International Journal of Mechanical Sciences 118 (2016) 205–218

Contents lists available at ScienceDirect

International Journal of Mechanical Sciences

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

Experimental and numerical study on shear-punch test of 6060 T6 extruded

aluminum profile
crossmark
Lingyun Qiana,b, , Marcelo Paredesa, Tomasz Wierzbickia, Yannik Sparrerc, Martin Feuersteind,
⁎

Pan Zengb, Gang Fangb

a
Impact and Crashworthiness Lab, Massachusetts Institute of Technology, Cambridge, MA 02139, USA
b
Department of Mechanical Engineering, Tsinghua University, Beijing 100084, China
c
Steel Institute, RWTH Aachen University, 52062 Aachen, Germany
d
Constellium, 78224 Singen, Germany

A R T I C L E I N F O A BS T RAC T

Keywords: In the present work a comprehensive experimental/numerical study on a 6060 T6 bumper profile subjected to
Aluminum extrusion combined shear-compression (shear-punch test) load is carried out. A set of key plasticity and fracture
Yld200-3d constitutive model parameters to predict correctly the mechanical response of the structural component is required.
Ductile fracture Fracture initiation tests on lab scale specimens provide the sufficient data needed to calibrate the newly
Modified Mohr-Coulomb
extended non-quadratic yield function Yld2000-3d and the isotropic Modified Mohr-Coulomb (MMC) fracture
Shear-punch test
model. The calibrated plasticity and fracture constitutive model is then validated upon disk-shaped specimen.
Moreover, test results are reported on full-scale extruded aluminum bumper subjected to shear-punch load. A
detailed FE model of the bumper and the supporting structure was built. The numerical simulation outcomes
demonstrate that the calibrated material model predicts correctly not only the global force-displacement
response but also the local fracture behavior at multiple initiation points. It was shown that various part of the
multi-cell cross-section of the extruded profile underwent different loading histories. Thus, in one single test a
wide range of stress triaxialities develop which requires to introduce a very complete plasticity and fracture
model.

1. Introduction Therefore, this solution was incomplete because the prediction of the
force-displacement curve was not based on material properties alone.
Shear test has been conducted in the labs around the world for Zhou and Wierzbicki [8] proposed a complete analytical model for the
decades to characterize plastic properties of metal and other materials. blanking process. Since 2000 most of work on shear-punch prediction
It has been serving for several purposes. On one hand, there was a need was done by means of finite element simulations.
to determine strength of structural elements subjected to shear loading. The automotive industry and sheet metal forming industry look
On the other, shear tests were conducted to verify varies hypothesis very carefully into the problem of the shear-punch loading and
regarding plastic deformation and flow in metals. According to the von blanking to improve the edge quality of sheets and reduce the amount
Mises yield condition, the ratio of the yield stress in tension to the yield of so-called “burr” and possible micro cracks formed during that
stress in shear is 3 . This hypothesizes has been verified by many process. Such micro-defects are potential of initiation points of major
investigators, see for example [1]. Such study also summarizes some of cracks in sheets during subsequent forming or crash. The hole
the most important experimental works done on shear tests up to 2005. expansion ratio (HER) is a simple but convenient industrial measure
Based on an extensive number of experiments conducted by Atkins [2– of the edge quality. One of the most recent and complete study of
4] and Dodd and Atkins [5,6], a simple computational shear model for shear- punch process and subsequent prediction of the remaining
blanking was proposed by Atkins [7]. In his model, an apparent plastic ductility in thin plates was conducted by Wang et al. [9] and Wang and
instability was attributed not only to the adiabatic heating but also to Wierzbicki [10]. A very fine FE mesh consisting of 160 elements
the decreasing of shear area defined as the current area of common through the thickness was able to capture most of the fracture
cross-section. The width of the shear zone was considered as a morphology typically observed in this type of experiments. With
parameter which was later found from the experimental data. increasing tool gap between the piece and press-die tool, the shear-

⁎
Corresponding author at: Department of Mechanical Engineering, Tsinghua University, Beijing, China.
E-mail address: [email protected] (L. Qian).

http://dx.doi.org/10.1016/j.ijmecsci.2016.09.008
Received 7 January 2016; Received in revised form 15 August 2016; Accepted 5 September 2016
Available online 15 September 2016
0020-7403/ © 2016 Elsevier Ltd. All rights reserved.
L. Qian et al. International Journal of Mechanical Sciences 118 (2016) 205–218

Fig. 1. (a) The extruded aluminum 6060-T6 bumper; and (b) dimensions of its cross-section (mm).

