基于Kriging代理模型的两类全局优化算法比较  

Comparison for two global optimization algorithms based on Kriging surrogate model

作　　者：周昳鸣 张君茹 程耿东[1] 

机构地区：[1]大连理工大学工程力学系工业装备结构分析国家重点实验室,大连116023

出　　处：《计算力学学报》2015年第4期451-456,共6页Chinese Journal of Computational Mechanics

基　　金：973项目(2014CB049000);国家自然科学基金(11372062;91216201);辽宁省高等学校优秀人才支持计划(LJQ2013005);高等学校学科创新引智计划(B14013);博士后基金(2014M551070)资助项目

摘　　要：代理模型在结构优化领域中的应用逐渐增多。相对传统优化方法,代理模型方法在处理带有噪音或仿真模拟十分耗时的问题时有明显优势。加点准则是代理模型技术的一个关键,为了避免陷入局部最优解,加点准则需要同时考虑局部搜索(exploitation)和全局搜索(exploration)两部分并加以平衡。本文在Kriging代理模型基础上提出一种基于几何全局搜索的全局优化算法MSG(Multi-start Local Search with Geometrical Exploration),通过数值算例将其与基于不确定性全局搜索的有效全局优化算法EGO(Efficient Global Optimization)进行比较,研究了MSG算法参数的影响,并讨论了MSG与EGO各自的特点和适用范围。Surrogate based algorithms have been applied increasingly in the field of structural optimiza-tion.Compared with traditional optimization algorithms,surrogate based algorithms have advantages in dealing with the problems which have noise or are very time-consuming in simulation.To avoid falling into local optima,surrogate based algorithms use infill criteria to balance exploitation and exploration. This paper presents a new global optimization algorithm based on Multi-start local search with geomet-rical exploration （MSG）,and compares it with efficient global optimization （EGO）by using several numerical problems.This paper analyzes the effects for MSG parameters and discusses the behaviors and applications for MSG and EGO.

关 键 词：全局优化算法 KRIGING EGO 代理模型 几何全局搜索 

分 类 号：O224[理学—运筹学与控制论]