Journal of Applied Mathematics and Mechanics

Zeitschrift für Angewandte Mathematik und Mechanik

Founded by Richard von Mises in 1921

Application of Continuum Damage Mechanics in discontinuous

crack formation: Forward extrusion chevron predictions
Celal Soyarslan1,3, A. Erman Tekkaya1,2 , and Ugurhan Akyuz3
1
Atılım University, Department of Manufacturing Engineering, 06836, Ankara, Turkey
2
University of Dortmund, Institute of Forming Technology and Lightweight Construction,
44227 Dortmund, Germany
3
Middle East Technical University, Department of Civil Engineering, 06531, Ankara, Turkey

Received 18 January 2008, revised 19 March 2008, accepted 26 March 2008

Published online 23 May 2008

Key words Fracture mechanisms (formation of central bursts), damage coupled elasto-plasticity, finite
strain, finite elements.
MSC (2000) 00-xx
Dedicated to Professor Peter Haupt on the occasion of his 70th birthday
Materializing Continuum Damage Mechanics (CDM), numerical modeling of discrete internal cracks,
namely central bursts, in direct forward extrusion process is presented. Accordingly, in a thermodynamically
consistent setting, a local Lemaitre variant damage model with quasi-unilateral evolution is coupled with
hyperelastic-plasticity. The formulations are constructed in the principal axes where simultaneous local inte-
gration schemes are efficiently developed. To this end, the framework is implemented as ABAQUS/VUMAT
subroutine to be used in an explicit FE solution scheme, and utilized in direct forward extrusion simulations
for bearing steel, 100Cr6. Discontinuous cracks are obtained with the element deletion procedure, where the
elements reaching the critical damage value are removed from the mesh. The periodicity of the cracks shows
well accordance with the experimental facts. The investigations reveal that, application of the quasi-unilateral
conditions together with the crack closure parameter has an indispensable effect on the damage accumulation
zones by determining their internal or superficial character.

Z. Angew. Math. Mech. 88, No. 6, 436 – 453 (2008) / DOI 10.1002/zamm.200800013
www.zamm-journal.org

NT
RI
EP
R
ZAMM · Z. Angew. Math. Mech. 88, No. 6, 436 – 453 (2008) / DOI 10.1002/zamm.200800013

Application of Continuum Damage Mechanics in discontinuous crack

formation: Forward extrusion chevron predictions
Celal Soyarslan1,3,∗ , A. Erman Tekkaya1,2 , and Ugurhan Akyuz3
1
Atılım University, Department of Manufacturing Engineering, 06836, Ankara, Turkey
2
University of Dortmund, Institute of Forming Technology and Lightweight Construction, 44227, Dortmund, Germany
3
Middle East Technical University, Department of Civil Engineering, 06531, Ankara, Turkey

Received 18 January 2008, revised 19 March 2008, accepted 26 March 2008

Published online 23 May 2008

Key words Fracture mechanisms (formation of central bursts), damage coupled elasto-plasticity, finite strain, finite
elements.
MSC (2000) 00-xx

Dedicated to Professor Peter Haupt on the occasion of his 70th birthday

Materializing Continuum Damage Mechanics (CDM), numerical modeling of discrete internal cracks, namely central
bursts, in direct forward extrusion process is presented. Accordingly, in a thermodynamically consistent setting, a local
Lemaitre variant damage model with quasi-unilateral evolution is coupled with hyperelastic-plasticity. The formulations
are constructed in the principal axes where simultaneous local integration schemes are efficiently developed. To this end,
the framework is implemented as ABAQUS/VUMAT subroutine to be used in an explicit FE solution scheme, and utilized
in direct forward extrusion simulations for bearing steel, 100Cr6. Discontinuous cracks are obtained with the element
deletion procedure, where the elements reaching the critical damage value are removed from the mesh. The periodicity
of the cracks shows well accordance with the experimental facts. The investigations reveal that, application of the quasi-
unilateral conditions together with the crack closure parameter has an indispensable effect on the damage accumulation
zones by determining their internal or superficial character.
c 2008 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim


1 Introduction
In metal forming processes, the aim is not to exhaust the formability reserve of the material. Once the reserve is exhausted,
the physical properties are degraded or even cracks occur, resulting in a defected product. The forming design aims, defining
appropriate kinematics, and predicting forces and stresses, which will not result in excessive damage accumulation. This
requires knowledge on micro-mechanism of the material weakening.
In the context of ductile damaging materials, the damage accumulation up to macro-crack occurrence requires a progres-
sive deterioration process, which shows itself in three steps, namely nucleation, growth and coalescence of micro-voids.
Nucleation of micro-voids occurs around secondary phase particles or impurities under plastic flow conditions 1, whereas
positive hydrostatic stresses cause growth of nucleated and/or already existing micro-voids to decrease material homoge-
nized stiffness and strength. Under increased loads, the enlarged micro-voids tend to coalesce to form unified macro-crack
(failure). This explains the fact that compressive stress fields promote deformation range whereas tensile triaxial stress
fields lead premature cracks, see e.g. Rice and Tracey [51] and Plancak [49]. In the literature, the summarized ductile
failure process up to rupture, has been studied by three distinct approaches, namely fracture mechanics (FM), micro-based
damage mechanics (MDM), and meso-based damage mechanics (CDM).
FM takes into account an accumulated plastic work threshold for material failure. There are various models proposed
in the literature, see e.g. [48] for Oyane criterion, [13] for Freudenthal criterion, [10] for Cockroft Latham criterion and
finally [4] for Brozzo criterion among others. Once the accumulated plastic work reaches a predefined critical value, frac-
ture occurs. Its uncoupled nature gives rise to practical implementation into existing FE software without any effort. No
additional equations to be satisfied incrementally/iteratively are introduced to the system, thus, mere burden becomes the
trivial explicit integration of the plastic work, which provides algorithmic efficiency. Besides, the non-softening material
∗ Corresponding author, E-mail: csoyarslan@atilim.edu.tr, Phone: +90 312 586 83 52, Fax: +90 312 586 80 91
1 The yield criteria may be assumed as a satisfactory criterion for micro-void nucleation.

c 2008 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim


ZAMM · Z. Angew. Math. Mech. 88, No. 6 (2008) / www.zamm-journal.org 437

behavior keeps the numerical setup well-posed. However, the same uncoupled character constitutes the disadvantage as
well. The deformation-damage uncoupling precludes mimicking the progressive structure of the deterioration. In other
words, throughout the deformation history, neither the load carrying capacity nor the stiffness of the material is lost up to
rupture, which is not realistic on physical grounds.
MDM models are derived from analysis on isolated unit cells involving idealized defects such as cracks, voids, or second
phase particles. A homogenization procedure is required to map gathered micro-mechanical behavior to continuum scale,
by which, it is possible to materialize the model in structural analysis. Gurson’s damage model [17] is a MDM model,
based on a voided rigid plastic matrix. With the proposed modified plastic potential, the homogenized behavior stands for
porous plasticity, where the physically obvious damage indicator variable is the void volume fraction (usually denoted by
f ). Tvergaard and Needleman [61]modified this model to give account for the increase in the void growth rate with void
coalescence, to create what is known today as Gurson-Tvergaard-Needleman model (GTN). Another MDM model, which
is thermodynamically consistent and being utilized largely in the literature, can be given as Rousselier model [54]. The
advantage of the MDM models is their clear micro-mechanical motivation reflecting the complete physical phenomena.
However, there are certain disadvantages noted in the literature. The determination of the material parameters are imprac-
tical 2. Their primarily hydrostatic stress dependent structure cannot predict shear-dominated failure; see e.g. Hambli [19].
Gologanu et al. [16] and more recently Nahshon and Hutchinson [47] attempt to extent Gurson’s damage model to shear-
dominated cases. In MDM models, although the shrinkage of the yield locus is reflected, which results in the decrease in
load carrying capacity with damage extent, the elastic stiffness degradation in unloading is not captured. Besides, the local
conventional setting of MDM formulations gives rise to spurious mesh dependence and nonphysical localization problems,
in numerical treatment of the post peak responses. Finally, the micro-mechanical construction of the formulation constitutes
a barrier in transferability of the model to the materials having different micro-structures.
In CDM, which is a mesoscopically phenomenological approach, irreversible micro-structural weakening is represented
by deformation coupled internal variables (i.e. damage variable). For isotropic damage, usually a scalar, D ∈ [0, 1], rep-
resents the extent of material deterioration, with a simple geometrical definition as the fraction of the damaged area to the
whole area in the plane of interest. The roots go back to early works of Kachanov [28] and Rabotnov [50], where effective
stress concept is proposed for creep rupture in metals. Lemaitre damage model [36] is one of the well-known CDM models.
Since then continuum damage theories to give account for finite strain formulations, anisotropy, thermal effects, nonlocal
formulations have been developed, see e.g. Steinmann et al. [60], Ju [26, 27], Lämmer and Tsakmakis [41], Mediavilla et
al. [45], Engelen [12] for finite strain formulations, Brünig [5], Menzel and Steinmann [46] for damage induced anisotropy,
and Brünig and Ricci [6], Voyiadjis et al. [65] for nonlocal extensions among others. The advantage of CDM is the existence
of consistent derivation through thermodynamics of irreversible processes. Besides, the models are capable of reflecting
both the shrinkage of the admissible stress space and the elastic stiffness degradation with material deterioration3. Like
in the case of MDM models, this softening behavior in turn creates an ill-posed initial boundary value problem, with the
loss of ellipticity for quasi-static cases, and hyperbolicity for dynamic problems, where spurious mesh dependence and
nonphysical localization problems are due for post peak responses.
For a detailed treatment of the phenomenological and micro-mechanical methods, the reader is referred to elsewhere,
Lemaitre [37], Krajcinovic [32], Skrzypek and Ganczarski [57], and Lemaitre and Desmorat [38].
In the present study, numerical damage analysis up to complete discrete crack propagation in the process of the direct
axi-symmetric extrusion with explicit FEM, is aimed. By definition, direct axi-symmetric extrusion, classified as a massive
metal forming process, is an area reduction method where a round billet is forced by a ram through a conical die. Two
deterioration zones, as possible crack emanation regions, are typical for an extrusion process, which are surface and center.
The cracks occurred accordingly are named as surface cracks (also named as snake skin, or fir-tree) or central bursts (also
named as chevrons), respectively. The driving parameters affecting the mechanism of failure through redistribution of the
mechanical and thermal fields are semi-cone die angle, area reduction, lubrication (friction), temperature, and mechanical
and micro-structural quality of the raw material (like grain size and morphology, non-metallic inclusion content, Ko and
Kim [29], see e.g. Lim and Dunne [39] for inclusion of particulate reinforcement). Resultant surface damage accumulations
are due to surface tearing/shearing in the die exit zones where unloading strains (elastic) form tensile residual pressures.
Central damage accumulations are due to positive central hydrostatic pressures in die reduction zones accompanied by

