You are on page 1of 4

Ductile fracture prediction in cold forging process chains

P.F. Bariani (1)

, S. Bruschi (2)
*, A. Ghiotti
, M. Simionato
DIMEG, University of Padova, Via Venezia 1, 35131 Padova, Italy
DIMS, University of Trento, Via Mesiano 77, 38050 Trento, Italy
1. Introduction
The onset of ductile fracture is the most severe constraint in
designing and optimising process chains for metal parts, which are
based on forming operations.
The separation of material in ductile fracture is only a nal
event, which is preceded by extensive plastic deformation and the
formation of tiny voids, which then growand coalesce, resulting in
cracks that lead to fracture. As is the case with other micro-
structural phenomena that occur in the steps of forming process
chains such as strain and strain-rate hardening, crystallographic
anisotropy and mechanical bering, recovery and recrystallization,
grain growth, and strain-induced ageing the onset of ductile
fracture is strongly affected by the thermal and mechanical
histories that have been generated in the previous steps [1].
Therefore, in ductile fracture studies, the analyses of the different
steps of the process chain must be effectively interconnected.
In cold forging process chains, the effects of the stress and strain
histories stored during the process largely prevail over those of
temperature- and time-related parameters in governing both the
material response to deformation and the ductile fracture
occurrence. On close examination, the loading and deformation
in the workpiece material zones that are exposed to fracture risk
may show complex paths (non-proportional and non-monotonic).
The stress triaxiality varies from zone to zone and, for the same
zone, during the deformation step, with stepwise variations
especially at the changes from one process stage to the next.
Ductile fracture phenomena in cold forging operations have
been extensively investigated by using different approaches and
models. Depending on the approach, the models can be classied
into three groups: energy-based models, void growth models and
Continuum Damage Mechanics (CDM) based models. The models in
the rst group provide satisfactory results whenthey are applied to
single-step operations where tensile stress states are predominant
[2]. They are easy to use and calibrate, but they lack accuracy when
they are applied to most cold forging operations where loading is
more complex. Void growth models incorporate the dependence of
damage on the stress triaxiality and are calibrated through tests
that have to reproduce the same strain and stress conditions of
the process. Although they provide greater validity than those in
the rst group, they are unable to predict fracture occurrence for
the intermediate regimes of stress triaxiality where competing
failure mechanisms work simultaneously [3,4]. CDM-based models
can effectively predict fracture occurrence because they model
damage evolution on a micro-scale and take account of rheological
behaviour [5]. Calibrating these models requires experimental and
numerical procedures that are complex and time-consuming,
making it quite difcult to implement them.
In this paper, a new approach is presented, which is capable of
accurately predicting ductile fracture occurrence in multi-stage
cold forging process chains. The approach combines the fracture
criterion proposed by Xue and Wierzbicky [6] with a linear damage
accumulation law. According to the fracture criterion, the
equivalent strain to fracture is evaluated as a function of both
the stress triaxiality factor and the deviatoric stress parameters. By
introducing the deviatoric parameter, the dependence of the strain
to fracture from the stress state is captured, making the model
applicable for any stress state and any failure mode. The restricted
number of calibration tests and ease of performing the relevant
standard experiments are a further advantage. The damage
accumulation law makes it possible to track the continuous
material damaging taking place within the individual deformation
steps in the process chain. These features allow the approach
presented in the paper to overcome the main limitations of the
above-cited models.
The paper is organized in two main parts. In the rst part, the
approach underlying the development of the model and its main
functional modules are illustrated. Then the paper focuses on the
industrial application case. The material tested and the experi-
ments carried out to calibrate the fracture criterion are presented
CIRP Annals - Manufacturing Technology 60 (2011) 287290
Cold forging
The paper presents a new approach for the prediction of ductile fracture occurrence in multi-stage cold
forging process chains. The approach combines the fracture criterion proposed by Xue and Wierzbicky
with a linear damage accumulation law. Thanks to this feature, the approach is capable of predicting both
the location where the failure events occur under the action of external loading and the time they take to
be generated. An application to the multi-stage cold forging of a C35 Torx-type socket screw carried out
on a double-blow header is presented and results of predictions are compared with experimental
2011 CIRP.
* Corresponding author.
E-mail address: (S. Bruschi).
Contents lists available at ScienceDirect
CIRP Annals - Manufacturing Technology
j ournal homepage: ht t p: / / ees. el sevi er. com/ ci rp/ def aul t . asp
0007-8506/$ see front matter 2011 CIRP.
together with the results from the application of the model and
their comparison with the experimental observations.
2. The approach
The major requirement that the predictive model has to meet is
to predict damage evolution and fracture occurrence in the multi-
stage cold forging processes with the complex loading and
deformation paths that operate in industrial applications. Non-
secondary feature of the model is the ease in its implementation
and application, particularly with regard to calibration tests. With
these requirements, the predictive model requires the joint use of
complementary techniques, both numerical and experimental,
organized in four functional modules which are devoted respec-
tively to: (i) application of the fracture criterion, (ii) calculation of
the damage accumulation, (iii) FE analysis of the process chain, and
(iv) generation of the material data for process simulation and
fracture criterion calibration. The four modules are described in the
rest of this section.
2.1. The fracture criterion
According to the ductile fracture criterion proposed by Xue and
Wierzbicky [5], fracture is postulated to occur when the
accumulated equivalent plastic strain, modied by the function
of the stress triaxiality T and the deviatoric state parameter X,
reaches a limiting value equal to one (for the denition of T and X
please refer to the original reference [6], where the two parameters
are named h and j, respectively). With the introduction of the two
stress parameters, the fracture criterion can be adequately applied
to deformation conditions with any stress state and to any failure
mode as well. Moreover, the fracture locus can be conveniently
represented in the stress domain.
In the case of material behaviour obeying the Hollomon
rheological law, the upper and lower limits of the fracture locus
can be correlated through the Tresca failure hypothesis [7]. The
equivalent strain at fracture e
, valid for any value of the stress
triaxiality T, can then be expressed by Eq. (1):

