You are on page 1of 19

Airow and thermal analysis of at and corrugated unglazed

transpired solar collectors


Siwei Li
a
, Panagiota Karava
b,
, Eric Savory
c
, William E. Lin
c
a
Purdue University, School of Civil Engineering, West Lafayette, IN, USA
b
Purdue University, School of Civil Engineering and Division of Construction Engineering and Management, West Lafayette, IN, USA
c
The University of Western Ontario, Department of Mechanical and Materials Engineering, London, Ontario, Canada
Received 19 October 2012; received in revised form 13 January 2013; accepted 31 January 2013
Available online 22 March 2013
Communicated by: Associate Editor Matheos Santamouris
Abstract
The ow structure and convective heat transfer mechanisms in Unglazed Transpired solar Collectors (UTCs) are crucial to their per-
formance. High-resolution, 3-dimensional, steady, Reynolds-Averaged NavierStokes (RANS), Computational Fluid Dynamics (CFD)
simulations have been used to analyze the convective heat transfer processes for both at and corrugated UTCs. The performance of ve
potentially suitable turbulence closure models: Standard ke, Renormalization Normal Group ke (RNG ke), Realizable ke, Shear
Stress Transport kx (SST kx) and Reynolds Stress Model (RSM), has been evaluated. Two scenarios have been considered: a at
UTC under free stream approaching ow and a corrugated UTC subjected to a plane wall jet ow. The results were compared against
experimental data from the literature and those obtained using a full-scale experimental set-up in a solar simulator, in terms of the veloc-
ity and turbulent kinetic energy of the airow as well as the plate surface and air temperature. The validation study showed that the RNG
ke model has the best overall performance for both UTC geometries at reasonable accuracy and computing cost. A parametric analysis
has been conducted using the RNG ke model at 3 m/s approaching ow velocity, for dierent suction ow rates (0.01 and 0.06 m/s) and
free stream turbulence intensities (0.1% and 20%). The results showed that in UTCs, it is the suction velocity, rather than the suction
ratio of V
s
/U
1
, that has the most profound impact on the boundary layer development and thermal eciency. For at UTCs, the pres-
ence of perforations is more signicant than the level of turbulence intensity in the approach ow. However, for corrugated UTCs, the
incident turbulence intensity plays a more important role in the system performance than the perforation dimensions due to the turbulent
interaction between the corrugated geometry and the incoming ow. Although a high suction rate can help maintain the asymptotic
boundary layer proles and, hence simplify the energy analysis of UTCs, it is not applicable in practice due to the requirement of large
fan power. These outcomes will not only further advance the development of UTCs and optimize their performance but also provide
insights for the development of simplied thermal analysis models for use in building energy simulation.
2013 Elsevier Ltd. All rights reserved.
Keywords: Computational Fluid Dynamics; Convective heat transfer; Corrugations; Unglazed transpired solar collectors; Solar buildings; Turbulence
models
1. Introduction
The idea of replacing conventional building cladding
with a system that generates electricity and heat is an area
that, until recently, has received only limited attention
although the potential energy and cost savings from inte-
grated and optimized solar technologies are high (Athieni-
tis et al., 2011). Unglazed transpired solar collectors,
known as UTCs, consist of dark porous metal sheets (at
or corrugated) installed as the exterior layer of the building
fac ade with a narrow gap beneath it. The cladding absorbs
0038-092X/$ - see front matter 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.solener.2013.01.028

Corresponding author. Tel.: +1 765 494 4573; fax: +1 765 494 0644.
E-mail address: pkarava@purdue.edu (P. Karava).
www.elsevier.com/locate/solener
Available online at www.sciencedirect.com
Solar Energy 91 (2013) 297315
solar radiation, thus heating up the suction fan-driven air
owing through the perforations. Corrugated UTCs can
be integrated with photovoltaic systems, thereby generat-
ing electricity and heat, resulting in overall eciencies that
can be up to 70% (Bambara, 2012). In UTC systems, the
hot air is collected at the outlet and used either to preheat
ventilation air, or as an input to heat pumps, thus satisfy-
ing a signicant part of the buildings heating, cooling, and
hot water requirements (Athienitis et al., 2011).
In corrugated UTCs, the complex airow features (sche-
matically illustrated in Fig. 1) due to irregular geometry,
i.e. the ow separation and recirculation in the valley and
the ow impingement on the windward slope, the discrete
suction (versus the homogeneous suction described later),
the incident and boundary layer turbulence and wide Rey-
nolds number (Re) range, together with the resulting con-
vective heat transfer on both sides of the plate, are
crucial to the performance. A simplied energy balance is
shown in Fig. 1 and details are provided in Appendix A.
The solar radiation ux (G) falling on the plate is equal
to the sum of the heat obtained (due to suction, radiation
and convection) by the system (Q
gain
) and the heat loss
(Q
loss
) with the thermal eciency g dened by Q
gain
/G. In
the gure, Q
conv.ext
stands for the convective heat ux
between the plate and ambient air, Q
conv.int
represents the
convective heat transfer between the plate and suction
air, Q
rad.ext
is the radiation between the heated plate and
the sky and Q
rad.int
is the radiation between the heated
plate and the back wall of the cavity, which, together with
Q
rad.ext
, are part of Q
loss
. It should be noted that Q
loss
is not
equal to Q
conv.ext
plus Q
rad.ext
and Q
rad.int
, as the heat con-
vected to the ambient air will be recaptured by the suction
process. The air temperature through the perforations is
higher than that of the ambient air as it has already been
heated up by a portion of Q
conv.ext
. Knowing this tempera-
ture rise is important for the thermal eciency calculation,
as this portion of Q
conv.ext
plus the Q
conv.int
together is the
Q
gain
. However, obtaining this fraction from experimental
methods is very dicult due to the small size of the perfo-
rations. Thus, there is a need to develop high-resolution
Nomenclature
b jet nozzle exit height, 0.055 m
C
p
specic heat capacity, 1.005 kJ/kg K
G solar radiation falling on the UTC plate per unit
area, W/m
2
Pr Prandtl number (0.71 for air)
Q
conv.ext
convective heat transfer between the plate and
ambient air, W/m
2
Q
conv.int
convective heat transfer between the plate and suc-
tion air, W/m
2
Q
gain
heat obtained by the UTC system, W/m
2
Q
loss
heat loss from the UTC system, W/m
2
Q
rad.ext
radiation from the plate to the background sky,
W/m
2
Q
rad.int
radiation from the plate to the back wall of the
cavity, W/m
2
T
c
cavity exit air temperature, K
T
p
plate surface temperature, K
T
1
free stream approaching ow air temperature, K
u(x, y) instantaneous stream-wise component of veloc-
ity, m/s
U
m
maximum stream-wise mean velocity within a
vertical prole, m/s
U
1
mean velocity of the uniform approach ow, m/
s
v(x, y) instantaneous vertical component of velocity,
m/s
V
s
mean suction velocity, m/s
x stream-wise distance, m
y vertical distance above the plate, m
y
1/2
the height where half the maximum stream-wise
velocity occurs, m
y
*
dimensionless wall (normal) distance
Greek symbols
d velocity boundary layer thickness, m; u = 0.99
U
1
at y = d
d
t
thermal boundary layer thickness, m; (T
p
-
T) = 0.99 (T
p
T
1
) at y = d
t
g thermal eciency of the UTC system; g = qV
s-
C
p
(T
c
T
1
)/G
m kinematic viscosity of air, 1.511 10
5
m
2
/s at
20 C, 1.697 10
5
m
2
/s at 40 C
q density of air, kg/m
3
Abbreviation
BIPV/TBuilding Integrated Photovoltaic/Thermal
CFD Computational Fluid Dynamics
CHTC Convective Heat Transfer Coecient
CHWA Crossed Hot-Wire Anemometer
HWA Hot-Wire Anemometer
LDV Laser Doppler Velocimetry
PRESTO! PREssure STaggering Option
PV/T Photovoltaic/Thermal
RANS Reynolds-Averaged NavierStokes
RNG ReNormalization Group methods
RSM Reynolds Stress Model
SIMPLE Semi-Implicit Method for Pressure-Linked
Equations
SST Shear Stress Transport
TI Turbulence Intensity
TKE Turbulent Kinetic Energy
UTCs Unglazed Transpired solar Collectors
298 S. Li et al. / Solar Energy 91 (2013) 297315
Computational Fluid Dynamics (CFD) models, which can
provide details of the thermal and airow eld as well as
useful insights for (a) innovative UTC designs that opti-
mize the extraction of heat; and (b) simplied thermal anal-
ysis models for use in building energy simulation.
The airow in UTCs is primarily driven by mechanical
fans, although the thermal buoyancy eects induced by the
solar radiation falling on the plate and the incident atmo-
spheric boundary layer create complex aerodynamic eects
with impingement, separation, reattachment, etc. The Re
for the airow through the perforations is generally on the
order of 10
3
, while that of the wind ow over the plate is rel-
atively high (O (10
6
)). Thus, the corresponding ow regime
may span fromlaminar, transition state, to turbulent. More-
over, the low porosity (0.52 %) formed by scattered small
perforations, results in a non-uniformowdistribution over
the plate and creates complex local airow patterns. These
dier from the classical case of homogeneous suction (it
has also been referred to as uniform or continuous suction
in the literature), e.g. Iglisch (1944), Kay (1948) and Schlich-
ting and Gersten (2000), in which the perforation spacing
distance is within few millimeters and the vertical velocity
is assumed constant at the plate surface, giving rise to the
asymptotic solution for the velocity eld.
Iglisch (1944) computed the velocity boundary layer on
a horizontally placed at plate with homogeneous suction.
Here, the boundary layer commences directly at the leading
edge, yielding the Blasius solution without suction:
d 5:0

mx
U
1
_
1
Further downstream, the suction acts, which leads to the
asymptotic suction prole,
ux; y U
1
1 e

