多目标进化算法的改进在齿轮减速器中的应用  

Improved Multi-Objective Evolutionary Algorithm and Application in Gear Reducer

作　　者：高淑芝[1] 任学鹏 张义民[1] GAO Shuzhi;REN Xuepeng;ZHANG Yimin(Institute of Equipment Reliability,Shenyang Chemical University,Liaoning Shenyang 110142,China;College of Information Engineering,Shenyang Chemical University,Liaoning Shenyang 110142,China)

机构地区：[1]沈阳化工大学装备可靠性研究所,辽宁沈阳110142 [2]沈阳化工大学信息工程学院,辽宁沈阳110142

出　　处：《机械设计与制造》2025年第4期190-193,197,共5页Machinery Design & Manufacture

基　　金：NSFC-国家自然科学重点基金—辽宁联合基金(U1708254);辽宁省特聘教授(No.[2018]3533)项目。

摘　　要：分解的多目标算法是利用一组权重向量将一个多目标优化问题分解为一组标量子问题。针对当帕累托前沿是一个多峰和断裂等其他较复杂的情况下,均匀分布的权重向量往往收敛效果较差的问题,提出了一种种群分区管理的自适应方法用来保持种群的多样性与收敛性之间的平衡。首先,采用了一种均匀随机的权重向量生成方式进行初始化;其次,采用Tchebycheff分解方法进行子代的更新;再次,将提出的自适应方法对分解的多目标进化算法进行了改进;最后,通过在标准测试函数和齿轮减速器的优化仿真,证明了提出的算法的有效性。The multi-objective decomposition algorithm is to decompose a multi-objective optimization problem into a set of sca⁃lar quantum problems by using a set of weight vectors.In view of the problem that the uniformly distributed weight vectors often have poor convergence effect when the Pareto front is a multi-peak and fracture,an adaptive method of population partition man⁃agement is proposed to keep the balance between diversity and convergence of population.Firstly,a uniform and random weight vector generation method is adopted for initialization.Secondly,the Tchebycheff decomposition method is used to update the off⁃spring.Thirdly,the proposed adaptive method improves the decomposed multi-objective evolutionary algorithm.Finally,the ef⁃fectiveness of the proposed algorithm is proved by the optimization simulation of standard test function and gear reducer.

关 键 词：多目标优化 分解算法 自适应 进化算法应用 

分 类 号：TH16[机械工程—机械制造及自动化] TP18[自动化与计算机技术—控制理论与控制工程]