punch process gives way to combine bending-tension loading mode. are initially subjected to compression which gives rise local buckles.
Luo and Wierzbicki [11] developed a computational tool for predicting Meanwhile, the upper flange undergoes a combined shear-tension and
fracture during stretch-bending operation. That tool relies on the bending load due to a very changing condition generated by the
application of Modified Mohr-Coulomb (MMC) fracture criterion, collapse of webs. Fracture may initiate at different locations and for
which should be calibrated from suitable tests. A unique dual-actuator the purpose of crash and weight optimization must be predicted with
loading frame and an experimental procedure for combined shear- high accuracy.
tension tests were developed by Mohr and Henn [12] and Dunand and The objective of the present work is to perform a bumper shear test
Mohr [13]. A new butterfly specimen was designed in such a way that at industrial scale and correlate the experimental results of global and
fracture never initiate from the edges, but always in the center. There local response with finite element simulations. Such simulations are
are several generations of the butterfly specimens and the most recent routinely carried out by automotive industry but rarely with the
version of the so called “smiley” shear specimen was described by Roth fracture option. Here, we have demonstrated that fracture initiation
and Mohr [14]. and propagation, in a large-scale aluminum automotive component,
Most of the shear experiments are carried out on small-scale can be predicted with good accuracy using the suitably calibrated MMC
coupons cut from sheets or thin sandwich panels with honeycomb model. There are several parameters in this model defining plastic
core. Extruded aluminum members do not fall into this category. The yielding and flow as well as fracture. Those parameters have been
wall-thickness of typical extrusion is around 3–4 mm and the cross- determined through a comprehensive testing and calibration program
section dimension could be an order of magnitude larger. Extruded involving flat specimens cut from the flanges of the extruded profile.
aluminum multi-cell hollow sections are typical structure elements The current study material is a 6060 T6 aluminum alloy whose
existing all around us from household appliances through lab equip- measured properties were different from those available in the open
ment to bumpers or entire car bodies. To the best of the author's literature. The test and calibration procedure is not new and closely
knowledge no large-scale shear test on extruded aluminum bumper has follows one developed at the Impact and Crashworthiness lab (ICL) at
not been described nor published in the open literature. The auto- MIT and describes the material plastic [15] and fracture [16] behavior
motive industry is routinely conducting crash, frontal crash test into a of an extruded profile of 6260 T6 aluminum alloy. What is new, it is the
pole of circular cross-section. Such test and corresponding numerical successful correlation of the global response history and prediction of
simulation involve three point bending with only small tension. At the fracture initiation at different points of the deforming bumper. There
same time shear loading using flat rectangular punch and relatively were several points in the bumper that underwent a complete different
small tool gap provides far more interesting information about material stress and strain histories. In order to tackle such a problem fracture
and cross sectional response of the structure. During such experiments predictive model must cover a very wide range of stress triaxialities
various cross sectional profiles follow different loading history that a including negative (compression), through zero (shear) all the way to
certain point could lead to fracture initiation. For example, the webs tension up to 2/3. We have demonstrated a successful prediction of

206
L. Qian et al. International Journal of Mechanical Sciences 118 (2016) 205–218

plastic flow and fracture for all of the above states and just one complex
experiment. To achieve this goal a rather complex plasticity model
(Barlat Yld2000-3d) and fracture model (MMC) had to be used. This
requires a careful and complete calibration procedure and the major
part of this study is devoted to that subject. In this way it is believed
that any interested reader will be able to reproduce the same type of
loading and simulation.

2. Material and plasticity model

Constellium Automotive Structures (Germany) provided a section

of extruded aluminum 6060 T6 bumper for the study, as shown in
Fig. 1(a).
Its length along the extrusion direction is 1480 mm. The cross-
section dimensions are shown in Fig. 1(b) (all dimensions are given in
mm). The material in the cross-section planes became inhomogeneous
Fig. 3. Engineering stress-strain curves of dog-bone specimens cut from eight different
after the extrusion process and quenching treatment. It is important to locations.
determine the amount of material inhomogeneity.

2.1. Evaluation of material inhomogeneity

According to the features of cross-section, the bumper can be

approximately divided into three parts: top flange, web plates and
bottom flange, as shown in Fig. 2(a). Eight locations were marked in
these three parts by T1, T2, W1, W2, W3, B1, B2 and B3. Eight
standard dog-bone specimens, Fig. 2(b), were cut from eight tagged
locations along the extrusion direction to get information on the
amount of material inhomogeneity. These specimens were tensioned
Fig. 4. Relative hardness measure at selected points in the bumper profile.
under quasi-static conditions. The digital image correlation (DIC)
technique was utilized to calculate strain evolution.
along three directions 0°, 45°, and 90° with respect to the extrusion
Fig. 3 shows the measured engineering stress-strain curves. The
direction, having their gauge sections lie in the width range of B2 part.
hardening behavior, yield stress and ultimate tensile strength varies
The specimens were ground down to 2 mm thickness after they were
marginally within small ranges. At the same time, the fracture strain in
cut from the bottom plate. The tensile tests were conducted under
uniaxial tensile tests differ one to another across the profile due to the
quasi-static conditions along with DIC system.
different cooling rates during extrusion. Therefore, no attempt was
Fig. 6(a) shows the engineering stress-strain curves of three dog-
made to capture this effect in the present computational model. The
bone specimens. The good repeatability of the engineering stress-strain
hardness of selected points on the cross-section were measured and
curves for different tests is observed. The yield stresses in three
their values were comparable, as depicted in Fig. 4. As an approxima-
material directions Y0, Y45 and Y90 are 220.8 MPa, 214 MPa and
tion, the bumper profile can be considered as homogeneous. In the
213.3 MPa, respectively as listed in Table 1. Despite the differences
following plasticity and fracture characterization, only the bottom plate
in stress-strain curves among specimens, the material does not exhibit
of bumper was employed for experimental tests.
fully planar isotropy. Thereupon, Lankford ratios are essential to
evaluate effects of anisotropy on initial yielding and subsequent plastic
2.2. Material anisotropy flow. Here, two in-plane plastic strains along extrusion direction (0°)
and transverse direction (90°) on the specimen surface were measured
The bottom plate B2 section was selected to test material aniso- through DIC technique. The thickness plastic strain was then deduced
tropy. Small dog-bone specimens (shown in Fig. 5) were extracted by assuming plastic incompressibility. The measured transverse and
thickness plastic strains for each direction are plotted in Fig. 6(b) with
their respective fitting lines. The gradient of these lines correspond to
the Lankford coefficients, whose values are: r0 = 0.5, r45 = 0.4 and r90 =
2.5. It is clear that this alloy displays a severe in-plane anisotropy.

