Methods Material Models and Crash Simul in LSDYNA
Methods Material Models and Crash Simul in LSDYNA
Methods Material Models and Crash Simul in LSDYNA
simulation in LS-DYNA
Methods -Material models and Crash
simulation in LS-DYNA
Contents
Contents
Contents ............................................................................................................. 9
Introduction ........................................................................................................ 1
Theory ................................................................................................................ 3
2.7 Optimization..........................................................................................10
3 ................................................................................................................................ 12
4................................................................................................................................ 17
Method .............................................................................................................. 17
6............................................................................................................................... 42
Conclusion ....................................................................................................... 42
7 ............................................................................................................................... 43
8 .............................................................................................................................. 46
Appendix 1 ....................................................................................................... 46
x
List of Figures
2.1 Von mises and Tresca yield criterion in σ1 and σ2 [8] ................................. 6
2.2 Tresca yield surface [8] ................................................................................. 6
2.3 Isotropic hardening [8] .................................................................................. 7
2.4 Kinematic hardening [8] ............................................................................... 8
2.5 Mixed hardening [10] .................................................................................... 9
2.6 Damage Mechanics [9]................................................................................. 10
2.7 Optimisation Process .................................................................................. 10
2.8 Optimisation Results................................................................................... 10
3.1 Yield locus correction (Image from CrachFEM User’s manual) [30] ........ 15
xi
List of Figures
xii
List of Tables
List of Tables
NOMENCLATURE
CAE - Computer Aided Engineering
PD - Product Development
DP - Dual Phase
FE - Finite Element
xi
ii
1
Introduction
1
1. Introduction
Based on available models, the variability of material properties and other conditions
cannot be mapped consistently and this has led to indeterministic models [6]. Hence,
the demand for precise material behaviour and material failure under different stress
states, strain rates, etc., have been found to be important and this requires extensive
experimental test results [7].
To calibrate a virtual material model, experimental test results (representing dif-
ferent multi-axial stress state behaviour) have been reported in the literature. The
main objective of this thesis was to contribute to the development of a virtual mate-
rial test laboratory. China Euro Vehicle Technology (CEVT), a major research and
development center in Sweden, has implemented the MFGenYld+CrachFEM ma-
terial model (commonly known as CrachFEM material model) for their simulations
involving structural analysis. This material model (developed by MATFEM) has been
based on the failure criteria of the material. These CrachFEM material mod- els have
versatile failure modelling capabilities using a phenomenological approach. However
these models have a high computational cost with an additional licensing expenditure.
The primary focus of the current work is to study the comprehensive material be-
haviour of Dual Phase steel can be captured in uniaxial tensile test through sim-
ulation. Here the material is assumed to be isotropic, homogeneous and free from
residual stresses. In this thesis the material model developed is focused on the
experimental tensile test results.
2
2
Theory
h2
u̇n+1 = u̇n + [ü + ün+1 ] (2.2)
4 n
While 0th order derivative uses,
h2
un+1 = un + hu̇n + [ün + ün+1] (2.3)
4
3
2. Theory
According to Equations 2.2 and 2.3, an equation system for ü is obtained in terms
of u̇n and & un .
The time step size (h) is bounded by the largest natural frequency of the structure
which in turn is bounded by the highest frequency of any individual element in the
FE mesh.
2.2 Steels
Dual phase steels are one of the important variants of high strength steels which are
comprised of a ferrite matrix and dispersed phase of martensite/restained austenite
and/or bainite. Development of these DP steels has resulted in steel grades with
high tensile strength and elongation properties to establish fatigue and crash resis-
tance. Cold rolling of low alloy steels followed by inter critical annealing are one of
the widely accepted methods to produce DP steels. These steels are mainly used
in the body panels, wheels, B pillars, bumpers and other important automotive
components.
Advantages of DP steels are as follows:
• Low yield strength
• High strain sensitivity
• Good fatigue resistance
• High deformation hardening that leads to high energy absorbing ability (Crash-
worthiness).
