ABSTRACT of the Doctoral Thesis

Study of Effects of Weld Seams in Finite Element Crash Simulations

by Christian Grohs (2008)


The aim of the present thesis is to study damage effects due to weld seams in finite element analyses of crash boxes for rail vehicles. Due to the high costs of real crash tests, on the one hand, and increasingly more stringent safety standards, on the other hand, computational simulations of the crashing behavior of rail vehicles are growing in importance. In the event of a collision welded regions in crash zones can be critical with respect to local material failure, which may lead to a reduction in the structures' ability to dissipate kinetic energy.

For modeling samples and structures containing weld seams a fully coupled damage model for metallic materials is used that is provided by the finite element software ABAQUS/Explicit. This approach allows to account for stress redistribution after damage or failure has occurred. Moreover, this model does not require identifying regions of the model for possible failure, but is capable of predicting damage anywhere in the structure in dependence on the local stress and strain states. In addition, it is not limited to some range of stress triaxialities but can be used for hydrostatic tension as well as for shear failure.

The main part of this work deals with the determination of modeling parameters for both the basis material and for weld seams in specimens and structures made of S355 steel or Al6008-T4 aluminum alloy, respectively. Different specimen shapes are used to extract information on the damage-related material parameters in dependence on the stress triaxiality. Especially for shear failure only the combination of simulations and experiments leads to the determination of proper sample geometries.

After studying specimen geometries discussed in the literature suitable samples are identified with respect to mainly constant stress triaxialities in the cross section until the onset of damage. The triaxiality of each of these specimens is determined by finite element simulations. Based on these calculations the modeling parameters are fitted for the pertinent stress triaxialities to obtain reaction forces that closely approach those of the corresponding experiments. Separate sets of damage parameters are determined for shell and solid models for each material, respectively, which are capable of predicting the reaction forces until fracture well for all the different investigated geometries. In contrast to steel, where a ductile criterion is sufficient, the aluminum shell models need an additional set of damage parameters to get correct predictions in case of shear failure.

With the established modeling parameters and following an additional calibration step welded specimens are defined and simulated, the results being compared to experiments. The tested steel specimens are made of two metal plates welded to a cross plate on either side with simple hollow weld seams, whereas the aluminum samples consist of two plates connected by through-welded butt joints. Both types of sample allow investigating different loading angles of the weld seams.

For the steel specimens the cross plate leads to increased stiffness in the welded regions, preventing the weld seams from failing unless the direction of tension is not perpendicular to the plate. To counteract this, slots are cut along the weld seams to weaken the remaining material, leading to failure in the welded regions in the experiments. The length and position of the slots are determined in advance by finite element simulations. The aluminum specimens have to be specially shaped to ensure failure in the welded region. For all specimens good agreement between finite element simulations and tests is obtained in terms of both the geometries of the failed samples and the load-displacement curves.

Finally, whole crash boxes are simulated with modeling parameters based on the previous parts of the study. Again two different geometries, for steel and aluminum, respectively, are investigated, containing the two different kinds of welds. The simulations are compared to quasi-static experiments, the results being not fully satisfactory. After a re-calibration of the material parameters improved agreement with respect to the reaction forces and deformation patterns is obtained. For the aluminum crash boxes failure of the welded regions is also predicted correctly, whereas the steel crash boxes show no material failure at all in the experiments. The described modeling concept offers a way for studying the influence of weld seams in structures subjected to crush and crash loading.


revised 080717 (hjb)