基于反向学习模型的多目标进化算法  

Multi-objective Evolutionary Algorithm Based on Opposition-based Learning Model

作　　者：王亚辉[1] 吴金妹[1] 贾晨辉[2] 

机构地区：[1]华北水利水电大学机械学院,郑州450011 [2]河南科技大学机电工程学院,洛阳471023

出　　处：《农业机械学报》2016年第4期326-332,342,共8页Transactions of the Chinese Society for Agricultural Machinery

基　　金：国家自然科学基金项目(51475142)

摘　　要：针对复杂多目标优化问题,提出一种基于分解机制和反向学习模型的多目标进化算法。该算法在基于分解机制的多目标进行算法的框架下,引入反向学习模型,该模型具有较好的局部寻优能力。在种群进化的过程中,反向学习模型和差分进化机制自适应的相互配合,能够较好地平衡算法的全局搜索与局部寻优能力。采用国际公认的具有复杂Pareto Set的LZ09系列测试问题进行实验验证,并与MOEA/D-DE、GDE3、NSGA-II和SPEA2等方法比较,实验结果表明,所提方法能够获得收敛性、分布性及延展性较好的Pareto最优解集。为了研究算法在求解约束问题的性能,将其应用于减速器多目标优化设计问题中,结果表明了该算法获得Pareto前端较均匀,说明其算法具有求解约束问题的能力和工程有效性。A multi-objective evolutionary algorithm cooperated with decomposition mechanism and opposition-based learning model was proposed for solving complex multi-objective optimization problems.Under the framework of multi-objective evolutionary algorithm based on decomposition,the oppositionbased learning model was introduced into the algorithm. The model improved the algorithm ＇s exploitation. During the evolution process,the opposition-based learning model facilitated the local optimization and the differential evolution strategy enhanced the global research for the new algorithm.The opposition-based learning strategy and differential evolution were in coordination to balance its exploration and exploitation. The benchmark LZ09 series of internationally recognized with complicated Pareto sets were adopted to verify its effectiveness. The proposed multi-objective evolutionary algorithm based on opposition-based learning model was compared with MOEA / D based on DE（ MOEA / D-DE）,the third evolution step of generalized differential evolution（ GDE3）,fast and elitist multi-objective genetic algorithm（ NSGA-II） and improving strength Pareto evolutionary algorithm（ SPEA2）,the results showed that the proposed algorithm can obtain Pareto fronts with good convergence,diversity and wild coverage. In order to analyze the algorithm to solve the problem of performance constraints,the proposed algorithm was applied to solve the multi-objective optimization design of speed reducer. The results showed that the Pareto front obtained by the algorithm was uniform, which demonstrated its good performance in solving practical problem with constraints and engineering effectiveness.

关 键 词：多目标优化 MOEA/D 反向学习模型 减速器 优化设计 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]