Publication

To meet increasing demands on reduced CO2 emissions, the automotive industry is currently very active in research to reduce vehicle weight by incorporating light-weight materials like laminated composites into structural components.

Historically, composite materials have mainly been used in the aerospace industry, whereby the virtual design and development tools for composite structures have been developed primarily to the specific needs and requirements in this industry. In general, the crashworthiness of aerospace structures is only assessed to a small extent. Especially, compared to that of automotive vehicles, where high rating in crash tests are a key selling argument. Consequently, no suitable virtual tools, capable of assessing the crashworthiness of composite automotive structures, have been developed. This lack of virtual tools is problematic since the development of modern automotive vehicles is almost exclusively driven by virtual developments. Without access to tools for crashworthiness assessment of composite materials, these materials will not be widespread in automotive vehicle structures. 

The fracture process of laminated composites is far more complicated than that of metals, the dominant class of materials used in automotive crash protection systems today. Thus, virtual tools developed for metals cannot be used to accurately predict the crashworthiness of composite materials. Instead, highly refined models that can resolve the complicated fracture process must be used. However, these models require excessive computational resources, making full-scale vehicle crash simulations infeasible. It is therefore crucial to develop computationally efficient virtual tools, which can accurately predict the crashworthiness performance of composite structures.

In this thesis, I will present a route towards full-scale vehicle crash simulations using an automatic-refinement method. The method is based on a computationally efficient shell model which, during the simulation, is automatically transformed to a highly refined model in areas where needed. This way, the increased computational cost, associated with the analysis of progressive damage in laminated composites, can be limited. 

The proposed method can successfully reproduce the same level of accuracy as a highly refined model, at lower computational cost. Consequently, this method can help to enable computationally efficient crash simulations of laminated structures, which in the long run will allow composite materials to have a widespread use in future automotive vehicles.