Scholarlink

<p>Effective lubrication is crucial for ensuring the reliability and longevity of rotating machinery. This study aims to assess the lubrication condition of bearings using vibration analysis and develop a predictive model to determine when lubrication is necessary. Experimental historical time-domain signals, spectral data, and envelope signals are analyzed to identify lubrication-related vibration characteristics. When lubrication is required, high-frequency components with significant amplitude appear in vibration spectra and envelopes. After applying grease, these frequencies diminish, indicating reduced friction. A deep learning approach based on long short-term memory (LSTM) networks is employed to predict lubrication necessity. The LSTM model is trained on time-domain vibration signals corresponding to both lubricated and non-lubricated conditions to enable accurate and automated prediction. The proposed method enhances predictive maintenance by enabling early detection of lubrication needs, reducing unplanned downtime, and optimizing lubrication schedules. This approach improves the efficiency and reliability of rotating machinery in industrial 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

Bearing Failure Prevention Through Predictive Lubrication Monitoring Using Vibration Signals and Deep Learning

  • 期刊论文

  • 发表日期：

    2025年9月10日

查看全文
  • Ali Hemati

摘要

Effective lubrication is crucial for ensuring the reliability and longevity of rotating machinery. This study aims to assess the lubrication condition of bearings using vibration analysis and develop a predictive model to determine when lubrication is necessary. Experimental historical time-domain signals, spectral data, and envelope signals are analyzed to identify lubrication-related vibration characteristics. When lubrication is required, high-frequency components with significant amplitude appear in vibration spectra and envelopes. After applying grease, these frequencies diminish, indicating reduced friction. A deep learning approach based on long short-term memory (LSTM) networks is employed to predict lubrication necessity. The LSTM model is trained on time-domain vibration signals corresponding to both lubricated and non-lubricated conditions to enable accurate and automated prediction. The proposed method enhances predictive maintenance by enabling early detection of lubrication needs, reducing unplanned downtime, and optimizing lubrication schedules. This approach improves the efficiency and reliability of rotating machinery in industrial applications.