基于CPSO-BP神经网络的柴油机排气门间隙故障诊断  

CPSO-BP neutral network based fault diagnose on diesel engine valve clearances

作　　者：李岩[1] 袁惠群[1,2] 梁明轩[1] 赵天宇[2] 

机构地区：[1]东北大学机械工程与自动化学院,沈阳110819 [2]东北大学理学院,沈阳110819

出　　处：《中国工程机械学报》2014年第1期56-61,共6页Chinese Journal of Construction Machinery

基　　金：国家高技术研究发展计划资助项目(2012AA040104);国家自然科学基金资助项目(51275081)

摘　　要：针对柴油机排气门间隙故障信号不易提取的特点,提出了将混沌粒子群神经网(Chaotic particle swarm optimization-Back Propagation,CPSO-BP)聚类模型应用于柴油机排气门间隙故障诊断.首先,采用经验模式分解(Empirical Mode Decomposition,EMD)对柴油机振动信号进行分解,将得到的前6个模态函数能量百分比作为反映故障状态的特征参数,重构BP神经网并用混沌粒子群算法对其结构和权值进行优化;最后,基于优化的神经网对排气门间隙为0.2mm,0.4mm,0.6mm等3种故障工况的信号进行聚类.结果表明:所有样本的测试结果均与实际状况一致,该方法可以较好地用于排气门间隙故障诊断.Pertaining to the uneasy feature extraction of fault signals from diesel engine valve clearances,a new fault diagnosis method is proposed using CPSO-BP.Firstly,the diesel engine vibration signals are decomposed via EMD.Then,the energy percentage of the first-six-order modal functions is treated as a feature parameter of fault states.Next,relevant structures and weights are recomposed integrating BP with CPSO.Finally,the fault signals under working conditions of 0.2,0.4 and 0.6 mm valve clearances are clustered based on the optimized neural network.Therein,owing that the sample testing results are coincided with practical conditions,this approach can be effectively applied for valve clearance fault diagnosis.

关 键 词：柴油机 故障诊断 经验模态分解 神经网络 混沌粒子群算法 

分 类 号：TK421.3[动力工程及工程热物理—动力机械及工程]