制約条件の感度解析と統合した多目的PSOの制約囲い込み法の改善

概要

論文の詳細を見る
In this study, a calculation efficiency for the authors' previously proposed reliability-based multiobjective optimization (RBMO) algorithm is discussed. The RBMO algorithm integrates the hybrid multiobjective particle swarm optimization (MOPSO) for handling the constraint satisfaction technique using gradient information of constraints with a concept of the single-loop-single-vector method (SLSV). Though the method requires iterative evaluations for moving the design candidate with constraint violation to the feasible region, the algorithms totally have a good convergence performance for the most design problems. For example, Pareto solutions that lie on the feasible boundary are found in the early stage. However, some design candidates require the large numbers of iterations until the candidate moves to feasible region under the case where the violated constraint conditions are shifted frequently during the constraint satisfaction process. This study proposes the improvement method by eliminating such deteriorating iterations based on an idea from the modified SLSV method or the conjugate mean value method. The key idea of the proposed method is to select a kind of average direction of the previous searching directions as the current moving direction, when the deteriorating iteration occurs. The efficiency of the proposed method is demonstrated through several benchmark design problems by comparing the convergence property of the improved method with that of the original hybrid MOPSO with SLSV method.

制約条件の感度解析と統合した多目的PSOの制約囲い込み法の改善

スポンサーリンク

概要

著者

関連論文

スポンサーリンク