Journal of Materials Processing Technology 215 (2015) 287298

Contents lists available at ScienceDirect

Journal of Materials Processing Technology

journal homepage: www.elsevier.com/locate/jmatprotec

Failure by fracture in bulk metal forming

C.M.A. Silva a , L.M. Alves a , C.V. Nielsen b , A.G. Atkins c , P.A.F. Martins a,
a

Instituto Superior Tcnico, Universidade de Lisboa, Av. Rovisco Pais, 1049-001 Lisboa, Portugal
Department of Mechanical Engineering, Technical University of Denmark, DTU Building 425, DK-2800 Kgs. Lyngby, Denmark
c
Department of Engineering, University of Reading, Box 225, Reading RG6 6AY, UK
b

a r t i c l e

i n f o

Article history:
Received 3 June 2014
Received in revised form 21 July 2014
Accepted 23 August 2014
Available online 2 September 2014
Keywords:
Bulk metal forming
Formability
Stress triaxiality
Finite element modelling
Experimentation

a b s t r a c t
This paper revisits formability in bulk metal forming in the light of fundamental concepts of plasticity,
ductile damage and crack opening modes. It proposes a new test to appraise the accuracy, reliability and
validity of fracture loci associated with crack opening by tension and out-of-plane shear under loading
conditions different from those found in conventional tests for bulk formability based on cylindrical,
tapered and anged specimens.
The new formability test consists of expanding rings of various wall thicknesses with a stepped conical
punch and allows investigating the onset of failure by cracking under three-dimensional states of stress
subjected to various magnitudes of stress triaxiality.
The presentation is supported by nite element analysis and experimentation in aluminium AA2030T4 and results show that failure by fracture under three-dimensional loading conditions can be easily
and effectively characterized in the space of equivalent strain to fracture and stress triaxiality.
2014 Elsevier B.V. All rights reserved.

1. Introduction
Failure by surface or internal cracking in bulk metal forming is
caused by the accumulation of ductile damage within regions that
are highly strained due to extensive plastic ow. Apart from special
purpose processes such as the shearing of bars and bar sections,
where cracks are needed to cut material, the occurrence of cracks
is generally undesirable and should be prevented during process
design.
Currently-available nite element computer programs may aid
this objective but appropriate input data regarding a relevant fracture locus is crucial for successfully predicting the onset of cracking
in bulk metal forming. The conditions at fracture depend on the
interaction between plasticity theory, the circumstances under
which each crack opening mode will develop and microstructural
ductile damage mechanics.
In a recent paper Martins et al. (2014) developed an analytical framework to characterize failure by fracture in metal forming
under plane stress conditions and concluded that surface cracking
in bulk forming can occur under two different opening modes (i)
tensile and (ii) out-of-plane shear, respectively the same as modes

Corresponding author. Tel.: +351 218419006.

E-mail addresses: carlos.alves.silva@ist.utl.pt (C.M.A. Silva), luisalves@ist.utl.pt
(L.M. Alves), cvni@mek.dtu.dk (C.V. Nielsen), a.g.atkins@reading.ac.uk (A.G. Atkins),
pmartins@ist.utl.pt (P.A.F. Martins).
http://dx.doi.org/10.1016/j.jmatprotec.2014.08.023
0924-0136/ 2014 Elsevier B.V. All rights reserved.

I and III of fracture mechanics. The framework allowed concluding

that the normalized version of the Cockcroft and Latham (1986)
ductile damage criterion is based on a hidden out-of-plane shearbased condition that explains its wide applicability to predict the
onset of cracking in opening mode III. The framework also conrmed the link between stresstriaxiality-based damage criteria
(McClintock, 1968) and fracture toughness in crack opening mode
I, which had been previously established by Muscat-Fenech et al.
(1996).
Martins et al. (2014) associated crack opening modes I and III to
linear fracture loci with slopes 1 and 1/2 observed in the pioneering bulk formability tests that were carried out from the late
1960s to the early 1980s. The investigations by Kudo and Aoi (1967)
and Kuhn et al. (1973), for example, showed that the limiting fracture strain pairs on the outside surfaces of upset test specimens fall
on a straight line of slope 1/2 in the principal strain space (that
is, a line parallel to uniaxial compression loading under frictionless
(homogenous) conditions refer to A in Fig. 1).
Martins et al. (2014) claimed this fracture locus is associated
with crack opening by out-of-plane shear (mode III) (Fig. 1a) in
close agreement with the observations by Kobayashi (1970) and
Atkins and Mai (1985) who concluded that both the vertical and
the inclined surface cracks that are commonly found in upset compression tests do not run radially. Vertical cracks occur in specimens
with low friction and larger aspect ratios whereas inclined surface
cracks are generally found in specimens with high friction and low
aspect ratios.

288

C.M.A. Silva et al. / Journal of Materials Processing Technology 215 (2015) 287298

Fig. 1. Schematic representation of fracture loci in bulk metal forming in the (a and c) principal strain space and in the (b and d) space of equivalent strain to fracture and
stress triaxiality. (a and b) Linear fracture locus associated to crack opening by out-of-plane shear (mode III); (c and d) bilinear fracture locus resulting from combination of
crack opening by out-of-plane shear (mode III) and by tension (mode I).

