3D High-resolution Flow Reconstruction Based on Deep Learning

doi:10.3901/JME.2025.16.338

Abstract

Abstract: Obtaining three-dimensional high-resolution flow field in mechanical structures is especially critical for mechanical design and optimization. However, due to the constraints on the size of mechanical structures and sensors, it is difficult to obtain the 3D high-resolution flow field data directly, and the deep learning method based on 2D convolution cannot handle the 3D time-varying flow field data directly. To address the above problems, this paper proposes a deep learning reconstruction method for 3D high-resolution flow field based on time-varying flow samples at finite measurement points. The high Reynolds number (Re=2.2×10⁴) flow around square cylinder were modeled and trained. A low-dimensional representation model of the 3D turbulent field is obtained, and then a high-precision coordinate-representation coding mapping is constructed to achieve a 3D high-resolution reconstruction model of the flow data, which is in good agreement with the target value. It is demonstrated that the proposed method can realize the high-resolution reconstruction of the 3D turbulent field, which can provide a new method for the analysis of time-history data based on the measurement sensors in mechanical flow problems.

Key words: 3D flow reconstruction, flow time-history deep learning, high resolution flow, flow representation model

CLC Number:

O375

ZHAN Qingliang, LIU Xin, BAI Chunjin, GE Yaojun. 3D High-resolution Flow Reconstruction Based on Deep Learning[J]. Journal of Mechanical Engineering, 2025, 61(16): 338-346.

