Rolling bearing fault diagnosis based on wavelet packet analysis and multi kernel learning

To diagnose the rolling bearing fault more accurately, a method based on wavelet packet analysis and multi kernel learning was proposed for rolling bearing fault diagnosis. Firstly, the three-level wavelet packet decomposition algorithm was used to decompose the vibration signals into different frequency bands. The relative energies of each frequency band were computed to form the feature vector. Secondly, multi kernel learning was employed to learn a kernel function and classifier from the training samples. Finally, the trained classifier was used to identify the fault type of rolling bearing. To validate the proposed method, an experiment of fault diagnosis for the rolling bearings was carried out. The results show that the fault diagnosis accuracy rate of the proposed method reaches 98.25%, higher than the traditional rolling bearing fault diagnosis method based on the wavelet packet and support vector machine. Since the problem of kernel function selection is avoided, the proposed method is more convenient for practical application.