You are on page 1of 19

International Journal of Impact Engineering 108 (2017) 370388

Contents lists available at ScienceDirect

International Journal of Impact Engineering


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

Quasi-static and high strain rate fracture behaviour of Ti6Al4V


TagedPD1X XP. VerleysenD2X X*, D3X XJ. PeirsD4X X
TagedPDepartment of Materials Science and Engineering, Faculty of Engineering and Architecture, Ghent University, Ghent, Belgium

TAGEDPA R T I C L E I N F O TAGEDPA B S T R A C T

Article History: Static and dynamic experiments on sheet and bar titanium Ti6Al4V alloy are presented. Aiming at a compre-
Received 15 December 2016 hensive set of data on the materials damage and failure properties, different sample geometries and loading
Revised 28 February 2017 modes are considered. For the Ti6Al4V sheet, tensile loading of in-plane shear, two different in-plane tensile
Accepted 1 March 2017
and plane strain samples are used, for the bar, tensile loading of cylindrical samples and torsion of thin-
Available online 2 March 2017
walled tubes. High strain rates, up to 1500 s¡1, are achieved in split Hopkinson bar tensile and torsion set-
ups. The evolution of the surface strain fields in the samples is obtained by a digital image correlation tech-
TagedPKeywords:
nique. Insight is gained in the underlying fracture mechanisms by post mortem evaluation of fracture
Fracture
Ti6Al4V
strains, fracture surface morphologies and voids in the material adjacent to the fracture surfaces, using opti-
Strain rate cal and SEM micrographs. Further analysis and interpretation of the test results is supported by finite ele-
Triaxiality ment (FE) simulations which give valuable information on the stress and strain state at and in the vicinity of
Damage the fracture location, including the loading path up to the onset of fracture. In the FE simulations, strain rate
and temperature dependent hardening is taken into account with the JohnsonCook hardening law. Two
models are used to describe the fracture behaviour of the Ti-alloy: for the sheet the phenomenological John-
sonCook damage initiation criterion combined with an energy-based ductile damage law, for the bar a
physically based Gurson-type failure model. Although, neither of the models is able to capture the full com-
plexity of the fracture process, the FE simulations provide key information for an in-depth understanding of
the fracture behaviour of Ti6Al4V.
In all samples and test conditions, a ductile fracture is observed which is mainly affected by the stress triaxi-
ality. At the lower triaxialities, the process of void nucleation and growth is abruptly ended by strain locali-
zation in narrow bands. The effect of strain rate is mainly felt through the induced thermal softening which
favours strain localization.
© 2017 Elsevier Ltd. All rights reserved.

1. Introduction TagedP ccidental use (impact, failure of components, crack propagation...)


a
or during various production processes (machining, punching, deep
TagedPTi6Al4V is the most widely used titanium alloy, due to the excel- drawing, incremental forming, cutting, ...). Because both the plastic
lent combination of low weight, good mechanical properties and and failure behaviour of titanium are significantly affected by the
resistance to hostile service conditions such as extreme tempera- strain rate, knowledge of the high strain rate behaviour is indispens-
tures and corrosive environments. The mechanical behaviour of able for users and manufacturers of titanium components. Therefore,
Ti6Al4V is determined by the multi-phase microstructure consisting it is not surprising that the scientific community has exhaustively
of a harder (and richer) b-phase, which gives its strength, and a studied the strain rate and temperature dependent mechanical
softer a-phase, which provides workability [1]. The a- and b-phases behaviour of Ti6Al4V [114]. Many of these studies report on test
act as incoherent phases, which results in a strengthening effect. So, programs aiming at the characterisation of plastic flow properties
these alloys have a high density of a/b-boundaries which are bar- using uniaxial tensile or compression, or torsion experiments
riers to slip transmission. On the other hand, the a/b-boundaries are [1,35,8,9]. Obviously, in between the lines, data can be found on
sites of strain accumulation and strain incompatibility, and might be the fracture behaviour, however, fracture is never systematically
preferential sites for void nucleation. analysed. In [2,6,7,1113] on the contrary, fracture of TI6Al4V is
TagedPTi6Al4V is used in the industry in applications such as aircraft addressed explicitly. However, the focus lies quasi-exclusively on
components, turbine blades or sports equipment. In these applica- the phenomenon of adiabatic shear banding. Up to the authors
tions, the material can be subjected to dynamic loads during knowledge, a broader, systematic study into the damage and frac-
ture behaviour of Ti6Al4V is still lacking.
*
Corresponding author.
TagedPAt the origin of fracture in ductile materials lies a process of void
E-mail address: Patricia.Verleysen@UGent.be (P. Verleysen). initiation, growth and coalescence. Plastic instabilities, caused by

http://dx.doi.org/10.1016/j.ijimpeng.2017.03.001
0734-743X/© 2017 Elsevier Ltd. All rights reserved.
P. Verleysen and J. Peirs / International Journal of Impact Engineering 108 (2017) 370388 371

TagedPlocalized plastic flow or damage softening [15], play an important tTagedP ests on sheet Ti6Al4V, and tensile and torsion tests on bar Ti6Al4V.
role in this process, because these instabilities generally involve one Post mortem, the fractured samples are thoroughly studied by light
or more shear bands which are a precursor to rapid failure. The fail- optical (LOM) and scanning electron microscopy (SEM). Next to the
ure process, including fracture, is determined by material properties fracture strain, also the fracture morphology is studied. Details are
and loading conditions. The latter includes imposed deformation given on the presence, size and distribution of voids in the material
and/or loading paths, temperature, strain rate and load repetitions adjacent to the fracture surface of all samples. Additionally, to come
[1619]. To evaluate the effect of the stress state on ductile damage to an in-depth interpretation and analysis of the test results, finite ele-
and fracture, two scalar descriptors are commonly used: the stress ment (FE) simulations are performed which give detailed information
triaxiality and Lode parameter (or Lode angle). While the triaxiality on the stress and strain distribution close to the fracture, including
is a measure for the mean normal or hydrostatic stress, the Lode the loading path up to the onset of fracture. In the FE model, J2-plas-
parameter is related to the position of the stress state in the devia- ticity with the JohnsonCook strain rate and temperature dependent
toric plane of the principal stress space [20]. The effect of hydrostatic hardening law [10,33] is used to describe the yield behaviour of
pressure on both plastic flow and fracture, was already recognised in Ti6Al4V. Two models are used for the fracture behaviour: 1) the phe-
the ’50s of previous century by Bridgman [21]. Based on void growth nomenological JohnsonCook damage initiation criterion combined
analysis, McClintock [22] and Rice [23] came to an exponential with an exponential damage evolution law for the sheet material
decrease of the fracture strain as a function of triaxiality. Since this [30,34] and 2) a physically based Gurson-type failure model in which
pioneering work, more complex and material specific relations growth and coalescence of voids is considered for the bar material
between fracture strain and triaxiality have been developed. For [29,35]. Based on simulations a link is establish between the fracture
some materials, the effect of triaxiality is not monotonic over a wide characteristics at the one hand and the stress states and strain rates at
range of stress triaxialities. For aluminium 2024-T351 for instance, the other. Unavoidably, test results are affected by the intrinsic behav-
Bao et al. [24] distinguished three triaxiality regions, each governed iour of the test material on the one hand, and structural effects related
by a different fracture mechanism, for which different relationships to sample geometry, boundary conditions and loading mode on the
between triaxiality and fracture strain are proposed. The apparent other hand. The simulations are used to distinguish between both.
non-monotonic relationship between the fracture strain and the
stress triaxiality is often indicative for a dependency of the Lode 2. Experimental techniques
parameter. Therefore, fracture loci have been developed taking both
triaxiality and Lode parameter into account [25]. Tests imposing pro- 2.1. Material and test samples
portional loading conditions, and consequently a constant triaxiality
and Lode parameter, to the material significantly simplify test analy- TagedPThe investigated material is Ti6Al4V (ASTM Grade 5) provided by
Ò
sis [26]. However, when high strain gradients arise due to plastic TIMET (Toronto, Canada) in two different product forms: bulk
instabilities, triaxiality and Lode parameter might deviate from their material (Timet, UK) and sheet material (Timet, USA). The bulk mate-
proportional values and fracture parameters can no longer be rial comes in the form of bars, obtained by a semi-continuous extru-
straightforwardly correlated with the stress state. sion process, in mill-annealed condition, with a diameter of
TagedPMany studies can be found linking the triaxiality and Lode param- 16 mm § 0.18 mm. The sheets, obtained by cold-rolling, come also in
eter to fracture strain, absorbed energy and fracture morphology. mill-annealed condition and have a thickness of 0.6mm § 0.06 mm.
Special attention is paid to whether or not a slant fracture surface is Although the bulk and sheet material are both classified as TIME-
Ò
obtained, because slant fractures detrimentally affect the energy TAL 6-4, they do not have exactly the same chemical composition
absorbed during failure. Although, slant ductile rupture is a com- nor the same microstructure. Table 1 shows the chemical composi-
monly reported failure mode, currently there are no comprehensive tion reported by the manufacturer for each of the products [36,37],
fracture criteria for quantitative prediction of slant fracture and as well as the limits defined in the AMS-4911 standard [38]. A differ-
many slant fracture issues still remain unresolved [2729]. Next to ent behaviour of the bulk and sheet material is expected, because
by these macroscopic features, a ductile fracture is also characterised the microstructure of Ti-6Al-4V strongly depends on the production
by microscopic features, such as dimples in the fracture surface and process parameters [39]. Ti6Al4V consists of a majority of hexagonal
traces of microdamage in the material adjacent to the crack, such as close packed a-phase and a finely dispersed body centred cubic
voids, shear bands and microcracks [29]. These features are not only b-phase (at room temperature approximately 4vol% b). More details
the result of the ductile damage process, they also affect the process. on the microstructure, including texture, can be found in [40].
Indeed, for Ti6Al4V experimental evidence is found that the softening TagedPThe effect of the stress state on the fracture properties is investi-
effect of voids favours failure of the material by localization [14]. gated using six different sample geometries. The geometries are
TagedPAlso the effect of strain rate and temperature on fracture was designed to cover a wide range of triaxialities h and Lode parameters
found to be strongly material dependent. For OFHC copper, Armco L defined as:
iron and 4340 steel, Johnson and Cook [30] concluded that both tem- sh
perature and strain rate had a positive effect on the fracture strain. hD ð1Þ
s eq
However, in literature also data can be found indicating a different
relationship between strain rate and temperature at the one hand ð2s II ¡s I ¡s III Þ
LD ð2Þ
and fracture strain at the other. For Weldox 460E steel, for example, ðs I ¡s III Þ
a decreasing fracture strain with strain rate and hardly any effect of
the temperature below 300 °C were found [31]. For Ti6Al4V, no sys-
Table 1
tematic studies can be found focussing on the effect of strain rate on Chemical composition (weight percent) of Ti6Al4V products reported by the supplier
fracture properties of Ti6Al4V. in the accompanying test reports [37,36] and defined in the ASM-4911 standard [38].
TagedPObviously, the prediction of the fracture characteristics from the
Fe V Al C O N
material properties and loading conditions demands for a multilevel
approach based on experimental data [32]. Aiming at a better under- TIMET Sheet [36] 0.16 3.98 6.27 0.009 0.19 0.009
standing of damage and fracture of Ti6Al4V, in present contribution, a Bar [37] 0.18 4.14 6.47 0.010 0.19 0.005
Standard [38] Min.  3.50 5.50   
comprehensive set of test results is presented. The influence of load-
Max. 0.30 4.50 6.75 0.080 0.20 0.050
ing conditions is studied by performing static and dynamic experi- Y, H and remaining elements: <0.42. Remainder: Ti
ments with different stress triaxialities: shear, tensile and plane strain
372 P. Verleysen and J. Peirs / International Journal of Impact Engineering 108 (2017) 370388

