4. LS-DYNA Anwenderforum, Bamberg 2005
© 2005 Copyright by DYNAmore GmbH
An Investigation on Spot Weld Modelling
for Crash Simulation with LS-DYNA
F. Seeger
a
, M. Feucht
a
, Th. Frank
a
, B. Keding
b
, A. Haufe
b
a
DaimlerChrysler AG, EP/SPB, HPC X411, 71059 Sindelfingen, Germany
{falko.seeger, markus.feucht, thomas.frank}@daimlerchrysler.com
b
DYNAmore GmbH, Industriestr. 2, 70565 Stuttgart, Germany
{bastian.keding, andre.haufe}@dynamore.de
Abstract:
Because of weight reduction necessities, high-strength steels are more and more widely used in body-
in-white (BiW) structures in recent years. Concerning the behaviour of such structures during a ca
r
crash the joints between high-strength materials are seen as critical points. Therefore the properties o
f
welded joints, especially the failure behaviour at high velocities have to be taken into consideration
during the development phase.
In this paper, we present an overview on the current activities at the DaimlerChrysler AG regarding
investigations on spot weld failure behaviour. A suitable model that is able to represent the failure
behaviour of spot welds in BiW structures and that is independent of mesh sensitivity has been
developed. An elasto-plastic material model based on von Mises plasticity (MAT_100) has been
further enhanced with a new failure criterion. The model has been implemented into LS-DYNA. The
material as well as the failure behaviour is verified and calibrated through precision experiments
conducted on specimen level and later validated on component level.
Keywords:
Spot weld modeling, constitutive model and failure, crashworthiness, connection modelling, finite
elements
Crash I - Spotweld / Bonding
B - I - 1
4. LS-DYNA Anwenderforum, Bamberg 2005
© 2005 Copyright by DYNAmore GmbH
1 Introduction
Since years various efforts have been made to identify suitable strategies to model spot weld or
connections in general in commercial FE-codes (see [1]-[5]). The rather simplistic modelling
techniques in crashworthiness using tied-contacts or rigid connections were improved by applying
deformable spot weld elements which take into account the stiffness of the connection itself. The latter
modelling technique uses a combination of tied contacts and structural elements such as beams or
one or more solid elements. This approach worked sufficiently well even though obvious
simplifications such as the negligence of anisotropy in the elements were made. One of the main
reasons for this satisfactory performance was the assumption that the spot weld itself does not fail in
case of crash impacts but rather the material which is connected by the spot weld. This assumption
was based on experimental testing and holds true for ductile steels. However, this picture changes
substantially with the use of high-strength steels. Now, rather the spot weld itself is failing than the
material surrounding the spot weld. As a consequence, the models using solely deformable structural
elements needed improvement in order to predict the correct failure behaviour of the connections in
complex car crash scenarios.
The present paper will investigate different modelling techniques used for connections in LS-DYNA [6].
The merits and limits of the widely used methods will be evaluated by using data from simple
experimental tests. A suitable model that is able to represent the failure behaviour of spot welds in
BiW structures and which is independent of mesh sensitivity, has been developed. An elasto-plastic
material model based on MAT_100 has been further enhanced with a new failure criterion. The model
has been implemented into LS-DYNA. The material as well as the failure behaviour is verified and
calibrated through precision experiments conducted on specimen level and later validated on
component level.
In addition, very recent developments in LS-DYNA will be discussed which enhance the modelling
capabilities in LS-DYNA for connections between high strength steel sheets dramatically. This
improved performance will be illustrated using experimental data from component testing.
2 Experimental basis
2.1 Motivation
As mentioned above, the goal of our investigation is to develop a suitable model to cover the failure
behaviour of spot weld connections in car crash simulations. As the result of continually rising
computational power the model size represented by the number of finite elements increased and at
the same time the element size in car models was getting smaller in recent years. Today a car model
consist of up to 1 200 000 finite elements with 5-15mm per element edge. Considering the over-all car
simulation the FE-model is relatively fine, but in comparison to the dimension of individual spot welds
the discretization is rather coarse. Typically the spot weld is to be modelled by one element, e.q. a
hexahedron or beam element, due to time step constraints. However, if only one element is available
to cover the material as well as the failure behaviour, specially adapted experiments which provide the
basis for investigation, verification, calibration and validation of new spot weld models are needed.
Such experiments should combine the realistic behaviour of the spot weld as observed in a real car
and the simplicity to model various failure mechanisms separately.
2.2 Specimen Level
On specimen level three different experimental setups were considered: First KSII-specimen, which
have been developed by [1], secondly simple lap shear specimen and finally a peel test specimen, see
Fig. 1, Fig. 2 and Fig. 3, respecively. The single parts are attached by only one spot weld.
The KSII-experiments, conducted with loading angles of 0°, 30°, 60° and 90°, are essential to derive
the relationship between normal and shear stress. With 90° only normal stresses occur in the spot
weld, whereas mainly shear stress is expected for the 0° load case. These experiments are used to
calibrate the material and failure parameters.
Furthermore, the peel test specimen is useful for checking the spot weld behaviour under bending
respectively peeling conditions. This experiment is also used for calibration.
Crash I - Spotweld / Bonding
B - I - 2
4. LS-DYNA Anwenderforum, Bamberg 2005
© 2005 Copyright by DYNAmore GmbH
A first validation of the numerical model can be performed on basis of the lap shear specimen. Due to
the load case a multi-axial stress state as a combination of shear, normal and bending stresses arises.
