Simulation strategies for complex turbulent flows

Abstract

Computational fluid dynamics in conjunction with the Reynolds-Averaged Navier-Stokes approach is nowadays routinely employed in a large variety of engineering and
industrial applications despite some well-known reliability issues in more complex flows. In this study, the performance of a state-of-the-art Explicit-Algebraic-Stress
Model (EASM) and a promising elliptic-blending approach is assessed on a range of test cases to predict complex turbulent flows. In an attempt to improve the quality of the predictions, near-wall-anisotropy modifications are introduced to the EASM, which provide better predictions for the Reynolds-stress and anisotropy tensor close to solid walls. In addition, a novel elliptic-blending RANS model is presented, which is based on the inverse turbulence time scale w, and which is equipped with a non-linear constitutive stress-strain relationship. The coefficients of the non-linear stress-strain relationship are obtained from the explicit solution of a Second-Moment Closure in the limit of weak equilibrium, and by imposing an internal consistency constraint and near-wall-anisotropy modifications, such that the highly anisotropic state of turbulence and the limiting two-component state is correctly reproduced at solid boundaries. The performance of the modified EASM and the novel elliptic-blending model are illustrated and assessed for a range of complex turbulent flows. It is expected that, due to ever increasing computational resources, unified or hybrid RANS/LES approaches will slowly penetrate into engineering applications where improved accuracy and reliability is needed. For this reason, a unified RANS/LES/DNS framework is presented, which is expected to provide the required amount of turbulence modelling for any mesh resolution and seamlessly operates between RANS and DNS mode. This is achieved by a revised Flow Simulation Methodology where the turbulence modelling contribution of a RANS model is rescaled using a damping function. The Flow Simulation Methodology is operating in conjunction with a newly developed damping function and a tailored convection discretisation scheme. In addition, a thorough calibration study is performed, which ensures proper turbulence resolving capabilities. It is conjectured that a sophisticated RANS model will also improve the overall quality of the predictions of any hybrid RANS/LES model. For this reason, the new elliptic-blending RANS model is incorporated, together with two successively simpler turbulence models, into the unified RANS/LES/DNS framework and the performance are assessed on a range of test cases, and compared to other widely used hybrid RANS/LES methods.

More information

Identifiers

ORCID for Richard D. Sandberg: orcid.org/0000-0001-5199-3944

Catalogue record

Export record