基于非支配排序遗传算法的多学科鲁棒协同优化方法

李海燕; 马明旭; 井元伟

引用本文:	李海燕,马明旭,井元伟.基于非支配排序遗传算法的多学科鲁棒协同优化方法[J].控制理论与应用,2011,28(4):561~566.[点击复制]
	LI Hai-yan,MA Ming-xu,JING Yuan-wei.Multidisciplinary robust collaborative optimization based on non-dominated sorting genetic algorithm[J].Control Theory and Technology,2011,28(4):561~566.[点击复制]

基于非支配排序遗传算法的多学科鲁棒协同优化方法

Multidisciplinary robust collaborative optimization based on non-dominated sorting genetic algorithm

摘要点击 2405 全文点击 2490 投稿时间：2009-11-29 修订日期：2010-06-01

查看全文查看/发表评论下载PDF阅读器

DOI编号 10.7641/j.issn.1000-8152.2011.4.CCTA091522

2011,28(4):561-566

中文关键词鲁棒协同优化 NSGA--II算法多目标学科间一致性

英文关键词 robust collaborative optimization NSGA--II algorithm multiobjective interdisciplinary consistency

基金项目国家“863”计划资助项目(2009AA04Z104).

作者	单位	E-mail
李海燕^*	东北大学辽宁省复杂装备多学科设计优化技术重点实验室东北大学流程工业综合自动化教育部重点实验室东北大学信息科学与工程学院	lihaiyan@ise.neu.edu.cn
马明旭	东北大学辽宁省复杂装备多学科设计优化技术重点实验室东北大学流程工业综合自动化教育部重点实验室
井元伟	东北大学信息科学与工程学院

中文摘要

针对鲁棒协同优化(robust collaborative optimization, RCO)具有两级优化结构和多目标形式的特点, 提出基于非支配排序遗传算法(non-dominated sorting genetic algorithm, NSGA--II)的RCO求解方法. 在NSGA--II非支配排序中, 根据个体的不可行度和不可行度阈值来决定其可行性, 并给出随进化过程逐渐减小的不可行度阈值. 在该阈值的作用下, 在进化初期, 保留较多的目标函数和标准差较小的个体, 以便优化向全局极值点附近靠近; 在进化后期, 保留较多的学科间一致性好的个体, 以便增强学科间的一致性. 该方法在保证各子学科间一致性的前提下, 可有效避免RCO优化结果易收敛到局部极值点的问题. 利用典型算例对该方法进行了验证, 结果表明该方法的优化性能良好.

英文摘要

To the robust collaborative optimization(RCO) scheme with two-level multiobjective optimization structure, a solution strategy employing the non-dominated sorting genetic algorithm(NSGA--II) is proposed. In the process of non-dominated sorting, the feasibility of an individual is determined by its infeasibility degree and the threshold of infeasibility degree. The threshold of infeasibility degree is reduced gradually in the process of evolution. At the initial stage of genetic evolution, the individuals with smaller values of objective function and standard deviation are more likely to be preserved to ensure the optimization process for reaching the neighborhood of the global extremum. In the following stages of genetic evolution，the individuals with smaller value of infeasibility degree are more likely to be preserved to enhance the interdisciplinary compatibility. The convergence of the results of RCO to the local extremum is usually avoided while keeping the desired interdisciplinary consistency. The results of validation by using typical examples show that the proposed approach is efﬁcient.