Solar Energy Vol. 62, No. 2, pp.

69–77, 1998
© 1998 Elsevier Science Ltd
PII: S0038-092X(97)00119-9 All rights reserved. Printed in Great Britain
0038-092X/98 $19.00+0.00

FULL-SCALE MEASUREMENTS OF WIND-INDUCED

CONVECTIVE HEAT TRANSFER FROM A ROOF-MOUNTED
FLAT PLATE SOLAR COLLECTOR
S. SHARPLES and P. S. CHARLESWORTH
Building Science Research Unit, School of Architecture, University of Sheffield, Sheffield S10 2TN, U.K.

Received 6 June 1997; revised version accepted 27 October 1997

Communicated by BRIAN NORTON

Abstract—A series of full-scale measurements in the real environment were made to assess the magnitude
and variability of wind-induced convective heat transfer from a raised heated surface mounted directly
on to the pitched roof of a domestic size building. The heated plate had dimensions similar to those of
a typical flat-plate solar collector (1.81×0.89 m2), giving a Reynolds number range of approximately
1×105 to 5×105. The measured forced convective heat transfer coefficient h was correlated against the
w
wind speed V and the wind direction prevailing at the site of the building during the periods of measure-
ment. For a range of wind directions it was found that both power and linear relationships between h
w
and V adequately represented the experimental results. Some sheltering effects were observed when the
heated plate was leeward to the prevailing wind direction. Values of h measured in this study were in
w
good agreement with data derived from previous wind tunnel and field work. © 1998 Elsevier Science
Ltd. All rights reserved.

1. INTRODUCTION showed that the choice of the value for h was

w
of major importance when the collector had no
The overall thermal performance of a flat-plate
cover, as often occurs with swimming pool
solar collector is strongly influenced by the
heating installations, or just one cover, as is the
energy exchanges which take place between the
case for many domestic systems. For wind-
external environment and the uppermost surface
sensitive collectors it is necessary to decide upon
of the collector. For design or modelling
the most appropriate values of h to input into
purposes these energy exchanges are usually w
the equations for calculating the top loss
described in terms of a top surface heat loss
coefficient U and hence the collector efficiency
coefficient U , which combines both the radia- t
t g.
tive and convective heat transfer processes. The
Fundamental forced convective heat transfer
calculation of U can be achieved using thermal
t theory expresses the relationships between fluid
network analysis, but this approach involves
flow and convective heat transfer in terms of
solution by cumbersome iterative analysis. A
correlations between the dimensionless Nusselt,
more empirically based technique to estimate
Reynolds and Prandtl numbers (Nu, Re and
U was first proposed by Hottel and Woertz
t Pr, respectively). For the simplest cases of lami-
(1942), who produced a semi-empirical equa-
nar and turbulent parallel flows over a flat plate
tion for U . This approach has since been used
t of length L the theoretically derived relation-
and refined by several other workers ( Klein,
ships take the form (Incropera and DeWitt,
1975; Agarwal and Larson, 1981; Malhotra
1996):
et al., 1981; Garg and Datta, 1984; Morrison
and Gilliaert, 1992; Samdarshi and Mullick, laminar flow: N =0.664Re0.5Pr0.33 (1)
1994). The wind-induced convective heat L L
transfer from the uppermost surface of a flat- turbulent flow: N =0.037Re0.8Pr0.33. (2)
plate solar collector can play a significant role L L
in determining the magnitude of U and, there- The parameter of interest to this study, h , can
t w
fore, the overall efficiency of the collector. Some be derived from rearranging eqns (1) and (2)
parametric analyses have been undertaken in and substituting the various fluid property
order to estimate the impact of the wind- values for air. For example, a collector upper
induced convective energy losses on U and a surface measuring 1.81×0.89 m2 (the dimen-
t
collector’s performance ( Watmuff et al., 1977; sions of the heated plate used in this study) and
Ramsey and Charmchi, 1980). The studies at an isothermal temperature 25 K above an
69
70 S. Sharples and P. S. Charlesworth

