> I'm using step-20 to calculate the dual-mixed formulation of the
> Laplace-equation in context of mechanics with zero boundary values for the
> scalar component. Here only the definitions change (pressure becomes
> displacement and velocities become stresses), the weak formulation stays
> nearly the same (only two algebraic signs change because of my classical
> formulation and substitution).
> 
> In step-20 the boundary value/ curve for the pressure is a polynomial
> function and the applied force/ right-hand-side is zero.
> 
> If I change the pressure boundary value to zero (i.e. g=0) and the force to
> non-zero as I need it for my formulation, the membrane breaks off at the
> boundary. This happens both if the force is applied on the hole domain
> [-1,1]^2 and if it is only applied on a subdomain like a circle around
> (0,0) with radius 0.5. I'm wondering about this because the natural
> boundary condition for the scalar component of the dual-mixed
> Laplace-problem is the zero value (cf. D. Braess, "Finite Elemente",
> Springer 2007, p. 140).

What exactly do you mean when you say the "membrane breaks off"?

I can't tell where the problem is (though what you describe should work at 
least in principle) but here are a few things you could try:

- Throw out the solver and replace it with SparseDirectUMFPACK (see step-29; 
the class can also deal with block matrices). This way you know that the 
problem is not in the solver but in assembling the matrix.

- Try a forcing function that has mean value zero. If the linear system is 
indefinite, you may get still get the right solution if you have such a rhs 
function, and knowing that may give you some insight where to search next.

- What happens as you refine the mesh -- does it converge to something?

Best
 W.

-------------------------------------------------------------------------
Wolfgang Bangerth                email:            [email protected]
                                 www: http://www.math.tamu.edu/~bangerth/
_______________________________________________
dealii mailing list http://poisson.dealii.org/mailman/listinfo/dealii

Reply via email to