Scholarlink

Bearing fault diagnosis is crucial for machine reliability and cost reduction. This project applies machine learning to classify bearing faults via vibration signal analysis. Data, prepared by Dr. Eric Bechhoefer for Machinery Failure Prevention Technology (MFPT), includes healthy bearings and those with inner and outer race faults under various loads. Key features like kurtosis, RMS frequency, and standard deviation were extracted, normalized, and ranked using a chi-squared test. Machine learning models, including decision trees (DT), Support Vector Machine (SVMs), and neural networks, were trained to classify bearing conditions. Performance was evaluated by accuracy, with discriminant analysis (DA) and ensemble models (EN) achieving high accuracy. This study demonstrates the effectiveness of combining vibration feature extraction with machine learning for precise bearing fault diagnosis, advancing predictive maintenance.

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

Precision Bearing Fault Diagnosis Using Advanced Machine Learning Models and Comprehensive Vibration Signal Feature Extraction

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Soufiane Aychour,
  • Ilias Ouachtouk,
  • Ahmed Mouchtachi

摘要

Bearing fault diagnosis is crucial for machine reliability and cost reduction. This project applies machine learning to classify bearing faults via vibration signal analysis. Data, prepared by Dr. Eric Bechhoefer for Machinery Failure Prevention Technology (MFPT), includes healthy bearings and those with inner and outer race faults under various loads. Key features like kurtosis, RMS frequency, and standard deviation were extracted, normalized, and ranked using a chi-squared test. Machine learning models, including decision trees (DT), Support Vector Machine (SVMs), and neural networks, were trained to classify bearing conditions. Performance was evaluated by accuracy, with discriminant analysis (DA) and ensemble models (EN) achieving high accuracy. This study demonstrates the effectiveness of combining vibration feature extraction with machine learning for precise bearing fault diagnosis, advancing predictive maintenance.