Journal of Jilin University(Engineering and Technology Edition) ›› 2024, Vol. 54 ›› Issue (4): 874-882.doi: 10.13229/j.cnki.jdxbgxb.20220611

Previous Articles     Next Articles

Reduced-order modelling of a bluff body turbulent wake flow field using hierarchical convolutional neural network autoencoder

Chao XIA1,2(),Meng-jia WANG1,2,Jian-yue Zhu3(),Zhi-gang YANG1,2,4   

  1. 1.School of Automotive Studies，Tongji University，Shanghai 201804，China
    2.Shanghai Automotive Wind Tunnel Center，Tongji University，Shanghai 201804，China
    3.Institute of Rail Transit，Tongji University，Shanghai 201804，China
    4.Beijing Aeronautical Science & Technology Research Institute，Beijing 102211，China
  • Received:2022-05-19 Online:2024-04-01 Published:2024-05-17
  • Contact: Jian-yue Zhu E-mail:chao.xia@tongji.edu.cn;zhujianyue@tongji.edu.cn

RICH HTML

 9   

PDF (PC)

338

Abstract:

In this study， a nonlinear hierarchical convolutional autoencoder （H-CNN-AE） is employed to sort the energy content of the latent vectors of AE， which is analogous to the method of proper orthogonal decomposition （POD）， and at the same time result in better performance in terms of reduced-order modelling within limits. This method is applied to a turbulent wake behind a three-dimensional circular cylinder bluff body at Re=20 000. We assess the ability of H-CNN-AE with L2 error and compares it with the results of POD. Furthermore， the concept of grouping AE-modes is extended. We observe the change of mean square error between the reconstructed and the original flow when adding the number of latent vectors of each group. It is demonstrated that when the number of subnetworks and low-dimensional vectors in latent space of each subnetwork is small， H-CNN-AE has better capability to restore the flow field than POD. However， it is also found that the strength of H-CNN-AE will weaken with the increase of the number of subnetworks and latent AE modes and will even be inferior to POD under certain conditions.

Key words: hierarchical autoencoder, convolutional neural network, bluff body turbulent wake, reduced order model

CLC Number: 

  • U462

Fig.1

Conventional convolutional neural network based autoencoder"

Fig.2

Hierarchical Autoencoder"

Fig.3

Computational case of a flow around 3D circular cylinder"

Table 1

Structure of CNN based hierarchical autoencoder"

网络层No.输出图形
Input1（192，384，2）
Input21（1）
Conv（f=16）1（192，384，16）
MaxPooling（192，384，16）
Conv（f=8）3（24，48，8）
MaxPooling（12，24，8）
Conv（f=4）1（12，24，4）
MaxPooling（6，12，4）
Conv（f=2）1（6，12，2）
Reshape1（144）
Dense4（64）
（32）
（16）
（1）
Concatenate1（2）
Dense4（16）
（32）
（64）
（144）
Reshape1（6，12，2）
Conv（f=2）1（6，12，2）
Upsampling1（12，24，2）
Conv（f=4）（12，24，4）
Upsampling3（96，192，8）
Conv（f=8）（96，192，8）
Upsampling1（192，384，8）
Conv（f=16）（192，384，16）
Conv（f=2）1（192，384，2）

Fig.4

Training error of network"

Table2

Time-averaged L2 error of different activation functions"

激活函数L2误差
softsign1.045
tanh1.094
Relu2.209

Fig.5

Comparison of reconstruction results between H-CNN-AE and POD"

Fig.6

Comparison of mode of streamwise velocity u between H-CNN-AE and POD"

Fig.7

POD processing of AE modes"

Fig.8

Reconstructed flow field L2 error at different n and ne"

Fig.9

Reconstructed flow field L2 error with four subnetworks"

