Building and Environment 36 (2001) 883–889

www.elsevier.com/locate/buildenv

The e"ect of location of a convective heat source on displacement

ventilation: CFD study
Hee-Jin Park ∗ , Dale Holland
8200 Normandale Blvd. Suite 500, Advanced Technologies Group, Dunham Associates, Inc., Minneapolis, MN 55437-1075, USA

Abstract

Two-dimensional computational simulations are performed to examine the e"ect of vertical location of a convective heat source on
thermal displacement ventilation systems. In this study, a heat source is modeled with seven di"erent heights from the 2oor (0.5, 0.75,
1.0, 1.25, 1.5, 1.75, 2:0 m) in a displacement ventilation environment. The 2ow and temperature 5elds in thermal displacement ventilation
systems vary depending on the location of the heat source. As the heat source rises, the convective heat gain from the heat source to an
occupied zone becomes less signi5cant. This e"ect changes the temperature 5eld and results in the reduction of the cooling load in the
c 2001 Elsevier Science Ltd. All
occupied zone. The strati5cation level is also a"ected by the heat source location at a given 2ow rate. 
rights reserved.

Keywords: Air 2ow; CFD; Displacement ventilation; Strati5cation; Temperature distribution; Plume; Convection

1. Introduction air through the strati5cation. This leads to the idea that en-
ergy saving as well as good indoor air quality can be ef-
A displacement ventilation system discharges air at low 5ciently controlled by the use of displacement ventilation.
velocity near the 2oor. The supplied cool, clean air spreads Many investigators have reported the advantages of dis-
and forms a pool of conditioned air over the 2oor. When placement ventilation theoretically and experimentally for
this air meets a heat source, a convective plume is gen- various HVAC applications [1,2]. It was also reported that
erated because of the temperature di"erence and resultant for 100; 000 ft2 oEce, the cooling load was reduced by 25
buoyant force. This plume acts as a channel through which –30% using displacement ventilation. Consequently, dis-
warmed and polluted air goes upward up to a ceiling area placement ventilation reduced the supply air volumetric 2ow
where it exits through the exhaust. Due to entrainment by rate to 70% of what is required in conventional mixing ven-
the surrounding air, the volumetric 2ow rate of the plume tilation in the same situation [3].
gets larger as the plume rises. When the 2ow rate of the Attention should be paid to vertical temperature distribu-
plume is equal to that of the supply of air, thermal and con- tion in the displacement ventilation. As the plume ascends,
tamination boundary levels form by which the upper level hot air in the plume warms surrounding air by convection.
(warm and polluted) and the lower level (cool and clean) Due to low entrainment, very stable strati5cation around the
are distinguished. plume forms. This results in a temperature gradient in a con-
This strati5cation is one of the most bene5cial factors ditioned space. Since this temperature gradient is an impor-
in the thermal displacement ventilation over conventional tant factor for comfort in displacement ventilation system
mixing type ventilation since, in the displacement ventila- design, the e"ect of the heat source location on this needs to
tion systems, only a portion of total loads considered in the be investigated. A temperature 5eld is also strongly related
mixing ventilation systems are satis5ed. More importantly, to the cooling load calculation in the displacement ventila-
the displacement ventilation improves indoor air quality in tion [4].
the lower level by separating contaminated air from clean There are two mechanisms through which a heat source
∗ Corresponding author. Tel.: +1-952-820-1437; fax: +1-952-
transmits energy; convection and radiation (Fig. 1). The con-
vection portion of total emitted energy, which is initiated by
820-2760.
E-mail addresses: [email protected] (H.-J. Park), the 2ow 5eld generated around the heat source, is warming
[email protected] (D. Holland). the surrounding air directly. In this case, most convective

0360-1323/01/$ - see front matter
c 2001 Elsevier Science Ltd. All rights reserved.
PII: S 0 3 6 0 - 1 3 2 3 ( 0 1 ) 0 0 0 1 4 - 2
884 H.-J. Park, D. Holland / Building and Environment 36 (2001) 883–889

