モジュール共通化における解空間構造の把握に基づく対話型遺伝的アルゴリズムによる解探索の提案

抄録

Product line design based on module commonization is a strategy to expand the number of product types while maintaining mass production effectiveness. Although there have been several studies on product line design, it is a multi-objective optimization problem, including evaluation of the commonization ratio and performance, which makes it difficult to find an optimal solution, and the number of optimal solutions tends to be large. Therefore, it is difficult to derive a single optimal design solution. In addition, as the design process progresses, uncertainties are introduced, and it is not appropriate to seek detailed design solutions at the stage of broadly defining a product line policy. In this study, we attempted to support the formulation of product family design strategies by automatically extracting and presenting trends in the structural characteristics of optimal solutions. In addition, we improved the solution search efficiency by limiting the solution search to a specific region while checking the trend of the solution space.