Subsequent investigations by Erman et al. (1983) disclosed the

possibility of the limiting fracture strain pairs on the outside surfaces of bulk formability test specimens to fall on a bilinear locus
resulting from combination of the previously mentioned straight
line of slope 1/2 (mode III) with a straight line of slope -1 , parallel to pure shear loading conditions (Fig. 1c). Martins et al. (2014)
claimed the fracture locus corresponding to a straight line of slope
1 parallel to pure shear loading conditions (refer to B) to result
from crack opening by tension (mode I). In connection to Fig. 1a and
c it is worth noting that C, D and E refer to loading paths corresponding to pure tension, plane strain and equal biaxial stretching
and that ellipses refer to isolines of constant effective strain .
The transformation of fracture loci in bulk metal forming from
principal strain space (Fig. 1a and c) to the space of equivalent strain
to fracture f and stress triaxiality m / is schematically shown in
Fig. 1b and d.
The space of equivalent strain to fracture and stress triaxiality
was originally proposed by Vujovic and Shabaik (1986) as a forming
limit criterion for bulk metalworking processes, based on the magnitude of the hydrostatic component and the effective stress of the
stress state (the rst invariant of the stress tensor and the second
invariant of the deviatoric stress tensor). However, its widespread
acceptance only came after Bao and Wierzbicki (2004) utilization
for investigating the capability of several well-known ductile fracture criteria to predict failure by fracture in metal forming. The

utilization of the space of equivalent strain to fracture and stress

triaxiality was further developed by Wierzbicki and co-workers in
several other publications that were focused on the development
of ductile fracture criteria for metal forming (for example, Li et al.,
2010).
In a subsequent publication Bai and Wierzbicki (2010) proposed
the utilization of an enhanced version of this space where the axes
are the equivalent strain to fracture, the stress triaxiality and the
normalized Lode angle parameter.
A review of the literature on the experimental determination
of fracture loci and on the development of non-coupled ductile
damage criteria for predicting the onset of cracking in metal forming reveals (i) that most of the investigations treat sheet and bulk
forming separately and (ii) that free surface normal strains causing cracking are exclusively characterized by means of plane stress
loading conditions.
The rst conclusion may be attributed to fundamental differences in plastic ow that result from two and three dimensional
stress loading conditions that are typical of sheet and bulk
metal forming and to its inuence on the circumstances under
which different processes fail by fracture. The need to treat sheet
and bulk metal forming separately also copes with the overall aims and scope of a recent keynote paper by Bruschi et al.
(2014) that is exclusively focused on formability in sheet metal
forming.

C.M.A. Silva et al. / Journal of Materials Processing Technology 215 (2015) 287298

The second conclusion raises the problem of accuracy, reliability

and validity of fracture loci under stress and strain loading conditions different from those of conventional bulk formability tests
based on cylindrical, ring, tapered and anged test specimens. In
fact, as far as the authors are aware, there is currently no simple bulk formability test aimed at assessing the onset of cracking
under three-dimensional stress loading conditions that are typical
of internal cracks found, for example, in open die forging of cast
ingots, or chevron cracks found in extrusion. This is probably due
to difculties in monitoring the onset and development of cracks
in locations away from the free surfaces where two-dimensional
plane stress conditions prevail.
Under these circumstances, the aim of this paper is threefold:
rst, to identify and characterize the transitions between tension
and out-of-plane shear dominated crack opening modes in bulk
metal forming, second to present a new and simple bulk formability test that permits the determination of limiting fracture strain
pairs in three-dimensional stress loading under a wide magnitude of stress triaxiality conditions. Furthermore it is demonstrated
that fracture forming limit diagrams in principal strain space that
successfully predict fracture in sheet forming, cannot characterize failure by fracture under three-dimensional loading conditions.
Instead, the space of equivalent strain to fracture and stress triaxiality must be employed. The presentation is supported by numerical
modelling using an in-house nite element computer program and
by experimentation in aluminium AA2030-T4.
The organization of the paper is the following. Section 2 summarizes the mechanical and friction characterization of the testpieces
and reviews conventional bulk formability tests. It goes on to
present the new bulk formability test that is capable of triggering cracks under three-dimensional stress loading conditions,
and describes the methods and procedures that were utilized to
determine strain loading paths and fracture strain pairs in the
experiments. Section 3 concerns the nite element modelling conditions that were utilized in the numerical simulations of the
bulk formability tests. Section 4 presents and discusses the results
obtained, namely the crack opening modes and the fracture locus,
in the light of the analytical framework that was recently proposed
by Martins et al. (2014). Section 5 presents the conclusions.

