Enhancing the shear-stress-transport turbulence model with symbolic regression: A generalizable and interpretable data-driven approach
Chenyu Wu, Yufei Zhang
Abstract
Data-driven Reynolds averaged Navier-Stokes (RANS) turbulence models for separated flows based on black-box machine learning models have been widely researched in recent years. However, they often lack generalizability and interpretability. In this work, field inversion and symbolic regression (FISR) are used to develop an interpretable and generalizable data-driven RANS turbulence model. The proposed turbulence model shows good accuracy in various test cases that are completely distinct from its training set.
Topics & Concepts
Reynolds-averaged Navier–Stokes equationsInterpretabilityTurbulenceGeneralizability theoryTurbulence modelingReynolds stress equation modelSymbolic regressionRegression analysisComputer scienceK-epsilon turbulence modelMachine learningMathematicsK-omega turbulence modelStatisticsPhysicsMechanicsGenetic programmingFluid Dynamics and Turbulent FlowsModel Reduction and Neural NetworksFluid Dynamics and Vibration Analysis