基于盲分离的电机故障诊断  

Electric Motor Fault Diagnosis Based on Blind Source Separation

作　　者：张敬春[1] 谷爱昱[1] 王战盟[1] 

机构地区：[1]广东工业大学自动化学院,广州510090

出　　处：《电力系统及其自动化学报》2006年第4期67-70,共4页Proceedings of the CSU-EPSA

摘　　要：电机故障发生时通常是多种故障同时发生,利用采集到的故障电机声音信号作为分析对象,基于盲分离理论,利用声源信号之间的相互独立性,对采集到的信号进行白化处理,采用一种已知较好的盲分离算法,从信号的联合概率分布密度出发,利用信号的联合概率的方向导数熵的最小值获得最佳旋转角度,对混合信号进行盲分离处理,得出令人满意的结果。成功分离了故障电机的各个单个故障的信号,验证了采用算法的有效性。Several faults may happen at the same time. This paper applies blind source separation（BSS） to analyze the noise of electric motor,makes use of the independence among the source signals to whiten the gathered signal ,adopts a well used algorithm based on the joint probability statistics of the signal ,and obtains the optimum rotation angle according to the direction derivative entropy of joint probability distribution minimum. The test shows that it succeeds in separating fault signal of electric motor and thus verifies the validity of the method.

关 键 词：盲分离 联合概率 白化 电机 故障诊断 

分 类 号：TH301.4[机械工程—机械制造及自动化]