Hi to all,
I would like to implement in Elmer a 3-D model coupling finite elasticity of a solid with diffusion of a fluid species. In particular, elasticity is defined on a 3-D domain and diffusion is defined on the boundary of that domain. Moreover, diffusion depends on the stress at the boundary of the elastic solid.
The first question is this: is it possible to couple 3-D physics with 2-D physics defined on the boundary of the 3-D domain? The 2-D physics must use the same boundary nodes of the 3-D mesh and must be able to use values of the degrees of freedom of the 3-D physics (stresses) computed at the boundary.
The second question, not related to the first, is: can I add custom finite elements to Elmer?
Thanks!!!
coupling 3-D physics with 2-D physics; custom finite element
-
raback
- Site Admin
- Posts: 4832
- Joined: 22 Aug 2009, 11:57
- Antispam: Yes
- Location: Espoo, Finland
- Contact:
Re: coupling 3-D physics with 2-D physics; custom finite element
Hi
Well, you can generally make any equation depend on another one using source terms and material laws. Technically this is achieved by dependency tables, MATC expressions or User Defined Functions (UDF).
Now to couple 3D and 2D leaves open some questions. I mean, the 3D mesh will have more nodes on the common boundary than the 2D mesh. Thus you need some restriction and projection operators. I guess you could use averaging in the 3D direction to obtain BCs for the 2D case?
I don't think that you need to add custom elements into Elmer. There are already 1st, 2nd degree nodal (Lagrange) elements, edge and face elements (Whitney forms), p-elements (up to 10th degree) etc. In Elmer the elements are most often separated from the physics and the equation can be solved with a number of elementtypes. Of course in some equations the formulation is limited to some elementtype. However, such deep connection with element and physics that at least historically was the tradition in Ansys, does not exist in Elmer.
What comes to implementing your own custom equations is in the heart of Elmer. In fact all physical equations in Elmer are just DLL:s that can be freely modified and compiled alone without touching the main program. So just choose your favourite starting point and start forking to your own physics (remembering of course the GPL licensing constraints).
-Peter
Well, you can generally make any equation depend on another one using source terms and material laws. Technically this is achieved by dependency tables, MATC expressions or User Defined Functions (UDF).
Now to couple 3D and 2D leaves open some questions. I mean, the 3D mesh will have more nodes on the common boundary than the 2D mesh. Thus you need some restriction and projection operators. I guess you could use averaging in the 3D direction to obtain BCs for the 2D case?
I don't think that you need to add custom elements into Elmer. There are already 1st, 2nd degree nodal (Lagrange) elements, edge and face elements (Whitney forms), p-elements (up to 10th degree) etc. In Elmer the elements are most often separated from the physics and the equation can be solved with a number of elementtypes. Of course in some equations the formulation is limited to some elementtype. However, such deep connection with element and physics that at least historically was the tradition in Ansys, does not exist in Elmer.
What comes to implementing your own custom equations is in the heart of Elmer. In fact all physical equations in Elmer are just DLL:s that can be freely modified and compiled alone without touching the main program. So just choose your favourite starting point and start forking to your own physics (remembering of course the GPL licensing constraints).
-Peter
Re: coupling 3-D physics with 2-D physics; custom finite element
Thanks for your reply, Peter.
Suppose, then, that I want to couple the two physics living on 3D and on its 2D boudary through some projection operator. Do I have to define two meshes, one for the 3D body and one for the 2D surface, and then the projection operator (maybe a subroutine) from one mesh on the other? Note that, in my problem, the two physics are strongly coupled in both ways, so I cannot solve for the 3D, project on the 2D and then solve for the 2D physics: I have to solve for both the physics at the same time. Is it possible to do that?
(If the problem is too abstract, I can provide further details).
As regards implementing custom elements, a part from the specific problem above for which I do not need new finite elements, it may be useful if one has to test new finite elements: is it possible to define new elements, then?
Suppose, then, that I want to couple the two physics living on 3D and on its 2D boudary through some projection operator. Do I have to define two meshes, one for the 3D body and one for the 2D surface, and then the projection operator (maybe a subroutine) from one mesh on the other? Note that, in my problem, the two physics are strongly coupled in both ways, so I cannot solve for the 3D, project on the 2D and then solve for the 2D physics: I have to solve for both the physics at the same time. Is it possible to do that?
(If the problem is too abstract, I can provide further details).
As regards implementing custom elements, a part from the specific problem above for which I do not need new finite elements, it may be useful if one has to test new finite elements: is it possible to define new elements, then?