Warning system for major railway disruptions in the works

Utrecht researchers publish in PLOS ONE

The development of the nation-wide railway disruption on 3 February 2012.

Researchers in Utrecht have used a large database of railway data to analyse the main causes of nation-wide railway network disruptions. This delay model will make it possible to calculate at any moment in time the likelihood of such a severe disruption occurring. The researchers earlier this month in the academic journal PLOS ONE.

For their publication, the researchers studied the development of nation-wide disruptions of the Dutch rail network. Mark Dekker, the publication鈥檚 lead author: 鈥淭hese are cases in which delays accumulate and expand, for example because several conductors find themselves stuck in a delayed train. In these cases, delays can spread through the rail network like an oil stain. It doesn鈥檛 happen very often; only around five times per year. But when it does happen, it鈥檚 very serious.鈥

Some methods I鈥檓 currently using for the railway network are also used in climate research.

Delay index

鈥淢y first challenge was to capture the macro-dynamics of the entire rail system in a single diagram鈥, Dekker explains. His analysis resulted in a two-dimensional 鈥榙elay index鈥, which reflects the total amount of delays on every route, as well as the general location of the delay, at a chosen moment in time. This allows Dekker to calculate the likelihood that the current situation will get out of control and cause a nation-wide rail service disruption at a given time. 鈥淔or example, the system can issue an alarm if the probability rises above a certain percentage. At that point, measures may have to be taken to prevent such an 鈥榦il stain鈥 effect from occurring.鈥

The two-dimensional delay index. Each point in the diagram is a snapshot of the total amount of delays at that moment: the closer to zero (bottom middle), the less trains are delayed.

Two main sources

Two railway trajectories appear to be main sources of national rail disruptions. 鈥淎 disruption around Groningen normally has little influence on the total amount of delays in the Netherlands. But the rail connections to our neighbouring countries appear to have much more of an effect.鈥 That means the routes from Amsterdam to Germany via Arnhem, and from Amsterdam to Belgium via Rotterdam. 鈥淎nd the problem can be exacerbated by the interplay between those two routes. The interaction between these two trajectories can have a decisive effect on delays elsewhere in the network.鈥

The effect of railway trajectories on major delays. The brighter the line, the greater effect it has on the total variability of train delays in the Netherlands.

Climate models

Dekker is currently conducting his PhD research at Utrecht 木瓜福利影视鈥檚 Centre for Complex Systems Studies, with PhD supervisors from the fields of Computer Science and Physics. 鈥淏efore this, I did a Master鈥檚 in Climate Physics. That might sound like a strange switch, but some methods I鈥檓 currently using for the railway network are also used in climate research. I made a deliberate choice to study this subject, because I wanted to do something with applications for society at large.鈥

Publication


Mark M. Dekker, Debabrata Panja, Henk A. Dijkstra, Stefan C. Dekker
PLOS ONE, 6 June 2019, DOI 10.1371/journal.pone.0217710
* All authors are affiliated with Utrecht 木瓜福利影视.

This project was financed under an NWO Complexity in Transport and Logistics grant. In December 2016, we published a news article about the project.