基于深度注意力网络的压气机流场重构方法

doi:10.7527/S1000-6893.2024.30580

Abstract

Abstract:

The rapid and accurate prediction of compressor performance and key flow field characteristics is critical for digital design， digital twin modeling， and virtual-real interaction of compressors. This study focuses on the first-stage stator of a transonic three-stage compressor. A flow field database was established using high-precision numerical simulation results for 205 different operating conditions at various speeds. We employed a symmetric convolutional neural network with attention mechanisms to reconstruct flow field parameters such as static temperature， static pressure， and Mach number at different radii of the stator blade. This network comprises a visual self-attention model and symmetric convolutional neural network. The former extracts geometric features of blade passages at different radii under various operating conditions， while the latter conducts deeper feature extraction of these features and other inputs （including inlet and outlet boundary conditions， flow field coordinates， and distance fields） through transposed convolution and convolution operations； consequently， this network predicts the flow field at different radial positions. Our findings demonstrate that the symmetric convolutional neural network based on deep attention can rapidly and effectively predict the internal flow field of compressors. The model accurately captures changes in physical parameters near the blade wall， blade wakes， and flow separation. Compared with high-precision numerical simulation， the model achieves rapid and accurate reconstruction of the flow field of compressor stators with an average relative error not exceeding 1%.

Key words: compressor stator, deep learning, visual self-attention mechanism, symmetric convolution, flow field reconstruction

CLC Number:

V231.3

Yueteng WU, Dun BA, Juan DU, Yunfei LI, Juntao CHANG. Compressor flow field reconstruction method based on deep attention networks[J]. Acta Aeronautica et Astronautica Sinica, 2024, 45(24): 630580.

