Accelerated degradation test and particle filter based remaining useful life prediction

This paper presents a particle filtering based long-term RUL prediction method that integrates two data resources: infield Accelerated Degradation testing (ADT) and field operation. This method improves the usage of historical information and makes accurate residual life prediction compared with conventional regression method. ADT data is used as prior information to establish dynamic system model by stochastic degradation process modelling, and more specifically drift Brownian motion. Then particle filtering is introduced to estimate system state or forecast residual life. Two stages are included which are on-line filtering and off-line prediction. The proposed method is validated through experimental data and operational data of Super Luminescent Diode.