2. Experimentation
2.1. Mechanical and friction characterization
The investigation was performed on commercial aluminium
AA2030-T4 in the as-supplied condition. The material was provided as solid rods with 35 mm diameter and the stressstrain
curve was determined by means of compression tests carried out on
Rastegaev cylindrical test specimens at room temperature (Fig. 2a),
using a hydraulic testing machine (Instron SATEC 1200kN). The at
recesses that were machined at the upper and lower ends of the
specimens (refer to the schematic cross section in Fig. 2a) were
lled with a molybdenum disulphide (MoS2) based lubricant with
the purpose of preventing barrelling by retaining the lubricant at
the contact interface between the tool and the specimens during
compression.
Friction at the contact interface between the tool and the various bulk formability test specimens that are described in Section 2.2
was estimated by means of ring compression tests. The ratio of the
outer diameter, the inner diameter and the thickness of each ring
specimen was 6:3:2. No lubricant was used in the ring compression tests because the bulk formability tests were mainly performed
under dry lubrication conditions.
Friction was characterized by means of the law of constant friction  = mk and the friction factor m was estimated as m = 0.2 after

289

Fig. 2. Mechanical and friction characterization of the aluminium AA2030-T4.

Stressstrain curve obtained by compression tests and detail of the Rastegaev cylindrical test specimen; experimental measurements and ring test calibration curves
relating the changes of the minimum internal diameter with the reduction in height
for several friction factors.

comparing experimental measurements and calibration curves

relating the changes of minimum internal diameter as a function of
the reduction in height that were previously determined by nite
element analysis (Fig. 2b).
2.2. Bulk formability tests
The bulk formability tests were performed on the same
hydraulic testing machine as used for the mechanical and friction
characterization. The experiments were carried out at room temperature and the crosshead velocity of the testing machine was set
equal to 10 mm/min (0.17 mm/s).
The work plan made use of two different groups of specimens
and tool setups. The rst group corresponds to conventional bulk
formability tests performed with axially-loaded cylindrical specimens, cylindrical specimens loaded across a diameter (rotated
cylindrical or Brazilian), tapered and anged test specimens that
were compressed between at parallel platens. The specimens
were machined according to the geometries provided in Table 1.
The second group corresponds to the new proposed bulk formability test (hereafter designated as ring expansion with a stepped
conical punch) that allows characterizing, for the rst time ever,
fracture strain pairs in three-dimensional stress loading under
varying magnitudes of stress triaxiality. The punch was made of
cold working tool steel (120WV4-DIN) hardened and tempered to a
Rockwell hardness of HRC 60 in the working region and exploratory
calculations (not given here) regarding the distribution of damage
and the required stroke of the punch led to its design. The specimens were machined from the supplied aluminium AA2030-T4
rods according to geometry provided in Table 2.
The new proposed bulk formability test may be seen as a modication of the expansion of thick-walled discs with a conical
punch that was employed by Tomkins and Atkins (1981). In fact,

25
30
25
5
Dry
t1
30
30

Dry
cr1
25
25

Dry
c2
25
25

MoS2
c1

Cylindrical

30
30

Dry
c3

Cylindrical rotated

Tapered

25
35
25
5
Dry
t2

20
35
25
5
Dry
f1

C.M.A. Silva et al. / Journal of Materials Processing Technology 215 (2015) 287298

Flanged

290

contrary to the expansion of thick-walled discs that lead to cracks

on the upper and lower surfaces at locations in-between the inner
and outer diameters, the expansion of thick-walled rings with a
stepped conical punch leads to cracks at the inner conical surface
edge and allows controlling the magnitude of stress triaxiality by
varying ring wall thickness. The physics behind the occurrence of
cracking in ring expansion with a stepped conical punch will be
comprehensively explained later in Section 4.
The methods and procedures to determine the experimental
strain pairs on the surface of the conventional bulk formability
test specimens that are shown in Table 1 are comprehensively
described in Gouveia et al. (1996). The technique made use of rectangular grids with 4 mm initial side length that were engraved at
the mid-height of the original specimens (Fig. 3a) and the measurements of grid spacing that resulted from plastic deformation
at several height reductions were carried out in a Mitutoyo TM111 toolmakers microscope equipped with an XY measurement
micrometre table.
The resulting circumferential  and axial z surface strains at
various stages of deformation were obtained from,
 = 1 = ln

w
,
w0

z = 2 = ln

h
h0

(1)

The procedure to determine the strain loading paths up to fracture in ring expansion with a stepped conical punch was performed
by inverse identication of strains at locations U and T of Fig. 3b
using nite element analysis. This was necessary because not only
it is difcult to engrave rectangular grids on the inner ring surfaces
but also they vanish during expansion by means of the stepped
conical punch.

Numerical simulation of the bulk formability tests was performed with the nite element computer program I-form. The
program was developed by the authors and is built upon the irreducible nite element ow formulation, which is based on the
following variational principle (extended to account for frictional
effects),

 dV + 12 K

H
D
d
t
Lubrication
Identication

Geometry

Table 1
Geometry and lubrication conditions utilized in conventional bulk formability tests. All dimensions are in mm.

3. Finite element modelling


V

2v dV


V

Ti ui dS +

  |ur |
Sf

f dur dS

(2)

The symbol  in (2) denotes the effective stress, is the effective strain rate, V is the volumetric strain rate, K is a large positive
constant imposing the incompressibility constraint, Ti and ui are
the surface tractions and velocities on ST , ur and  f are the relative
velocity and the friction shear stress (according to the law of constant friction  f = mk) on the contact interface Sf between tools and
test specimens, and V is the control volume limited by the surfaces
SU and ST . Further information on the nite element ow formulation and the computer program I-form can be found elsewhere
(Nielsen et al., 2013).
The numerical simulation of the bulk formability tests based
on cylindrical, tapered, anged and ring specimens made use of
two dimensional nite element models and axisymmetric linear
quadrilateral elements that took advantage of rotational symmetry
conditions whereas the numerical simulation of the bulk formability tests based on rotated cylindrical test specimens made use
of three dimensional nite element models and linear hexahedral
elements (Fig. 4).
The numerical simulation of the transitions between tension
and out-of-plane shear dominated crack opening modes in bulk
metal forming that will also be presented in Section 4 involved
three dimensional nite element modelling and discretization of
the anged test specimen by means of linear hexahedral elements.

C.M.A. Silva et al. / Journal of Materials Processing Technology 215 (2015) 287298

291

Table 2
Geometry and lubrication conditions utilized in ring expansion with a stepped conical punch. All dimensions are in mm.

Geometry

Specimens

H
D
d
Lubrication
Identication

15
12
10
Dry
r1

w0

t0

(a)

15
15
10
Dry
r2

15
17.5
10
Dry
r3

15
20
10
Dry
r4

15
25
10
Dry
r5

15
30
10
Dry
r6

(b)

Fig. 3. Experimental strain pairs in bulk formability tests. Schematic representation of the grids that were utilized for measuring the major and minor surface strains in case
of conventional bulk formability specimens; Locations U and T to perform inverse nite element identication of the major and minor strains in case of ring expansion
with a stepped conical punch.

Fig. 4. Finite element discretization of the bulk formability test specimens that are identied (from a to f) as c3, cr1, t1 and f1 in Table 1 and r5 and r1 in Table 2. All nite
element models apart from (b) took advantage of rotational symmetry conditions.

292

C.M.A. Silva et al. / Journal of Materials Processing Technology 215 (2015) 287298

Fig. 5. Experimental and nite element predicted evolution of the load with displacement for the (a) cylindrical and (b) anged test specimens that are identied
as c3 and f1 in Table 1.

The volume integrals in (2) were evaluated by means of

complete and reduced integration schemes in order to ensure
incompressibility of the material ow. The rightmost surface
integral in (2) required tools to be discretized by means of contactfriction linear or spatial triangular elements in the case of two
and three-dimensional modelling, respectively, and made use of
ve-Gauss point integration rules. The friction factor was taken
as m = 0.2 after calibration by means of ring compression testing
(Section 2.1).
The numerical simulations of the bulk formability tests were
performed through a succession of displacement increments each
one modelling approximately 0.1% of the initial height of the specimens. No remeshing operations were performed, and the overall
computing time for a typical two-dimensional analysis containing 2500 elements and 2700 nodal points was below 2 min on a
standard laptop computer equipped with an Intel i7 CPU (2.7 GHz)
processor.
4. Results and discussion
4.1. Failure by fracture on free surfaces
Fig. 5 shows the experimental and nite element predicted evolution of the load with displacement for the cylindrical and anged
test specimens that are identied as c3 and f1 in Table 1.
As seen in the gure, both load-displacement curves show an
initial stage (labelled as A) characterized by a steep increase of
the forming load followed by a second stage (labelled as B) in
which the load grows at a lower rate, as deformation progresses.
The experimental load at the onset of cracking is labelled as C and
major differences between the two load-displacement curves are
only found afterwards.
The tests performed with cylindrical specimens c3 raise the
issue of crack propagation stability because the decrease in load
after the peak C is so fast that it is not possible to stop crack
propagation and restart at will (Fig. 5a). The difference between

nite element estimates and experimental values after C can be

explained by the fact that numerical simulation was exclusively
focused on nding the instant and location where cracks were
triggered instead of being focused on modelling the propagation
of cracks inside the specimen which would decrease the overall
strength of the specimen. This explains the reason why there is no
decrease in the nite element predicted load after reaching a tool
displacement corresponding to point C.
In contrast, the experimental evolution of the load with displacement for the anged test specimen f1 presents a near
constant value from point C to D followed by a rapid decrease
after reaching D (Fig. 5b). The onset of cracking is located at the
ange and takes place at an amount of displacement correspondent
to point C.
Cracks are opened by tension (mode I) and the near plateau in
load from C to D results from competition between the reduction
in load caused by crack propagation by tension inside the ange,
and the increase in load caused by strain hardening and by the
growing contact area between the end surfaces of the inner cylindrical core and the parallel compression platens. The ange ensures
the necessary constraints for crack stability in the sense that propagation can be stopped and restarted, and the inner cylindrical core
provides the conditions for transition and subsequent crack opening by out-of-plane shear at point D. Under these conditions, point
D corresponds to the onset of crack opening and propagation by
mode III but it is worth mentioning that at this stage of deformation
the cracks in the ange are already fully propagated.
Crack opening modes and associated transitions can be further
understood by analyzing nite element estimates of accumulated
damage according to the normalized Cockcroft and Latham (1986)
DCL and stress triaxiality based (McClintock, 1968) DST damage
functions,


