Eventos Anais de eventos
ENCIT 2020
18th Brazilian Congress of Thermal Sciences and Engineering
Strategies to mitigate the error propagation of explicit Reynolds Stress Tensor closures
Submission Author:
Bernardo Brener , RJ
Co-Authors:
Bernardo Brener, Roney Thompson, Matheus Cruz
Presenter: Bernardo Brener
doi://10.26678/ABCM.ENCIT2020.CIT20-0446
Abstract
Reynolds-averaged-Navier-Stokes is still the most applied approach in turbulent flow simulations for industrial applications. Despite the well-known lack of accuracy of the commonly used linear eddy viscosity models, the higher costs of more accurate approaches, such as LES (Large Eddy Simulation) and DNS (Direct Numerical Simulation), makes the cost/benefit relation of closures for the Reynolds Stress Tensor (RST) still competitive. Recent works employed DNS databases to analyze the ability of RANS equations to recover the mean velocity field, by plugging explicitly the DNS RST, as a source term. For the plane channel flow, it was shown that small errors in the RST could lead to large discrepancies in the mean velocity field. This result raised a concern about the conditioning of the RANS equations, i.e. whether this set of equations amplify small errors present in the RST closure. Some contributions in the literature proposed an implicit treatment of the RST as a way to obtain smaller error propagation in the recovered mean velocity field. These studies are of great importance for the emerging field of data-driven turbulence modeling as well as for the conventional turbulence modeling. The present work conducted an analysis into the strategies to mitigate this error amplification. Two different approaches in solving RANS equations with DNS data closure are presented, in order to elucidate the error amplification nature. Both result from decoupling the two main aspects in this study, namely the implicit treatment of the linear part of the RST with respect to the rate-of-strain tensor and the use of information of the DNS mean velocity field. It is shown that, although the implicit treatment process leads to smaller errors in the mean velocity profile, the major factor for accuracy gain is the use of information from the DNS mean velocity field. This analysis is confirmed for two problems, the flow through a square duct and another over a periodic hill.
Keywords
Error Amplification, Reynolds Stress Closure, Data-driven Turbulence Modeling, Explicit/Implicit Treatment
DOWNLOAD PDF VIEW PRESENTATION