ambient temperature of 300 K gives small rectangular cassettes containing naphtha-

lene to form a smooth flat surface. The cassettes
laminar flow: h #4.01V0.5 (3)
w 2 were placed in a wind tunnel and subjected to
turbulent flow: h #6.36V0.8 . (4) air speeds, at various attack angles, from 4.5
w 2
In reality, the application of fundamental flat to 24 m s−1, corresponding to values of the
plate convective heat transfer formulae to real Reynolds numbers Re from 2×104 to 9×104.
L
solar collectors must proceed with caution. To Convective heat transfer relationships were
draw an analogy between a uniformly heated derived by using the analogy between mass
flat plate and the upper surface of a solar transfer and heat transfer. A single ‘‘global’’
collector may be misleading on several counts. correlation, within a deviation of ±10%, was
The upper surface of the collector may not be found to describe adequately the experimental
isothermal and its average temperature will data for the range of cassette geometries, wind
change with time. In terms of air flow a solar incidence angles and Reynolds numbers used
collector is not a flat plate in a parallel air during the measurements. This correlation used
stream. Rather, it is a three-dimensional paral- the Colburn analogy factor j and took the form
lelepiped of finite thickness presenting a bluff j=0.86Re−0.5 (7)
L
surface to the wind, which gives rise to flow
where the characteristic length in eqn (7) is
phenomena such as separation and reattach-
given by
ment. Finally, the natural wind will contain
free-stream turbulence, which has been shown L=(4×plate area)/(plate circumference).
to enhance convective heat transfer from a flat
(8)
surface (McCormick et al., 1984). Therefore, it
would seem unwise to extrapolate from funda- Since
mental convective heat transfer theory to real j=(h /rc V )Pr0.67=0.86Re−0.5 (9)
collectors without the benefit of corroborating w p 2 L
evidence from empirical studies. then a direct relationship for h may be derived
w
by rearranging eqn (9). For a mean plate–air
temperature of 313 K this rearrangement gives
2. PREVIOUS EXPERIMENTAL WORK
h =5.1V0.5 L−0.5. (10)
One of the earliest experimental studies of w 2
forced convective heat transfer was that of Sparrow and his co-workers applied the mass
Jurges, as described by McAdams (1954), who, transfer analogy technique to a large number
in 1924, heated a 0.5 m2 copper plate mounted of conditions of relevance to convective heat
vertically and flush with the side of a wind transfer from solar collectors. Much of this
tunnel. The air speed was measured at the centre work was summarised in the paper by Sparrow
of the tunnel and an analysis of the results et al. (1982).
yielded a correlation for smooth surfaces of Although the body of work carried out by
Sparrow and his co-workers is extensive it still
h =3.8V +5.7 W m−2 K−1 V ≤5 m s−1. leaves many issues unresolved. None of their
w 2 2
(5) experiments were truly similar to the conditions
which exist when a real collector is mounted
Jurges’s results are still widely quoted by many onto the roof of a building. As well as the
sources as a means of calculating h . The stan-
w geometric differences in the model shapes used,
dard work by Duffie and Beckman (1991), for there are also questions relating to the effect of
example, cites Jurges, together with the sugges- turbulence in the natural wind and the influence
tion from Watmuff et al. (1977) that Jurges’s on average convective heat transfer coefficients
equations may include free convection and radi- of collector dimensions measured in metres
ation effects and so be overestimating the value rather than the millimetres used in the model
of h . The amended version of eqn (5) by
w studies. Some wind tunnel studies have used
Watmuff et al. (1977) is heated elements mounted flush with the pitched
h =3.0V +2.8 W m−2 K−1 V ≤5 m s−1. roofs on actual three-dimensional building-
w 2 2
shaped models. Kind et al. (1983) worked on a
(6)
1:32 scale model of a single-family residence
Duffie and Beckman (1991) also refer to work placed in a wind tunnel airflow that was both
by Sparrow et al. (1979). These workers used shear and turbulent. His results, extrapolated
 Full-scale measurements of wind-induced convective heat transfer 71

up to full-scale roof-mounted collectors, indi- 3. EXPERIMENTAL DETAILS

cated that h values would be two to three times
w 3.1. Apparatus
lower than those predicted by standard relation-
ships such as eqn (5). Sensitivity of h to turbu- A flat-plate solar collector was represented
w by a layered construction held within a wooden
lence and building details was low, although
there was some variation found for different frame (see Fig. 1). The top layer, which simu-
wind directions. Shakerin (1987) carried out a lated the top cover plate exposed to the external
similar study to Kind et al. (1983), but was environment, consisted of a 2 mm thick alumin-
able to vary the roof pitch of his models between ium sheet which had an exposed heat transfer
0° and 90°. Shakerin (1987) found that for roof surface measuring 1.81 m×0.89 m. The top sur-
pitches below 40° the convective heat transfer face of this sheet was sprayed with matt black
coefficient h was higher than that predicted by paint which had laboratory-measured values of
w solar absorptance a and thermal longwave
the ‘‘global’’ eqn (7) of Sparrow et al. (1979). p
Above this angle the value of h was found to emittance e of 0.97 and 0.90 respectively.
w p
be approximately independent of roof pitch and Attached to, and flush with, the underside of
to be well represented by a Sparrow-type equa- this plate were nine copper–constantan thermo-
tion. Francey and Papaioannou (1985) placed couples, arranged in a 3×3 array so as to be
centred on nine equal areas of plate. The alu-
a real solar collector with a front plate area
minium sheet’s top surface was not perfectly
of 1.85 m2 in a wind tunnel of cross-
isothermal, as small-scale wind flow patterns
section 4 m×3 m and pumped hot water
over the large surface produced localised
through the collector. Wind speeds between 1
differential cooling. The range of surface tem-
and 5 m s−1 were used to produce convective
peratures between the nine thermocouples was
cooling and allow top plate losses to be esti-
typically 5 to 10 K The mean value of these
mated and compared with the theoretical pre-
nine temperatures was taken as the plate tem-
dictions of Agarwal and Larson (1981).
perature T . This mean plate temperature varied
Measured top losses were found to be up to p
with the prevailing climatic and experimental
40% higher than predicted values. Changes in
conditions, but was usually in the range 300 to
collector tilt and wind direction were also found
330 K. Ambient air temperatures during the
to produce large changes in experimental
several months of the measurement programme
results. One of the few studies to investigate
were between 280 and 295 K. For most experi-
wind-induced heat transfer from surfaces
mental runs the mean plate–air temperature
exposed to the real wind was performed by Test difference lay between 10 and 40 K. Beneath
et al. (1981). A heated plate measuring the plate was a heating element, based upon a
1.22 m×0.81 m was placed in the natural envi- commercially available radiant ceiling heating
ronment and the wind speed at a height of 1 m system, which consisted of a flexible thin sheet
above the plate was recorded. A linear relation- of glass cloth impregnated with a conducting
ship between h and V was established of the PTFE dispersion. Copper strip electrodes were
w
form stitched along each long edge of the glass cloth
and an electrically insulating film laminated to
h =2.56V +8.55 W m−2 K−1. (11) each side of the sheet. The heating element had
w 2
an overall thickness of 2 mm and a similar area
The basic conclusion to be drawn from this to that of the aluminium plate. The heater was
review is that as experiments come closer to powered by a solid state phase controlled ac
resembling ‘‘real’’ collector situations so more power regulator, which was capable of supply-
discrepancies and inconsistencies are found both ing nine discrete power levels Q . Laboratory
e
between measured results from different experi- calibration of the heater using an ac wattmeter
ments and with standard flat-plate forced con- showed the power levels to range from 165 to
vection relationships for h . The major area for 2220 W m−2. Beneath the heating element was
w
corroborating data appears to be for measure- placed a sheet of bright aluminium foil to reflect
ments made on collector-sized heating elements heat back in the direction of the exposed alu-
placed on the roofs of real buildings and minium top plate. Back losses from the appara-
exposed to the real wind environment. This tus were minimised by having beneath the
study set up just such an experimental arrange- aluminium foil a 100 mm thick sheet of
ment in order to provide data of this type. expanded polystyrene insulation. Thermo-
72 S. Sharples and P. S. Charlesworth

