基于机器学习的印刷电路板式换热器流动换热预测与仿真

doi:10.11949/0438-1157.20240095

摘要/Abstract

摘要：

基于Zigzag形通道印刷电路板式换热器内跨临界甲烷的流动换热数值模拟结果开展通道内局部对流传热系数与压降机器学习与预测。采用微元分段法提取各通道内局部多物理场参数构建数据库，通过互信息法筛选输入参数，并根据验证集预测效果确定最佳网络结构和超参数。预测结果表明，人工神经网络模型表现最佳，预测对流传热系数的平均绝对百分比误差为2.228%，预测压降则为5.009%。利用机器学习对流动换热参数的预测开发了一种Zigzag形通道印刷电路板式换热器一维仿真方法，实现了通道内流体温度、壁温、对流传热系数和压降的快速准确预测，为换热器设计提供了新的方法。

关键词: 印刷电路板式换热器, 传热, 计算流体力学, 神经网络, 一维仿真

Abstract:

Based on the numerical simulation results of transcritical methane flow in a Zigzag-channel printed circuit plate heat exchanger, machine learning models were used to predict the local convection heat transfer coefficient and pressure drop in the channel. The local multiple physical parameters along the channel were obtained by the microsegment method to create a database. The input parameters are screened by Mutual Information method, and the optimal network structure and hyper parameters are determined according to the predicting effect of validation set. The predicting results show that the artificial neural network model performs best, with a mean absolute percentage error of 2.228% for predicting heat transfer coefficient and 5.009% for predicting pressure drop. Using machine learning to predict flow heat transfer parameters, a one-dimensional simulation method for Zigzag-shaped channel printed circuit board heat exchangers was developed to achieve rapid and accurate prediction of fluid temperature, wall temperature, convective heat transfer coefficient and pressure drop in the channel, providing a new method for heat exchanger design.

Key words: printed circuit plate heat exchanger, heat transfer, computational fluid dynamics, neural network, one-dimensional simulation

中图分类号:

TK 172

李倩, 张蓉民, 林子杰, 战琪, 蔡伟华. 基于机器学习的印刷电路板式换热器流动换热预测与仿真[J]. 化工学报, 2024, 75(8): 2852-2864.

Qian LI, Rongmin ZHANG, Zijie LIN, Qi ZHAN, Weihua CAI. Prediction and simulation of flow and heat transfer for printed circuit plate heat exchanger based on machine learning[J]. CIESC Journal, 2024, 75(8): 2852-2864.

图/表 16

图1 通道几何模型

Fig.1 Channel geometry model

图2 模拟网格

Fig.2 Mesh of simulation domain

图3 不同流量下局部传热系数与压降随温度的变化

Fig.3 Changes of local heat transfer coefficient and pressure drop with temperature at different flow rates

表1 数据库参数及范围

Table 1 Parameters and the range in database

参数	范围
进口温度T_in/K	100.0~349.01
进口压力P_in/Pa	5007448.4~10049058.0
进口速度V_in/(m/s)	0.26~23.83
进口密度 ρ_in/(kg/m³)	35.02~445.58
进口黏度 μ_in/(kg/(m·s))	1.02×10^-5~1.73×10^-4
进口热导率 λ_in/(W/(m·K))	0.03~0.21
质量流率G/(kg/(m²·s))	107.0~954.0
Reynolds数Re	1340.0~78300.0
Prandtl数Pr	0.403~7.92
段长L/m	24.6~28.6
对流传热系数h/(W/(m²·K))	826.94~12097.01
压降 Δp/Pa	79.0~9227.7

图4 基于ANN、DTR、AdaBoost、Catboost模型的示意图

Fig.4 Schematic views of the models based on ANN,DTR,AdaBoost,CatBoost

图5 互信息特征选择方法的结果

Fig.5 Results of mutual information feature selection method

图6 一维仿真模型示意图

Fig.6 One-dimensional simulation diagram

表2 预测h的神经网络结构选择

Table 2 Neural network structure selection for predicting h

序号	网络结构	R²	RMSE	MAPE/%
1	9-18-1	0.9714	456.287	7.864
2	9-18-9-1	0.9913	251.180	3.210
3	9-27-9-1	0.9906	261.555	3.293
4	9-18-36-9-1	0.9938	213.042	2.580
5	9-27-36-9-1	0.9948	194.579	2.456
6	9-18-36-18-1	0.9958	173.984	2.430
7	9-27-36-18-1	0.9972	142.832	1.900
8	9-27-36-18-9-1	0.9967	156.229	2.095

