化学工业与工程
 首页 |  在线投稿 |  期刊介绍 |  编 委 会 |  投稿指南 |  期刊订阅 |  下载中心 |  出版伦理 |  联系我们 |  English
化学工业与工程 2025, Vol. 42 Issue (1) :130-137    DOI: 10.13353/j.issn.1004.9533.20220317
化工模拟与计算 最新目录 | 下期目录 | 过刊浏览 | 高级检索 << | >>
基于神经网络和CFD的规整填料塔流体力学计算
李邢1, 曾爱武1,2
1. 天津大学化工学院, 天津 300072;
2. 化学工程联合国家重点实验室, 天津 300354
Hydrodynamic calculation of structured packing column based on neural network and CFD
LI Xing1, ZENG Aiwu1,2
1. School of Chemical Engineering and Technology, Tianjin University, Tianjin 300072, China;
2. State Key Laboratory of Chemical Engineering, Tianjin 300354, China

摘要
参考文献
相关文章
Download: PDF (2774KB)   HTML ()   Export: BibTeX or EndNote (RIS)      Supporting Info
摘要 提出一种结合计算流体力学(CFD)和BP (Back Propagation)人工神经网络的多尺度计算方法,计算规整填料塔上的流体力学行为。根据最小特征单元实际尺寸建立小尺度三维CFD模型,研究了填料塔单气相和气液两相流体流动分布方式,弥补了在研究塔壁单元和层间转换单元产生压降方面的缺陷。建立了结点网络模型,计算全塔的流体分布等宏观信息。以CFD计算收集到的数据集训练了2组神经网络模型,分别以结点流量为输入计算干塔压降和持液量。计算结果与相关实验数据比较,干塔压降计算模型的平均相对偏差为8.63%,最大相对偏差为14.02%。持液量计算模型的平均相对偏差约为9.63%,最大相对偏差为13.97%。这表明该训练好的人工神经网络模型具备较好的预测能力,结果较为可信。
Service
把本文推荐给朋友
加入我的书架
加入引用管理器
Email Alert
RSS
作者相关文章
李邢
曾爱武
关键词规整填料   计算流体力学   BP神经网络   干塔压降   持液量     
Abstract: A multi-scale calculation method combining computational fluid dynamics (CFD) and back propagation neural network (BPNN) was proposed to calculate the hydrodynamic behavior of structured packing column. According to the actual size of the representative elementary units, a small-scale 3D CFD model was established, and the flow distribution of single gas phase and gas-liquid two-phase fluid in the packing column was studied, which made up for the defects in the pressure drop of the column wall unit and the interlayer conversion unit. A node network model was established to calculate the macroscopic information such as the fluid distribution of the whole column. Two neural network models were trained with the data set collected by CFD calculation, and the dry pressure drop and liquid holdup were calculated respectively with the node flow as input neuron. Compared with the relevant experimental data, the average relative deviation of the prediction model for dry pressure drop was 8.63% and the maximum relative deviation was 14.2%. The average relative deviation and maximum relative deviation of the prediction model for liquid holdup were 9.63% and 13.97%, respectively. The trained artificial neural network model was proved to be prospective to determinate dry pressure drops and liquid holdup of structured packing columns.
Keywordsstructured packing columns   computational fluid dynamic   back propagation neural network   dry pressure drops   liquid holdup     
Received 2022-04-11;
Corresponding Authors: 曾爱武,副研究员,awzeng@tju.edu.cn     Email: awzeng@tju.edu.cn
About author: 李邢(1996—),女,硕士研究生,现从事规整填料塔流体力学研究。
引用本文:   
李邢, 曾爱武.基于神经网络和CFD的规整填料塔流体力学计算[J].  化学工业与工程, 2025,42(1): 130-137
LI Xing, ZENG Aiwu.Hydrodynamic calculation of structured packing column based on neural network and CFD[J].  Chemcial Industry and Engineering, 2025,42(1): 130-137
Copyright 2010 by 化学工业与工程