V sy
m
_ _
; vx; y V
s
2
which will emerge at about:

V
s
U
1

U
1
x
m
_
2 3
where d is the velocity boundary layer thickness, m; m is the
kinematic viscosity of air, equal to 1.511 10
5
m
2
/s at
20 C and 1.697 10
5
m
2
/s at 40 C; x is the stream-wise
distance, m; U
1
is the uniform approach velocity, m/s; u
and v are the stream-wise and vertical component of veloc-
ity, respectively; V
s
is the suction velocity, m/s and y is the
vertical distance above the plate, m.
Kay (1948) found that if the boundary layer was in an
undisturbed laminar condition at the start of the suction
region, this state could be maintained despite the small dis-
turbances, which would normally be sucient to promote
transition in the absence of suction. Schlichting and Ger-
sten (2000) observed a similar regulation, showing that if
the Re, based on the boundary layer thickness and free
stream velocity, is below 7 10
4
, transition to turbulence
on a smooth plate with homogeneous suction will not
occur and the boundary layer thickness will remain con-
stant after the starting length, regardless of the plate length.
By adopting Schlichtings homogenous suction theory
for at UTCs, Kutscher (1992) and Kutscher et al. (1993)
demonstrated that since there is no net ux of heat and
momentum into the boundary layer (due to the homoge-
neous suction), the velocity and thermal boundary layer
thicknesses will remain constant along the plate after the
starting length with the ratio of d/d
t
% Pr, where d
t
is the
thermal boundary layer thickness, m; Pr is the Prandtl
number of air and is equal to 0.71. As a result, the convec-
tive heat loss from the plate due to the approaching ow
only occurs over the starting length because in the asymp-
totic region all the heat transferred into the boundary layer
will be recaptured when the ambient air is drawn into the
cavity. In other words, the total amount of heat loss from
the plate into the boundary layer along the starting length
will be the same as the heat loss of the entire plate. This
solution implies that for large area UTC applications the
convection heat loss Q
conv.ext
is negligible since the starting
length is typically only few millimeters and the only loss of
the system is radiative (Q
rad.ext
and Q
rad.int
), which will
result in an underestimation of Q
loss
and, consequently,

