拉索方法对脉冲折流板萃取柱存留分数的预测  

Prediction of hold-up in pulsed disc-doughnut column based on least absolute shrinkage and selection operator regression

作　　者：金磊 高杨[1] 何辉 谢滨骏龙[1] 周羽 张萌[1] JIN Lei;GAO Yang;HE Hui;XIE Bingjunlong;ZHOU Yu;ZHANG Meng(College of Nuclear Science and Technology,Harbin Engineering University,Harbin 150001,China;China Institute of Atomic Energy,Beijing 102413,China)

机构地区：[1]哈尔滨工程大学核科学与技术学院,黑龙江哈尔滨150001 [2]中国原子能科学研究院,北京102413

出　　处：《哈尔滨工程大学学报》2023年第12期2151-2156,共6页Journal of Harbin Engineering University

基　　金：乏燃料后处理科研专项.

摘　　要：为了获得适用性好,且便于分析的脉冲折流板萃取柱内存留分数变化的预测方程,本文使用拉索回归方法拟合分析了通过体积置换法测试的脉冲折流板萃取柱中煤油-水体系下的脉冲强度、分散相流速和连续相流速3个操作参数对存留分数的影响。并对比分析了相关文献中的多组经验关系式重新拟合的方程与使用拉索回归方法拟合出的多元二次形式的预测方程。结果表明:拉索回归方法拟合出的预测方程能够很好的贴合实验数据,并且通过分析该方程对应3个操作参数的偏导数,定量地分析不同参数在具体的操作范围内对存留分数的影响大小。本文方法同时能够很好地适用于其他文献的数据当中,拟合结果与实验数据的平均相对误差都在20%以内,能够简单直观地分析各个参数对存留分数的作用大小。To obtain a predictive equation that provides high applicability and analytical convenience,based on the least absolute shrinkage and selection operator(LASSO).Investigate the effects of pulse intensity,disperse phase velocity,and continuous phase velocity on the hold-up in a pulsed disc-doughnut column with a kerosene-water sys-tem using the volume replacement method.Subsequently,several sets of empirical equations from related literature were compared and analyzed with new multivariate quadratic prediction equations fitted using the LASSO method.The results revealed that the prediction equations fitted using the LASSO method fit the experimental data extremely well,and the effect of the parameter on the hold-up in the specific operating range could be quantitatively analyzed by calculating the partial derivatives of the equation for each of the three operating parameters.The results also a-dapted very well to other literature data,and the average relative deviation of the fitted results from the experimental data was within 20%.This demonstrates that LASSO is a simple and intuitive way to analyze the effect of each pa-rameter on the hold-up.

关 键 词：拉索方法 存留分数 脉冲折流板萃取柱 液液两相流 煤油-水体系 水力学性能 机器学习 体积置换法 

分 类 号：TL24[核科学技术—核燃料循环与材料]