This task increasingly involves use of AI tools, with several trials rolled out recently to assess and counter risks to journeys and passengers. These include Radar (Rail Anomaly Detection And Reaction), one of several rail-focused projects to receive a portion of £32m from UK Research and Innovation (UKRI) earlier this month.
Led by JR Dynamics, Network Rail, First Trenitalia West Coast Rail, Angel Trains and Komodo Digital, the work is focused on safety issues and disruption caused by faults between electric pantographs and the overhead lines they draw power from.
AI and machine learning techniques will be used in a ‘self-learning, automated’ anomaly detection system, enhancing the existing Pandas-V pantograph and overhead line monitoring solution, which combines high-definition video footage with accelerometer and GPS data to pinpoint defects.
With up to 50 trains per hour passing every anomaly and exacerbating them with each passage, Radar aims to quickly identify and mitigate risks, preventing further damage and disruption to the network. The project will also develop an AI-driven interface for workers, adapting to user requirements, engineering knowledge levels and roles within the supply chain to deliver precise and understandable information.
Oxford firm Purple Transform also received a share of the £32m for its Levelling Up Crossings project, which will develop and trial AI models to monitor level crossings and reduce the number of safety incidents. The company’s ‘intelligent platform’ SiYtE will scan CCTV footage in real-time, alerting relevant workers if it detects dangerous or unexpected behaviour such as objects on the track or trespassing pedestrians.
“Level crossings are a major safety concern for the entire rail industry; not only do they represent a serious risk to the public, they are often the root cause of costly and frustrating service delays,” said Gregory Butler, CEO of Purple Transform.
“As it’s simply not feasible for rail employees to monitor all of these crossings all of the time, radical new solutions are required. That’s where AI comes in.
“Our Levelling Up Crossings project will help identify safety incidents before they develop, empowering relevant personnel to rectify the situation, keeping the railway moving and people out of harm’s way.”
Other projects, such as InteGraph from DZP Technologies, KS Composites, Network Rail, Nprime and the University of Sheffield, are focused on structural health monitoring of infrastructure and engineering structures. The scheme will combine graphene sensors systems with AI and machine learning to improve the collection, processing and analysis of large amounts of data.
Living assets
The recent UKRI-funded projects are far from the only ones using AI for rail safety. Last month, Hitachi Rail announced a 12-month collaboration with Network Rail, LNER and digital supplier CrossTech to automate detection of hazards including overhanging trees, leaves on the track and embankment subsidence.
The system uses a forward-facing CCTV camera installed inside the driver's cabin of a LNER Azuma train operating on the East Coast Main Line. The AI camera sensor technology monitors areas in real-time to pinpoint where maintenance work is needed.
“Vegetation is the only living asset on the railway network, and as such, understanding the potential risk to trains is ever changing,” said Johanna Priestley, route engineer at Network Rail.
“Using forward-facing footage allows us to 'see' from the driver's perspective. We can use this technology to understand where vegetation is encroaching on the operational railway and at risk of making contact with either trains or fixed infrastructure, such as overhead electrified wires. We can also identify where vegetation growth has compromised the driver's view, such as on the approach to signals or level crossings.”
The approach should make passengers' journeys more reliable and help minimise the risk of disruption on the network, she added.
Network Rail previously estimated that vegetation-related incidents cost up to £3m annually in the Southern region alone. A previous trial in that area helped Network Rail reduce the number of incidents involving trees blocking the line and trees being struck by a train. The trials also reduced the number of times foliage obscured signals and the number of ‘vegetation entanglements’ with train pantographs.
Want the best engineering stories delivered straight to your inbox? The Professional Engineering newsletter gives you vital updates on the most cutting-edge engineering and exciting new job opportunities. To sign up, click here.
Content published by Professional Engineering does not necessarily represent the views of the Institution of Mechanical Engineers.