Dirichlet bc with embed.h - not reached

[Cyprien jsc]

Dear all,

I’m interested in simulating a convection driven by an inflow of a hot fluid.

I’d like to set up a Dirichlet condition in speed and temperature for the inflow, and a Neumann condition on the same face for the surrounding (for an open boundary around the inflow).

I understand that the best way is to use the embedded boundaries as karman.c example.

But I have trouble to enforce the boundary condition on embed. Without embedded boundary, it works correctly. Once activated, it still have the information of the boundary condition, but it doesn’t reach the expected value ( here T_insitu[embed]  =dirichlet (100.) ). I tried to play on partial cells ratio.

 I also tried the list of fixes suggested in this discussion with no success:  https://groups.google.com/g/basilisk-fr/c/SLnKb8U0BXs.

Would someone have a hint of a solution?

I enclose the code, my configuration is a box with a scalar to define the exit, and a embed plan in xy. I use a Paraview output from Acastillo sandbox (also attached).

Cheers,

Cyprien

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

Hi,

I do not if this is your case but I have found  difficulties in the convergence of the Poisson solver with embed  in 2D if a minlevel is not set.

...., minlevel = 4);

Cheers

Pepe

[Cyprien jsc]

Hi Pepe ,

Thank you for your answer,

The convergence is reached properly (and in my configuration I did specified the minlevel). I think the issue here is linked to the flux computed only for partial cells.

Cyprien

[Bruno Deremble]

Hi Cyprien,

Yes, embedded boundaries and Neumann conditions are tricky to setup. check the
discussion here: https://groups.google.com/g/basilisk-fr/c/NxPmr50E0KM


Best,
Bruno

[Gong CHEN]

Hello, I would like to ask what is the meaning of the _k parameter in output_vtk_foreach.h. I always get an error when I try to get vtk through this file, showing that _k is not decleared, thanks~

[Cyprien jsc]

Hi  Bruno,

Thank you for the tip,

I wasn't aware the embedded boundary is not rotated. But shifting the domain and assigning the correct orientation for velocity didn't resolve the issue. Did you manage to enforce the neumann(1) on embedded boundary in the channel?

Like Antoon said, if the no-penetration condition is always enforced then I guess any condition other than dirichlet(0) won't work.

The workaround I see would be to create a hole in the embed part to assign the inflow velocity on a back standard boundary, and a neumann(0) in the embed part.

Cheers,

Cyprien

[Bruno Deremble]

I must admit, I did not finished my homework for this problem (yet)