Nnomenclature

English symbols up mean component of a velocity at yp

Ar Archimedes number u shear velocity (frequently termed wall fric-
B model constant tion velocity)
C1 model constant V characteristic velocity
C2 model constant xi ; xj coordinate variable
C3 model constant yp distance of the adjacent cell center from solid
C model constant surface
Cp constant pressure speci5c heat of air
d characteristic length scale Greek symbols
F buoyant force  coeEcient of thermal expansion
g gravitational acceleration  model constant
k thermal conductivity ij Kronecker delta function
ke turbulent kinetic energy k model constant
kp turbulent kinetic energy at yp  turbulent dissipation
kt turbulent conductivity p turbulent dissipation at yp
p time mean pressure  Karman constant
Pr Prandtl number  viscosity
Prt turbulent Prandtl number t turbulent viscosity
Q cooling load  kinematic viscosity, =
t time  air density
Ht time interval 0 reference density
T local mean temperature and characteristic tem- w wall shear stress
perature ! time mean heat dissipation
T0 reference temperature
HT temperature di"erence Subscripts
u mean component of a velocity i; j coordinate index
u
2uctuation component of a velocity oz occupied zone

Fig. 1. Energy exchange between a heat source and surrounding.

Fig. 2. Cooling load collected by occupied zone.

energy out of the heat source goes to the upper region di-
rectly following strong plume. And a portion of convective portion of heat from all heat sources to the occupied zone and
energy follows the 2ow 5eld within the occupied zone. Q2, a secondary convective portion from warmed or cooled
The radiation portion of total emitted energy out of the surfaces in the lower level which results from direct radiation
heat source, which radiates in all directions, is directly reach- heat exchange between heat sources and those surfaces. Note
ing the colder or hotter surfaces and causes the surface tem- that Q3 is another secondary convection that is coming from
perature to be changed as a result. Then some of the sur- hot air in the upper level and ceiling surface. The important
faces of which temperature is beyond the air temperature fact here is that each portion of heat gains is changing with
start re-radiating energy into the space in the form of sec- the heat source location since radiation and convection heat
ondary convection. exchanges vary. For example, Q1 changes due to the 2ow
The cooling load (Qoz ) in the occupied zone (Fig. 2) con- 5eld generated for speci5c source location and Q2 due to
sists of two heat gains; primary and secondary convection. geometry con5guration and Q3 due to both e"ects [5 –7].
In Fig. 2, Q1 is designated as a primary convective Therefore, it is necessary to understand the e"ect of the
 H.-J. Park, D. Holland / Building and Environment 36 (2001) 883–889 885

equation which is derived from the continuity equation. This

process updates new pressure 5eld and face mass 2ow rate.
Finally, other scalars are sought. Once every 5eld variable
is calculated, these values are compared to those obtained
at the previous iteration. If the di"erences between them
are smaller than the speci5ed criteria, this set of data is
considered as a 5nal solution set. Otherwise, these data be-
come initial data for the next iteration. The details of this
Fig. 3. Geometry under consideration.
algorithm are well described in Patankar [10].
Various vertical locations of the heat source including
location of the heat source to estimate cooling load e"ect- 0.50, 0.75, 1.00, 1.25, 1.50, 1.75, and 2:00 m are simulated
ively for the displacement ventilation. to see the e"ect of vertical locations of the heat source.
Despite numerous investigations on the e"ect of a supply The criteria of termination residuals for this simulation
2ow rate and a supply air temperature on the displacement are 0.5% of overall mass, heat 2ux, and momentum 2ux
ventilation [6 –8], there are few reports [9] regarding the for pressure, temperature, and velocity, respectively. Less
e"ect of the heat source locations on the displacement venti- than 500 iterations are taken for a complete convergence
lation systems. In this study, the e"ects of a primary convec- for each case.
tive heat gain as the vertical location of the source changes
on displacement ventilation systems are investigated by 2.1. CFD formulation
using computational 2uid dynamics (CFD).
It is assumed that 2ow is steady, turbulent, Newtonian
and incompressible with constant physical properties. The
2. CFD approach mean (time-averaged) continuity, momentum and energy
conservation equations are
Fig. 3 shows the two-dimensional geometry (9:0 m × @
3:0 m) under consideration. A heat source (0:5 m × 0:5 m) uj = 0; (1)
@xj

