Optimal condition based maintenance using attribute Bayesian control chart

Condition-based maintenance (CBM) has been emerged as a relatively new trend in maintenance management. Instead of conducting preventive maintenance actions in specified time intervals, the CBM program collects information through condition monitoring, then recommends maintenance actions based on the observed data. On the other hand, Bayesian control charts use the posterior probability of being the system in an unhealthy state as the chart statistic. An attribute Bayesian control chart is employed in this study to monitor a deteriorating system and plan CBM actions based on a continuous-time homogeneous Markov chain. The system consists of three states: healthy, unhealthy, and failure states. A partially observable Markov decision process (POMDP) is developed, which optimally determines the sample size, sampling interval, and warning limit to minimize the long-term expected cost per time unit. Numerical examples and sensitivity analyses are conducted to clarify the performance of the proposed attribute control chart. To the best of the authors’ knowledge, this is the first study of the applications of attribute Bayesian control charts in condition-based maintenance.