Giovanni Giorgi
2010-05-27 16:30:06 UTC
Dear ff3d users,
I am trying to solve with ff3d a time harmonic scattering problem
in the case of an impenetrable sound soft obstacle. And I need help...
Being dD the boundary of the obstacle and being S a surface
posed enough far from it, the problem is stated as follows:
Laplacian(US)+k^2*US=0 (between dD and S)
US=-UI (on dD)
dn(US)=-i*k*US (on S)
where k is a constant, UI a known function and US our unknown.
(In attachment there are the .ff and .pov files I am using).
My problem is that, when I arrive to solve the linear system, cg and bicgstab don't
converge (the problem could be not coercive). gmres seems to minimize the
residual but in an extremely slow way and all the other functions (fgmres and
ilufact) seem not to work... (In ilufact case, ff3d gives me an "unexpected error" of
this type: KrylovSolver.cpp:139: not implemented).
In particular, I would be very interested in being able to use ilufact: before
passing to the 3D problem, I solved the 2D one using Freefem++. Here, methods
as ilufact or Crout are available and work very well for this problem (I can't use
Freefem++ also for the 3D problem because of the grid dimension which is too big).
My questions are: is it possible to use ilufact in this case? And, if not, one way
could be to export matrices: how to do it in ff3d?
Thanks a lot in advance.
Best regards,
Giovanni Giorgi
I am trying to solve with ff3d a time harmonic scattering problem
in the case of an impenetrable sound soft obstacle. And I need help...
Being dD the boundary of the obstacle and being S a surface
posed enough far from it, the problem is stated as follows:
Laplacian(US)+k^2*US=0 (between dD and S)
US=-UI (on dD)
dn(US)=-i*k*US (on S)
where k is a constant, UI a known function and US our unknown.
(In attachment there are the .ff and .pov files I am using).
My problem is that, when I arrive to solve the linear system, cg and bicgstab don't
converge (the problem could be not coercive). gmres seems to minimize the
residual but in an extremely slow way and all the other functions (fgmres and
ilufact) seem not to work... (In ilufact case, ff3d gives me an "unexpected error" of
this type: KrylovSolver.cpp:139: not implemented).
In particular, I would be very interested in being able to use ilufact: before
passing to the 3D problem, I solved the 2D one using Freefem++. Here, methods
as ilufact or Crout are available and work very well for this problem (I can't use
Freefem++ also for the 3D problem because of the grid dimension which is too big).
My questions are: is it possible to use ilufact in this case? And, if not, one way
could be to export matrices: how to do it in ff3d?
Thanks a lot in advance.
Best regards,
Giovanni Giorgi