Rotating turbulence provides a simple configuration to study characteristic features and turbulent model performance in anisotropic turbulence. It has been observed in experiments and Direct Numerical Simulation (DNS) that rotating turbulence has many distinctive qualities, such as reduced kinetic energy dissipation, reverse kinetic energy transfer from small scales to large scales, quasi 2D flow at large scales, and cyclone/anti-cyclone asymmetry. A successful subgrid scale (SGS) model should be able to capture these features, and the challenge is to simultaneously reflect the anisotropic 3D nature of small scales and the primarily 2D nature of larger scales. A-priori tests and a-posteriori tests were carried out to examine model performance.
A-priori tests of models were performed using DNS results for forced isotropic and rotating turbulence. A range of models were tested varying from algebraic, gradient, and scale similarity, to one-equation viscosity and non-viscosity dynamic structure models. Anisotropy and Material Frame Indifference (MFI) requirements for models in rotating systems were reviewed and used to help construct new models based on the dynamic structure approach. The models were evaluated primarily using correlation and regression coefficients of individual components of the SGS tensor, components of the divergence of the SGS stresses, and the SGS energy production term. For all measures examined, the MFI-consistent dynamic structure models perform significantly better, especially for rotating turbulence.
At the a-posteriori test level, we evaluated models using two flow configurations: (i) homogeneous decaying turbulence; and (ii) rotating turbulence forced at large or intermediate scales. Testing was done for the first configuration through a systematic comparison between DNS results and large eddy simulation results at lower resolutions. The results were then analyzed in terms of several representative characteristics, including resolved kinetic energy, SGS energy production, molecular dissipation, and kinetic energy spectrum. The new models showed more accurate results than traditional models for almost all of these characteristics. The second configuration concerns the characteristic features of rotating turbulence. Again, compared to traditional models, the new models were better able to capture essential features of rotating turbulence.