Tese: Locally stress-constrained topology optimization with continuously varying loading direction and amplitude: Toward large-scale problems
Aluno(a) : Fernando Vasconcelos da SenhoraOrientador(a): Ivan Menezes
Área de Concentração: Mecânica Aplicada
Data: 13/05/2022
Link para tese/dissertação: https://doi.org/10.17771/PUCRio.acad.59650
Resumo: Inside the field of structural optimization, Topology Optimization (TO) is one of the most general techniques because it is able to generate incredibly complex structures with intricate details for a wide range of problems. However, most of the works in TO have focused on compliance-based design that does not consider structural integrity in the design process leading to structures that do not satisfy material failure requirements. In this work, we focus on the stress-based design approach. We introduce stress constraints in the optimization procedure to guarantee the structural integrity of the final optimized design. This approach leads to a more natural formulation that addresses a simple engineering question: what is the lightest structure able to withstand its loads? We developed a large-scale GPU-based parallel stress-constrained TO framework considering a continuous range of varying load directions to answer this question and close the gap between TO and practical application. The developed GPU-based C++/CUDA framework efficiently addresses the main challenges of large-scale TO, such as filtering, optimization algorithm, and the solution of the equilibrium equations, only requiring a moderately affordable GPU hardware. At the same time, we obtain more robust and suitable designs for engineering applications by considering a continuous variable range of load directions that more closely resemble real-life service loads using a worst-case analytical approach. We present several numerical results, including 3D problems with over 45 million constraints providing detailed optimal structures that demonstrate the capabilities of the techniques developed in this work. The large-scale GPU framework, combined with the analytical solutions for continuously varying load directions, has the potential to expand the applications of TO techniques leading to new and improved engineering designs.