2.3. Plasticity model characterization

In order to characterize material plastic behavior, one has to

determine the yield function, hardening law and plastic flow rule.
The shape of the yield surface is defined by a non-quadratic anisotropic
function in which the equivalent stress is defined. The strain hardening
response of the material is related to the equivalent plastic strain
obtained from the plastic work conjugacy definition, while co-direc-
tionality principle is assumed between plastic flow and stresses to
define the plastic strain increment tensor. The yield condition is
defined by
f (σ, k)=σ−k = 0 (1)
Fig. 2. (a) Cross-section configuration and its constitutive parts; (b) Specimen geometry
of uniaxial tensile test (all measures are in mm). where k is the deformation resistance which is controlled by the

207
L. Qian et al. International Journal of Mechanical Sciences 118 (2016) 205–218

Fig. 5. Dog-bone specimens aligned along 0°, 45°, and 90° respective to the extrusion direction.

hardening law Table 1

Eight experimental measurements for calibrating Barlat Yld2000-3d model.
dk =H(εp )dεp (2)
Uniaxial yield stress Shear yield stress Lankford ratio
where the equivalent plastic strain increment dεp is defined from the
plastic work-conjugacy as Y0 [MPa] Y45 [MPa] Y90 [MPa] τ0 [MPa] τ45 [MPa] r0 r45 r90

σ ∙dεp=σ dεp (3) 220.8 214.3 213.3 160.0 127.4 0.5 0.4 2.5

The hardening function H(εp ) is calibrated based on the stress-

strain curve determined form uniaxial tension test. Here, the Voce
exponential hardening law is used
p
k =k 0+Q (1 − e−βε ) (4)

The non-quadratic anisotropic yield function Barlat Yld2000-3d is

employed to predict the plastic flow behavior. This model was recently
modified by Dunand et al. [15] from the original plane stress version
(Barlat Yld2000-2d) to incorporate the out-of-plane stress components.
The basic equations are detailed in Appendix A. Fig. 7. The butterfly specimen for determining shear yield stress.

2.4. Experimental programs for calibrating plasticity model 2.2 on basis of uniaxial tension tests. The shear yield stress was
determined by conducting pure shear tests on butterfly specimens
A number of at least eight experiments are required to calibrate the using the custom dual-actuator loading apparatus. The dimensions of
eight independent coefficients of the Barlat Yld2000-3d model. In this butterfly specimens are shown in Fig. 7. The specimens were cut from
study, three uniaxial yield stresses of Y0, Y45 and Y90; three Lankford B2 plate in the direction of 0° and 45° with respect to the extrusion
ratios of r0, r45 and r90; and two shear yield stresses of τ 0 and τ 45 with direction. They were tested under monotonic quasi-static loading
respect to the extrusion axis are required to adjust the constitutive condition. The procedures of calculating shear yield stress using
model . The above mentioned parameters were calculated in Section

Fig. 6. Uniaxial tensile tests of 6060 T6 for three angular measures 00, 450 and 900; (a) Engineering stress-strain curves; (b) Evolution of thickness to transverse plastic strains.

208
L. Qian et al. International Journal of Mechanical Sciences 118 (2016) 205–218

Table 2 Table 3
Eight coefficients of Barlat Yld2000-3d model. Parameters of Voce exponential hardening law.

α1 α2 α3 α4 α5 α6 α7 α8 k0 Q β

0.476 1.431 1.310 0.482 0.946 0.555 0 1.414 214.5 56 24

butterfly specimens are explained in great detail by Mohr and Oswald exponential type function, Eq. (4), was employed to fit the plastic
[17]. The values of τ 0 and τ 45 were found to be 160.0 MPa and response in the pre and post-necking stages by simply extrapolating the
127.3 MPa, as listed in Table 1. function for large deformations. The fitting coefficients are listed in
A suitable minimization function is implemented to reduce the Table 3.
error derived from the experimental/numerical correlation given by the
following expression 3. MMC fracture model and calibration
8 ⎛ p ⎞2
f ( pi )= ∑ ⎜⎜ mi −1⎟⎟ The MMC fracture model originally proposed by Bai and Wierzbicki
p
i =1 ⎝ i ⎠ (5) [18] is fully described in Appendix B for the sake of space. It is
noteworthy that there exists a certain inconsistency between the strain
where pi is the predictive value of a given variable for calibration, pim
hardening exponent (n) from Swift hardening curve , Eq. (B.3), used to
represents its experimental measured value. The variables pi consid-
describe the local fracture criterion and the Voce exponential type, Eq.
ered for optimization were: Yield stresses (Yi ), Lankford ratios (ri ) and
(4), employed for predicting plastic behavior. It is believed that such
shear stress at yielding (τi ) for different material orientations (0°, 45°
inconsistency does not lead to large errors because the plasticity and
and 90°) as described earlier. The resulting set of coefficients α are
fracture models are uncoupled. In the range of strain before necking,
listed in Table 2.
the differences between experimental flow curve and Swift-Voce fitting
Fig. 8(a) shows a comparison of the von Mises yield locus with the
results are indistinguishable. Large differences can occur in the post-
Barlat yield envelopes in plane stress condition for the current
necking range. We have found through extensive numerical simula-
aluminum alloy and 6260 T6 tested by both Dunand et al. [15] and
tions that the predicted load-displacement response of all type of
Luo et al. [16]. The three yield loci show overlapping in uniaxial tension
specimens follow the measured curves only using Voce model. The
and compression (σ2 principal axis) and some overlapping in shear
swift formula over predicted the global response. For that reason, the
quadrants, but there is a remarkable difference in equi-biaxial loading
Voce model was used for global plasticity simulation. The Swift
condition in tension and compression quadrants. Because of the strong
equation was only used to transform the stress based fracture criterion
material orientation dependence given by Lankford coefficients (see
into mixed stress-strain space. This transformation was enriched by
Fig. 6(b)), yielding become larger as biaxial ratio increases. For metal
introducing the Lode angle dependence on the hardening curve
sheet forming, the main deformation areas are subjected to stress
according to the following equation
states ranging from shear to biaxial tension. It is worth mentioning
that, the Barlat Yld2000-3d constitutive model predicts the shrinkage ⎡ 3 ⎛θπ ⎞ ⎤
of the yield locus in tension-tension (compression-compression) quad- σ = A (εp )n ⎢c3 + (1 − c3)(sec ⎜ ⎟ −1)⎥
⎣ 2− 3 ⎝ 6 ⎠ ⎦ (6)
rant although any equi-biaxial experiment was not used for calibration.
The equi-biaxial tension (compression) will be a valuable option to be Here, the term involving the Lode angle θ and the associated new
considered for further validation tests. fracture constant c3 is another way of introducing the non-quadratic
Likewise, plastic flow behavior of both aluminum alloys are also yield surface into the local fracture response. In this work, we have
compared in Fig. 8(b) in terms of true stress vs. plastic strain. It turns followed the formulation of MMC fracture model [18], in which c1, c2
out that despite similar yield stresses between them, a markedly and c3 are three free parameters to be determined. The above fracture
discrepancy can be observed in strain hardening behavior. A simple criterion is only used for predicting initiation without post-failure

Fig. 8. Comparison between 6060 T6 and 6260 T6 [15] aluminum alloys: (a) plane stress yield surfaces and (b) true stress-plastic strain curves.

