Copyright Nafems 2025 Licensed solely to Georgiy Makedonov (gmakedonov@vias3d.

com) for single use only

© NAFEMS 2021 REPRODUCTION AND REDISTRIBUTION PROHIBITED www.nafems.org

New Concept of Geometry-Based Finite Element Model

Generation for Crash Simulation within the Graph-
and Heuristic-Based Topology Optimisation

T. Pohl, MEng 1,2

M. Sc. F. Beyer 2

Prof. Dr.-Ing. A. Schumacher 2

1
Stellantis, Opel Automobile GmbH, Rüsselsheim, Germany
2
Bergische Universität Wuppertal, Germany

Abstract

Due to electrification and autonomous systems, the concept of vehicles and their
structure will potentially change considerably. New structural layouts may be
needed. To design safe but also light vehicles, it is necessary to include
optimisation procedure for crash. A new method has been developed to allow a
topology optimisation based on design rules – heuristics. The credibility of the
simulation results also depends on realistic simulation models. Based on the
results of validated crash simulation models the needs for geometric
representations have been defined, and a model creation tool is being developed
to allow the inclusion of credible simulation models in the topology
optimisation.

1. Introduction

Over the past five years, passenger vehicles have changed more than in several
decades before, and the change is still ongoing. The main driver for this change
is the increasing need to reduce emissions, especially CO2, and the electrification
is a key element to achieve this goal. Another trend is the increased use of
automated driving and assistance systems. But above all, the changes of the
propulsion system and energy storage result in a change of the vehicle layout.
Batteries or hydrogen tanks must be protected, the mass distribution is different,

Presented at the NAFEMS World Congress 2021 1 25th-29th October 2021

Copyright Nafems 2025 Licensed solely to Georgiy Makedonov ([email protected]) for single use only
Copyright Nafems 2025 Licensed solely to Georgiy Makedonov ([email protected]) for single use only

© NAFEMS 2021 REPRODUCTION AND REDISTRIBUTION PROHIBITED www.nafems.org

and the propulsion components can potentially be arranged in a different way –

electric motors at the rear axle being one good example. In case of fully
automated and even driverless vehicles, the impact will be even bigger.

This means that traditional layouts of the structure may well be not the optimum
anymore. Since there is a continuous pressure on mass and cost – driven
especially by the battery – there is a need to optimise the structure.

The biggest impact to the design to reduce mass or cost can be achieved in the
early phases of the development, as indicated in figure 1. Having chosen a certain
layout will have fixed most of the mass and product cost. Hence it is beneficial
to start the utilisation of numerical optimisation procedures as early as possible.

Figure 1: Cost commitment over development time [1]

CAE based results will then guide the concept selection and CAD design. The
focus of this work will be on the development of the vehicle structure, both body
and chassis, since they are the foundation for a safe and comfortable vehicle. It
is obvious that a multidisciplinary approach is necessary, as both stiffness on the
one hand and energy absorption through deformation on the other are required.

There are different methods for structural optimisation, starting with the
optimisation of parameters such as the sheet metal thickness. Secondly, there is
the shape optimisation. This may modify local and potentially small geometric
features such as a radius or a hole, but also large changes such as the cross-
section of a profile or the position of a structural member. Third, it is possible to
optimise the material, such as fibre direction. Finally, there is the optimisation
of the topology of the structure. This will determine the layout of the structure
in the design space. The latter is the method that will have the biggest impact in
the development of a structural concept [2].

Presented at the NAFEMS World Congress 2021 2 25th-29th October 2021

Copyright Nafems 2025 Licensed solely to Georgiy Makedonov ([email protected]) for single use only
Copyright Nafems 2025 Licensed solely to Georgiy Makedonov ([email protected]) for single use only

© NAFEMS 2021 REPRODUCTION AND REDISTRIBUTION PROHIBITED www.nafems.org

Shape Sizing Material Topology

Optimisation Optimisation Optimisation Optimisation

Figure 2: Types of optimisation [2]

Topology optimisation is a well-established method for linear loadcases. The

design space is filled with material, and this initial structure is optimised for
minimum compliance, internal elastic energy, with a given restriction on e.g. the
stiffness at a given point and volume fraction of the design space. The density of
the elements filling the design space are the design variables, and so the matrial
is distributed in this space in the optimum way.

Figure 3: Linear topology optimisation: design space and remaining loadpaths [3]

But in the case of the crash loadcases this procedure cannot be applied in the
same way. Non-linearities need to be accounted for, such as non-linear material
behaviour, components coming into contact in the course of the crash, buckling
of structures, or the fracture of material. Hence the use of gradient-based
mathematical optimisation procedures is problematic.

