Crash topology optimization (TO) is important to derive vehicle concepts in early design phases. Here, gradient-based approaches are not directly applicable and heuristic or simplified methods are used, limiting the range of addressable problems. Hence, this thesis proposes a generic, non-gradient TO via evolutionary algorithms and low-dimensional level-set representation to optimize arbitrary criteria based on high-fidelity explicit crash simulations. The high costs of evolutionary search are reduced via machine learning approaches.
«
Crash topology optimization (TO) is important to derive vehicle concepts in early design phases. Here, gradient-based approaches are not directly applicable and heuristic or simplified methods are used, limiting the range of addressable problems. Hence, this thesis proposes a generic, non-gradient TO via evolutionary algorithms and low-dimensional level-set representation to optimize arbitrary criteria based on high-fidelity explicit crash simulations. The high costs of evolutionary search are r...
»