Optimization of replacement times using imperfect monitoring information

We propose a state-based PM policy based on a stopping rule for an imperfectly monitored two-unit parallel system consisting of s-dependent units. The observed failure rate of the system is proposed as an efficient tool to integrate the imperfect monitoring information in the maintenance decision process : the possible nondetection of the failure of a unit and the monitoring quality are explicitly taken into account when optimizing the maintenance decisions. Using classical martingale results, a stochastic cost model is developed to assess the performance of the monitoring-maintenance policy for this system. Numerical experiments show that the PM policy based on the observed failure rate of the system is more robust to defective monitoring information and that it performs better than a policy which discards completely this imperfect information.