σf = kε˙m (2.5)
where σf is flow stress, ε˙ is strain rate, k is a constant and ’m’ represents the strain
rate sensitivity (varies with temperature and grain size)
4
2. Theory
Predominant theories used in engineering that define yield surfaces are as follows:
1. von Mises surface
This theory is based on the assumption that yield of the material begins when the
principal stresses induced in the material satisfies the following relation:
3
σy = √(2) (σ 1− σ m )2 + (σ 2 − σ m )2 + (σ 3 − σ m )2 (2.8)
5
2. Theory
Figure 2.1: Von mises and Tresca yield criterion in σ1 and σ2 [8]
6
2. Theory
plastic deformation. Increase in yield strength lead to a change in the yield surface
because the yield surface is a generalised term of yield strength. This phenomenon of
subsequent transformation of yield surface (yielding behaviour) due to plastic load-
ing is called the hardening rule [12]. Generally a hardening behaviour is predicted
using phenomenological models while there are many other models like Bingham
viscoplastic model, Maxwell rheological model, Perzyna’s formulation etc. that also
describes the behaviour of the material [17]. Based on the response of the yield
surface during the plastic loading and elastic unloading, hardening behavior is often
broadly classified into three main types namely
1. Isotropic hardening plastic behavior
2. Kinematic hardening plastic behavior
3. Distortional hardedning or Mixed isotropic and kinematic hardening plastic
behavior
where,
• κ = micro hardening stress or drag stress representing the hardening state of
the microstructure of the material. = -Hk
• H = Hardening modulus that is associated with plastic strain and H>0
• k = internal hardening variable
• |σ| = Applied load
The flow stresses are described as a function of equivalent plastic strain.
7
2. Theory
8
2. Theory
evaluating the structural integrity of the material. Failure can be broadly classified
as microscopic failure and macroscopic failure.
Microscopic failure mainly deals with crack initiation and propagation. Conse-
quently, failure leads to fracture i.e. when the material looses the capacity to bear the
applied load, cracks are generated which in-turn finally leads to separation of the
material. Failure of material might even be due to material and geometric in-
stabilities. Fracture is defined as material disintegration into two or more fragments
under the action of external loads. Fracture can be classified into two types namely
ductile fracture and brittle fracture. The phenomenon of fracture mainly depends
on the type of material, temperature, state of stress, rate of loading etc. [15].
Ac Atotal − Areduced
D= =
Atotal Atotal
When damage tends towards unity, the material will fail as the damage area is same
as nominal area. Reduction of cross section has a direct effect in the stress values,
thereby stiffness of the material is reduced according to Equation 2.13.
σ∗ = σ[1 − D] (2.13)
where σ∗ = Effective stress
D = Damage
9
2. Theory
σ = Instantaneous stress
2.7 Optimization
Optimisation is defined as the process of identifying the most suitable solution under
prescribed conditions by maximising the required objective function and minimising
the errors that might exist. Cost effectiveness, efficiency and accuracy are some of
the major parameters considered while performing optimisation. Optimisation tech-
niques are also used in stochastic analysis and the solution of system identification
problems.
There are many techniques to solve arbitrary nonlinear optimisation tasks. A few
of the generally used methods are Successive Response Surface Method, Multidisci-
plinary optimisation, Sequential with Domain reduction method etc. According to
Figures 2.7 and 2.8, optimisation techniques have been broadly carried out in two
stages namely: Optimisation process and Optimization results.
material properties and governing equations are applied over these elements and
expressed in terms of unknown values. A set of equations are generated by combin-
ing all the constraints along with the loads applied. Solutions for set of equations
is achieved by various numerical techniques. The solutions obtained predict the
approximate behaviour of the systems [14]. Applications of FEM range from stress
analysis of a complex system like aircraft, automotive, building, etc to thermal anal-
ysis of heat flux, magnetic flux, and other flow problems. There are two main types
of elements used in this thesis, shell and solid elements.
1. Solid elements are three dimensional finite elements used to model structures
without geometrical simplifications. The use of 3D finite solid elements is proved to
be advantageous as it doesn’t require geometric, constitutive and loading assump-
tions. Pragmatic boundary conditions are applied. FE model developed using 3D
solid elements replicates the actual CAD model [22]. Solid elements are usually
modelled in tetrahedron or hexahedron shapes according to the CAD model. But
FE modelling of solid elements makes a design computationally expensive.
