Regarding the following question:

> Slanted windows may only be of LVO type 1, 4 and 5. For slanted
> windows (e.g. windows in a roof)
> the Own Height Factor (see figure 4.8), which is the cosine of the
> angle between the window plane
> and a vertical plane, must be given in &-NET-LINks (see page
> 100).With this factor the program will
> calculate the start height and the height of the opening in order to
> get the correct stack pressure
> difference profile. For calculating the area of the opening, the Own
> Height Factor will not be
> taken into account because the flow is assumed to be perpendicular to
> the opening plane. With
> this definition it is even possible to input an opening which is
> horizontal (Own Height Factor = 0).
>
>    That means it is possible to model horizontal windows in COMIS and
> I did my modelling in that way to simulate the atrium and the answer
> from COMIS seems reasonable, how do you think?
>
>>   I am now trying to simulate the air-flow in an office building with
>> an big atrium in the middle, with TRNSYS and COMIS. Because of the
>> high
>> height and the temperature difference within the atrium, I divided
>> the
>> atrium vertically into 3 zones, linking them with big openings as if
>> they are fully linked and my concentrations for temperature will be
>> on the
>> lowest level only.
>>   But I got some strange answers from COMIS when I use the "large
>> opening" air flow model between the zones of atrium.

As I understand it, to model the atrium, you want to draw an imaginary
plane at some height, and then use that plane as a division between two
imaginary sub-zones inside the atrium.

In order to force the software to recognize the imaginary plane, you
want to model those sub-zones as zones, and then connect them via a
window element. You have selected the window element because it has an
input parameter that allows you to tell COMIS that it's horizontal.

It is important to realize that, while the COMIS software will
calculate a result when you model the atrium this way, that result will
not be the "answer" you want. The window model does not include any of
the relevant physics.

The window model, like most COIMIS flow element models, is based on
analogy to an orifice. It calculates the flow through an orifice,
based on the pressure drop across it. The window model has some extra
provisions that allow for two-way flow, based on the "stack effect" on
either side of the window. The idea is that, if air has a different
density on opposite sides of the window, then the pressure drop across
the window will not be constant across its whole height. In some
cases, the pressure difference may change sign, allowing flow in both
directions.

Here are three reasons the window model does not capture the physics
you want:

(1) There is no "orifice" inside the atrium. There is only an
imaginary plane, whose boundary is drawn on smooth walls. If you chase
the orifice model back to its experimental roots, you'll see that the
orifice is defined by a smaller cross-sectional area than the tube
(duct) in which the fluid is flowing. The orifice model that results
from a "Bernoulli" analysis of that physical setup includes the
following factor in its denominator:
(1 - A1/A2).
Thus, when the cross-sectional areas are the same, the engineering
version of the orifice model "blows up" numerically. This is a sign
that it is being taken out of the context in which it applies. You
don't see this "blow-up" in COMIS because the correlations programmed
into the software account for this term indirectly. Nevertheless, it
is still true that the model is being used out of its engineering
context.

(2) The "stack effect" is zero when applied across a horizontal
opening. In COMIS, the stack effect only has meaning when applied
across an orifice with some height component. With a horizontal
opening, the pressure calculated across each face of the opening will
be uniform, since it is unaffected by the pressure change due to
height. Therefore the pressure drop through the opening will be
uniform across the entire opening. Therefore the presumed point of
using a window-- modeling a horizontal opening, in order to allow
two-way flows-- is pointless. You might as well use a plain orifice
model. Of course, I object to a plain orifice as well, see (1) above.

(3) The "stack effect" is not actually the physics that interest you
in an atrium. What makes an atrium hard to model is that, with its
large height, buoyancy effects are important. We are accustomed to
thinking of the "stack effect" as the same as buoyancy effects, but it
is only a part of the story. Buoyancy affects the flows internal to a
space, and the density distribution through the height of a space.
However, the "stack effect" in COMIS does not address these issues.
When you hear about the "stack effect" in COMIS, that really means
that, given the density distribution, the software calculates the
resulting change in pressure with height. That pressure profile only
gets used to calculate the pressure drop through flow paths that
connect two zones. In other words, COMIS has no mechanism for
estimating the density-height distribution and (coupled) internal
flows, which are the effects in an atrium that force people to use
computational fluid dynamics when they really want to resolve what's
going on. All COMIS can do is use an assumed density distribution to
estimate the pressures, and then use those estimated pressure to find
the zone-to-zone flows.

Taking all these objections together, I conclude that your best plan
for getting reasonable answers is as follows: (1) model the atrium as a
single zone; (2) use COMIS to find the zone-to-zone flows; (3) let
TRNSYS determine the heat balance in the atrium; (4) if TRNSYS does not
provide it, use a CFD or correlation model to estimate the resulting
density distribution in the atrium; then (5) enter the new density
distribution in the COMIS input file, then re-run COMIS. Repeat until
you converge to a solution (which probably will take four or five major
iterations at most).

More generally, the approach you've asked about has been attempted by
many people. The core idea is to estimate the flows within a space, by
dividing that space up into lots of sub-zones. By itself, COMIS cannot
give meaningful answers to this type of model. Fundamentally, COMIS is
built around the idea of calculating pressure-drive airflows through
orifices. You can divide up a large space if you want, but there are
no "orifices" there except in your mind. There are only invisible,
imaginary planes, whose physics do not in the least resemble the
engineering model of orifice flow. This is doubtless the reason for
the "strange results" that you mentioned in your original question.

There are some efforts to put some of the physics in there; the
literature calls these "zonal" or "sub-zonal" models. There are also
efforts to combine a multizone airflow tool with a full-fledged CFD
program. However, in your case, the density distribution, rather than
the flow pattern inside the zone, is of primary importance to your
problem domain (natural ventilation). Therefore I think you can avoid
the complexity of trying to estimate those internal flows. Just focus
on the energy balance, and hope the static density distribution is good
enough to give you the answer you want.

-Dave