Formulating turbulence closures using sparse regression with embedded form invariance
Sarah Beetham, Jesse Capecelatro
Abstract
In recent years, machine-learning techniques have been leveraged to improve closures for the Reynolds-average Navier-Stokes equations. The efficacy of a sparse regression-based method that results in compact, algebraic models and ensures key physical properties such as form invariance is demonstrated. Further, success for learning accurate models from data spanning full high-fidelity sets to experimental (sparse and noisy) data is shown.
Topics & Concepts
FidelityTurbulenceRegressionKey (lock)MathematicsRegression analysisAlgebraic numberApplied mathematicsComputer scienceReynolds-averaged Navier–Stokes equationsArtificial intelligenceAlgorithmMachine learningMathematical analysisStatisticsPhysicsTelecommunicationsComputer securityThermodynamicsFluid Dynamics and Turbulent FlowsModel Reduction and Neural NetworksProbabilistic and Robust Engineering Design