[overture] Re: Oges' compatibility constraint for a system of equations

Hi Petr,

Have you applied boundary conditions to the extra ghost lines you added?
If you have, then they are considered "used" by the equation solver and not
available for extra equations.  When we need extra equations, we use a commands 
like

ghost points
 all
 2 2 2 3

in ogen when creating the overlapping grid.  This sequence of commands
puts 2 ghost lines on three sides of the overlapping grid and 3 on one side.
We only apply boundary conditions to two ghost lines, leaving the third as 
unused.
Oges takes care of any unused points so you don't need to worry about them
in the equation solver; these are the points used for the extra equations.

Regards,
Kyle

On Wednesday 25 February 2009 12:59, Petr Hotmar wrote:
> Hi,
> 
> I have a problem with implementing an Oges' compatibility constraint for a
> system of equations (namely time-independent linear 2D Navier-Stokes eqn (2
> vel. components, 1 pressure) with Neumann boundary conditions for pressure
> everywhere and a linear divergence damping term). I have followed the
> relevant examples (tcm.C, tcm3.C), yet these solve a single (Poisson)
> equation only.
> 
> 
>  If I run my program on a single grid (pipe: inlet, outlet, 2 walls), I get
> the following error after solver.initialize():
> 
> Oges:findExtraEquations:ERROR unable to find locations for extra equations
> 
> This application is requesting numberOfExtraEquations =1
> 
> Extra equations are placed at unused points on the grid
> 
> You could add an extra ghostline to one of the grids
> 
> 
>  I have 2 ghost lines (4th order accuracy), adding more doesn't help.
> 
> 
>  If I run my program on a composite grid, I get a conformability error after
> solver.initialize():
> 
> Lengths in each dimension =
> 
> Along axis=3 -- length = 34 length of X = 34
> 
> Along axis=3 -- length = 14 length of X = 14
> 
> Along axis=3 -- length = 1 length of X = 1
> 
> Along axis=3 -- length = 1 length of X = 3
> 
> Along axis=3 -- length = 1 length of X = 1
> 
> Along axis=3 -- length = 1 length of X = 1
> 
> 
>  My coefficient matrix is dimensioned the same way as for a non-singular
> problem (when the pressure is specified explicitly, which works OK). It
> appears as if the compatibility constraint needs to be implemented
> differently for multi-component grid functions?
> 
> Petr Hotmar
> 

-- 
Kyle K. Chand
mailto:chand1@xxxxxxxx
phoneto:  (925)  422 7740

Other related posts: