采用混合粒子群算法实现匹配追踪算法  

Matching pursuit based on hybrid particle swarm optimization algorithm

作　　者：张建军[1] 王仲生[1] 余汇[1] 

机构地区：[1]西北工业大学航空学院,西安710072

出　　处：《振动与冲击》2010年第1期143-147,共5页Journal of Vibration and Shock

基　　金：国家自然科学基金(50675178)

摘　　要：针对匹配追踪信号稀疏分解的巨大计算量问题,在具有全局优化能力的粒子群算法基础上,提出了一种结合BFGS(Broyden、Fletcher、Goldfarb和Shanno)方法和变异操作的混合粒子群算法实现信号匹配追踪分解。利用BFGS方法增强了算法的局部开发能力,加快了信号特征提取速度;通过变异操作控制种群多样性以避免早熟收敛,增强了算法全局探测能力,提高了信号特征提取精度。通过与单一粒子群算法和遗传算法实现仿真信号匹配追踪分解的结果进行对比,证明了使用混合粒子群算法的匹配追踪分解能够快速准确提取信号特征参数。最后,将该算法应用于某内圈损伤轴承振动信号中的冲击特征提取,结果表明该算法在工程应用中具有一定的准确性和实用性。A hybrid particle swarm optimization algorithm(HPSO) to implement matching pursuit was developed, where BFGS(Broyden,Fletcher,Goldfarb and Shanno)method was combined with particle swarm optimization algorithm (PSO) to speed up the local search,and mutation operation was embedded to avoid premature convergence.The HPSO could overcome the disadvantages existing in traditional optimization algorithms:poor convergence rate and decomposition accuracy.Compared with results of using the single PSO and genetic algorithm to implement matching pursuit in impulse atoms dictionary,the identification accuracy and speed of signal characteristics were improved through HPSO computation simulation.Meanwhile,the periodic impulses were extracted in joint time-frequency domain,and the single point defect in inner race of a rolling element bearing was identified in a rotation machine test rig accordingly.Results showed that the matching pursuit using HPSO is applicable and effective.

关 键 词：粒子群算法 匹配追踪 BFGS方法 变异 

分 类 号：TP277[自动化与计算机技术—检测技术与自动化装置]