Discussion Forum

Hi

I do not have a running COMSOL here, but mostly you hit some topological limitations of the quad/block mesher.

A simple block will be meshed OK, but a block with one boundary split by a line will not, this extra boundary gives problems for the mesher to understand how to continue. If you split your block by an internal boundary into two blocks it will mesh both OK.

First why not try "thets", COMSOL has another way to consider classical "higher order elements":. This is via the "shape functions" that is applied with the Physics to the Entities, and not to the mesh element.

Meshing is the last (still important but not as important for older FEM programmes) operation performed before solving in COMSOL, this is not always understood correctly by many users coming from more traditional FEM tools

--
Good luck
Ivar

Hi,

Ivar, thank you for your reply!
Could you explain in detail. "split your block by an internal boundary into two blocks" What's the meaning? In my model the air domain is cylinder and the plate is a block. Attached is the geometries.

Thank you very much!

Hi
First of all, if you are in 3D with such a nice symmetrical model, do consider symmetry (you have a 1/4 symmetry at least there, that is roughly 1/16 times more rapid to solve !

I would anyhow use tets and not bother about quads for your case, at least to start with. Then I would cut the model into the main symmetry planes (adding internal boundaries). This will force the mesher to put nodes on these internal boundaries, and your mesh will be more symmetric.

then do not forget, if you are in MEF to set some conductivity to the air to avoid numerical instabilities and convergence issues, somewhere 1-1000[S/m] should do, then check that the total dissipation in the air part is not too high compared to the material part, Another point, the coil excitation, MEF solves for V, so if you block fully the current you miss the back-EMF potential drop part

--
Good luck
Ivar

Hi,

Ivar, thank you for your detailed explanation.

Several weeks ago, I made the 2D axisymmetrical simulition about the same object and the results are nice. The resluts is very similar to the experimental results. Now, I want to make a 3D simulation to compare with the 2D results. The reason of using hexahedral element for the plate is that I want to make split into the plate in the near futrue. So the meshed element of plate should be structured. I have meshed the model succesfsully by the command of convert. Even though the element of plate are not hexahedral, I still can get the shape-structured split because I can set the five tetrahedral elements with the same material characteristic. Unfortunately, I encount another error about solving.

Error in automatic sequence generation.
Expression:(((0)+(0)+(0))+()
subexpression:()
position:17

What does this mean?

Attached is my model. Thank you in advance!

Hi

To start with, if you set your "coil" with something more conductive than "air" it will help.

But you have still something wrong with the model.

I'm in 4.1.0.112 (and I strongly suggest to you to ask your local rep to get updated) and I cannot analyse your geometry.

Try to rebuild your model

--
Good luck
Ivar

Hello,

two problems, my friend:

1) You must define the the z-compnent of the external current density, e.g. J_z=0.

2) Defining an external current density in a domain AND defining the conductivity of the very same domain with zero is physically wrong and therefore probably a problem for COMSOL.

I corrected these two things by dclaring domain 3 as copper and setting J_z=0 and the model solved with an error message (relative error is greater than the relative tolerance), but correct looking plots.

Bye
MV

Hello,

Marco, thank you for your help. It's very kind of you. Now I am clear about the error.

Best wishes for you:-)

Dear Ivar,
I am using COMSOL to build a 1D model of Dielectric Barrier Discharge. I have encountered a problem
Error in Automatic Sequence Generation
Model parameter
name: e
property: tlist
Would you please suggest me how can I solve this problem?
Thanking you,
Deepak