[RabiArya]

I want to understand how the KW-SST model converts into K-E model and K-W model according to required conditions. What changes (like functions, variables, etc) are used to directly convert from one system of equation to another Or by using some if statements to switch between the two different models according to different situations? Basically how does this KW-SST model alone accomplishes work of both equations alone?

[LuckyTran]

There's a (actually several) blending function(s) that blends the turbulent viscosity in space. The gist of it is... you calculate the turbulent viscosity everywhere as if you were using k-epsilon. You also calculate the turbulent viscosity everywhere as if it were a k-omega. So now you have two values for the turbulent viscosity. Then you use the blending function to smoothly choose the regions in space where you prefer the k-epsilon one versus the k-omega one. In principle you could hard-switch between models. But in practice and for commercial implementations you use a smooth blending function.

As for how this blending function works. It's a hyperbolic tangent (tanh). There's also a bunch of sanity checks involved and the max function gets called several times. If blending function sounds too fancy then you can think of it as interpolating between the two turbulent viscosities using y*. That's not the end of the story of course. There are a few more modeled terms (like the production of k and dissipation) that also needs blending.

If you read literature on the topic, these are the so-called F1 and F2 blending functions.

[piu58]

The blending makes sense because k-w works well in the near of boundary but gets instable in the free stream far away form any boundary. k-e woks not so well at boundaries because you cannot have a fine mesh here. But it is stable in free stream.

[RabiArya]

Originally Posted by LuckyTran

There's a (actually several) blending function(s) that blends the turbulent viscosity in space. The gist of it is... you calculate the turbulent viscosity everywhere as if you were using k-epsilon. You also calculate the turbulent viscosity everywhere as if it were a k-omega. So now you have two values for the turbulent viscosity. Then you use the blending function to smoothly choose the regions in space where you prefer the k-epsilon one versus the k-omega one. In principle you could hard-switch between models. But in practice and for commercial implementations you use a smooth blending function.

As for how this blending function works. It's a hyperbolic tangent (tanh). There's also a bunch of sanity checks involved and the max function gets called several times. If blending function sounds too fancy then you can think of it as interpolating between the two turbulent viscosities using y*. That's not the end of the story of course. There are a few more modeled terms (like the production of k and dissipation) that also needs blending.

If you read literature on the topic, these are the so-called F1 and F2 blending functions.

To use a linear(or whatever maybe) interpolation based on the y* value, there must be a boundary designated for certain values of y* for viscosity calculated using kw and certain for value obtained using kE. I mean like there must be a value of y* where it should be pre described that for that value only KW method is to be used and below some KE should be used. Thus, interpolation can only be done within the intermediate values of y*(lies between threshold value of y* for kw and ke) . Hence I want to know what are these exact values of y* between which the interpolation is done ? What are the threshold(y* at extremes) for kw and ke to be used?

[LuckyTran]

Other implementations sure but in menter's kw-SST there isn't a hard cutoff threshold where it switches. Menter's SST uses a blending function to interpolate smoothly. Just look up the F1 and F2 blending functions if you want to know exactly what is going on because it's not simply switching based only on the y*.

Also you have it backwards. For low y* you want to use k-omega and high y* you want to use k-epsilon.

Google it. I's on our wiki. You can find it on NASA website. You can find it in the original publication. You ca find it in the help documents of Fluent, Star. Lots of places. It's no secret. I have assumed we are talking about Menter's version since that's the one most commonly referred to as "the kw-SST model." If you are looking for a non-Menter one, then please specify. Speaking of which, you are referring to an entire family of models. Difficult to tell you exactly what the "threshold" is if you don't tell us which particular implementation you are referring to.

[sbaffini]

There are no such things as boundaries and thresholds in SST model. Look here:

https://turbmodels.larc.nasa.gov/sst.html

It's just two blending functions, F1 = tanh(arg1^4) and F2 = tanh(arg2^2), where the arguments are function of k, omega, wall distance and several other constants.

Of course, there is a rationale, see for example here:
https://pdfs.semanticscholar.org/767...405.1613660237

[RabiArya]

Thank you both, got a lot to learn ☺☺