Optimizing the re-profiling policy regarding metropolitan train wheels based on a semi-Markov decision process

Abstract

In this article, we present a maintenance model for metropolitan train wheels subjected to diameter or flange thickness overruns that includes condition monitoring with periodic inspection. We present a dynamic ( $x_{\theta }$ , $r$ ) policy based on condition monitoring information, where $x_{\theta }$ is the wheel flange thickness threshold that triggers preventive re-profiling and $r$ is the recovery value for the wheel flange thickness after preventive re-profiling. The problem is modelled as a semi-Markov decision process that considers wear in terms of the diameter and flange thickness simultaneously. The problem is formulated in a two-dimensional state space; this space is defined as a combination of the diameter state and the flange thickness state. The model also considers imperfect wheel maintenance. The model’s objective is to minimize the maintenance cost per unit time that is expected in the long run. We apply a policy-iteration algorithm as the computational approach to determine the optimal re-profiling policy and use an example to demonstrate the method’s effectiveness.