Hi all! I'm trying to simulate the flow over a car in 2D. I'm using Tri elements for the mesh (run in Gambit). My doubt is that I don't know the convenience or not in using boundary layers to define the mesh in those regions closer to the car's surface. Can anybody give me advise about it and, in that case, the criterion to select the thicknes of that boundary layer?

Thanks all, again

Fer

hi Fer, If u r not comfortable with boundary layers..u can go ahead with meshing the edges and then proceed towards meshing the face. The best thing is to divide the geometry into sections and then mesh. Using boundary layer is always best because it gives u total control over the thickness of the cells near the boundary. For estimating the grid spacing..it depends on the turbulence models...that is either u r going for y+ <1 or y+>30 U can estimate it by using the following link http://geolab.larc.nasa.gov/APPS/YPlus/

hope this helps

bharat

Thanks very much bharat,

I have made some things you told me in your post: dividing the whole geometry into different pieces (to control better the mesh) and calculate the ds distance (function of y+, Reynolds,...). I have used a K-e+RNG turbulence model and Standard Wall Treatment (and also Non-Equilibrium), so in that case y+ is recommended to be in (30,300) (better in (30,150)). After simulating this y+ parameter plotting fits quite well with that recommendation. The doubt is if the shape of the boundary layer (with edges more aligned at the flow near the walls than the Tri elements) works better than meshing with those Tri elements until the wall surface.

Thanks with your help again,

Fer

Try to use a boundary layer mesh using quad elements around the car body (walls). The distance upto which you should use a B.L. mesh depends on the problem, specially velocity. You can choose a couple of distances. For the rest of the domain use a tri mesh as you are using. Also you can use y+ adaption if your initial B.L. y+ is higher or lower than what you need. As you are doing a 2D case, trying a few different grids with different B.L. thickness would not take much time.

The last question,

I suppose that using boundary layer of quad elemets aruound the car body and Tri mesh for the rest supplies better results than Tri mesh for the whole fluid area, but does it improve the convergence process?

Thanks bharat and Anindya for your help

Fer

Hi Fer, I think using quad elements at the boundary would reduce the number of cells as compared to tri elements. I think it should not affect the convergence (open to doubt).

hope this helps bharat