You are on page 1of 9

An Experimental Study of Natural

Convection in Vertical, Open-


Ended, Concentric, and Eccentric
Annular Channels
G. H. Choueiri
The effects of eccentricity on the natural convection heat transfer from a vertical open-
S. Tavoularis1 ended cylindrical annulus with diameter ratio of 1.63 and aspect ratio of 18:1 have been
e-mail: stavros.tavoularis@uottawa.ca investigated experimentally. Within the range of present conditions, and with the possible
exclusion of the highest eccentricities, it was found that the flow was thermally fully
developed in a considerable section of the apparatus, as indicated by the linear variation
Department of Mechanical Engineering, of wall temperature with height. This made it possible to estimate the mass flow rate from
University of Ottawa, the wall temperature gradient in the mid-section of the annulus, and use it to calculate
161 Louis Pasteur, the bulk Reynolds number, which was found to be weakly sensitive to eccentricity for a
Ottawa, ON K1N 6N5, Canada constant wall heat flux and to increase with increased wall heat flux. With the exception
of the very low eccentricity range in which it was insensitive to eccentricity, the overall
heat transfer rate diminished monotonically with increasing eccentricity. Plots of the
local azimuthal variation of the Nusselt number showed that, at low eccentricities, the
heat transfer rate increased near the wider gap but decreased near the narrower gap.
The average Nusselt number was found to decrease measurably with increasing eccen-
tricity and to increase slightly with increasing heat flux within the examined range. In
contrast, the Grashof number was found to be much more sensitive to changes in heat
flux and only had a weak dependence on eccentricity. [DOI: 10.1115/1.4004428]

Keywords: natural convection, annular channel, eccentricity, thermocouples

1 Introduction ity of the narrow-gap, making it relatively insensitive to gap width


within a certain range. The possible formation of such vortices in
Heat exchangers are essential parts of many engineering sys-
vertical eccentric annular channels in natural convection has not yet
tems, including heating, ventilating, and air conditioning systems
been investigated.
and power generation plants of all sorts. Among the many heat
The heat transfer literature contains many studies of the effect
exchanger designs that are available, a relatively simple one is the
of eccentricity on the natural convection heat transfer in concen-
double-pipe configuration, in which hot and cold fluids flow
tric and eccentric annuli. A lot of this work focused on the hori-
through coaxial tubes. In industrial heat exchangers, forced con-
zontal annular geometry; it includes studies of two-dimensional
vection is the usual mode of heat transfer, nevertheless, in some
natural convection in horizontal cylindrical annuli with isothermal
applications, such as solar water heating systems, natural convec-
boundary conditions for both the concentric [11] and eccentric
tion is employed to eliminate the need of pumps. Some advanced
[12] configurations and three-dimensional natural convection in
nuclear reactor concepts use natural convection as the main driv-
concentric and eccentric annuli with open ends [13]. Studies of
ing force for coolant circulation, which has the additional advant-
natural convection in vertical annuli include some in closed annuli
age of serving as an emergency cooling system in the event of
[14–17] and others in open-ended ones [18,19]. In nearly all previ-
power failure [1].
ous studies, only one wall of a vertical annulus was heated. In
Most common designs of simple double-pipe heat exchangers
such cases, undesirable flow reversal near the outlet may occur as
consist of annular ducts, which would ideally be concentric but in
a result of low outlet pressure associated with flow acceleration
practice may have some eccentricity introduced by manufacturing
[20]. Moreover, when only one wall is heated, it is impossible to
tolerances, imperfections in construction, or thermal stresses. Ec-
isolate the effects of heat convection from those of radiation,
centricity may be nonuniform along the duct, as the result of pipe
which could be significant and even dominant. A thorough litera-
bowing, end misalignment or creep, however, interest in the pres-
ture survey has revealed no previous experimental studies of natu-
ent study focuses on uniform eccentricity.
ral convection in vertical open-ended annuli with both the inner
Forced convection in strongly eccentric annular channels is domi-
and outer walls heated with a constant wall heat flux; this is the
nated by large-scale vortices that form on either side of the narrow
subject of the present work.
gap and therefore cannot be predicted by empirical analyses, for
The primary objective of this article is to document the sensitiv-
example, those based on hydraulic diameter scaling, or steady-flow
ity of natural convection heat transfer to eccentricity in an open-
computational fluid dynamics (CFD) simulations [2–5]; see also
ended vertical annulus formed between a cylindrical core and a
Refs. [6–10] for forced convection in rectangular ducts containing a
surrounding cylindrical cavity. Both walls were heated electrically
cylinder. These vortices generally enhance heat transfer in the vicin-
providing the air in the annulus with an approximately constant
heat flux and the eccentricity was adjusted by traversing the core
1
Corresponding author. laterally. Heating both walls to approximately the same tempera-
Contributed by the Heat Transfer Division of ASME for publication in the JOUR-
NAL OF HEAT TRANSFER. Manuscript received April 30, 2010; final manuscript
ture reduced radiation effects to a negligible level in the mid-sec-
received June 13, 2011; published online October 5, 2011. Assoc. Editor: Patrick H. tion of the annular duct. Local wall temperature was measured at
Oosthuizen. many locations along and around the annular channel and the

