Equipment Degradation Diagnostics and Prognostics Under a Multistate Deterioration Process

Abstract

The increasing level of system complexity in the current competitive market implies that efficient asset management is of paramount importance, particularly for systems with costly downtime and failure. Timely detection of faults and failures through an efficient reliability and health management framework allows for appropriate maintenance actions to be scheduled proactively to avoid catastrophic failures and minimize unnecessary maintenance actions. This thesis employs a general stochastic process - the Nonhomogeneous Continuous-Time Hidden Semi-Markov Process - to model a condition-monitored degradation process with hidden states. This thesis also proposes an unsupervised learning process, which can be used to estimate the characteristic parameters of the degradation and observation processes. It then develops dynamic diagnostic and prognostic measures for online health monitoring. Finally, it introduces a condition-based replacement policy that can be used as an online tool to determine when to replace a degraded device under condition monitoring.