hi,
i want to do an error estimation for temperature and flow analysis, to get refinement indicators for the improvement of FEM meshes.
and for this im searching for a special formula, but i don´t found it in the documentation.
the formula look like this:
||u_h -u|| <= D(h^alpha+tau^beta)
im especially interested in alpha and betta.
regards
martin
Error estimation
-
raback
- Site Admin
- Posts: 4838
- Joined: 22 Aug 2009, 11:57
- Antispam: Yes
- Location: Espoo, Finland
- Contact:
Re: Error estimation
Hi martin
I'm no expert at this but I would say that rigorous error estimation is not really realistic in your coupled case. Except for some linear equations this is still a topic of ongoing research. Elmer has just some local error indicators implemented for some of the equations (heat and N-S included). Indicators, however, are only intended as a guidance for the mesh generators, and they do not really help you in saying that the error is +/- 3% for example.
The normal way to estimate error is to perform some sort of grid convergence analysis. I.e. increasing mesh density should not too much influence the results.
-Peter
I'm no expert at this but I would say that rigorous error estimation is not really realistic in your coupled case. Except for some linear equations this is still a topic of ongoing research. Elmer has just some local error indicators implemented for some of the equations (heat and N-S included). Indicators, however, are only intended as a guidance for the mesh generators, and they do not really help you in saying that the error is +/- 3% for example.
The normal way to estimate error is to perform some sort of grid convergence analysis. I.e. increasing mesh density should not too much influence the results.
-Peter