Scholarlink

Objective <p>To address the issues of poor noise resistance, limited generalization capability, and high computational complexity in existing deep learning methods for industrial bearing fault diagnosis, this paper proposes a Dual-Path Wavelet Attention Dynamic Network (TF-DPWN).</p> Methods <p>The model incorporates three innovative mechanisms: 1) An adaptive wavelet convolution mechanism that dynamically optimizes Gaussian envelope parameters and center frequencies through backpropagation to autonomously identify fault characteristics; 2) A lightweight cross-dimensional attention mechanism that reduces computational complexity from O(N<InlineEquation ID="IEq1"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="42417_2025_2087_Article_IEq1.gif" Format="GIF" Height="10" Rendition="HTML" Resolution="72" Type="Linedraw" Width="8" /> </InlineMediaObject> <EquationSource Format="TEX">\(^{2}\)</EquationSource> <EquationSource Format="MATHML"><math> <mmultiscripts> <mrow /> <mrow /> <mn>2</mn> </mmultiscripts> </math></EquationSource> </InlineEquation>) to O(N) while maintaining sensitivity to subtle faults; 3) A dynamic channel gating mechanism that enhances fault-related features while suppressing irrelevant ones.</p> Results <p>Experimental results on an eyeglass lens edger platform show that TF-DPWN achieves average accuracies of 99.94% (±0.00%) and 98.42% (±0.17%) under noise-free and white Gaussian noise conditions, respectively, while reducing parameter count by 13.13% and computational load by 69.76% compared to baseline models.</p> Conclusions <p>The proposed TF-DPWN not only enhances noise robustness and accuracy in industrial fault diagnosis, but also provides an intelligent maintenance solution with both interpretability and efficiency for practical industrial applications through adaptive feature learning and lightweight design.</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 Dual-Path Wavelet Dynamic Network for Intelligent Bearing Fault Diagnosis

  • 期刊论文

  • 发表日期：

    2025年9月10日

查看全文
  • Qi Jin,
  • De Gu,
  • Jianchu Liu

摘要

Objective

To address the issues of poor noise resistance, limited generalization capability, and high computational complexity in existing deep learning methods for industrial bearing fault diagnosis, this paper proposes a Dual-Path Wavelet Attention Dynamic Network (TF-DPWN).

Methods

The model incorporates three innovative mechanisms: 1) An adaptive wavelet convolution mechanism that dynamically optimizes Gaussian envelope parameters and center frequencies through backpropagation to autonomously identify fault characteristics; 2) A lightweight cross-dimensional attention mechanism that reduces computational complexity from O(N \(^{2}\) 2 ) to O(N) while maintaining sensitivity to subtle faults; 3) A dynamic channel gating mechanism that enhances fault-related features while suppressing irrelevant ones.

Results

Experimental results on an eyeglass lens edger platform show that TF-DPWN achieves average accuracies of 99.94% (±0.00%) and 98.42% (±0.17%) under noise-free and white Gaussian noise conditions, respectively, while reducing parameter count by 13.13% and computational load by 69.76% compared to baseline models.

Conclusions

The proposed TF-DPWN not only enhances noise robustness and accuracy in industrial fault diagnosis, but also provides an intelligent maintenance solution with both interpretability and efficiency for practical industrial applications through adaptive feature learning and lightweight design.