表3 预测Δp的神经网络结构选择

Table 3 Neural network structure selection for predicting Δp

序号	网络结构	R²	RMSE	MAPE/%
1	9-18-1	0.9955	126.071	6.455
2	9-18-9-1	0.9962	113.682	4.722
3	9-27-9-1	0.9973	96.325	4.563
4	9-18-36-9-1	0.9972	96.929	4.495
5	9-27-36-9-1	0.9985	71.435	3.978
6	9-18-36-18-1	0.9976	90.092	4.476
7	9-27-36-18-1	0.9982	77.919	4.054
8	9-27-36-18-9-1	0.9981	80.123	3.546

表4 不同模型预测h选取的超参数

Table 4 Different models predict hyperparameters selected by h

模型	超参数	数值
ANN	activation function	leaky-relu
	learning rate	0.005
	solver	adam
	weight decay	0.01
DTR	max depth	40
	min samples split	14
	min samples leaf	6
AdaBoost	N estimators	56
	max depth	8
	learning rate	0.254
	base estimator	DTR
CatBoost	depth	7
	learning rate	0.126
	L2 leaf reg	1
	iterations	577

表5 不同模型预测Δp选取的超参数

Table 5 Different models predict the hyperparameters selected by Δp

模型	超参数	数值
ANN	activation function	leaky-relu
	learning rate	0.002
	solver	adam
	weight decay	0.01
DTR	max depth	46
	min samples split	9
	min samples leaf	1
AdaBoost	N estimators	89
	max depth	9
	learning rate	0.372
	base estimator	DTR
CatBoost	depth	8
	learning rate	0.033
	L2 leaf reg	5
	iterations	303

图7 传热系数预测结果与测试数据对比

Fig.7 Comparison of heat transfer coefficient prediction results with test data

表6 对比所用的关联式

Table 6 Correlation used for comparison

引用

通道形状

关联式

Chen等^[34]

Zigzag

$N u = (0.05516 ± 0.00160) R e 0.69195 ± 0.00559 (1400 ≤ R e ≤ 2200) (0.09221 ± 0.01397) R e 0.62507 ± 0.01949 (2200 < R e ≤ 3558)$

$f = 17.639 R e 0.8861 ± 0.017 (1400 ≤ R e ≤ 2200) 0.019044 ± 0.001692 (2200 < R e ≤ 3558)$

Kim等^[35]

Zigzag

$N u = 4.089 + 0.00497 R e 0.95 P r 0.55 (0 < R e < 3000,0.66 < P r < 13.41)$

$f p R e = 15.78 + 0.0557 R e 0.82 (0 < R e < 3000)$

表6 对比所用的关联式

Table 6 Correlation used for comparison

引用

通道形状

关联式

Chen等^[34]

Zigzag

$N u = (0.05516 ± 0.00160) R e 0.69195 ± 0.00559 (1400 ≤ R e ≤ 2200) (0.09221 ± 0.01397) R e 0.62507 ± 0.01949 (2200 < R e ≤ 3558)$

$f = 17.639 R e 0.8861 ± 0.017 (1400 ≤ R e ≤ 2200) 0.019044 ± 0.001692 (2200 < R e ≤ 3558)$

Kim等^[35]

Zigzag

$N u = 4.089 + 0.00497 R e 0.95 P r 0.55 (0 < R e < 3000,0.66 < P r < 13.41)$

$f p R e = 15.78 + 0.0557 R e 0.82 (0 < R e < 3000)$

图8 压降预测结果与测试数据的对比

Fig.8 Comparison of pressure drop prediction results with test data

图9 一维仿真结果与CFD模拟结果对比

Fig.9 Comparison of one-dimensional simulation and CFD simulation results

图10 不同流量下传热系数与压降随温度变化的预测结果

Fig.10 Prediction results of heat transfer coefficient and pressure drop with temperature at different flow rates

