Artificial Intelligence Enhanced Analysis of Eddy Current Testing and Remote-Field Testing Scans for Heat Exchanger Assessment

Abstract Heat exchanger inspections are critical for ensuring operational safety and efficiency in industries such as power generation, petrochemical, and manufacturing. Non-destructive testing methods, including eddy current testing and remote field testing, enable the detection of corrosion, wear, and structural flaws in heat exchanger tubes. Despite their effectiveness, these methods present the challenges associated with complex signal interpretation requiring significant expertise and can be time-consuming to analyze. This paper introduces two artificial intelligence-based processes that address these limitations by automating key aspects of eddy current and remote-field testing signal analysis. The first process applies advanced algorithms to eddy current testing data to detect structural landmarks, such as tube sheets and support plates, and identifies potential anomalies in non-ferromagnetic tubes. The second process, tailored for remote-field testing signals, focuses on landmark detection in ferromagnetic tubes and incorporates an efficient, automated approach to signal normalization to enhance analysis reliability. Both methods employ advanced deep learning techniques in their training and validation pipelines ensuring robustness to noise and defect variability. Key benefits include increased probability of detection, reduced false detection rates, faster analysis, and minimized reliance on human expertise, thereby addressing critical industry challenges. Field and simulated data validate these results, highlighting the operational benefits of integrating artificial intelligence into heat exchanger inspection. This work represents a significant advancement in nondestructive testing practices, paving the way for safer and more efficient industrial maintenance protocols.