airfoil optimizer chain with drag problem
to say it first: I'm a beginner in OF (vers 1.6).
Now my Problem: My aim ist to build a small optimizer chain to find the optimum low drag airfoil for casing of struts with a constant angle of attack of 0°. To reach this aim every optimizer iteration is done by following steps:
perlscript write_blockmesh_edgegrading.pl -> move blockMeshDict to /constant/polymesh/ -> blockMesh -> simpleFoam -> read out cd_value
This steps work automatic, but the result is not so nice :(
for example for a NACA0018 airfoil to case a diameter of 0.18m (=> chord 1m) the result of the simulation is approx: cd = 0,0155 with a look to the measured data of this airfoil it should be around 0.0075
So the result is with a faktor of 2 to high.
As far as I analysed the Problem the friction drag looks good, but the pressure part is to high. Resulting from this the optimizer converges to a unphysical result (NACA0004 looks a bit wrong. expected result around 0028), because it has a to high attitude to search for a low pressure drag and not for the combinition of pressure drag and friction drag.
what i already tried to solve the problem:
checking convergence: the residuals are <= 10^-6 and the drag value ist at it's final value after 400 of the 1000 iterations
changing turbulencemodell: i found in the forum that the SpalartAllmaras modell is used frequently for airfoil calculation. with this modell the cd converged to 0.04. with the RNGkEpsilon modell i had the above result of 0.0155. with all other modells i had worse results.
changing mesh resolution: there was no big change in results on rising the number of meshcells. (approx +- 2%)
searching the forum to find a solution: I found many hints, but they didn't give me a solution for my problem.
changing velocity: with the higher Re the cd value got a little bit smaller, but not in a significant size.
So my question is: can you please take a look on my casefiles attached?
As i worte above I'm a beginner with OF, so i fear there could be some "stupid" mistakes. ;)
thank you very much for your help.
Sorry, made a mistake with the uploaded case files. They had a mixed up in turbulence modell.
Sorry for that.
These results don't seem so wrong to me. You will rarely obtain the good drag values using RANS models.
On an airfoil at a low angle, an important part of the boundary layer is laminar, hence the friction drag is smaller than in a turbulent boundary layer. Traditional RANS models (SA, k-epsilon...) predict a fully turbulent boundary layer resulting in a higher drag than experimental data.
Concerning the bad result with the SpalartAllmaras model, did you have a mesh with y+ around 1 ? Unless you use wall functions (which are optional with SpalartAllmaras) you need a good resolution to resolve the near-wall region.
In my experience, near wall resolution gives a little better drag prediction than wall functions.
Hope this helps.
|All times are GMT -4. The time now is 15:20.|