遺伝的アルゴリズムによる構造物の信頼性評価

概要

論文の詳細を見る
To evaluate reliability indices is an important step in the structural reliability analysis. It is, however, difficult to evaluate reliability indices by conventional optimization methods in case local optimal solutions exist or the limit state function is not differentiable. In order to solve these problems, this paper presents genetic algorithm-based optimization methods to evaluate structural reliability indices. Two methods, the penalty function method and the backtracking method, are applied to the GA to deal with constraints. A local GA is also added to the program to accelerate convergence to the optimal solution. Through the application of the developed program to typical structural reliability problems, it is revealed that the GA-based optimization methods can evaluate the reliability indices even when conventional methods fail, and the local GA is quite effective to reduce CPU time and to increase the stability of convergence for both the penalty function method and the backtracking method.
一般社団法人日本機械学会の論文
2001-05-25