A novel framework for bearing fault diagnosis across working conditions based on time-frequency fusion and multi-sensor data fusion

Abstract The condition of bearings significantly impacts the healthy operation of rotating machinery. However, bearings are prone to failure under a harsh working environment and alternating load. Integrating time-domain, frequency-domain, and multi-sensor data information has been a remarkable way to improve the accuracy and robustness of bearing fault diagnosis. How to combine these pieces of information remains a significant challenge. A novel network architecture called time-frequency multi-sensor fusion network is developed to address this issue. Firstly, a multi-scale feature extraction module based on a one-dimensional convolutional neural network is proposed for extracting multi-scale information from time-domain signals. Secondly, a multi-sensor data fusion strategy based on scaled dot product attention is applied to facilitate feature interaction among multi-sensor data. Thirdly, a time-frequency fusion module is designed to fuse the time-domain and frequency-domain features from multi-sensor. Finally, the effectiveness and superiority of the proposed method are validated on the Paderborn dataset.