Ensemble of bootstrapped models for the prediction of the remaining useful life of a creeping turbine blade

• Published in: 2012 IEEE Conference on Prognostics and Health Management pp. 1 - 8
• Main Authors: Baraldi, Piero, Mangili, F., Zio, E.
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01-06-2012
• Subjects: Blades; bootstrap ensemble; Creep; Data models; Degradation; Predictive models; prognostics; Strain; turbine blades; Turbines
• ISBN: 1467303569; 9781467303569
• DOI: 10.1109/ICPHM.2012.6299506

The problem of predicting the Remaining Useful Life (RUL) of a creeping turbine blade is considered in this work. We assume to have a sequence of direct measurements of the blade creep strain; failure is declared with respect to a threshold value of maximum creep strain, beyond which the blade cracks. An ensemble of bootstrapped models is developed to predict the RUL and estimate its variance. This latter quantity combines the stochastic uncertainty in the future evolution of the creep strain with the error made by approximating the actual creep growth process with the empirical model. The diverse models of the ensemble are obtained by considering the independent creep strain increments between consecutive measurements instead of the sequence of correlated creep strain measurements. The performance of the ensemble of models is evaluated by considering the accuracy in the prediction of the blade RUL and in the estimation of the corresponding prediction interval. The results point out the importance of generating diverse bootstrapped models in order to quantify the modeling error.