Académique Documents
Professionnel Documents
Culture Documents
To cite this article: H Hooputra, H Gese, H Dell & H Werner (2004): A comprehensive failure model for crashworthiness
simulation of aluminium extrusions, International Journal of Crashworthiness, 9:5, 449-464
This article may be used for research, teaching, and private study purposes. Any substantial or systematic
reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form to
anyone is expressly forbidden.
The publisher does not give any warranty express or implied or make any representation that the contents
will be complete or accurate or up to date. The accuracy of any instructions, formulae, and drug doses
should be independently verified with primary sources. The publisher shall not be liable for any loss, actions,
claims, proceedings, demand, or costs or damages whatsoever or howsoever caused arising directly or
indirectly in connection with or arising out of the use of this material.
A comprehensive failure model for
crashworthiness simulation of aluminium
extrusions
doi:10.1533/ijcr.2004.0289
1 2 2 1
H Hooputra , H Gese , H Dell and H Werner
1
BMW Group, Forschungs- und Innovationszentrum, Knorrstrasse 147, D-80788 Munich, Germany
2
MATFEM Partnerschaft Dr. Gese & Oberhofer, Nederlingerstrasse 1, D-80638 Munich, Germany
Abstract: A correct representation of the plastic deformation and failure of individual component
parts is essential to obtaining accurate crashworthiness simulation results. The aim of this paper is to
Downloaded by [McGill University Library] at 21:59 18 November 2012
© Woodhead Publishing Ltd 0289 449 IJCrash 2004 Vol. 9 No. 5 pp. 449–463
H Hooputra, H Gese, H Dell and H Werner
ε eq
*
Equivalent plastic strain at onset of instability However, this kind of model is limited to representing
γ
ε eq Equivalent plastic strain at onset of instability ductile fracture and as such ignores the fracture mechanism
for one orientation angle γ of the initial neck based on shear [1, 2]. In FEM analysis, failure prediction
relative to rolling direction using the Gurson model is highly dependent on mesh
refinement as this model causes strain softening prior to
ε eq
**
Equivalent plastic strain at onset of fracture
fracture. Considering these shortcomings prevalent in
ε S+ , ε S– Equivalent plastic strain in equibiaxial today’s numerical codes, there emerges a strong need for
tension/compression at shear fracture the development of a comprehensive approach for failure
ε T+ , ε T– Equivalent plastic strain in equibiaxial prediction coupled with a numerically robust implementa-
tension/compression at ductile fracture tion. This approach must be flexible enough to make use
η Stress triaxiality of the available discretisation used in today’s automotive
η+, η– Stress triaxiality in equibiaxial tension/ crash simulations (i.e. the use of shell elements with typical
compression at ductile fracture edge length between 5 and 15 mm). As localised necking
θ Shear stress parameter of thin sheets or profiles (necking in the direction of sheet
θ+, θ– Shear stress parameter for equibiaxial thickness) cannot be modelled with this discretisation, a
tension/compression. θ+ = 2 – 4kS, θ– = 2 + criterion for instability must be introduced in addition to
4kS the fracture models.
Downloaded by [McGill University Library] at 21:59 18 November 2012
ν Angle between extrusion direction and major Sheets and thin-walled extrusions made from aluminium
principal strain rate alloys generally fail due to one or a combination of the
σ 1, σ 2, σ 3 Principal components of stress tensor following mechanisms (Figure 1):
σeq Equiv. stress • ductile fracture (based on initiation, growth and
σm σm = (σ1 + σ2 + σ3)/3 coalescence of voids),
σ x, σ y Stress components in extrusion direction and • shear fracture (based on shear band localisation).
transverse direction • instability with localised necking (followed by ductile
τxy In plane shear stress component or shear fracture inside the neck area),
φ Ratio of maximum shear stress to equivalent
stress Instability is necessary as a third failure criterion if the
structure is discretised with shell elements. A more in-
depth explanation will be given in the section “Numerical
INTRODUCTION Model for Instability”.