pffiffiffi
TagedPWith s h the hydrostatic stress, s eq the Mises equivalent stress, and TagedPsheet sample (Fig. 1c), a triaxiality of 1/ 3 is aimed at, the Lode
s I, s II and s III the principal stresses ordered following s I  s II  s III. parameter is 0.
Important to note is that for plane stress conditions the Lode param- TagedPFor the bar material (Fig. 1, right), the thin-walled tubular sample
eter is a function of the stress triaxiality; under axisymmetric load- (Fig. 1d), when deformed in torsion, is designed to obtain homoge-
ing the Lode parameter is either ¡1 or C1 [15]. neous stresses and strains in the gauge section, with a triaxiality and
TagedPTo characterise the basic, static properties of the sheet Ti6Al4V, Lode parameter 0 [43]. In the cylindrical bar tensile sample
tensile tests are conducted on a geometry based on the ASTM-E8 (Fig. 1e) again very soon after the onset of plastic deformation triax-
standard with a length of the parallel gauge section of 50 mm and ialities higher than 1/3 appear in the gauge section.
width of 12.5 mm [9]. This sample is denoted by GL50. The geometry TagedPThe sheet samples are cut by means of electrical discharge
guarantees that -before necking- homogeneous stresses with a triax- machining. For the tensile, plane strain and shear tests, samples are
iality of 1/3 and Lode parameter of ¡1 are reached in a sufficiently taken in the rolling direction. Tensile and torsion samples are
large region around the centre. machined out of the extruded Ti6Al4V bar along the axial direction.
TagedPThe influence of the stress state and strain rate is studied using
the geometries presented in Fig. 1. The three sheet samples (Fig. 1, 2.2. Experimental programme
left) are subjected to a tensile deformation. The in-plane shear sam-
ple geometry (Fig. 1a) is optimized for Ti6Al4V aiming at a homoge- TagedP2.2.1. Mechanical tests
neous strain and triaxiality in the shear zone [8,41]. Key feature is TagedPThe experimental programme consists of static and high
the eccentric position of the notches which gives rise to an almost (5001500 s¡1) strain rate tests at room temperature. The static
Ò
pure shear stress state up to large strains, i.e. triaxiality and Lode tests are carried using a universal Instron test machine with a
parameter 0. In the in-plane sheet tensile sample (Fig. 1b) on the 50 kN loadcell (Fig. 2a). Split Hopkinson bar (SHB) tensile [44] and
contrary, starting from their respective uniaxial tensile values of 1/ torsion setups [45] are used for the dynamic experiments. The small
3 and ¡1, the triaxiality and Lode parameter in the centre very rap- dimensions of the samples in Fig. 1 guarantee quasi-static equilib-
idly increase during plastic deformation [8,42]. In the plane strain rium from the early stage of loading in the dynamic experiments. To

Fig. 1. Overview of sample geometries used for static and high strain rate experiments. Left, the three sheet samples are presented denoted as in-plane shear (a), in-plane tensile
(b) and plane strain (c), right the two bulk samples torsion (d) and cylindrical tensile (e). Samples (a), (b), (c) and (e) are loaded in tension, sample (d) in torsion. The shaded zones
are used to glue the sheet samples in the setup.
P. Verleysen and J. Peirs / International Journal of Impact Engineering 108 (2017) 370388 373

TagedPexclude effects related to geometry or dimensions of the samples TagedPmeasurements the standard deviation rises up to a strain of 7%. For
[42,46], the samples presented in Fig. 1 are used for both the static the cylindrical samples, the standard deviation is limited to 2.7%.
and dynamic experiments. TagedPFor the in-plane shear and torsion samples, the fracture strain is
TagedPFor the dynamic tests, the sheet samples are glued in slots made calculated based on the grain morphology: the shape, or more spe-
in the Hopkinson bars, the bar tensile samples are screwed between cifically the orientation, of the elongated grains is used to quantify
the Hopkinson bars. To have the same boundary conditions in the the shear strain [50]. Indeed, as schematically presented in Fig. 3a,
static and dynamic tests, in the static tests the samples are also glued during a simple shear deformation the tensile strain axis rotates.
in slots of or screwed between short bars with the same diameter Denoting the angle between the major strain axis and the shear ori-
and material as the Hopkinson bars. A typical setup for the static entation by u, the relation between the engineering shear strain
experiments can be seen in Fig. 2b where an in-plane tensile sample g D tan(a) and u can be written as (Fig. 3b):
Ò
in the Instron test bench is presented. The torsion samples have !
hexagonal flanges which fit in sockets machined in the Hopkinson 2
u D arctan pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ð3Þ
bars. 4 C g 2 ¡g
TagedPDuring the static experiments the sample deformation is mea- TagedPFor small strains u is 45°. The method is applied to a SEM image of
sured using three LVDTs, see Fig. 2. Additionally, a speckle pattern is an in-plane shear and torsion sample as shown in respectively Fig. 3c
applied to the sample surface which is recorded during the tests and d. Close to the fracture shear strains of g D 0.50.7 are found
with two cameras: one captures images at a relatively high spatial corresponding with equivalent strains between 0.28 and 0.37 or log-
resolution (1024*1024), but at a low framerate (1 fps), the other arithmic shear strains between 0.24 and 0.32 according to [41]:
camera is a high speed camera that captures the deformation during  pffiffiffiffiffiffiffiffiffiffiffiffiffiffi  pffiffiffiffiffiffiffiffiffiffiffiffiffiffi
the short period the sample is failing at a high framerate (between ln 2 C g 2 C g 4 C g 2 ¡ ln 2 C g 2 ¡g 4 C g 2
25,000 and 70,000 fps), but with a much lower resolution (typically ðe12 ÞLog D pffiffiffiffiffiffiffiffiffiffiffiffiffiffi ð4Þ
2 4 C g2
512*128).
TagedPFor the dynamic tests, Kolsky's theory is used to obtain the force- TagedPThe size and shape of the dimples in the fracture surfaces are
displacement or torque-rotation histories of the samples respec- studied by SEM images. Two different techniques are used to esti-
tively subjected to a tensile [44] or torsional [47] loading. As for the mate the size of the dimples: 1. Jeffries planimetric method recom-
static tests, a speckle pattern is applied to the sample which is mended by ASTM E112 [51] for which the dimples within a circle of
recorded by high speed cameras. The images are processed by the 0.5 mm2 are manually counted and 2. automatic dimple detection
Ò
commercial digital image correlation (DIC) software MatchID to and size calculation by the image processing software ImageJ. SEM
obtain the evolving strain field in the samples during the tests [48]. micrographs with a resolution <0.5 mm are used to characterize the
Close to the fracture large strain gradients occur which give rise to voids in the material close to the fracture surfaces.
heavily distorted speckle patterns. As a result, the obtained DIC
strain might underestimate the actual strain value. 3. Numerical modelling

