How to define a pretty high thermal zone?

Hello! I'm modelling a large atrium with natural ventilation to test the cooling effects of the natural ventilation. For this atrium, to prevent the smoke spreading in fire, it sets a floor in the middle and connects upper part with lower part by some large horizontal openings. As shown in the picture. image description

For this problem, as the I/O reference mentioned, it cannot use the airflow network, because it cannot calculate the temperature stratification. So, it needs to use a Roomair:node: airflowNetwork. However,there is a confusion.

How can I define this space? If I define this atrium as a single zone and use Roomair:node to simulate temperature stratification, it will ignore the effects of floor slab in middle. But If I define it as two zones, as mentioned in the reference, it cannot calculate buoyant effects between these two zones correctly. Any suggestions are very helpful.

You can define an AFN model that treats the two spaces as (1) two separate zones or as (2) part of one single zone with an internal RoomAir network. But neither option is great.

For option (1), you could have that as just two nodes in the network or as multiple nodes (each zone with an internal RoomAir AFN network). The buoyant effects will be "correctly" calculated between the two spaces (see below) and the floor slab in the middle will be accounted for.

For option (2), with one RoomAir AFN setup, the floor slab in the middle doesn't really exist as far as E+ is concerned, so that won't be accounted for. But the buoyant effects will be "correctly" calculated between the two spaces (see below).

The problem is that for both of these options, the goings on in the two spaces is a problem. Treating the spaces as single nodes violates the well-mixed assumption that underlies the pressure network model (and to a certain extent the rest of E+ as well). So while flows between the spaces are correctly calculated for the temperature and pressure differences, the temperature and pressure differences are likely wrong (so it's a garbage in, garbage out situation). For the RoomAir AFN approach, unless you know what the flow patterns are supposed to look like, then definition of the linkages is pretty difficult. In all likelihood the temperature and pressure differences between the two spaces are again wrong (so again it's a garbage in, garbage out situation).

The best approach I know of is to use CFD or measurements to establish what the flow patterns should be, then you have something to compare to your results (using either option). That's obviously a lot more work than just defining a couple of nodes and simulating, but atria are hard.