On-Line Fault Diagnosis and Failure Prognosis Using Particle Filters

This work introduces an on-line particle-filtering-based framework for fault diagnosis and failure prognosis in nonlinear, non-Gaussian systems. This framework considers hybrid state-space models of the system under analysis (with unknown time-varying parameters) and particle-filtering (PF) algorithms to estimate the current probability density function (pdf) of the state, enabling on-line computation of the conditional fault probability (fault diagnosis module) and the pdf of the remaining useful life (RUL) in the case of a declared fault condition (failure prognosis module). The proposed method allows to use the state pdf estimate of the diagnosis module as initial condition for the prognosis module, improving the accuracy of RUL estimates at the early stages of the fault condition. This framework provides information about precision and accuracy of long-term predictions, RUL expectations, and 95% confidence intervals for the condition under study. Ground truth data from a seeded fault test are used to validate the proposed approach.