The failure strains of the different mechanisms depend
Most of today’s crashworthiness simulation codes offer primarily on strain rate, temperature, anisotropy, state of
an incomplete selection of failure models. Often this stress and strain path.
selection is limited to simple fracture models based on This paper describes the derivation of three failure
the maximum strain criterion (i.e. the true fracture strain criteria for Instability, Ductile, and Shear fracture (IDS
is constant for all stress states). The problem with these failure criteria). The failure criteria are based on
models is that they do not take the dependence of the macroscopic stresses and strains. The criteria include the
fracture strain on the complete state of stress in a effect of anisotropy, state of stress and strain path. One
component into account. The result being, that these set of parameters is valid for one temperature and one
simplified approaches have resulted in inaccurate fracture strain rate regime (quasi-static or dynamic).
predictions [1]. On a more advanced level, there exist The IDS failure criteria have been integrated into the
some theoretical models based on mesoscopic effects. One software code CrachFEM1. CrachFEM is an add-on
example is the Gurson model, which accounts for the module to FEM codes which use an explicit-dynamic time
evolution of material porosity using a special yield criterion. integration scheme. CrachFEM transiently predicts failure
Ductile fracture Shear fracture Sheet instability
Localised neck
Figure 1 Visualisation of ductile fracture, shear fracture and sheet instability (localised necking).
1
CrachFEM is a trademark of MATFEM (GER)
AW-7108 T6 was examined in detail first (strain hardening, strains below about 20%, while above this value, the
yield locus). material shows a negative strain rate sensitivity due to the
adiabatic heating of the material.
CHARACTERIZATION OF MATERIAL PLASTICITY
Tube-shaped specimens extruded from the same batch
under similar process parameters were used to determine
Short tensile specimens have been cut out of the outer the initial yield locus of EN AW-7108 T6. Testing was
walls of a double chamber extrusion, at 0°, 45° and 90° to performed on a multifunctional testing machine with
the extrusion direction. The plastic anisotropy r (necessary hydraulic gear for the following load cases
for the IDS failure criteria calculation) and the strain
1. Axial tension
hardening parameters were derived experimentally using
2. Axial compression
these samples. Test results are summarised in Table 1.
3. Torsion
The results show that the plastic orthotropy is very
4. Axial tension (compression) with internal pressure
pronounced, r-values change from 0.327 in extrusion
5. Torsion with internal pressure
direction to 1.378 in diagonal direction.
6. Torsion with axial tension (compression) and internal
Static and dynamic, tensile and compression tests of
pressure.
prismatic specimens cut in the extrusion direction have
been performed to quantify the strain rate sensitivity of Additionally, ring specimens for uniaxial compression
the material and adiabatic flow stress curves. These are tests were cut from tubes. The yield locus was measured
required as input for commercial FEM crash codes with for an equivalent plastic strain of 2%, see Figure 3. The
an explicit-dynamic time integration scheme. A fully symmetric yield locus of Barlat et al. [3] in connection
coupled thermo-mechanical solution procedure is not with an associated flow rule has been used to describe
supported by these codes. The method used to obtain plastic deformation. It gives a good approximation in the
adiabatic flow stress curves from these experiments is tension-tension regime. However, it can be seen from
described in [1]. The adiabatic flow stress curves for Figure 3 that the case of pure shear (σx = –σy) is not well
aluminium alloy EN AW-7108 T6 are shown in Figure 2. represented. The experimental results display a slightly
There is no significant strain rate sensitivity for strain concave shape in this area. Strain hardening and strain
rates between quasi-static and 1 s–1. At higher strain rates, rate effects have been taken into consideration by an affine
2
PAM-CRASH is a trademark of ESI Group (F)
–0.5
sheet thickness, the necking process and subsequent
Sig–x fracture cannot be modelled directly in industrial crash
simulation (shell elements have a typical edge length of 3
Approximation–2% experiment–2% to 10 times the sheet thickness). The instability strain ε eq *
σy σy + ∆σy
Onset of
N3 fracture
N4 Lo
N2 ca
ls
Local strain
tra
N1 in Onset of
σx
instability
Glo
bal 45°
stra σ x + ∆σ x ε eq
*
in
Global strain
(a) Onset of instability (b) Onset of fracture (c) Local and global strain
Figure 4 (a) The stress distribution in a sheet may just induce the onset of instability; the strain distribution is still
homogeneous. (b) Slightly increasing the stresses leads to a local neck, immediately followed by fracture. The strain distribution
is inhomogeneous, showing a peak within the neck (local strain). In a numerical model, discretised with shell elements, only
the global strain can be evaluated from nodal displacements N1 to N4. Since the global strain does not change significantly
from onset of instability to onset of fracture, the former may be used as a slightly conservative fracture criterion. For the shell
elements, a length to thickness ratio greater than 5 should hold.
