Using AI to evaluate cooling towers structures

The automatised visual analysis tool for inspections of cooling towers (AVIATOR) project is using computer vision to create a platform to assess the condition of the concrete of nuclear power plant cooling towers.

Evaluating the structural conditions of our assets is crucial for ensuring their safety, prolonging their lifetime and planning effective maintenance. Across EDF Group, inspections are performed in a variety of ways for each of our assets. Teams throughout EDF Group have been working to automate our approaches.

Collaboration 

To better coordinate efforts, the AVIATOR project was created as a collaboration between EDF UK R&D Digital Innovation, EDF R&D PRISME and EDF R&D China on behalf of DTG (Division Technique Générale) to bring together expertise from across EDF Group to push our development forward. Together, the AVIATOR project is developing a platform that uses computer vision models to evaluate the structural condition of the concrete in cooling towers of nuclear power plants by identifying cracks and other defects, to aid inspections. 

Benchmarking

 In 2024, the project researched current technologies, models, techniques and public datasets used to assess structural damage in concrete and other civil engineering applications. In 2025, we built upon this research, benchmarking multiple AI models for the platform and have begun developing algorithms to extract accurate dimensions of the defects detected by the AI model. We are currently evaluating the results of this initial development to identify the best models to build upon as we continue the project. 

Next steps 

The AVIATOR platform will continue development over the next couple of years, improving our models and creating additional functionality, such as defect differentiation, automated reports and automated defect dimension calculations before delivering the tool to the DTG team.