2. Shell elements are the simplified formulations of solid elements in a finite element
approach to reduce the model and computational time. They are one of the most
sophisticated elements as they exhibit both membrane and bending deformations.
Shell elements are formulated on the principle of virtual work for 3D continuum
models [13].
11
3
Material Modelling
3.1.1 *MAT_ADD_EROSION
Failure and erosion of the elements are not supported by many of the consti-tutive
models in LS-DYNA. Determinination of integrity of the structure is as- sessed based
on material failure. Hence to achieve failure and erosion in com-plex load cases,
LS-DYNA provides an additional supportive material model called
"*MAT_ADD_EROSION" which has to be coupled with other constitutive models.
The main theory behind the model is to define failure criteria and once the ele-
ment satisfies the necessary criterion, the element is deleted. This model provides two
failure models namely "Generalized Incremental Stress-State dependent dam- age
MOdel" (GISSMO) and Damage Initiation and Evolution Model (DIEM)
models. These models have options used to define the failure of the material based
on different conditions.
3.1.1.1 GISSMO
Generalized Incremental Stress-State dependent damage MOdel" (GISSMO) is
a damage model available in LS-DYNA that allows incremental softening and fail-
ure, that leads to mapping the process of damage accumulation. GISSMO primely
uses a constitutive model to predict the uniform plastic behavior. It is intended to
provide possible inputs of a damage definition. An instability curve, failure curve,
damage exponent and fading exponent etc. are the necessary parameters to define the
GISSMO model
12
3. Material Modelling
DMGEXP × D(1 −( ))
1
DM GEXP
∆D = ∆ϵp (3.1)
ϵf
where,
The rate of damage is integrated to obtain a relation between plastic strain and
failure strain.
ϵp
D=( )DM GEXP (3.2)
ϵf
where ϵf = constant
σ∗ = σ(1 − D) (3.3)
DCRIT (function of triaxiality and equivalent plastic strain to failure) and FADEXP
(function of triaxiality and instability strain) are the parameters which controls post
critical behavior that governs softening and damage accumulation progressively.
n
F (1 − )∆εp
1
∆F = (3.4)
n
∗
ε(η )
!F ADEXP
D−D
= [1 ] (3.5)
CRIT
σ∗ σ −
1 − DCRIT
13
GISSMO allows for calculation of pre-damage for forming and crashworthiness sim-
ulations.
Triaxiailty (η) is one of the other important parameters that GISSMO depends on.
This triaxiailty is the measure of current stress state and mathematically is theratio of
hydrostatic stresses and von Mises stress. Lode parameter (L) is also an additional measure
of current stress state.
σH
η=
σV M
where ,
14
3. Material Modelling 3. Material Modelling
3.1.1.2 DIEM
Damage Initiation and Evolution Model (DIEM) is a damage model which is de-
fined by specifying different damage initiation and evolution criteria. Initiation/Evo-
lution types are defined at each integration point consisting of a damage initiation
variable (ωDi ) and evolution history variable (Di) such that
i
ωD ∈ [0, ∞)Di ∈ [0, 1] (3.8)
where i=1,2,. .... n
Damage evolution governs the damage of the material. Damage initiation might be
caused with evolution types as mentioned below:
1. Ductility based on stress triaxiality (Shear initiation, Forming Limit Diagram
(FLD) initiation criteria)
2. Ductility based in normalised principal stress
Advancements in CAE industry, has facilitated robust material models that repli- cate
the physical test results effectively and efficiently. Many of the automotive OEMs
(Original Equipment Manufacturer) found the process of defining material models
(using the available information) in virtual environment to be streneous and time
consuming. Based on the solver profile and with the coordination of material
suppliers and the industry, this requirement was fulfilled by an external distributor.
where,
• k = stress-state dependent correction factor
• F (σij) = symmetrical reference yield locus
CrachFEM consists of three different criteria for localised necking, shear fracture and
ductile fracture. It usually considers the equivalent plastic strain (ϵeq) and failure
plastic strain (ϵf ) to deduce the fracture. Therefore fracture risk (ψ) can be defined
as
ϵeq
ψ= (3.10)
ϵf
15
which is usually based on kinematic hardening model. Ductile fracture is computed
Figure 3.1: Yield locus correction (Image from CrachFEM User’s manual) [30]
by considering triaxiality and strain rate while the shear fracture is computed by
considering the shear stress parameter and plastic strain rate.
