Dear all
I am trying to use MappingFEField on a parallel::distributed::Triangulation
with TrilinosWrappers::MPI::Vector.
In this case, how should the euler vector be created ?
Should the euler vector use locally owned dofs or locally relevant dofs ?
I tried the following
unsigned int degree = 4;
const FE_Q<dim> fe(QGaussLobatto<1>(degree+1));
const FESystem<dim> fesystem(fe, dim);
DoFHandler<dim> dh_euler (triangulation);
dh_euler.distribute_dofs (fesystem);
IndexSet locally_owned_dofs;
IndexSet locally_relevant_dofs;
locally_owned_dofs = dh_euler.locally_owned_dofs ();
DoFTools::extract_locally_relevant_dofs (dh_euler,
locally_relevant_dofs);
TrilinosWrappers::MPI::Vector euler_vector (locally_owned_dofs,
MPI_COMM_WORLD);
compute_mapping (degree, triangulation, dh_euler, euler_vector);
MappingFEField<dim> map(dh_euler, euler_vector);
When “map" is destroyed, I get some error
An error occurred in line <128> of file
</Users/praveen/Applications/deal.II/git/source/base/subscriptor.cc> in
function
void dealii::Subscriptor::check_no_subscribers() const
The violated condition was:
counter == 0
The name and call sequence of the exception was:
ExcInUse (counter, object_info->name(), infostring)
Additional Information:
Object of class N6dealii6VectorIdEE is still used by 1 other objects.
Thanks
praveen
On 30-May-2016, at 11:32 AM, Praveen C <[email protected]> wrote:
Dear all
I have problems involving airfoil geometry where the cells have high aspect
ratio in the boundary layer. Due to this, it is not enough to just use
curved boundaries but the interior faces also need to be curved. See the
attached figure gridq1.pdf
I have implemented a method based on winslow equations, see
http://persson.berkeley.edu/pub/fortunato15winslow.pdf
http://dl.acm.org/citation.cfm?id=2884302
Using this winslow method, I can generate an euler vector that gives valid
cells, see gridqk.pdf which uses degree=4.
Using this euler vector, I want to create a MappingFEField and use that in
my Navier-Stokes DG code.
I have some doubts.
When I use a MappingFEField, do I still need to attach manifold
descriptions to the curved boundaries ? Assume that I am not doing any grid
adaptation.
If I want to adapt the grid, is there a way to use the MappingFEField to
place the new points ?
Are there any functions in deal.II to measure the quality of the mesh ?
The Winslow eqn approach seems quite useful and one can write a general
class that may be useful for many people. At present I only have a serial
implementation.
The approach is as follows:
Start with a Q1 mesh
Attach manifolds
Solve the winslow eqn to get the euler vector
Create a MappingFEField
Thanks
praveen
<gridq1.pdf><gridqk.pdf>
--
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 [email protected].
For more options, visit https://groups.google.com/d/optout.
