Improving railway maintenance planning using a matheuristic approach based on random search and decomposition methods

The quality of railway transportation involves a complex interaction of systems and use of resources. Optimized planning of maintenance activities for assets is crucial for this process, increasing service availability and reducing operational costs. However, the development of enhanced techniques to solve combinatorial problems in the planning of maintenance activities for the rolling stock in passenger railway companies should be more explored. Therefore, a model-based algorithmic approach is developed in a metaheuristic fashion, which can be adapted to other maintenance problems. The novelty of the proposed method of hybrid nature is bridging the gap between purely heuristics and exact methods. Random problem decompositions iteratively find an initial solution to be considered in solving the global problem. The performance of directly applying a commercial solver, which uses branch and bound exact algorithms and other techniques, is compared with the proposed matheuristic approach in a real-world problem. Results show evidence that the matheuristic is a reliable and efficient approach in the experiments tested. The computation time to achieve optimality can significantly be reduced. Time savings are greater in larger and more difficult problems, where feasible solutions are found earlier. By performing the matheuristic approach to support the decision-making in the railway company, costs with maintenance operations can be reduced, asset life cycles can be extended, while labour force and time are maximized.