Scholarlink

<p>In bearing fault diagnosis, deep learning has gained substantial attention; however, the increasing complexity of models results in a proliferation of hyperparameters, intensifying the difficulty of manual tuning. This study proposes an innovative diagnostic strategy based on improved high-dimensional Bayesian optimization for convolutional neural networks (CNNs). Leveraging the efficiency, scalability, and rapid convergence of the enhanced optimization algorithm, the proposed method automates and simplifies the hyperparameter tuning process for CNNs. Experiments conducted on the Case Western Reserve University (CWRU) bearing dataset show that the optimized CNN model achieves superior fault diagnosis performance for wind turbine bearings. Compared with manual or default configurations under the same architecture, the proposed approach improves diagnostic accuracy by 1–23.5 % and significantly enhances generalization ability and robustness. These results confirm the effectiveness and practical value of the proposed strategy, offering an efficient, reliable, and automated solution for fault diagnosis and predictive maintenance.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Fault diagnosis strategy for wind turbine bearings based on improved high-dimensional Bayesian optimized CNN

  • 期刊论文

  • 发表日期：

    2025年11月8日

查看全文
  • Juan Guan,
  • Yanhua Wang

摘要

In bearing fault diagnosis, deep learning has gained substantial attention; however, the increasing complexity of models results in a proliferation of hyperparameters, intensifying the difficulty of manual tuning. This study proposes an innovative diagnostic strategy based on improved high-dimensional Bayesian optimization for convolutional neural networks (CNNs). Leveraging the efficiency, scalability, and rapid convergence of the enhanced optimization algorithm, the proposed method automates and simplifies the hyperparameter tuning process for CNNs. Experiments conducted on the Case Western Reserve University (CWRU) bearing dataset show that the optimized CNN model achieves superior fault diagnosis performance for wind turbine bearings. Compared with manual or default configurations under the same architecture, the proposed approach improves diagnostic accuracy by 1–23.5 % and significantly enhances generalization ability and robustness. These results confirm the effectiveness and practical value of the proposed strategy, offering an efficient, reliable, and automated solution for fault diagnosis and predictive maintenance.