TagedP2.2.2. Post mortem fracture characterisation 3.1. Finite element model


TagedPAfter mechanical testing, the fracture surfaces are extensively
studied. The cross-section of the fracture surface Af of the cylindrical TagedPTo come to an enhanced interpretation and in-depth understand-
tensile, in-plane tensile and plane strain samples is measured ing of the test results, coupled thermo-mechanical finite element
according to the ASTM-E8M standard to calculate the fracture strain simulations are carried out. The FE models are also used for the
ɛf using equation ɛf D lnðA0 =Af Þ, in which A0 is the initial cross-sec- inverse identification of the material model parameters following a
tion. It should be noted that the thus obtained fracture strain is an procedure similar to the one described in [8].
averaged value which ignores the large strain gradients in the frac- TagedPAll simulations are carried out using the dynamic explicit solver
Ò
ture surface [49]. Unfortunately, because of the small dimensions of provided by the finite element program Abaqus . As mentioned in
the sheet samples in the thickness direction, relatively large errors Section 2.1, in all tests quasi-static conditions prevail and modelling
on the determined fracture strain are obtained: for all fracture strain of the waves propagating in the entire SHB system can consequently

Fig. 2. Picture of the electromechanical tensile test device used for the static tests (a) and close-up of the specimen glued between bars. Three LVDTs measure the relative dis-
placement of two circular plates. A high resolution camera is positioned in front of the specimen, a high speed camera at the back.
374 P. Verleysen and J. Peirs / International Journal of Impact Engineering 108 (2017) 370388

Fig. 3. Deformation of a circle by simple shear (a), orientation of the major strain axis u in simple shear as a function of the shear strain g D tan (a) (b), SEM image of the shear
region near the fracture from in-plane (c) and torsional (d) shear specimen after high strain rate loading. The measured angles u are indicated together with the strain g calculated
using Eq. (3).

TagedPbe omitted. The computational efficiency is furthermore improved TagedPIn which s y is the yield stress, ɛp the true equivalent plastic strain,
by exploiting the sample symmetry while maintaining the possibil- ɛ_ p the plastic strain rate, ɛ_ 0 a reference value of the strain rate, T the
ity of an asymmetric failure mode: only one half along the width of temperature, Tmelt the melting temperature and Troom the room tem-
the tensile and plane strain samples are modelled applying appropri- perature. A, B, C, n and m are material parameters. At high strain
ate symmetry conditions. To model through-thickness deformation, rates the experiments are considered adiabatic and plastic work con-
including necking, for the sheet samples, along the thickness 6 solid verted into heat results in a significant increase of the sample tem-
elements of type C3D8R (i.e. a 8-node linear brick making use of perature. At sufficiently low strain rates, heat is dissipated to the
reduced integration with hourglass control) are used instead of com- environment and isothermal conditions can be assumed. The transi-
putationally more efficient plane stress elements. For the torsion and tion between isothermal/non-isothermal and adiabatic deformation
cylindrical tensile samples, CAX4R elements (axisymmetric quadri- occurs at relatively low strain rates compared with other materials
lateral elements with reduced integration) are used. For all samples because of the low heat conductivity and mass density of the
similar element sizes are used: typically 0.1 mm in the gauge section Ti6Al4V. Sample heating is taken into account in the parameter fit-
gradually evolving to 0.15 mm in the transition zones or shoulders ting procedure using the thermal model presented in [9].
of the sample. It has been checked that the mesh size is sufficiently TagedPIn Table 2 two parameter sets -one for the sheet and one for the
small to have mesh independent stress-strain curves up to the point bar- are presented determined using the procedure described in [8].
of necking. More details on mesh densities can be found in [8]. The The thus obtained JC parameters lie within the range of values found
experimental, linear or rotational velocity histories are imposed at in [5,6,9,52].
one end of the sample. TagedPThe fracture behaviour of the Ti-sheet is modelled using the
phenomenological JohnsonCook damage initiation criterion

Table 2
3.2. Material models Elastic properties and parameters used for the JohnsonCook hardening law for
the Ti6Al4V sheet and bar material.
TagedPTo describe the mechanical behaviour of Ti6Al4V, J2-plasticity is E (GPa) n ɛ_ 0 (1 s¡1) Troom (°C) Tmelt (°C)
adopted with strain rate and temperature dependent, isotropic hard-
ening given by the JohnsonCook (JC) hardening law [33]: Sheet & bar 117 0.3 1 20 1653
A (MPa) B (MPa) n C m
h i ɛ_
 
T¡Troom m
 
Sheet 951 892 0.70 0.015 0.71
s y D A C Bnp 1 C C ln p 1¡ ð5Þ Bar 895 477 0.38 0.011 0.86
ɛ_ 0 Tmelt ¡Troom
P. Verleysen and J. Peirs / International Journal of Impact Engineering 108 (2017) 370388 375

TagedPcomplemented with a damage evolution law. For the bar material, TagedP olume fraction f*:
v
the measured void densities are taken into consideration and a !  
s 2eq q2 s h
gðs ij ; f Þ D 
¡1¡q3 ðf  Þ D 0
physically based Gurson-type failure model is used. 2
C 2q f cosh ð9Þ
s 2y 2 sy
1

TagedP3.2.1. JohnsonCook damage criterion with s y the yield stress of the material without voids, q1, q2 and q3
TagedPThe phenomenological JohnsonCook ductile damage initia- are model parameters which determine its shape [32].
tion criterion, based on the model used to predict fracture strain TagedPThe rate of void growth is obtained by assuming mass conserva-
presented in [30], has certain aspects in common with the JC tion and depends on the volume change part of the plastic strain.
hardening model. Indeed, the effect of three important parame- Consequently, there is no void growth in pure shear deformation.
ters -stress triaxiality h, strain rate and temperature- on the frac- The void nucleation depends on the equivalent plastic strain ɛp, here
ture strain is described independently in a multiplicative fashion. a normal distribution A is used:
Damage is initiated when condition (6) is met, using the equiva-
lent plastic strain at onset of damage ɛf0 under the current con- f_ D f_ growth C f_ nucl
ditions of strain rate, temperature and stress triaxiality f_
growth D ð1¡f Þ_ɛ p
calculated with expression (7):
ii
  ! ð10Þ
fN 1 ɛp ¡ɛN 2
Z f_ nucl D Aɛp with A D exp ¡
dɛp sN 2 sN
D1 ð6Þ
ɛf 0
where єN and sN are respectively the average nucleation strain and
    the standard deviation. The amount of nucleating voids is controlled
ɛ_ p T¡Troom
ɛf 0 D ½d1 C d2 expð¡d3 hÞ 1 C d4 ln 1 C d5 ð7Þ by the parameter fN.
ɛ_ 0 Tmelt ¡Troom
TagedPThe last phase in ductile fracture comprises void coalescence into
TagedPThe JC criterion given by Eqs. (6) and (7) only predicts initia- the final fracture. Here again, different models exist to predict when
tion of damage. An additional progressive material degradation voids coalesce. An often used simple model uses a critical void frac-
model is implemented which considers damage induced soften- tion fc: as long as the real void fraction is below the critical void frac-
ing of the yield stress as well as degradation of the elastic stiff- tion, the virtual f* and real void fraction f are equal, once the critical
ness proportional with (1-d). For the damage parameter d, an void fraction is reached, the virtual void fraction f*, used in Eq. (9) is
exponential damage accumulation evolution based on the energy changed into a much higher value given by Eq. (11). This causes soft-
Gf dissipated per unit area during the damage process is adopted ening of the material and accelerated growth of the void fraction f*
[53,54]: until the fracture void fraction fF is reached. At this moment the
 Z up 
s y dup material is fractured.
d D 1¡exp ¡ ð8Þ 8 qffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Gf
<f
> if f f c
0
q1 C q21 ¡q3
in which up D Lɛp is the equivalent plastic displacement defined with f D f F ¡f c with f F D ð11Þ
>
:fc C if f c < f < f F q3
the characteristic length of the element and s y the yield stress of the f F ¡f c
damaged material. Using a stress-displacement response rather than
TagedPApplication of this model requires knowledge of 9 parameters:
a stress-strain response reduces the mesh dependency of the results.
three model parameters (q1, q2 and q3), the initial void fraction f0,
The equivalent plastic displacement up is the fracture work conju-
three void nucleation parameters (єN, sN and fN), two failure parame-
gate of the yield stress s y after the onset of damage. Eq. (8) guaran-
ters (fc and fF). The value of the parameters for the Ti6Al4V bar,
tees that the energy dissipated during the failure process is equal to
together with their meaning, can be found in Table 4. These values
Gf [54].
are based on both literature data, microscopic observations and test
TagedPThe parameters d1, d2, d3, d4 and d5 needed for Eq. (7) can be
results.
found in Table 3. Starting from values for Ti6Al4V found in [34], the
parameters are optimized following an iterative scheme similar to
the one described in [8] using the static and dynamic test results for 4. Experimental observations
the in-plane shear, tensile and plane strain sheet samples. The same
values for the reference strain rate ɛ_ 0 and room temperature Troom 4.1. Sheet material
are used in Eqs. (5) and (7), see Table 2. It is found that the value of
the fracture energy Gf has an important influence on the fracture
TagedP4.1.1. Force-displacement
morphology, more specifically whether or not a slant or straight
TagedPRepresentative static and dynamic stress-strain or stress-dis-
fracture is obtained in the tensile experiments. A satisfactory agree-
placement curves for the shear, tensile and plane strain tests on the
ment between the simulations and the experiments is obtained
sheet Ti6Al4V are shown in Fig. 4. Engineering stresses are obtained
using GfD50mJ/mm2.
by dividing the measured force by the initial gauge section, the

