[sturgeon]

Hi all

I'm using buoyantBoussenesqPimpleFoam to simulate a relatively simple case, where I have a 2D rectangle, with an inlet left, outlet right, wall bottom and the top should be open to the atmosphere, with a heat source in the middle added by fvOptions. Essentially it is the circuitboard cooling tutorial, but rather than a wall, it has an open top. The inlet conditions are just standard fixedValue for U and fixedFluxPressure for p_rgh, for the outlet I've been using inletOutlet or pressureInletVelocityOutlet for U and totalPressure for p_rgh. Floor is noSlip, fixedFluxPressure.

The only other difference is that the domain is significantly larger than the circuitboard cooling case - I am trying to do a rudimentary model of the urban heat island. I am aware that atmosphere stratification effects would be an issue, and I intend to use a modified version of the solver to work around this - just now I'm trying to use the existing solver to fix boundary conditions rather than possibly encounter a bug from me poorly coding the modified solver, then switch to the modified solver.

The ceiling boundary condition is the one I can't figure out. I've tried matching the inlet conditions, as the ceiling should be far enough away from the thermal disturbances that it should match the freestream velocity, but the case eventually settles with a persistent vertical downwards velocity over the heat source, which is clearly wrong, and some upwards velocity at the outlet. (I can post more details and screenshots from paraView if necessary - the upshot is that this boundary condition doesn't seem to work)

The next thing I tried was slip, which I'd seen used in literature for modelling flow over small hills. But since that makes it a wall, it seems to accelerate the flow over the heated area, and downwind I have an upper layer of accelerated flow, and a lower of decelerated. This is sort of expected, but it continues all the way throughout the domain, whereas I thought it would disperse and mix together to converge back on the inlet conditions. (Domain subsequent of the thermal source is very long)

The last thing I tried was removing the inlet and making every side except the floor a zeroGradient in U and totalPressure in p_rgh, matching the damBreak case, to see how this would work. But the simulation settles on this:



So the air at the bottom is getting drawn into the heat source, which is expected, but at the top it seems to flow outwards, rather than just passing out of the domain, as if the top is a wall.

Is anyone able to give me some advice as to what the ceiling boundary condition should be to achieve this? I can post more information and case files if necessary.

Cheers

[piu58]

I don't understand fully what you try to simulate. A sketch of the problem would have eben more useful than the paraFoam results.

Nevertheless, I have some hints:
- You should have an outlet somewhere. The numerics get more stable if you have one instead of a circulating flow. In many problems it does not count whether there is a large volume ore an open space.
- The boundary conditions for p_rgh are quite critical. Especially the reference area, where you set it to zero. The outlet is a good point to set it. The area where you need the result is the worst.

[sturgeon]

Sorry, here is a sketch of the problem:



As you can see, it's fairly simple, just like the circuitboard cooling tutorial case, except I want the ceiling to open to the atmosphere, instead of a fixed wall.

- You should have an outlet somewhere. The numerics get more stable if you have one instead of a circulating flow. In many problems it does not count whether there is a large volume ore an open space.

I do have an outlet - at the far right of the domain

- The boundary conditions for p_rgh are quite critical. Especially the reference area, where you set it to zero. The outlet is a good point to set it. The area where you need the result is the worst.

The outlet is set to 0 with a totalPressure boundary condition

Thank you for your response

[piu58]

I don't know what you want to have at the ceiling. You may use a two component simulation, with air on top. I don't believe that is necessary in your case. You may try if a zeroGradient for U is adequate to your question.

[sturgeon]

Originally Posted by piu58

I don't know what you want to have at the ceiling. You may use a two component simulation, with air on top. I don't believe that is necessary in your case. You may try if a zeroGradient for U is adequate to your question.

I have tried zeroGradient - I get strange results where the vertical plume appears over the heat source as expected, but then the plume seems to travel downwind and the simulation settles on a strange solution where the vertical velocities are fluctuating randomly.

Of course it is possible that there is another error in my simulation, but that would seem strange since the case is just modified from the circuit board cooling case, and all I've changed is the upper boundary, enlarging the domain, adjusted turbulence values and added a heat source in fvOptions.

[bullmut]

HI Sturgeon


did you come right in the end?
I am also using the circuit board cooling as my base to a simulation on heating in a building. But i keep getting negative densities.
Hoping you could help me here with the boundary conditions.