[keitaro7_14]

Hello everyone,

I'm trying to simulate a wind velocity field in a scenario where there is a porous wall. However, I'm not totally convince, since I think that the results are not realistic.

If I simulate the velocity field in a scenario with a wall (no porous), using simpleFoam (steady-state turbulent k-epsilon simulation), the behaviour seems reallistic.




The initial U profile is recovered near the right boundary (Yes, I know that I should enlarge the domain to achieve a quasi totally recovered profile).




The nut field is also coherent. The turbulence increases near the wall due to the vortex formation, and it begins to disappear near the right boundary (again, I should enlarge the domain)

However, If I use Porousbafflepressure BC to add the porous wall, or if I use porousSimpleFoam to add the porous wall implicitly (I have checked that the results are identical), the result is strange for me. With a moderate pressure drop (Dcoef = 10^5), this is the result:



The U profile is recovered (a little bit) near the right boundary. But it is needed 1km practically. Well, probably it is true, I don't know.



However, the corresponding nut field is very strange. Why is the vortex created so far of the wall? Is it normal? How can it be formed at 800m from the source of discontinuity? I'm freaking out!

The boundary conditions are very simple. No slip in the atmosphere, a uniform profile in the inlet, a zero gradient in the outlet. I think it is not the problem.

Can someone with more experience than me enlighten me? I only want to know if it seems realistic. My main doubt is related to the nu field in the last picture. Can the nu field be formed so far from the wall?

Thank you so much!!

[keitaro7_14]

Well, I solved the problem. So, I'm going to explain in case it helps someone.

Firstly, we have to know that porous fences behaviour is different to the solid one. In a porous fence, a bleed flow appears, while porosity is low. In this paper is clearly explained: Xu, Y., & Mustafa, M. Y. (2015). Investigation of the structure of airflow behind a porous fence aided by CFD based virtual sensor data. Sensors & Transducers, 185(2), 149.

This bleed flow is disturbed when it goes through the pores, and it can increase the turbulence in consequence. In this paper a LES simulation is performed, and the turbulence can be seen at a great distance from the fence (Figure 6): Maruyama, T. (2008). Large eddy simulation of turbulent flow around a windbreak. Journal of Wind Engineering and Industrial Aerodynamics, 96(10), 1998-2006.

To be sure that you have a physically correct solution, my recommendation is to validate the model in a wind tunnel or in other experiment. If it is not possible, as in my case, it is highly recommended reproduce some scientific paper. I tried to reproduce this: Bourdin, P., & Wilson, J. D. (2008). Windbreak aerodynamics: is computational fluid dynamics reliable?. Boundary-Layer Meteorology, 126(2), 181-208.

In this case, I noticed that my initial values for k and epsilon fields were horrible and, in consequence, my results too. When I introduced the boundary conditions correctly, I could reproduce the paper, and "validate" my model.

Now, I know that I must take many precautions, specially with k and epsilon.

PD: Previously, I revised that the discretization was correct and that the mesh was smooth enough.

[RicardoLB]

Hello!
I am trying to do something similar (a net in water), but I am getting very different U profiles... Would it be possible for you to send me the case file?

Greetings,
Ricardo

[keitaro7_14]

Which is exactly your problem? I cannot send you my case exactly because it depends on external (and confidential) files (I call several subprocesses before the solver, and the boundary conditions are also customized), but I can try to help you of course.

I noticed that the results were very dependent on the values of k and epsilon.

Regards

[RicardoLB]

Hey,
As you mentioned, the k and epsilon values are fundamental...

I was getting this behaviour and no convergence found in the first image.


Now I get a solution which matches (at least qualitatively) my sources with convergence (second image).

I would say, problem solved. Thanks!