Maximum Liklihood Deterministic Particle Filter for State Estimation and Fault Detection in Stochastic Hybrid Systems

In this paper we present a deterministic particle method for estimating the joint continuous/discrete state (x, i), of a class of stochastic hybrid systems, where the discrete state obeys a Markov chain, while noisy measurements of continuous states are taken. We focus on the problem of fault detection. A turbojet engine system example is used to demonstrate this approach, in fault detection and estimation