Re: [ff3d-users] problem with hydrostatic sample

[Stephane Del Pino]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On Wednesday 04 February 2004 18:35, Robert Li wrote:
> Dear Stephane Del Pino,
>   I found a hydrostatic sample on the internet that
> you wrote on Mar.,21,2003. The title is Domain
> Decomposition for the Couplex Exercise.
>
>   I have some problem with it.
>
> 1. When I run it, the error is as following
> Generating Surface Meshes
> ./solver\Domain.hpp:138: Trying to access Object in
> case of R^3 domain
Yes the syntaxe has changed a little since that time.
Since domain definition has no inside or outside part, ff3d considers that the 
computational domain is equivalent to the mesh. But we impose a condition on 
the border <1,0,0>. So use:

domain D= domain(S,outside(<1,0,0>));

The second remark is that you need to remove the 'memory(matrix=none)' solver 
option to use the incomplete Choleski factorization preconditioner.

Third, this computation will require a lot of memory so change N=30 to 20 or 
less.

Finally, this computation is quite difficult and you may have difficulties to 
see anything hapen, so I changed a few the geometry: a bigger object.

> 2.I don't clear about the concepts of computational
> domain and the fictitious domain. I think
> computational domain is domain D, fictitious domain is
> mesh, right?

Yes. The principle of FDM is to use a mesh that do not match with 
computational domain! Look at the documentation, the 'FEM and EFDM' chapter 
is quite complete.

> 3.We usually only set boundary conditions on border of
> computational domain D. Here you also set boundary
> condition on Mesh border, except for computational
> domain<1,0,0>. Is this necessary or right?
Has I said above. This sample is not standard and quite old now. This sample 
uses explicitly the domain characteristic function (penal and a part of K) 
which is no more required (this is performed automatically --- so that ff3d 
file is simplified). So I removed them.

Concerning computational domain and mesh: if the computational domain is an 
external problem for instance you *may have* to impose conditions on the mesh 
border as well as on POV-objects. There is no rule: it is completely related 
to your objectives ...

Regards,
Stephane.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.4 (GNU/Linux)

iD8DBQFAITTlbZyhKhZwdc0RAirJAJ9bdROeHf2W2r1hqHzFYn5owWnaXACfSNIb
I6ep8u/YO5ID4fwwVvmuwL4=
=uiMQ
-----END PGP SIGNATURE-----

dechet2.pov
Description: Text document

hydrostat.ff
Description: Text document