steel grades. However, the model does not cover the whole
π x′
FLC and is limited to isotropic materials. Hora et al. [8] h˜0 = h0 1 – d cos [3]
have applied a modified force criterion to non-linear strain l
paths. The model does not include kinematic hardening
and will therefore be limited in the quantitative prediction x′ indicates the local direction normal to the necking line
of necking in non-linear cases. This review shows that a according to Figure 5. The initial thickness h0 has no
need for a complete description of sheet instability for all influence on the numerical problem and is fixed to 1. The
strain paths still exists. ratio x′/l changes from 0 (neck center) to 1/2 (region of
The newly developed algorithm CRACH has been used the sheet with homogeneous deformation). d is the
in this study which includes a refined description of inhomogeneity parameter and its initial value is calibrated
material behaviour on a macroscopic level, in conjunction from the limit strain out of one experiment with the
with the effects of the material microstructure. The individual sheet. The parameter d increases with
mechanical problem is given in Figure 5. The basic concept deformation in the CRACH algorithm in order to account
is derived from the Marciniak model [4]. In this model for the roughening of the sheet during plastic deformation.
the localised neck is triggered by an initial imperfection The strain outside the neck can be increased
of the sheet. The sheet has an initial thickness of h0. The incrementally according to the strain history of a finite
thickness inside the infinitely small imperfection is given element. The strain inside the neck is calculated by CRACH
Downloaded by [McGill University Library] at 21:59 18 November 2012
in equation (1). using equation (3) and a strain rate dependent plasticity
model with isotropic-kinematic hardening.
h˜0 = h0 (1 – δ) [1] The equivalent flow stress of the material is defined by
δ is a very small number. Here and in the following chapters, ε˙eq
all values inside the localised neck are indicated with a σ eq = a1 ( ε 0 + ε eq ) n [4]
( ε˙eq ) ref
tilde (~). As the neck is infinitely small, the increase in
strain parallel to the neck (defined as local y′-direction in with strain hardening exponent n and strain rate sensitivity
Figure 5) is identical inside and outside the neck according parameter m, where n and m can differ between the
to equation (2). quasistatic and dynamic loading regimes. The plastic
orthotropy is defined by the Lankford coefficients r0, r45,
dε˜ ′y = dε ′y [2] and r90. An anisotropic yield locus according to Hill-1948
This model is practicable for linear strain paths in the is combined with a model for anisotropic strain hardening
region of ε2 < 0 (ε1 and ε2 are the principal strains in the according to Backhaus [9] to account for the Bauschinger
plane of the sheet with ε2 ≤ ε1), where the neck width is effect in the CRACH algorithm.
very small. In the region ε2 > 0 the neck can have a The global strain around the neck is increased
significant width. Therefore, the CRACH algorithm uses incrementally as long as the force equilibrium is fulfilled
an approximation of the neck cross section according to according to equations (5a) and (5b).
equation (3). σ˜ h˜ = σ h
x′ x′ [5a]
y σ˜ x ′y ′ h˜ = σ x ′y ′ h [5b]
The first increment without equilibrium indicates the
σy instability with the start of the localised necking. This
y′ σxy
mechanical problem has to be solved for different
x′
orientation angles γ according to Figure 5 of the initial
neck relative to the rolling direction of the sheet. The
γ
limit strain ε eq
*
for one deformation path is derived through
σx
x optimisation according to equation (6).
