The current document has been compiled to provide information about the optimal selection and usage of turbulence models. It was pointed out that not all differences between CFD simulations and experimental data are the responsibility of the turbulence model, but that many different set-up details play a crucial role.

The turbulence modeling strategy at Ansys, designed around models was discussed and different sub-models of that family have been presented in comparison with different types of models. It was shown that the based models offer significant advantages in terms of integration to the wall, as well as increased accuracy and robustness for complex flows, especially flows with separation from smooth walls.

The motivation behind the tunable GEKO model has been presented. The influence of the different parameters has been shown for some building-block flows. A more detailed Best Practice document for the GEKO model is available ([Generalized k-omega Two-Equation Turbulence Model (GEKO) in Ansys CFD]), in addition to a Best Practice Document for Scale-Resolving Simulations [1].

Numerous sub-models and model extensions have been discussed and their effect on the simulation has been presented.

An area of active research in turbulence modeling is Machine Learning (ML). The topic has not reached the maturity required to be included in a Best Practice document. However, Ansys is currently developing a comprehensive infrastructure for optimization of GEKO parameters using ML. This infrastructure is based on an adjoint solver for the turbulence model in Ansys Fluent including non-linear EARSM coefficients. The infrastructure will enable users to employ ML concepts to their own CFD applications in an integrated and convenient fashion.