基于WOA-CNN-GRU-ATT的起伏振动水平和垂直管气液两相流截面含气率预测

doi:10.11949/0438-1157.20250581

摘要/Abstract

摘要：

海上漂浮核电站被认为是解决海洋能源短缺问题的有效措施之一，但在风浪的作用下处于运动状态，导致蒸汽发生器内气液两相流截面含气率变化，影响运行安全。针对起伏振动气液两相流截面含气率计算模型适用性有限的问题，基于卷积神经网络、门控循环单元、注意力机制和鲸鱼算法建立了起伏振动气液两相流截面含气率预测模型。预测结果表明，该模型对起伏振动水平和垂直上升管截面含气率有良好的适用性，且对流型具有良好的鲁棒性。研究结果为起伏振动下气液两相流截面含气率的精准预测提供了新的方法，为海上核电站蒸汽发生器的参数设计提供理论参考。

关键词: 气液两相流, 气含率, 神经网络, 起伏振动, 优化算法

Abstract:

Floating offshore nuclear power plant is considered a promising solution to address marine energy shortages. However, their oscillatory motion under wind and wave conditions causes fluctuations in the void fraction of gas-liquid two-phase flow within steam generators, posing challenges to operational safety. To overcome the limited applicability of existing models for calculating cross-sectional void fraction under fluctuating vibration, this study proposes a novel prediction model integrating convolutional neural networks (CNN), gated recurrent units (GRU), attention mechanisms, and the whale optimization algorithm (WOA). Experimental results demonstrate that the proposed model has good applicability for predicting void fractions in both horizontal and vertical upward pipes under fluctuating vibration. Additionally, the model shows robust performance across varying flow patterns. This research provides a new method for accurate void fraction prediction in fluctuating gas-liquid two-phase flows, offering theoretical support for design and operational optimization of steam generators of floating nuclear power plant.

Key words: gas-liquid two-phase flow, gas holdup, neural networks, fluctuating vibration, optimization algorithm

中图分类号:

TL 334

刘起超, 张世博, 李昱庆, 周云龙, 冉议文. 基于WOA-CNN-GRU-ATT的起伏振动水平和垂直管气液两相流截面含气率预测[J]. 化工学报, 2025, 76(11): 5900-5910.

Qichao LIU, Shibo ZHANG, Yuqing LI, Yunlong ZHOU, Yiwen RAN. Prediction of void fraction in gas-liquid two-phase flow under fluctuating vibration in horizontal and vertical pipes based on WOA-CNN-GRU-ATT[J]. CIESC Journal, 2025, 76(11): 5900-5910.

图/表 11

图1 水平管实验系统图示意图1—空压机;2—储气罐;3—水箱;4—离心泵;5—减压阀;6—截止阀;7—电磁流量计;8—气体质量流量计;9—球阀;10—针阀;11—相混合器;12—电磁阀;13—实验段;14—高速摄影仪;15—计算机;16—起伏振动台;17—旋风分离器

Fig.1 Schematic diagram of experimental system with horizontal pipe

表1 实验测量参数的相对不确定度

Table1 Relative uncertainty of experimental measurement parameters

测量参数	相对不确定度
水流量	1%~6.667%
空气流量	0.556%~5%
截面含气率	0.47%~7.155%

图2 预测模型结构图

Fig.2 Structure diagram of the prediction model

图3 不同因素对截面含气率Sobol敏感性分析

Fig.3 Sobol sensitivity analysis of different factors on void fraction

表2 模型误差对比

Table2 Comparison of model errors

模型	RMSE	MAPE/%	R²
BPNN	0.0976	17.67	0.7465
SVM	0.0737	13.47	0.8552
CNN-GRT-ATT	0.0389	7.17	0.9676
WOA-C-G-A	0.0368	6.61	0.9709

图4 优化前后预测结果相对误差对比

Fig.4 Comparison of relative errors in prediction results before and after optimization

图5 已有模型和本文新建模型对水平管和垂直管截面含气率预测结果

Fig.5 Comparison of void fraction predictions between existing models and the newly proposed model in this study for horizontal and vertical pipes

图6 水平管预测误差

Fig.6 Prediction errors in horizontal pipes

图7 竖直管预测误差

Fig.7 Prediction errors in vertical pipes

表3 倾斜管截面含气率测试集相对误差

Table 3 Train set error of void fraction in inclined tube

振频/Hz	振幅/m	气相折算速度/(m/s)	液相折算速度/(m/s)	截面含气率实验值/%	截面含气率预测值/%	相对误差/%
4	0.025	0.5	1	16.05	26.82	67.08
4	0.075	1	0.1	48.27	52.25	8.23
0	0	1	0.1	56.03	52.16	-6.90
4	0.025	1	0.5	46.60	42.39	-9.04

