1111 最適設計問題の多階層表現による定式化と最適化戦略の構築

概要

論文の詳細を見る
Recently, the design problems for machine products such as machine tools tend to have a large number of evaluative characteristics, and are almost always multi-objective. In such cases, as the optimization problems are composed of a great variety of interdependent characteristics, optimization processes often fall into unsatisfactory local optima. In order to overcome this problem, this paper proposes a gradual optimization method based first on decomposition and the hierarchical ordering of characteristics, and then focusing on the interrelationships among the ordered design characteristics. This method begins by carefully examining the characteristics and decomposing each one into simpler sub-characteristics. Next, based on an analysis of the interrelationships among all characteristics, all characteristics are ordered hierarchically and grouped into a number of basic optimal units (BOU). Then, from the lowest level of the hierarchy, each BOU is optimized and Pareto optimum solutions are formed. BOUs at successively higher levels of the hierarchy are optimized using the Pareto optimum solution from the level immediately below. By sequentially optimizing the BOUs from the lowest level, Pareto optimum solutions of the objective functions are eventually formed. Finally, this method is verified by applying it to the design problem for a simple machine product.
2002-11-27