Uncertain Generalized Remaining Useful Life Prediction-driven Predictive Maintenance Decision

Predictive maintenance based on generalized health assessment is a flourishing approach, which can decrease maintenance cost and increase operational availability efficiently. Remaining useful life prediction (RULP) is a kind of health assessment methods, which is fundamental and vital for predictive maintenance. But current predictive maintenance suffers the disconnection between RULP and maintenance decision, the aim of this paper is to solve uncertain RULP-driven predictive maintenance decision. Firstly the uncertainty indices of RULP such as false alarm rate and missed detection rate are presented. Secondly a maintenance behavior evolution process is described, and an optimal generalized RULP-driven predictive maintenance decision model is proposed for perfect replacement case. Furthermore the effect of RULP period and maintenance threshold on predictive maintenance decision is drawn using Monte Carlo simulation. Finally this approach is demonstrated with a case study on gearbox bearings. The results show that RULP-driven predictive maintenance policy can decrease average cost rate and increase average operational availability.