Figures/Tables 24

Fig.1

Fig.2

Table 1

Fig.3

Fig.4

Fig.5

Fig.6

Fig.7

Fig.8

Fig.9

Fig.10

Table 2

Fig.11

Fig.12

Table 3

Fig.13

Table 4

Fig.14

Table 5

Fig.15

Fig.16

Fig. 17

Fig. 18

Fig. 19

References 28

1	张伟伟，寇家庆，刘溢浪. 智能赋能流体力学展望［J］. 航空学报， 2021， 42（4）： 524689.
	Zhang W W， Kou J Q， Liu Y L. Intelligent empowerment in fluid dynamics perspective［J］. Acta Aeronautica et Astronautica Sinica， 2021， 42（4）： 524689 （in Chinese）.
2	张伟伟，朱林阳，刘溢浪，等. 机器学习在湍流模型构建中的应用进展［J］. 空气动力学学报， 2019， 37（3）： 444-454.
	ZHANG W W， ZHU L Y， LIU Y L， et al. Progresses in the application of machine learning in turbulence modeling［J］. Acta Aerodynamica Sinica， 2019， 37（3）： 444-454 （in Chinese）.
3	陈海昕，邓凯文，李润泽. 机器学习技术在气动优化中的应用［J］. 航空学报， 2019， 40（1）： 522480.
	CHEN H X， DENG K W， LI R Z. Utilization of machine learning technology in aerodynamic optimization［J］. Acta Aeronautica et Astronautica Sinica， 2019， 40（1）： 522480 （in Chinese）.
4	LUI H F S， WOLF W R. Construction of reduced-order models for fluid flows using deep feedforward neural networks［J］. Journal of Fluid Mechanics， 2019， 872： 963-994.
5	陈皓，郭明明，田野，等.卷积神经网络在流场重构研究中的进展［J］.力学学报， 2022， 54（9）： 2343-2360.
	Chen H， Guo M M， Tian Y， et al. The progress of convolutional neural networks in flow field reconstruction research. Chinese Journal of Theoretical and Applied Mechanics， 2022， 54（9）： 2343-2360 （in Chinese）.
6	YANG Z S， DONG Y D， DENG X G， et al. AMGNET： multi-scale graph neural networks for flow field prediction［J］. Connection Science， 2022， 34（1）： 2500-2519.
7	CAO W B， SONG J H， ZHANG W W. A solver for subsonic flow around airfoils based on physics-informed neural networks and mesh transformation［J］. Physics of Fluids， 2024， 36（2）： 027134.
8	WANG J， HE C， LI R Z， et al. Flow field prediction of supercritical airfoils via variational autoencoder based deep learning framework［J］. Physics of Fluids， 2021， 33（8）： 086108.
9	SABATER C， STÜRMER P， BEKEMEYER P. Fast predictions of aircraft aerodynamics using deep-learning techniques［J］. AIAA Journal， 2022， 60（9）： 5249-5261.
10	JIN X W， CHENG P， CHEN W L， 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）： 047105.
11	SEKAR V， JIANG Q H， SHU C， et al. Fast flow field prediction over airfoils using deep learning approach［J］. Physics of Fluids， 2019， 31（5）： 057103.
12	HUI X Y， BAI J Q， WANG H， et al. Fast pressure distribution prediction of airfoils using deep learning［J］. Aerospace Science and Technology， 2020， 105： 105949.
13	DURU C， ALEMDAR H， BARAN O U. A deep learning approach for the transonic flow field predictions around airfoils［J］. Computers & Fluids， 2022， 236： 105312.
14	THUEREY N， WEIßENOW K， PRANTL L， et al. Deep learning methods for reynolds-averaged navier–stokes simulations of airfoil flows［J］. AIAA Journal， 2020， 58（1）： 25-36.
15	ZUO K J， YE Z Y， ZHANG W W， et al. Fast aerodynamics prediction of laminar airfoils based on deep attention network［J］. Physics of Fluids， 2023， 35（3）： 037127.
16	LI Y F， CHANG J T， KONG C， et al. Recent progress of machine learning in flow modeling and active flow control［J］. Chinese Journal of Aeronautics， 2022， 35（4）： 14-44.
17	LI Y F， CHANG J T， WANG Z A， et al. Inversion and reconstruction of supersonic cascade passage flow field based on a model comprising transposed network and residual network［J］. Physics of Fluids， 2019， 31（12）： 126102.
18	杜周，徐全勇，宋振寿，等. 基于深度学习的压气机叶型气动特性预测［J］. 航空动力学报， 2023， 38（9）： 2251-2260.
	DU Z， XU Q Y， SONG Z S， et al. Prediction of aerodynamic characteristics of compressor blade profile based on deep learning［J］. Journal of Aerospace Power， 2023， 38（9）： 2251-2260 （in Chinese）.
19	JIN Y， LI S， JUNG O. Prediction of flow properties on turbine vane airfoil surface from 3D geometry with convolutional neural network［C］‍∥ Proceedings of ASME Turbo Expo 2019： Turbomachinery Technical Conference and Exposition， 2019
20	LI J X， LI Y Z， LIU T Y， et al. Multi-fidelity graph neural network for flow field data fusion of turbomachinery［J］. Energy， 2023， 285： 129405.
21	冯云阳，宋西镇，袁巍，等. 基于物理嵌入神经网络的叶栅损失模型［J］. 航空动力学报， 2023， 38（7）： 1615-1625.
	FENG Y Y， SONG X Z， YUAN W， et al. Physics-informed neural networks based cascade loss model［J］. Journal of Aerospace Power， 2023， 38（7）： 1615-1625 （in Chinese）.
22	YANG G A， YANG Y， LU Z Z， et al. STA-TSN： spatial-Temporal Attention Temporal Segment Network for action recognition in video［J］. PLoS One， 2022， 17（3）： e0265115.
23	LI J N， WONG Y， ZHAO Q， et al. Attention transfer from web images for video recognition［C］∥ Proceedings of the 25th ACM international conference on Multimedia. New York： ACM， 2017： 1-9.
24	ZHAO Y C， ZHANG J P， DUAN R， et al. Lightweight target-aware attention learning network-based target tracking method［J］. Mathematics， 2022， 10（13）： 2299.
25	WANG Y Q， LIU T Y， ZHANG D. Aerodynamic prediction on the off-design performance of a S-CO₂ turbine based on deep learning［C］‍∥ Proceedings of ASME Turbo Expo 2021： Turbomachinery Technical Conference and Exposition， 2021.
26	POST P， WINHART B， DI MARE F. Investigation of physics-informed neural networks based solution techniques for internal flows［C］‍∥Proceedings of ASME Turbo Expo 2022： Turbomachinery Technical Conference and Exposition， 2022.
27	CHEN C F R， FAN Q F， PANDA R. CrossViT： cross-attention multi-scale vision transformer for image classification［C］‍∥2021 IEEE/CVF International Conference on Computer Vision （ICCV）. Piscataway： IEEE Press， 2021： 347-356.
28	DOSOVITSKIY A， BEYER L， KOLESNIKOV A， et al. An image is worth 16x16 words： transformers for image recognition at scale［DB/OL］. arXiv preprint： 2010.11929， 2020.

