Using AI in nuclear decommissioning

Above: Lyra has shown there is far greater scope for automation in the nuclear industry

The decommissioning of legacy nuclear plants has long presented challenges for the industry. It requires a significant investment of both time and money, with workers having to make multiple trips into contaminated areas wearing air-fed suits. It has been estimated that each suited entry costs £10,000.

Because of these challenges and costs the nuclear industry has been a long-term area of interest for the team at the Manchester Centre for Robotics and AI. This is partly because the team has always felt there is far greater scope for automation in the nuclear industry more widely – unfortunately there aren’t many examples of robots being used at nuclear sites. This position led to the creation of Lyra, a robot created for use in areas that are inaccessible or unsafe for humans to enter.

When starting out, no-one imagined that the work would be among Time magazine’s Best Inventions of 2022. Nonetheless, it has already been a long journey to get to this point, but Lyra is an important first step towards accelerating the pace of decommissioning legacy nuclear facilities.

How it works

Lyra has five radiation detectors measuring gamma, neutron, beta and x-rays. As it is driven along it collects radiometric measurements, from which a full map of the levels of radioactivity along the length of a duct or other environment can be derived. It also has a manipulator which can be used to take swabs at various locations. These can subsequently be sent to the laboratory for analysis.

It is not always possible to use a tether as this would potentially collect contamination, so it was important to make sure Lyra could be controlled wirelessly. However, a fishing wire was connected to the robot which enabled it to be winched back in the event of an emergency. This recovery line was very thin and therefore not considered a contamination problem.

Indeed, a major problem with deploying robots at nuclear sites is that if there is any mobile contamination where the robot is used, the robot itself becomes contaminated. As it is not easy to effectively decontaminate, the likelihood is that the robot will then need to be treated as waste.

This makes it difficult for the nuclear industry to justify the deployment of robots as it can be quite expensive – even more so than visits with air-fed suits. To address this, it was decided to design a low-cost robot that, if necessary, could be sacrificed, but could still perform all the necessary functions to a high standard.

After several years of development, the first iteration of the device was a machine called Vega, which was deployed at the legacy site in Dounreay. While Vega was able to perform the tasks needed effectively, it didn’t have the power required to navigate the unexpected levels of rubble in the duct where it was deployed.

Unleashing Lyra

The team, which included Matthew Nancekievill and Keir Groves, took the learnings from this Dounreay experience to create Lyra. It operates in the same way, but has the extra power needed to handle the challenging terrain. After a test in a mock-up duct at The University of Manchester, it was clear that Lyra would be up to the job in Dounreay.

Lyra was used to survey one of the radiologically contaminated ducts running beneath the central corridor between the plant’s laboratories. It performed the equivalent of more than 400 air-fed suited entries into the site, equal to 2,250 man-hours. This capability reduced costs by an estimated £5m (US$6.2m).

Unfortunately, the first edition of Lyra had to be disposed of due to long-term exposure to contaminated material. It was a shame – the team had become very attached to Lyra who became like family. However, it was also clear that Lyra was only the first step.

Since than it has been possible to develop three further robots that are ready for future deployments. Matthew Nancekievill has also set up a company called Ice Nine Ltd. to invest in further development and make these robots commercially available.

The Manchester Centre for Robotics and AI will continue to work with the team at Dounreay, and know there are countless sites in the UK and worldwide where this kind of technology can make a real difference. As nice as it is to earn recognition from Time, ultimately Lyra’s success will come from a lasting impact on the industry.

Author: Professor Barry Lennox, Co-director of the Manchester Centre for Robotics and AI at The University of Manchester