Bearing Intelligent Fault Diagnosis Under Complex Working Condition Based on SK-ES-CNN

At present, most of the existing bearing fault diagnosis methods focus on a single working condition. However, it is far from the complex working condition with changeable motor speed, environmental noise interference and the weakness of early feature in the real industrial applications. Therefore, it is very significant to determine appropriate features for intelligent fault diagnosis of rolling element bearings (REBs) under complex working conditions. To solve this problem, an intelligent diagnosis method of bearing faults based on spectrum kurtosis (SK), envelope spectrum (ES) and convolutional neural net (CNN) is proposed in this paper under variable rotational speed and multiple fault states. In this method, SK and bandpass filtering are firstly used to improve the signal-to-noise rate (SNR) of fault from the original vibration signals. Then the rich information of fault characteristic frequencies related to the rotating speed is extracted by ES analysis. Subsequently, a CNN model is built to identify bearing defects by automatically extracting these representative features. Four experiments are performed on the Case Western Reserve University (CWRU) bearing dataset to demonstrate the effectiveness of this method. By comparing experiment results with others, the superiority and effectiveness of this method are illustrated.