EUVIMEDEUVIMED entdecken
Crossref

Physics based synthetic data generation for nondestructive testing applications

Christophe Reboud, Roberto Miorelli, Anthony Touron, Nicolas Leymarie

e-Journal of Nondestructive Testing · 2026-07 · 31 (7)

DOI: 10.58286/33448 ↗

Kurz gefasst

Fachartikel in der Zeitschrift e-Journal of Nondestructive Testing, 2026-07. Verfasst von Christophe Reboud und 3 weiteren Autor:innen.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

The performance and accuracy of AI models depend heavily on the quality and relevance of the datasets used to train them. The use of poor or biased data can lead to unreliable, if not flawed, results. Unfortunately, for many non-destructive testing applications, the access to relevant and sufficiently large experimental datasets often remains a challenge. This can be due to confidentiality issues between industrial players, lack of standardization of acquisition practices, or simply lack of examples: signatures of critical flaws are fortunately not so often observed. Physics based modelling is a reliable solution for synthetic data generation that provides accurate and quantitative signals, as well as the ability to sample properly over flaws classes, dimensions and locations and to get automatic labelling of the data. A key issue, however, is related to the representativeness of simulated data with respect to real life experimental acquisitions. This communication proposes an innovative fusion technique between simulation and experiments to produce representative synthetic data. Its performance with respect to other strategies like generative AI is discussed and some illustrative results are presented for popular NDT techniques: ultrasonic testing, eddy current testing and radiographic testing.

Zitieren

APA

Reboud, C., Miorelli, R., Touron, A., & Leymarie, N. (2026). Physics based synthetic data generation for nondestructive testing applications. e-Journal of Nondestructive Testing, 31(7). https://doi.org/10.58286/33448

Vancouver

Reboud C, Miorelli R, Touron A, Leymarie N. Physics based synthetic data generation for nondestructive testing applications. e-Journal of Nondestructive Testing. 2026;31(7). doi:10.58286/33448

BibTeX
@article{Reboud2026,
  author = {Reboud, Christophe and Miorelli, Roberto and Touron, Anthony and Leymarie, Nicolas},
  title = {Physics based synthetic data generation for nondestructive testing applications},
  journal = {e-Journal of Nondestructive Testing},
  year = {2026},
  volume = {31},
  number = {7},
  doi = {10.58286/33448}
}
Kostenlos starten →