Axisymmetric channel

[tannhorn]

Hello,

Not sure if this belongs here or to the nekRS forum, but here is the question. I would like to use nekRS for fluid dynamics solution in a multiphysics set-up, in particular solving the RANS equations + energy conservation, the fluid domain in question being a cylindrical flow channel. With Nek5000, I could just simulate in two dimensions with axisymmetric setting, but according to the nekRS documentation, it can only do 3D meshes -- what would be the preferred strategy to using nekRS (while avoiding excessive mesh size)?

Thanks!

[aprilnovak]

Hi @tannhorn! Great question - you are correct, NekRS only supports 3D geometries. To solve physically 2-D flows (like axisymmetric cases or symmetry cases), you can create a mesh which is just 1 element deep in the symmetry direction (so for your case, this would be like a 1-degree cylindrical wedge). On the two faces in the symmetry direction, you can use sym boundary conditions for velocity and I boundary conditions for temperature (the zero-gradient conditions).

[tannhorn]

Hey, thanks for the quick answer! I presume that instead of a symmetry axis, I should just have a finite-length boundary at the inner edge of the domain? Any recommendation about the value of the inner radius $r_0$ in terms of the overall radius of the channel? Do you anticipate any problems with the elements having significantly different width near the symmetry axis and at the outer edge of the channel? I will go ahead and try but just wanted to see if there are any established best practices beforehand.

[aprilnovak]

Hi @tannhorn, do you mean the geometry you are intending to model is corresponding to an annular cylinder (inner and outer radius), or a full cylinder (fully described in terms of just a radius)?

[tannhorn]

Full cylinder. Sorry for not being precise.

[aprilnovak]

In either case, I think what I suggested should work - let me know if you run into any issues! I don't think there should be issues with different widths, since the physics in the transverse direction is entirely symmetric (no gradients).

If you want to send me your email (you can find mine here: https://npre.illinois.edu/people/profile/ajnovak2), I can add you to the NekRS users slack, which can be a good place to ask future questions as well.

[tannhorn]

Thanks, will do!