Journal of Heat Transfer Copyright V


C 2011 by ASME DECEMBER 2011, Vol. 133 / 122503-1

Downloaded From: https://heattransfer.asmedigitalcollection.asme.org on 06/30/2019 Terms of Use: http://www.asme.org/about-asme/terms-of-use


results were analyzed to determine relevant average and local H=Do  18. Vertical position z along the test section was meas-
dimensionless groups and their sensitivity to eccentricity and heat ured starting from its lowest end. The eccentricity is defined as
flux. Besides helping us understand an important aspect of natural
convection, the results of our work could serve as a benchmark Dx
e¼ (1)
for validation of future CFD studies of natural convection in simi- ro  ri
lar geometries.
such that it is equal to 0 when the two cylinders are coaxial and
equal to 1 when the two cylinders are in contact. Dx is the distance
between the axes of the inner and outer cylinder, and ro and ri are
2 Experimental Facility and Procedures the radii of the outer and inner cylinders, respectively, as shown
in Fig. 2.
2.1 Apparatus. The experimental apparatus consisted of an Each of the inner and outer cylinders was covered with a single
annular test section with a traversable core and a plenum, all held sheet of stainless steel foil having a thickness of 51 lm. The inner
together by a welded steel frame (Fig. 1). foil was fastened on the G-11 tube by epoxy, whereas, the outer
The test section was positioned vertically in the central section foil was first wrapped on a polyvinyl chloride (PVC) cylinder that
of the frame. The plenum was sealed from all directions with the was subsequently removed and then it was covered externally by
exception of a rectangular inlet having glass walls. The inlet paper sheets that provided good electrical and fair thermal insula-
cross-sectional area was adjusted by trial-and-error, such that, it tion and a galvanized steel sheet that provided additional stiffness.
was the smallest possible to affect negligibly the flow rate through To avoid local overheating, the two sides of the foil were not con-
the test section at the highest heat flux encountered in our experi- nected as they met around each cylinder, but left a narrow
ments. The top section of the apparatus was open and allowed air unheated gap which was less than 1 mm wide. The upper and
to exit the test section freely. Two precision screws were mounted lower ends of the foil on the inner cylinder were pressed on two
on the bottom and top of the frame. The core of the test section copper rings, so that the electrical current supplied to the foil
was supported on these screws by coaxial rods approximately 0.5 would be evenly distributed azimuthally. Similarly, the two ends
m in length and 13 mm in diameter and could be traversed hori- of the foil of the outer cylinder were pressed on aluminium plates.
zontally with an uncertainty of 0.05 mm. The external surfaces of these rings and plates were electrically
The test section consisted of a core cylinder, which was a sec- and thermally insulated using 6.35 mm thick Teflon plugs for the
tion of a tube made of G-11 glass reinforced epoxy (P&A Plastics inner cylinder and Lexan plates of comparable thickness for the
Inc., Hamilton, Ontario) with a wall thickness of 6.35 mm, sur- outer one. Except for its upper and lower ends, the test section
rounded by a cylindrical cavity. In the following, we shall refer to exterior was packed within thick fiberglass insulation and then
the outer surface of the core as the inner cylinder and to the inner wrapped with aluminium-coated bubble wrap to keep heat losses
surface of the cavity as the outer cylinder. The outer cylinder had to the surroundings as low as possible. The frame supporting the
a diameter of Do ¼ 83 mm and the inner cylinder had a diameter entire system was sealed air tightly with sheet metal. The core
of Di ¼ 51 mm, so that the diameter ratio was Di=Do ¼ 0.61 and tube was also packed with fiberglass insulation to prevent the for-
the hydraulic diameter was Dh ¼ Do  Di ¼ 32 mm. The height of mation of internal convection currents.
the test section was H ¼ 1.5 m, corresponding to an aspect ratio of
2.2 Instrumentation. Two identical 10 kW electric power
supplies (Sorensen, DCR-T series, San Diego, CA), one for each

