Re: [Libmesh-users] Navier-Stokes with obstacles

[<ti...@ce...>]

Dear John,

On Tue, 16 Jan 2007, John Peterson wrote:

> Upon closer look at your code I finally noticed this part:
> What does it do exactly?

Well, it's the obstacle.  I.e., in this case, a (large) square area is 
considered to be solid and fixed, i.e. u=0 in this area.  The idea is 
that this acts as a no-slip boundary condition at the boundary of the 
obstacle.  Within the obstacle, I'm not interested in the 
Navier-Stokes quantities anyway.  However, removing these cells from 
the grid might be problematic in my real application for two reasons: 
First, I have a second equation that has to be solved in the whole 
domain, including the obstacles.  Second, the size, number, and shape 
of obstacles changes in time.

> It can be quite tricky trying to
> work with two different "penalty" values, in case they accidentily
> end up in the same row.

In my opinion, they can't happen to be in the same row in this case 
(except for extremly rough grids) since the obstacle is away from the 
boundary of the domain.  Also, the behaviour of the program does not 
change if 1e7 is used for *all* penalties.  However, if 1e10 is used 
for all penalties, then the computation with obstacle fails already on 
the 20x20 grid.

Best Regards,

Tim