sin2p arccos X=3

where e
is the upper limit of the fracture locus. Assuming e
to be
an exponential function of the stress triaxiality factor T [8], Eq. (1)
becomes the fracture locus represented by Eq. (2):
sin2p arccos X=3

where C
and C
are material constants to be calculated through
tests aimed to calibrate the fracture criterion (see Section 2.4) and
n is the exponent in the Hollomon power law. Therefore, the
material fracture locus can be fully determined as a function of any
stress state expressed in terms of the stress triaxiality factor T and
the deviatoric stress parameter X.
2.2. The damage accumulation law
The accumulation of the material damage in the deformation
steps of the process chain is tracked by a linear damage
accumulation law. According to this law, the additional damage
introduced at each time step in the process chain numerical
simulation is dened by [6]:
T; X
d e (3)
where e
T; X represents the material fracture locus of Eq. (2). The
time derivative of Eq. (2) is implemented into the numerical code
used for the process chain simulation by compiling a user routine
whose ow chart is shown in Fig. 1. The material failure is
postulated to occur when the damage parameter D reaches the
unit value.
2.3. The FE model of the process chain
The main task of the FE model of the cold forging process chain
is to provide an accurate analysis of the stress- and strain-related
parameters inside the workpiece material in the different stages of
the process chain. The damage law described in Section 2.2 is
implemented in the FE model. In this way, the damage
accumulated inside the individual stages of the process chain is
calculated and the accumulated damage is transferred from one
step to the next.
2.4. Tests for model calibration
Rheological and workability tests are utilized to identify the
material constants C
, C
and n in Eq. (1). In particular, compression
tests are used to determine the material ow behaviour and to
calculate the strength coefcient and the exponent of the Hollomon
power law. Tensileandtorsiontests areusedtoidentifytheconstants
and C
. The two constants are calculated by comparing the results
measured in the experimental tests with the corresponding values
calculated through numerical simulations of the same tests, and
therefore are independent fromthe particular reference process, but
only material dependent. Details are given in Section 3.1, where the
tests carried out for the application case are presented.
3. Application case
The application case refers to the multi-stage cold forging of a
Torx-type socket screw carried out on a double blow header (Fig.
2). The material is a 6.5 mm diameter wire of carbon steel C35.
After straightening on a roll straightening unit, the wire is cropped
to the required length and then forged. The two forging steps
consist in a preliminary heading followed by the piercing of the
Torx socket combined with the calibration of the screw head. A
ductile fracture regularly occurs in the last forming stage at one of
the six points of the star-shaped cavity (Fig. 2d and lower part of
Fig. 6). It is important to note that the material zone where fracture
occurs has already been damaged and separated during the
cropping operation. In the forging sequence, the same material
undergoes complex loading with medium-high values of stress
triaxiality, which shows stepwise changes when the blank moves
from one forming stage to the following.

Fig. 1. Flow-chart of the damage accumulation routine.

