Problem with Turbulent Heat Flux Temperature boundary condition
 

Dear All,

I have a problem with the compressible::turbulentHeatFluxTemperature boundary condition. I am performing an heat transfer computation with some heat source surfaces. On these surfaces I want to impose fixed heat fluxes (in the range 500-30000 W/m2). The problem I encounter is that, starting from the first iteration, the temperature diverges approaching unphysical values on the corresponding surfaces. I tried to bound the temperature (bounding value 2000 K). Unfortunately this does not help the temperature to approach realistic values and furthermore, using the wallHeatFlux utility I checked that the "real" heat fluxes acting on these surfaces are different from the imposed ones.
Have you got any idea about the reason why this problem happens??

Thank you.

Manuel

What turbulent model are you using? What are the y+ values at the heated wall? Corresponding on the wall functions for kappa (or alpha) it is based on viscosity (mu_turbulent/Pr_turbulent). And for some wall functions of mu_turbulent, it might not be calculated if y+ is too small. So mu_turbulent might be only mu_laminar, so the heat transfer becomes much too low => temperature too high

Dear jherb,

I am using the kEpsilon turbulence model, with wall functions only for k and epsilon (no wall functions for mut and alphat, they are simply calculated on the walls). I am using the wall functions even if the mesh is coarse at walls and yPlus is outside the log-law region. I can't avoid to have large yPlus values since my mesh is already very large and making it finer is not feaseable for me.
Should I use another treatment for turbulent quantities at walls? I tried to bypass the wall functions using directly fixed boundary conditions for k and epsilon (k, epsilon = 0 at walls). Anyway the solver automatically assigns the wall functions and does not consider my fixed value boundary conditions.
Have you any suggestions??

Thank you,

Manuel

Hej,

you need to set wall functions for alphat and mut as well. For mut you probably want to set mutkWallFunction and for alphat you can either use alphatWallFunction, which is very basic or alphatJayatillekeWallFunction, which is a more advanced wall function which gives good values over a large range of Reynolds and Prandtl numbers.

Dear romant,

Ok. I am going to try in accordance with your suggestions. Anyway, don't you think that applying wall functions outside their region of validity (I mean log-law layer) could be improper and/or lead to erroneous results?

Thank you,

Manuel

Yes, outside the region it will lead to erroneous results. However, the validity region is quite large 30<y+<150 (sometimes even larger). So if your mesh does not fall within this region, one option is to adapt the mesh.

You won't be able to have anything else at the wall, there is just no models (to my knowledge) that would support larger y+ at the wall and going with something that is laminar is definitely wrong.

Dear romant,

Ok. I'll try with these 2 wall functions and I'll let you know back.
Thank you,

Manuel

Dear romant,

I tried to switch the turbulence model from k-epsilon to realizableKE. The latter one automatically assigns the wall functions for mut and alphat.
Unfortunately this has not brought up any improvement.
What do you suggest to do??

Manuel

Try to set up a smaller case, similar in physics to the one you have and see if you can get it to work.

Also, you need to have your y+ in the range of the wall functions, this is then the coarsest you can go at the wall, otherwise your results will most likely be incorrect.

At the moment I would stay with just k-epsilon.

Which wall function do you use for mut, k, epsilon, alphat? Which boundary conditions do you have for p_rgh, U and T?

You still have to specify the wall functions for mut and alphat in the corresponding files (mut and alphat in the 0 directory).

Dear All,

I thank you for your suggestions.
I remeshed the domain in order to guarantee a yPlus being in the validity range. In this way results quality improved significantly. The maximum temperature value is decreased drastically but it still remains unphysical. I would try to switch the turbulence model form realizableKE to k-w SST. Do you think that this change could produce better results? Do you have other suggestions?

Thank you,

Manuel

I am faced with the same problem here.