AI and Machine Vision: A Strategic Asset for Innovation at the Cattenom Nuclear Power Plant

In the Grand Est region, EDF is deploying an innovative solution at its Cattenom power plant: an AI-powered machine vision system designed to assist operators during nuclear fuel reloading. This is a concrete example of technological integration that enhances safety and industrial performance.

The Cattenom nuclear power plant is one of the cornerstones of France’s nuclear fleet. With its four 1,300 MWe reactors, it alone covers 70% of the Grand Est region’s electricity needs, accounting for 8% of EDF’s national production. Every 18 months, one of the units is shut down to replace one-third of its 193 fuel assemblies contained within the reactor core.

                                      “Optimizing nuclear safety and production”

These refueling operations are subject to an extremely strict protocol established by the French Nuclear Safety and Radiation Protection Authority (ASNR). Each fuel assembly, depending on its wear, geometry, and energy potential, must be precisely repositioned within the reactor core. This is a highly precise operation carried out in a confined space.

                                               “There is absolutely no room for error” 

To ensure this precision, operators must visually identify the reference marks engraved on the assemblies—which are submerged nearly six meters deep in the reactor’s high-thermal-flux pool—in order to swap them out with those stored in the storage pool. This task requires sharp eyesight, extreme concentration, and consistency. This is where the Arrea solution developed by Aprex Solutions—a company specializing in machine vision and image analysis—comes in to assist the operators.

The system implemented is based on three main technologies. The first is an automatic character recognition module capable of interpreting reference numbers even under severely degraded visual conditions. The second technology involves an intelligent matching function that automatically compares the read data with the instructions transmitted by the ASNR, while integrating alerts, reminders, and a digital map of the operations performed. The third, finally, involves an inter-assembly measurement tool that verifies repositioning with a tolerance of less than 1 mm, preventing a cumulative error of 7 mm on a line.

“AI doesn’t replace the operator. It assists the operator, who then validates and logs the data.” 

This solution does not replace human workers; rather, it supports them, improves operational reliability, and ensures optimal compliance within a highly secure environment.

MULTIPLE BENEFITS

The benefits are technical, human, and economic. Well-tailored technological support can therefore serve as a strategic lever for EDF. Overall, recharging time has been nearly cut in half. In addition, operational traceability has been enhanced and the radiation exposure of workers reduced.

AN INNOVATION SETTING THE STANDARD ACROSS THE NUCLEAR FLEET

This project won the EDF 2024 DPNT Innovation Challenge in the New Technologies and People’s Choice categories. With its philosophy of partnering with local providers of technical solutions, and through this successful integration of artificial intelligence, EDF demonstrates that innovation and synergy can be combined with the high standards of the nuclear industry. A concrete showcase of AI-assisted maintenance in critical environments. ●

APREX / EDF Cattenom Case Study
Article published in Production & Maintenance magazine
No. 90 July August September 2025 - WNE Special Report