t
Flow separation
and recirculation
in the valley
Impingement on the
windward slope
Suction in various
directions
Q
loss
Q
conv.ext
Q
gain
G
Q
conv.int
Windward
slope
Valley
Crest
Leeward
slope
Q
rad.ext
Adiabatic wall
Separation
point
Q
rad.int
Fig. 1. Schematic diagram of the thermal and airow eld in corrugated UTC.
S. Li et al. / Solar Energy 91 (2013) 297315 299
lead to an overestimation of the system thermal eciency
g, dened by the following simplied equation:
g qV
s
C
p
T
c
T
1
=G 4
where q is the density of air, kg/m
3
; C
p
is the specic heat
capacity of air, kJ/kg K; T
c
and T
1
are the cavity exit air
temperature and free stream air temperature, respectively,
and G is the solar radiation falling on the UTC, W/m
2
.
Although this assumption greatly simplies the thermal
analysis of UTC systems, it is only valid for laminar, uni-
form approaching ow conditions and homogeneous suc-
tion, which may be an oversimplication in two aspects.
Firstly, the suction process in UTCs cannot be considered
as being uniform, due to the large spacing of the perfora-
tions and low porosity. Secondly, the UTCs are mounted
on the surfaces of buildings, which are blu bodies im-
mersed in the atmospheric boundary layer and so the
approaching ow is highly turbulent.
Most previous convective heat transfer studies of UTCs
(Kutscher, 1992; Golneshan, 1994; Van Decker et al., 2001;
Gawlik, 1993) were focused on experimental investigations
due to the lack of advanced computational techniques.
Kutscher (1992, 1994) conducted experiments to evaluate
the overall convective heat transfer process for several
0.3 m 0.5 m UTCs subjected to uniform approaching
ow with a turbulence intensity of 0.8%. An empirical cor-
relation was obtained for the average Nusselt number (Nu)
of the entire plate by assuming uniform plate temperature
after the starting length. Other experimental studies
extended Kutschers research to discrete suction with slots
(Golneshan, 1994), various plate materials and thicknesses
(Van Decker et al., 2001) and dierent plate conductivity as
well as a sinusoidal shaped plate (Gawlik, 1993). The most
signicant limitation of all these studies is the small collec-
tor plate (approximately 0.15 m
2
) compared to realistic sys-
tems installed in buildings (typically 50010,000 m
2
) due to
the experimental restrictions and the fact that, based on the
homogeneous assumption, the airow and thermal eld is
the same after the starting length and so there is no need
to test a large plate.
There are also a few numerical studies following the
work by Kutscher (1992) which limited the geometry to a
single pitch (Arulanandam et al., 1999) or one corrugation
(Gawlik, 1993; Gawlik and Kutscher, 2002; Abulkhair and
Collins, 2010) by assuming that UTCs are a combination
of multiple pitches with similar thermal performance.
Other recent numerical studies focused on the thermal e-
ciency and energy modeling aspects, e.g. Delisle (2008)
developed a model for a prototype Photovoltaic/Thermal
(PV/T) UTC which was coupled with TRNSYS. Abulkhair
(2011) used numerical simulations with a laminar model
for ow over a UTC with one trapezoidal-shaped corruga-
tion to investigate the heat loss in the developing region
due to the wind and reported that the heat loss was about
three times larger than that from a at UTC. However, to
the authors best knowledge, no previous study in the liter-
ature has investigated the convective heat transfer process
for the entire perforated plate subjected to turbulent
approaching ow conditions, due to previous signicant
limitations in experimental and computing resources. In
addition, the homogeneous suction assumption may not
be applicable for corrugated UTCs due to the complex ow
features in separated ow regions.
CFD is a powerful alternative tool for investigating con-
vective heat transfer on UTCs, eliminating the diculties
associated with the precise control of the experimental con-
ditions. By solving the conservation equations of mass,
momentum and energy, CFD simulations, when properly
validated, can provide quantitative data for various airow
and heat transfer parameters in the system. This approach
also oers a higher degree of exibility and detail, for
example allowing the distinction of convective heat transfer
from radiation and providing detailed local convective heat
uxes (Q
conv.ext
and Q
conv.int
).
The present study builds on advancements in computing
power that enable the use of high resolution grids and full-
scale simulations as well as developments in turbulence
modeling and employs, for the rst time, realistic length
scales and approaching ow conditions in CFD simula-
tions for UTCs, validated with measured data in a state-
of-the-art solar simulator facility. The study provides an
in-depth analysis of the thermal boundary layer develop-
ment and convective heat transfer process in UTCs that
leads into guidelines for eciency estimation and collector
design and sets a solid foundation for future research on
this topic. In addition, ndings related to the use of existing
turbulence closure models in predicting specic ow fea-
tures and thermal elds may be applicable to other research
areas, such as cooling of electronics.
The ultimate goal of the present study is to develop
models for the analysis of UTCs and UTCs integrated with
PV/T systems that can be compiled within building simula-
tions to enable optimized design and coordinated operation
with HVAC systems and thermal storage mechanisms. In
this paper, experimental data are used to assess the accu-
racy of dierent CFD models in evaluating the complex
convective heat transfer process on UTCs. Then, the con-
vective heat transfer between the plate and the ambient
air is analyzed by computing the thermal boundary layer
development along the plate. The main parameters consid-
ered are the geometry (at and corrugated plate), suction
velocity and approaching ow.
2. Methodology
Two cases, based on the limited access to previous
experimental data and the restrictions of the full-scale
experiment setup, are considered to evaluate dierent
CFD models: a at UTC under uniform approach ow
and a corrugated UTC subjected to a plane wall jet. The
selected testing conditions (approaching ow velocity, sim-
ulated solar radiation, jet ow, etc.) are chosen according
to the available data. All radiation heat transfer has been
300 S. Li et al. / Solar Energy 91 (2013) 297315
excluded in the simulations to enable a focused and in-
depth study of the convective heat transfer process.
The commercial CFD software, ANSYS (version 13.0)
was used to solve the 3-D, steady state conservation equa-
tions of mass, momentum and energy and GAMBIT 2.4.6
was used for all geometry models. Five Reynolds-Averaged
NavierStokes (RANS) turbulence closure models, namely
the Standard ke, Renormalization Group Methods
(RNG) ke, Realizable ke, Shear Stress Transport (SST)
kx with low Re correction and the Reynolds Stress Model
have been evaluated. The basic RANS equations are
described in Appendix B. It should be noted that the Rey-
nolds Stress Model (RSM) was only used for isothermal
conditions since initial investigations showed that it
becomes unstable and does not converge for non-isother-
mal conditions. A second order upwind spatial discretiza-
tion scheme was adopted for all the variables except
pressure, which used PRESTO!, whilst SIMPLE was used
to couple the pressure and momentum equations. Conver-
gence was considered acceptable if the normalized residuals
were less than 10
6
for the energy and velocity components
and 10
4
for the other variables. In order to ensure that
grid resolution would not have a notable impact on the
results, grid independence for each case was checked using
three dierent grids with the renement considered accept-
able when the velocity and temperature changes with a
change of grid were within 5%. In all the simulations
high-resolution grids were used in the near wall region,
especially for modeling a corrugated UTC subjected to
plane wall jet, in which a ner grid was applied in the jet
core region and near the corrugated plate in order to better
capture the jet characteristics and the interaction between
the jet and the corrugated-shape plate. The height of the
rst cell layer was set to be 0.1 mm to maintain y
*
% 1, thus
fullling the required condition of enhanced wall
treatment.
2.1. Flat UTC under uniform approach ow
Experimental data from the literature (Van Decker,
1996) has been used to validate the at UTC model. A per-
forated plate with an area of 0.6 m 0.6 m and a cavity
with 0.15 m height were considered. The plate has a pitch
length of 16.89 mm and a perforation diameter of
1.59 mm. Considering the small size of the perforations,
their shape was assumed to be square, instead of the circu-
lar shape used in the experiments, for model simplicity.
Other control parameters in the experiments: i.e. the
600 W/m
2
simulated solar radiation, 298 K ambient air
temperature and the uniform, 0.8% turbulence intensity
of the oncoming ow, were used as inputs in the CFD
model. The simulated approach ow velocity ranges from
0 to 1 m/s and the suction velocity spans from 0.045 to
0.077 m/s, which were the tested conditions in the
experiments.
The model conguration for the at UTC, along with
the computational domain which includes 416,352 hexahe-
dral cells, is shown in Fig. 2. In the region above the plate,
the left side of the domain was dened as the velocity inlet
with a specied approach ow velocity and the right side of
the domain was dened as the pressure outlet with zero
gauge pressure. The plate was set to be a wall with heat ux
of 600 W/m
2
and thickness of 0.86 mm and the perfora-
tions on the plate were set to be interior. The bottom of
the cavity was also dened as a velocity inlet, but with
reversed ow direction and a specied suction velocity.
The left and right side of the cavity were dened as adia-
batic walls. A symmetry boundary condition was used for
all other faces, enforcing zero velocity and temperature
gradients. The upper domain is 0.3 m height, which is suf-
ciently high to not aect the boundary layer growth along
the collector surface and to eliminate any possible inuence
due to the size of the computational domain.
2.2. Corrugated UTC subjected to a plane wall jet
An experimental set-up was designed in a solar simula-
tor facility, as shown in Fig. 3. The solar simulator consists
of a lamp eld which uses a set of eight metal halide global
(MHG) lamps, with a total peak power output of 36.8 kW.
The MHG lamp eld produces a dense multiline spectrum
of rare earth metals similar to the air mass 1.5 spectrum
dened by EN 60904-3. This provides a spectral distribu-
tion very close to natural sunlight and fullls the specica-
tions of relevant standards EN 12975:2006 and ISO 9806-
1:1994 (PSE, 2009). An articial sky is installed in front
of the lamp eld in order to eliminate long-wave infrared
radiation emitted by the hot lamps to the solar system
0.3
0.5
0.055
0.15
Velocity inlet
Pressure outlet
Adiabatic wall
Wall with constant
heat flux
y
x
z
Solid row
Perf orated row
(a)
(b)
Fig. 2. (a) Flat UTC model conguration (unit: m). (b) Sketch of
perforated and solid row.
S. Li et al. / Solar Energy 91 (2013) 297315 301
tested (as the glass in front of the MHG lamps can reach
temperatures over 100 C during operation), while emulat-
ing sky temperatures. The articial sky consists of two
panes of low iron glass with an anti-reection coating, posi-
tioned parallel to the lamp eld, creating a cavity in
between where cold air is circulated in a closed loop and
it is cooled down by a waterair heat exchanger. More
information about the solar simulator can be found in
Bambara (2012). During the experiments, the collector area
was scanned with an automated pyranometer that was pro-
grammed to measure irradiance on a 150 mm grid. The dis-
tribution of solar irradiation had 5.5% uniformity, and the
average irradiance measured was 1029 W/m
2
. The articial
sky temperature during the measurements was about
10 C. Considering that the temperature dierence between
the plate surface and ambient air is small (less than 20 K)
and the solar simulator and the back wall of the cavity is
parallel and close to the UTC plate, the maximum corre-
sponding radiation heat loss is less than 10% of the total
convective heat ux. Therefore, the radiative heat loss from
the plate to the ambient environment during the experi-
ment is negligible.
A plane wall jet was created by a blower with a rectan-
gular nozzle outlet height of 0.055 m and width of 1.58 m.
A wood plate with a height of 0.55 m was placed above the
fan exit to reduce the entrainment of ambient air, thereby
resulting in a wall jet without a co-ow. The jet ow devel-
opment over a solid, at, smooth wood plate was rst eval-
uated to test the delity of the experimental equipment and
to serve as an additional evaluation of the CFD models.
Then, a 1.73 m 1.58 m corrugated UTC plate was tested.
A calibrated Crossed Hot-Wire Anemometer (CHWA) was
used for recording the time series of the three velocity com-
ponents. A constant temperature anemometry unit (Dantec
Dynamics MiniCTA Type 54T30) was used and the voltage
range was maximized by signal conditioning. Prior to the
experiment, the directional response of the 55P61 probe
was assessed (i.e. yaw and pitch factors were determined
following Bruun et al., 1990) using a smooth-ow, free
jet calibration facility (Dantec 90H02). Within the solar
simulator facility the wind speed calibration was completed
at the wall jet exit, before and after each unique test condi-
tion. By this in situ calibration procedure, the acquired
CHWA voltages were related to the wind speed determined
from a Pitot-static probe connected to an electrolyte micro-
manometer (Dwyer M1430 with accuracy of 0.4 m/s) and
a barometric pressure reading. A power law t (King, 1914)
to the calibration points yielded a high coecient of deter-
mination (R
2
= 0.998 or better) irrespective of whether the
MHG lamps were on or o. A temperature correction
based on thermocouple readings, which were coincident
with the CHWA samples, was applied as needed when tem-
perature drift/uctuation during sampling was signicant.
During the experiment, the air velocity was measured at
a rate of 10,000 samples per second for a duration of 30 s
at each measurement point. The CHWA probe was posi-
tioned by software control of a two-axis traverse, which
yielded a positioning accuracy of 0.2 mm per 254 mm
of travel (or better). After the calibration, the total uncer-
tainty was estimated as being 10% for the velocity com-
ponent measurements and 14% for the turbulence
quantities.
The parameters which have been measured for the
model validation are air temperatures in the cavity, at the
jet exit and representative of the laboratory ambient envi-
ronment; surface temperature of the UTC plate; vertical
proles of air velocity at various sites along the plate
length; suction ow rate through the UTC perforations;
and incident simulated solar radiation, wherein the spatial
distribution is assessed for uniformity over the UTC sur-
face. T-Type Thermocouples (with accuracy of 0.5 K),
connected to a data acquisition system, were used for the
air and surface temperature measurements. Air was drawn
through the cavity with a centrifugal fan (New York
Blower Company 1406 aluminum) coupled with a vari-
able speed motor (Baldor Electric Company 3 hp), and
the suction ow rate was measured with a laminar ow ele-
ment (accuracy: 5% of reading). All the design and con-
trol parameters considered in the experiments, i.e.
0.58 mm plate thickness, 0.095 m cavity height, 297 K
ambient air temperature and 5.6% incoming ow turbulent
intensity, were used as inputs in the CFD model.
Fig. 4 shows the model conguration and its computa-
tional domain, which contains 821,364 hexahedral cells.
The general boundary condition settings are the same as
those in the at UTC model except for the height of the
Solar simulator
HWA probe
HWA frame
Suction fan
Wood plate
Honeycombs & grids
Supply fan
Jet exit
Thermocouples
Fig. 3. Experimental setup for corrugated UTC (ow direction: left to right).
302 S. Li et al. / Solar Energy 91 (2013) 297315
upper domain that is extended to 0.6 m for consistency
with the experimental setup.
3. Model validation
The CFD model validation is based on two aspects,
namely the airow and the thermal eld. For the at
UTC model, only the non-isothermal case is considered
due to the limitation in the availability of experimental
data. For the corrugated UTC model, both isothermal
and non-isothermal conditions have been modeled. Also
an initial generic case of an isothermal at solid plate sub-
jected to a 2-D plane wall jet has been investigated to pro-
vide a more complete validation.
3.1. Thermal eld validation for at UTC under uniform
approach ow
The surface temperature at the center of the plate and
the average air temperature at the cavity exit are computed
using dierent RANS models and compared with the
experimental data, as shown in Fig. 5. For the plate surface
temperature, the agreement between the CFD results and
experimental data are satisfactory for all the models tested
(error within 2 K). The Standard ke and the RNG ke
models give similar results to the Realizable ke model
but are more stable in terms of convergence and have a
more consistent performance for all cases considered (error
within 1.2 K). The performance of the SST kx model is
dierent from all the ke models, with the computed sur-
face temperature slightly overestimated, but still accept-
able. For the cavity outlet temperature, all the models
provided identical results, except the Realizable ke model,
particularly for congurations with low suction ow rate.
Thus, based on this comparison, both the Standard ke
and the RNG ke model can provide suciently accurate
results for thermal modeling.
3.2. Flow eld validation for at solid plate subjected to the
plane wall jet
Unlike cases with uniform approach ow conditions, the
interaction of large-scale turbulence eddies in the outer
layer with smaller scale eddies in the inner layer create a
complicated ow that determines the development of the
wall jet. Therefore, an initial evaluation is conducted to
ensure that the wall jet is properly developed in the
stream-wise direction. The average stream-wise velocity at
the jet exit is 7.2 m/s, with the uniformity in the span-wise
and vertical direction being within 4%. The uncertainty
of surface temperature measurement using thermocouples
is 0.5 K. Using air at an approximate room temperature,
a nominal inlet Re around 2.6 10
4
is calculated based on
the jet exit height (b). The main measurement series are
taken at the following stream-wise positions: x = 18, 438,
Exit of wall jet,
velocity inlet
(Non-isothermal:
8.59m/s;
Isothermal: 5.89m/s)
Wood plate,
adiabatic wall
Back wall of the
cavity, adiabatic wall
Top of the domain,
symmetry
Outlet of the cavity,
velocity inlet
(392kg/hr)
Outlet of the exterior domain,
pressure outlet
Corrugated plate, wall with heat
flux
(1029W/m
2
)
Perforations, interior
Fig. 4. Corrugated UTC model conguration.
311
313
315
317
319
311 313 315 317 319
P
l
a
t
e

