Opened 17 years ago
Last modified 16 years ago
#252 new defect
Transmissive_Momentum_Set_Stage boundary may cause numerical instabilities
| Reported by: | ole | Owned by: | steve |
|---|---|---|---|
| Priority: | normal | Milestone: | |
| Component: | Functionality and features | Version: | |
| Severity: | major | Keywords: | |
| Cc: | stephen.roberts@… |
Description
This boundary condition was implemented to allow momentum to be automatically derived from the underlying numerical scheme and it was instrumental in getting the Okushiri validation to work.
However, it appears that numerical instabilities may arise from this in cases where reflected waves flow back into this boundary. This is not unlike the instabilities we encountered with the fully transmissive boundary. The example https://datamining.anu.edu.au/anuga/browser/anuga_work/debug/bad_time_step/run_dam.py has been boiled down to the basics and demonstrates this phenomenon. The instability occurs at the point where the reflected wave interacts with the boundary. The diagnostic output confirms that huge momentum values build up and that the large speeds resulting are not due to very shallow water.
Coarsening the mesh seems to help, but it would be good to address the underlying issue.
Good luck!
This does not cause problems with the Hinwood study.
https://datamining.anu.edu.au/anuga/browser/anuga_work/development/Hinwood_2008/run_dam.py (This code has to be uncommented to use Transmissive_Momentum_Set_Stage_boundary).