TagedP3.2.2. GursonTvergaardNeedleman model


Table 4
TagedPIn the GursonTvergaardNeedleman (GTN) model [55], a mate- Dimensionless parameters for the GTN model for the bar Ti6Al4V.
rial with randomly distributed voids is idealised as a porous material
in which the hydrostatic part of the stress also affects yielding [56]. Parameter Meaning Source Value

A strong coupling between deformation and damage is introduced q1 Shape of yield surface Ref. [57] 1.5
[29] by a plastic potential function which is dependent on the void q2 Shape of yield surface Ref. [57] 1
q3 Shape of yield surface Ref. [57] 2.25
f0 Initial void fraction Microscopy 0.0005
Table 3
fN Void nucleation rate Optimised 0.008
Values used for the JohnsonCook damage initiation criterion for the Ti6Al4V
fc Critical void fraction Microscopy 0.02
sheet.
fF Failure void fraction Estimation 0.2
d1 d2 d3 d4 d5 eN Void nucleation strain Microscopy 0.25
SN Standard deviation on void nucleation Estimation 0.025
¡0.078 0.282 0.479 0.029 3.87 strain
376 P. Verleysen and J. Peirs / International Journal of Impact Engineering 108 (2017) 370388

Fig. 4. Representative static and dynamic stress-strain or stress-displacement curves for shear (a), tensile (b) and plane strain (c) tests on the Ti6Al4V sheet, together with the
local strain obtained by DIC. The DIC fracture strain is indicated by a dot.

TagedPaverage engineering strain in the tensile test by dividing the relative TagedP4.1.2. Fracture morphology and mechanism
displacement of the bar/sample interfaces by the initial gauge TagedPIn the static tensile experiments on the GL50 samples, diffuse
length. The local, logarithmic DIC strain in the centre of the gauge necking is followed by localized necking, as schematically repre-
section is represented on the secondary vertical axis of the graphs. sented in Fig. 5a [58]. The diffuse necking involves contraction in
From Fig. 4 it is clear that the strain rate significantly affects both both the width and thickness directions. During localised necking
stress and deformation levels. At high strain rates, engineering the sample thins without further width contraction and the material
curves of Fig. 4 all show higher yield stresses, though lower global surrounding the localised neck remains quasi-rigid. The width of the
fracture elongations. Local strain values, however, give a slightly dif- localised neck is roughly equal to the sample thickness, conse-
ferent view. Indeed, the strain is more localized in the dynamic tests: quently the total sample elongation after localised necking is very
for a certain displacement or engineering strain, a larger local strain limited [58,59]. Fracture follows soon because of the considerable
is reached at higher strain rates. As a result, for the plane strain sam- increase of the equivalent strain within the localised neck. In most
ple of Fig. 4c, even though the fracture elongation is lower at high GL50 samples, the entire fracture follows the localised band. How-
strain rates, a higher fracture strain is obtained. For the shear and ever, in few GL50 samples (Fig. 5b) and also in all the in-plane tensile
tensile experiment the local fracture strains are still lower at high samples of Fig. 1 (Fig. 5c), only at the edges the fracture surface is
strain rates. aligned with the localised neck, in the centre of the sample the frac-
TagedPIn Table 5 average values of the fracture strains for all tests are ture surface is oriented in the transverse direction. The localized
given, measured following the procedures described in Section 2.2.2. necks in Fig. 5b and c are inclined over §54.7° which corresponds
The values given in Table 5 confirm the trends observed in Fig. 4. with the angle reported for isotropic materials [58]. The majority of
the tensile samples exhibits a slant fracture in the transverse crack
and a flat fracture in the inclined cracks. In some samples, the trans-
Table 5 verse crack is V-shaped over the thickness. In [29] a list of conditions
Average equivalent fracture strains for the sheet and bar Ti6Al4V promoting slant fracture is given including (1) low strain hardening
observed under different test conditions. of the material, (2) small sample thickness and (3) high loading rates.
Sheet Bar The first two are applicable to the Ti6Al4V sheet considered here, the
last one also for the high strain rate tests.
Shear Tensile Plane strain Shear Tensile
TagedPIn Fig. 6 high speed images (framerate of 42,000 fps) are pre-
Static 0.44 0.48 0.21 0.48  sented in which the fracture initiation is captured in a static (left)
Dynamic 0.3 0.42 0.24 0.33 0.62 and dynamic (right) experiment on in-plane tensile samples. The
P. Verleysen and J. Peirs / International Journal of Impact Engineering 108 (2017) 370388 377

Fig. 5. Schematic presentation of diffuse and localised necking in a sheet tensile sample [58] (a), GL50 tensile samples after static testing (b) and small tensile samples (Fig. 1b)
after high strain rate testing (c).

TagedPhigh speed images show the two stage fracture process: first, a TagedP4.1.3. Fracture surface
transverse crack is formed in the centre of the sample, during some TagedPRegardless of the loading mode, all SEM images of the fracture
time this crack remains stable while inclined localised necks develop surfaces, clearly reveal dimples indicative of ductile fracture. In the
at both sides of the crack, finally, also the material in the localised in-plane tensile samples, the entire fracture surface is covered by
neck fails. There is no significant difference in fracture mechanism equiaxed dimples and no differences are observed between the dim-
between the statically and dynamically loaded samples. In all tensile ples in the transverse part of the fracture and in the inclined fracture
tests, fracture propagation takes less than 0.35 ms which corre- at the borders.
sponds with a crack speed higher than 5 m s¡1. TagedPIn Fig. 8 the fracture surface, including details with a higher
TagedPFracture propagation in the in-plane shear and plane strain sam- amplification, of a dynamically loaded shear sample is presented.
ples is faster than in the in-plane tensile samples. High speed camera Compared to the tensile samples, larger and elongated dimples are
measurements at a framerate of 70,000 fps show that in the shear present in the centre of the sample. Obviously, the elongated shape
tests, fracture is completed in between two frames, so within maxi- is the result from the imposed shear deformation. The larger dimple
mum 14 ms, which corresponds with a fracture propagation speed of size indicates a lower amount of void nucleation sites in shear. On
at least 200 m s¡1. The shear fracture surface is oriented in the direc- the other hand, near the left border the dimples are only slightly
tion of the load and is perpendicular to the sample plane (Fig. 7a). elongated and have approximately the same size as in the tensile
The fracture surface in a plane strain tests is always perpendicular to test. This observation supports the hypothesis that fracture initiates
the tensile direction (Fig. 7b). in the centre of gauge section. Indeed, a fracture in the centre give
TagedPAlthough, the location of fracture initiation is not experimentally rise to a dominant tensile stress state near the edges. The dimple-
identified in the shear and plane strain tests, it is reasonable to free zone in the bottom right Fig. 8 is the result of rubbing of the
assume that also in these samples fracture is initiated in the centre fracture surfaces against each other. Fig. 9 shows the fracture surface
of the gauge section. Indeed, in the shear sample, significantly higher of a statically loaded shear sample. Compared to the dynamically
strains are reached in the centre of the gauge section compared to loaded sample of Fig. 8, the dimples are smaller and less elongated.
the notches [41]. In the plane strain sample, both the plastic strain At first sight, it seems contradictory that the dimples are smaller and
and triaxiality in the centre of the gauge section are significantly less elongated in the static sample while a higher fracture strain is
higher than at the notch boundaries. established. The formation of an adiabatic shear band (ASB) in which

Fig. 6. Fracture initiation in the in-plane tensile sample presented in Fig. 1b during a static (a) and a dynamic tensile (b) test.
378 P. Verleysen and J. Peirs / International Journal of Impact Engineering 108 (2017) 370388

Fig. 7. Fracture in dynamically loaded shear sample of Fig. 1a (a) and plane strain sample of Fig. 1c (b).