Fig. 2 Sketch of the annular duct cross-section showing the


definitions of the coordinates and the positions of the thermo-
Fig. 1 Schematic diagram of the experimental apparatus couples and the foil gaps

122503-2 / Vol. 133, DECEMBER 2011 Transactions of the ASME

Downloaded From: https://heattransfer.asmedigitalcollection.asme.org on 06/30/2019 Terms of Use: http://www.asme.org/about-asme/terms-of-use


cylinder were used to power the foils. These power supplies were
controlled remotely using a program written in LabVIEW 8.5 and
a data acquisition board (National Instruments, Model NI PCI-
6024E) with a 12-bit resolution and a maximum sampling rate of
200,000 samples=s. The voltage drop and the current were
recorded every 10 s using the data acquisition board by means of
remote voltage drop sensors and an internal shunt, both of which
were parts of the power supplies.
48 K-type thermocouples were positioned on the wall of each
cylinder, for a total of 96 thermocouples for the entire test section.
These thermocouples were electrically insulated and placed as
closely as possible to the back face of the stainless steel foil with-
out perturbing into the annulus. The thermocouples in the core
were inserted into the G-11 tube through notches machined on its
surface and all wiring was led through the core tube and the sup-
port rods at its two ends, so that it did not interfere with the flow
or the heat transfer. Thermocouples were spaced by 100 mm verti-
cally, starting at the inlet and ending at the exit, so that the test Fig. 3 Variations of the room temperature and the difference
between the average annulus wall temperature and the room
section had 16 wall temperature measurement planes. On each temperature during a 50 h time interval
horizontal measurement plane, three thermocouples were placed
on each of the inner and outer cylinders, as shown in Fig. 2. These
were labeled as Si0 , Si90 , and Si180 for the core, and So0 , So90 , and So180 ied by less than 1 K during the same period, with significant
for the surrounding cavity. The thermocouples were positioned so stretches of time during which this temperature difference varied
as to measure the temperature in one half of the annulus, assuming by less than 0.2 K. Because of room temperature fluctuations, an
that the wall temperature distribution on the opposite half would ideal steady state, during which all temperature measurements
be identical due to symmetry and disregarding the effect of the would be time-independent, could not be reached. For the pur-
gap in the foil, where no thermocouple was placed. In addition to poses of this study, we defined a quasi-steady state as a state dur-
the 96 thermocouples, another K-type thermocouple was placed in ing which no recorded thermocouple reading changed by more
the inlet duct of the plenum to measure the ambient air tempera- than 0.5 K during a 2 h period. Under these conditions, the tem-
ture. Two additional thermocouples were placed within the insula- perature difference T w  Ta varied by less than 0.2 K, which is
tion surrounding the test section to verify its effectiveness. considered to be the measurement uncertainty of temperature dif-
Thermocouple readings were recorded using a second data acqui- ferences. All measurements reported here were taken under such
sition system with a 16-bit resolution and a maximum sampling conditions.
rate of 500 samples=s for each channel (Sensoray Co. Inc., 2600 A second preliminary experiment was performed to measure
Series, Tigard, OR). Six thermocouple modules (Model 2608) the sensitivity of the electrical resistivity of the stainless steel foil
with 16 channels each were used. to variations in local wall temperature, because a change in the re-
All 96 thermocouples were calibrated at room temperature using sistivity of the foil would directly affect the uniformity of wall
a digital hand-held thermocouple thermometer (OMEGA Model heat flux q00 . The resistivity was calculated at room temperature
