Discussion Forum

Hi,

if I assume that your (x,y,z) direction are defined with respect to the (80x4x0.055)[mm] bar size then I notice that you have a shape ratio of 1:>8000 which I consider as big.

You must certainly take some car with the meshing to get reasonable mesh quality (have you checked the quality ?)

I would try a mapped mesh of the 4x0.055 with a 1:100 scale ratio (check the advanced tab) with some 200-400 x 4-8 elements on the section and then take some 100 to 200 elements with a sweep mesh along the 80 mm length

Then solve the system and analyse the results, start with the minimum, then increase the number of elements to check that the results do not change by more than 5-10%

--
Good luck
Ivar

Hi,

I tried to do your advice but I can't find the tabs to choose the scale ratio and elements.

I did one folder with mapped and swept mesh I still have problems :(

I'm a begginer with comsol and if you could check my file and comment me what is wrong I'll be happy.

Best Regards

Ana

Hi

when you have such large size differences (aas well as for MEMS, it's worth to work without the "View" "Preserve Aspect ratio" on. Then you will notice that you have only 1 element across the 55um, you should have at least 3, I prefere 5-7 minimum (the more the better), but then their real shape should be suqare or close to, with a high quality (you have now 6% "quality" even if this is >2% (my lower limit) you have too few elements across the beam.

Try to add a "Distribution" to your Mesh/Malla Mapped/Mapeado 1 , select edges 1,2,4,6 and use the default 5 items, then remesh and resolve and look at the differences. To improve the quality you should add many many more along edges 4&6

Then you might also need to add a "Distribution" subnode to your Sweep/Barrido (I'm learning Spanish like this ;) to improve the "quality". careful, you end easily up with 100k elements so if you are to heavy on the nodes, you will end up with RAM issues.

Take a look how the results vary when you change the number of elements.

This indicates anyhow that you could consider 2D or even a 3D shell elements for your thin "plate"/beam

Last thing, your "fixed constraint are probably too severe, you need probably some thermal expansion matching fixed BC's to minimise the edge effects, that is another story, some simple but interesting equation writing to get that better defined, without overstress

--
Good luck
Ivar

The displacement in the Z-direction should be small because you have a very small thickness in that direction. It should be a near constant positive displacement on the top face and the negative of that amount on the bottom face. Is that what you expect? The reason you are getting a different solution is your choice of solver. Iterative solvers, with default convergence parameters, are not the best option for such very thin structures. When you switch to direct solver you get a more accurate solution.

Nagi Elabbasi
Veryst Engineering

Sorry for my spanish version...

Thank for your reply. I've done your advice and I've noticed the changes in the mesh and solvers but I continued with problems such as residual error or relative tolerance error. So, I've changed my bar and now I "comsoling" as plate (the conditions are the same, fixed ends and thermal expansion). With this way I don't have problems with the mesh but the results are a little bit wrong :(

Regards.
Ana

Hi

come on no reason to be "sorry" for having a "spanish" version, we have all "national" versions in some way, it's just funny and instructive to see and learn a mixed half translated model ;)

So what is "wrong" with your model, are you sure you have the same boundary conditions, or that all other hypothesis are the same with the model you are comparing too ?

But bugs exist, also in COMSOl, that is why all of us should re-check our results with approached analytical values, to verify the coherehnce of our results, beautiful rainbow views are not enough today, almost everyone can make them, making them correct all the time is far better ;)

--
Good luck
Ivar