Title :
Multiple model moving horizon estimation approach to prognostics in coupled systems
Author :
Pattipati, Bharath ; Sankavaram, Chaitanya ; Pattipati, Krishna ; Zhang, Yilu ; Howell, Mark ; Salman, Mutasim
Author_Institution :
Dept. of Electr. & Comput. Eng., Univ. of Connecticut, Storrs, CT, USA
Abstract :
The key objectives of this paper are to analyze and implement a novel moving horizon model predictive estimation scheme based on constrained nonlinear optimization techniques for inferring the survival functions and residual useful life (RUL) of components in coupled systems. The approach employs a data-driven prognostics framework that combines failure time data, static and dynamic (time-series) parametric data, and the Multiple Model Moving Horizon Estimation (MM-MHE) algorithm for predicting the survival functions of components based on their usage profiles. Validation of the approach has been provided based on data from an electronic throttle control (ETC) system. The proposed prognostic approach is modular and has the potential to be applicable to a wide variety of systems, ranging from automobiles to aerospace.
Keywords :
automotive electronics; failure analysis; life testing; nonlinear programming; remaining life assessment; time series; ETC system; MM-MHE algorithm; constrained nonlinear optimization technique; coupled system; data-driven prognostics framework; dynamic parametric data; electronic throttle control system; failure time data; moving horizon model predictive estimation scheme; multiple model moving horizon estimation; residual useful life prediction; static parametric data; survival function prediction; time series data; Computational modeling; Data models; Estimation; Hazards; Hidden Markov models; Mathematical model; Predictive models; electronic throttle control (ETC) system; model predictive control (MPC); multiple model moving horizon estimation (MM-MHE); nonlinear programming; proportional hazard model (PHM); survival function estimation;
Conference_Titel :
AUTOTESTCON, 2011 IEEE
Conference_Location :
Baltimore, MD
Print_ISBN :
978-1-4244-9362-3
DOI :
10.1109/AUTEST.2011.6058750
