Multi-state degradation analysis for a condition monitored device with unobservable states

The degradation process associated with a mechanical device can be represented by a discrete type of multi-state degradation model. Each discrete state in a multi-state degradation model corresponds to a certain health status of the device. In this paper, a general type of multi-state degradation model is considered for a device for which the health states are not directly observable and instead, indirect information is available through condition monitoring. We first demonstrate how nonhomogeneous continuous-time hidden semi-Markov process can be used to model a multi-state degradation process and then briefly review an unsupervised estimation procedure, which can be used to estimate the unknown characteristics of the stochastic processes associated with the degradation process and the observation process of a multi-state device. A simple numerical example is provided to demonstrate the application of the presented model.