to the particular formulation
of the equivalent stress adopted in the model. Equations (3.37) and (3.38) lead to
the following expressions of the yield criterion and flow rule (see also Eqs. (3.27)
and (3.28)):
t
σ1 · F t ζ = t Y t ε , t
ζ = t σ2
    &t t
σ1 , σ1 > 0, (3.39)

t
ε̇α = t ε̇ · Gα t ζ , t
ζ = t σ2
  &t t
σ1 , σ1 > 0, α = 1, 2. (3.40)
(A)
The linear strain paths defined as in Eq. (3.30) fulfil the condition t σ1 > 0.
Under these circumstances, Eq. (3.40) can be applied to region A:
 4
t (A) (A) (A) t (A) t (A)
ε̇α = t ε̇ · Gα t ζ (A) , t (A)
ζ = t σ2 σ1 , σ1 > 0, α = 1, 2. (3.41)

Equations (3.41) and (3.30) allow to obtain a relationship between ρ (A) and t ζ (A) :
 
G2 t ζ (A) = ρ (A) · G1 t ζ (A) . (3.42)

It is again noticeable that, for linear strain paths ρ (A) = const. , Eq. (3.42)
 

implies the constancy of the principal stress ratio, i.e.

4
t (A) (A) t (A)
ζ = t σ2 σ1 = ζ (A) = const. (3.43)

At the level of region A, Eqs. (3.39) and (3.40) can thus be written in the
particular forms
 
t (A)
σ1 · F ζ (A) = t Y t ε (A) , (3.44)

t (A) (A)
ε̇α = t ε̇ · Gα ζ (A) , α = 1, 2. (3.45)

(B)
Because the stress state in region B also fulfils the condition t σ1 > 0, we can
define the corresponding ratio
4
t (B) (B) t (B) t (B)
ζ = t σ2 σ1 , σ1 > 0. (3.46)
3.4 Theoretical Predictions of the Forming Limit Curves 189

As we shall see below, t ζ (B) generally varies even if the strains in the non-
defective zone evolve along a linear path. Due to this fact, Eqs. (3.39) and (3.40)
should be written as follows when making reference to region B:
 
t (B)
σ1 · F t ζ (B) = t Y t ε(B) , (3.47)

t (B) (B)
ε̇α = t ε̇ · Gα t ζ (B) , α = 1, 2. (3.48)

As in the classical formulation of the H–N model, two sets of constraints will be
enforced at the interface between the regions A and B (see Fig. 3.50):

• Continuity of the strain-rate along the necking band

t (A) (B)
ε̇2′ 2′ = t ε̇2′ 2′ (3.49)

• Equilibrium of the normal and tangential loads acting on the interface from both
sides

t (A) (B)
σ1′ 1′ · t s(A) = t σ1′ 1′ · t s(B) (3.50)

t (A) (B)
σ1′ 2′ · t s(A) = t σ1′ 2′ · t s(B) (3.51)

By making use of the thickness non-homogeneity parameter t f (see Eq. (3.24)),

we write Eqs. (3.50) and (3.51) in the equivalent forms

t (A) (B)
σ 1′ 1 ′ = t f · t σ1′ 1′ (3.52)

t (A) (B)
σ 1′ 2 ′ = t f · t σ1′ 2′ (3.53)

The rotated tensor components involved in Eqs. (3.52) and (3.53) can be also
expressed in terms of the principal stresses, thus obtaining

t (A) (A) (B) (B)
σ1 cos2 ϕ + t σ2 sin2 ϕ = t f · t σ1 cos2 ϕ + t σ2 sin2 ϕ (3.54)
 
t (A) (A) (B) (B)
σ1 − t σ2 sin ϕ · cos ϕ = t f · t σ1 − t σ2 sin ϕ · cos ϕ (3.55)

Because 0◦ ≤ ϕ < 45◦ , we are allowed to rewrite the above relationships as

follows:

t (A) (A) (B) (B)
σ1 + t σ2 tan2 ϕ = t f · t σ1 + t σ2 tan2 ϕ (3.56)
 
t (A) (A) (B) (B)
σ1 − t σ2 tan ϕ = t f · t σ1 − t σ2 tan ϕ (3.57)
190 3 Formability of Sheet Metals

At last, with the help of the principal stress ratios associated to regions A and B
(see Eqs. (3.43) and (3.46)), Eqs. (3.56) and (3.57) become
 
t (A) (B)
σ1 · 1 + ζ (A) tan2 ϕ = t f · t σ1 · 1 + t ζ (B) tan2 ϕ (3.58)

 
t (A) (B)
σ1 · 1 − ζ (A) tan ϕ = t f · t σ1 · 1 − t ζ (B) tan ϕ (3.59)

In general, Eq. (3.58) cannot degenerate to the trivial case 0 = 0. Under such
circumstances, it is possible to divide Eq. (3.59) by Eq. (3.58). After some simple
manipulations, we obtain the following relationship between the principal stress
ratios associated to regions A and B:

ζ (A) − t ζ (B) sin ϕ = 0. (3.60)

For the strain paths characterized by the condition −1<ρ (A) <0, Eq. (3.36)
defines an angular parameter 0◦ < ϕ < 45◦ . In this case, Eq. (3.60) enforces
t ζ (B) = ζ (A) = const. The principal stress ratios associated to regions A and B

are thus rigorously coincident and constant when −1 < ρ (A) < 0.
The plane-strain path ρ (A) = 0 needs a separate discussion, as in this case
Eq. (3.36) defines an angular parameter ϕ = 0◦ and Eq. (3.60) degenerates to the
trivial form 0 = 0. Anyhow, when ϕ = 0◦ , the local frame associated to the groove
is superimposed to the plastic orthotropy frame (1 = 1′ and 2 = 2′ ). The constraints
(A) (B)
given by Eqs. (3.32) and (3.49) now reduce to t ε̇2 = t ε̇2 = 0, meaning that region
B evolves along the same plane-strain path and enforcing again the constancy of the
principal stress ratio: t ζ (B) = ζ (A) = const. We are able to conclude that

t (B)
ζ = ζ (A) = const., if − 1 < ρ (A) ≤ 0. (3.61)

For all the strain paths characterized by the condition 0 < ρ (A) ≤ 1, Eq. (3.36)
defines an angular parameter ϕ = 0◦ . In this case, Eq. (3.60) also degenerates to the
trivial form 0 = 0, but Eq. (3.49) will not enforce the constancy of the stress ratio in
region B as it takes the more general form t ε̇2(A) = t ε̇2(B) .
One may notice that, whatever is the value of the parameter ρ (A) in the range
−1 < ρ (A) ≤ 1, the equilibrium constraint given by Eq. (3.58) reduces to

t (A) (B)
σ1 = t f · t σ1 , (3.62)

due to Eqs. (3.61) and (3.36). For all the strain paths characterized by the condition
−1 < ρ (A) ≤ 0, the above relationship becomes even simpler when combined with
Eqs. (3.44), (3.47) and (3.61):
 
t
Y t ε (A) = t f · t Y t ε (B) , if − 1 < ρ (A) ≤ 0. (3.63)
3.4 Theoretical Predictions of the Forming Limit Curves 191

Equation (3.63) makes redundant the second equilibrium constraint expressed by

Eq. (3.59). In fact, Eq. (3.63) has been deduced using Eq. (3.61) which is a corollary
of Eq. (3.59).
In the case 0 < ρ (A) ≤ 1, Eqs. (3.44) and (3.47) can be exploited to reformulate
Eq. (3.62) as follows:

4  4 
t
Y t ε (A) F ζ (A) = t f · t Y t ε (B) F t ζ (B) , if 0 < ρ (A) ≤ 1. (3.64)

Again, Eq. (3.64) should not be accompanied by Eq. (3.59) because the second
equilibrium constraint now degenerates to the trivial form 0 = 0.
The strain-compatibility enforced by Eq. (3.49) also deserves a discussion. In the
case −1 < ρ (A) ≤ 0, this constraint becomes trivial (0 = 0) and redundant due to
Eqs. (3.36) and (3.61) already included in the model. For the remaining strain paths
0 < ρ (A) ≤ 1, Eq. (3.49) reduces to the simpler formulation (see also Eqs. (3.35),
(3.45) and (3.48))

 
t (A) (B)
ε̇ · G2 ζ (A) = t ε̇ · G2 t ζ (B) , if 0 < ρ (A) ≤ 1. (3.65)

Equation (3.65) is non-trivial and accompanies Eq. (3.64) in the model used to
calculate the right branch of the forming limit curve.
We shall focus now on the presentation of the computational strategy. The evolu-
tion of the sheet metal up to the necking is analyzed for individual strain paths.
Each of these paths is defined by a constant value of the parameter ρ (A) in the
range −1 < ρ (A) ≤ 1. The straining process is analyzed in an incremental man-
ner. Let [T, T + T] be the discrete time interval corresponding to one of the
steps performed in the analysis. All the parameters associated to the T moment are
known quantities both for the non-defective area and the groove. The corresponding
configuration of the sheet metal is thus taken as a reference state. In particular,
the parameters associated to the moment T = 0 are defined by the conditions
0 ε (A) = 0 ε (B) = 0, and 0 ε (A) = 0 ε (B) = 0 (α = 1, 2). The initial value of the thick-
α α
ness ratio 0 < 0 f < 1 is also prescribed. As concerns the parameters corresponding
to the T + T moment, they are unknown quantities and should be evaluated.
The computation is conducted by applying small increments of the equivalent
strain to region A. In order to obtain sufficiently accurate results, these increments
should remain small. During the numerical tests performed by the author, ε(A) =
10−3 ÷ 10−4 has proven to be a good selection range.
Due to the fact that ρ (A) uniquely defines the ratio of the principal stresses in
region A, the parameter ζ (A) should be evaluated only once, namely at the beginning
of each strain path. This task is accomplished by solving the equation (see Eqs.
(3.42) and (3.43))

 
ρ (A) · G1 ζ (A) − G2 ζ (A) = 0 (3.66)
192 3 Formability of Sheet Metals

with respect to the unknown ζ (A) . In general, numerical procedures must be used
to evaluate ζ (A) . During the tests performed by the author, the bisection method has
worked very well, especially when combined with a bracketing strategy.
As soon as ζ (A) is known, the increments of the principal strains in region A can
be evaluated from Eq. (3.45) rewritten as
 
εα(A) = ε(A) · Gα ζ (A) , α = 1, 2. (3.67)

(A)
One may also notice that, for a given strain path, εα (α = 1, 2) are constant
quantities and should be computed only once.
At this stage, the parameters associated to the non-defective area of the sheet
metal can be updated using the formulae
T+T (A)
ε = T ε(A) + ε(A) , T+T (A)
εα = T εα(A) + εα(A) , α = 1, 2. (3.68)
We are now prepared to evaluate the groove parameters corresponding to the
T + T moment. If −1 < ρ (A) ≤ 0 (left branch of the forming limit curve), the
principal stress ratios are the same in regions A and B (see Eq. (3.61)). In this case,
only the increment of the equivalent strain ε(B) should be found as a solution of
Eq. (3.63) written for the T + T moment:
 
T+T
Y T+T ε(A) = T+T f · T+T Y t ε(B) + ε(B) , if − 1 < ρ (A) ≤ 0,
(3.69)
where the current thickness ratio T+T f is expressible from Eqs. (3.24) and (3.29)

T+T s(B) 
T+T f = T+T = 0 f exp T+T ε (B) − T+T ε3
(A)
=
3
s(A)
 (3.70)
0f exp T+T ε (A) (A)
+ T+T ε2
(B)
− T ε1
(B)
− T ε2
(B)
− ε1
(B)
− ε2
1

(B)
with εα (α = 1, 2) resulting from Eqs. (3.48) and (3.61):
 
εα(B) = ε(B) · Gα ζ (A) , if − 1 < ρ (A) ≤ 0, α = 1, 2. (3.71)

Equation (3.69) can be solved only in a numerical manner. During the tests per-
formed by the author, the bisection method has proven excellent performances in
combination with a bracketing strategy. After ε(B) is determined, the increments
of the principal strains in region B can be easily evaluated from Eq. (3.71).
In the case 0 < ρ (A) ≤ 1 (right branch of the forming limit curve), the prin-
cipal stress ratio associated to region B is no longer constant. As a consequence,
two unknown quantities should be determined. They are the current principal stress
ratio T+T ζ (B) and the increment of the equivalent strain ε(B) . Fortunately, the
strain-rate along the necking band does not vanish if 0 < ρ (A) ≤ 1. Under such
circumstances, Eq. (3.65) can be put in an incremental form and used to express
3.4 Theoretical Predictions of the Forming Limit Curves 193

ε(B) as a dependency on T+T ζ (B) (see also Eq. (3.67)):

(A)
ε2
ε(B) = T+T ζ (B)
, if 0 < ρ (A) ≤ 1. (3.72)
G2

ε(B) given by Eq. (3.72) should be replaced in Eq. (3.64) written for the T +T
moment. We thus obtain
T+T Y T+T ε (A) F ζ (A) =
&
 (A)
9
T+T f · T+T Y T ε (B) + ε2
G2 [T+T ζ (B) ]
F T+T ζ (B) , if 0 < ρ (A) ≤ 1.
(3.73)
The current thickness ratio T+T f is still defined by Eq. (3.70), but the principal
(B)
strain increments εα (α = 1, 2) result now from a more complicated flow rule
(see Eqs. (3.48) and (3.72)):

 G1 T+T ζ (B)
(B) (A) (B) (A)
ε1 = ε2 T+T ζ (B)
, ε2 = ε2 , if − 1 < ρ (A) ≤ 0. (3.74)
G2

In conclusion, Eqs. (3.70) and (3.74) will bring Eq. (3.73) to a formulation
involving only T+T ζ (B) as unknown. Again, the numerical solution can be found
using the bisection method combined with a bracketing strategy. After T+T ζ (B) is
determined, Eqs. (3.72) and (3.74) allow the evaluation of the increments ε(B) and
(B)
εα (α = 1, 2), respectively.
At this stage, the parameters associated to the defective area of the sheet metal
can be updated using the formulae

T+T (B)
ε = T ε(B) + ε(B) , T+T (B)
εα = T εα(B) + εα(B) , α = 1, 2. (3.75)

The procedures described above with reference to the left and right branches of
the forming limit curve are simple and efficient. In all cases, the problem consists in
solving a unique non-linear equation. At the level of region A, it is always possible
to find a solution by numerical techniques. Region B needs a more careful treat-
ment from this point of view. Generally, strains accumulate faster in the groove. As
described above, the model tries to enforce the equilibrium of the tractions along the
interface with the non-defective area of the sheet metal. At higher strain levels, the
bearing capability of the groove can be limited by the hardening law. In such cases,
it is not possible to find the solution at the level of region B. The bearing limitation
can be trapped by testing the value of the equivalent strain increment ε(B) during
the bracketing procedure. If the search for an initial guess fails even for very large
increments ε(B) , we can be sure that region B has already attained its bearing limit.
From the mechanical point of view, this situation corresponds to the occurrence of
the necking phenomenon in the groove. As a consequence, the current values of the
194 3 Formability of Sheet Metals

Fig. 3.51 H–N prediction

versus experiments [159] for
AA5182-T4 aluminium alloy

principal strains in region A should be considered as defining the limit state of the
sheet metal.
The occurrence of the necking must be also checked & after finding a numeri-
cal solution for the groove. Normally, the ratio ε(B) ε(A) should be tested. If
this quantity becomes very large (ε(B) ε(A) > 100, for example), we may con-
&

clude that the necking has been initiated. The inspection of the strain path should
be stopped as the current& values of the principal strains in region A define the limit
state. If the ratio ε(B) ε(A) is not great enough, the computation will continue
after applying a new increment of the equivalent plastic strain ε(A) to region A.
Different formulations of the equivalent stress (von Mises, Hill48, Barlat89, and
BBC 2005) and hardening laws (Hollomon, Swift, Voce, Ghosh, Hockett-Sherby,
and AutoForm) have been implemented in the strain localization model presented
above. In all cases, the numerical tests have shown a very good stability and
robustness of the solution procedure. In order to validate the performances of the
computational algorithm, its predictions have been compared with the experimen-
tal data corresponding both to steel and aluminium alloys [158]. As an example,
Fig. 3.51 shows the comparison between the numerical results and the experimental
data included in the Benchmark 1 of the NUMISHEET 2008 conference [159] for
the case of the AA5182-T4 aluminium alloy.

3.4.5 Linear Perturbation Theory

Dudzinski and Molinari [145] proposed the use of the linearized theory of sta-
bility in order to determine the limit strains. Boudeau [160] has used this theory
for calculating the FLC’s. Localized necking is considered in this analysis as a
local instability problem associated with local equilibrium equations and cotitutive
equations of the material for a given state of strain and stress.
3.4 Theoretical Predictions of the Forming Limit Curves 195

3.4.6 Modified Maximum Force Criterion (MMFC)

The ‘Modified Maximum Force Criterion’ (MMFC) for diffuse necking proposed
by Hora et al. [142] is based on Considère’s maximum force criterion. The idea
behind the MMFC-Model is to factor in an additional increase in hardening, which
is triggered by the deviation from the initial, homogeneous stress condition—e.g.
uniaxial tension—to the stress condition of local necking and with this to the point
of plane strain (Fig. 3.52) [161].
The mathematical expression of the criterion is:

∂σ11 ∂σ11 ∂β
+ = σ11 (3.76)
∂ε11 ∂β ∂ε11

Herein, β represents the strain rate ratio given by

ε̇22
β= (3.77)
ε̇11
The MMFC model will be written ina form independent of the yield criterion, i.e.
it can accommodate any yield criterion. According to Hora et al. [142] the following
relations are defined:
σ22 σ11
α= , σ̄ = , ε̄ = g (β) ε11 . (3.78)
σ11 f (α)

The stress ratio α takes the values 0 ≤ α ≤ 1, i.e. it ranges from uniaxial tension
(α = 0) to equibiaxial tension (α = 1). σ̄ is the equivalent stress defined by the
yield criterion which is utilized in the necking analysis, see below. ε̄ is the equivalent
plastic strain.

g (α) = f (α) [1 + α · β (α)] (3.79)

Fig. 3.52 Basic principle

of the MMFC criterion [161]
196 3 Formability of Sheet Metals

The function f (α) is obtained from:

1
f (α) = (3.80)
σ̄ (σ11 = 1, σ22 = α)

Assuming the instantaneous yield stress is represented by the Swift hardening law
Hora’s necking criterion then reads [142]

f ′ (α)·g(α)·β(α)
Y ′ (ε̄) · f (α) · g (α) − Y (ε̄) · = f (α) Y (ε̄) (3.81)
β ′ (α)ε̄

with β ′ = dβ/dα, f ′ = df /dα, Y ′ = dY/dε̄.

The primary unknown variable ε̄ can easily be calculated as the root of the neck-
ing criterion given in Eq. (3.81) (which is, in general, a non-linear equation) by
means of Newton’s method. Once the equivalent plastic strain at the onset of neck-
ing for a chosen linear strain path is calculated from Eq. (3.81) the major and minor
plastic in-plane strains corresponding to the onset of necking are found from

∗ ε̄ ∗ ∗ ∗
ε11 = , ε22 = β · ε11 (3.82)
g

ε̄∗ is the root of the necking criterion Eq. (3.81).

In order to take into account the influence of the thickness on the limit strains,
an enhanced MMFC (eMMFC) has been proposed recently by Hora and his co-
workers [162]. A term is added to the original formulation (3.76). The eMMFC is
expressed as

∂σ11 t  ∂σ ∂β
11
1+ + e(E, t) + ≥ σ11 (3.83)
∂ε11 2r ∂β ∂ε11

 tp is the thickness, r is the sheet curvature radius and e(t, E = const) =
where,
E0 tt0 represent the influence of the thickness. The parameters E0 , p and t0 are
determining using experimental data [161].
Banabic and Soare [163] make more precise statements about the nature of the
numerical instability of the MMFC model, asses the predictive capabilities of the
criterion, and introduce a fitting parameter for its plane strain calibration. In order
to improve the prediction of limit strains using MMFC model, Paraianu et al. [164]
chose to introduce two fitting coefficients in the original model.
The advantage of the MMFC criteria can be found in their independence of the
inhomogeneity assumption. This criterion could be used to calculate FLC for non-
linear strain path. A drawback of the MMFC models is the fact that it contains
a singularity that emerges if the yield locus contains straight line segments, like
Barlat 2000 or BBC 2005, respectively [165]. Comsa and Banabic [166] removed
this limitation of the MMFC criterion by modifying the initial formulation. As an
3.5 Commercial Programs for FLC Prediction 197

Fig. 3.53 FLC of the

AA2090-T3 aluminium alloy
predicted by classic and new
MMFC models

example, the singularity noticed by Aretz [165] in the case of the AA2090-T3 alu-
minum alloy is no more present when using the new formulation proposed in [166]
(see Fig. 3.53).

3.5 Commercial Programs for FLC Prediction

Based on the above mentioned models have been developed more commercial
programs for the limit strains prediction.
Hora and his co-workers (http://www.forminnotech.ethz.com) have developed
MATFORM code based on the MMFC model [142, 161, 162]. This code is able to
calculate and plot the limit strains and also the visualization of the strain hardening
curve and yield loci using Hill 1948, Hill 1979, Hill 1990 and Barlat 1989 crite-
ria. The program is useful for evaluation of most common experiments like tensile,
bulge, Miauchi, torsion dilatometer and tube hydroforming tests. The program is
very well documented and is able to export the constitutive models in FEM spe-
cific form for the application in the mostly spread FEM-codes like AutoForm or
PamStamp.
Using the CRACH algorithm (based on the Marciniak–Kuczynski model), Gese
and Dell [167] have developed two software: CrachLAB, a product for prediction
of the initial FLC and CrachFEM a product for coupling with the FEM codes.
Criteria for ductile and shear fracture have been included in CrachFEM to cover
the whole variety of fracture modes for sheet materials. The material model used to
calculate instability describes: the initial anisotropy (using Hill 1948 criterion), the
combined isotropic-kinematic hardening and the strain rate sensitivity. CrachFEM
is now included in the FEM codes PamStamp and PamCrash of ESI Group.
198 3 Formability of Sheet Metals

3.5.1 FORM-CERT Program

Based on a Marciniak–Kuckzynski model, Jurco and Banabic [168–170] have devel-
oped so-called FORM-CERT commercial code. The BBC 2005 yield criterion is
implemented in this model. This yield criterion can be reduced to simpler formula-
tions (Hill 1948, Hill 1979, Barlat 1989, etc). In this way, the yield criterion can be
also used in the situations when only 2, 4, 5, 6, or 7 mechanical constants are avail-
able. The program consists in four modules: a graphical interface for input, a module
for the identification and visualization of the yield surfaces, of the strain hardening
laws and a module for calculating and visualizing the forming limit curves. The
numerical results can be compared with experimental data, using the import/export
facilities included in the program. FORM-CERT can be used as a standalone appli-
cation for calculating FLC’s and comparing them with experimental data, or as an
auxiliary tool for the finite-element simulation of sheet metal forming processes. In
its current structure, the program offers useful functionalities both for research and
industrial laboratories. A short description of this program will be presented in the
next sections.
The program FORM-CERT developed in the CERTETA research centre consists
in the following modules:

• Identification module associated to the yield criterion (responsible for evaluating

the coefficients of the yield criterion, as well as for the graphical output of the
yield locus and planar distribution of the yield stress and r-coefficient)
• Module for calculating and displaying the strain hardening law
• Module for calculating and displaying the forming limit diagram.

Figure 3.54 shows a structural diagram of the program. This diagram presents
the modules mentioned above, as well as their interaction. We shall describe next
the functionality of each module.

Fig. 3.54 Structure of the FORM-CERT program

3.5 Commercial Programs for FLC Prediction 199

3.5.1.1 Identification Module Associated to the Yield Criterion

The yield surface is described in a general manner using the BBC 2005 yield crite-
rion [2.53]. The BBC 2005 yield criterion can be reduced to simpler formulations
(Hill 1948, Barlat 1989, etc.) by enforcing equalities between some of its coeffi-
cients (see Sect. 2.5.3). In this way, the general yield criterion can be also used in the
situations when only 2, 4, 5, 6, or 7 mechanical constants are available. The perfor-
mances of the BBC 2005 yield criterion for different identification strategies have
been evaluated and presented in [171, 172]. Figure 3.55 shows the user graphical
interface of this module. Here is the place where the user can choose the identi-
fication strategy by specifying the number of mechanical parameters available for
input. The input boxes can be filled from keyboard but the program also offers the
possibility to import data from an ASCII file. The module performs the identifica-
tion, displays the values of the coefficients and stores them into an ASCII output
file. Optionally, the values of the coefficients can be passed to the FLC calculation
module.
The identification module also offers the possibility to plot the computed yield
locus. Different yield loci obtained from different identification strategies (with
more or less coefficients) can be superimposed on the same diagram. A separate
panel of the graphical interface (‘Experimental data’) can be accessed for importing
experimental points from ASCII files. This data is also shown on the same dia-
gram. In this way, the user has the possibility to evaluate the performances of the

Fig. 3.55 Graphical user interface of the identification module

200 3 Formability of Sheet Metals

yield criterion in different identification cases and to select the best formulation. The
identification module also allows studying the sensitivity of the yield locus to the
variation of the input data.

3.5.1.2 Calculation and Displaying the Planar Distribution of the Uniaxial

Yield Stress and r-Coefficient
As it is well known (see Chap. 2), the knowledge of the yield locus is not suffi-
cient for evaluating the performances of a yield criterion. The planar distribution
of the uniaxial yield stress and r-coefficient offers useful information related to this
aspect. The accuracy of the constitutive models is strongly dependent on their capa-
bility to describe such variations. In fact, the residual stresses, the springback and
the thickness distribution obtained from FE calculation is closely connected to the
performances of the yield criterion. As a consequence, the program developed by
the authors includes a module that computes and displays the distribution of the uni-
axial yield stress and r-coefficient in the plane of the sheet metal (see Fig. 3.56). Of
course, the diagrams are calculated using the type of the yield criterion selected in
the identification module. A detailed description of the computational procedure is
given in Ref. [53] from Chapter 2. Again, experimental points can be superimposed
on the same diagram by accessing the panel ‘Experimental data’. The module also
offers the possibility to perform a sensitivity analysis of the curves to the variation
of the input data (coefficients of the yield criterion).

