发动机轴承并发故障信号盲源分离方法研究

Blind Source Separation Method of Engine Bearing Concurrent Fault Signal

针对发动机轴承并发故障信号微弱、相互耦合且易被其他信号淹没的问题,提出了基于平行因子分析的轴承并发故障信号盲源分离方法,利用平行因子分析预测了信号传播路径响应函数,并以此构造了虚拟观测信号,结合实际观测信号,将盲源分离从欠定变为适定或超定,采用自适应平行因子分析算法,得到了反映故障特征的分离信号;实际应用中针对曲轴轴承与连杆轴承并发故障,分离信号的特征频率分布分别对应曲轴轴承和连杆轴承的故障频率,因此,所提出的盲源分离方法能够准确诊断出轴承并发故障。

For the weak, mutual coupling and easily-submerged problem of engine bearing concurrent fault signal, the blind source separation method was proposed based on the parallel factor analysis. The response function of propagation path was predicted with the parallel factor and the virtual observation signal was built. Combining the real observation signals, the blind source separation became determined or overdetermined from underdetermined. Through the adaptive parallel factor analysis method, the signals reflecting fault feature were acquired. In the practical application to crankshaft bearing and connecting rod bearing concurrent fault signals, the feature frequencies of separation signal were corresponding to crankshaft bearing and con- necting rod bearing fault frequency respectively. Accordingly, the proposed blind source separation method could diagnose the bearing concurrent fault correctly.