A systematic approach to describe the air terminal device in

CFD simulation for room air distribution analysis

Y. Huo
a
, F. Haghighat
a,
*, J.S. Zhang
b
, C.Y. Shaw
b
a
Department of Building, Civil and Environmental Engineering, Concordia University, 1445 de Maisooneuve Blvd W, Montreal, Canada
b
Institute for Research in Construction, National Research Council, Canada
Received 22 January 1999; received in revised form 7 April 1999; accepted 12 July 1999
Abstract
Proper specication of air terminal device boundary conditions is the essential element in accurate prediction of the air
distribution in a ventilated room. The conventional method of describing the air terminal device boundary conditions requires
large computing time, and some of the assumptions made are far from reality.
This paper proposes a systematic approach that correctly and simply describes the air terminal device boundary conditions for
CFD simulation. Based on the detailed study of the air terminal device characteristics, the specication of complicated air
terminal device boundary conditions is transferred to the specication of a volume around the diuser. One surface of the
volume will be located inside the jet main region. The boundary conditions of the volume are calculated using the diuser jet
characteristic equations. This method was proved to be easy to use, more ecient, applicable to any type of diusers and, above
all, can correctly predict the airow in a ventilated room.
The predictions of the newly proposed method were compared with the predictions of the conventional model as well as with
the measured data. It is shown that the prediction of the new model is signicantly more accurate than the prediction of the
conventional model. 7 2000 Elsevier Science Ltd. All rights reserved.
Keywords: Diuser; CFD; Air distribution; Indoor air quality
1. Introduction
Recent advances in computational uid dynamics
(CFD) and computer power have provided tools that
can be used to accurately predict some features of air-
ow within ventilated spaces [2]. The CFD method has
been successfully applied for airow analysis in rela-
tively complicated conditions, like non-isothermal,
three-dimensional and with furniture inside the room.
Haghighat et al. [12] numerically studied the eect of
door locations in a partition on contaminant removal
eciency and thermal comfort in a two-zone enclo-
sure. Haghighat et al. [13] also numerically studied the
indoor air quality in a newly painted oce and
assessed the eects of ventilation airow rate and par-
tition layout on the pre-ventilation time required for
the contaminant concentration level to drop to an
acceptable level. Zhang et al. [26] evaluated a numeri-
cal simulation model for predicting room air velocity,
turbulence kinetic energy and temperature. Chen et al.
[7] numerically studied the indoor air quality and ther-
mal comfort in a classroom lled with people and fur-
niture. In a ventilated room, the supply air ow
condition and type of diuser is an essential parameter
aecting the contaminant distribution in the room [14].
Experts have successfully applied commercial and
non-commercial CFD codes to analyze practical pro-
blems in a research environment. Conversely, HVAC
engineers do not generally nd these codes useful in
supporting alternative design assessments. Particularly
Building and Environment 35 (2000) 563576
0360-1323/00/$ - see front matter 7 2000 Elsevier Science Ltd. All rights reserved.
PII: S0360- 1323( 99) 00047- 5
www.elsevier.com/locate/buildenv
* Corresponding author. Tel.: +1-514-848-3192; fax: +1-514-848-
7965.
E-mail address: haghi@cbs-engr.concordia.ca (F. Haghighat).
limiting are the diculties in modeling the complex
geometry of the air terminal device in a room for IAQ
analysis. The geometry of some air terminal device in
a room is extremely complicated and it is hard to
describe the boundary condition of the air terminal
device properly (boundary conditions) in numerical
methods. The inaccurate simplication of the air term-
inal boundary conditions could lead to errors.
The conventional method for describing the air
terminal device (i.e. air supply diuser) boundary con-
ditions is as follows:
1. Consider the supply diuser as a simple free open-
ing.
2. The supply velocity component perpendicular to the
free opening plane is calculated as the ratio of the
supply air ow rate and the area of the free open-
ing, Q
s
/A
c
.
3. The supply velocity component parallel to the free
opening plane is neglected or calculated based on
manufacturer's data.
Although many validations of the conventional
method have been performed, the cases studied used
only diusers with a simple geometry, that is the free
opening area equals to the eective diuser area. This
method may cause errors for diusers that have a
complicated geometry.
The complicated geometry of the supply air diusers
may have a louvered or perforated face, with vanes,
curved surfaces, etc. Fig. 1 shows a typical supply air
diuser [24]. As shown in Fig. 1, the eective supply
air diuser area, A
0
, is smaller than the diuser open-
ing area, A
c
. While the indoor airow in a ventilated
room is simulated using the CFD method, the supply
area is usually assumed as the diuser opening area,
A
c
, instead of the diuser eective area, A
0
. Since A
c
is
bigger than A
0
, some problems can occur while
describing the supply air conditions (diuser boundary
conditions) in CFD simulation. The supply airow
rate, Q
s
, equals the supply air velocity, V
0
, multiplied
by the supply area, A
0
. i.e., Q
s
=V
0

A
0
. When the
supply air ow rate equals to the actual amount, the
supply air velocity calculated based on V
0
=Q
s
/A
c
will
be smaller than in reality. When the supply air velocity
equals to the actual speed, the supply airow rate cal-
culated based on Q
s
=V
0