There are several new methods under development to address these issues. One
of them is the “Equivalent Static Loads ESL” method developed by G.J. Park
[4]. It is a stepwise linearization of the initially non-linear crash problem. For
each step, the displacement is calculated using the non-linear crash solver.
Subsequently, a set of linear loads is determined that gives the same
displacement. For these linear loadcases, a gradient-based optimisation is
performed. This method has been extended to a difference based ESL [5]. One
alternative method is the method of Hybrid Cellular Automata, as introduced for
crashworthiness optimisation by Patel [6] and continuously developed, starts

Presented at the NAFEMS World Congress 2021 3 25th-29th October 2021

Copyright Nafems 2025 Licensed solely to Georgiy Makedonov ([email protected]) for single use only
Copyright Nafems 2025 Licensed solely to Georgiy Makedonov ([email protected]) for single use only

© NAFEMS 2021 REPRODUCTION AND REDISTRIBUTION PROHIBITED www.nafems.org

with a design space filled by cells – similar to linear topology optimisation – and
derives a final topology by reducing the density of the cells if the internal energy
density is low. This resulting set of remaining cells needs to be converted into
actual 3D geometry for the continuation of the design process.

2. Graph- and Heuristic-based Topology Optimisation

The graph and heuristic-based topology optimisation GHT uses design rules –
heuristics – to generate new proposals for the structural layout. The change is
applied to a graph description of the structure in the first place, and a
corresponding finite element model is derived from this. This approach as
initially been developed by Schumacher and Ortmann [7] for 2D sections that
can be extruded into the 3D space.

The next step was to evaluate in how far this approach can be extended into the
3D design space, the work being performed at the Bergische Universität
Wuppertal, namely Florian Beyer, Dominik Schneider, Miriam Kick and Axel
Schumacher, and supported by the Forschungsvereinigung Automobiltechnik
FAT in the Project group for CAE in the Concept Definition, with participation
of Ford, Porsche, Volkswagen, Opel, ZF, Benteler and Altair [8, 9].

2.1 Heuristics

As mentioned above, heuristics are design rules which are based on experience
and expert knowledge. Already for the development of the 2D, vehicle safety
engineers have been involved to give input to come up with a set of typical
strategies to counteract undesirable structural behaviour in the crash event. There
are two sets – one the one hand competing heuristics, and on the other
complementary non-competing heuristics. The latter are the scaling of mass, to
make up for additional mass due to additional structural members, and the
straightening of the structure in case of unnecessary angles between members.
Based on an initial simulation, the former are used to come up with new
topologies which are then tested and compared in their effectivity. One heuristic
is the support of fast deforming edges. Buckling is a typical behaviour of
structures under axial load, but it is often not desired since the energy absorption
is low. Hence it is preferred to keep the member in place to allow crushing. The
automatic modification of the structure requires the automatic detection of the
structural response. For the fast-deforming edges, the member is spilt into
sections, and the centroid of these section is taken as reference. If there is a high
velocity normal to the axis describing the member it is reasonable to assume
buckling. Another rule is to add members between parts of the structure that
reduce or increase their distance, to add rods in compression or tension. Here,

Presented at the NAFEMS World Congress 2021 4 25th-29th October 2021

Copyright Nafems 2025 Licensed solely to Georgiy Makedonov ([email protected]) for single use only
Copyright Nafems 2025 Licensed solely to Georgiy Makedonov ([email protected]) for single use only

© NAFEMS 2021 REPRODUCTION AND REDISTRIBUTION PROHIBITED www.nafems.org

also the centroids of the sections are used, and checked against the position of
these of all other members of the structure.

Figure 4: Heuristics used in the optimisation procedure [8, 9], Support fast deforming
edges (1), Use deformation space tension and compression (2), Split long edges (3),
Balance energy density (4)

A third rule is to split long structures as they are prone to buckling at some stage.
A fourth rule is to balance the energy density of the structure. A member with
high energy density, i.e. carrying a high load, is connected to a member carrying
little load. In addition, if members are found that have a very low energy density,
and there is no way to make good use of them by linking them to loadbearing
structure, they can just as well be removed at all.

