Robust Selective Maintenance Strategy Under Imperfect Observations

Selective maintenance, as a pervasive maintenance policy in military and industrial environments, aims at choosing a subset of feasible maintenance actions to achieve the success of the subsequent mission under limited maintenance resources. Nevertheless, the existing work on selective maintenance optimization all assumed that components' statuses can be perfectly observed. However, such premise may not be always true in reality. To fill this gap, a new robust selective maintenance model is put forth in this paper by considering the uncertainty produced by imperfect observations. The resulting optimization problem is to maximize the expectation of the probability of a repaired system successfully completing a mission and simultaneously to minimize its variance. The Pareto front offering a set of non-dominated selective maintenance strategies is identified. An illustrative example shows that the selective maintenance strategy with the maximal expected probability of successfully completing the next mission for a repaired system may not be desirable as it usually possesses a huge variance. Several comparative studies are also conducted to examine the effects of the imperfection of observations and maintenance budget on the results. It concludes that the proposed approach can identify robust selective maintenance strategies that can reduce the uncertainty associated with the success of the next mission.