A
c
will be bigger than in rea-
lity. Both of these two conditions can lead to the
wrong prediction of the air velocity in the whole room.
Kurabuchi et al. [18] studied the airow pattern in a
ventilated room. The complicated geometry diuser
was simplied as an opening using the conventional
method. The measured supply velocity was used in the
simulation. The supply area was considered as the area
of the free opening. The simulation results showed
that the supply air went straight to the opposite wall.
Contrary to the measurement result that showed, the
airow started to turn back in the middle of the room.
The reason for this dierence is that a higher supply
airow rate was used in the simulation than what
occurred in reality.
Heikkinen [15] simulated a complex diuser with 84
round nozzles. When the diuser opening area was
used, the simulation result was far from the measured
data since the eective diuser area was much smaller
than the diuser opening area in this case. He devel-
oped a so-called basic model for such a diuser; the
diuser was modeled as a rectangular slot that has the
same eective ow area as the complex diuser. The
basic model provides reasonably good predictions of
the airow pattern in a room under isothermal con-
ditions. Chen and Moser [5] found that this approach
is not suitable for non-isothermal ow.
The problem of diuser boundary simplication
does not only occur for a diuser whose supply open-
ing is dierent from its eective supply area. The
boundary conditions for square diusers with compli-
cated geometry, widely used in oce buildings, are
also dicult to specify. Furthermore, the connecting
duct to the square diuser could be dierent for the
same diuser in dierent applications. This means that
the same diuser under the same ow rate could have
dierent supply conditions. From diuser manufac-
turer's catalogues, we can nd two diuser's airow
rate and supply faces are exactly the same while the
connecting duct size, eective supply air velocity and
their directions are dierent. The conventional method
denitely cannot handle this kind of diuser.
The swirl diuser is also widely used in oce
buildings. Christianson et al. [9] found that the baf-
e, placed in either the lower or the upper position
for the same diuser, could cause the dierent air-
ow patterns. It also indicated that the same supply
diuser could have dierent supply conditions. At
the same time, the round geometry of the swirl dif-
Fig. 1. A supply diuser with vanes.
Y. Huo et al. / Building and Environment 35 (2000) 563576 564
fuser increases the complexity of boundary condition
specications, especially when simplied rectangular
grids are used.
Several simplied modeling methods have been
developed to describe the complicated supply diu-
ser boundary conditions in recent years.
Nielsen [20,21] proposed a box method (box
located around the diuser), in which the descrip-
tion of the diuser boundary conditions is trans-
ferred to the description of the box boundary
conditions. The boundary conditions at the box sur-
face parallel to the supply-opening surface were
measured data. The boundary conditions at the box
surface perpendicular to the supply opening surface
were considered as @F/@n zero. In this equation, n
is the direction parallel to the studied surface, F is
velocity u, v, kinetic energy k and the dissipation
rate of kinetic energy e etc. Compared with the
measured data good results were obtained, neverthe-
less detailed measurements are needed for each dif-
fuser. Therefore, this method is not very practical.
Nielsen [19] proposed another method called the
velocity prescription method. In this method, the
diuser boundary conditions were specied using
the conventional method; nevertheless, one com-
ponent of the velocity was measured inside the
assumed box area. The velocity component values
inside the box were prescribed as extra boundary
conditions to correct the predicted velocity around
the diuser area. The results were good for the dif-
fuser studied. Applying this method, measurements
for each diuser studied is still needed. The descrip-
tion of boundary conditions for the complicated
supply diuser cannot be avoided.
Chen and Moser [5] proposed a new momentum
method in which the velocity vector was calculated
based on the eective area, not the opening area, to
have the correct velocity description. To keep the
appropriate supply airow rate and to introduce the
same amount of air into the room, the boundary con-
ditions for the continuity equation and the momentum
equations are separately described. Chen et al. [6] and
Jiang et al. [17] simulated diusers with complicated
geometry using the momentum method. The results
showed that this method is applicable for the diusers
studied. Most of the commercial CFD software does
not support the separate description of boundary con-
ditions for continuity and momentum equations. This
method also needs to be validated for more diusers.
Chen and Jiang [8] simulated a two-dimensional dif-
fuser with complex geometry. The diuser was pre-
sented in detail in the simulation. They used the nite
volume method, tried dierent coordinates and grid
systems and demonstrated that it is possible to
describe in detail the boundary of a complex diuser
using state-of-the-art techniques. They also reported
the diculties of this kind of diuser presentation in
CFD simulation and the great demand for computer
capacity.
Emvin and Davidson [10] reviewed the dierent dif-
fuser description methods, and found that the full rep-
resentation of the supply diuser is useful but
expensive and very time consuming. The momentum
method can give qualitative results but cannot prop-
erly represent the entrainment of the supply diuser
and may only work on coarse meshes. The box model
is consistent and supposes to perform as well as the
full representation, but it requires measurements.
2. Model development
This paper begins by providing a brief overview of
the diuser characteristic, classication, and character-
istic equations. It then describes a simple method for
describing the supply air diuser boundary conditions
for CFD simulation using the diuser characteristic jet
equations. Finally, it examines the relative merit of the
newly proposed method compared with the conven-
tional method as well as the measured data.
2.1. Diuser characteristics and air jet classication
In a ventilated room, turbulent air jets distribute the
air supplied into the room through various types of
diusers (e.g., grille-like, ceiling diusers, and perfo-
rated panels). These air jets are the primary factors
aecting room air motion [3].
The supply air diuser's characteristics need to be
carefully studied and summarized before trying to cor-
rectly and practically describe its boundary conditions
in CFD simulation.
Diuser air jets have dierent characteristics when
they are supplied from dierent types of diusers or
under dierent conditions (initial air temperature,
room geometry and size, supply direction, etc.). When
the temperature of the supplying air jet is equal to the
temperature of the air in the room, the jet is called an
isothermal jet. When there is a temperature dierence
between the incoming air jet and the room air, the air
jet is called a non-isothermal jet. When the jet is dis-
charged into a large open space and not inuenced by
walls and ceilings, the jet is called a free jet. Whereas,
when the incoming air jet is attached to a ceiling or
wall, it is called an attached jet. Furthermore, if the
incoming air jet is inuenced by the reverse airow in
the room caused by the jet itself; it is called a conned
jet.
Depending on the types of diusers, diuser air jets
can be classied as follows [3].
Linear jets: formed by slots or rectangular openings
with a large aspect ratio. These jet ows are approxi-
Y. Huo et al. / Building and Environment 35 (2000) 563576 565
mately two-dimensional. Air velocity is symmetric in
the plane at which maximum air velocities are in the
cross section area.
Compact jets: formed by cylindrical tubes, rectangu-
lar or square openings with a small aspect ratio. Com-
pact air jets are three-dimensional and axis-symmetric.
The maximum velocity occurs on the axis.
Radial jets: formed by the ceiling cylindrical or
other air diusers with the air horizontally directed in
all directions.
There are other jets called conical jets, incomplete
radial jets and swirling jets, etc.
2.2. Diuser air jet expansion regions
The full length of an air jet (compact, radial, linear,
or conical), can be divided into four regions in terms
of the maximum or centerline velocity and temperature
dierential at the cross section [3].
Initial region: A short core region. The maximum
velocity (temperature) of the air stream remains practi-
cally unchanged.
Transition region: A short region. The centerline vel-
ocity and temperature are predictable. The prole of
the velocity in this region cannot be normalized. No
predictable velocity and temperature proles can be
obtained in this region.
Main region: A region of fully established turbulent
ow with velocity prole similarity. The velocity and
the temperature can be determined with accuracy from
the characteristic equations. The velocity and the tem-
perature prole can be expressed by a single curve in
terms of dimensionless coordinates. Temperature and
density dierences have little eect on cross-sectional
velocity proles [3].
Terminal region: A region of diuser jet degra-
dation. It is relatively far from the diuser and it will
not be considered in the diuser boundary conditions
of this study.
The initial and transitional regions are small. The
focus of the study in this paper is on the main region.
The characteristic equations of the main region are dis-
cussed next. It is adequate to apply the diuser charac-
teristic equations only in the jet main region for the
newly proposed diuser description method in CFD
simulation. Fig. 2 shows the rst three regions [1].
2.3. Diuser characteristic equations
2.3.1. Velocity decay of isothermal free jets
The centerline velocity decay of linear jets in the
main region can be described by Eq. 1, and the vel-
ocity decay of compact and radial jets can be described
by Eq. 2 [23]:
V
m
V
0
K
1

