Non-linear Subgrid Scale Model

[hgopalan]

The calculation of surface stresses has a significant effect on the profile of the mean wind and stresses. Smagorinsky model is known to over-predict the wind speed profile. The non-linear sub-grid scale models have been proposed to improve the wind speed profiles.

I have added the non-linear sub-grid scale model of Kosovic to the AMR-Wind turbulence model list. The model also includes a hybrid RANS-LES blending term at the surface. This significantly improves the mean wind speed profile when compared with log-law for the neutral boundary layer (stable and unstable under testing). Finally, there is an extra option for turning of the sub-grid scale model at certain height above the ground and the model will operate in hybrid RANS-LES-ILES mode.

[moprak-nrel]

Can we add documentation similar to the other SGS models here: https://exawind.github.io/amr-wind/theory/theory.html#turbulence-models ?

[hgopalan]

Added unit test.

[hgopalan]

Added documentation for the model.

[hgopalan]

Added unit tests for the nonlinear sum operator and non-linear model. The model has been tested for neutral, unstable and stable stratification. The model has not been tested with actuator line model and multiphase flow. Future work can include the modification of the One-Equation model to add the extra turbulence production/destruction of TKE due to the non-linear term.

Neutral: Surface RANS length scale correction improves the results Weakly convective: Surface RANS length scale correction improves the results Convective and Stable: Surface RANS length scale correction does not offer any advantage.

[mbkuhn]

I confirmed that the unit tests pass on GPUs and the new reg test runs on GPUs.