Multicriteria optimization and evaluation in complex products design

This study is related to the application of artificial intelligence approaches to the design of complex products. The purpose is to help designers to choose design parameters according to technical specifications and customer's needs. Because of the high number of parameters and possible combinations, two heuristic based approaches are proposed for optimization activity when the task is a multiple objective combinatorial one. A first algorithm based on Tabu search guided by an objective function, and retaining the non dominated solutions is effective when a preference order can be defined for criteria to optimize. When those criteria cannot be weighted, a hybrid search approach based on some heuristic mechanisms is proposed for optimization without resuming the multiple objective search to a single-objective one. The two specified algorithms derive the not dominated solution set: the Pareto front, from the possible solutions set according to the design parameters. Self-organizing maps are then used to analyze and evaluate the obtained front. Our approach is fully illustrated in the case of the design of a 2-degrees of freedom robot.