Fig. 2. Forging sequence for a C35 Torx-type socket screw.
P.F. Bariani et al. / CIRP Annals - Manufacturing Technology 60 (2011) 287290 288
3.1. Material testing and data generation
A set of tests at room temperature was carried out on samples
cut from the C35 wire in order to generate the rheological and
workability data, which are required to calibrate both the FE model
and fracture criterion. The testing program includes compression,
tensile and torsion tests.
The compression tests were carried in order to generate the true
stress-true strain curve of the C35 steel. The tests were carried out
on an MTS 322 servo-hydraulic machine with a ram speed of
0.1 mm/s. After compensation for the press frame stiffness, the
ow curve was modelled by using Hollomon power law [9]. The
strain-hardening exponent n and strength coefcient K were
calculated through linear regression analysis and their values are
shown in Table 2.
The tensile tests were carried out on both smooth (un-notched)
and notched specimens in order to evaluate the inuence of stress
triaxiality on the material workability limits. Three different values
of the notch radius were utilized: 2.00, 5.00 and 11.00 mm, which
generated a suitable range of variation for the stress triaxiality. The
tensile tests were carried out on an MTS 322 servo-hydraulic
machine with a ram speed of 0.1 mm/s. The equivalent strain at
fracture was calculated from the area reduction to fracture,
measured on a coordinate measuring machine.
The torsion tests were carried out in order to generate material
workability data under plane strain conditions. The torsion tests
were carried out on a torsion machine equipped with a special grip
at a revolution speed of 10 rpm. To determine the equivalent strain
to fracture, the material ow curves were calculated according to
the FieldsBackofen theory.
3.2. Finite element analysis
To provide accurate analyses of the stress- and strain-related
parameters inside the workpiece material in the different forming
stages of the Torx-type socket screw, a 3D FE model of the whole
process chain from wire drawing to socket piercing was
developed using the commercial FE code Forge 2008
. The main
assumptions andinput parameters in the FE model were: (i) elasto-
plastic behaviour of the wire material according to Hollomon
power law, (ii) perfect rigid tooling, (iii) a Tresca friction factor
equal to 0.07, according to measurements in double cup extrusion
tests [9], and (iv) constant values of 10,000 and 10 W/(m
K) for the
heat transfer coefcient at the material-tooling interfaces and with
the environment, respectively. For accurate simulation of the
material separation in the cropping stage, very ne mesh was used
in combination with the killing element technique, which deletes
the elements when the critical damage value is reached.
3.3. Fracture locus identication
Fig. 3 shows the stress triaxiality factor T vs. the equivalent
strain e for the un-notched and notched specimens of the tensile
tests. The values of T and e were calculated in the neck zone by
using an FE model of the tests where local Lagrangian sensors were
implemented. Because the parameters T and e vary during the tests
and in the necking section, average values were dened [10], which
are given in Table 1 for the four tensile tests together with those
calculated for the torsion test.
The material constants C
and C
were then calculated from the
data in Table 1 using linear regression analysis applied to Eq. (2),
where current values of T and e
were replaced by average values.
The calculated values of the two material constants are given in
Table 2. Finally, the material fracture locus for the C35 steel is
shown in Fig. 4, where the data of the tensile and torsion tests are
3.4. Results
This section presents the FE predictions computed with the
proposed fracture criterion and damage accumulation law. These
were then compared with the measurements from the industrial
trials and with the FE predictions obtained by implementing the
conventional Oyane damage criterion, which was calibrated on
the basis of the torsion tests described in Section 3.1. The
comparison was made with reference to: (i) the shape and
extension of the sheared zone in the cropping stage, (ii) the
damage evolution during the heading and piercing steps, and (iii)
the location of the damaged areas on the screw head in the nal
piercing stage.
Fig. 5 shows the comparison of the sheared zone obtained in the
industrial trials with those predicted by both the proposed model
and the Oyane damage criterion. The shape of the sheared surfaces
computed in the FE simulations is strongly dependent on the
implemented fracture criteria, as was evidenced in [10]. According
to the proposed model, the average height variation of the sheared
surfaces was estimated to be about 170 mm, while for the bar it
was around 200 mm, with a deviation of less than 10% compared to
experimental results. The deviation of the same prole simulated
using the Oyane criterion was signicantly higher (35% larger than
that in the experiments). Fig. 6(a) shows the evolution of the

Fig. 3. Stress triaxiality at a Lagrangian sensor location for the different tensile
Table 1
Workability data of the C35 steel.
Specimen geometry T
X e
RNB2 (tensile) 0.91 1 0.38
RNB5 (tensile) 0.67 1 0.46
RNB11 (tensile) 0.54 1 0.59
Smooth (tensile) 0.45 1 0.69
Torsion 0 0 0.69

