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.

Working with multiple geometries

Please login with a confirmed email address before reporting spam

Dear Ivar,


I had some problems regarding the size of geometries ( Size of smallest part to largest). After I got Instructions from you I decreased the size of large parts.Now the model is working ok. But I do need a large boundary which I couldnot use because of this limitation. I thought to use a couple geometry as you instructed. But my question is If I use a small boundary in one gemetry and If I connected ( through Identity conditions) it to a larger geometry would I get my solution correct. If there is something I might read Could you please let me know that. I am greatly appreciating your help.


Thanks

manjula

4 Replies Last Post 24.04.2010, 11:01 GMT-4
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 23.04.2010, 15:10 GMT-4
Hi

I cannot really tell, as I have never tried it out (coupling two very different size items), but as Comsol tends to consider the geometries separately (carefull with the variable names though) I would assume it^'s a good approach.

The easiest, and what I do when I have such fundamental questions, is to try it out on a (very) simple case, coupled cantlevers or something you have/can solve easily analytically, to better verify and validate your model.

You also have the PML (see previous thread), if you are doing RF models (as you have not stated what you are doing in some details, I'm just guessing here).

In a general way, if we all start our "Subject" field with a i.e [v3.5a RF] ... or something like that it would be easier to understand each others ;)

Have fun Comsoling
Ivar
Hi I cannot really tell, as I have never tried it out (coupling two very different size items), but as Comsol tends to consider the geometries separately (carefull with the variable names though) I would assume it^'s a good approach. The easiest, and what I do when I have such fundamental questions, is to try it out on a (very) simple case, coupled cantlevers or something you have/can solve easily analytically, to better verify and validate your model. You also have the PML (see previous thread), if you are doing RF models (as you have not stated what you are doing in some details, I'm just guessing here). In a general way, if we all start our "Subject" field with a i.e [v3.5a RF] ... or something like that it would be easier to understand each others ;) Have fun Comsoling Ivar

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 23.04.2010, 15:22 GMT-4
Dear Ivar,

Thanks for your information. Sorry for not explaning much about my modeling. I am working with a diffusion problem with small items. Chemical Engineering-Mass transport-diffusion. The reason I am looking for a large boundary is to get the correct solution. If you can give me more instructions with these delails It would be great. I will try with multiple geometry with a simple case where I know the answer already. Thanks again.

Manjula
Dear Ivar, Thanks for your information. Sorry for not explaning much about my modeling. I am working with a diffusion problem with small items. Chemical Engineering-Mass transport-diffusion. The reason I am looking for a large boundary is to get the correct solution. If you can give me more instructions with these delails It would be great. I will try with multiple geometry with a simple case where I know the answer already. Thanks again. Manjula

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 24.04.2010, 02:38 GMT-4
Hi

well then its not that easy as PML are for RF and electromagnetic waves, probably not implementable as is for diffusion, but there must be a way to apply the same principle for your case, if comparable, it should be easier as you have no reflections do deal about. Needs some thinking, try to make a very simple case.

By the way, one thing is that probably you do not need the same mesh density all around. When you have very different mesh densities you get also problems (again I use the 1:10'000 rule). This is mostly easy to overcome, just box your fine model with a few intermediate squares/circles/spheres, "à la poupée russe" then you mesh from the inner to the outer, with a mesh ratio of at most 1:10'000 in each region

Finaly if your gobal behaviour can be analysed by a detail view in a small region, there should be also a possibility to have 2 geometry regions, one with the fine "details" and one with the overall response, to separate the two scales, but this need some analytical tweaking of your model equations and then to distribute them correctly on the two models and to use extrusion or other coupling variables to link the 2 models

Hope this helps on the way
Ivar
Hi well then its not that easy as PML are for RF and electromagnetic waves, probably not implementable as is for diffusion, but there must be a way to apply the same principle for your case, if comparable, it should be easier as you have no reflections do deal about. Needs some thinking, try to make a very simple case. By the way, one thing is that probably you do not need the same mesh density all around. When you have very different mesh densities you get also problems (again I use the 1:10'000 rule). This is mostly easy to overcome, just box your fine model with a few intermediate squares/circles/spheres, "à la poupée russe" then you mesh from the inner to the outer, with a mesh ratio of at most 1:10'000 in each region Finaly if your gobal behaviour can be analysed by a detail view in a small region, there should be also a possibility to have 2 geometry regions, one with the fine "details" and one with the overall response, to separate the two scales, but this need some analytical tweaking of your model equations and then to distribute them correctly on the two models and to use extrusion or other coupling variables to link the 2 models Hope this helps on the way Ivar

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 24.04.2010, 11:01 GMT-4
Dear Ivar,

Thank you so much for all the information. I will try to proceed with your advices.

Thanks Again.

Manjula
Dear Ivar, Thank you so much for all the information. I will try to proceed with your advices. Thanks Again. Manjula

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.