NUCLEAR TECHNIQUES, Volume. 48, Issue 6, 060010(2025)
Three-dimensional reduced-order analysis method for flow and heat transfer performance of PCHE based on Bayesian optimized GPR model
Figures & Tables(17)
Fig. 5. Comparison of temperature field (a), velocity field (b) between FVM calculation results (upper) and reconstruction results (bottom) (color online)
Fig. 6. Comparison of reconstruction results of longitudinal section of temperature field (a), velocity field (b) between FVM calculation results (upper) and reconstruction results (bottom) (color online)
Fig. 7. Comparison of reconstruction results of cross section of temperature field (a),velocity field (b) between FVM calculation results (left) and reconstruction results (right) (color online)
Fig. 8. Comparison of prediction results of interpolation (a) and extrapolation (b) temperature field (color online)
Fig. 9. Comparison of results of interpolation (a) and extrapolation (b) longitudinal section temperature field (color online)
Fig. 10. Comparison of prediction results of interpolation (a) and extrapolation (b) cross section temperature field (color online)
Table 1. Boundary parameter
View tableView in ArticleTable 1. Boundary parameter
流体域
Fluid domain
进口温度
Inlet temperature / K
进口压力
Inlet pressure / MPa
进口流量
Inlet flowrate / kg·s-1
热侧 Hot side 730 7.6 4.82×10-4 冷侧 Cold side 380 20.2
Table 2. Grid verification
View tableView in ArticleTable 2. Grid verification
网格量Grid number 压降Pressure drop / Pa 51万 6 512 78万 6 197 96万 5 990 107万 5 978
Table 3. Value range of sample parameters
View tableView in ArticleTable 3. Value range of sample parameters
流体域Fulid domian 进口温度Inlet temperature / K 进口流量Inlet flowrate / kg·s-1 热侧 Hot side 640~780 (4.82~14.45)×10-4 冷侧 Cold side 300~350
Table 4. Comparison of decomposition algorithms
View tableView in ArticleTable 4. Comparison of decomposition algorithms
降阶模型
Reduced-order model
分解模型
Decomposition model
基函数数量
Number of basis
functions / order
分解耗时
High decomposition time-consumption / s
重构误差(RMSE)
Reconstruction error
数据量
Data size
融合场降阶模型
Fusion field reduction model
rPOD 46 6 0.001 1 130.2×104×60 tPOD 46 0.9 0.001 1 基础降阶模型
Basic reduced-order model
rPOD 39 (T)+50 (V) 4.23 (T)+0.72 (V) 0.001 3 (T)+0.000 2 (V) 94.7×104×60
+35.5×104×60
Average: 0.001 0 tPOD 39 (T)+50 (V) 0.67 (T)+0.21 (V) 0.001 3 (T)+0.000 2 (V) Average: 0.001 0
Table 5. Proxy model effect
View tableView in ArticleTable 5. Proxy model effect
代理模型
Delegation model
MAE RMSE R2 SVM 0.031 1 0.061 1 0.78 可优化GPR
GPR can be optimized
0.018 1 0.026 8 0.94 可优化树集成
Optimized tree integration
0.049 2 0.064 9 0.76 BP神经网络
BP neural network
0.041 2 0.066 1 0.82 RBF神经网络
RBF neural network
0.035 5 0.065 7 0.75
Table 6. GPR model hyperparameters
View tableView in ArticleTable 6. GPR model hyperparameters
模型
Model
标准化
Standardization
核函数
Kernel function
核尺度
Kernel scale
Sigma GPR Yes Nonisotropic Matern 5/2 3.380 1 1.205 2
Table 7. Performance of reduced-order model
View tableView in ArticleTable 7. Performance of reduced-order model
计算模型
Calculation model
研究对象
Research object
均方根误差MSE 计算效率提升倍数
The calculation efficiency is
increased by multiples
重构
Reconstruction
内推
Interpolate
外推
Extrapolate
融合场降阶模型
Fusion field
reduction model
温度场
Temperature field
4.58×10-4 0.001 9 0.008 270 速度场
Velocity field
4.6×10-4 0.002 2 0.008 评价因子
Evaluation factor
0.004 6 0.022 5 0.138
Tools
Get Citation
Copy Citation Text
Ziyan ZHAO, Congyi WEN, Pengcheng ZHAO, Zijing LIU. Three-dimensional reduced-order analysis method for flow and heat transfer performance of PCHE based on Bayesian optimized GPR model[J]. NUCLEAR TECHNIQUES, 2025, 48(6): 060010
Paper Information
Category: Special Topics of Academic Papers at The 27th Annual Meeting of the China Association for Science and Technology
Received: Aug. 15, 2024
Accepted: --
Published Online: Jul. 25, 2025
The Author Email: Zijing LIU (刘紫静)
© Copyright 2018-2021 | Chinese Laser Press.
All Rights Reserved 沪ICP备15018463号-20