Hello all

I have a simple question about channel height in my simulation. I have a channel with two wall in top and bottom boundary channel and I use mid-plane bouncebackB.Cin these walls.If upper wall located on lattice 40 and lower wall located on lattice 0(zero), what is the height of my channel? 38 unit lattice or 39 unit lattice?

Many thanks in advance and with kind regards

Hello,

let us do it like this:

0 W

1 F

2 F

3 F

4 F

5 F

6 F

7 F

8 W

Two walls at 0 and 8. This makes 7 fluid nodes in between. The physical walls are midway between W and F, resulting in a width of 7, in your case 39.

Timm

thank you very much timm

I tried both channel heights a while ago using the fullway bounceback in a Poiseuille flow problem in combination with a Zou/He condition at the inlet. Using N nodes across the channel (including the wall nodes), defining channel height to N-2 gave me horrible results, whereas N-1 seemed rather good.

Is it correct then, to say that for midway bounceback the channel height should be N-2, and for fullway bounceback it is N-1? This seems to contradict what is said about this in the book of Wolf-Gladrow.

Elrohir

Hi

@Elrohir

As far As I know it should be the reverse

in midway bounceback the channel height should be N-1, and for fullway bounceback it is N-2

try to fix the corner boundary. it is always the souce of most problems in Poiseuille flow problem.

regards

I’m not sure if it is a good idea to use Zou and He AND bounce-back in the same code. Bounce back should place the wall half way between grid points and Zou and He is an on-node scheme. I imagine this will mean problems at the corners, as SaS says, and could lead to difficulties in stability and convergence with grid-refinement, as well as accuracy. It will probably be better not to combine them.

Note also that if you want to implement no-slip with bounce-back you should really use a TRT or MRT collision operator because BGK introduces a non-zero numerical “slip” velocity at the wall, which is a function of the relaxation time. Note that this is purely a numerical artefact - it depends on the resolution (quadratic in grid spacing) so it is NOT Knudsen or Navier slip. It is an error term.