wallShearStress for compressible cases is off - possibly by density rho
Arcadia197 opened this issue · 3 comments
Hey Timofey,
I noticed something odd when postprocessing results for compressible simulations with the NASA CRM-HL using your library. It seems like the field estimated from the wallShearStress needs to be multiplied with the density. Since my current setup runs with a density of around 0.2 this becomes rather noticeable in comparison to air at standard conditions.
I did not check any of the formulations yet, but for the OF version they do differentiate between compressible and incompressible cases with differences also hinting on the density. You can find it here with the corresponding part in lines 182 to 208.
I added some experimental results for assessing the correction. The first is a view from the top of the airplane while the other is from the bottom. They have been computed in ParaView using:
abs(mag(wallShearStressMukha) / mag(wallShearStress) * rho - 1)
Fuselage or pressure side looks well, but suction side shows quite some differences, linked to #8 I guess. Flow is largely separated on suction side since I run on a very coarse mesh.
- Julius
Hi! I'm pretty sure you are right! What is computed in the incompressible case is, in fact, the stress divided by rho, since the whole RHS is divided by rho in the derivation of the equations. The same thing is now done for compressible, i.e. in the standard TBL case the we have (nu + nut)du/dy, whereas it should be mu+mut. See line 426 here
We need some clever if statement there, which would tell us if we are in a compressible solver. What would be a good proxy for that? Maybe check if there is a rho volScalarField in the registry or something like that?
Nice CRM pics though :-)
Hi @Arcadia197 ! This should now be fixed on the develop branch hopefully. What I do is grab muEff at the wall from the turbulenceModel class and use that to compute the stress. This should work for both the compressible and incompressible case.