Conference article

Simulation is one of the powerful means within the toolset of railway operations research. In contrast to timetabling and to queuing theory, it supports a precise representation of interdependencies and has thus a large field of application. Since in today’s railway operation many timetable concepts and even big investment-decisions are based on studies conducted with simulation tools, a focus should be set to the sound evaluation of simulation results, too. Nevertheless, the aggregation, validation and interpretation of simulation (raw) data can barely be found in literature. This fundamental task is subject of this paper. A simulation consists of the following steps: model design, parametrisation and calibration, simulation, processing of raw data, interpretation and visualisation of results. First, various input parameters are manipulated and simulation results are manually evaluated in a simple closed-loop principle. As each simulation is subject to outliers, runs affected by dubious conflict solutions have to be identified and excluded automatically. In most cases, a special focus is on the comparison of different scenarios and the necessity of establishing comparability by forming intersections between the simulation runs. The remaining subset of simulation runs per scenario can be considered (statistically) representative, as soon as the key figure of each scenario series converges. Finally, the raw data can be processed for the evaluation of simulation results. Results of simulations are mostly complex but by producing results for different target groups the complexity has to be reduced without losing important details or provoking misinterpretation. For this reason, it is necessary to choose key figures which comprehensively represent the simulation results.

simulation, evaluation, calibration, intersectioning, interpretation