多邻域链式结构的多目标粒子群优化算法

Optimization of Multi-objective Particle Swarm Algorithm Based on Multi-neighborhood Cycle-chain Structure

为了提高多目标粒子群算法求解多目标问题的性能,改善算法的收敛性,提出一种多邻域链式结构的多目标粒子群优化算法。首先,以一种环形链式拓扑结构,将种群划分为多个邻域,每个邻域之间相互交叉重叠,并针对不同位置的粒子,进行不同的速度和位置更新策略。其次,对所有粒子采用速度钳制策略,并引入差分进化策略对粒子进行扰动,从而进一步提高算法的多样性。通过14个无约束和3个有约束函数仿真实验,表明该算法相对于NSGA-II、SPEA2、MOEA/D-DE、SMPSO和OMOPSO算法,获得Pareto解集分布更加均匀,算法的收敛性和多样性也更好。为了进一步验证算法的可行性和有效性,将其应用于72杆桁架结构尺寸设计,并与其他优化方法进行了比较,结果表明该算法获得的Pareto前端更均匀,收敛性更好。

In order to enhance the performance and convergence of multi-objective particle swarm optimization （MOPSO） algorithm for multi-objective optimization, a multi-neighborhood cycle-chain structure of multi-objective particle swarm optimization （MNCS-MOPSO） was proposed. Firstly, the population was divided into many neighborhoods. The mutual overlaps were existed between the adjacent neighborhood, and updating strategy was used for different velocity and position aimed at particles of different positions. In addition, velocity control strategy was adopted for all particles and differential evolution strategy was introduced to make disturbance. Comparing with NSGA-II, SPEA2, MOEA/D- DE, SMPSO and OMOPSO by testing 14 uneonstraint and 3 constrain benchmark functions, simulation experiments showed that the proposed algorithm could obtain a more uniform distribution of Pareto solution set, and better convergence as well as diversity than those state-of-the-art multi-objective metaheuristics. In order to verify the performance of MNCS-MOPSO algorithm, classical 72-bar truss sizing optimization problems were used to demonstrate the feasibility and effectiveness of this algorithm, and the results were compared with other optimization methods. The results indicate that the MNCS-MOPSO provides better performance in the diversity, the uniformity and the convergence of the obtained solution than other methods.