基于核函数主元分析的机械设备状态识别

Machine condition recognition based on kernel principal component analysis

研究了核函数主元分析在机械故障模式分类中的应用 .通过计算原始空间的内积核函数实现原始数据空间到高维数据空间的非线性映射 ,再对高维数据作主元分析 ,求取更易于分类的核函数主元 .实验表明 ,核函数主元分析更适于提取故障信号的非线性特征 ,能有效区分不同的故障模式 ,可以应用于机械设备的状态识别 .

Describes a study of kernl Principal Component Analysis in Mechanical faults feature extraction and clessification. By using integral operator kernel functions, the nonlinear principal components in high dimensional feature spaces related to input space by some nonlinear map were computed. Industrial gearbox vibration signals measured from different operating conditions were analyzed using the above method. Experiment results indicate that kernel PCA has good performance in extracting the nonlinear feature from fault signals, and it is able to identify clearly a gearbox operating condition with fatigue spalling or wear compared with the normal condition. It can be applied in machine condition dynamic recognition.