Accuracy Improvement Technique of DNN for Accelerating CFD Simulator

  title={Accuracy Improvement Technique of DNN for Accelerating CFD Simulator},
  author={Yukitoshi Tsunoda and Toshihiko Mori and Hisanao Akima and Satoshi Inano and Tsuguchika Tabaru and Akira Oyama},
  journal={AIAA SCITECH 2022 Forum},
There is a Computational fluid dynamics (CFD) method of incorporating the DNN inference to reduce the computational cost. The reduction is realized by replacing some calculations by DNN inference. The cost reduction depends on the implementation method of the DNN and the accuracy of the DNN inference. Thus, we propose two techniques suitable to infer flow field on the CFD grid. The first technique is to infer the flow field of the steady state from the airfoil shape. We use the position on the… 

Figures from this paper

Multi-Fidelity Machine Learning Applied to Steady Fluid Flows

Predictive capabilities of the machine learning model are demonstrated in steady-state flow of incompressible fluid around a cylinder and a Joukowski airfoil, and the predicted flow field is used to warm-start CFD simulations to achieve acceleration in solver convergence.

Towards high-accuracy deep learning inference of compressible turbulent flows over aerofoils

The proposed deep learning method speeds up the predictions of flow fields and shows promise for enabling fast aerodynamic designs.



CFDNet: a deep learning-based accelerator for fluid simulations

This paper introduces CFDNet - a physical simulation and deep learning coupled framework, for accelerating the convergence of Reynolds Averaged Navier-Stokes simulations, and demonstrates the generalization capacity of CFD net by testing its prediction on new geometries unseen during training.

Machine learning methods for turbulence modeling in subsonic flows around airfoils

Reynolds-Averaged Navier-Stokes(RANS) method will still play a vital role in the following several decade in aerospace engineering. Although RANS models are widely used, empiricism and large

Global Approximation and Optimization Using Adjoint Computational Fluid Dynamics Codes

It is shown that an adjoint-based approximation model can provide increased accuracy over traditional nongradientbased approximations at comparable cost, at least for modest numbers of design variables.

Efficient Optimization Design Method Using Kriging Model

The present method is applied to a two-dimensionalAirfoil design and the prediction of flap’s position in a multi-element airfoil, where the lift-to-drag ratio (L/D) is maximized.

Self-adaptive-grid method with application to airfoil flow

A self-adaptive-grid method that is suitable for multidimensional steady and unsteady computations is described. Based on variational principles, a spring analogy is used to redistribute grid points

Transonic Axial-Flow Blade Optimization: Evolutionary Algorithms/Three-Dimensional Navier-Stokes Solver

The development of a high-fidelity aerodynamic design optimization tool based on evolutionary algorithms for turbomachinery is attempted and a three-dimensional Navier-Stokes solver was used for aerodynamic analysis, showing the superiority of the present method over the conventional design approach.

Multi-Objective Aeroacoustic Design Exploration of Launch-Pad Flame Deflector Using Large-Eddy Simulation

This study explores possible design of an aeroacoustic flame deflector for a rocket launch pad. The design objectives are 1) minimization of the overall sound pressure level near the payload fairing,

Drag Reduction of Transonic Wings with Surrogate-Based Optimization

Two different approaches to drag minimization in the design of transonic wings are investigated: (1) optimize all the section shapes and their twists independently (2) scale all the wing sections together and optimize them with their twists.

A Review of Surrogate Modeling Techniques for Aerodynamic Analysis and Optimization: Current Limitations and Future Challenges in Industry

A comprehensive and state-of-the art survey on numerous surrogate modeling techniques for aerodynamic analysis and surrogate-based optimization (SBO) is presented, with an emphasis on models selection and validation, sensitivity analysis, infill criterion, constraints handling, etc.