Precise and accurate train run data: Approximation of actual arrival and departure times

On the most intensively used parts of the Danish railway network, registration of arrivals and departure times are based on occupation of main track circuits and block sections. These measurements are precise. However, due to the nature of track circuits, they do not register the actual time when trains have come to a halt nor when trains have set in motion again.

Thus the measurements are inaccurate and do not express the experience of the passengers. A commonly accepted method to make this measurement possible is to construct a correction function to the track circuit based measurement. This function estimates the inaccuracy or bias of the measurement and thus which offset is needed to approximate the actual arrival and departure times.

The development of such a function is described in this paper. The development is based on international best practices combined with what is technically possible with the present systems. GPS data from a major Danish Railway Undertaking is used as an alternate data source with more accurate arrival and departure times.

The offset is based on the median of the time difference between these two sources. Factors taken into consideration when constructing the correction function, are location, message type, platform used and train type. The approximated correction values are then analysed to ensure that interquartile range is within the defined criteria.

The practical implementation is an additional column in the train run history database tables with the approximated actual arrival and departure times. As a result, all future statistics can now either be based on track circuit data with high precision or approximated actual arrival times with a high accuracy.

Consequently, performance analysis will be more accurate, punctuality statistics more correct, KPI more acceptable, and bonus / malus calculations more justifiable.