In this dissertation a method for automatically calculating the damage of spot welds using notch stresses is presented. Therefore data from tests with constant or variable amplitudes are collected. There are over 800 tests with various steels and over 150 tests with aluminium in this collection. As a reference, current methods using structural stresses are evaluated and their advantages and drawbacks are discussed, in particular the force-based Rupp-method and the stress-based method used from ECS Femfat.

The presented method called BMWeld uses the ability of the finite element (FE) solver MSC.Nastran to condense a model into a superelement containing its stiffness and mass matrices. These superelements are automatically generated for all used diameter and sheet thickness combinations. After calculating the whole model with the included superelements the displacements and stresses in the notch can be exported and evaluated for the durability assessment. The whole process of superelement generation, implementation in the FE-model, evaluating the results and creation of a Wöhler-curve from a given dataset is implemented in Matlab.

To include the superelement in the FE-model a direct and an indirect connection as well as 8 and 16 connection points are evaluated. In a first step it is shown that the superelement reduction and connection is valid. The direct connection yields more precise results with a maximum deviation of displacement and stresses from a fine model of 5 percent while the indirect connection shows a deviation up to 24 percent. The stiffness comparisons show slightly more accurate results than the reference methods. Comparing the results of the structural durability analysis the scatter of the Wöhler-curves are reduced up to 15 percent. The usage of an effective stress is evaluated but the improvement is marginal while the calculation time is increased significantly. Finally, the robustness of the method is assessed by changing the surrounding mesh size and alignment of the spot weld element edge. The BMWeld-method with direct connection to the mesh shows only small deviations and is therefore very robust. The indrect connection with 16 points is on par with the Femfat method while 8 connection points with indirect connection is not recommended.

Therefore it is shown that the BMWeld method is better than the reference methods while maintaining the modelling requirements. First results from testing automotive assemblies emphasize these improvements. Currently the process is implemented in common FE pre- and post-processors.