t
e
m
p
e
r
a
t
u
r
e

-
C
F
D

[
K
]
Plate temperature - Experiment [K]
Experimental data ERROR: +/-1K ERROR: +/-2K
STD k- RNG k- SST k-
Realizable k-
303
305
307
309
311
303 305 307 309 311
C
a
v
i
t
y

e
x
i
t

a
i
r

t
e
m
p
e
r
a
t
u
r
e

-
C
F
D

[
K
]
Cavity exit air temperature - Experiment [K]
Experimental data ERROR: +/-1K ERROR: +/-2K
STD k- RNG k- SST k-
Realizable k-
(a)
(b)
Fig. 5. Comparison between CFD results and experimental data from the
literature for at UTC model: (a) plate surface temperature and (b) cavity
exit air temperature.
S. Li et al. / Solar Energy 91 (2013) 297315 303
1033, and 1633 mm (that is, x/b = 0.4, 8.0, 18.8, and 29.7),
corresponding to the crests of the corrugated UTC.
Fig. 6 shows the stream-wise mean velocity proles from
x/b = 8.0 to x/b = 29.7. The classical outer scaling is used,
where U
m
is the maximum velocity in the vertical prole,
m/s; and y
1/2
is the height where the velocity is half of
U
m
, m. Two sets of experimental data from the literature
(Wygnanski et al., 1992, using Hot-Wire Anemometry
(HWA) and Eriksson et al., 1998, using Laser Doppler
Velocimetry (LDV)), at similar Re, have been used to
assess the delity of the experimental results. Based on
the comparison, it may be seen that, except for the location
x/b = 8.0, which may still be within the development region
of the wall jet, the present experimental data match very
well with those published in the literature. The normalized
CFD results at position x/b = 18.8 are shown in Fig. 6,
while satisfactory agreement with the experimental data
(from the present study as well as those found in the liter-
ature) was obtained for all positions, with the modeling
error calculated to be within 5%.
In most of the wall jet literature, prole comparisons
have only been made after normalization due to the vari-
able experimental conditions and the limited amount of
data provided. However, the actual velocity proles
obtained from the CFD simulations and the measurements
are plotted in Fig. 7 without scaling in order to achieve a
more robust comparison. Near the jet exit (x/b = 0.4) the
satisfactory agreement between the simulation results and
the experimental data reects the proper boundary condi-
tion settings in the simulation. Generally, the simulated
vertical proles of stream-wise velocity decay faster with
stream-wise distance than the measured proles, but the
predictions improve at greater distance from the nozzle.
Among all the models that have been tested, the highest
modeling error is about 20% and, in most cases, the error
is within the 10% measurement uncertainty. The Standard
ke model provides the best prediction, but the results from
the other ke models are also suciently accurate.
The turbulent kinetic energy (TKE) predicted from the
CFD simulations has also been compared with the experi-
mental data, as shown in Fig. 8. It can be seen that the sim-
ulation results for the TKE are underestimated in the near
wall region, but overestimated in the outer part of the wall
jet. The Standard ke model is slightly better than the other
models in terms of accuracy with an average modeling
error of 20%. A similar dierence between the CFD and
experimental results was also observed in the study by
Balabel and El-Askary (2011), in which two ke models
and one t
2
f model were selected for computing the nor-
malized Reynolds stresses in a wall jet and the average dis-
crepancy was around 30%. Pajayakrit and Kind (2000) also
tested the Standard ke and the SST kx model for wall
jets and reported that the modeling error for TKE was
around 30%. Therefore, it may be concluded that the Stan-
dard ke model provides better predictions in simulating
the plane wall jet compared to the other models. This is
consistent with the ndings of the study by Klinzing and
Sparrow (2009), which tested the same ve turbulent mod-
els for a plane wall jet and a cylindrical wall jet.
3.3. Airow and thermal eld validation for a corrugated
UTC subjected to the plane wall jet
Both isothermal and non-isothermal conditions were
considered in the simulation for the corrugated UTC
beneath the plane wall jet. In the isothermal case, the core
region speed at the jet exit is 5.9 m/s with a turbulence
intensity of 5.6% and a suction velocity of 0.033 m/s. The
vertical proles of stream-wise velocity and turbulence
kinetic energy at six crests are computed with all the RANS
models and compared with the experimental results, as
shown in Figs. 9 and 10. In Fig. 9, the agreement between
the CFD simulation results and experimental data is very
good in the near wall region (within 0.03 m), which governs
the heat transfer process from the plate, especially for the
RNG ke model that provides predicted values within the
10% measurement uncertainty range. Further away from
the wall, as in the simulation of the wall jet over a solid at
plate, all of the models underestimate the jet growth in the
vertical direction. The possible explanation could be that
air entrainment due to the lab air movement (it was an
unconned experiment) cannot be captured in the more
idealized CFD model. Also, inherent deciencies exist in
the steady RANS models when it comes to modeling jet
growth rates, e.g. Boussinesqs isotropic eddy viscosity
assumption is not valid in shear layers, while additional
complexities are associated with the turbulence generated
at corrugated permeable surfaces (Greig et al., 2012).
Therefore, the prediction with 20% modeling error, given
by the tested models except the Realizable ke model, is
considered acceptable.
0.0
0.2
0.4
0.6
0.8
1.0
0.0 0.2 0.4 0.6 0.8 1.0 1.2
y
/
y
1
/
2
U/U
m
Experimental data x/b = 8.0
Experimental data x/b = 18.8
Experimental data x/b = 29.7
Standard k- at x/b = 18.8
Realizable k- at x/b = 18.8
SST k- at x/b = 18.8
RNG k- at x/b = 18.8
RSM at x/b = 18.8
Wygnanski 1992 (30b < x < 140b)
Eriksson 1998 (40b < x < 150b)
Fig. 6. Comparison between CFD results, experimental data and litera-
ture data for the jet ow: stream-wise scaled mean velocity proles of the
jet.
304 S. Li et al. / Solar Energy 91 (2013) 297315
Fig. 10 shows that the general shapes of TKE proles
predicted by the CFD simulations match with the experi-
mental data, although there are quantitative dierences.
Among all the RANS models tested, none of the turbu-
lence models can provide predictions within the measure-
ment uncertainty (14%), but the prediction from the
RNG ke model is almost within the 20% modeling error
range. The results from the Standard ke and SST kx
models deviate more than the RNG ke model (up to a
30% dierence from the experimental data). In the
stream-wise direction, the RSM prediction diered greatly
from the experimental data (more than 30% for most posi-
tions). Therefore, although RSM involves a more elaborate
treatment of turbulence (namely the transport of Reynolds
stresses) at a much higher computational cost compared to
the two-equation models, it does not result in more accu-
rate predictions, for the cases considered here. The abnor-
mal behavior of the Realizable ke model could be due to
two reasons. Firstly, this model was developed for high Re
ow and so, it may not be able to resolve ows with suc-
tion, which have a wide Re range. Secondly, there is an
extra non-physical rotation term in the turbulent viscosity
equation, which makes the model more adaptive for ows
with rotation or strong adverse pressure gradients. So for
the ow in the corrugated UTC, which has some curvature
due to the geometry but with zero or small adverse pressure
gradients, this extra rotation term will result in errors (Shih
et al., 1995).
In the non-isothermal case, the core region air velocity
at the jet exit is 8.59 m/s, with turbulence intensity of
5.8%, and 0.03 m/s suction velocity. The surface plate tem-
perature at each crest (3rd10th) and the cavity outlet air
temperature were computed and compared with the exper-
imental data as shown in Fig. 11. Based on the investiga-
tions for the isothermal case, the Realizable ke model
was deemed to be inaccurate and so it was not considered
0.00
0.02
0.04
0.06
y

[
m
]
U [m/s]
x/b=0.4
Experimental data STD k- RNG k-
SST k- Realizable k- RSM
0.00
0.02
0.04
0.06
y

[
m
]
U [m/s]
x/b=8.0
Experimental data STD k- RNG k-
SST k- Realizable k- RSM
0.00
0.02
0.04
0.06
y

[
m
]
U [m/s]
x/b=18.8
Experimental data STD k- RNG k-
SST k- Realizable k- RSM
0.00
0.02
0.04
0.06
0 2 4 6 8 10 0 2 4 6 8 10
0 2 4 6 8 10 0 2 4 6 8 10
y

