Evaluation Method of Rotating Machinery Health State Based on TPE-XGBoost

Aiming at the problems of imbalanced data samples and complex parameter optimization in the existing health assessment tasks of rotating machinery method, this paper develops a health assessment method for rotating machinery based on the TPE-XGBoost model. The method consists of the following steps: (1) Extract 15 commonly used time-frequency domain parameters as input; (2) Label the samples with three categories: healthy, degradation and failure; (3) Use SMOTE-ENN method to balance the labeled samples; (4) With the balanced samples as input, TPE-XGBoost model is trained offline to get the best parameters and structures for online evaluation. (5) Evaluate the health status according to the probability obtained. The applications show that this method can be effectively trained and give early warnings of equipment degradation. It is superior than other commonly used methods.