The Rail Safety and Standards Board (RSSB), in collaboration with the University of Sheffield, is developing a tool that uses AI to predict poor track conditions.
In particular, there will be a focus on the seasonal challenge of how to deal with leaves on the line.
The goal of the project is to deliver an online “railhead friction estimation” tool by fall 2023, which will allow users to input data and receive network-related friction predictions. This project will involve the use of AI to analyze data and high-resolution images on friction between the wheel and the rail.
These detailed images will be combined with sideline imagery and topographic/local environmental data.
Low-grip track conditions, such as sheets on the line, cause safety and operational problems, costing the industry around £350m a year.
These delays affect train performance and station overflows. In addition to leaves on the track, temperature and humidity can affect the level of adhesion between the train wheels and the rail.
Project viability
Paul Gray, professional engineering and research lead at RSSB, notes that pattern recognition and AI will provide an estimate of localized railhead sticking conditions.
“Railroad works hard to manage low-stick conditions using a suite of technologies and processes that can generally deal with a variety of stick conditions that can vary throughout the year and are generally more problematic in the fall,” says Gray.
“However, sustained very low adhesion conditions can still occur and can cause trains to be unable to slow down and stop in a controlled manner, which can cause station platform overflows, endangered passing signals and, very occasionally, more serious incidents such as the collision that happened near Salisbury in October 2020.”
“Sustained very low adhesion conditions can lead to serious incidents, such as the collision that occurred near Salisbury in October 2020.”
This project is part of ADHERE, which is the cross-industry adhesion research program that aims to provide research to achieve adhesion conditions that are unaffected by weather and climate.
RSSB is building on its previous research, which found that railhead friction conditions can be estimated using environmental images, railhead images, and other sensor data.
“Having detailed estimates of localized sticking conditions can help improve operational decision-making and near-term rail head treatment planning and could offer a way to reliably compare one route against another to help focus investments. added to mitigate low adherence,” explains Gray.
“This capability would help reduce passenger delays caused by poor adhesion.”
Once the feasibility of this project has been approved, work is underway to extend the new tool to other types of infrastructure. There is also a focus on developing a prototype adherence estimation tool for initial testing.
The future of AI in the railway
Britain’s rail network contains around 20,000 track miles, and artificial intelligence technology can assist in the efficient monitoring and management of this network. This can prevent disruptions, from problems like leaves on the line to passenger and cargo trips.
Management strategies are often based on weather forecasts as well as information on the state of local infrastructure. The information provided by AI technology on this subject is expected to be more accurate.
Gray offers a positive outlook for the future use of AI regarding the prediction of low-grip track conditions.
“Low-adhesion conditions can occur quickly and be highly variable in different locations and from one route to another,” he says.
“Using images that are routinely captured and then using an AI pattern recognition approach offers a realistic way to estimate low-adherence conditions to help prioritize treatments and guide investments to reduce disruption caused due to low adherence.