CL3512A calibrator=thermometer). The hand-held device was first by measuring the voltage and current passing through the inner
checked against an ice point reference to ensure its accuracy. Read- and outer cylinder immediately after the power supply was turned
ings of the inlet temperature and the exit temperature of the on and while the annulus was still at room temperature. The corre-
unheated apparatus were used to define a linear temperature profile sponding values of the resistivity of the foil in the outer and inner
(accounting for a linear density stratification in the room and, corre- cylinders at room temperature were 7.3  107 and 7.8  107
spondingly, in the unheated test section), against which the readings X m. The temperature coefficient of resistivity was calculated by
of all the wall thermocouples were compared. Appropriate correc- monitoring the change in resistivity in the annulus (calculated
tions were applied to each thermocouple to compensate for any off- from the voltage drop) while holding the current constant for an
sets (this correction also compensated for errors introduced during extended period of time as the annulus slowly heated up. The tem-
signal discretization). The temperature at the exit of the annulus perature coefficients of resistivity of the stainless steel foil in the
was found to be 0.5 K higher than that at the inlet. Removed offsets inner and outer cylinders at 293 K were measured to be
of the 96 thermocouples were in the range from 0.2 to 1.8 K. 1.0  103 and 1.1  103 K1, respectively. Having calculated
All effects considered the estimated uncertainty in the measurement the effect of temperature on the resistivity of the stainless steel,
of any temperature by any thermocouple was roughly 1 K. Never- we were then able to calculate the effects of temperature variation
theless, following calibration and offset removal, the uncertainty in on the local heat flux within the annulus. The maximum local var-
the measurement of a temperature difference by any two thermo- iation of wall heat flux for the 30 W=m2 case was 2.1%, while
couples was estimated to be less than 0.1 K. that for the 80 W=m2 was 4.5%, corresponding to maximum wall
temperature differences of 19.5 and 41.3 K, respectively. The dif-
2.3 Experimental Procedures and Uncertainties. Preliminary ference in the resistivities of the two foils is attributed to manufac-
tests were performed to determine the variation in room tempera- turing uncertainty, as the foils used for the inner and outer
ture and its effect on the test section wall temperature distribution. cylinders were purchased at different times and likely belonged to
First, the current and voltage in both the inner and outer cylinders different batches.
were adjusted to provide a constant wall heat flux of 80 W=m2. To take into account the dependence of the physical and ther-
Following this, the experiment was left running for 76 h; in the modynamic properties of air upon temperature, sixth-order poly-
last 50 h, measurements of wall and ambient temperature were nomial expressions were fitted to tabulated air properties in the
recorded at a rate of one value every 5 min, with each recorded range between 150 and 500 K to facilitate calculations. This was
value being the average of 500 samples taken within an interval of done for the density q, the specific heat cp, the thermal conductiv-
0.5 s. The results of this test, shown in Fig. 3, demonstrated that ity k, the kinematic viscosity , the thermal expansion coefficient
the room temperature, Ta at the inlet of the annulus varied during b, the Prandtl number Pr, and the thermal diffusivity a.
the 50 h test period by up to a maximum difference of 2.3 K. In The main experiments consisted of two lengthy runs, having
turn, the room temperature variation affected the average annulus wall heat fluxes equal to q00 ¼ 30 and 80 W=m2, respectively. For
wall temperature T w , but the temperature difference T w  Ta var- each heat flux, tests with seven different eccentricities, e ¼ 0, 0.1,

