Meta-model assisted multi-objective optimization for non-steady 3D metal forming processes

This paper studies efficient techniques to find the optimal set of solutions (Pareto front) for multiobjective optimization problems in the context of time expensive evaluation of functions. These techniques make use of a meta-model based on the Meshless Finite Difference Method (MFDM) coupled with evolutionary Multi-Objective algorithm (here: NSGA-II) in order to minimize the time consuming evaluations and to achieve a faster convergence to the Pareto front. The different studied methods differ in the choice of master points, the evolution of the meta-model, and the updating of elitism. They are studied and compared on several analytical functions, with only 100 exact evaluations of the objective function. The obtained results show the efficiency of these techniques.