Airfoil aerodynamic loads calculation by vortex element method using neural networks


Аuthors

Kamenev N. D., Shcheglov G. A.*

Baumann Moscow State Technical University, 105005, Moscow, 2nd Baumanskaya St., b. 5, c. 1

*e-mail: shcheglov_ga@bmstu.ru

Abstract

The prospect of using a pre-trained artificial neural network in vortex element method to reconstruct velocity fields based on the known distribution of point vortices is investigated. Models such as ResNet, U-Net and Attention U-Net are considered. In addition, the method of calculating the aerodynamic loads on the basis of the obtained vortices is considered. The results of the solution of a test problem on the wing profile flowing with a plane-parallel flow of a inviscid incompressible fluid are presented. The advantages in acceleration of calculations are analyzed and the error of the obtained results is estimated. It is shown that the use of neural network models to simultaneously solve two interrelated tasks — predicting the image of vortex wake and predicting aerodynamic loads using a common error function led to a significant improvement in the accuracy of the model. These collaborative optimization allows the model to better capture key dependencies in the data and learn more efficiently. Vortex wake prediction improves spatial perception of flow dynamics, while load prediction helps the model capture important patterns. The use of a common error function combines these two processes, which allows the model to simultaneously improve its abilities both in segmentation of vortex wake and in predicting aerodynamic loads. As a result, the model becomes more informative and accurate, as it can be trained on additional signals and correlations, which led to improvements in the RMSE and DICE metrics. Performance analysis showed that OpenVINO-optimized models on Intel CPUs provide a balance between computing speed  and accuracy, while maintaining acceptable forecast quality over short time intervals.

Keywords:

vortex element method, computational fluid dynamics, airfoil, 2D flow, artificial neural network, ResNet, U-Net, Attention U-Net

References

  1. Cottet G.-H., Koumoutsakos P.D. Vortex methods: theory and practice. 2nd ed. Cambridge University Press, 2000, 328 p.
  2. Rodionov, N.G. Experimental Studies of Airfoil Cascades with High Velocity Vector Circulation around the Airfoil. Thermal Engeneering, 2022, Vol. 69, P.42–50. DOI:10.1134/S0040601521110033.
  3. Varaksin, A.Y., Ryzhkov, S.V. Vortex Flows with Particles and Droplets. Symmetry, 2022, Vol. 14, Iss. 10, Paper ID 2016 DOI:10.3390/sym14102016.
  4. Egorov I.A. Trudy MAI, 2022, no. 123. URL: https://trudymai.ru/eng/published.php?ID=165481 
  5. Klimina L., Dosaev M., Selyutskiy Y. Asymptotic analysis of the mathematical model of a wind-powered vehicle. Applied Mathematical Modelling, Vol. 46, 2017, pp. 691-697, DOI: 10.1016/j.apm.2016.06.022.
  6. Andronov P.R., Dosaev M.Z., Dynnikova G.Y. et al. Modeling of oscillating wind turbine. J. Mach. Manuf. Reliab., 2009, No. 38, P. 383–387. DOI:10.3103/S1052618809040153.
  7. Kotsur, O.S., Shcheglov, G.A., Marchevsky, I.K. Approximate Weak Solutions to the Vorticity Evolution Equation for a Viscous Incompressible Fluid in the Class of Vortex Filaments. Russian Journal of Nonlinear Dynamics, 2022, Vol. 18, Iss. 3, P. 423-439. DOI:10.20537/nd220307
  8. Ryatina, E., Marchevsky, I., Kolganova, A. Boundary integral equation solving in vortex method using the Barnes-Hut/multipole algorithm. Proceedings - Ivannikov ISPRAS Open Conference 2022, P. 74-80.
  9. Kochov D., Smith J.A., Alieva A., Hoyer S. Machine learning-accelerated computational fluid dynamics. PNAS, 2021, Vol. 118(21), Art.No.e2101784118. DOI: 10.1073/pnas.2101784118.
  10. Loktev, D.A., Loktev, A.A., Makarov, A.A., Shepelev, D.N.  The System Simulating the State of the Objects and the Process of Their Monitoring with the Help of the Convolutional Neural Network. International Review on Modelling and Simulations 2022, Vol. 15, Iss. 1, P. 1-9.
  11. Senthilraj, S., Raja, R., Cao, J., Fardoun, H.M. Dissipativity analysis of stochastic fuzzy neural networks with randomly occurring uncertainties using delay dividing approach. Nonlinear Analysis: Modelling and Control 2019, Vol. 24, Iss. 4, P. 561-581. DOI: 10.15388/NA.2019.4.5
  12. Dorofeev F.E., Vasilenko D.A., Dorofeev E.A. Trudy MAI, 2021, no. 119. URL: https://trudymai.ru/eng/published.php?ID=159784
  13. Sokolov D.Y. Trudy MAI, 2022, no. 123. URL: https://trudymai.ru/eng/published.php?ID=165563 
  14. Xiaowei Jin, Shujin Laima, Hui Li. Physics-enhanced deep learning methods for modelling and simulating flow fields. Chinese Journal of Theoretical and Applied Mechanics, 2021, No.53(10) P. 2616-2629. DOI: 10.6052/0459-1879-21-373
  15. Son P.T. Trudy MAI, 2025, no. 140. URL: https://trudymai.ru/eng/published.php?ID=184057 
  16. Hao Yan, Yangwei Lui, Lipeng Lu. Turbulence anisotropy analysis in a highly loaded linear compressor cascade. Aerospace Science and Technology, 2019, Vol. 91, P. 241-254. DOI: 10.1016/j.ast.2019.05.029
  17. Xiao-Dong Bai, Wei Zhang. Machine learning for vortex induced vibration in turbulent flow. Computers & Fluids, 2022,Vol. 235, PaperID 1052666. DOI:10.1016/j.compfluid.2021.105266
  18. Marchevsky I., Sokol K., Ryatina E., Izmailova Y. The VM2D Open Source Code for Two-Dimensional Incompressible Flow Simulation by Using Fully Lagrangian Vortex Particle Methods. Axioms, 2023, No. 12(3), Iss. 248, 33 p. DOI:10.3390/axioms12030248.
  19. Kaiming He, Xiangyu Zhang, Shaoqing Ren, Jian Sun. Deep Residual Learning for Image Recognition. Computer Vision and Pattern Recognition. Proceedings of 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR). DOI:10.1109/CVPR.2016.90
  20. Ronneberger, O., Fischer, P., Brox, T. U-Net: Convolutional Networks for Biomedical Image Segmentation. Medical Image Computing and Computer-Assisted Intervention – MICCAI 2015. Lecture Notes in Computer Science 2015, vol 9351. Springer, Cham. DOI: 10.1007/978-3-319-24574-4_28
  21. Hinton G.E., Salakhutdinov R.R. Reducing the dimensionality of data with neural networks. Science. 28 Jul 2006. Vol. 313, Iss. 5786, P. 504-507. DOI: 10.1126/science.1127647
  22. Lindemann B., Muller T., Vietz H., Jazdi N., Weyrich M. A survey on long short-term memory networks for time series prediction. Procedia CIRP, 2021 Vol. 99, P. 650-655. DOI: 10.1016/j.procir.2021.03.088
  23. Ashish Vaswani, Noam Shazeer. Attention Is All You Need. Proceedings of 31st Conference on Neural Information Processing Systems (NIPS 2017). URL: https://proceedings.neurips.cc/paper_files/paper/2017/file/3f5ee243547dee91fbd053c1c4a845aa-Paper.pd... (accessed 23.07.2025).


Download

mai.ru — informational site MAI

Copyright © 2000-2025 by MAI

 Вход