TagedPlocally very high strains are reached in the high strain rate test is a TagedP ndeformed grain size, measured with Jeffries method, is 10.9 mm.
u
plausible explanation for this observation. However, since the shear The tensile and plane strain dimples are thus smaller than the
band is very narrow [10], the strain in the band is not captured by grains.
the techniques described in Section 2.2.2 used to calculate the frac-
ture strain given in Table 5. Indeed, the strain in Table 5 can be inter- TagedP4.1.4. Voids
preted as the macroscopic strain just before the shear band TagedPThe influence of strain rate and loading mode on the occurrence
initiation. and distribution of voids is studied by SEM images taken close to the
TagedPThe fracture surface of a dynamically loaded plane strain sample fracture surface. In Fig. 11 micrographs of in-plane shear, in-plane
is shown in Fig. 10. In comparison with the tensile test, the dimples tensile and plane strain samples close to the fracture surface are pre-
are less elongated and subdimples can be distinguished within the sented. Clearly, the number of voids is significantly affected by the
larger dimples. stress state obtained in the different samples: almost no voids are
TagedPIn Table 6 the measured average size and area of the dimples are observed in the shear sample, a lot of voids in the in-plane tensile
given. The smallest dimples are observed in the tensile sample, sample and a limited number in the plane strain sample.
slightly larger dimples in the plane strain sample. The largest dim- TagedPAlthough, in the shear samples, because of their elongated shape,
ples appear in the dynamic shear sample, however, smaller dimples not all voids might have been revealed by the SEM micrographs,
might be erased during deformation. It is interesting to note that the the lower amount of voids corresponds with the observed larger

Fig. 8. Fracture surface of a sheet shear sample (Fig. 1a) after dynamic loading with higher magnification images taken near the edge (bottom left), closer to the centre (bottom
centre) and rubbed region at the other edge (bottom right). The arrows indicate the shear direction; the displacement direction of the pictured specimen part is opposite to the
arrows.
P. Verleysen and J. Peirs / International Journal of Impact Engineering 108 (2017) 370388 379

Fig. 9. Fracture surface of a sheet shear sample (Fig. 1a) after static loading: near the centre (1), closer to the edge (2) and (3).

TagedPdimples in the fracture surface. Additionally, the SEM micrographs of ITagedP ndeed, in the clusters, a limited growth of the voids is sufficient for
dynamically loaded shear samples also show a narrow band below coalescence. Many examples of interacting voids are found in the
the fracture surface in which no grains can be distinguished. This micrographs of the tensile and plane strain samples. The orientation
observation supports the hypothesis that failure in the dynamic of the cracks formed by void coalescence is mostly perpendicular to
shear samples is preceded by strain localisation in an ASB. the loading direction.
TagedPNo qualitative differences between the voids occurring in the TagedPTo come an in-depth understanding of the effect of the stress
quasi-static and dynamically loaded samples are noticed. Most of state, on the SEM images voids >§0.3 mm are scanned and system-
the voids nucleate at the a/b boundaries. Many examples of hetero- atically analysed, using the image processing software ImageJ. The
geneously distributed voids are found. In Fig. 12 clusters of 4 or shear tests are not considered in the analysis, because the limited
more voids within an area of approximately 10 mm are marked. number of voids in these tests does not allow a reliable statistical
These voids nucleate at one or two grain boundaries. The heteroge- interpretation. Values for the void fraction, number of voids per area
neous void distribution might encourage early void coalescence. unit and average void size within a distance of 100 mm to the frac-
ture surface are presented in Fig. 13a for in-plane tensile versus
plane strain tests and in Fig. 13b for static versus dynamic samples.
TagedPAn average void fraction of §0.5% is reached in the 100 mm zone
adjacent to the fracture surface of the tensile samples. As is already
clear in Fig. 12, in the plane strain samples the void fraction, and
since the average void size is very similar to the one in the tensile
samples, also the number of voids is much smaller. The strain in the
plane strain sample at fracture is approximately half the value in
tensile samples. Between the static and dynamic tests only limited
differences in void characteristics are found, see Fig. 12b: the num-
ber of voids is slightly lower in the dynamic tests, though the

Table 6
Average dimple size dav in the fracture surfaces of the dynamic shear, in-plane
tensile and plane strain samples measured with Jeffries planimetric method and
ImageJ.

Shear  centre Shear  edge Tensile Plane strain

dav (mm) Jeffries 14 7.7 5.5 6.0


dav (mm) ImageJ 14.9/9.4  4.7/3.3 6.2/3.7
average/median
Fig. 10. Fracture surface of a plane strain sample (Fig. 1c) after dynamic loading.
380 P. Verleysen and J. Peirs / International Journal of Impact Engineering 108 (2017) 370388

Fig. 11. SEM micrographs of sheet Ti6Al4V samples close to the fracture loaded in shear (a), tensile (b) and plane strain (c). The magnification used for the shear sample is higher
than for the other samples.

TagedPaverage void size is larger. In the dynamic tests, the voids are more sTagedP train rate is not very pronounced. Only the void distribution is
concentrated around the fracture. slightly more localised in dynamically loaded samples.
TagedPMore information on the local void distribution is obtained by con-
sidering strips of 25 mm width, parallel to the fracture surface. In each
of these band-like regions, the number, density and average size of 4.2. Bar material
the voids is calculated. In order to increase the statistical significance
of the results, again average values for all the available samples are TagedPIn Fig. 15 (top) a representative high strain rate tensile curve is
calculated. Fig. 14 show the void characteristics as a function of the presented for the cylindrical sample of Fig. 1e. Compared to the in-
distance from the crack in the in-plane tensile (a) and plane strain plane tensile sheet samples (Fig. 4b), the stress levels are slightly
samples (b). For both the tensile and plane strain samples, the void lower, which is not that surprising since the chemical composition
fraction near the fracture is §0.60.7%. The highest void fraction mea- and the microstructure of the sheet and bar material are not exactly
sured in an in-plane tensile sample is 1.1%. Also for the plane strain the same. In all samples a cup-and-cone fracture is observed; corre-
samples relatively low failure void fractions are found. This might be sponding values of the fracture strain can be found in Table 5.
related to the technique used to measure the voids which does not Because of rebound of the bars used to grip the cylindrical tensile
capture voids with dimensions smaller than 0.3 mm. However, as men- sample in the static tensile test device (Section 2.2.1), most of the
tioned earlier, because of the heterogeneous void distribution, locally fracture surfaces are damaged after testing and no reliable fracture
void fractions significantly higher than the average values are found strains are obtained for the static tests. From Table 5, it is clear that
which might give rise to early void coalescence at these locations. the fracture strain in the cylindrical tensile samples is significantly
TagedPIn the tensile samples next to large voids, many small voids are higher than in the in-plane tensile sample: 0.62 vs. 0.45. The cup-
observed. In the plane strain samples fewer, though larger voids are and-cone fracture surfaces of statically (0.001 s¡1) and dynamically
present. Further away from the fracture, the number of voids and (750 s¡1) loaded tensile samples are shown in Fig. 16. Dimples are
void fraction decreases rapidly. This is most pronounced for the formed in both samples, however, differences can be noted between
plane strain sample. Beyond 100 mm, the number of voids is below the two fracture surfaces: in the dynamic sample, the shape of the
one fourth of the maximum and the void fraction is under 0.1%. If dimples is better outlined and the dimples are larger. The dimples in
the spatial distribution of void density, void fraction and void size is the fracture surfaces of the dynamically loaded bar and sheet sam-
compared between static and dynamic samples, the effect of the ples are very similar.

Fig. 12. Examples of heterogeneously distributed voids in tensile samples: front view of sheet sample (a), side view of sheet sample (b) and front view of cylindrical sample (c).
Red markers indicate void clusters. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
P. Verleysen and J. Peirs / International Journal of Impact Engineering 108 (2017) 370388 381

Fig. 13. Comparison of fracture strain and void characteristics averaged over a region within a distance of 100 mm from the fracture. Void fraction, number of voids per unit area
and average void size are represented in tensile and plane strain sheet samples (a) and comparison between all static and dynamic tests, shear tests excluded (b).

