Two new mixed‐integer programming models for the integrated train formation and shipment path optimization problem. Issue 3 (5th December 2022)

• Record Type: Journal Article
• Authors: Arsalani, Pouria; Reisi‐Nafchi, Mohammad; Dardashti, Vahid; Moslehi, Ghasem
• Is Part Of: Networks. Volume 81:Issue 3(2023)
• Page Start: 359
• Page End: 377
• Publication Date: 2022-12-05
• Subjects: mixed‐integer linear programming -- railcar blocking -- railway transportation -- shipment path routing -- single‐block train -- train formation; Network analysis (Planning) -- Periodicals; 658.4032
• DOI: 10.1002/net.22133
• Languages: English
• ISSNs: 0028-3045

Abstract

Abstract: Railcars are known as the heart of the freight rail transportation industry. Hence, any improvements in their operations can lead to sharp reductions in various operating costs. One of the most critical operations on railcars is blocking and routing their transportation. Railway companies face continuous challenged about what blocks should be formed to carry shipments across different origin–destination pairs (O–D pairs) and reclassify them in intermediate yards to minimize transportation reclassification costs. In addition, it is necessary to determine train service between O–D pairs and the number of trains. Along with the shipment routing plans, this problem is called train formation and shipment path optimization (TFSP). In TFSP, some substructure and rail network operational constraints should be considered, including link capacity, classification capacity, the number of sorting tracks, and path length. This paper presents two arc‐based mixed‐integer linear programming (MILP) models to formulate the TFSP problem. To the best of the authors' knowledge, no MILP arc‐based model has been published for the problem that does not need any preprocess before solving. Computational results of solving models on the datasets showed that the first model could obtain a feasible solution with a maximum 0.05% gap up to 48 yards instance. The second model also could find a solution with a small gap compared to the optimal solution in a reasonable time for instances up to 128 yards. Also, the proposed models were compared to the best methods in the literature, and their superiority was shown.