I'm trying to evaluate the drag of a concept vehicle, but I can see a true sensivity to the mesh size. Yet I've done 3 simulations.
1) Auto mesh - approx. 360k cells
2) Auto mesh + BL refinement(10 layers, 1.15 gradation, 1.05 factor) - 1.7 mln cells
3) Auto mesh + mesh refinement(around and wake) - 2 mln cells
In case 1) the cd = 0.31
In case 2) the cd = 0.23 - 23% difference to 1)
In case 3) the cd = 0.13 - 23% difference to 1)
Basically in case 3) I can see a clear problem, which I suppose is an application problem, the mesh apperas to be "cracked" in the volume and visible all times, even when I'm in shaded view or outline view. What's the reason of that?
I can't make more than 2 mln cells study because I'd get stucked with the time - I've got only 6GB RAM so the last case takes over 10 hours of running...
PS: I can't make an adaptive cycle from case 1) because it fails in the next steps on surface meshing...always, co it also might be a problem with the geometry.
Ok, fine, after all I came up that this could be an application internal error, maybe connected to the GPU rendering because when I switch between other scenarios and go back to the "cracked" one, then everything is fine, the weired mesh is no longer a problem. But the problem with adaptive meshing for the geometry still cannot be resolved, I think that's truely bad geometry.
Because I'm writing me thesis based on this CFD analysis,another question is, how can I model the yaw angle wind stream if I'm using a half-body model. When the body is full, then there is no problem, "turning" created external volume at an angle, but when you have half body, then you cannot turn it, it's obvious why. It could help me because as I said, I can make a better flow study using half body, because then I can use more mesh on less volume, so my results are more accurate. When I have a full body, then I don't even try to simulate 3 or 4mln cells because I would wait days...