Journal of Heat Transfer DECEMBER 2011, Vol. 133 / 122503-3

Downloaded From: https://heattransfer.asmedigitalcollection.asme.org on 06/30/2019 Terms of Use: http://www.asme.org/about-asme/terms-of-use


0.3, 0.5, 0.7, 0.8, and 0.9, were conducted. Starting from the con- of the three thermocouples on each cylinder (denoted by symbols
centric case, the wall heat fluxes for the inner and outer cylinders corresponding to different eccentricities in these figures) were
were adjusted as precisely as possible to the desired value by con- used to estimate the azimuthal variations as fourth-order polyno-
trolling the current. The eccentricity was then changed after at mials, which have a minimum at the wide-gap and a maximum at
least 24 h had passed, allowing more than sufficient time for a the narrow-gap location. The small temperature nonuniformity
quasi-steady state to be reached. In general, a quasi-steady state present in the concentric cases may plausibly be attributed to geo-
was reached within approximately 4 h after changes in the metric imperfections of the apparatus.
eccentricity. The azimuthally averaged temperatures, T i ðzÞ and T o ðzÞ for
Besides heat convection to the flowing air, there was additional inner and outer cylinders at each elevation, were computed by
heat transfer from the heated surfaces in the form of radiation to integrating the corresponding fitted polynomials. Then, the azimu-
the surroundings and conduction to the supports. It is obvious that thally averaged wall temperature of the annulus at a given eleva-
these undesirable heat transfer modes would have more significant tion was computed as the average of the wall temperatures of the
effects on the wall temperature near the two ends. The decrease of two cylinders, weighted by their surface areas, namely as
bulk fluid temperature due to end conduction losses was found to
be less than 0.1 K, which is lower than the overall temperature T o Ao þ T i Ai
uncertainty and will be neglected. Radiation losses along the test Tw ¼ (4)
Ao þ Ai
section were estimated roughly by approximating the cylindrical
walls of the annulus by parallel flat plates having a spacing com- The temperature variation in the annulus was apparently influ-
parable to the average spacing between the two cylindrical surfa- enced by end effects, caused largely by radiation losses. To avoid
ces at different eccentricities, and using published expressions for the difficult task of accounting for such effects, we have decided
the view factors [21]. Although it was found that radiation losses to focus on the mid-section of the annulus, where end effects were
increased slightly with increasing eccentricity, the eccentricity estimated to be insignificant. The vertical variation of the azimu-
effect on these losses resulted in temperature corrections that were thally averaged wall temperature has been plotted in Fig. 6. These
lower than the overall temperature measurement uncertainty. An results show that the average annulus wall temperature increased
analysis of the conduction and radiation heat transfer and addi- with increasing eccentricity over the entire measured range of ec-
tional details of the experimental procedures have been presented centricity for both the wall heat fluxes. The general effect of ec-
by Choueiri [22]. Based on this analysis, a correction to the fluid centricity on wall temperature for the two wall heat fluxes appears
bulk temperature to account for radiation losses in the lower half to be consistent; however, the vertical temperature gradient was
of the test section was applied to the estimates that will be presented higher for the higher wall heat flux.
in Sec. 3. This correction is 0.11 6 0.03 K for q00 ¼ 30 W=m2
and 0.18 6 0.04 K for q00 ¼ 80 W=m2.
The heat flux uncertainty was estimated from the uncertainties in 4 Analysis and Discussion of the Results
the temperature measurements, the foil resistivity, and the foil
thickness. It was roughly 1% for q00 ¼ 30 W=m2 and up to 2% for 4.1 Mass Flow Rate and Reynolds Number. For steady
q00 ¼ 80 W=m2. In these experiments, the heat flux is the driving flow, the mass flow rate through any cross-section of the annulus
force for the fluid motion and its magnitude should have a direct would be
effect on the flow. For a discussion of the results in terms of dimen- p 2  p 
sionless parameters, one may define a dimensionless heat flux as m_ ¼ qb Vb D  D2i ¼ qin Vin D2o  D2i (5)
4 o 4
Do þ Di
q00 ¼ q00 (2) where qb is the bulk air density, Vb is the bulk air velocity, qin is
kTa the air density at the inlet, and Vin is the air velocity at the inlet,
assumed to be uniform.
where the thermal conductivity is evaluated at the ambient tem- Consider a control volume (CV) in the annulus bounded by two
perature. This definition is independent of the eccentricity and horizontal planes with distance dz. Considering that the fluid in
other variables. The dimensionless heat flux values corresponding the CV does not produce any work, and treating air as an ideal
to the two heat fluxes considered in this study were 0.53 and 1.39. gas, one can write the first law of thermodynamics for this CV as

3 Measurements _ p dTb ¼ q00 pðDo þ Di Þdz  dqL


mc (6)
Figure 4 presents the variations of local wall temperature meas-
ured by the different thermocouples in the inner and outer cylin- where dTb is the change in the bulk temperature Tb of the air from
ders of the annular duct. These results are presented as the inlet to the outlet of the CV and dqL represents heat losses due
temperature increases to conduction and radiation. As mentioned previously, losses due
to conduction were negligible. Radiation losses from the inlet
DTw ¼ Tw  Ta (3) would have an effect on the bulk fluid temperature at least in the
where the ambient air temperature, measured at the inlet of the lower half of the test section, whereas, in the upper half of the test
plenum, was typically 297.0 6 1.6 K during the 3 week period, section, radiation losses from the outlet would also play a more
during which the reported measurements were taken. It is note- significant role. Nevertheless, analysis has shown that heat losses
worthy that DTw in the reported results fluctuated by less than 0.2 were negligible in the mid-section of the annular duct. Therefore,
K following small changes in Ta . in the mid-section, and in the limit of vanishing dz, one may com-
In the concentric case, the temperature readings of all thermo- pute the bulk fluid temperature gradient as
couples at a given vertical position (with the exception of the out-
let region) essentially coincided. As eccentricity increased, these dTb q00 pðDo þ Di Þ
¼ (7)
temperature readings deviated more and more, with thermocou- dz _ p
mc
ples facing the narrow gap (Si0 and So0 ) having the highest tempera-
tures and thermocouples facing the wide gap (Si180 and So180 ) Neglecting the temperature dependence of cp for the relatively
having the lowest temperatures. narrow temperature range of the present experiments and assum-
To further illustrate this, we have plotted the azimuthal temper- ing the wall heat flux to be constant, one can see that the bulk fluid
ature variations at the mid-height of the annulus for the inner and temperature gradient in the annulus would also be constant, so
outer cylinders in Figs. 5(a) and 5(b), respectively. The readings that the bulk fluid temperature would increase linearly with