209
L. Qian et al. International Journal of Mechanical Sciences 118 (2016) 205–218

Fig. 9. Specimen configurations for tensile (NT-R20, NT-R10, NT-R6.67 and CH) and shear (BS) tests.

softening at the element integration point. It is believed that the magnitude about 100 µm) is observed by DIC, as plotted by vertical
fracture process (initiation and propagation) in metal sheet problems dash lines in Fig. 10.
occurs almost without distinction since the reduced wall-thickness of Fig. 11 shows the loading paths of five specimens. The stress
the structural components (see illustrative examples in [11,19]). triaxiality and Lode angle parameter of central hole (CH) specimen
exhibits a nearly proportional loading path all the way to fracture
(around 0.4 and 0.87, respectively). This stress state resembles to the
3.1. Experimental/numerical program to calibrate the MMC model one developed by uniaxial tensile test, where η and θ equals 1/3 and 1,
respectively. Likewise, an almost proportional relation between the
The testing and calibration procedures follow the one described by stress state and applied load is observed in shearing deformation (BS).
Mohr and Dunand [20]. Therefore it is only briefly summarized here. This special loading mode develops zero triaxiality and Lode parameter
Three notched flat specimens with different radii of 20 mm (NT-R20), as well. On the other hand, circular notched (NT's) specimens display a
10 mm (NT-R10), and 6.67 mm (NT-R6.67); along with central hole severe non-proportional dependence to loading history ranging from
(CH) and butterfly (BS) specimens were cut from the bottom plate B2 0.4 ≤ η ≤ 0.8 and −0.5 ≤ θ ≤0.8. The results of the three notched
along the extrusion direction. The cut outs and central hole were specimens were similar.
introduced by milling (CNC) process having a final thickness of 2mm as The three parameters c1, c2 and c3 are calibrated using the above
shown schematically in Fig. 9. These geometries cover the following loading paths by minimizing the mean squared errors (MSE) between
stress state range: the stress triaxiality of 0 ≤ η ≤ 1 and Lode angle predicted and measured fracture strains. The optimized values are
parameter of −0.5 ≤ θ ≤ 1; and were used to calibrate the MMC model. listed in Table 4. The MSE value of 0.5% indicates the good correlation
The first four specimens were tensioned and the last one was loaded by of the measured data with the fitting surface.
shearing all the way to fracture. The displacement fields developed in Fig. 12(a) depicts the calibrated 3D fracture envelope for 6060 T6
the flat gauge section were recorded successively (several snapshots) by aluminum alloy which includes selected stress histories of some of the
DIC system, whereby strains and stresses are inferred. geometries used for calibration. The loading paths for NT-R20, CH and
The combined experimental-numerical approach was utilized to BS are plotted having the experimental strain to fracture measures and
calibrate the MMC model. To this purpose, a set of very detailed finite estimated errors marked by solid circles and colored bars, respectively.
element models of five specimens were built. The characteristic length The magenta color line corresponds to the plane stress condition. In
of the critical element was 0.1 mm. The calibrated plasticity model with order to compare this results with those performed by Luo et al. [16],
Yld2000-3d was encoded into ABAQUS/Explicit [21] via user-defined two fracture surfaces were plotted on the top of each other in
VUMAT subroutine. The global force-displacement curve and the local Fig. 12(b). The difference is apparent.
logarithmic axial strain (E1) of the central point on the surface between The FE simulation of five specimens were performed again by
experiments and simulations were compared to verify the robustness of encoding the calibrated MMC fracture model into the material model
the present FE models. Fig. 10 displays the results of NT-R20, CH and definition in the subroutine. The numerical simulations were stopped
BS ; where the adjusted constitutive model provides a good predict- when the damage indicators in the critical elements reached unity. The
ability in terms of global force-displacement (red solid line) and local predicted fracture initiation strains are compared with their experi-
surface strain (blue colored line) responses. The evolution of the mental values in Fig. 13. To help the visual comparative analysis a
equivalent plastic strain (PEEQ) corresponds to the load history of theory line is added, where experiments and predictions are ideally
the critical element, which experiences the highest plastic deformation identical. These results indicate that the fracture model provides good
gradient during straining. Failure initiation is assumed that oc- prediction for all tested configurations. Moreover, two additional
curs when the first surface discontinuity (macro crack in the order of

210
L. Qian et al. International Journal of Mechanical Sciences 118 (2016) 205–218

Fig. 10. Comparisons of force-displacement curves as well as central logarithmic axial strain on the surface between experiments and simulations for three specimens: (a) NT-R20; (b)
CH and (c) BS.

Fig. 11. Loading paths at the critical material points of five specimens: (a) in the space of stress triaxiality and equivalent plastic strain; (b) in the space of Lode angle parameter and
equivalent plastic strain.

211
L. Qian et al. International Journal of Mechanical Sciences 118 (2016) 205–218

Table 4
Three parameters of MMC model.

c1 c2 c3 MSE (%)

0.097 201.902 0.968 0.5

Fig. 13. Fracture strain comparisons between experimental results and predicted values
using MMC fracture model.