16
3. Material Modelling
4
Method
The thesis work was broadly classified into two tasks. The first task involved mod-
elling of coupon test specimens with fine meshes of solid elements and validation
of those results against the experimental test results. The second task involved
conversion of the solid element coupon test models to shell element models (in-
cluding damage model) also validated against the experimental test results. These
results are also compared to results obtained from CrachFEM material models. The
schematic work flow is illustrated in the Figure 4.1.
All FE models were based on the experimental test procedures conducted on DP steel
by an external supplier - MATTEST. Nine different experiments were con- ducted
by the supplier to characterise the material. The detailed description of the specimens
and test procedures are discussed below. The tests fall under two cate- gories, tests
to describe plastic behavior and tests that describe fracture.
17
4. Method
List of Experiments:
• Uniaxial tension
• Torsion
• Layered Compression
• Tension of specimen with groove
• Tension of specimen with hole
• Unibiaxial test
• Shear test
• Bend test
17
4. Method
18
4. Method
Figure 4.7: Schematic representation of Tension test (specimen with a hole) [30]
19
4. Method
Figure 4.8: Schematic representation of Tension test (specimen with a waist) [30]
Figure 4.9: Schematic representation of Tension test (specimen with a groove) [30]
Shear test
Since the torsion test was prone to instability inside the circular groove before frac-
ture, another test was performed to determine the effective plastic strain at fracture
under pure shear. This rectangular specimen with two parallel grooves was loaded in-
plane at the centre of the specimen which causes shear deformation in the groove.
Note: Exact experimental setup and boundary conditions were unknown.
20
4. Method
21
4. Method
From Figure 4.13, it was observed that the initial elastic region and plastic region
until the ultimate strength limit might not match with the experimental test results.
Hence in the current study, this method was found to be not feasible to proceed in the
development of virtual test lab. Another approach to overcome this problem, was to
consider the true stress-strain curve (from the experimental test result) until the
ultimate strength point and extrapolating it using different hardening rules.By
this method, experimental results could be matched exactly up to the ultimate strength
limit.
A new methodology has been implemented with the help of an optimiser and same is
explained in the following section. A meta model-based optimization setup in LS-
OPT is shown in Figure 4.14. The two stage optimization process is explained below.
1. Setup : In the global problem setup stage, the end points of the true stress-
strain curve at the ultimate strength i.e stress and stain values defined as
parameters in the ".key" file were automatically detected. These parameters
were then defined as a continuous variables with lower and upper boundaries
assigned to the initial value.
2. Sampling : Different metamodel settings were defined in this stage. A polyno-
mial, quadratic order Successive Response Surface Method (SRSM) was used
in the optimization study with Latin Hypercube point selection for sampling.
3. Flowcurve : User defined python script computes and extrapolates the input
true stress-strain curve from the experiment using the parameters defined in the
setup stage and the extrapolated flow curves in each iteration were copied to
the respective iteration folders in the ’Tensile’ stage for the main simulation.
22
4. Method
The program extrapolates the curve with strictly increasing slope condition to
avoid negative slope in the flow curve.
4. Tensile : The flow curve from the previous stage was used as an input curve
to the material model *MAT_24 and simulation of uniaxial tensile test was
performed. Time histories like nodal output (NODOUT) and section forces
(SECFORC) were defined to plot the stress-strain curve from the simulation. A
curve matching objective was defined as a composite in this stage. This
composite matches the history with the response and combines the response
surface into a single expression.
5. Optimization : After the simulation, the metamodel, which is nothing but an
approximation to the design response, starts to build according to the meta-
model setting. An optimization process starts with a partial curve mapping
algorithm to match the simulation curve with experimental curve as per the
objective defined in the previous stage.
