Particle Filter-Based Model Fusion for Prognostics

Predictive maintenance is an emerging technology which aims at increasing availability of systems, reducing maintenance cost, and ensuring the safety of systems. There exist two main issues in predictive maintenance. The first challenge is the system operation region definition, detection and modelling; and another one is estimation of the remaining useful life (RUL). To address these issues, this paper proposes a particle filter (PF)-based model fusion approach for estimating RUL by classifying the system states into different operation regions in which a data-driven model is developed to estimate RUL corresponding to each region, and combined with PF-based fusion algorithm. This paper reports the proposed approach along with some preliminary results obtained from a case study.