With over 60 km of tunnels and more than 80 caverns at depths ranging from 50 to 175 metres, CERN’s underground infrastructure is one of the most complex in the world. Most of these tunnels are more than 60 years old, and CERN’s geology of moraines, molasse and limestone requires continuous risk assessment as any movement could disrupt or even halt the operation of the accelerator complex.

During LS2, the Future Studies (FS) section of the Site and Civil Engineering (SCE) department inspected 60 km of underground tunnels and subsurface infrastructure. They found 550 defects, mostly minor. However, of the 8% of faults that were severe, cracks were the most common issue.

Traditionally, an engineer would inspect and manually document cracks and other issues in the tunnels, which is a meticulous and slow process. But a new project being carried out in collaboration with other departments is aimed at finding time-efficient solutions to improve the safety and efficacy of the inspections using new technologies that allow automated analysis, remote inspection and digitalisation.

One of these solutions is the CERN Inspection Tool (TIC – “Collector”), a fully digital, mobile-based app that is fully integrated into CERN’s GIS portal and allows users to record a fault, attach photos, measure distances and locate the fault on a map. After an inspection, all the records are uploaded wirelessly to the GIS servers, where they can be viewed immediately on the “Tunnel Inspection” thematic map.

But most recently, artificial intelligence devices native to CERN – like the CERNbot and the TIM robot – are being used to acquire data and photos from the tunnels. These remotely operated robots, developed by the Controls, Electronics and Mechatronics (BE-CEM) group, take inspection photos that are then processed to identify cracks and automatically locate them. Using photogrammetry and deep learning to analyse CERN's underground infrastructure, a team of experts from the Future Studies section and University College Cork (UCC) has developed a real-time crack and feature recognition algorithm. “This remote collection of photos and data obtained using robots promises to allow more regular inspections and less risk to inspectors, although further testing is needed”, said John Osborne, Future Studies section leader.

Another PhD research project involving the SCE department and the UCC explores the use of fibre optic cables to remotely measure underground movements, allowing the tunnels to be continuously monitored even during accelerator operation.