相对换算转速	转速/（r·min^-1）	S1安装角/（°）
1.00	9 687.0	18.87
0.99	9 590.1	17.11
0.90	8 718.3	3.63
0.80	7 749.6	-5.42

模块	层名称	输出尺寸
嵌入层	重塑层	196×768
嵌入层	全连接层	196×512
Transformer 编码器×3	归一化层	197×512
	多头注意力	197×512
	残差连接	197×512
	归一化层	197×512
	多层感知机	197×512
	残差连接	197×512
特征输出	层归一化	512
特征输出	全连接层	8

模块	层名称	卷积核/步长/填充	输出尺寸
线性模块	全连接层1		5 632
线性模块	全连接层2		1 536
塑形	塑形层		256×2×3
转置卷积×7	转置卷积1	2×2/2/0	128×4×6
转置卷积×7	ELU		128×4×6
转置卷积	转置卷积8	2×2/2/0	1×512×768
转置卷积	ELU		1×512×768
卷积×7	卷积1	2×2/2/0	2×256×384
卷积×7	ELU		2×256×384
卷积	卷积8	2×2/2/0	256×2×3
	ELU		256×2×3
	平均池化层	2×2/2/0×1	256×1×2
塑形	扁平化层		512
线性模块×5	全连接层3		1 024/512/128/32/1
线性模块×5	ELU		1 024/512/128/32/1

名称	相对换算转速	半径/m
训练集	0.8、0.83、0.85、0.87、0.89、0.91、0.93、0.95、0.97、0.99、1.015	0.18、0.21、0.24、0.27、0.30、0.33、0.36
验证集	0.86、0.90、0.94、0.98以及训练集转速上其他半径	0.18、0.21、0.24、0.27、0.30、0.33、0.36（部分工况增加了其他半径，如0.195、0.285等）
测试集	0.88、0.92、0.96、1.00以及训练集转速上其他半径	0.18、0.21、0.24、0.27、0.30、0.33、0.36（部分工况增加了其他半径，如0.255、0.345等）

名称	版本	时间/h
平台	Windows
CPU	Intel（R） Xeon（R） Platinum 8375C CPU @ 2.90 GHz
GPU	NVIDIA GeForce RTX4090	44
Cuda	11.8
Pytorch	2.0.0+cu118

Compressor flow field reconstruction method based on deep attention networks

RichHTML

PDF (PC)

