Filters and parameter estimation for a partially observable system subject to random failure with continuous-range observations (Q4664089)

The authors consider a deteriorating system subject to random failure. The state of the system is modelled by a partially observed continuous-time Markov chain with a finite state space. Observations on the system are made at discrete time points and have conditionally normal distributions given the state, except that the state variable representing the failure state of the system (\textit{true} or \textit{false}) is observable. A new measure is introduced such that under it the observation process becomes a sequence of i.i.d. multivariate standard normal variables and two related processes become martingales. This allows for developing a general recursive filter from which filters for the quantities of interest for estimation are derived. The application of the methodology is illustrated with an example from engineering.