基于KPCA与RVM感应电机故障诊断研究  

Research on fault diagnosis of induction motor based KPCA and RVM

作　　者：阳同光[1,2] 桂卫华[2] 

机构地区：[1]湖南城市学院机械与电气工程学院,湖南益阳413000 [2]中南大学信息科学与工程学院,湖南长沙410083

出　　处：《电机与控制学报》2016年第9期89-95,共7页Electric Machines and Control

基　　金：国家高技术研究发展计划项目(2009AA11Z217);国家自然科学基金(61273158);湖南教育厅科学研究项目(11C0725)

摘　　要：针对感应电机非线性、强耦合、时变的特点,提出一种将核主成分分析(KPCA)和相关向量机(RVM)相结合的感应电机故障诊断方法。首先,对感应电机定子电流进行小波分解,并采用核主元分析方法有效去除信息的冗余,得到能反映感应电机运行状态的特征向量。然后,利用相关向量机对故障特征向量进行故障分类,识别感应电机的运行状态。通过对不同运行状态下感应电机进行识别分析,验证了此方法的可行性和实用性,并和其他3种方法比较,结果表明基于KPCARVM方法的故障诊断方法有较好的分类效果和泛化能力,是一种有效的感应电机故障诊断方法。According to the characteristics of induction motor, such as nonlinear, strong coupling and timevarying, a fault diagnosis method based on kernel principal component analysis （KPCA） and relevance vector machine （RVM） was proposed. Firstly, the induction motor stator current was decomposed using wavelet, and the KPCA approach was adopted to extract the feature vector and remove the redundant infor- mation effectively. Secondly, the relevance vector machine was used to classify the fault feature vectors and to identify the states of induction motor. The experiments were setup to verify the feasibility and prac- ticability of this method under different running condition. The results show that the method based on KP- CA- RVM has better classification effectively and better ability of generalization than other three methods and is an effective method for induction motor fault diagnosis.

关 键 词：核主成分分析 相关向量机 感应电机 故障诊断 

分 类 号：TP206[自动化与计算机技术—检测技术与自动化装置]