Optimizing Railroad Bridge Networks Management Using Mixed Integer Linear Programming and Genetic Algorithm

Railroad management entities in the U.S. are developing new tools to improve the management of railroad bridge networks, in order to comply with new federal regulations on bridge safety and to increase their profitability. Decisions about maintenance, repair, and replacement (MRR) actions are currently prioritized by rating the bridges based on structural inspections and predictions about the estimated costs of operations. This study proposes a framework for the management of railroad bridge networks that: (1) utilizes a consequence-based management approach that considers relationships between displacements, serviceability levels and bridge MRR decisions; and (2) minimizes the expected value of total network costs by determining the best MRR decisions based on an annual MRR budget. Through this study, two different optimization methods are explored in two different scenarios: (I) mixed integer linear programming (MILP) when the impact of bridge location on costs is insignificant resulting in linear objective function and constraints; and (II) genetic algorithm (GA) when the impact of bridge location on costs is significant resulting in nonlinear objective function and constraints. A case study of a network comprised of 100 railroad bridges is used to demonstrate the proposed framework. The results show that scenario I leads the optimum MRR decision to replace more bridges. On the other hand, scenario II leads the optimum MRR decision to more repair or maintain groups of bridges which are closer to each other.