Decision Support for Predictive Maintenance Planning on the Freight Corridor From Rotterdam to Genoa

The central task of any rail infrastructure manager is to maintain the track properly and make it available for operation. To achieve this, maintenance management must meet both safety and economic challenges. The assessment criteria to be applied for network maintenance are defined in regulations and standards. In detail, the considered limit values for the superstructure are defined by the standard EN13848 [1] at European level and by DB Group Guideline (KoRil) 821.2001 [2] at the national level. On this basis, the condition assessment of the superstructure is determined by the measured values obtained during prescribed regular inspections (e.g., of the track geometry).
Due to the nature of the underlying optimization problem, which is characterized by a high complexity due to the individual development of the condition of each 25 m segment, the diversity of the machine types, the variation on the effect of maintenance and the large range of possible train free time windows an exact solution of the resulting linear mixed-integer program would require a high numerical effort and thus an infeasible computation time. Therefore, we present a heuristic solution approach to solve the described optimization problem with sufficient accuracy in a significantly lower runtime. The solution approach is based on a branch and bound algorithm, where the local optimization is limited by the respective machine characteristic.