6. Domain reduction : After each iteration, SRSM reduces the domain with
all points within a sub-region. It moves the region of interest towards the
optimal solution within the design space by ignoring the points belonging to the
previous iterations. If the approximated solution converged, SRSM writes the
final solution. If not, the loop continues with the reduction of domain towards
the optimum until it meets the termination criteria.
7. Termination : Each design terminates after a specific number of iterations
specified in this stage along with the tolerance between the design, objective
and metamodel accuracy.
The schematic representation of optimization process is shown in Figure 4.15. Simu-
lation history, optimization history and the metamodel response surface was plotted
in the post processing. Based on the results obtained from the optimization process,
suitable design and corresponding input flow curve was selected. From Figures 4.16
and 4.17, the simulation history and the metamodel response surface can be viewed.
23
4. Method
24
4. Method
Figure 4.18: Tension test model Figure 4.19: Groove in shear model
25
4. Method
Different failure options and damage models could be defined through *MAT_ADD_EROSION
model. Some of the important parameters to define and calibrate damage model
have been given:
• MID : Material ID
• IDAM : Flag for damage model
• DMGTYP : Damage coupling to stress
• LCSDG : Load curve defines equivalent plastic strain to failure vs. triaxiality
• ECRIT : Critical plastic strain
• DMGEXP : Exponent for non linear damage accumulation
• FADEXP : Exponent for damage related stress fadeout
• LCREDG: Load curve ID defining element size dependent regularization fac-
tors for equivalent plastic strain to failure
The MID of *MAT_ADD_EROSION should be same as the MID of *MAT_24 in
order to link the damage model to the constitutive model. The GISSMO damage
model was initialised by switching the flag IDAM to 1.
26
4. Method
Figure 4.24: Failure strain vs tri- Figure 4.25: Critical strain vs tri-
axiality axiality
27
4. Method
The combination of DMGEXP, FADEXP and EPSF plays a significant role in the
overall behaviour of a damage model in predicting the fracture in different stress
states. The optimisation technique with LS-OPT was used to understand the influ-
ence of these parameters on each other. Sensitivity of parameters was measured by
defining each parameter as a variable having the other two as constants.
28
5
Results and Discussion
The optimization technique was used to extrapolate the hardening curve (used in the
simulations), to predict the failure in the material and the optimal design (hardening
curve) from the uniaxial tension test was used to carry out all other load cases.
The solid FE models developed in this study have exhibited good correlation with the
experimental test results results and predicted the different stress states close
proximity to the theoretical triaxiality.
All the necessary information like the triaxiality, failure strain and critical strain was
extracted from the solid FE models and the optimization was extensively carried
out to identify the parameters DMGEXP and FADEXP to calibrate the GISSMO
damage model. Three good designs (set of parameters) were selected from the
optimization which resulted in satisfying results in the uniaxial tension test. The three
set of parameters are included in Table 5.1.
The tension tests with different element sizes ranging from 0.5mm to 5mm were
executed for mesh regularization and the failure strain was scaled across the element
size range. Figure 5.1 is the scaling curve for mesh regularization and Figure 5.2
shows the interaction of the post critical region to element size.
29
5. Results and Discussion
Figure 5.3 shows the variation of fracture stain is scaled for different element sizes
so that result is identical for all element size between 0.5mm to 5mm. Figure 5.4
shows the von Mises stress fringe plot at the time of necking for different element
sizes. The calibrated material model was used in all load cases and the results are
presented in the following section.
30
5. Results and Discussion
Eight tests were conducted virtually on DP steels and the results are presented
in the current section. The plots mainly include experimental test results, solid model
results (using *MAT_24) and shell model results (using CrachFEM and
*MAT_24+GISSMO). Specially, the uniaxial tension test plot in Figure 5.5, com-
prises shell model results from all 3 sets of values in Table 5.1, out of which the
second set showed promising results not only in tension but in other tests as well.
Hence, only the Set 2 results of shell element models have been presented in rest of
the plots.
31
5. Results and Discussion
Both solid and shell models overestimated the force level in the tension test of a
specimen with a groove. This was due to the instability in the groove and lack of
information about the physical test procedure. On the contrary, it can be observed
in Figure 5.8, the results from the CrachFEM material model effectively captured the
material failure in the specimen with groove. The same CrachFEM material model in
other tension tests showed early softening.
