An application of neural networks to the prediction of aerodynamic coefficients of aerofoils and wings

Abstract This work proposes a novel multi-output neural network for the prediction of aerodynamic coefficients of aerofoils in two dimensions and wings in three dimensions. Contrary to existing neural networks that are often designed to predict aerodynamic quantities of interest, the proposed network considers as output the pressure at a number of selected points on the aerodynamic shape. The proposed multi-output neural network is compared with other approaches found in the literature. Furthermore, a detailed comparison of the proposed neural network with the popular proper orthogonal decomposition method is presented. The numerical results, involving high dimensional problems with flow and geometric parameters, show the benefits of the proposed approach.

[1]  Juan J. Alonso,et al.  Airfoil design optimization using reduced order models based on proper orthogonal decomposition , 2000 .

[2]  Yao Zhang,et al.  Application of Convolutional Neural Network to Predict Airfoil Lift Coefficient , 2017, ArXiv.

[3]  Bernhard Schölkopf,et al.  Kernel Principal Component Analysis , 1997, ICANN.

[4]  Francisco Chinesta,et al.  Non-intrusive proper generalized decomposition involving space and parameters: application to the mechanical modeling of 3D woven fabrics , 2019, Adv. Model. Simul. Eng. Sci..

[5]  A. Kharal,et al.  Neural networks based airfoil generation for a given Cp using Bezier–PARSEC parameterization , 2012 .

[6]  Bento Silva de Mattos,et al.  Artificial Neural Networks Applied to Airplane Design , 2015 .

[7]  Gianluigi Rozza,et al.  Reduced Order Methods for Modeling and Computational Reduction , 2013 .

[8]  Lijuan Cao,et al.  A comparison of PCA, KPCA and ICA for dimensionality reduction in support vector machine , 2003, Neurocomputing.

[9]  Ning Qin,et al.  Fast dynamic grid deformation based on Delaunay graph mapping , 2006 .

[10]  Herbert Martins Gomes,et al.  An airfoil optimization technique for wind turbines , 2012 .

[11]  Paris Perdikaris,et al.  Physics-informed neural networks: A deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations , 2019, J. Comput. Phys..

[12]  M. Rai.  Three-dimensional aerodynamic design using artificial neural networks , 2002 .

[13]  K. Morgan,et al.  Reduced order modelling for unsteady fluid flow using proper orthogonal decomposition and radial basis functions , 2013 .

[14]  Jan S. Hesthaven,et al.  Non-intrusive reduced order modeling of nonlinear problems using neural networks , 2018, J. Comput. Phys..

[15]  James E. Steck,et al.  An exploratory application of neural networks to airfoil design , 1994 .

[16]  A. Chatterjee.  An introduction to the proper orthogonal decomposition , 2000 .

[17]  Simona Perotto,et al.  A POD‐selective inverse distance weighting method for fast parametrized shape morphing , 2017, International Journal for Numerical Methods in Engineering.

[18]  H. Sobieczky,et al.  Geometry Generator for CFD and Applied Aerodynamics , 1997 .

[19]  L. Herrmann.  Laplacian-Isoparametric Grid Generation Scheme , 1976 .

[20]  Charbel Farhat,et al.  POD-based Aeroelastic Analysis of a Complete F-16 Configuration: ROM Adaptation and Demonstration , 2005 .

[21]  Weiwei Zhang,et al.  Uncertainty Quantification and Sensitivity Analysis of Transonic Aerodynamics with Geometric Uncertainty , 2017 .

[22]  Richard J. Beckman,et al.  A Comparison of Three Methods for Selecting Values of Input Variables in the Analysis of Output From a Computer Code , 2000, Technometrics.

[23]  Jens K. Fassbender,et al.  MEGAFLOW - Numerical Flow Simulation for Aircraft Design , 2005 .

[24]  Gunnar Rätsch,et al.  Kernel PCA pattern reconstruction via approximate pre-images. , 1998 .

[25]  Han Zhao,et al.  Learning Neural Networks with Adaptive Regularization , 2019, NeurIPS.

[26]  Francisco Chinesta,et al.  PGD for solving multidimensional and parametric models , 2014 .

[27]  Richard P. Lippmann,et al.  An introduction to computing with neural nets , 1987 .

[28]  Les A. Piegl,et al.  The NURBS Book , 1995, Monographs in Visual Communication.

[29]  Ruben Sevilla,et al.  Numerical integration over 2D NURBS-shaped domains with applications to NURBS-enhanced FEM , 2011 .

[30]  M. O. Tokhi,et al.  Training neural networks: backpropagation vs. genetic algorithms , 2001, IJCNN'01. International Joint Conference on Neural Networks. Proceedings (Cat. No.01CH37222).