γ
l ε eq
*
= min {ε eq (γ )} [6]
γ
For some angles γ, the equivalent limit strain ε eq
can be
γ
high or even infinite. Therefore, an upper limit of ε eq =
1.2 has been defined. In summary, the CRACH algorithm
solves the plastic flow problem inside and outside the
initial neck area. Instability occurs if there is no common
X
h h̃
solution for the flow problem inside and outside of the
neck.
d = (h – h˜ )/h Additional features of the CRACH algorithm are:
Figure 5 Schematic representation of the imperfection • introduction of a heat treatment effectivity parameter
triggering a localised neck in the CRACH algorithm. ch(0 < ch < 1) to represent interstage annealing; ch > 0
reduces the effect of strain hardening and resets all for equibiaxial stress (η ≅ 2) can be higher than the
micro stresses to zero according to [10]; equivalent plastic strain at fracture in plane strain loading
• inclusion of the possibility to introduce new flow curves (η ≅ 3 ). Equation (10) represents a more general
between stages due to heat treatment (i.e. solution heat formulation and includes a non-monotonic decrease of
treatment and age hardening) or significant changes in the fracture strain with increasing stress triaxiality.
strain rate (i.e. for history deep drawing and crash).
ε eq
**
= d0 exp (–cη) + d1 exp (cη) [10]
PHENOMENOLOGICAL MODEL FOR DUCTILE AND where d1 is an additional material parameter. Of course,
SHEAR FRACTURE equation (10) includes the special case of equation (9) for
a monotonic decrease in fracture strain vs. stress triaxiality.
Two main mechanisms can cause the fracture of a ductile
However, equations (9) and (10) remain limited to
metal:
describing isotropic materials.
• void nucleation, void growth and void coalescence; A more general formulation of equation (10), which
• shear fracture due to shear band localisation. includes the orthotropy of fracture, must also include the
boundary conditions of the equivalent fracture strain ε T+
Most of the phenomenological fracture models are based
for the equibiaxial tension condition which must not be
on a fracture diagram which gives the equivalent plastic
Downloaded by [McGill University Library] at 21:59 18 November 2012
Shear fracture
0.2
For shear fracture, it is assumed that the equivalent strain
eps–1
at fracture ε eq
**
is a function of the variable θ given in 0.15
equation (15). Quasistatic
0.1
1 – ks η Dynamic
θ= [15] 0.05
φ
where ks is a material parameter and φ is the ratio of the 0
–0.1 0 0.1 0.2 0.3
maximum shear stress and the equivalent stress (von Mises) esp–2
according to equation (16).
Figure 6 Initial forming limit curves (FLC) predicted
τ with CRACH for the quasistatic and dynamic cases
φ = max [16]
σ eq (approx. 250 s–1) for specimens cut in the extrusion
Downloaded by [McGill University Library] at 21:59 18 November 2012
direction.
Analogous to equation (12), the equivalent plastic strain
for shear fracture with respect to θ is given in equation correlation of the flow stress curves at a strain rate of
(17). 1 s–1 and 100 s–1 yields the strain rate sensitivity of mdynamic
= 0.006 for the dynamic regime.
ε S+ sinh[ f (θ – θ – )] + ε S– sinh [ f (θ + – θ)] The CRACH algorithm (input: r-values and dynamic
ε eq
**
= [17]
sinh [ f (θ + – θ – )] strain hardening parameters as cited above) was used to
derive the inhomogeneity parameter d for the limit strains
where θ+ and θ– are the values of the parameter θ for of the dynamic tensile tests in the different orientations.
equibiaxial tension und compression. Equation (17) has A mean value of d = 0.0025 for the three orientations was
two constant parameters ε S+ , ε S– and one orientation used to calculate the dynamic FLC (see Figure 6). It may
dependent parameter f. No significant orthotropy of the be observed from the results that the static and mean
shear fracture curve has been found up to now for different dynamic inhomogeneity parameters are very similar, and
sheets and extrusions. Therefore, it is assumed that f is a that the dynamic FLC is somewhat lower than the quasi-
constant, independent of the orientation to extrusion static one.
direction.
Ductile fracture limit
DERIVATION OF FRACTURE PARAMETERS Different specimen geometries are used to define different
deformation states (i.e. plane strain, equibiaxial strain etc.).