2 Calibration of material parameters is relatively easy for 3 parameter Rousselier damage model.
3 A noteworthy point is that, although CDM provides a strong coupling environment where the progressive deterioration of the material is resolved,
for high strength and low ductility materials (like high carbon steels), critical damage values (which denotes the local material failure) can be
considerably small (Dcr can even be in the order of 0.05, see e.g. Venkata Reddy et al. [62]. The micro-mechanical picture is in correlation to this
fact, the critical void volume fraction for element failure is taken as f = 0.05 to f = 0.2, where f = 1.0 is never practically reached, Gelin [14]).
This may lead to a FM-like uncoupled application. However, the model produced should be applicable to larger extent of materials, together with
less mathematical restrictions, and provide soundness on physical grounds. Thus, in this study a fully coupled progressive deterioration mechanism
is presented in chevron prediction.

www.zamm-journal.org c 2008 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim


438 C. Soyarslan et al.: CDM applications for discontinuous crack formation

plastic flow. This internal cracks result in a drastic reduction in the service capacity of the product. Investigation of these
hidden cracks, to avoid the faulty product transfer to market, requires nondestructive ultrasonic testing on every one of
products, Zimerman et al. [64], which is neither practical nor cheap. Thus, proper investigation is required on the governing
mechanisms of damage accumulation in extrusion.
The defect formations in extrusion processes have been of interest since early experimentally works of Jennison [25].
Analytical works of Avitzur [2] materialized upper bound theorem to separate safe and unsafe regions on the plane of semi-
cone die angle and area reduction ratio. Zimerman and Avitzur [63] expanded this work to give account for hardening.
Without specific reference to the material Sriram and Van Tyne [59] created a similar diagram for spherical dies making
use of numerical analysis (see also e.g. Lim and Dunne [39] in which the Zimerman et al. [64] diagram is reproduced for
reinforced metal matrix composites where strain hardening and friction is neglected, or the envelope curves plotted in Ko
and Kim [29] utilizing the Cockroft Latham model which is an FM model). According to these studies the favorable central
bursting conditions are the ones with low area reductions and high semi-cone angles, see e.g. Ko and Kim [29], Aravas
[1], Zimerman et al. [64] among others. Usually, increasing area reduction together with friction has an effect of damage
transfer from interior to the billet face, motivating superficially dominated deterioration and surface cracks Lim and Dunne
[39].
For a practising engineer such forming limit diagrams are quite useful in avoiding defected products. In this context,
the macro-crack emanation and its propagation may seem to be of second interest compared to investigating damage
accumulation mechanisms, to define barriers between safe and unsafe, Ko and Kim [29]. However, the resolution and the
control of crack mechanisms gain importance, when expanding the application area to include deteriorative processes such
as machining, orthogonal cutting or blanking, Ceretti et al. [8]. In the literature, considerable labor is devised for numerical
prediction of damage in forward extrusion. Table 1 summarizes a number of up to date chronologically ordered numerical
damage analysis of extrusion processes. In Table 1, the listed references are evaluated according to the investigated process
(extrusion or drawing4, abbreviated as ext. or drw.), damage model utilized, inclusion of the thermal and friction effects
in the analysis, and finally macro-crack investigation and the numerical method used to carry out cracks. These methods
include element deletion (Elem. Del.), element degradation (Elem. Deg.) and node separation (Node Sep.), each of which
have relative advantages and disadvantages.
In presenting the periodic central bursts, the present work utilizes a Lemaitre variant local isotropic damage strongly
coupled to a physically sound finite strain hyperelastic-plastic framework, through a single J2 yield potential. A principal
axes formulation, which provides convenience in active-passive damage evolutionary forms, besides in formulation of finite
plasticity5, is constructed in an Euclidean setting. It is shown that the active-passive evolutionary conditions controlled by
the crack closure parameter have an important effect on the damage accumulation zones. In the case of complete crack
closure, the internal character is amplified rather than the superficial one.
An explicit FE solution scheme is conducted, where crack propagation and contact conditions are efficiently resolved.
Following Lemaitre [36], strain equivalence principle and effective stress concept, Kachanov [28] and Rabotnov [50], are
selected as the fundamental phenomenological hypotheses. The discrete crack formations are carried out applying element
deletion procedure, where the elements whose Gauss points have reached the critical damage threshold are eliminated from
the analysis. This procedure is supplied in many widely used commercial packages (see e.g. ABAQUS/Explicit, MSC Marc,
and Deform2D among others).
The outline of the rest of this paper is as follows. In Sect. 2, governing equations are derived in a thermodynamically
consistent setting in principal stress space. Numerical treatment of the algorithmic forms in the explicit finite element
solution environment is given in Sect. 3. A detailed analysis of the discrete chevron predictions under the effect of certain
process and material parameters is given in Sect. 4.
Throughout the paper, following notations will be used. Assuming a and b as two second order tensors, together with the
Einstein’s summation convention on repeated indices, c = a • b represents the product with cik = aij bjk where c comes
out to be a second order tensor. d = a : b represents the inner product with d = aij bij where d is a scalar. E = a ⊗ b,
F = a ⊕ b, and G = a ⊖ b represent tensor products with Eijkl = aij bkl , Fijkl = aik bjl , and Gijkl = ail bjk , respectively,
where E, F , and G are fourth order tensors. [•]T and [•]−1 denote the transpose and the inverse of [•] respectively. [•]sym
and [•]skw are associated with the symmetric and skew-symmetric parts of [•], respectively, with [•] = [•]sym + [•]skw .

4
Drawing is very similar to extrusion, where the deriving force is applied to the front tip of the billet creating a tension dominated zone, unlike
extrusion, where the force is applied from the rear tip, to result in mainly a compression dominated zone. This gives rise to higher possibility of
central crack formation, increasing the extent of material regions subjected to positive hydrostatic stresses.
5 For a detailed evaluation of the principal stress space formulation of finite plasticity, the reader is referred to Ibrahimbegović [24].

c 2008 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

 www.zamm-journal.org
ZAMM · Z. Angew. Math. Mech. 88, No. 6 (2008) / www.zamm-journal.org 439

Table 1 Numerical damage investigation for direct extrusion/drawing.

Macro-
ID Reference Process Model Temperature Friction Crack / Method

1 Aravas [1] Ext. MDM − − − / −

2 Liu and Chung [40] Ext. FM −  − / −
3 Giardini et al. [15] Ext. FM −  − / −
4 Ceretti et al. [8] Ext. FM    / Elem. Del.
5 Lim and Dunne [39] Ext. MDM − − − / −
6 Lee and Hahm [35] Ext./Drw. MDM −  − / −
7 Gelin [14] Ext. MDM − − − / −
8 Choi et al. [9] Ext. FM − − − / −
9 Srikanth and Zabaras [58] Drw. MDM   − / −
10 Komori [30] Drw. MDM −   / Node Sep.
11 Ko and Kim [29] Ext./Drw. FM −  − / −
12 Hambli and Badie-Levet [18] Ext. CDM −  − / −
13 Reddy et al. [62] Drw. FM/CDM −  − / −
14 Komori [31] Drw. FM/MDM −   / Node Sep.
15 Saanouni et al. [55]† Ext. CDM −   / Elem. Del.
16 McAllen and Phelan [43] Drw. MDM+FM −   / Elem. Deg.
17 Chamacho et al. [7] Drw. FM − − − / −
18 McVeigh and Liu [44] Ext. MDM    / Elem. Del.
19 McAllen and Phelan [42] Drw. MDM+FM −   / Elem. Deg.
†
In Saanouni et al. [55], although a thermo-mechanical framework is presented, the extrusion process is simulated for
isothermal conditions, thus, the thermal effects on the chevron predictions are not investigated.

