Hi,
I have Cahn-Hilliard system - scalar equation for
phase field and chemical potential. I tried to modify only
the phase field part of solution in distributed code like
step-40:
template
> void Main::postprocess_solution(){
>
> nsSystem.computing_timer.enter_subsection("Solution postprocessing");
>
>
>
> std::vector
> selected_dofs(dofHandlerCrate.dof_handler_phase->n_locally_owned_dofs());
>
>
> //here I choose to modify only the phase-field part of solution
> std::vector component_mask_vec = {true,false};
>
> ComponentMask component_mask(component_mask_vec);
>
> //modified_phase_solution is the nonghosted trilinoswrappers mpi vector
> modified_phase_solution.reinit
> (dofHandlerCrate.dof_handler_phase->locally_owned_dofs(), MPI_COMM_WORLD);
>
> IndexSet locally_owned
> =dofHandlerCrate.dof_handler_phase->locally_owned_dofs();
>
> DoFTools::extract_dofs(*dofHandlerCrate.dof_handler_phase,component_mask,selected_dofs);
>
> for (types::global_dof_index i=0;
> in_locally_owned_dofs(); ++i){
> int global_index=locally_owned.nth_index_in_set(i);
> double value = globalCrateCurrent.solution_phase_n(global_index);
>
> if(selected_dofs[i]==true){
>
//modify if it is phase field part of solution
>
modified_phase_solution(global_index)=truncate_conc(value);
> }
> else
>
//else let it be
>
modified_phase_solution(global_index)=value;
>
> }
>
> modified_phase_solution.compress(VectorOperation::insert);
>
> //globalCrateCurrent.solution_phase_n is ghosted vector of solution
> globalCrateCurrent.solution_phase_n=modified_phase_solution;
>
> nsSystem.computing_timer.leave_subsection("Solution postprocessing");
>
> }
>
However the solution does not get modified. It appears to me, that
selected_dofs vector
is filled with false values.Maybe I use wrongly the function
DoFTools::extract_dofs for the distributed case.
Could You please provide me guidance, how to use it properly?
Thanks
Marek
Dne pondělí 28. srpna 2017 18:55:30 UTC+2 Wolfgang Bangerth napsal(a):
>
> On 08/27/2017 11:53 AM, Marek Čapek wrote:
> >
> >
> > I am getting the solution in
> >
> >PetscWrappers::MPI::Vector solution_phase_n;
> >
> > I am interested in modifying the vector of the solution - I
> > want to cut the undershoots and overshoots in phase-field component
> > of the solution. Namely I have values of phase field like -1.005 or
> 1.005,
> > however they should be only in <-1,1>.
> > Is it possible to modify only the relevant parts of the
> > PetscWrappers::MPI::Vector ?
> >
> > I want to get afterwards the solution values and gradients using
> >
> > fe_v_phase.get_function_values
> > fe_v_phase.get_function_gradients
> >
> > to get the "repaired" values and gradients of phase field.
> >
> > Or should I split the system, ie. firstly solve for the phase
> > field, then modify the phase field solution and afterwards
> > solve for chemical potential?
>
> The easiest solution is probably to use a
> PETScWrappers::MPI::BlockVector with two blocks that corresponds to the
> two variables. You would then just apply the operation on one of the
> blocks. This may require you to also substructure your matrix into
> blocks, but all of your solvers should continue to work. In particular,
> just because your matrix is substructured does not imply that you have
> to solve one equation at the time (though you can).
>
> The alternative is to figure out which elements of your vector
> correspond to the phase-field. Namespace DoFTools has functions that
> give you a mask or IndexSet that indicates which variables belong to one
> particular vector component. You would then just operate on those vector
> elements that are listed in the mask/IndexSet.
>
> Best
> W.
>
>
> --
>
> Wolfgang Bangerth email: bang...@colostate.edu
>
> www: http://www.math.colostate.edu/~bangerth/
>
--
The deal.II project is located at http://www.dealii.org/
For mailing list/forum options, see
https://groups.google.com/d/forum/dealii?hl=en
---
You received this message because you are subscribed to the Google Groups
"deal.II User Group" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to dealii+unsubscr...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.
