The Gas-Liquid Flow Rate Measurement Based on Multisensors and Machine Learning

Gas-liquid two-phase flow non-separation measurement plays a crucial role in industrial production and two-phase flow theory research. In this paper, an intelligent multi-sensing system is designed for horizontal gas-liquid flow measurement, which integrates near-infrared (NIR), differential pressure (DP) and acoustic emission (AE) technology. The interaction and disturbance information of the gas-liquid phase are detection by the AE and NIR sensors, and then transformed into the form and margin factor which reflect the void fraction and flow rate of gas liquid flow respectively. Both of them are input as feature variables into the least absolute shrinkage and selection operator (LASSO) machine learning model for predicting the liquid phase volume fraction, and the best fusion solution is proposed. The Analysis and evaluation have been conducted in several classic correlations, and the Collins correlation is selected to optimize by LASSO optimization algorithm. The prediction accuracy is improved by 7.14% compared to the Collins correlation, and the MAPE for liquid mass flow rate of slug flow was 8.75% and 89.47% of the predictions were within 20% relative error.