2 Theory
2.1 Thermodynamic framework for the damage coupled finite plasticity
Let B0 ⊂ R3 and B ⊂ R3 denote the reference and deformed configurations. For any material point X at the reference
configuration, the respective position at the deformed configuration, i.e. x, is supplied by the map ϕ(X, t) : B0 × T → B,
where T denotes the time interval. The deformation gradient, F , is then given by, F := ∂X ϕ. According to the kinematics
of Lee [34], a local multiplicative factorization, F = F e • F p , can be postulated. Consequently, be = F e • F e, T and
ǫe = 1/2 log [be ] become the elastic left Cauchy Green deformation tensor and the elastic logarithmic strain tensor, which
constitute the base deformation measures used in the current multiplicative framework. The spectral representations of
these strain measures read
3
 3

e
b = beA A A
ν ⊗ν , ǫ =e
ǫeA ν A ⊗ ν A , (1)
A=1 A=1

where beA and ǫeA are the eigenvalues of respective tensors and ν A denote the corresponding eigenvectors with A = 1, 2, 3.
The principal elastic logarithmic strains in terms of the elastic principal stretches, λeA , read, ǫeA = log[λeA ], with beA = λe,
A .
2

For details of the underlying nonlinear continuum mechanics in the context of finite deformations, the reader is referred to
Haupt [20].
To set the stage, in formulation of the damage coupled multiplicative hyperelastic plasticity, we start by defining a scalar
D ∈ [0, 1], as damage internal variable, which denotes the ratio of damaged area to the total area at the plane of interest,
Kachanov [28]. At the two extremes, D → 0 represents the undamaged material whereas D → 1 stands for the complete
rupture. Following the thermodynamics of internal variables, the Helmholtz free energy, Ψ, can be given in an additively
decoupled damage deteriorated elastic portion, Ψe = (1 − D) Ψ̃e , and the plastic portion, Ψp ,

Ψ(be , α, D) = (1 − D) Ψ̃e (be ) + Ψp (α) , (2)

www.zamm-journal.org c 2008 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim


440 C. Soyarslan et al.: CDM applications for discontinuous crack formation

which constructs the state coupling between elasticity and damage. Ψ̃e (be ) represents damage unaffected free elastic en-
ergy which is assumed to be quadratic in elastic logarithmic strains and has the following decoupled representation into
deviatoric and volumetric parts,
1
Ψ̃e := Ψ̃e,vol + Ψ̃e,dev , where Ψ̃e,vol = H [ǫe1 + ǫe2 + ǫe3 ]2 , Ψ̃e,dev = µ [ǭe,2 e,2 e,2
1 + ǭ2 + ǭ3 ] , (3)
2
with H and µ representing the bulk modulus and the shear modulus, respectively. ǭeA are associated with the principal com-
ponents of the deviatoric elastic logarithmic strain tensor. Ψp (α) denotes the plastic isotropic hardening potential with the
argument α giving account for the equivalent plastic strain, i.e. isotropic hardening strain like internal variable. Following
the 2nd principle of thermodynamics, assuming isothermal conditions, non-negativeness of the mechanical dissipation, Ω,
is postulated as follows,
0 ≤ Ω = τ : d − ∂t Ψ , (4)
where, verbally, Ω is associated with the difference between the stress power and the rate of the free energy. Using the chain
rule of differentiation one carries out,
1
0 ≤ Ω = [τ − 2∂be Ψ • be ] : d − 2∂be Ψ • be : [ Lv be • be,−1 ] + [−∂α Ψ] ∂t α + [−∂D Ψ] ∂t D , (5)
   2
=0

where the objective Lie derivative of be is denoted by Lv be , which, for this specific case, has the following Oldroyd rate
form,
Lv be := ∂t be − l • be − be • lT . (6)
l denotes the spatial velocity gradient with l := ∂t F • F −1 . After proper derivations, one obtains the following state laws,
τ = 2 ∂be Ψ • be = 2 (1 − D) ∂be Ψ̃e • be , q = −∂α Ψ = −∂α Ψp , Y = −∂D Ψ = Ψ̃e , (7)
where τ is the homogenized Kirchhoff stress tensor with its effective counterpart as, τ̃ = τ /(1 − D). The isotropic
hardening stress like internal variable, q, stands for the yield locus expansion. Y represents the damage conjugate variable,
which stands for the total elastic strain energy release rate. With the proposed additively decoupled elastic strain energy
definitions, the explicit representation for the Kirchhoff stress tensor may be given as, in a spectral form,
3

τ = τA ν A ⊗ ν A , τA = sA + p, with sA = (1 − D) 2µ ǭAe , p = (1 − D) H [ǫe1 + ǫe2 + ǫe3 ] , (8)
A=1

with τA , sA , and p denoting the homogenized principal total Kirchhoff stress, the homogenized principal deviatoric Kirch-
hoff stress, and the homogenized Kirchhoff type hydrostatic stress, respectively. Due to isotropy of the material, which
preserves validity with the isotropy of damage, the Kirchhoff stresses and the elastic logarithmic strains are coaxial. For the
isotropic hardening plasticity, a saturation form is adapted in the current setting, with q = K ′ (α),
K ′ (α) = K α + (τ∞ − τ0 ) (1 − exp [−δ α]) , (9)
where K is the linear hardening parameter. τ∞ and τ0 stand for the level off stress and the virgin yield stress, respectively.
δ denotes the saturation parameter in the form of a hardening exponent. Making necessary substitutions, the dissipation Ω
turns into
1
0 ≤ Ω = τ : [− Lv be • be,−1 ] + q ∂t α + Y ∂t D . (10)
2
Following the Hill-Mandel maximization of the inelastic dissipation postulate, Hill [21], utilizing similar steps with Le-
maitre [37], we propose a total dissipation potential decoupled into plastic and damage dissipation potentials, which reads,
φt (τ , q, Y ; D) = φ̃(τ̃ , q) + φd (Y ; D) . (11)
The problem of maximization of the dissipation, Ω, which is subjected to the given constraint, φt , is transformedinto a
mathematical unconstrained minimization problem by postulating a Lagrange functional, FL , as FL := −Ω + γ̇ φt , with
γ̇ being the Lagrange multiplier. Accordingly, with the method of Lagrange multipliers, the stationarity conditions for FL ,
such as ∂τ FL = 0, ∂q FL = 0, and ∂Y FL = 0, give the following evolutionary equations, in the form of plastic flow rule
and the rates of α and D, respectively,
γ̇
Lv be = −2 ∂τ̃ φ̃ • be , ∂t α = γ̇ ∂q φ̃, ∂t D = γ̇ ∂Y φd . (12)
(1 − D)

c 2008 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

 www.zamm-journal.org
ZAMM · Z. Angew. Math. Mech. 88, No. 6 (2008) / www.zamm-journal.org 441

γ̇ is also called the plastic multiplier. Through a singular plastic multiplier, the kinematic coupling between plasticity and
damage is constructed. The plastic dissipation potential is specified as a J2 type yield potential, defined in the principal
effective stress space,

2  2 2 2
 21 2
φ̃(τ̃ , q) := τ̃ + τ̃2 + τ̃3 − τ̃1 τ̃2 − τ̃1 τ̃3 − τ̃2 τ̃3 − y(q) ≤ 0 , (13)
3 1 3
where y(q) represents the hardening function with y(q) = [τ0 + q]. The damage dissipation potential is selected to be
quadratic in Y , in its simplest possible form,
a
φd (Y ; D) := Y2, (14)
2(1 − D)
where a is damage related material parameter. With this proposed set of potentials, the evolutionary forms given in Eq. (12)
read, using ν A ⊗ ν A ≡ ν̃ A ⊗ ν̃ A , i.e. the coaxiality of the homogenized and the effective Kirchhoff stresses, respectively,

3 
e γ̇  s̃A
A A e 2 a
Lv b = −2 ν ⊗ν • b , ∂t α = γ̇ , ∂t D = γ̇ Y. (15)
(1 − D)
s̃
3 (1 − D)
A=1

