You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Finding adequate solver settings is not a trivial task. I.e. given some accuracy, which combination of the number of axial cells, absolute/relative tolerances etc. is optimal s.t. computation time is minimized.
As a (naive) first step, functionality should be added to automatically create regular convergence plots (along given dimensions). However this becomes infeasible if the number of hyper parameters increases (e.g. for DG).
For this purpose, we should add a tool that automates this task. The final workflow could look something like:
Detect all hyper parameters in Process
Generate reference solution as ground truth (or use external one if provided)
Theres some code for convergence analysis in CADET-Verification, which I'd like to see cleaned up and outsourced to CADET-Process, see CADET-Verification/#4.
As stated in the issue, this would be a good task for and nice contribution from a student. This would also make a good basis to continue with the optimization issue as described above in the issue description.
Finding adequate solver settings is not a trivial task. I.e. given some accuracy, which combination of the number of axial cells, absolute/relative tolerances etc. is optimal s.t. computation time is minimized.
As a (naive) first step, functionality should be added to automatically create regular convergence plots (along given dimensions). However this becomes infeasible if the number of hyper parameters increases (e.g. for DG).
For this purpose, we should add a tool that automates this task. The final workflow could look something like:
Process
The text was updated successfully, but these errors were encountered: