cylindrical flow (metric + embedded + adaptivity)

[alexandr...@gmail.com]

Hello,

I wanted to simulate Rayleigh-Bénard instabilities (RBI) in a cylinder with embedded boundaries.

I already have a code for RBI in a cylinder without embed :

http://www.basilisk.fr/sandbox/Vierron/Natural-convection/cylindrical/rbcyl.c

(I will improve the video)

Now with embed.h :

http://www.basilisk.fr/sandbox/Vierron/Natural-convection/cylindrical/rbcyl-emb.c

For the moment I'm not sure at 100% that my conditions are respected at the embedded boundaries. I'm still working on it but my code runs.

So following the link below

http://www.basilisk.fr/src/COMPATIBILITY

It's said that metric are maybe compatible with Solid. For the moment in my case (radial.h) I'd say yes.

I then tried to use the adaptive mesh, but I immediately get an error:

/home/alexandre/basilisk/src/embed-tree.h:243: Assertion `nv > 0.' failed.
../sysdeps/unix/sysv/linux/raise.c:50:error: Program received signal SIGABRT, Aborted.

But the compatibility matrix said "yes"...

Do you have any idea why ?

(It would be interesting to have a small simulation of a flow in cylindrical coordinate).

Last question  : 

It's necessary to write boundaries conditions on "y" direction and "x" direction ?

because we have u.r :

for example :

u.n[top] = dirichlet(0.);

u.n[right] = dirichlet(0.);

Thanks,

Best regards,

Alexandre

[jose.lopez...@gmail.com]

Hi Alexandre,

The problem is described here...

https://groups.google.com/g/basilisk-fr/c/tSfpvPXR1Q4/m/VXkD8RLDBAAJ

The problem is in the adaption. The list of variables must be reordered before the adaption, 

Here I sent you a patch I think it should fix it.

Please, let's me know if it works for you.

Cheers

[alexandr...@gmail.com]

Hello,

Thanks for your reply.

I apply this patch

(darcs apply 23-08......patch)

and re-run my code :

http://www.basilisk.fr/sandbox/Vierron/Natural-convection/cylindrical/rbcyl-emb.c

But :

/home/alexandre/basilisk/src/grid/tree-common.h: In function ‘adapt_wavelet’:
/home/alexandre/basilisk/src/grid/tree-common.h:180:27: error: ‘cs’ undeclared (first use in this function); did you mean ‘cm’?
/home/alexandre/basilisk/src/common.h:50:26: note: in definition of macro ‘is_constant’
/home/alexandre/basilisk/src/grid/tree-common.h:180:27: note: each undeclared identifier is reported only once for each function it appears in
/home/alexandre/basilisk/src/common.h:50:26: note: in definition of macro ‘is_constant’
/home/alexandre/basilisk/src/grid/tree-common.h:181:65: error: expected expression before ‘{’ token
make: *** [rbcyl2.s] Error 1

The first error say: ‘cs’ undeclared  but cs should be a global variable ?

Thanks

Best regards,

Alexandre

[alexandr...@gmail.com]

Even with this code

http://www.basilisk.fr/sandbox/Vierron/Natural-convection/cylindrical/rbcyl.c

I get the same error

[Yi Dai]

Hi Alexandre,

do you use the latest version of basilisk, i treid this example rbcyl.c without any problem. 

Dai

[alexandr...@gmail.com]

Hello,

alexandre@alexandre-HP-EliteBook-x360-1030-G8-Notebook-PC:~/basilisk$ darcs pull
Pulling from "http://basilisk.fr/basilisk"...
patch dfa701bfa5976f6ddabf45d7c52b6628378f84f4
Author: Stephane Popinet <pop...@basilisk.fr>
Date:   Sun Jun 25 15:35:09 CEST 2023
  * Z-buffer is simply an array of floats
 
  This should fix the display when using libfb_tiny with MPI
Shall I pull this patch? (1/2)  [ynW...], or ? for more options: y
patch d6f41fa60b1444cf7dd17073b6b62ce09619a1da
Author: Stephane Popinet <pop...@basilisk.fr>
Date:   Wed Jun 28 15:46:24 CEST 2023
  * The normal is always the latest one
Shall I pull this patch? (2/2)  [ynW...], or ? for more options: y
Do you want to Pull these patches? [Yglqk...], or ? for more options: y
Finished pulling.

do you use the latest version of basilisk ? almost yes ...

My code : rbcyl.c works fine without the  jose's patch 23-08-02_list_reordering_adapt (see above)

The problem comes from this code with adapt_wavelet :

http://www.basilisk.fr/sandbox/Vierron/Natural-convection/cylindrical/rbcyl-emb.c

Have you tried ?

How can I temporarily remove your patch Jose ?

Thanks,

Alexandre

[Yi Dai]

I tried adding cs to adapt_wavelet and it run. (without the patch)

Dai

[alexandr...@gmail.com]

Yes ! you're right, it run.

First : I remove jose's patch (darcs rollback -p "embed default variables before metric ones reorderedin lists for adaption")

and add cs to adpt_wavelet !

That's good news ! So we can use radial.h + embed.h + adapt in 3D

Thanks,

Alexandre

[Alexis Berny]

By any chance, does your simulation works with adaptivity, embed boundary AND two phases?

[alexandr...@gmail.com]

Hi Alexis !

I think it can only work with two phases if one of them touches the embedded boundary (I'm not sure).

Best regards,

Alexandre

[Alexis Berny]

Which would explain why my code is not working. I made a small simulation that show the problem (in the bug report)

Interestingly, we can use two-phase and embed together if there is no adaptivity

[Youssef Saadeh]

Dear all,

Please check Pepe's mail in this same thread. It should work if you apply the patch in here.

Best,

Youssef