Prediction of aerodynamic flow fields using convolutional neural networks

  • Saakaar Bhatnagar
  • Yaser AfsharEmail author
  • Shaowu Pan
  • Karthik Duraisamy
  • Shailendra Kaushik
Original Paper


An approximation model based on convolutional neural networks (CNNs) is proposed for flow field predictions. The CNN is used to predict the velocity and pressure field in unseen flow conditions and geometries given the pixelated shape of the object. In particular, we consider Reynolds Averaged Navier–Stokes (RANS) flow solutions over airfoil shapes as training data. The CNN can automatically detect essential features with minimal human supervision and is shown to effectively estimate the velocity and pressure field orders of magnitude faster than the RANS solver, making it possible to study the impact of the airfoil shape and operating conditions on the aerodynamic forces and the flow field in near-real time. The use of specific convolution operations, parameter sharing, and gradient sharpening are shown to enhance the predictive capabilities of the CNN. We explore the network architecture and its effectiveness in predicting the flow field for different airfoil shapes, angles of attack, and Reynolds numbers.


Aerodynamics Deep learning Convolutional neural networks Airfoils RANS 



This work was supported by General Motors Corporation under a contract titled “Deep Learning and Reduced Order Modeling for Automotive Aerodynamics.” Computing resources were provided by the NSF via grant 1531752 MRI: Acquisition of Conflux, A Novel Platform for Data-Driven Computational Physics (Tech. Monitor: Stefan Robila).


