Hello,

I'm considering the SST k-omega model for my finite volume solver for the compressible Navier-Stokes eqs. Thinking about the implementation of this model, I came up with some details that I don't understand. I hope you can help:

(1) Boundary conditions:

-------------------------

I bought (yes, bought, I don't have access to journals) the AIAA paper 93-2906 by Menter which explains the model. It does not mention which boundary conditions to apply at various types of boundaries. I sought help in the CFD wiki article "Turbulence free-stream boundary conditions" but this seems to be more about how to specify initial conditions. In particular, I would like to know what to do at no-slip walls and inflows/outflows.

For the moment, my solver implements rather simple BC's. At no slip walls I specify zero flux of momentum and I intended to do the same for k and omega. At inflows and outflows I use a flux vector splitting in the spirit of Steger and Warming to convect ingoing and outgoing information according to the characteristics. I was also planning on using a similar approach for k and omega. But I lack the insight to assess if this is ok or just plain stupid.

(2) Computation of blending functions

--------------------------------------

Referring to the expressions here:

http://www.cfd-online.com/Wiki/SST_k-omega_model

there seems to me to be a potential for numerical problems whenever omega or y approach zero, because they occur as denominators (fx see F1). Would you normally in an implementation introduce some mechanism that detects values of omega close to zero and then switch to some better behaved expression (e.g. by replacing omega with a larger "safe" value) ?

For y (the wall distance) to actually become zero I would have to evaluate the blending function at wall points. Presumably, then, this means I should not do that. This makes sense in as much as the blending function should be 1 in the near wall region anyway, so there should be no need to evaluate it. It does make me wonder, however, if I really understand what is going on. After all, shouldn't the SST model be valid right down to the wall, and hence it should be possible to evaluate all its expressions at the wall ? Or is there some tacid assumption that whenever you need to evaluate at the wall you replace the blending function expression with a hard-coded 1 ?

Thanks, Martin

Just as a clue refer to Peric codes (consult his book too), there u find implementation of both of k-eps and k-omega, maybe helps.

peric code URL: ftp://ftp.springer.de/pub/technik/peric/peric.tar.gz

good luck

Thanks for your reply.

This code uses the original Wilcox model with wall functions though, so it's not directly applicable.

I'm still not sure what to do at walls and inlets/outlets. I spent a great deal of time examining what happens to omega as y approaches zero. It seems that omega approaches a value inversely proportional to the square of y. So now I'm thinking about just setting omega to this value at walls, and setting other turbulent quantities (k, and turbulent viscosity) to zero there. But again, I lack the insight to know if that makes sense-

Niels,

For Menter's papers (92 and 93), they are free : http://ntrs.nasa.gov/search.jsp

Regards