TagedPIn the cylindrical tensile samples, many void clusters appear in tTagedP he FE model described in Section 3 are presented, together with the
columns oriented in the direction of the elongated grains, see simulated and experimental fracture morphology. An excellent
Fig. 12c. These voids are not likely to coalesce because of their rela- quantitative and qualitative agreement between the experiments
tive orientation to the loading direction, which might contribute to and simulations is obtained. Indeed, next to the stress-strain
the higher fracture strain of the bar material compared to the sheet response also the specific features of the fracture process are cap-
material. Analysis of the voids close to the fracture surface of the tured by the simulation: the X-shaped strain localisation which pre-
cylindrical tensile samples shows the maximum void fraction is cedes fracture in the GL50 sample (Fig. 21a), the mixed transverse-
around 2%, which is two times higher than in the sheet sample. inclined fracture in the in-plane tensile samples (Fig. 21b) and the
TagedPIn Fig. 17 (top) a torsion curve is given. Here again, lower stress straight crack aligned with the tensile direction in the in-plane shear
levels are obtained compared to the dynamic in-plane shear test sample (Fig. 21c).
(Fig. 4a). The bar shear fracture strains are, for the static as well as
for the dynamic tests, slightly higher compared to the in-plane shear
tests. Fig. 18 shows the fracture surface of a dynamically deformed 5.2. Bar material with JohnsonCook hardening and
torsion sample. The location where the fracture has initiated is indi- GursonTvergaardNeedleman damage
cated with white arrows. Large regions of the fracture surface are
destroyed by rubbing of the two fractured parts; in the non-dam- TagedPFig. 15 shows an experimental and two simulated stress-strain
aged regions dimples are present. curves for a dynamic (average strain rate of 420 s¡1) tensile test on a
cylindrical sample. In both simulations, for the material JC hardening
is considered with the parameters given in Table 2, in one simulation
5. Numerical simulations additionally the GTN failure model is activated with the parameters
of Table 4. Results of the latter correspond very well with the tests.
5.1. Sheet material with JohnsonCook hardening and damage Also contour plots of the plastic strain distribution at different stages
in the experiment are shown: both the strain at the onset of fracture
TagedPIn Fig. 21 simulated stress-strain or force-displacement curves for and the fracture pattern are well-predicted.
a statically loaded GL50 sample (a), a dynamically loaded in-plane TagedPSimulated and experimental stress-strain curves for a dynamic
tensile sample (b) and a dynamically loaded shear sample (c) using (average g_ D § 1400) torsion test on a thin-walled tube are shown

Fig. 14. Distribution of void density, void fraction and void size as a function of the distance from the fracture surface for the sheet tensile (a) and plane strain (b) samples.
382 P. Verleysen and J. Peirs / International Journal of Impact Engineering 108 (2017) 370388

Fig. 15. Experimental and simulated stress-strain curves of a dynamic tensile test on a cylindrical sample (top) and distribution of the plastic equivalent strain in the sample at the
stages indicated on the top graph (bottom). For the simulations JohnsonCook hardening is used with the GTN failure model, however, also the stress-strain curve without GTN
damage is presented.

TagedPin Fig. 17. Also, the plastic strain distribution in the torsion sample at sTagedP tress and axisymmetric samples considered in this study, the Lode
three different moments is represented: the strain gradually local- parameter can either be written as a function of the stress triaxiality
ises around the central section of the gauge until, finally, only the or is constant, further analysis of the results will focus on the triaxi-
central elements are further deforming. Here again, simulations ality only. Fig. 20 shows the simulated evolutions of the plastic
with and without GTN failure model are performed. In both models, equivalent strain and stress triaxiality in the centre of the gauge sec-
the strain eventually localises, however, without the GTN failure tions -where fracture is assumed to initiate- of the samples pre-
strain localisation occurs at higher levels of deformation. As is clear sented in Fig. 1, together with the experimental fracture strains.
from Fig. 17, both simulations overestimate the strain: even if GTN From the figure it is clear that not only the triaxiality at fracture, but
damage is activated, locally, plastic strains of almost 500% are also the triaxiality evolution affects the strain reached at fracture.
obtained in the shear band. The low triaxiality in the torsion samples Indeed, at fracture similar values of the stress triaxiality are reached
retards void growth, the increase of void fraction is mainly attrib- in the tensile and plane strain samples, however, the tensile fracture
uted to void nucleation. The critical value of 2% voids to start void strain is more than double the fracture strain in plane strain condi-
coalescence is not reached in the simulation. Consequently, the GTN tions.
model itself does not predict failure of the torsion sample. In the TagedPUsing the parameters of Table 3, also the JohnsonCook damage
simulation, the torsion sample loses its load carrying capacity due to initiation strain (Eq. (7)) is represented in Fig. 20. Two curves can be
strain localization caused by thermal softening. Voids in the mate- found: a static curve calculated for a strain rate of 0.0007 s¡1 and
rial, considered by the GTN model, accelerate the strain localization. temperature of 293 K, and a dynamic curve considering a strain rate
of 1000 s¡1 and temperature of 350 K. The temperature of 350 K is
6. Analysis and discussion used to account for the temperature increase in dynamically
deformed samples due to adiabatic heating. Obviously, the tempera-
TagedPIn Section 4 clear evidence is given that the fracture properties ture evolution till the onset of damage differs from sample to sam-
are affected by the stress state which can be quantified by the stress ple, however, a value of 350 K gives an average in line with data
triaxiality and Lode parameter, see Section 2.1. Since, for the plane found in [9]. Compared with the experimental fracture strains, the
P. Verleysen and J. Peirs / International Journal of Impact Engineering 108 (2017) 370388 383

Fig. 16. Fracture surface of cylindrical tensile sample after static (a) and dynamic test (b).

TagedPJohnsonCook fracture initiation strain curves do not follow the TagedP xist close to the fracture. The simulations reveal that this is also the
e
same trends. Indeed, the significantly higher fracture strains for the case in the in-plane tensile samples, as can be seen in Fig. 21 which
tensile samples are not captured. Also, the influence of the strain shows a light optical micrograph of a dynamic in-plane tensile sam-
rate is not properly described over the entire triaxiality region. Of ple together with the simulated strain distribution at the crack front.
course, and as opposed to the interpretation in the original paper of At the crack, an equivalent plastic strain of 0.50 is reached, at a dis-
Johnson and Cook [30], the JC strain values in Fig. 20 only predict tance of 350 mm from the crack, the strain drops to 0.25. The insert
onset of damage and are as such not equal to the fracture strain. In of Fig. 21 presents the measured void fraction as a function of the
the FE models for the sheet samples, after onset of damage, the dam- strain. From these microscopic observations it follows that only very
age evolution is described by Eq. (8) in which, as mentioned in Sec- few voids are present beyond 350 mm. Obviously, for the in-plane
tion 3.2, Gf as a crucial parameter. A good agreement between tensile samples, a strain of §0.25 seems to be required for the nucle-
simulations and experiments, including the fracture morphology, ation of voids. The existence of a void incubation strain with a value
can only be achieved if a relatively high value for Gf (50mJ/mm2) is similar or even lower than 0.25 can explain the very low and rapidly
used. The high value for Gf results in a long path from damage initia- decreasing void fractions observed in the material adjacent to the
tion to fracture. fracture in plane strain samples. Indeed, plane strain samples fail at
TagedPThe FE simulations shed new light on the microscopic observa- a strain of §0.25 and, because of the strong strain gradients, the
tions. The lower number of voids close to the fracture in the plane material next to the fracture is subjected strains significantly lower
strain samples compared to the tensile samples seems to conflict strains than 0.25. The notion of incubation strain is in line with the
with the observation of similar dimple sizes in all fracture surfaces, analysis presented in [60], and is also reflected in the JC damage ini-
see Table 6. However, because of the constantly changing width of tiation strain (Eq. (7)) and the GTN void nucleation rate (Eq. (10)).
the gauge section of the plane strain samples, high strain gradients Moreover, Eq. (10) gives a statistical distribution for the void
384 P. Verleysen and J. Peirs / International Journal of Impact Engineering 108 (2017) 370388

Fig. 17. Experimental and simulated stress-strain curves of a dynamic torsion test on a thin-walled tubular specimen (top) and distribution of the plastic equivalent strain in the
sample at the stages indicated on the top graph (bottom). For the simulations JohnsonCook hardening is used with the GTN failure model, however, also the stress-strain curve
without GTN damage is presented.

Fig. 18. Fracture surface of torsion sample after dynamic deformation.


P. Verleysen and J. Peirs / International Journal of Impact Engineering 108 (2017) 370388 385

Fig. 19. Comparison of experimental and simulated stress-strain (or force-displacement) curves and fracture patterns in a static tensile test with GL50 specimen (only half the
specimen is simulated, the other part is mirrored) (a), dynamic in-plane tensile test (b) and dynamic in-plane shear test (c). The simulation contour plots represent plastic equiva-
lent strain (PEEQ) values.