2.2 Graph Representation

It is not straightforward to apply these design rules directly to a finite element

model of the structure, and especially to add members in a sensible way. Hence,
the structure is first represented by a graph description, and this graph is then
translated into a finite element model. Changes in the topology are imposed on
the graph, and thus a good new finite element representation can be generated.
The graph consists of “vertices”, describing the end points of the structural
members, and “edges”, describing how these vertices are connected and also
carrying information about cross section, material and thickness.

Figure 5: Example of Graph Syntax with two Vertices and one Edge

The structures that can be described this way are frame type of structures, as they
are typical for rails and frames of vehicles. A graph can also be checked for
certain features, such as the angle between edges or the proximity of two
vertices. If these are below certain thresholds, the design can be seen as invalid.
Manufacturing constraints can be taken into account this way. Symmetry can

Presented at the NAFEMS World Congress 2021 5 25th-29th October 2021

Copyright Nafems 2025 Licensed solely to Georgiy Makedonov ([email protected]) for single use only
Copyright Nafems 2025 Licensed solely to Georgiy Makedonov ([email protected]) for single use only

© NAFEMS 2021 REPRODUCTION AND REDISTRIBUTION PROHIBITED www.nafems.org

also be enforced, and restrictions on the movement of the vertices can be

imposed.

2.3 Optimisation Procedure

The optimisation process consists of two loops, an inner loop to optimise the size
and shape of the structure, and an outer that modifies the topology. For a given
starting design, the inner loop is performed first. If this reached its termination
criterion, the outer loop is started. The response of the structure is analysed with
respect to the above criteria. All potentially applicable heuristics are being
activated in parallel, in order to assess competing designs. After the topological
modification, the inner loop size and shape optimisation is performed again. The
best topology is then used for the next iteration. An internal meshing tool
generates the finite element model based on the graph.

The structural members are meshed with shell elements and the joints are
represented by rigid spiders from the end of the member to a common node at
the location of the vertex. The crash simulation itself has been performed using
the LS-Dyna code.

Figure 6: Frame structure and joint created by internal model builder [8]

2.4 Example Applications

This optimisation procedure has been applied to a couple of loadcases.

A cubic frame with members at the edges and fixed at one side is impacted by a
pole from below. The optimisation target is to minimise the intrusion. The area
inside the initial frame is the design space where additional structures can be
created.

Figure 7: Deformation of initial structure, at t0 = 0ms, T = 15ms and t = 30 ms [8]

Presented at the NAFEMS World Congress 2021 6 25th-29th October 2021

Copyright Nafems 2025 Licensed solely to Georgiy Makedonov ([email protected]) for single use only
Copyright Nafems 2025 Licensed solely to Georgiy Makedonov ([email protected]) for single use only

© NAFEMS 2021 REPRODUCTION AND REDISTRIBUTION PROHIBITED www.nafems.org

Initially, there is a large parallelogram type deformation, with an intrusion of

174mm. Already the shape optimisation is able to reduce this to 30.5mm. The
subsequent application of heuristics resulted in a further reduction, such that the
final value was 19.7mm. The subsequent shape optimisation resulted in an
additional 5.1mm reduction of the deformation, such that the optimised structure
showed 15.6mm of intrusion at the same mass as the original design.

Figure 8: Optimised structure, deformation reduced, at t0 = 0 ms, t = 3 ms

and t = 10 ms [8]

The same loadcase has been optimised but with a different target – to minimise
the contact force between the structure and the cylinder, with a restriction of the
maximum intrusion. The force of the initial design was 10kN, and after seven
heuristic and the initial and final shape optimisation this is reduced to 5.4kN.

Figure 9: Structure optimised for minimum contact force at t0 = 0 ms, t = 10 ms and

t = 30 ms, and contact force vs. time for initial and final design [8]

In a second loadcase, the same initial structure was impacted by a rigid wall of
65kg and a velocity of 32 km/h.

Figure 10: Impact of rigid wall, deformation at t0 = 0 ms, t = 20 ms and t = 50 ms [8]

The first optimisation target was again to minimise the intrusion. The initial
shape optimisation reduced the intrusion from 210mm to 116.5mm. The
application of heuristics resulted in a further reduction to 45.7mm. The effect of
the heuristics was even larger than in the first example. The final shape

Presented at the NAFEMS World Congress 2021 7 25th-29th October 2021

Copyright Nafems 2025 Licensed solely to Georgiy Makedonov ([email protected]) for single use only
Copyright Nafems 2025 Licensed solely to Georgiy Makedonov ([email protected]) for single use only

© NAFEMS 2021 REPRODUCTION AND REDISTRIBUTION PROHIBITED www.nafems.org

optimisation resulted in a further reduction by only 1.6mm. Figure 11 shows an

intermediate step to indicate how the heuristics are applied in the process.

