Академический Документы
Профессиональный Документы
Культура Документы
= t t h q
ws
&
The driving force for heat transfer ( q&
) is the temperature difference between the wall
surface (t
ws
) and the free fluid stream (t
).
t
t
ws
t
ws
c
c
ws
c
c
ws
c
ws
Flow
v
q
&
m&
( )
= c c h m
ws m
&
The conductance h
m
is essentially a fluid mechanic property of the system, whereas the
concentration is a thermodynamic property.
This second equation has the same form as the first one resulting to the rise of the heat-
and mass- transfer analogies.
The form of these equations is in fact a special case of the general form of a convection
coefficient as given by :
Force Driving x T COEFFICIEN FLUX =
2.2.2 Hydraulic, thermal and concentration boundary layer
Figure 2 : Heat and mass transfer from a surface in contact with a fluid
In 1904, Ludwig Prandtl stated : At high Reynolds number the effect of fluid friction is
limited to a thin layer near the boundary of the body , hence the term THE BOUNDARY
LAYER came into engineering practice.
Figure 2 shows the boundary layer developing over a flat plate under forced convection,
meaning there is an external velocity
=
x v
Re
x
In general is the Reynolds number is lower than a certain value, depending on the
geometry flow, is laminar. Above a second value for the Reynolds number the flow is
fully turbulent. In between transitional flow occurs. In general a Reynolds number is
defined as
vL
= Re
with v a characteristic velocity in the flow and L a characteristic length.
By solving the Navier-Stokes equations for a two dimensional flow for this geometry, as
discussed in Kays WM, Crawford ME, (1993), the hydraulic boundary layer thickness as
function of the position along the plate can be found. The hydrodynamic or momentum
boundary layer may be defined as the region in which the fluid velocity changes from its
99% free stream value to zero at the body surface. This is not a precise definition of the
boundary layer thickness. It only means that the boundary layer thickness is the distance
from the wall in which most of the velocity change takes place.
Out of this analysis follows the drag coefficient also known as the friction coefficient (c
f
):
2 /
2
=
v
c
f
with the fluid density and the fluid friction or shear stress.
When there is heat or mass transfer between the fluid and the surface, it is also found
that in most practical applications the major temperature and concentration changes
occur in the region very close to the surface. This gives rise to the concept of the thermal
boundary layer and the concentration boundary layer, and again the relative thinness of
these boundary layers permits the introduction of boundary-layer approximations similar
to those introduced for momentum. Solving the Navier-Stokes equations for the energy
or concentration transport equations results in a thermal boundary layer thickness and a
concentration boundary layer thickness as function of the coordinate x.
In the solution of the diffential equations the Prandtl number
k
c
p
= Pr appears, relating
the viscous boundary layer to the thermal boundary layer. For mass transfer this is
expressed by the Schmidt number
AB
D
Sc
= relating the viscous boundary layer to the
concentration boundary layer.
If the ratio is taken of the Prandtl number to the Schmidt number the Lewis number is
found, relating mass to thermal diffusion
Sc
Le
Pr
= . As this number relates the thermal to
the mass transfer boundary layer it will determine the analogy between heat and mass
transfer.
In the 19
th
century Reynolds was the first to report on the analogous behaviour of heat
and momentum transfer (Welty et al. 2001). He presented results on frictional resistance
to fluid flow in conduits which made the quantitative analogy between the two transport
phenomena possible. Out of these observations the Reynolds analogy was stated. The
Reynolds analogy relates the heat transfer coefficient (h) to the skin friction coefficient
using the free stream velocity and the free stream density and heat capacity (c
p
):
2
f
p
c
c v
h
St =
This relation can be deduced out of the boundary layer equations for laminar forced flow
across a solid boundary under the conditions that the Prandtl number (Pr) is equal to
one and no form drag is present.
