多种改进选择算子的遗传算法性能比较与应用研究  

A Comparison and Application on Performances of Multiple Genetic Algorithms Adopting Improved Selection Operators

作　　者：蒋艳[1] 黎向锋[1] 左敦稳[1] 焦光明[1] 薛善良[1] 

机构地区：[1]南京航空航天大学机电学院,江苏南京210016

出　　处：《中国制造业信息化（学术版）》2010年第6期46-50,共5页

基　　金：国防科工委十一五重大基础科研项目

摘　　要：为了比较针对遗传算法的早熟收敛和后期收敛速度慢而提出的改进选择算子的性能,利用3种权威测试函数对基于排名的轮盘赌选择、基于上限的确定式采样选择、无回放最大值选择以及基于小生境技术的(μ+λ)选择这4种改进的选择算子进行了收敛速度、收敛可靠性和运算成本3方面的测试比较,并对测试结果作了详细的分析。最后,将采用基于小生境技术的(μ+λ)选择算子的遗传算法应用于一个混合装配流水线上最小装配时间协同优化的实际问题中,获得了较优的优化结果。Aiming at the premature convergence and weak local optimization problems of genetic algorithm （GA）, many researches propose improved selection operators. In order to compare the performances of these improved selection operators, it researches four kinds of improved selection operators, which are roulette wheel selection operator based on ranking, deterministic sampling based on upper- limit, biggest value selection without replacement and （μ＋λ） selection operator based on niche technology. It tests these three kinds of authoritative test functions from three aspects of convergence rate, convergence reliability and computational cast, analyzes the test results in detail. Finally it applies this method in a case of co - optimization for mixed -model assembly line to minimize the makespan and obtains the better result.

关 键 词：改进的选择算子 收敛可靠性 收敛速度 运算成本 混流装配线 

分 类 号：TH165[机械工程—机械制造及自动化] TP301.6[自动化与计算机技术—计算机系统结构]