located at the center of the geometry produces a total of @ @p @ @ui @uj
1000 W. Its vertical location changes from 0.5 to 2:0 m  ui uj = − +  +
@xj @xi @xj @xj @xi
above the 2oor. Supply air enters from a wall-mounted,
low-velocity di"user with mass 2ux of 0:17 kg=s and tem- @
◦
perature of 18 C. A return air outlet is installed near a ceiling − u
u
 + Fi ; (2)
@xj
i j 
in the opposite wall side. The size of the modeled supply air @ @ @T @
di"user and outlet is 1.2 and 0:3 m in height, respectively. Cp Tuj = −k + Cp uj
T
 + !; (3)
@xj @xj @xj @xj
The 5nite volume method [10] is used to solve the
time-averaged Navier–Stokes equations with a non-uniform where ui , T , and p are the mean components, and uj
, T
,
grid network. The 5nite volume method converts by dis- and p
(not seen in the equation since it is averaged out) are
cretization governing partial di"erential equations to a set 2uctuation components of an instantaneous velocity, tem-
of algebraic equations that are solvable with a digital com- perature and pressure, respectively. The mean and 2uctu-
puter. The line-by-line iterative method is used to solve ation quantities are de5ned as (here take a velocity as an
the set of algebraic equations. A conventional turbulent example)
 t+Ht
k– model [11] with a universal logarithmic variation of 1
velocity near any solid surface is adopted. For the heat 2ux ui = Ui dt;
Ht t
and mean temperature pro5les near the solid surface, the
ui
= Ui − ui ;
analogy for eddy viscosity and eddy conductivity with tur-
bulent Prandtl number is used. In this study, the Boussinesq where Ui is an instantaneous quantity. The derivation of
model, that assumes that density is only a function of tem- Eqs. (2) and (3) from instantaneous Navier–Stokes equa-
perature through the computational domain, is used for the tions is well described in Tennekes and Lumley [12].
buoyant force term in the momentum equation. This model The angled bracket is used to represent the time-averaged
uses constant reference density with temperature variation quantity of a product of two 2uctuation quantities. The
instead of calculating local density using ideal gas law. terms, ui
uj
, Cp uj
T
 in Eqs. (2) and (3) are de-
The semi-implicit method for pressure-linked equations rived during the time averaging process with instantaneous
(SIMPLE) is used to reach a convergent solution set. Based Navier–Stokes equations. Physically, these terms represent
on a set of initialized data, the 2uid properties are updated. additional stress (Reynolds stress) to a 2uid element and
With these temporal 2uid properties, momentum equations additional heat 2ux due to turbulent phenomena. ! is a
are solved with pre-described pressure 5eld for velocities. mean dissipation term which is not considered in this study
These new velocities should satisfy the pressure-correction due to low velocity scale.
886 H.-J. Park, D. Holland / Building and Environment 36 (2001) 883–889

Fi is the ith component of buoyant force due to tempera- The third term in the right hand side of the Eq. (10) in-
ture di"erence which is expressed in dicates the in2uence of buoyancy e"ect on turbulent dis-
−gi ( − 0 ): (4) sipation. However, the buoyancy e"ect on turbulent dissi-
pation is not well understood. In this study, therefore, this
If we apply the Boussinesq approximation, Eq. (4) for buoy- term is neglected due to its uncertainty of empirical constant
ancy becomes C3 . The existing constants are employed as in Launder
0 gi (T − T0 ): (5) et al. [14].
 is the coeEcient of thermal expansion de5ned as C1 = 1:44; C2 = 1:92; $k = 1:0; $ = 1:3:


1 @
=− :
 @T p 2.3. Near wall region and boundary conditions
Since HT 1 in this study, the Boussinesq approximation
In the near wall region, wall functions are expressed as
is appropriate for this simulation.
The Reynolds stress in Eq. (2) is related to the mean strain up 1  u yp 
= ln + B: (12)
5eld through the Boussinesq hypothesis: u  




2 @ui @uj Here, up is mean velocity at yp ;  is Karman’s constant and
ui uj  = ke ij − t + : (6) is assumed to be 0.41. And the universal constant B is 5.0 in
3 @xj @xi
this study [15]. u is a shear velocity (frequently termed as
ke , turbulence kinetic energy de5ned as
wall friction velocity) produced by friction along the wall
1
ke = (ui
ui
) surface, and de5ned as
2
1=2
and the turbulent heat 2ux term in Eq. (3) can be expressed w
u = :
by Boussinesq analogy with turbulent viscosity 




