Hierarchical HMMs for Autonomous Diagnostics and Prognostics

The ability to effectively control maintenance and support costs under ever increasing global competition not only requires timely identification of impending failures of critical equipment but also calls for accurate estimation of remaining-useful-life (RUL). This is especially important for critical systems with high maintenance and logistics cost such as military systems and nuclear plants. Accurately forecasting the failure time (i.e., prognostics) can lead to failure prevention and diminish the time and cost associated with unnecessary maintenance. This paper presents a prognostic method that relies on hidden Markov models (HMMs) for health-state forecasting. HMMs present an opportunity to estimate unobservable health-states using observable sensor signals. In particular, implementation of HMM based models as dynamic Bayesian networks (DBNs) improves the design and realization of flexible HMM structures (such as auto-regressive HMMs and hierarchical HMMs) for representing complex systems. Hierarchical HMMs are employed here to estimate on-line the health-state of drill-bits as they deteriorate with use on a CNC drilling machine. Hierarchical HMM is composed of sub-HMMs in a hierarchical fashion such that health-states can be represented as distinct nodes at the top of the hierarchy. Monte Carlo simulation, with state transition probabilities derived from a hierarchical HMM, is employed for RUL estimation. Results on health-state and RUL estimation are very promising and are reported in this paper.