Application of Data Driven Method Combining LBP with RANet in Rolling Bearing Fault Diagnosis

Rolling bearing is the core part of rotating machinery, its health affects the daily operation and maintenance of mechanical equipment directly. In the actual working condition, the detected fault signal is often accompanied by excessive background noise, therefore, in order to enhance the fault diagnosis accuracy of rolling bearings, this paper proposes a rolling bearing fault diagnosis method based on the combination of LBP(Local Binary Patterns) and RANet(Residual Attention Network) under data driven. Firstly, the LBP feature extraction method has been introduced to transform the one-dimensional time-series fault signal into two-dimensional feature images to improve the characterization capabilities of the fault features. On the basis of ResNet(Residual Network), attention mechanism and soft threshold function are introduced to obtain RANet to improve the diagnostic performance of neural network in strong noise background. Migration learning is introduced to optimize the weight values of RANet to reduce training time of the model and improve the fault diagnosis accuracy. At last, the real bearing fault signals have been analyzed, and the results show the proposed method has high fault diagnosis accuracy in the noisy background.