[
m
]
U [m/s]
x/b=29.7
Experimental data STD k- RNG k-
SST k- Realizable k- RSM
Fig. 7. Comparison between CFD results and experimental data for the jet ow: stream-wise mean velocity proles without scaling of the jet (error bars:
10%).
S. Li et al. / Solar Energy 91 (2013) 297315 305
here. From Fig. 11, it can be concluded that all the three
RANS models tested perform well (errors < 2 K), among
which the RNG ke model shows the best performance
(error < 1 K). One explanation for the superior perfor-
mance of the RNG ke model would be the extra term in
the e-transport equation (Choudhury, 1993), which makes
the model more responsive to the eect of rapid strain
and streamline curvature. Also, the turbulent viscosity is
calculated dierently for high and low Re ows, which
makes the model more adaptive to this specic problem.
Hence, it may be concluded that, although the Standard
ke model has slightly better performance in plane wall jet
simulation, for the two cases considered, the RNG ke
model provides more accurate results for UTC modeling,
which is the main subject of this study. Also, it has previ-
ously been shown that the RNG ke model is more capable
of accurately modeling the characteristics of airow and
heat transfer within the cavities formed by closely spaced
heated horizontal plates (Turgut and Onur, 2007), which
is another feature of UTC system modeling. Therefore,
the RNG ke model has been selected for the following dis-
cussion and analysis.
4. Airow and thermal analysis of at and corrugated UTCs
This section presents results from CFD simulations for
at and corrugated UTCs which aim to provide insights
into the convective heat transfer process and to further
advance the development of innovative UTC designs and
performance prediction models. To better represent the
actual approaching ow conditions experienced by build-
ing-integrated UTCs without dealing (at this stage) with
the complicated local airow features (the building is a
blu body immersed in the atmospheric boundary layer),
the following discussion is based on CFD simulations for
uniform approach ow conditions. The main parameters
0.00
0.02
0.04
0.06
y

[
m
]
TKE [m
2
/s
2
]
x/b=0.4
Experimental data STD k- RNG k-
SST k- Realizable k- RSM
0.00
0.02
0.04
0.06
y

[
m
]
TKE [m
2
/s
2
]
x/b=8.0
Experimental data STD k- RNG k-
SST k- Realizable k- RSM
0.00
0.02
0.04
0.06
y

[
m
]
TKE [m
2
/s
2
]
x/b=18.8
Experimental dat STD k- RNG k-
SST k- Realizable k- RSM
0.00
0.02
0.04
0.06
0 1 2 3 0 1 2 3
0 1 2 3 0 1 2 3
y

[
m
]
TKE [m
2
/s
2
]
x/b=29.7
Experimental data STD k- RNG k-
SST k- Realizable k- RSM
Fig. 8. Comparison between CFD results and experimental data for the jet ow: turbulent kinetic energy proles without scaling of the jet (error bars:
14%).
306 S. Li et al. / Solar Energy 91 (2013) 297315
considered in the present study are the UTC geometrical
conguration (at or corrugated), the suction velocity
and the turbulence intensity of the approaching ow.
Examining the impact of corrugation geometry and perfo-
ration dimension is beyond the scope of this study. Li and
Karava (2012) carried out a boundary layer analysis for a
0.5 m long at UTC, with a 6 m/s approach ow and suc-
tion velocities from 0.0448 to 0.0688 m/s, and the results
0.00
0.02
0.04
0.06
0 1 2 3 4 5 6 7
y

[
m
]
U [m/s]
x/b=2.6
Experimental data STD k- RNG k-
SST k- Realizable k- RSM
0.00
0.02
0.04
0.06
0 1 2 3 4 5 6 7
y

[
m
]
U [m/s]
x/b=5.3
Experimental data STD k- RNG k-
SST k- Realizable k- RSM
0.00
0.02
0.04
0.06
0 1 2 3 4 5 6 7
y

[
m
]
U [m/s]
x/b=8.0
Experimental data STD k- RNG k-
SST k- Realizable k- RSM
0.00
0.02
0.04
0.06
0 1 2 3 4 5 6 7
y

[
m
]
U [m/s]
x/b=16.0
Experimental data STD k- RNG k-
SST k- Realizable k- RSM
0.00
0.02
0.04
0.06
0 1 2 3 4 5 6 7
y

[
m
]
U [m/s]
x/b=24.1
Experimental data STD k- RNG k-
SST k- Realizable k- RSM
0.00
0.02
0.04
0.06
0 1 2 3 4 5 6 7
y

[
m
]
U [m/s]
x/b=29.7
Experimental data STD k- RNG k-
SST k- Realizable k- RSM
Fig. 9. Comparison between CFD results and experimental data for the corrugated UTC model: stream-wise mean velocity proles isothermal condition
(error bars: 10%).
S. Li et al. / Solar Energy 91 (2013) 297315 307
showed that the thermal boundary layer thickness and the
local convective heat transfer between the plate and ambi-
ent air tends to remain constant within the tested plate
length (0.5 m). In order to better understand the thermal
performance, the present study includes a longer, at
UTC model, which is extended to 2 m, and a 1.73 m long
0.00
0.02
0.04
0.06
0 1 2 3 4
y

[
m
]
TKE [m
2
/s
2
]
x/b=2.6
Experimental data STD k- RNG k-
SST k- Realizable k- RSM
0.00
0.02
0.04
0.06
0 1 2 3 4
y

[
m
]
TKE [m
2
/s
2
]
x/b=5.3
Experimental data STD k- RNG k-
SST k- Realizable k- RSM
0.00
0.02
0.04
0.06
0 1 2 3 4
y

[
m
]
TKE [m
2
/s
2
]
x/b=8.0
Experimental data STD k- RNG k-
SST k- Realizable k- RSM
0.00
0.02
0.04
0.06
0 1 2 3 4
y

[
m
]
TKE [m
2
/s
2
]
x/b=16.0
Experimental data STD k- RNG k-
SST k- Realizable k- RSM
0.00
0.02
0.04
0.06
0 1 2 3 4
y

[
m
]
TKE [m
2
/s
2
]
x/b=24.1
Experimental data STD k- RNG k-
SST k- Realizable k- RSM
0.00
0.02
0.04
0.06
0 1 2 3 4
y

[
m
]
TKE [m
2
/s
2
]
x/b=29.7
Experimental data STD k- RNG k-
SST k- Realizable k- RSM
Fig. 10. Comparison between CFD results and experimental data for the corrugated UTC model: turbulent kinetic energy proles isothermal condition
(error bars: 14%).
308 S. Li et al. / Solar Energy 91 (2013) 297315
corrugated UTC model with the same 0.95 m width cavity.
The simulated free stream approaching ow velocity is
3 m/s with two values of turbulence intensity (TI): 0.1%,
to represent low turbulence intensity and uniform
approach ow, and 20%, which is closer to that in the
atmospheric boundary layer. The suction velocity is set to
be 0.01 m/s and 0.06 m/s to represent low and high suction
conditions, respectively. Although the ow may remain
attached to the corrugated plate under certain conditions
with low approaching velocity (less than 1 m/s), these cases
are not considered in the present study. The heat ux on
the UTC plate is set to be 1000 W/m
2
and the temperature
of the approaching ow is 300 K. The cases studied are
listed in Table 1.
Due to the lowporosity (0.7%for the at UTCmodel and
0.8%for the corrugated UTCmodel) the plate is divided into
two parts in the stream-wise direction: a perforated row and
a solid row, as shown in Fig. 2b, in order to better examine
the impact of the perforations. The perforated row can be
considered as a discrete suction surface, which has widely
spaced perforations, so that the results may be compared
with those for a classic case with homogeneous suction.
The solid row covers more than 90% of the plate area and
is similar to a smooth, impermeable plate.
4.1. Thermal boundary layer development
The dierence between the perforation and solid row will
be discussed in the next section and therefore, Fig. 12 only
shows the thermal boundary layer development of the perfo-
rated rows along the stream-wise direction. As displayed in
Fig. 1, the corrugations can be divided into four regions:
crest, valley, windward and leeward slope. The vertical
velocity and temperature proles used to calculate the
boundary layer thickness are computed along the crest
regions due to the fact that the corrugation amplitude is
smaller than the boundary layer thickness, which makes
the corrugated plate more like a rough at plate, and the
crests are longer than the other corrugation regions. Also,
the simulation results show that the air trapped in recircula-
tion regions (valleys) will help the approaching air to ow
smoothly over the valleys, instead of causing sudden velocity
drops, as occurs in the classical cavity skimming ow
regime that is, for example, associated with narrow street
canyons (Grimmond and Oke, 1999).
In the literature, it has been noted that the asymptotic
boundary layer is laminar or turbulent depending on the
suction rate, but also on the surface porosity and whether
the upstream ow is laminar or turbulent. Dutton (1958)
tested both nylon fabric and a perforated plate, which
are the closest examples of homogeneous suction, and
found that the turbulent boundary layer reaches a constant
thickness after a longer starting length for a suction ratio
V
s
/U
1
> 0.0073. Also, Moat and Kays (1984) indicated
that for homogeneous suction with V
s
/U
1
> 0.004 the tur-
bulent boundary layer would revert to an asymptotic lam-
inar boundary layer. It can be seen from Fig. 12a with low
suction velocity (Cases 1-1 and 1-2 for which the velocity
ratio is 0.0033), that there is no sign that the boundary
layer will develop into an asymptotic prole after the mod-
eled length for either the at or the corrugated UTC. How-
ever, the results dier greatly when suction velocity is
increased by a factor of six (Cases 2-1 and 2-2). One may
expect that it will be easier for the boundary layer to
develop into the asymptotic prole for at UTCs, but the
results show that the boundary layer thickness keeps
306
308
310
312
314
316
306 308 310 312 314 316
T
e
m
p
e
r
a
t
u
r
e