1 Muralidhar S D, Podvin B, Mathelin L, et al. Spatio-temporal proper orthogonal decomposition of turbulent channel flow[J]. Journal of Fluid Mechanics, 2019(864): 614-639.
2 Krizhevsky A, Sutskever I, Hinton G E. Imagenet classification with deep convolutional neural networks[J]. Advances in Neural Information Processing Systems, 2012(25): 1097-1105.
3 Hinton G, Deng L, Yu D, et al. Deep neural networks for acoustic modeling in speech recognition: The shared views of four research groups[J]. IEEE Signal Processing Magazine, 2012, 29(6): 82-97.
4 Karpathy A, Toderici G, Shetty S, et al. Large-scale video classification with convolutional neural networks[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition,Columbus,USA,2014: 1725-1732.
5 Silver D, Schrittwieser J, Simonyan K, et al. Mastering the game of go without human knowledge[J]. Nature, 2017, 550(7676): 354-359.
6 叶舒然, 张珍, 王一伟,等. 基于卷积神经网络的深度学习流场特征识别及应用进展[J]. 航空学报, 2021, 42(4):185-199.
Ye Shu-ran, Zhang Zhen, Wang Yi-wei, et al. Progress in deep convolutional neural network based flow field recognition and its applications[J]. Acta Aeronautica et Astronautica Sinica, 2021, 42(4):185-199.
7 Kutz J N. Deep learning in fluid dynamics[J]. Journal of Fluid Mechanics, 2017(814):1-4.
8 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.
9 Lecun Y, Bottou L, Bengio Y, et al. Gradient-based learning applied to document recognition[J]. Proceedings of the IEEE, 1998, 86(11): 2278-2324.
10 Hinton G E, Salakhutdinov R R. Reducing the dimensionality of data with neural networks[J]. Science, 2006, 313(5786): 504-507.
11 Kai F, Fukagata K, Taira K. Super-resolution reconstruction of turbulent flows with machine learning[J]. Journal of Fluid Mechanics, 2019( 870):106-120.
12 叶舒然, 张珍, 宋旭东,等. 自动编码器在流场降阶中的应用[J]. 空气动力学学报, 2019, 37(3):498-504.
Ye Shu-ran, Zhang Zhen, Song Xu-dong, et al. Applications of autoencoder in reduced-order modeling of flow field[J]. Acta Aerodynamica Sinica, 2019, 37(3):498-504.
13 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):No.015006.
14 Murata T, Kai F, Fukagata K. Nonlinear mode decomposition with convolutional neural networks for fluid dynamics[J]. Journal of Fluid Mechanics, 2020, 882: No.A13.
15 Kai F, Hasegawa K, Nakamura T, et al. Model order reduction with neural networks: application to laminar and turbulent flows[J].SN Computer Science, 2021, 2(6):No.467.
16 Kai F, Nakamura T, Fukagata K. Convolutional neural network based hierarchical autoencoder for nonlinear mode decomposition of fluid field data[J]. Physics of Fluids, 2020, 32(9):No.95110.
17 杜祥波, 陈少强, 侯靖尧,等. 基于卷积神经网络的钝体尾迹识别研究[J]. 力学学报, 2022,54(1):59-67.
Du Xiang-bo, Chen shao-qiang, Hou jing-yao, et al. Wake recognition of a blunt body based on convolutional neural network[J]. Theoretical and Applied Mechanics, 2022,54(1):59-67.
18 Le Q, Ngiam J, Chen Z, al et, Tiled convolutional　neural networks [J]. Adv Neural Inf Proc Syst,2010,23:1279-1287.
19 Saegusa R, Sakano H, Hashimoto S. Nonlinear principal component analysis to preserve the order of principal components[J]. Neurocomputing, 2004,61:57-70.
20 Chu S J, Xia C, Wang H F, et al. Three-dimensional spectral proper orthogonal decomposition analyses of the turbulent flow around a seal-vibrissa-shaped cylinder[J]. Physics of Fluids, 2021, 33(2):No.25106.
21 Sirovich L. Turbulence and the dynamics of coherent structures. I. Coherent structures[J]. Quarterly of Applied Mathematics, 1987, 45(3): 561-571.
[1] Guo-jun YANG,Ya-hui QI,Xiu-ming SHI. Review of bridge crack detection based on digital image technology [J]. Journal of Jilin University(Engineering and Technology Edition), 2024, 54(2): 313-332.
