Multi-Scale CNN based on Attention Mechanism for Rolling Bearing Fault Diagnosis

Abstract

In recent years, deep learning has shown great vitality in the field of intelligent fault diagnosis. However, most diagnostic models are not yet capable enough to capture the rich multi-scale features in raw vibration signals. Therefore, a multi-scale, attention-mechanism based, convolutional neural network (MSAM-CNN), is proposed to automatically diagnose health states of rolling bearings. The network is one-dimensional, and the information of the original vibration signal on different scales is processed by a parallel multi-branch structure. Then the learned complementary features from different branches are fused. Meanwhile, the attention mechanism can automatically select the optimal features. The MSAM-CNN is evaluated on the bearing dataset that is provided by Case Western Reserve University (CWRU). Experimental results indicate that the proposed network can greatly improve the fault recognition ability of the convolutional neural network, and the MSAM-CNN is superior to four forefront deep learning fault diagnosis networks under strong noise interference.