応答曲面法による多峰性問題の最適設計法の開発

概要

論文の詳細を見る
A great number of functional evaluations may be required until convergence in the process of optimization. Although approximation models constructed by the response surface methodology, which may greatly save the function evaluation, is usually used to get optimal designs, it is thought that the design accuracy may be dependent on the type of activation functions and the design region of interest. In this paper, techniques to search the region of interest containing the global optimal design with designs selected by random seeds, and techniques for finding more accurate approximation using Holographic Neural Network (HNN) is investigated. Furthermore, the mapping method of extrapolation is proposed to make the technique available to general application in structural optimization and formulas to estimate the necessary function evaluations for functions under certain condition is presented. Application examples show that HNN may be expected an potential activate and feasible surface functions in response surface methodology than the polynomials in function approximations. Finally, the real design example of a vehicle component crash worthiness is used to show the effectiveness of the proposed method.
1999-02-25