An Improved Anti-Noise Network for Bearing Fault Diagnosis under Noisy Environment

Abstract

Recently, deep learning (DL) has been widely used in vibration-based rolling bearing fault diagnosis. However, the collected vibration signals always contain a lot of noise in real industries. It is a huge challenge to ensure the accuracy and stability of the DL-based fault diagnosis methods. Thus, an improved Anti-Noise network (IANet) is proposed to address this problem. Firstly, the K-singular value decomposition (K-SVD) method is employed to reduce noise. Then, an adaptive residual block (ARB) with a 1D convolutional block attention module (1D-CBAM) is introduced to extract the representative fault features from the channel and spatial dimensions. An IANet based on the ARB is designed to achieve intelligent bearings fault diagnosis. Finally, experiments are carried out on rolling bearing data to verify the effectiveness of the IANet. Results demonstrate the excellent performance of the IANet under noisy environment.