Knowledge

Abstract

Cite this article

share this article

Figures/Tables 24

References 28

Related Articles 15

Recommended Articles

Metrics

Comments

[1]	Jianyu XU, Li ZHOU, Zhanxue WANG, Jie SHI, Hao SHI. Calculation method for hypersonic plume infrared radiation based on a fast line-by-line calculation model [J]. Acta Aeronautica et Astronautica Sinica, 2025, 46(8): 630778-630778.
[2]	Lingjie MENG, Hongguang LI, Xinjun LI. SAR image simulation method guided by geomorphic category information [J]. Acta Aeronautica et Astronautica Sinica, 2025, 46(7): 331003-331003.
[3]	Zhihao ZHAO, Zhaohua YANG, Yun WU, Yuanjin YU. Single-photon counting imaging denoising method based on deep learning in low-light environment [J]. Acta Aeronautica et Astronautica Sinica, 2025, 46(3): 630531-630531.
[4]	Yiquan WU, Kang TONG. Research advances on deep learning-based small object detection in UAV aerial images [J]. Acta Aeronautica et Astronautica Sinica, 2025, 46(3): 30848-030848.
[5]	Xiaowei JIANG, Yiquan WU. Research progress of UAV aerial image mosaic methods [J]. Acta Aeronautica et Astronautica Sinica, 2025, 46(17): 331799-331799.
[6]	Lin CHEN, Xiwen GU, Zhiying CHEN, Zhuo ZHANG, Xiaoliang SUN. High-precision monocular vision pose measurement for large distance span in carrier landing guidance [J]. Acta Aeronautica et Astronautica Sinica, 2025, 46(15): 331568-331568.
[7]	Bin SUN, Hang YOU, Wenbo LI, Xiangrui LIU, Jiayi MA. Dual-band payload image fusion and its applications in low-altitude remote sensing [J]. Acta Aeronautica et Astronautica Sinica, 2025, 46(11): 531343-531343.
[8]	Fanteng MENG, Yong QIN, Jing CUI, Yunpeng WU, Zicheng ZHANG, Shaowei WEI. Unknown risk detection in external environment of railroad using UAV images [J]. Acta Aeronautica et Astronautica Sinica, 2025, 46(11): 531262-531262.
[9]	Weishi CHEN, Hongchuang NIU, Xin WANG, Jian WAN, Xianfeng LU, Jie ZHANG, Qingbin WANG. Review on multi-source detection technologies for birds and drones in airport clearance area [J]. Acta Aeronautica et Astronautica Sinica, 2025, 46(10): 31251-031251.
[10]	Jie LIN, Zhigong TANG, Weiqi QIAN, Yueqing WANG, Peng ZHANG, Weixia XU, Jie LIU. Research progress and prospects of aircraft aerodynamic design based on generative models [J]. Acta Aeronautica et Astronautica Sinica, 2025, 46(10): 631679-631679.
[11]	Yonghai WANG, Haoge LI, Jiaxin LI, Yi DUAN, Chuan TIAN, Lingxi GUO, Xusheng WU. Rapid aircraft shape generation based on deep learning [J]. Acta Aeronautica et Astronautica Sinica, 2025, 46(10): 631614-631614.
[12]	Jiaqi LIU, Rongqian CHEN, Jinhua LOU, Xu HAN, Hao WU, Yancheng YOU. Aerodynamic shape optimization of high-speed helicopter rotor airfoil based on deep learning [J]. Acta Aeronautica et Astronautica Sinica, 2024, 45(9): 529828-529828.
[13]	Xudong LUO, Yiquan WU, Jinlin CHEN. Research progress on deep learning methods for object detection and semantic segmentation in UAV aerial images [J]. Acta Aeronautica et Astronautica Sinica, 2024, 45(6): 28822-028822.
[14]	Haiqiao LIU, Meng LIU, Zichao GONG, Jing DONG. A review of image matching methods based on deep learning [J]. Acta Aeronautica et Astronautica Sinica, 2024, 45(3): 28796-028796.
[15]	Zi WANG, Jinghao WANG, Yang LI, Zhang LI, Qifeng YU. Non-cooperative target pose estimation from monocular images based on lightweight neural network [J]. Acta Aeronautica et Astronautica Sinica, 2024, 45(22): 330248-330248.