The Reynolds analogy can also be applied to mass transfer in case the Schmidt number
(Sc) is equal to one:
2
f
c
m
c
v
h
p
=
In case both Pr and Sc numbers are equal to one, and hence the Lewis number (Le) is
one. Comparing both equations, a relation between the mass transfer coefficient and the
heat transfer coefficient is found, hence the analogy between heat and mass transfer
was founded :
m
p
h
c v
h
=
In general the convection heat transfer coefficient is made dimensionless through the
definition of a Nusselt-number and the mass transfer convection coefficient through the
definition of the Sherwood-number
k
hL
Nu =
AB
m
D
L h
Sh =
For forced convection the heat and mass transfer coefficients can be expressed as the
Nusselt number as function of the Reynolds and Prandtl number :
( ) Pr Re, F Nu =
( ) ,Sc F Sh Re =
For natural convection the flow is driven by buoyancy as a result of density differences in
the air volume. The dimensionless number characterising this flow type is the Grashof
number given by
2
3
=
L g
Gr
For natural convection the Grashof number takes over from the Reynoldsnumber to
determine the convection coefficients :
( ) Pr , Gr F Nu =
( ) Sc Gr F Sh , =
For the convective heat transfer coefficient a lot of data is available. For several, relative
simple geometries and different flow conditions (laminar, transitional, turbulent, forced
and buoyancy driven convection) an analytical solution of the Navier-Stokes equations
applied to a boundary layer exists. (See eg Kays WM, Crawford ME, 1993)
For more complex geometries correlations have be determined by curve fitting
dimensionless numbers to large data sets.
As there are many different correlations available care has to be taken in selecting the
suitable correlation. For analytical derived correlations the validity of assumptions and
simplifications should be checked. For experimentally derived correlations the range and
accuracy of the data set should be taken into consideration.
For the mass transfer coefficient boundary layer analysis leads again to analytical
solutions. Due to the fact that the differential equations for heat and mass transfer
resulting from boundary layer analysis are analogues, the solutions obtained for heat
transfer can be transformed into mass transfer solutions, by using the correct
dimensionless number cited earlier (Welty et al (2001)).
Furthermore it is very difficult to determine the convective mass transfer coefficient
experimentally. Therefore this analogy is applied in a lot of cases for calculating the
convective mass transfer coefficient, starting from the thermal measurements that where
done. Validity of the thus obtained mass transfer coefficients is by consequence even
more limited and great care should again be taken in selecting the proper correlation for
the studied geometry. (See eg Kays WM, Crawford ME, 1993)
For flow around buildings very little information was found about mass transfer
determination, both experimentally or numerically. For flows inside buildings, most
research is focussing on flows over building materials or porous materials. Wadso,L ,
1993 gives a very broad literature review.
During the progress of the Annex 41 new experiments were proposed to determine the
mass transfer coefficient. Often these experiments were found to have limited validity.
Secondly numerical methods, based on CFD, were used to determine mass transfer
from a fluid to a porous material. Finally the heat and mass transfer analogy was looked
into.
2.2.3 Overview of useful dimensionless numbers and nomenclature
Name Symbol Definition Meaning
Reynolds number Re
vL
= Re
inertial forces compared to
viscous forces
Grashof number Gr
2
3
=
L g
Gr
buoyancy forces compared to
viscous forces
Nusselt number Nu
k
hL
Nu =
dimensionless heat transfer
coefficient
Sherwood number Sh
AB
m
D
L h
Sh =
dimensionless mass transfer
coefficient
Prandtle number Pr
k
c
p
= Pr
viscosity to thermal diffusion
Schmidt number Sc
AB
D
Sc
=
viscosity to mass diffusion
Stanton Number St
Pr Re
Nu
c v
h
St
p
=
ratio of heat transferred into a
fluid to the thermal capacity of
fluid
Lewis number Le
Sc
Le
Pr
=
thermal diffusion to mass
diffusion
Archimedes number Ar
( )
2
3
=
L g
Ar
g l l
the motion of fluids due to density
differences
Rayleigh number Ra Ra = Gr Pr
Richardson number Ri
2
Re
Gr
Ri =
Measure for comparing forced
convection to buoyancy driven
convection (mixed convection)
Units
Heat capacity c
p
J /kgK
Diffusion coefficient D
AB
m/s
Gravitational
constant
g 9.81 m/s
Heat transfer
convection
coefficient
h W/mK
Mass transfer
convection
coefficient
h
m
m/s
Characteristic Length L M
Flow velocity v m/s
Free stream velocity
v
m/s
Boundary layer
thickness
m
Thermal conductivity W/mK
Dynamic viscosity Pa s
Kinematic viscosity m/s
Density kg/m
Table 1 : Dimensionless numbers and nomenclature
2.3 HEAT TRANSFER
2.3.1 Flow over and around buildings (experimental data)
Air flows around buildings are mainly of a forced nature as they are caused by wind.
