A novel rolling bearing modelling method and its application in bearing intelligent fault recognition

Traditional dynamic models reveal a vibration response that is difficult to accurately represent the bearing fault state due to the fewer degrees of freedom (DOFs) and insufficient consideration of influencing factors. Therefore, this paper first establishes a multiple degree-of-freedom (MDOF) rolling bearing dynamic model considering multiple factors to deduce the onset and evolution of bearing fault. Secondly, to enhance the fidelity of the model, grey wolf optimiser (GWO) is employed to optimise principal dynamic parameters, thereby reducing discrepancy between simulated and measured vibration responses. The effects of friction coefficient and oil film lubrication on bearing vibration response are revealed. The variation rule of bearing vibration response under different speeds and fault sizes is investigated. Finally, two case experiments are performed to assess the established model. The results demonstrate that the optimised model can accurately reproduce bearing vibration features under different fault conditions. Furthermore, the simulation data generated from the established dynamic model is utilised to augment samples, which increased the bearing fault recognition accuracy to over 99%.