Bearing fault prognosis based on Cyclical Remaining Useful Life (CRUL)

Over the last decade, prognosis has become an interesting line of research in preventive maintenance for a variety of industrial applications. It generally focuses on estimating the Remaining Useful Life (RUL) of mechanical systems. In this context, this paper proposes a new prognosis approach based on cyclical prediction of RUL. The main idea lies in the integration of cyclical prediction of the degradation based on an appropriate adjustment of the data. The proposed degradation model can monitor continuously the evolution of degradation and finally provide an instantaneous estimation of RUL throughout the bearing life. First, the Simple Moving Average (SMA), Cumulative Moving Average (CMA), and Exponential Moving Average (EMA) filters are compared. Then, an exponential degradation model is constructed and fitted to the final <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML"> <mml:mi>n</mml:mi> </mml:math> data using both Principal Component Analysis (PCA) and model fitting. Finally, the developed model is evaluated in cyclical estimation of the bearing RUL using vibration data acquired from an endurance test rig. The results demonstrate an accurate prediction throughout each phase of bearing life, which confirms its ability to be applied to the prognosis of bearing failure.