Fault diagnosis of rolling bearings based on improved empirical wavelet transform and IFractalNet

Considering that the traditional methods for rolling bearing fault diagnosis largely depend on expert prior knowledge and easily to be disturbed by noise, a method based on improved empirical wavelet transform (IEWT) and improved FractalNet (IFractalNet) was proposed. Firstly, the segmentation method of Fourier spectrum of empirical wavelet transform was improved, and the raw vibration signals of bearings were adaptively decomposed into several intrinsic modal functions (imfs). The imfs which can best reflect the fault characteristics of the raw signals were selected using the comprehensive evaluation index based on kurtosis, correlation coefficient, and energy ratio. Secondly, the loss function and activation function of FractalNet were improved to solve the imbalance problem of sample set. Finally, the selected imfs were reconstructed and fed into IFractalNet for automatic feature extraction and fault recognition. The experimental results indicate that the proposed method can effectively identify the bearings with multiple fault types and multiple fault severities, which is less affected by prior knowledge and subjective knowledge and avoids complex artificial feature extraction process. The generalization ability, feature extraction ability, and recognition ability of proposed method are superior to other methods. © 2020, Editorial Office of Journal of Vibration and Shock. All right reserved.