Fig. 3.56 Graphical user interface of the module used for displaying uniaxial yield stresses and
r-coefficient distributions
3.5 Commercial Programs for FLC Prediction 201

3.5.1.3 Calculation and Displaying the Strain Hardening Law

A separate module of the program has been developed for calculating and displaying
to types of hardening laws: Swift and Voce, respectively.
The expression of the hardening laws are as follows:

• Swift:

Y(ε̄p ) = K(ε̄0 + ε̄ p )n (3.84)

• Voce:
p)
Y(ε̄p ) = B − (B − A)e(−mε̄ (3.85)

In Eqs. (3.84) and (3.85), K, n, ε̄0 , B, A and m are material parameters. In these
formulations, Y is chosen to be the uniaxial yield stress associated to the rolling
direction.
The graphical interface allows the user to choose the type of the hardening law
and plots the associated curve (see Fig. 3.57). The predictions offered by different
hardening laws can be superimposed on the same graph and also compared with
experimental data (acquired via the ‘Experimental data’ panel). This module offers
the possibility to adopt the most accurate hardening law.

Fig. 3.57 Graphical user interface of the module used for calculating and displaying the strain
hardening rule
202 3 Formability of Sheet Metals

Fig. 3.58 Graphical user interface of the FLD module

3.5.1.4 Calculation and Displaying the FLC

This is the main module of the program. The Hutchinson-Neale model (see Sect.
3.4.4) of the strain localization process has been implemented in the FORM-CERT
program.
The graphical user interface provided by this module is divided in two regions
(Fig. 3.58). The first one receives the input data: coefficients of the yield criterion
and strain hardening law calculated by the modules mentioned bellow, a parameter
specifying the thickness non-homogeneity factor and also the value of the strain
increment used for computing the FLC.
At present, the FLD module works only for linear strain paths. The second
region of the graphical user interface is used for plotting the FLC predicted.
Several curves can be superimposed on the same diagram and also compared
with experimental data (imported from ASCII files via the ‘Experimental data’
panel).

3.5.1.5 ‘Experimental Data’ Module

This module is structured as a panel for acquiring input data (Fig. 3.59). The user
has the possibility to type this data or to import it from ASCII files. The data
can be plotted on diagrams or exported to other modules of the FORM-CERT
3.6 Semi-empirical Models 203

Fig. 3.59 Graphical user interface of the ‘experimental data’ module

program. All the diagrams generated by the modules mentioned above can be pro-
cessed and also exported in different graphical formats (Bitmap, Windows Metafile,
GIF, JPEG, Postscript, PDF, etc.). In addition, the results of the computations
can be exported in a numerical format (via ASCII, XML, Excel, and HTML
files).

3.6 Semi-empirical Models

Despite all the recent enhancements of the computational models (see Sect. 3.4),
they are not able to give very accurate predictions of the limit strains in all the cases
encountered in practical applications (different materials, thickness, forming rates,
temperatures, strain paths, etc.). Due to this fact, the commercial finite-element
codes still make use of experimental FLD’s or FLD’s calculated with semi-empirical
models. Some of the widely-used semi-empirical models will be presented next.
Keeler and Brazier [104] proposed an empirical relationship for calculating the
limit strains corresponding to plane strain, ε10 :
n
ε10 (%) = (23.3 + 14.13 · t) (3.86)
0.21
where t is the sheet thickness (t ≤ 3 mm).
204 3 Formability of Sheet Metals

Assuming that the shape of the FLD remains the same and having determined
the value of ε10 , it is possible to obtain the FLD by translating the Keeler–Goodwin
curve along the vertical coordinate axis.
Cayssials [173, 174] developed the Keeler–Brazier model by including both the
coefficient of strain-rate sensitivity m and the ‘internal damage’ parameters. The
limit strain is the solution of the equation

a (ε10 − n)3 + b (ε10 − n)2 + c (ε10 − n) − 10 · mt = 0 (3.87)

where a, b and c are material constants. As a first approximation, ε10 can be

expressed as follows:

ε10 = n + 5mt (3.88)

Cayssials and Lemoine [175] have extended the formulation (3.87) by including
the anisotropy coefficient and has been obtained:
√
(2+4r)
a(ε10 − n)3 + b(ε10 − n)2 + c(ε10 − n) − −14 √(r+1)(r+2) mt = 0 (3.89)

where, r is the anisotropy coefficient.

More, by coupling the former model with the Stören and Rice model [144] has
been possible the extension of the new model also for drawing and stretching areas.
So, the new model is able to predict the FLC for the complete domain, both for lin-
ear and non-linear strain paths, using only the mechanical parameters (yield stress,
strain hardening coefficient, strain rate sensitivity index and anisotropy coefficient)
and thickness of the material. The results obtained are in very good agreement with
the experimental data for new grades of steel alloys (UHSS, DP, TRIP etc.) [175].

References
1. Eary DF (1974) Techniques of press working sheet metals. Prentice Hall, London
2. Dieter GE, Kuhn HA, Semiatin SL (2003) Handbook of workability and process design.
ASM International, Metals Park, OH
3. Kumpulainen J (1984) Factors limiting the formability of sheet metals. PhD Thesis, Helsiki
University of Technology, Helsinki
4. Marciniak Z (1965) Stability of plastic shells under tension with kinematic boundary
condition. Archiwum Mechaniki Stosorwanej 17:577–592
5. Marciniak Z (1971) Limits of metal sheets formability. WNT, Warsaw (in Polish)
6. Marciniak Z (1978) Sheet metal forming limits. In: Koistinen DP, Wang NM (eds)
Mechanics of sheet metal forming. Plenum Press, New York/London, 215–235
7. Marciniak Z (1984) Assessment of material formability. In: Advanced Technology of
Plasticity. Proceeding of the 2nd ICTP, Tokyo, 685–694
8. Marciniak Z, Duncan JL, Hu SJ (2002) Mechanics of sheet metal forming. Butterworth-
Heinemann, Oxford
9. Cockroft MG, Latham DJ (1968) Ductility and the workability of metals. Journal of the
Institute of Metals 96:33–39
References 205

10. Banabic D, Dörr RI (1992) Formability of thin sheet metals. OIDICM, Bucharest (in
Romanian)
11. Pearce R (1978) Sheet metal testing – From the 19th century until now. Proceedings of the
Biennial Congress of the IDDRG, Warwick, 355–362
12. Pearce R (1982) 4000 Years of sheet metal forming. In: Newby JR, Niemeier BA (eds)
Formability of metallic materials 2000 AD, ASTM, Chicago, 3–18
13. Siebel E, Beisswanger H (1955) Deep-drawing. Carl Hanser, München (in German)
14. Considere A (1985) Application of iron and steel in constructions. Annales des Ponts et
Chaussees 9:574–575 (in French)
15. Erichsen AM (1914) A new test for thin sheets. Stahl und Eisen 34:879–882 (in German)
16. Pomey G (1976) Formability of thin sheet, Vols 1–3, Collection I.R.S.I.D.-O.T.U.A, Paris
(in French)
17. Averkiev AI (1985) Formability methods for metal sheets. Masinostroenie, Moscow (in
Russian)
18. Banabic D, Dörr RI (1995) Modeling of the sheet metal forming. Transilvania Press, Cluj
Napoca (in Romanian)
19. Banabic D et al. (2000) Formability of metallic materials. Springer, Heidelberg
20. Col A (2000) Forming limit diagrams: A survey. In: Banabic D (ed) Proceedings of the Cold
Metal Forming Conference, Printek, Cluj Napoca, 100–118
21. Ghosh AK, Hecker SS, Keeler SP (1985) Sheet metal forming and testing. In: Dieter FE (ed)
Workability testing technique. ASM, Metals Park, OH, 135–195
22. Hayashi H (1997) Co-operative research on forming limit diagram in JDDRG. IDDRG WG
Meeting, Haugesund
23. Hasek V (1973) On the strain and stress states in drawing of large un-regular sheet metal
components. Berichte aus dem Institut für Umformtechnik, Universität Stuttgart, Nr. 25,
Girardet, Essen (in German)
24. Miles M (2006) Formability testing of sheet metals. In: Semiatin SL (ed) ASM hand-
book, Vol 14B, Metalworking: Sheet metal. ASM International, Materials Park, OH,
673–698
25. Narashimhan K, Nandedkar VM (1996) Formability testing of sheet metals. Transactions of
the Indian Institute of Metals 49:659–676
26. Parniere P, Sanz G (1976) Formability of thin sheet. In: Baudelet B (ed) Metal forming of
the metals and alloys. Editions CNRS, Paris, 305–330 (in French)
27. Petrasch W (1951) Deep-drawing test by impact. Mitt. Forsch.-ges. Blechbearbeitung,
209–211 (in German)
28. Pöhlandt K (1989) Materials testing for the metal forming industry. Springer, Heidelberg
29. Pomey G, Parniere P (1980) Formability of thin sheet. Techniques de l’ingenieur, M
695:1–16, M 696:1–19 (in French)
30. Ragupathi RS (1985) Sheet metal properties and testing methods. In: Lange K (ed)
Handbook of metal forming. SME, Dearborn, MI
31. Taylor B (1988) Evaluation of formability for secondary (sheet) forming. In: Semiatin SL
(ed) Metals handbook, Vol 14, Forming and forging, 9th edn. ASM, Metals Park, OH,
877–900
32. Xu Y (2006) Modern formability. Measurement, analysis and applications. Hanser Gardner
Publications, Cincinnati, OH
33. Esser H, Arend H (1940) Anz. Machinenwesen 87:140–144
34. Kaftanoglu B, Alexander JM (1961–62) An investigation of the Erichsen test. Journal of the
Institute of Metals 90:457–470
35. Kayseler H (1934) On the proprieties of sheet steels. Mitt. Forsch. 2:39–42 (in German)
36. Kokkonen V, Hygren G (1959) Investigation into the accuracy of the Erichsen cupping test.
Sheet Metal Industries 36:167–178
37. Yokai M, Alexander JM (1967) A further investigation of the Erichsen test. Sheet Metal
Industries 44:466–475
206 3 Formability of Sheet Metals

38. Olsen TY (1920) Machines for ductility testing. Proceedings of the American Society for
Testing and Materials 20:398–403
39. Hecker SS (1974) A cup test for assessing stretchability. Metals Engineering Quarterly
14:30–36
40. Ghosh AK (1975) The effect of lateral drawing-in on stretch formability. Metals Engineering
Quarterly 15:53–64
41. Jovignot C (1930) Method and testing device for the study the fracture of the sheet metals
(in French). Revue de Metallurgie 27:287–291
42. Siebel E, Pomp A (1929) A new test for thin sheets. Mitt. K.W.I. 11:287–291 (in German)
43. Sachs G (1930) A new testing device for deep-drawing. Metallwirtschaft 9:213–218 (in
German)
44. Loxley EM, Swift HW (1945) The wedge drawing test. Engineering 159:38–40, 77–80,
136–138
45. Guyot J (1962) Deep-drawability of thin sheets. Dunod, Paris (in French)
46. Chung SY, Swift HW (1951) Cup-drawing from a flat blank. Proceedings of the Institution
of Mechanical Engineers 165:199–223
47. Swift HW (1954) The mechanism of a simple drawing operation. Engineering 178:431–435
48. Swift HW (1954) The mechanism of a simple deep-drawing operation. Sheet Metal
Industries 31:817–828
49. Willis J (1954) Deep-drawing: A review of the practical aspects of professor H. W. Swift’s
researches. Butterworths Scientific Publications, London
50. Fukui S, Yoshida K, Abe K (1960) Correlation among experimental values obtained in var-
ious formability tests. Scientific Papers of the Institute of Physical and Chemical Research
54:199–205
51. Yoshida K (1959) Classification and systematization of sheet-metal press-forming process.
Scientific Papers of the Institute of Physical and Chemical Research 53:126–187
52. Yoshida K, Hayashi Y (1979) Developments in research into sheet-metal forming process in
Japan. Sheet Metal Industries 55:14–17, 56:261–270
53. Sachs G (1934) Sheet metal testing with the Erichsen device. Metallwirthschaft 13:78–91
(in German)
54. Eisenkolb F (1932) Research on the assessment of the formability of thin sheets. Stahl und
Eisen 52:357–364 (in German)
55. Eisenkolb F (1949) Testing of thin sheets. Stanzereischriften 5 (in German)
56. Engelhardt W (1961) Development and first evaluation of a new method for testing deep-
drawability. PhD Thesis, TU Dresden (in German)
57. Drewes EJ et al. (1972) Forming limit diagrams and application to actual press forming.
IDDRG Meeting Group 1, Amsterdam
58. Miles MP, Siles JL, Wagoner RH, Narasimhan K (1993) A better formability test.
Metallurgical Transactions 24A:1143–1151
59. Keeler SP (1961) Plastic instability and fracture in sheet stretched over rigid punches, PhD
Thesis, Massachusetts Institute of Technology, Boston
60. Gensamer M (1946) Strength and ductility. Transactions of the American Society for Metals
36:30–60
61. Keeler SP, Backofen WA (1963) Plastic instability and fracture in sheets stretched over rigid
punches. Transactions of the American Society for Metals 56:25–48
62. Goodwin GM (1968) Application of strain analysis to sheet metal forming problems in the
press shop. Society of Automotive Engineers No. 680093, 380–387
63. Havranek B (1977) The effect of mechanical properties on wrinkling in conical shells.
Journal of Mechanical Working Technology 1:115–129
64. Arrieux R (1981) Contribution to the determination of forming limit curves of titanium and
aluminum. Proposal of an intrinsic criterion. PhD Thesis, INSA, Lyon (in French)
65. Simha CHM, Gholipour J, Bardelcik A, Worswick MJ (2007) Prediction of necking in
tubular hydroforming using an extended stress-based FLC. ASME Journal of Engineering
Materials and Technology 129:136–147
References 207

66. Simha CHM, Grantab R, Worswick MJ (2007) Computational analysis of stress-based

forming limit curves. International Journal of Solids and Structures 44:8663–8684
67. Janssens K, Lambert F, Vanrostenberghe S, Vermeulen M (2001) Statistical evaluation of
the uncertainty of experimentally characterised forming limits of sheet steel. Journal of
Materials Processing Technology 112:174–184
68. Brozzo P, De Luca B, Rendina R (1972) A new method for the prediction of the formability
limits of metal sheets. Proceedings of the 7th Biannual Congress of the IDDRG, Amsterdam
69. Hecker S (1972) A simple forming limit curve technique and results on aluminum alloys.
Proceedings of the IDDRG Congress, Amsterdam, 5.1–5.8
70. Nakazima K, Kikuma T (1967) Forming limits under biaxial stretching of sheet metals.
Testu-to Hagane 53:455–458 (in Japanese)
71. Nakazima K, Kikuma T, Hasuka K (1971) Study on the formability of steel sheets. Yawata
Technical Report No. 284, 678–680
72. ISO 12004 (2006) Metallic materials. Sheet and strip. Determination of the forming limit
curves. Part 2: Determination of forming-limit curves in the laboratory. ISO, Geneva
73. Hasek V (1978) Research and theoretical description concerning the influences on the FLDs.
Blech Rohre Profile 25:213–220, 285–292, 493–499, 617–627 (in German)
74. Montfort G (1997) Summary of experimental FLD work. BDDRG Working Group Report,
Haugesund
75. Novello F, Montfort G (1999) Summary of experimental FLD work-BDDRG working group,
IDDRG Meeting, WGIII, Birmingham, 3.1–3.14
76. Col A (2005) Forming limit curves: Are we at a turn? In: Boudeau N (ed) Proceedings of
the 24th IDDRG Congress, University of Franche Comte, Besancon, 25.1–25.10
77. Geoffrey JL (2003) Standardization of forming limit curves. IDDRG Conference, Bled
78. Hotz W (2006) European efforts in standardization of FLC. In: Hora P (ed) Numerical
and experimental methods in prediction of forming limits in sheet forming and tube
hydroforming processes. ETH Zürich, Zürich, 24–25
79. Hotz W, Timm J, Keller S (2001) Measurement of forming limit diagrams: Activities
regarding the harmonization in the aluminum industry. IDDRG German Group Meeting,
Düsseldorf
80. Hora P, Tong L (2006) Prediction of forming limits in virtual sheet metal forming-yesterday,
today and tomorrow. In: Hora P (ed) Numerical and experimental methods in prediction
of forming limits in sheet forming and tube hydroforming processes. ETH Zürich, Zürich,
8–17
81. Till ET (2001) Parameters affecting forming limits in steel sheet. Nr. 575, VDI Verlag,
Düsseldorf (in German)
82. Till ET, Walch C, Pauli H (2006) A note on forming limits and forming limit stress diagrams.
In: Santos AD, Barata da Rocha A (eds) Proceedings of the 25th IDDRG Congress, Porto,
11–18
83. Takashina K et al. (1968) Relation between the manufacturing conditions and the average
strain according to the scribed circle tests in steel sheets. La Metallurgia Italiana 8:757–765
84. Veerman C et al. (1971) Determination of appearing and admissible strains in cold-reduced
sheets. Sheet Metal Industries 687–694
85. Veerman C, Neve PF (1972) Some aspects of the determination of the FLD-onset of localized
necking. Sheet Metal Industries 421–423
86. Bragard A, Baret JC, Bonnarens H (1972) A simplified technique to determine the FLD at
onset of necking. Centre for Research in Metallurgy 33:53–63
87. d’Hayer R, Bragard A (1975) Determination of the limiting strains at the onset of necking.
Centre for Research in Metallurgy 42:33–35
88. Hecker SS (1975) Simple technique for determining forming limit curves. Sheet Metal
Industries 52:671–676
89. Kobayashi T, Ishigaki H, Tadayuki A (1972) Effect of strain ratios on the deforming limit
of steel sheet and its application to the actual press forming. Proceedings of the IDDRG
Congress, Amsterdam, 8.1–8.4
208 3 Formability of Sheet Metals

90. ∗∗∗ (1983) Methods of determining the forming limit curve. Proceedings of the IDDRG WG
III Meeting, Zurich
91. Marron G et al. (1997) A new necking criterion for the forming limit diagrams, Proceedings
of the IDDRG 1997 WG Meeting, Haugesund
92. Volk W (2006) New experimental and numerical approach in the evaluation of the FLD
with the FE-method. In: Hora P (ed) Numerical and experimental methods in prediction
of forming limits in sheet forming and tube hydroforming processes. ETH Zürich, Zürich,
26–30
93. Liebertz H et al. (2004) Guideline for the determination of forming limit curves. Proceedings
of the IDDRG Conference, Sindelfilgen, 216–224
94. Lewison DJ, Lee D (1999) Determination of forming limits by digital image process-
ing methods. Proceedings of International Body Engineering Conference and Exposition
(IBEC), Detroit, MI (Paper 01–3168)
95. Brunet M, Morestin F (2001) Experimental and analytical necking studies of anisotropic
sheet metals. Journal of Materials Processing Technology 112:214–216
96. Schatz M, Keller S, Feldmann P (2005) Experimental determination of the FLD for sheet
thickness from 2.5 to 5.0 mm. UTF Science III:1–8 (in German)
97. Feldmann P, Schatz M (2006) Effective evaluation of FLC-tests with the optical in-process
strain analysis system AUTOGRID. In: Hora P (ed) Numerical and experimental methods in
prediction of forming limits in sheet forming and tube hydroforming processes. ETH Zürich,
Zürich, 69–73
98. Friebe H et al. (2006) FLC determination and forming analysis by optical measure-
ment system. In: Hora P (ed) Numerical and experimental methods in prediction of
forming limits in sheet forming and tube hydroforming processes. ETH Zürich, Zürich,
74–81
99. Haberfield AB, Boyles MW (1973) Laboratory determined the FLC of sheet steel. Sheet
Metal Industries 50:400–411
100. Romano G, Rault D, Entringer M (1976) Utilization des courbes limites de formage et du
mode de repartition des deformations comme critere de jugement de l’amptitude d’un acier
extradoux a l’emboutissage. Mem. Sci. Rev. Mettalurgie 73:372–383
101. Hiam J, Lee A (1978) Factors influencing the FLC of sheet steel. Sheet Metal Industries
50:400–411
102. Kleemola HJ, Kumpulainen JO (1980) Factors influencing the FLD. Influence of sheet
thickness. Journal of Mechanical Working Technologies 3:303–311
103. Fukui Y, Nakanishi K (1988) Effects of sheet thickness on the in-plane sterch forming limit
in an aluminium sheet. JSME International Journal 31:679–685
104. Keeler SP, Brazier WG (1975) Relationship between laboratory material characterization
and press-shop formability. Proceedings of the Micro alloying Conference, New York, NY,
21–32
105. Grumbach M, Sanz G (1972) Influence of various parameters on forming limit curves. Revue
de Metallurgie 61:273–290 (in French)
106. Kikuma T, Nakazima K (1971) Aspects of deforming conditions and mechanical properties
in the stretch forming limits of sheet metals. Transactions of the Iron and Steel Institute of
Japan 11:827–830
107. Woodthorpe J, Pearce R (1970) The effect of the r and n upon the FLD of sheet steel.
Proceedings of the ICSTIS Conference, Tokyo, 822–827
108. Pearce R (1971) A user guide to FLD. Sheet Metal Industries 48:943–949
109. Conrad H, Demeri MY, Bhatt D (1978) Effects of material parameters including strain-rate
sensitivity of the flow stress on the stretch formability of sheet metal. In: Hecker SS, Ghosh
AK, Gegel HL (eds) Formability: Analysis modeling and experimentation. Metal Society of
AIME, New York, NY, 208–231
110. Rault D (1976) Description of the deep-drawing end connected problems (in French). In:
Baudelet B (ed) Metal forming of the metals and alloys. CNRS, Paris, 297–303
References 209

111. Ghosh AK, Hecker SS (1975) Stretching limits in sheet metals: In-plane versus out-of-plane
deformations. Metallurgical Transactions 5A:2161–2164
112. Charpentier PL (1975) Influence of the punch curvature on the stretching limits of sheet
steel. Metallurgical Transactions 6A:1665–1669
113. Shi MF, Gerdeen JC (1991) Effect of strain gradient and curvature on forming limit diagrams
for anisotropic sheets. Journal Material Shaping Technology 9:253–268
114. Lange E (1975) Die Bedeutung von Kennwerten und Verfahren zur Beurteilung
des Umformverhaltens beim Tiefziehen von Feinblechen (I). Baender Blech Rohre
5:511–514
115. Ayres RA, Wenner ML (1978) Strain and strain-rate hardening effect on punch stretching of
5182-0 aluminium at elevated temperature. Sheet Metal Industries 55:1208–1216
116. Kumpulainen JO, Ranta-Eskola AJ, Rintamaa RHO (1983) Effects of temperature an deep-
drawing of sheet metals. Transactions of ASME. Journal of Engineering Materials and
Technology 105:119–127
117. Li D, Ghosh AK (2004) Biaxial warm forming behavior of aluminum sheet alloys. Journal
of Materials Processing Technology 145:281–293
118. Abedrabbo N, Pourboghrat F, Carsley J (2006) Forming of aluminum alloys at elevated tem-
peratures – Part 2: Numerical modeling and experimental verification. International Journal
of Plasticity 22:342–373
119. van den Boogaard AH (2002) Thermally enhanced forming of aluminium sheet. Modeling
and experiments. PhD Thesis, University of Twente, Enschede
120. Drewes EJ, Martini A (1976) Einfluss der Umformgeschwindigkeit auf die
Grenzformaenderungen und die Formaenderungsverteilung von Feinblech. Archiv
fuer Eissenhuettenwessen 47:167–172
121. Percy JH (1980) The effect of strain rate on the FLD for sheet metal. Annals of CIRP
29:151–152
122. Balanethiram VS, Daehn GS (1994) Hyperplasticity-Increased forming limits at high
workpiece velocities. Scripta Metallurgica 31:515–520
123. Gerdooei M, Mollaei Dariani B (2009) Strain rate-dependent forming limit diagrams for
sheet metals. Journal of Engineering Manufacture 222:1651–1659
124. Keeler SP (1970) La formabilité est améliorée par pression hydrostatique. Machine Moderne
43–45
125. Padwal SB, Chaturvedi RC, Rao US (1992) Influence of superimposed hydrostatic tension
on void growth in the neck of a metal sheet in biaxial stress fields (I, II). Journal of Materials
Processing Technology 32:91–107
126. Gotoh M, Chung T, Iwata N (1995) Effect of out-of-plane stress on the forming limit strains
of sheet metals. JSME International Journal 38:123–132
127. Smith LM et al. (2003) Influence of transverse normal stress on sheet metal formability.
International Journal of Plasticity 19:1567–1583
128. Matin PH, Smith LM (2005) Practical limitations to the influence of through-thickness
normal stress on sheet metal formability. International Journal of Plasticity 21:671–690
129. Banabic D, Soare S (2008) On the effect of the normal pressure upon the forming limit
strains. In: Hora P (ed) Proceedings of the 7th International Conference and Workshop on
Numerical Simulation of 3D Sheet Metal Forming Processes, Interlaken, 199–204
130. Wu PD et al. (2008) Effects of superimposed hydrostatic pressure on sheet metal formability.
International Journal of Plasticity 25:1711–1725
131. Allwood JM, Shouler DR (2008) Generalised forming limit diagrams showing increased
forming limits with non-planar stress states. International Journal of Plasticity 25:
1207–1230
132. Bleck W et al. (1998) A comparative study of the forming limit diagram models for steel
sheets. Journal of Materials Processing Technology 83:223–228
133. Bressan JD (1997) The influence of material defects on the forming ability of sheet metal.
Journal of Materials Processing Technology 72:11–14
210 3 Formability of Sheet Metals