TagedPnucleation strain whereby voids can nucleate from the early stages rTagedP ather than damage. The simulations show that voids favour strain
of deformation till fracture. The presence of many small voids in the localization and thus affect failure of the material, which confirms
tensile samples shows that this is indeed the case in real deforma- the experimental observations published in [14].
tion processes. TagedPIn the plane strain sample, on the contrary, no localization of the
TagedPThe simulations also demonstrate the important role of strain strain in bands is observed in both the experiments and simulations.
localization phenomena in the fracture processes. In Fig. 22 the After a strain driven void nucleation, rapid void growth, driven by
simulated plastic strain and triaxiality distributions over the the high triaxiality, results in a void coalescence dominated failure.
thickness around the centre of the in-plane tensile sheet sample The shallow dimples in the fracture surface, together with the rela-
of Fig. 1b just before fracture are presented. Around the centre, tively low fracture strain, indicate a limited plastic deformation of
two 45° inclined bands with localised strains and high triaxial- the ligaments between the voids.
ities are formed. Fracture occurs along these bands which
explains the experimentally observed V-shaped and slant frac-
tures in the tensile samples. In the centre an equivalent strain
close to 0.65 is obtained, which is much higher than the experi-
mental values given in Table 5, which are averaged over the frac-
ture surface. Also the 54.7° localized neck formation at the origin
of fracture in the GL50 samples and in the ligaments next to the
transverse fracture in the samples of Fig. 1b, is captured by the
simulations, see Fig. 19a and b. The fact that, despite the higher
triaxiality (see Fig. 20), significantly higher fracture strains are
observed in the cylindrical tensile samples compared to the
sheet tensile samples, is mainly attributed to the absence of
localised necking in the cylindrical samples. Indeed, in the sheet
tensile samples, strain localization abruptly ends the ductile void
nucleation and growth process which involves high deforma-
tions. The void fraction in the sheet material next to the fracture
is half the value observed in the cylindrical samples.
TagedPAlso in the in-plane shear and torsion samples, strain localization
in ASBs is key to understand the lower fracture strains in the Fig. 20. Simulated evolutions of the plastic equivalent strain and triaxiality at the
dynamic samples. Although, the FE model of the torsion samples location of fracture initiation for the different tests. The experimentally established
fracture strains are indicated with a square. The solid lines represent the damage initi-
could not fully capture the experimental observations (Fig. 17), the
ation strain predicted by the JohnsonCook damage model for static (strain rate of
simulations are in agreement with the experimental observation 0.0007 s¡1, temperature of 290 K) and dynamic conditions (strain rate of 1000 s¡1,
that strain localisation in a shear band is the precursor to failure temperature of 350 K).
386 P. Verleysen and J. Peirs / International Journal of Impact Engineering 108 (2017) 370388

Fig. 21. Front view of a sheet tensile specimen with indication of the simulated plastic equivalent strain as a function of the distance from the fracture surface. The simulation
shows a large plastic strain gradient near the fracture. Insert with the measured void fraction as a function of the plastic strain.

TagedPDetailed analysis of thermal effects in tensile tests on the Ti6Al4V TagedP nd fracture. This can linked with the experimental observation that
a
sheet also considered in this study showed that the strain at the at high strain rates, the voids are slightly larger (Fig. 13b) and more
onset of necking is only indirectly affected by the strain rate via the concentrated around the fracture. Indeed, earlier onset of necking
sample heating due to plastic work [9]. Due to the high production or, more generally, more pronounced strain localization in the
of heat, the low heat capacity and low thermal conductivity com- dynamic tests, results in higher triaxialities and thus faster void
bined with the pronounced thermal softening of Ti-6Al-4V, sample growth. Also, other observations in the dynamic test such as the ear-
heating cannot be neglected. The temperature increase in a sample lier localization of strain in bands in the torsion, in-plane shear and
is highly dependent on the imposed strain rate, mainly because the in-plane tensile tests, can be attributed to thermal softening. All this
heat transferred to the environment is strongly dependent on the might led to the conclusion that also for fracture, the strain rate only
duration of the test and thus on the strain rate. In [9] it is shown that has an indirect impact via the sample heating. However, further
isothermal conditions can only be assumed for strain rates below research is needed to confirm this hypothesis.
0.0001 s¡1, adiabatic conditions are met from strain rates as low as TagedPThe tests show that test setup related parameters such as sample
0.01 s¡1. Therefore, in the FE simulations described in Section 5 ther- geometry have a significant effect on the test results. Different
mal effects are taken into account. The simulation results show that expressions of the same material behaviour are observed even in
also here sample heating has a major effect on strain localization identical loading conditions. Indeed, the three different geometries

Fig. 22. Comparison of the simulated distribution of the plastic strain and stress triaxiality with experimental fracture morphology in a side view of the sheet tensile specimen.
P. Verleysen and J. Peirs / International Journal of Impact Engineering 108 (2017) 370388 387

TagedPused for tensile testing give rise to three different fracture morphol- TagedP elgium and the partners of IUAP-VII-project P7/21 ‘Multiscale
B
ogies: a mixed transverse-54.7° inclined fracture for the small sam- mechanics of interface dominated materials’.
ples of Fig. 1b, a fracture dominated by the 54.7° neck for GL50
samples and a cup-and-cone fracture for the cylindrical samples of References
Fig. 1e. The FE models capture these fracture morphologies and thus
allow to distinguish between intrinsic material behaviour and struc- TagedP [1] Majorell A, Srivatsa S, Picu RC. Mechanical behavior of Ti6Al4V at high and mod-
erate temperatures - Part I: experimental results. Mater Sci Eng A 2002;326
tural effects related to sample geometry, boundary conditions and (2):297–305.
loading mode. TagedP [2] Xu YB, Bai YL, Meyers MA. Deformation, phase transformation and recrystalliza-
TagedPMore generally, all the above-mentioned considerations confirm tion in the shear bands induced by high-strain rate loading in titanium and its
alloys. J Mater Sci Technol 2006;22(6):737.
that the prediction of the fracture characteristics from the material TagedP [3] Follansbee PS, Gray GT. An analysis of the low temperature, low and high strain-
properties and loading conditions demands for a multilevel approach rate deformation of Ti6Al4V. Metall Trans A 1989;20:863–74.
based on experimental data and simulations, as stated in [32]. TagedP [4] Khan AS, Kazmi R, Farrokh B. Multiaxial and non-proportional loading responses,
anisotropy and modeling of Ti6Al4V titanium alloy over wide ranges of strain
rates and temperatures. Int J Plast 2007;23:931–50.
TagedP [5] Khan AS, Suh YS, Kazmi R. Quasi-static and dynamic loading responses and con-
7. Conclusions
stitutive modelling of titanium alloys. Int J Plast 2004;20:2233–48.
TagedP [6] Lee WS, Lin CF. Plastic deformation and fracture behaviour of Ti6Al4V alloy
TagedPStatic and dynamic experiments are carried out on sheet and bar loaded with high strain rate under various temperatures. Mater Sci Eng A
Ti6Al4V. Aiming at different stress states, several sample geometries 1998;241(1-2):48–59.
TagedP [7] Peirs J, Verleysen P, Degrieck J, Coghe F. The use of hat-shaped samples to study
and loading modes are used: in-plane shear, in-plane tensile and the high strain rate shear behaviour of Ti6Al4V. Int J Impact Eng 2010;37
plane strain sheet samples, and cylindrical tensile and thin-walled (6):703–14.
tubular torsion samples from bar material. Post mortem, the fracture TagedP [8] Peirs J, Verleysen P, Van Paepegem W, Degrieck J. Determining the stressstrain
behaviour at large strains from high strain rate tensile and shear experiments.
surfaces and material adjacent to the fracture are extensively stud- Int J Impact Eng 2011;38:406–15.
ied. Analysis and interpretation of the experimental results is sup- TagedP [9] Gala n J, Verleysen P, Degrieck J. Thermal effects during tensile deformation of
ported by finite element simulations by which detailed information, Ti6Al4V at different strain rates. Strain 2013;49(4):354–65.
TagedP[10] Peirs J, Tirry W, Amin-Ahmadi B, Coghe F, Verleysen P, Rabet L, et al. Microstruc-
complementary to the test results, is obtained on the stress and ture of adiabatic shear bands in Ti6Al4V. Mater Charact 2013;75:79–92.
strain distribution close to the fracture, including the loading path TagedP[11] Macdougall DAS, Harding J. A constitutive relation and failure criterion for
up to the onset of fracture. Ti6A14V alloy at impact rates of strain. J Mech Phys Solids 1999;47(5):1157–85.
TagedP[12] Rittel D, Wang ZG. Thermo-mechanical aspects of adiabatic shear failure of
TagedPIn all samples, a ductile fracture is obtained. From the analysis it
AM50 and Ti6Al4V alloys. Mech Mater 2008;40(8):629–35.
is clear that strain localization phenomena play a major role in the TagedP[13] Shih-Chieh L, Duffy J. Adiabatic shear bands in a Ti6Al4V titanium alloy. J
fracture process at lower triaxialities. Indeed, in all sheet samples, Mech Phys Solids 1998;46(11):2201–31.
TagedP[14] daSilva MG, Ramesh KT. The rate-dependent deformation and localization of
except the plane strain, fracture is preceded by localization of the
fully dense and porous Ti6Al4V. Mater Sci Eng A 1997;232(1-2):11–22.
strain in narrow bands. Also the torsion sample fails by strain local- TagedP[15] Pineau A, Benzerga AA, Pardoen T. Failure of materials I: Brittle and ductile frac-
ization. Simulations show that thermal softening induced by con- ture. Acta Mater 2016;107:423–83.
version of plastic work into heat, is a key element in this TagedP[16] Lecarme L, Tekoglu C, Pardoen T. Void growth and coalescence in ductile solids
with stage III and stage IV strain hardening. Int J Plast 2011;27:1203–23.
localization process. It is even believed that the strain rate has no TagedP[17] Lou Y, Yoon JW, Huh H. Modeling of shear ductile fracture considering a change-
direct effect on the fracture properties, but only via the tempera- able cut-off value for stress triaxiality. Int J Plast 2014;54:56–80.
ture increase and thermal softening it brings along. However, fur- TagedP[18] Malcher L, Andrade Pires FM, Ce sar de Sa JMA. An assessment of isotropic consti-
tutive models for ductile fracture under high and low stress triaxiality. Int J Plast
ther research is needed to prove this hypothesis. Analysis of the 2012;3031:81–115.
dimples in the fracture surface and the voids in the material next TagedP[19] Scheyvaerts F, Onck PR, Tekoglu C, Pardoen T. The growth and coalescence of
to the fracture, show that void nucleation is mainly driven by strain ellipsoidal voids in plane strain under combined shear and tension. J Mech Phys
Solids 2011;59(2):373–97.
and void growth by triaxiality. TagedP[20] Lode W. Versuche u € ber den einfuss der mittleren hauptspannung auf das fliessen
TagedPThe obtained test results cannot straightforwardly be linked with der metalle eisen kupfer und nickel. Ztg Phys 1926;36:913–39.
the material behaviour. Indeed, the results reflect the combined TagedP[21] Bridgman PW. Studies in large plastic flow and fracture. Cambridge, MA: Har-
vard University Press; 1952.
response of test setup and material. Certainly the sample geometry
TagedP[22] McClintock FA. A criterion of ductile fracture by the growth of holes. J Appl Mech
has a distinct influence on the test results. FE models in conjunction 1968;35:363–71.
with appropriate material models, fed and validated by high quality TagedP[23] Rice JR, Tracey DM. On the ductile enlargement of voids in triaxial stress fields. J
Mech Phys Solids 1969;17:201–17.
experiments, are needed to distinguish between intrinsic material
TagedP[24] Bao Y, Wierzbicki T. On fracture locus in the equivalent strain and stress triaxial-
behaviour and setup related response. In present study, for the sheet ity space. Int J Mech Sci 2004;46:81–98.
Ti6Al4V, the phenomenological JohnsonCook hardening and dam- TagedP[25] Bai Y, Wierzbicki T. A new model of metal plasticity and fracture with pressure
age initiation criterion together with an energy-based law describing and Lode dependence. Int J Plast 2008;24:1071–96.
TagedP[26] Roth C, Mohr D. Ductile fracture experiments with locally proportional loading
the progressive damage evolution are used in the FE models. For the histories. Int J Plast 2016;79:328–54.
bar material, the porous GursonTvergaardNeedleman plasticity TagedP[27] Lan WM, Deng XM, Sutton MA, Cheng CS. Study of slant fracture in ductile mate-
model is adopted, complemented with phenomenological laws for rials. Int J Fract 2006;141(3-4):469–96.
TagedP[28] Huang HC, Xue L. Prediction of slant ductile fracture using damage plasticity the-
void nucleation, growth and coalescence. Although both material ory. Int J Pressure Vessels Pip 2009;86(5):319–28.
models do not capture the full complexity of the fracture process, TagedP[29] Pardoen T, Pineau A. Failure of metals. In: Milne I, Ritchie RO, Karihaloo B, edi-
the FE simulations reproduce the sample dependent competition tors. Comprehensive structural integrity. Elsevier; 2008. p. 686–783.
TagedP[30] Johnson GR, Cook WH. Fracture characteristics of three metals subjected to vari-
between a pure, ductile void coalescence fracture and a failure by ous strains, strain rate, temperatures and pressures. Eng Fract Mech
macroscopic plastic strain localization, whether or not affected by 1985;21:31–48.
damage. Also, the interaction between material and sample response TagedP[31] Børvik T, Hopperstad OS, Berstad T, Langseth M. A computational model of visco-
plasticity and ductile damage for impact and penetration. Eur J Mech - A/Solids
and the importance of thermal softening are revealed by the
2001;20(5):685–712.
approach based on experimental data and simulation. TagedP[32] Li H, Fu MW, Lu J, Yang H. Ductile fracture: Experiments and computations. Int J
Plast 2011;27(2):147–80.
TagedP[33] Johnson GR, Cook WH. A constitutive model and data for metals subjected to
Acknowledgements large strains, high strain rates and high temperatures. In: 7th international sym-
posium on ballistics; 1983. p. 541–7.
TagedP[34] Kay G. Failure modeling of titanium 6Al-4V and aluminium 2024-T3 with the
TagedPThe authors would like to acknowledge the Interuniversity Johnson-Cook material model. Lawrence Livermore National Laboratory; 2003.
Attraction Poles Program (IUAP) of the Federal Science Policy of p. 24.
388 P. Verleysen and J. Peirs / International Journal of Impact Engineering 108 (2017) 370388

