Ultrasonic flowmeter fault early warning method based on Bayesian optimization XGBoost

To accurately identify the abnormal operating status of ultrasonic flowmeters in advance, a fault warning method for the key parameters of ultrasonic flow meters based on Bayesian optimization and XGBoost algorithm is proposed. The method involves effective data preprocessing techniques applied to historical operational data from the remote online diagnostic system of the measurement system. A prediction model for the key parameters of ultrasonic flow meters is built using the Bayesian optimization approach combined with XGBoost algorithm. The fault warning limits for the key parameters of ultrasonic flow meters are determined based on the three-sigma criterion. The results of the experimental tests demonstrate that the proposed method can detect the abnormal operating status of the key parameters of ultrasonic flow meters in advance. Comparative analysis with lightGBM and CatBoost algorithms reveals that the XGBoost model optimized by Bayesian optimization exhibits significant advantages in terms of computational time and model regression evaluation.