A Multi-Level Selective Maintenance Strategy Combined to Data Mining Approach for Multi-Component System Subject to Propagated Failures

In several industrial fields like air transport, energy industry and military domain, maintenance actions are carried out during downtimes in order to maintain the reliability and availability of production system. In such a circumstance, selective maintenance strategy is considered the reliable solution for selecting the faulty components to achieve the next mission without stopping. In this paper, a novel multi-level decision making approach based on data mining techniques is investigated to determine an optimal selective maintenance scheduling. At the first-level, the age acceleration factor and its impact on the component nominal age are used to establish the local failures. This first decision making employed K-means clustering algorithm that exploited the historical maintenance actions. Based on the first-level intervention plan, the remaining-levels identify the stochastic dependence among components by relying upon Apriori association rules algorithm, which allows to discover of the failure occurrence order. In addition, at each decision making level, an optimization model combined to a set of exclusion rules are called to supply the optimal selective maintenance plan within a reasonable time, minimizing the total maintenance cost under a required reliability threshold. To illustrate the robustness of the proposed strategy, numerical examples and a FMS real study case have been solved.