[sturgeon]

I'm trying to make a solver similar to bBPF, but with some additional forces. I'm having difficulty, so I'm trying to start out by modifying pimpleFoam to match bBPF, reasoning that if I can get these modifications to work, I can understand how to add additional forces.

To do this I'm referencing the source code, along with the detailed explanation of the buoyantBoussenesqPisoFoam (http://openfoamwiki.net/index.php/Bu...sinesqPisoFoam).

My first question is about the buoyancy force term. In the piso version, it is added to the momentum predictor as:

fvc::interpolate(rhok)*(g & mesh.Sf())

and in the corrector:

rUAf*fvc::interpolate(rhok)*(g & mesh.Sf());

But in bBPF, the terms are:

- ghf*fvc::snGrad(rhok)

and

phig(-rAUf*ghf*fvc::snGrad(rhok)*mesh.magSf());

Why is the term negative in pimple, but not in piso?

My other question is about adjustPhi. As I understand it, adjustPhi will adjust mass outflow to ensure continuity across adjustable outlets like boundaries of zeroGradient for U within a certain tolerance. pimpleFoam contains this adjustPhi, however, bBPF does not. Why is this?

I'm using the hotRoom tutorial as a testcase to compare my own version of the code to bBPF. If I leave the adjustPhi term in, I immediately get a mass discontinuity error. I've discovered the error is linearly proportional to the temperature advantage of the faces set by setFields, and the solver executes fine if I remove this temperature advantage, so clearly it is something to do with that. If I remove adjustPhi from the code entirely, the solver executes, albeit with nonphysical results, but I guess that may be to do with subsequent errors.

I also tried modifying bBPF to include an adjustPhi term, and when I used it to execute hotRoom, it also met with similar mass discontinuity errors.

Apologies for the long post. In short, my questions are:

1) Why is the buoyancy term in bBPF negative, whereas in piso, it is not?
2) Why is adjustPhi not present in bBPF?

Thank you

[alexeym]

Hi,

1.



so



and so, due to formulation, your pressure gradient has extra term, which should be removed. I.e. on openfoamwiki you add buoyancy term, in bBPF you have buoyancy term automatically but have to remove extra term.

2. There is constrainPressure call with comment

Update the pressure BCs to ensure flux consistency

Since pressure in bBPF is not pure pressure, you cannot simply redistribute phi between patches.

[sturgeon]

Thank you!

Since pressure in bBPF is not pure pressure, you cannot simply redistribute phi between patches.

This makes sense. Does this mean that solvers which cannot use adjustPhi are inherently less stable?

[alexeym]

Hi,

I do not think so. You can compare stability of the solvers: take pimpleFoam case, suppress buoyancy (either by setting beta to 0, or avoiding temperature gradients), run it with buoyantBoussinesqPimpleFoam, and compare.