Comparative Analysis on Optimal Maintenance Policies for Markovian Deteriorating Systems under Uncertain Repair

摘要

We consider a system whose deterioration follows a discrete state and discrete time Markov chain. When the system is put into practice, we may select wait (operation), repair, or replacement. Since the system deteriorates under operation and the result of repair is uncertain, the true state cannot be identified after completion of operation or repair. On the other hand, replacement always returns the system to an initial new state. For this system, we construct two partially observable Markov decision process models. The first model assumes that the system is not monitored after any action. The second model assumes that the system is monitored after operation or repair to obtain some incomplete information on the degree of the deterioration. For these models, we derive sufficient conditions that the optimal maintenance policy satisfies the monotone property. Then, we compare the sufficient conditions and explore the relationship between the optimal maintenance policies and the sufficient conditions.