Scholarlink

Bearing fault diagnosis is crucial for ensuring the reliability and efficiency of industrial machinery. This paper amalgamates state-of-the-art methodologies derived from recent research efforts to enhance the accuracy and efficiency of bearing fault diagnosis. It integrates state-of-the-art techniques such as Swin Transformer-based architectures, Variational Mode Decomposition (VMD) for feature extraction, and hybrid models that merge one-dimensional CNNs with two-dimensional Swin Transformers. Evaluated on benchmark datasets from Case Western Reserve University, the proposed model shows superior accuracy over 95%, outperforming traditional convolutional neural networks (CNNs) and other machine learning algorithms like K-Nearest Neighbors (KNN) and Extreme Learning Machines (ELM). Moreover, by integrating unsupervised learning, optimized Swin Transformers, and adaptive feature fusion, it can address challenges in low-speed operations, data imbalance, and minor faults, ensuring better reliability and efficiency in complex industrial settings.

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

Bearing Fault Diagnosis with a Parallel Classification Model of CNN and Swin Transformer

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Jinze Li,
  • Yuan Liu,
  • Ying Wang,
  • Tongxin Zheng

摘要

Bearing fault diagnosis is crucial for ensuring the reliability and efficiency of industrial machinery. This paper amalgamates state-of-the-art methodologies derived from recent research efforts to enhance the accuracy and efficiency of bearing fault diagnosis. It integrates state-of-the-art techniques such as Swin Transformer-based architectures, Variational Mode Decomposition (VMD) for feature extraction, and hybrid models that merge one-dimensional CNNs with two-dimensional Swin Transformers. Evaluated on benchmark datasets from Case Western Reserve University, the proposed model shows superior accuracy over 95%, outperforming traditional convolutional neural networks (CNNs) and other machine learning algorithms like K-Nearest Neighbors (KNN) and Extreme Learning Machines (ELM). Moreover, by integrating unsupervised learning, optimized Swin Transformers, and adaptive feature fusion, it can address challenges in low-speed operations, data imbalance, and minor faults, ensuring better reliability and efficiency in complex industrial settings.