Create basic buoyancy term as node kernel, named buoyancy_density #1274
Conversation
|
The code looks good to me. Are there any regression tests that use it just to make sure it has coverage? We could just modify one of the multiphase tests to use buoyancy_density, for example |
That's a good idea, it should be covered in the reg tests. Should I changed the droplet test? Btw, does it affect your work on the balanced buoyancy at all? |
|
The droplet case would be perfect. The balanced buoyancy introduces a buoyancy_source field variable that gets dropped in instead of rho * g. It should be simple to modify what you have done to pull the field when called for and use it or rho * g as necessary. I think we get this in first and then I can rebase on top of it, putting in the small modifications that it would need as part of the balanced buoyancy PR |
|
Compared the results from the updated VOFDroplet case to the current version, and the reg test passes. I still updated the gold file, you can see that the differences are very small. I haven't confirmed that it works on GPUs or anything like that, but I would expect it to, given how I mimicked the buoyancy_boussinesq implementation. |
There was a problem hiding this comment.
Seems fine to me. @wjhorne I'll merge if you are satisfied.
|
@psakievich Go ahead and merge. It all looks good to me. |
Motivation
Though the more commonly used "buoyancy_boussinesq" source term is implemented as a node kernel and therefore has no issue with hypre solvers, the more basic "buoyancy" source term is not. This source term is more common for multiphase flows, though, and it shouldn't be limited to only tpetra solvers.
Description
I set up a node kernel version of the buoyancy term and named it "buoyancy_density" (as it is density-based) to differentiate it from the other implementation. This implementation has no issues with hypre solvers. Regression tests will be valuable to check against the existing buoyancy source term -- haven't done that yet.
Additional comments
I know Wyatt has been doing a lot of work with the balanced buoyancy implementation that is not yet merged, and I hesitate to disrupt that. Prior discussion leading to this PR is in #1265.