Fig. 1. Schematic section through convective heat transfer equipment.

couples held either side of the insulation, 3.3. Instrumentation

together with a knowledge of the thermal con- Wind speed V and wind incidence angle i
ductivity of the insulation (0.037 W m−2 K−1 r
were measured using a lightweight three-cup
at 353 K ) enabled an estimate of the conductive vane anemometer mounted on a small mast
losses Q from the back and sides of the frame 1.5 m above the ridge line of the roof at the
c
to be made. For most of the experiments the west end of the building i.e. 6.1 m above ground
heater output was set at 625 W m−2. Side and level. This height was an arbitrary choice, to
back losses from the system were calculated to give the anemometer a secure fitting without
be typically 25 W m−2. Therefore, the heat flow excessive structural support. However, the value
through the top plate was typically is not dissimilar to the height chosen by Test
600 W m−2. The entire system described above et al. (1981) for their wind speed measurements
was held rigidly and slightly compressed (to above an inclined heated plate exposed to the
give good thermal contact) by a wooden frame external environment. No attempt was made to
arrangement with a wooden backing panel. All measure the wind speed just above the plate for
joints were sealed with a silicone mastic. comparison with the roof speed. This decision
was taken for three reasons. Firstly, the support
3.2. Building and plate location structure required to hold any anemometer
The heater plate apparatus was attached above the plate would have altered the wind
centrally and parallel to the south facing pitched flow patterns over the plate. Secondly, the flows
roof of an isolated single storey building sur- close to the surface of the plate would be very
rounded mostly by open fields. The building low speed and would have required sensitive
measured 9.60 m×5.03 m in plan, with its long instruments, such as hot wire anemometers,
axis running approximately east–west. The which would have been susceptible repeatedly
height to the top of the roof ridge was 4.59 m to damage from rain, hail or birds. Finally, to
and the roof had a pitch of 35°. Owing to know the relationship between the wind speed
stringent safety requirements the heater plate at one height above the plate and at one height
system had to be secured to the roof using a above the roof is not particularly useful as it is
rigid scaffolding arrangement which formed a so location specific. Of more interest is the
frame onto which the heater plate could be actual magnitude of the convective heat transfer
bolted. This meant that the exposed top plate under field conditions. External air temperature
of the system was some 330 mm above the plane T was measured using a shielded copper–con-
a
of the roof. Although this is somewhat higher stantan thermocouple. Incoming global short
than a standard collector’s top plate would be, wave irradiance G was measured by a newly
this was a necessary compromise within the calibrated Kipps solarimeter, whilst the incom-
experimental design. Figures 2 and 3 show the ing long wave irradiance R was measured using
building and the equipment in position on the a Funk-type net radiometer with a black cavity
roof of the building. on one side and the other surface recording
Full-scale measurements of wind-induced convective heat transfer 73

Fig. 2. Single storey building used in study.

Fig. 3. Heat transfer equipment in position on roof.

