Title Detection and staging of hydrogen induced damage in heat exchanger shells using machine learning techniques
Authors Samaitis, Vykintas ; Raišutis, Renaldas ; Asokkumar, Aadhik
DOI 10.58286/33110
Full Text Download
Is Part of e-Journal of nondestructive testing: 14th European conference on non-destructive testing (ECNDT 2026), 15-19 June 2026, Verona, Italy.. Mayen : NDT.net. 2026, vol. 31, spec. iss. 7, p. 1-2.. ISSN 1435-4934
Keywords [eng] Hydrogen induced cracking ; non-destructive testing ; phase coherence imaging ; machine learning ; HIC defect staging
Abstract [eng] This study presents a hybrid artificial intelligence framework for the automated detection and classification of hydrogen-induced cracking (HIC) defetcs in heat exchanger shells. The proposed methodology integrates supervised and unsupervised machine learning techniques to classify HIC evolution into four stages: isolated voids, clusters of voids, mixed void–crack regions, and fully developed cracks. Multi-plane wave ultrasonic imaging enhanced by Circular Coherence Factor (CCF) reconstruction served as the input modality. A dedicated training dataset containing 29,000 HIC cases from heat exchanger shell with naturally occurring damage, labeled by non-destructive testing (NDT) experts, was used for model development. Eighteen statistical and morphological descriptors capturing geometrical, textural, and relational characteristics of defects were extracted and analyzed. Feature selection algorithms were applied to identify the most discriminative features, improving model interpretability and robustness. Multiple classifiers, including ensemble and deep learning architectures, were evaluated for sensitivity, specificity, and generalization capability against unsupervised clustering approaches. Validation on unseen samples demonstrated strong generalization across different HIC stages. The proposed technique aims to assist in the detection of HIC damage, automate defect classification, and serve as a supporting approach for estimating the stage of HIC progression, emphasizing the defect classes that are more related to the advanced damage progression.
Published Mayen : NDT.net
Type Conference paper
Language English
Publication date 2026
CC license CC license description