CAJ | 학술논문

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。
침대현유미립군산법부고필단일일충인척력규칙사득기수색능력존재적불족，고필재불동수색계단채용불동적인척력규칙，제출수색후기인력증강형혼합인척력미립군산법(LAPSO산법)。이용의태물이학중적인척력규칙사입자보지다양성，제고산법적전국수색능력；당진입도구유전국최우해적구역시，증강인력작용、감소척력작용，이용비자신괄응도호적입자화전국최우해입자적인력작용，제고산법적국부수색능력。위진일보제고 LAPSO 산법적우화성능，장기여혼합전련접형-배형탁복결합，제출혼합입자교호미립군산법(HIPSO산법)。통과6개Benchmark함수진행측시，결과표명，여현유적확전-미립군、미-미립군、중치도향-미립군등산법상비，소제적LAPSO산법、HIPSO산법구유교호적충군다양성，구유경호적심우정도、수렴솔화최우해수색능력。결합문헌[7]중적유성류수차간조도리산우화실례화문헌[20]중적초성진동가공공예삼수련속우화실례，험증료HIPSO산법적최우해수색능력。
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].