09.06.17
New tool uses ‘incontrovertibly accurate’ big data algorithm to avert SPADs
A new software has used big data involving more than a hundred million incidences of trains approaching signals on the UK rail network in order to help operators ensure that the likelihood of trains passing red lights are vastly reduced.
Created by the University of Huddersfield’s Institute of Railway Research (IRR) in partnership with the RSSB, the software – called RAATS, or Red Aspect Approach to Signals – analysed Network Rail’s live feeds to develop an algorithm that provides, for the first time ever, an “incontrovertibly accurate figure” for the number of times on which the signals are at red – meaning the driver must brake.
IRR experts explained that the data is important in itself for understanding the overall likelihood of SPADs and working to prevent them. But the software will also be of huge value to rail companies, which will now be able to focus on individual signals and the number of occasions on which they have been red when approached by a train.
This can therefore have significant implications for driver training on specific routes, timetable planning and other strategic issues, argued Julian Stow, IRR’s assistant director.
Before RAATS, operators used information from driver surveys to determine an overall estimate for the number of times on which trains approached signals at red.
“What our work allows operators to do is to look not only at the overall red approach rate for the system but to query individual signals, looking at factors such as time of the day or week, so you can start exploring how the red approach rate is influenced by how the railway is running,” explained Stow, “and that has the potential for a much better estimation of SPAD risk at individual signals.”
The algorithm behind RAATS, which will be made freely available online to UK rail companies, was established after a year’s worth of data covering around 137 million situations in which trains approached signals.
An initial study of seven signalling areas showed that 3.3% of all approaches are to red signals, but there was a large variation in the approach rates between areas and between individual signals.
SPAD risk assessment of individual signals could thus be “significantly enhanced” by the tool’s ability to estimate red approach rates for individual signals, as explained in a new article co-written by Stow, IRR’s Yunshi Zhao and RSSB’s Chris Harrison.
Prof George Bearfield, director of system safety and health at the RSSB, said: “The risk from signals passed at danger has reduced substantially since the turn of the millennium, so the rail industry is keen to understand where to turn to now, to make the next step change in risk management. The revolution in big data offers us that opportunity to explore risk in more depth, and we’re looking forward to continuing the work with Huddersfield to enhance and then promote the RAATS tool more widely.”
The project is the latest result of the £5m partnership between the IRR and the RSSB, which have worked together to develop a series of innovative tools in the past.