On the value of information from inspections in condition-based maintenance of degrading units

This paper addresses the problem of investigating the value of information gained via inspections in condition-based maintenance of degrading units. Namely, the maintenance model is constructed by assuming that the degradation process of the units of interest can be described by a gamma process with random effect. It is also assumed that inspections allow to know the exact level of degradation of the units. The study is performed by using a novel aperiodic sequential maintenance policy, where at each inspection a decision is made about whether to immediately replace the unit, to define its future replacement time, or to plan another inspection at a future time. All decisions are made adaptively based on the distribution of the remaining useful life, conditional to all the available measurements. The optimal policy is determined by adopting as performance index the long-run average cost rate. The study is replicated by considering several scenarios characterized by different values of inspection cost, logistic costs, and degradation model parameters. Under each scenario, once determined the optimal policy, the value of inspections is evaluated by: a) computing the probability that the decision to replace the unit will be made by performing only one or two inspections at most and b) evaluating the difference between the values of the long-run average cost rate obtained under the optimal policy, which allows for multiple inspections, and those obtained by using similar policies that allow either for only one or two inspections at most.

Keywords: Condition-based maintenance, Value of Information, random effect, adaptive decision-making, long-run average cost rate.