混合粒子交互微粒群算法  

Hybrid-particle Interaction Particle Swarm Optimization Algorithm

作　　者：姚成玉[1] 王斌[1] 陈东宁[2,3] 张瑞星[2,3] 

机构地区：[1]燕山大学河北省工业计算机控制工程重点实验室,秦皇岛066004 [2]燕山大学河北省重型机械流体动力传输与控制重点实验室,秦皇岛066004 [3]先进锻压成形技术与科学教育部重点实验室(燕山大学),秦皇岛066004

出　　处：《机械工程学报》2015年第6期198-207,共10页Journal of Mechanical Engineering

基　　金：国家自然科学基金(51405426);河北省自然科学基金(E2012203015);河北省教育厅资助科研(ZH2012062)资助项目

摘　　要：针对现有微粒群算法仅考虑单一一种引斥力规则使得其搜索能力存在的不足,考虑在不同搜索阶段采用不同的引斥力规则,提出搜索后期引力增强型混合引斥力微粒群算法(LAPSO算法)。利用拟态物理学中的引斥力规则使粒子保持多样性,提高算法的全局搜索能力;当进入到具有全局最优解的区域时,增强引力作用、减少斥力作用,利用比自身适应度好的粒子和全局最优解粒子的引力作用,提高算法的局部搜索能力。为进一步提高LAPSO算法的优化性能,将其与混合全连接型-环形拓扑结合,提出混合粒子交互微粒群算法(HIPSO算法)。通过6个Benchmark函数进行测试,结果表明,与现有的扩展-微粒群、微-微粒群、中值导向-微粒群等算法相比,所提的LAPSO算法、HIPSO算法具有较好的种群多样性,具有更好的寻优精度、收敛率和最优解搜索能力。结合文献[7]中的柔性流水车间调度离散优化实例和文献[20]中的超声振动加工工艺参数连续优化实例,验证了HIPSO算法的最优解搜索能力。To overcome the searching shortages of the existing particle swarm optimization algorithms only considered a single kind of attraction and repulsion rules, different attraction and repulsion rules should be considered in different searching stages, later-stage attraction-enhanced hybrid attraction and repulsion particle swarm optimization algorithm（LAPSO algorithm） is proposed： At the early stage, the diversity of particles are maintained by the rules of attraction and repulsion in artificial physics, to improve the global searching ability; When the particles move to the global optimal solution area, enhanced the effect of attraction and reduced the effect of repulsion, using the attractions of other particles with better fitness values and the global optimal solution particle to improve the local searching ability. In order to further improve the optimal performance of LAPSO algorithm, hybrid-particle interaction particle swarm optimization algorithm （HIPSO algorithm） is proposed by combining LAPSO algorithm and hybrid fully connected-ring topology. The test results of six Benchmark functions show that the proposed LAPSO algorithm and HIPSO algorithm have better population diversity, better optimization precision, convergence rate and optimal solution searching ability than the existing extended-particle swarm optimization algorithm, micro-particle swarm optimization algorithm and median-oriented particle swarm optimization algorithm. The optimal solution searching ability of HIPSO algorithm is verified by the discrete optimization example of flexible flow-shop scheduling in the reference [7] and continuous optimization example of ultrasonic vibration process parameters in the reference [20].

关 键 词：微粒群算法 引斥力 LAPSO算法 混合粒子交互 HIPSO算法 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程] N945[自动化与计算机技术—控制科学与工程]