@Ti @Tj A no 2ow condition is imposed on all 5xed wall surfaces.
Cp uj T  = kt + : (7)
@xj @xi For the calculation of the turbulent kinetic energy near 5xed
The t and kt in Eqs. (5) and (6) are the properties of 2ow walls, the wall boundary layer is assumed to be in an equi-
not of 2uid. These two turbulent properties are connected librium state, that is, the production of turbulence is equal
through the turbulent Prandtl number that is de5ned as to dissipation in the wall region. The kp , turbulence energy
Cp t at yp , near the wall can be expressed as
Prt ≡ : (8) u2
kt kp =   :
◦
Since Pr of air at 20 C is around 0.72, Prt in this study C
it is assumed as 0.9 [13]. The p , turbulent dissipation at yp , near the wall can be
written as
2.2. Conventional k– model
C3=4 kp3=2
p = :
yp
The distribution of the turbulent viscosity is obtained from
k2 Adiabatic condition is imposed on walls, which means that
t = C e (9) there is no heat exchange between the computational do-

where C = 0:09 [14]. main and the wall. For the turbulent conductivity and mean
The kinetic energy of turbulence ke its dissipation rate  temperature pro5les, Eqs. (3), (7), and (8) are used.
are obtained by solving two additional transport equations,
@ @ t @ke 3. Results and discussions
uj ke =
@xj @xj $k @xj

 Typical velocity 5elds are shown in the Fig. 4 (source
@uj t @T
+ −ui
uj
 + gi −  ; (10) height of 0.5 and 1:5 m). A plume is generated by the heat
@xi Prt @xi
source. As the plume goes upward, the volumetric 2ow rate
of the plume is increased by entrainment of surrounding
@ @ t @
uj  = air. Finally, rising air impinges upon the ceiling, which is
@xj @xj $ @xj directed to the wall and circulation 2ow is formed in the

 upper region. It is observed that while hot polluted air is


@ui t @T 
+ C1 −ui uj  + C3 gi circulated in the upper region, another large circulation 2ow
@xj Prt @xi ke
is also created in the lower region. These two 2ow regions
2 yield strati5cation (momentum based separation) between
− C2  : (11) the upper and lower regions since each circulation contains
ke
 H.-J. Park, D. Holland / Building and Environment 36 (2001) 883–889 887

Fig. 5. Strati5cation levels as heat source rises.

Fig. 4. Velocity vector plots for a heat source at (a) 0:5 m and (b) 1:5 m.

properties to its region such that two 2ow regions have little
momentum interaction. Therefore, the strength and size of
those circulation 2ows are main factors in characterizing the
strati5cation level. Another interesting observation is that
while gravity 2ow exists around the di"user when the heat
source is at 0:5 m, it almost becomes an isothermal jet with
the heat source at 1:5 m. The change of 2ow 5eld around
the di"user will be discussed in detail later in this section.
It is observed in Fig. 4 that as the location of a heat
Fig. 6. Velocity distribution of the 2ow region o" the di"user by 1 m.
source changes (the cases of 0.25, 0.75, 1.25, 1.75, 2:0 m are
not shown here), the characteristics of the circulation 2ows
formed in the upper region vary. This change of 2ow 5elds is increasing as the plume rises, the higher location of the
results in di"erent levels of strati5cation. While a larger heat source yields a higher strati5cation level. The variation
circulation 2ow is generated in the upper region when the of strati5cation level with source locations is shown in Fig.
source is at a low location ((a) in Fig. 4), a smaller one is 5. It indicates that vertical location of the convective heat
formed in the upper region when the source is at a higher source alters the strati5cation level linearly over the heights
position ((b) in Fig. 4). This is related to plume strength. (0.5 –1:75 m). When heat source is located at 2:0 m (not
2
If a plume pressure, plume Vplume , which is built up as a seen here), no distinct strati5cation level is observed. This
static pressure in the upper region becomes strong enough is because the rising plume is directed into the wall before
to overcome buoyant force, H(g)lower zone air , then hot air it develops to a level where its 2ow rate equals to that of
in the upper region pushes room air downward until it loses supply air. In this case it can also be observed that air from
its momentum. When it loses vertical momentum, it 2ows the lower region enters the upper region and mixes directly
horizontally making a circulation in the upper region. It is with the air in the upper region.
deduced that when the source is at low location and the The source height not only changes the strati5cation level
plume develops fully, a large circulation is created yielding but also the 2ow pattern of the di"user 2ow. Fig. 6 shows an
a lower strati5cation level. The reverse is applied to the case e"ect of the heat source height on the 2ow around the dif-
of higher locations of the heat source where a plume jet does fuser. It has been known that the 2ow out of a low-velocity
not develop enough to build strong jet pressure. In this case di"user has characteristics of gravity current rather than that
momentum based strati5cation is formed at a higher level. It of a jet [17]. One of the gravity current behaviors is the
should be noted that no strati5cation level would be created de2ection of direction and spreading over an area (cascad-
with jet pressure that is strong enough to push buoyant air to ing) due to a temperature di"erence. The result shows that
the 2oor level. In this case, only one large circulation 2ow when a heat source is at heights equal to or below the ver-
would be observed in whole conditioned space [16]. tical location of a di"user (in this case under 1:3 m), the
Dotted lines in Fig. 4 show 2ow rate based strati5cation source location a"ects the behavior of gravity current sig-
levels where the volumetric 2ow rate of the plume is equal to ni5cantly. In the case of source above di"user height, how-
that of supply air. Since the volumetric 2ow rate of the plume ever, the e"ect of the source height is less consequential.
888 H.-J. Park, D. Holland / Building and Environment 36 (2001) 883–889

