Early Fault Detection of Rolling Elememt Bearings Based on Visibility Graph Modeling of Vibration Signals

In this paper, the relevant information hidden in the frequency components of the vibration signal in the rolling element bearing (REB) is an important indicator of the REB status. In order to make better use of this information, visibility graph is introduced into REB failure analysis. The original signal is firstly converted into the time-frequency domain via short-time Fourier transform (STFT), and then the periodogram method is adopted to extract the power spectrum. Afterwards, visibility graph is utilized to model the power spectrum to extract relevant information between each frequency component. Next, anomaly score will be obtained based on specific distance metric. Finally, faults will be detected through a retrospective analysis method. The theory crafting and experiments implemented in a REB fault scenario have proved the feasibility and superiority of this method.