Turbulence Models

Modeling is currently THE greatest bottleneck in high Reynolds number CFD solutions.  Everything else done in CFD research will be largely a waste of time if current turbulence models are not improved dramatically.  And No - LES (Large Eddy Simulation) alone will not be the answer to this problem.  

Collision Models

• These models treat turbulence as a collection of particles or a colloidal suspension of particles (eddies). 

• These models are more expensive than RANS models but are exact in many limits (and therefore tend to be far more accurate) and contain far fewer tunable model parameters.

Universal RANS/LES Models

• RANS and LES are not separate modeling approaches.  They can be seamlessly merged. 

• This is not a Hybrid approach where two models are smashed together.  It is a single model that automatically does RANS, URANS, LES, or DNS based on the resolution of the mesh provided and without any user intervention.  

Turbulent Potential Model

• This model uses a Helmholtz decomposition for tensors (which probably has been done before - but we have never found a reference).

• This simplifies the Reynolds Stress Transport equations into uncoupled scalar and vector transport equations.  The resulting system has fewer variables, is easier to solve numerically (because the equations are uncoupled), and is easier to model. 

Dissipation Modeling

• In the limit of strong inhomogeneity the dissipation is dominated by the gradients in the turbulence and can be modeled exactly.

• It is also possible to create models which exactly capture homogeneous dissipation in the high and low Re limits. 

Back to top