Analysis of Reactive Maintenance Strategies on a Multi-component System Using Dynamic Bayesian Networks

Recently effective planning and management of maintenance activities have gained great importance. Although proactive maintenance approach is preferred against reactive maintenance approach in some highly critical systems, it is inevitable to give up the latter one because of the probabilistic nature of faults. Hence, reactive maintenance is still widely applied. Therefore the aim of this study is to develop an effective reactive maintenance strategy for a dynamic system consisting of four components. Components are hidden and degrading over time. However it is possible to receive partial observations about the condition of the system. One can replace components at any time period. Our aim is to minimize the total number of replacements in a given planning horizon. We propose several approaches within the framework of reactive maintenance for this system and compare their performances by simulating these using Dynamic Bayesian Networks.