Optimizing Task Reliability with Predictive Maintenance Under Incomplete Information and Limited Spare Resources

Preventive maintenance actions driven by health information are of great significance in reducing the failure risk during the operation of industrial equipment. However, existing studies have failed to propose effective maintenance action planning methods for addressing the challenges of limited resources and incomplete information received by industrial equipment during the operation. This study proposes a task system predictive maintenance behavior planning model that can simultaneously consider resource constraints and the incomplete acquisition of health information. Unlike previous models, it combines partial health information with resource constraints to guide sequential replacement actions, aiming to maximize task reliability. This model performs dynamic planning for subsequent state monitoring and maintenance behaviors based on belief states and has been proven to have an optimal strategy with a fixed structure and properties. Based on this property, we improved the traditional dynamic programming algorithm, significantly improving the computational efficiency for generating the optimal strategy. Case study on a radar system validate the theoretical approach and demonstrate its effectiveness in improving task reliability under challenging conditions of data scarcity and resource limitations.