The project encompasses two goals: assessing the reasons for passengers choosing to travel by train, and upgrading railway maintenance infrastructure. While the University of Leeds is more heavily involved in the former objective, Huddersfield’s Institute of Railway Research, which is based at the University of Huddersfield, is playing a ‘central role’ in the latter, according to the university.
“The conventional approach to rolling stock maintenance is to carry out scheduled (interval-based) maintenance, which includes visual inspection to check the condition of components,” said the university.
Yet this approach requires significant human effort to function, as individuals are required to visit sites, submit reports, and then read submitted reports; and the scale is growing. Peter Hughes, of Rail Technology Magazine, commented that in 2014, there were 180 reports logged per day, compared to 650 now.
To ensure all reports are responded to, SMaRTE aims to unite human experience and AI software.
“A safety analyst interacts with the computer and teaches it the words and phrases that are meaningful to safety management,” said Hughes. “The computer learns from the analyst’s actions and updates its search results based on what the analyst has told it.”
Called ‘smart maintenance’ by the University of Huddersfield, the program will reduce costs, improve safety and improve vehicle availability, as rail vehicles will spend less time out of service, effectively increasing the capacity of the rail service.
The software has been trialled on accident reports from railways in Switzerland; these reports were written in German, French and Italian, and were understood with 98% accuracy by a computer. A simplified version of the software has been applied to Network Rail, and a ‘full implementation’ will be underway in the Rail Safety and Standards board by the end of 2018.
SMaRTE is a two-year project that is funded as part of the EU’s Horizon 2020 scheme, an initiative to invest €77 bn into research and innovation projects between 2014 and 2020.