Recirculation bubble and wall shear stress
 

hi
I'm simulating an atmospheric boundary layer with a perturbation (an obstacle) on the bottom of domain. Downstream of the obstacle there is a Recirculation bubble, and i want to calculate the wall shear stress on the bottom. But the wss on the bottom has an overshooting when start the Recirculation bubble. How can i try to eliminate it? it's very important

Thanks

No idea for this problem?

I'm trying to use this Grad scheme in my fvSchems
gradSchemes
{
default Gauss upwind;

}

But I have this error when simulation start


--> FOAM FATAL IO ERROR:
attempt to read beyond EOF

file: /u/acconcia/OpenFOAM/acconcia-1.6/run/blayer/system/fvSchemes::gradSchemes::default at line 26.

From function ITstream::read(token& t)
in file db/IOstreams/Tstreams/ITstream.C at line 83.

Hi Daniele,

You may have a parenthesis problem in your file. Can you check it and share it if you didn't solve the problem ?

Thanks
If I use Gauss linear there isn't error

Have you tried this ?

gradSchemes
{
default Gauss linearUpwind;
}

Does Gauss linear works ? Because it is second order so more precise.

Yes I used it, but I have the same error, I'm traying to eliminate the stress overshooting but I don't understand where is th problem

Ok, so the problem is in your file, can you share it?

I don't see the problem on the results, you have a recirculation so the flow is stalled after the obstacle so the stress become null at the stall point and then negative. Because your shape is not smooth (not derivable actually), the velocity might have a discontinuity or a jump anyway so its gradient will have an overshoot and consequently she wall shear stress too.

Yes, but if you consider the backward-facing step the wall shear stress haven't overshooting

That would more be linked with your mesh I think. If it is sufficiently refined, then the singularity should disappear. It seems to be coarse when I see the x variations of the wss which leads to the discontinuity that I pointed in the previous post. Well, that is what i think.
Could you post a screen shot of your mesh please ?

I tried to refine the mesh but the overshooting don't disappear. I used 5000 cell over a 500 m domain.

Are you using 5000 cells in the X direction or 5000 cells on the whole domain ? Is your mesh 2D ? How many cells do you use globally ? Did you refine the mesh close to the obstacle ? Can you post a screenshot of your mesh ?

The correct data are:
2D domain along x (main flow direction) 1000 cells
along y 100 cells
along z 1 cells

Have you checked the y+ values ? Are they in the correct range for your turbulence model ?

How can do it? What is the correct range for the k-eps model?

Thanks you

There is an utility for that, yPlusRAS, and the correct range depends if you are using wall functions or not, which you should as you use a coarse mesh. Then the first cell point must be in the logarithmic area so with y+ between 50 and 200.

I use wall function, with the mesh that I post. The y+ was more or less 800, now I change the mesh and it is 150

Ok, you can then play with that to check if your results can be improved but that should be ok like that.
I think that you could use much more cells in the region which you are interested in as you only use about 20 cells in it (if I had understood well your screen shot) and the gradients are here much more important than anywhere else.

Refined mesh. Then I'll post the simulation results

I used a vertical step in my profile and this is the wall shear stress at bottom, They seem numerical error, because if I refine the mesh on chenge of shape don't change anything.

I try to use with my case the mesh of pitzDaily tutorials, I only change mesh, and in this case a I haven't overshooting on step. Why on my mesh is there this overshooting?

I can see two jumps in the size of the cells in the x directions, of a factor 2 or 3, which is not a good meshing practice. You should use very smooth variations in the cell sizes, in all directions. Variations of 10 % between two consecutive cells, especially in critical areas such as the boundary layer, is sometimes cited, which is not stupid in the experience I had.

Cheers

Francois

I eliminate the overshooting I change mesh add a cell where the shape change slope

That seems to be better, did you keep the same cell height at the wall because of the y+ criteria ? Any improvement in the results ?

No because the three cell near the slope has different areas the solution is bad

I am sorry but I don't understand.

My English is very bad!:p
In my previous mesh near the slope i have only two cell. In my new cell three. Three cell with at least a vertex shared with the vertex where th obstacle changes its slope.

I don't understand why the mesh configuration could change the geometry. This is more a problem of how you do the mesh.

Uhmm I don't understand your question. Sorry

I think I did not understand what you said before.

-In my previous mesh near the slope i have only two cell. In my new cell three.
Here you say that you didn't use the same number of cells in the slope, is it that ?


- Three cell with at least a vertex shared with the vertex where th obstacle changes its slope.

And here that this change in the mesh configuration changes the slope of the obstacle.

This is not normal. The geometry should be fixed and then you do the mesh on it. If the way you do the mesh changes the geometry it is because you don't do the mesh in an appropriate manner.

No. I didn't change the shape of th obstacle, I divide my domain i four hex in my blockMeshDict, so near the vertex where start the step I have more cell. The cell number, two or three, aren't the number of cell on slope, but the cell number that surrounding the vertex where start the step

hi
I return on this problem. I must try to solve it with a monoblock grid instead of a multiblock grid. The problem isn't the separation a use obstacle with high aspect ratio, and I haven't overshooting. I think that the problem is the point with differet value of derivate (left and right). Is possible resolve it?