探索空間と遺伝子長の動的縮小を考慮した遺伝的アルゴリズムによる電磁機器の形状最適化(<論文特集>電気機器・電子素子における最適設計・逆解析)

概要

論文の詳細を見る
For inverse shape optimization of electromagnetic devices, the gradient of the objective function usually cannot be calculated explicitly, therefore searching methods based on the Gentic Algorithms (GAs) together with the Finite Element Method (FEM) have generally proven very useful, robust and effective. However, for such optimizations, in order to obtain the optimal solution, it is necessary to perform numerous FEM analyses proportional to the number of generated populations, which consequently results in a lengthily computation time. In this paper, a new GA procedure is proposed in order to decrease the computation time and increase the accuracy of the results. This procedure is based on dynamically reduction of the length of genes and the width of the searching space as the optimization process evolves, or, as the number of generated populations increases. As a result, more accurate results were obtained with less number of generation, i. e. faster. The proposed procedure is applied for inverse shape optimization of several electromagnetic devices successfully.
日本シミュレーション学会の論文
1999-03-15