TagedP[35] Pardoen T. Numerical simulation of low stress triaxiality ductile fracture. Com- TagedP[49] Bacha A, Daniel D, Klocker H. On the determination of true stress triaxiality in
put Struct 2006;84(26-27):1641–50. sheet metal. J Mater Process Technol 2007;184(1-3):272–87.
TagedP[36] Certificate of conformity. 80135196. USA: Timet; 2002. TagedP[50] Ohashi M. Quantitative metallographic study for evaluation of fracture strain.
TagedP[37] Certificate of conformity. H9421-B01. Canada: Timet; 2002. Exp Mech 1998;38(1):13–7.
TagedP[38] SAE International. AMS-4911M: titanium alloy, sheet, strip and plate, 6AL-4V, TagedP[51] ASTM E112  13: Standard test methods for determining average grain size
annealed. 1977. TagedP[52] Nemat-Nasser S, Guo WH, Nesterenko VF, Indrakanti SS, Gu YB. Dynamic
TagedP[39] Properties and Processing of TIMETAL 6-4. Material datasheet. USA: Timet; 1998. response of conventional an hot isostatically pressed Ti-6Al-4V alloys: experi-
TagedP[40] Galan-Lopez J. Crystal plasticity based modelling of the strain rate dependent ments and modeling. Mech Mater 2001;33(8):425–39.
mechanical behaviour of Ti-6Al-4V. Ghent University; 2014. TagedP[53] Hillerborg A, Mode er M, Petersson PE. Analysis of crack formation and crack
TagedP[41] Peirs J, Verleysen P, Degrieck J. Novel technique for static and dynamic shear growth in concrete by means of fracture mechanics and finite elements. Cem
testing of Ti6Al4V sheet. Exp Mech 2011;52:729–41. Concr Res 1976;6(6):773–81.
TagedP[42] Verleysen P, Degrieck J, Verstraete T, Van Slycken J. Influence of sample geome- TagedP[54] Abaqus 6.14 User Manual.
try split Hopkinson tensile bar on sheet materials. Exp Mech 2008;48:587–93. TagedP[55] Tvergaard V. Influence of voids on shear band instabilities under plane-strain
TagedP[43] Peirs J. Experimental characterization and modeling of the dynamic behavior of conditions. Int J Fract 1981;17(4):389–407.
the titanium alloy Ti6Al4V. Ghent University; 2012. TagedP[56] Dixit PM, Dixit US. Modeling of metal forming and machining processes: by
TagedP[44] Kolsky H. An investigation of the mechanical properties of materials at very high finite element and soft computing methods. London: Springer; 2008.
rates of loading. Proc Phys Soc London B 1949;62:676–700. TagedP[57] Perrin G, Leblond JB. Analytical study of a hollow sphere made of plastic porous
TagedP[45] Baker W, Yew CH. Strain rate effects in the propagation of torsional plastic material and subjected to hydrostatic tension-application to some problems in
waves. J Appl Mech 1966;33:917–23. ductile fracture of metals. Int J Plast 1990;6(6):677–99.
TagedP[46] Gilat A, Schmidt TE, Walker AL. Full field strain measurement in compression TagedP[58] Hosford WF, Caddell R. Metal forming: mechanics and metallurgy. Cambridge
and tensile split Hopkinson bar experiments. Exp Mech 2009;49(2):291–302. University Press; 2007.
doi: 10.1007/s11340-008-9157-x. TagedP[59] Yingbin B. Dependence of ductile crack formation in tensile tests on stress triaxi-
TagedP[47] Kobayashi T, Simons JW, Brown CS, Shockey DA. Plastic flow behavior of Inconel ality, stress and strain ratios. Eng Fract Mech 2005;72(4):505–22.
718 under dynamic shear loads. Int J Impact Eng 2008;35(5):389–96. TagedP[60] Chu CC, Needleman A. Void nucleation effects in biaxially stretched sheets. J Eng
TagedP[48] Peirs J, Verleysen P, Degrieck J. High speed DIC for Hopkinson tests. Dymat con- Mater Technol ASME 1980;102:249–56.
ference high-speed imaging for dynamic testing of materials and structures. Lon-
don, 1820 November 2013; 2013.

You might also like