Probabilistic deep learning for reliable diagnostics using guided wave
e-Journal of Nondestructive Testing · 2026-07 · 31 (7)
DOI: 10.58286/33003 ↗Kurz gefasst
Fachartikel in der Zeitschrift e-Journal of Nondestructive Testing, 2026-07. Verfasst von Jaebeom Lee und 4 weiteren Autor:innen.
Abstract (Originalsprache)
Structural health monitoring (SHM) using guided-wave (GW) techniques has gained significant attention in civil, aerospace, and energy infrastructures due to its ability to inspect large areas with relatively few sensors. With the growing availability of data and computational resources, deep learning (DL) methods have been increasingly adopted for GW-based SHM, showing strong potential in feature extraction and damage characterization. However, conventional deep neural networks (DNNs) provide deterministic predictions even when inputs deviate significantly from the training distribution—a behavior known as the “always-predict” problem. This limitation underscores the need for SHM models that can not only make predictions but also quantify their confidence. Probabilistic deep learning (PDL) offers a promising framework to address this issue by explicitly modeling predictive uncertainty. Despite its rapid advancement in other fields, PDL has been scarcely explored in nondestructive evaluation (NDE) and SHM applications. Moreover, uncertainty quantification and interpretation require domain-specific considerations, particularly for guided-wave data. This study investigates PDL approaches tailored for GW-based SHM and demonstrates their effectiveness through experiments on aluminum plates with hole-type defects, showing improved reliability and interpretability in diagnostics.
Zitieren
APA
Lee, J., Fisher, C., Lee, S., PARK, C., & Calmon, P. (2026). Probabilistic deep learning for reliable diagnostics using guided wave. e-Journal of Nondestructive Testing, 31(7). https://doi.org/10.58286/33003
Vancouver
Lee J, Fisher C, Lee S, PARK C, Calmon P. Probabilistic deep learning for reliable diagnostics using guided wave. e-Journal of Nondestructive Testing. 2026;31(7). doi:10.58286/33003
BibTeX
@article{Lee2026,
author = {Lee, Jaebeom and Fisher, Clément and Lee, Seungjun and PARK, Choon-su and Calmon, Pierre},
title = {Probabilistic deep learning for reliable diagnostics using guided wave},
journal = {e-Journal of Nondestructive Testing},
year = {2026},
volume = {31},
number = {7},
doi = {10.58286/33003}
}
EUVIMED