Rolling stock rostering and maintenance scheduling optimization

Abstract

This thesis addresses identification and analysis of frameworks for optimizing medium-term maintenance planning and rolling stock rostering. Rolling Stock Management (RSM) is the main cost factor for Rail Undertakings. For example, for high-speed trains, more than 30% of the lifecycle costs is spent for maintenance operations. In order to reduce the costs due to railway operations, every company should address the joint problem of rolling stock rostering and maintenance scheduling since they are strongly related parts of the same problem. Maintenance optimization can be a key factor to increase the productivity of railway companies. At the same time, in a competitive globalized and multimodal market, RSM is one of the competitiveness key factors because services quality level depends on it. The strategic relevance of RSM, in particular of maintenance scheduling, is thus due to the reduction of needs (such as platforms and human resources) and to the enhancement of quality standards (such as vehicle reliability and cleaning). From our point of view the literature is focused on manufacturing setting in order to reduce the occurrence of a failure while unfortunately the coordination of maintenance and rolling stock scheduling is still underinvestigated. A key problem in railway planning process requires to cover a given set of services and maintenance works with a minimum amount of rolling stock units. Additional objectives are to minimize the number of empty runs and to maximize the kilometres travelled by each train between two maintenance operations of the same type. First,the rostering and maintenance optimization problems are formulated by graph theoretical approaches that involve medium-term maintenance operations, the scheduling tasks related to train services and empty rides. The constraints of the maintenance optimization problem require that the different types of maintenance operations must be carried out for each train periodically. The various maintenance tasks can only be done at a limited number of dedicated sites. Starting from the solutions of the rostering and maintenance optimization problems, we developed another graph theoretical approach to optimize workshop management and in particular to minimize the number of drivers involved and to verify the feasibility of the maintenance plan at each site. For a set of timetables and rolling stock categories, we compare flexible versus rigid plans regarding the number of empty rides and maintenance kilometres. For different feasible frameworks and different kinds of timetables, we provide new mixed-integer linear-programming formulations for train rostering and maintenance scheduling problems and we also show how the proposed scheduling formulations could be used as effective tools to absorb real-time timetable perturbations while respecting the agreed level of service. The specific objective of our research is to related to the following questions: “How can the timetable be executed by an efficient use of resources such that the overall railway company costs are reduced? Which is the maximal improvement that can be achieved? At which cost?". In this thesis, we give an answer to these questions by performing an assessment of key performance indicators. The computational evaluation presents the efficiency of the new solutions compared to the practical solutions. Experimental results on real-world scenarios from Trenitalia show that these integrated approaches can reduce significantly the number of trains and empty rides when compared with the current plan. We use a commercial MIP solver for developing a decision support tool that computes efficient solutions in a short time.