Thermal boundary non-cubic domain

[alexandr...@gmail.com]

Hello,

I'm trying to simulate a simple Rayleigh-Bénard convection inside a rectangular domain.

my code : 

http://basilisk.fr/sandbox/Vierron/Natural-convection/rbvisco.c

but I don't understand why my top thermal boundary doesn't work ? 

and I have confident in the code ... 

Any suggestions are welcome.

Thanks,

Alexandre Vierron

[Francesco Picella]

Hello Alexandre,

To sort it out, I'd suggest you _de-couple_ your problem first.

Have you tried to solve _only_ for the diffusive problem?

In a purely diffusive regime, with fixed Dirichlet BC on top/bottom, you should get a linear profile...

Best,

F

[Edoardo Cipriano]

Hello Alexandre,

I think the problem is the line:

size (npe());

which sets up a domain with length = 4 (due to the 4 processors used).

Then you initialize the temperature field to the value -y. Therefore, the top initial top temperature, considering the origin shifting of (-0.5,-0.5), will be -3.5.

When the simulation starts, it correctly sets the BCs to -0.5 on top, thus leading to a sort of heating on top compared to the initial condition.

If you avoid the line size (npe()), the problem looks reasonable.

Or maybe you can directly use dirichlet (y) and dirichlet (-y) for the temperature boundary conditions.

Hope this helps.

Cheers,

Edoardo

[alexandr...@gmail.com]

Hello,

Thanks for yours replies : 

I tried size (4) but the problem remain

I found a solution : 

dimensions(nx = npe(), ny = 1);

I set the x dimension with nx, but in the past, dimensions(ny=1) worked.

Thanks,

Alexandre

[Stephane Popinet]

> I set the x dimension with nx, but in the past, dimensions(ny=1) worked.

Yes, this has changed in the latest release, see:

http://basilisk.fr/src/NEWS#non-squarecubic-domains-on-multigrids