122503-4 / Vol. 133, DECEMBER 2011 Transactions of the ASME

Downloaded From: https://heattransfer.asmedigitalcollection.asme.org on 06/30/2019 Terms of Use: http://www.asme.org/about-asme/terms-of-use


Fig. 4 Wall temperature measurements along the annulus for different eccentricities; *, ^,
and h denote the readings of thermocouples S0i , S90
i i
, and S180 , respectively, whereas 1, 3, and
are for the corresponding thermocouples on the outer cylinder; longer dashed lines corre-
00 00
spond to q * 5 0.53 and shorter ones to q * 5 1.39

increasing elevation. In this mid-section, the thermal field may be observation provides support for the approximate validity of this
assumed to be fully developed [23] and heat transfer to take place analysis and allows one to compute the mass flow rate as
entirely by convection so that
 q00 pðDo þ Di Þ
q00 ¼ h T w  Tb (8) m_ ¼ (10)
cp ðdT w =dzÞ
where h is the convective heat transfer coefficient, which would The wall temperature gradient was calculated by differentiation of
be constant in this section. Then, it follows that T w  Tb should sixth-order polynomials fitted to the azimuthally averaged wall
also be constant along this section, which implies that T w should temperature measurements. The results are shown in Fig. 7, in
increase linearly with elevation at the same rate as Tb , namely that which they have been normalized by the corresponding values at
z=H ¼ 0.5, shown in Fig. 8. It is clear that, at low eccentricities,
dT w dTb
¼ (9) there is a significant region of the test section that has an approxi-
dz dz mately constant wall temperature gradient. As eccentricity
As Fig. 6 shows, T w in the present apparatus increased fairly line- increases, however, this region diminishes in width and at the
arly with increasing elevation in the range 0.2 < z=H < 0.8. This highest eccentricities it ceases to exist. Moreover, the fact that the

Journal of Heat Transfer DECEMBER 2011, Vol. 133 / 122503-5

Downloaded From: https://heattransfer.asmedigitalcollection.asme.org on 06/30/2019 Terms of Use: http://www.asme.org/about-asme/terms-of-use


Fig. 7 Variation of normalized average wall temperature gradi-
ent with height; the vertical axis scale corresponds to the con-
centric case (e 5 0), but each curve for the increasingly
eccentric cases has been shifted upwards by three units com-
pared to the previous one (symbols same as in Fig. 6)

This expression shows that the Reynolds number would decrease


slowly with height along the annulus, due to the corresponding
increase in the viscosity of air. As shown in Fig. 9, the Reynolds
number at the inlet of the annulus increased with increasing heat
flux and generally decreased, although very slightly, with increas-
ing eccentricity. Its values were in the range between 670 and
1030, which for pipe flows extends within the laminar flow re-
gime. For the range of eccentricities between 0.1 and 0.9, the
Fig. 5 Azimuthal temperature variation for the inner (a) and maximum deviation of the Reynolds number from the one in the
outer (b) cylinders at z=H 5 0.5. , h, 3, 1, ^, *, and ^ repre- concentric case was 3% for q00 * ¼ 0.53 and 15% for q00 * ¼ 1.39.
sent eccentricities of 0, 0.1, 0.3, 0.5, 0.7, 0.8, and 0.9, respectively One may plausibly speculate that, if the heat flux were increased
further, transition to turbulence would occur. For a certain range of
mid-section wall temperature gradient decreases with increasing
heat fluxes and high eccentricities, it may even be possible that the
eccentricity for the two highest eccentricities brings the applic-
flow would be laminar in the narrow-gap and transitional or turbulent
ability of Eq. (10) to these cases into question. As the present
in the wide-gap. These speculations cannot be confirmed by the pres-
work focuses on the relative effects of eccentricity, we shall
ent experiments and there are no other experiments on natural con-
assume that Eq. (10) is approximately valid for all eccentricities
vection which may be used as evidence. Nevertheless, they are
as long as the wall temperature gradient is evaluated at z=H ¼ 0.5,
consistent with the results of experiments [8] and simulations [5] in
keeping in mind that the high-eccentricity results would be sub-
forced convection through eccentric annuli and similar geometries.
jected to additional uncertainty due to possible flow development,
which is not easy to estimate.
A Reynolds number for this flow may be defined as 4.2 Bulk Air Temperature, Heat Transfer Coefficient, and
Nusselt Number. Integration of Eq. (7) provides the bulk air
qb Vb Dh 4 m_ temperature increase in the thermally fully developed section of
Re ¼ ¼ (11) the apparatus as
lb p ðDo þ Di Þlb
q00 pðDo þ Di Þ
DTb ðzÞ ¼ Tb ðzÞ  Ta ¼ z  TL ðzÞ (12)
_ p
mc

