Fault Feature Extraction and Diagnosis Method Based on Multi-Channel Feature Fusion Residual Network

Abstract

Density equalization for multichannel features is a research priority, especially considering the complexity of the signal features generated by industrial rotating parts. To balance the density of complex features in different channels, we developed a new deep learning model named a residual network (ResNet) with multichannel weighting (ResNet-MCW). We applied it to feature extraction and fault diagnosis of bearing vibration signals. The results show that the proposed method obtains fairly high diagnostic accuracy and is superior to the traditional deep learning methods for the rolling bearing datasets.