embed + axi update

748 views
Skip to first unread message

jmlopez...@gmail.com

unread,
Mar 15, 2023, 1:40:38 PM3/15/23
to basilisk-fr
Hi all,

I have updated the patch I did with my implementation of the embed with the axi in order to  work with the last stable release (tag 'release 22-07-05').

The patch contains strictly the changes done to deal with embed + axi problems. I have dismissed my attempt to remove the small cell problem in vof variables (It was wrong) as well as the slip boundary condition on embed.

I also add some (very few) missing metric factors in all-mach.h and momentum.h that bother me a lot.

I hope you find it  useful.

Cheers
2023_03_01_axi+embed_over_tag_22-07-05.patch

Stephane Popinet

unread,
Mar 29, 2023, 1:08:02 PM3/29/23
to basil...@googlegroups.com
Hi Pepe,

Thanks for the patch. I am in the process of merging it with the latest
release.

There are a few missing files in your patch.

Can you please send me the .ref files for the new test cases:

neumann-axi.ref
dirichlet-axi.ref
rising-axi-momentum.ref
lamb.ref
poiseuille-embed.ref

cheers,

Stephane

jmlopez...@gmail.com

unread,
Mar 30, 2023, 7:28:57 AM3/30/23
to basilisk-fr
Hi Stephane,

Sorry, I always forget that a proper test case needs a ref file to compare with!

I have amended the patch incorporating these ref files.

In the other hand, I did not modify vof.h although I think it is worth to adapt your trick in

 patch 21d01dbdc2e01a23dc6de47ff17c6b835ba0deed
Author: Stephane Popinet <pop...@basilisk.fr>
Date:   Mon Oct 25 14:22:31 CEST 2021
  * A simple implementation of VOF + embedded boundaries

to the metric. It simply requires  to consider just the metric factor (not the solid) what it is just multiply by cs[]/cm[].
I have put it in a different pach (and file) in case you wish to consider it. I did not merge both because I have not prepared any test despite it is a minuscule change.

Cheers

Pepe
2023-03-30_stephane_trick_with_metric.patch
2023_03_01_axi+embed_over_tag_22-07-05.patch

Stephane Popinet

unread,
Apr 2, 2023, 1:37:52 PM4/2/23
to basil...@googlegroups.com
Hi Pepe,

Thanks for the files. You will see that the latest release includes two
new patches:

2023-04-02 19:02:51 CEST (J.M. Lopez-Herrera): Embedded solids can be
combined with metric

2023-04-02 18:04:47 CEST (J.M. Lopez-Herrera): Added missing metric
terms in the all-Mach and momentum solvers

With some substantial changes compared to yours in the embed + axi
patch. Please check.

I have only included the neumann-axi.c and dirichlet-axi.c test cases
for now. For the other two cases:

- lamb.c : I don't think it is really finished when I compare with:

http://gerris.dalembert.upmc.fr/gerris/tests/tests/axi.html

There is no convergence study etc. and you should not need to run it for
so long (up to t = 50)

- poiseuille-embed.c : the results do not look very convincing?

Please tell me if you agree and we can think of how to go further. It
would be nice to have a few more tests. I also think there probably are
issues with adaptive refinement etc.

cheers,

Stephane

jmlopez...@gmail.com

unread,
Apr 2, 2023, 4:51:07 PM4/2/23
to basilisk-fr
Hi Stephane,

I fully agree with you. Either the lamb and the poiseuille-embed are more a proof I did in a rush  that all the viscous terms were considered correctly. They need more work. I will go back to lamb test and see if I am able to turn it in a real test.

Cheers

Pepe

jmlopez...@gmail.com

unread,
Jun 9, 2023, 6:22:04 AM6/9/23
to basilisk-fr
Hi Stephane,

After the warning arisen in

https://groups.google.com/g/basilisk-fr/c/HuBi2hRkMwE

I notice that the default order of the 'all' list is...

1: fm.x
2: fm.y
3: cm
4: cs
5: fs.x
6: fs.y
7: p
....

which gives up to problems with the refining/coarsening of cells.  Note that cm & fm depends on  cs & fs. Therefore, cs & fs  must be updated first. The 'all' list could  be reordered in the metrics events (in embed.h) mimic to what is is done in axi.h.

Also, a reordering is performed in grid/tree-common.h to assure the order in adapt_wavelet() (although it might be unnecessary if the default ordering is kept).  Also in  tree-mpi.h a reordering is done.

Do you wish I propose a patch about it ?

cheers 

Jose

El domingo, 2 de abril de 2023 a la(s) 19:37:52 UTC+2, Stephane Popinet escribió:

Youssef Saadeh

unread,
Jun 9, 2023, 6:28:23 AM6/9/23
to jmlopez...@gmail.com, basilisk-fr
-- 
You received this message because you are subscribed to the Google Groups "basilisk-fr" group.
To unsubscribe from this group and stop receiving emails from it, send an email to basilisk-fr...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/basilisk-fr/1edf9108-8ba9-4a1d-aa45-9eb4a0563474n%40googlegroups.com.

Reply all
Reply to author
Forward
0 new messages