CAJ | 학술논문

借助"记忆"已搜索过的可行空间并将其中状态与当前随产生状态的对比、分析，以及引入局部寻优算法，本文给出了一种多极值点连续变量函数全局优化的智能化新算法-改进的区域消除法。典型函数验证和应用计算实例表明，改进算法的收敛速度优于模拟退火算法。因此，本文为电磁场逆问题的求解计算提供他一种可供选择的全局优化算法。
차조"기억"이수색과적가행공간병장기중상태여당전수산생상태적대비、분석，이급인입국부심우산법，본문급출료일충다겁치점련속변량함수전국우화적지능화신산법-개진적구역소제법。전형함수험증화응용계산실례표명，개진산법적수렴속도우우모의퇴화산법。인차，본문위전자장역문제적구해계산제공타일충가공선택적전국우화산법。
An improved domainelimination method for global optimization ofmultimodal functions with continuous variables is presented in this paper.The method incorporate the ideas of both the memorization of the searched parameter spaces and compare the states compromising the memorized history with the newly generated ones as well as to introduce of a local phase.The advantage of the present method over the traditional simulated annealing algorithm is validated by using a mathematical test function and also by solving the geometry optimization of the end region of a typical power transformer.Thus an alternative for the global optimization of inverse problems in electromagnetics is introduced.