H
0
x
_
1
Fig. 2. Diuser jet regions.
Y. Huo et al. / Building and Environment 35 (2000) 563576 566
V
m
V
0
K
1

A
0
p
x
2
where, V
m
is the centerline velocity, V
0
is the average
velocity at discharge, x is the distance to the diuser
face on the jet centerline, K
1
is the centerline velocity
decay constant, H
0
is the eective width of the linear
diuser jets, A
0
is the eective area of the diuser,
A
0
=C
d

A
c
, C
d
is the discharge coecient (usually
between 0.65 and 0.90), and A
c
is the diuser free
opening area.
2.3.2. Velocity decay of non-isothermal free jets
When the supply air temperature is dierent from
the room air temperature, the diuser air jet behavior
is inuenced by the thermal buoyancy (due to the
dierence of air density). The centerline velocity of a
non-isothermal jet introduced vertically and the trajec-
tory of a non-isothermal jet introduced horizontally
are inuenced by the Archimedes number [4]:
Ar
gDt
0
L
0
TV
2
0
3
where, g is the acceleration due to gravity, Dt
0
is the
temperature dierence between the supply and return
air, L
0
is the length scale of hydraulic diameter of the
diuser outlet, T is the mean absolute temperature of
the room air, and V
0
is the average velocity at dis-
charge.
2.3.3. Vertically discharged non-isothermal free jets
For vertically discharged linear diuser jets, it was
found that the calculation of velocity decay can be per-
formed using the following equations [23]:
V
m
V
0
K
1

H
0
x
_
K
n
4
K
n

_
12
1:8K
2
K
2
1
Ar
_
x
H
0
_
3=2
_
1=3
5
where, K
2
is the temperature decay constant of the
jets. The `` 2'' sign should be positive when the buoy-
ant force is in the same direction of the initial force,
and negative when they are not in the same direction.
For other vertically discharged non-isothermal free
jets, such as compact and radial, the following
equations can be applied [23]:
V
m
V
0
K
1

A
0
p
x
K
n
6
K
n

_
12
2:5K
2
K
2
1
Ar
_
x

A
0
p
_
2
_
1=3
: 7
2.3.4. Horizontally discharged non-isothermal free jets
For horizontally declined compact free jets, with a
decline angle, a, of not more than 458, the trajectory
of the centerline can be described as [23]:
z
A
0

A
0
p tga2c
K
2
K
1
Ar
_
x

A
0
p
_
3
8
where, z is the distance below the ceiling at which the
maximum velocity in the jet occurs, x is the horizontal
distance to the jet, and c is a coecient determined by
diuser type, size etc. c=0.4720.06.
Zhang et al. [27] and Zhivov [28] studied in detail
the trajectory and characteristics of inclined jets.
Their work made it possible to apply the character-
istic equations to inclined diuser jets.
2.3.5. Velocity prole of the jets in the main region
The velocity at other points within the jet main
region can be calculated by Eq. (9) [3]:
_
r
r
0:5
_
2
3:3 log
V
m
V
9
where, V is the velocity at the point being considered,
V
m
is the centerline velocity in the same cross-sectional
plane, r is the radial distance of the point under con-
sideration from the centerline of the jet, and r
0.5
is the
radial distance in the same cross-sectional plane from
the axis to the point where the velocity is one-half of
the centerline velocity, i.e. V=0.5 V
m
.
2.3.6. Temperature decay of non-isothermal free jets
The temperature decay in the centerline of the jet
was calculated using the following equations [23].
For linear jets, horizontally projected:
T
m
T
r
T
0
T
r
K
2