DCL =
0

1
d ,



DST =
0

m
d


(3)

where  1 is the largest principal stress,  m is the mean stress,  is

the effective stress and d is the increment of effective strain.
Fig. 6 shows the distribution of accumulated damage at the onset
crit and Dcrit ) that were determined by nite element
of fracture (DCL
ST
analysis. Results do not account for negative terms in the integrals
of the damage functions (3) because the accumulation of negative
material damage is associated to material healing, which is a concept that only makes sense to be taken into consideration if nite
element modelling is performed with coupled damage approaches.
The onset of fracture for the cylindrical c3 and anged f1 test
specimens corresponds to point C in Fig. 5a and b and to similar
locations in the experimental evolution of the load with displacement in the case of tapered t1 and t2 test specimens.
As seen in Fig. 6a, the largest value of accumulated damage at
crit occurs in the region of the cylindrical c3
the onset of fracture DCL
test specimen where cracks were triggered by out-of-plane shear
(mode III). The agreement is also good for the tapered t1 test specimen but becomes fuzzy in case of the tapered t2 and anged f1
test specimens that fail by tension (mode I) at the outer diameter and ange before propagating into the core and experiencing
transitions to out-of-plane shear opening modes (mode III).
In contrast, there is no accumulation of damage based on
stresstriaxiality in the case of the cylindrical c3 test specimens
and the agreement between the largest value of accumulated damcrit and the locations where cracks are
age at the onset of fracture DST
triggered only become reasonable in case of tapered t2 and anged
f1 test specimens (Fig. 6b).
The experimental and nite element strain loading paths resulting from the entire set of bulk formability tests that are listed in
Table 1 are shown in Fig. 7a. As seen, the limiting fracture strain

C.M.A. Silva et al. / Journal of Materials Processing Technology 215 (2015) 287298

293

Fig. 6. Finite element estimates of accumulated damage at the onset of fracture for cylindrical c3, tapered t1 and t2 and anged f1 test specimens according to (a)
normalized CockcroftLatham and (b) McClintock stresstriaxiality based ductile fracture criteria.

pairs on the free surfaces of the specimens where cracks are triggered fall on a bilinear fracture locus consisting of two straight lines
of slopes 1/2 and 1 related to crack opening by out-of-plane
shear (mode III) and by tension (mode I). The transitions from crack
opening mode I to mode III that were observed in the bulk formability tests performed with tapered and anged specimens correspond
to strain loading paths in the region of the light grey circle imprinted
at the intersection between the two linear loci (Fig. 7a).
The transformation of the bilinear fracture locus from the principal strain space to the space of equivalent strain to fracture and
stress triaxiality is shown in Fig. 7b. The hyperbolic curve corresponding to fracture locus in mode III was determined by tting
the points resulting from combinations of the strain pairs at fracture, which are almost identical to those obtained by nite elements
(refer to Fig. 7a), with the corresponding values of stress triaxiality that were obtained by nite element analysis. This procedure
avoided assumptions on proportional loading and allowed plotting
the hyperbolic black solid curve that is shown in Fig. 7b.
The hyperbolic grey solid curve corresponding to fracture locus
in mode I was determined by means of an alternative analyticalexperimental procedure that assumes proportional, isotropic strain
loading paths ( = dz /d = z / ) at the free surface locations of the
specimens where cracks are triggered.
The proposed procedure makes use of the Y-intercept where the
experimental fracture locus in mode I crosses the  -axis (Fig. 7a)

and the main reason for not using a procedure similar to that
employed in the hyperbolic curve corresponding to fracture locus
in mode III was due to the fact that the number of available fracture
points corresponding to crack opening by mode I was below critical to perform an adequate curve tting. In fact, only two points
corresponding to tapered t2 and anged f1 test specimens are
available.
Under these circumstances, by expressing the effective stress 
and the increment of effective strain d under plane stress conditions ( r = 0) as follows,


 =

2 + z2  z

2
d =
3


d2 + d2z + d dz

(4)

and subsequently applying the LevyMises constitutive equations

as proposed by Martins et al. (2014), it is possible to determine the
effective strain and stress triaxiality m / as a function of the
slope as follows,
2 
=
1 + + 2 1 ,
3

(1 + )
m
= 

3 1 + + 2

(5)

Then, by combining the above equations with the experimental value of the Y-intercept in Fig. 7a, one obtains the following

294

C.M.A. Silva et al. / Journal of Materials Processing Technology 215 (2015) 287298

where f and zf are the circumferential and axial strain pairs at

fracture, respectively. Eq. (6) is plotted in Fig. 7b as the grey hyperbolic curve labelled as Mode I.
4.2. Transitions between crack opening modes

