Adaptive mission risk control under incomplete health information and resource limitation: A constrained multi-state predictive maintenance model

Abstract

Information-empowered online predictive maintenance (PdM) is essential to mitigating unplanned failure risks of safety-critical industrial equipment during mission executions, whose effectiveness, however, is increasingly challenged by data inadequacy and resource limitation. This study investigates an innovative predictive maintenance model for multi-state mission-oriented systems under limited maintenance resources, where the system's health evolution is only partially revealed through collected monitoring data. As opposed to previous studies, we synthesize incomplete system health information and resource reservation conditions to inform sequential replacement actions under resource constraints, so as to maximizing system mission reliability. In particular, we establish an adaptive belief-state-based maintenance decision model based on belief states, and delve into a series of structural properties with respect to the model. The optimization problem of interest is shown to constitute a dynamic control limit structure that substantially improves decision robustness; by exploiting this structure, we present an efficient heuristic algorithm to alleviate computational burden in acquiring optimal maintenance solutions. Numerical experiments conducted on radar driver demonstrate the theoretical feasibility and practical implications of our approach.