OT? FEM comparison

Hello,

I am working at an university and we are using two different
FEM-programs. ANSYS and CosmosWorks (conected with SOlidWorks). I am
very interested to make a comparison with both and may be other
programs. Because of this I made a test with a part where I changed the
position of the supporting points. The solutions for the displacement
were almost equal (less than 1% difference) but the stress was
significantly different (over 60%). Now I am trying to get informations
about "weak" points of both programs and their solvers.
Has somebody some hints or sources or links to get an overview over
these programs and compared results?

Many Thanks

M. Lotz

 

Hello,

1) I use the High Quality mesh at COSMOSWorks with tetrahedrons. Because
of the mesh "problems" I made tests with COSMOS. I changed the density
of the mesh in some steps and looked at the maximum stress v. mises and
the resulting displacement of the whole model and four different points.
Normally I would expect that stress and displacement values will
converge. The displacement values converge like I expected to a value
which I also reached with ANSYS. Unfortunately I can not see a
convergence in the stress values.
The mesh in ANSYS consists of ET187, also tetrahedron. I will also make
some tests with its mesh density. Unfortunately it takes much longer to
solve this model in ANSYS because it is a "selfwritten" parametric model
to change supporting points.

2) That is true. I used at COSMOS two models, one with and one without
local refinement. But both showed a convergence of the displacement but
no convergence of the stress.

I want to test how much can I trust solutions from both programs because
we want to design some parts which sould be lightweight. If we calculate
the stress to high parts will have more weight, if it is too low the
parts can be destroyed. Unfortunately the parts are too difficult to
make only estimates with analytic calculations.

I think so too. But I found one comparison between COSMOS and
DesignSpace (little ANSYS) were COSMOS showed better behaviour also in
convergence...

Thank you very much!

M. Lotz
Technical University of Ilmenau
Germany

 

Are you trying to analyze parts or assemblies?
-Ram