Scholarlink

<p>Accurate and efficient bearing fault diagnosis technology is crucial for ensuring the long-term stable operation of rotating machinery. In real industrial environments, sensor data often exhibit varying timescales and heterogeneous feature distributions, which make it difficult for conventional CNN-based methods to extract sufficient discriminative features. To address this issue, this paper proposes a novel fault diagnosis framework: adaptive dilated convolution with improved channel attention and multi-scale feature parallel fusion (ADCF). To begin with, a gating mechanism is designed to adaptively adjusts dilation rates to capture fault features across multiple time scales, addressing the issue of scale variability. Then, an improved channel attention module is employed to emphasize critical fault-related channels while suppressing redundant information, thereby mitigating feature distribution imbalance. In addition, a multi-scale feature parallel fusion module is designed to efficiently fuse the learned multi-scale features. Finally, ADCF is applied to multiple datasets. Experimental results demonstrate that this framework significantly outperforms existing state-of-the-art fault diagnosis methods across multiple evaluation metrics.</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

A novel multi-scale adaptive feature fusion framework for accurate and efficient bearing fault diagnosis

  • 期刊论文

  • 发表日期：

    2025年10月25日

查看全文
  • Yandong Hou,
  • Huige Huang,
  • Zhengquan Chen,
  • Xiaoao Cai,
  • Yankun Han,
  • Xiaodong Zhai

摘要

Accurate and efficient bearing fault diagnosis technology is crucial for ensuring the long-term stable operation of rotating machinery. In real industrial environments, sensor data often exhibit varying timescales and heterogeneous feature distributions, which make it difficult for conventional CNN-based methods to extract sufficient discriminative features. To address this issue, this paper proposes a novel fault diagnosis framework: adaptive dilated convolution with improved channel attention and multi-scale feature parallel fusion (ADCF). To begin with, a gating mechanism is designed to adaptively adjusts dilation rates to capture fault features across multiple time scales, addressing the issue of scale variability. Then, an improved channel attention module is employed to emphasize critical fault-related channels while suppressing redundant information, thereby mitigating feature distribution imbalance. In addition, a multi-scale feature parallel fusion module is designed to efficiently fuse the learned multi-scale features. Finally, ADCF is applied to multiple datasets. Experimental results demonstrate that this framework significantly outperforms existing state-of-the-art fault diagnosis methods across multiple evaluation metrics.