Figure 11: Impact of rigid wall, step from iteration 1 (undeformed and deformed state
left and centre) to iteration 2, heuristic “support fast deforming edges” to remediate
the buckling of the two middle beams [8]

As before, the reduction of contact force was a second optimisation objective.

The maximum allowed intrusion was again 150mm. After the initial shape
optimisation, the contact force had been reduced from 89.8kN to 76.4kN. The
application of heuristics brought this value down to 53kN, with a final shape
optimisation giving a contact force of 30.8kN. The corresponding force-
displacement curve shows a nearly constant force all the way, giving a very high
structural efficiency. Most of the structural members are being deformed during
the impact, also indicating this high efficiency.

Figure 12: Structure optimised for minimum contact force at t0 = 0 ms, t = 10 ms and
t = 40 ms, and contact force vs. time for initial and final design [8]

In all these cases, the Graph- und Heuristic-based Topology Optimisation

procedure successfully came up with new structural solutions, achieving to
improve the desired objective considerably while respecting the restrictions. The
work on this method is continuing in order to turn this feasibility study into a
process that can be applied in industrial context.

3. Simulation Model Idealisation

3.1 Simulation Credibility

Presented at the NAFEMS World Congress 2021 8 25th-29th October 2021

Copyright Nafems 2025 Licensed solely to Georgiy Makedonov ([email protected]) for single use only
Copyright Nafems 2025 Licensed solely to Georgiy Makedonov ([email protected]) for single use only

© NAFEMS 2021 REPRODUCTION AND REDISTRIBUTION PROHIBITED www.nafems.org

For such simulation and optimisation results to be accepted to guide the

development of the system and its components, the simulation must be credible.
Simulation model of early concept development phases tend to be regarded as
too much simplified by crash development engineers and CAD designing
engineers. But very detailed models take too long to create and are not flexible
enough to be modified quickly or even to be used in optimisation procedures as
shown above. Hence, it is necessary to find a good compromise between required
geometric detail and flexibility of model generation.

In order to find such a compromise, it needs to be laid out where potential errors
in the simulation may come from, and ideally to quantify them. It was shown
that on the way from reality to simulation result, there are several steps which
all might introduce errors [10, 11]. First, the real situation – the problem entity
– needs to be transformed into a physics model. Only the physical phenomena
of interest are taken into account, e.g. structural mechanics as opposed to thermal
analysis, and time dependency. This process is called abstraction. Next, this
physics model is transformed into a mathematical model – the idealisation. The
decision to use beam theory, shell mesh model or a solids is taken in this step,
as well as the definition of loads and boundary conditions. Finally, this is turned
into a computational model.

Figure 13: Simple (left) and more detailed (right) over view of the abstraction of the
modelling and simulation process [10]

The choice of the representation is an iterative process, and the appropriate

representation may vary along the development process as different types of
decisions may be taken based on the simulation results.

In the context of the structural optimisation, it needs to be assessed whether the

finite element model is suitable for this purpose, and above all whether the
resulting topology is robust with respect to changes in the idealisation.

3.2 Analysis of a Front Crash Model

In order to find out the impact of geometric details on a crash simulation, a

detailed model has been simplified in a number of steps [12]. The starting model

Presented at the NAFEMS World Congress 2021 9 25th-29th October 2021

Copyright Nafems 2025 Licensed solely to Georgiy Makedonov ([email protected]) for single use only
Copyright Nafems 2025 Licensed solely to Georgiy Makedonov ([email protected]) for single use only

© NAFEMS 2021 REPRODUCTION AND REDISTRIBUTION PROHIBITED www.nafems.org

was a fully validated and verified full vehicle front crash model representing a
final tooling release status.

Figure 14: Crash of a vehicle model (complete Body-in-white present in the model but
not shown in the picture)

This has been modified in a first step by removing some sub-systems and
modifying some component material and thickness assignments and introducing
some geometric changes – but with all details that are typical for such a late
phase. This was done so the model does not reveal the actual crash behaviour of
the vehicle as it is on the market. But the mentioned changes do not impact the
fact that the model can be assumed to show a validated and verified behaviour.

Figure 15: Stepwise simplification of reference model

The reduction of geometric detail was performed in several steps. First, the
secondary load path of the front subframe extensions and the brace wheelhouse
have been modelled with sharp corners and without cut-outs and holes, and small
holes (for positioning during the welding of the sub-assembly) in the main load
path, the front rail parts, have been closed. In the last and most simplified model
the entire front rail had also been modelled with sharp corners and no holes.