表4 垂直上升管截面含气率测试集相对误差

Table 4 Train relative error of void fraction in vertical upward tube

管径/mm	气相折算速度/(m/s)	液相折算速度/(m/s)	截面含气率实验值/%	截面含气率预测值/%	相对误差/%
101.6	5.57059	0.7	71.80	70.92	-1.23
152	4.08715	0.5	76.51	72.68	-5.01
12.7	0.92647	6.32	11.68	14.24	21.95
12.7	0.67281	6.32	9.17	8.97	-2.20
12.7	1.50396	6.32	18.11	17.24	-4.81
101.6	0.30113	0.7	20.28	22.58	11.36
25.4	0.8976	2	24.02	24.38	1.49
52.3	2.57203	1	42.45	55.68	31.15
101.6	10.42862	0.7	82.21	73.06	-11.12
101.6	8.09252	0.7	78.12	72.64	-7.02
52.3	0.57219	1	8.03	9.09	13.23

参考文献 33

[6]	Zhou Y L, Zhao P, Yang N. Experimental study on flow pattern transition of gas liquid two-phase flow in horizontal tubes under vibration condition[J]. Journal of Engineering for Thermal Energy and Power, 2017, 32(6): 17-22.
[7]	周云龙, 李珊珊. 起伏振动状态下倾斜管气液两相流型实验研究[J]. 原子能科学技术, 2018, 52(2): 262-268.
	Zhou Y L, Li S S. Experiment investigation on gas-liquid two-phase flow pattern in inclined pipe under fluctuant vibration condition[J]. Atomic Energy Science and Technology, 2018, 52(2): 262-268.
[8]	周云龙, 汪俊超, 刘起超. 起伏非线性振动下倾斜上升管内气液两相流流型转变分析[J]. 原子能科学技术, 2020, 54(10): 1787-1794.
	Zhou Y L, Wang J C, Liu Q C. Analysis of gas-liquid two-phase flow pattern transition in inclined rising pipe under fluctuant nonlinear vibration condition[J]. Atomic Energy Science and Technology, 2020, 54(10): 1787-1794.
[9]	刘起超, 张世博, 周云龙, 等. 起伏振动水平管气液两相流型及转变机理[J]. 化工学报, 2024, 75(2): 493-504.
	Liu Q C, Zhang S B, Zhou Y L, et al. Gas-liquid two-phase flow regimes and transformation mechanism in horizontal tube under fluctuating vibration[J]. CIESC Journal, 2024, 75(2): 493-504.
[10]	Chen S W, Lee W C, Chang Y H, et al. Experimental investigation on stable/unstable flow behaviors of parallel boiling channels under forced vertical vibrations[J]. Applied Thermal Engineering, 2022, 202: 117840.
[11]	Zhou Y L, Chang H, Lv Y Z. Gas-liquid two-phase flow in a horizontal channel under nonlinear oscillation: flow regime, frictional pressure drop and void fraction[J]. Experimental Thermal and Fluid Science, 2019, 109: 109852.
[12]	Wang L, Zhang Y, Bao Y Y, et al. Numerical study of transient flow characteristics of gas-liquid two-phase flow in inclined upward tube under periodic vibration[J]. Ocean Engineering, 2023, 282: 115024.
[13]	刘起超, 周云龙, 陈聪. 起伏振动垂直上升管气液两相流截面含气率分析与计算[J]. 化工学报, 2023, 74(6): 2391-2403.
	Liu Q C, Zhou Y L, Chen C. Analysis and calculation of void fraction of gas-liquid two-phase flow in vertical riser under fluctuating vibration[J]. CIESC Journal, 2023, 74(6): 2391-2403.
[14]	Zhou Y L, Ran Y W, Liu Q C, et al. Investigation on void fraction of gas-liquid two-phase flow in horizontal pipe under fluctuating vibration[J]. Nuclear Engineering and Design, 2025, 431: 113710.
[15]	Osman E S A. Artificial neural network models for identifying flow regimes and predicting liquid holdup in horizontal multiphase flow[J]. SPE Production & Facilities, 2004, 19(1): 33-40.
[16]	Shippen M E, Scott S L. A neural network model for prediction of liquid holdup in two-phase horizontal flow[J]. SPE Production & Facilities, 2004, 19(2): 67-76.
[17]	Xiao R G, Li K, Sun L Y, et al. The prediction of liquid holdup in horizontal pipe with BP neural network[J]. Energy Science & Engineering, 2020, 8(6): 2159-2168.
[18]	Li C F, Liu M M, Zhao N, et al. Void fraction measurement using modal decomposition and ensemble learning in vertical annular flow[J]. Chemical Engineering Science, 2022, 247: 116929.
[19]	Han B B, Ge B, Wang F, et al. Void fraction detection technology of gas-liquid two-phase bubbly flow based on convolutional neural network[J]. Experimental Thermal and Fluid Science, 2023, 142: 110804.