134. Hiroi T, Nishimura H (1997) The influence of surface defects on the forming-limit diagram
of steel sheet. Journal of Materials Processing Technology 72:102–109
135. Sowerby R, Johnson W (1975) A review of texture and anisotropy in relation to metal
forming. Materials Science and Engineering 20:101–111
136. Keeler SP (1988) Fifty years of sheet metal formability – Has science replaced the art?
Proceedings of the SAE Conference, Detroit, 197–204
137. Taylor B (1985) Sheet formability testing. In: Newby JR (ed) Metals handbook, Vol 8,
Mechanical testing, 9th edn. ASM, Metals Park, OH, 547–570
138. Chatfield DA, Keeler SP (1971) Technology for using sheet steel designing for formability.
Metal Progress 99:60–63
139. Lee D, Majlessi SA, Vogel JH (1988) Process modelling and simulation for sheet forming.
In: Semiatin SL (ed) Metals handbook, Vol 14, Forming and forging, 9th edn. ASM, Metals
Park, OH, 911–927
140. Swift HW (1952) Plastic instability under plane stress. Journal of the Mechanics and Physics
of Solids 1:1–18
141. Hill R (1952) On discontinuous plastic states with special reference to localized necking in
thin sheets. Journal of the Mechanics and Physics of Solids 1:19–30
142. Hora P, Tong L (1994) Prediction methods for ductile sheet metal failure using FE-
simulation. In: Barata da Rocha A (ed) Proceedings of the IDDRG Congress, Porto,
363–375
143. Marciniak Z, Kuckzynski K (1967) Limit strains in the process of stretch-forming sheet
metal. International Journal of Mechanical Sciences 9:609–620
144. Storen S, Rice JR (1975) Localized necking in thin sheets. Journal of the Mechanics and
Physics of Solids 23:421–441
145. Dudzinski D, Molinari A (1991) Perturbation analysis of thermoviscoplastic instabilities in
biaxial loading. International Journal of Solids and Structures 5:601–628
146. Considère M (1885) L’emploi du fer et Lacier Dans Les Constructions. Annales Des Ponts
et Chausses 9:574–775
147. Banabic D, Dannenmann E (2001) The influence of the yield locus shape on the limits
strains. Journal of Materials Processing Technology 109:9–12
148. Marciniak Z (1965) Stability of plastic shells under tension with kinematic boundary
condition. Archiwum Mechaniki Stosorwanej 17:577–592
149. Marciniak Z, Kuczynski K, Pokora T (1973) Influence of the plastic properties of a mate-
rial on the FLD for sheet metal in tension. International Journal of Mechanical Sciences
15:789–805
150. Hutchinson RW, Neale KW (1978) Sheet necking I. In: Koistinen DP, Wang NM (eds)
Mechanics of sheet metal forming. Plenum Press, New-York/London, 111–126
151. Hutchinson RW, Neale KW (1978) Sheet necking II. In: Koistinen DP, Wang NM (eds)
Mechanics of sheet metal forming. Plenum Press, New-York/London, 127–153
152. Hutchinson RW, Neale KW (1978) Sheet necking III. In: Koistinen DP, Wang NM (eds)
Mechanics of sheet metal forming. Plenum Press, New York/London, 269–285
153. Banabic D et al. (2004) FLD theoretical model using a new anisotropic yield criterion.
Journal of Materials Processing Technology 157–158:23–27
154. Butuc MC, Barata da Rocha A, Gracio AA, Ferreira Duarte J (2002) A more general
model for forming limit diagrams prediction. Journal of Materials Processing Technology
125–126:213–218
155. Ganjiani M, Assempour A (2008) Implementation of a robust algorithm for prediction of
forming limit diagrams. Journal of Materials Engineering and Performance 17:1–6
156. Banabic D, Barlat F, Cazacu O, Kuwabara T (2007) Anisotropy and formability. In:
Chinesta F, Cueto E (eds) Advances in material forming – ESAFORM 10 years on. Springer,
Heidelberg, 143–173
157. Banabic D, Barlat F, Cazacu O, Kuwabara T (2010) Advances in anisotropy and formability.
International Journal of Materials Forming (accepted)
References 211

158. Bichis I, Dragos G, Paraianu L, Comsa DS, Banabic D (2010) Robust algorithm for the
solution of the Marciniak–Kuczynski strain localization model (in preparation).
159. Volk W, Illig R, Kupfer H, Wahlen A, Hora P, Kessler L, Hotz W (2008) Benchmark 1 –
Virtual forming limit curves. In: Hora P (ed) Proceedings of the 7th International Conference
and Workshop on Numerical Simulation of 3D Sheet Metal Forming Processes, Part 2,
Interlaken, 3–9
160. Boudeau N (1995) Prediction of instability in local elasto-plastic instabilities. PhD Thesis,
University of Franche-Compte, Besancon (in French)
161. Hora P, Tong L (2006) Numerical prediction of FLC using the enhanced modified maxi-
mum force criterion (eMMFC). In: Hora P (ed) Numerical and experimental methods in
prediction of forming limits in sheet forming and tube hydroforming processes. ETH Zürich,
Zürich, 31–36
162. Hora P, Tong L, Reissner J (2003) Mathematical prediction of FLC using macro-
scopic instability criteria combined with micro structural crack propagation models.
In: Khan A (ed) Proceeding of Plasticity’03 Conference, Neat Press, Fulton, MD,
364–366
163. Banabic D, Soare S. (2009) Assessment of the modified maximum force criterion for
aluminum metallic sheets. Key Engineering Materials 410–411:511–520
164. Paraianu L, Dragos G, Bichis I, Comsa DS, Banabic D (2009) An improved version of the
modified maximum force criterion (MMFC) used for predicting the localized necking in
sheet metals. Proceedings of the Romanian Academy 10:237–243
165. Aretz H (2004) Numerical restrictions of the modified maximum force criterion for predic-
tion of forming limits in sheet metal forming. Modelling and Simulation in Materials Science
and Engineering 12:677–692
166. Comsa DS, Banabic D (2010) A new formulation of the modified maximum force criterion
to avoid the numerical instabilities (in preparation)
167. Gese H, Dell H (2006) Numerical prediction of FLC with the program CRACH. In: Hora P
(ed) Numerical and experimental methods in prediction of forming limits in sheet forming
and tube hydroforming processes. ETH Zürich, Zürich, 43–49
168. Banabic D (2006) Numerical prediction of FLC using the M-K-Model combined with
advanced material models. In: Hora P (ed) Numerical and experimental methods in pre-
diction of forming limits in sheet forming and tube hydroforming processes. ETH Zürich,
Zürich, 37–42
169. Jurco P, Banabic D (2005) A user-frienldy program for calculating forming limit diagrams.
In: Banabic D (ed) Proceedings of the 8th ESAFORM Conference on Material Forming,
Cluj Napoca, 423–427
170. Jurco P, Banabic D (2005) A user-friendly program for analyzing the anisotropy and forma-
bility of sheet metals. In: Boudeau N (ed) Proceedings of the IDDRG 2005 Conference,
Besancon, 26.1–26.8
171. Banabic D, Paraianu L, Cosovici G, Jurco P, Comsa DS, Vos M (2003) Performances of
the BBC 2003 yield criterion when using data obtained from different mechanical tests. In:
Gyenge C (ed) Proceedings of the MTeM 2003 Conference, Cluj-Napoca, 23–26
172. Banabic D, Comsa DS, Paraianu L, Cosovici G, Jurco P (2003) Prediction of the yield loci
for anisotropic materials using uniaxial and plane-strain tensile tests. Proceedings of the
MSE2003 Conference, Sibiu, 11–15
173. Cayssials F (1998) A new method for predicting FLC. Proceedings of the IDDRG Congress
Meeting Working Group III, Brussel, 1–6
174. Cayssials F (1999) The version of the ‘Cayssials’ FLC model. Proceedings of the IDDRG
Meeting Working Group III, Birmingham, 1–7
175. Cayssials F, Lemoine X (2005) Predictive model for FLC (Arcelor model) upgraded to
UHSS steels. In: Boudeau N (ed) Proceedings of the IDDRG Conference, Besancon,
17.1–17.8
Chapter 4
Numerical Simulation of the Sheet Metal
Forming Processes

List of Special Symbols

a, m, p exponents in the Swift-Hockett/Sherby strain hardening law
C coefficient in the Swift-Hockett/Sherby strain hardening law
M exponent in the BBC 2005 yield criterion
Rm tensile strength
Rp0.2 yield stress
r0 , r45 , r90 anisotropy coefficients in the rolling, diagonal and transversal
directions
rb biaxial anisotropy coefficient
α coefficient in the Swift/Hockett/Sherby strain hardening law
ε0 pre-strain
ε pl plastic strain
σ equivalent stress
σb biaxial yield stress
σi initial stress
σ Sat saturation stress
σ 0 , σ 45 , σ 90 yield stresses in the rolling, diagonal and transversal directions

4.1 AutoForm Solutions

4.1.1 The Role of Simulation in Process Planning

Much attention is given to the theoretical and technological methods behind stamp-
ing simulation models; however the applicability of these models in the productive
completion of a comprehensive digital process plan is often overlooked. The role
that simulation models play in the development process must consider the alignment
of software usage, the available inputs, and the desired outputs supporting the work
process, in addition to the software capabilities.
Each of the finite element analysis technologies discussed in previous chapters
of this book describes the technological differences between the available stamping

D. Banabic, Sheet Metal Forming Processes, DOI 10.1007/978-3-540-88113-1_4, 213


C Springer-Verlag Berlin Heidelberg 2010
214 4 Numerical Simulation of the Sheet Metal Forming Processes

simulation solvers. Beyond the technology of the solvers, the time at which these
tools are applied greatly changes the utility of these tools in achieving the desired
outcome. Using finite element analysis to determine if a sheet metal part can be
formed requires inputs and boundary conditions for the definition of the mathemat-
ical model. As a digital process layout evolves the availability of information and
assumptions to be used in the pre-processing of the simulation model greatly deter-
mines if the output from the simulation code can be trusted or useful. Certainly, no
software code will provide any result if the inputs are incomplete and some solver
methodologies require fewer inputs to converge. To this end, we should perhaps not
discuss solver technology but rather the types of desired simulation outcomes:

• Part Feasibility (geometry check)

• Manufacturability (process validation)
• Process Capability (production robustness)

Each of these outcomes does potentially benefit from the application of differ-
ent simulation tools, but the most distinguishing characteristics are the assumptions
and included inputs to the simulation model. The broader the assumptions, regard-
less of the selected solver methodology, the more likely it is that reality will differ
from the simulation predictions. The earlier simulation is attempted, the broader the
assumptions must be; as the process matures assumptions lead to design decisions
and production process parameterization, thereby allowing for improved accuracy
in the simulation. The more realistic the alignment of simulation assumptions with
the reality of the stamping environment, the more accurate the simulation results
can be—this is not a solver technology constraint but instead a condition of defining
the modeled environment.
Simulation results viewed outside the context of the simulation inputs and
assumptions made in compiling without inputs cannot be trusted to fulfill any of
the above desired outcomes. It is the existence of the results in light of the assumed
conditions of the simulation that make the simulation results valuable. The avail-
ability of input and output data varies with time in the design process. The known or
assumed parameters at the time of simulation define and differentiate the following
levels of maturity of simulation outcome

• Part feasibility: Feasibility Solution evaluating part geometry and material

specification to deliver the product function
• Manufacturability: Tooling and Tryout Solution evaluating part and tooling
geometry, planned production process, and material specification
• Process capability: Robustness Solution demonstrating consistent manufactura-
bility in a variable production environment; a sustainable process can be defined
and maintained amid the expected production variability

Within the realm of possible outcomes for a given product design, we can iden-
tify that some but not all will be recognized as feasible, as illustrated in Fig. 4.1. Part
4.1 AutoForm Solutions 215

Fig. 4.1 Potential process

layout outcomes

Feasibility demonstrates that the part could be made safely—within the forming lim-
its of the material—without regard to the manufacturing process. Manufacturability
indicates that the part can be successfully made in a well defined manufacturing
process. Process Capability shows that the selected process will produce acceptable
parts over the range of variation likely for the selected manufacturing environment
and specifications. Process Capability is a subset of those process layouts that pro-
duce manufacturability. Manufacturability is often a subset of what is feasible, but
as shown in the Fig. 4.1 it can occur that a product design that is at first recog-
nized as infeasible will later prove—through the ingenuity and diligence of skilled
tool makers and designers—to be manufacturable. While this is a possibility that
those designs that are deemed infeasible will later prove manufacturable, via process
means not recognized during feasibility (geometry) checks, it is a limited potential
that is usually recognized easily with some input from manufacturing personnel.
As with any recognizable subset relationship, Part Feasibility does not guarantee
either manufacturability or process capability, nor does manufacturability guarantee
process capability. It is most likely that a capable process will be based upon a man-
ufacturable process and that manufacturable process is often a result of a recognized
feasible product. Therefore it is easy to state that any effort to reduce the number of
infeasible products early in the design process will, greatly improve the likelihood
of the desirable outcome of a fully capable manufacturing process.

4.1.2 Material Data in Digital Process Planning

Part function, as determined by the product designer/engineer, typically speci-
fies a combination of 3-dimensional product geometry and material grade. The
selected combination of product geometry and material grade serve the part
function—structural strength, assembly, packaging, crash worthiness, stiffness, ride
216 4 Numerical Simulation of the Sheet Metal Forming Processes

and handling, etc.—and are presumed to have been specified for sound reasons and
are considered to be a requirement of the part design.
To achieve the product function, a near infinite number of possibilities exist for
the process layout—manufacturing tool and process design—however, not all the
possible process layouts would deliver the specified product while achieving opti-
mal quality, cost, and production delivery (lead time) goals. Certainly, nearly any
part design can be made, given enough time and money; but as the concept of
inexhaustible time and resources is not reasonable, it must be considered that com-
promises to the acceptable quality or the desired part function must be considered.
The alternative is that the product is deemed impractical or too expensive; and never
realized.
At the time a product is first being designed its effect on cost, quality, and delivery
are often not considered by the product designer. The sole concern of the designer
is attainment of the functional or styling requirements of the product. However, to
design blindly is a waste of resources and time: to that end a simple set of feasibility
assumptions can be made in order to weed out many truly infeasible design con-
cepts. Later, as manufacturing decisions regarding the process layout are made, the
prior assumptions and therefore feasibility assessment may no longer be valid. This
has, in the past, been used as an indictment of the technology applied during the
feasibility solution. This could not be farther from the truth; it is merely testimony
to the fact that, as the process matures the rules of the game change.
At the time the product is first designed, the limitations that the product geom-
etry places on quality, cost, and lead time are not quantified. Without feedback the
product designer will create products that suit their function only. However, as the
product is prepared for manufacturing—digital process layout planning—it may
become apparent that the part is not practical perhaps due to forming difficulties,
process costs or quality reasons. At this time the product will either have to be
changed, or the process or product is sub-optimized.
In Fig. 4.2 note the comparison between a stamped component for a premium
quality niche vehicle and that same component in a quality high production vehicle.
In the figure, the outer ring (5) represents optimal attainment of the defined target
criteria, as the points shift toward the center (1) they are either compromised or
rendered sub-optimal. In the premium quality niche vehicle, note that no compro-
mise was taken in regard to part quality characteristics or function characteristics,
however this is likely the result of compromises in the area of cost and lead time. It
takes more time and money to achieve the highest level of product performance and
quality to design intent.
We see for the quality high production vehicle; compromises are made regard-
ing part geometry and function, as well as some measures of quality—Fig. 4.2.
This may be the result of an uncompromised pursuit of optimal production costs
and minimized lead time. During early feasibility, a component design for the two
vehicles would most likely be recognized as feasible for their original product
functions and specified material. As the manufacturing targets are more accurately
defined, a component for the higher production vehicle—with its optimized targets
for production costs—will have different tooling and production restraints, which
4.1 AutoForm Solutions 217

Fig. 4.2 Comparison of

impact of digital process
plans on function, quality,
cost, and lead time for:
(a) high quality niche vehicle;
(b) quality mass production
vehicle
218 4 Numerical Simulation of the Sheet Metal Forming Processes

might render the original product design poorly manufacturable. Is the simulation
tool, or the simulation analyst, to blame when the process planning forces prod-
uct concessions to assist in manufacturing? Neither. At the time, the part seemed
feasible for the selected geometry and material grade. But in the light of the new
information and limitations placed on the tooling and tryout solution, the analysis
outcome is different.
Assessing the forming of the part, without considering production requirements,
may lead to unreliable results; later when the reality of these constraints is placed
upon the part, or the process, the resulting prediction will seemingly contradict an
earlier outcome. However, there is still great value in assessing the forming of the
part based upon whatever reasonable assumptions can be made at the time. This
value should be tempered with an understanding that the subsequent evolution of
the process layout may alter the initial predictions. Finite element analysis of the
stamping process layout is but one piece of the larger question: can a particular
component be made with the required function and acceptable quality, at a reason-
able cost, on time? The future of Digital Process Planning relies on a push towards
the comprehensive analysis of quality, cost, and production metrics.

4.1.3 Feasibility (Part Feasibility)

4.1.3.1 Applied Technology
The earliest application of simulation in the evolution of process layout is the eval-
uation of the part feasibility. Traditionally, Inverse One-Step methods have served
as the primary early feasibility tool. The solver technology and the assumptions that
MUST be made to utilize the solver relegate inverse one-step solvers to this out-
come. The minimum user inputs for an inverse one-step code are a 3D model of the
sheet metal part and a material grade specification. Some argue that, with additional
boundary conditions added, the inverse one-step can perform advanced feasibility
analyses and even tooling and tryout solutions (manufacturability). However, even
with added boundary conditions, the solution is but a potential reality, because the
effect of physical responses such as tool timing, gravity induced sag, and realistic
stamping reactions cannot be fully modeled in the inverse one-step model.
While inverse one-step solver technology has been associated primarily with fea-
sibility; a broader definition of feasibility could include full incremental analysis.
The richer feedback that is available from incremental analysis can be highly useful
when attempting to narrow a given product design to the most feasible geometry;
but rapid incremental simulations performed during the feasibility stage may later be
contradicted by simulations later performed during the Tooling and Tryout solution
phase. This is not due to limitations of the rapid incremental simulation technology,
but instead recognition of the fact that simulation outcome validity is primarily a
function of the modeled process inputs rather than solver technology. Simulations
run at later stages in product development benefit from a more complete definition
4.1 AutoForm Solutions 219

of production variables, such as the blank shape and precise tooling geometry, and
that is what differentiates feasibility from manufacturability.
The application of different incremental simulation solver methods—Dynamic
Explicit and Quasi-static Implicit solvers—and the presumed accuracy of each
method, have relegated these codes to highly specialized usage in some organiza-
tions. Simulation outcomes have been deemed manufacturable based solely on the
solver methodology used, with little discussion manufacturing environment or the
inputs used. It must be understood that even the most accurate solver will provide
improbable outcome if using improbable assumptions—safe or failing results based
on incorrect process inputs do not make the design any safer. The accuracy of the
simulation has to be matched with the accuracy of the available input data. If we
do not consider the accuracy of our process inputs, discussion over solver accuracy
becomes a moot point.
We propose to define part feasibility analyses not based on the simulation tech-
nology used, but on the precision of the simulation inputs. For example, if one runs
incremental simulations with a blank that does not reflect design intent, then those
results can only be used to determine feasibility at best—not a proof of manufac-
turability or a tooling and tryout solution. Conversely, if a competent solver is used
to simulate a stamping process with the precise process inputs matching the intent
of the tools, then the resultant prediction can indicate the manufacturability of the
part.
This differentiation of part feasibility from process feasibility (manufacturabil-
ity) is analogous to the use of prototype ‘soft’ tools versus production ‘hard’ tools.
While the existence of parts that were produced from prototype tools can be taken
as a proof of part feasibility, the means by which the prototypes are produced is
often not acceptable for full scale production. The prototype tooling may use tooling
geometry that is not optimal for production stamping, lubrication may be non-
standard, trimming conducted using lasers, and final flanges hand-bent. Similarly,
we must distinguish between simulations that faithfully represent the production
environment (Manufacturability or Process feasibility) from those simulations that
make less accurate although reasonable assumptions (Part feasibility or geometry
check).

4.1.3.2 Input Data

3D CAD data and the material grade specification comprise the minimum required
inputs for early feasibility. The material model requires a hardening curve, yield
surface model, and forming limit. The CAD data should be the latest design
intent data fully faced and complete. Other optional, but value added, inputs can
include desired manufacturing orientation, boundary conditions (sheet restraint and
friction) and addendum (tool layout geometry). If an incremental software code
is to be used for an advanced feasibility solution, then additional inputs such
as the 3D tooling proposal, applied external forces and initial blank outline are
required.
220 4 Numerical Simulation of the Sheet Metal Forming Processes

One unfortunate reality of the part feasibility analysis is that the reliability of
mechanical properties for the specified material grade varies. In some cases, the
material selection is made primarily on the product function of the finished part with
little or no regard to the method of manufacture for the part. Also, the information
shared with the product designer regarding their material selection may be limited
to stiffness, minimum initial yield and tensile strengths, and total elongation—
properties pertinent to product function but not sufficient to run finite element
forming analysis. Broad assumptions, regarding the specifics of the material model,
result in predictions of feasibility that are often times overly optimistic. Product
designers select material grades that, to their knowledge, suit the final performance
requirements of their design. There may be a host of material grade specifications
that might deliver the similar initial yield and tensile strengths, but the manufactur-
ing behavior of the available materials may not be comparable, or even compatible.
Additionally, the necessary methods to make the part from the specified material
grade may not be in balance with the process layout goals for the part in regards to
cost, lead time, or the attainable product quality.
External boundary conditions can be input—such as restraint of the sheet edge,
friction, symmetry, etc.—but the correct values to use in determination of feasibil-
ity can be difficult to ascertain, as the tooling design used to attain such restraint
is still to be defined. Moreover, the physical conditions required to achieve such
boundary conditions may be unrealistic. A limitation of the use of these boundary
conditions is that the application and distribution of these forces may be unknown
by the product feasibility engineer. At best, the product designer may be reduced
to assuming a uniform distribution of some amount of restraint, or a variable dis-
tribution of restraints about the sheet boundary. It is easy to see how arbitrary the
resultant feasibility assessment can become.
For parts where drawing is the likely forming method, one potential method-
ology for ensuring part feasibility is to run a number of iterations with uniform
restraint conditions varied between free (no restraint applied) to full (boundary is
fully restrained), the resultant outputs could then be reviewed to assess at which
forming condition an observed formability anomaly is recognized as an issue for
escalation. This technique which substitutes the arbitrary constraints added to the
sheets boundary for the development of full process layout tooling surfaces—binder
and addendum—has been successfully applied by a number of Automotive OEMs.
These OEMs have shown the use of an inverse one-step as a product feasibility test
can dramatically increase the ability to define formability concerns. The formability
concerns are ultimately addressed through direct product intervention by the product
designer, or through escalation to advanced feasibility support from manufacturing
engineering.
Inverse one-step results of drawn parts can be further improved if the input is
geometry representative of a potential draw addendum and binder. The addition
of the draw die geometry can improve the validity of the resultant feasibility as
well as blank outline; however the creation of such geometry requires some tool-
ing knowledge and process awareness, as well as some ability to digitally model
reasonable tooling geometry—see Fig. 4.3 below. The additional effort can pay
dividends through improved precision in the predicted material utilization and blank
4.1 AutoForm Solutions 221

Fig. 4.3 Prediction of part feasibility could be run to consider plausible tooling

nesting, in addition to highlighting forming issues that may be a result of the

product geometry’s influence on the tooling layout; however, results can be skewed
if the process layout geometry proposed is inappropriate. The ability to augment the
analyzed product geometry with some tooling geometry—that follows some typical
norms and rules for stamping—without the product engineer having to wait oth-
ers, will allow multiple product geometry to be assessed prior to the release of the
product data.
If the solver method is an inverse one-step, then there is another input implicit to
the applied solver methodology—the part geometry is to be achieved in one action.
This limitation—that the plastic deformation will be achieved in one-step—is clear
from the description of the solver method. In cases where the part is to be formed
in multiple operations, e.g. draw with a secondary trim then reform, the results
of the inverse one-step simulation may indeed be very misleading. Consultation
with appropriate manufacturing personnel may indicate that an alternative process
is required. This too should bring about an escalation to advanced feasibility and the
use of an incremental solver to ascertain the part feasibility.
In the case of an incremental ‘advanced-feasibility’ solution the selection of ele-
ment formulation as an input becomes relevant. For accuracy, it may be argued that a
solid ‘brick’ element is potentially the most accurate available element formulation,
but most agree that it is computationally inefficient for forming analysis of a typi-
cal stamped part—the gain in result ‘accuracy’ is not proportional to the increased
modeling, computing, and post processing requirements. Additionally, at the time of
an advanced feasibility solution, the accuracy of some of the selected process inputs
is in doubt. An ‘accurate’ simulation of a process condition that is incomplete in
its assumptions is not an accurate model at all. Therefore, at the time of feasibility
determination it is possible to benefit from computational efficiencies in the element
222 4 Numerical Simulation of the Sheet Metal Forming Processes

formulation: provided that the use of the alternate element formulation does not con-
tradict the physical conditions of the assumed stamping environment. Significant
gains in computational speed can be achieved by using situation optimized element
formulations, such as Bending-Enhanced Membrane elements for draw operations
that use a full blank holder to constrain the sheet edge. Such adaptations, made in the
light of an understanding of the other inherent assumptions of the feasibility solu-
tion, are acceptable if forming concerns can be identified during the design phase,
when the greatest opportunity for cost effective correction exists.

