Kriging和距离因子辅助的全局优化方法  

A Global Optimization Method Assisted by Sequential Kriging and Distance Factor

作　　者：白政民[1] 李耀辉[1] BAI Zheng-min;LI Yao-hui(School of Mechanical and Electrical Engineering,Xuchang University,He’nan Xuchang461000,China)

机构地区：[1]许昌学院电气(机电)工程学院,河南许昌461000

出　　处：《机械设计与制造》2021年第4期15-18,共4页Machinery Design & Manufacture

基　　金：国家自然科学基金面上项目(51775472)—基于Kriging模型的仿真优化方法关键技术研究。

摘　　要：基于Kriging元模型的优化方法通常存在优化效率较低、收敛速度较慢、昂贵估值次数较多且难以有效平衡Kriging模型的局部与全局搜索行为等弊端。为此,提出一种序列Kriging和距离因子辅助的全局优化方法。其实现过程主要包含两个阶段:一是利用Kriging模型近似复杂昂贵的黑箱函数;二是利用Kriging模型与距离因子的乘积构造加点采样法则,并通过免于求导的DIRECT算法优化该法则,以获取新的昂贵估值点。六个数值函数测试和一个摆线泵仿真实例验证所提出方法的有效性。Aiming at the disadvantages of low efficiency,slow convergence,more expensive-evaluation number and inability to balance the local and global search behavior of Kriging model in the unconstrained optimization process,a global optimization assisted by sequential Kriging and distance factor is proposed.The implementation process mainly includes two stages:one is to approximate the expensive black box function by using Kriging model;the other is to construct the infill sampling method by the product of distance factor and Kriging model.The rule is optimized by the DIRECT algorithm free of iteration to obtain new expensive-evaluation points.Six numerical function tests and a cycloid pump simulation example verify the effectiveness of the proposed method.

关 键 词：Kriging元模型 距离因子 加点采样法则 全局优化 

分 类 号：TH16[机械工程—机械制造及自动化] TPH137[自动化与计算机技术]