EUVIMEDEUVIMED entdecken
Crossref

CAFE: Condition-Attention Feature Extractor for Multicondition Fault Diagnosis of Rotating Machinery

Hongkai Xue, Baokang Zhang, Qi Wu, Caoyuan Gu, Hangfeng Guo, Wen-An Zhang

IEEE Sensors Journal · 01.07.2026 · 26 (13) · 19952-19961

DOI: 10.1109/jsen.2026.3691708 ↗

Kurz gefasst

Fachartikel in der Zeitschrift IEEE Sensors Journal, erschienen am 01.07.2026. Verfasst von Hongkai Xue und 5 weiteren Autor:innen. Die Arbeit nennt 33 Literaturangaben. Gefördert von: Fundamental Research Funds for the Provincial Universities of Zhejiang, Project of Huzhou Science and Technology Bureau, Key Research and Development Program of Zhejiang Province.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Variations in rotational speed and load cause significant changes in the statistical distribution of machinery signals, presenting a major challenge for multicondition fault diagnosis (MCFD). A common strategy is to align condition-invariant features to mitigate cross-condition variability. However, the coupling of condition and fault representations persists, limiting generalization across conditions. To address these challenges, a condition-attention feature extractor is proposed for MCFD of rotating machinery, namely CAFE. In this work, stator current signals are utilized for MCFD instead of vibration signals. A learnable notch module (LNM) is employed to decompose the three-phase current into a fundamental component and a residual component, thereby decoupling condition and fault representations. A condition-guided representation module (CGRM) is developed to improve the robustness and adaptability of fault diagnosis under diverse conditions. In addition, an adaptive weighting strategy (AWS) is introduced to dynamically balance condition recognition and fault diagnosis. Experimental results on three datasets show that the proposed method achieves diagnostic performance comparable to vibrationbased approaches, reaching a peak fault diagnosis accuracy of 99.33%, surpassing most existing state-of-the-art MCFD approaches.

Förderung

  • Fundamental Research Funds for the Provincial Universities of Zhejiang
  • Project of Huzhou Science and Technology Bureau
  • Key Research and Development Program of Zhejiang Province
  • Zhejiang Provincial Major Science and Technology
  • National Natural Science Foundation of China

Zitieren

APA

Xue, H., Zhang, B., Wu, Q., Gu, C., Guo, H., & Zhang, W. (2026). CAFE: Condition-Attention Feature Extractor for Multicondition Fault Diagnosis of Rotating Machinery. IEEE Sensors Journal, 26(13), 19952-19961. https://doi.org/10.1109/jsen.2026.3691708

Vancouver

Xue H, Zhang B, Wu Q, Gu C, Guo H, Zhang W. CAFE: Condition-Attention Feature Extractor for Multicondition Fault Diagnosis of Rotating Machinery. IEEE Sensors Journal. 2026;26(13):19952-19961. doi:10.1109/jsen.2026.3691708

BibTeX
@article{Xue2026,
  author = {Xue, Hongkai and Zhang, Baokang and Wu, Qi and Gu, Caoyuan and Guo, Hangfeng and Zhang, Wen-An},
  title = {CAFE: Condition-Attention Feature Extractor for Multicondition Fault Diagnosis of Rotating Machinery},
  journal = {IEEE Sensors Journal},
  year = {2026},
  volume = {26},
  number = {13},
  pages = {19952--19961},
  doi = {10.1109/jsen.2026.3691708}
}
Kostenlos starten →