Reliability of railway operations becomes more and more demanding with increasing train traffic, which asks for stable and robust timetables capable of neutralizing deviations from scheduled time-distance paths and stabilizing delay propagation. Timetable performance evaluation is thus a crucial aspect in the railway timetable design process to guarantee and maintain reliability of operations. Feedback from realized railway operations is essential to evaluate the performance of the timetable in practice and to find and improve structural shortcomings in the timetable design. This thesis presents the developed software TNV-Prepare which recovers the infrastructure utilization of train traffic from train detection data of the safety and signalling systems based on daily records of the train describer systems, the so-called TNV-logfiles. A case-study at the Dutch railway station Eindhoven demonstrates the potential of a statistical analysis of this train traffic data to identify structural sources of delays. Railway operations are characterized by many network interdependencies resulting from the timetable and shared usage of railway infrastructure. Therefore, a railway timetable must be carefully tested on stability and robustness before implementation using a mathematical model of the scheduled railway operations. This thesis proposes an analytical model for evaluating timetable stability and robustness based on max-plus algebra. The max-plus linear system approach gives a formal stability test based on a max-plus eigenvalue problem, identifies critical events and processes, quantifies robustness in terms of recovery times, and computes the propagation of delays over time and space. The method has been implemented in the software PETER (Performance Evaluation of Timed Events in Railways) which enables users to analyse large-scale network timetables in real-time. A case-study of the Dutch national railway timetable illustrates the developed methodology.