基于非线性盲提取的柴油机振动诊断研究  

Study on Vibration Diagnosis of Diesel Engine Based on Nonlinear Blind Sourse Processing

作　　者：周勃[1] 陈长征[1] 张宇[1] 勾轶[1] 

机构地区：[1]沈阳工业大学振动噪声研究所,沈阳110023

出　　处：《噪声与振动控制》2008年第6期43-46,共4页Noise and Vibration Control

基　　金：高等学校博士学科点专项科研基金(20060142002)

摘　　要：提出一种在振动源个数未知的情况下,采用自然梯度算法分离环境噪声信号的同时提取柴油机早期多故障的方法。通过分析信号的非平稳特性,应用具有自适应时变特征的非线性激活函数,根据归一化峭度判断信号统计特性,从而高效真实地反映设备运行状态。为了有效提取信号特征,探讨不同传感器数量对信号分离精度和故障识别的影响,应用实例诊断出柴油机的磨损和撞击故障。结果表明,该方法可有效消除振动信号采集过程中混入的噪声,同时分离未知个数的故障源,为柴油机多故障识别诊断提供理论依据。A method is proposed by separating ambient noise signal and withdrawing the early-time multi-breakdowns of diesel engine based on natural gradient algorithm in the case that the number of vibration sources is unknown. Through analyzing the non-steady characteristic of the signals, the nonlinear activation function with self-adapted time-variable characteristic is applied to judge the signal＇s statistical property according to the normalized kurtosis. In this way, the equipment operating status can be reflected effectively. To extract the signal characteristic effectively, the influence of the number of sensors on separation precision of signals and fault recognition is discussed, and the diesel engine＇s attrition and hitting breakdowns are diagnosed in a practical example. Results indicate that the mixed noise in the vibration signals in the collecting process can be eliminated by this method, and the unknown sources can be separated. This method provides a theory base for multi-fault recognition and diagnosis of diesel engines.

关 键 词：振动与波 盲提取 非线性 多故障 柴油机 

分 类 号：TK428[动力工程及工程热物理—动力机械及工程] TH165.3[机械工程—机械制造及自动化]