Rao-Blackwellized particle filtering for fault detection and diagnosis

This paper deals with the problem of fault detection and diagnosis of multiple failures in a nonlinear dynamic system. By modeling the multiple failures as different models of jump Markov nonlinear systems, a Rao-Blackwellized particle filter is developed by combining with the unscented transform technique, in which the posterior model probability is used as an indicator of a failure occurrence. Two numerical examples are provided to illustrate the effectiveness of the proposed method, and simulation results show that the fault can be detected quickly and reliably.