Fig. 7. Vertical temperature gradient according to heat source location at

x = 3 m from the di"user.

It is observed that for a source height above 1:5 m the cas-

cading e"ect of the gravity current disappears and shows
same velocity pro5les (like a isothermal 2ow) even though
source location changes furthermore. Low velocities below
0:2 m are attributed to 2oor surface friction. The Archimedes
number (Ar), a non-dimensional number that governs the
non-isothermal gravity current, is de5ned as
Buoyance Force g HTd
Ar = = ; (13)
Inertia Force TV 2
where g is gravitational acceleration, T is characteristic tem-
perature, HT is the temperature di"erence between supply
air and characteristic temperature, V is supply air velocity
at the face of di"user, and d is characteristic length which
is usually taken by height of the di"user. The higher the Ar
number, the more the gravity current is de2ected.
Fig. 7 shows how the vertical temperature distribution
changes depending on the various heat source heights. It Fig. 8. Distributions of temperature due to di"erent vertical locations of
the heat source.
is shown that temperature does not vary linearly over the
room height and it can be divided into three parts; lower,
middle, and upper parts. The lower part is designated as the air from top to bottom. It is shown that the region below the
region from the 2oor up to the heat source height while the source height maintains supply air temperature while in the
upper part is the region in which temperature is maintained upper region the temperature gradients changed with the
uniformly in the upper level. While the height of the lower source locations. This is because most of convective heat
part is changing with the source height, the upper part is from the heat source which is carried by the plume and is
well 5xed for various source locations. While temperature transported and accumulated at the upper region. It is also
remains constant in the lower and the upper parts, (except interesting to note that temperature distribution in the upper
the 2:0 m case) temperature in the middle part changes lin- region shows a similar trend as heat source rises. This leads
early. It is interesting to note that when the location of the to an idea that temperature distribution in the upper level is
source gets higher, a larger temperature gradient is created somewhat independent of the location of heat source unlike
(compared to lower source location) in the region above the the lower level.
heat source. Fig. 9 shows how the average temperature changes in the
Fig. 8 shows temperature distributions with source upper and lower regions. It should be noted that the average
locations (5gures of 0.75, 1.25, 1:75 m are not shown). This temperature of the lower region is taken either in a region,
data shows that the heat source only a"ects the temperature 0.2–1:8 m or 0.2–2 m regardless of the strati5cation level. It
5eld from its own level up. In other words, the heat source reveals that as the source ascends, the average temperature of
◦
has no e"ect on the temperature 5eld below its location the upper region is gradually increasing (about 1 C change),
because a hot surface suspended in the air cannot draw the while in the lower region it is decreasing more rapidly (about
 H.-J. Park, D. Holland / Building and Environment 36 (2001) 883–889 889

this case radiation heat gain into lower level is more im-
portant in calculating the cooling load of the displacement
ventilation.

