A Multiobjective Evolutionary Algorithm For Car Front End Design

The aim of this study is to find the optimal structural geometry of the front crash member of a car of minimal mass that optimally satisfies all operational conditions. The mechanical domains that have been considered are crash, acoustic (dynamic) and static. They are summed up by 9 objective functions, resulting in a 10-objective optimization problem. However, this problem is further turned into minimizing the mass while maximizing the internal energy (crash objective), subject to constraints on the 8 objectives that arise from the acoustic and static domains. The dimension of the objective space of this cons' trained problem is much lower than that of, the original 10-objective problem. This significantly reduces convergence time, while decreasing decision making efforts among solutions obtained though pareto-based multiobjective optimization.Nevertheless, since the computation of the structural responses is based on a very time-consuming FEM crash analysis, direct computation of the fitness within an evolutionary algorithm is impossible: The response of car front members is computed using an approximative evaluation that had been identified during the BE96-3046 European project (CE)(2):Computer Experiments for Concurrent Engineering. Thanks to this approximation, very good results are obtained in a reasonable time using a Pareto elitist evolutionary algorithm based on NSGA-II ideas, combined with an infeasibility objective approach for constraint handling.