Scholarlink

Background <p>Rolling bearings are essential components in rotating machinery, but their operation in harsh environments makes them susceptible to failure. Diagnosing these faults is challenging due to the non-stationary nature of vibration signals and interference from strong background noise. Furthermore, traditional decomposition methods often suffer from mode mixing and endpoint effects, hindering accurate feature extraction.</p> Purpose <p>This study aims to propose a novel integrated fault diagnosis approach to effectively extract weak fault features from noisy signals and address the specific limitations of traditional Empirical Mode Decomposition (EMD) methods.</p> Methods <p>A hybrid diagnostic framework is introduced. First, signals are preprocessed using an improved wavelet threshold denoising method that innovatively combines cross-correlation coefficient filtering, sample entropy-based threshold optimization, and an improved threshold function to suppress noise. Second, the denoised signals are decomposed using a novel Complementary Ensemble High-Frequency Harmonic-Assisted Empirical Mode Decomposition (CEHFHA-EMD) algorithm. This algorithm utilizes high-frequency harmonic assistance to suppress mode mixing and a new boundary extreme point determination criterion to mitigate endpoint effects.</p> Results <p>Simulation results demonstrated that the improved denoising method achieved a Signal-to-Noise Ratio (SNR) of 12.2215, significantly outperforming traditional hard and soft thresholding methods. The CEHFHA-EMD algorithm reduced the Orthogonality Index to 0.0521, representing a 52.7% reduction compared to standard CEEMD, thereby effectively suppressing mode mixing. Experimental validation using the Case Western Reserve University dataset and a custom test rig confirmed that the proposed method produced clearer envelope spectra with more distinct characteristic frequencies compared to the ICEEMDAN method, particularly for rolling element faults.</p> Conclusion <p>The proposed approach, integrating improved wavelet threshold denoising and CEHFHA-EMD, successfully enhances signal decomposition accuracy and fault feature extraction. It provides a robust solution for identifying bearing fault types in complex, real-world noise environments.</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 Approach Integrating Improved Wavelet Threshold Denoising and CEHFHA-EMD

  • 期刊论文

  • 发表日期：

    2026年1月14日

查看全文
  • Xia Yang,
  • Ruixiang Hou,
  • Changfan Liu,
  • Shuangshuang Liu,
  • Yafeng Ji,
  • Cunlong Zhou

摘要

Background

Rolling bearings are essential components in rotating machinery, but their operation in harsh environments makes them susceptible to failure. Diagnosing these faults is challenging due to the non-stationary nature of vibration signals and interference from strong background noise. Furthermore, traditional decomposition methods often suffer from mode mixing and endpoint effects, hindering accurate feature extraction.

Purpose

This study aims to propose a novel integrated fault diagnosis approach to effectively extract weak fault features from noisy signals and address the specific limitations of traditional Empirical Mode Decomposition (EMD) methods.

Methods

A hybrid diagnostic framework is introduced. First, signals are preprocessed using an improved wavelet threshold denoising method that innovatively combines cross-correlation coefficient filtering, sample entropy-based threshold optimization, and an improved threshold function to suppress noise. Second, the denoised signals are decomposed using a novel Complementary Ensemble High-Frequency Harmonic-Assisted Empirical Mode Decomposition (CEHFHA-EMD) algorithm. This algorithm utilizes high-frequency harmonic assistance to suppress mode mixing and a new boundary extreme point determination criterion to mitigate endpoint effects.

Results

Simulation results demonstrated that the improved denoising method achieved a Signal-to-Noise Ratio (SNR) of 12.2215, significantly outperforming traditional hard and soft thresholding methods. The CEHFHA-EMD algorithm reduced the Orthogonality Index to 0.0521, representing a 52.7% reduction compared to standard CEEMD, thereby effectively suppressing mode mixing. Experimental validation using the Case Western Reserve University dataset and a custom test rig confirmed that the proposed method produced clearer envelope spectra with more distinct characteristic frequencies compared to the ICEEMDAN method, particularly for rolling element faults.

Conclusion

The proposed approach, integrating improved wavelet threshold denoising and CEHFHA-EMD, successfully enhances signal decomposition accuracy and fault feature extraction. It provides a robust solution for identifying bearing fault types in complex, real-world noise environments.