Using simulation to reduce the impact of railway network disruptions

Anyone who has ever used public transportation knows how frustrating it is to run late due to a delay or even cancellation of your train, tram, bus or metro train. One disruption can have a huge impact on a major part of the transit schedule, affecting a large number of public transport system users. A very undesirable situation. That is why RET, the operator of the Rotterdam metro network, enlisted the help of TU Eindhoven to predict the impact of infrastructural disruptions on the timetable, and to limit its effects.

To avoid unexpected infrastructural disruptions, maintenance on the network is performed. But doing this too soon is unnecessary and expensive. On the other hand, the financial consequences of unexpected disruptions due to overdue maintenance, are unknown. That was the reason why RET sought a method to evaluate planned and unplanned disruptions in their metro network. Led by Marko Boon, Assistant Professor in the Stochastics section of the Department of Mathematics and Computer Science, Richard Both worked on this challenge during the graduation project of his Master’s degree program.

Modeling disruptions

The method sought by RET resulted in a simulation that models the precise effects of infrastructural disruptions within a network. Think of switch failures, section interruptions — where parts of the trajectory are not given free while they are supposed to be free — and power outages. But the simulation also takes into account bridge openings and passing freight trains.

To model the simulation, a large amount of data was used that was available at RET. For example, the number of metro trains of each type and their specifications, the dwelling time — the time between the arrival and the departure of a train at a station, the full timetable, all logs automatically generated by the traffic management system, and the time needed for passengers to board and alight at each station. Using various mathematical techniques such as graph theory for network modeling and stochastic simulation for the simulation of disruptions, the simulation was built.

Example of a part of a trajectory, displayed in a graph

An important part of the graduation project was the comparison of the modeled simulation with RET’s metro simulator. This simulator is used by RET to train metro drivers and is therefore very suitable for simulating travel times very accurately. However, disruptions cannot be included in RET’s metro simulator, so a comparison of both simulators has been made without the disruptions. The result? Richard Both’s simulation was very accurate. That paved the way for the further use of the simulation.

One of the objectives of the graduation project was to provide a functional graphical user interface. This enables RET to take a look at certain scenarios. For example, what diversion a metro train can take when a part of the trajectory is not passable due to a switch failure as shown in the image, how much time this will cost, and its effects on other trajectories in the metro network. Based on these scenarios, all critical points of the infrastructure of the metro network have been mapped.

Using the simulation in practice

Efficiently managing and controlling a metro network is a very complex process. Using the simulation and the provided user interface, RET can now evaluate this process very accurately, predict the effects of disruptions and minimize their impact. Nowadays, Richard Both’s simulation is used by RET in order to determine travel times of alternative routes, and selecting the most favorable timetable when a disruption occurs. As a result, the effects of planned and unplanned disruptions have the least possible impact on the people using the Rotterdam metro network every day.