Abstract:

The partially observable Markov decision processes model is proposed to solve the problem of condition based maintenance in the instance of the degraded system with partially observable information. Firstly, the hidden Markov model is used for state evaluation to get the transition probability and observation probability of the system. Secondly, the proportion hazards model is used for reliability analysis to get the failure rate and reliability function of the system, which considers not only the working time, but also the degradation state of system. Finally, aiming at the rate of the minimum average cost in a long term operation, the best inspection interval and optimal replacement strategy are gotten. A case study indicates that the method can provide a scientific reference of maintenance decision for maintenance personnel.