Remaining useful life estimation methods for predictive maintenance models: defining intervals and strategies for incomplete data
A. Delmas, M. Sallak, W. Schon and L. Zhao
https://doi.org/10.19124/ima.2018.001.09
Abstract
Predictive maintenance allows the drawbacks of corrective and preventive maintenance to be overcome by estimating the Remaining Useful Life (RUL) of components. This value helps undertake required tasks at the right time: when the component is deteriorated but right before a failure occurs.
We propose an imprecise RUL estimation method with the use of Hidden Markov Models (HMM). HMMs have been used for prognosis in several studies. Here, we fit the training data by a polynomial and then we frame it with 2 other polynomials of the same degree. This results in 3 distinct trained HMMs: the first is used for the RUL estimation and the other 2 provide the lower and higher bounds. Finally, we propose a strategy that can be adopted when dealing with incomplete or imprecise data. This work has been applied to the IEEE 2008 PHM challenge which was a competition aiming at evaluating prognosis methods.