基于模式识别的复杂二十四脉波可控整流电路故障诊断方法

Fault diagnosis of complicated 24-pulse wave controllable rectifier circuit based on pattern recognition

对带有多个平衡电抗器的复杂二十四脉波可控整流电路在各种故障情况下的整流电压波形进行了详细分析和归类,定义“伪面积向量”以建立故障特征向量,然后根据故障特征向量的个数M进行分类,确定使用多个并联子网络中的一个,再利用BP子网络识别出故障类。最后根据三值逻辑编码中首次出现异常信息的位置诊断出故障类中的具体故障元。提出一种基于模式识别的故障诊断方法,并总结了本方法的特点。仿真实验结果表明,该方法能够快速、准确地同时诊断出各种故障。

The rectifier voltage waveforms are analyzed in details and classified under different circumstance for the complicated 24-wave controlled circuit with many balanced reactance.Firstly, ＂fake area vector＇expressions are defined to built fault eigenvector, then it confirms which one to use among these parallel BP sub-networks and identifies the fault category. Finally, based on the occurrence of exceptional information in the three numerical values code for the first time,it diagnoses the specified fault state of fault category. A fault diagnosis method based on pattern recognition is put forward,and the characteristic of this method is summarized. The simulation experimental results show that the algorithm can diagnose various faults fast, accurately and simultaneously.