Fig. 7. Failure by fracture in aluminium AA2030-T4. Fracture loci in the principal

strain space; fracture loci in the space of equivalent strain to fracture and stress
triaxiality.

hyperbolic variation of the effective strain at fracture f with the

stress triaxiality m / for crack opening by tension (mode I),
f =

2
2 Y
(f + zf ) =
3 m /
3(m /)

(6)

Fig. 8 shows crack opening by mode III in the cylindrical c3 test

specimen and the transitions from crack opening mode I (tension)
to crack opening mode III (out-of-plane shear) in tapered t2 and
anged f1 test specimens.
The leftmost white arrow in the magnication of the tapered
t2 test specimen indicates the transition from mode I to mode III
as the crack propagates from the equatorial region into the core of
the specimen.
The white arrow in the magnication of the anged f1 test specimen indicates the region of the ange where cracks are triggered
by circumferential tensile stresses. Some of these cracks will propagate throughout the ange (as can be seen in the fully propagated
crack that is located sideways) before developing into the core of
the specimen. It can, therefore, be concluded that cracks propagate
by tension (mode I) in the outer region of the anged test specimens before propagating by out-of-plane shear (mode III) in the
core.
The transitions from crack opening mode I to mode III that were
observed in the tapered and anged f1 test specimens were further
investigated by analysing the morphology of the cracked surfaces in
the light of three-dimensional nite element modelling. The result
is shown in Fig. 9 for the anged f1 test specimen.
Observation of the surfaces of fracture in Fig. 9a allowed the
identication of a brous pattern with dimples as a result of cracking in opening mode I due to tensile circumferential stresses acting
in the ange normal to the plane of the crack, and an inclined
pattern characterized by extensive slip along planes on which the
effective strain rate is maximum as a result of cracking in opening
mode III due to shear stresses.
The nite element simulation of the anged test specimen
shown in Fig. 9b and c was carried out until the end of crack
opening by tension (mode I). In other words, up to the instant
before cracks develop into the core of the specimen by out-of-plane
shear (mode III). The simulation made use of an element deletion
(removal) technique based on the critical value of the stress triaxiality (McClintock, 1968) DST damage function (refer to Fig. 6b) in
order to account for stress relaxation.

Fig. 8. Cylindrical c3 test specimen showing a crack opening mode III (out-of-plane shear) and tapered t2 and anged f1 test specimens showing transitions from crack
opening mode I to mode III. The black arrows stand for magnication, the leftmost white arrow indicates transition from crack opening mode I to mode III in case of the
tapered t2 test specimen and the rightmost white arrow indicates the location of crack initiation in case of the anged f1 test specimen.

C.M.A. Silva et al. / Journal of Materials Processing Technology 215 (2015) 287298

295

Fig. 9. Failure of the anged f1 test specimen under crack opening modes I and III. (a) Photographs showing the surfaces of fracture corresponding to opening modes I and
III. (b) Three dimensional nite element estimate of accumulated damage according to McClintock stresstriaxiality based ductile fracture criteria at the onset of fracture
(mode I). The white arrow indicates the location of the crack initiation. (c) Three dimensional nite element estimate of the effective strain rate (s1 ) at the onset of cracking
by out-of-plane shear (mode III) into the core of the specimen after nishing crack propagation by tension (mode I) in the ange.

As seen in Fig. 9b, the rst elements to be removed are placed in

a region similar to the location of crack initiation that was observed
in the experiments (refer to the magnication of the anged f1 test
specimen in Fig. 8).
The result in Fig. 9c not only corroborates the propagation path
of the cracks in the ange by tension (mode I) as the distribution
of effective strain rate allows anticipating the location of the shear
planes where cracks are subsequently triggered by out-of-plane
shear (mode III) into the core. This last result conrms the link
between ductile damage functions based on McClintock (1968) and
Cockcroft and Latham (1986) and crack opening modes I and III that
was proposed by Martins et al. (2014).
4.3. Failure by fracture under three-dimensional stress states
Fig. 10 shows the experimental and nite element predicted
evolution of the load with displacement for the ring expansion with
a stepped conical punch performed with test specimens identied
as r2 and r6 in Table 2.
The nite element estimates were obtained using the law of constant friction  f = mk with a friction factor m = 0.8 in order to match

the experimental results. The explanation for the actual value of

the friction factor being much larger than that previously determined by means of the ring compression test (m = 0.2) is attributed
to differences in deformation and contact pressure of ring compression and ring expansion with a stepped conical punch. In case of
ring compression testing, material is subjected to surface expansion
under moderate pressures whereas in case of ring expansion, material is subjected to very high pressures and little surface expansion
along the contact interface with the stepped conical punch. This
leads to near-sticking friction conditions and to a much larger than
expected value of the friction factor m that had been previously
determined by means of ring compression testing.
As seen in Fig. 10a and b, both load-displacement curves show
a monotonic increase of the load during the rst stage (labelled as
A) followed by a noticeable change of slope and a decrease of the
load growth rate as a result of cracks being triggered at the inner
conical surface edge of the rings during the second stage (labelled
as B). The white arrows in the pictures indicate several cracks that
were triggered at the inner conical surface edge of the rings and the
inclined propagation path of these cracks is typical of opening by
out-of-plane shear stresses (mode III).

