[Libmesh-users] Re: Is this correct for applying periodic boundary condition?

[Wout R. <wou...@gm...>]

ShengliXu

// you forgot the all important line:
dof_map.constrain_element_matrix_and_vector(Ke, Fe, dof_indices);
before
system.matrix->add_matrix (Ke, dof_indices);
system.rhs->add_vector    (Fe, dof_indices);

So all the other explanations I came up with for the program not
working are far-fetched or nonsense.

You can choose the solver on the command line, for example:
time /usr/bin/mpirun -np 1 $(target) -ksp_type gmres -pc_type jacobi >> log=
.txt

Now, one more time:
If you apply a strain in x-direction the bounds are going to move
apart, this CONFLICTS with the boundary conditions you just put up (as
you're telling them they have the same displacement).
As you will see, if you couple all dofs on all opposing sides as you
propose you will only get some noise as an answer regardless of the
loading.
Read my earlier post on how to deal with that (by adding dofs to the
system in a dummy element that represent a displacement of a whole
side, keeping the sides in the same shape).
Secondly, the code you put in to deal with corners does not do
anything, as I said, the corners are in the sides and are dealt with
automatically (because "couples" is a std::map and keys are unique in
those).

Cheers
W