H
0
x
_
: 10
For linear jets, vertically projected:
T
m
T
r
T
0
T
r
K
2

H
0
x
_
1
K
n
: 11
For compact, radial and conical jets, horizontally pro-
jected:
T
m
T
r
T
0
T
r
K
2

A
0
p
x
: 12
For compact, radial and conical jets, vertically pro-
jected:
T
m
T
r
T
0
T
r
K
2

A
0
p
x
1
K
n
13
Y. Huo et al. / Building and Environment 35 (2000) 563576 567
where, T
0
is the supply temperature at the diuser, T
r
is the return air temperature, T
m
is the temperature
along the centerline of the jet main region, and K
2
is
the temperature decay constant. K
n
is a coecient cal-
culated using Eqs. 5 and 7.
2.3.7. Temperature prole in the jet main region
The relation between the velocity distribution and
temperature distribution in the cross section of non-
isothermal jets is expressed by [3]:
_
r
r
0:5
_
2
4:7 log
T
m
T
r
T T
r
14
where, T is the actual air temperature at the point
being considered, T
m
is the centerline air temperature
in the same cross-sectional plane, and T
r
is the return
air temperature.
2.3.8. Eects of ceilings and walls
Jets discharge parallel to a surface with one edge of
the outlet coinciding with the surface take the form of
one-half of an axial jet discharging from an outlet
twice as large, similar to radial jets from ceiling pla-
ques. Entrainment takes place almost only along the
surface of a half cone, and the maximum velocity
remains close to the surface [3].
Ceiling jets were found to attach to the ceiling and
ow along it due to the ``Coanda'' eect if the initial
jet axis was close to the ceiling [25]. The spread of the
jet in the traversing direction was reduced when the
axis of a long jet was too close to the ceiling and par-
allel to it. The angle of divergence of the jet away
from the wall was slightly less than one-half the angle
of a free conical jet [25].
It was found that if an edge of the nozzle was in
contact with the plane, as long as the axis of the noz-
zle formed an angle less than 408458 with the plane,
the jet would cling to the plane and spread over it.
However, if the edge of the jet was shifted away from
the plane, air entrainment would occur on all sides of
the jet, and the jet did not cling anymore [4]. If the jet
attached to a ceiling or a wall, K
1
became larger than
free jets. The values of K
1
were those of the free jets
multiplied by 1.4 [3]. When the temperature of the
attached air jet is lower than the temperature of the
ambient air, this jet will remain attached to the ceiling
until the downward buoyancy force becomes greater
than the upward static pressure (``Coanda'' force).
2.3.9. Eect of connement
The conned centerline velocity V
mc
and the tem-
perature dierential Dt
mc
caused by the reverse ow in
the room can be modied using the coecient K
c
[11]:
V
mc
V
m
K
c
15
Dt
mc
Dt
m

