[bgp723]

Hello,


I am coming back to CFD world after a long hiatus and am doing some turbulence modeling. One thing I'm trying to do is to quantify the system level CFD model uncertainty by varying various inputs. One thing that piqued my interest was this paper:



Schaefer, J., et al., 2017, "Uncertainty Quantification of Turbulence Model Closure Coefficients for Transonic Wall-Bounded Flows" where Table 3 gives the SST constant default values (these defaults agree with Menter, 1994 paper provided values and CFD-Online wiki's constants).


I wanted to also apply similar ranges provided in the paper into my system level model to see their impact on my variables of interest (heat transfer coefficients on some of my surfaces), but what surprised me was that it seems that ANSYS Fluent defines four coefficients inversely. Namely, sigma_k1, sigma_k2, sigma_w1, and sigma_w2 :
https://www.afs.enea.it/project/nept.../th/node67.htm


Can someone smarter than me please confirm that ANSYS Fluent defines these inversely than the established formulation of the SST models? It seems very odd to me that they would use the same variable name as the original paper but define them inversely.


Thankfully, changing these variables didn't influence my results significantly, though I did notice that changing ANSYS Fluent's sigma_w2 from default of 2 to 0.5 did significantly improve my solution stability (I had slight oscillations before), though I understand that it's very dangerous to tweak these parameters without fully understanding what these are supposed to do, so I would like to get some confirmation.



I appreciate any feedback.
I couldn't find any previous forum post talking about the inversed definition for these constants, but if I missed any, please let me know. I'll keep looking around in the meantime.



Thank you!
Dan B

[LuckyTran]

Firstly, it would be absolutely tragic if Menter implemented his own model wrongly in Fluent...

This irony being put aside, yes they are defined inverses/recipriocal compared to one another. In older literature you follow in the forward fashion of the eddy viscosity approach and multiply the gradient by the effectively turbulent conductivity. The implementation in Fluent follows the new meta of defining additional transport parameters via turbulent Prandtl/Schmidt numbers (which means you divide the turbulent viscosity by the turbulent Prandtl/Schmidt number). Otherwise, sigma is not the turbulent Schmidt number and should be written using a more appropriate symbol. Beyond simply a notation difference, it gives you coding fleixbility because now you can instantly swap out the molecular/turbulent Prandtl/Schmidt numbers and you have a solver that easily switches between laminar and turbulent cases with no additional hacking of the code needed.


tl;dr don't label things sigma if they're not sigma

[bgp723]

Thank you for the quick reply with explanation and for the additional insight!

I will make sure to highlight the differences in the labeling of the variables when referencing these.



Sincerely,
Dan B