Reasonable Mesh resolution #284
Description
Hello,
Thank you very much for FluidX3d software. I can see a problem which I would like to discuss.
The problem is that one has to have a GPU with enormous amount of GPU memory to get a fine enough mesh for his model.
For airfoils - a high lift small wing - for example, I've used up to 14 GB GPU memory and the mesh is very very bad - the curvature is absolutely damaged - while FluidX3D shows many hundreds of millions of cells which is totally unreasonable for me!
I've also seen the same for your examples of space shuttle, cow and others.
Actually I could achieve a much better mesh and accuracy and finish the CFD with very reasonable time - ~ 45 minutes - with few millions of cells using snappyhexmesh and openfoam-based CFD.
I think the problem is that you're using the same cell size for the whole domain? While on snappyhexmesh we use adaptive meshing with refinement towards the walls, and the walls itself with much high refinements with layers and smaller cells in between flow and walls.