改进策略的粒子群算法及其实验测试

摘要

针对研究敷设薄吸声层的多层结构材料特性参数反演中,目标函数复杂、具有多极值的问题,在研究了现有粒子群算法陷入局部最优原因及各项影响因素的基础上,提出一种改进的动态混合粒子群算法.该算法引入了动态跟踪邻域项,改进了惯性权重和学习因子的调整策略,并对新算法收敛性和有郊性进行了分析.通过实验测试和反演算例仿真表明,该算法可以有效提高算法的搜索成功率、收敛速度和准确性,全局搜索性能得到显著提高.%The physical parameters inversion of multilayer structure covered with thin sound absorption material had complex and multimodal target function. A improved dynamic mixed PSO is provided. The new algorithm inducted part dynamically tracking neighbour and impoved the adjustment strategies of inertia weight and study factors. Then analysis and experiment on astringency and effectiveness of algorithm are conducted. The test and simulation results show that the improved algorithm is asymptotic convergence and has better performance in global search ability, convergence rate and accuracy.