Optimization of Wagon Maintenance Mix Allocation in Freight Railways Using Integer Linear Programming

This paper presents an integer linear programming model for multi-period wagon mainte-
nance planning in Brazilian freight railways. The model allocates regional demand among heteroge-
neous maintenance posts while considering wagon type, technical eligibility, effective person-hour
capacity, composite costs, and backlog. An anonymized, reproducible instance comprises 1,004
interventions over twelve months, three operating regions, two gondola types, five maintenance-
scope families, and three candidate posts. A public gross-weight ratio differentiates the maintenance
workloads of the GDT and GDU gondola types, while heavy maintenance scopes (VR4 and VRG)
are restricted to the single technically equipped post. Compared with a local-first heuristic under
the same constraints, the optimized plan reduced the final backlog from 39 to 9 wagons, reduced
the cumulative backlog by 54.4%, and lowered total system cost by 25.4%. The 1,152-variable
instance was solved to optimality within the prescribed gap tolerance in less than one second, and
experiments with up to 11,520 variables were solved in under six seconds. These results indicate
that the model remains practical under realistic monthly capacity disruptions and supports timely
tactical maintenance planning decisions.