利用SVM和相关山形聚类分析识别滚动轴承状态

Identification of Rolling Bearing Status Using SVM and Relational Mountain Clustering Method

状态识别是机械设备故障诊断的关键环节。提出了一种结合支持向量机、主成分分析及相关山形聚类分析的滚动轴承状态识别方法。首先利用支持向量机对轴承特征参数进行分类,辨别是否异常;再利用主成分分析得到独立的、包含原参数主要信息的主成分;最后对这些主成分作聚类分析,达到诊断故障部位的目的。实例分析的结果表明,该方法可以有效地识别轴承状态,正确率达到96%。

The status identification is key step in the fsult diagnosis of mechanical equipment. A new method combining support vector machine （SW）,principal components analysis （PCA） and relational mountain clustering method （RMCM） is put forwards to identify rollins bearing staus. Firstly, the SVM is used to classify the data of characteristic parameters into two cluster, normal or abnormal; ,secondly, the independent principal components contained main information of origin parameters are obtained by PCA; lastly, the clustering analysis is made for principal components, so the failure position is diagnosed. The practical analysis example shows that this method can effectively identify bearing status with the correct rate up to 96%.