1
K
c
16
Where V
m
is the centerline velocity calculated from
Eqs. 1, 2, 4 or 6, and Dt
m
is the temperature dierence
between the jet centerline and the occupied region cal-
culated from Eqs. 1012 or 13. The values of K
c
are
available from graphs [11].
2.4. Diuser characteristic equation application
All these above equations are developed for the
main region of the diuser jet. Since the rst two
regions are very small, the beginning of the main
region is still close to the jet surface compared to the
room size. The parameters such as the eective supply
area A
0
and the average air supply velocity u
0
can be
obtained from the diuser's manufacturer product in-
formation data. The velocity and temperature decay
coecient K
1
and K
2
for dierent diusers are given in
ASHRAE Fundamentals. As long as the supply air-
ow rate and the supply air temperature are given, the
velocity and temperature in the jet main region can be
calculated using the diuser jet characteristic
equations.
2.5. The new proposed diuser specication method
The newly proposed diuser boundary conditions
specication method is a jet main region specication
method. It takes advantage of the existing diuser
characteristic equations. The specication of the com-
plicated diuser boundary is transferred to the speci-
cation of the surfaces of a volume around the diuser.
One volume surface is located inside the main region
of the diuser jet [16].
Fig. 3 shows the boundary conditions considered for
the diuser using the new proposed method. A two-
dimensional air jet (linear air jets can be considered
two-dimensional) is applied in this gure. The air is
introduced into the room at an angle a to the vertical
direction. A
1
B
1
is the jet opening. O
1
is the center of
the jet opening and O
1
O
2
is the centerline of the sup-
plying jet. A
1
A
2
and B
1
B
2
are the jet region borders.
B
2
is selected such that it is located at the beginning of
the jet main region. A
2
is chosen to make A
2
B
2
paral-
lel to the ceiling. Points C and D are selected to make
CA
2
DB
1
a rectangular volume.
Point B
2
dictates the height of the volume and
points B
1
and A
2
specify the width of the volume. As
long as A
2
B
2
is located inside the jet main region, the
volume should be selected as small as possible to mini-
mize the inaccuracy caused by the simplication.
For three-dimensional diusers (radial, compact
etc.), the volume can be selected in a similar manner.
The approach is that, part of the volume surfaces need
Y. Huo et al. / Building and Environment 35 (2000) 563576 568
to be located at the beginning of the jet main region
and the other part of the volume surfaces should be
selected to make the volume as small as possible. This
will reduce the inaccuracy of the simplication.
The uniqueness of the volume selection is that the
jet main region concept is introduced and applied. The
surfaces of the volume around the diuser are divided
into two parts, inside the jet main region (A
2
B
2
) and
outside the jet main region (CA
2
, B
2
D and B
1
D), for
the boundary conditions specication. One surface of
the volume (A
2
D) could be partly inside the jet region
and partly outside the jet region (Fig. 3).
2.5.1. Boundary conditions of the selected volume
2.5.1.1. Velocity boundary conditions. The velocity
boundary conditions of the volume surface inside the
jet main region will be calculated using the jet main
region characteristic equations.
For two-dimensional cases, as shown in Fig. 3, the
velocity components, temperature, k and E must be
dened at the boundary. The velocity on the bound-
ary, such as point E on line A
2
B
2
(inside the jet main
region), is obtained by determining the velocity of its
orthogonal projection point, O
E
, on the centerline
using the velocity decay equations (Eqs. 1, 2 or 4). The
average velocity at discharge, V
0
, is obtained based on
the supply conditions and the diuser data from man-
ufacturer's catalogue.
The velocity V
E
, then is calculated using the velocity
prole equations based on the centerline velocity at
point O
E
using Eq. 9. The velocity direction for the
points on A
2
B
2
is considered the same as the velocity
direction of the centerline since the velocity component
perpendicular to the centerline is very small compared
with the velocity component parallel to the centerline
[4].
For the three dimensional cases, the velocity on the
surface located inside the jet main region can be calcu-
lated in a similar manner. When the velocity of a point
on that surface, V, is needed, the velocity of its orthog-
onal projection point on the centerline, V
m
, is calcu-
lated rst using Eqs. 2 or 6. Then the velocity, u, can
be calculated using Eq. 9. The distance, r, should be a
three dimensional distance in Eq. 9.
When considering the boundary conditions of the
other part of the volume surfaces, a zero gradient is
assumed based on the fact that the velocity outside the
jet main region does not have big changes [21]. The
velocity component parallel to the volume surfaces is
assumed as:
@V
@n
0 17
where, n is the direction perpendicular to the volume
surface; and V is the velocity component parallel to
the volume surface. The velocity component perpen-
dicular to the volume surface is determined by using
the continuity equation in the simulation. Fig. 3 shows
the equations for a two dimensional case. The velocity
of a three dimensional case can be similarly derived.
2.5.1.2. Temperature boundary conditions. Like the vel-
ocity boundary conditions, the temperature boundary
conditions are also decided depending on whether the
surface of the volume is located inside or outside the
jet main region.
The jet main region temperature characteristic
equations are used for the temperature boundary con-
ditions of the volume surface inside the jet main
Fig. 3. Boundary conditions in the jet main region.
Y. Huo et al. / Building and Environment 35 (2000) 563576 569
region. The temperature on the diuser centerline is
calculated rst according to the supply air temperature
and the diuser type. When the temperature at a point
on the surface inside the jet main region, T, is needed,
the temperature of that point's orthogonal projection
on the centerline, T
m
, is calculated rst using Eqs. 10
12 or 13, depending on the diuser type. Then the
temperature at the point, T, can be calculated using
the temperature prole (Eq. 14).
Zero temperature gradient is also assumed for the
temperature boundary conditions of the volume sur-
faces outside the jet main region; considering that the
temperature variation between the inside and outside
jet main region is negligible.
@y
@n
0 18
where, y is the temperature dierence, and n is the
direction perpendicular to the volume surface.
2.5.1.3. Boundary conditions for k and E E. The k and E
boundary conditions for the volume surface inside the
jet region can be determined based on Rodi and Spald-
ing's [20] work. For the same type of diusers, the
same dimensionless curve can be obtained in the jet
main region. They also suggested that E could be calcu-
lated using the following equation:
E C
D
k
3=2
L
19
where, C
D
is a constant, equal to 0.09 for a plane jet
and 0.06 for a round jet; L is a length scale. Its value
approximately equals to 0.075 d for a radial jet, 0.052 d
for a plane jet and 0.033 d for a round jet [22]. d is the
distance from the jet centerline to the edge of the jet
region at the location where E is specied.
The boundary conditions, k and E, for the other part
of the volume surfaces are described in a similar man-
ner as the velocity and temperature boundary con-
ditions. They are assumed as:
@k
@n
0,
@E
@n
0 20
where, n is the direction perpendicular to the volume
surface, as shown in Fig. 3.
3. Validation of the proposed method
3.1. Simulation software
EXACT3 [18] was modied for this study. The k-E
two-equation turbulence model is applied, and an
energy equation is included to consider the buoyancy
eect in the code.
EXACT3 uses the following equations to simulate
the air ow pattern:
@u
j
@x
j
0 21
@u
i
@t