-
C
F
D

[
K
]
Temperature -Experiment [K]
Experimental data ERROR:+/-1K
ERROR:+/-2K STD k- - Plate temperature
RNG k- - Plate temperature SST k- - Plate temperature
STD k- - Exit air temperature RNG k- - Exit air temperature
SST k- - Exit air temperature
Fig. 11. Comparison between CFD results and experimental data for the
corrugated UTC model: crest surface temperature and cavity exit air
temperature (non-isothermal condition).
Table 1
Details of the cases studied.
Case
number
Approaching ow Suction velocity
V
s
(m/s)
Suction ratio
(V
s
/U
1
)
Temperature
(K)
Heat ux on
plate (W/m
2
)
Velocity
U
1
(m/s)
Turbulence
intensity (%)
1-1
300 1029
3
0.1
0.01 0.003
1-2 20
2-1 0.1
0.06 0.020
2-2 20
3-1 1 0.1
0.02
0.020
4-1 6 0.1 0.003
S. Li et al. / Solar Energy 91 (2013) 297315 309
increasing even for the high suction velocity case
(V
s
/U
1
= 0.02), which is approximately an order of magni-
tude larger than in previous studies of homogeneous suc-
tion. For the corrugated UTC, the boundary layer
development with low incident TI (Cases 1-1 and 2-1)
approaches a stable asymptotic prole; whereas, with high
incident TI, asymptotic behavior is only seen with an
accompanying high suction ratio (Case 2-2). The possible
explanation for the dierences in boundary layer proles
between the at and corrugated UTCs could be that the
valleys between the corrugations act in a similar manner
to a large perforation, creating recirculation regions
which trap the air, resulting in an asymptotic prole (as dis-
cussed later in Section 4.4). Also, the suction process would
be more eective due to the dierent suction directions
caused by the perforations on the corrugation slopes.
Therefore, considering the general suction velocity range
in UTC systems, typically 0.010.06 m/s, the homogeneous
suction assumption is not applicable for either at or cor-
rugated UTCs, which implies that the convection heat loss
cannot be ignored in the energy simulations. Two more
cases for corrugated UTCs (also listed in Table 1) have
been investigated, with the results shown in Fig. 12b, to
further examine the impact of the suction ratio. It can be
seen that in Case 4-1, which has a small suction ratio of
0.033 and a suction velocity value of 0.02 m/s, the bound-
ary layer thickness is thinner due to the high approaching
velocity and keeps increasing with distance. However, for
Case 3-1, the suction ratio is 0.02, the same as for Case
2-1, but the boundary layer prole is similar to that for
Case 1-1, which keeps growing with distance due to the
low suction velocity. Thus, in UTCs it is the suction veloc-
ity itself, rather than the suction ratio typically considered
in studies with homogeneous suction, that has a more pro-
found impact on the boundary layer proles.
4.2. Plate surface temperature
Fig. 13 shows the crest surface temperature of the perfo-
rated row along the stream-wise direction. Due to the
unique geometry of the corrugated UTC, the plate temper-
ature may vary in dierent corrugation regions, i.e. crests,
valleys, windward and leeward slopes. For this particular
corrugation geometry, the crest surface temperature is
more dominant in the surface-averaged temperature for
the whole corrugation as the crest length is much longer
than the other regions and it directly faces the approaching
ow. Based on the plate temperatures shown in Fig. 13 for
the corrugated UTC, the variation of temperature is within
1 K after x = 1.1 m under low suction rate and for high
suction rate the variation of temperature becomes negligi-
ble after x = 0.5 m, both irrespective of the level of incident
TI; thus it can be assumed to be uniform along the plate.
Also, the simulation results show that the average surface
(a)
(b)
0
0.02
0.04
0.06
0.08
0.1
0.12
0 0.5 1 1.5 2

t
[
m
]
x [m]
Case 1-1
Case 1-2
Case 2-1
Case 2-2
0
0.02
0.04
0.06
0.08
0.1
0.12
0 0.5 1 1.5 2

t
[
m
]
x [m]
Case 1-1
Case 1-2
Case 2-1
Case 2-2
Case 3-1
Case 4-1
Fig. 12. Thermal boundary layer development along the stream-wise
direction of the perforated rows: (a) at UTC and (b) corrugated UTC.
(a)
(b)
315
320
325
330
335
340
345
350
0 0.5 1 1.5 2
P
l
a
t
e

t
e
m
p
e
r
a
t
u
r
e

[
K
]
x [m]
Case 1-1
Case 1-2
Case 2-1
Case 2-2
315
320
325
330
335
340
345
350
355
0 0.5 1 1.5 2
P
l
a
t
e

t
e
m
p
e
r
a
t
u
r
e

[
K
]
x [m]
Case 1-1
Case 1-2
Case 2-1
Case 2-2
Fig. 13. Plate surface temperature along the stream-wise direction of the
perforated rows: (a) at UTC and (b) corrugated UTC.
310 S. Li et al. / Solar Energy 91 (2013) 297315
temperature of windward and leeward slopes and valleys
can be maintained constant, after the same distance. More-
over, the results reveal that when the suction velocity is
low, the average surface temperature of the crest and lee-
ward slope is quite close, with the dierence being within
1 K, and the average valley surface temperature is about
34 K higher than the crest temperature. However, when
the suction velocity is high, the average crest temperature
is closer to the valley temperature (the dierence is within
1 K) and the leeward slope surface temperature is about
5 K lower than the crest. For both suction velocities, the
average windward slope temperature is about 10 K lower
than the crest due to the strong impingement.
In at UTCs, although the boundary layer will not
revert to an asymptotic prole in any of the cases studied
and the overall temperature uniformity is not as pro-
nounced as for the corrugated UTC conguration, the
plate temperature after x = 1 m remains almost constant
when the suction velocity is high. The possible explanation
is that along the stream-wise direction, the increasing suc-
tion ow rate in the cavity beneath the plate will increase
the convective heat ux from the interior surface to the
air in the cavity, compensating for the decreasing convec-
tive ux on the exterior side and, therefore, resulting in a
constant surface temperature. In other words, if the suction
ow rate in the cavity reaches a certain value, a constant
surface temperature is maintained. This can also be shown
by the temperature proles for low suction velocity, which
approach a constant value after a much longer distance
than in the high suction velocity cases. The capability of
maintaining a relatively constant surface temperature
along the plate for at UTCs or within dierent corruga-
tion regions for corrugated UTCs is an important feature,
unlike cavity walls with a single inlet, in which the surface
temperature will become too high and eventually the collec-
tor will lose its heating ability. This feature makes the
UTCs very suitable for integration with BIPV/T systems
to cool down the PV panels.
4.3. Cavity exit air temperature and thermal eciency
Fig. 14 shows the cavity exit air temperature and the
corresponding thermal eciency of the system. Since the
radiation has been excluded in all the CFD simulations,
the heat loss of the UTCs is mainly due to convection.
The simulated solar radiation falling on the UTC plate is
1023 W/m
2
and the plate absorptance is 0.95. Close obser-
vation of Figs. 1214, reveals that changing the suction
velocity from 0.01 to 0.06 m/s can eectively reduce the
thermal boundary layer thickness by about 33% for the at
UTC and 50% for the corrugated UTC, which indicates
more heat captured by the suction and less net ux of heat
into the free stream and, consequently, lower plate and exit
air temperatures (by 2025 K on average for both UTC
types), and a higher system thermal eciency g (by about
40%). Considering the fact that the thermal boundary layer
almost reaches the asymptotic prole at high suction rates,
the high thermal eciency (nearly 90%) can be explained
by the homogeneous suction theory, in which the convec-
tive heat loss can be considered negligible. The results indi-
cate that the convective heat loss is actually a large
contributor to the overall thermal eciency, especially
when the suction rate is low. This is consistent with the
nding that the at UTC generally has a higher thermal
eciency than the corrugated UTC for the same suction
rate, as the corrugation geometry will enhance the exterior
convective heat transfer on the plate surface and therefore
result in higher heat loss. Also, the dierence between the
at and corrugated UTCs tends to decrease with the
increase of suction velocity, i.e. it is more than 10% when
the suction velocity is 0.01 m/s and diminishes to 3% for
0.06 m/s suction velocity. Therefore, optimizing the suction
velocity for dierent surface geometry can lead to better
overall thermal performance.
4.4. Airow eld
Fig. 15 shows the vertical velocity proles around the
center point (x = 1.0 m) for all the test cases for both the
at and corrugated UTCs, in which P stands for perforated
row and S represents the solid row. Generally, increasing
the incident TI will result in an enhancement of convective
heat transfer rate, due to the rise in the velocity, and a cor-
respondingly higher temperature gradient in the near wall
region, as shown in Fig. 15. Simonich and Bradshaw
(1978) reported that a variation of free stream turbulence
intensity (incident TI) from 0% to 6% led to a relative
increase of the heat transfer coecient by up to 30% for
an impermeable, at plate. A similar observation has also
been reported by Karava et al. (2011, 2012) for the forced
convective heat transfer from the inclined windward roof
of an isolated low-rise building. Also, one may expect that
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0
15
30
45
60
75
90
105
120
135
150
Case 1-1 Case 1-2 Case 2-1 Case 2-2
T
h
e
r
m
a
l

