基于主分量子空间的设备状态诊断研究

Machine condition diagnosis based on principal component subspace-based methods

结合主分量分析(Principal Component Analysis,PCA)和子空间法研究了基于主分量子空间的设备状态诊断,探讨了压缩子空间和类属子空间两种主分量子空间结构来表达和分类设备的状态。所提出的设备状态诊断方法依靠PCA可以提取稳定有效的设备状态低维特征表示,依靠子空间法能够以低代价有效辨识设备状态。以汽车变速齿轮箱的疲劳状态诊断为例,分析表明两种主分量子空间方法都获得了良好的结果,且具有各自的优点,可以有效地用于设备的状态监测和诊断中。

Machine condition diagnosis was studied by combining the principal component analysis（PCA） and the subspace methods.With the principal component subspace-based methods,two subspace structures,called information compression subspace and class-specific subspace,were presented to represent and classify machine condition patterns.The proposed methods could extract effective and stable low-dimensional features of a machine condition with PCA,and effectively identify a machine condition with lower computational cost based on the subspace methods.Experimental analysis for fatigue condition diagnosis of an automobile transmission gearbox showed that both of two principal component subspace-based methods can obtain better results with the corresponding advantages,and can be effectively applied to machine condition monitoring and diagnosis.