Fig. 14. Illustration of tools and sheet in punch test (mm).

symmetric features of punch test, only one quarter of disk was modeled
by meshing with eight-node reduced-integration solid elements C3D8R
[21]. The punch and die was set as rigid in the FE model. The outer
surface of the disk was fixed in the FE model to replace the holder effect
which precludes the inward material flow during the punch process.
The punch tests were conducted under two different friction
conditions between the plate and the hemispherical punch. In the first
case, a grease lubricated sandwich (teflon layers stack up) was utilized
and the contact can be assumed to be frictionless. In the last case,
friction was generated by removing the teflon stack up. The friction
coefficient (fc) was determined by an inverse engineering method. After
several attempts, the fc was determined as 0.1, where the simulated
Fig. 12. Tridimensional fracture surfaces in the space of stress triaxiality, Lode angle
fracture morphology coincided with the observed fracture localization
parameter and equivalent plastic strain for (a) 6060 T6; and (b) 6260 T6.
in the experiment of the punch disk. The friction coefficient poses a
negligible effect on the global load-displacement behavior, but become
fracture points of punch disk (PU-0 and PU-0.1) tests are added to
significant for the local fracture criterion by shifting initiation from the
further validate the current fracture criterion, which is discussed in
vertex to somewhere else in the disk.
detail in Section 4.
Validation will be addressed in three different aspects. Firstly, the
global force- displacement responses of two punch tests were predicted
4. Punch test on circular disks and compared with the measured data, as depicted in Fig. 15. The
measured and calculated load-displacement curves were almost iden-
The punch tests and simulations serve as a validation of the present tical. The onset of fracture was not easily identified without reliable
uncoupled fracture model at the size of lab specimens. The disk inspection methods. The instant of measured peak force was normally
specimen with a diameter of 70 mm and thickness of 2 mm was cut deemed as the moment of fracture onset. However, at this moment of
from B2 section. It was clamped between a circular die and a circular simulated peak force, there was already significant localization in the
holder, and then loaded monotonically using a hemispherical punch sheet and its damage indicator was higher than unity. The predicted
with a speed of 5 mm/min. The radius of the punch is 22.2 mm and the onset of fracture occurs earlier than the peak force is attained in the
inner radius of die is 24.5 mm, as illustrated in Fig. 14. Considering the

212
L. Qian et al. International Journal of Mechanical Sciences 118 (2016) 205–218

Fig. 15. Comparison between measured and simulated force-displacement curves for punch tests: (a) PU-0 (fc = 0); (b) PU-0.1 (fc = 0.1).

Fig. 16. Experimental fracture locations and corresponding predicted results using damage indictor fields for two punch tests: (a) PU-0 (fc = 0); (b) PU-0.1 (fc = 0.1).

213
L. Qian et al. International Journal of Mechanical Sciences 118 (2016) 205–218

approximately estimated by considering the measured final and initial

thicknesses of the disk. For two conditions with fc = 0 and fc = 0.1, the
experimental fracture strains were calculated as 0.60 and 0.64,
respectively. The fracture strains predicted by MMC model at the onset
of fracture (marked by red squares in Fig. 15) are 0.58 and 0.61,
respectively. The measured and predicted fracture strains for two
punch experiments are compared in Fig. 13 (PU-0 and PU-0.1).
During the experiments the virtual onset of fracture was difficult to
determine, so the tests were carried out all the way until the force drops
sharply. Therefore, the final measured thickness in failure areas are
inevitably smaller than those corresponding to initiation. That is the
reason why the measured fracture strains were larger than their
Fig. 17. Experimental system for punch shear test of an extruded bumper. predicted values. The loading path of punch test with fc = 0 is added
in the fracture surface plot (Fig. 12(a)). It is clear that its final point
numerical simulation. The fracture initiation predicted by the cali- located near the biaxial tension with stress triaxiality of 2/3 and Lode
brated MMC model were marked by red squares in Fig. 15. angle of −1. Above all, the calibrated material model provided a fairly
Secondly, the predicted failure locations for two cases were also good prediction of plasticity and fracture behaviors for punch tests at
compared with experimental observations, as illustrated in Fig. 16. specimen size.
From Fig. 15, it is clear that the friction conditions have a negligible
effect on global force-displacement responses, but failure locations
were mainly affected by it. As mentioned above, matching the
numerical simulations with experimental observations of fracture
localization was the criterion to determine the friction coefficient. For
the test with fc = 0, the crack locates at the center of disk specimen and
propagates linearly along TD axis (perpendicular to extrusion direc-
tion) as shown in the plain view in Fig. 16(a). The damage indicator is
color coded to exhibit the failure feature. In the profile view, the
predicted necking is quite well captured by the model, which is
compatible with experimental observations. On the other hand, the
failure location in the friction punch test (shown in Fig. 16(b)) shifts a
few millimeters from the center. The friction coefficient needed to
mimic the observed fracture morphology was fc = 0.1. It can be seen
that the crack initiate and propagate angularly at certain radius away
from the disk center as illustrated in the plain view. Meanwhile, slant
fracture feature is also well captured by the FE model as displayed in
the profile view.
Lastly, the punch disk undergoes equi-biaxial tension during
loading in which the in-plane plastic strain components are analogous
to the measured logarithmic through-thickness strain. It can be Fig. 19. True stress -plastic strain curve of bolt material.

Fig. 18. (a) Half tridimensional FE model of the bumper profile shear system; (b) Detailed cross-section mesh.

