基于集成学习和信息融合的段塞流分相流量测量  

Phase separation flow measurement of plug flow based on ensemble learning and information fusion

作　　者：温佳祺 杨叙宁 李金硕 丁振君[1,3] 董芳 WEN Jiaqi;YANG Xuning;LI Jinshuo;DING Zhenjun;DONG Fang(School of Quality and Technical Supervision,Hebei University,Baoding 071002,China;Engineering Research Center of Zero-carbon Energy Bulidings and Measurement Techniques,Ministry of Education,Baoding 071002,China;National and Local Joint Engineering Research Center for Measuring Instruments and Systems,Baoding 071002,China)

机构地区：[1]河北大学质量技术监督学院,河北保定071002 [2]零碳能源建筑与计量技术教育部工程研究中心,河北保定071002 [3]计量仪器与系统国家地方联合工程研究中心,河北保定071002

出　　处：《河北大学学报（自然科学版）》2024年第5期541-550,共10页Journal of Hebei University(Natural Science Edition)

基　　金：国家自然科学基金资助项目(62173122);河北省自然科学基金资助项目(F2022201034)。

摘　　要：段塞流是气液两相流中典型流型,准确测量其分相流量有利于实时监控生产过程,优化工艺控制,确保系统在安全、经济的工况下运行.本文在改进长喉文丘里管的基础上,设计了一种集近红外(NIR)、声发射(AE)技术于一体的水平气液流量智能多传感系统.利用AE传感器和NIR传感器检测气液两相的流动噪声信息和截面信息,采用经验模态分解法(EMD)提取气体体积分数的特征变量.通过集成学习算法进行特征级融合,融合后的段塞流体积含气率预测模型平均绝对百分比误差(MAPE)为4.11%,92.45%的预测结果偏差在±10%以内.在Collins模型的基础上,提出了基于梯度提升决策树(GBDT)的段塞流质量流量预测模型,其MAPE值为0.96%,全部预测结果的偏差在±20%以内.本研究为气液两相流段塞流参数混合不分离测量提供了一种新方法,为气液两相流动机理研究奠定了基础.Plug flow is a typical flow pattern in gas-liquid two-phase flow,and accurate measurement of plug flow is conducive to real-time monitoring and optimizing process of production process,ensuring safe and economical operation of the system.Based on the improved Venturi tube with long throat diameter,an intelligent multi-sensor system for horizontal gas-liquid flow is designed,which integrates near infrared(NIR)and acoustic emission(AE)technology.AE sensor and NIR sensor were used to detect the gas-liquid phase interaction and disturbance information,and empirical mode decomposition(EMD)was used to extract the characteristic variables of gas volume fraction.The integrated learning algorithm was used for feature-level fusion.The mean absolute percentage error(MAPE)of the fused plug flow volume gas content prediction model was 4.11%,and the deviation of 92.45%of the predicted results was within±10%.On the basis of Collins model,a mass flow prediction model of plug flow based on gradient lifting decision tree(GBDT)is proposed.The MAPE value of GBDT is 0.96%,and the deviation of all prediction results is within±20%.This study provides a new method for measuring the parameters of gas-liquid two-phase plug flow,which provides a research basis for sensing mechanism and measurement of multiphase flow.

关 键 词：气液两相流 数据融合 段塞流 多传感器 集成学习算法 

分 类 号：TH814[机械工程—仪器科学与技术]