32
5. Results and Discussion
33
5. Results and Discussion
The solid and shell models are in good agreement with the experiments in three- point
bending test. However, both models demonstrated slightly stiff behavior. This is due
to the variation of physical test (consisting of roller supports) compared tothe
numerical test (consisting of fixed supports). Also the friction values used inthe
experimental test were unknown. CrachFEM, on the other hand, was found to
underestimate the force levels.
The Figure 5.10 shows the results obtained for the biaxial stress state in the Erichsen
test where the solid element and shell element models (through simulations), match
well with the experimental test result. The solid model revealed a crack quite
later, when compared to the experiment in the biaxial test or Erichsen test. The shell
model was in concurrence with the experimental test result by predicting the crack at
the right displacement of the punch. The simulation involving CrachFEM material
model, depicted an early failure in the test.
Figure 5.11 shows the correlated results for the layered disc compression test. The
test was conducted using solid elements to calculate the material hardening behavior
at higher strains. The discs were compressed close to half of its initial height and the
increase in the discs diameter was measured and was found to be in accordance with
the experimental test result. The shear model with a butterfly notch illustrated
right deformation behaviour despite an early failure. The force levels were also
underestimated when compared to the experimental test result. Lack of information
about the experimental test procedure might be the reason for inaccurate results in the
shear test simulation.
34
5. Results and Discussion
35
5. Results and Discussion
36
5. Results and Discussion
The dimensions of the cracks after the failure of solid specimens were measured in
all load cases and were compared against the dimensions of cracks measured from
the physical tests. Overall, a 95% accuracy was achieved with the solid element
specimens in comparison with physical tests. The width and thickness at the crack
centre was measured in all tension test models. The ductile shear fracture was ob-
served in all tension tests.
The shear test demonstrated good qualitative behavior in comparison with the ex-
periment but the quantitative results failed to do so (due to early failure). This is
due to the fact that the complete information on the physical test was unknown.
The bending test (involving both tension and compression stress states) exhibits
no fracture in the simulation similar to the physical test. Deformation behaviour
of unibiaxial test specimen exhibited good correlation with experimental test result
but with a crack initiation slightly later in the simulation.
A sheet with a grid was used on the test specimens in bending and unibiaxial tests to
measure the strain in the experiments. The same grid dimensions were represented
in the simulations using null shell elements which can be viewed in the Figure 5.13.
The grids were also used to understand the crack initiation distance from the locus.
Crack dimensions of all specimens have been presented in the Table 5.3.
37
5. Results and Discussion
The accuracy of the solid element models was also measured by triaxiality in all stress
states. The triaxiality of all solid element models was plotted against their respective
failure strains. All solid models were validated with theoretical triaxiality values
which can be seen in the Figure 5.14. The uni-biaxial test (involving complexstress
states) also indicated the correct triaxiality at failure with 0.667.
38
5. Results and Discussion
Front side members of the vehicle structure play a vital role in crash manage-
ment in frontal crashes. The new material card was assigned to the front mem-
bers of Lynk & Co 01 vehicle for validation. Two tests were conducted, one with
*MAT_24+GISSMO assigned to the front side members and another with Crach-
FEM.
Figures 5.16 and 5.17 illustrate the BIW of Lynk & Co 01 pre and post deformation,
respectively. Both models were superimposed to check the deviation in deformation
behaviour by visual inspection. Similar deformations were observed in both the cases.
This indicates that the front side members of the vehicle structure, assigned with
different material models exhibited similar performance during the crash. De- tailed
deformation of BIW front end are observed in Figure 5.17.
39
5. Results and Discussion
A closer observation of the front side members of the vehicle structure are shown in
the below figures (Fig. 5.18, Fig. 5.19, Fig. 5.20, Fig. 5.21, Fig. 5.22, Fig. 5.23).The
left side members of the vehicle structure were engaged in the test and the bending
modes between the two tests were found to be similar and this is observed inthe Figure
5.19. This indicated the calibrated material card (*MAT_24+GISSMO) was in good
agreement with CrachFEM that can also be justified by the section
40
5. Results and Discussion
forces in the side member and can be viewed in the Figures 5.22 and 5.23.