296

C.M.A. Silva et al. / Journal of Materials Processing Technology 215 (2015) 287298

Fig. 10. Experimental and nite element predicted evolution of the load with displacement for ring expansion with a stepped conical punch performed in test
specimens that are identied as r2 and r6 in Table 2. Details show cracks at an
intermediate stage of deformation.

In connection to this, it is worth mentioning that the abovementioned decrease in the load growth rate at point B does not
occur for the nite element predicted evolution of the load with displacement because the two-dimensional numerical models used in
this investigation do not include crack triggering and propagation
features.

The third stage (labelled as C) is characterized by a second

monotonic increase of the load up to a peak value labelled as D.
During the third stage, propagation of cracks is stable and facilitates
ring expansion by means of the stepped conical punches (refer to
the photographic details in Fig. 10a and b). Once the peak value D
is reached, cracks propagate very fast and the load experiences an
abrupt drop.
Fig. 11 presents the nite element distribution of accumulated
crit and Dcrit ) corresponding
damage at the onset of fracture (DCL
ST
to region B in Fig. 10a and b, for the ring expansion specimens
identied as r2 and r6 in Table 2. As seen in Fig. 11a, the largest
crit occurs
value of accumulated damage at the onset of fracture DCL
simultaneously at the inner conical surface edge (location U in
Fig. 3b) and at the transition between the inner conical and cylindrical surfaces (location T in Fig. 3b) of the specimens. The results
obtained for location U are in good agreement with the experimental observations because all the cracks are triggered at the inner
conical surface edge of the specimens but the results obtained for
location T are not correct. This conclusion is further strengthened
by analysing the associated loading paths in the space of equivalent
strain to fracture and stress triaxiality that is shown in Fig. 12a. In
fact, all the loading paths corresponding to locations U of the ring
expansion specimens that are listed in Table 2 intersect the hyperbolic fracture locus in Fig. 12a for different magnitudes of stress
triaxiality. In the case of ring expansion specimens identied as r2
and r6 the aforementioned intersection occurs for a given amount
of punch displacement corresponding to region B in Fig. 10a and
b. This result agrees with the experimental observation that cracks
are triggered at the inner conical surface edge and propagate downward and into the material as the punch is pushed down in order to
expand the ring test specimens (refer again to photographic details
in Fig. 10a and b).
In contrast, the strain paths corresponding to locations T
(refer to T r2 and T r6 in Fig. 12a) fail to intersect the fracture locus and are located towards high negative stress triaxiality
values.
crit at the onset
The largest value of accumulated damage DST
of fracture occurs at the transition between the inner conical

Fig. 11. Finite element estimates of accumulated damage at the onset of fracture for ring expansion with a stepped conical punch (specimens r2 (top) and r6 (bottom) of
Table 2) according to (a) normalized CockcroftLatham and (b) McClintock stresstriaxiality based ductile fracture criteria.

C.M.A. Silva et al. / Journal of Materials Processing Technology 215 (2015) 287298

Fig. 12. Failure by fracture in ring expansion with a stepped conical punch (locations
U and T). Fracture loci in the space of equivalent strain to fracture and stress
triaxiality; fracture loci in the principal strain space.

and cylindrical surfaces of the ring expansion specimens (location T in Fig. 11). This result is different from experimental
observations and may be attributed to the fact that cracks are
not opened by tension (mode I) but, instead, by out-of-plane
shear (mode III). In fact, only the thinnest ring test specimen
(with 1 mm wall thickness and identied as r1) seems able
to crack under opening mode I due to its loading path in
the space of equivalent strain to fracture and stress triaxiality
(Fig. 12a). The result obtained for the thinnest ring test specimen r1 is also consistent with the radial crack propagation path
that was observed along the wall thickness of the deformed test
specimens.
Fig. 12b shows the strain loading paths corresponding to locations U in the principal strain space for the entire set of ring
expansion specimens that are listed in Table 2. As seen, there is no
correlation between the fracture locus determined from conventional bulk formability tests and the strain loading paths resulting
from ring expansion with a stepped conical punch. This is because
the straight lines with slopes 1 and 1/2 corresponding to
modes I and III require cracks to be triggered on free surfaces
under plane stress loading conditions. In fact, ring expansion with a
stepped conical punch or open die forging or extrusion fail by fracture under three-dimensional loading and different magnitudes of
stress triaxiality.
As a result of this, the principal strain space and the fracture
forming limit diagrams cannot be utilized to analyze these bulk
forming processes. Instead, the space of equivalent strain to fracture and stress triaxiality should be employed.
5. Conclusions
Failure by fracture in bulk metal forming was analyzed
by combining the fundamental concepts of plasticity, ductile

