小波包能量谱和RVM在自动机故障诊断中的应用

Application of Wavelet Packet Energy Spectrum and RVM in Automaton Fault Diagnosis

针对传统自动机维修保障模式操作繁琐、维修周期长的问题,提出了一种应用小波包能量谱信息和相关向量机(Relevance vector machine,RVM)相结合的故障诊断方法。对每一组自动机振动信号进行小波包分解,得到不同频率成分的子频带分量,计算子分量占原信号能量的百分比,实现自动机状态信息表征,最后将特征输入RVM中进行分类识别。自动机故障诊断实例表明,该方法能较理想的实现自动机故障诊断,达到较高的诊断准确率。此外,通过对比支持向量机(SVM)的诊断结果,验证了RVM可以在很大程度上提升故障诊断的稀疏性与实时性。

Aiming at the drawbacks that the traditional maintenance mode of automaton is operated complicatedly and the maintenance cycle is too long,a method that based on the combination of wavelet packet energy spectrum and RVM was proposed.Wavelet packet is used to decompose the vibration signals.Then,the sub-band components of different frequency are obtained.The representation of state information is achieved by calculated the energy percentages of each band with the original signal.Finally,the characteristic matrix was put into RVM to recognize the different fault types.The experimental result of automaton shows that the method can classify usual fault types of automaton exactly and can achieve a higher recognition accuracy.In addition,by compared with the diagnostic result of SVM,the conclusion that RVM can improve the sparsity and real-time of fault diagnosis can be verified.