  1. 1.
    Aksoy S, Haralick RM (2000) Feature normalization and likelihood-based similarity measures for image retrieval. Pattern Recognit Lett 22:563–582CrossRefzbMATHGoogle Scholar
  2. 2.
    Amidror I (2002) Scattered data interpolation methods for electronic imaging systems: a survey. J Electron Imaging 11(2):157–176CrossRefGoogle Scholar
  3. 3.
    Aranake A, Lakshminarayan V, Duraisamy K (2012) Assessment of transition model and cfd methodology for wind turbine flows. In: 42nd AIAA fluid dynamics conference and exhibit. American Institute of Aeronautics and Astronautics, p 2720Google Scholar
  4. 4.
    Bengio Y (2009) Learning deep architectures for ai. Found Trends Mach Learn 2(1):1–127CrossRefzbMATHGoogle Scholar
  5. 5.
    Carr JC, Beatson RK, Cherrie JB, Mitchell TJ, Fright WR, McCallum BC, Evans TR (2001) Reconstruction and representation of 3d objects with radial basis functions. In: Proceedings of the 28th annual conference on computer graphics and interactive techniques. ACM, New York, NY, USA, SIGGRAPH ’01, pp 67–76Google Scholar
  6. 6.
    Chollampatt S, Tou Ng H (2018) A multilayer convolutional encoder–decoder neural network for grammatical error correction. arXiv e-prints arXiv:1801.08831
  7. 7.
    Duraisamy K (2005) Studies in tip vortex formation, evolution and contro, dept of aerospace engineering, Univ of Maryland. PhD thesis, University of MarylandGoogle Scholar
  8. 8.
    Duraisamy K, Iaccarino G, Xiao H (2019) Turbulence modeling in the age of data. Annu Rev Fluid Mech 51(1):357–377CrossRefzbMATHGoogle Scholar
  9. 9.
    Fernando B, Karaoglu S, Saha SK (2015) Object class detection and classification using multi scale gradient and corner point based shape descriptors. arXiv e-prints arXiv:1505.00432
  10. 10.
    Foley JD, van Dam A, Feiner SK, Hughes JF (1995) Computer graphics: principles and practice in C, 2nd edn. Addison-Wesley, ReadingzbMATHGoogle Scholar
  11. 11.
    Fuhrmann S, Goesele M (2014) Floating scale surface reconstruction. ACM Trans Graph 33(4):46:1–46:11CrossRefzbMATHGoogle Scholar
  12. 12.
    Guo X, Li W, Iorio F (2016) Convolutional neural networks for steady flow approximation. In: Proceedings of the 22Nd ACM SIGKDD international conference on knowledge discovery and data mining. ACM, KDD ’16, pp 481–490Google Scholar
  13. 13.
    Hennigh O (2017) Lat-Net: Compressing Lattice Boltzmann flow simulations using deep neural networks. arXiv e-prints arXiv:1705.09036
  14. 14.
    Hoppe H, DeRose T, Duchamp T, McDonald J, Stuetzle W (1992) Surface reconstruction from unorganized points. SIGGRAPH Comput Graph 26(2):71–78CrossRefGoogle Scholar
  15. 15.
    Jürgen S (2015) Deep learning in neural networks: an overview. Neural Netw 61:85–117CrossRefGoogle Scholar
  16. 16.
    Kazhdan M, Bolitho M, Hoppe H (2006) Poisson surface reconstruction. In: Polthier K, Sheffer A (eds) Symposium on geometry processing. The Eurographics Association, pp 61–70Google Scholar
  17. 17.
    Koren B (1993) A robust upwind discretization method for advection, diffusion and source terms. Vieweg, Decatur, pp 117–138zbMATHGoogle Scholar
  18. 18.
    Lakshminarayan VK, Baeder JD (2010) Computational investigation of microscale coaxial-rotor aerodynamics in hover. J Aircr 47(3):940–955CrossRefGoogle Scholar
  19. 19.
    Lecun Y, Bottou L, Bengio Y, Haffner P (1998) Gradient-based learning applied to document recognition. Proc IEEE 86(11):2278–2324CrossRefGoogle Scholar
  20. 20.
    Lee S, You D (2017) Prediction of laminar vortex shedding over a cylinder using deep learning. arXiv e-prints arXiv:1712.07854
  21. 21.
    Lee S, You D (2018) Data-driven prediction of unsteady flow fields over a circular cylinder using deep learning. arXiv e-prints arXiv:1804.06076
  22. 22.
    Ling H, Jacobs DW (2007) Shape classification using the inner-distance. IEEE Trans Pattern Anal Mach Intell 29(2):286–299CrossRefGoogle Scholar
  23. 23.
    Mathieu M, Couprie C, LeCun Y (2015) Deep multi-scale video prediction beyond mean square error. arXiv e-prints arXiv:1511.05440
  24. 24.
    Medida S, Baeder J (2011) Application of the correlation-based gamma-re theta t transition model to the Spalart–Allmaras turbulence model. American Institute of Aeronautics and AstronauticsGoogle Scholar
  25. 25.