@u
i
u
j
@x
j

1
r
@P
@x
i

@
@x
j
_
n n
t

_
@u
j
@x
i

@u
i
@x
j
__
bg
i
y
22
where, P=p +2rk/3
@y
@t

@yu
j
@x
j

@
@x
j
_
_
K
n
t
s
y
_
@y
@x
j
_
hx
j
, t 23
@k
@t

@ku
j
@x
j

@
@x
j
_
_
n
n
t
s
k
_
@k
@x
j
_
n
t
_
@u
j
@x
i

@u
i
@x
j
_
@u
i
@x
j
bg
i
n
t
s
y
@y
@x
i
E 24
@E
@t

@Eu
j
@x
j

@
@x
j
__
n
n
t
s
E
_
@E
@x
j
_

E
k
_
C
1
n
t
_
@u
j
@x
i

@u
i
@x
j
_
@u
i
@x
j
C
2
E C
3
bg
i
n
t
s
y
@y
@x
i
_
25
Eddy viscosity:
n
t
C
D
k
2
E
26
where, x
i
is the Cartesian coordinate, t is the time, u
i
is
the velocity component in the x
i
direction, y is the
temperature dierence, k is the turbulent kinetic
energy, E is the dissipation rate of the turbulent kinetic
energy, p is the static pressure dierence, r is the den-
sity of air, n is the kinematic viscosity of air, n
t
is the
eddy viscosity of air, b is the volumetric coecient of
expansion, g
i
is the gravitational acceleration in the x
i
direction, K is the thermal diusivity, and h is the heat
generation rate.
Empirical constants are:
C
D
0:09, C
1
1:44, C
2
1:92, C
3
C
1
,
s
k
1:0, s
E
1:3 and s
y
0:5:
The density of air r is considered as a constant in
EXACT3 to simplify the calculation process.
Finite dierence numerical method is applied in
Y. Huo et al. / Building and Environment 35 (2000) 563576 570
EXACT3, and the ``Marker and Cell'' (MAC) method
is implemented to solve these equations [18]. A pseudo
time step is used and the simulation reaches a steady
state result when the accumulated time step converges.
EXACT3 can only be used to simulate a rectangular
region. To apply the new diuser description method,
the boundary condition subroutine of the EXACT3
code was modied to assign the boundary conditions
of the volume around the diuser.
3.2. Case studies
Three cases were studied to verify the validity of the
new proposed jet main region specication method for
diuser boundary conditions description in CFD simu-
lation. The rst two cases were designed to validate
whether the new method is applicable and whether the
new method can improve the prediction and conver-
gence. The third case was designed to compare the per-
formance of the new method with the experimental
results and with the prediction of conventional
method.
3.2.1. First case
In the rst case, a simple linear diuser under a two
dimensional isothermal condition was studied. Since
the conventional method provides correct predictions
for simple geometry diusers, such as the one applied
in this case, the prediction of the conventional method
was treated as a correct solution.
The simulation results of the new jet main region
specication method and the conventional method for
this case are shown in Fig. 4(a) and (b). The supply
diuser is located in the middle of the ceiling. The
return air grill is located on the right of the ceiling.
There are no velocity vectors inside the small volume
around the supply diuser in Fig. 4(a) when the newly
proposed method is used. For comparison purposes,
the velocity vectors were removed in the same volume
position when the conventional method was applied in
Fig. 4(b). The results indicate that the airow pattern
predicted using the new method is similar to the air-
ow pattern predicted by the conventional method.
The velocity along several vertical lines was com-
pared for further detailed study. Fig. 5(a) shows the
comparison along a vertical line at the supply position.
Fig. 5(a) also shows the air velocity along the same
line using the diuser characteristic equations. The
gure indicates that there is a good agreement between
the prediction using the conventional method and with
the results obtained using the jet main region
equations. Acceptable results were also observed
between the prediction made by the conventional
model and the new model (jet main region).
The comparison of the velocity along a vertical line
0.75 m from the diuser is shown in Fig. 5(b). The
simulation predictions using the new jet main region
specication method is in good agreement with the
results obtained using the conventional method in the
working area. There are some dierences in the region
close to the ceiling. This is due to the inaccuracy
caused by the simplication of the volume surfaces,
which were not located in the jet main region. The lar-
gest dierence that occurred in the small region close
to the ceiling is less than 20%. The comparison also
showed that the air velocities in the near oor point
has a bigger discrepancy. This happened in a very
small area of the room as shown in Fig. 4. The vel-
ocity dierence at the breathing height (1.52 m away
from the ceiling) of the two methods is very little.
A comparison of the velocity along a vertical line at
the return position is shown in Fig. 5(c). The dier-
ences between these two methods are negligible.
The simulation results of this case indicate that the
new jet main region specication method could be
applied. The results are in good agreement with the
other methods as shown in Fig. 5, and no convergence
diculties were encountered.
Fig. 4. Simulation velocity vectors in the rst case: (a) jet main
region specication method; (b) conventional method.
Y. Huo et al. / Building and Environment 35 (2000) 563576 571
3.2.2. Second case
The second case was similar to the rst one and the
only dierence was that the direction of the supply air
was not perpendicular to the ceiling but at an angle of
608. In such a condition, there is diculty in selecting
the volume around the diuser as well as setting the
appropriate boundary conditions for the selected
volume. If such a situation can be simulated correctly,
we will know how to simulate supply diusers, which
supply air neither horizontally or vertically. This case
was designed to examine whether the simplications
under this critical condition would cause any pro-
blems. In this case, two and a half edges of the volume
were located outside the jet main region. Simplication
of the boundary conditions had to be applied to all
these edges. Furthermore, the simplications were
applied to more than one direction. The boundary
conditions of two edges perpendicular to the ceiling
and one edge parallel to the ceiling all needed to be
simplied. In the simplied part of the volume bound-
aries, the air may move out the volume in one position
and move into the volume in another position (see
Fig. 3). In short, many more simplications were
required for this case.
The velocity vector distribution of the simulation
using the new jet main region specication method is
Fig. 5. Detailed velocity comparison of the rst case: (a) air velocity in a vertical line (supply position); (b) air velocity in a vertical line (0.75 m
from the supply); (c) air velocity in a vertical line (return position).
Y. Huo et al. / Building and Environment 35 (2000) 563576 572
shown in Fig. 6(a). It indicates that the airow pattern
inside the room is reasonable. No large discrepancies
can be seen directly. No convergence problem and
other diculties were encountered in the simulation
when the jet main region specication method was
applied for such a case.
The simulation results using the conventional
method are shown in Fig. 6(b). It can been seen from
Fig. 6(a) and (b) that the new proposed jet main
region specication method provided a reasonable air-
ow pattern prediction.
Although no measurement data strongly support the
results in this case, the results still indicate that the
simplication at the three edges of the volume for the
new proposed jet main region specication method
was acceptable. The new proposed method could, at
least, provide the correct air ow pattern inside the
room.
3.2.3. Third case
Data from a detailed measurement which was car-
ried on a test chamber [9] was used to study the val-
idity of the new diuser description method. This case
was also designed to compare the model predictions
made by the new method to those made by the
measurement and the conventional method, as well as
with the measured data. The diuser used in this study
was a square diuser and the simulation was per-
formed for a three-dimensional and non-isothermal
situation.
The geometry of the test room and the square diu-
ser layout are shown in Fig. 7(a) and (b), respectively.
There were 15 21 16 measurement points in the
room to study the velocity and temperature distri-
butions.
A 16 26 22 grid was dened in the simulation
while the jet main region specication method was
applied for this case. The four surfaces of the volume
perpendicular to the ceiling around the diuser were
located in the jet main region. The boundary con-
ditions on these four surfaces were calculated using jet
main region characteristic-equations. The boundary
condition on the surface of the volume parallel to the
ceiling was simplied using zero gradients (Eqs. 17, 18
and 20) for velocity, temperature, k and E specication.
The simulation took about 12 h on an IBM system/
6000 workstation. The velocity vector and contour
along a center vertical plane using the new method are
Fig. 7. Test room layout of the third case: (a) geometry of test room;
(b) location of the square diuser.
Fig. 6. Simulation velocity vectors for the second case: (a) jet main
region specication method; (b) conventional method.
Y. Huo et al. / Building and Environment 35 (2000) 563576 573
shown in Figs. 8(a) and 9(a), respectively, and the
measured velocity contour in the same plane is shown
in Fig. 10.
In the case of conventional method, the following
boundary conditions were used:
u
0

