106 ロバスト最適設計における計算効率の適応的累積関数近似による改善法の構想  [in Japanese] A Proposal of Computational Cost Improvement of Robust Optimal Design by Cumulative Function Approximation  [in Japanese]

This paper discusses the role of cumulative function approximation in computational cost improvement of robust design, which is typical of complicated design optimization problems. Robust optimal design is often expensive in computational cost for considering the distribution under the deviation of design variables by means of substitutional models. The proposed method uses the cumulative function approximation, which has been configured based on Voronoi diagram by the authors, to efficiently construct such models instead of whenever referring exact values. Its effectiveness is demonstrated in comparison between nominal optimal design and robust optimal design and between direct optimization and cumulative approximation based optimization through numerical examples.