E
f
f
i
c
i
e
n
c
y
E
x
i
t

a
i
r

t
e
m
p
e
r
a
t
u
r
e

r
i
s
e

[
K
]
Flat UTC exit air temperature
Corrugated UTC exit air temperature
Flat UTC efficiency
Corrugated UTC efficiency
Fig. 14. Cavity exit air temperature and UTC thermal eciency (bars
represent the exit air temperature and dots represent the thermal
eciency).
S. Li et al. / Solar Energy 91 (2013) 297315 311
the boundary layer thickness will be larger for higher inci-
dent TI.
As shown in Figs. 1215, for the at UTC, minor
impacts of varying the incident turbulence intensity from
0.1% to 20% can be detected in the thermal boundary
thickness, plate and exit air temperature, thermal eciency
and vertical velocity proles, but these are not signicant
when compared to those of the suction velocity and perfo-
rations, especially for the high suction velocity cases. How-
ever, when referring to the corrugated UTC, a signicant
impact in the thermal and airow eld can be observed
when varying the incident turbulence intensity. Stel et al.
(2010) carried out numerical and experimental analysis of
turbulent ow in d-type corrugated pipes and the results
showed that the core ow squeezes the vortices conned
inside the d-type corrugations, producing a signicant tur-
bulent interaction between the core ow and cavity vortex
along the corrugated interface line, which increased the
TKE, the Reynolds shear stress and, consequently, the
overall shear stress along the corrugated plate. The contour
plots of TKE from the same study, as well as the measure-
ments of normal velocity and Reynolds stress at dierent
corrugation locations, also conrmed that the turbulent
intensities increased near the corrugations. Also, Greig
et al. (2012) investigated the ow dynamics in a channel
with a corrugated surface and reported that, even at Rey-
nolds numbers within the conventional laminar range,
strong turbulence was observed in the channel, which was
almost entirely produced around the valleys and was
greatly intensied in the vicinity of the corrugation wave-
form, due to the interaction between the bursting ow orig-
inating from the valley, sweeping ow from the bulk region
and the vortex shedding o the crest; in addition, the heat
ux from the corrugated plate further enhances the turbu-
lence intensity in the free stream as well as in the near wall
region (Greig et al., 2013). Therefore, the corrugations,
together with the plate heating, will result in turbulence
production at scales of the order of the corrugation height
that enhances the overall level of boundary layer turbu-
lence near the wall. Although a higher suction velocity
can overcome the eects of turbulence to a certain extent,
the thermal eciency is still about 4% lower for both the
low and high suction velocity cases, due to the high turbu-
lence intensity.
Regarding the impact of perforations, Fig. 15 shows that
compared to the exponential curve fromEq. (2) for homoge-
neous suction, the velocity boundary layer thickness is much
larger for both UTCs. For the at UTC, there is almost no
dierence between the perforated row and the solid row
for low suction velocity, but this is not the case for higher
suction velocity. However, for the corrugated UTC, the
velocity proles are identical for both the perforated row
and the solid row. This observation implies that the existence
of perforations has a more profound impact on the at UTC
than on the corrugated UTC. This nding is consistent with
previous studies, in which the perforation dimensions are
included as a key parameter in every Nu-correlation devel-
oped for at UTC (Kutscher, 1992; Van Decker et al.,
2001; Arulanandam et al., 1999), but are disregarded in the
Nu-correlations for corrugated UTC (Gawlik, 1993).
5. Conclusions
High-resolution, 3-dimensional, steady, RANS CFD
models have been developed for both at and corrugated
UTCs around 2 m long. Experimental data from the liter-
ature and a specially designed experimental set-up in a
solar simulator have been used to evaluate the performance
of ve turbulence closure models. It was found that the
RNG ke model provides the most accurate and consistent
results, with economic computing cost, for all the tested
cases, namely a heated at UTC under uniform approach
(a)
(b)
0
0.01
0.02
0.03
0.04
0.05
0 0.2 0.4 0.6 0.8 1
y

[
m
]
U/U

Flat-case 1-1-P
Flat-case 1-1-S
Flat-case 1-2-P
Flat-case 1-2-S
Exponential curve from Eqn. 2
Corrugated-case 1-1-P
Corrugated-case 1-1-S
Corrugated-case 1-2-P
Corrugated-case 1-2-S
0
0.01
0.02
0.03
0.04
0.05
0 0.2 0.4 0.6 0.8 1
y

[
m
]
U/U

Flat-case 2-1-P
Flat-case 2-1-S
Flat-case 2-2-P
Flat-case 2-2-S
Exponential curve from Eqn. 2
Corrugated-case 2-1-P
Corrugated-case 2-1-S
Corrugated-case 2-2-P
Corrugated-case 2-2-S
Fig. 15. Vertical velocity proles for at and corrugated UTC at
x = 1.0 m: (a) suction velocity = 0.01 m/s and (b) suction veloc-
ity = 0.06 m/s (P: Perforated row and S: Solid row).
312 S. Li et al. / Solar Energy 91 (2013) 297315
ow, an isothermal at impermeable plate and a corru-
gated UTC, isothermal rst and then heated, both sub-
jected to a plane wall jet ow. The Standard ke and SST
kx model can provide moderate accuracy, but not as
good as the RNG ke model. The Realizable ke and the
RSM are not recommended for simulating UTCs because
of their unsatisfactory performance in this study.
Based on the CFD model validation results, a paramet-
ric study has been conducted with the RNG ke model for
dierent suction ow rates and free stream turbulence
intensity for a constant 3 m/s approach ow velocity.
Approach ow velocities of 1 and 6 m/s are studied in addi-
tion to the reference case, in order to examine the inuence
of the suction ratio. The combined impacts of the perfora-
tions and corrugations have also been considered. Valuable
conclusions can be drawn which are useful for the design
process and performance analysis of UTCs:
Based on the thermal boundary layer analysis, it is shown
that increasing the suction rate can help maintain an
asymptotic prole, thus allowing application of the
homogeneous suction theory to UTC systems and
thereby, greatly simplifying the performance analysis
process. However, the necessary suction rate may be too
high for practical applications, considering the large fan
power that will be needed and the fact that electricity is
more valuable than heat. Moreover, the results showthat
the corrugated UTCcan maintain the asymptotic bound-
ary layer prole at relatively lower suction rates com-
pared to the at UTC.
An important feature is the capability of maintaining a
relatively constant surface temperature along the plate
for at UTCs or within dierent corrugation regions
for corrugated UTCs.
Based on the thermal eciency analysis presented, the
convection heat loss is a large contributor, i.e. approxi-
mately 60% for a suction velocity of 0.01 m/s and 10%
for a suction velocity of 0.06 m/s. The at UTC can pro-
vide higher thermal eciency than the corrugated UTC
for the geometry considered here.
The suction velocity, rather than the suction ratio of V
s
/
U
1
, has been found to be the most important parameter
in the present study. The impact of other parameters, i.e.
incident turbulence intensity and surface geometry (at
or corrugated), can be potentially controlled by optimiz-
ing the suction velocity.
In addition to the suction velocity, another important
parameter for the at UTC is the existence of perfora-
tions, which is more important than the turbulence inten-
sity of the approach ow. For the corrugated UTC, the
incident turbulence intensity is more crucial than the per-
forations, as the interaction between the corrugations and
the incoming owwill result in an enhancement of bound-
ary layer turbulence near the wall.
In the present study, one type of corrugation geometry
has been considered while the impact of perforation size
and corrugation aspect ratio should be included in future
work. Based on the outcomes, appropriate modeling reso-
lution for system design and performance prediction can be
established in the development of energy models for UTCs.
Acknowledgements
The authors are grateful to Professor Andreas Athienitis
(Concordia University, Canada) for providing access to the
solar simulator and to Mr. James Bambara for providing
valuable insights during the experimental set-up design.
Appendix A. Energy balance equations in Fig. 1
The energy balance shown in Fig. 1 can be described by:
G Q
loss
Q
gain
A:1
where Q
loss
is:
Q
loss
Q
rad:ext
Q
rad:int
Q
conv:ext
qV
s
C
p
T
s
T
1
A:2
where T
s
is the air temperature through the perforations,
K. The last term on the right hand side represents the heat
captured by the suction and T
s
is higher than the ambient
air temperature as the air has been already heated up by the
convective heat transfer on the exterior side of the UTC
plate. Correspondingly, Q
gain
is calculated by:
Q
gain
Q
conv:int
_ mC
p
T
s
T
1
A:3
Appendix B. The Reynolds-Averaged NavierStokes
equations used in CFD simulations
In RANS-based modeling approach, the solution vari-
ables in the instantaneous NavierStokes equations are
decomposed into the mean (ensemble-averaged or time-
averaged) and uctuating components. For the velocity
components:
u
i
u
i
u
0
i
B:1
where u
i
and u
0
i
are the mean and uctuating velocity com-
ponents (i = 1, 2, 3).
Likewise, for pressure and other scalar quantities:
/

/ /
0
B:2
where / denotes a scalar such as pressure, energy or species
concentration. Substituting expressions of this form for the
ow variables into the instantaneous continuity and
momentum equations and taking a time average yields
the ensemble-averaged momentum equations in Cartesian
tensor form as:
@q
@t

@
@x
i
qu
i
0 B:3
S. Li et al. / Solar Energy 91 (2013) 297315 313
@
@t
qu
i

@
@x
j
qu
i
u
j

@p
@x
i

@
@x
j
l
@u
i
@x
j

@u
j
@x
i

2
3
d
ij
@u
k
@x
k
_ _ _ _

@
@x
j
qu
0
i
u
0
j
_ _
B:4
where q is the density of the uid, kg/m
3
; u
i
, u
j
and u
k
are
the mean velocity components with the average over bar
dropped in the equations to keep them simple (j = 1, 2, 3;
k = 1, 2, 3); l is the dynamic viscosity, kg/m s; d
ij
is the
Kronecker delta function, which is equal to 0 when i j
and equal to 1 when i = j. (B.3) and (B.4) are the Rey-
nolds-Averaged NavierStokes (RANS) equations. The
additional terms, qu
0
i
u
0
j
, are the Reynolds stresses that dis-
tinguish the RANS equations from the instantaneous Na-
vierStokes equations and must be modeled in order to
close Eq. (B.4). The Boussinesq hypothesis is commonly
used to relate the Reynolds stresses to the mean velocity
gradients:
qu
0
i
u
0
j
l
t
@u
i
@x
j