Presented at the NAFEMS World Congress 2021 10 25th-29th October 2021

Copyright Nafems 2025 Licensed solely to Georgiy Makedonov ([email protected]) for single use only
Copyright Nafems 2025 Licensed solely to Georgiy Makedonov ([email protected]) for single use only

© NAFEMS 2021 REPRODUCTION AND REDISTRIBUTION PROHIBITED www.nafems.org

Figure 16: Model “Detail -2” (dark cyan) vs. reference run (blue) at onset of buckling
and final deformation, showing very similar behaviour

The results show that the lightly simplified models still have a similar
deformation pattern. But once the profile cross-section does not have any radius
any more along the edges and along the weld flanges, the deformation is very
much different.

Figure 17: Left: Model “Detail -2” (dark cyan) vs “Detail -4” only outer radii but
sharp corners at flange)(green) shows reduced buckling at centre subframe support;
Right: model “Detail -6”(all sharp corners)(pink) shows reduced buckling at front of
the rail

Two modified base models have been created, one having a larger depression
for the front wheel envelope, and the other with a reinforcement removed at the
connection between the front rail and the dash. Both have been evaluated with
the same steps of geometric simplifications. In the case of the larger depression
for the wheel envelope, it shows that this depression dominates the deformation
behaviour. There is no such a big difference in results between the levels of
simplification. The removed reinforcement does have a local impact on the
results, but in general, the same pattern as for the in initial version can be
observed, the main radii along the load bearing rails play an important role.

Hence it can be deduced that a model without the main radii of the cross-section
appears to be an over-simplification. The physics model is not adequately
represented by the mathematical model. Further investigations and
quantification of the results will be required to be able to come to a conclusive
result.

4. Graph-based SFE-Concept Model

Since the success of the GHT3D optimisation does not only depend on good
heuristics and optimisation procedures but also on a credible simulation model,

Presented at the NAFEMS World Congress 2021 11 25th-29th October 2021

Copyright Nafems 2025 Licensed solely to Georgiy Makedonov ([email protected]) for single use only
Copyright Nafems 2025 Licensed solely to Georgiy Makedonov ([email protected]) for single use only

© NAFEMS 2021 REPRODUCTION AND REDISTRIBUTION PROHIBITED www.nafems.org

the above results indicate that a more realistic representation of the graph may
be required.

In the software SFE-Concept the geometry is based on Influence Points, Lines

and Sections to create Beams and Joints as shown by Hans Zimmer et al., fig.1
[13].

Figure 18: SFE-Concept entities [13]

This conceptually resembles the geometry description of the graph

representation as shown in figure 9. SFE-Concept model can be saved in two
formats, the binary .SFECmod and the ascii .SFECmac format. This ascii format
can also be generated outside of the SFE software.

Since SFE-Concept is capable of creating considerably more realistic geometric

representations and hence simulation models, including flanges with connecting
welds or radii at the corners of the section, this was seen as a potential tool to
create simulation models of the desired level of realism.

Figure 19: Relation between Graph Description and SFE-Concept entities

For each Beam, the Influence Points (IPs), Lines, Base Sections, and the Beam
description itself must be created. The IP and the Line can directly created from
the graph description, it is merely a re-formatting. The SFECmac description of

Presented at the NAFEMS World Congress 2021 12 25th-29th October 2021

Copyright Nafems 2025 Licensed solely to Georgiy Makedonov ([email protected]) for single use only
Copyright Nafems 2025 Licensed solely to Georgiy Makedonov ([email protected]) for single use only

© NAFEMS 2021 REPRODUCTION AND REDISTRIBUTION PROHIBITED www.nafems.org

the Base Sections and Beams in considerably more complicated such that a
creation ‘on the fly’ is not possible. Instead, it was deemed to be more efficient
to create a library of potential Base Sections and corresponding Beams, and to
position these according to the position and orientation of the IPs and Lines. The
Base section is named in the graph. The library needs to contain variables which
are then replaced by the actual values given in the graph. Hence the “Graph-to-
Model-Generator” reads the graph in the GHT3D format as input and generates
a complete SFE-Concept model on its basis.

Figure 20: Libraries for different cross-sections and respective beams

The most complex task though is the creation of realistic structural joints. The
type of the joint depends on the number of beams attached, but also on the
direction they come from. This will be the next step in the process to create a
tool that can offer an alternative and more realistic model in the GHT3D
optimisation procedure.

