Knowledge-Driven Intelligent Defect Detection for Glass-Encapsulated Electrical Connectors Using Hybrid MS-ViT-CNN
International Journal of Knowledge Management · 26.06.2026 · 22 (1) · 1-17
DOI: 10.4018/ijkm.413884 ↗Kurz gefasst
Fachartikel in der Zeitschrift International Journal of Knowledge Management, erschienen am 26.06.2026. Verfasst von Yuhong Du und 2 weiteren Autor:innen. Die Arbeit nennt 32 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 3.0).
Abstract (Originalsprache)
Effective knowledge extraction and intelligent decision-making are critical for manufacturing quality control. This study develops a knowledge-driven defect detection framework for glass-encapsulated electrical connectors by integrating optical prior knowledge and a hybrid multi-scale vision transformer-convolutional neural network (MS-ViT-CNN). The framework embeds differentiable transmission matrix modeling to simulate refraction-scattering interferences and generate high-reliability training samples. The MS-ViT-CNN backbone combines convolutional features and window self-attention to capture both local defect details and long-range interference patterns. Evaluations on 12,736 real industrial images show that the model significantly reduces missed and false detection rates, with strong interpretability verified via Grad-CAM. Deployment on Jetson NX achieves real-time edge inference for production lines. This approach provides a practical solution for knowledge-based intelligent manufacturing and automated defect detection.
Zitieren
APA
Du, Y., Zhang, S., & Hou, S. (2026). Knowledge-Driven Intelligent Defect Detection for Glass-Encapsulated Electrical Connectors Using Hybrid MS-ViT-CNN. International Journal of Knowledge Management, 22(1), 1-17. https://doi.org/10.4018/ijkm.413884
Vancouver
Du Y, Zhang S, Hou S. Knowledge-Driven Intelligent Defect Detection for Glass-Encapsulated Electrical Connectors Using Hybrid MS-ViT-CNN. International Journal of Knowledge Management. 2026;22(1):1-17. doi:10.4018/ijkm.413884
BibTeX
@article{Du2026,
author = {Du, Yuhong and Zhang, Songqi and Hou, Shouming},
title = {Knowledge-Driven Intelligent Defect Detection for Glass-Encapsulated Electrical Connectors Using Hybrid MS-ViT-CNN},
journal = {International Journal of Knowledge Management},
year = {2026},
volume = {22},
number = {1},
pages = {1--17},
doi = {10.4018/ijkm.413884}
}
EUVIMED