Imagine trying to see through a foggy window; standard ultrasound often gets 'confused' by the grain of thick metal, making defects appear in the wrong place. This project uses AI to create a map of the metal's internal structure so the ultrasound knows exactly how the sound waves bend. It's like giving the inspection tool a pair of glasses that corrects the distortion, allowing engineers to find cracks exactly where they are.
Standard ultrasonic inspections often misplace defects in thick welds because they assume the material is uniform. This leads to expensive unnecessary repairs or dangerous missed cracks in critical high-pressure infrastructure.
An AI model and a comprehensive weld library that predicts internal material structures to correct ultrasound imaging paths. This includes numerical inversion methodologies and practical demonstrations on manufactured welds.
If you are a plant operator dealing with fatigue cracks in high-pressure components — this project developed an AI-driven imaging process that prevents unnecessary repairs and costly disruptions by accurately locating defects in thick welds.
If you are a manufacturer dealing with complex anisotropic welds in high-temperature environments — this project developed structure-informed imaging that improves the signal-to-noise ratio for more reliable safety certifications.
If you are an offshore firm dealing with the risk of integrity loss in thick-section welds — this project developed a weld library and AI model to predict internal material structures for better in-situ inspection.
Based on available project data, specific pricing or commercial costs are not mentioned; the project was funded with a EUR 1,161,136 EU contribution.
Yes, the project aims to implement imaging and characterization methods in-situ and broaden application across several industrial branches beyond nuclear energy.
The project ambition includes delivering open tools to support the development of these ultrasonic inspection methods.
It provides a comprehensive inspection planning workflow that uses AI-predicted internal material structures to refine inversion and imaging algorithms.
The project period runs from 2022-10-01 to 2026-09-30, suggesting the technology is currently in the development and demonstration phase.
The consortium is well-balanced for technology transfer, consisting of 8 partners across 4 countries. With a 38% industry ratio (3 industrial partners, including 2 SMEs), the project blends academic research from 4 universities and 1 research center with practical industrial application led by EDF.
Contact ELECTRICITE DE FRANCE (EDF) regarding the iWeld project deliverables.
Contact us to find the specific AI-weld library tools developed in this project.
