Fault Diagnosis of Rolling Bearing Based on KPCA and Stacking Algorithm

Aiming at the nonlinear relationship between bearing fault signals, a fault diagnosis method based on KPCA and stacking algorithm is proposed to realize the common fault identification of rolling bearing. Firstly, Empirical Mode Decomposition (EMD) is conducted to decompose the bearing signal and extract the features to obtain the running state information of the bearing in different states. Then, Kernel Principal Component Analysis (KPCA) is applied to fuse features and reduce the dimension of bearing signals to reduce the influence of nonlinear correlation on fault identification. Finally, the stacking algorithm is used to identify the bearing fault signal, and the test data is used to validate it. The results show that the stacking algorithm based on KPCA can effectively identify the types of bearing fault.