Cold attic room under the eaves - unexpected flows
 

Hi.
I’m working om a project for a building faculty, where we try to simulate ventilation of cold attic rooms. Cold attic rooms are attic rooms where insulation is placed on the interior surfaces and therefore they have a temperature close to outdoor temperature. The project is both based on full scale experiments combined with modellings and simulations in ANSYS FLUENT. The purpose of this model is to gain insight in outdoor temperature and wind conditions influence on ventilation of the attic room, and thereby also moisture conditions.

The geometry of the cold attics in the experimental setup is approached by generating a right-angled triangular prism, with an extra plane placed at some distance from the sloped plane. Thereby two volumes are generated; an attic room and a ventilation cavity between the roof plates and roofing underlay. Battens are subtracted from the ventilation cavity. In the plane, which makes up the roofing underlay, two openings are made with a circular plane in front to model valves in the roofing underlay.
At the boundary opening near the eaves and at the ventilation gap in top of the model, an inlet pressure boundary condition is placed. Variations in pressures at the pressure boundary conditions are based on a model the wind influence on the pressures on the surface of the building.

Before doing an actual parameter analysis of the outdoor conditions influence on the ventilation conditions of the attic room, I have made some tests to see if the results make sense. The first question is related to some unexpected results from the tests. The second question is related to how the model might be improved/expanded.

Therefore, I have two questions. Please let me know if you need abbreviation of any of them.

1. What causes a downwards directed flow in the cavity between the roof plates and roofing underlay and how can this flow be calculated accurately?

a. The problem: In one test, the gauge pressure in both pressure boundary conditions are set to 0 Pa. Outdoor conditions are set to a winter situation at 1°C. The velocity inlet at the attic floor is set to 0 m/s, meaning there is no infiltration. This setup results in a flow directed downwards in the ventilated cavity between roof tiles and roofing underlay. This flow is about 0,01 kg/s or about 7 liters/s.

b. What I have tried:
i. The flow downwards becomes smaller for higher positive pressures in the bottom (eaves of roof) or higher negative pressures at the top (cowl of roof), where after it eventually becomes upwards.
ii. The flow downwards also becomes smaller for higher outdoor temperatures, and shifts to be directed up at about 9 °C.
iii. When gravity is switched off, the flow direction is always upwards for positive pressures at the bottom pressure inlet.
iv. When the temperature of the roofing underlay between the attic room and the ventilated cavity is fixed at a certain temperature, the flowrate in the cavity becomes higher for higher temperatures of the roofing underlay.

c. What I expect the reason/solution to be: Since the flow is clearly related to gravity and temperature, I suspect that the flow downwards is caused by a negative stack effect. This could be caused by the slightly warmer attic room. Is this the reason, or could it have something to do with my settings for the roofing underlay plane?
I would calculate such a negative stack in the ventilated cavity like a chimney that experience two different temperatures on either side, by using the average the two as an ‘outdoor’ temperature in the stack calculation. From here I calculate the buoyancy caused by such a difference in density. Is this a correct way to calculate the ‘negative stack’?

2. How can the model be expanded?
To ensure that stack is simulated correctly, I was considering expanding the model to include the ventilated cavity in full height, and not just the section in front of the attic room. (See area encircled by blue line in the cross-section drawing.) Any comments on whether this would be a more correct way to model the setup, with regards to stack effect and magnitude of flow?
At the top (cowl) and bottom (eaves) of the ventilated cavity channel there is a single pressure loss. The pressure loss curve has already been estimated theoretically. What is the correct way to include this in the model? Is it by using the loss coefficient option under the inlet-vent boundary condition?

I think you should also explain what formulation are you using...is it a RANS solution? It seems from the velocity plot that the solution is oscillating therefore this is yet not converged...

It is a RANS solution. I have used a RNG k-epsilon model. I have included full buoyancy effects.
Could it be beneficial to use a realizable k-epsilon model, to converge easier than RNG?
All scaled residuals have met the criteria for a converged solution.