Instability For the derivation of the fracture parameters, the stress
The limit strain ε 1* of the initial FLC in uniaxial tension state parameters η (defined in equation (8)) and φ for the
specimens is derived from the specimen thickness and shear fracture curve (defined in equation (16)) must be
width measured at a distance of twice the sheet thickness known, and thus a material model must be introduced for
from the fracture line (outside the localised neck), assuming this purpose. For this analysis, an isotropic von Mises
volume constancy during the plastic deformation. The yield locus was used. This yield locus does not exactly fit
mean value of three quasi-static experiments in extrusion the materials plasticity results, but it does make the fracture
direction resulted in a strain limit of 0.125 . The algorithm model more robust and more general. This does not cause
CRACH (input: r-values r0 = 0.327, r45 = 1.378, r90 = an incompatibility during the calculation because the
0.965 and strain hardening coefficients astatic = 596 MPa, material plasticity in the FEM module and the fracture
ε0,static = 0.02, nstatic = 0.143 in extrusion direction; quasi- models are uncoupled.
static strain rate sensitivity of mstatic = 0) was used to Erichsen test (equibiaxial stress with η = 2), three point
derive the inhomogeneity parameter d for this limit strain bending of sheet coupons (width/thickness > 4 with plane
(d = 0.0028). The quasi-static FLC predicted by CRACH strain tension and η = 3 ; tests under 0°, 45° and 90° to
is shown in Figure 6. extrusion direction) and waisted tensile specimens with
An approximation of the flow stress curves from tensile fracture at the notch root (uniaxial tension with η = 1)
tests at 250 s–1 was used to derive the parameters of the have been used to obtain the ductile fracture limit. The
Swift equation for CRACH in the dynamic regime (adynamic local fracture strains have been derived from a grid on the
= 572.9 MPa, ε0,dynamic = 0.032, ndynamic = 0.122)3. A surface of the specimens (Erichsen test and three point
3
This approximation is only used for the prediction of instability with algorithm CRACH. The Swift parameters must not be used for the
extrapolation up to higher strains since this function type cannot account for softening effects due to adiabatic heating.
1.5
Experiment–0 shown in Figure 8 together with the experimental data.
Experiment–45
1 Experiment–90 Table 3 Material parameters in equation (17) for the shear
fracture limit in quasistatic and dynamic case (EN AW-
7108 T6)
0.5
Quasi-static Dynamic
0
0 0.5 1 1.5 2
ε S+= 0.26 ε S+ = 0.35
Stress triaxiality eta
ε S–= 4.16 ε S– = 1.2
f = 4.04 f = 2.05
Figure 7 Ductile fracture limit curves ε **
eq ( η ) together
with experimental data for quasi-static case in orientation The shear fracture limit curves are plotted in Figure 9 for
of 0°, 45° and 90° to extrusion direction. the quasistatic case and in Figure 10 for the dynamic case
together with the instability limit and the ductile fracture
For the characterization of the loading path for all three limit. For the characterization of the loading path for all
types of limit curves, α = ε˙2 /ε˙1 (α = ε2/ε1 for linear three types of limit curves, α = ε˙2 /ε˙1 (α = ε2/ε1 for
strain paths) is used as a common measure. For purposes linear strain paths) is used as a common measure. The
of comparing these results with the other fracture limits, derivation of α from the stress triaxiality η and the
the ductile fracture curves are shown in Figure 9 for the parameter θ is given in Appendix A.
1.5
Instability
Ductile fracture 0
0.5
0
–2 –1 0 1
alpha = phi–2/phi–1
Figure 9 Quasistatic failure diagram for extrusion EN AW-7108 T6. Plotted limit curves are valid for linear strain paths
and membrane loading.
1.5
Ductile fracture 0
Ductile fracture 45
Equivalent plastic strain
Ductile fracture 90
1 Shear fracture
Instability
0.5
0
–2 –1.5 –1 –0.5 0 0.5 1
Alpha = Phi–2/phi–1
Figure 10 Dynamic failure diagram for extrusion EN AW-7108 T6: Plotted limit curves are valid for linear strain paths
and membrane loading.
(a) Quasistatic three point bending test (b) Simulation of three point bending with IDS
failure criteria. (5 mm edge length of
shell elements)
Figure 12 Three point bending test - Fracture pattern from tests and simulation.
