Flat 2D surface immersed in 3D flow, is it possible?

Hello everybody.
This is my problem: I need to run a CFD simulation of a ventilation system, and the problem is that an important piece of the system is made from metal sheets. The only really problematic part is a metal sheet caping an extreme of an inlet with one of its edges free, and I cannot avoid its modelling so easily, but if I account for the sheet thickness it would lead to an unnecessary over refinement on that zone and maybe if I do it the run will not converge.
I would like to know if I can represent this sheet as a simple 2D surface, that is a flat shape with no thickness, and then set it as a wall. Both sides of the sheet are wetted by the flow. I plan to use Code_Saturne.
To show better my problem, please look at this picture www.absolutely-clean.com/cms_images/extractor_hood_after.jpg . The sheet I'm referring too is the big square surrounding the extractor system (it's thin, don't get fooled by the bottom part which is a trough). It's not my geometry but rather similar.
Thank you in advance.

César

Hi César,


In absence of someone more knowledgeable than me to answer this , I would like to start a conversation with this.

In my view, what you want to do is not possible with code_saturne or any CFD software that is base on finite volume. Even a simple 2D problem with this type of software requires to have thickness.

I might be wrong, but when I tested a similar case, I did not get a good result although there was no error issued.

my one cent

florante

Thanks for the reply. And yeah, I'm afraid you're right... unfortunately I had to waste a couple of hours to see it was not possible, after I saw a beautiful message in the console saying my run didn't converge
Perhaps it would be convenient if I give a slightly larger thickness than in real, that would allow a less compromised meshing and certainly won't change much the flow.

Hello,

It is in fact possible to model surfaces with zero thickness in Code_Saturne, but this requires extra precautions when meshing:

A face sharing 2 cells is automatically an interior face, and faces sharing the same vertices are merged during the preprocessing step. So to model a zero sheet, matching "top" and "bottom" faces of the sheet must have at least on vertex which is topologically different (the top and bottom vertices lay have the same coordinates, but their vertex id's must be different).

One way to build this would be to build the sheet with a small thickness everywhere except at its rim, and then to "flatten" it by moving vertices once merged. Another way would be to mesh the top and bottom fluid volumes separately, and then merge them, making sure only the vertices at the edge/rim of the sheet are merged.

Best regards,

Yvan

Now that's great news, and it gave me a better idea perhaps): keeping the surface, extruding it with the sheet thickness and then beveling so I finally get a sharp edge sheet. In that way I should get practically the same number of elements, no complications in meshing a thin border. Let's see how it goes. I tried again doing the run in the first way I thought about but effectively Saturne ignored the zero thickness surface and did not assign it any condition, it was part of the volume.
Yvan, let me ask you two other questions:
1) I have some doubts on the difference between the "norm" velocity and the different option to define the direction at the inlet. In 2D flows it's ok, but on 3D flows I have two inlets whichs velocity directions are 45º oblique respect to the inlet plane. I set the "norm" field as the module of the velocity vector and then in direction I use a user profile and set something like
x="norm"*sqrt(2)/2;
y="norm"*sqrt(2)/2;
z=0;
Where it says "norm" i put the same value I entered in the norm field; a similar profile is given for the other inlet. However, despite my input is practically equal to my 2D case, I'm getting in one of the inlets a smaller velocity module and normal to the inlt instead of oblique as I specified. Any suggestions?
2) How should I request to join the official CS forum? I sent an e-mail to the address given in code_saturne.info but got no answer and it didn't go to spam folder. Thank you.

Well, it seems both of my issues got solved magically. I just needed to let the simulation run longer and the vecor more or less accomodated to the desired direction... yet my doubt persist. And I got the forum activation mail. Thanks.