Title Feature-based machine learning approach for adhesive joint integrity assessment
Authors Jasiūnienė, Elena ; Samaitis, Vykintas
DOI 10.58286/33309
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] Adhesive Joints ; defect classification ; machine learning ; defect ; ultrasonic testing ; feature analysis
Abstract [eng] Ensuring the integrity of adhesively bonded joints is essential for maintaining structural reliability in aerospace applications. This work presents an approach that integrates ultrasonic pulse-echo testing with machine learning (ML) to enable robust defect detection and bonding-quality classification. Aluminium and composite joints exhibiting a range of bonding conditions—including perfect bonds, interfacial defects such as disbonds and delaminations, as well as weak bonds arising from contamination or improper curing—were inspected using pulse-echo immersion ultrasonic techniques. Following signal preprocessing and time gating, a comprehensive set of ultrasonic features was extracted from both the time and frequency domains. Feature engineering and dimensionality-reduction methods— including tree-based selection, recursive and sequential algorithms, and Linear Discriminant Analysis (LDA)—were applied to optimize model performance. Support Vector Machine (SVM) classifiers demonstrated high accuracy in assessing bonding strength and detecting interfacial defects: over 90% for binary defect detection, up to 99% for weak-bond identification, and up to 97% for defect-depth classification. These results highlight the strong potential of ML-enhanced ultrasonic inspection as an automated, accurate, and efficient solution for assessing the integrity of adhesive joints.
Published Mayen : NDT.net
Type Conference paper
Language English
Publication date 2026
CC license CC license description