Statistical method for the evaluation of railway systems modifications

Abstract

This paper describes how railway production plans daily face random events. Therefore, before setting up modifications for the railway system (changing the network, new signaling system, new timetable, etc.), its robustness to stochastic events must be estimated. Tests cannot actually be carried out, which is why the SNCF uses a traffic simulator developed by its research department. Until now, the timetable robustness has been estimated by considering a few small incidents and their simulated consequences on train delays. The new approach we are working on consists of taking into account the stochastic dimension in the evaluation of robustness. The methodology is divided into three parts that: (1) identify the randomness to be taken into account when evaluating robustness; (2) study the robustness of the railway system to: determine experimental design, carry out simulations, aggregate results; and (3) evaluate the lack of precision in the results. The objective of this paper is to develop the required statistical methods and to design a decision tool following the three steps above. It will help the infrastructure managers to evaluate, in a statistical manner, the robustness of various railway system configurations.