    Miyanawala TP, Jaiman RK (2017) An efficient deep learning technique for the Navier–Stokes equations: application to unsteady wake flow dynamics. arXiv e-prints arXiv:1710.09099
  26. 26.
    Pandya S, Venkateswaran S, Pulliam T (2003) Implementation of preconditioned dual-time procedures in overflow. American Institute of Aeronautics and AstronauticsGoogle Scholar
  27. 27.
    Prantl L, Bonev B, Thuerey N (2017) Generating liquid simulations with deformation-aware neural networks. arXiv e-prints arXiv:1704.07854
  28. 28.
    Pulliam T, Chaussee D (1981) A diagonal form of an implicit approximate-factorization algorithm. J Comput Phys 39(2):347–363MathSciNetCrossRefzbMATHGoogle Scholar
  29. 29.
    Raissi M, Karniadakis GE (2018) Hidden physics models: machine learning of nonlinear partial differential equations. J Comput Phys 357:125–141MathSciNetCrossRefzbMATHGoogle Scholar
  30. 30.
    Raissi M, Yazdani A, Karniadakis GE (2018) Hidden fluid mechanics: a Navier–Stokes informed deep learning framework for assimilating flow visualization data. arXiv e-prints p arXiv:1808.04327
  31. 31.
    Raissi M, Perdikaris P, Karniadakis G (2019) Physics-informed neural networks: a deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations. J Comput Phys 378:686–707MathSciNetCrossRefGoogle Scholar
  32. 32.
    Ramachandran P, Zoph B, Le QV (2017) Searching for activation functions. arXiv e-prints arXiv:1710.05941
  33. 33.
    Roe PL (1986) Characteristic-based schemes for the euler equations. Annu Rev Fluid Mech 18(1):337–365MathSciNetCrossRefzbMATHGoogle Scholar
  34. 34.
    Sethian JA (1996) A fast marching level set method for monotonically advancing fronts. Proc Natl Acad Sci USA 93:1591–1595MathSciNetCrossRefzbMATHGoogle Scholar
  35. 35.
    Shijie J, Ping W, Peiyi J, Siping H (2017) Research on data augmentation for image classification based on convolution neural networks. In: 2017 Chinese automation congress (CAC). pp 4165–4170Google Scholar
  36. 36.
    Somers D (1997a) Design and experimental results for the s805 airfoil. Tech. Rep. NREL/SR-440-6917Google Scholar
  37. 37.
    Somers DM (1997b) Design and experimental results for the s809 airfoil. Tech. Rep. NREL/SR-440-6918Google Scholar
  38. 38.
    Somers DM (2004) S814 and s815 airfoils: October 1991–july 1992. Tech. repGoogle Scholar
  39. 39.
    Spalart P, Allmaras S (1992) A one-equation turbulence model for aerodynamic flows. American Institute of Aeronautics and AstronauticsGoogle Scholar
  40. 40.
    Taylor GW, Fergus R, LeCun Y, Bregler C (2010) Convolutional learning of spatio-temporal features. Comput Vis ECCV 2010:140–153Google Scholar
  41. 41.
    Tompson J, Schlachter K, Sprechmann P, Perlin K (2016) Accelerating Eulerian fluid simulation with convolutional networks. arXiv e-prints arXiv:1607.03597
  42. 42.
    Turkel E (1999) Preconditioning techniques in computational fluid dynamics. Annu Rev Fluid Mech 31(1):385–416MathSciNetCrossRefGoogle Scholar
  43. 43.
    van Leer B (1979) Towards the ultimate conservative difference scheme. v. a second-order sequel to godunov’s method. J Comput Phys 32(1):101–136CrossRefzbMATHGoogle Scholar
  44. 44.
    Xu K, Kim VG, Huang Q, Kalogerakis E (2015) Data-driven shape analysis and processing. arXiv e-prints arXiv:1502.06686
  45. 45.
    Zhang D, Lu G (2004) Review of shape representation and description techniques. Pattern Recognit 37(1):1–19CrossRefGoogle Scholar
  46. 46.
    Zhang Y, Sung WJ, Mavris D (2017) Application of convolutional neural network to predict airfoil lift coefficient. arXiv e-prints arXiv:1712.10082
  47. 47.
    Zhao H (2005) A fast sweeping method for Eikonal equations. Math Comput 74(250):603–627MathSciNetCrossRefzbMATHGoogle Scholar
  48. 48.
    Zuo Z, Shuai B, Wang G, Liu X, Wang X, Wang B, Chen Y (2015) Convolutional recurrent neural networks: learning spatial dependencies for image representation. In: 2015 IEEE conference on computer vision and pattern recognition workshops (CVPRW). pp 18–26Google Scholar

Copyright information

© Springer-Verlag GmbH Germany, part of Springer Nature 2019

Authors and Affiliations

  • Saakaar Bhatnagar
    • 1
  • Yaser Afshar
    • 1
    Email author
  • Shaowu Pan
    • 1
  • Karthik Duraisamy
    • 1
  • Shailendra Kaushik
    • 2
  1. 1.Department of Aerospace EngineeringUniversity of MichiganAnn ArborUSA
  2. 2.General Motors Global R&DWarrenUSA

Personalised recommendations