Method for merging scales in finite element analyses - Framework for automated global/local analyses

Abstract

The work of a CAE-engineer is a balance between accuracy and efficiency for the analyses. The outcome of the analyses must be accurate enough to be used for evaluation of the analysed component design and provide trustworthy recommendations when needed. Furthermore, it becomes increasingly important that the analyses are carried out fast to minimise lead time and cost. In today's work this balance is often met by choosing to rely on a coarser model together with engineering experience. The aim of this thesis is to develop a method and provide guidelines for multiscale analyses that can be used to analyse critical areas in a global model with higher accuracy. This is done by adopting the technique of submodelling and developing a script to perform the steps of the method in an automatic manner. Using the submodelling technique local models, also known as submodels, are created for critical areas found in the case studies. The new models are treated independently, only connected to the global analysis by applying new boundary conditions. Several analyses where the submodel responses are evaluated using different modelling configurations, such as submodel size and element size, are performed. The outcome showed that, by using the submodelling approach, a more accurate response prediction can be made. Also, due to that the volumetric size is drastically reduced, this is achieved in a reasonable amount of time. Furthermore, is it concluded that the submodel volume must be chosen in such way that the boundaries of the volume are located where the strain concentration has subsided and that the submodel element size preferably is chosen even smaller than the recommended size used at Volvo Cars today. Finally, correlation to a physically tested component indicated that the submodels to a greater extent predicts damage as it showed up in the test, both in size and intensity. The submodelling approach is automated in such way that the user of the script could interact with the submodelling process and make adjustment based on the specific geometry and load case. Keywords: Abaqus, Ansa, Meta, Submodelling, Multiscale analysis, Polymers