[Sereff]

Hi Foamers,

I am trying to study the thermal and particle interactions of a ABL, with buoyantBoussinesqPimpleFoam and DPMFoam respectively. To do this I firstly simulate a
fully-developed neutral boundary (periodic doamin) flow with kEqn model of LES. But as I intergrate the fields to the buoyancy and particle solver. An unphysical area,
where the pressure cell pRefCell was set to have reference pRefValue (see attachment "testBCs_old"), appears on pressure field adn would lead to divergence. This
seems to be a rather common issue when no fixed value was given to p_rgh boundary conditions for density based solvers. However, not many clear solutions were
presented.

I have read that one of the solution is to set p_rgh at upper boundary to be a fixed value. But this is likely to lead to instability or even divergence to the boundary flow,
(see attachment "ABL_diverge"). Another potential solution is to apply the follwing BCs. The patches not mentioned in the following are set to be "cyclic" and the "upperWall"
is actually not a wall since it's a boundary flow. And i am getting the following result (see attachment "testBCs_Prgh/P/U"). Do these screenshots look physical? If not, what
could be the practical way of setting up boundary conditions for p_rgh and U field for density based transient solvers?

P_rgh boundaries:

upperWall{
	type		totalPressure;
	p0		uniform 1e5;
	value		uniform 1e5;
}
lowerWall{
	type		fixedFluxPressure;
	rho		rhok;
	value		uniform 0;
}

U boundaries:

upperWall{
	type		inletOutlet;
	inletValue	uniform (25 0 0);
}
lowerWall{
	type		noSlip;
}

Kind regards,
sereff,

[petros]

Hi Sereff,

Do you have any updates on that?
I am dealing with a similar issue, using the buoyantBoussinesqSimpleFoam instead. When using periodic boundary conditions ('cyclic') I cannot reach convergence. In particular, I notice that p_rgh residuals are stuck at 0.01.

Have you managed to converge your case when using the 'cyclic' boundary condition?

[Sereff]

Hi Petros,

I think one simple way to avoid this is to set one of your boundary to have a Dirichlet boundary condition to avoid the solver assigning a cell to be reference. As you can see from the post above, use the combination:
> P: totalPressure; U: inletOutlet;
at the free surface (atmosphere for example) is able to avoid divergence, and as the attached figures suggested, there would be some fluctuations of pressure. Another way is to just used:
> P: fixedValue; U: inletOutlet;

They worked fine for my simulations, but as we all know CFD is an evil black box, so I can't really tell for sure if it would fix your issue. Give it a try and se if they works.

kind regards,

[petros]

Hi Sereff,

Thank you for your quick reply.
I'm using a constant value for pressure and constant velocitty gradient on top. The problem is that I don't get convergence below 0.01 for pressure. I've seen people in the forum using an additonal momentum source term to maintain the flow, although I'm not entirely convinced about this. Did you encounter anything similar in your case?

[Sereff]

May I ask what type flow are you simulating?

If you are trying to simulate boundary flow or free surface channel flow, that combination of p/U boundary condition is not an optimal combo and would most likely leads to divergence. Like I said, if you need fixedValue for pressure at the free surface, then use inletOutlet for U.

Kind regards,

[petros]

It's a relatively rough neutral ABL.
Weird thing is that simulation isn't diverging. It seems to converge but only around a value of 0.01 for pressure and can't go below. I will try your suggestion and report back in.
Another guess could be that everything is due to the increased roughness.

Best,
Petros

[Sereff]

hmmm... I have experienced an issue of RAS simulation on cyclic ABL, where p/U are all zeroGradient. Basically the solver decided that the pressure field has no more flutuations and the velocity solver just stuck in iterations... unfortunatley i didn't manage to fix that. I just call it a day since the equalibrium is acheived...