Re: [ff3d-users] Stationary Navier-Stokes equations / Flow past cylinder problem

[Stephane Del Pino]

Hi Dominique.

Le lundi 21 mai 2007, Dominique Deloison a écrit :
> Hi Stéphane,
>
> Thank you for your quick answer. I have however still some problems.
>
> To answer your question,  I found the out-of-date" navier-stokes"
> example here: http://www.freefem.org/ff3d/examples.html
Ooops. These are really obsolete. The same applies for the online 
documentation...
I have just updated the files.

> Concerning the files you sent me, I have the following comments:
>  1)      - during the resolution, the following warning appears:
> "implementation not finished! use of dofposition required". Is is a
> problem?
This is not important. It is just a reminder for fictitious domain 
development...

> 2)      - the size of each element (0.1) is the same than the 
> radius of the cylinder, so it is very difficult to recognize anything and
> to be able to decide whether the solution looks reasonable or not. I guess
> I have to increase the number of elements.
Yes. This is just an example of use, it might not be anyway physical...
By the ways I don't know why the cylinder size changed...

I just have been looking to it and it seems to me that it is not a good 
example. So I slightly changed the example. You can download it from the web 
site...

>  3)     -  the syntax of the "convect" operator is different than the
> one used in the documentation.
>           According to the documentation, the syntax is:
> "convect(phi,ux,uy,uz)".
>          In the file there is: "convect([u0,v0,w0],dt,u0)" which is
> identical to the freefem++ syntax.
Yes. The online manual was also obsolete. I just put a newer version, sorry 
for the confusion...

>  4)     - I have difficulties to figure out the mathematical description
> of the problem described in the example file. For me it looks like a
> transient problem. I would be interested in a stationnary solution as
> described by the equations I wrote in my previous mail. Is there a non
> time dependent version of the convect operator ?
Yes. This is a transient resolution...
There is no stationary convect operator, but, you can write a fix point 
algorithm. I join you an example.
> 5) I tried to postprocess the streamlines with medit (I wrote an  .iso
> file to describe the seeding points).  I was unsuccessful (segmentation
> fault). have you ever tried this with this example?
The reason is that medit is not very good for hexahedra.
In the joined file I use a tetrahedrization of the mesh and an alternative VTK 
output.

I hope that helps. Thanks for your reports.

Best regards,
Stéphane.

stationary-ns.ff
Description: Text Data