[2D4] Using machine learning tools to enhance signal analysis efficiency in heat exchanger tube inspection
É Provencal, V Shahsavari, S Loffredo and D Aubé
Eddyfi Technologies, Canada
The non-destructive testing (NDT) industry is facing a shortage of skilled workers, particularly in heat exchanger tube inspection, where there is a high demand for experts who can analyse collected data. As new technologies emerge and the industry shifts to array probe configurations, task complexity and data volume continue to increase. This highlights the need for tools that simplify data analysis, improve efficiency and provide reliable results. Although there are a few assisted analysis tools for tube inspection, most of the solutions available are adapted to eddy current testing (ECT) for steam generator tube inspection. Traditional detection methods, relying on phase angle, amplitude thresholds and probe positioning, become limited under less-than-ideal conditions, such as uneven pulling speed, incomplete scans or missing encoder data. Additionally, incomplete tube bundle information, such as missing landmark tables or re-tube section details, complicates analysis. These challenges are especially common outside the nuclear sector and can affect data reliability. To address these issues, Eddyfi Technologies has developed an artificial intelligence (AI)-powered assisted analysis tool. This tool enhances several stages of the tube inspection process, such as landmark and defect detection and localisation, without needing detailed prior knowledge of tube configuration.