214
L. Qian et al. International Journal of Mechanical Sciences 118 (2016) 205–218

Fig. 1(b). The punch and die had the same fillet radii of 2 mm. The tool
gap between punch and die was 20 mm on either sides. The bumper
was firmly attached to the flat base using bolts and screws at two sides,
which otherwise the end of the bumper would easily lift with relatively
large tool gap. During the shear test, the punch drives the bumper
down at low speed of 5 mm/min until the punch force response
gradually drops by ductile fracture and subsequent plastic collapse.
Due to symmetry a half of the bumper assembly was modeled as
shown in Fig. 18. The punch and die were set as rigid. The bumper was
fixed by one bolt on each side. The radius of bolts were 4 mm and went
through three web holes supported by two fixed brackets. The bolt itself
was subject to plastic deformations due to combined bending-shear
effect during the test and therefore it was considered as elastic-plastic
material in the simulation. The Young's modulus and Poisson's ratio of
such material are 200 GPa and 0.33 with a measured true stress-plastic
strain curve given in Fig. 19. Large plastic deformation of the bumper
occurs between the punch and the die as well as in the three holes
around the bolts. These two locations are marked as ‘shear effect zone’
and ‘bolt zone’ in Fig. 18. The bumper was meshed using solid elements
of C3D8R [21]. Finer meshes were used in the above two zones having
a non-cubic isoparametric element with 1 mm length along the
extrusion (in-plane) direction and 0.3×0.3 mm length through thick-
ness direction, as shown in Fig. 18(b). Similar mesh size was used for
the deformable zone around the supporting pins. The mesh size effect
becomes more significant after the onset of necking. Necking was
observed neither around the pins nor in the shear zone under punch.
For that reason, the assumed mesh size was considered to be adequate
for large -scale simulations without affecting significantly the strain
level at which fracture occurs, initially calibrated with a much finer
mesh size (0.1 mm) as described above. The element was deleted once
damage indicator reached unity.

6. Results and discussion

Fig. 20. Comparison of force-displacement curves between experiments and simulations

Fig. 20 depicts the comparison of the measured and simulated load-
for bumper shear-punch test and damage evolutions of (a) bolt zone and (b) shear effect displacement curves of bumper shear test. As a comparison, the
zone. plasticity model with von Mises yield condition and MMC fracture
criterion were used to simulate such an experiment. The predicted
5. Shear-punch test of bumper in industrial application force-displacement curve using a simple J2 plasticity setting is plotted
in Fig. 20(a) by a red colored line. It is clear that the von Mises model
The input file for the material model consists in a total of 16 significantly underpredicts the global experimental behavior. This
parameters. Those are: 8 coefficients α1, …, α8 of the anisotropy yield underpins a necessity of conducting a very detailed calibration to take
condition; 3 coefficients of the Voce exponential type function, 2 account of plastic anisotropy. The simulations of Yld2000-3d model
coefficients of the Swift power law equation and 3 coefficients of the accurately predict the global force-deflection response and the magni-
fracture initiation criterion. The determination of these parameters tude of the peak force, which demonstrate not only the reliability of the
through tests and calibration was described in the previous section. We FE modeling but also the adequacy of the present material constitutive
provided sufficient details so that interested readers will be able to re- model. Slight differences are mainly attributed to the approximate
create these data for any other materials. The ground is now set for the description of the bumper support system by both brackets. The
full scale validations. damage evolution of the bolt and shear effect zones are represented
Fig. 17 shows the experimental set-up which consists of four main by color fringes in Fig. 20. The material around the hole area (W3
parts: punch, die, bumper and fixing structures. The length of the plate) initially fails at very small displacement around 5 mm. The hole
bumper is 1480 mm. The dimension of the cross-section is shown in areas on W1 and W2 fail subsequently. In the shear affected zone, the
fracture first appears in the middle web frame at displacement around

Fig. 21. Final shape of deformed bumper with its predicted damage evolution color fringe.

215
L. Qian et al. International Journal of Mechanical Sciences 118 (2016) 205–218

Fig. 22. Comparisons of fracture locations between experimental and predicted simulation results for bumper shear-punch test.

critical points to better understand the fracture process. Six material

points were selected from typical failure locations, as marked by P1–P6
in Fig. 22. The loading history in terms of stress triaxiality and
equivalent plastic strain at these six critical points (the point where
fracture initiates) were extracted from the FE model and plotted in
Fig. 23. The zero stress triaxiality corresponds to pure shear (verical
dash line), which serves as a division line between compression and
tension sides. During the whole process, P1 and P2 are mainly
subjected to compression with negative stress triaxiality levels.
Meanwhile, the state of stresses of the remaining points (P3-P6)
undergo tension mostly in the late deformation stage when the
equivalent plastic strain exceed 0.1. Only P4 and P6 experience tension
load all the way to fracture. It is clear that all material points follow a
non-proportional loading path. However, P1 fails at a relatively high
equivalent plastic strain with a negative triaxiality value, such as 1.2
and -0.22, respectively. Likewise, it occurs for P2 where failure initiates
at very low (positive) triaxiality level around 0.18. This demonstrates
Fig. 23. Stress triaxiality evolution of six material points extracted from failure zones. that the fracture occurrence in the bumper at negative stress triaxiality
is possible. Other important observation is that the stress states of P2,
13 mm and then develops to the left and right webs simultaneously. P3 and P5 go first into compression for then shift to tension passing
Plastic collapse finally occurs between bottom and upper layers of the through pure shear. In P3 the triaxiality varies in a small range around
profile as shown in Fig. 22. 0.28, while P5 changes drastically from fully compression (η = -1.25) to
Fig. 21 shows the deformed and the simulated profiles of the equi-biaxial tension (η = 0.66). Finally, one of the selected fracture
bumper. Large plastic deformation and fracture in several locations locations develops a quasi-proportional loading during straining. P4
were observed from the tested bumper in the three holes of the passing history resembles to the one observed in transverse plane strain
bolt and shear affected zones mentioned above. These two failure zones experiments (η = 0.58). Above all, the stress states of critical points
are also captured by the FE model, where the fracture initiation is of the component cover a wide range of stress triaxiality, which goes
highlighted by colored contours of damage indicator. Detailed failure from −0.22 to 0.6. To the best knowledge of the authors, the shear-
locations of these two zones between experiment and simulation are punch test on bumper is a rare example of industrial application in
compared in Fig. 22. The three holes are elongated along the extrusion which all possible state of stresses are generated at the same time.
direction. The material on top and bottom surfaces flows horizontally Compared with the effect of stress triaxiality on fracture, the effect of
under shear loading. Meanwhile, the lateral hole surfaces contacting Lode angle parameter was even more complicated, which is not
with the bolt are gradually folded with some out-of-plane wrinkles as discussed in the present study.
the contact pressure increases. In the shear affected zone, the flange B1
section fails due to the bending load. The bumper section between 7. Conclusion
punch and die are subjected to shear-punch loading and undergoes
large plastic deformation along with ductile fracture. The three webs The plasticity and ductile fracture behavior of an extruded alumi-
are forced into local buckling while fracture occurs in the outer surface num 6060 T6 bumper was investigated by means of a combined
of W1 and W3. Another fracture point takes place in the contact area experimental-numerical approach. The established material model
between top flange and punch die initiating in T1 upper flange. In considering anisotropic plasticity and fracture criterion were verified
summary, locally developed plastic zone and ductile failure locations of at the scale of lab tests and then applied to a scale of industrial process.
the bumper have been successfully predicted by the current fracture The key conclusions are drawn as follows:
criterion. The overall agreement of the shear-punch test is acceptable
considering the inherent complexity of modeling the loading/support 1. The amount of material inhomogeneity of extruded bumper was
system and material inhomogeneity. assessed and the bumper can be approximately taken as homo-
Since the present fracture model depends entirely on the stress geneous.
state and load history, it is of interest to analyze the loading path of 2. The uniaxial yield stresses, Lankford ratios, shear yield stresses

