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

[Stephane Del Pino]

Hello.

Le mercredi 23 mai 2007, Dominique Deloison a écrit :
> Hi Stéphane,
> ...
> 2) I should change something else (P,p0 or eps or dt ...). I have to say
> I am not at all familiar with this resolution scheme and that I do not
> really understand it. It is clear for me that the dt should be small
> because it defines the time discretisation. The role of eps, P and the
> definition of p0 are far from being obvious to me.
In fact this example is not very good. It is just here to have an idea of how 
ff3d can be used to write projection algorithms...

The main difficulty with this example is the treatment of the pressure. The 
pressure is treated using a penalty trick to avoid spurious modes. If I am 
right (but this should be checked) this epsilon must be of the form C*h, 
where C is a constant and h the mesh step size.

P is important : it is the penalty term that forces the velocity to be 0 in 
the cylinder.

I modified again the algorithm using a stable element : so called Q2-isoQ1/Q1 
which means that the pressure grid is twice coarser than the velocity grid. 
This does not require anymore 'eps'. I join you a modified version where I 
also initialized p0 to 0 (probably more reasonable).

Just tell me if you get better results. This is not the best projection 
algorithm. You can for example have a look to the one proposed by Jean-Luc 
Guermond.

Best regards,
Stéphane.

navier-stokes.ff
Description: Text Data