On 21/set/09, at 11:53, Till Heinemann wrote:
Hello,
Working on my project (see mailinglist entry [deal.II] "Combining
BEM and
FEM") I have come to an unexpected obstacle:
I don't seem to be able to create a FESystem<1,2> BEM_fe
(FE_Q<1,2>(1), 2)
in the main class' constructor
template <int dim>
ElasticProblem<dim>::ElasticProblem ()
:
FEM_dof_handler (FEM_triangulation),
FEM_fe (FE_Q<dim>(1), dim),
BEM_dof_handler (BEM_triangulation),
BEM_fe (FE_Q<dim-1,dim>(1), dim+1) // create dim+1 BEM
element with dim+1 degree of // one for potential xi and three
displacement variables.
// FESystem<dim-1,dim> BEM_fe;
{}
But that's only in 2D case, 3D runs fine.
The exception is due to the FESystem constructor's attempt to
dealii::FESystem<1,2>::initialize_unit_face_support_points(), as their
dimension is dim-1 referred to the FESystem's dimension, which is
already
dim-1=1, so we get a point with dim=0, which is not valid.
I see.
As mentioned above, it's no problem in 3D or if I don't build a
FESystem,
but use a single FE_Q object in dim=spacedim-1 (2D and 3D). Still I
want to
solve a vector valued problem with BEM in 2D (and 3D)...
--------------------------------------------------------
An error occurred in line <391> of file
</home/user/Public/deal.II/base/include/base/tensor_base.h> in
function
dealii::Tensor<1, dim>::Tensor(bool) [with int dim = 0]
The violated condition was:
dim>0
The name and call sequence of the exception was:
ExcDimTooSmall(dim)
Additional Information:
dim must be positive, but was 0
Stacktrace:
-----------
#0 /home/user/Public/deal.II/lib/libdeal_II_1d.g.so.6.2.0:
dealii::Tensor<1, 0>::Tensor(bool)
#1 /home/user/Public/deal.II/lib/libdeal_II_1d.g.so.6.2.0:
dealii::Point<0>::Point()
#2 /home/user/Public/deal.II/lib/libdeal_II_1d.g.so.6.2.0:
dealii::FESystem<1, 2>::initialize_unit_face_support_points()
#3 /home/user/Public/deal.II/lib/libdeal_II_1d.g.so.6.2.0:
dealii::FESystem<1, 2>::initialize()
#4 /home/user/Public/deal.II/lib/libdeal_II_1d.g.so.6.2.0:
dealii::FESystem<1, 2>::FESystem(dealii::FiniteElement<1, 2> const&,
unsigned int)
#5 /home/user/Public/deal.II/examples/step-8/step-8:
ElasticProblem<2>::ElasticProblem()
#6 /home/user/Public/deal.II/examples/step-8/step-8: main
--------------------------------------------------------
Looking at the fe_system.cc I understand the 1D case is to be caught
by a
preprocessor condition (line 1833), checking deal_II_dimension, which
appears to coincide with the spacial dimension. But it doesn't catch
the
case of 1D objects in 2D space.
So I wanted to somehow change that condition in fe_system.cc.
But not being very experienced with C++ and preprocessor stuff, I
have not
been able to find out, where (and how!) exactly this variable is
created/defined!! (I use a small project makefile as in step-8
tutorial). So
this would be my first (general) question.
The next one is whether this exception should be caught by an altered
preprocessor condition, or to implement another if(dim==1){// no
faces, do
nothing} condition inside the general template <int dim, int
spacedim> void
FESystem<dim,spacedim>::initialize_unit_face_support_points (){...}
routine
(which I've done for simplicity in my library, seems to do the job,
but
that's hardly professional I guess)
Till, I think you just spot a bug in FESystem for codimension one
problems. I'll try to look into that and get back to you as soon as
possible. All exceptions you see are correct, except that FESystem
should not try to initialize initialize_unit_face_support_points().
Thanks for spotting this one.
Luca.
_______________________________________________
dealii mailing list http://poisson.dealii.org/mailman/listinfo/dealii
