Hej Markus,
I kicked these parts out. Instead I just applied zero boundary
conditions using interpolate_boundary_values on indicator 0 (the outer
left node) and indicator 1 (the outer right node), ending up with:
constraints.clear();
//dirichlet boundaries on left and right surface
std::vector<bool> comps(2,true);
comps[1] = false;
//in 1d indicator 0 gives most left node
dealii::VectorTools::interpolate_boundary_values(dofHandler, 0,
dealii::ZeroFunction<1>(2),
constraints,
comps);
//indicator 1 gives most right one
dealii::VectorTools::interpolate_boundary_values(dofHandler, 1,
dealii::ZeroFunction<1>(2),
constraints,
comps);
constraints.close();
Still the code is executed just fine until it comes to solving the
system. Then the solver exists after step 1 with the given exception (no
convergence, residual nan).
Best regards,
Timo
Markus Bürg wrote:
Hello Timo,
then the problem could be the boundary values:
//set surface flags to determine which to apply dirichlet and
periodic
tria.begin_active()->face(0)->set_boundary_indicator(1); //left
surface
tria.begin_active()->face(1)->set_boundary_indicator(1); //right
surface
tria.begin_active()->face(2)->set_boundary_indicator(2); //bottom
surface
tria.begin_active()->face(3)->set_boundary_indicator(2); //top
This does not work in 1d, because cells do not have faces in 1d.
Please have at look at the corresponding FAQ and tutorial step-15.
Best Regards,
Markus
Am 30.03.11 12:44, schrieb Timo Koellner:
I already implemented a GFEM in Matlab solving these equations, which
works just fine. So I'd assumed, it must be possible to solve these
equations also using deal.II.
I'm now switching to deal.II due to performance issues. The given
eqns are actually linearized eqns of the ones I later want to solve
and here Matlab performs very badly.
Markus Bürg wrote:
Hello Timo,
you are solving a first-order PDE. Usually these are not stable in
Galerkin formulation. Have you checked the stability?
Best Regards,
Markus
Am 30.03.11 11:15, schrieb Timo Koellner:
Hi folks,
I just started working with deal.II and do experience some problems
with solving the following equations:
\nabla n = 0
\nabla \vec{E} = -e(n - \Theta(x))
For simplicity, I started off in 1d where I can solve the equations
analytically. The problem I'm stuck with is
"Exception on processing:
Iterative method reported convergence failure in step 1 with
residual nan"
I attached the program so you can have a look. What I actually did
is to take the Step-by-Step example on periodic boundary conditions
(since in higher dimensions I want to apply them) and modified it
using a FESystem and of course inserted my equations and left out
the periodics as I don't need them in 1d.
I tried using different Solver classes ending up with the same
error. So, I guess there's something wrong with setting up the
system, but after rereading the examples and the explanations on
the module for solving vector-valued eqns, I have no idea what's
going wrong.
Am I missing something?
I would appreciate any help on this. Thanks.
Timo
_______________________________________________
dealii mailing list http://poisson.dealii.org/mailman/listinfo/dealii
------------------------------------------------------------------------
_______________________________________________
dealii mailing list http://poisson.dealii.org/mailman/listinfo/dealii
_______________________________________________
dealii mailing list http://poisson.dealii.org/mailman/listinfo/dealii
_______________________________________________
dealii mailing list http://poisson.dealii.org/mailman/listinfo/dealii
_______________________________________________
dealii mailing list http://poisson.dealii.org/mailman/listinfo/dealii