2.2 Quasi-unilateral damage evolution

The selected damage rate form is assumed to be linear in the the damage conjugate variable Y . In the principal Kirchhoff
stress space, following representation can be derived,
1+ν  2  9ν 2
Y (τ̃A ) = τ̃1 + τ̃22 + τ̃32 − p̃ , (16)
2E 2E
with p̃ = (τ̃1 + τ̃2 + τ̃3 )/3. This representation does not distinguish between the tensile or compressive characters of the
stress terms processed. However, micro-mechanical facts reveal the acceleration of void growth in the tensile hydrostatic
stress zones, whereas compressive stress fields delay crack occurrence. Thanks to principal axes formulation, Eq. (16)
can be refined to include the so-called micro-crack closure effect without additional labor. Accordingly, assuming a full
contribution of tensile (positive) principal stresses and partial contribution of compressive (negative) principal stresses,
where the extent of this contribution is controlled by the crack closure parameter, h ∈ [0, 1], in the total elastic strain energy
release rate expression, quasi-unilateral damage evolution can be mimicked. The refined damage conjugate variable, i.e.
Y + , then reads,
1+ν   9ν
Y + (τ̃A ) = τ̃1 2 + τ̃2 2 + τ̃3 2 − p̃2
2E 2E
h (1 + ν)   9h ν
+ −τ̃1 2 + −τ̃2 2 + −τ̃3 2 − −p̃2 , (17)
2E 2E
with · denoting the Macauley bracket where x := 1/2 (x + |x|). For h → 1, the crack closure effect and the distinction
between Eq. (17) and Eq. (16) is neglected, whereas h → 0 gives account for a complete crack closure and zero damage
evolution under compressive stress states.
The current formulation exploits the ideas on the crack closure and the effective stress, Ladeveze and Lemaitre [33], and
reflects them in a simplified manner to the damage evolutionary equations. Application of the quasi-unilateral conditions
with anisotropic damage is prone to certain anomalies, which is beyond the purpose of the current study. For a general
perspective, the reader is referred to Chaboche [11], Lemaitre [37] and the references therein.

3 Numerical implementation
3.1 Explicit FE formulation of the coupled IBVP
The generalized momentum balance in a Eulerian description reads,

div σ + ζ = ργ , (18)

where div[•] stands for the spatial divergence operator. σ denotes the Cauchy stress tensor defined at current configuration.
ρ is the density, whereas ζ and γ are the body forces and accelerations, respectively. Eq. (18) constitutes the so called
strong form. Multiplying both sides by a sufficiently continuous (i.e. smooth) virtual velocity field, η, integrating at the

www.zamm-journal.org c 2008 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim


442 C. Soyarslan et al.: CDM applications for discontinuous crack formation

current configuration and applying the divergence theorem, one carries out the following weak statement of the (global)
equilibrium of the body
   
g(ϕ, η) = σ : grad η dv − ζ • η dv − t̄ • η da + ρ γ • η dv = 0 , (19)
B B ∂Bσ B

where grad[•] denotes the spatial gradient operator. η satisfies the condition, η = 0, at ∂Bu , where ∂Bu ⊂ ∂B denotes
the part of the boundary, ∂B, on which the Dirichlet boundary conditions are specified with the prescribed displacements,
ū. Neumann type boundary conditions act on ∂Bσ ⊂ ∂B with the tractions, t̄ = σ• n, where n denotes the spatial
normal vector. The boundary parts satisfy the conditions, ∂Bu ∂Bσ = ∂B and ∂Bu ∂Bσ = Ø. In a compact setting,
Eq. (19) stands for the principle of virtual power,

δP = δP int − δP ext + δP kin = 0 , (20)

where δP int , δP ext , and δP kin denotes virtual internal, virtual external, and virtual kinetic (inertial) power, respectively,

int sym
δP = σ : [grad η] dv , (21)
B
 
δP ext = ζ • η dv + t̄ • η da , (22)
B ∂Bσ

kin
δP = ρ γ • η dv , (23)
B

skw
with σ : [grad η] = 0, due to symmetry of σ. Application of the finite element discretization,
with the subdivision of
the domain, i.e. B, into non-overlapping subdomains (elements), i.e. Be , where B = ne el Be , and representing the vector
of approximation functions with N e , and discrete gradient operator (strain displacement matrix) with B e , one obtains the
following semi-discrete momentum equation,

Fint − Fext + M • ü = 0 , (24)

in which Fint , Fext , and M denote the internal loads, externally applied loads, and the mass matrix, respectively. ü stands
for the acceleration vector. Explicit definitions for Fint and Fext can be given as follows,
nel  
Fint = A
e=1 Be
B e, T • σ e dv e , (25)

nel

  
F ext
= A
e=1 Be
N e, T e
• ζ dv +
∂Beσ
N e, T
• t̄ da e
, (26)

where A is the assembly operator, Hughes [23]. Finally M reads,

nel  
M= A
e=1 Be
ρ N e, T • N e dv e . (27)

Solution of the resultant problem is possible through implicit or explicit procedures. Making use of an explicit FE procedure,
serves advantages for the problems involving severe discontinuities, where an implicit solution scheme may fail to converge.
Besides, the resolution of the complicated contact conditions are more efficiently handled compared to implicit procedures.
The memory saving behavior for very large problems and amenability to parallelization are other apparent advantages of
the explicit codes. Moreover, since the momentum iterations are irrelevant, the procedure does not require computation of
the material tangent matrix. This is advantageous for especially the mathematically complicated material models, where a
closed form expression for the consistent material moduli is not available. Although there exists a minimum for the time
step which preserves conditional stability, Hughes [22], in the context of slow (quasi-static) processes, where the inertia
effects can be discarded, increasing the density (mass scaling), or the loading rate, artificially, may help expanding this
limitation, thus, accelerating the process solution. For explicit FE applications in bulk metal forming, reader may refer to
Rojek et al. [52, 53].

c 2008 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

 www.zamm-journal.org
ZAMM · Z. Angew. Math. Mech. 88, No. 6 (2008) / www.zamm-journal.org 443

3.2 Algorithmic treatment of the time discrete forms

For the temporal discretization, we follow a notation where [•]n−1/2 , [•]n , [•]n+1/2 , and [•]n+1 give the definitions of any
variable denoted by [•], at times tn−1/2 , tn , tn+1/2 , and tn+1 , respectively. Using an explicit time integration scheme, the
kinematics can be integrated with central differences as

un+1 = u̇n + ∆tn+1 u̇n+1/2 , (28)

where
∆tn+1 + ∆tn
u̇n+1/2 = u̇n−1/2 + ün , (29)
2
 
ext int
ün = M−1n • Fn − Fn , (30)

with u and u̇ representing the displacement and the velocity vectors, respectively. The algorithmic setting requires the
determination of the current state at time tn+1 , from a presumably known state at time tn . As a first step, in a strain driven
procedure, un+1 , i.e. the deformation at ∆t = tn+1 − tn , is integrated using Eq. (28), where the relative deformation
gradient, f n+1 , can be computed using f n+1 = Fn+1 • Fn−1 . Assumption of a fully elastic step (elastic prediction) will
yield,
3

b e, tri e T
n+1 = f n+1 • b n • f n+1 , b e, tri
n+1 = (λe, tri 2 tri, A tri, A
n+1, A ) νn+1 ⊗ νn+1 . (31)
A=1

In the current damage coupled setting, according to the strain equivalence principle, the utilized effective trial elastic
Kirchhoff stresses are independent of damage. BOX 1 involves the formal steps of elastic prediction.

BOX 1. Elastic predictor

1. Database at x ∈ B,
ben , αn , Dn .
2. For a given Fn+1 , compute the relative deformation gradient,
fn+1 = Fn+1 • Fn−1 .
3. Compute the trial elastic left Cauchy Green deformation tensor,

b e, tri e T
n+1 = f n+1 • b n • f n+1 .

4. Apply the spectral decomposition,

3

b e, tri
n+1 = b e, tri tri, A tri, A e, tri e, tri
n+1, A νn+1 ⊗ νn+1 , with λ n+1, A = (b n+1, A )
1/2
.
A=1

5. Compute the trial elastic logarithmic strains,

 
ǫ e, tri
n+1, A = log λ e, tri
n+1, A .

6. Compute the effective principal Kirchhoff stress predictions,

tri tri
τ̃ n+1, A = p̃n+1 + s̃ n+1, A ,

p̃n+1 = H log [Jn+1 ] ,

e, tri
s̃ tri
n+1, A = 2µ ǭ n+1, A .

