Unsupervised Learning for Opportunistic Maintenance Optimization in Flexible Manufacturing Systems

Abstract

Large scale manufacturing systems with a high degree in automation and the ability to produce several product variants in parallel meet current requirements of a highly flexible and at the same time productive manufacturing process. In practice, however, the non-transparency as well as the complexity of these systems overwhelm the maintenance department in the effective planning and implementation of maintenance tasks. As a result, major maintenance tasks are postponed to non-production times which causes increased maintenance cost as well as a decrease in system availability. This research explores a method that uses unsupervised learning algorithms to analyze type mixes and related process performances inside the system. The information is used to determine the optimal master production schedule prior to maintenance activities which leads to more frequent and extended time windows for maintenance activities during production time and thus to an increase in system availability.