基于聚类分析的转子碰摩声发射特征提取

摘要

根据碰摩声发射信号的特点,选用波形幅值、脉冲指标、裕度指标和峭度指标作为声发射信号聚类分析的特征参数.利用聚类分析,提取转子试验有、无碰摩两种情况下的聚类中心声发射信号,并采用伪Wigner-Ville分布对各中心信号进行分析和对比,从而得到转子碰摩故障的声发射信号特征频率.试验结果表明:聚类分析方法能有效区分出碰摩声发射信号,为碰摩故障诊断提供了一种新方法.%The amplitude of acoustic emission( AE) waveform,impulse index,margin index and kurtosis index were selected as parameters for cluster analysis based on the characteristic of AE signal. AE signals of cluster center were extracted in two cases of rotor experiment with rubbing and without rubbing using cluster analysis method. These signals were analyzed and contrasted by pseudo Wigner-ville distribution to obtain characteristic frequency of AE signals with rub fault of rotor. The result shows that the rubbing AE signals are efficiently discriminated by cluster analysis that is a new method for diagnosing rubbing fault.