The flow rule given in Eq. (15.a), is integrated for a time step ∆t with ∆γ := ∆t γ̇ n+1 , using an exponential mapping
approximation,

3

∆γ  s̃ n+1, A
b en+1 = exp −2 ν A ⊗ νn+1
A
• b e, tri
n+1 . (32)
(1 − Dn+1 )
s̃n+1
n+1
A=1

www.zamm-journal.org c 2008 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim


444 C. Soyarslan et al.: CDM applications for discontinuous crack formation

tri, A tri, A
The coaxiality of b en+1 and τ̃n+1 due to elastic isotropy, together with Eq. (32) reveals, νn+1
A A
⊗ νn+1 ≡ νn+1 ⊗ νn+1 .
Then it is possible to reexpress Eq. (32) as
 
∆γ s̃n+1, A
e
λ n+1, A = exp − λ e, tri
n+1, A . (33)
(1 − Dn+1 )
s̃n+1

A handy plastic/damage correction in strain space is carried out by taking the natural logarithms of both sides,
∆γ
ǫ en+1, A = ǫ e, tri
n+1, A − n n+1, A , (34)
(1 − Dn+1 )
with the substitution, n n+1, A := s̃n+1, A /
s̃n+1
. Once the corrections in the effective stress space are concerned, just
like in the case of conventional J2 plasticity, the plastic/damage stress correction will only act on the effective deviatoric
Kirchhoff stresses, whereas the effective trial Kirchhoff type pressure will stay unchanged. Its resulting homogenized value
however will experience a deterioration.
∆γ e, tri
s̃ n+1, A = s̃ tri
n+1, A − 2µ n n+1, A , with s̃ tri
n+1, A = 2µ ǭ n+1, A ,
tri
τ̃ n+1, tri
A = p̃ + s̃ n+1, A . (35)
(1 − Dn+1 )
In the context of finite deformations, time integration of the scalar internal variables is straightforward compared to tensorial
ones. An implicit backward Euler method will be used for damage and equivalent plastic strain integrations,

2 a
αn+1 = αn + ∆γ , Dn+1 = Dn + ∆γ Yn+1 , (36)
3 (1 − Dn+1 )
which finalizes the local integration equations of the strain like internal variables. At the current stage, the coupled nonlinear
equation set can be cast in a residual setting, with rn+1 = rn+1 (xn+1 ) = 0, as follows,
⎧ ⎫
tri
⎪
⎪ τ n+1, 1 − τ n+1, 1 + 2µ∆γ n n+1,1 ⎪
⎪ ⎧ ⎫
⎪
⎪ tri
⎪
⎪ ⎪ τn+1, 1 ⎪
⎪ τn+1, 2 − τn+1, 2 + 2µ∆γ nn+1,2
⎪
⎪ ⎪
⎪ ⎪
⎪ ⎪
⎪
⎪ ⎪ ⎪
⎪
⎪ tri
⎪
⎪ ⎪
⎪ τn+1, 2
⎪
⎪
⎪ τn+1, 3 − τn+1,
⎪
⎪ 3 + 2µ∆γ nn+1,3
⎪
⎪
⎪
⎪
⎪
⎪
⎪
⎪
⎪
⎪
⎨ 2
⎪
⎬ ⎪
⎨ τn+1, 3 ⎪
⎬
rn+1 = α n+1 − αn − 3 ∆γ ; xn+1 = ∆γ . (37)
⎪ ′ ⎪ ⎪ ⎪
⎪
⎪
⎪ q n+1 − K (α n+1 ) 
⎪
⎪
⎪
⎪
⎪ D
⎪ n+1 ⎪
⎪
⎪
⎪
⎪
⎪ ⎪
⎪ ⎪
⎪ ⎪
⎪ φ̃n+1 :=
s̃ n+1
− 23 y (qn+1 )
⎪
⎪
⎪
⎪
⎪ ⎪ αn+1 ⎪
⎪
⎪ ⎪
⎪
⎪
⎪ a ⎪
⎪ ⎪
⎩ ⎪
⎭
⎪
⎩ Dn+1 − Dn − ∆γ
⎪ Yn+1 ⎭
⎪
⎪ q n+1
(1 − Dn+1 )
For the local problem, Yn+1 = Yn+1 (αn+1 , Dn+1 ). Accordingly, 7 equations in Eq. (37), will be solved for 7 unknowns,
xn+1 . Following similar arguments as Simó and Taylor [56], this local equation set can be substantially reduced to two
coupled equations, which read,
⎧  ⎫
  ∆γ  
⎨ 23 {τ0 + K ′ (αn + 23 ∆γ )} − s̃ tri
⎪
n+1 + 2µ
 ⎪
⎬
(1 − D n+1 ) ∆γ
rn+1 = a ; xn+1 = , (38)
⎩ Dn+1 − Dn − ∆γ
⎪ Yn+1 ⎪
⎭ Dn+1
(1 − Dn+1 )
with rn+1 (∆γ , Dn+1 ) = 0. In Eq. (38) the hardening function is defined in terms of the plastic multiplier. Application of
the Newton-Raphson solution scheme with the linearized form will yield, drn+1 = Drn+1 • dxn+1 , where the Jacobian,
Drn+1 , of the system is given as,
⎡ ⎤
∂r1, n+1 ∂r1, n+1
⎢ ∂∆γ ∂D ⎥
Drn+1 = ⎢ n+1 ⎥
⎣ ∂r2, n+1 ∂r2, n+1 ⎦ . (39)
∂∆γ ∂Dn+1
−1
Derivations for the components of the local Jacobian, i.e. Drn+1 , are listed in Appendix A. Using Drn+1 • drn+1 =
dxn+1 , the solution for xn+1 , which constitutes the local return mapping realized at each Gauss point, can be carried out
with the following iterative scheme,
(k+1) (k) −1, (k) (k)
xn+1 = xn+1 − δ (k) Drn+1 • rn+1 , (40)

c 2008 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

 www.zamm-journal.org
ZAMM · Z. Angew. Math. Mech. 88, No. 6 (2008) / www.zamm-journal.org 445

where δ (k) ∈ (0, 1] is the proper line-search parameter. BOX 2 summarizes the numerical implementation of the local
return map algorithm for a damage coupled finite isotropic hyperelastic plasticity, which concludes this section.

BOX 2. Return-mapping algorithm with two-step operator-split

(Elastic predictor-plastic/damage corrector type algorithm)
1. Process BOX 1. (Elastic predictor),
2. Check for plastic-damage loading on effective Kirchhoff stress space,
 1/2 2
tri tri 2 tri 2 tri 2
φ̃n+1 = (s̃n+1, 1 ) + (s̃n+1, 2 ) + (s̃n+1, 3 ) − [τ0 + K(αn )] .
3
IF φ̃tri
n+1 ≤ 0 THEN
Set (•)n+1 = (•)tri
n+1 , EXIT. (Elastic step)
ELSE
GOTO 3. (Plastic-damage corrector)
ENDIF
(0)
3. Find the plastic multiplier and the damage, ∆γ (0) = 0, Dn+1 = Dn , with line-search,
⎡ $ %(k) $ %(k) ⎤
    ∂r2, n+1 ∂r1, n+1  
− (k)
∆γ (k+1) ∆γ (k) δ (k) ⎢⎢ ∂Dn+1 ∂Dn+1
⎥
⎥ r1, n+1
(k+1) = (k) − (k) ⎢ $ %(k) $ %(k) ⎥ • (k) ,
Dn+1 Dn+1 ∆n+1 ⎣ ∂r2, n+1 ∂r1, n+1 ⎦ r2, n+1
−
∂∆γ ∂∆γ
(k) (k)
with ∆n+1 = det(Drn+1 ) and δ (k) ∈ (0, 1] which denotes the proper line-search parameter.
4. Return map with the plastic/damage correction,

∆γ s̃ tri
s̃ n+1, A = s̃ tri
n+1, A − 2µ
 n+1, A ,
(1 − Dn+1 ) s̃ tri
n+1


2
αn+1 = αn + ∆γ ,
3
a
Dn+1 = Dn + ∆γ Yn+1 .
(1 − Dn+1 )
5. Find the homogenized total Kirchhoff stresses,
τ n+1, A = (1 − Dn+1 ) (s̃ n+1, A + p̃n+1 ) .
6. Update the intermediate configuration,
  3
2 s̃ n+1, A  tri, A tri, A
b en+1, A = J 3 exp , with b en+1 = b en+1, A νn+1 ⊗ νn+1 .
µ
A=1

4 Application problems: Chevron predictions

The given framework is implemented as a user defined material subroutine, VUMAT, for ABAQUS/EXPLICIT. The code
is materialized in the simulation of a set of damage accumulation studies in axi-symmetric forward extrusion, for a single
pass reduction of the billet. A large strain explicit solution procedure is followed. CAX4R elements, which stand for 4-
node bilinear axi-symmetric quadrilateral elements with reduced integration are utilized together with a combined stiffness-
viscous hourglass control (where the weight factor is selected as 0.5). The friction at the interfaces is modeled with Coulomb
friction with µf denoting the coefficient of friction. A double precision computation is sought, with second order accuracy
and distortion control. The conducted simulations aim to investigate the effects of certain phenomena such as friction and
crack closure parameter on damage accumulation zones, and frequency and morphology of the obtained discrete cracks.
The problem geometry and the boundary conditions are given in Fig. 1, where a mesh size of 0.2 mm is utilized in the
simulations. A velocity controlled loading is applied.

www.zamm-journal.org c 2008 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim


446 C. Soyarslan et al.: CDM applications for discontinuous crack formation

Table 2 Material parameters for

100Cr6.
Parameter Symbol Unit Magnitude
Bulk modulus H GPa 175.0
Shear modulus µ GPa 80.7692
Linear hardening K MPa 115.1
Saturation stress τ∞ MPa 987.9
Flow stress τ0 MPa 752.0
Fig. 1 Problem dimensions, mesh and the boundary con- Saturation parameter δ - 14.3
ditions of the axi-symmetric extrusion problem, mesh size Damage parameter a - 0.3
= 0.2 mm.

