A new method for predicting failures of a partially observable system is presented. System deterioration is modeled as a hidden, 3-state continuous time homogeneous Markov process. States 0 and 1, which are not observable, represent good and warning conditions, respectively. Only the failure...

A parameter estimation problem for a condition-based maintenance model is considered. We model a failing system that can be in a healthy or unhealthy operational state, or in a failure state. System deterioration is assumed to follow a hidden, three-state continuous time Markov process. Vector...