Heat Transfer between two solids separated by an air filled cavity

First of all, I would not use a default heat transfer flux on all the walls, unless you know for a fact (derived experimentally) that is the case in this scenario. You chose a very specific number as well, where does that number come from?

As for the air, use the geometry check tool and click "show fluid" to make sure you have air where you think you have air. You mention a subdomain, but this is an internal flow problem with only 1 domain, so there might be some confusion here.

What boundary condition do you want to apply to the side walls of this cavity? Are they adiabatic or constant temperature? How do they interact with the energy transfer between the hot/cold plates?

Can you upload the model?

As far as the walls, you still need to think about the boundary condition on the outer edges. The question is, do you want heat to be able to escape through those walls, or will it all be transferred to the cold plate? If it's the latter, make sure you create an adiabatic condition on the outer surfaces of the walls. This will trap all the heat inside the calculation domain and still enable heat conduction through the walls as well as convection via the air.

Dan,

Your files do not include a Flow SImulation project. Try to "pack and go" again and make sure you have the setup in your files.

OK, I see you used a constant temperature of 1000K for your heat source and ideal wall conditions for your cavity walls. Why the ideal walls though? I'm not sure what you were going for there.

You had the "exclude cavities without flow conditions" option turned on, that's why there wasn't any fluid in the cavity.

The real issue here is that you don't have a heat sink. Without anywhere for the heat to go, the entire solid and fluid domain will simply heat up to 1000K and stay there.

Here is what I did instead:

Added volumetric heat generation sources (equal and opposite on the hot/cold plates)

Suppressed the wall conditions. It seems like there's an assumed adiabatic condition on the boundary of the domain, so that's fine.

Here's a snapshot of some quick 'n dirty results, proving that the boundary conditions work:

You'll note I used 20W instead of 20,000W like you have in the original posting. I'm assuming that was a typo, since that much heat will completely incinerate this steel assembly.

By the way, this simulation would be a good candidate for symmetric domain conditions. You could simulate just 1/4 of the domain with symmetry in the X and Z axes and have a much finer mesh.