Discussion Forum

Hi

in FEM there are no default contacts information around, hence nothing prevents two volumes to overlap each others. In your case probably you apply a force so big theat even the non-linear range of the material is not cabable to resist it.

Contacts can be implemented in COMSOL but is is very CPU resource demanding, and should be used with care

Goos luck
Ivar

Dear ivar

I want to say thank you for your replay.

I was trying in a cylinder elastomer where I fix the bottom side and apply a pressure on the top side. I was doing frequency analysis and at the eigen frequencies the deformation a point on the top plate is longer that the cylinder hight( it goes beyond the fixed plan). is there something error in my model, any suggestion?

best regards

Hi

if you are in eigenfrequency/eigenmodes then NO this is normal.

Comsol normalises the eigenmodes shapes in one way, but you must notice that the absolute values are NOT to be considered s "real values", only the relative shape is relevant.

This comes from the fact that there is no "energy limit" used in an eigenfrequency simulation, so deformations could go infinite for that sake, they are renormalised to "1" individually if I remember right.

There are other ways to normalise the eigenfrequencies, but this does NOT make the abolute displacement more or less representative. Among other I'm still waiting for the total mass normalisation, which allows to sort eigenmodes by their relative "energy" or modal mass participation factor hence their relative importance for further modal analysis and modal reduction. Currently this is not done/implemented, not even as postprocessing option in COMSOL.
This mass participation factor method/norm is the standard one for most other FEM programmes COSMOS, NASTRAN, ANSYS, ABACUS ...

So most probably your simulation is OK

Hope this helps on the way
Good luck
Ivar

Dear ivar

thanks again for your replay.

Now I got the idea of the displacement in eigenfrequency mood.

But when I work also at frequency response mode, the displacement has a problem still. I was plotting the the displacement of a point on the top plan as a function of frequency ( including the first 3 eigenfrequencies ) and at the eigenfrequncies the displacement is very very large...as I have said beyond the height of the cylinder I assumed.

Is it still an implementation issue in comsol ?

waiting your answer




best regards

Hi

No its not an implemental issue, it all depends on how you define your damping in the model, you should ask yourself were does the energy go ?

In eigenfrequency, as in frequency sweeps, when you hit a resonance, and have not inlcuded into the physics any damping (i.e. something proportional to the velocity, or to the stress), then you should get (mathematically) an infinite amplitude. This is the way you are setting up your physics.
Now COMSOL, as most other FEM programmes managed to pas "through" a resonance without crashing because there are mostly some "numerical dissipation" term in the algorithm to avoid/survive (mostly) with 0/0 or infinity.

Add in some "structural damping" and you will see that your model will rapidly loose amplitude at these resonances, and the peak will broaden.

This you discover with hardware too, I remember a case where we had a steel part surviving >40G accelerations, from our linear FEM calculations, but then it broke by the first "sinus sweep" test on the shaker at less than 0.1G, far before we were to put up the 2 minutes of (satellite) launch load shaker run.
We had missed a non linear coupling to a resonance under certain load cases that made the part fail by buckling more than 100 times below specified dynamic vibrational load case. Such failures are easy to get into your design, when you do "compliant mechnisms". FEM does not remove the need to THINK your physics through before, and to validate and verify each step, also by hand calculations AND physical tests.

Have fun Comsoling ;)
Ivar