The Smith Institute worked with RSSB to undertake a data-driven investigation into train driving variability in different adhesion conditions.

Adhesion at the wheel-rail interface can be compromised by moisture, grease, diesel fuel and leaf fall. Drawing on anecdotal evidence, RSSB hypothesised that driving under low adhesion conditions leads to greater variability in drivers taking power, train speed and braking applications than when driving under good adhesion conditions. If evidence is found to support this hypothesis, then there may be an opportunity for the rail industry to seek more consistency in driving, and ultimately to improve performance and safety.

During autumn 2018, On Train Data Recorder (OTDR) data was collected by West Midlands Trains from trains operating on the Birmingham Cross-City Line. This OTDR data, in combination with timetable and infrastructure data, allowed RSSB to begin to explore quantitatively, driving variability under different adhesion conditions. The Smith Institute undertook the creation of a database and the statistical analysis required to test RSSB’s hypothesis.

We extracted the data, grouped it into comparable journey types and applied filters to remove repeated, erroneous or anomalous data. The data received included over 100 million rows of OTDR data for over 10,000 journeys. Metrics were created to summarise driving behaviour, examples of which were:

  • average speed between adjacent stations,
  • initial power level upon departure from a station,
  • time before brake is applied for approach to a station.

Data for comparable journeys was split into ‘normal’ and ‘compromised’ adhesion groups to allow a comparison of the variability of a given driving metric in each group using a statistical test. Finally, where variability in a metric was found to be greater in the presence of low adhesion conditions, we attempted to find statistically significant influences by exploring the relationship with variables such as time of day, weather and lateness.

One of the key project findings is that location-dependent evidence of greater driving variability in low adhesion conditions was found on the Cross-City Line. Our analysis gives meaningful insight to the rail industry about which aspects of driving behaviour may be most affected and will enable RSSB to plan future research programmes focused on improving performance and safety in compromised adhesion conditions.

The full research report can be accessed through the RSSB’s repository, Spark.