Deep Learning to advance the Eigenspace Perturbation Method for Turbulence Model Uncertainty Quantification
This work addresses uncertainty quantification in turbulence simulations for engineering applications, representing an incremental improvement by integrating machine learning into an existing method.
The paper tackles the problem of large error and uncertainty in Reynolds Averaged Navier Stokes (RANS) turbulence model predictions by developing a machine learning approach to aid the Eigenspace Perturbation Method for uncertainty quantification. The result is a neural network that predicts discrepancies in the Reynolds stress ellipsoid shape, successfully identifying regions of modeling errors in turbulent flows compared to DNS, LES, or experimental data.
The Reynolds Averaged Navier Stokes (RANS) models are the most common form of model in turbulence simulations. They are used to calculate Reynolds stress tensor and give robust results for engineering flows. But RANS model predictions have large error and uncertainty. In past, there has been some work towards using data-driven methods to increase their accuracy. In this work we outline a machine learning approach to aid the use of the Eigenspace Perturbation Method to predict the uncertainty in the turbulence model prediction. We use a trained neural network to predict the discrepancy in the shape of the RANS predicted Reynolds stress ellipsoid. We apply the model to a number of turbulent flows and demonstrate how the approach correctly identifies the regions in which modeling errors occur when compared to direct numerical simulation (DNS), large eddy simulation (LES) or experimental results from previous works.