AIAA Journal · 2019 · 572 citations · 55 references
EngineeringAerospace EngineeringPhysicsComputational Fluid DynamicsFluid DynamicsFluid MechanicsComputational AerodynamicsDeep LearningNavier-stokes EquationsNeural NetworksPartial Differential EquationsPressure-velocity CouplingAirfoil AerodynamicsCartesian GridsNavier-stokes SolversNeural Network PredictionReynolds-averaged Navier - Stokes ModelingReynolds-averaged Navier-stokes EquationsReynolds-averaged Navier-stokes ModelingVelocity
Reynolds‑averaged Navier–Stokes (RANS) simulations are widely used for airfoil flow modeling, and the neural‑network approach explored here is broadly applicable to many PDE boundary‑value problems on Cartesian grids. The study evaluates how accurately deep learning models can infer RANS solutions for airfoil flows. The authors use a modernized U‑net architecture, training many neural networks while varying training data size and weight count to assess their impact on pressure and velocity prediction accuracy. The best models achieve mean relative pressure and velocity errors below 3% on unseen airfoil shapes, and the authors provide the source code publicly for reproducibility.
This study investigates the accuracy of deep learning models for the inference of Reynolds-averaged Navier–Stokes (RANS) solutions. This study focuses on a modernized U-net architecture and evaluates a large number of trained neural networks with respect to their accuracy for the calculation of pressure and velocity distributions. In particular, it is illustrated how training data size and the number of weights influence the accuracy of the solutions. With the best models, this study arrives at a mean relative pressure and velocity error of less than 3% across a range of previously unseen airfoil shapes. In addition all source code is publicly available in order to ensure reproducibility and to provide a starting point for researchers interested in deep learning methods for physics problems. Although this work focuses on RANS solutions, the neural network architecture and learning setup are very generic, and applicable to a wide range of partial differential equation boundary value problems on Cartesian grids.
55
Adam: A Method for Stochastic Optimization
Diederik P. Kingma, Jimmy Ba · UvA-DARE (University of Amsterdam) · 2014 · 84.8K citations · Full text
Densely Connected Convolutional Networks
Gao Huang, Zhuang Liu, Laurens van der Maaten et al. · 2017 · 44.7K citations
Unpaired Image-to-Image Translation Using Cycle-Consistent Adversarial Networks
Jun-Yan Zhu, Taesung Park, Phillip Isola et al. · 2017 · 21.9K citations · Full text
Maziar Raissi, Paris Perdikaris, George Em Karniadakis · Journal of Computational Physics · 2018 · 16.9K citations · Full text