Railway track maintenance is a critical problem for any railway administrator. More precisely, preventive maintenance scheduling is a nondeterministic polynomial time (NP)-hard problem, which additionally involves multiple objectives such as economic cost, maximum capacity, serviceability, safety, and passenger comfort. This paper proposes a multiobjective optimization approach to this problem, combined with a track deterioration model that takes into account the degradation caused by maintenance operations. The track behavior is simulated by an exponential deterioration model based on a two-level segmentation. The maintenance schedule is built using a Pareto-based algorithm with two objectives (cost and delay) and three constraints, on top of an initialization heuristic based on expert knowledge. The proposed approach has been tested with two different algorithms (NSGA-II and AMOSA) over a model of a real track to create schedules for different horizons ranging between 3 and 20 years. The solutions obtained by AMOSA outperform those designed by human experts both in terms of time delay and economic cost, demonstrating the capability of the proposal to produce near-optimal long-term maintenance schedules.
|Number of pages||11|
|Journal||Journal of Computing in Civil Engineering|
|Publication status||Published - 1 May 2018|