关键词: 主轴承, 自动机器学习, AutoGluon, 故障诊断

Abstract: The failure of the main bearing will have a serious impact on the reliability and operational efficiency of generator. To improve the accurate diagnosis efficiency of main bearing faults, the proposes a real-time and fast fault diagnosis method based on the AutoGluon algorithm model of Automated Machine Learning ( AutoML) . Feature engineering processing was performed on the raw signal data of the main bearings of the generator collected; based on manual experience, label the sample dataset with fault and non fault labels, and divide the samples into training and testing sets in an 8 ∶ 2 ratio; The AutoGluon model was used for training, including automated model selection and hyperparameter tuning, and its effectiveness was verified by comparing it with traditional machine learning models such as random forests and extreme gradient boosting trees. The experimental results show that the main bearing fault diagnosis method using the
AutoGluon model has achieved significant results in accuracy and recall, reaching 83. 14% and 60. 74%, respectively, higher than the random forest and extreme gradient boosting tree models optimized by hyperparameters. This method can quickly and accurately diagnose the current state of the main bearing, with high diagnostic accuracy and robustness, and has certain application prospects in the field of generator main bearing fault diagnosis.

Key words: main bearings, automatic machine learning( AutoML), AutoGluon, fault diagnosis