Re: [ff3d-users] Periodic BCs

[Stephane Del Pino]

Hello.

Le Jeudi 6 Janvier 2005 02:00, Bharat Rangan a écrit :
> Hi,
> Thanks for the detailed reply. As you suggested, I am going through the NS
> example and trying my own case. A couple of questions in this are
>
> 1. The projection method in the example has -div(eps*grad(p)) + eps*p +
> dx(u1)+dy(u2)+dz(u3). The methods that I have come across are
> usually -div(grad(p)) + large_number*div(u) =0. Is there a difference in
> these implementations.
Well, I am not a specialist of stabilization, but you can omit the eps*p term 
(if you then multiply the equation by 1/eps = large_number you will get back 
to your formulation). Obviously you can use your favorite projection 
algorithm and couple it with the stabilization procedure of your choice... I 
would also be interest in references and feed-back if your algorithm gives 
better results. I used the one that is in the documentation during my PhD 
thesis, I picked it up from an old freefem+ example, so I do not have precise 
reference to give to you...

> 2. Is there an external way to implement periodic BCs. Like some penalty on
> the difference of the solution on the two faces and then iterating ?
I am not sure if this is possible, and I do not see how to do it this way, 
maybe using Robin-Robin like condition as it is done in domain 
decomposition ...
But, even if it is possible, to compute Navier-Stokes, you will need it to 
work also for convection operator (characteristics method).

Best regards,
Stephane.