Fig. 4. Fracture locus of the C35 steel.
Table 2
Rheological and workability constants of the C35 steel.
K n C
881 0.22 1.8224 2.377
P.F. Bariani et al. / CIRP Annals - Manufacturing Technology 60 (2011) 287290 289
damage as predicted by the two criteria, and the evolution of the T
and X parameters computed at the point P plotted vs. the time
during the heading and piercing steps.
At the beginning of the heading stage, the values of the damage
calculated with both the criteria, are not null since they derive
from the computation of the previous cropping step. The proposed
model proved to be accurate in predicting both the accumulated
damage and localization of the damage areas. In fact, the critical
value D
is reached in the piercing stage, as was recorded in
the industrial trials. On the contrary, the time needed to reach the
Oyane critical damage, indicated with D
in Fig. 6(a), is
signicantly longer. The lower part of Fig. 6 shows the maps of the
accumulated damage as calculated by the two criteria. The
maximumvalues of the scales indicate the critical values according
to the two damage models. Due to the stress state of the piercing
stage, characterized by high levels of both the T and X parameters,
the initial calibration of the Oyane criterion through torsion test
data proves to no longer be valid. On the contrary, thanks to its
formulation, the proposed damage model does not require any
additional calibration and can predict the fracture start and
propagation fromthe points of the Torx star, as was recorded in the
industrial trials.
4. Summary
A newapproach for the prediction of ductile fracture occurrence
in multi-stage cold forging process chains was presented. The
approach combines a proper fracture criterion with a linear damage
accumulationlawandit provedtobecapable of accuratepredictions
of both the location where the failure events occur under the action
of external loadingandthe time theytake tobe generated. Thanks to
a fracture criterion that is reliable with any stress state and failure
mode, the approach proves to be general enough to be applied to
most of cold forging process chains. Ease in calibration and
application is a further advantage of the approach. Further research
work in progress at the time of writing deals with the implementa-
tion of non-linear damage accumulation law, which should be
considered for damage evolution description in complex stress
paths, such as those induced by reverse loading.
[1] Pietrzyk M, Madej L, Kuziak R (2010) Optimal Design for Manufacturing Chain
Based on Forging for Copper Alloys, with Product Properties Being the Objec-
tive Function. CIRP Annals 59(1):319322.
[2] Ceretti E, Taupin E, Altan A (1997) Simulation of Metal Flow and Fracture
Applications in Orthogonal Cutting, Blanking, and Cold Extrusion. CIRP Annals
[3] Gouveia BPPA, Rodrigues JMC, Martins PAF (2000) Ductile Fracture in Metal-
working: Experimental and Theoretical Research. Journal of Materials Proces-
sing and Technology 101:5263.
[4] Barsoum I, Faleskog J (2007) Rupture Mechanisms in Combined Tension
and Shear-Micromechanics. International Journal of Solids and Structure 44:
[5] Soyarslan C, Tekkaya E (2009) Chevron Cracks in Extrusion: Modelling
and Prevention. Proceedings of 5th JSTP Seminar of Precision Forging, Kyoto,
[6] Wierzbicki T, Bao Y, Lee YW, Bai Y (2005) Calibration and Evaluation of Seven
Fracture Models. International Journal of Mechanical Sciences 47:719743.
[7] Coppola T, Cortese I, Folgarait P (2009) The Effect of Stress Invariants on Ductile
Fracture Limit in Steels. Engineering Fracture Mechanics 76:12881302.
[8] Rice JR, Tracey DM (1969) On the Ductile Enlargement of Voids in Triaxial
Stress Fields. Journal of Mechanics and Physics of Solids 17:201217.
[9] Simionato M(2010), Damage Modelling inCold Forging Process Chains, DIMEG
PhD Thesis.
[10] Breitling V, Chernauskas E, Tupin. Altan T (1997) Precision Shearing of Billets:
Special Equipment and Process Simulation. Journal of Materials Processing and
Technology 71:119125.

Fig. 5. Shape and extension of the sheared zone: experimental measurements (a);
calculated by the Oyane criterion (b) and by the proposed model (c).

Fig. 6. Damage evolution according to the Oyane criterion and the proposed model
(a); maps of damage according to the proposed model (b) and to the Oyane criterion
(c). Cracks in the industrial trials (d).
P.F. Bariani et al. / CIRP Annals - Manufacturing Technology 60 (2011) 287290 290