参考文献 35

1	Zhang G J, Dou L Z, Xu Y. Opportunities and challenges of natural gas development and utilization in China[J]. Clean Technologies and Environmental Policy, 2019, 21(6): 1193-1211.
2	沈妍, 刘锐, 王子, 等. LNG微通道气化器四段设计方法及数值模拟[J]. 油气储运, 2022, 41(3): 324-331.
	Shen Y, Liu R, Wang Z, et al. Four-segmented design method and numerical simulation of LNG micro-channel vaporizer[J]. Oil & Gas Storage and Transportation, 2022, 41(3): 324-331.
3	Huang C Y, Cai W H, Wang Y, et al. Review on the characteristics of flow and heat transfer in printed circuit heat exchangers[J]. Applied Thermal Engineering, 2019, 153: 190-205.
4	黄志豪, 李光熙, 唐桂华, 等. 单侧加热方形通道内超临界水传热研究[J]. 化工学报, 2022, 73(4): 1523-1533.
	Huang Z H, Li G X, Tang G H, et al. Numerical investigation on heat transfer of supercritical water in a side-heated square channel[J]. CIESC Journal, 2022, 73(4): 1523-1533.
5	Kwon D, Jin L X, Jung W, et al. Experimental investigation of heat transfer coefficient of mini-channel PCHE (printed circuit heat exchanger)[J]. Cryogenics, 2018, 92: 41-49.
6	李德凯, 刘乐, 韩辉, 等. 不同通道结构下印刷电路板式换热器流动换热特性[J]. 油气储运, 2023, 42(2): 206-214.
	Li D K, Liu L, Han H, et al. Flow and heat transfer characteristics of printed circuit heat exchangers with different channel structures[J]. Oil & Gas Storage and Transportation, 2023, 42(2): 206-214.
7	谢瑶, 李剑锐, 胡海涛. 印刷电路板式换热器内超临界甲烷流动换热特性模拟[J]. 化工学报, 2021, 72(S1): 203-209.
	Xie Y, Li J R, Hu H T. Simulation of supercritical methane flow and heat transfer characteristics in printed circuit heat exchanger[J]. CIESC Journal, 2021, 72(S1): 203-209.
8	余智强, 吴建泽, 任亚涛, 等. 印刷板式微通道换热器流动与传热特性的理论模型[J]. 化工学报, 2022, 73(12): 5324-5342.
	Yu Z Q, Wu J Z, Ren Y T, et al. Calculation model of flow and heat transfer characteristics of printed microchannel heat exchanger[J]. CIESC Journal, 2022, 73(12): 5324-5342.
9	许婉婷, 许波, 王鑫, 等. 方形微通道内超临界CO₂流动换热特性研究[J]. 化工学报, 2022, 73(4): 1534-1545.
	Xu W T, Xu B, Wang X, et al. Heat transfer characteristics of supercritical CO₂ in square microchannels[J]. CIESC Journal, 2022, 73(4): 1534-1545.
10	Chu W X, Li X H, Ma T, et al. Study on hydraulic and thermal performance of printed circuit heat transfer surface with distributed airfoil fins[J]. Applied Thermal Engineering, 2017, 114: 1309-1318.
11	Yoon S H, No H C, Kang G B. Assessment of straight, Zigzag, S-shape, and airfoil PCHEs for intermediate heat exchangers of HTGRs and SFRs[J]. Nuclear Engineering and Design, 2014, 270: 334-343.
12	Cai W H, Li Y, Li Q, et al. Numerical investigation on thermal-hydraulic performance of supercritical LNG in a Zigzag mini-channel of printed circuit heat exchanger[J]. Applied Thermal Engineering, 2022, 214: 118760.
13	房贤仕, 李秋英, 陈杰, 等. 管内气液两相流换热关联式研究现状[J]. 东北电力大学学报, 2022, 42(1): 71-78.
	Fang X S, Li Q Y, Chen J, et al. A review on heat transfer correlation of gas-liquid two-phase flow in tubes[J]. Journal of Northeast Electric Power University, 2022, 42(1): 71-78.
14	Lan C F, Song B W, Zhang L, et al. State prediction of hydro-turbine based on WOA-RF-Adaboost[J]. Energy Reports, 2022, 8: 13129-13137.
15	王玉坤, 刘吉营, 吴小虎. 电子元件中板翅式散热器梯形翅片的优化研究[J]. 东北电力大学学报, 2023, 43(5): 85-92, 101.
	Wang Y K, Liu J Y, Wu X H. Optimization of trapezoidal fin of plate-fin radiator in electronic components[J]. Journal of Northeast Electric Power University, 2023, 43(5): 85-92, 101.
16	Zhang R Z, Xia Y, Dong J M, et al. Comprehensive analysis of m7G-related genes and chronic hepatitis B: diagnostic markers, immune microenvironment regulation, disease progression[J]. Journal of Immunology Research, 2023, 2023: 9471520.
17	Lan C F, Zhang L, Zhang Y Y, et al. Attention mechanism combined with residual recurrent neural network for sound event detection and localization[J]. EURASIP Journal on Audio, Speech, and Music Processing, 2022, 2022(1): 29.
