Data-Driven Approach for Condition-Based Maintenance of Freight Train Wheelsets using Markov Decision Process

This article presents a data-driven model based on the Markov decision process approach applied to freight train wheelsets to provide a way to support a conditionbased maintenance for freight wheelsets.This study analyses observed wear data of freight wheelsets and developed a Markov decision process model.A comparison between key operating variables is also analysed namely the mileage since last maintenance and the gross ton mileage since last maintenance to determine which parameter is more appropriate for developing a two-dimensional state space along with wheel tread diameter.A Markov transition matrices are estimated for various actions, and an optimal strategy is provided, with a decision map for the best actions depending on the current state of the wheelset.