Scholarlink

Purpose <p>This study aims to develop an adaptive and efficient fault diagnosis method to promptly capture various fault modes of rolling bearings and provide new ideas for the future real-time application. The research enhances fault feature extraction and classification accuracy while minimizing computational time compared to existing methods.</p> Methods <p>A fault diagnosis method, termed RTH-VMD and E-RTH-LSSVM, integrates the Red-Tailed Hawk (RTH) optimization algorithm, optimized Variational Mode Decomposition (VMD), Dispersion entropy (DispEn)-based feature selection, and optimized Least Squares Support Vector Machine (LSSVM). Initially, RTH optimizes VMD parameters to adaptively decompose bearing vibration signals into intrinsic mode functions (IMFs). Subsequently, the two IMFs with the highest DispEn values are selected to construct feature vectors based on spectral kurtosis, peak index, and kurtosis index. The RTH algorithm then optimizes LSSVM hyperparameters to minimize classification error. The method is validated using the Case Western Reserve University (CWRU) bearing dataset.</p> Results <p>The proposed method achieves 99.68% classification accuracy on the training set and 97.5% on the test set, with a computational time of 62&#xa0;s. Compared to other optimization methods (average test accuracy of 96.27%, an improvement of approximately 1%, computational time of 100–400&#xa0;s, an improvement of approximately 47.5% to 84.1%), it offers higher accuracy and faster processing.</p> Conclusion <p>The RTH-VMD-LSSVM method effectively extracts fault features, enhancing diagnostic accuracy and efficiency. Its robustness and reduced computational time make it promising for practical bearing fault diagnosis applications.</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 Rolling Bearing Fault Diagnosis Method Based on E-RTH Multi-parameter Optimization of LSSVM

  • 期刊论文

  • 发表日期：

    2025年12月16日

查看全文
  • Jiahao Zhong,
  • Wenyi Liu,
  • Di Song,
  • Jianbin Cao,
  • Taskeen Zahra,
  • Guohui Xu

摘要

Purpose

This study aims to develop an adaptive and efficient fault diagnosis method to promptly capture various fault modes of rolling bearings and provide new ideas for the future real-time application. The research enhances fault feature extraction and classification accuracy while minimizing computational time compared to existing methods.

Methods

A fault diagnosis method, termed RTH-VMD and E-RTH-LSSVM, integrates the Red-Tailed Hawk (RTH) optimization algorithm, optimized Variational Mode Decomposition (VMD), Dispersion entropy (DispEn)-based feature selection, and optimized Least Squares Support Vector Machine (LSSVM). Initially, RTH optimizes VMD parameters to adaptively decompose bearing vibration signals into intrinsic mode functions (IMFs). Subsequently, the two IMFs with the highest DispEn values are selected to construct feature vectors based on spectral kurtosis, peak index, and kurtosis index. The RTH algorithm then optimizes LSSVM hyperparameters to minimize classification error. The method is validated using the Case Western Reserve University (CWRU) bearing dataset.

Results

The proposed method achieves 99.68% classification accuracy on the training set and 97.5% on the test set, with a computational time of 62 s. Compared to other optimization methods (average test accuracy of 96.27%, an improvement of approximately 1%, computational time of 100–400 s, an improvement of approximately 47.5% to 84.1%), it offers higher accuracy and faster processing.

Conclusion

The RTH-VMD-LSSVM method effectively extracts fault features, enhancing diagnostic accuracy and efficiency. Its robustness and reduced computational time make it promising for practical bearing fault diagnosis applications.