Bugfix inc hyd
This fixes issues #972 (closed) The solution is to allow vertical diffusion even in cells that are distorted. If not, effects of bottom friction are correctly taken into account and velocities are completely wrong. I think that for prisms, it is not a problem to allow vertical diffusion, as this happens in a vertical profile only, so it should not be affected by hydrostatically inconsistent elements.
A few notes: I only tested for prismatic elements, not for tetraëders (so i do not know whether also changes are needed to mt02_tt.f). I only tested with hydrostatic flow.
I found that the V test case is affected. Now some spurious currents (with velocities 1E-7 m/s) are generated. That's why I changed the coefficients in the vnv.py file. the currents are weaker though than without the use of the hydrostatic inconsistencies filter.
Merge request reports
Activity
added module:telemac-3d type:bug labels
assigned to @chi-tuan.pham
Hello Alexander,
Thanks for you suggestion. I have to think deeper but I am quite busy for the next 10 days.
If your suggestion modifies the results, it is normal that the values of the epsilons have changed. You should first update the reference files and then the epsilons should better work (in particular the comparisons between results in scalar or parallel and the new reference file). When updating the reference file, do not forget to only keep the last time step (see run_telfile.py alter -s -1 etc).
Maybe other reference files of the TELEMAC-3D examples should be updated as well (and epsilons also). You should run the validation with tag "telemac3d" to check: validate_telemac.py --tags='telemac3d' as described in the Software Quality Plan (see e.g. the checklist in appendix C).
Chi-Tuân
Hi Chi-Tuan,
I have had a further look at this case. I wrote out the matrix with and without the fix in the V case, and found no differences in that, so the results should be the same. I think the different epsilons were probably due to a compiler setting (-ffast-math). I will revert the epsilons, rerun the validation for telemac3d and if it is then ok, update the merge request.
Note that there are some interesting things in the V test case. The test case is intended to show, that no flow develops in this configuration, but actually there is some flow. Strangely, the flow disappears by setting horizontal diffusion for tracers to zero.
Kind regards
Alexander
P.s. Do you know how much disk space a full test run takes? I recently had a crash because a partition on our cluster was completely filled after trying to run a test.
added 76 commits
-
39b56ec3...ef067927 - 75 commits from branch
main - 36c25ad3 - adding bugfix for wrong velocities/water levels with inchyd
-
39b56ec3...ef067927 - 75 commits from branch
- Resolved by Alexander BREUGEM
- Resolved by Alexander BREUGEM
- Resolved by Alexander BREUGEM
- Resolved by Alexander BREUGEM
- Resolved by Alexander BREUGEM
- Resolved by Alexander BREUGEM
- Resolved by Alexander BREUGEM
- Resolved by Alexander BREUGEM
added 18 commits
-
917db02b...db3408ae - 9 commits from branch
main - c633c832 - adding bugfix for wrong velocities/water levels with inchyd
- c272d9b4 - Apply 1 suggestion(s) to 1 file(s)
- fd4c0840 - Apply 1 suggestion(s) to 1 file(s)
- 0faab2fa - Apply 1 suggestion(s) to 1 file(s)
- a50f508b - Apply 1 suggestion(s) to 1 file(s)
- 6bba890f - Apply 1 suggestion(s) to 1 file(s)
- c0b4fddc - Apply 1 suggestion(s) to 1 file(s)
- 70214f84 - adding suggestions chi-tuân
- 8ee2dcfc - adding news
Toggle commit list-
917db02b...db3408ae - 9 commits from branch
mentioned in commit 521e08b7