Scholarlink

Purpose <p>In fluid transportation, pipe elbows are a primary source of vibration. Their low-frequency vibration signals reflect critical fluid-structure interaction characteristics. However, these signals are weak and non-stationary, often causing the low-frequency components to overlap with broadband noise. Most existing noise reduction methods focus on removing high-frequency narrowband noise, making it difficult to effectively separate broadband noise from low-frequency fluid-structure coupling features. This leads to potential loss of useful information, affecting subsequent pipeline vibration analysis.</p> Method <p>Therefore, this study proposes a feature extraction method combining GWO-VMD-WD. First, the VMD algorithm decomposes the original vibration signal, while the Grey Wolf Optimizer (GWO) adaptively optimizes the VMD parameters using a composite fitness function. Next, the Wasserstein Distance (WD) classifies the decomposed components, identifying noise modes based on their natural frequency characteristics. Then, wavelet thresholding extracts potential valid vibrations from the noise modes. Finally, the method combines the extracted feature modes with the recovered valid components to reconstruct the signal, achieving effective elbow vibration feature extraction.</p> Results <p>Compared to conventional denoising methods, the GWO-VMD-WD approach delivers better pipeline vibration feature extraction, achieving higher signal-to-noise ratio and lower mean square error. The experimental data reveals three distinct phases: (1) Below 16&#xa0;m/s, the amplitude grows linearly with airflow speed, (2) Between 16 and 18&#xa0;m/s, induced resonance occurs, (3) Above 18&#xa0;m/s, the amplitude shows nonlinear increase. Frequency analysis identifies characteristic bands for different flow regimes: low-speed (2–10&#xa0;Hz), medium-speed (60–80&#xa0;Hz), and high-speed (160–180&#xa0;Hz).</p> Conclusion <p>The GWO-VMD-WD algorithm developed in this study demonstrates excellent applicability in pipeline vibration research. The results indicate that the increase in elbow amplitude follows a nonlinear pattern, with vibration intensity continuing to strengthen as airflow velocity increases until reaching a saturated state. The dominant vibration frequency shifts with changes in airflow velocity. This phenomenon stems from the unique vibration modes present within the pipeline at different velocities, with each natural frequency exhibiting sensitivity to changes in airflow velocity. This study provides insights into vibration reduction for large-scale industrial pipeline systems.</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

VMD-WD Noise Reduction and Feature Extraction for Vibration Signals of Hydrogen Pipeline Elbows

  • 期刊论文

  • 发表日期：

    2025年9月3日

查看全文
  • Lihua Huang,
  • Dongliang Zheng,
  • Zhonglin Ma,
  • Shuting Wan,
  • Meixin Feng,
  • Mengjie Jiang

摘要

Purpose

In fluid transportation, pipe elbows are a primary source of vibration. Their low-frequency vibration signals reflect critical fluid-structure interaction characteristics. However, these signals are weak and non-stationary, often causing the low-frequency components to overlap with broadband noise. Most existing noise reduction methods focus on removing high-frequency narrowband noise, making it difficult to effectively separate broadband noise from low-frequency fluid-structure coupling features. This leads to potential loss of useful information, affecting subsequent pipeline vibration analysis.

Method

Therefore, this study proposes a feature extraction method combining GWO-VMD-WD. First, the VMD algorithm decomposes the original vibration signal, while the Grey Wolf Optimizer (GWO) adaptively optimizes the VMD parameters using a composite fitness function. Next, the Wasserstein Distance (WD) classifies the decomposed components, identifying noise modes based on their natural frequency characteristics. Then, wavelet thresholding extracts potential valid vibrations from the noise modes. Finally, the method combines the extracted feature modes with the recovered valid components to reconstruct the signal, achieving effective elbow vibration feature extraction.

Results

Compared to conventional denoising methods, the GWO-VMD-WD approach delivers better pipeline vibration feature extraction, achieving higher signal-to-noise ratio and lower mean square error. The experimental data reveals three distinct phases: (1) Below 16 m/s, the amplitude grows linearly with airflow speed, (2) Between 16 and 18 m/s, induced resonance occurs, (3) Above 18 m/s, the amplitude shows nonlinear increase. Frequency analysis identifies characteristic bands for different flow regimes: low-speed (2–10 Hz), medium-speed (60–80 Hz), and high-speed (160–180 Hz).

Conclusion

The GWO-VMD-WD algorithm developed in this study demonstrates excellent applicability in pipeline vibration research. The results indicate that the increase in elbow amplitude follows a nonlinear pattern, with vibration intensity continuing to strengthen as airflow velocity increases until reaching a saturated state. The dominant vibration frequency shifts with changes in airflow velocity. This phenomenon stems from the unique vibration modes present within the pipeline at different velocities, with each natural frequency exhibiting sensitivity to changes in airflow velocity. This study provides insights into vibration reduction for large-scale industrial pipeline systems.