Q
s
A
c
v
0
, w
0

V
2
0
u
2
0
_
27
where, u
0
, v
0
and w
0
are the supply air velocity com-
ponent in the x, y and z direction, respectively. A
c
is
the area of the diuser opening, Q
s
is the supply air
ow rate, and V
0
is the supply air velocity at ow rate
Q
s
, provided by the diuser manufacturer's manual.
The prediction results for the 16 26 22 grid
using the conventional method were very poor com-
pared with the measured data. Some diculties with
convergence were also observed in this case. Therefore,
the total number of grids for the conventional method
were increased to (16 30 26). A half time step as
applied in the jet main region specication method,
was applied to obtain the converged results. At the
end, the simulation time under this condition was ve
times longer than that using the newly proposed jet
main region specication method.
The velocity vector in the same vertical plane for the
simulation results using the conventional method is
shown in Fig. 8(b). The velocity contour at this plane
is shown in Fig. 9(b).
The simulation results obtained by the jet main
region specication method (Fig. 9(a)) yield a similar
airow pattern as the measurement data (Fig. 10). The
ow pattern from the conventional method did not
match well with the measurement data.
A comparison of the velocity at the breathing height
along the middle vertical plane for both diuser speci-
cation methods and the measured data are shown in
Fig. 11. It shows that the prediction made by the
newly proposed jet main region specication method is
in good agreement with the measured data which also
indicates that the assumptions made regarding the
boundary conditions are reasonable. Poor agreement
was observed between the prediction made by the con-
ventional model and the measured data.
When comparing with the contours, Figs. 9(a) and
Fig. 8. Simulation velocity vectors for the third case: (a) jet main
region specication method; (b) conventional method in the middle
vertical plane.
Fig. 9. Simulation velocity contour in the middle vertical plane: (a)
jet main region specication method; (b) conventional method.
Y. Huo et al. / Building and Environment 35 (2000) 563576 574
10, air velocity dierences between the newly proposed
method and the measured data exist if the entire room
is considered. The air velocity in the jet main region,
calculated using the jet characteristic equations was a
little higher than the measured data. This might cause
some inaccuracy. It could be improved by developing
more accurate jet characteristic equations. In spite of
this, the predicted results using the new jet main region
specication method was much better than the conven-
tional method when compared to the measured data.
4. Conclusions
It can be concluded from the case studies that, the
newly proposed jet main region specication method
has proved to be an applicable approach and a more
accurate way to study the airow pattern in a venti-
lated room.
The jet main region specication method is devel-
oped based on the box method. However, they are not
the same. The main dierences are as follows.
1. The use of the jet main region concept: the jet main
region specication method applies the analytical
data only to the volume boundaries inside the jet
main region. The volume boundary could be fully
or partially located inside the jet main region. In
Nielson's box method [19], the boundary conditions
of the box use fully analytical data or fully
measured data.
2. The selection of the box: the jet main region speci-
cation method accurately describes how to select a
volume around the diuser. The volume boundary
is accurately set after the diuser is located and its
jet main region is calculated. The original box
method did not specify how to improve the accu-
racy by selecting the box. Recently, Nielson [20]
suggested a procedure for selecting the box, result-
ing in a suggested range of possible selection rather
than a unique selection.
3. The tting for dierent kinds of diusers and room
situations: the jet main region specication method
can be applied conveniently to dierent kind of dif-
fusers and room situations. For example, when the
air jet has an angle with the diuser, the volume
boundary can be partially located inside the jet
main region. Such a situation will be a problem in
the box method. There is no analytical data avail-
able on any side of the box.
The advantages of the newly proposed method are as
follows.
1. The user can avoid describing the complicated diu-
ser geometry by using data from manufacturer's cat-
alogs.
2. It provides an accurate prediction of the air ow
patterns in a ventilated room.
3. The method is applicable for dierent diusers.
4. It reduces the simulation time appreciably since a
lower grid density is required.
It should also be noted that, when the new jet main
region specication method is applied, it is very im-
portant to nd the suitable diuser characteristic
equations. Calculation is required for the jet main
region boundary conditions specication.
Although the new method is not a perfect one, it
leads to a much better airow pattern prediction in
air-conditioned rooms than the conventional method,
while a complicated geometry supply air diuser is