216
L. Qian et al. International Journal of Mechanical Sciences 118 (2016) 205–218

along 0°, 45°, 90° to extrusion direction were used to calibrate eight terms of force-displacement curve and failure locations. Fracture
parameters of Barlat Yld2000-3d yield function. The Voce exponen- initiationoccurs at different triaxiality values from negative (η =
tial hardening law as well as associated flow rule were adopted to -0.22) to higher (η = 0.66) levels, which are all well predicted by the
define material plastic model. MMC model.
3. The MMC fracture model in space of stress triaxiality, Lode angle
parameter, and equivalent plastic strain was adopted to analyze the
failure behavior. The loading paths of notch, center hole and Acknowledgment
butterfly shear specimens were extracted from simulations to
calibrate the MMC model. The support of this work comes primarily from the MIT Industrial
4. The punch tests with and without friction were conducted to validate Fracture Consortium. The involvement of Constellium Automotive
the material model at the specimen size. The predicted force- Structures from Germany is highly appreciated in providing material
displacement curves and fracture strains show a good agreement and experimental data. Thanks are also due to Dr. M. Luo, Dr. C. Roth
with experimental results. The different fracture locations of two and members of ICL for their valuable contributions. The first author
tests are also captured by the FE modeling. also acknowledges her partial financial support from the Tsinghua
5. The established material model predicts the deformation and Scholarship for Overseas Graduate Studies as well as the National
fracture response in the shear-punch test of extruded bumper in Natural Science Foundation of China under the Project no. 51375256.

Appendix A

The backbone of Yld2000-3d model is

1 1
σ= (∅′(S′) + ∅′ ′ (S′ ′)) α
21/ α (A.1)
α α
where ∅′(S′)= SI′ − SII′ α , ∅′ ′ (S′ ′)= 2SII′ ′ + SI′ ′ + 2SI′ ′ + SII′ ′ . The exponent α is assumed to be equal 8, which is based on texture (crystal plasticity)
observations. The stress deviator vectors S′and S′ ′ are linearly transformed from the Cauchy stress vector σ in the following form
S′=L′σ (A.2)
′ ′ (A.3)
S′ =L′ σ
′
where ={σxx σyy σzz σxy σyz σzx} , S = {sxx ′ syy
′ sxy
′ syz
′ szx
′ } and
S′ ′ = {sxx
′ ′ syy
′ ′ sxy
′ ′ syz ′ ′}. The two functions ∅′(S′) and ∅′ ′ (S′ ′) can also be re-written in terms of the components of S′and S′ ′ as
′ ′ szx
α
∅′(S′)=[(sxx ′ 2 + sxz
′ )2 +4(sxy
′ − syy ′ 2 + syz
′ 2 )] 2 (A.4)

⎡3 ′ 1 2 ⎤α ⎡ 3 ′ 1 2 ⎤α
∅′ ′ (S′ ′)=⎢ (sxx ′′ ) +
′ − syy ′ ′ − syy
(sxx ′ ′ ) +4(sxy
′ ′2 + sxz
′ ′2 + syz
′ ′2 ) ⎥ +⎢ (sxx ′′ ) −
′ − syy ′ ′ − syy
(sxx ′ ′ ) +4(sxy
′ ′2 + sxz
′ ′2 + syz
′ ′2 ) ⎥
⎣2 2 ⎦ ⎣ 2 2 ⎦ (A.5)
The linear transformation L′ and L′ ′read
⎡ 2α −α1 −α1 0 0 0 ⎤
⎢ 1 ⎥
−α2 2α2 −α2 0 0 0
1⎢ ⎥
L′= ⎢ 0 0 0 3α7 0 0 ⎥
3⎢ 0 0 0 3α 9 0 ⎥
0
⎢ 0 3α10 ⎥
⎣ 0 0 0 0 ⎦ (A.6)

⎡ −2α +2α +8α −2α −4α +4α +2α −4α α3+2α4−4α5−2α6 0 0 0 ⎤

3 4 5 6 4 6 3 5
⎢ ⎥
⎢ 4α3−4α4−4α5+α6 −2α3+8α4+2α5−2α6 −2α3−4α4−2α5+α6 0 0 0 ⎥
1
L′ ′= ⎢ 0 0 0 9α8 0 0 ⎥
9⎢ 9α11 0 ⎥
⎢ 0 0 0 0 ⎥
⎢⎣ 0 0 0 0 0 9α12 ⎥⎦ (A.7)
where αi , i=1–12, are the parameters of anisotropy. For α9 ,α10 ,α11,α12 is set equal unity according to the suggestion of Dunand et al. [15]. Thus, the
Yld2000-3d yield function is determined by eight coefficients of α1,α2, …, α8 .

