Torsion problem with ASTER

I'm trying to understand how to solve a torsion problem using ASTER (boundary conditions, applied force/torque), but the results seem to be rather different, comparing with other calculations.

The system is depicted in the attached figure. Briefly, it is a plate fixed at one end and on the other end there are 2 opposing forces applied (along y) in order to create a torsion torque. Only half of the system is showed and also modeled.
So I've constructed a mesh using hexahedra (read about the inaccuracy of tetrahedra) and defined the group of facets of fixation (clamping), the A-B plane on the figure. There is an another group of facets declared for the antisymmetry condition and also a group of nodes, containing one node to apply the force, exactly as in the image.

The boundary conditions:
- fixation: DX, DY, DZ = 0
- antisymmetry: DY, DZ = 0
Imposed force:
- node at the corner: 1000 N

The obtained maximum displacement is 645 mm.
The reference problem can be found here (www.scribd.com/doc/8091542/Modelisation-...s-Par-Elements-Finis the last example), along with results of an FE study and also a semi-analytical problem.

Where is my mistake ?

Thanks in advance for your answers,
Francisc

PS: As I work in millimeters, I defined E = 73e3. (N/mm^2).

Thanks for your suggestions Kees. To answer you quickly,
1) The antisymmetry conditions are over a surface (of 10 mm)
2) Only one layer of elements was used in the thickness of the plane, I'll try to put more elements and see how will it change the outcome.

Thanks again.

I've put 4 elements in the thickness and the maximum displacement is 502 mm. It is still far from the seeked value of 35.37 mm.

Any ideas ?

Kees Wouters wrote:
No ideas for now. Can you upload your part, mesh, command and astk files pls to have a look at? Maybe we come up with a solution then.

** kind regards - kees[/quote]

I will do this a bit later, because I made the same model in a commercial FEM program and I got almost the same results (about 600 mm), so I don't think it is my model. So I'm going to the source of the problem and check if it is correct.

Do you know a book or internet page with a simple torsion problem having an analytical solution in order to compare the results of FEM with it ? It seems pretty difficult to find examples where the cross-section of the bar isn't circular.

Thanks anyway for your help,
Francisc