References

[1] 宁超,袁寿其,李亚林,等. 相同叶栅稠密度下叶片数对复合式叶轮内流场的影响[J]. 机械工程学报, 2022, 58(14): 320-327. NING Chao, YUAN Shouqi, LI Yalin, et al. Effect of blade number on flow field in compound impeller with same cascade density[J]. Journal of Mechanical Engineering, 2022, 58(14): 320-327.
[2] 周伟,朱鑫宁,连云崧,等. 质子交换膜燃料电池的三维流场技术研究进展[J]. 机械工程学报, 2021, 57(8): 2-12. ZHOU Wei , ZHU Xinning, LIAN Yunsong, et al. Research progress of three-dimensional flow field technology for proton exchange membrane fuel cells[J]. Journal of Mechanical Engineering, 2021, 57(8): 2-12.
[3] 张文武,余志毅,李泳江,等. 叶片式气液混输泵全流道内流场特性分析[J]. 机械工程学报, 2019, 55(10): 168-174. ZHANG Wenwu, YU Zhiyi, LI Yongjiang, et al. Analysis of flow field characteristics in full channel of vane gas-liquid mixed transport pump[J]. Journal of Mechanical Engineering, 2019, 55(10): 168-174.
[4] 蔡华闽,张继业,李田. 高速列车转向架区域气动特性及流场规律研究[J]. 机械工程学报, 2018, 54(12): 49-57. CAI Huamin, ZHANG Jiye, LI Tian. Research on regional aerodynamic characteristics and flow field of high-speed train bogies[J]. Journal of Mechanical Engineering, 2018, 54(12): 49-57.
[5] 胡岳,张涛. 分离涡流场数值仿真的参数影响研究[J]. 机械工程学报, 2016, 52(12): 165-172. HU Yue, ZHANG Tao. Research on parameter influence of numerical simulation of separation eddy current field[J]. Journal of Mechanical Engineering , 2016 , 52(12) : 165-172.
[6] 王松涛,丁骏,姜斌,等. 三维叶片技术对离心风机流场结构的影响[J]. 机械工程学报 , 2014 , 50(4) : 178-184. WANG Songtao, DING Jun, JIANG Bin, et al. Effect of three-dimensional blade technology on flow field structure of centrifugal fan[J]. Journal of Mechanical Engineering, 2014, 50(4): 178-184.
[7] BERKOOZ G, P HOLMES A, LUMLEY J L. The proper orthogonal decomposition in the analysis of turbulent flows[J]. Annual Review of Fluid Mechanics, 1993, 25(1): 539-575.
[8] SIROVICH L. Turbulence and the dynamics of coherent structures. I. Coherent structures[J]. Quarterly of Applied Mathematics, 1987, 45(3): 561-571.
[9] SCHMID P J , SESTERHENN J. Dynamic mode decomposition of numerical and experimental data[J]. Journal of Fluid Mechanics, 2010, 656(10): 5-28.
[10] ROWLEY C W, MEⅡĆ I, BAGHERI S, et al. Spectral analysis of nonlinear flows[J]. Journal of Fluid Mechanics, 2009, 641: 115-127.
[11] 金晓威,赖马树金,李惠. 物理增强的流场深度学习建模与模拟方法[J]. 力学学报, 2021, 53: 2616. JIN Xiaowei, LAI Mashujin, LI Hui. Physically-enhanced deep learning modeling and simulation of flow fields[J]. Chinese Journal of Mechanical Mechanics, 2021, 53: 2616.
[12] LING J, KURZAWSKI A, TEMPLETON J. Reynolds averaged turbulence modelling using deep neural networks with embedded invariance[J]. Journal of Fluid Mechanics, 2016, 807: 155-166.
[13] MURATA T, FUKAMI K, FUKAGATA K. Nonlinear mode decomposition with convolutional neural networks for fluid dynamics[J]. Journal of Fluid Mechanics, 2020, 882: A13.
[14] RAISSI M , KARNIADAKIS G E. Hidden physics models: Machine learning of nonlinear partial differential equations[J]. Journal of Computational Physics, 2018, 357: 125-141.
[15] KIM H, KIM J, WON S, et al. Unsupervised deep learning for super-resolution reconstruction of turbulence[J]. Journal of Fluid Mechanics, 2021, 910: A29.
[16] MURATA T, FUKAMI K, FUKAGATA K. Nonlinear mode decomposition with convolutional neural networks for fluid dynamics[J]. Journal of Fluid Mechanics, 2020, 882: A13.
[17] OMATA N , SHIRAYAMA S. A novel method of low-dimensional representation for temporal behavior of flow fields using deep autoencoder[J]. AIP Advances, 2019, 9(1): 15006.
[18] KAI F , FUKAGATA K , TAIRA K. Machinelearning-based spatio-temporal super resolution reconstruction of turbulent flows[J]. Journal of Fluid Mechanics, 2021, 909: A9.
[19] LIU B, TANG J, HUANG H, et al. Deep learning methods for super-resolution reconstruction of turbulent flows[J]. Physics of Fluids, 2020, 32(2): 25105.
[20] JIN X, CHENG P, CHEN W, et al. Prediction model of velocity field around circular cylinder over various Reynolds numbers by fusion convolutional neural networks based on pressure on the cylinder[J]. Physics of Fluids, 2018, 30(4): 47105.
[21] CALLAHAM J, MAEDA K, BRUNTON S L. Robust flow field reconstruction from limited measurements via sparse representation[J]. Phys. Rev. Fluids, arXiv.1810. 06723, 2018.
[22] FUKAMI K, MAULIK R, RAMACHANDRA N, et al. Global field reconstruction from sparse sensors with Voronoi tessellation-assisted deep learning[J]. Nature Machine Intelligence, 2021, 3(11): 945-951.
[23] ERICHSON N B, MATHELIN L, YAO Z, et al. Shallow neural networks for fluid flow reconstruction with limited sensors[J]. Proceedings of The Royal Society A Mathematical Physical and Engineering Sciences, 2020, 476(2238): 20200097.
[24] DENG Z, CHEN Y, LIU Y, et al. Time-resolved turbulent velocity field reconstruction using a long short-term memory (LSTM)-based artificial intelligence framework[J]. Physics of Fluids, 2019, 31(7): 75108.
[25] HAN R , WANG Y , ZHANG Y , et al. A novel spatial-temporal prediction method for unsteady wake flows based on hybrid deep neural network[J]. Physics of Fluids, 2019, 31(12): 127101.
[26] FUKAMI K, FUKAGATA K, TAIRA K. Assessment of supervised machine learning methods for fluid flows[J]. Theoretical and Computational Fluid Dynamics, 2020, 34(4): 497-519.
[27] 战庆亮,白春锦,葛耀君. 基于时程深度学习的流场特征分析方法[J]. 力学学报, 2022, 54(3): 822-828. ZHAN Qingliang, BAI Chunjin, GE Yaojun. Flow field characteristics analysis method based on time-history deep learning[J]. Chinese Journal of Theoretical and Applied Mechanics, 2022, 54(3): 822-828.
[28] 战庆亮,白春锦,张宁,等. 基于时程卷积自编码的机翼绕流特征识别方法[J]. 航空学报, 2022, 43(11): 126531. ZHAN Qingliang, BAI Chunjin, ZHANG Ning, et al. Feature recognition method for wing flow around a plane based on time-history convolutional self-coding[J]. Acta Aeronautica Sinica, 2022, 43(11): 126531.
[29] 战庆亮,葛耀君,白春锦. 基于深度学习的流场时程特征提取模型研究[J]. 物理学报, 2022, 71(7): 225-234. ZHAN Qingliang, GE Yaojun, BAI Chunjin. Research on time-history feature extraction model of flow field based on deep learning[J]. Acta Physica Sinica, 2022, 71(7): 225-234.
[30] 战庆亮,白春锦,葛耀君. 基于时程深度学习的桥面绕流表征与重构方法[J]. 工程力学, 2022, 40(9): 13-19. ZHAN Qingliang , BAI Chunjin , GE Yaojun. Characterization and reconstruction method of flow around bridge deck based on time-history deep Learning[J]. Engineering Mechanics, 2022, 40(9): 13-19.
[31] 战庆亮,周志勇,葛耀君. Re=3 900 圆柱绕流的三维大涡模拟[J]. 哈尔滨工业大学学报, 2015, 47(12): 75-79. ZHAN Qingliang , ZHOU Zhiyong , GE Yaojun. Three-dimensional large eddy Simulation of flow around a Re=3 900 cylinder[J]. Journal of Harbin Institute of Technology, 2015, 47(12): 75-79.
[32] LYN D A, EINAV S, RODI W, et al. A laser-doppler velocimetry study of ensemble-averaged characteristics of the turbulent near wake of a square cylinder[J]. Journal of Fluid Mechanics, 2006, 304(304): 285-319.
[33] LYN D A, RODI W. The flapping shear layer formed by flow separation from the forward corner of a square cylinder[J]. Journal of Fluid Mechanics, 1994, 267: 353-376.
[34] GRIGORIADIS D G E, BARTZIS J G, GOULAS A. LES of the flow past a rectangular cylinder using the immersed boundary concept[J]. International Journal for Numerical Methods in Fluids, 2003, 41(6): 615-632.
[35] LUO S C, YAZDANI M G, CHEW Y T, et al. Effects of incidence and afterbody shape on flow past bluff cylinders[J]. Journal of Wind Engineering and Industrial Aerodynamics, 1994, 53(3): 375-399.