EUVIMEDEUVIMED entdecken
Crossref

A Novel Graph Deviant-Associate Discrepancy Attention Network for Condition Monitoring With Wind Turbine Cyber-Physical Systems

Wenhe Chen, Hanting Zhou, Yuwang Sun, Zongxiang Chen, Zhipeng Chang, Min Xia

IEEE Transactions on Industrial Cyber-Physical Systems · 2026 · 4 · 562-574

DOI: 10.1109/ticps.2026.3703013 ↗

Kurz gefasst

Fachartikel in der Zeitschrift IEEE Transactions on Industrial Cyber-Physical Systems, 2026. Verfasst von Wenhe Chen und 5 weiteren Autor:innen. Die Arbeit wurde bisher 1-mal zitiert (Crossref) und nennt 27 Literaturangaben.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Wind-turbine (WT) cyber-physical systems (CPSs) stream multivariate sensor data whose strong noise and nonlinear couplings often cause false or missing alarms. However, the scarcity of labelled faults limits supervised solutions. To address these challenges, a semi-supervised graph deviant-associate discrepancy attention network (GDTAN) is proposed that learns only from healthy data to accurately monitor turbine conditions. GDTAN embeds each sensor into a graph and applies graph deviant attention to capture global spatial deviations. Meanwhile, the proposed method employes the temporal association discrepancy attention to highlight dynamic temporal patterns, which utilizes a symmetric Kullback–Leibler divergence–driven associate discrepancy loss to amplify reconstruction errors of anomalous signals. In addition, an online correction module is used to further smooth the resulting condition index to suppress spurious fluctuations. Experiments on two real-world WT datasets, named blade-icing and toothed-belt fracture, show that GDTAN attains precision, recall, F1-scores, and accuracy of {0.9837, 0.9696, 0.9766, 0.9604} and {1.0000, 0.9114, 0.9537, 0.9514}, respectively. The proposed method can achieve 4–27% higher performance than other state-of-the-art baselines on F1-scores, which validates its robustness, interpretability, and practical value for large-scale wind-turbine condition.

Zitieren

APA

Chen, W., Zhou, H., Sun, Y., Chen, Z., Chang, Z., & Xia, M. (2026). A Novel Graph Deviant-Associate Discrepancy Attention Network for Condition Monitoring With Wind Turbine Cyber-Physical Systems. IEEE Transactions on Industrial Cyber-Physical Systems, 4, 562-574. https://doi.org/10.1109/ticps.2026.3703013

Vancouver

Chen W, Zhou H, Sun Y, Chen Z, Chang Z, Xia M. A Novel Graph Deviant-Associate Discrepancy Attention Network for Condition Monitoring With Wind Turbine Cyber-Physical Systems. IEEE Transactions on Industrial Cyber-Physical Systems. 2026;4:562-574. doi:10.1109/ticps.2026.3703013

BibTeX
@article{Chen2026,
  author = {Chen, Wenhe and Zhou, Hanting and Sun, Yuwang and Chen, Zongxiang and Chang, Zhipeng and Xia, Min},
  title = {A Novel Graph Deviant-Associate Discrepancy Attention Network for Condition Monitoring With Wind Turbine Cyber-Physical Systems},
  journal = {IEEE Transactions on Industrial Cyber-Physical Systems},
  year = {2026},
  volume = {4},
  pages = {562--574},
  doi = {10.1109/ticps.2026.3703013}
}
Kostenlos starten →