[OldUnixGuy49]

I am modeling a piece of equipment that will be mounted on a roof. I am trying to calculate the surface temperature of roofing below the equipment. This is an image of the simplified physical model.
AirSpaceCFD02.jpg
The top surface is very hot with some insulation separating the heat from an air space (white area). The bottom surface has a cooler bottom boundary. Initially I want to know if natural convection from the open sides is sufficient to keep the roof waterproofing cool. I have the following questions/issues:

1. Is there a single boundary condition that would give me bi-directional flow at the perimeter of the air space? I expect that hot air will exit near the top and cooler air will enter near the bottom.

2. Must I split the faces into two pieces and make one an inlet and the other an outlet?

3. I split the long sides in the middle and assigned a small flow to the bottom inlet. That got the model to run with chtMultiRegionalFoam if I use a deltaT of 0.1, but it is likely to take days of calculation to approach a steady state temperature condition and I still need boundary conditions that will create a natural convection rather than forced.

4. I tried initially to use chtMultiRegionalSimpleFoam to get a steady state solution. With a deltaT of 1.0 or 0.1 the calculation starts but the temperatures of the air layers drop below the coldest boundary and rise above the hottest boundary until the calculation crashes. I can suspect that a shorter time step could stabilize the temperatures, but I haven't seen any references about steady state time step size for calculations.

Thanks in advance for any illumination...