Exterior convective heat and mass transfer coefficients at building surfaces are to a large
extent determined by the local wind speed. Usually, empirical formulae are used to relate
the reference wind speed at a meteorological station to the local wind speed near the
building surface and to relate the local wind speed to surface transfer coefficients. These
formulae however are based on a limited number of measurements.
Practical correlations given by J rges (1924) give a relation between free stream wind
(
V ) speed and the thermal convection coefficient :
s m V V h
s m V V h
/ 5 ; 6 . 5 1 . 7
/ 5 ; 6 . 5 0 . 4
78 . 0
> + =
< + =
Charples (1984) presented the following algorithm :
1 . 5 7 . 1 + =
loc
V h
with V
loc
the local wind speed measured at 1 m distance from the surface. It is expressed
as a simple function of the reference wind speed U
10
:
leeward U V
windward U V
loc
loc
; 7 . 1 4 . 0
; 2 . 0 8 . 1
10
10
+ =
+ =
Also ASHRAE (ASHRAE 1975) proposed practical correlations of a similar nature :
605 . 0
886 . 1
loc
U h =
3 . 0 05 . 0
/ 2 5 . 0
/ 2 25 . 0
10
10
10 10
+ =
< =
> =
U U leeward
s m U if
s m U if U U windward
loc
loc
These type of correlations are proven to be very sensitive to errors. First of all the
definition of the free stream wind speed around a building is not clear and secondly the
relation between wind speed and the actual heat transfer coefficient is not clearly stated.
This is discussed in Annex paper A41-T3-B-05-5.
2.3.2 Flow over and around buildings (CFD data)
Little is known about the actual value and the variability of local wind speed and surface
transfer coefficients across facades of different building geometries. In Annex paper
A41-T3-B-05-5 a validated Computational Fluid Dynamics (CFD) model (FLUENT) is
used to calculate the local wind speed near the exterior surface of a cubic building model
as a function of wind speed, wind direction and the position on the facade.
It is shown that the variation of the local wind speed across the facade is very
pronounced and that using the available empirical formulae can yield large errors in
HAM calculations.
In Annex paper A41-T3-Br-07-2 (and Emmel M, Abadie M, Mendes N (2007)) similar
conclusions were drawn using in essence the same approach. Using CFD calculations
with CFX, correlations for the heat transfer coefficient were determined for the BESTEST
reference case J udkoff R.D., Neymark J .S. (1995). De correlations presented in the
previous paragraph were compared to the CFD calculations and both over and under
predictions (to about a factor 4) of these correlations were found in relation to the CFD
solutions. The paper ends with a list of new correlations determined by doing several
calculations with CFD on the BESTEST geometry. These are copied here. More
information about validity and boundary conditions can be found in the paper.
Table 2 : Data according to A41-T3-Br-07-2
2.3.3 Flow inside buildings (experimental data)
Air flows inside buildings occur due to two main reasons. Firstly there are air streams
caused by ventilation systems (jets) or pressure differences between adjacent rooms
(draught). These are thus of forced nature as the flow is not driven by the temperature or
density fields it creates. Secondly temperature and concentration (vapour) differences
inside a room cause density differences and thus buoyancy.
Inside buildings both forced and natural convection will occur. Sometimes they will even
operate at the same place and time. This is what is called mixed convection.
2.3.3.1 Forced convection inside buildings
Spitler et al. (1991b) designed a full-scale experimental facility, with internal dimensions
of 4.57 x 2.74 x 2.74 m and a fan system delivered air to one of the two room inlets over
a range of 5 to 100 air changes per hour (ACH). The walls, floor and ceiling were
covered by heated panels, each with an independent electrical resistance heater.
Figure 3 Experimental facility of Spitler et al. 1991a
Spitler et al. correlated the convective heat transfer coefficients to the jet momentum
number J :
5 . 0
2 1
J C C h + = with
room
gV
U m
J
=
0
&
(U
0
jet inlet velocity, V
room
room volume)
The correlations from Spitler et al. (1991b) are listed in the Table below.
Table 3.Heat transfer coefficient correlations of Spitler et al. (1991b)
=
2
0
U
TL g
Ar
Surface Inlet h Limits
Ceiling Ceiling 11.4 + 209.7 J
0.5
0.001<J <0.03
Vertical walls Ceiling 4.2 + 81.3 J
0.5
0.001<J <0.03
Floor Ceiling 3.5 + 46.8 J
0.5
0.001<J <0.03
Ceiling Side wall 10.6 + 59.4 J
0.5
0.002<J <0.011 and Ar<0.3
Vertical walls Side wall 1.6 + 92.7 J
0.5
0.002<J <0.011
Floor Side wall 3.2 + 44.0 J
0.5
0.002<J <0.011 and Ar<0.3
Fisher extended Spitlers work by investigating buoyant, wall and free jets over a range
of room inlet conditions using the same enclosure. The room was also isothermal in
most of the experiments. However, a single wall was chilled in one group of ceiling-
diffuser experiments to examine the combined influence of buoyancy and forced effects.