74 S. Sharples and P. S. Charlesworth

incoming long and short wave irradiances. The convective heat transfer coefficient h and the
w
value of the external convective heat transfer roof wind speed V . This relationship has pre-
r
coefficient h was derived from the steady-state viously been expressed in the literature as both
w
energy balance equation at the external plate linear and power functions of h on wind speed,
w
surface: and Table 1 shows the linear and power regres-
sion fits to the measured data from this study,
Q +a G+e R=e sT4 +Q +h (T −T )
e p p p p c w p a together with the values of the correlation
(12) coefficient r (significance level for all fits was
P<0.001). For the linear regression the 95%
where Q was the heating element’s output and
e confidence limit values are also shown. The
Q the conductive losses from the sides and
c magnitude of these limits, when applied to the
back of the wooden frame holding the heating
system. Rearranging eqn (12) gives mean values of h measured in the field, suggest
w
that percentage error associated with the mea-
h =[Q +a G+e (R−sT4 )−Q ]/(T −T ). sured value is approximately ±20%. This figure
w e p p p c p a
(13) reflects the difficulty of making field measure-
ments involving rapidly varying variable such
3.4. Procedure as the wind speed and direction and incoming
All transducer signals (25 in all ) were solar radiation.
recorded every 2 min onto magnetic tape car- It is apparent from Table 1 that both the
tridges via a data logger housed in the test linear and power regressions on the measured
building. Data were collected for periods of 8 data are giving reasonable fits, with the power
to 26 h, depending upon prevailing wind condi- fit being, in general, slightly better. Table 2
tions. In general, measurement periods were shows the derived magnitude of h from the
w
characterised as having overcast skies with mod- linear and power fits to the measured data for
erate to strong winds. Solar radiation levels three values of roof wind speed V (1, 3 and
r
were low. No differences were observed between 5 m s−1), these values being representative of
data collected during day and night time the range of V measured during the experi-
r
periods. In all, 29 experimental runs were per- ments. The differences between the predicted
formed, giving a total of over 12 000 scans of values of h from the linear and power fits,
w
the transducer signals. averaged over the eight wind directions, are
9.5% for V at 1 m s−1, 1.4% for V at 3 m s−1
r r
and 5.0% for V at 5 m s−1. Obviously, the
4. RESULTS AND DISCUSSION r
power fit cannot reflect the natural convective
The large amount of data collected during heat transfer coefficient as V tends to zero,
r
the field measurements had to be restructured whereas the linear fits give, at their y-axis
to allow some meaningful analyses to be per- intercepts, values of h which are representative
w
formed. This was done in two ways: firstly, of the measured values of h at very low wind
w
hourly mean values of the measured parameters speeds (an average of 8.1 W m−2 K−1 at
were calculated from the raw data; secondly, V =0 m s−1). The conclusion to be drawn from
r
these mean results were then grouped by wind these field measurements is that, for a given
direction into eight wind incidence angle sets. wind direction, either a linear or power relation-
The wind incidence angle i was defined as the ship will give an adequate estimate of the forced
angle between the normal to the south-facing convective, wind-related heat transfer coeffi-
collector side of the test building and the wind cient. However, algorithms which seek to
direction, with values clockwise running from extrapolate down to very low wind speeds can
0° to +180° and anticlockwise from 0° to use the greater generality of the linear relation-
−180°. Each of the eight sectors represented by ships without a serious loss of accuracy.
i (0°, 45°, 90°, 135°, 180°, −135°, −90°, −45°) The field measurements indicated a slight but
related to a specific wind incidence angle centred distinct variation in the convective heat transfer
on the sector value with a spread of values of coefficient with wind direction, particularly at
±22.5°. Such a grouping recognises the fact the higher values of wind speed observed during
that wind direction itself is a fluctuating quan- the experiments. Convective heat transfer from
tity. Lacy (1977) demonstrated that wind direc- the plate, for a given wind speed, was at its
tions typically oscillate about their mean values lowest value when the incidence angle i was
by approximately ±15°. The relationship of 180° (plate on leeward surface). Heat transfer
most interest was that between the wind-induced was generally highest for winds blowing either
 Full-scale measurements of wind-induced convective heat transfer 75