Appendix B

The phenomenological fracture model can be interpreted by using a damage indicator definition as
εf dεp
D= ∫0 εf (η, θ ) (B.1)
where dε p is the equivalent plastic strain increment, η is the stress triaxiality and θ is the Lode angle parameter. The weighting function εf (η , θ )
defines the fracture strain under different stress states, and is served as “fracture envelope”. It is postulated that the fracture initiates when the
damage indicator reaches unity for any non-proportional and proportional loading. For the latter, the integration of Eq. (B.1) can be written as the
following expression for fracture initiation
εf =εˆf (η, θ ) (B.2)

217
L. Qian et al. International Journal of Mechanical Sciences 118 (2016) 205–218

The weighting function used in the present study corresponds to the modified Mohr-Coulomb (MMC) fracture criterion which was proposed by
Bai and Wierzbicki [18]. The original stress based Mohr-Coulomb model is transformed into the mixed strain-stress based fracture criterion by
means of a simple power law function (Swift) as follows
σ =A (ε0 + εp )n (B.3)
where A, n and ε0 are material parameters to be fitted during calibration procedure. The MMC model is then expressed in terms of stress triaxiality
and lode angle parameter
1
⎧ ⎡⎡ ⎫− n
⎪A 3 ⎛ ⎛ θ π ⎞ ⎞ ⎤ ⎡ 1+c 2 ⎛θπ ⎞ ⎛ 1 ⎛ θ π ⎞⎞⎤⎤⎪
εf =⎨ ⎢ ⎢c3 + (1−c3) ⎜sec ⎜ ⎟ −1⎟ ⎥ × ⎢ 1
cos ⎜ ⎟ + c1 ⎜η + sin ⎜ ⎟ ⎟ ⎥ ⎥ ⎬
⎪ c2 ⎢⎣ ⎢⎣ 2− 3 ⎝ ⎝ 6 ⎠ ⎠ ⎥⎦ ⎣⎢ 3 ⎝ 6 ⎠ ⎝ 3 ⎝ 6 ⎠ ⎠ ⎥⎦ ⎥⎦ ⎪
⎩ ⎭ (B.4)
The corresponding fitted Swift parameters A and n are 363.57 MPa and 0.131, respectively.

References [12] Mohr D, Henn S. Calibration of stress triaxiality dependent crack formation
criteria: a new hybrid experimental-numerical method. Exp Mech
2007;47(6):805–20.
[1] Guduru RK, Darling KA, Kishore R, Scattergood RO, Koch CC, Murty KL. [13] Dunand M, Mohr D. Optimized butterfly specimen for the fracture testing of sheet
Evaluation of mechanical properties using shear-punch testing. Mat Sci Eng A materials under combined normal and shear loading. Eng Fract Mech
2005;395:307–14. 2011;78:2919–34.
[2] Atkins AG. Surfaces produced by guillotining. Philos Mag A 1981;43:627. [14] Roth C, Mohr D. Ductile fracture experiments with locally proportional loading
[3] Atkins AG. Mechanics of the transition from plastic flow to ductile fracture in histories. Int J Plast 2016;79:328–54.
guillotining. Proc 7th Eur Conf Fract (ECF-7) 1988;1:81. [15] Dunand M, Maertens AP, Luo M, Mohr D. Experiments and modeling of
[4] Atkins AG. On the mechanics of guillotining ductile metals. J Mater Proc Tech anisotropic aluminum extrusions under multi-axial loading – Part I: plasticity. Int J
1990;24:245. Plast 2012;36:34–49.
[5] Dodd B. Shear instabilities in blanking and related processes. Met Tech 1983;10:57. [16] Luo M, Dunand M, Mohr D. Experiments and modeling of anisotropic aluminum
[6] Dodd B, Atkins AG. Flow localization in shear deformation of void-containing and extrusions under multi-axial loading – Part II: ductile fracture. Int J Plast
void-free solids. Acta Metall 1983;31:9. 2012;32–33:36–58.
[7] Atkins AG. On cropping and related processes. Int J Mech Sci 1980;22:215. [17] Mohr D, Oswald M. A new experimental technique for the multi-axial testing of
[8] Zhou Q, Wierzbicki T. A tension zone model of blanking and tearing of ductile metal advanced high strength steel sheets. Exp Mechs 2007;48:65–77.
plates. Int J Mech Sci 1996;38(3):303–24. [18] Bai Y, Wierzbicki T. Application of extended Mohr–Coulomb criterion to ductile
[9] Wang K, Greve L, Wierzbicki T. FE simulation on edge fracture considering pre- fracture. Int J Fract 2009;161:1–20.
damage from blanking process. Int J Solids Struct 2015;71:206–18. [19] Pack K, Luo M, Wierzbicki T. Sandia fracture challenge: blind prediction and full
[10] Wang K, Wierzbicki T. Experimental and numerical study on the plane-strain calibration to enhance fracture predictability. Int J Fract 2014;186:155–75.
blanking process on an AHSS sheet. Int J Fract 2015;194(1):19–39. [20] Dunand M, Mohr D. Hybrid experimental–numerical analysis of basic ductile
[11] Luo M, Wierzbicki T. Numerical failure analysis of a stretch-bending test on dual- fracture experiments for sheet metals. Int J Solids Struct 2010;47(9):1130–43.
phase steel sheets using a phenomenological fracture model. Int J Solids Struct [21] Abaqus/CAE/Analysis User's Manual, Abaqus Inc. 2016.
2010;47:3084–102.

218