Several challenges will have to be solved to make that vision a reality, according to Puri and other speakers at IMechE’s Automation of Inspection and Maintenance in Railways conference last week (6 November) – but, attendees heard, some smart technologies are already preventing delays and disruption in 2025.
Getting innovation on track
Designed for technology developers, engineers, maintenance professionals and researchers, the event at the Institution’s Westminster headquarters explored the ways that automated and connected systems are spotting and solving issues.
Systems included the Trains with Brains range from Northumberland firm Transmission Dynamics, comprising the AI-enhanced Pandas-V camera system for pantograph and overhead line monitoring, wireless accelerometers, smart oil plugs and small sensors for vibration, temperature and more, all of which can be installed on in-service trains.
By combining hardware expertise with analytics and automated reporting, founder Jarek Rosinski said the company can highlight issues with rolling stock – damage to pantographs from vegetation or bird strikes, for example – and tracks, such as the sway caused by rails shifting before landslides.
“We can transfer a lot of infrastructure monitoring tasks directly to the standard, routinely operating trains,” Rosinski said. “The advantage there is not only that it costs very little to effectively exploit the data… but the advantage is that we are scanning the infrastructure every few minutes or every half an hour.”
Other technology highlighted at the event included Transport for London’s (TfL) automatic track monitoring systems (ATMS), installed on trains on six Underground lines to provide unattended monitoring of track geometry and alignment, and noise and vibration.
AI asset monitoring innovator Cordel, which is working with partners including TfL and Network Rail, highlighted its ‘fit and forget’ Lidar systems. The technology, which can be mounted to track maintenance machines or in-service trains, creates point clouds of assets and the trackside environment, providing 3D mapping, positioning and localisation information in an accessible format.
“After we generate the point cloud itself… we slice it in two-dimensional slices,” said principal engineer Ajtony Farkas. “What this achieves is that we can extract engineering information in easily digestible CSV formats. We have different modules that do that, and this is where machine learning applies as well.”
Once the AI has classified different objects such as tracks, ballast, vegetation or structures, the system can provide data about parameters including structure gauge, cant deficiency and information about level crossings.
Featured trackside technologies included fibreoptic systems from Austria-headquartered Sensonic, which provides distributed acoustic sensing. The company’s sensing units send laser pulses down fibreoptic cables alongside up to 40km of track in both directions, with reflected pulses revealing the time and location of vibrations.
“You can rapidly roll out long-distance monitoring without having to put a load of assets on the trackside,” explained marketing lead Daniel Pyke. The AI-enhanced systems have picked up incidents from potentially “catastrophic” landslides and road vehicles stuck on the track to pedestrian intruders trespassing next to the line.
‘Stop talking the railway down’
While some speakers at the event stressed the challenges of widespread deployment, the new technologies offer significant opportunities to prevent disruption.
Almost half (49%) of delays to rail journeys are caused by infrastructure issues, said Timothy Mangozza, transport innovation lead at PA Consulting, which could be reduced by predictive maintenance.
AI tools also offer a significant financial opportunity, according to a 2024 report by the International Union of Railways, in partnership with McKinsey. “Greater adoption could unlock an estimated $13bn to $22bn in impact a year, globally,” a McKinsey announcement said, representing a 14% uplift in the total value of railway companies.
“There’s a market there, and there is opportunity to do things better,” Mangozza said. “My challenge is: why haven’t we unlocked that value already?”
One reason is that the UK sector is set up for continuous improvement instead of large transformational changes, he said. Measures to fix this could include industry conversations taking a systems-level approach, the adoption of investment portfolio mindsets and a renewed focus on success criteria.
Robots and AI technologies have so far been applied in “pockets”, said Puri, with a complex web of regulations holding deployment back.
The UK industry is “very, very good at implementing specialised things for very specific areas, and we do that incredibly successfully,” he said. “The problem comes when we want to do that elsewhere. Suddenly, all kinds of problems occur – whether that’s capital costs, whether that’s industrial relations, whether that’s issues to do with not having data and the infrastructure to support it, all of those things become challenging.”
He later added: “The railway sector has to give signals to the market. We are not a big market. The AI people, the people building virtual reality or augmented reality, do not care about the rail industry… We need to give signals to the sector that we’re open for innovation. Be ready for it, and we’re ready to support it. And leaders have to stop talking the railway down.”
Data management also emerged as a key issue throughout the day. Lydia Parsons, senior account manager at software specialist One Big Circle, discussed the challenges involved in making its video-based fault identification data available to remote users at partners such as Network Rail.
“Aggregating all of that data, making it accessible in an effective way and a usable way… is a bit of a challenge when we’re in a very ‘boots on ballast’ environment,” she said. “[We’re] trying to build that kind of user-friendly approach that means people aren’t scared to use data. It’s there to support your existing roles and build into your existing systems.”
The sheer volume of data provided by connected systems can cause both unexpected hurdles and opportunities, said TfL engineer Emil Tschepp. “Hardware isn’t normally the biggest limitation. The biggest issue is usually the data and how we use it,” he said.
“You’ll almost always produce or come up with other use cases for data after you start collecting it… a monitoring system will be set up for a particular purpose, and you find out [that] it starts giving you useful information on a completely different area.”
Rich and extensive data from across the capital’s network can help the operator decide where to limit operating speeds, for example, to reduce damage being done to tracks.
Solid foundations
The event gave delegates the opportunity to hear about the latest developments firsthand and to discuss challenges and opportunities with industry peers, said Jason Groombridge, chair of the organising committee.
“It was positive to hear how companies are collaborating to integrate solutions and how real value is being delivered through better asset availability and reliability, reduced maintenance and improved asset knowledge,” the strategic rail consultant said.
“Moving from automated asset condition monitoring to fully automated maintenance and inspection is still some way off, but the topics presented clearly showed the foundations are being laid across the rail industry for more sophisticated automation in the future, with all the benefits in safety and performance that could bring to the industry.”
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.