Table 1. Linear and power regressions of h on V by wind incidence angle i

w r
Wind incidence No. of data Regression equations for Correlation Wind speed range 95% confidence limits to
angle i points N h ( W m−2 K−1) coefficient r (m s−1) linear fit ( Wm −2 K−1)
w
0° 65 h =2.2V +8.3 0.767 0.8<V <6.7 ±4.3
w r r
h =9.3V0.44 0.821
w r
45° 136 h =2.6V +7.9 0.675 0.6<V <6.2 ±6.8
w r r
h =9.5V0.46 0.747
w r
90° 42 h =3.3V +6.5 0.878 0.8<V <6.2 ±4.7
w r r
h =9.5V0.48 0.804
w r
135° 42 h =2.2V +7.9 0.716 0.8<V <6.4 ±5.6
w r r
h =8.3V0.51 0.751
w r
180° 36 h =1.3V +8.3 0.439 0.6<V <4.9 ±6.6
w r r
h =9.2V0.25 0.559
w r
−135° 123 h =2.3V +7.8 0.618 0.3<V <7.2 ±8.2
w r r
h =9.6V0.41 0.754
w r
−90° 39 h =2.2V +11.9 0.651 0.5<V <6.7 ±7.9
w r r
h =11.7V0.41 0.777
w r
−45° 32 h =3.9V +6.0 0.834 0.5<V <6.7 ±7.8
w r r
h =9.1V0.57 0.600
w r

Table 2. Comparative values of h for linear and power regressions for three roof wind speeds V
w r
Wind angle i 0° 45° 90° 135° 180° −135° −90° −45°

h (V =1)
w r
linear fit 10.5 10.5 9.8 10.1 9.6 10.1 14.1 9.9
power fit 9.3 9.5 9.5 8.3 9.2 9.6 11.7 9.1
% diff. 11.1 9.5 3.1 17.8 4.2 5.0 17.0 8.1
h (V =3)
w r
linear fit 14.8 15.7 16.3 14.6 12.3 14.5 18.4 17.5
power fit 15.0 15.8 16.2 14.6 12.1 15.1 18.3 17.1
% diff. −1.4 −0.6 0.6 0.0 1.6 −4.1 0.5 2.3
h (V =5)
w r
linear fit 19.3 20.9 23.0 18.9 14.8 19.3 22.9 25.5
power fit 18.9 19.9 20.6 18.9 13.8 18.6 22.6 22.8
% diff. 2.1 4.8 10.4 0.0 6.8 3.8 1.3 10.6

parallel to the plate (i=±90°) or obliquely side wall of a wind tunnel; with Sparrow et al.
when the plate surface was windward (i= (1979), who held small mass transfer cassettes
±45°). For example, Table 2 shows that for in a wind tunnel air flow, and with Test et al.
V =5 m s−1 the value of h is 16.2 W m−2 (1981), who exposed an isolated large
r w
K−1 at i=180° but raises to 25.5 W m−2 K−1 (1.22×0.81 m2) heated plate to the natural
at i=45°. wind. The Watmuff modified version of the
well-known relationship from Jurges has
4.1. Comparison with previous work
already been given in eqn (6).
The wind-induced convective heat transfer Sparrow’s ‘‘global’’ relationship, given in
field measurements of this study were compared eqns (7) and (10), converts approximately, for
with existing collector convective heat loss the plate dimensions used in this study (with
design algorithms. There are some major prob- L=1.193) to
lem with any such comparison:
$ the representative wind speed used in each h =4.7V0.5 W m−2 K−1. (14)
w 2
experiment may be different;
$ the structure and turbulence in the incident The relationship of Test et al. (1981), for a
wind may vary; wind speed measured 1 m above the plate, was
$ the physical dimensions of the heat transfer given in eqn (11).
surfaces can vary from a few millimetres to The above relationships have been compared
a few metres; with the derived field measurement relationships
$ the range of Reynolds numbers used in each from this study for two wind directions (i=0°
experiment may be different. and 90°).
The current study’s results were compared
h =2.2V +8.3 W m−2 K−1 (i=0°)
with the results of Jurges [in McAdams (1954)] w r
who used a small heated plate attached to the (15)
76 S. Sharples and P. S. Charlesworth

Fig. 4. Comparison of previous convection relationships with current study.

