Uncertainty in Remaining Useful Life Prediction
Stephen Ekwaro-Osire1; Haileyesus B. Endeshaw1; Duc H. Pham1; Fisseha M. Alemayehu2
1 Texas Tech University; 2 Penn State Hazleton
doi:10.20906/CPS/COB-2015-2705
Resumo
Often wind turbine gearboxes do not meet the expected operating lifetime due to premature failure in the drive-train system. The gearboxes fail after three to seven years of operation although the expected lifetime of the wind turbine is twenty years. The estimation of the remaining useful life (RUL) has emerged as an important tool to schedule critical maintenance and repairs to prevent failure of gearboxes. The RUL depends on the observed condition monitoring, operational environment, and the current state of the gearbox. It has been shown that understanding and accurately quantifying the impact of uncertainty are critical in estimating RUL of gearboxes. Existing methods for quantifying uncertainty in prognostics and RUL can be classified as testing-based and condition-based prognostics. Methods for testing-based prognostics are based on rigorous testing before and/or after operating an engineering system, whereas methods for condition-based prognostics are based on monitoring the performance of the system during operation. In the condition-based prognostics, the uncertainty in RUL is subjective. In this context, RUL is a random variable whose probability distribution is sought by casting the task as an uncertainty propagation problem. Several statistical methods have been proposed to solve this uncertainty propagation problem. However, each method introduces its own set of challenges. For this study, the following research question was developed: Does Monte Carlo method improve the quantification of uncertainty in RUL prediction? To answer this research question, Monte Carlo method was used in RUL prediction. The results of this study showed that the Monte Carlo method improves the quantification of uncertainty in RUL prediction. Moreover, the probabilistic sensitivity analysis showed random variables that had the largest contribution to the RUL. The impact of the proposed framework on the uncertainty in RUL of wind turbine gearboxes is highlighted and advocated.
Palavras-chave: condition monitoring; prognostics; remaining useful life; uncertainty; wind turbine gearbox