It is therefore a simple yet useful setup to validate calibrated models and evaluate the chosen
strategy.
Fig. 1: Different load cases for the KSII-test
Fig. 2: Peel test specimen
Fig. 3: Lap shear specimen
2.3 Test on component level: T-section
The test cases that were shown in the last paragraph are simple specimens with one spot weld only.
However, in real structures interactions by several spot welds and multi-axial stress states are
dominating the behaviour. To validate the present spot weld model and to study the interaction of spot
welds experiments on component level were performed. Here the so called T-section component (see
F
F
F
F
F
F
F
90° loading
60° loading
30° loadin
g
0° loading
5
3
F
Crash I - Spotweld / Bonding
B - I - 3
4. LS-DYNA Anwenderforum, Bamberg 2005
© 2005 Copyright by DYNAmore GmbH
Fig. 4) is used. The structure is supported via clamping on the back side. Furthermore, it is loaded by
the force F
T
in transversal and F
L
in longitudinal direction. The load was applied quasi-static and
dynamic with a velocity of 5m/s.
Fig. 4: T-section component
3 Numerical Investigation
The numerical model of the spot weld failure behaviour is investigated for application in full car crash
simulations. The FE-models of the specimen and component experiments have to fulfil the restrictions
concerning the modelling techniques of full car crash models, namely element size and type and
contact formulation options used. These restrictions lead to the conclusion that the numerical model
for the spotweld will consist of one element.
By fulfilling these restrictions the spot welds modelled, calibrated and validated by specimen and
component tests can be directly transferred to the car model.
3.1 Base model
The simple experiments on specimen level provide the basis for further investigations on numerical
modelling. Therefore the simulation models of the KSII-, peel and lap shear specimen have to cover
the experiment as good as possible while still fulfilling the modelling restrictions stated above. Fig. 5
shows the finite element model of the KSII-specimen. For further investigation we will focus on the
external forces
e
Y
F and
e
Z
F , representing the forces determined by the experiment, and on the internal
forces
i
Y
F and
i
Z
F , representing the forces in the spot weld. Both forces will be calculated in the
simulation, separately.
Fig. 5: External and internal forces for checking the mesh sensitivity
e
Y
F
e
Z
F
i
Z
F
i
Y
F
i
Z
f
i
Y
f
A
F
T
F
L
Crash I - Spotweld / Bonding
B - I - 4
4. LS-DYNA Anwenderforum, Bamberg 2005
© 2005 Copyright by DYNAmore GmbH
As described above the force-displacement characteristic of single spot welds is determined by
experiments. Here the force is measured locally close to the clamping of the KSII fixation and the
displacement is measured globally including the flexibility of the testing machine. Thus the force-
displacement characteristic covers the overall stiffness of the testing machine and of the specimen.
Therefore, the stiffness of the testing machine had to be included in the model. This could be achieved
by simple beam elements on top and bottom of the specimen, where the stress-strain behaviour of the
beams represents the stiffness of the testing machine.
3.2 Modelling technique
Among many different possibilities to model spot welds in crashworthiness applications, three
modelling techniques have been investigated: A single beam element, a single hexahedron element
and four hexahedron elements. Obviously, using a beam element is the easiest discretization effort to
model spot welds with finite elements. Here we found that the overall stiffness depends strongly on the
position where the beam nodes are connected to the shell surfaces. Secondly, the torsion stiffness of
the beam cannot be activated, since the corresponding degree of freedom is not available in the
attached shell formulation. Considering the spot weld discretization with four hexahedron elements,
the small element size would decrease the time step too much and the additional mass to
counterbalance this is not acceptable. The third modelling technique consists of one single
hexahedron. However, the available options to model failure were found not accurate enough in most
cases.
3.3 Material behaviour of MAT_100
Concerning the experiments on KSII-level we consider the 0° load case and the 90° load case to
calibrate the material behaviour. In the 90° load case bending deformations of the flanges dominate
the force-displacement curves, whereas the characteristic shear deformations of the spot weld
dominate the behaviour of the 0° load case. Therefore the 0° load case will be taken to determine the
material properties of the spot weld material model.
For modelling the material behaviour a bilinear elastic-plastic material law based on the classical
vonMises-criteria is used, see Fig. 6. As input data only Young’s modulus and Poisson’s ratio are
required for elasticity, and both yield stress and the hardening modulus for plastic range, respectively.
In most cases this relatively simple material law is able to reproduce the experimental data.
The calculation of the forces/moments or stresses in the spot weld are also performed in the MAT_100
subroutine and is handled in several steps. On the basis of the strain increment the element stresses
are calculated by the bilinear elastic-plastic material law. Since the MAT_SPOTWELD can only be
used with reduced integration, element stresses exist only for one Gaussian point. With the element
stresses the nodal forces at the bottom of the spot weld hexagon are calculated. Corrected by the
hourglass forces the stress components are computed back to gain the maximum edge-stresses of an
equivalent beam with circular cross section. These modified stresses not only contain shear and
normal components but also a bending component. The main reason for calculation of residual
stresses is the presence of a bending component which is important if we want to distinguish a
symmetric normal load from a asymmetric peeling load.
Crash I - Spotweld / Bonding
B - I - 5