18	Tian Z, Gu B, Yang L, et al. Performance prediction for a parallel flow condenser based on artificial neural network[J]. Applied Thermal Engineering, 2014, 63(1): 459-467.
19	张仲彬, 李煜, 郭进生, 等. 预测板式换热器污垢热阻的新方法[J]. 东北电力大学学报, 2014, 34(2): 1-6.
	Zhang Z B, Li Y, Guo J S, et al. A new forecasting method of fouling resistance for the plate heat exchanger[J]. Journal of Northeast Dianli University, 2014, 34(2): 1-6.
20	Wu H W, Bagherzadeh S A, D'Orazio A, et al. Present a new multi objective optimization statistical Pareto frontier method composed of artificial neural network and multi objective genetic algorithm to improve the pipe flow hydrodynamic and thermal properties such as pressure drop and heat transfer coefficient for non-Newtonian binary fluids[J]. Physica A: Statistical Mechanics and Its Applications, 2019, 535: 122409.
21	Kwon B, Ejaz F, Hwang L K. Machine learning for heat transfer correlations[J]. International Communications in Heat and Mass Transfer, 2020, 116: 104694.
22	Mehrabi M, Pesteei S M, Pashaee G T. Modeling of heat transfer and fluid flow characteristics of helicoidal double-pipe heat exchangers using Adaptive Neuro-Fuzzy Inference System (ANFIS)[J]. International Communications in Heat and Mass Transfer, 2011, 38(4): 525-532.
23	Hughes M T, Chen S M, Garimella S. Machine-learning-based heat transfer and pressure drop model for internal flow condensation of binary mixtures[J]. International Journal of Heat and Mass Transfer, 2022, 194: 123109.
24	Li Q, Zhan Q, Yu S P, et al. Study on thermal-hydraulic performance of printed circuit heat exchangers with supercritical methane based on machine learning methods[J]. Energy, 2023, 282: 128711.
25	Li Q, Lin Z J, Yang L, et al. Micro segment analysis of supercritical methane thermal-hydraulic performance and pseudo-boiling in a PCHE straight channel[J]. Petroleum Science, 2024, 21(2): 1275-1289.
26	McCulloch W S, Pitts W. A logical calculus of the ideas immanent in nervous activity[J]. Bulletin of Mathematical Biophysics, 1943, 5(4): 115-133.
27	Feng C X J, Gowrisankar A C, Smith A E, et al. Practical guidelines for developing BP neural network models of measurement uncertainty data[J]. Journal of Manufacturing Systems, 2006, 25(4): 239-250.
28	Myles A J, Feudale R N, Liu Y, et al. An introduction to decision tree modeling[J]. Journal of Chemometrics, 2004, 18(6): 275-285.
29	Collins M, Schapire R E, Singer Y. Logistic regression, AdaBoost and bregman distances[J]. Machine Learning, 2002, 48(1): 253-285.
30	Hancock J T, Khoshgoftaar T M. CatBoost for big data: an interdisciplinary review[J]. Journal of Big Data, 2020, 7(1): 94.
31	Huang D, Chow T W S. Effective feature selection scheme using mutual information[J]. Neurocomputing, 2005, 63: 325-343.
32	Nie F, Wang H C, Zhao Y X, et al. A universal correlation for flow condensation heat transfer in horizontal tubes based on machine learning[J]. International Journal of Thermal Sciences, 2023, 184: 107994.
33	张义飞, 刘舫辰, 张双星, 等. 超临界二氧化碳用印刷电路板式换热器性能分析[J]. 化工学报, 2023, 74(S1): 183-190.
	Zhang Y F, Liu F C, Zhang S X, et al. Performance analysis of printed circuit plate heat exchanger for supercritical carbon dioxide[J]. CIESC Journal, 2023, 74(S1): 183-190.
34	Chen M H, Sun X D, Christensen R N, et al. Pressure drop and heat transfer characteristics of a high-temperature printed circuit heat exchanger[J]. Applied Thermal Engineering, 2016, 108: 1409-1417.
35	Kim I H, No H C. Thermal-hydraulic physical models for a printed circuit heat exchanger covering He, He-CO₂ mixture, and water fluids using experimental data and CFD[J]. Experimental Thermal and Fluid Science, 2013, 48: 213-221.