4.1.3.3 Output Data

In the instance of using inverse one-step solvers, sometimes the best the user can
do is use the software to identify forming issues to escalate to those who run the
advanced-feasibility models. Again, this is due to the nature of the early feasibil-
ity solution. Based on the inputs used for the inverse one-step solution, the likely
combination of inputs and results may not be fully known. Attempts to resolve all
the issues via product modification may be unnecessary and only add cost to the
engineering process while not achieving the improvement in process layout that is
desired.
Recognition of these escalation issues is easily achieved via an FLD based output
differentiating the forming modes of the part and risk of failure. Formability plots
with such summary views are attainable in most commercially available inverse one-
step solvers. The forming tendencies of the part for safe, marginal, splitting failure,
compression, wrinkling, or lack of stretch will often be shaded on the simulation
mesh in different colors, coded for easy recognition.
A significant shortcoming in the inverse one-step feasibility method is recogniz-
ing at which boundary condition the issues should be identified. At free boundary
restraint very few splits should be found—those that can be immediately recognized
as critical geometry issues or indication that process modification may be needed.
Eventually for many parts complete splitting failure can be attained by increasing
the restraint until the sheet edges are locked. Does this mean that every part is safe
and unsafe at the same time? At which condition would one pronounce that the
forming issue exists?
The challenge is not only to identify that forming issues exist, but more impor-
tantly which restraints was the most likely stamping condition. The condition in the
real production tool is likely to be something in between the ‘free’ and ‘locked’
restraints. The tool and die makers would tune the die to ensure that the part was
stretched enough to avoid wrinkles but not so much that the part will split. A system
can be devised to allow the feasibility engineer to identify a minimum threshold
of stretching—between free and locked—that should be achieved. As the boundary
restraint is increased from free to fully constrained the amount of compression and
thickening related issues will decrease while the amount of thinning and splitting
issues will increase, see Fig. 4.4 above. If splitting issues arise and compression
and thickening issues still exist then one could conclude that the tools that would
4.1 AutoForm Solutions 223

Fig. 4.4 With different restraint conditions the ‘safety’ of the part changes

later be built to manufacture the product would likely induce such failure potential.
The question of forming feasibility is never a simple yes or no proposition, instead
it must be proposed with many conditions. The result should be presented in the
context of the assumptions that lead to the outcome—i.e. the part is feasible at a
stretching threshold of 2% thinning on a majority of the part surface, the part is not
feasible at the minimum threshold of stretch, the part is feasible if produced from a
fully developed blank, the part is not feasible in a single forming operation, etc.
The feasibility prediction attained can be used to highlight the geometry based
forming issues and potential resolution (i.e. product concessions, larger radii, open
walls, shallow geometry), however, it should be noted that failure to qualify the
part geometry as ‘feasible’ does not automatically disqualify product geometry as
being non-manufacturable. In the case of a feasibility result from an inverse one-
step solver, if the part is manufactured in a series of drawing or forming operations
there is a strong likelihood that the product could be successfully made if the proper
tooling and tryout solution is applied. Therefore the numerical results gathered dur-
ing the feasibility assessment may need to be taken with some skepticism. Similarly,
some parts that are deemed feasible should be dismissed when reviewed for tooling
and tryout, because the methods required to achieve the feasible product design are
undesirable. To put it simply: just because a product can be made, does not mean
that it should be made.
Feasibility, as much as we might wish it would be, is not a binary output. That a
part has passed feasibility is a statement that should be highly qualified with a com-
plete list of conditions—assumptions, inputs, and boundary conditions. The part is
feasible if these conditions are met. In the case of inverse one-step technology the
conditions include the assumption that the part is formed in only one forming oper-
ation, that the blank will be a fully developed shape, that the tool contact is uniform,
consistency of material parameters exactly as modeled, that the external forces don’t
change over time or distance. Knowing that the actual delivered parts will very likely
be manufactured differently than modeled should inform our decisions on how to
use the available results—see Table 4.1. In the case of advanced feasibility, assump-
tions will include but are not limited to the user defined blank outline, the binder
pressure, friction (lube), material parameters exactly as modeled, draw bead effect
and location, and process layout geometry.
224 4 Numerical Simulation of the Sheet Metal Forming Processes

Table 4.1 Feasibility solution outputs

Reliability

Advanced
Output Feasibility feasibility Caveats Action

Formability Moderate Moderate Some splits may be Escalate if immediate

splitting failure correctable via geometry adjustment
issues process does not resolve or
adjustment (i.e. improve. Obtain
different blank review from process
outline, holding or method planning
condition,
multi-stage
forming)
Formability Low Moderate Wrinkling behavior Escalate all known
compres- (process strongly wrinkling tendencies
sion/wrinkling layout controlled by so process layout can
issues dependent) process layout plan accordingly
Blank outline High (crash Moderate to Blank outline from Run inverse one-step
form) High inverse one-step is using tooling
Low (draw (process a reasonable start geometry and desired
parts) layout for process boundary of formed
dependent) determination sheet. Embed or nest
(tooling and and escalate to
tryout). advanced feasibility
Part only blanks to achieve higher
from one-step risk confidence in the
severe under quote blank size estimate.
material Nesting and
requirements embedding into shear
blank improves
material utilization
prediction (advanced
feasibility or tooling
and tryout solution)
Thickness (mm) Low Low to Thinning Inverse one-step
and thinning Moderate distribution will thinning results are
(%) (process change with the better than assumed
layout process layout nominal for
dependent) FEA/CAE
functionality testing,
however advanced
feasibility results will
achieve superior
results
Springback None Low Stresses from No actionable steps
feasibility based regarding springback
inputs are highly can be taken
questionable—
dependence on
process layout
4.1 AutoForm Solutions 225

4.1.4 Manufacturability (Process Validation)

4.1.4.1 Applied Technology
What is feasible is not always possible given circumstances limited by budget or
manufacturing capabilities. This may sound like semantic double speak, but instead
speaks to the difference between what we can model, and what can actually be
accomplished in today’s world. Further along in the manufacturing development
of the sheet metal stamped part is the process definition, tooling development and
design stage. Simulations performed during this stage of the product life cycle are
presumably defined to determine the manufacturability of the production intent
sheet metal stamped part. At this point in the product development process some
simulation solver technologies are more viable than others. Inverse one-step has for
the most part outlived its usefulness when moving on to consider manufacturing
feasibility—incremental solvers are the generally accepted technology to determine
manufacturability.
Incremental simulations may have already been performed on the component, but
if the process inputs used were not production intent then that resultant prediction
of forming was only a validation of the part feasibility. Once the appropriated inputs
are defined and used for conduct of an incremental simulation we can then recognize
that the outcome of the simulation is a validation to the manufacturing feasibility—
manufacturability—of the part.
What truly distinguishes part manufacturability from part feasibility is not the
simulation technology to be used, but instead the precision required in the selec-
tion and definition of the pre-processing inputs provided. The goal of a tooling and
tryout solution is to provide timely feedback on the viability of the manufacturing
process layout plan. In this regard, we can see that many process layouts that are
deemed manufacturable were a subset of those product geometries that are deemed
feasible, though some exceptions may exist. However, not all feasible parts will be
deemed manufacturable. The characteristic that separates these two outcomes is the
flexibility of the constraints in the manufacturing process layout.
Now that the inputs have been improved to emulate the production intent of
the manufacturing environment, some of the earlier assumptions made during the
advanced-feasibility also may need to be amended. Using the better inputs with
speed-optimized FEA parameters—like coarse mesh and situational optimized ele-
ment formulations—should make way for use of finer mesh and more specific
element formulations. Indeed, we still see the use of a solid element for most sheet
stamping simulations as overkill, but now the use of a full shell element can be
seen as value added, provided that the process being simulated has already shown
reasonable convergence potential with the adapted inputs. It would be unwise to
immediately throw out the speed benefit of bending enhanced membrane, when a
passing result is still several iterations away; however, once a reasonable outcome is
found, a final validation run using an Elastic-Plastic shell is a prudent step to ensure
the highest level of confidence as we move forward.
226 4 Numerical Simulation of the Sheet Metal Forming Processes

4.1.4.2 Input Data

The use of production intent input data is the key distinguishing factor that separates
a Tooling and Tryout Solution for manufacturability from a Feasibility Solution for
part feasibility. Without recognition of the inputs used as the production intent the
value of the simulation outcome is limited—part feasibility. Instead we propose
that the goal of a Tooling and Tryout solution for manufacturability is to validate
the production tooling prior to tool build. This end can only be achieved by means
of simulation that emulates in a very accurate manner the conditions that will be
present in the tooling tryout and eventual production.
At this time we start to see the larger influence of the decisions made in regard
to the comprehensive digital process plan. The blank that may have been used for
an advanced feasibility simulation may have been run without full knowledge of the
planned production line, stamping process, or material utilization targets.
Listed in Table 4.2 are many of the required inputs that must be determined—
with the goals set out for the digital process plan in consideration, at present it is
not possible to enter directly into the simulation software the cost per blank goals
of the part, nor production robustness goals and have the simulation set itself up to
achieve these ends. Instead the human operator makes value decisions regarding the
attainment of these goals and selects the inputs that appear to further the pursuit of
those goals. Returning to the goals of the digital process plan we can assume that
for lower production vehicles one-time costs such as initial tooling costs may have
greater influence on the selected blank shape than a high volume-low cost vehicle,
where material utilization may far exceed the initial tooling cost for importance. It
is no longer acceptable to determine if the part CAN be made, we now seek to prove

Table 4.2 Inputs for a manufacturability analysis-tooling and tryout solution

Tooling Geometry inputs Blank inputs Environment variables

3D tooling surface data Full material description Effective friction data

•N Lube, coating, substrate
•R combination
• YS (min-max)
• TS (min-max)
• Yield Surface
• Rolling direction
• FLC
3D trim lines Blank shape and location Applied tool forces
• In coil (pad/blankholder/binder)
• In die
3D flanging tools Initial Thickness Die closing timing
3D net points for springback Cushion travel/blank
analysis support
Bead Profile (effective
restraint)
Bead centerline
4.1 AutoForm Solutions 227

that the part will be made in our plants, using our tooling standards, to our accepted
level of part quality.

Example Case
Using a feasibility solution an automotive b-pillar was deemed a feasible product
design, the feasibility assessment was performed using inverse one-step technol-
ogy and advanced feasibility was performed using quasi-static implicit incremental
solver. As inputs to the advanced feasibility a simple draw development was
designed and a blank outline was calculated using the inverse one-step blank pre-
diction. Formability assessment and thinning strain predictions indicated that the
b-pillar appeared to be feasible—Fig. 4.5.
Upon release for tooling and tryout, the draw development used in advanced fea-
sibility was re-worked to support the following tooling operations in the process
layout plan (i.e. trimming without use of cams and flanging) which required some
alterations to the draw development. Additionally, the blank used in the tooling and
tryout solution was planned as a sheared blank with a trapezoidal profile—Fig. 4.6.
When simulation was completed no passing result could be attained without further
compromising on compressive and wrinkling behaviors in critical areas—Fig. 4.7.
The part was determined to be poorly manufacturable and sent back to product
design for concessions in the latch area. The reason for the change in blank was
the manufacturing cost of the shaped blank in production. Shaped blanks require

Fig. 4.5 Advanced feasibility—arbitrary blank

228 4 Numerical Simulation of the Sheet Metal Forming Processes

Fig. 4.6 Blank nesting to

achieve target costs

Fig. 4.7 Manufacturability using shear blank

blanking dies, blanking dies often require offline production of blanks, and offline
blank-production may drive the use of special pallets for the blanks. The product
engineer responsible for the feasibility analyses may lack the foreknowledge to
select the appropriate blank shape, or the decision to use a shear blank had not
been made yet by the manufacturing department.
The manufacturability of the sheet metal part is often compromised by goals
made for the production stamping environment: such as production cost controls,
corporate guidelines for tool processing, material utilization goals, production plant
limitations, or other factors. Earlier recognition of this fact allows for potential
to alter other contributing factors, such as minor product concessions, additional
4.1 AutoForm Solutions 229

process steps (draw-redraw), or lowered expectations for finished part quality. This
difference exists regardless of the incremental simulation technology utilized, there-
fore it is clear that the ‘accuracy’ of a given simulation solution is driven by the
simulation input precision and the predictive reliability of the solver results.
With this is mind, the difference between feasibility and manufacturability—
tooling and tryout solutions—can be seen as less of a technological barrier but due
to a limitation of process knowledge and information. The ‘quality’ of the output
from sheet metal stamping simulation therefore lies in the ability of the operator to
use the software to best emulate the intended production environment. This may
entail ensuring that production intent inputs are used, accurate depiction of the
blanks mechanical properties (i.e. hardening curve, yield surface, failure criteria),
and appropriate element formulation. If the condition of having realistic produc-
tion intent inputs is not met then solver technology is a moot argument. Given the
definition of the production intent environment, the results from the solver can be
considered reasonable proofs of the likelihood of safe manufacturing of the part, for
the production variables as modeled.
An outcome of predicted manufacturability is, as with feasibility a conditional
assertion: if the production environment delivers the tools as engineered and the
blanks shipped to match the mechanical properties used in the incremental analy-
sis, and the set-up of the tool is the same as modeled, the production tool should
deliver results matching the simulation. That being said, it is most likely that when
a discrepancy exists between the prediction from the Finite Element Analysis and
the physical tooling that immediate reactions should first be to look at the as built
condition of the die, the blank used in the die, and the set-up of the die in the press
environment before looking to blame the quality of the analysis outcome.

4.1.4.3 Output Data

The types of outputs that are expected from the tooling and tryout solution for man-
ufacturability are much the same as we see in feasibility analysis, only now with the
use of improved inputs the reliability of those results is much improved. An outcome
of predicted manufacturability validates that a desirable outcome will be reached in
the built tool if built to the engineering specifications as modeled.
Strain distributions, and therefore the nearness to failure, resultant from tool-
ing and tryout solutions are far more reliable, and can provide to the tool and die
designer and eventually the tool maker a reasonable target for tool build conditions.
These strain distributions if used in CAE modeling of vehicle structural perfor-
mance can provide enhanced realism when compared to results from CAE models
that assume conditions based upon the nominal designed stock thickness and initial
material strengths.
With improved strain distribution as well as strain path, the stress distribution of
the part is now more realistic than achievable when compared to any stress result
from advanced feasibility. Again it is important to recognize that this expected
improvement in the stress distribution is not so much a factor of the simulation
technology used, but that we expect that the inputs used to run the tooling and
230 4 Numerical Simulation of the Sheet Metal Forming Processes

tryout solution for manufacturability to be used when the dies are actually built
for production. The resultant stress distribution should result in improved spring-
back prediction reliability, and possibly the ability to predict shape fixation in the
finished part and begin the process of engineering countermeasures for the elastic
deformation of the part.
As we saw with feasibility the predicted outcome is conditional upon the inputs
that were used. It is meaningless to state that the part can be manufactured with-
out defining the boundary conditions that are required to deliver that outcome. For
many output variables that would be used to establish the outcome of manufac-
turability (process feasibility) establishing that the process yields a ‘safe’ result is
fairly straightforward—one can over engineer for avoidance of splits or wrinkles in
the finish product. However, other quality attributes such as springback and surface
quality cannot simply be defined as ‘safe’. These attributes can be highly variable
with even minor fluctuations in the process inputs (i.e. material, lube, in-die forces)
and there is no ‘over-engineering’ to avoid negative ramifications of this variability.
To this end we see that merely validating that the single set of process inputs used to
validate the manufacturability of the stamping process is insufficient to assure that
production of the part can proceed with confidence.

4.1.5 Capability (Robustness)

4.1.5.1 Applied Technology
The ultimate goal of the use of computer aided engineering tools is the virtual pro-
duction of the stamped product so as to eliminate the need for time consuming and
expensive tryout. At present there are many examples where this expectation has
already been met, many examples can be found where the outcome was not so reli-
able and the tryout of the tool progressed slowly and may have never achieved a
‘safe’ panel. When the computer aided engineering models yield results different
from the physical tools, the difference is not due to the failings of technology, but
perhaps a failure to recognize a fundamental issue: The world is variable while deter-
ministic computer simulation models are fundamentally fixed—run from a single
set of selected inputs that represent singular instance, a specific sheet of metal with
a single idealized set of mechanical parameters, blank shape, tonnage application,
friction, bead effect, etc—based on one possible process state.
The result of an incremental tooling and tryout solution for manufacturability is
constrained by the concept that the simulated part and process is formable if exact
conditions are achieved in the manufacturing environment. The outcome is highly
conditional, and often we may find that if the input conditions change even slightly
the result of the simulation may change from acceptable to unacceptable, safe to
failing, within specification or out of specification. The results of the simulation after
many hours of modeling by the analyst and perhaps multiple iterations achieves a
passing result, however once the passing result is reached it is unclear how stable
and repeatable that result may be.
4.1 AutoForm Solutions 231

During physical tryout this fact is evident, few tool and die makers would sell a
tool on the basis of a single successful panel produced during tryout; instead mul-
tiple successful panels must be made and proven to be acceptable to the customer.
However, many users of simulation will pass off on a simulation of a single success-
ful iteration or ‘hit’. That single passing simulation is a result of a perfect collection
of events, which may or may not be achievable in the actual production environment.
To retain the value brought to the engineering of the sheet metal stamped part
from computer simulation, the simulation analyst must consider and include in
the analysis the very real potential for variation in the assumptions made during
the analysis. Much debate could be made over the claim that many make when
defending the validity of their own feasibility or manufacturability analyses, ‘the
simulation was run using the worst case inputs’—the necessary question to this
comment is what are the ‘worst case’ inputs and does that by default illustrate the
worst case outcomes. The answer to this query is that what appears to be ‘worst case’
cannot be assumed in a consistent manner nor can the effect of that assumption be
recognized in advance—see Fig. 4.8. If the effect of these changes was predictable
in advance computer, simulation would be entirely trivial.
The fact is that in a variable environment any difference from the assumed
inputs will yield some difference, and whether or not that difference is significant
is the root of the problem. A simple design of experiments using multiple manu-
facturability simulations with user adjusted inputs to represent the variation could
potentially yield valuable feedback, but this feedback will be skewed again by the
users’ assumptions and predetermined suspicions over which variables are ‘worst
case’. One could attempt to derive some design of experiments to seed multiple iter-
ations of the simulation model to attain results that could represent the outcome of
the variable stamping environment. However, the data collected would be only as
reliable as the design of experiments, which for the reasons listed above may be
difficult to design as one must already have a clear idea which variable should be
changed and the potential for influence of these changes in order to select the com-
bination of variables to test. Also interpretation of such vast amounts of data would
be overwhelming.
The perception that we can produce an accurate simulation result from finite ele-
ment analysis software now must always address the range of inputs that were used,
and are those ranges of inputs a likely condition that we might see in production. For

Fig. 4.8 Thinning results from small set of analyses with different but acceptable inputs
232 4 Numerical Simulation of the Sheet Metal Forming Processes

if the ranges used are not a precise representation of the probable production ranges
then the outcome of the simulation(s) will not predict the likely or probable results.
The outcome of process capability from a robustness solution is instead an entirely
new paradigm for simulation where the user can input ranges of inputs that are
probable, and the simulation solver automatically and statistically randomizes the
variables, combines the variables into reasonable input decks, manages the running
of the multiple simulations, and most importantly is able to represent the outcome
in a meaningful way. Stochastic analyses—based upon larger sets of incremental
results—can provide the technological answer to the issue at hand when trying to
ascertain the capability of the designed part in the approved process.

4.1.5.2 Input Data

Similar to feasibility and manufacturability outcomes the capability of a part and
process is constrained by the available inputs. In a capability analysis the input vari-
ables are now represented as ranges of potential values. The recognition of these
possible ranges is typically not available at the time of early feasibility or manufac-
turability, nor would it be reasonable to attempt a full capability analysis at the time
of product conception and design. These variables are a direct result of production
decisions and assumptions that may not be known or available until closer to the
time of production. These variables can be the result of the steel coil, the selected
blanking process, the design of the forming tool, press selection, and compromises
made in the 3D process layout throughout the planning process.
The material inputs used to determine manufacturability we most likely a ‘typi-
cal’ or minimum acceptable value. But it is known that the material shipped to the
stamping plant, like any other manufactured good is subject to variation potential
from the manufacturing source. The goal of the robustness solution is to determine if
the defined process is still capable of producing parts within this variation potential.
Figure 4.9 illustrates how over a 23 lot production run the material at one stamp-
ing plant varied greatly in terms of the delivered materials mechanical properties.
If the robustness solution is conducted using reasonable ranges of material prop-
erties that emulate the expected normal distribution of the expected values, then
the resulting simulation outcome can illustrate the potential for variation within the
manufacturing environment.
Other assumptions that can be considered questionable from the previous solu-
tions is the assumption of the blanks initial location, the distribution and effective
ness of the lubrication, the size and shape of draw beads, the amount of pressure
applied by the binder or pad systems, and many other critical design elements that
up to this point are considered design decisions and constants. But in the production
environment, the attainment and maintenance of these ‘constants’ is not certain.

4.1.5.3 Output Data

The resulting outcome from running so many simulations with the varied inputs
and assumptions will be a very rich amount of data that, if reviewed properly,
can indicate which variables have influence over which output results. From one
4.1 AutoForm Solutions 233

Fig. 4.9 Material property Frequency of R-value in production run

variation potential gathered 9
from 23 real production lots 8
7
6
5
4
3
2
1
0
1.2 1.3 1.4 1.5 1.6 1.7 1.8 1.9 2 2.1 2.2 2.3
Measured R-bar

Frequency of R-value in production run

9
8
7
6
5
4
3
2
1
0
1.2 1.3 1.4 1.5 1.6 1.7 1.8 1.9 2 2.1 2.2 2.3
Measured n-value

simulation to another, if a single attribute of the material changes, the influence

of variation in the inputs and their effect on the outcome can be derived; however
when more than one variable changes from run to run, it is not obvious which of
the inputs influence the outcome—Fig. 4.10. To manually inspect the simulation
for prediction of process capability, it may be nearly impossible to arrive at any
reasonable conclusion. However, using advanced techniques to interrogate the
response surface generated from the battery of simulations it will be possible to
derive relationships between the observed changes in the output variables and the
associated changes in the design inputs.
Ultimately, the goal is to arrive at a capable process; not always a likely event.
For this reason, another useful outcome from a seemingly failing robustness analysis
is the ability to recognize relationships between the output variables and the input
factors that may have led up to the undesirable result (Table 4.3). Provided there
are enough simulations in the response set to be statistically sound, it is possible to
derive which input variables are influential for generating any response. If an input
is influential then the sensitivity to the input variation we can be gauged—Fig. 4.11.
This information can be used to conduct a ‘robust engineering’ of the process—an
engineering process by which the robustness solution is used to identify not only
234 4 Numerical Simulation of the Sheet Metal Forming Processes

Fig. 4.10 Springback results from two slightly different materials in same process

Table 4.3 Input variation possible for robustness analysis to capture production noise

Input Production variation Range Influence/effect

Yield strength Varies with the mill and Suppliers normal High influence on
processor: chemistry, distribution or ± hardening behavior,
mechanical processing, 20 MPa yield surface,
and annealing effectiveness of
beads and pads,
springback
Tensile strength Variable with chemistry, Suppliers normal Influences hardening
and mechanical distribution or ± (with YS)
processing 20 MPa
N value Variable with chemistry, Fluctuates with Hardening behavior,
mechanical processing, variation of yield and forming limit,
and annealing tensile strength
R value Hot roll vs cold roll, ± 20% (steel) Yield surface,
mechanical processing, ± 10% (aluminium) strain/stress
coil rolling direction distribution
Lube (friction Varies over sheet surface, ± 10% Strain distribution,
coef.) varies over blank batch, material flow, stress
performance is variable distribution
in production
(beginning to end)
Blank location May shift as result of ± 1 mm or tolerance Binder force
automation and distribution, bead
gauging, varies with coil effect, lubrication
width (shear blank) effect, tooling contact
Binder/pad Varies through the stroke, ± 10% Force distribution,
pressure varies with die setting, bead effect (full set),
varies during production tooling contact
run
Blank thickness Varies with coil/blank ± 10% Tool contact, force
distribution, contact
pressure

that a process is capable or not, but to point the designer towards the design inputs
that may result in a more capable outcome.
Through the application of a robustness solution, the designer can engineer a
design that ensures that not only has the part and process been designed to achieve a
4.1 AutoForm Solutions 235

Fig. 4.11 Determination of which inputs influenced thinning result: (a) roof rail; (b) influences

‘safe’ and feasible result, but furthermore that the process will continue to produce
acceptable results throughout production, if the inputs are held within the window of
variation provided by the robustness solution. With such data at the readily available,
the designer will be able to predict if any changes are forced to the process layout,
or in some cases merely look up from the body of simulation the expected results

Fig. 4.12 Sampling the large result set allows for recognition of process capability
236 4 Numerical Simulation of the Sheet Metal Forming Processes

from some specific set of input variables and their on the overall forming safety, or
capability, of an alternative process—Fig. 4.12.
Changes to the blank, stamping process layout or product specifications will have
quantifiable ramifications for the ‘quality’ of the part. A ‘production map’ of sorts
can be developed to be referenced as a guide during tooling tryout or part produc-
tion, so that in the event of some system change (new splitting failures, parts out of
tolerance) the robustness results can be polled to find which inputs are most likely
to have slipped outside of acceptability to identify a list of ‘usual suspects’ in the
loss of ‘quality’.

4.1.6 Simulation Result ‘Quality’

As forming simulation technology has matured, most of the attention to the qual-
ity of the process layout has been measured in how closely the simulation results
matched the results in press, often forgetting that the manufacturing process inputs
cannot be assumed to match exactly the simulated inputs. Also, the focus has been
on the attainment of simulation results that accurately depict some already known
result of a single fixed and fully measured input data set.
A great deal of energy has been dedicated to discussion and determination of
appropriate element formulations, contact algorithms, solver methodologies, and
material models which surely has advanced the science of sheet metal stamping.
August institutions devote time and resources to the pursuit of improved simula-
tion of the effect of sheet metal stamping. Certainly, the yield surface model can
deeply influence the reliability of the forming simulation. However the observable
differences from material modeling subtleties and finite element nuance, pales in
comparison to the differences yielded when the engineering environment does not
match the production environment—varied blank shape, bead location, or applied
pressure. Little consideration is given to whether or not the result came from a pro-
cess layout that could be approved for cost, and even more difficult to measure has
been whether the result could be reliably achieved in the ever changing production
environment. Only by implementing a robustness solution in pursuit of a capable
process can a truly comprehensive assessment of the part function, quality, cost, and
delivery capability be made.

4.1.7 Comprehensive Digital Process Planning

Many significant decisions need to be made throughout the development of the digi-
tal process layout that will directly affect the ‘quality’ of the part—the safety margin
from splitting or wrinkling, the shape repeatability, assembly fit, and surface quality.
These changes are made in the pursuit of meeting cost targets for material utiliza-
tion, tool cost, or process simplification. The effect of these changes is rarely, if
ever, quantified. Computer Aided Engineering, if it is to truly achieve the goal of
4.1 AutoForm Solutions 237

eliminating the need for physical modeling, will have to overcome this limitation by
creating a platform in which this data is tabulated and compared. Stamping simula-
tion cannot live isolated from the measurement calculation of cost or repeatability.
The quality of a panel cannot be blindly pursued over the real world implications
of tooling manufacturing costs, and the need to ensure rapid and on time delivery
of tools and parts of those tools. Similarly every cost initiative must be assessed for
potential impact on the resulting part quality.
Using the computer simulation to prove that part geometry is possible (feasibil-
ity) for the selected material, while providing value during product development,
is not enough. Production concerns that constrain the blank process layout, such
as blank shape, number and type for stamping operations are made for concerns
of cost, quality, and lead time goals. Figures 4.13 and 4.14 illustrate the poten-
tial relationship between the desire to achieve ‘better’ resulting process layout and
risk to other attributes. These process layout changes reduce the number of feasi-
ble designs to those that are recognized as manufacturable. Without great cost the
production environment cannot assure that all process variables will be constant
or stable over time—die maintenance, material variation, production environment
all influence the process layout reducing the number of viable designs to those
that may be capable. Recognizing whether a process layout can deliver the prod-
uct as designed on time, at cost, and to the required level of quality is the ultimate
goal in stamping simulation. Accuracy in simulation is the ability to predict pro-
cess capability in a reliable and timely manner, such that production issues can
be averted.

