A novel bearing fault diagnosis method based on principal component analysis and BP neural network

As a critical component in rotating machinery field, rolling bearings are prone to damage under the working conditions of high speed, heavy load and strong impact, resulting in the reduction of production efficiency or even production outage. Therefore, fault diagnosis of rolling bearing plays a significant role in improving the availability of the rotating machinery equipment. The fault recognition rate of the existing fault diagnosis techniques of wavelet packet energy feature is low. Thus, a fault diagnosis approach of bearing is proposed to address this issue in this paper. First, in view of the non-stationary and non-linear properties of the rolling bearing vibration signals, namely, time domain features, frequency domain features, time-frequency domain features and entropy feature are selected to form high-dimensional feature vectors. Second, principal component analysis (PCA) technique with dimension reduction ability is adopted to process high-dimensional features to further remove noise and redundant features and prevent over-fitting. Third, BP neural network is utilized to perform fault diagnosis. Finally, the rolling bearing vibration data of Case Western Reserve University (CWRU) is applied to verify the proposed approach. The diagnosis results shows that the proposed approach has higher fault recognition rate than the traditional wavelet packet energy features based fault diagnosis approach.