h =3.3V +6.5 W m−2 K−1 (i=90°). that the magnitude of the prevailing h values
w r w
shown in Fig. 4 is similar for high wind speeds
(16)
where forced convection dominates, and that it
The results of this comparison are shown in is only the way in which the prevailing wind
Fig. 4. It is apparent from Fig. 4 that there is speeds are defined that alter the heat transfer
good agreement between the current field relationships presented in the literature. This
study’s values and those derived by Test et al. observation was also made by Sharples (1984)
(1981). The agreement with Watmuff/Jurges is after his field study of convective heat transfer
reasonable at higher wind speeds, but becomes from external building surfaces. Figure 4
poorer with reducing wind speeds. The appears to show poor agreement between the
agreement with Test et al. (1981) is not surpris- other three studies and the ‘‘global’’ equation
ing, as they was measuring values of h from a of Sparrow et al. (1981). However, Sparrow
w et al. (1981) did observe during their studies
plate of similar dimensions to that used in this
study. He also exposed his plate to the real that if the inclined heat transfer cassettes was
placed above a passive host surface then there
wind and measured the wind speed at 1 m above
was an up to twofold enhancement of the
the plate surface, compared with the 1.6 m
convective heat transfer from the cassette. This
height of this study’s anemometer. It is, how-
was attributed to a rapidly recirculating flow
ever, surprising that the presence of a single
over the cassette induced at the step created
storey building beneath this study’s plate did where the cassette met the host surface. Such a
not produce a more obvious discrepancy with step exists in this study, where the collector
the results of Test et al. (1981). The reasonable rests on the roof. If this twofold enhancement
agreement with Watmuff/Jurges at higher is applied to eqn (13) of Sparrow et al. (1981)
speeds is interesting, as there appears to be so it produces for this study’s plate
little in common between Jurges’s small
plate/wind tunnel arrangement and this study’s h =9.4V0.5 W m−2 K−1. (17)
w 2
large plate/real wind set up. It can only be Values of h from this equation are similar
speculated that, whatever the prevailing forced w
to those given by the other studies in Fig. 4.
convective heat transfer flat plate conditions
and geometries, the fundamental heat transfer
process at the plate surface is essentially still 5. CONCLUSION
similar to that observed by Jurges. The differ- The results from a field study of wind-induced
ence with Watmuff/Jurges at low speeds may convective heat transfer from a large heated flat
be linked to the greater influence of natural plate attached to the pitched roof of a single
convection at these velocities. It is suggested storey building have been presented for a range
 Full-scale measurements of wind-induced convective heat transfer 77

