Scholarlink

<p>Journal bearing failures in centrifugal compressors can lead to severe operational disruptions, making early fault detection essential for predictive maintenance. This study presents a comprehensive failure analysis and prediction framework using vibration analysis and machine learning techniques. Over an extended period, extensive vibration data were collected and analyzed to identify journal bearing failures. The root cause of the failures was determined to be stress-induced deformation due to pressure variations in the suction and discharge lines. Leveraging this dataset, a hybrid model combining support vector machine (SVM) and random forest was developed to enable early fault detection. The proposed model demonstrated 98.8998% accuracy in predicting journal bearing failures, providing a reliable tool for condition monitoring and preventive maintenance. The findings emphasize the significance of integrating vibration-based diagnostics with machine learning for enhancing the reliability and efficiency of centrifugal compressors 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

High-Accuracy Prediction of Journal Bearing Failures in Centrifugal Compressors Using a Hybrid SVM–Random Forest Approach

  • 期刊论文

  • 发表日期：

    2025年7月15日

查看全文
  • Ali Hemati

摘要

Journal bearing failures in centrifugal compressors can lead to severe operational disruptions, making early fault detection essential for predictive maintenance. This study presents a comprehensive failure analysis and prediction framework using vibration analysis and machine learning techniques. Over an extended period, extensive vibration data were collected and analyzed to identify journal bearing failures. The root cause of the failures was determined to be stress-induced deformation due to pressure variations in the suction and discharge lines. Leveraging this dataset, a hybrid model combining support vector machine (SVM) and random forest was developed to enable early fault detection. The proposed model demonstrated 98.8998% accuracy in predicting journal bearing failures, providing a reliable tool for condition monitoring and preventive maintenance. The findings emphasize the significance of integrating vibration-based diagnostics with machine learning for enhancing the reliability and efficiency of centrifugal compressors in industrial applications.