4. Summaries and conclusions

The e"ect of source location on convective heat gain into

the lower region of displacement ventilation systems is in-
vestigated by using CFD simulation. When a location of
the heat source is higher, a convective heat gain from the
heat source to the lower level decreases signi5cantly, which
results in change in temperature 5eld and in a reduction of
the cooling load in that region. The results also show that the
level of strati5cation is altered depending on the source lo-
Fig. 9. Average temperature of lower and upper regions as heat source cation, because the characteristics of circulation 2ows gen-
rises.
erated in the upper region changes with source heights. The
source location also signi5cantly a"ects the behavior of the
gravity current produced by a low velocity di"user when
the heat source is in a location lower than a di"user height.
A larger temperature gradient is created in the region above
the heat source as the heat source ascends.

References

[1] Mathisen HM. Case studies of displacement ventilation in public

hall. ASHRAE Transactions 1989;95(2):1018–27.
[2] Sandberg M, Blomqvist C. Displacement ventilation systems in oEce
rooms. ASHRAE Transactions 1989;95(2):1041–9.
[3] Dunham Inc. HVAC Comparison Study at 225 Bush Street San
Francisco, Minneapolis, MN, USA. 1999.
[4] Li Y, Sandberg M, Fuchs L. Vertical temperature pro5les in
rooms ventilated by displacement: full-scale measurement and nodal
Fig. 10. Normalized cooling load. modeling. Indoor Air 1992;2:225–43.
[5] Halton Inc. Displacement ventilation design guide, Finland: Halton
Oy, 1999.
◦ [6] Zhivov AM. Design guide for displacement ventilation. Savoy, USA:
3 C change). It is apparent that the location of the heat IAT, 1997.
source has a more signi5cant e"ect in the lower level than [7] Chen Q, Glicksman L. Performance evaluation and development of
in the upper level, which should be considered in the design design guidelines for displacement ventilation. Boston, USA: MIT,
of the displacement ventilation systems. It is interesting to 1998.
[8] Yuan X, Chen QC, Glicksman LR. A critical review of displacement
note that as the heat source rises, the exhaust air temperature ventilation. ASHRAE Transactions 1998;104(1A):78–90.
range increases. For example, when the heat source is at [9] Nielsen PV. Displacement ventilation-theory and design. Denmark:
◦
1:75 m, the temperature range of exhaust air is 1:2 C while Aalborg University, 1993.
◦
at 0:5 m, it is 0:2 C. It is also shown that in all cases, there [10] Patankar SV. Numerical heat transfer and 2uid 2ow. New York,
USA: Hemisphere, 1980.
is a region where the temperature is beyond the average
[11] Launder BE, Spalding DB. The numerical computation of turbulent
exhaust temperature and it is shown in Fig. 8. 2ows. Comp. Meth. Appl. Mech. Energy 1974;3:269–89.
Fig. 10 shows how the cooling load in the occupied zone, [12] Tennekes H, Lumley JL. A 5rst course in turbulence. Cambridge
that is de5ned as Q = Cp × mass 2ow rate × HT , is reduced MA: The MIT Press, 1972.
with source height. HT is the temperature di"erence be- [13] White FM. Viscous 2uid 2ow. New York: McGraw-Hill, Inc., 1991.
[14] Launder BE, Spalding DB. Lectures in mathematical models of
tween supply and occupied zone temperature. The cooling turbulence. New York: Academic, 1972.
load is normalized based on the cooling load at a source [15] Coles DE, Hirst EA. Computation of turbulent boundary layers-1968
height of 0:5 m. It is shown that the cooling load decreases AFOSRIFP Stanford Conference, Proceedings of 1968 Conference,
signi5cantly with the source height. It indicates that if the Vol. 2, Stanford University, Stanford, CA.
height of the heat source is higher than the strati5cation [16] Shilkrot EO. Determination of design loads on room heating and
ventilation systems using the method of zone by zone balances.
level, practically all its convective heat contribution into ASHRAE Transactions 1993;99(1):987–91.
the lower zone can be neglected for cooling load calcula- [17] Li Y, Sandberg M. Ventilation by displacement-its characteristics,
tions unlike the conventional ventilation (mixing type). In design and other related developments. AIRAH J 1998;52(3):31–8.