[1]	朱子良, 王爽, 姜宇昂, 林梅, 王秋旺. 欧拉-拉格朗日迭代固-液相变算法[J]. 化工学报, 2024, 75(8): 2763-2776.
[2]	罗正航, 李敬宇, 陈伟雄, 种道彤, 严俊杰. 摇摆运动下低流率蒸汽冷凝换热特性和气泡受力数值模拟[J]. 化工学报, 2024, 75(8): 2800-2811.
[3]	王倩倩, 李冰, 郑伟波, 崔国民, 赵兵涛, 明平文. 氢燃料电池局部动态特征三维模型[J]. 化工学报, 2024, 75(8): 2812-2820.
[4]	毛宇飞, 曹飞, 上官燕琴. 超临界压力流体管内湍流对流传热的计算方法[J]. 化工学报, 2024, 75(8): 2821-2830.
[5]	曹佳蕾, 孙立岩, 曾德望, 尹凡, 高子翔, 肖睿. 双流化床化学链制氢反应器的数值模拟[J]. 化工学报, 2024, 75(8): 2865-2874.
[6]	金虎, 杨帆, 戴梦瑶. 基于格子Boltzmann方法的液滴在圆柱壁面上运动过程研究[J]. 化工学报, 2024, 75(8): 2897-2908.
[7]	吕方明, 包志铭, 王博文, 焦魁. 气体扩散层侵入流道对燃料电池水管理影响研究[J]. 化工学报, 2024, 75(8): 2929-2938.
[8]	杨锦蕊, 郑宏飞, 马兴龙, 金日辉, 梁深. 两级叠置式加湿除湿海水淡化装置性能研究[J]. 化工学报, 2024, 75(7): 2446-2454.
[9]	余清杰, 杨洪海, 刘玉浩, 方海洲, 何伟琪, 王军, 卢心诚. 脉动热管温度信号的小波分析及流型识别[J]. 化工学报, 2024, 75(7): 2497-2504.
[10]	周文轩, 刘珍, 张福建, 张忠强. 高通量-高截留率时间维度膜法水处理机理研究[J]. 化工学报, 2024, 75(7): 2583-2593.
[11]	张香港, 常玉龙, 汪华林, 江霞. 废弃秸秆等生物质低能耗非相变秒级干燥[J]. 化工学报, 2024, 75(7): 2433-2445.
[12]	罗小平, 侯云天, 范一杰. 逆流相分离结构微细通道流动沸腾传热与均温性[J]. 化工学报, 2024, 75(7): 2474-2485.
[13]	王芝安, 兰忠, 马学虎. 喷嘴参数对超临界水热燃烧特性影响的模拟[J]. 化工学报, 2024, 75(6): 2190-2200.
[14]	李新泽, 张双星, 杨洪海, 杜文静. 基于电池冷却用新型脉动热管性能的实验研究[J]. 化工学报, 2024, 75(6): 2222-2232.
[15]	张晗, 张淑宁, 刘珂, 邓冠龙. 基于慢特征分析与最小二乘支持向量回归集成的草酸钴合成过程粒度预报[J]. 化工学报, 2024, 75(6): 2313-2321.