Discussion Closed This discussion was created more than 6 months ago and has been closed. To start a new discussion with a link back to this one, click here.

Meshing of thin geometry

Please login with a confirmed email address before reporting spam

hi,
I'm using comsol to solve my model. My model geometry consists of a block of 6.7*6.7*6.7 m at the centre of which is another block of 0.10*0.60*5e-5 m. As second blocks thickness is too small, i'm unable to form a mesh for this geometry in 3D. I tried mapped mesh by dividing 1st block in 6 subdomains. By doing this i could even form the mesh. But when i tried to solve the problem, it's showing error as "unable to evaluate the expression". I checked out mesh statistics. It's showing mesh points (which i generated through mapped mesh), but mesh elements as zero.
Please help me out. How should i form the mesh? I want to solve diffusion and Navier stoke's equation together in this geometry.
I'm waiting for your reply.
Thank You.

7 Replies Last Post Jul 27, 2010, 11:54 a.m. EDT

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago Jul 9, 2010, 2:41 p.m. EDT
I had the same problem and in order to circumvent it and still keep some physical fidelity in the model, I used the free mesh function in the drop down menu and meshed the "thin" part of the model with the coarsest mesh available. good luck.
I had the same problem and in order to circumvent it and still keep some physical fidelity in the model, I used the free mesh function in the drop down menu and meshed the "thin" part of the model with the coarsest mesh available. good luck.

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago Jul 10, 2010, 6:08 a.m. EDT
Thanks for the reply.
I had tried what you said. I had used extremely course mesh size. But it was showing an error of out of memory while meshing. Also i don't think by using such coarse mesh size, I would be able to get the feasible solution.
Can you suggest how mesh can be formed without loosing much accuracy?
Thanks for the reply. I had tried what you said. I had used extremely course mesh size. But it was showing an error of out of memory while meshing. Also i don't think by using such coarse mesh size, I would be able to get the feasible solution. Can you suggest how mesh can be formed without loosing much accuracy?

Ivar KJELBERG COMSOL Multiphysics(r) fan, retired, former "Senior Expert" at CSEM SA (CH)

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago Jul 10, 2010, 3:39 p.m. EDT
Hi

if you are in 3.5 (in V4 the values are different and located elsewhere but basically it's the same story) you must ensure that your relative tolerance is small enough to catch the ration smalles to largest diameter of your items. by default ists around 10E-6 in 3.5. (there are other discussions about this and meshing largley different shapes in comsol on the forum, try a search).

Then I use a rule of thumb that I never have a ratio between two embedded parts greater than about 1E-4. So if you have a large sphere (or block) with a very small ones inside, then I make one or two extra spheres of radius such that the ratio is alsways respected, this adds a few interiour boundaries, but eases or even make possible the meshing, in a simple way.

Good luck
Ivar
Hi if you are in 3.5 (in V4 the values are different and located elsewhere but basically it's the same story) you must ensure that your relative tolerance is small enough to catch the ration smalles to largest diameter of your items. by default ists around 10E-6 in 3.5. (there are other discussions about this and meshing largley different shapes in comsol on the forum, try a search). Then I use a rule of thumb that I never have a ratio between two embedded parts greater than about 1E-4. So if you have a large sphere (or block) with a very small ones inside, then I make one or two extra spheres of radius such that the ratio is alsways respected, this adds a few interiour boundaries, but eases or even make possible the meshing, in a simple way. Good luck Ivar

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago Jul 12, 2010, 11:35 a.m. EDT
hi,
Thanks fr ur reply
I could form the mesh. But when i solved, its showing the error that "streamline plot is failed. All equations can not be evaluated on one domain".
Can you tell me how to tackle this problem?
hi, Thanks fr ur reply I could form the mesh. But when i solved, its showing the error that "streamline plot is failed. All equations can not be evaluated on one domain". Can you tell me how to tackle this problem?

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago Jul 15, 2010, 11:55 a.m. EDT
Hi Ivar,

I couldn´t understand very well the rule of thumb you mentioned. How would you mesh the model attached? Could you help me with that?

Thanks a lot,
Andre
Hi Ivar, I couldn´t understand very well the rule of thumb you mentioned. How would you mesh the model attached? Could you help me with that? Thanks a lot, Andre


Ivar KJELBERG COMSOL Multiphysics(r) fan, retired, former "Senior Expert" at CSEM SA (CH)

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago Jul 16, 2010, 6:04 a.m. EDT
Hi

Well your model should be better "cut" up. I see a thin layer between block 2&3 of material 1. I would really make this as a set of boxes from the front surface to the bottom one, and then add a frame around. Quads and sweep meshing is very easy for regular rectangular shapes. It takes far more time to tweak it for complex geometries. And cutting up your geometry "1" into 2-4 volumes (with interiour continuous boundaries) does not REALLY make the model more complex.

Try it out
Good luck
Ivar
Hi Well your model should be better "cut" up. I see a thin layer between block 2&3 of material 1. I would really make this as a set of boxes from the front surface to the bottom one, and then add a frame around. Quads and sweep meshing is very easy for regular rectangular shapes. It takes far more time to tweak it for complex geometries. And cutting up your geometry "1" into 2-4 volumes (with interiour continuous boundaries) does not REALLY make the model more complex. Try it out Good luck Ivar

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago Jul 27, 2010, 11:54 a.m. EDT
Hi

It is very easy , as suggested by Ivar . just change the geometry tolerance .

In comsol 3.5a u can find it in the ... draw- create composite object ..

In comsol4 --- if u click on union .. then u can see the geometry tolerance.. just change it to for example 1e-6 to 1e-15.

have a look at attached file too.

Rajesh
Hi It is very easy , as suggested by Ivar . just change the geometry tolerance . In comsol 3.5a u can find it in the ... draw- create composite object .. In comsol4 --- if u click on union .. then u can see the geometry tolerance.. just change it to for example 1e-6 to 1e-15. have a look at attached file too. Rajesh

Note that while COMSOL employees may participate in the discussion forum, COMSOL® software users who are on-subscription should submit their questions via the Support Center for a more comprehensive response from the Technical Support team.