Figure 5.22: Section force - brace Figure 5.23: Section force - TWB
41
6
Conclusion
The primary focus of this study was to develop a virtual test laboratory and method-
ology to characterise the material in different stress states. Calibration of any ma-
terial card required detailed data collection from extensive experimental tests which
are time consuming and expensive. A new approach was outlined in this study, where
an optimisation based method demonstrated the feasibility of capturing ma- terial
behaviour in different stress states using a single hardening curve obtained from a
uniaxial tension test. The optimisation technique using the response surface method
was found to be more efficient in extrapolating the hardening curve of the material
using a solid element specimen. The solid element models developed in thisstudy
showed promising results in capturing the failure behaviour similar to thatof
experimental test results. A quantitative result accuracy of 94.5% was achieved using
the solid test specimens in comparison with the physical tests. This was ob- served
for the particular cases of triaxiality and crack dimensions.
The method of extracting the maximum data from the solid element models was a
key step in calibrating a GISSMO material definition. The parameters like triaxi-
ality, failure strain, and critical strain values were carefully chosen from the results
obtained from the solid element models. Extensive optimisation played a vital role
in tuning the parameters like DMGEXP (Damage Exponent), FADEXP (Fading
stress exponent), LCSDG (Curve: Triaxiality v/s failure strain), ECRIT (Critical
plastic strain), and EPSF (Plastic failure strain) in GISSMO card. Mesh regularisa-
tion was done for the GISSMO damage model so that the failure strain was scaled
over a mesh range of 0.5 mm to 5 mm. A calibrated GISSMO card with *MAT_24
was found to be very effective in predicting the failure of material in different stress
states. Most complex stress states like biaxial test and bending test were accurately
captured with this method.
The calibrated material card (*MAT_24+GISSMO) was also tested on a full vehi- cle
model to test its accuracy in a complex load case. The results from the ODB crash
test revealed that the deformation behavior and section forces in the frontside
member of the vehicle was found to be quite similar when GISSMO and Crach-FEM
material models were used. Hence the calibrated material card was in good agreement
with CrachFEM. The methodology outlined here significantly led to the reduction in
the computational time and was found to be cost-effective. The pro- posed approach
clearly illustrated that a single hardening curve from a uniaxial tension test led to the
complete calibration of material model with good correlation with experimental data.
Hence the huge cost of material testing could be minimised.Another advantage of this
method was its consistency and repeatability, as the same methodology holds good
for other grades of steel and the virtual test laboratory de- veloped in this thesis
consisting of eight different specimens could be used for the comprehensive
development and calibration of material and damage models.
42
7
Future Scope
The provided methodology and results obtained were promising for the given ma-
terial and also constitute a fixed framework in the development and calibration of
necessary material models from the tensile test. Since the behavior of the material
was versatile and varies with their properties, in the future factors like strain rate and
lode parameters could be included in the above-mentioned methodology. Over the
last few years, polymers have played a vital in automotive applications, hence this
methodology could also be evaluated for the characterisation of polymers.
In the current method, the flow curve was calibrated using a uniaxial tension test
whereas the flow curve might be calibrated for complex stress states i.e bend test
or biaxial test using the test data. Also, different sampling methods during the op-
timisation process may be used to get optimum samples closer to the experimental
data. The reliability of the material model (developed using this current method)
should also be verified against weld failure analysis.
Manufacturing histories may be included in the damage model to consider the ef- fect
of the manufacturing process on the material properties. Other damage models like
DIEM, e-GISSMO could be coupled with different constitutive material models
(*MAT_33,*MAT_36E, *MAT_37 etc.) and verified ( in terms of both time and
accuracy). Finally, a potential toolbox could be designed to automate the current
methodology, where a calibrated material model is obtained within a few steps.
43
8
Appendix 1
8.1 Supporting results
46
Figure 8.3: Tension test(specimen with a 90°groove)
47
8.2 Post failure images
Figure 8.7: Tensile specimen with Figure 8.8: Tensile specimen with
waist hole
48
Figure 8.9: Tensile specimen with a groove
49
Figure 8.11: Unibiaxial test
50
51