Vision-based bearing fault diagnosis under non-stationary conditions using optimized short-time concentrated transform method

The condition of rolling bearings is closely related to the economy and safety of industrial production. The fault diagnosis of bearing under time-varying speed can realize the state analysis more comprehensively and deeply. However, due to the influence of a complex industrial field environment, there are many problems in equipment signal acquisition, and it’s difficult to achieve efficient real-time monitoring and acquisition. Limited by the difficulty of image matching and the extremely weak amplitude, there are still few research results on visual fault diagnosis of bearings. Therefore, in this paper, visual vibration measurement is introduced into the field of bearing fault diagnosis. Combined with LK optical flow method, the vibration signal is collected and extracted by an industrial high-speed camera. An enhanced time-frequency (TF) resolution method based on improved short-time centralized transform is proposed to effectively improve TF resolution and extract ridge line, to realize bearing fault diagnosis under unsteady conditions through video signal. A numerical simulation signal and rotating machinery fault simulation experiment system are used to verify the method. The results show that the vision-based signal acquisition method is feasible, and the proposed method is effective for TF analysis of bearing faults under unstable conditions based on video signals.