5731 2015 2015 eng conferenceobject 0 2015-07-23 -- -- Railway transportation and in particular train timetabling is one of the basic and source application areas of combinatorial optimization and integer programming. We will discuss two well established modeling techniques for the train timetabling problem. In this paper we focus on one major ingredient - the bounding by dual relaxations. We compare two classical dual relaxations of large scale time expanded train timetabling problems - the Lagrangean Dual and Lagrangean Decomposition. We discuss the convergence behavior and show limitations of the Lagrangean Decomposition approach for a configuration based model. We introduce a third dualization approach to overcome those limitations. Finally, we present promising preliminary computational experiments that show that our new approach indeed has superior convergence properties. Proceedings of Conference on Advanced Systems in Public Transport 2015 yes urn:nbn:de:0297-zib-56068 Frank Fischer Thomas Schlechte Thomas Schlechte Mathematical Optimization Schlechte, Thomas KOSMOS MODAL-RailLab Trassenbörse MODAL-Gesamt Applied Optimization