On-Line Fault Diagnosis and Failure Prognosis Using Particle Filters

This work introduces an on-lineparticle-filtering-based framework for faultdiagnosis and failure prognosis in nonlinearnon-Gaussian systems. This framework considers hybridstate-space models of the system under analysis (withunknown time-varying parameters) andparticle-filtering (PF) algorithms to estimate thecurrent probability density function (pdf) of thestate enabling on-line computation of theconditional fault probability (fault diagnosismodule) and the pdf of the remaining useful life(RUL) in the case of a declared fault condition(failure prognosis module). The proposed methodallows to use the state pdf estimate of the diagnosismodule as initial condition for the prognosis moduleimproving the accuracy of RUL estimates at the earlystages of the fault condition. This frameworkprovides information about precision and accuracy oflong-term predictions RUL expectations and 95%confidence intervals for the condition under study.Ground truth data from a seeded fault test are usedto validate the proposed approach.