Prediction of the interfacial disturbance wave velocity in vertical upward gas-liquid annular flow via ensemble learning. (1st March 2022)

• Record Type: Journal Article
• Authors: Li, Chaofan; Song, Yajing; Xu, Long; Zhao, Ning; Wang, Fan; Fang, Lide; Li, Xiaoting
• Is Part Of: Energy. Volume 242(2022)
• Publication Date: 2022-03-01
• Subjects: Annular flow -- Near-infrared measurement sensors -- Disturbance wave velocity -- Ensemble learning -- Cross-correlation analysis; Power resources -- Periodicals; Power (Mechanics) -- Periodicals; Energy consumption -- Periodicals; 333.7905
• DOI: 10.1016/j.energy.2021.122990
• Languages: English
• ISSNs: 0360-5442

Abstract

Abstract: Interfacial disturbance wave velocity is an important parameter for the study of momentum transfer between the gas core and the liquid film at the two-phase interface, which directly affects the calculation of frictional pressure drop. Since an exact analytical solution of the interfacial disturbance velocity cannot be derived by the two-phase flow theory equations, an ensemble learning framework for the disturbance wave velocity is constructed and a new model is proposed, which is appropriate for predicting different flow conditions in vertical two-phase flow. The dimensionless velocity-related parameters of interfacial disturbance waves are obtained by feature selection based on the interfacial shear force model, and the grid search method is equipped to tune the important parameters. By comparing the current and literature data prediction results, the extrapolation and applicability of the ensemble learning model are further verified. For the Extra Tree model, the Mean Absolute Percentage Error of the optimized Extra Tree model is less than 20%, and the relative measurement uncertainty is within ±25% for 95.67% of the results. It shows that the proposed ensemble learning framework provides a novel approach in the study of interfacial wave spatiotemporal parameters. Highlights: The disturbance wave velocity is measured by a near-infrared sensor combined with the cross-correlation principle. The dimensionless parameters are used for the ensemble learning model based on the interfacial shear model. The study develop an ensemble learning framework for predicting the disturbance wave velocity. The prediction models incorporates the effects of pipe diameter and pressure, extending the capability of empirical models.