Figure 13 Static and dynamic compression tests of double chamber extrusions (center) compared to simulations with and
without IDS failure criteria (left and right, respectively). Shell elements with 5 mm edge length were used.
Shear fracture is initiated at the T-joint in the plane of failure of individual component parts is essential to
symmetry of the profile. In the dynamic case, see Figure obtaining accurate crashworthiness simulation results. A
13 bottom, nearly no folding occurs. The wall segments comprehensive approach for predicting failure in structural
are separated completely in the corners and at the T-joint components based on macroscopic strains and stresses
between the middle wall and the outer walls, which is in using the CrachFEM code has been presented. This
good agreement with the experiment. approach fits to the state-of-the-art in discretisation of
The predominating fracture mode occurring in the automotive crash simulation models (shell elements with
quasi-static and dynamic compression tests is shear and edge lengths of 5 to 15 mm). An edge length of 5 mm is
ductile fracture. It seems likely that aluminium alloys recommended in areas of high strain gradients. Due to
generally tend to be highly sensitive for shear loading in the absence of adaptive meshing procedures, these critical
the dynamic case (Figure 13 – dynamic compression test). areas have to be identified in advance. CrachFEM includes
all relevant failure mechanisms, such as Instability (localised
CONCLUSIONS necking), Ductile and Shear fracture (IDS failure criteria).
All failure criteria are implemented in a way to include
A correct representation of the plastic deformation and the influence of non-linear strain paths.
Failure risk
parameter
0.2
0.4
0.6
0.8
1
Figure 14 Quasi-static compression of EN AW-7108 T6 double chamber extrusions. Left: arrows point to ductile fracture
sites of inward bent folds in the experiment. The fringe plots display the failure risk parameter for the three modes.
Failure is to be expected if the failure risk parameter is greater or equal to 1. The failure modes and their respective crack
Downloaded by [McGill University Library] at 21:59 18 November 2012
location are in good agreement with the experiment. NOTE: element elimination is suppressed in the simulation; all
displacements are scaled by a factor of 0.25 to “unfold” the profile.
All studies presented in this paper were carried out on excellent experimental work on the measurement of the
extrusions made from aluminium alloy EN AW-7108 T6. yield locus and Dr. Andrew Heath for programming the
The plasticity of these extrusions has been examined interface in PAM-CRASH to CrachFEM.
experimentally. A Barlat yield criterion has been used to
model the plastic orthotropy. The IDS failure curves of REFERENCES
EN AW-7108 T6 are strongly dependent on the stress
state and strain rate of the material. The ductile fracture 1. EL-MAGD, E, GESE, H, THAM, R, HOOPUTRA, H and
limit curves and the instability curves show a strong WERNER, H. ‘Fracture Criteria for Automobile
orthotropic behaviour, whereas the shear fracture limit Crashworthiness Simulation of wrought Aluminium Alloy
Components’, Mat.-wiss u Werkstofftech, 2001 32 712–724.
curves show no significant dependence on the extrusion
2. SCHMITT, W, SUN, D Z, BLAUEL, J G and CHRISTLEIN, J.
orientation direction. To obtain an accurate failure
‘Improved Description of the Material Behaviour of
prediction, this anisotropic behaviour has to be taken into Aluminium Automobile Components by the Gurson
account in the simulation. Model’, Proceeding of the 31st International Symposium on
The comparison of numerical results to test data for Automotive Technology and Automation, Düsseldorf, 1998.
the three point bending and axial compression tests of 3. BARLAT, F, LEGE, D J and BREM, J C. ‘A six-component
double chamber extrusions demonstrates a comprehensive yield function for anisotropic materials’, Int J Plasticity,
approach to accurately predict component failure, both 1991 7 693.
in terms of the mode and the location of cracks. Due to 4. MARCINIAK, Z, KUCZYNSKI′ , K and POKORA, T. ‘Influence of
the loading conditions in all of the examples, instability the plastic properties of a material on the forming limit
did not show up as a dominating failure mode. However, diagram for sheet metal in tension’, Int. J. of Mechanical
Sciences, 1973 15 789–805.
