Capacity and reliability on railway networks: a simulative approach

An efficient train operation is a primary success factor for all infrastructure managers, since it allows operating a higher number of trains without significant infrastructure investments. As known, a trade-off exists between capacity and punctuality, forcing planners to find an equilibrium allowing the highest number of slots to be operated with satisfying punctuality indicators. This is particularly challenging in nodes, where the combination of different stochastic parameters on various lines and for different trains dramatically increases modelling tasks. In the last years, railway simulators have become a very powerful instrument to support the different steps of the planning process: from the layout design to capacity investigations and offer model validations. More recently, the possibility of an automatic import of infrastructure layouts and timetables widened the application spectrum of micro-simulators to large nodes and to more detailed stochastic stability evaluations. Stochastic micro-simulators can reproduce most processes involved in rail traffic and comprehend not only its deterministic aspects, but also human factors. This is particularly relevant in order to simulate traffic under realistic conditions, considering variability at border, various driving styles and stop times. All these parameters have to be calibrated using real-world collected data for single trains or train families, considering their different behaviour in the network and at its border. Since a perfect representation of all stochastic and deterministic parameters involved in rail traffic is not possible, a calibrated model must be validated to evaluate its precision before using it in practice. Calibration has been tested on the Palermo - Punta Raisi single-track line, on the Trieste - Venice double-track line and in the node of Turin. The model is first used to forecast reliability of the operations after infrastructure and timetable changes. Results have been compared ex-post with real traffic data, showing remarkable reliability. An approach is then presented, in which stochastic micro-simulation is used to represent the relationship between robustness, capacity and a number of other important factors, such as traffic variability or running time supplements. The approach can be used to estimate the buffer times, and the running time supplements to obtain a given reliability level. First, micro simulation with its advantages and weaknesses is presented; then, after a presentation of the most common reliability measures, the a new indicator is explained. Third, calibration, validation and application of the case studies is described; in the last part, an approach to evaluate the trade-off between different parameters is presented.