EUVIMEDEUVIMED entdecken
Crossref

DWT-MWCGAN: A Data Enhanced Model for Diesel Engine Bearing Bush Fault Diagnosis Under Small Sample

Bingwu Gao, Jing Xu, Zhenrui Zhang, Changwen Yang, Xinyang Sun, Xiaonan Chang, Jianfeng Liu

IEEE Transactions on Instrumentation and Measurement · 2025 · 74 · 1-20

DOI: 10.1109/tim.2025.3618431 ↗

Kurz gefasst

Fachartikel in der Zeitschrift IEEE Transactions on Instrumentation and Measurement, 2025. Verfasst von Bingwu Gao und 6 weiteren Autor:innen. Die Arbeit wurde bisher 3-mal zitiert (Crossref) und nennt 26 Literaturangaben. Gefördert von: Natural Science Foundation of Jiangsu Province, China Postdoctoral Science Foundation, Jiangsu University of Science and Technology Doctoral Research Start-Up Funds.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

As a critical component of marine diesel engine, the accurate diagnosis of bearing bush faults is crucial. However, in complex real-world scenarios, the acquisition of fault data for bearing bush is challenging and often hindered by data scarcity. Accurately diagnosing bearing bush faults under small-sample conditions remains a significant obstacle. Therefore, this article proposes a method based on the combination of discrete wavelet transform (DWT) and multikernel weight learning maximum mean difference (MWMMD) conditional generative adversarial network (CGAN), DWT-MWCGAN, for the fault diagnosis of diesel engine bearing bush. The proposed method employs multilevel wavelet decomposition to effectively extract both low- and high-frequency components of the signal. By capturing the time- and frequency-domain features, the accuracy of information modeling and the capability for frequency-based information analysis are significantly enhanced. Furthermore, the integration of a multikernel weight learning strategy with the maximum mean difference function enables the model to automatically adjust and optimize the contribution of each kernel. This approach not only mitigates the excessive smoothing of specific scales or features but also improves the adaptability of the model to varying data distributions. The experimental results demonstrate that a high diagnostic accuracy exceeding 98% has been achieved by DWT-MWCGAN under various rotational speed conditions, compared to traditional and other advanced methods. These findings indicate that the proposed method effectively addresses the challenges posed by small-sample data scarcity and exhibits excellent performance in real-world application scenarios.

Förderung

  • Natural Science Foundation of Jiangsu Province
  • China Postdoctoral Science Foundation
  • Jiangsu University of Science and Technology Doctoral Research Start-Up Funds
  • Basic Science (Natural Science) Research Project of Higher Education in Jiangsu Province
  • Marine Equipment and Technology Institute of Jiangsu University of Science and Technology

Zitieren

APA

Gao, B., Xu, J., Zhang, Z., Yang, C., Sun, X., Chang, X., & Liu, J. (2025). DWT-MWCGAN: A Data Enhanced Model for Diesel Engine Bearing Bush Fault Diagnosis Under Small Sample. IEEE Transactions on Instrumentation and Measurement, 74, 1-20. https://doi.org/10.1109/tim.2025.3618431

Vancouver

Gao B, Xu J, Zhang Z, Yang C, Sun X, Chang X, et al. DWT-MWCGAN: A Data Enhanced Model for Diesel Engine Bearing Bush Fault Diagnosis Under Small Sample. IEEE Transactions on Instrumentation and Measurement. 2025;74:1-20. doi:10.1109/tim.2025.3618431

BibTeX
@article{Gao2025,
  author = {Gao, Bingwu and Xu, Jing and Zhang, Zhenrui and Yang, Changwen and Sun, Xinyang and Chang, Xiaonan and Liu, Jianfeng},
  title = {DWT-MWCGAN: A Data Enhanced Model for Diesel Engine Bearing Bush Fault Diagnosis Under Small Sample},
  journal = {IEEE Transactions on Instrumentation and Measurement},
  year = {2025},
  volume = {74},
  pages = {1--20},
  doi = {10.1109/tim.2025.3618431}
}
Kostenlos starten →