in loading situations where membrane tensile strains
5. DELL, H, GESE, H, KEßLER, L, WERNER, H and HOOPUTRA,
prevail, instability will be of great importance. An example
H ‘Continuous Failure Prediction Model for Nonlinear
for such a case is shown in the publication from Pickett et Load Paths in Successive Stamping and Crash Processes’,
al. [12]. New Sheet Steel Products and Sheet Metal Stamping (SP-
The presented failure approach, however, can only 1614), SAE 2001 World Congress, Michigan, SAE-Paper
predict the crack initiation. The element elimination used 2001-01-1131, 2001.
in the simulation, after the onset of fracture, represents 6. HILL, R. ‘On discontinuous plastic states with special
only a preliminary approach for simulating crack reference to localised necking in thin sheets’, J Mech Phys
propagation. A suitable criterion for crack propagation in Solids, 1952 1 19–30.
combination with a numerical implementation which is 7. CAYSSIALS, F. ‘A new method for predicting FLC’, IDDRG
mesh independent to the greatest possible extent remains Conference, Geneva, 1998 443–454.
8. Hora, P, Tong, L, Reissner, J. ‘A Failure Criterion for Prediction
a challenge for future development work.
of Strain Path Dependent Failures for Quadratic and Non-
Quadratic Yield Loci’, Proceedings of Numisheet, 1996.
ACKNOWLEDGEMENT 9. BACKHAUS, G. ‘Plastic deformation in form of strain
trajectories of constant curvature – Theory and comparison
The authors would like to thank Dr. V. Yelisseyev and his with experimental results’, Acta Mechanica, 1979 34 193–
co-workers at company TEST in Voronezh (RUS) for the 204.
10. DELL H and ELISEEV‚ W W. ‘Materialmodell für 12. PICKETT, A, PYTTEL, T, PAYEN, F, LAURO, F, PETRINIC, N,
mehrstufige Umformung mit Wärmebehandlung zwischen WERNER, H and CHRISTLEIN, J. ‘Failure prediction for
den Stufen’, Iswestija AN SSSR Metalli 1991 4 171–174. advanced crashworthiness of transportation vehicles’,
11. KOLMOGOROV, W L. ‘Spannungen Deformationen Bruch’, Int. J. of Impact Engineering, 2004 Vol. 30 Issue 7
Metallurgija, 1970 230. 853–872.
Downloaded by [McGill University Library] at 21:59 18 November 2012
The plane stress condition yields –2 ≤ η ≤ + 2. Figure A1 illustrates the dependencies of the parameters
α, η and θ for an arbitrary state of plane stress. For a given
η 2 – 6 + η 12 – 3η 2 state of plane stress, σ1, σ2, σ3 = 0, the corresponding value
α= with the special cases of of the parameter is displayed normal to the σ1 – σ2 plane
2 (3 – η 2 )
of the von Mises yield locus. It is evident from the left side
of Figure A1 that all parameters show a plane of symmetry
α=0 for η= 3 which is normal to the line σ1 = – σ2. Therefore, the right
hand side of Figure A1 shows a side view along the line σ1
α ⇒ –∞ for η= – 3 = – σ2 in the plane of the von Mises yield locus.
(a) Parameter α α α
σ2
1
1
0.5
σ1
von Mises
yield locus –0.5
von Mises –1
yield locus –1
(b) Parameter η η η
σ2
2
2
1
1
σ1
2
–2
(c) Parameter θ θ θ
σ2
2.5 2.5
2 2
1.5 1.5
2 2
0.5 0.5
σ1
von Mises
yield locus
von Mises
yield locus
Figure A1 Parameters α, η and θ displayed as a function of principal stresses σ1 and σ2. As explained in the text,
Downloaded by [McGill University Library] at 21:59 18 November 2012
each parameter is plotted normal to the von Mises yield locus. The parameter α is only shown in the relevant area for
instability. The dark shaded areas of parameter θ indicate that the in-plane shear stresses are most critical. The light
shaded areas indicate that the out-of-plane shear stresses are most critical. A value kS = 0.1 is used to construct the
θ-dependency.