计算机工程与应用 ›› 2019, Vol. 55 ›› Issue (8): 250-263.DOI: 10.3778/j.issn.1002-8331.1809-0183

• 工程与应用 • 上一篇    下一篇

自适应简化粒子群优化算法及其应用

张  鑫1,邹德旋1,肖  鹏1,喻  秋2   

  1. 1.江苏师范大学 电气工程及自动化学院,江苏 徐州 221116
    2.江苏师范大学 机电工程学院,江苏 徐州 221116
  • 出版日期:2019-04-15 发布日期:2019-04-15

Self-Adjusted Simplified Particle Swarm Optimization Algorithm and Its Application

ZHANG Xin1, ZOU Dexuan1, XIAO Peng1, YU Qiu2   

  1. 1.School of Electrical Engineering & Automation, Jiangsu Normal University, Xuzhou, Jiangsu 221116, China
    2.School of Mechatronic Engineering, Jiangsu Normal University, Xuzhou, Jiangsu 221116, China
  • Online:2019-04-15 Published:2019-04-15

摘要: 针对粒子群优化算法(Particle Swarm Optimization,PSO)寻优速度慢、收敛精度不高且搜索结果波动性较大的缺点,提出了一种自适应简化粒子群优化算法(Self-Adjusted Simplified Particle Swarm Optimization,SASPSO)。在每次迭代过程中,粒子只受全局最优解影响,且加入按一定规律分布的锁定因子,令粒子受影响的程度有规律性。同时,利用锁定因子和当前粒子位置令惯性权重自适应配置,更有效地利用惯性权重对粒子群优化算法的影响。引入4种近期提出的改进粒子群算法同时搜索不同维度时的18个基准函数,与SASPSO的搜索结果对比,并使用T-test进行差异性分析。为了进一步分析算法性能,统计5个改进算法搜索100维函数达到期望值时的成功率与平均迭代次数。实验结果证明,SASPSO在无约束问题寻优中的收敛速度、寻优精度有了明显提升,且搜索结果异常值较少,波动性弱。将SASPSO应用于机床主轴结构参数优化问题,结果显示SASPSO优化性能更好。

关键词: 粒子群优化算法, 自适应简化粒子群算法, 群体智能, 基准函数, 无约束问题, 优化设计, 机床主轴

Abstract: In order to overcome the disadvantages of slow search speed, low convergence accuracy and high search volatility of particle swarm optimization, a Self-Adjusted Simplified Particle Swarm Optimization(SASPSO) algorithm has been proposed. During the evolutionary process, the particles are only affected by the global optima, and the locking factor distributed according to a certain rule is added so that the degree of influence of the particles can be traced. At the same time, the inertia weight is adaptively adjusted by using the locking factor and the current particle position, which makes the influence of inertia weight on the particle swarm optimization algorithm more effective. Four improved particle swarm algorithms proposed recently are introduced to simultaneously search for eighteen benchmark functions with different dimensions. The results are compared and T-test is used to analyze their differences. The success rates and the average iteration times are recorded when five improved algorithms search for 100-dimensional functions, which is beneficial for analyzing the performance of the proposed algorithm. Experimental results show that the convergence speed and optimization accuracy of SASPSO have been improved significantly, and the search results have lower objective function values and smaller standard deviations. The SASPSO is applied to the optimum design of machine-tool spindle parameters. The results show that the SASPSO optimization performance is better.

Key words: particle swarm optimization algorithm, self-adjusted simplified particle swarm optimization, swarm intelligence, benchmark function, unconstrained problem, optimum design, machine-tool spindle