presented. This method can be enhanced with the
development of more accurate diuser characteristic
Fig. 10. Measurement velocity contour in the middle vertical plane.
Fig. 11. Detailed velocity comparison of the third case.
Y. Huo et al. / Building and Environment 35 (2000) 563576 575
equations and the development of the CFD simulation
software.
Acknowledgements
Dr Zhenhai Li of the University of Illinois kindly
provided the measured data. His support is highly ap-
preciated
References
[1] Abramovich GN. The theory of turbulent jets. Moscow:
Fizmatgiz, 1960.
[2] Albright LD. Research needs. In: Christianson LL, editor.
Building system: room air and air contaminant distribution.
USA: ASHRAE, 1989. p. 810.
[3] ASHRAE Fundamentals. Space air diusion. USA: ASHRAE,
1993 Chapter 31.
[4] Baturin VV. Fundamentals of industrial ventilation, 3rd ed.
New York: Pergamon Press, 1972.
[5] Chen Q, Moser A. Simulation of a multiple nozzle diuser. In:
Proceedings of the Twelfth AIVC Conference, Ottawa, Canada,
1991a. Vol. 2. p. 114.
[6] Chen Q, Suter P, Moser A. A database for assessing indoor air
ow, air quality and draught risk. ASHRAE Transactions
1991;97(2):15063.
[7] Chen Q, Jiang Z. Evaluation of air supply method in a class-
room with a low ventilation rate. In: Proceedings of Indoor Air
Quality, Ventilation and Energy Conservation Conference,
Montreal, Canada, 1992. p. 4129.
[8] Chen Q, Jiang Z. Simulation of a complex air diuser with
CFD technique. In: Fifth International Conference on Air
Distribution in Rooms, RoomVent'96, Yokohama, Japan, 1996.
Vol. 1. p. 22734.
[9] Christianson LL et al. Low temperature air distribution: jet of
low temperature air, Final Report for ASHRAE Project ]]705,
ASHRAE, USA, 1994.
[10] Emvin P, Davidson L. A numerical comparison of three inlet
approximations of the diuser in Case E1 Annex20. In: Fifth
International Conference on Air Distribution in Rooms,
RoomVent'96, Yokohama, Japan, 1996. Vol. 1. p. 21926.
[11] Grimitlyn M, Pozin G. Determination of parameters for the jet
in a conned space following blocked or through ow pattern.
In: Proceedings of Research Institutes for Labor Protection
VTsSPS 91, Moscow: Prozdat, 1973.
[12] Haghighat F, Jiang Z, Wang JCY. A CFD analysis of venti-
lation eectiveness in a partitioned room. Indoor Air
1991;1(3):60615.
[13] Haghighat F, Jiang J, Zhang Y. The impact of ventilation rate
and partition layout on the VOC emission rate: time-dependent
contaminant removal. Indoor Air 1994;4:27683.
[14] Haghighat F, Huo Y, Zhang J, Shaw CY. The inuence of
oce furniture, workstation layouts, diuser types and location
on indoor air quality and thermal comfort at workstations.
Indoor Air 1996;6:188203.
[15] Heikkinen J. Modeling of a supply air terminal for room air
ow simulation. In: Twelfth AIVC Conference, Ottawa,
Canada, 1991. Vol. 3. p. 21330.
[16] Huo Y, Zhang JS, Shaw CY, Haghighat F. A new method to
specify the diuser boundary conditions in CFD simulation. In:
Fifth International Conference on Air Distribution in Rooms,
RoomVent'96, Yokohama, Japan, 1996. Vol. 2. p. 23240.
[17] Jiang Z, Chen Q, Moser A. Comparison of displacement and
mixing diusers. Indoor Air 1992;2(3):16879.
[18] Kurabuchi T, Fang JB, Rechard AG. A numerical method for
calculating indoor air ows using a turbulence model, US
Department of Commerce, NISTIR 89-4211, 1989.
[19] Nielsen PV. Description of supply openings in numerical models
for room air distribution. ASHRAE Transaction
1992;98(1):96371.
[20] Nielsen PV. The box Method a Practical procedure for intro-
duction of an air terminal device in CFD calculation, Institute
for Bygningsteknik, Aaborg University, Denmark, 1997.
[21] Nielsen PV, Restivo A, Whitelaw JH. The velocity character-
istics of ventilated rooms. Journal of Fluid Engineering
1978;100:2918.
[22] Rodi W, Spalding DB. A two-parameter model of turbulence
and its application to free jets. In: Numerical prediction of ow,
heat transfer, turbulence, and combustion: selected works of
Professor D. Brian Spalding. Oxford; New York: Pergamon
Press, 1983. p. 2232.
[23] Shepelev I. Air supply ventilation jets and air fountains. In:
Proceedings of the Academy of Construction and Architecture
of the USSR 4, Moscow, 1961.
[24] Skovgaard M, Nielsen PV. Modeling complex inlet geometry in
CFD applied to air ow in ventilated rooms. In: Proceedings
of the Twelfth AIVC Conference, Ottawa, Canada, 1991. Vol.
3. p. 183200.
[25] Tuve GL. Air velocity in ventilating jets, ASHRAE researches
report No. 1467, ASHRAE Transaction Vol. 59, p. 261, 1953.
[26] Zhang JS, Christianson LL, Wu GJ, Zhang RH. An experimen-
tal evaluation of a numerical simulation model for prediction
room air motion. In: Proceedings of Indoor Air Quality,
Ventilation and Energy Conservation Conference, Montreal,
Canada, 1992. p. 36071.
[27] Zhang G, Strom JS, Morsing S. Jet drop models for control of
cold air jet trajectories in ventilated livestock buildings. In:
Fifth International Conference on Air Distribution in Rooms,
RoomVent'96, Yokohama, Japan, 1996. Vol. 1. p. 30516.
[28] Zhivov A. Theory and practice of air distribution with inclined
jets. ASHRAE Transaction 1993;99(1):11529.
Y. Huo et al. / Building and Environment 35 (2000) 563576 576