[sa har]

I try to simulate natural convection in channel that has rectangular fins in ansys fluent.i use boussinesq model but i dont know which turbulence model is appropriate for it.
1.k epsilon model with wall function (y+>30)
2.k epsilon model with enhanced wall treatment(y+=1)
3.standard k omega (y+=1)
4.sst k omega(y+=1)

which model is good for channel with fins?and how can i reach to intended y+?

when i calculate y+,i should choose one of the turbulence model or choose one it and try to make a fine mesh. which of this way right?

[LuckyTran]

Make sure you are supercritical, it's fairly difficult to achieve turbulent flow in a natural convection problem. You generally need more than 1m of length to reach the critical Reynolds number.

The k-epsilon with enhanced wall treatment and kw-sst are fairly reasonable general purpose models.

You need a sufficiently fine mesh to achieve a desired y+. You do not know y+ ahead of time and you should try the calculation a few times to see what y+ you get from your solution, and then update your mesh finer/coarser as necessary.

The actual value of y+ will depend on the solution and also the turbulence model used (you'll get different y+ using different turbulence models).

[sa har]

Originally Posted by LuckyTran

Make sure you are supercritical, it's fairly difficult to achieve turbulent flow in a natural convection problem. You generally need more than 1m of length to reach the critical Reynolds number.

The k-epsilon with enhanced wall treatment and kw-sst are fairly reasonable general purpose models.

You need a sufficiently fine mesh to achieve a desired y+. You do not know y+ ahead of time and you should try the calculation a few times to see what y+ you get from your solution, and then update your mesh finer/coarser as necessary.

The actual value of y+ will depend on the solution and also the turbulence model used (you'll get different y+ using different turbulence models).

thanks alot foryour answer.
you are right,my channel length is more than 1 m and rayleigh number more than 1*10^9 then flow is turbulent.(i solve problem with steady state assumption)
if i use sst k-omega,should be y+ less than 1? a friends said to me that i cant use enhanced wall treatment (k-epsilon) when channel has fins is it right?

[LuckyTran]

Originally Posted by sa har

if i use sst k-omega,should be y+ less than 1?

SST needs a fine mesh. Otherwise it is equivalent to running a high Reynolds standard k-epsilon model.

The y+ does not have to be 1, what's more important is getting a sufficient number of points to resolve the boundary layer with at least 1 or 2 points in the linear regime. You can have y+ 2 or 3 or 4 and still get great results.

Originally Posted by sa har

a friends said to me that i cant use enhanced wall treatment (k-epsilon) when channel has fins is it right?

"can't" is a strong and one of the most biased words to use in modelling. You "can" use it, just click the stupid button in Fluent. Maybe your friend can justify why other models are not appropriate. You may ask someone and they will tell you that you cannot model real substances as a fluid with locally uniform properties. Then what do you do? Give up?