Fig. 2 (online colour at: www.zamm-journal.org) Flow curve

for 100Cr6. Fig. 3 Damage-equivalent plastic strain curve.

Bearing steel 100Cr6 is selected as a material, for which the experimental chevron outputs are available. For the material
plasticity parameters, the outcomes of the standard upsetting experiments are utilized, as given in Table 2. Neglecting
damage accumulation with void nucleation as a consequence of shear decohesion under compressive stresses with the
presumption of perfect frictionless upsetting conditions, damage is assumed not to affect the parameter identification for
plasticity, with the aid of complete crack closure effect. Coming to damage, for the simplest possible damage rate form, the
single parameter, a, is selected as, a = 0.3, as also given in Table 2. Damage coupled and uncoupled flow curves are given in
Fig. 2 for the selected material parameters for the simple plane strain loading of a unit square shown on the figure, utilizing
a single finite element having a single Gauss point. Fig. 3 shows the concave-up nature of the damage-equivalent plastic
strain curve for the whole range of damage accumulation. Admittedly, at this stage of preliminary studies, in the selection
of the damage parameter a, the obtained cracks patterns are utilized assuming complete crack closure effect with h = 0,
where the cracks are supplied with removal of the elements with integration points, satisfying D ≥ Dcr . Without detailed
reference to the underlying microstructure, Dcr = 0.27 is selected, taking into account the listed critical damage values
of certain materials, Lemaitre and Desmorat [38]. Proper selection of h, a, and Dcr requires further experimental studies.
Besides, more complicated damage evolutionary forms, which are nonlinear in Y , may be adapted into the developed
framework easily.
Figs. 4 and 5 belong to the frictionless extrusion simulations where the critical damage value is relaxed to Dcr = 1.0
which is never actually reached. Neglecting friction at this stage, aims, isolating the system from additional effects of
external agencies in addition to precluding any additional punch force demands. Accordingly selecting the crack closure
parameter, h, as h = 0, the damage deriving stress mechanisms are investigated for complete crack closure. Fig. 4 shows the
positive portions of the principal stresses which contribute to the evolution of damage. The central damage accumulation
in Fig. 5 reveals that, although at the die exit surface involves considerable amount of tensile residual stresses, damage
is centrally located where equivalent plastic strain rate at the reduction zone meets with positive tensile principal stress

c 2008 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

 www.zamm-journal.org
ZAMM · Z. Angew. Math. Mech. 88, No. 6 (2008) / www.zamm-journal.org 447

Fig. 5 (online colour at: www.zamm-journal.org)

Damage and equivalent plastic strain accumulation.
Final step of simulations without crack formation,
Fig. 4 (online colour at: www.zamm-journal.org) From left to
mesh size = 0.2 mm, µf = 0.00.
right, tensile portions of max, mid and min principal and hydro-
static stresses. An intermediate step of simulations without crack
formation, mesh size = 0.2 mm, µf = 0.00.

Fig. 6 (online colour at: www.zamm-journal.org) Damage ac- Fig. 7 (online colour at: www.zamm-journal.org) Equivalent
cumulation that generate cracks throughout the process history, plastic strain accumulation throughout the process history, mesh
mesh size = 0.2 mm, µf = 0.00. size = 0.2 mm, µf = 0.00.

components. In the current geometry, the maximum principal stress seems to be the mere damage source. The hydrostatic
stress field plots also show the central location of the tensile triaxial stress field.
The given damage accumulation is utilized to produce cracks by simply reducing Dcr , from 1.0 to 0.27, and applying
element deletion for the elements where the damage accumulation on the Gauss point comes out to be higher then the
threshold. Since the reduced integration CAX4R elements are utilized, an element has a single Gauss point to monitor and
track the state variables. Removal of a finite element from the existing mesh will affect the mass matrix and the internal
force vector with elimination of the regarding Gauss point from the computational stack. Accordingly, the mass and the
internal force contributions of the deleted element will not be assembled. Adaptive element sizing gains importance at this
point for a smooth solution, where the destabilizing effects of the element removable are scaled by the element size. The
obtained cracks, together with the damage and the equivalent strain fields are listed in Figs. 6 and 7, throughout the process
history. As seen in Fig. 6, after the occurrence of any free surface by production of a crack, the accumulation of the central
damage to produce another crack requires a certain period. This time to accumulation, creates the periodicity of the discrete

www.zamm-journal.org c 2008 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim


448 C. Soyarslan et al.: CDM applications for discontinuous crack formation

Fig. 9 Extent of deleted elements, shown on the undeformed

Fig. 8 (online colour at: www.zamm-journal.org) Punch force mesh, mesh size = 0.2 mm, µ = 0.00.
as a function of (normalized) punch displacement, mesh size =
0.2 mm, µ = 0.00.

crack formation. Fig. 7 shows that, the distribution of the equivalent plastic strains is affected by the central bursts. The
crack tip gives account for considerable plastic strain accumulation.
A comparison between the simulations with crack formations and the ones without crack formations is quite intuitive.
Fig. 8 shows the punch displacement versus punch force demand curves for crack producing and non-producing simulations.
Comparison of the plotted curve trends reveals that the frequency of the oscillations of the force-displacement curve of the
crack involving simulation is in accordance with the crack periodicity. The emanation of the force reduction after every
oscillatory peak is coincident with the discrete central macro-crack nucleation. The force-displacement curve of the other
simulation, on the other hand, follows a constant trend at the steady state, which is typical for a frictionless axi-symmetric
extrusion test.
The element deletion procedure is criticized for its violation of mass conservation. Fig. 9, which shows the deleted
elements mapped on the undeformed mesh, for the frictionless analysis, is used to clarify the extent of this violation. This is
to show that, in the meridian plane the large crack gap is not completely due to the deleted elements, but primarily due to the
separation (displacement, falling apart) of the produced free surfaces, under the redistribution of stress fields. Consequently,
together with the efficiency served by element deletion in crack propagation problems, one concludes that, this violation
is justifiable. Besides, the deleted material percent can be calibrated through proper damage controlled adaptive remeshing
schemes, which allocate the smallest possible element size to the maximum damaged area, see e.g. Borouchaki et al. [3]
and Saanouni et al. [55].
Fig. 10 shows the periodic character of the radial displacement response of the surface, for the frictionless simulation.
This phenomenon stands for the bamboo lines observed on the extruded specimen surface which involve internal cracks
Without internal cracks such radial surface oscillations are irrelevant.
The effect of the crack closure parameter, h, is investigated utilizing 0.4 mm mesh, and frictionless conditions. Simula-
tions are conducted for h = 0.0, h = 0.1, h = 0.2, h = 1.0 where the resultant radial extrudate damage accumulations are
plotted on a logarithmic scale. To observe the full extent of the closure parameter, the damage rate multiplier a is selected
as a = 0.003, where damage accumulation is significantly decelerated. The results show the major effect of crack closure
parameter on damage accumulation zones. For complete crack closure, i.e. h = 0, central accumulation governs whereas
making use of standard Lemaitre form, i.e. h = 1.0, surface damage dominates. For h = 0.10 the central and surfacial
damage accumulations are nearly equal. Hambli and Badie-Levet [18] includes the applications of standard Lemaitre model
where significant surface damage accumulation is observed for even not very steep die angles. The main reason, as reveals
this observation, is dematerialized crack closure effect. It should be reminded that, in obtaining the cracks, carried out in
the current study, a complete crack closure, with h = 0, is assumed. For h = 0.20, which is the proposed crack closure
parameter for most of the metals, Lemaitre [37], surfacial rather than central cracks can be carried out as shown in the plot.
Thus, proper determination of the crack closure parameter as a material constant requires significant research.
Finally the effect of friction at the die-billet interface is investigated. For this purpose, a set of simulations for a set of
friction coefficients, such as µf = 0.00, µf = 0.02, µf = 0.04, µf = 0.06, and µf = 0.08, are utilized. The resultant
crack patterns are given in Fig. 12. Accordingly, it can be concluded that, wider and more frequent central cracks are due,

c 2008 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

 www.zamm-journal.org
ZAMM · Z. Angew. Math. Mech. 88, No. 6 (2008) / www.zamm-journal.org 449

Fig. 11 Effect of the crack closure parameter on extrudate ra-

Fig. 10 (online colour at: www.zamm-journal.org) Radial dis-
dial damage distribution (steady state), mesh size = 0.4 mm,
placement at the billet surface, mesh size = 0.2 mm, µf = 0.00.
µf = 0.00.