Fig. 4.13 Improvements to

‘fit’ may drive material costs
from optimum
238 4 Numerical Simulation of the Sheet Metal Forming Processes

Fig. 4.14 Reducing tooling

cost may affect quality or
ramp-up time adversely

4.2 Simulation of the Elementary Forming Processes

4.2.1 Simulation of the Bulge Forming Process

In an European research project concerning inverse modeling [1], an experimental
program was conducted consisting of tensile tests and cruciform tests for mate-
rial characterization, complemented by bulge tests for validation. In the following,
results for DC04-IF (0.81 mm sheet thickness) are presented.

4.2.1.1 Hardening Description

Hardening parameters for a combined Swift-Hockett/Sherby law, Eq. (4.1), were
fitted to the results of tensile tests in rolling direction and are given in Table 4.4. All
stress units are in MPa.
  m     −aεp 
σ = (1 − α) C εpl + ε0 + α σSat − σSat − σi e pl . (4.1)

Table 4.4 Hardening parameters of DC04-IF

Material α ε0 m C (MPa) σ i (MPa) σ Sat (MPa) a p

DC04-IF 0.50 0.0044 0.27 580 140 432 5.47 0.76

4.2 Simulation of the Elementary Forming Processes 239

4.2.1.2 Yield Surface Description

From the tensile tests, the three yield stresses σ 0 , σ 45 , σ 90 and the three r-values r0 ,
r45 , r90 are known (see Table 4.5). From the cruciform tests, the biaxial yield stress
σ b can be derived. In these tests, no strain measurement was performed. Therefore,
the rb -value is not known.
The strategy used in AutoForm to identify the coefficients in the BBC 2005 cri-
terion is described in Table 4.6. Grey fields denote given experimental data. White
fields denote values that are defined by the model. For example, for the Hill 1948
model in Table 4.6, r0 , r45 , r90 and σ 0 are input and rb , σ 45 , σ 90 and σ b are output
parameters. The exponent M is chosen to be M = 6.
In Fig. 4.15, the yield surfaces of the above yield functions are displayed in the
principal stress plane. The BBC 2005-8 formulation is missing because a measured
rb value is not available. The data points correspond to an equivalent plastic strain
(evaluated with Hill 1948 yield criterion) of 0.05.
Within the Hill 1948 model, the biaxial yield stress is fully determined by the
r-values. This fact results in a biaxial yield stress that is too large compared to the
measured value, see Fig. 4.15. On the other hand, the BBC 2005-7 model can take
the measured biaxial yield stress into account exactly, and also the rest of the cruci-
form test data is described much better by the BBC 2005-7 model than by the BBC
2005-6, the BBC 2005-4 and the Hill 1948 models, respectively.

4.2.1.3 Bulge Tests

In the frame of the European project ‘Material parameters for sheet metal forming
simulations by means of optimization algorithms’ [1], the bulge equipment had an
internal diameter of 100 mm and a tool radius of 2 mm. The sheets were gridded
to allow optical strain measurements. For the DC04-IF sheets, the bulge height of
20 mm was reached at an internal pressure of 7.75 MPa.

Table 4.5 Yield stresses and r-values of DC04-IF

Material σ 0 (MPa) σ 45 (MPa) σ 90 (MPa) σ b (MPa) r0 r45 r90 rb

DC04 137 142 142 154 1.9 1.8 2.5 –

Table 4.6 Strategy to define the coefficients of the BBC 2005 model for steel for cases where not
all input is available: missing experimental input (white fields) for BBC model is filled up with
Hill 1948 data

Availability of test results Model M σ0 σ45 σ90 σb r0 r45 r90 rb

All data available BBC 2005-8 6
Biaxial test available but rb not given BBC 2005-7 6
Biaxial test not available BBC 2005-6 6
Only r0, r45, r90 given, σ0 arbitrary BBC 2005-4 6
Only r0, r45, r 90 given, σ0 arbitrary Hill 1948 2
240 4 Numerical Simulation of the Sheet Metal Forming Processes

400 400
Hill48 4,6 Hill48
4, 6
300 350
7

S22 [MPa]
S22 [MPa]

200 Tension, 0.05 300

Crucif., 0.05
Hill48 7
100 BBC2005-4 250
BBC2005-6
BBC2005-7
0 200
0 100 200 300 400 200 250 300 350 400
S11 [MPa] S11 [MPa]

Fig. 4.15 Yield surfaces of the DC04-IF material in principal stress space. Yield functions: Hill
1948, BBC 2005-4, BBC 2005-6 and BBC 2005-7

4.2.1.4 Bulge Simulations

The bulge simulations are run with the ‘Hydromech’ functionality of the module
AutoForm-Incremental of AutoForm 4.1 [2], both with the Hill 1948 and the BBC
2005 models. Three node shell elements with 5 integration points through the thick-
ness are used. The initial element size is set to 4 mm in all simulations, and the
adaptive refinement settings are chosen according to the AutoForm recommenda-
tions for final validation simulations [2]. In the simulation, the final internal pressure
of 7.75 MPa is prescribed, and the strain distribution at the final (unknown) bulge
height is computed.
In Fig. 4.16, measured minor and major strain values are plotted as data points
over the distance s from the center of the specimen. Simulations are run with the Hill
1948, the BBC 2005-4, BBC 2005-6 and the BBC 2005-8 yield surface models, and
the results are displayed with lines.
At the distance s ≈ 60 mm the sheet was clamped in the experiments with help
of a step bead. In the simulation, this clamping is not modeled in detail, so that for
s > 60 mm the computed and measured strain values should not be compared.

4.2.1.5 Discussion
In a bulge test, the material near to the center of the specimen is in a biaxial stress
state. On the other hand, the material at the edge where the sheet is clamped is in a
plane strain state. Therefore, in a bulge test all material points are between a state of
plane strain and biaxial stress, and it is to be expected that the simulation results are
very sensitive to the choice of the biaxial stress point in the yield surface model. In
fact, the simulation with the BBC 2005-7 model differs from the simulations with
the other yield surface models.
Only the BBC 2005-7 model is able to describe the measured strain values accu-
rately, whereas the other models give strain values that are too low, see Fig. 4.16.
4.2 Simulation of the Elementary Forming Processes 241

0.15 0.15
Measured
Minor strain Major strain
Hill48 7 Meas.
0.12 0.12
BBC2005-4
7
Minor strain [–]

Major strain [–]

BBC2005-6
Meas. 0.09
0.09 Hill48
BBC2005-7

0.06 0.06
4 4, 6
Hill48
6 0.03
0.03

0.00
0.00
0 20 40 60 80
0 20 40 60 80
s [mm] s [mm]

a) b)
Fig. 4.16 Measured and computed principal strain for DC04-IF: (a) minor strain; (b) major strain.
Yield functions: Hill 1948, BBC 2005-4, BBC 2005-6 and BBC 2005-7

Table 4.7 Computed final bulge height for an internal pressure of 7.75 MPa from Hill 1948, BBC
2005-4, BBC 2005-6 and BBC 2005-7 model

Hill 1948 BBC 2005-4 BBC 2005-6 BBC 2005-7 Meas.

Final bulge height (mm) 17.1 17.2 17.3 20.2 20

Also the final bulge height is matched well by the BBC 2005-7 model and not by
the others, see Table 4.7.
It is concluded that for hydroforming processes where an important fraction of
the part is loaded near the equibiaxial stress state the usage of the BBC 2005-7
model can significantly improve the prediction of material flow, localization and
final failure.

4.2.2 Simulation of Stretch Forming of Spherical Cup

In order to verify the accuracy of the determined parameters for a material model, we
need information from some experiment the produces strain paths that are deviating
from those paths used to determine the parameters, i.e. uni-axial and equi-biaxial
strain paths. It is also an open question which values the exponent M should take for
each material. Therefore, the idea with the current experiment is to achieve a strain
path somewhere between plane strain and equi-biaxial stretching (Vegter H, 2006–
2008, private communication). In order to achieve this, a circular blank is stretch
formed with spherical punch. The diameter of the punch is 100 mm and during the
experiment the edge of the blank is locked with a lock bead and a very high blank
holder force. The blank is cleaned before the test and then no other lubrication is
added to the blank.
242 4 Numerical Simulation of the Sheet Metal Forming Processes

Table 4.8 Experimentally determined anisotropy parameters of H180BD

σ 0 (MPa) σ 45 (MPa) σ 90 (MPa) σ b (MPa) r0 r45 r90 rb

188 205 193 229 2.01 1.02 2.72 0.97

The material used in this example is H180BD bake hardening material. The
experimentally determined anisotropy parameters are presented in Table 4.8. These
parameters have been determined at Volvo Cars with experimental data from tensile
test and viscous bulge test [3].
The opinion among almost every researcher and sheet metal forming simulation
engineer in the world is that the exponent M is equal to six for all steel grades.
This is based on results from Hosford [4]. But since he based his conclusions
from studies on an isotropic material, one could question if this is also valid for an
anisotropic material like the one used in this example. Therefore simulations have
been performed with BCC 2005 material model using all parameters in Table 4.8
and two values of the exponent, namely M equals to five and M equals to six. For
comparison the simulations have also been performed with the Hill 1948 material
model.
Figure 4.17 presents several interesting results. First of all, the Hill 1948 material
model overestimates both the punch force and the formability of the material since
the simulation doesn’t start to localise. The overestimate of the punch force is due
to the fact that Hill 1948 overestimates the equi-biaxial yield stress for all materials
with high Lankford coefficients. In order to explain why the localisation is delayed
we also need to look at the BBC 2005 results. With the exponent M equal to five
we observe that the simulation punch force curve slope changes almost at the same
depth as the experimental curve slope changes. On the other hand, with the exponent
M equal to six the change in slope is earlier in the simulations than in the experi-
ments, i.e. an increase of the exponent reduces the punch depth where localisation
starts in the simulation, see Mattiasson et al. [5]. The value of exponent M is also
the explanation for why the Hill 1948 model localises too late, since the Hill 1948
model is exactly the same as using the BBC 2005 model with only the longitudinal
yield stress together with the three uni-axial r-values as input and an exponent M
equal to two. The final conclusion is therefore that it is important to have an accu-
rate value of the exponent in the material model if it is too small the localisation in
the simulations will start too late and if it is too large the simulation will localise
too early.
In order to further analyse the difference between the material models, the major
and minor strains at 27 mm punch depth are compared in a cross section through
the centre of the punch. These results are displayed in Figs. 4.18 and 4.19. For both
major and minor strains, the Hill 1948 predicts larger strains than was measured in
the experiments. On the other hand, with an exponent M equals to six, the simula-
tions predict smaller major and minor strains than in the experiments, while with
an exponent M equals to five the agreement between simulations and experiments is
very good.
4.2 Simulation of the Elementary Forming Processes 243

Fig. 4.17 Punch force versus punch displacement in both experiments and simulations with
different material models

0.18
0.17
0.16
0.15
0.14
0.13

True Major Strain [log]

0.12
0.11
0.10
0.09
0.08
0.07
0.06
0.05
Experimental results 0.04
Hill `48 0.03
BBC2005, M=5 0.02
BBC2005, M=6
0.01
0.00
–40 –36 –32 –28 –24 –20 –16 –12 –8 –4 0 4 8 12 16 20 24 28 32 36 40
x coordinate [mm]

Fig. 4.18 Major strains at 27 mm punch depth

This example illustrates a few important facts that must be taken into considera-
tion in the sheet metal forming simulations. First of all, the exponent M is not equal
to six for all steel grades. In fact, studies at Volvo Cars have shown that the exponent
M is different for different type of materials and in some cases also different for the
same material from different suppliers. It has the lowest values for mild steel grades
244 4 Numerical Simulation of the Sheet Metal Forming Processes

0.10

0.09

0.08

0.07

True Minor Strain [log]

0.06

0.05

0.04
Experimental results
Hill `48 0.03
BBC2005, M=5
BBC2005, M=6 0.02

0.01

0.00
–40 –36 –32 –28 –24 –20 –16 –12 –8 –4 0 4 8 12 16 20 24 28 32 36 40
x-coordinate [mm]

Fig. 4.19 Minor strains at 27 mm punch depth

with high formability and as the formability of the material diminishes, the value
of exponent M increases. For some AHSS materials it is close to six and for some
other AHSS material it is even larger than six. Furthermore, in order get the highest
accuracy of the simulation one must use all parameters in Table 4.8 and perform
simulations of some experiment to determine the appropriate value of the exponent
M. If the simulation engineer uses the standard values, i.e. M equals to six for steel
grades and M equals to eight for aluminium alloys, he or she will predict a different
strain state than the real one and, what is even more important, predict localisation
earlier than in reality, i.e. a too conservative prediction. But it is also important to
emphasize that the BBC 2005 model with an exponent equal to six in this case has
better agreement with experiments than the Hill 1948 material model.

4.2.3 Simulation of Cross Die

The measurements used in this section were performed in the European project
‘Forming of metallic materials’ (FOMM). In this project, the cross die geom-
etry was used to assess the simulation accuracy of commercial finite element
packages [6].
Raw data for a 0.79 mm thick DC04 material and a 1.01 mm thick Ac121-T4
material is available from these material characterization tests:

• uniaxial tensile tests under 0, 45 and 90◦ from RD

• bulge tests
• plane strain tests
4.2 Simulation of the Elementary Forming Processes 245

For the simulation of the cross die, CAD data of the real tool and blank geometry
is available, together with information about the process conditions (lubrication,
blank holder force, tool movement etc.)

4.2.3.1 Hardening Description

The parameters of a combined Swift-Hockett/Sherby approach (Eq. 4.1) are approx-
imated from hardening curves measured in the 0◦ tensile tests. The approximated
hardening parameters are listed in Table 4.9:

Table 4.9 Hardening parameters of DC04 and Ac121-T4

Material α ε0 m C (MPa) σ i (MPa) σ Sat (MPa) a p

DC04 0.50 0.0061 0.26 561 153 415 6.13 0.8

Ac121-T4 0.50 0.0070 0.29 492 130 330 9.08 0.96

The yield stress σ 0 computed from these parameters is identical with the value
used in the yield surface description.

4.2.3.2 Yield Surface Description

From the tensile tests, the three yield stresses σ 0 , σ 45 , σ 90 and the three r-values
r0 , r45 , r90 are known (see Table 4.10). From the load-deflection curves measured
in the bulge tests, the biaxial yield stress σ b can be derived. In the bulge tests, no
strain measurement was performed. Therefore, the rb -values of the materials are not
known. The plane strain test data is not used.
The different strategies used in AutoForm to identify the coefficients in the BBC
2005 criterion are described in Tables 4.11 and 4.12. The exponent M is M=6 for
steel and M=8 for aluminum.
In these tables, grey fields denote given experimental data. White fields denote
values that are defined by the model. For example, for the Hill 1948 model in
Table 4.3, r0 , r45 , r90 and σ 0 are input and rb , σ 45 , σ 90 and σ b are output parameters.
In the next figures, the yield surfaces of the above yield functions are displayed
in the principal stress plane. The BBC 2005-8 formulation is missing because a
measured rb value is not available.
For the DC04 material, the yield surface of the Hill 1948 model is plotted in
addition to the various forms of the BBC 2005 model (Fig. 4.20). Uniaxial yield
stresses σ 0 and σ 90 from the tensile tests and biaxial yield stress σ b from the bulge

Table 4.10 Yield stresses and r-values of DC04 and Ac121-T4

Material σ 0 (MPa) σ 45 (MPa) σ 90 (MPa) σ b (MPa) r0 r45 r90 rb

DC04 151 166 163 192 1.83 1.39 2.11 –

Ac121-T4 126 122 121 137 0.65 0.40 0.77 –
246 4 Numerical Simulation of the Sheet Metal Forming Processes

Table 4.11 Strategy to define the coefficients of the BBC 2005 model for steel for cases where
not all input is available: missing experimental input (white fields) for BBC model is filled up with
Hill 1948 data

Availability of test results Model M σ0 σ45 σ90 σb r0 r45 r90 rb

All data available BBC 2005-8 6
Biaxial test available but rb not given BBC 2005-7 6
Biaxial test not available BBC 2005-6 6
Only r0, r45, r90 given, σ0 arbitrary BBC 2005-4 6
Only r0, r45, r90 given, σ0 arbitrary Hill 1948 2

Table 4.12 Strategy to define the coefficients of the BBC 2005 model for aluminium for cases
where not all input is available: missing experimental input (white fields) for BBC model is filled
up with Hill 1948 data

Availability of test results Model M σ0 σ45 σ90 σb r0 r45 r90 rb

All data available BBC 2005-8 8
Biaxial test available but rb not given BBC 2005-7 8
Biaxial test not available BBC 2005-6 8
Only r0, r45, r90 given, σ0 arbitrary BBC 2005-4 8
Only r0, r45, r90 given, σ0 arbitrary Hill 1948 2

250 220
6, 7
200
BBC2005-7
200
Hill48, 4
S22 [MPa]
S22 [MPa]

150
tension test

100 bulge test BBC2005-4

Hill48 180
BBC2005-4
50
BBC2005-6 BBC2005-6
BBC2005-7 Hill48
0 160
0 50 100 150 200 250 160 180 200 220
S11 [MPa] S11 [MPa]

Fig. 4.20 Yield surfaces of the DC04 material in principal stress space. Yield functions: Hill 1948,
BBC 2005-4, BBC 2005-6 and BBC 2005-7

test are plotted as data points. All yield functions use the given r-values as input.
This is why the slopes of the yield surfaces in the uniaxial points are the same for
all models.
The BBC 2005-7 model describes all measured yield stresses exactly because all
of them are used as input parameters for the model. The BBC 2005-6 model matches
only the measured σ 90 value but not the σ b value. Finally, the BBC 2005-4 curve
and the Hill 1948 curve do not pass σ 90 and σ b but only σ 0 .
4.2 Simulation of the Elementary Forming Processes 247

200 160
4
150
150 BBC2005-7
BBC2005-4

S22 [MPa]
6, 7
S22 [MPa]

140
100
tension test
130
bulge test
BBC2005-4 BBC2005-6
50
BBC2005-6 120
BBC2005-7
0 110
0 50 100 150 200 110 120 130 140 150 160
S11 [MPa] S11 [MPa]

Fig. 4.21 Yield surfaces of the Ac121-T4 material in principal stress space. Yield functions: BBC
2005-4 (identical with Barlat 1989), BBC 2005-6 and BBC 2005-7

The Hill 1948 model is a special case of the BBC 2005-4 model with M = 2
instead of M = 6. Therefore, these models have the same uniaxial and biaxial yield
stresses. The biggest differences between the Hill 1948 curve and the BBC 2005-4
curve is in the plane strain region.
For the yield surfaces of the Ac121-T4 material shown in Fig. 4.21, the previous
discussion of the data in Fig. 4.20 holds as well. The only difference is that only the
BBC 2005 models are displayed. This is because the Barlat 1989 model is identical
with the BBC 2005-4 model, see Table 4.12.

4.2.3.3 Cross Die Experiments

The geometry of the cross die example is shown in Fig. 4.22 and Table 4.13.
In the experiments the sheet dimensions and the blank holder force were chosen
such that a punch stroke of 60 mm could be reached without fracture, see Table 4.14.

Fig. 4.22 Left: Punch geometry from above. Right: Sheet at punch stroke 60 mm, with sections
for draw in measurements and thickness measurements
248 4 Numerical Simulation of the Sheet Metal Forming Processes

Table 4.13 Geometrical parameters of punch and die. The clearance between punch and die is
2.3 mm. Thus, on the die (not shown in Fig. 4.22) each radius has the value 22.3 mm

apunch (mm) bpunch (mm) rpunch (mm) adie (mm) bdie (mm) rdie (mm)

59.4 179.4 20 64 184 22.3

Table 4.14 Experimental parameters for DC04 andAc121-T4

Material Blank holder force (kN) Sheet dimensions (mm) Lubrication

DC04 350 330 × 330 Grease

Ac121-T4 116 260 × 260 Grease/teflon

After stopping the forming after 60 mm punch stroke, the draw in and the thickness
were measured along the two sections displayed in Fig. 4.22.

4.2.3.4 Cross Die Simulations

The simulations of the cross die forming experiments were run with the module
AutoForm-Incremental of AutoForm 4.1, employing three node shell elements with
5 integration points through the thickness. The initial element size was set to 11 mm
in all simulations. The adaptive refinement settings were chosen following the AF
recommendations for final validation simulations [2]. Typically, this resulted in 35
solution increments with 2,200 elements at the beginning and 12,000 elements at
the end of the simulation.
First of all, in simulations using the BBC 2005-7 yield surface model the coeffi-
cient of Coulomb’s friction law and the elastic stiffness of the tools were adjusted
to measured draw in values along the diagonal and the meridian cut. A value of
0.05 for Coulomb’s friction coefficient was found to give a satisfactory agreement
between measured and computed draw in both for DC04 and Ac121-T4. Note that
the experiments were carried out with special lubrication, see Table 4.14. The elas-
tic stiffness of the blankholder was increased by a factor of 20 compared to the
AutoForm default value. This was necessary because in the experiments the binder
was not ground (or spotted) for sheet thickening. The following simulations with
different yield surface models were run with these settings fixed.
In Figs. 4.23 and 4.24, thickness measurements are compared with computed
thickness distributions for the yield surfaces displayed in Figs. 4.20 and 4.21. The
distance s runs from the centre of the cross specimen outwards along the diagonal
cut and the meridian cut as displayed in Fig. 4.22.
For the DC04 material, the simulation with the BBC 2005-7 model matches the
thickness measurements very well. Especially, the minimum thickness is predicted
accurately. All the other yield surface descriptions yield higher deformations and,
thus, overestimate the risk of failure for that part.
For the Ac121-T4 material, the BBC 2005-7 simulation matches the thickness
measurements rather well, especially in the meridian cut. In the diagonal cut, the
4.2 Simulation of the Elementary Forming Processes 249

0.90 0.90
diagonal meridian
7 Hill48
0.80 0.80
Thickness [mm]

Thickness [mm]
Meas
6
0.70 0.70

Meas 7
Hill48 4
0.60 0.60
6
4
0.50 0.50
0 50 100 150 200 0 50 100 150 200
s [mm] s [mm]

Fig. 4.23 Measured and computed thickness for DC04. Yield functions: Hill 1948, BBC 2005-4,
BBC 2005-6 and BBC 2005-7

1.20 1.20
diagonal meridian
1.10 1.10
Thickness [mm]

Thickness [mm]

Meas
1.00 1.00
7
7

0.90 0.90
Meas
0.80 0.80
6 6
4 4
0.70 0.70
0 50 100 150 200 0 50 100 150
s [mm] s [mm]

Fig. 4.24 Measured and computed thickness for Ac121-T4. Yield functions: BBC 2005-4
(identical with Barlat 1989), BBC 2005-6 and BBC 2005-7

agreement between measured and computed thickness is not as good as it is for the
DC04 material in Fig. 4.23. Anyway, the data is described much better by the BBC
2005-7 model than by BBC 2005-6 and BBC 2005-4 (which is identical with the
Barlat 1989 model). The widely used Barlat 1989 model largely overestimates the
risk of failure for that part.

4.2.3.5 Discussion
Although the differences between the various yield surface models do not seem to
be very large at first sight (see Figs. 4.20 and 4.21), the cross die test is by far best
described with help of the BBC 2005-7 model. The results demonstrate that for an
accurate failure prediction it is crucial to take not only the uniaxial yield stresses and
250 4 Numerical Simulation of the Sheet Metal Forming Processes

Fig. 4.25 Measured and computed thickness for DC04, diagonal cut, s = 80 mm. Left: Variation
of σ b . Right: Variation of rb

r-values into account but also the biaxial yield stress σ b . The question remains if the
prediction could be further improved by taking also the rb value into account with
help of the BBC 2005-8 model. Since no measured rb value is available, this ques-
tion was tackled with a purely numerical sensitivity analysis using the the module
AutoForm-Sigma of AutoForm 4.1.
Two series of simulations were performed for the DC04 material. In the first
series, only the σ b value was varied while keeping the other material parameters
fixed. In the second series, the same procedure was applied to the rb value. The
computed thickness is evaluated in the diagonal cut at the position of minimum
thickness (s ≈ 80 mm). The results are compared in Fig. 4.25.
The dependency between the computed thickness and the biaxial yield stress σ b
is nearly linear, especially in the range 192 MPa ± 4.8 MPa that is indicated by
the vertical blue band in the left part of Fig. 4.25. The value of 4.8 MPa was input
as standard deviation for the SIGMA analysis. Since no statistical information was
available for the DC04 material, the value was arbitrarily chosen to be 2.5% of the
value 192 MPa.
The assumed standard deviation for the biaxial anisotropy parameter rb is
0.02175 (2.5% of the value 0.87). The effect on the result is hardly visible. This
means that even if rb would have been measured and used in the simulation, it had
no significant impact on the computed minimum thickness.
In this light, for the cross die simulation the usage of the BBC 2005-7 model is
proven to be sufficiently accurate.

4.3 Simulation of the Industrial Parts Forming Processes

The first example studies the different material models for a H180BD bake hard-
ening material. The experimentally determined anisotropy parameters are presented
in Table 4.15. These parameters have been determined at Volvo Cars with experi-
mental data from tensile test, viscous bulge test [3] and stretch forming test with a
spherical punch. This is the same material as in Sect. 4.2.2
4.3 Simulation of the Industrial Parts Forming Processes 251

Table 4.15 Experimentally determined anisotropy parameters of H180BD

σ 0 (MPa) σ 45 (MPa) σ 90 (MPa) σ b (MPa) r0 r45 r90 rb M

188 205 193 229 2.01 1.02 2.72 0.97 5.0

Generally, not all data in Table 4.15 are always available for all materials in the
industry. Therefore, it is interesting to compare three different set-ups:

• Hill 1948 material model in which measured σ 0 , r0 , r45 and r90 are used as input
and M equals 2. This is how the majority of industrial simulations are done today.
All other parameters, i.e. σ 45 , σ 90, σ b and rb , are in this case predicted by the
material model. For this particular material, the Hill 1948 material model over-
estimates both the σ 45 and σ 90 values while rb is lower than the measured value,
see Table 4.16.
• BBC 2005 with six parameters measured in tensile tests: σ 0 , σ 45 , σ 90, r0 , r45 and
r90 . Here σ b and rb are then values predicted by AutoForm which in this case
means the same values as for Hill 1948. In this set-up M is equal to 6 for steel
grades and 8 for aluminium alloys. This set-up is called AF in Table 4.16.
• BBC 2005 with all data in Table 4.16.