The experiments only examined room cooling, the incoming air stream always being
colder than the air within the room and the room surfaces.
Correlations were developed for three classes of flows: (1) ceiling jets in isothermal
rooms, (2) free horizontal jets in isothermal rooms and (3) ceiling jets in non-isothermal
rooms.
For the first class, Fisher (1995) found surface convection to be independent of the inlet
velocity of the ceiling jet, but rather to depend upon the jets volumetric flow rate. He also
found the buoyancy forces of the cold jet to be negligible relative to the viscous Coanda
effect, adhering the jet to the ceiling and walls. The form of the correlations, expressed in
dimensionless parameters, can be described as follows: (applicable 3<ACH<100)
3
Re
2 1
C
e
C C Nu + = where
=
3 / 1
room
hV
Nu
and
3 / 1
Re
room
diffuser
e
V
V
=
&
(
diffuser
V
&
jet volumetric flow rate m/s)
With the free horizontal jets in isothermal rooms the buoyancy forces of the cold jet also
had a negligible impact on convection from the walls and floor. Therefore the same type
of equation was also used to correlate these data. Convection from the ceiling, though,
was affected by buoyancy.
Consequently, an alternate equation to correlate the ceiling data: (applicable for 3<
ACH<12)
e
C
e
Ar
C C Nu
3
Re
2 1
+ =
Table 4 : Convective heat transfer correlations of Fisher (1995)
Fisher and Pedersen (1997) correlated the same data as Fisher (1995) using a different
functional form. The correlations are applicable for forced convection and work in similar
conditions as the correlation developed by Fisher (1995). The main difference is the
position of the diffuser. These correlations are applicable in rooms with ceiling diffusers
where the diffuser jet is attached to the ceiling surface. As with previous correlations,
supply air temperature is used as the reference temperature. The correlations are given
in Table 5.
Table 5 :Convective heat transfer correlations of Fisher and Pedersen (1997)
2.3.3.2 Natural convection or buoyancy driven flow inside buildings
Alamdari and Hammond (1983) are one of the first to develop correlations dedicated to
building applications. Correlations that cover laminar, transitional and turbulent flow
regimes for the following three configurations are given: (1) vertical surfaces, (2) stably-
stratified horizontal surfaces (e.g. warm air above a cool floor) and (3) buoyant flow from
horizontal surfaces (e.g. cool air above a warm floor). The correlations cover the full
range of temperature and dimensions that appear in buildings. But they are not
Surface type Configuration h
Walls
( )
8 . 0
190 . 0 199 . 0 ACH +
Floor
( )
8 . 0
116 . 0 159 . 0 ACH +
Ceiling
Forced convection with ceiling
diffuser in isothermal rooms
( )
8 . 0
484 . 0 166 . 0 ACH +
Walls
( )
8 . 0
132 . 0 110 . 0 ACH +
Floor
( )
8 . 0
168 . 0 704 . 0 ACH +
Ceiling
Forced convection with wall
diffuser (free jet)
( )
8 . 0
00444 . 0 064 . 0 ACH +
Surface type Configuration h
Walls
( )
8 . 0
19 . 0 ACH
Floor
( )
8 . 0
13 . 0 ACH
Ceiling
Forced convection with ceiling
diffuser in isothermal rooms
( )
8 . 0
49 . 0 ACH
applicable for cases where buoyancy is created by thermal devices (e.g. radiator
baseboard heater or fan coil) and mechanically driven jets as experienced in
mechanically ventilated buildings. They are only valid for purely buoyant flow, in cases
where buoyancy is caused by the temperature difference between a surface and the
surrounding room air (T). Alamdari and Hammond (1983) did not perform new
experiments, the correlations are based on collected experimental data reported in the
literature. All data is derived from experiments conducted with free standing surfaces
(surfaces not part of the room), which limits the applicability of the correlations.
