Noconforming mesh

Numerical methods and mathematical models of Elmer
Post Reply
MrBenson
Posts: 35
Joined: 31 Oct 2014, 19:44
Antispam: Yes

Noconforming mesh

Post by MrBenson »

Hi everybody,

I'd like to know if Elmer can work with nonmatching meshes ? Is there some DDM inside like Mortar method ? I'd like to handle a complex geometry divided into components without meshing the all geometry.

Best regards.
raback
Site Admin
Posts: 4812
Joined: 22 Aug 2009, 11:57
Antispam: Yes
Location: Espoo, Finland
Contact:

Re: Noconforming mesh

Post by raback »

Hi MrBenson

Yes, sort off. There are quite a few (39) tests involving mortar FEs:

Code: Select all

 
elmeruser@ElmerCSC-VM ~/Source/elmerfem/fem/tests $ grep  -i 'apply mortar bcs' */*.sif | wc -l
39
Mortar FEs have been primarily used to allow continuity of solution over moving interfaces. By default the resulting system then becomes a saddle-point problem making the convergence slower. Hence this has not been too much applied to very large problems where this kind of meshing problem usually occurs as there the convergence is often critical.

Now for some applications the elimination of the mortar dofs works quite well, so there it might not be too much additional cost for convergence.

Also, we did some testing using the mortar conditions appliying standard DMM schemes (such as Dirichlet-Neumann & Robin-Robin) using the mortar FEs. However, this didn't give satisfactory convergence at our target problem at the time (rotating machines) so the tests didn't carry to the official distribution.

-Peter
MrBenson
Posts: 35
Joined: 31 Oct 2014, 19:44
Antispam: Yes

Re: Noconforming mesh

Post by MrBenson »

Hi Peter,

Thank you a lot for your answer. Do you think i can use mortar to solve heat equation with 10^4/6 mesh elements (and maybe around 10^3/5 mortar elements) ? I'm not exigent about the temperature relative error, only 10^-2. I have a 2D (but also 3D after) global domain composed of many nonconforming/nonmatching (but nonoverlapping) subdomains, and i don't want to mesh the whole domain but only the subdomains. I can't see the way to prescribe boundary conditions (Natural and radiation also) on some lines (or surfaces) that i do not know yet, cause it may be a partial-joined line/face. Not sure to be clear about Mortar sorry, cause I'm new with that method i just discovered it, i know a little bit more about DDM with conforming meshes.

Thanks.
Post Reply