Benders decomposition to accelerate determination of optimal railway intervention programs

An important task of railway asset managers is to develop intervention programs. These interventions need to be developed considering network-level synergies and constraints, in addition to the condition of the assets and their optimal intervention strategies. Considering these concerns may lead to executing interventions earlier or later than specified in asset intervention strategies to reach optimality. Synergies include the fact that the simultaneous execution of more than one intervention only disrupts train movements once. Constraints include budget limits and not closing parallel lines simultaneously. Although many railway asset managers currently determine intervention programs in a rather qualitative iterative fashion, there is an increasing interest in exploiting digitalisation to improve the process. This interest has led to a rise in research focused on the development of mixed-integer linear programs to determine optimal programs more efficiently and effectively. These powerful models, however, still have issues with complicated intervention planning problems, making their use slower than desired. This paper investigates the potential use of Benders decomposition to accelerate the determination of optimal railway intervention programs for 2.2 km of the Irish Rail network. It is found that the optimal intervention program is determined up to 30% faster for the studied example.