Discussion Forum

I have the same problem.
I actually changed the model quite a bit according to my design but whenever I run sim, I get the above error. I saw the same errors appear on the example model without any change.
There are other posts on this issue but,
as a solution, I defined the problem as nonlinear in stationary tab (solve parameters) and then it worked.
But later on it didn't work. So refined the mesh then I had 'out of memory' error. I think that moving frame and electrostatics don't go well together.
I have been struggling for a quite some time as well...

I am using version 3.5a.
I have tried the following:
1) Refine mesh.
2) Get Initial Value. Update Model.
3) In the Stationary tab (solver parameters), chose Nonlinear. Then Damped Newton. I also played with the damping parameters.

I get the following errors:
1) Inverted Mesh.
2) When there is no inverted mesh, the solver quits and says, 'Out of memory during LU factorization'

I would really appreciate someone looking at this model. My error is similar to the above user's.
Thank you,

Hi

Some comments:

Inverted mesh does not necesarily mean full failure, one have to take a closer look each time.

Inverted mesh mostly comes from to large displacements or too restrictive boundary conditions especiall for the ALE

I do not beleive you gain anything by:

"2) Get Initial Value. Update Model."

The "get initial value" runs a first iteration to populate your matrices, after that you can do checks of the coherence of you BC's and you can perform all postporcessing tasks, but on the starting "condtions"

"Update model" is only rerunning all the formulas on your existing model, that is particularly usefull when you add a integration coupling variables do do further postprocessing and you do not want to rerun th full solution as these are not use during the solving but only for postprocessing (ifthis is NOT the case you MUST SOLVE again the full model, all depends what you are doing.

Hope this helps a little, ALE is tricky and require often some exercices and different trials

Have fun Comsoling
Ivar

Hi

Looking at your model I see some differences with the comb drive !
if you do not get ALL BC's right it would most probably not converge, and anyhow the results would be wrong, pls check again (you can normally run two comsol sessions side by side on the same PC, useful to compare models)

In the structural, on the force boundaries select Edge load Force/Area using thickness (otherwise your units are wrong and the results as well).

Most imporant: in the Solver Parameters, you must decide which one is "parametric", you should not leave both "static" nothing will change, I would select the structural, and have the ALE left static

In the electrostatic domain remains for me still an open question: should one define the tickness as 1[m] as is, or use the same thickness as in the structural ? or does the units and thickness calculation in the structural take care of this ? in one case we could get forces some 500'000 times too big. A quick hand calculation should resolve that, I leave that one to you.

Then tick on "Postprocessing plot parameters deformed mesh" to "see" the deformation, reduce perhaps the scale factor maually to 100 or 200 (I would have expected 1 !?) to avoid an visual overlap (check the first fixed comb to the right), you can also use the plot while solving, and have the mesh visualisation turned on the observe betterwhat is happening (but where is that thick again ;).

The only thing I cannot understand is why the air domain remains "white" on my model, when I plot the potential V I should see it everywhere, one more of these ticks crossed badly in one of the GUI's ;)

have fun comsoling
Ivar

Ivar,
thank you for your detailed explanation. This will be useful for me. I am also attaching my model which should be similar to Kristen's. I believe I am following the exact procedure as in the example model with the exception of symmetry plane.
Thanks,

Here is my model. Thanks again.

Hi Ivar and Suketu,

Thank you! I changed the force boundary condition as well as the thickness of the model in electrostatic simulation, and the combination of the two fixed my model. Changing the simulation to nonlinear also seemed to help.

I now get an inverted mesh situation at the pull in voltage of the comb drive, but I suppose that is to be expected. Thanks for your help.

Hi

pls check carefully if you should really adapt the thickness of the electrostatic model, as there might be a unit issue thereby, in one case you calculate the force per length = depth and it's corrected by the use of the tickness in the structural model, the other cas is you have an absolute force and you distribute that as a pressure.

Pls double check the values you get, as in the COMSOL model they use 1[m] depth in the DC calcuation

have fun Comsoling
Ivar

thank you Ivar for your help