Table 6 : Alamdari and Hammond convection correlations (Beausoleil-Morrison, 2000)
(H is room height en D
h
the room hydraulic diameter)
Surface type Ventilation regime h
Wall
[ ]
6 / 1
6
3 / 1
6
4 / 1
23 . 1 5 . 1
T
H
T
Floor (T
surface
>T
air
)
Ceiling (T
surface
<
T
air
)
[ ]
6 / 1
6
3 / 1
6
4 / 1
63 . 1 4 . 1
T
D
T
h
Floor (T
surface
<T
air
)
Ceiling (T
surface
>T
air
)
Natural convection
(system is off)
5 / 1
6 . 0
h
D
T
Khalifa and Marshall (1990) performed experiments in a room sized test cell to produce
correlations specific to internal convection within buildings. Convection correlations are
developed based on measurements in an experimental chamber with room sizes: 2.95 x
2.35 x 2.08 m (l x w x h). The correlations for vertical surfaces are defined for surfaces in
the vicinity of a terminal device and for other surfaces. To assess a number of common
convection regimes, the test cells configuration was varied. Different heating systems
(e.g. radiator, in-floor heating, convective heating) were analyzed, as was the placement
of the heating device (e.g. underneath a window or facing a window).
Figure 4 : Experimental test room of Khalifa and Marshall (1990)
Khalifa (1989) used the average room air temperature as the reference temperature to
calculate the convective heat transfer coefficient. But Khalifa and Marshall (1990)
measured the air temperature outside the thermal boundary layer at a distance of 60 mm
from the interior surface of the wall, which is used as the reference temperature.
Khalifa generated a total of 36 correlations. Data from similar correlations were
combined together in order to obtain new and more general correlations which can be
applied in more than one configuration (Khalifa and Marshall 1990). By combining these
similar results the data were collapsed into a series of 10 equations (Tables 7 and 8).
Table 7 Khalifa convection correlations (Beausoleil-Morrison 2000)
Surface type Ventilation regime h
Wall
In the vicinity of the
terminal device
Rooms heated by radiator
Radiator not located under window
Only surfaces adjacent to radiator
32 . 0
98 . 1 T
Rooms heated by radiator
Radiator located under window
Wall
Rooms with heated walls
Not applicable for heated walls
24 . 0
3 . 2 T
Wall Rooms heated by circulating fan
heater
Only for surfaces opposite to fan
25 . 0
92 . 2 T
Rooms heated by circulating fan
heater
For surfaces not opposite to fan
Rooms with heated floor
Wall
Rooms heated by radiator
Radiator not located under window
For surfaces not next to radiator
23 . 0
07 . 2 T
Window
Rooms heated by radiator
Radiator located under window
11 . 0
07 . 8 T
Window
Rooms heated by radiator
Radiator not located under window
06 . 0
61 . 7 T
Rooms heated by radiator
Radiator located under window
Ceiling
Rooms with heated walls
17 . 0
1 . 3 T
Rooms heated by circulating fan
heater
Rooms with heated floors
Ceiling
Rooms heated by radiator
Radiator not located under window
13 . 0
72 . 2 T
Table 8 Khalifa and Marshall (1990) convection correlations
Surface type Ventilation regime h
Wall
Large isolated vertical surface
14 . 0
03 . 2 T
Floor
Large heated surface facing upward
24 . 0
27 . 2 T
Calay et al. (1998) performed an experimental study of buoyancy-driven convection in
rectangular enclosures. The enclosure was one-quarter scale model of a typical room. It
was based on hot box arrangement, in which two opposing walls are heated and cooled
while others are insulated and act as adiabatic walls. Four sets of experiments were
performed to simulate the following convective heat-flow configurations: (1) enclosure
heated from side, (2) large vertical walls as hot and cold plates, (3) small vertical walls
as hot and cold plates, (4) enclosure heated from below, (5) stably stratified convection
(enclosure heated from ceiling).
The convective heat transfer correlations are given in terms of dimensionless
parameters: Nusselt, Prandtl and Grashof number. The correlations recommended by
ASHREA (1985) and CIBSE (1986) and other correlations derived from tests with full
size enclosures and similar configurations are used for comparing the experimental
results (Table 9)
Table 9 Equations employed for comparison (Calay et al. 1998)
Equation Correlation, Nu Gr range Flow
condition
Configuration: stably
stratified, T
w
=c
te
CIBSE (1986)
ASHRAE (1985)
Alamdari and Hammond
(1983)
Min et al. (1956)
0.236Gr
1/4
0.218Gr
1/4
0.56Gr
1/5
0.065Gr
0.255
10
8
<Gr<10
1
0
10
8
<Gr<10
1
0
10
8
<Gr<10
1
0
Not
specified
Laminar
Laminar
Laminar
Not specified
Awbi and Hatton (1999) conducted experiments in two experimental chambers, with
different size in order to assess scale effects (Figure 5). The first chamber had a typical
room size of 2.78 x 2.78 x2.3 m. The second was considerably smaller 1 x 2.78 x 2.3 m.
