EUVIMEDEUVIMED entdecken
Crossref

Anomaly Detection Reliability in Consumer Electronics Manufacturing

Chuen-Wun Pai, Hung-Wei Hsueh, Shu-Han Hsu

IEEE Transactions on Consumer Electronics · 2025-11 · 71 (4) · 9287-9300

DOI: 10.1109/tce.2025.3626738 ↗

Kurz gefasst

Fachartikel in der Zeitschrift IEEE Transactions on Consumer Electronics, 2025-11. Verfasst von Chuen-Wun Pai und 2 weiteren Autor:innen. Die Arbeit wurde bisher 2-mal zitiert (Crossref) und nennt 32 Literaturangaben. Gefördert von: National Science and Technology Council, Taiwan, Google Research Award.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

In consumer electronics manufacturing, deep neural networks (DNNs) are increasingly employed for anomaly detection, yet current research predominantly emphasizes model accuracy at the expense of reliability, posing significant costs and potential safety issues. Despite the critical importance of reliability, existing research offers limited insight into DNN reliability for anomaly detection and practical methods for enhancement. To address these gaps, this study provides an in-depth analysis of the reliability of autoencoders, a representative anomaly detection model, at various levels of granularity. We employ software based fault injection (FI) to systematically assess the reliability of autoencoders on an industrial dataset, revealing substantial variability in resilience across network layers and bit positions, with minimal variation among object categories. We further propose a workflow to enhance model robustness, achieving an average of ~48% improvement in resilience with only additional ~4% memory overhead for convolutional autoencoders. Based on the findings, the study presents guidelines for systematic reliability assessment and targeted protective measures, providing a practical path toward safer, more dependable anomaly detection models in consumer electronics manufacturing.

Förderung

  • National Science and Technology Council, Taiwan
  • Google Research Award

Zitieren

APA

Pai, C., Hsueh, H., & Hsu, S. (2025). Anomaly Detection Reliability in Consumer Electronics Manufacturing. IEEE Transactions on Consumer Electronics, 71(4), 9287-9300. https://doi.org/10.1109/tce.2025.3626738

Vancouver

Pai C, Hsueh H, Hsu S. Anomaly Detection Reliability in Consumer Electronics Manufacturing. IEEE Transactions on Consumer Electronics. 2025;71(4):9287-9300. doi:10.1109/tce.2025.3626738

BibTeX
@article{Pai2025,
  author = {Pai, Chuen-Wun and Hsueh, Hung-Wei and Hsu, Shu-Han},
  title = {Anomaly Detection Reliability in Consumer Electronics Manufacturing},
  journal = {IEEE Transactions on Consumer Electronics},
  year = {2025},
  volume = {71},
  number = {4},
  pages = {9287--9300},
  doi = {10.1109/tce.2025.3626738}
}
Kostenlos starten →