297

damage and crack opening modes of fracture mechanics. Experiments with aluminium AA2030-T4 combined with nite element
modelling and analysis of the morphology of the fracture surfaces showed that bilinear fracture loci promotes transitions
between tension and out-of-plane shear opening modes as cracks
propagate from the outer surfaces to the inward volume of the
specimens.
The new proposed ring expansion test performed with a stepped
conical punch proved adequate to replicate three-dimensional
stress loading under a wide magnitude of stress triaxiality by varying the wall thickness of the specimens. The resulting loading paths
in the space of equivalent strain to fracture and stress triaxiality showed excellent agreement with the fracture loci obtained
by means of independent bulk formability tests performed with
cylindrical, rotated cylindrical, tapered and anged specimens.
This points to the suitability of the space of equivalent strain
to fracture and stress triaxiality to model the onset of internal
cracks in bulk forming parts subjected to three-dimensional stress
states.
In contrast, the principal strain space and the fracture forming
limit diagram are not suitable to investigate the onset of cracking
under three-dimensional stress loading conditions because they
require cracking to occur on free surfaces under plane stress loading
conditions.
Ductile damage criteria also proved adequate to reveal the location and the amount of deformation for a crack to be triggered
in conventional bulk formability tests performed with cylindrical, rotated cylindrical (CockcroftLatham), tapered and anged
(McClintock) specimens but provides unclear and dubious predictions in the case of ring expansion tests. This justies the need to
further analyze the accuracy, validity and reliability of the space of
equivalent strain to fracture and stress triaxiality to provide information of the onset of cracking in case of open die forging and
extrusion, among other processes.
Acknowledgment
The work was partially supported by the Portuguese Foundation for Science and Technology under the research contract
PEst-OE/EME/LA0022/2011.
References
Atkins, A.G., Mai, Y.W., 1985. Elastic & Plastic Fracture. Chichester, Ellis Horwood,
UK.
Bai, Y., Wierzbicki, T., 2010. Application of extended MohrCoulomb criterion to
ductile fracture. Int. J. Fract. 161, 120.
Bao, Y., Wierzbicki, T., 2004. A comparative study on various ductile crack formation
criteria. J. Eng. Mater. Technol.Trans. ASME 126, 314324.
Bruschi, S., Altan, T., Banabic, D., Bariani, P.F., Brosius, A., Cao, J., Ghiotti, A.,
Khraisheh, M., Merklein, M., Tekkaya, A.E., 2014. Testing and modelling of material behaviour and formability in sheet metal forming. CIRP Ann. Manuf. Technol.
63, 727749.
Cockcroft, M.G., Latham, D.J., 1986. Ductility and the workability of metals. J. Inst.
Metals 96, 3339.
Erman, E., Kuhn, H.A., Fitzsimons, G., 1983. Novel test specimens for workability
testing. In: Compression Testing of Homogeneous Materials and Composites
ASTM STP 808. ASTM International, Baltimore, USA, pp. 279.
Gouveia, B.P.P.A., Rodrigues, J.M.C., Martins, P.A.F., 1996. Fracture predicting in bulk
metal forming. Int. J. Mech. Sci. 38, 361372.
Kobayashi, S., 1970. Deformation characteristics and ductile fracture of 1040 steel
in simple upsetting of solid cylinders and rings. J. Eng. Ind. Trans. ASME 92,
391398.
Kudo, H., Aoi, K., 1967. Effect of compression test condition upon fracturing of a
medium carbon steel. J. Jpn. Soc. Technol. Plast. 8, 1727.
Kuhn, H.A., Lee, P.W., Erturk, T., 1973. A fracture criterion for cold forming. J. Eng.
Mater. Technol. Trans. ASME 95, 213218.
Li, Y., Luo, M., Gerlach, J., Wierzbicki, T., 2010. Prediction of shear-induced fracture
in sheet metal forming. J. Mater. Process. Technol. 210, 18581869.
Martins, P.A.F., Bay, N., Tekkaya, A.E., Atkins, A.G., 2014. Characterization of fracture
loci in metal forming. Int. J. Mech. Sci. 83, 112123.
McClintock, F.A., 1968. A criterion for ductile fracture by the growth of holes. J. Appl.
Mech. Trans. ASME 35, 363371.

298

C.M.A. Silva et al. / Journal of Materials Processing Technology 215 (2015) 287298

Muscat-Fenech, C.M., Arndt, S., Atkins, A.G., 1996. The determination of fracture
forming limit diagrams from fracture toughness. Sheet Metal 1996. In: Proceedings of the 4th International Conference, University of Twente 1, The
Netherlands, pp. 249260.
Nielsen, C.V., Zhang, W., Alves, L.M., Bay, N., Martins, P.A.F., 2013. Modeling of Thermo-Electro-Mechanical Manufacturing Processes with

Applications in Metal Forming and Resistance Welding. Springer-Verlag,

London, UK.
Tomkins, B., Atkins, A.G., 1981. Crack initiation in expanded fully plastic thick-walled
rings and rotating discs. Int. J. Mech. Sci. 23, 395412.
Vujovic, V., Shabaik, A.H., 1986. A new workability criterion for ductile metals. J. Eng.
Mater. Technol. Trans. ASME 108, 245249.