[31]  Antonio Huerta,et al.  Nonintrusive proper generalised decomposition for parametrised incompressible flow problems in OpenFOAM , 2019, Comput. Phys. Commun..

[32]  Mengqi Zhang,et al.  Inverse Design of Airfoil Using a Deep Convolutional Neural Network , 2019, AIAA Journal.

[33]  Dimitris Drikakis,et al.  An artificial neural network-based multiscale method for hybrid atomistic-continuum simulations , 2013, Microfluidics and Nanofluidics.

[34]  Andrzej Cichocki,et al.  Kernel PCA for Feature Extraction and De-Noising in Nonlinear Regression , 2001, Neural Computing & Applications.

[35]  Gunnar Rätsch,et al.  Input space versus feature space in kernel-based methods , 1999, IEEE Trans. Neural Networks.

[36]  George E. Karniadakis,et al.  Hidden Fluid Mechanics: A Navier-Stokes Informed Deep Learning Framework for Assimilating Flow Visualization Data , 2018, ArXiv.

[37]  Nateri K. Madavan,et al.  AERODYNAMIC DESIGN USING NEURAL NETWORKS , 2000 .

[38]  K. Willcox,et al.  Aerodynamic Data Reconstruction and Inverse Design Using Proper Orthogonal Decomposition , 2004 .

[39]  R. Hibbeler.  Statics and Mechanics of Materials , 1995 .

[40]  A. Patera,et al.  Reduced basis approximation and a posteriori error estimation for affinely parametrized elliptic coercive partial differential equations , 2007 .

[41]  N. Weatherill,et al.  Efficient three‐dimensional Delaunay triangulation with automatic point creation and imposed boundary constraints , 1994 .

[42]  Sundaram Suresh,et al.  Lift coefficient prediction at high angle of attack using recurrent neural network , 2003 .

[43]  Dimitris Drikakis,et al.  Machine-Learning Methods for Computational Science and Engineering , 2020, Comput..

[44]  Antonio Huerta,et al.  Solution of geometrically parametrised problems within a CAD environment via model order reduction , 2020, Computer Methods in Applied Mechanics and Engineering.

[45]  Anders Krogh,et al.  A Simple Weight Decay Can Improve Generalization , 1991, NIPS.

[46]  Geoffrey E. Hinton,et al.  On the importance of initialization and momentum in deep learning , 2013, ICML.

[47]  P. Holmes,et al.  The Proper Orthogonal Decomposition in the Analysis of Turbulent Flows , 1993 .

[48]  Weiwei Zhang,et al.  Multi-fidelity modeling framework for nonlinear unsteady aerodynamics of airfoils , 2019 .

[49]  Bento Silva de Mattos,et al.  Aerodynamic Coefficient Prediction of Airfoils Using Neural Networks , 2008 .

[50]  Benjamin Peherstorfer,et al.  Projection-based model reduction: Formulations for physics-based machine learning , 2019, Computers & Fluids.

[51]  Robert F. Stengel,et al.  Identification of aerodynamic coefficients using computational neural networks , 1993 .

[52]  Gang Sun,et al.  Artificial neural network based inverse design: Airfoils and wings , 2015 .

[53]  Adrien Leygue,et al.  The Proper Generalized Decomposition for Advanced Numerical Simulations: A Primer , 2013 .

[54]  Jaime Peraire,et al.  The computation of three-dimensional flows using unstructured grids , 1991 .

[55]  Levent Burak Kara,et al.  A data-driven investigation and estimation of optimal topologies under variable loading configurations , 2014, Comput. methods Biomech. Biomed. Eng. Imaging Vis..

[56]  Peter Dabnichki,et al.  Effect of the wing shape on the thrust of flapping wing , 2011 .

[57]  Manas Khurana,et al.  Application of swarm approach and aritficial neural networks for airfoil shape optimization , 2008 .

[58]  Chris Lacor,et al.  CFD modeling and multi-objective optimization of cyclone geometry using desirability function, artificial neural networks and genetic algorithms , 2013 .

[59]  J. Peraire,et al.  Balanced Model Reduction via the Proper Orthogonal Decomposition , 2002 .

[60]  K. A. Sørensen,et al.  A multigrid accelerated hybrid unstructured mesh method for 3D compressible turbulent flow , 2003 .

[61]  Michele Conti,et al.  A nonintrusive proper generalized decomposition scheme with application in biomechanics , 2018 .

[62]  Yoshua Bengio,et al.  Understanding the difficulty of training deep feedforward neural networks , 2010, AISTATS.

[63]  Geoffrey E. Hinton,et al.  Learning representations by back-propagating errors , 1986, Nature.