Thanks Stephane, that fixed it.
I thought my formulation was equivalent to the one you wrote,
As I understand it, my formulation includes the natural boundary
conditions which is unnecessary but I thought that there was no harm
(other than inefficiency) to include them, anyway thanks for the
pointers to the wikipedia stuff, I will look through them and try to see
why including the natural boundary codtions in the surface integral
won't work.
I'm about to try using periodic BC, I note from the mailing list that
you were playing around with this a couple of years ago, do you have any
sample files or other documentation or pointers to the source code?
Do you have any infinite elements? I'm thinking of a radiation condition
decaying to zero at infinity on an open boundary.
regards,
Tom
Stephane Del Pino wrote:
Dear Thomas,
thanks for your interest in ff3d.
In fact, you did not compute the right variational formulation associated to
your PDE. You have to be careful in the integration by part
Have a look to
http://en.wikipedia.org/wiki/Finite_element_method#Variational_formulation
and to
http://en.wikipedia.org/wiki/Green%27s_identities
for some details.
In your case, you should have written:
solve(phi) in M
{
test(w)
int(grad(phi)*grad(w))
+int[M zmax](w*alpha*phi)
=int[M xmin](w*f)
-int[M xmax](w*f)
=0;
};
Do not hesitate to ask me if you get troubles with that.
PS
One small documentation suggestion; the method of defining Robin type
boundary conditions be added to the documentation or README file.
The documentation is far from being completed. I not this on the TODO list.
Best regards,
Stéphane.
_______________________________________________
ff3d-users mailing list
[email protected]
http://lists.nongnu.org/mailman/listinfo/ff3d-users
_______________________________________________
ff3d-users mailing list
[email protected]
http://lists.nongnu.org/mailman/listinfo/ff3d-users