Table 4.16 Anisotropy parameters used in the simulations

σ 0 (MPa) σ 45 (MPa) σ 90 (MPa) σ b (MPa) r0 r45 r90 rb M

Hill 1948 188 246 197 248 2.01 1.02 2.72 0.74 2.0
BBC 2005-6 188 205 193 248 2.01 1.02 2.72 0.74 6.0
BBC 2005-8 188 205 193 229 2.01 1.02 2.72 0.97 5.0

4.3.1 Simulation of an Outer Trunklid

The differences between the yield loci with these three different settings are small,
it is only in the plane and equi-biaxial strain part of the yield locus that the three
set-ups are different. It is therefore easy to come to the conclusion that they should
produce similar results. In order to test this conclusion, simulations of the forming
of an outer trunklid with all three set-ups have been performed in AutoForm 4.1.1.
In these simulations all other inputs, e.g. friction coefficients, draw bead restraining
forces and hardening curves are identical. The results from the simulations are also
compared with results from trials of the same material in Volvo Cars production.
The forming experiments revealed no major forming problems but in two areas
on the addendum surfaces the comparisons between experimental and numerical
results are extremely interesting. The first area was at the rear upper corner of the
part. Here a small failure occurred in the experiments, but this fracture could eas-
ily be removed with a small modification of the addendum in this area. This area
252 4 Numerical Simulation of the Sheet Metal Forming Processes

is called Area A in Fig. 4.26. The second area is the transition area between the
horizontal part and the vertical part of the trunklid. In this area the material was
subjected to very large strains but there was no failure. This area is called Area B in
Fig. 4.26.
Generally, the simulation results with the three different set-ups are very similar
when compared. But in the two areas mentioned above, the three different set-ups
produce quite different results.
The strain signatures for Area A with the three different simulation set-ups are
displayed in Fig. 4.27. In this area, the Hill 1948 and the BBC 2005 material model
will all experimental parameters produces similar results and with both set-ups the
maximum principal strains are far above the FLC curve. The strain state is more
severe with the BBC 2005 model, but both material models clearly predict a fracture
in the simulation is this area which corresponds well with experimental results. In
the BBC 2005 model using only tensile test data, the so called AutoForm set-up,
all major strains in this area are below the FLC and therefore no fracture in the part
according to this set-up.
In the Area B, the two BBC 2005 set-ups are producing similar results, see
Fig. 4.28. It also seems that BBC 2005 using all data from Table 4.15 predicts a

Fig. 4.26 The two studied areas

Fig. 4.27 Strain signatures for the Area A with Hill 1948 (left), BBC 2005 with tensile test data
(middle) and BBC 2005 with all parameters (right)
4.3 Simulation of the Industrial Parts Forming Processes 253

Fig. 4.28 Strain signatures for the area B with Hill 1948 (left), BBC 2005 with tensile test data
(middle) and BBC 2005 with all parameters (right)

slightly more severe strain state than BBC 2005 using only tensile test data. These
results are also in good agreement with experimental observations where very large
deformations were observed without any fracture. The Hill 1948 set-up yields a
completely different strain signature. The major strains are far above the FLC curve
and the strains are so large that the software removes elements in this area. There is
therefore no doubt that there would be a fracture in this area according to the Hill
1948 material model.
The final conclusion is that although the yield loci are similar in this case, the
BBC 2005 model predictions are much closer to the test results than the Hill 1948
predictions in the studied areas. The explanation for the difference is that the two
strain states studied are close to plane strain see Figs. 4.27 and 4.28, and here the
differences between the three yield loci are large.
The BBC 2005 model with all parameters as input models the equi-biaxial and
plane strain part of the yield locus with high accuracy since it is using the measured
equi-biaxial stress as input. The choice of M-value then determines stress state at
plane strain. By comparing experimental and simulation results from e.g. stretch
forming experiments with friction, an appropriate value of M for each material could
be determined, see Sect. 4.2.2.
The Hill 1948 material model overestimates in this case the equi-biaxial stress
and since M is low, the stress state at plane strain is incorrect. This then yields the
poor agreement with experiments for Area B in this case. In Area A, the agreement
with experiments is good, which is due to the fact that this strain state is more biaxial
than the strain state in Area A.
The example also shows the major improvement of the simulation accuracy
changing from Hill 1948 material model to BBC 2005 material model. Based on
the results from Hill 1948, the part would be classified as not feasible with this
material, but with both settings of BBC 2005 material model one problem area dis-
appears. The recommendation therefore is to always use the BBC 2005 material
model. If only tensile test data are available it would still be a major improvement
of the simulation accuracy.
254 4 Numerical Simulation of the Sheet Metal Forming Processes

Finally, a comment on these results compared to the results from the stretch form-
ing experiment of the same material in Sect. 4.2.2. In the present case we see that in
Area B is the formability of the material increases going from the Hill 1948 to the
BBC 2005 material models, while in the stretch forming experiment the formability
of material was reduced going from Hill 1948 to the BBC 2005 material model. At
first these results seems to be in contradiction to each other, but what they really
shows is the complexity of material modeling. One cannot make a general state-
ment that going from Hill 1948 to BBC 2005 material models would give a certain
effect for all grades and for all strain states. But, it is also clear that the agreement
between simulation and experimental results always would be improved going from
Hill 1948 to BBC 2005 material models, which this example have shown.

4.3.2 Simulation of a Sill Reinforcement for Volvo C30

The second example is a Sill Reinforcement for Volvo C30 (Fig. 4.29) [7]. In this
case has a 1.1 mm thick DP600 is used. In this example two different material mod-
els are compared, the Hill 1990 and the BBC 2005 material models and the purpose
of the study was to evaluate the influence of material and process scatter on spring-
back predictions. The used material data in the stochastic simulations are presented
in Table 4.17. For the Hill 1990 material model only σ 0 , r0 , r45 and r90 used as input
while for BBC 2005 are all data in Table 4.17. used as input.
As well as this, the hardening curve, Young’s modulus, Draw bead restraining
forces, friction coefficient and blank centre were allowed to vary in the AutoForm-
Sigma simulations (see Sect. 4.4), but these variations were the same for both
material models. For more information about the set up and also the results,
see [7].
The springback after forming was evaluated in twenty different points that corre-
sponded to the measuring point in running production. The min, mean, max and
standard deviations from the simulations were then compared with results from

Fig. 4.29 The Sill reinforcement for Volvo C30

Table 4.17 Material data for Sill reinforcement study

σ 0 (MPa) σ 45 (MPa) σ 90 (MPa) σ b (MPa) r0 r45 r90 rb

Min 331 339 345 361 0.53 0.70 0.92 0.80

Median 383 386 402 416 0.74 0.87 1.16 0.90
Max 430 417 439 451 0.95 1.14 1.40 1.00
Std. Dev. 18.0 14.2 12.1 16.4 0.07 0.09 0.08 0.03
4.4 Robust Design of Sheet Metal Forming Processes 255

Table 4.18 Root mean square errors for the springback predictions

Hill 1990 BBC 2005

Median value 1.21 (mm) 0.91 (mm)

Std. dev. 0.12 (mm) 0.07 (mm)

Fig. 4.30 Measured and predicted values obtained in one measuring point with BBC 2005 material
model

production. Generally, the agreement between predicted and measured standard

deviations was quite good, while the agreement between predicted and measured
mean-values was poorer. Another interesting observation was that the BBC 2005
model has a better agreement with production measurements than the Hill 1990.
This was determined by computing the root mean square error for the results from
the two material models, see Table 4.18. In Fig. 4.30, all production measurements
are presented together with predicted and measured min, median and maximum
spring back values for the BBC 2005 material model for one of the measuring points.
In this case the agreement is excellent.

4.4 Robust Design of Sheet Metal Forming Processes

Even with the extreme high level of today’s simulations it still happens that in real
life production scrap has been produced. On the one hand this is caused by the
difference between reality and the simulation. On the other hand a simulation is
only one simulation prescribing one specific set of properties and settings.
256 4 Numerical Simulation of the Sheet Metal Forming Processes

We all know that in reality variability and noise exist. The applied forces of the
press are not as constant as we want them to be. The film of oil is not always as thick
as we want it to be. And also the mechanical properties are not always as exact as
we want them to be.
This uncertainty has been accommodated by introducing safety margins and
worst case scenarios. Hence, this engineering practice has reduced the incorpora-
tion of the variability and noise into a deterministic problem. This approach however
doesn’t say anything about the variability of the result. It can result in much too con-
servative solutions which might be too expensive. On the other hand it doesn’t even
guarantee a reliable process.
So, in order to improve the prediction of the reliability of the production we have
to incorporate the variability and noise and solve a stochastic problem instead of a
deterministic one [8].

4.4.1 Variability of the Material Parameters

In this section we will consider the variability of the material properties incorporated
by the parameters of the constitutive equations. Next we will have a closer look
at the mechanical properties of a high strength steel, in this case HCT600X. The
mechanical properties of the HCT600X according to DIN EN 10336 are listed in
Table 4.19.
The norm data shows a relatively large tolerance range. 132 batches of mate-
rial delivered according to the above mentioned specifications have been evaluated.
Table 4.20 gives a summary of the mechanical properties Rp0.2 and Rm , and the
thickness variation from the nominal thickness.
The average values of the mechanical properties perfectly fit with the above men-
tioned norm. But looking at the minimum and maximum values one can see that a

Table 4.19 Mechanical properties for HCT600X according to DIN EN 10336

Yield strength Tensile strength Elongation

Rp0.2 Rm min A80 min

MPa MPa %
340–410 600 20

Table 4.20 Mechanical property and thickness variation for HCT600X

Rp0.2 (MPa) Rm (MPa) Dt (mm)

Average 378 631 0.00

Min 330 518 –0.09
Max 430 697 0.07
Stand. dev. 20.4 29.7 0.03
No. of Samples 132 132 132
4.4 Robust Design of Sheet Metal Forming Processes 257

Table 4.21 Variation of the r0 , r45 and r90 values for HCT600X

r0◦ r45◦ r90◦

Average 0.73 0.88 0.94

Min 0.50 0.72 0.07
Max 1.05 1.05 1.35
Stand. dev. 0.09 0.06 0.09
No. of Samples 122 122 122

large window of both the yield stress and tensile strength exits. It even turns out that
some of the samples fall outside the norm prescriptions. The standard deviation of
the Rp0.2 and Rm are both roughly 5% of the average value, meaning that 68% of the
samples are varying from 358 to 398 for the Rp0.2 and from 601 to 661 for the Rm .
Although not in the norm, for simulation purpose later, we will have a look at the
r-values. Table 4.21 gives a summary of the three r-values, r0 , r45 and r90 . Studying
this data, one can see that also the variation of the r-values is relatively large. The
standard deviation is roughly 10% of the average value. As an example we can have
a look at the most extreme of the three r-values, the r0 . The average value is 0.73
and 68% of the samples have a value between 0.64 and 0.82.
In summary we can say that the properties of the HCT600X are not a constant
set of properties. In general one can say that for every material grade mechanical
properties vary. In order to incorporate this variation into a simulation model, we
have to step into stochastic analysis.

4.4.2 AutoForm-Sigma
Today’s simulations are mainly applied to evaluate the feasibility of a part and its
forming process. The outcome is a virtual prototype saying that it is possible to
produce the part. In fact one process point has been defined. However, when going
into production a process window must be known to guarantee a stable production
process. In order to achieve the latter condition, we are suggesting a stochastic anal-
ysis. Based on multiple simulations, the influence and sensitivity of various process
parameters on the forming process can be identified. By combining the analysis with
the statistical process control evaluation, the process capability (Cpk-values) can be
defined. The result of the stochastic analysis is the identification of the process win-
dow and process capability before any tool has been manufactured. So, we have to
solve the stochastic problem instead of the deterministic one.
Looking up the word stochastic in the dictionary one finds a description like:
being or having a random variable. The mechanical properties as described in Sect.
4.4.1 clearly have a variability. So, if we want to incorporate this variability we have
to enter the world of stochastics.
How should such an analysis look like? The stochastic analysis consists of
multiple simulations. In the various simulations some parameters are varied. The
258 4 Numerical Simulation of the Sheet Metal Forming Processes

parameters which are varied are called the design variables. In our case the design
variables are the mechanical properties like Rp0.2 , Rm and the three r-values. The
thickness variation can also be defined as a design variable.
The exact values of the design variables are chosen randomly but it can be defined
that this random distribution represents the normal distribution. The combination or
pairing of the various design variables is also chosen randomly. In this case we are
making use of Latin Hypercube techniques. The set of simulations are all solved
automatically generating several result files. The number of simulations strongly
depends on the number of design variables. The total number of simulations can
exceed 100.
It is relatively impossible and very time consuming to evaluate these simulations
individually. How can we evaluate the set of stochastic simulations? For that reason,
special result variables have been defined to ease the evaluation of multiple simu-
lations at once. These result variables are based on statistic algorithms. Several of
these result variables will be described below.
The variation of the result is described with the help of the Standard Inter
Quantile Range (Standard IQR). This measure can be applied on random distri-
butions and is equivalent to the standard deviation in case of a normal distribution.
So, this result variable indicates how much the result will vary. In worst cases this
result variation will yield in production failure.
The influence is a measure for the extent to which a parameter influences the
result. The influence can vary from 0 to 1. A value of 0 means no influence at all so
the result is absolutely independent of the parameter. A value of 1 means a strong
influence so the result is completely dependent on the parameter.
The sensitivity is a measure of the capability of the parameter to control the
result. Its value shows how much the result will change in case that the parameter
will be changed.
Statistical process control is normally applied during production in order to val-
idate the productivity and capability of the production process. Since we have the
possibility to perform multiple simulations taking into account the real life variabil-
ity it is only a small step to apply the statistical process control on the simulation
results. By doing so, we can evaluate the success rate of the proposed production
process in an early stage without manufacturing any of the tools.
For statistical process control, the process precision (also called process capa-
bility) Cpk has often been used. The process precision indicates the controllability
of the process around the given specification limit. It indicates the probability of
the result exceeding the specification limit because of the given variation of the
input [9].

4.4.3 Robust Design: Case Studies

The methodology described in Sect. 4.4.2 is applied on two cases of automotive
parts currently in production: a Front Side Member Inner and a Hood Inner. The
application examples show how the stochastic analysis can be applied today in
engineering practice.
4.4 Robust Design of Sheet Metal Forming Processes 259

4.4.3.1 Case 1: Front Side Member Inner

The first case presents the results from a project carried out by Volvo Car
Corporation and AutoForm Engineering. The objective of this project was to study
the influence of the scatter in mechanical properties on the sheet metal forming
simulation results.
The project has studied the forming of the front side member inner (FSMI) made
from a tailor welded blank (HCT600X and HCT260P) incorporated in the Volvo
S40/V50 model. This part was chosen since it was known to have splitting problems
in production that were not visible in the final forming simulation [10].
An effective way to visualize splitting is using the result variable failure. This
variable is defined as the ratio between the major strain at a specific point and the
major strain of the FLC at the same minor strain value. This implies that splits occur
at a value of 1.0 or more since at that value the point reaches the FLC. Values below
1.0 indicate that the point is still below the FLC, i.e. a value of 0.8 indicates the point
is 20% below the FLC. In engineering practice one works with values between 0.6
and 0.8 as limit introducing a 40–20% safety margin.
Figure 4.31 shows the failure values of the FSMI as obtained in the basic simula-
tion. One can clearly distinguish that all failure values are below the 0.6 mark (red)
being at the very safe side.
As mentioned before, we studied this part since it was known to have splitting
problems in production that were not visible in the final forming simulation. The
hypothesis was that these differences come from the scatter of the blank mechanical
properties. To investigate this, a stochastic analysis has been performed based on the

Fig. 4.31 Maximum failure results of the basic simulation

260 4 Numerical Simulation of the Sheet Metal Forming Processes

scatter values as indicated in Tables 4.20 and 4.21. The variation of the mechanical
properties Rp0.2 and Rm has been introduced by manipulating the basic flow curve.
The variations of the r-values are incorporated in the constitutive equation, in this
case we used the Hill 1948 model. The thickness variation is simply introduced by
varying the initial thickness.
For this stochastic analysis, 100 simulations have been performed automatically
while varying the design variables, which are actually noise variables.
It is a relatively impossible and very time consuming task to evaluate these 100
simulations individually to check whether and how much of the simulations are
critical. So, the special result variables will be used to ease the evaluation of the 100
simulations at once.
In order to evaluate the sensitive zones, the result variable Standard IQR of fail-
ure is used as shown in Fig. 4.32. The larger the value of the Standard IQR, the
more variation of the failure values due to the variation of the input exists. So, in the
yellow zones, the most important variation of the failure value is seen, meaning that
in those regions the failure value is sensitive to the variation of the mechanical prop-
erties. In these sensitive zones the failure value can easily exceed critical values due
to the variation of the input. Two sensitive zones can be distinguished as indicated
with white circles in Fig. 4.32.
The easiest way to check whether a sensitive zone exceeds a critical value is
to define a limit value and calculate the probability whether the result will exceed
this value. In this case we define an Upper Specification Limit (USL) of 0.8 for the

Fig. 4.32 Standard IQR of failure of the stochastic analysis to indicate the sensitive zones as
indicated with the white circles
4.4 Robust Design of Sheet Metal Forming Processes 261

failure value and evaluate the process precision Cpk. The process precision is used
very often in quality process control and is evaluated according to the colour scheme
as given in Table 4.22. The traffic light color scheme directly expresses the process
precision according to DIN 55319.
This process precision classification can directly be plotted on the part
(Fig. 4.33). Most of the part is green, meaning that the process is reliable. A reliable
process has an expected reject rate smaller than 0.004%.
But in zone 1 a red spot can be distinguished. This red spot indicates that the pro-
cess is unacceptable resulting in more than 2.25% rejects. This spot indeed coincides
with the spot which has been indicated as a critical area during production.
For a more detailed evaluation of the process precision, the histogram shows the
frequency of how often a failure result has been obtained due to the variation of the
input (Fig. 4.34). The vertical black line shows the defined specification limit of 0.80
maximum failure. It can be seen that in some cases the 0.80 mark is passed which
indicates risk of failure. With help of the histogram one can more precisely analyze
how reliable or unreliable the process is. In zone 1 three from the 100 simulations

Table 4.22 Colour scheme of process precision according to DIN 55319

Green Reliable No more than 0.004% of the results are outside the limits
Yellow Control required Between 0.004 and 0.14% of the results are outside the limits
Orange Unreliable Between 0.14 and 2.25% of the results are outside the limits
Red Unacceptable More than 2.25% of the results are outside the limits

Fig. 4.33 Process precision Cpk of failure plotted on the front side member inner
262 4 Numerical Simulation of the Sheet Metal Forming Processes

Fig. 4.34 Histogram for

zone 1 showing the frequency
of failure

result in a failure value higher than the USL of 0.8, which in quality process control
measures means unacceptable.
The application example showed that realistic scatter of blank mechanical
properties clearly influence the sheet metal forming simulation.

4.4.3.2 Case 2: Hood Inner

The second case shows a hood inner [11]. The basic parameters of the hood inner
production have been listed in Table 4.23. The hood inner has been produced with
uncoated material as well as zinc coated material. The material specification SPCD
(Japanese norm) is equivalent with DC03 (European norm) and SPCD GA (Japanese
norm) is equivalent with DX53D (European norm).
The optimized stamping process showed a feasible part in the simulation. The
formability evaluation shows a ‘green’ result. The tolerated thinning has been
defined on a demanding –0.20. Unfortunately, the production of the part didn’t go as
easily as the optimized simulation predicted. During a 10 weeks monitoring phase,
splitting at two areas of the part have been observed resulting in unacceptable reject
rates (Fig. 4.35).
During the monitoring phase the mechanical properties of all coils have been
determined as well. An overview of the mean values and its extremes is listed

Table 4.23 Basic parameters

of the hood inner production Thickness 0.62 mm
Material SPCD, SPCD GA
Coil width 1,439 mm
Coefficient of friction 0.15
Blank holder force 60 tons
Press type Single action
4.4 Robust Design of Sheet Metal Forming Processes 263

Fig. 4.35 Necking of the hood inner

in Table 4.24. We have to state that no difference in mechanical properties of the

uncoated and zinc coated material has been found.
The surface structure of both materials (the uncoated and zinc coated variant) are
different. Because of this, strip draw tests have been performed in order to determine
the coefficient of friction. Since the amount of conservation oil also varies, strip
draw tests with material as delivered and the oil wiped off have been performed.
The results of the strip draw tests are summarized in Table 4.25. One can clearly
see that the coefficient of friction for the zinc coated material is very sensitive to the
lubricant condition.
The stochastic analysis has been performed in order to investigate the influence
of the above mentioned parameter variations on the stamping result. In our case
the noise variables are the mechanical properties like yield strength and tensile
strength. The thickness variations as well as the variation of the coefficient of fric-
tion have been defined as noise variables too. The noise variable input is defined in

Table 4.24 Mechanical property variation

Mean Min Max

Yield strength (MPa) 154 142 165

Tensile strength (MPa) 311 303 321
Thickness (mm) 0.620 0.615 0.629
r-value 1.89 1.76 2.06
n-value 0.232 0.224 0.237

Table 4.25 Coefficient of friction for different lubricant conditions for both materials

Lubricant condition Uncoated Zinc coated

Conservation oil 0.141 0.139

Wiped off 0.144 0.189
264 4 Numerical Simulation of the Sheet Metal Forming Processes

Table 4.26. A theoretical normal distribution has been assumed, which represents
the real distribution very well [9].
The nominal simulation has been evaluated with respect to thinning. As can be
seen in Fig. 4.36, several areas have thinning values close to the critical limit of
–0.20, although none of them exceeds this critical limit. In order to evaluate the
stochastic analysis of 100 simulations, the special result variables have been applied.
The easiest way to check whether a result might exceed a critical value is to
define a limit value and calculate the probability whether the result will exceed this
value. In this case we define a Lower Specification Limit (LSL) of –0.2 thinning and
evaluate the process precision Cpk.
The process precision classification (Cpk) can be plotted directly on the part
(Fig. 4.37). Most of the part is green, meaning that the process is reliable. But also

Table 4.26 Noise variable input

Noise variable Median Standard deviation Min Max

Yield strength (MPa) 154 3 142 165

Tensile strength (MPa) 311 3 303 321
Thickness (mm) 0.620 0.0023 0.615 0.629
Coefficient of friction 0.15 0.0067 0.13 0.18

Fig. 4.36 Thinning distribution of the hood inner

4.4 Robust Design of Sheet Metal Forming Processes 265

Fig. 4.37 Identification of unacceptable process precision

red spots can be distinguished. These red spots indicate that the process is unac-
ceptable. These spots indeed coincide with the spots which have been indicated as a
critical area during the monitoring phase.
The robustness analysis clearly indicates that production problems might occur.
The analysis even managed to point exactly the areas where these problems would
occur. Without incorporating the uncertainty of the property variations we would
never have been able to judge the reliability or robustness of the production process.
For the purpose of this study we will focus on the result evaluation of two areas.
These two areas are indicated in Fig. 4.36 as well as in Fig. 4.37. The first area
(zone 1) shows a thinning value of –0.17 in the nominal simulation and is identified
as having unacceptable process precision in the robustness analysis. The second area
(zone 2) shows a thinning value of –0.18 in the nominal simulation and is identified
as having reliable process precision in the robustness analysis.
The Pareto chart shows the influences of the considered noise variables sorted
in decreasing order (Fig. 4.38). For zone 1, one can clearly see that the thinning is
mainly influenced by the coefficient of friction (Lub). The other parameters hardly
influence the thinning result. For zone 2, the coefficient of friction is again the dom-
inant variable. But the yield strength still has a significant influence on the thinning
result.
266 4 Numerical Simulation of the Sheet Metal Forming Processes

Fig. 4.38 Pareto chart for zone 1 and zone 2 showing the influence of the noise variables on the
thinning

So, to control the result in zone 1 the coefficient of friction must be controlled. In
zone 2 the coefficient of friction must be controlled as well as the yield strength. But
before starting any action to control these parameters, the sensitivity of the thinning
behavior to the variations must be evaluated.
So, for a more detailed analysis we will have a look at the relationship between
the coefficient of friction and thinning. Figure 4.39 shows the scatter plot for zone
1 and zone 2. The scatter plot shows the raw result variable value and the selected
noise variable value in an xy-scatter plot for all simulations. The x-axis represents
the value for the coefficient of friction; the y-axis represents the resulting thinning.

Fig. 4.39 Scatter plot for zone 1 and zone 2 showing the relationship between the thinning and
the variation of the coefficient of friction
4.5 The Springback Analysis 267

The critical limit of –0.20 thinning is indicated with help of the lower black hori-
zontal line. The upper black horizontal line indicates a thinning value of –0.15 and
has just been used for scaling purposes
The scatter plots of Fig. 4.37 clearly show that zone 1 is much more sensitive to
the variation in friction than zone 2. So, the robustness analysis shows that zone 1
will cause more problems in production although the nominal simulation shows less
thinning for zone 1 than for zone 2.
Since zone 2 is insensitive to the variations it is not needed to take any measures
on the control of the yield strength. Since zone 1 is very sensitive for the variation
of the coefficient of friction this parameter must be controlled carefully to achieve a
robust production process.
This result coincides with experiences in the real production. The zinc coated
material caused problems. Strip draw test gave that the coefficient of friction varies
considerably depending on the oiling condition. The robustness analysis showed
that the production reliability is strongly dependent on the value of the coefficient
of friction.
So, in case of controlling the oiling condition of the zinc coated material the
production reliability will increase.

4.4.4 Conclusion
The results of the two cases presented here unmask the common practice of evalua-
tions with fixed safety margins as either non-effective and hence very dangerous or
as extremely conservative and thus costly. The application of stochastic simulation
methods reduces the need for wide safety margins while at the same time increas-
ing the reliability of the process by incorporating uncertainty into the simulation
itself.

4.5 The Springback Analysis

4.5.1 Introduction
When stamped sheet components are removed from the forming tools, the residual
internal stresses will relax, and a new equilibrium state will be reached. As a result,
the final shape of the drawn part will deviate from the shape imposed by the forming
tool. This phenomenon is known as springback.
Springback is the major quality concern in the stamping field. The final shape
of a part is determined by the springback deviation. If the shape deviation due
to springback exceeds the given tolerance, it can create serious problems for
the subsequent assembly operations. In recent years, the trends of applying high
strength steel and aluminium to automobile components have widely emerged due
268 4 Numerical Simulation of the Sheet Metal Forming Processes

to their desirable low weight-to-strength ratio, which leads to better fuel efficiency.
Consequently, springback deviations became more severe due to the higher yield
strength-to-modulus ratio.
Modern finite element codes for sheet metal forming simulation have been
shown to be able to produce excellent results regarding the formability pre-
diction. The accurate simulation of springback, however, has still been proven
difficult. Springback simulation is the last step of numerical simulation of sheet
metal forming, consequently, any calculation errors resulting from previous sim-
ulation of forming processes will be accumulated and influence the springback
analysis. Therefore, the accuracy of springback simulation is not only related to
springback analysis itself, but also strongly dependent on the accuracy of forming
processes.
There are many papers [12–16] published to investigate the numerical fac-
tors which influence the accuracy of springback simulation, however, most of the
researchers have concentrated on the improvement of dynamic explicit FE codes.
The static implicit FE codes with different strategies of contact, matrix solutions
and element formulation etc., are not often discussed and no practical result is
concluded.
In the present subchapter, the static implicit FE code AutoForm is consid-
ered, and the main factors which influence the accuracy of springback simulation,
for instance, element formulation, time step, material model, drawbead model
and other numerical parameters, are discussed. Optimized numerical parame-
ters, so-called final validation settings, which aim at obtaining a robust and
accurate springback result, are suggested. For verification and comparison, exam-
ples of Numisheet benchmarks are used and stable and accurate results are
achieved.

4.5.2 Example Description

To simplify the investigation, a well known U-bending benchmark of Numisheet’93
is used during the optimization of numerical parameters [17]. Figure 4.40 illus-
trates the tool information and measurement method of the U-bending processes.
The blank dimension is 0.81 mm × 35 mm × 350 mm, with a value of
Young’s modulus of 71 GPa, a Possion’s ratio of 0.33, a density of 2.7 × 103
kg/m3 , a friction coefficient of 0.162 and the blank holder force of 2.45 kN.
The blank material is Aluminium and its stress-strain relation is expressed as
σ = 570.4 (0.01502 + εP )0.3469 , the anisotropic coefficients are 0.71, 0.58, 0.70 at
0, 45 and 90◦ with respect to the rolling direction. Due to the symmetry, only the
strip in one half is analyzed; therefore, the blank-holder force is taken as 1.225 kN.
Moreover, the gap between die and punch is additionally extended to 3 mm instead
of 1 mm only for the investigation of the influence of element size, time step and the
integration scheme, which aims at eliminating the combined influence from other
factors, e.g. contact treatment etc. For other factors’ investigation, the real gap of
1 mm is used.
4.5 The Springback Analysis 269

Fig. 4.40 Tool dimensions (a) and measurement method (b) of U-Bending benchmark of
Numisheet’93

4.5.3 The Influences on the Accuracy of Springback Simulation

4.5.3.1 The Influence of the Element Formulation
The element formulation is the core technology in numerical simulation of sheet
metal forming, which has significant influence on the simulation result, particularly
on springback simulation. Two classes of elements are provided in AutoForm, bend-
ing enhanced membrane (BEM) and elastic-plastic shell (EPS), respectively. Both
are based on a 3-node constant strain element formulation.
270 4 Numerical Simulation of the Sheet Metal Forming Processes

Fig. 4.41 Comparison of calculated springback between EPS and BEM

Fig. 4.42 Comparison of

bending behavior between
EPS and BEM

Bending enhanced membrane based on the scheme of uncoupled bending and

stretching solution [18], is the simplest one, which is very effective and cheap for
the numerical simulation of industrial parts, and has been verified with good accu-
racy on formability prediction [17, 19]. For springback analysis, however, BEM
provides inadequate accuracy, especially in the area of bending/inverse bending,
due to relative insufficient ability of description of bending effects.
Elastic-plastic shell element based on Mindlin-Reissner shell assumption is well
accepted in numerical simulation of sheet metal forming, which has good accuracy
in wrinkling prediction and springback analysis due to the better description of stress
distribution. Figure 4.41 shows the comparison of springback simulations obtained
by EPS and BEM. The reason why BEM has less accuracy as compared with EPS
can be explained in Fig. 4.42, in which, EPS demonstrates a better ability of bending
behavior description than BEM.

4.5.3.2 The Influence of the Element Size

Linear elements with 3 or 4 nodes are widely used in numerical simulation of sheet
metal forming due to cheaper cost. Their accuracy has been well verified for the
simulation in flat regions, in which the relative curvature (t/R) is smaller than 1/6.
4.5 The Springback Analysis 271

Fig. 4.43 The calculated springback varies with different element sizes

However, the calculated error cannot be neglected in small bending/inverse bending

regions (t/R > 1/6 [20]) where the linear shape function is insufficient. To improve
the accuracy of simulation in bending/inverse bending regions, a smaller element
size is required from theoretical point view. Figure 4.43 shows the tendency of
calculated springback varying with element size, which can verify that a smaller
element size will result in improved accuracy of calculated springback. When the
element size reaches 1/2∼1/3 of relative bending radii, the magnitude of calcu-
lated springback is starting to be robust and no further improvement is introduced
any more.
It should be kept in mind that the further decrease of element size will not only
notably increase the time consumption, but also increase the cost of contact treat-
ment and matrix solution etc., which might inversely introduce additional error of
springback simulation. As a result, 4 to 5 elements per radii are suggested for the
purpose of both accuracy and robustness of springback simulation.

4.5.3.3 The Influence of the Integration Scheme and Integration Points

The integration scheme and the number of integration points through sheet thickness
have been the key topics in springback simulation. The magnitude of springback
depends on the bending moment which in turn depends on the stress distribution
through sheet thickness. Most shell elements require numerical integration of stress
and strain distribution through sheet thickness in order to obtain bending moments
and tensile force, which will inevitably introduce numerical error in the simulation
results.
There are three types of integration schemes commonly recommended in element
formulation; which are, Gauss integration scheme, Gauss-Simpson combined inte-
gration scheme and Lobatto integration scheme. The Lobatto integration scheme
is employed by the AutoForm element model due to its accurate description of
numerical integration.
272 4 Numerical Simulation of the Sheet Metal Forming Processes

For the influence of the number of integration points through thickness (Nip),
there are many articles published [12–15, 20, 21], however, different number of
through thickness integration points (varying from 5 to 51) are specified, which
implies that the number of integration points through thickness is still an open issue
in springback simulation. In this subchapter, the influence of the number of integra-
tion points based on triangular element is investigated as well. Figure 4.44 shows
how simulated springback varies with the number of integration points. It is no doubt
that poor springback result will be obtained when the Nip is less than 3, however,

Fig. 4.44 The influence of the number of integration points through thickness

8
T(Min.)

3
Fig. 4.45 Time consumption Nip
versus number of integration 2
points 1 3 5 7 9 11 13 15 17
4.5 The Springback Analysis 273

when the Nip is larger than 5, the simulated springback starts to be stable, and there
is nearly no further improvement introduced after Nip is over 9.
Figure 4.45 shows how time consumption varies with the number of integra-
tion points through thickness. Larger Nip will result in more time consumption,
when the Nip is over 13, the time consumption will notably increase, however, no
further improvement of springback simulation is introduced (shown in Fig. 4.44).
Considering complicated bending/inverse bending cases (e.g. drawbead), and the
efficiency and accuracy, 11 Nip is recommended in AutoForm for springback
analysis.

4. 5. 3. 4 T h e I n fl u e n c e o f t h e T i m e Step
Different from the concept in dynamic explicit FE software [12, 13, 15], the time
step in the static implicit FE code, AutoForm, denotes the time interval to update
the strain and stress information of elements, or the tool displacements. The time
step can influence the description of strain and stress history (e.g. bending/inverse
bending history) and the treatment of friction etc. It is no doubt that a smaller time
step will result in more accurate simulation results, however, the time consumption
will be increased as well. Figure 4.46 shows the simulation results using various
time steps. The smaller time step will result in more stable springback. When the
time step is less than 2.5 s (based on a tool velocity of 1 mm/s) or maximum tool
displacement of 2.5 mm per increment, the magnitude of simulated springback is
starting to concentrate in a narrow band and no further improvement is introduced.
To track the bending/inverse bending history and the treatment of friction, the time
step of 1.6 s (or 1.6 mm) per increment is suggested in this subchapter.

Fig. 4.46 The calculated springback varies with time step

274 4 Numerical Simulation of the Sheet Metal Forming Processes

4.5.3.5 The Influence of the Material Model

Material modeling has been another key topic in numerical simulation of sheet metal
forming, where, the hardening model is widely thought to have an obvious influence
on springback simulation. Figure 4.47a shows the comparison of calculated spring-
back between isotropic hardening model and kinematic hardening model, while
Fig. 4.47b shows the transient softening compression curve and the kinematic hard-
ening parameters of Aluminum. With the kinematic hardening model introduced
in Sect. 2.8, an obvious improvement of calculated springback is obtained in this
benchmark.
However, it should be kept in mind that improvement of calculated springback
with kinematic hardening model is not guaranteed. Some examples with nega-
tive influence have been found as well during the investigation. The reason can

Fig. 4.47 The comparison of calculated springback between kinematic hardening model and
isotropic hardening model. (a) The comparison of calculated springback between isotropic and
kinematic hardening model; (b) The kinematic hardening parameters
4.5 The Springback Analysis 275

be explained as follows. There are two main factors in the kinematic hardening
model which may influence the final springback result; these are the transient soft-
ening of the hardening curve and the reduction of Young’s modulus. While the
factor of Young’s modulus reduction will cause larger springback deformation, on
the other hand, the factor of transient softening of the unloading hardening curve
will inversely result in smaller springback deviation. The final springback result is
decided by the combined influence of these two factors. In reality, the measure-
ment of kinematic hardening parameters is not so easy, in many cases, the error of
measurement is so large that even springback simulation becomes worse.
For an accurate springback simulation, well-measured experimental data related
to kinematic hardening parameters are required.

4.5.3.6 The Influence of the Drawbead Model

Drawbeads are widely used in stamping processes which aim at controlling the
material flow and improving the formability. There are two types of drawbead mod-
els in numerical simulation, which are, namely, physical drawbead and equivalent
drawbead model.
Physical drawbead model can simulate the real behavior of bending and inverse
bending history. However, it requires about 2.5∼8 times more computation time
than the equivalent drawbead model. Moreover, the radii of drawbeads are com-
monly so small that the application limit of shell theory is often exceeded, which
will introduce great challenges in numerical simulation of sheet metal forming.
Due to the above-mentioned reasons, the equivalent drawbead model is widely
used in numerical simulation, and it has been verified to provide good accuracy
in formability prediction [22, 23]. Springback analysis, however, is still in discus-
sion due to the neglection of bending and inverse bending history in the equivalent
drawbead model.
To investigate the influence of drawbeads on springback behavior, the well-
known S-Rail benchmark with drawbead provided in Numisheet 2008 [23] is used
in this investigation. Figure 4.48 shows the comparison of springback result between
a physical drawbead model and an equivalent drawbead model. Obvious differ-
ences in behavior of calculated springback are obtained after drawing (Fig. 4.48a),
however, nearly the same springback is observed after trimming (Fig. 4.48b).
To analyse the reasons why the springback calculated with the physical bead
and the equivalent bead is different after the processes of drawing and trim-
ming, the history of material flow over the whole forming processes is tracked
in Fig. 4.49, where, the highlighted curve on the blank surface denotes the bead
impact lines. As shown in Fig. 4.49, the material which flows through the drawbead
was finally cut after the trimming process, that means that the difference of bend-
ing and inverse bending path between physical drawbead and equivalent drawbead
model was finally trimmed, and the stress and strain distribution of the final part
is nearly identical, which gives the reason why the final springback is nearly the
same.
276 4 Numerical Simulation of the Sheet Metal Forming Processes

S1

Equivalent bead

S1 Physical bead
a)

S1

S1
b)
Fig. 4.48 The comparison on springback simulation between physical drawbead model and equiv-
alent drawbead model. a) The springback comparison with/without physical drawbead; b) The
springback comparison with/without physical drawbead(after trimming)

Fig. 4.49 Movement of a bead impact line during the forming process
4.5 The Springback Analysis 277

For the purpose of time consumption, the equivalent drawbead model is com-
monly recommended, because it makes no sense to handle complicated geometric
bead models if the equivalent drawbead can do the same. However, when the
material passing a drawbead remains on the final part, the difference of bend-
ing/inverse bending history resulting from drawbead model cannot be neglected and
the physical drawbead model is necessary.

4.5.4 The Optimized Numerical Parameters of Springback

Simulation: Final Validation Settings

Based on the above-mentioned investigations, the optimized numerical parameters,

which aim at obtaining a robust and accurate springback simulation, are suggested.
Table 4.27 shows the optimized numerical parameters suggested in this subchapter,
named as final validation settings.
With these settings, a robust and accurate springback simulation is expected to
be obtained.

4.5.5 The Simulation of Numisheet 2005 Benchmark #1: Decklid

Inner Panel

4.5.5.1 Description of the Benchmark

To check the behaviour and the accuracy of springback with the optimized numer-
ical parameters described above, the Numisheet 2005 Benchmark #1, Springback
prediction of decklid inner panel [22], which was provided by the GM Metal
Fabricating Division, US Steel, and Troy Design Manufacturing, is used. The

Table 4.27 The optimized numerical parameters for springback simulation

Element formulation Elastic-plastic shell

Radius penetration 0.22
Maximum element angle 15∼22.5
Initial element size Min (2 Rmin , 20 mm)
Maximum refinement level 3
Maximum time step or maximum 1.6∼2.2 mm
displacement
Tangential refinement ON
The number of integration points 11
through thickness (Nip)
Drawbead model Equivalent bead: material passing
bead is finally cut
Physical bead: material passing bead
remains on part

where, Rmin denotes minimum bending radius.

278 4 Numerical Simulation of the Sheet Metal Forming Processes

process involves pre-bending of the sheet, forming, trimming and constrained

springback. The objective of this study is to benchmark the springback prediction
capability of various software and skills of users on a complex inner body panel.
Lower punch, binder and upper die are illustrated in Fig. 4.50. This tooling geom-
etry has a symmetry plane at x=0. The process is a three-piece air draw. The punch
is stationary. The upper die is a solid one-piece tool. The binder is supported by
hydraulic cylinders, which provide the binder holding force. One metal thickness
equalizer [22] is used to provide a constant binder gap. Therefore, no binder pressure
is applied through the sheet during the drawing process.
Aluminum 6111 T4P is used in this subchapter with a value of Young’s modulus
70.9 of GPa, a Possion’s ratio of 0.3, a density of 2.7 × 103 kg/m3 , a friction coef-
ficient of 0.09, and the blank thickness is 0.9 mm. The BBC 2005 model is used in
this benchmark research which aims at an accurate description of the yield locus.
Table 4.28 illustrates the material parameters of AL 6111-T4P.
Due to the fact that there is no experimental data available for the kinematic hard-
ening model, the isotropic hardening model is still used in this benchmark research.
The equivalent drawbead model is used in this benchmark due to the fact that the
material passing drawbead is finally cut.
In the end, the optimized numerical parameters shown in Table 4.27, are used in
this benchmark, which aims to verify whether a reasonable springback simulation
can be obtained.

Fig. 4.50 The Benchmark #1 in Numisheet 2005

Table 4.28 Material parameters of AL 6111-T4P

True stress–true plastic

strain descriptions:
σ = C(εp − ε0 )m
σ0 σ 45 σ 90 σb
Material ε0 m C (MPa) r0 (MPa) r45 (MPa) r90 (MPa) rb (MPa) M

Al 6111 0.0044 0.268 558.6 0.616 127 0.646 132 0.778 136 0.58 128 8
T4P
4.5 The Springback Analysis 279

4.5.5.2 The Result of Numerical Simulation

Figure 4.51 illustrates the simulation of the entire forming processes of this bench-
mark, including pre-bending of the sheet, gravity, closing, drawing, trimming, rigid
body rotation in car position and constrained springback.
Figure 4.52 shows the simulation result of constrained springback based on the
optimized numerical parameters. Figure 4.53 gives the comparison between sim-
ulation and experimental data at the pre-defined measuring points. The statistical

Fig. 4.51 The simulation of the entire forming processes of a decklid inner panel
280 4 Numerical Simulation of the Sheet Metal Forming Processes

Fig. 4.52 The simulation result of constrained springback

6
Springback deviation(mm)

0
P1 P2 P3 P4 P5 P6 P7 P8 P9 P10 P11 P12 P13 P14 P15 P16 P17 P18 P19 P20

Measuring pionts

-2

-4 Experiment
Simulation

-6

Fig. 4.53 The springback comparison between simulation and experiment

4.5 The Springback Analysis 281

Fig. 4.54 The comparison on the statistical result of springback deviation between simulation and
experiment

analysis shows that the standard deviation between the calculated springback and
the experimental results is well-controlled under the value of 0.65 mm.
Figure 4.54 shows the comparison of the statistical result of absolute error for
springback in normal direction, based on the published data of Numisheet 2005
[22], where, AutoForm V4.1.1 is the statistical result of calculated springback in
this subchapter based on the final validation settings, BM1.18 is the published result
of the author, and BM1.20 is the published result of the AutoForm user in Numisheet
2005 [22], both are based on AutoForm V4.0 and similar settings of the numerical
parameters shown in Table 4.27.
All these results verify that a robust and accurate springback simulation can be
guaranteed with so-called final validation settings.

4.5.6 Conclusion
The springback simulation is a well-known to be a sensitive process, which is not
only influenced by springback computation itself, but also strongly depends on the
accuracy of previous forming simulation. There are so many numerical parameters
influencing the accuracy of springback calculation, that it is not easy to obtain a
robust and accurate springback simulation.
In this subchapter, the influences on springback simulation are thoroughly inves-
tigated, and the optimized numerical parameters, so-called final validation settings,
are provided. The verification of experiment has shown that the robust and accurate
springback simulation can be guaranteed with these settings. With the robust and
282 4 Numerical Simulation of the Sheet Metal Forming Processes

accurate springback simulation, further developments, for instance, computer aided

springback compensation etc. can be realized.

4.6 Computer Aided Springback Compensation

4.6.1 Introduction
Springback is an inevitable problem in the field of die face engineering. Although it
is impossible to prevent the springback, it can be minimized by some techniques, for
instance, reinforcing part by smaller radii or additional folding, raising the stretch-
ing deformation of the sheet etc. [24]. Even so, there are still many cases where
the springback deviation exceeds the given tolerance. Where the minimized spring-
back deviation is still so large that the subsequent assembly operation is seriously
influenced, the additional geometric modification of the tool surface, the so-called
springback compensation, has to be introduced in order to reduce the shape devia-
tion between the drawn part and desired product. With the increasing usage of high
strength steel and aluminum, springback deviation becomes more and more severe.
Therefore, the geometric compensation of springback is generally necessary to be
taken into account.
In the past, springback compensation was done manually by doing extensive
measurements on prototype or even production tools, and altering tool geometry by
hand, which is a time consuming and cost-prohibitive process. It was reported [24]
that a single correction loop of springback compensation for a hood inner like part,
takes about 5 weeks and costs about C70,000. If additional operations are involved,
one extra-iteration may additionally take about 10 weeks and cost about C150,000.
In many cases, numbers of iterations are needed during the compensation, which
seriously increases the cost of tool development. With complex part shapes and new
materials, it is difficult or even impossible, to rely on such kind of experience to
estimate shape deviations and compensate the die surfaces. Therefore, how to con-
trol the cost of compensation and shorten the period of tool development becomes
one of the key issues in current die manufacturing industry.
With the development of computer technologies and the finite element method,
particularly with recent improvements of springback prediction, a new compensa-
tion method based on calculated springback has been developed. With the help of
this method, the springback can be compensated easily and effectively and the cost
of tool development can be significantly decreased. In this subchapter, the basic
methodologies of springback compensation are presented, and the factors which
influence the springback compensation, for instance, the accuracy of springback
simulation, the robustness of springback responding to the variation of material and
process parameters in manufacturing process, are discussed, and a guide line which
aims at fulfilling a successful springback compensation is recommended. As a dis-
cussion subject, the example of Numisheet 2005 Benchmark #1 [22] is used again
in this subchapter.
4.6 Computer Aided Springback Compensation 283

4.6.2 The Basic Methodologies of Computer-Aided Springback

Compensation
There are two methodologies used to realize the geometric compensation of
springback. They are the so-called spring-forward compensation and spring-back
compensation. Both are based on an iterative procedure.
In the approach of spring-forward compensation, the stress tensor of the part,
which is used to calculate springback, is multiplied with a negative factor, there-
fore, the shape after ‘springback’ simulation is already the subsequent compensated
surface. This approach has been thought as an effective method. However, only few
successful stories are reported, as compared with the approach of springback com-
pensation. The reason can be explained as follows. The spring-forward approach is
based on the assumption that the sign change of stress tensor correspondingly results
in the response of spring forward deformation instead of spring backward, which is
not 100% true in reality due to the complicated stress distribution in drawn part.
In the approach of springback compensation, the geometric compensation is
based on the real calculated springback. Therefore, there is no artificial error intro-
duced as compared with spring-forward approach. Figure 4.55 illustrates the basic
work flow of computer-aided springback compensation [25]. First, the forming and
springback simulation must be run based on the original CAD geometry and process
parameters. After that, the shape deviation between the sprung part and the desired
shape has to be analysed in order to determine whether the dimension of the drawn
piece can meet the needs of the desired shape. If the shape deviation exceeds the

Fig. 4.55 The work flow of

computer-aided springback
compensation
284 4 Numerical Simulation of the Sheet Metal Forming Processes

given tolerance, the geometric compensation of springback based on pre-defined

strategies of compensation and compensation factor, must be implemented, in order
to decrease the shape deviation with respect to the desired shape. In the end, the
drawing and springback processes have to be assessed again, because the spring-
back might have been significantly changed due to the modified tool geometry. This
procedure is repeated until the desired shape is obtained.
In this subchapter, the second approach of springback compensation is used due
to the fact that it is closer to the compensation routine used in practice.

4.6.3 The Influences of the Quality of Computer Aided

Springback Compensation

The quality of computer aided springback compensation (CASC) is critical for tool
development. Unreliable compensated data or unstable springback behavior will sig-
nificantly increase the cost of tool development, thereby influence the time schedule
of car development. There are two main factors which will influence the quality of
computer aided springback compensation, namely the robustness and accuracy of
springback prediction, and the robustness of springback responding to the variation
of material and process parameters.

4.6.3.1 The Robustness and Accuracy of Springback Calculation

The robust and accurate springback simulation is critical for computer-aided
springback compensation. It is well-known that the springback simulation is sen-
sitive to many numerical parameters, for instance, element formulation, element
size, number of integration point through thickness, time step, material model etc.
To ensure the reliable compensated data being introduced, the optimized numeri-
cal parameters, so-called final validation settings, are strongly required during the
computer aided springback compensation. It has been well-verified by many exam-
ples [22, 23] that both accuracy and robustness of springback simulation can be
guaranteed with such a kind of final validation settings.

4.6.3.2 The Robustness of Springback Responding to Variation of Material

and Process Parameters
It is also well known that the springback can sensitively respond to the variation of
the material properties and process parameters existing in the manufacturing pro-
cess. Small relative variations of the parameters might result in significant changes
of springback, in some cases, even the direction of springback, which introduces
great challenges for springback compensation. Particularly with the increasing
usage of high strength steel and aluminum, such problems become more serious.
To ensure the quality of springback compensation, a robustness analysis is recom-
mended before and after compensation. If the scatter of springback exceeds the
allowed tolerance of springback compensation, the process parameters or tooling
conception have to be optimized in order to obtain a repeatable result.
4.6 Computer Aided Springback Compensation 285

4.6.4 The Recommended Work Flow of Computer-Aided

Springback Compensation
To obtain a successful springback compensation, a recommended work flow of
computer-aided springback compensation [25] (shown in Fig. 4.56) is suggested
as follows.

Fig. 4.56 The recommended

work flow of CASC

4.6.4.1 The Optimization of Process Layout

When the first tooling conception is generated, the layout of deep drawing
stage should be validated with respect to splits, wrinkles, thinning and potential
skid/impact lines etc. which aims at obtaining a feasible process layout in the
286 4 Numerical Simulation of the Sheet Metal Forming Processes

first optimization. Moreover, the optimization of material usage based on nest-

ing has to be considered as well, because any change of sheet outline later might
cause a significant change of springback deviation, therefore, all efforts following
might be irrevocably lost. It should be kept in mind as well that the springback
compensation is only the last measure to reduce the shape deviation between the
drawn part and desired product, due to the fact that it is a time consuming and
cost-prohibitive process. Trying our best to reduce the springback in the stage of
geometry and process improvement is strongly recommended before springback
compensation.

4.6.4.2 Compensability Analysis and Optimization of Tooling Concept

After the optimization of process layout, the first survey of springback has to be
done in order to clarify whether the springback can be compensated geometrically
or whether the tooling concept needs to be modified. If not, additional optimization
loops have to be introduced.

4.6.4.3 Robustness Analysis and Optimization Before Compensation

Based on the optimization of process layout and compensative analysis, the robust-
ness of springback responding to the variation of manufacturing processes is
recommended to be analyzed before compensation, because a successful compen-
sation is closely connected to a robust forming process. It has been shown in
practice that more dimensional deviation due to springback occurred despite the
compensated tools, when a non-robust springback is introduced [25].

4.6.4.4 Computer Aided Springback Compensation

Based on a robust springback response, the geometric compensation can be intro-
duced when springback deviation exceeds the given tolerance. Three different
areas within the tool are required to be defined during the compensation. First,
the area of the part which will be compensated directly based on the node vec-
tor of springback displacement and the compensation factor. Second, the area of
binder surface which remains fixed in order to make sure that no wrinkles occur
during the closing of tools. Third, a transition area corresponding to rest part of
tool is used for generating the smooth transition surface between direct compen-
sation and fixed region. With the definition of compensation areas, the computer
aided springback compensation based on the above mentioned springback compen-
sation approach (see Fig. 4.55) will be introduced and the compensated tool will be
generated.

4.6.4.5 Robustness Analysis and Optimization After Compensation

Due to springback compensation, material flow might be changed, in turn, the
springback behavior might become non-robust again if there is a great amount
4.6 Computer Aided Springback Compensation 287

of geometric change of tool surface. Therefore, the robustness analysis after the
springback compensation is necessary in order to ensure that the final scheme of
springback compensation is repeatable.
Based on the above guideline, the robust springback compensation can be guar-
anteed and the cost of tool development related to the springback compensation can
be significantly decreased.

4.6.5 The Springback Compensation of Numisheet 2005

Benchmark #1

To demonstrate the application of the new method outlined above, the Numisheet
2005 Benchmark#1 Decklid inner panel [22] is used. The process parameters and
material properties are described in the previous subchapter in detail.