[2] Xiang-jiu CHE,Huan XU,Ming-yang PAN,Quan-le LIU. Two-stage learning algorithm for biomedical named entity recognition [J]. Journal of Jilin University(Engineering and Technology Edition), 2023, 53(8): 2380-2387.
[3] Zhen-hai ZHANG,Kun JI,Jian-wu DANG. Crack identification method for bridge based on BCEM model [J]. Journal of Jilin University(Engineering and Technology Edition), 2023, 53(5): 1418-1426.
[4] Pei-yong LIU,Jie DONG,Luo-feng XIE,Yang-yang ZHU,Guo-fu YIN. Surface defect detection algorithm of magnetic tiles based on multi⁃branch convolutional neural network [J]. Journal of Jilin University(Engineering and Technology Edition), 2023, 53(5): 1449-1457.
[5] Feng-le ZHU,Yi LIU,Xin QIAO,Meng-zhu HE,Zeng-wei ZHENG,Lin SUN. Analysis of hyperspectral image based on multi-scale cascaded convolutional neural network [J]. Journal of Jilin University(Engineering and Technology Edition), 2023, 53(12): 3547-3557.
[6] Hai-long GAO,Yi-bo XU,De-zao HOU,Xue-song WANG. Short⁃term traffic flow prediction algorithm for road network based on deep asynchronous residual network [J]. Journal of Jilin University(Engineering and Technology Edition), 2023, 53(12): 3458-3464.
[7] Jun-jie WANG,Yuan-jun NONG,Li-te ZHANG,Pei-chen ZHAI. Visual relationship detection method based on construction scene [J]. Journal of Jilin University(Engineering and Technology Edition), 2023, 53(1): 226-233.
[8] Xuan-jing SHEN,Xue-feng ZHANG,Yu WANG,Yu-bo JIN. Multi⁃focus image fusion algorithm based on pixel⁃level convolutional neural network [J]. Journal of Jilin University(Engineering and Technology Edition), 2022, 52(8): 1857-1864.
[9] Ming-hua GAO,Can YANG. Traffic target detection method based on improved convolution neural network [J]. Journal of Jilin University(Engineering and Technology Edition), 2022, 52(6): 1353-1361.
[10] Huai-jiang YANG,Er-shuai WANG,Yong-xin SUI,Feng YAN,Yue ZHOU. Simplified residual structure and fast deep residual networks [J]. Journal of Jilin University(Engineering and Technology Edition), 2022, 52(6): 1413-1421.
[11] Xue-zhi WANG,Qing-liang LI,Wen-hui LI. Spatio⁃temporal model of soil moisture prediction integrated with transfer learning [J]. Journal of Jilin University(Engineering and Technology Edition), 2022, 52(3): 675-683.
[12] Xiang-jun LI,Jie-ying TU,Zhi-bin ZHAO. Validity classification of melting curve based on multi⁃scale fusion convolutional neural network [J]. Journal of Jilin University(Engineering and Technology Edition), 2022, 52(3): 633-639.
[13] Xian-tong LI,Wei QUAN,Hua WANG,Peng-cheng SUN,Peng-jin AN,Yong-xing MAN. Route travel time prediction on deep learning model through spatiotemporal features [J]. Journal of Jilin University(Engineering and Technology Edition), 2022, 52(3): 557-563.
[14] Wen-zhi GAO,Yan-jun WANG,Xin-wei WANG,Pan ZHANG,Yong LI,Yang DONG. Real⁃time diagnosis for misfire fault of diesel engine based on convolutional neural network [J]. Journal of Jilin University(Engineering and Technology Edition), 2022, 52(2): 417-424.
[15] Long ZHANG,Tian-peng XU,Chao-bing WANG,Jian-yu YI,Can-zhuang ZHEN. Gearbox fault diagnosis baed on convolutional gated recurrent network [J]. Journal of Jilin University(Engineering and Technology Edition), 2022, 52(2): 368-376.
Viewed
Full text


Abstract

Cited
  Shared   
  Discussed   
No Suggested Reading articles found!
Copyright © Journal of Jilin University(Engineering and Technology Edition), All Rights Reserved.
Tel: +86-431-85095297　Fax: +86-431-85094074　E-mail: xbgxb@jlu.edu.cn
Powered by Beijing Magtech Co. Ltd