Scholarlink

<p>Accurate fault diagnosis of spinning frame bearings holds significant engineering value in ensuring continuous production, within the context of intelligent transformation in the textile industry. Addressing three core challenges in industrial settings: (1) difficulty in extracting weak fault signatures; (2) insufficient characterization capability under multi-operational conditions; (3) lack of diagnostic robustness amidst complex interference, this study proposes the innovative multi-condition diagnostic model PMC-RAN-MD. The innovation of the proposed model manifests in three technical dimensions: First, the pyramid multi-scale convolution (PMC) module based on convolution kernel stacking achieves deep coupling of cross-scale fault features. Second, embedding channel attention mechanisms into parallel residual architectures establishes the residual attention network (RAN) with dynamic feature weighting to enhance critical fault signatures. Finally, the multimodal downsampling (MD) architecture enhances adaptability to complex operational conditions through collaborative mapping in heterogeneous feature spaces. Experimental validation demonstrates that the proposed model achieves 98.52% recognition accuracy on synthetic fault datasets and attains 100% diagnostic reliability in practical industrial scenario testing. The proposed model demonstrates superior diagnostic performance compared to existing fault diagnosis models.</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 of Spinning Frame Bearings Based on PMC-RAN-MD

  • 期刊论文

  • 发表日期：

    2025年6月21日

查看全文
  • Zhen Lei,
  • Tian Chen

摘要

Accurate fault diagnosis of spinning frame bearings holds significant engineering value in ensuring continuous production, within the context of intelligent transformation in the textile industry. Addressing three core challenges in industrial settings: (1) difficulty in extracting weak fault signatures; (2) insufficient characterization capability under multi-operational conditions; (3) lack of diagnostic robustness amidst complex interference, this study proposes the innovative multi-condition diagnostic model PMC-RAN-MD. The innovation of the proposed model manifests in three technical dimensions: First, the pyramid multi-scale convolution (PMC) module based on convolution kernel stacking achieves deep coupling of cross-scale fault features. Second, embedding channel attention mechanisms into parallel residual architectures establishes the residual attention network (RAN) with dynamic feature weighting to enhance critical fault signatures. Finally, the multimodal downsampling (MD) architecture enhances adaptability to complex operational conditions through collaborative mapping in heterogeneous feature spaces. Experimental validation demonstrates that the proposed model achieves 98.52% recognition accuracy on synthetic fault datasets and attains 100% diagnostic reliability in practical industrial scenario testing. The proposed model demonstrates superior diagnostic performance compared to existing fault diagnosis models.