getfem-users
[Top][All Lists]

## Re: [Getfem-users] Problems to build a matrix for a boundary term

 From: Iago Barbeiro Subject: Re: [Getfem-users] Problems to build a matrix for a boundary term Date: Sun, 6 Sep 2009 17:23:10 -0300

Dear Ronan and Yves,

I think you are right, the Normal constant term for affine geometric transformation is probably not taken into account for exact computations.
I have just tried  the Newton-Cotes method ('IM_NC(2,1)') and the error message is gone.

Iago

On Sun, Sep 6, 2009 at 1:29 PM, Renard Yves wrote:

Dear Iago,

I think this is the term "Normal" which leads to this error (can you verify it ?). The normal is indeed constant for an affine geometric transformation but I a not sure that this term is taken into account for exact computations. The simplest is to use a cubature method !

Yves.

Iago Barbeiro <address@hidden> a écrit :

Dear Ronan,

You are right, circular domain and holes may sound higher order
transformation, but it is not the case here.
I'm using an affine triangular mesh, GT_PK(2,1), and also affine continuos
shape functions FEM_PK(2,1).
Any other clues?

Regards,

Iago

Dear Iago,

the exact integration is mainly linked to the use of an affine mesh.
Is it the case here? You speak about a circular hole and maybe you use
second order transformation for the elements around the hole...

I hope it helps.

Best regards,
Ronan

Iago Barbeiro a écrit :

Dear Getfem++ users,

I am facing problems to build an elementary matrix using the generic
assembly procedure and I hope anyone could give me a hand.

It is a boundary integral in a two dimensional mesh (It is a circular
domain with a circular hole inside and I want to integrate over both
boundaries):

So it is a bilinear form in "p" and "dp" where both are scalars defined
with the same fem (mf_p).

For that I have tried the following block:

*template<typename MAT> *
*void asm_P_Lap_boundary(const MAT &K_,*
* **                   const mesh_im &mim,*
* **                   const mesh_fem &mf_p, *
* **                   const mesh_region &rg) {*
*
*
*     MAT &K = const_cast<MAT &>(K_);*
*
*
*     getfem::generic_assembly assem;*
*   *
*          *
*     assem.push_mi(mim);*
*     assem.push_mf(mf_p);*
*     assem.push_mat(K);*
*     assem.assembly(rg);*
* }*

Where the mesh_region was built using:

*getfem::mesh_region boundary;*
*getfem::outer_faces_of_mesh(mymesh, boundary);*

But then I found an error message, which is:

*============================================*
*|      An error has been detected !!!      |*
*============================================*
*Error in getfem_mat_elem.cc, line 121 : *
*Exact integration not allowed in this context*

Why exact integration would not be allowed here? Am I doing something
wrong? Any light?

Iago C. Barbeiro
------------------------------------------------------------------------

_______________________________________________
Getfem-users mailing list
https://mail.gna.org/listinfo/getfem-users