of wind directions. The relationship between Duffie J. A. and Beckman W. A. (1991) Solar Engineering
of Solar Processes, 2nd edition. Wiley Interscience, New
the convective heat transfer coefficient h and
w York, pp. 173–174.
the prevailing roof top wind speed V was well Francey J. L. A. and Papaioannou J. (1985) Wind-related
r
described by either power or linear relation- heat losses of a flat-plate collector. Solar Energy 35,
ships. There was evidence that the convective 15–19.
Garg H. P. and Datta G. (1984) The top loss calculation
heat transfer rate was lower, for a given wind for flat-plate solar collectors. Solar Energy 32, 141–143.
speed, when the plate surface was on the leeward Hottel H. C. and Woertz B. B. (1942) The performance of
side of the building. A comparison with other flat-plate solar-heat collectors. Transaction of the ASME
studies, both wind tunnel and field, showed 64, 91–104.
Incropera F. P. and DeWitt D. P. (1996) Fundamentals of
good agreement with the predicted values of Heat and Mass Transfer, 4th edition. Wiley, New York,
h , even though the means of defining the wind pp. 348–358.
w Kind R. J., Gladstone D. H. and Moizer A. D. (1983)
speed varied from one case to another.
Obviously, the results from this study are strictly Convective heat losses from flat-plate solar collectors in
turbulent winds. Transactions of the ASME J. Solar
only applicable to the particular experimental Energy Engng 105, 80–85.
conditions existing at the site of the measure- Klein S. A. (1975) Calculation of flat-plate collector loss
ments. However, the results from this investiga- coefficients. Solar Energy 17, 79–80.
Lacy R. E. (1977) Climate and Building in Britain.
tion are encouraging for those involved with
HMSO, London.
solar energy simulation. In general, the order Malhotra A., Garg H. P. and Palit A. (1981) Heat loss
of magnitude values of h derived from earlier calculation of flat plate solar collectors. J. Thermal
w Engng 2, 59–62.
studies have been shown to be similar to those
McAdams W. H. (1954) Heat Transmission, 3rd edition.
observed from the real world measurements of
McGraw-Hill, New York.
this study. McCormick F. L., Test F. L. and Lessmann R. C. (1984)
The effect of free-stream turbulence on heat transfer
NOMENCLATURE from a rectangular prism. Transactions of the ASME
J. Heat Transfer 106, 268–275.
c specific heat capacity of air (J kg−1 K−1) Morrison G. L. and Gilliaert D. (1992) Unglazed solar
p
G total (direct and diffuse) solar radiation incident upon collector performance characteristics. Transactions of the
collector ( W m−2) ASME J. Solar Energy Engng 114, 194–200.
h wind-induced top plate convective heat transfer Ramsey J. W. and Charmchi M. (1980) Variance in solar
w
coefficient ( W m−2 K−1) collector performance predictions due to different meth-
i wind incidence angle (deg) ods of evaluating wind heat transfer coefficients.
j Colburn analogy factor (h /rc V )Pr0.67 (dimension- Transactions of the ASME J. Heat Transfer 102,
w p 2
less) 766–768.
k thermal conductivity ( W m−1 K−1)
Samdarshi S. K. and Mullick S. C. (1994) Generalized
L characteristic plate length, 4×plate area/plate circum-
analytical equation for the top heat loss factor of a flat-
ference (m)
Nu Nusselt number h L/k (dimensionless) plate solar collector with N covers. Transactions of the
w ASME J. Solar Energy Engng 116, 43–46.
Pr Prandtl number n/a (dimensionless)
Q conductive losses from back and sides of collector Shakerin S. (1987) Wind-related heat transfer coefficient for
c flat-plate solar collectors. Transactions of the ASME
frame ( W m−2)
Q electrical power input into heater element ( W m−2) J. Solar Energy Engng 109, 108–109.
e Sharples S. (1984) Full-scale measurements of convective
R incoming longwave radiation ( W m−2)
Re Reynolds number V L/n (dimensionless) energy losses from exterior building surfaces. Building
2
T ambient air temperature ( K ) and Environment 19, 31–39.
T mean top plate temperature ( K ) Sparrow E. M., Ramsey J. W. and Mass E. A. (1979) Effect
p
U overall top plate heat loss coefficient ( W m−2 K−1) of finite width on heat transfer and fluid flow about an
t
V roof height wind speed (m s−1) inclined rectangular plate. Transactions of the ASME
r
V free stream wind speed (m s−1) J. Heat Transfer 101, 199–204.
2
Sparrow E. M., Somie F. and Lau S. C. (1981) Heat transfer
Greek letters
from a plate elevated above a host surface and washed
a thermal diffusivity (m2 s−1) by a separated flow induced by an elevation step.
a solar absorption coefficient of top plate (dimension- Transactions of the ASME J. Heat Transfer 103,
p
less) 441–447.
e thermal emittance of plate (dimensionless) Sparrow E. M., Nelson J. S. and Tao W. Q. (1982) Effect
p
g solar collector efficiency (dimensionless) of leeward orientation, adiabatic framing surfaces, and
n kinematic viscosity (m2 s−1) eaves on solar-collector-related heat transfer coefficients.
r density of air (kg m−3) Solar Energy 29, 33–41.
s Stefan–Boltzmann constant, 5.67×10−8 W m−2 K−4 Test F. L., Lessmann R. C. and Johary A. (1981) Heat
transfer during wind flow over rectangular bodies in the
natural environment. Transactions of the ASME J. Heat
REFERENCES Transfer 103, 262–267.
Agarwal V. K. and Larson D. C. (1981) Calculation of the Watmuff J. H., Charters W. W. S. and Proctor D. (1977)
top loss coefficient of a flat plate collector. Solar Energy Solar and wind induced external coefficients solar collec-
27, 69–71. See also errata 30 (1983) 86; 36 (1986) 479 tors. Int. Revue d ’Hellio-technique 2, 56.