Development of a Fast Approximate Solver for Robust Optimization Problems Including Various Uncertainties in Design Variables and Conditions

2018 Fiscal Year Final Research Report

• Project/Area Number: 15H03600
• Research Category: Grant-in-Aid for Scientific Research (B)
• Allocation Type: Single-year Grants
• Section: 一般
• Research Field: Computational science
• Research Institution: Tohoku University
• Principal Investigator: Shimoyama Koji 東北大学, 流体科学研究所, 准教授 (80447185)
• Co-Investigator(Kenkyū-buntansha): パラル プラムヂタサトリア 東北大学, 流体科学研究所, 学術研究員 (30792163)
• Research Collaborator: Kawai Soshi
• Project Period (FY): 2015-04-01 – 2019-03-31
• Keywords: 近似最適化 / 不確かさの定量的評価

Outline of Final Research Achievements

This study has developed numerical methods, which are fast and accurate to find robust optimization solutions whose objective functions are insensitive to various uncertainties in design variables and conditions. For this purpose, this study has proposed a method to reduce high-dimensional design space into a simplified space, a method to approximate various behaviors of the objective function as a non-stationary representation, and a method to adaptively update the approximated objective function with additional evaluations of the true objective function in proper cases. These methods have been applied to real-world design problems, which demonstrate superior effectiveness compared to conventional methods.

Free Research Field

最適設計

Academic Significance and Societal Importance of the Research Achievements

工学製品の形状や使用環境が変わったりしても，当初の製品性能をある程度維持できる，ものづくりのための手法を開発した．手元にある限られた評価データを有効活用して，ありとあらゆる不測の事態に対する性能の変化を正しく予測できるように，独自の数学モデルを考案した．これを航空機・宇宙機の設計問題に適用した結果，不測の事態における設計の安全性がしっかり担保されることを実証できた．以上より，本研究の成果は，安心・安全なものづくり社会の確立に貢献するものである．