Performance Data Prognostics Based on Relevance Vector Machine and Particle Filter

Abstract

Prognostics of equipment state is one of the most important and difficult issue in Prognostics and Health Management (PHM) technology. In this paper, Relevance Vector Machine (RVM) and Particle Filter (PF) are used to analyse equipment state and predict its trend. RVM regression is used to extract the trend information of historical measurement data (trajectories). Then RVM based PF method is proposed to figure out the particle sampling distribution using RVM regression, also a new weight calculation method of particle is introduced based on the likelihood sum of whole trajectories. The proposed RVM based PF method can handle the situation that state-transition function of equipment state is unavailable, and it can provide satisfied prognostics of equipment state.