Figure 21: Different joint types will require respective library

5. Conclusions

• A new method has been developed to allow the topology optimisation for
crash loaded structures. The topology changes are based on design rules –
heuristics – and a graph description of the structure.

Presented at the NAFEMS World Congress 2021 13 25th-29th October 2021

Copyright Nafems 2025 Licensed solely to Georgiy Makedonov ([email protected]) for single use only
Copyright Nafems 2025 Licensed solely to Georgiy Makedonov ([email protected]) for single use only

© NAFEMS 2021 REPRODUCTION AND REDISTRIBUTION PROHIBITED www.nafems.org

• The credibility of a simulation result depends on the correct choice of the

correct idealisation.

• The analysis of a front crash model shows that certain geometric features,
namely the radii at the edges, need to be included in the model.

• A model creation tool can be used to generate SFE-Concept models in the

required level of geometric simplification.

• Further work needs to be done to allow the creation of structural joints in

the same way.

6. References

[1] Savair M, Newnes L, Mileham A R, Goh Y M, (2008), “Estimating cost at

the conceptual design stage to optimize design in terms of performance and
cost”; Collaborative Product and Service Life Cycle Management for a
Sustainable World - Proceedings of the 15th ISPE International Conference
on Concurrent Engineering, pp. 123 – 130

[2] Schumacher A, (2020) “Optimierung mechanischer Strukturen - Grundlagen

und industrielle Anwendungen. 3rd edition Berlin Heidelberg New York:
Springer, ISBN: 978-3-662-60328-4

[3] Siegmann J, (2017). “Lightweight Body Design by Consequent Use of

Optimization Techniques”. 9th European Altair Technology Conference,
Frankenthal, Germany

[4] Choi W S, Park G J, (2002). “Structural Optimization using Equivalent Static

Loads at all the Time Intervals”. Comput. Methods. Appl. Math., 191.19, S.
2077–2094

[5] Triller J, Immel R, Timmer A, Harzheim L, (2021), ”The difference-based

equivalent static load method: an improvement of the ESL method’s
nonlinear approximation quality”, Structural and Multidisciplinary
Optimization, 63:2705–2720

[6] Patel N M, Renaid J E, Tovar A, (2009) “Crashworthiness Design Using

Topology Optimization”. ASME. J. of Mech Design., 132.6

[7] Ortmann C, Schumacher A, (2014) “Mathematical Description and

Algorithmization of Expert Knowledge for the Support of the Topology
Optimization of Crashworthiness Structures”. PAMM Proc. Appl. Math.
Mech. 14, S. 1023–1026

Presented at the NAFEMS World Congress 2021 14 25th-29th October 2021

Copyright Nafems 2025 Licensed solely to Georgiy Makedonov ([email protected]) for single use only
Copyright Nafems 2025 Licensed solely to Georgiy Makedonov ([email protected]) for single use only

© NAFEMS 2021 REPRODUCTION AND REDISTRIBUTION PROHIBITED www.nafems.org

[8] Beyer F, Schumacher A, (2020) “Untersuchung zu den Einsatzmöglichkeiten

der Graphen- und Heuristikbasierten Topologieoptimierung zur
Entwicklung von 3D-Rahmenstrukturen in Crashlastfällen“, FAT-
Schriftenreihe 329

[9] Beyer F, Schneider D, Schumacher A, (2021), “Finding three-dimensional

layouts for crashworthiness load cases using the graph and heuristic based
topology optimization“, Structural and Multidisciplinary Optimization,
63:59.73

[10] Smith J W, (2020), “Engineering Simulation Quality Management

Guidelines”, NAFEMS; R0129

[11] Frerichs D, Schulz S, Hahn K M, Lage Places S D, (2021), “Model Based

Quality Management for Virtual Prototypes considering Uncertainties”,
30th Aachen Colloquium Sustainable Mobility 2021

[12] Reihs J, (2019), “Konzeption einer Methodik für die Beurteilung von
Idealisierungen in der FEM-Crashsimulation”, Master Thesis at DiK,
Technische Universität Darmstadt

[13] Zimmer H, Prabhuwaingankar M and Duddeck F, (2009), “Topology- and

geometry-based structure optimization using implicit parametric models
and LS-OPT”, Proc 7th Europ LS-DYNA Conf, Salzburg, Austria

Presented at the NAFEMS World Congress 2021 15 25th-29th October 2021

Copyright Nafems 2025 Licensed solely to Georgiy Makedonov ([email protected]) for single use only