This chamber was kept at a low temperature, so that the wall that connected the two
chambers acted as a heat sink. The main chamber was conditioned by electrically
heated plates affixed to the surfaces. A single surface (wall, floor or ceiling) was heated
in each experiment. All the walls were aluminium plated and long-wave radiation was
taken into account.
Figure 5 Environmental chamber of Awbi and Hatton (1999)
Natural convection from heated room surfaces was characterized by a correlation of the
mean convection heat transfer coefficient for whole wall heated surfaces.
Table 10 Awbi and Hatton (1999) natural convection correlations
Khalifa (2001) gives an extensive review of studies about natural convective heat
transfer coefficients on surfaces in two- and three-dimensional enclosures with primary
focus on those with a direct application to heat transfer in buildings. Figures 6 to 8 give a
comparison of the different correlations mentioned in Khalifa (2001).
Figure 6 : Convective heat transfer coefficient correlations for vertical surfaces (Khalifa
2001)
Surface type Ventilation
regime
Nu
Walls
( 9 x 10
8
<6 x 10
10
)
( )
293 . 0
Gr 289 . 0
Floors
( 9 x 10
8
<7x 10
10
)
( )
308 . 0
Gr 269 . 0
Ceilings
( 9 x 10
8
<1 x 10
11
)
( )
133 . 0
Gr 78 . 1
Partly heated
ceilings
Buoyant with
heated surface
( )
16 . 0
Gr 517 . 3
Figure 7 Convective heat transfer coefficient correlations for heated plate facing upward
(heated floor/ cold ceiling) (Khalifa 2001)
Figure 8 Convective heat transfer coefficient correlations for heated plate facing
downward (heated roof/cold floor) (Khalifa 2001)
2.3.3.3 Mixed convection inside buildings
Beausoleil-Morrison (2000) developed a suitable method for solving mixed flow. He
created his correlations by combining the correlations for natural convection (Alamdari
and Hammond 1983) and for forced convection where the air is supplied by a ceiling
diffuser (Fisher 1995). In some cases the mechanical and buoyant forces will assist (act
in same direction, Figure 9) while in others they will oppose (act in opposite directions) or
act transversely (act in perpendicular directions). It is difficult (usually impossible) to
predetermine whether a configuration will be dominated by buoyant forces or mechanical
forces. Beausoleil-Morrison solves this problem by selecting and combining the
appropriate correlations for forced and natural convection.
Figure 9 Assisting mechanical and buoyant forces
Table 11 Convective heat transfer coefficient correlations of Beausoleil-Morrison (2000)
for mixed flow
Surface type h
Assistin
g forces
[ ] ( ) [ ]
1
3
8 . 0
6
1
3
6
3 / 1
6
4 / 1
190 . 0 199 . 0 23 . 1 5 . 1
ACH
H
d s
Wall
Opposin
g forces
[ ] ( ) [ ]
[ ]
( ) [ ]
8 . 0
6 / 1
3 / 1
6
4 / 1
3
8 . 0
6
1
3
6
3 / 1
6
4 / 1
190 . 0 199 . 0 of % 80
23 . 1 5 . 1 of % 80
190 . 0 199 . 0 23 . 1 5 . 1
max
ACH
H
ACH
H
d s
d s
Floor
Buoyant
[ ] ( ) [ ]
3 / 1
3
8 . 0
6
1
3
6
3 / 1
6
4 / 1
116 . 0 159 . 0 63 . 1 4 . 1
ACH
D
d s
h
Stably
stratified
( ) [ ]
3 / 1
3
8 . 0
3
5 / 1
116 . 0 159 . 0 6 . 0
ACH
D
d s
h
Ceilin
g
Buoyant [ ] ( ) [ ]
1
3
8 . 0
6
1
3
6
3 / 1
6
4 / 1
484 . 0 166 . 0 63 . 1 4 . 1
ACH
D
d s
h
The experiments of Awbi and Hatton (2000) were carried out in the same enclosure as
the natural convection experiments. They only placed an air handling unit onto the
ceiling of the small (cold) compartment to cool the dividing wall that separates the two
compartments. The fan and heating plates were positioned on a wall, the floor and the
ceiling to investigate the effect of a 3D wall jet on the surface convective heat transfer
coefficient (Figure 10). The flow regime was a combination of natural convection,
caused by the heated plates and forced convection, due to the fan.
