The Coupling of Robust Metamodel and Heuristic Methods in Reliability Based Design Optimization

The context of this paper is the establishment of help tools to aid in the decisions for robust design of mechanical structures in an uncertain environment. The problem considered is how to determine an optimized structure with regard to the security level criteria. This problem is called reliability optimization; that means the optimization is with regard to the constraint reliability. With a falling probability level, the engineer can decide if the structure has sufficient reliability. This paper presents a coupling between the methods of reliability approximation (FORM, SORM) and a set of modern techniques from the domain of the artificial intelligence which are characterized by their importance in the process of collection and analysis of the uncertainties. Among these techniques, the particular swarm, the genetic algorithms and the ant colony techniques should be mentioned. In this paper, a coupling of the reliability method and dynamics with metamodels (condensed models) to guide reliability optimization of such systems is proposed. Two numerical examples are presented to illustrate the performance of the proposed approach.