@u
j
@x
i
_ _

2
3
qk l
t
@u
k
@x
k
_ _
d
ij
B:5
The advantage of this approach is the relatively low
computational cost associated with the computation of
the turbulent viscosity l
t
. In the case of the ke and kx
models, two additional transport equations (for the turbu-
lence kinetic energy k, and either the turbulence dissipation
rate e, or the specic dissipation rate x) are solved and l
t
is
computed as a function of k and e or x. The disadvantage
of this approach is that it assumes l
t
is an isotropic scalar
quantity, which is not strictly true.
The alternative approach, embodied in RSM, is to solve
transport equations for each of the terms in the Reynolds
stress tensor. An additional scale-determining equation is
also required. Detailed information for each model can
be found in: Launder and Spalding (1972), Choudhury
(1993), Shih et al. (1995), Menter (1994) and Gibson and
Launder (1978).
References
Abulkhair, H., 2011. Thermal analysis of unglazed transpired solar
collectors. M.A.Sc. Thesis, Department of Mechanical Engineering,
University of Waterloo, Waterloo, Canada.
Abulkhair, H., Collins, M., 2010. Investigation of wind heat loss from
unglazed transpired solar collectors with corrugation. In: 5th Annual
International Green Energy Conference, Waterloo, Ontario, June 13,
2010.
Arulanandam, S.J., Hollands, K.G.T., Brundrett, E., 1999. A CFD heat
transfer analysis of the transpired solar collector under no-wind
condition. Solar Energy 67, 93100.
Athienitis, A.K., Bambara, J., ONeill, B., Faille, J., 2011. A prototype
photovoltaic/thermal system integrated with transpired collector. Solar
Energy 85, 139153.
Balabel, A., El-Askary, W.A., 2011. On the performance of linear and
nonlinear ke turbulence models in various jet ow applications.
European Journal of Mechanics B/Fluids 30, 325340.
Bambara, J., 2012. Experimental study of a fac ade-integrated photovol-
taic/thermal system with unglazed transpired collector. M.A.Sc.
Thesis, Department of Building, Civil and Environmental Engineering,
Concordia University, Montreal, Canada.
Bruun, H.H., Nabhani, N., Fardad, A.A., Al-Kaylem, H.H., 1990.
Velocity component measurements by X hot-wire anemometry.
Measurement Science and Technology 1 (12), 13141321.
Choudhury, D., 1993. Introduction to the Renormalization Group
Method and Turbulence Modeling. Fluent Inc. Technical, Memoran-
dum TM-107.
Delisle, V., 2008. Analytical and experimental study of a PV/thermal
transpired solar collector. M.A.Sc. Thesis, Department of Mechanical
Engineering, University of Waterloo, Waterloo, Canada.
Dutton, R.A., 1958. The eects of distributed suction on the development
of turbulent boundary layers. Aeronautical Research Council Techni-
cal Report, R&M No. 3155, London, England.
Eriksson, J.G., Karlsson, R.I., Persson, J., 1998. An experimental study of
a two-dimensional plane turbulent wall jet. Experiments in Fluids 25
(1), 5060.
Gawlik, K.M., 1993. A numerical and experimental investigation of
heat transfer issues in the practical utilization of unglazed,
transpired solar air heaters. Ph.D. Thesis, Department of Civil,
Environmental, and Architectural Engineering, University of Colo-
rado, Colorado, USA.
Gawlik, K.M., Kutscher, C.F., 2002. Wind heat loss from corrugated,
transpired solar collectors. Journal of Solar Energy Engineering 124,
256261.
Gibson, M.M., Launder, B.E., 1978. Ground eects on pressure uctu-
ations in the atmospheric boundary layer. Journal of Fluid Mechanics
86, 491511.
Golneshan, A.A., 1994. Forced convection heat transfer from low porosity
slotted transpired plates. Ph.D. Thesis, Department of Mechanical
Engineering, University of Waterloo, Waterloo, Canada.
Greig, D., Siddiqui, K., Karava, P., 2012. An experiment investigation of
the ow structure over a corrugated wave form in a transpired air
collector. International Journal of Heat and Fluid Flow 38, 133144.
Greig, D., Siddiqui, K., Karava, P., 2013. The inuence of surface heating
on the ow dynamics within a transpired air collector. International
Journal of Heat and Mass Transfer 56, 390402.
Grimmond, C.S.B., Oke, T.R., 1999. Aerodynamic properties of urban
areas derived from analysis of surface form. Journal of Applied
Meteorology 38, 12621292.
Iglisch, R., 1944. Exact calculation of laminar boundary layer in
longitudinal ow over a at plate with homogeneous suction. Schriften
der Deutschen Akademie der Luftfahrtforschung, Band 8B, Heft 1.
Translation: NACA TM No. 1205.
Karava, P., Jubayer, C., Savory, E., 2011. Numerical modelling of forced
convective heat transfer from the inclined windward roof of an isolated
low-rise building with application to photovoltaic/thermal systems.
Applied Thermal Engineering 31, 19501963.
Karava, P., Jubayer, C., Savory, E., Li, S., 2012. Eect of incident ow
conditions on convective heat transfer from the inclined windward
roof of a low-rise building with application to photovoltaicthermal
systems. Journal of Wind Engineering and Industrial Aerodynamics
104106, 428438.
Kay, J.M., 1948. Boundary Layer along a Flat Plate with Uniform
Suction. ARC-RM-2628.
King, L.V., 1914. On the convection of heat from small cylinders in a
stream of uid: determination of the convection constants of small
platinum wires, with applications to hot-wire anemometry. Proceed-
ings of the Royal Society A: Mathematical, Physical and Engineering
Sciences 90, 563570.
Klinzing, W.P., Sparrow, E.M., 2009. Evaluation of turbulence models for
external ows. Numerical Heat Transfer 55, 205228.
Kutscher, C.F., 1992. An investigation of heat transfer for air ow
through low porosity perforated plates. Ph.D. Thesis, Department of
Mechanical Engineering, University of Colorado at Boulder, Colo-
rado, USA.
314 S. Li et al. / Solar Energy 91 (2013) 297315
Kutscher, C.F., 1994. Heat exchanger eectiveness and pressure drop for
air ow through perforated plates with and without crosswind. Journal
of Heat Transfer 116, 391399.
Kutscher, C.F., Christensen, C., Barker, G., 1993. Unglazed transpired
solar collectors: heat loss theory. ASME Journal of Solar Engineering
115, 182188.
Launder, B.E., Spalding, D.B., 1972. Lectures in Mathematical Models of
Turbulence. Academic Press, London, England.
Li, S., Karava, P., 2012. Numerical study of convective heat transfer for
at unglazed transpired solar collectors. In: Proc. of High Performance
Buildings Conference, Purdue University, July 2012.
Menter, F.R., 1994. Two-equation eddy-viscosity turbulence models for
engineering applications. AIAA Journal 32, 15981605.
Moat, R.J., Kays, W.M., 1984. A review of turbulent-boundary-layer
research at Stanford, 19581983. Advances in Heat Transfer 16, 242
363.
Pajayakrit, P., Kind, R.J., 2000. Assessment and modication of two-
equation turbulence models. AIAA Journal 38, 955963.
PSE (Projects in Solar Energy), 2009. Indoor Test Stand for Solar
Thermal Collectors Technical Description. Freiburg, Germany.
Schlichting, H., Gersten, K., 2000. Boundary Layer Theory, eighth ed.
Springer, New York.
Shih, T.H., Liou, W.W., Shabbir, A., Yang, Z., Zhu, J., 1995. A new ke
eddy-viscosity model for high reynolds number turbulent ows
model development and validation. Computers Fluids 24 (3), 227238.
Simonich, J.C., Bradshaw, P., 1978. Eect of free-stream turbulence on
heat transfer through a turbulent boundary layer. Journal of Heat
Transfer 100, 671677.
Stel, H., Morales, R.E.M., Franco, A.T., Junqueira, S.L.M., Erthal, R.H.,
Gonc alves, M.A.L., 2010. Numerical and experimental analysis of
turbulent ow in corrugated pipes. Journal of Fluids Engineering 132,
071203-1-13..
Turgut, O., Onur, N., 2007. An experimental and three-dimensional
numerical study of natural convection heat transfer between two
horizontal parallel plates. International Communications in Heat and
Mass Transfer 34, 644652.
Van Decker, G.W. E., 1996. Asymptotic thermal eectiveness of unglazed
transpired plate solar air heaters. M.A.Sc. Thesis, Department of
Mechanical Engineering, University of Waterloo, Waterloo, Canada.
Van Decker, G.W.E., Hollands, K.G.T., Brunger, A.P., 2001. Heat-
exchange relations for unglazed transpired solar collectors with
circular holes on a square or triangular pitch. Solar Energy 71, 3345.
Wygnanski, I., Katz, Y., Horev, E., 1992. On the applicability of various
scaling laws to the turbulent wall jet. Journal of Fluid Mechanics 234,
669690.
S. Li et al. / Solar Energy 91 (2013) 297315 315

You might also like