EUVIMEDEUVIMED entdecken
Crossref

Segmentation-Model–Based Detection of Hairline Cracks in Forged Steel Components

Steve Simmert, Dominik Braun, Ingmar Jakobi, Feli Schanz, Mathias Hauptvogel

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

DOI: 10.58286/33387 ↗

Kurz gefasst

Fachartikel in der Zeitschrift e-Journal of Nondestructive Testing, 2026-07. Verfasst von Steve Simmert und 4 weiteren Autor:innen.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Reliable detection of hairline cracks and forging laps in safety-relevant forged components such as ball joint housings remains a challenge for automated non-destructive testing. Established methods like magnetic particle testing are sensitive but labour-intensive, operator-dependent, and difficult to integrate into inline processes. This work presents an automated approach combining lock-in thermography with machine-learning-based segmentation. A segmentation network, trained with transfer learning on a dataset of approximately 300 phase images, performs pixel-wise localisation of hairline cracks and forging laps. The model achieves high true positive rates for both defect types and provides spatially resolved indication maps rather than a simple pass/fail result. Evaluation on previously unseen components demonstrates good generalisation, with missed detections primarily linked to defect orientations underrepresented in the training data. The results confirm the feasibility of this approach as an objective, scalable complement to conventional surface inspection methods.

Zitieren

APA

Simmert, S., Braun, D., Jakobi, I., Schanz, F., & Hauptvogel, M. (2026). Segmentation-Model–Based Detection of Hairline Cracks in Forged Steel Components. e-Journal of Nondestructive Testing, 31(7). https://doi.org/10.58286/33387

Vancouver

Simmert S, Braun D, Jakobi I, Schanz F, Hauptvogel M. Segmentation-Model–Based Detection of Hairline Cracks in Forged Steel Components. e-Journal of Nondestructive Testing. 2026;31(7). doi:10.58286/33387

BibTeX
@article{Simmert2026,
  author = {Simmert, Steve and Braun, Dominik and Jakobi, Ingmar and Schanz, Feli and Hauptvogel, Mathias},
  title = {Segmentation-Model–Based Detection of Hairline Cracks in Forged Steel Components},
  journal = {e-Journal of Nondestructive Testing},
  year = {2026},
  volume = {31},
  number = {7},
  doi = {10.58286/33387}
}
Kostenlos starten →