Figure 10 : Different positions of the fan in case of heated ceiling
Table 12 Awbi and Hatton (2000) Convective heat transfer coefficient correlations for
forced convection
Novoselac (2005) investigated the validity of the existing correlations from different
authors for the airflow regimes in buildings. Afterwards, he developed new convection
correlations for surface types and airflow regimes where validation of the existing
correlations failed by experimental measurements. The measurements were conducted
Stably
stratified
( ) [ ]
3 / 1
3
8 . 0
3
5 / 1
2
484 . 0 166 . 0 6 . 0
ACJH
D
d s
h
Surface type Ventilation
regime
h
cf
h
cf
/h
cn
Walls
( 9 x 10
8
<6 x 10
10
)
( )
873 . 0 536 . 1
U W 79 . 3
( )
293 . 0
873 . 0
536 . 1
T
U
W 3165 . 2
Floors
( 9 x 10
8
<7x 10
10
)
( )
557 . 0 575 . 0
U W 248 . 4
( )
308 . 0
557 . 0
595 . 0
T
U
W 06 . 2
Ceilings
( 9 x 10
8
<1 x 10
11
)
J et over a heated
surface
( )
772 . 0 074 . 0
U W 35 . 1
( )
133 . 0
772 . 0
074 . 0
T
U
W 45 . 3
in an experimental chamber with typical room size (6 x 4 x 2.7 m) and typical positions
for diffusers and radiant panels (Figure 11). Adjacent to the environmental chamber is a
climate chamber to simulate external conditions. For the correlations developed with a
displacement ventilation system, air was supplied by displacement diffusers. For the
forced convection correlations with mixing ventilation systems, a high aspiration diffuser,
located at the ceiling, discharge jets along the long side of the radiant panels. The
cooling panels occupied 50% of ceiling space and they were integrated into the
suspended ceiling structure.
Figure 11 Experimental facility for the development of convection correlations
(Novoselac 2005)
2.3.4 Flow inside buildings (CFD data)
Awbi (1998) compared experimental results for natural CHTC of heated room surfaces
with CFD calculations. Two turbulence models were used: (1) a standard k- model
using wall functions and (2) a low Reynolds number k- model.
The logarithmic standard wall functions describe the momentum and heat transfer from
the internal surfaces of a room. But these functions are empirically derived for forced
convection in pipes and over flat plates. Awbi (1998) concluded that prediction of the
convective heat transfer coefficient using wall functions is extremely sensitive to the
distance of the point from the surface (y
p
) at which the wall function is applied. But CFD
analysis, which uses wall functions, proved to be useful in the investigation of the airflow
over the heated plates and the air movement within the chamber (Awbi (1998)).The
more accurate prediction of the heat transfer from room surfaces, using a low Reynolds
number turbulence model, is very time consuming.
An alternative is to use an experimental determined expression for the convective heat
transfer coefficient for room surfaces in a CFD code. This is what Beausoleil-Morrison
(2000) and Novoselac (2005) have done with their ACA and respectively MACA
algorithms.
During the Annex 41 CFD was used to evaluate the possibilities of determining heat
transfer coefficients with CFD. In [A41-T3-C-06-5 and A41-T3-C-06-6] CFD was used to
determine the heat transfer coefficient for flow between two infinite plates. A good
agreement was found in laminar flow between analytical solutions for different cases and
the CFD results if the bulk fluid temperature was used as a reference temperature (error
< 10
-2
%). For turbulent forced convection a good agreement was also found and
different turbulence models give limited deviations in the developed zone. For a natural
convection case the CFD calculations did not result in velocity profiles consistent with
experimental data when using law-of-the-wall equations for forced convection. Using a
Low-Re model resulted in good agreement between experiment and simulation.
2.4 MASS TRANSFER
2.4.1 Heat and mass transfer analogy
As little direct experimental data is available on vapour transfer coefficients the heat and
mass analogy is used to calculate the vapour transfer coefficient. This is the basis for the
prEN 15061, 2004. The heat and mass transfer analogy is applicable for very specific
cases.
