Abstract
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.
Similar content being viewed by others
References
Aksoy S, Haralick RM (2000) Feature normalization and likelihood-based similarity measures for image retrieval. Pattern Recognit Lett 22:563–582
Amidror I (2002) Scattered data interpolation methods for electronic imaging systems: a survey. J Electron Imaging 11(2):157–176
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 2720
Bengio Y (2009) Learning deep architectures for ai. Found Trends Mach Learn 2(1):1–127
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 67–76
Chollampatt S, Tou Ng H (2018) A multilayer convolutional encoder–decoder neural network for grammatical error correction. arXiv e-prints arXiv:1801.08831
Duraisamy K (2005) Studies in tip vortex formation, evolution and contro, dept of aerospace engineering, Univ of Maryland. PhD thesis, University of Maryland
Duraisamy K, Iaccarino G, Xiao H (2019) Turbulence modeling in the age of data. Annu Rev Fluid Mech 51(1):357–377
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
Foley JD, van Dam A, Feiner SK, Hughes JF (1995) Computer graphics: principles and practice in C, 2nd edn. Addison-Wesley, Reading
Fuhrmann S, Goesele M (2014) Floating scale surface reconstruction. ACM Trans Graph 33(4):46:1–46:11
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 481–490
Hennigh O (2017) Lat-Net: Compressing Lattice Boltzmann flow simulations using deep neural networks. arXiv e-prints arXiv:1705.09036
Hoppe H, DeRose T, Duchamp T, McDonald J, Stuetzle W (1992) Surface reconstruction from unorganized points. SIGGRAPH Comput Graph 26(2):71–78
Jürgen S (2015) Deep learning in neural networks: an overview. Neural Netw 61:85–117
Kazhdan M, Bolitho M, Hoppe H (2006) Poisson surface reconstruction. In: Polthier K, Sheffer A (eds) Symposium on geometry processing. The Eurographics Association, pp 61–70
Koren B (1993) A robust upwind discretization method for advection, diffusion and source terms. Vieweg, Decatur, pp 117–138
Lakshminarayan VK, Baeder JD (2010) Computational investigation of microscale coaxial-rotor aerodynamics in hover. J Aircr 47(3):940–955
Lecun Y, Bottou L, Bengio Y, Haffner P (1998) Gradient-based learning applied to document recognition. Proc IEEE 86(11):2278–2324
Lee S, You D (2017) Prediction of laminar vortex shedding over a cylinder using deep learning. arXiv e-prints arXiv:1712.07854
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
Ling H, Jacobs DW (2007) Shape classification using the inner-distance. IEEE Trans Pattern Anal Mach Intell 29(2):286–299
Mathieu M, Couprie C, LeCun Y (2015) Deep multi-scale video prediction beyond mean square error. arXiv e-prints arXiv:1511.05440
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 Astronautics
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
Pandya S, Venkateswaran S, Pulliam T (2003) Implementation of preconditioned dual-time procedures in overflow. American Institute of Aeronautics and Astronautics
Prantl L, Bonev B, Thuerey N (2017) Generating liquid simulations with deformation-aware neural networks. arXiv e-prints arXiv:1704.07854
Pulliam T, Chaussee D (1981) A diagonal form of an implicit approximate-factorization algorithm. J Comput Phys 39(2):347–363
Raissi M, Karniadakis GE (2018) Hidden physics models: machine learning of nonlinear partial differential equations. J Comput Phys 357:125–141
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
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 378:686–707
Ramachandran P, Zoph B, Le QV (2017) Searching for activation functions. arXiv e-prints arXiv:1710.05941
Roe PL (1986) Characteristic-based schemes for the euler equations. Annu Rev Fluid Mech 18(1):337–365
Sethian JA (1996) A fast marching level set method for monotonically advancing fronts. Proc Natl Acad Sci USA 93:1591–1595
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 4165–4170
Somers D (1997a) Design and experimental results for the s805 airfoil. Tech. Rep. NREL/SR-440-6917
Somers DM (1997b) Design and experimental results for the s809 airfoil. Tech. Rep. NREL/SR-440-6918
Somers DM (2004) S814 and s815 airfoils: October 1991–july 1992. Tech. rep
Spalart P, Allmaras S (1992) A one-equation turbulence model for aerodynamic flows. American Institute of Aeronautics and Astronautics
Taylor GW, Fergus R, LeCun Y, Bregler C (2010) Convolutional learning of spatio-temporal features. Comput Vis ECCV 2010:140–153
Tompson J, Schlachter K, Sprechmann P, Perlin K (2016) Accelerating Eulerian fluid simulation with convolutional networks. arXiv e-prints arXiv:1607.03597
Turkel E (1999) Preconditioning techniques in computational fluid dynamics. Annu Rev Fluid Mech 31(1):385–416
van Leer B (1979) Towards the ultimate conservative difference scheme. v. a second-order sequel to godunov’s method. J Comput Phys 32(1):101–136
Xu K, Kim VG, Huang Q, Kalogerakis E (2015) Data-driven shape analysis and processing. arXiv e-prints arXiv:1502.06686
Zhang D, Lu G (2004) Review of shape representation and description techniques. Pattern Recognit 37(1):1–19
Zhang Y, Sung WJ, Mavris D (2017) Application of convolutional neural network to predict airfoil lift coefficient. arXiv e-prints arXiv:1712.10082
Zhao H (2005) A fast sweeping method for Eikonal equations. Math Comput 74(250):603–627
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 18–26
Acknowledgements
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).
Author information
Authors and Affiliations
Corresponding author
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Saakaar Bhatnagar and Yaser Afshar have the Co-First/Equal authorship.
Appendix: Governing equations
Appendix: Governing equations
The RANS equations are derived by ensemble-averaging the conservation equations of mass, momentum and energy. These equations, for compressible flow are given by:
where the overbar indicates conventional time-average mean, \(u_i\) is the fluid velocity, \(\rho \) is the density, p is the pressure, \(\tau _{ij}\) is the Reynolds stress term, \(c_P\) is the heat capacity at constant pressure, and \(\kappa \) is the kinetic energy of the fluctuating field (local turbulent kinetic energy). The density weighted time averaging (Favre averaging) of any quantity \(\xi \), denoted by \({{\hat{\xi }}}\) is given as \({{\hat{\xi }}}=\overline{\rho \xi }/{{\bar{\rho }}}\), where,
To provide closure to the above equations, we use the model proposed by Spalart and Allmaras [39]. In this closure, the Boussinesq hypothesis relates the Reynolds stress and the effect of turbulence as an eddy viscosity \(\mu _t\). Employing the Boussinesq approach, and Reynolds Analogy a transport equation for a working variable \({{\tilde{\nu }}}\) is solved to estimate the eddy viscosity field at every iteration.
The turbulent eddy viscosity is computed as \(\mu _t={{\bar{\rho }}}{{\tilde{\nu }}} f_{v1}\), where,
The first term on the right hand side of this Eq. 19 is the production term for \({{\tilde{\nu }}}\) while the second term represents dissipation.
Rights and permissions
About this article
Cite this article
Bhatnagar, S., Afshar, Y., Pan, S. et al. Prediction of aerodynamic flow fields using convolutional neural networks. Comput Mech 64, 525–545 (2019). https://doi.org/10.1007/s00466-019-01740-0
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s00466-019-01740-0