Hello fellow Elmers,
i wanted to ask if you guys know how hard it would be to develop your own solver in the field of contact resistance.
I basically would implement an infinitesimal small Area Resistance (Contact Resistance) by following equation.
n * J = h (V1-V2)
, where n = Normal Vector, J = Current density (A/m2), h = specific contact conductivity (S/m2), and V1 V2 as the potentials in V.
I am referring to the following page:
https://doc.comsol.com/5.5/doc/com.coms ... 7.057.html
Is it maybe possible that the equations in the link can be used in the matrix stiffness matrix composition of the electrostatic poisson equation?
In case of two potentials,ie V1 and V2, that are located in the same node at the interface.
Would it be hard, would it even make sense?
Your thoughts about that are apreciated.
Cheers,
Domi
Contact Resistance
-
raback
- Site Admin
- Posts: 4832
- Joined: 22 Aug 2009, 11:57
- Antispam: Yes
- Location: Espoo, Finland
- Contact:
Re: Contact Resistance
Hi Domi,
This is done for the heat equation:
https://github.com/ElmerCSC/elmerfem/bl ... c.F90#L374
Now the jumps are nothing new for Elmer. The 1st approach was to modify the mesh so that the nodes are duplicated at the interface. This becomes cumbersome if there are other equations since for them you need to create additional constraints that ensure the continuity. So you open a pandoras box. Such constraints make the linear system more challenging etc. So ideally you only want to add the jump to the one equation and keep the mesh intact. You can still find discussion and examples of this 1st approach.
The above approach uses discontinous Galerkin basis where the jumps are maintained only where requested. It is implicit i.e. no nonlinear iterations are required to find the solution.
-Peter
PS. The may be a ready implementation for this. Mail me if interested.
This is done for the heat equation:
https://github.com/ElmerCSC/elmerfem/bl ... c.F90#L374
Now the jumps are nothing new for Elmer. The 1st approach was to modify the mesh so that the nodes are duplicated at the interface. This becomes cumbersome if there are other equations since for them you need to create additional constraints that ensure the continuity. So you open a pandoras box. Such constraints make the linear system more challenging etc. So ideally you only want to add the jump to the one equation and keep the mesh intact. You can still find discussion and examples of this 1st approach.
The above approach uses discontinous Galerkin basis where the jumps are maintained only where requested. It is implicit i.e. no nonlinear iterations are required to find the solution.
-Peter
PS. The may be a ready implementation for this. Mail me if interested.