In case both Pr and Sc numbers are equal to one, and hence the Lewis number (Le) is
one, the relation between the mass transfer coefficient and the heat transfer coefficient
is given by :
m
p
h
c v
h
=
The Reynolds analogy is limited in its application because of the strict conditions under
which it is valid. (It was deduced for laminar forced flow across a solid boundary under
with no form drag.) Yet this analogy inspired researchers to seek for better analogies
which are more generally applicable. Prandtl developed an analogy for heat and
momentum transfer and for mass and momentum transfer considering the turbulent core
and the laminar sublayer in the boundary layer equations. The effect of Pr and Sc
numbers different from one is taken into account in this analogy. This led to the following
equations for the heat and mass transfer coefficients:
( ) 1 Pr 5 1
2
+
=
S
c
c
p
f
f
c v
h
St
( ) 1 5 1
2
+
=
Sc
v
h
S
c
c
m
f
f
Von Karman extended Prandtls work and took the effect of the transition layer between
the laminar sublayer and the turbulent core into account. This led to an extra correction
term as function of respectively Pr and Sc in the two previous equations. The application
of the Prandtl and Von Karman analogies is restricted to cases with negligible form drag.
Both the Prandtl analogy as the Von Karman analogy reduce to the Reynolds analogy
for Pr and Sc number equal to one.
While Prandtl and Von Karman adapted the Reynolds analogy by considering the
transfer equations in the boundary layer, Chilton and Colburn sought modifications to the
Reynolds analogy using experimental data (Colburn 1933, Chilton & Colburn 1934).
They suggested a simple modification for situations with Pr and Sc numbers different
from unity. This was done by defining the j factor for heat transfer and the j factor for
mass transfer:
2
Pr
3 / 2
f
h
c
St j = =
2
Pr
3 / 2
f
m
m
c
v
h
j = =
Colburn applied the j factor for heat transfer to a wide range of data for flow on different
geometries and found it to be quite accurate for conditions where no form drag exists
and for Pr between 0.5 and 50. The complete Chilton-Colburn analogy is found when
equations are combined:
3 / 2 3 / 2
Pr Sc h
c
h
m
p
=
-
When form drag is present neither j
H
or j
m
equals c
f
/2, yet it has been found that equation
remains valid. It is clear that the Chilton-Colburn analogy also reduces to the Reynolds
analogy for Pr and Sc numbers equal to unity. Unlike the Prandtl or Von Karman analogy
the relation between the heat and mass transfer coefficients is no longer function of the
skin friction coefficient.
The analogy between heat and momentum transfer and between mass and momentum
transfer is based on the assumption that respectively the dimensionless velocity and
temperature profiles and the dimensionless velocity and mass concentration profiles are
similar. This is the case for forced convection flow over a solid surface without form drag.
All the analogies mentioned in this paragraph, as well the theoretical deduced ones as
the experimental Chilton-Colburn analogy, were developed for this case.
In A41-T3-C-04-7 D. Derome presented a limited and preliminary set of experiments
through which the author suggests that the analogy between heat and mass transfer
gives an overprediction of the mass transfer. The author claims more research is needed
into the use of the analogy.
In A41-T3-B-07-4 Steeman et al. investigated the influence of conditions in which the
analogy is valid and determined accuracy. CFD simulations are performed to compare
simulated vapour transfer coefficients with coefficients predicted out of the heat and
mass analogy. It is found that dissimilarity of the boundary conditions induces the largest
differences between predicted and simulated transfer coefficients: for the average
transfer coefficient the largest difference seen in the simulations is an under prediction
with a factor 0.41 while for local transfer coefficients differences up to a factor 10 are
found. Hence the heat and mass analogy gives a reasonable estimate of the average
vapour transfer coefficient, but can lead to large errors for local coefficients.
2.4.2 Flow over and around buildings (experimental data)
Little information can be found in literature about experimentally determined mass
transfer coefficients for building applications. Swartz (1972) presented some
experimental results for mass transfer to building surfaces under different wind speeds.
Swartz reported for V
wind
>1m/s
face leeward V h
face windward V h
wind m
wind m
43 . 0
58 . 0
20
27
=
=
Worch (2004) published data for vapour transfer resistance for flows in natural en forced
convection, in and around buildings. He determined vapour transfer resistances which
are related to the mass transfer coefficient as :
m
d
h
s
=