Materials testing is usually done in the laboratory. This involves using small-to-medium-sized samples of a material and testing them using apparatus designed for an indoor lab space.
But what if what you want to study is much larger? What if you wanted to examine surface damage through an entire railway tunnel, for example?
This question is exactly the issue that Jon Hughes, senior scientist at the National Physical Laboratory (NPL), has been working to solve. An expert with more than 15 years’ experience in materials testing, Hughes is now focused on bringing non-contact materials testing techniques out of the lab, using them to examine large-scale structures such as railway tunnels, bridges and nuclear infrastructure.
At the Advanced Materials Show 2025, Hughes presented his current work on adapting Digital Image Correlation (DIC) from a laboratory technique into something that can be used in much more challenging environments.
What is Digital Image Correlation?
Digital Image Correlation (DIC) is a non-contact optical technique that can be used to measure strain and displacement in a material. By comparing high-quality digital photographs of a material at different points in time and tracking blocks of notable pixels, a DIC system can measure any surface displacement.
“Anything that has got a decent texture on its surface, you can apply a DIC-type technique to, because you have some feature there that you can track,” Hughes told Technology Networks.
Depending on its use case, these displacements can either be flagged for further visual review or the system can use the surface displacement data to build up 2D and 3D deformation vector fields and strain maps. The DIC technique is also not exclusively limited to high-quality photography – it can also be used to analyze high-speed video, microscopy images and surface roughness maps.
“DIC is a technique that we have in the laboratory, and over time, as computing power has increased, we now have the capability to expand it up to be applied onto larger and larger items,” Hughes said. “My colleague Dr Nick McCormick really pioneered using DIC outside of the laboratory on other stuff. He started on testing concrete, but things have really blown out of proportion over the last decade and now we are able to have projects like this, where we’ve been starting to scan entire railway tunnels for Network Rail. It is really pushing the science boundary forward.”
Out of the lab, onto the railway
For this project, Hughes and his team have built a rig that can be run on standard railway tracks, featuring a high-intensity light box and a high-resolution dual linescan camera mounted on the end of a long boom. The rig is wheeled along the length of the railway tunnel, repositioning the boom after each run,collecting long image stripes which are stitched together to form a large panoramic image of the inside of the tunnel. Using DIC on these images, the team can identify areas of interest in the tunnel.
“Primarily, we are looking for changes over time. Ideally, you could take one scan, then come back a year or so later and take a second scan, and then you can apply DIC to compare these two images. It will highlight any areas of low correlation, which means that there has been a change compared to how this area looked before,” Hughes explained.
“That might be something as simple as some soot that has come down, but it could also be where brickwork has started to degrade, or where mortar is falling out. It could mean spalling is occurring, which is where the brick face is flaking and coming off. It could be water ingress or a crack or an area where the bricks have been changed during maintenance.”
Thanks to a slight overlap in the dual linescan camera, the team is also able to triangulate and measure height – meaning that any areas where bricks or mortar are sticking out from the wall can also be flagged.
An image generated from scans of a railway tunnel wall. The image shows irregularities in some of the brickwork, which may require a follow-up investigation from a trained tunnel inspector. Credit: Bethan Nye / NPL.
“These giant images just capture an awful lot of data that you can’t really see when you’re inside the tunnel; the tunnel is pitch black, so all an inspector can see is what is lit up by their head torch,” Hughes said.
“The added advantage of these huge panoramic images is that you can see everything in incredible detail on your screen or on a giant print-out on the floor. Essentially, the tunnel inspector doesn’t need to be in the tunnel to carry out most of their work anymore – they can just look at these images and see everything that they need to see without having to be in a hazardous environment.”
The future of large-scale engineering inspection
While the NPL camera rig can help inspectors to get a better look at the tunnel’s interior surface, the research team is not stopping there.
“There is another layer to it that we added earlier in the year,” Hughes said. “We invited a company that does very large ground-penetrating radar systems to take a scan of the tunnel at the same time we were doing it, and that has given us lots of sub-surface data about what is going on through the brickwork and in the underlying geology as well.”
The curved tunnel wall image with ground-penetrating radar scan data inserted at different height levels. The data reveals an area of irregularity that may indicate where the top layer of brick has become debonded. Credit: Bethan Nye / NPL.
The team has already begun to apply their DIC-based technology to other use cases. The High Accuracy Inspection System (HAIS) is another NPL project, designed for inspecting nuclear waste stores at nuclear sites. The HAIS operates using a camera arm that can be lowered up to 16 meters deep into a nuclear waste store inspection port. Using NPL’s DIC technology, the HAIS can detect very small changes in the waste store environment – such as corrosion or structural movement– which can be noted and used to help infer future waste behavior that may need addressing.
Challenges in large-scale structure assessment
While this technology could be transformative in terms of improving how large-scale structures and engineering projects are monitored and serviced, there are still a number of important challenges that must be overcome in order to make these solutions more accessible.
Chief among these is data management, Hughes explains: “We generate images that are nearly 500 gigapixels in size – that is roughly 10–25 terabytes. They are massive files, so one of our big challenges is the size of the data and how we handle that.
“It’s a case of working out how you process this, how you move it around, how you store it and back it up without corrupting anything – because massive data sets can go wrong very quickly,” he added.
Working with these data on NPL’s internal computer systems is one thing. But for future applications, the data needs to be able to be sent out to clients for their review. This presents another set of challenges for the team.
“There are bottlenecks in terms of how you get the data in and out of storage to get processed, and that’s just with us working in-house. Getting that to a client is a whole other challenge because these files are so large you simply cannot upload them to the cloud,” Hughes said. “The challenge is going to be in working out how we can transfer data over to somebody like Network Rail so that they can analyze it themselves, because I think clients would ultimately prefer to be able to do it themselves.”
“At the moment, the easiest way for us to do this is to use large numbers of hard disks and physically take them over to the customer. That is the most reliable way of doing things at the moment, and it is quite old school, but even large hard disks can be quite unreliable – they have their own limits of what they can do, so we always have multiple copies of everything just in case,” Hughes said.
Despite these current obstacles, Hughes is confident that this technology will develop over the next 5–10 years to find use in even more industry sectors, while the NPL scientists continue to refine the analytical and data-related systems behind the technology.
“I don’t think you could fully automate this technology right now – there just aren’t readily available data sets out there that could be used to teach it – but you could certainly semi-automate it and make the feature recognition system more advanced. Obviously, you don’t want to go full AI with an unreliable data set on something that is going to be making critical safety decisions,” Hughes commented.
“Looking to the future with us at NPL, for the National Rail project, I’d like to see it so that they can do the data collection on one of their service trains with a system built by a big contractor that has lots of experience in implementing machinery directly onto trains, then we can work more on the data site of things,” he continued.
“We have already demonstrated that you can add in sub-surface data to this system, which gives you very high confidence in what you are looking at. And we should be able to transfer this over to any of the other projects that we are working on. As an organization, we’re hoping to be able to offer this to other sectors now that we’re very confident in the data that we have.”