Fig. 6 Azimuthally averaged temperature variation along the


annulus for various eccentricities. , h, 3, 1, ^, *, and ^ rep-
resent eccentricities of 0, 0.1, 0.3, 0.5, 0.7, 0.8, and 0.9, respec-
tively; longer dashed lines correspond to q00 * 5 0.53 and shorter Fig. 8 Variation of average wall temperature gradient with ec-
ones to q00 * 5 1.39. centricity at z=H 5 0.5; q00 * 5 0.53 (*) and 1.39 ()

122503-6 / Vol. 133, DECEMBER 2011 Transactions of the ASME

Downloaded From: https://heattransfer.asmedigitalcollection.asme.org on 06/30/2019 Terms of Use: http://www.asme.org/about-asme/terms-of-use


Fig. 9 Variation of inlet Reynolds number with eccentricity for
q00 * 5 0.53 (*) and 1.39 ()

where TL ðzÞ is the bulk temperature reduction due to cumulative


radiation and conduction losses to the surroundings up to the ele-
vation considered. It may be plausibly assumed that TL ðzÞ would
reach a constant value TL1 after some distance from the inlet, pro-
vided that exit loss effects are negligible at that elevation. Obser-
vation of the wall temperature measurements indicates that a
thermally fully developed state seemed to have been reached by
the flow at approximately z=H ¼ 0.3, except possibly for the two
highest eccentricities.
The average wall temperature and the bulk fluid temperature
have been plotted in Fig. 10. It can be seen in this figure that, as
the eccentricity increases, the average wall temperature increases Fig. 10 Variations of average wall temperature rise (n) and
significantly (by approximately 5 K for q00 * ¼ 0.53 and 9 K for bulk fluid temperature rise (~) with eccentricity at z=H 5 0.5;
q00 * ¼ 1.39), whereas the bulk fluid temperature changes very little q00 * 5 0.53 (a) and 1.39 (b)
(by 0.3 K for q00 * ¼ 0.53 and 2 K for q00 * ¼ 1.39). This means that,
although the heat power transferred from the wall to the fluid did
not change much with eccentricity for this geometry and range of bulk temperature, for the two heat fluxes at a height z=H ¼ 0.5 has
conditions, the heat transfer coefficient decreased substantially, so been plotted in Fig. 12. These figures show that at low eccentric-
that the wall temperature rose on the average and local hot spots ities a decrease in the heat transfer rate occurred in the narrow-
appeared in the narrow gap. In this respect, eccentricity dimin- gap region, but this was compensated for by an enhancement of
ishes the efficiency of annular channels as heat exchangers. heat transfer in the wide-gap region; as a result, the azimuthally
As explained in Sec. 2.3, an asymptotic bulk temperature cor- averaged heat transfer did not change significantly with increasing
rection due to inlet losses was estimated to be TL1 ¼ 0:11K for eccentricity. At high eccentricities, however, the heat transfer rate
q00 * ¼ 0.53 and 0.18 K for q00 * ¼ 1.39. With this correction, decreased throughout the annulus, leading to the overall decrease
Eq. (12) was used to determine the fluid bulk temperature in the in the average Nusselt number.
middle of the test section, and then Eq. (8) was used to estimate
4.3 Grashof and Rayleigh Numbers. The Grashof number
the average heat transfer coefficient in the thermally fully devel-
is a dimensionless parameter that is meant to represent the relative
oped section of the apparatus. A Nusselt number for the thermally
importance of buoyancy with respect to the viscous force. In cases
fully developed section of the annular duct may be defined as
in which heat is introduced in the form of uniform wall heat flux,
hDh
Nu ¼ (13)
k

