Cross-domain intelligent fault diagnosis of rotating machinery under data-scarce scenarios: Wasserstein distance-guided joint domain adaptation
Measurement Science and Technology · 30.01.2026 · 37 (4) · 046118
DOI: 10.1088/1361-6501/ae3a07 ↗Kurz gefasst
Fachartikel in der Zeitschrift Measurement Science and Technology, online erschienen am 30.01.2026. Verfasst von Kangning Du und 4 weiteren Autor:innen. Die Arbeit wurde bisher 5-mal zitiert (Crossref) und nennt 56 Literaturangaben. Gefördert von: National Key Research and Development Program of China.
Abstract (Originalsprache)
Domain adaptation (DA) based on statistical distance metrics often neglects the geometric features of data distributions; adversarial learning-based DA typically employs the adversarial loss function within the generative adversarial network, which can lead to the domain discriminator’s gradient vanishing during training. To address these issues, this paper proposes a Wasserstein distance (WD)-guided joint DA approach for cross-domain fault diagnosis scenarios. First, we establish a DA method based on the 2-WD, employing the covariance matrix and second-order metric structure to perform geometric encoding on the distribution of high-dimensional fault data. This enables the recovery of intrinsic geometric relationships within the data distribution on high-dimensional manifold spaces, achieving sensitive capture of higher-order correlations, local geometric curvature, and manifold shape. Second, we introduce an adversarial DA method based on WD to provide a more reasonable cost function for diagnostic models when learning low-dimensional manifold distribution features. This adversarial DA method no longer enforces Lipschitz constraints through weight clipping but instead indirectly controls the discriminator’s functional space via gradient penalty terms, enabling it to satisfy more relaxed Lipschitz conditions. The proposed method was experimentally validated on six datasets, with results demonstrating its diagnostic performance significantly outperforms existing popular methods.
Förderung
- National Key Research and Development Program of China
Zitieren
APA
Du, K., Wu, D., Bao, K., Li, M., & Zhu, B. (2026). Cross-domain intelligent fault diagnosis of rotating machinery under data-scarce scenarios: Wasserstein distance-guided joint domain adaptation. Measurement Science and Technology, 37(4), 046118. https://doi.org/10.1088/1361-6501/ae3a07
Vancouver
Du K, Wu D, Bao K, Li M, Zhu B. Cross-domain intelligent fault diagnosis of rotating machinery under data-scarce scenarios: Wasserstein distance-guided joint domain adaptation. Measurement Science and Technology. 2026;37(4):046118. doi:10.1088/1361-6501/ae3a07
BibTeX
@article{Du2026,
author = {Du, Kangning and Wu, Dinghui and Bao, Ke and Li, Mengwei and Zhu, Bingxian},
title = {Cross-domain intelligent fault diagnosis of rotating machinery under data-scarce scenarios: Wasserstein distance-guided joint domain adaptation},
journal = {Measurement Science and Technology},
year = {2026},
volume = {37},
number = {4},
pages = {046118},
doi = {10.1088/1361-6501/ae3a07}
}
EUVIMED