DescriptionOperating delays and network propagation are inherent characteristics of railway operations. Train detection systems collect large amounts of data in operation every day and recurring delay patterns can be spotted to improve the timetable design against delay propagation.
We propose multivariate statistic and computational data analysis tools to analyze railway delays from historical records. The trains paths are partitioned through different clustering methods to spot typical delay patterns, following the spatial profiles of absolute delays and changes in delay. The relations between the delay of the clusters and impacting factors, such as rolling stock compositions, time of the day, and of the year, are investigated and reported.
Data from Danish Railway is analyzed, and criticalities in data collection are highlighted. The tools presented can easily be transferred to other countries and other means of transport with sufficient data granularity.
Understanding the delay development and propagation on railway lines allows an improved allocation of time supplements, and results in smaller overall aggregate timetable supplement, reduced transport travel times, and higher productive utilization of train rolling stock. The study results will lead eventually to both better allocation of time supplements in timetable structures, and identification of areas that should be a high priority for correction.
|Period||22 Feb 2017|
|Event title||FOR 2083 meets IPTOP|
|Degree of Recognition||International|