[20]	Kan Z, Liu X Y. The study on void fraction prediction of gas-liquid two phase flow based on convolutional neural network[J]. Journal of Physics: Conference Series, 2021, 2121(1): 012029.
[21]	Gao Z K, Hou L H, Dang W D, et al. Multitask-based temporal-channel wise CNN for parameter prediction of two-phase flows[J]. IEEE Transactions on Industrial Informatics, 2021, 17(9): 6329-6336.
[22]	Gao Z K, Li M Y, Hou L H, et al. Stage-wise densely connected network for parameter measurement of two-phase flows[J]. IEEE Sensors Journal, 2021, 21(16): 18123-18131.
[23]	Chu W J, Liu Y, Pan L Q, et al. Study on measure approach of void fraction in narrow channel based on fully convolutional neural network[J]. Frontiers in Energy Research, 2021, 9: 636813.
[24]	Zhai H F, Zhou Q, Hu G H. Predicting micro-bubble dynamics with semi-physics-informed deep learning[J]. AIP Advances, 2022, 12(3): 035153.
[25]	Shi Y Y, Sun Y X, Wang M, et al. Flow pattern recognition for horizontal gas-liquid two-phase flow with a hybrid deep-learning model[J]. Flow Measurement and Instrumentation, 2025, 102: 102829.
[1]	Buongiorno J, Jurewicz J, Golay M, et al. The offshore floating nuclear plant concept[J]. Nuclear Technology, 2016, 194(1): 1-14.
[2]	Orr R S, Dotson C. Offshore nuclear power plants[J]. Nuclear Engineering and Design, 1973, 25(3): 334-349.
[3]	Ghadimi P, Bankhshandehrostami A. Determination of the heave and pitch motions of a floating cylinder by analytical solution of its diffraction problem and examination of the effects of geometric parameters on its dynamics in regular waves[J]. International Journal of Applied Mathematical Research, 2012, 1(4):611-633.
[4]	Li S L, Jiang Y Q, Cai W H, et al. Numerical study on condensation heat transfer and pressure drop characteristics of methane upward flow in a spiral pipe under sloshing condition[J]. International Journal of Heat and Mass Transfer, 2019, 129: 310-325.
[5]	Ren Y, Cai W H, Jiang Y Q. Numerical study on shell-side flow and heat transfer of spiral-wound heat exchanger under sloshing working conditions[J]. Applied Thermal Engineering, 2018, 134: 287-297.
[6]	周云龙, 赵盘, 杨宁. 振动状态下水平管内气液两相流流型转变的实验研究[J]. 热能动力工程, 2017, 32(6): 17-22.
[26]	Qiao W L, Guo H, Huang E Z, et al. Two-phase flow pattern identification by embedding double attention mechanisms into a convolutional neural network[J]. Journal of Marine Science and Engineering, 2023, 11(4): 793.
[27]	Baroczy C J. A systematic correlation for two-phase pressure drop[J]. Chemical Engineering Progress Symposium Series, 1966, 62: 234-249.
[28]	Cengel Y A, Cimbala J M. Fluid Mechanics: Fundamentals and Applications[M]. McGraw-Hill Higher Education, 2013: 231-259.
[29]	Massel S R. Ocean Surface Waves: Their Physics and Prediction[M]. Singapore: World Scientific, 1996.
[30]	Moffat R J. Describing the uncertainties in experimental results[J]. Experimental Thermal and Fluid Science, 1988, 1(1): 3-17.
[31]	Mirjalili S, Lewis A. The whale optimization algorithm[J]. Advances in Engineering Software, 2016, 95: 51-67.
[32]	Wang L J, Liu J Y, Yan Y, et al. Gas-liquid two-phase flow measurement using Coriolis flowmeters incorporating artificial neural network, support vector machine, and genetic programming algorithms[J]. IEEE Transactions on Instrumentation and Measurement, 2017, 66(5): 852-868.
[33]	Liu H, Ma Z Y, Zhang L T, et al. Minimum entropy production principle-based semi-empirical model for void fraction prediction in vertical upward dispersed gas-liquid flows[J]. International Journal of Heat and Mass Transfer, 2025, 247: 127120.