Fig. 13 (online colour at: www.zamm-journal.org) Compari-

Fig. 12 Effect of friction on discrete crack morphologies and son of the experimental and the numerical discrete crack peri-
frequencies, mesh size = 0.2 mm. odicities, mesh size = 0.2 mm, µf = 0.04.

with the reduction of friction. These results are in complete accordance with those in Saanouni et al. [55]. Fig. 13 shows
that the periodicity of the discrete cracks, for µf = 0.04, are compatible with the experimental observations.
The effect of friction on central and surfacial damage formations may more clearly seen in Figs. 14 and 15, where
observations for µf = 0.00 and µf = 0.08 are compared. Also, for these simulations, the variations of the observed
damage distributions follow a similar frequency pattern with the observed cracks. Besides, with increased friction, an
anticipated increase in the surface damage formations occurs.

5 Closure
Numerical prediction of the periodic central bursts in the direct axi-symmetric extrusion is presented. For this purpose, a
framework, based on a Lemaitre variant (in its simplest form) local isotropic damage strongly coupled to a physically sound
finite strain hyperelastic-plastic formulation through a single J2 yield potential, is derived. A principal axes formulation
is constructed in an Euclidean setting. The derived forms are implemented as VUMAT user defined material subroutines,
for ABAQUS/Explicit code, which supports element deletion procedure to form discrete crack formations. The conducted
single pass axi-symmetric forward extrusion simulations, utilizing bearing steel 100Cr6, supplied periodic central bursts,
which are in correlation with the experimental facts. The investigations show that the crack closure parameter, controlling
the quasi-unilateral evolutionary forms of damage, has an important effect on the damage accumulation zones, by determin-
ing their central or superficial character. Further micro-mechanical and experimental research will be conducted for proper
determination of the material parameters together with more advanced damage rate forms.

www.zamm-journal.org c 2008 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim


450 C. Soyarslan et al.: CDM applications for discontinuous crack formation

Fig. 15 Effect of friction on surfacial damage distribu-

Fig. 14 Effect of friction on central damage distributions, tions, mesh size = 0.2 mm.
mesh size = 0.2 mm.

Acknowledgements This study is conducted with financial support of Julich Research Center and Turkish Science and Technology
Comission, under joint research project MISAG JULICH 04. This support is greatly acknowledged.

Appendix A. Derivations for the local tangent

Components for the local Jacobian, for the simultaneous local integration scheme, can be given as

∂r1, n+1 2 1
= K ′′ (αn+1 ) + 2µ , (41)
∂∆γ 3 (1 − Dn+1 )
∂r1, n+1 ∆γ
= 2µ , (42)
∂Dn+1 (1 − Dn+1 )2
 
∂r2, n+1 a ∂Yn+1
=− Yn+1 + ∆γ , (43)
∂∆γ (1 − Dn+1 ) ∂∆γ
 
∂r2, n+1 a ∂Yn+1
= 1 − ∆γ Yn+1 + (1 − Dn+1 ) , (44)
∂Dn+1 (1 − Dn+1 )2 ∂Dn+1

where K ′′ := ∂ 2 K(α)/∂α2 .

c 2008 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

 www.zamm-journal.org
ZAMM · Z. Angew. Math. Mech. 88, No. 6 (2008) / www.zamm-journal.org 451

+
Appendix B. Derivations for the refined damage conjugate variable Yn+1 = Yn+1
In the present context, the specific damage conjugate variable is selected as the one with the crack closure effect, given in
(17). Hence any of the following,

+ +
∂Yn+1 ∂Yn+1
, , (45)
∂∆γ ∂Dn+1
will exploit the chain rule as follows,
+ 3 +
∂Yn+1  ∂Yn+1 ∂ τ̃n+1, A
= , (46)
∂[•] ∂ τ̃n+1, A ∂[•]
A=1

where [•] represents one of ∆γ or Dn+1 . Accordingly one may derive the following,

+
∂Yn+1 (1 + ν) 3ν (1 + ν) 3ν
= τ̃n+1, A  − p̃ + h −τ̃n+1, A  − h −p̃ , (47)
∂ τ̃n+1, A E E E E

which concludes the formulations.

References
[1] N. Aravas, The analysis of void growth that leads to central bursts during extrusion, J. Eng. Ind. 34, 55–79 (1986).
[2] B. Avitzur, Analysis of central bursting defects in extrusion and wire drawing, J. Mech. Phys. Solids 90, 79–91 (1968).
[3] H. Borouchaki, P. Laug, A. Cherouat, and K. Saanouni, Adaptive remeshing in large plastic strain with damage, Int. J. for Numer-
ical Meth. Eng. 63, 1–36 (2005).
[4] P. Brozzo, B. Deluca, and R. Rendina, A new method for the prediction of formability limits in metal sheets, sheet metal forming
and formability, Proceedings of the Seventh Biannual Conference of the International Deep Drawing Research Group (1972).
[5] M. Brünig, An anisotropic ductile damage model based on irreversible thermodynamics, Int. J. Plasticity 19, 1679–1713 (2003).
[6] M. Brünig and S. Ricci, Nonlocal continuum theory of anisotropically damaged metals, Int. J. Plasticity 21, 1346–1382 (2005).
[7] A. M. Camacho, C. Gonzáles, E. M. Rubio, and M. A. Sebastián, Influence of geometrical conditions on central burst appearance
in axisymmetric drawing processes, J. Mat. Proc. Technol. 177, 304–306 (2006).
[8] E. Ceretti, E. Taupin, and T. Altan, Simulation of metal flow and fracture applications in orthogonal cutting, blanking, and cold
extrusion, Ann. CIRP 46, 187–190 (1997).
[9] S. Choi, Y. Lee, and H. Oh, Ductile fracture in axisymmetric extrusion, J. Mater. Proc. Technol. 74, 263–267 (1998).
[10] M. G. Cockroft and D. J. Latham, Ductility and the workability of metals, J. Inst. Metals 96, 33–39 (1968).
[11] J. L. Chaboche, Chapter 2.04, Damage Mechanics, in: Comprehensive Structural Integrity, Volume 3: Numerical and Computa-
tional Methods, edited by B. Karihaloo (Elsevier Ltd., Amsterdam, 2003) pp. 213–284.
[12] R. A. B. Engelen, Plasticity induced damage in metals, PhD thesis, Technische Universiteit Eindhoven, 2005.
[13] A. M. Freudenthal, The Inelastic Behavior in Solids (Wiley, New York, 1950).
[14] J. C. Gelin, Modelling of damage in metal forming simulations, J. Mater. Proc. Technol. 80–81, 24–32 (1998).
[15] C. Giardini, E. Ceretti and G. Maccarini, Formability in extrusion forging: the influence of die geometry and friction conditions,
J. Mater. Proc. Technol. 54, 302–308 (1995).
[16] M. Gologanu, J.-B. Leblond, G. Perrin, and J. Devaux, Recent extensions of Gursons model for porous ductile metals, in: Con-
tinuum Micromechanics, CISM Lecture Series, edited by P. Suquet (Springer, New York, 1997) pp. 61–130.
[17] A. L. Gurson, Continuum theorie of ductile rupture by void nucleation and growth: Part I. Yield criteria and flow rules for porous
ductile media, J. Eng. Mater. Technol. Trans. ASME 99, 2–15 (1977).
[18] R. Hambli and D. Badie-Levet, Damage and fracture simulation during the extrusion process, Comput. Meth. Appl. Mecha. Eng.
186, 109–120 (2000).
[19] R. Hambli, Comparison between Lemaitre and Gurson damage models in crack growth simulation during blanking process, Int.
J. Mech. Sci. 43, 2769–2790 (2001).
[20] P. Haupt, Continuum Mechanics and Theory of Materials (Springer, Berlin, 2002).
[21] R. Hill, A variational principle of maximum plastic work in classical plasticity, Q.J. Mech. Appl. Math. 1, 18–28 (1948).
[22] T. J. R. Hughes, Analysis of transient algorithms with particular reference to stability behavior, in: Computational Methods for
Transient Analysis, edited by T. Belytschko and T. J. R. Hughes (North-Holland, Amsterdam, 1983), pp. 67–155.

www.zamm-journal.org c 2008 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim


452 C. Soyarslan et al.: CDM applications for discontinuous crack formation

[23] T. J. R. Hughes, The Finite Element Method–Linear Static and Dynamic Finite Element Analysis (Dover Publishers, New York,
2000).
[24] A. Ibrahimbegović and F. Gharzeddine, Finite deformation plasticity in principal axes: from a manifold to the Euclidean setting,
Comput. Meth. Appl. Mech. Eng. 171, 341–369 (1999).
[25] H. C. Jennison, Certain types of defects in copper wire caused by improper dies and drawing practice, in: Proc. Inst. Metal Div.
AIME 89 (1930).
[26] J. W. Ju, Energy-based coupled elastoplastic damage models at finite strains, J. Eng. Mech. 115, 2507–2525 (1989).
[27] J. W. Ju, Isotropic and anisotropic damage variables in continuum damage mechanics, J. Eng. Mech. 116, 283–287 (1990).
[28] L. M. Kachanov, Time of the rupture process under creep conditions, Izv. Akad. Nauk. SSR 8, 26–31 (1958).
[29] D. Ko and B. Kim, The prediction of central burst defects in extrusion and wire drawing, J. Mater. Proc. Technol. 102, 19–24
(2000).
[30] K. Komori, Simulation of chevron crack formation and evolution in drawing, Int. J. Mech. Sci. 41, 1499–1513 (1999).
[31] K. Komori, Effect of ductile fracture criteria on chevron crack formation and evolution in drawing, Int. J. Mech. Sci. 45, 141–160
(2003).
[32] D. Krajcinovic, Damage Mechanics (Elsevier, Amsterdam, 1996).
[33] P. Ladeveze and J. Lemaitre, Damage effective stress in quasi-unilateral conditions, IUTAM Congress, Lyngby (Danemark), 1984.
[34] E. H. Lee, Elasto-plastic deformation at finite strains, ASME J. Applied Mechanics 36, 1–6 (1969).
[35] Y. Lee and S. Hahm, Mechanical property changes in drawing/extrusion of hardening viscoplastic materials with damage, Int. J.
Mech. Sci. 39, 565–573 (1997).
[36] J. Lemaitre, Evaluation of dissipation and damage in metals, in: Proceedings of I.C.M. 1 (Kyoto, Japan.,1971)
[37] J. Lemaitre, A Course on Damage Mechanics (Springer-Verlag, Berlin, 1996).
[38] J. Lemaitre and R. Desmorat, Engineering Damage Mechanics (Springer-Verlag, Berlin Heidelberg, 2005).
[39] L. G. Lim and F. P. E. Dunne, Modelling central bursting in the extrusion of particulate reinforced metal matrix composite mate-
rials, Int. J. Mach. Tools Manufact. 37, 901–915 (1997).
[40] T. S. Liu and N. L. Chung, Extrusion analysis and workability prediction using finite element method, Comput. Struct. 36, 369–
377 (1990).
[41] H. Lämmer and Ch. Tsakmakis, Discussion of coupled elastoplasticity and damage constitutive equations for small and finite
deformations, Int. J. Plast. 16, 496–523 (2000).
[42] P. J. McAllen and P. Phelan, Ductile fracture by central bursts in drawn 2011 aliminium wire, Int. J. Fract. 135, 19–33 (2005).
[43] P. J. McAllen and P. Phelan, Numerical analysis of axisymmetric wire drawing by means of a coupled damage model, J. Mater.
Proc. Technol. 183, 210–218 (2007).
[44] C. McVeigh and W. K. Liu, Prediction of central bursting during axisymmetric cold extrusion of a metal alloy containing particles,
Int. J. Solids Struct. 43, 3087–3105 (2006).
[45] J. Mediavilla, R. H. J. Peerlings, and M. G. D. Geers, A nonlocal triaxiality-dependent ductile damage model for finite strain
plasticity, Comput. Meth. Appl. Mech. Eng. 195, 4317–4634 (2006).
[46] A. Menzel and P. Steinmann, A theoretical and computational framework for anisotropic continuum damage mechanics at large
strains, Int. J. Solids Struct. 38, 9505–9523 (2001).
[47] K. Nahshon and J. W. Hutchinson, Modification of the Gurson Model for shear failure, Eur. J. Mech. – A/Solids 27, 1–17 (2008).
[48] M. Oyane, Criteria of ductile fracture strain, Bull. JSME 15, 1507–1513 (1972).
[49] M. Plancak, Stress distribution within the specimen in cold forward extrusion of steel, J. Mater. Proc. Technol. 24, 387–394
(1990).
[50] Y. Rabotnov, Creep rupture, in: Applied Mechanics. Proceedings of the 12th International Congress of Applied Mechanics, edited
by M. Hetenyi and W. Vincenti (Springer-Verlag, Berlin, 1968), pp. 342–320.
[51] J. R. Rice and D. M. Tracey, On the ductile enlargement of voids in triaxial stress fields, J. Mech. Phys. Solids 17, 201–217 (1969).
[52] J. Rojek, O. C. Zienkiewicz, E. Oñate, and E. Postek, Advances in FE explicit formulation for simulation of metalforming pro-
cesses, J. Mater. Proc. Technol. 119, 41–47 (2001).
[53] J. Rojek, E. Oñate, and E. Postek, Application of explicit FE codes to simulation of sheet and bulk metal forming processes, J.
Mater. Proc. Technol. 80–81, 620–627 (1998).
[54] G. Rousselier, Ductile fracture models and their potential in local approach of fracture, Nucl. Eng. Design 105, 97–111 (1987).
[55] K. Saanouni, J. F. Mariage, A. Cherout, and P. Lestriez, Numerical prediction of discontinuous central bursting in axisymmetric
forward extrusion by continuum damage mechanics, Comput. Struct. 82, 2309–2332 (2004).
[56] J. C. Simó and R. L. Taylor, Consistent tangent operators for rate-independent elastoplasticity, Comput. Meth. Appl. Mech. Eng.
48, 101–118 (1985).
[57] J. J. Skrzypek and A. Gancarski, Modelling of Material Damage and Failure of Structures (Springer-Verlag, Berlin, 1999).
[58] A. Srikanth and N. Zabaras, A computational model for the finite element analysis of thermoplasticity coupled with ductile
damage at finite strains, Int. J. for Numerical Meth. Eng. 45, 1569–1605 (1999).
[59] S. Sriram and C. J. Van Tyne, Criterion for prevention of central bursting in forward extrusions through spherical dies using the
finite element method, J. Manufact. Sci. Eng., ASME 124, 65–70 (2002).
[60] P. Steinmann, C. Miehe, and E. Stein, Comparison of different finite deformation inelastic damage models within multiplicative
elastoplasticity for ductile material, Comput. Mech. 13, 458–474 (1994).
[61] V. Tveergaard and A. Needleman, Analysis of the cup-cone fracture in a round tensile bar, Acta Metall. 32, 157–169 (1984).

c 2008 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

 www.zamm-journal.org
ZAMM · Z. Angew. Math. Mech. 88, No. 6 (2008) / www.zamm-journal.org 453

[62] N. Venkata Reddy, P. M. Dixit, and G. K. Lal, Ductile fracture criteria and its prediction in axisymmetric drawing, Int. J. Mach.
Tools Manufact. Design, Res. Appl. 40, 95–111 (2000).
[63] Z. Zimerman and B. Avitzur, Analysis of central bursting defects in extrusion and wire drawing, J. Eng. Ind. 90, 135–145 (1970).
[64] Z. Zimerman, H. Darlington, and E. H. Kottcamp, Selection of operating parameters to prevent central bursting defects during
cold extrusion, in: Metal Forming – Interrelations between Theory and Practice, edited by A. L. Hoffmanner (Plenum Press, New
York, London, 1971), pp. 47–62.
[65] G. Z. Voyiadjis, R. K. Abu Al-Rub, and A. N. Palazotto, Thermodynamic formulations for non-local coupling of viscoplasticity
and anisotropic viscodamage for dynamic localization problems using gradient theory, Int. J. Plast. 20, 981–1038 (2004).

Book Review

Leonid P. Lebedev and Michael J. Cloud, Tensor Analysis, manner permitted by the material. (4) Full proofs are pro-
World Scientific, Singapore, 2003, 204 pp., US $ 58.00, vided. A particularly nice feature is a well-organized ap-
£ 43.00, ISBN: 978-981-238-360-0 pendix containing compact listings of all the principal for-
This is a textbook on tensor analysis, a tool without which mulas required in applications – these 18 pages alone are
subjects such as continuum mechanics and relativity the- worth the price of the book.
ory cannot be understood. The book gives a brief, clear, The book covers the algebra of tensors as well as the cal-
and comprehensive introduction to the theory. It assumes culus of tensor fields. It includes a brief yet comprehensive
as background only the basics of linear algebra and calcu- introduction to differential geometry that presents the basic
lus that are known to any engineering student. It is a book theory of curves and surfaces. This part is written in the same
for those who must understand and apply tensor analysis in manner as the main text.
their work. The book is useful for studying, understanding, and learn-
The book opens with elementary discussions of vectors ing to apply tensors. It should be of interest to both under-
and of linear operators represented by matrices with respect graduate and graduate students of physics and engineering,
to some basis. It provides an understanding of tensors as a as well as to practicing physicists and engineers.
generalization of these notions. The theory is presented in The book consists of five chapters: Chap. 1. Preliminar-
three-dimensions and therefore geared toward engineering ies; Chap. 2. Transformations and Vectors; Chap. 3. Tensors;
applications. The book presents a selection of results from Chap. 4. Tensor Fields; Chap. 5. Elements of Differential
tensor analysis that could be characterized as follows. (1) Geometry. Appendix A. Formulary. Appendix B. Hints and
The results are necessary for practical purposes. (2) They Answers.
cover the needs of the majority of workers in continuum
mechanics. (3) They are presented in the most elementary Rostov on Don Victor A. Eremeyev

www.zamm-journal.org c 2008 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim