This paper proposes a method for effectively extracting fault features and identifying fault patterns from the early impact vibration signals of the rolling bearings of complex equipment which is non-stationary, nonlinear and has strong background noise.

First, the fault features of the original vibration signals are extracted via fast spectral correlation analysis and quantified via multi-scale permutation entropy (FSC-MPE). The fault feature data is then input into a BP neural network for fault diagnosis model training and testing. Finally, fault identification research is carried out on the rolling bearing fault simulation experimental data under variable speed and the public bearing fault test dataset of Case Western Reserve University.

The results show that the proposed method has high identification accuracy for different types of faults, reaching more than 97%.

The feasibility and superiority of the proposed rolling bearing fault diagnosis method based on FSC-MPE and BP neural network are verified, and it can provide technical support for rolling bearing health evaluation.