4.6.5.1 The Feasibility Analysis of the Forming Processes and Springback

The numerical simulation of the entire processes of this benchmark has been
presented in the previous subchapter. The verification of the experiment has well-
testified that the accurate and robust simulation result can be obtained with the
optimized so-called final validation parameters.
Based on the simulation, the deep drawing process has been analyzed and
well-verified with respect to splits, wrinkles and thinning etc.. Therefore, the fea-
sibility of the tooling conception and process parameters of this benchmark can
be taken as guaranteed and the subsequent possibility of the geometric springback
compensation can be discussed.
From the springback simulation shown in Figs. 4.52 and 4.53 in the pre-
vious Sect. 4.5, we can find that the maximum shape deviations between the
sprung piece and the desired shape are –5.54 mm at the measuring point P4, and
4.14 mm at the measuring points P12 and P16, which exceeds the required toler-
ance. Therefore, the geometric compensation of the tool surface is necessary for this
benchmark.

4.6.5.2 The Robustness Analysis of Springback Before Compensation

The stable springback responding to the variation of the manufacturing noise
is critical for the successful springback compensation. Therefore, the robustness
analysis of springback is necessary in order to guarantee that the dimension of
the drawn part is repeatable for the stamping process. According to the pro-
cess layout of this benchmark, the variation of the following parameters is
considered:

• Sheet thickness
• Sheet position
288 4 Numerical Simulation of the Sheet Metal Forming Processes

• Friction coefficient
• Yield strength and tensile strength of the material at 0, 45 and 90◦ with respect
to the rolling direction
• Drawbead restraining force resulting from friction and wear, etc.

With the help of AutoForm-Sigma, these parameters are automatically varied

according to the defined variation range and the individual simulations are evaluated
based on statistical methods.
Figure 4.57 shows the distribution of the standard IQR [26] which denotes the
variation of springback along the normal direction (the meaning of standard IQR is
similar to the standard deviation in statistics). From Fig. 4.57, we can see that the
value of the standard IQR is well-controlled under 0.374. Thus one may consider
that the springback is still robust.
Figure 4.58 shows the scatter of springback deviation in Zone 1 and Zone 2 (the
positions of Zone 1 and Zone 2 are marked in Fig. 4.57), which approximately cor-
respond to the largest negative and positive springback respectively. From Fig. 4.58,
we can see that the springback along the normal direction varies between –5.44 and
–6.76 mm in Zone 1, and between 3.45 and 4.69 mm in Zone 2. This fact proves
that the springback directions of both zones are stable. Therefore, the quality of
springback compensation can be guaranteed.

Fig. 4.57 The variation of springback along the normal direction

4.6 Computer Aided Springback Compensation 289

Fig. 4.58 The scatter of springback deviation of Zone 1 (left) and Zone 2 (right) before
compensation (Miu: friction coefficient)

4.6.5.3 Computer-Aided Springback Compensation

Using the module AutoForm—Compensator, the compensation strategy of this
benchmark is defined. In this compensation strategy, the tool geometry is subdi-
vided into the areas of ‘Fixed region’, ‘Direct compensation region’, and ‘Transition
region’, which are shown in Fig. 4.59. The binder surface is defined as fixed in order
to make sure that no obvious change of the deformed shape will be introduced after
closing.

Fig. 4.59 The definition of springback compensation regions

290 4 Numerical Simulation of the Sheet Metal Forming Processes

Fig. 4.60 The tool geometry after springback compensation

With the well-defined compensated regions, the springback compensation is

automatically done by AutoForm—Compensator. Figure 4.60 shows the tool geom-
etry after springback compensation, where, section A-A and section B-B are
the comparison between the compensated surface and the part before and after
springback.
After only two iteration loops of springback compensation, the geometric devia-
tion with respect to desired shape is well controlled under the required tolerance
of 0.2 mm. Figure 4.61 shows the geometric deviation between the sprung part
and the desired shape, where, section A-A and section B-B in this figure are the
comparison between the drawn part and the desired shape. Figure 4.62 shows
a comparison of shape deviation before and after compensation at measuring
points.

4.6.5.4 Robustness Analysis of Springback After Compensation

Due to the change in tool geometry, it cannot be assumed that the robust form-
ing process before compensation is still robust after compensation. Therefore, a
final robustness analysis based on the compensated tool surface is necessary to be
4.6 Computer Aided Springback Compensation 291

Fig. 4.61 The geometrical deviation between the sprung part and the desired shape after
springback

4
Distance from normal [mm]

0
P1 P2 P3 P4 P5 P6 P7 P8 P9 P10 P11 P12 P13 P14 P15 P16 P18 P17 P19 P20

–2 Measurement points

–4
Before compensation(Exp)
–6 Before compensation(Sim)
After compensation itr. 2(Sim)
–8

Fig. 4.62 Comparison of the shape deviations before and after compensation
292 4 Numerical Simulation of the Sheet Metal Forming Processes

checked again. The variation of the following parameters is considered during the
robustness analysis:

• Sheet thickness
• Sheet position
• Friction coefficient
• Yield strength and tensile strength of the material at 0, 45 and 90◦ with respect to
the rolling direction
• Drawbead restraining force resulting from friction and wear, etc.

Figure 4.63 shows the distribution of the standard IQR [26] which denotes the
variation of distance from the desired shape along the normal direction. It can be
seen that no values of standard IQR greater than 0.3 mm are identified. Therefore,
the process after compensation can be considered as robust and the quality of this
compensation scheme can be taken as guaranteed.
Figure 4.64 shows the Cp [26] assessment for the distance with repect to the
desired shape along the normal direction. From Fig. 4.64 we can see that most areas
of the drawn part belong to the reliable region (marked with the green colour),
which denotes that the compensation is reliable and the compensated tool can be
manufactured.

Fig. 4.63 Variation of the distance from the desired shape along the normal direction
4.6 Computer Aided Springback Compensation 293

Fig. 4.64 The Cp assessment for distance from the desired shape along the normal direction

4.6.6 Conclusion
Springback compensation is a challenging and cost prohibitive process in current
die manufacturing. With help of computer-aided springback compensation, this pro-
cess can be easy and effective. To obtain a successful springback compensation,
additional analysis and optimization have to be done before and after compensation.

• The robust and accurate calculated springback is critical for computer aided
springback compensation. Therefore, the use of optimized numerical parameters,
the so-called final validation settings, is strongly recommended in springback
simulation in order to ensure the reliable compensated values.
• The repeatability of springback deviation in the real stamping process is the cen-
tral premise to fulfill the springback compensation. Therefore, the robustness
analysis and optimization before and after compensation is necessary in order
to make sure that the springback compensation is reliable.
• In the end, a guideline for the sucessful computer aided springback compen-
stion is presented in this subchapter. Using this guideline, a robust compensation
scheme can be guaranteed and the cost of tool development can be significantly
decreased.
294 4 Numerical Simulation of the Sheet Metal Forming Processes

Acknowledgements Finally, I would like to thank all my colleagues, especially Thomas

Schoenbach, who were involved in the preparation of this subchapter.

References
1. Kleiner M et al. (2002) Material parameters for sheet metal forming simulations by means of
optimization algorithms. Final Report of the European Project 7210-PR-244, Research Fund
for Coal and Steel (RFCS)
2. AutoForm 4.1 Software Manual. (2007) AutoForm Engineering GmbH, Zurich
3. Sigvant M, Mattiasson K, Vegter H, Thilderkvist P (2009) A viscous pressure bulge test for the
determination of plastic hardening curve and equibiaxial material data. International Journal
of Material Forming, 2:235–242
4. Hosford WF (1996) On the crystallographic basis of yield criteria. Texture and
Microstructures 26–27:479–493
5. Mattiasson K, Sigvant M (2004) Material characterization and modeling for industrial sheet
forming simulations. In: Ghosh S, Lee JK, Castro JC (eds) Proceedings of the NUMIFORM
2004, AIP, New York, NY, 875–880
6. Forming of new metallic materials. Brite Euram Project, 1996–1999.
7. Sigvant M, Mattiasson K, Skogsgårdh A (2007) Industrial experiences of stochastic simula-
tions of sheet metal forming In: Hora P (ed) Proceedings of the Application Stochastics and
Optimization Methods, Zürich, 23–31
8. Bonte MHA et al. (2006) Optimizing towards robust metal forming processes. In: Juster
N, Rosochowski A (eds) Proceedings of the International Conference on Material Forming
ESAFORM 2006, Glasgow, 47–51
9. Dietrich E, Schulze A (2005) Statistische Verfahren zur Maschinen- und Prozessqualifikation,
Hanser Verlag, Munchen
10. Carleer B (2006) Applicability of stochastical methods in the control of the scatter influ-
ence of constitutive parameters. Proceedings of the FLC Conference, March 15th–16th 2006,
Zürich
11. Lee CH (2006) Application of AutoForm Sigma on a hood inner. Proceedings of the 5th Press
Forming Symposium, Korean Society of Technology for Plasticity, Jeju Island, June 2006
12. Xu WL, Ma CH, Li CH, Feng WJ (2004) Sensitive factors in springback simulation for sheet
metal forming. Journal of Materials Processing Technology 151:217–222
13. Valente F, Li XP, Messina A (1997) Springback prediction for stamping tools compensation
by numerical simulation. Centro Ricerche Fiat, Torino
14. Chung WJ, Cho JW, Belytschko T (1996) A study on dynamic effects of dynamic explicit
FEM in sheet metal forming analysis. In: Lee JK Kinzel JL, Wagoner RH (eds) Proceedings
of the NUMISHEET’96 Conference, Columbus, OH, 414–426
15. Lin Z, Liu G (2000) Study on the effects of numerical parameters on the precision of spring-
back prediction. Proceedings of the 6th International LS_DYNA User’s Conference, Session
13C, Dearbon, MI
16. Huang M, Gerdeen JC (1994) Springback of doubly curved developable sheet metal surface –
An overview. SAE Technical Paper No. 940938, 718–731
17. Makinouchi A, Nakamachi E, Onate E, Wagoner RH (eds) (1993) Proceeding of the 2nd
International Conference and Workshop on Numerical Simulation of 3D Sheet Metal Forming
Processes, Isehara
18. Kubli W, Anderheggen E, Reissner J (1991) Nonlinear solver with uncoupled bending
and stretching deformation for simulating thin sheet metal forming. Proceedings of the 1st
International Conference and Workshop on Numerical Simulation of 3D Sheet Metal Forming
Processes, VDI, Dusseldorf, 325–343
19. Yang DY, Hug H, Oh SI, Kim YH (eds) (2002) Proceedings of the 5th International
Conference and Workshop on Numerical Simulation of 3D Sheet Metal Forming Processes-
Verification of Simulation with Experiment, Jeju Island
References 295

20. Li K, Wagoner RH (1998) Simulation of springback. In: Huetink J, Baaijens FPT (eds)
Simulation of Materials Processing: Theory, Methods and Applications. Proceedings of the
6th International Conference on NUMIFORM 1998, Enschede, 21–31
21. Wagoner RH, Li M (2005) Advances in springback. In: Smith LM, Pourboghrat F, Yoon
JW, Stoughton TB (eds) Proceedings of the 6th International Conference and Workshop on
Numerical Simulation of 3D Sheet Metal Forming Processes-Verification of Simulation with
Experiment, Numisheet 2005, Detroit, 209–214
22. Smith LM, Pourboghrat F, Yoon JW, Stoughton TB (eds) Proceedings of the 6th International
Conference and Workshop on Numerical Simulation of 3D Sheet Metal Forming Processes-
Verification of Simulation with Experiment, Numisheet 2005, Detroit
23. Hora P (ed) (2008) Proceedings of the 7th International Conference and WorkShop on
Numerical Simulation of 3D Sheet Metal Forming Processes-Verification of Simulation with
Experiment, Numisheet 2008, Interlaken
24. Roll K, Lemke T, Wiegand K (2005) Possibilities and strategies for simulation and com-
pensation for springback. In: Smith LM, Pourboghrat F, Yoon JW, Stoughton TB (eds)
Proceedings of the 6th International Conference and Workshop on Numerical Simulation of
3D Sheet Metal Forming Processes-Verification of Simulation with Experiment, Numisheet
2005, 295–302
25. Schoenbach T, Bauer T (2008) New method to calculate and compensate springback. In: Hora
P (ed) (2008) Proceedings of the 7th International Conference and Workshop on Numerical
Simulation of 3D Sheet Metal Forming Processes-Verification of Simulation with Experiment,
Numisheet 2008, Interlaken, 515–520
26. AutoForm Release Notes 4.0 (2005) AutoForm Engineering. GmbH, Zurich
Index

A 2004 yield criterion, 79–80, 116

ABAQUS/Explicit, 23–24 Bauschinger effect, 121–135
Advanced feasibility, 218–227, 229 Beisswänger test, 150
Anisotropy, 27, 29, 30–39, 45–46, 51–53, Bending enhanced membrane elements, 24,
56–60, 62, 64, 66, 68–71, 73, 75, 79, 222
82–83, 89–91, 94, 98, 101, 106, 108, Biaxial
113, 116–117, 120, 126, 144, 170, 181, anisotropy, 36–37, 79, 82, 118, 250
197, 204, 242, 250 balanced stress, 37
‘anomalous’ behavior, 51–52, 54, 58–60, tensile testing, 37, 115
76 yield stress, 27, 29, 38, 56, 66, 79, 81, 86,
‘anomalous’ behavior of second order, 52, 90, 92–93, 106–107, 111, 118, 120,
58–60, 118 213, 239, 242, 245, 247, 250
biaxial coefficient, 38–39, 66, 81, 90 Blank holding pressure, 142
coefficient, 27–29, 30–39, 46–48, 51–53, Bragard method, 162–164
55–60, 62, 64–65, 67, 70, 73–74, 79, Budiansky yield criterion, 74–75
82, 85–89, 115, 117–118, 144, 170, Bulge
181, 204, 213 bulge tests, 250
planar, 27, 35, 53, 61–62, 75, 87 forming process, 238–241
principal axes, 45–46
uniaxial coefficient, 30–36, 38, 58, 70, 83, C
93, 113–114, 116, 120 CAD data, 219
Aretz–Barlat yield criterion, 79 Capability, 214–215, 230–236, 257–258, 278
Associated flow rule, 9, 34, 39, 54, 73, 75, 127 Cauchy stress, 9
AutoForm Cayssials model, 204
model, 131–135, 134 Cazacu–Barlat 2001 yield criterion, 115
Sigma, 257–258 Cazacu–Plunckett–Barlat yield criterion, 87
Springback compensator, 289–290 Central difference method, 19
Chaboche model, 128–129
B Coefficient
Banabic, Balan, Comsa (BBC) biaxial anisotropy, 36–39, 79
2000 yield criterion, 82, 113–115, 117–120 geometrical non-homogeneity, 183
2005 yield criterion, 91–106, 198–199, normal anisotropy, 35
213, 278 plastic anisotropy, 33–38, 51–52, 56, 58,
2008 yield criterion, 106–112, 116, 120 63–64, 66, 76–77, 80–83, 87, 113–114,
Barlat 116, 120
1989 yield criterion, 27–28, 61–66, 106 uniaxial anisotropy, 56
1994 yield criterion, 27, 29 Combined
1996 yield criterion, 73, 77 deep-drawing methods, 150–151
2000 yield criterion, 27–29, 82, 113–115 drawing, 147

D. Banabic, Sheet Metal Forming Processes, DOI 10.1007/978-3-540-88113-1, 297


C Springer-Verlag Berlin Heidelberg 2010
298 Index

Compression test, 37, 169 F

Computer aided springback compensation, Feasibility, 214–216, 218–224, 237, 287
282–293 prediction, 223
Comsa yield criterion, 89–90 solution, 216, 219, 221–222, 224, 226–227
Considère criterion, 181 Ferron yield criterion, 75
Consistency condition, 9 Finite deformations, 13–16
Consistent mass matrix, 12, 14 Flow approach, 16–18, 23
Co-rotational rate, 9 Flow rule, 9, 34, 39, 46, 54, 57, 67–68, 73, 75,
Courant condition, 20 92–94, 96, 99, 107, 127, 129, 188, 193
CrachFEM code, 197 Formability, 141–204, 220, 222, 224, 227, 242,
CrachLAB code, 197 262, 268, 270, 275
Critical time step, 21 index, 144, 146, 148–151
Cross die, 244–250 tests, 147
Current configuration, 11 FORM-CERT code, 198
Forming Limit Band (FLB), 156
D Forming Limit Curve (FLC), 152–156, 161,
Decklid inner panel, 277–281, 287 165, 173, 180, 182, 184, 194, 197–203
Deep drawing, 149–151, 175, 286–287 Forming Limit Diagram (FLD), 147, 152–179,
methods, 150 198
Defects, 142–143 Frame invariant, 4
Deformation gradient, 5 Front side member inner, 258–262
Digital process planning, 215–218, Fukui test, 150
236–238
G
Displacements, 4
Gauss integration scheme, 271
Double profile method, 162
Geometric
Drawbead model, 268, 275–282
compensation of springback, 282–284
Drawing ratio, 142, 150
stiffness matrix, 15
Drucker yield criterion, 43–44
Geometrical non-homogeneity, 183–184
Dynamic
Geometry check, 214–215, 219
equilibrium equations, 19
Gotoh yield criterion, 75, 88, 90
explicit, 18–21, 24,
Green strain, 6, 8
219, 268, 273
Grid size, 166–168
Guyot test, 149
E
Effective H
plastic strain, 9 Hardening
plastic strain rate, 10 isotropic, 49, 125–126, 129–130, 274, 278
stress, 9 kinematic, 125–126, 128–130, 132, 134,
Eisenkolb test, 150 197, 274–275, 278
Elastic energy of distortion, 42 Hasek Test, 161
Elastic-plastic Hecker
constitutive relation, 22 method, 163
shell, 225, 269–270, 277 test, 148, 159
Engelhardt test, 150 Hencky yield criterion, 42–43
Erichsen Hershey yield criterion, 44, 66, 81, 83
index, 148 Hill
test, 148 model, 173
Eulerian 1948 yield criterion, 52–61, 78, 88, 98, 99,
description, 3 125, 126, 135–137, 141, 143, 214
FE-formulations, 16–18 1979 yield criterion, 59–62, 65
Explicit method, 18–21 1990 yield criterion, 62
Extended stress-based limit curve, 154 1993 yield criterion, 68
External force vector, 12, 14 Hood inner, 258, 262–264
Index 299

Hosford yield criterion, 61–63, 72, 88, 120, M

242 Manufacturability, 214–215, 218–219,
Huber yield criterion, 42–43 225–232
Hutchinson–Neale model, 182–185 Manufacturing noise, 287
Hu yield criterion, 89 Marciniak–Kuckzynski model, 182–185, 198
Hydroforming processes, 241 Marciniak test, 159–162
Material
I data, 215–218, 254
Implicit inputs, 232
alghoritm, 185 model, 3, 9–11, 22, 122, 197, 219, 236,
method, 19 242, 244, 252–255, 268, 274, 284
Incremental simulations, 229 properties, 232–233, 256, 284, 287
Influence stiffness matrix, 15
normal pressure, 172–175 MATFORM code, 197
strain rate, 171–172 Maximum shear stress criterion, 40
temperature, 171 Mechanical
Initial stress stiffness matrix, 15 parameters, 37, 47, 51, 57–60, 79–81,
In-process measurement, 164 112–113, 115–116, 118–119, 170, 199,
Integration 204, 230
points, 240, 248, 271–273, 284 properties, 30, 121, 170–171, 220, 229,
scheme, 271 232, 256–260
Internal Methods based on simulating tests, 147–151
force, 12 Mindlin-Reissner shell, 270
force vector, 13 Modified Maximum Force Criterion (MMFC),
Inverse one-step method, 218 195–197
Myauchi test, 161
J
N
Jovignot test, 149
Nakazima test, 160–161
Necking, 142, 152, 154, 163–166, 168,
K 175–176, 179, 182–183, 187, 193–196,
Karafillis–Boyce yield criterion, 67–68, 71–74, 263
81, 120 Nodal displacements, 12
Keeler–Brazier model, 204 Nodal velocities, 18
Keeler test, 158–159 Normal anisotropic coefficient, 49
Kirchhoff stress, 9 Normal anisotropy, 28, 35, 48, 62, 106
Normal direction (ND), 30, 45, 70, 91, 107,
L 281, 288, 292–293
Lagrangian Normality condition, 11
descrition, 3
strain, 16 O
Lankford parameter, 30, 242 Objective stress rate, 10
Limit Olsen test, 148, 157
dome height test, 151 Optimization of process layout, 286
strain, 152, 154, 156, 162–165, 168–170, Orthotropy, 30, 75, 81, 84, 91, 96, 107, 120
181, 196–198, 204 axes, 30, 120
Linear planar axes, 59
elasticity, 12 OSUFT test, 151
perturbation theory, 194 Outer trunklid, 251–254
stiffness matrix, 13
Lobatto integration scheme, 271 P
Localized necking, 179, 182, 194 Pareto chart, 265–266
Logarithmic strains, 15 Part feasibility, 214–215, 218–226
Lower specification limit, 264 Particle, 3
 300 Index

Performances of the yield criteria, 116–118 Scatter plot, 266–267

Petrasch test, 150 Second Piola-Kirchhhoff stress, 8
Plane stress, 10 Semi-empirical models, 203–204
Plastic Severity index, 176–177
strain, 33, 127–128, 131–132 Siebel test, 149
strain rates, 11, 34 Sill reinforcement, 254
Point, 4 Simulation
Polynomial yield criteria, 88–91 inputs, 214
Prager model, 127–128 test, 147
Process Small deformation, 6, 11–13
capability, 214–215, 232–233, 235, 237, tangent stiffness matrix, 15
257–258 Soare yield criterion, 90–91
feasibility, 219, 230 Spherical cup, 241–244
layout, 214–216, 218, 220–225, 227, 232, Spin, 6
237, 286–287 Springback
layout geometry, 221, 223 accuracy, 268–277, 284
layout outcomes, 215 analysis, 267–270, 275
parameters, 177, 257, 282–285, 287 calculated, 275, 281, 284
planning, 213–215, 218, 236–238 compensation, 282–291, 293
validation, 214, 225–230 computer aided compensation (CASC),
Production robustness, 214, 226 284–285
Punch constrained, 279–280
curvature, 171 prediction, 122, 131, 230, 254–255,
stretching methods, 148–149 277–287, 284
stretching test, 157–158 simulation, 268, 274, 276–278, 281–284,
287, 293
R
Spring-forward compensation, 283
Rate
S-Rail benchmark, 275
deformation, 6, 9
Stamping
plastic strain, 10
environment, 214, 222, 228, 231
Reference configuration, 13
plant, 232
Reliability, 220, 224, 229–230, 236, 256, 265
Representation theory of tensor functions, 84, process, 262, 270, 275, 280, 287, 293
86 simulation, 23–24, 213, 225, 229, 237
Rigid-plastic Standard Inter Quantile Range (IQR), 258,
approach, 22–23 260, 288, 292
model, 17 Static
theory, 11 explicit approach, 23
Robust implicit approach, 22–24
analysis, 287, 290–293 implicit FE code, 268, 273
compensation, 293 Statistical process control, 257–258
design, 255–267 Stochastic analysis, 257–260,
springback compensation, 282 263–264
Robustness, 230–236, 265, 267, 271, 282, Stören–Rice model, 179–180, 204
284–289, 293 Strain
analysis, 233, 265, 267, 285–287, 293 energy criterion, 40
springback simulation, 268, 271, 277 matrix, 13
Rolling direction, 30, 32–35, 45, 50, 57, 63, measures, 4–8
67, 79, 81–82, 85, 87, 89, 91, 107, 113, path, 186–192, 202–204, 229, 241
117, 201, 226, 234, 238 signatures, 253
Strain-rate method, 164
S Stress
Sachs test, 149 force vector, 12
Safety margin, 176, 236, 256, 259, 267 forming limit diagram, 155
Index 301

Stretch Vegter yield criterion, 29, 87–88

forming, 241, 250, 253–254 Velocities, 4
values, 15 gradient, 7
Strip stretch test, 151 Viscoplastic material, 17
Structural non-homogeneity, 183 von Mises yield criterion, 40, 43, 49, 52
Surface quality, 230, 236
Swift W
model, 179 Wang yield criterion, 89
test, 194 Wedge drawing test, 149
Swift/Hockett/Sherby law, 213 Working flow, 285–286
Wrinkling, 161
T limit diagram, 177
Tangent matrix, 15
Tearing, 142, 149–150, 159
Y
Tensile test
Yield
equibiaxial, 60, 76
condition, 9–11, 44, 127
uniaxial, 30, 52, 60, 67, 69, 74, 90, 112,
criterion, 35, 38–39, 41–47, 49–54, 58–61,
157, 244
63–64, 66–68, 71, 74–76, 79–94,
Tension-compression test, 169
Three circle method, 162 105–106, 109, 112–115, 126, 128, 182,
Time step, 19–23, 268, 273 186, 195, 198–202, 213, 239
Total strain rates, 11 function, 39, 44–46, 51, 54, 56–58, 60–62,
Total stress, 11 65, 67–68, 70–75, 77, 79–80, 83–90,
Transverse direction, 30, 59, 81, 91, 107, 111 107, 115, 126, 141, 181, 239–241,
Tresca 245–247, 249
yield criterion, 41, 44 locus, 44, 48–50, 60–61, 64–65, 86–88,
yield surface, 44 113, 115–117, 198–200, 251, 253, 278
True stress, 8 point, 39
Tryout solution, 214, 218–219, 223–227, shear stress, 54, 72
229–230 strength, 121–122, 125, 146, 148, 234, 256,
263–268, 288, 292
U stress, 18, 27–30, 42, 44–47, 49–52, 54–60,
U-bending benchmark, 269 62–64, 66–68, 74–77, 79, 83, 85–87,
Uniaxial 89–95, 98–99, 101, 106–107, 109–111,
coefficient of plastic anisotropy, 33–35, 38, 113–120, 124, 196, 198, 200–201, 213,
51, 56, 63, 83, 113–114, 116–120 239, 250, 257
yield stress, 27–30, 49–52, 54, 60, 62–64, stress in equibiaxial tension, 55, 74
66–68, 74–77, 79, 82–93, 85, 86–87, surface, 9–10, 40–41, 43–44, 49–50, 52, 58,
89, 92–93, 106–107, 110, 113–114, 60–61, 69–70, 72, 75, 77, 81–84, 87,
120, 200–201, 245 90–93, 106–108, 125–129, 198–199,
Updated Lagrangian (UL) formulations, 15 219, 234, 236, 239–240, 245–249
Upper specification limit, 260 Yoshida–Uemori model, 129–131
Young modulus, 131–132, 254, 268, 278
V
Variability of the material properties, 256–257 Z
Veerman method, 162 Zurich Nr.5 method, 163