A dual attention module and convolutional neural network based bearing fault diagnosis

Abstract

Vibration signals of rolling bearings are affected by changing operating conditions and environmental noise, so they are characterized by a high degree of complexity. Although deep learning fault diagnosis methods have achieved considerable success in practical applications, the high complexity characteristics are ignored. To address this issue, we propose a dual attention module and convolutional neural network (DAM-CNN) for rolling bearing fault diagnosis. In this method, we designed a dual-attention module (DAM) by using a channel-attention module and a spatial-attention module. DAM can recode feature information in channel and spatial dimensions, so as to achieve adaptive enhancement of effective network information and suppression of interference information. In addition, to enhance the extraction of long-range features of the convolutional network, we introduce the non-local feature extraction module. This module can significantly expand the perceptual field of convolutional operations and enhance the generalization ability of the network. By verifying the effectiveness of the method in CWRU datasets, the results show that the method in this paper not only has good noise immunity in strong noise environment, but also has high diagnostic accuracy and good generalization performance in different load condition domains.