ミニ・マックス型ロバスト最適設計の適切化変換を用いた高精度化法(機械要素,潤滑,工作,生産管理など) Enhancement of Mini-Max Type Robust Optimal Design using Function Regularization
This paper proposes a method for enhancing the quality of mini-max type robust optimal design by using the concept of function regularization. Since robust optimal design considers variations under various noises, the quality of a solution is affected by the intermediate model for considering variations of the objective function and constraints within a distribution region. The mini-max type robust optimal design has been proposed by the authors for considering the bounding points of the objective and constraints within the distribution region as a definition of robust optimality. The function regularization proposed in this paper enhances its accuracy by filtering the functions so as to improve fidelity of quadratic approximation, which is used for obtaining the bounding points. The filter is formulated as the form of Fourier series and is implemented for the mini-max type robust optimal design scheme. Then, numerical experiments, in which second-order Fourier series is used as the filter, are demonstrated with two numerical sample problems; a two-dimensional algebraic problem and a simple structural optimal design problem.