Training 3D AI models for PAUT: Performance, Data Needs, and Deployment Considerations
e-Journal of Nondestructive Testing · 2026-07 · 31 (7)
DOI: 10.58286/33438 ↗Kurz gefasst
Fachartikel in der Zeitschrift e-Journal of Nondestructive Testing, 2026-07. Verfasst von Richard Rhéaume.
Abstract (Originalsprache)
This paper examines the training of a three-dimensional (3D) deep learning model for detecting indications in phased-array ultrasonic testing (PAUT) data. We evaluate the use of a 3D model in terms of accuracy, computational cost, and adaptability to real-world data. We analyze the model performance on 3D indications, highlighting how 3D models leverage volumetric context. The study also examines the quantity, diversity, and preprocessing of training data required for robust detection, as well as the trade-offs between recall and precision. Finally, we discuss deployment considerations, including web-based platform requirements and integration into an inspection workflow.
Zitieren
APA
Rhéaume, R. (2026). Training 3D AI models for PAUT: Performance, Data Needs, and Deployment Considerations. e-Journal of Nondestructive Testing, 31(7). https://doi.org/10.58286/33438
Vancouver
Rhéaume R. Training 3D AI models for PAUT: Performance, Data Needs, and Deployment Considerations. e-Journal of Nondestructive Testing. 2026;31(7). doi:10.58286/33438
BibTeX
@article{Rhaume2026,
author = {Rhéaume, Richard},
title = {Training 3D AI models for PAUT: Performance, Data Needs, and Deployment Considerations},
journal = {e-Journal of Nondestructive Testing},
year = {2026},
volume = {31},
number = {7},
doi = {10.58286/33438}
}
EUVIMED