where the thermal conductivity k is taken as that of air at the bulk


temperature.
Figure 11 shows the dependence of Nusselt number upon wall
heat flux and eccentricity. It can be seen that the Nusselt number
increased somewhat, but not very significantly, with increasing
wall heat flux. In contrast, the dependence of Nusselt number on
eccentricity was quite strong, as Nu decreased monotonically with
increasing eccentricity, although rather imperceptibly for very
low eccentricities. At higher eccentricities (0.3 < e < 0.9), the rate
at which the Nusselt number decreased with eccentricity appeared
to be constant for both wall heat fluxes. Remarkably, Nu for the
concentric case was about 1.9 times its value for the case with the
highest eccentricity considered (e ¼ 0.9) when q00 * ¼ 0.53, and 1.6
times the corresponding value when q00 * ¼ 1.39.
The azimuthal variation of the local Nusselt number, defined
using the difference between the local wall temperature and the Fig. 11 Nusselt number versus eccentricity for q00 * 5 0.53 (*)
bulk fluid temperature and the thermal conductivity of air at the and 1.39 ()

Journal of Heat Transfer DECEMBER 2011, Vol. 133 / 122503-7

Downloaded From: https://heattransfer.asmedigitalcollection.asme.org on 06/30/2019 Terms of Use: http://www.asme.org/about-asme/terms-of-use


it has been suggested [23] that a modified Grashof number,
defined as
gbq00 D4h
Gr ¼ (14)
k 2
should be used. This parameter is proportional to the heat flux and
insensitive to eccentricity, only subject to the weak effect of ec-
centricity on the values of the various thermophysical properties
that appear in Eq. (14). Figure 13 shows that, for eccentricities
between 0.1 and 0.9, this modified Grashof number deviated from
the one in the concentric case by up to 4% for q00 * ¼ 0.53 and
4.5% for q00 * ¼ 1.39. An increase of the average wall heat flux by
165% led to a 131% increase in the modified Grashof number.
The modified Rayleigh number, defined as Ra** ¼ Gr** Pr, in
which the Prandtl number was calculated at the bulk fluid temper-
ature, has the same dependence on the various parameters as the
modified Grashof number, because in the temperature range of the Fig. 13 Modified Grashof number versus eccentricity for
present experiments Pr was essentially constant; Ra** was q00 * 5 0.53 (*) and 1.39 ()
approximately 104,700 for q00 * ¼ 0.53 and 241,400 for q00 * ¼ 1.39.
Equation (14) does not expose readily the physical significance 5 Conclusions
of the Grashof number. We have derived the following alternative Natural convection in vertical concentric and eccentric annular
expression as a direct estimate of the ratio of the buoyancy and channels with approximately uniform wall heat flux has been stud-
the viscous force ied experimentally. Within the range of present conditions, and
with the possible exclusion of the highest eccentricities, it was
qb gbðTb  T1 ÞD3h pðDo þ Di Þ found that the flow was thermally fully developed in a consider-
Gr ¼ (15)
b m_ 16 able section of the apparatus, as indicated by the linear variation
of the wall temperature with height. This made it possible to esti-
Although, the values of Grashof number computed from Eq. (15) mate the mass flow rate from the wall temperature gradient at the
are several orders of magnitude smaller than those computed from center of the annulus, and use it to calculate the bulk Reynolds
Eq. (14), both expressions show that eccentricity effects on the number, which was found to be relatively insensitive to eccentric-
Grashof number are weak. ity for the geometry and range of heat flux used in our study.
It was found that the azimuthally averaged Nusselt number gen-
erally diminished with increasing eccentricity, although at low
eccentricities this decrease was negligible. It was also observed
that at low eccentricities, a decrease in the heat transfer experi-
enced in the narrow-gap region was compensated by an enhance-
ment of heat transfer in the wide-gap region. At higher
eccentricities, the heat transfer rate diminished, by comparison to
the concentric case, throughout the annulus.
The Grashof and Rayleigh numbers were found to be relatively
insensitive to eccentricity, which was attributed to the relatively
small change in the bulk fluid temperature as well as the Reynolds
number. However, an increase of the average wall heat flux led to
a nearly proportional increase in the modified Rayleigh number
and a weaker but still significant increase in the Grashof number.
In summary, the value of eccentricity under the current experi-
mental conditions did not affect significantly either the mass flow
rate or the bulk fluid temperature. Consequently, the eccentric
annulus has no advantage over the concentric one in terms of
pumping power and heat transfer effectiveness. On the contrary,
eccentricity results in local and average wall overheating, which
is normally an undesirable effect. On the positive side, the present
results show that heat transfer in eccentric annuli is insensitive to
eccentricity for small eccentricities. This implies that high preci-
sion is not necessary in manufacturing concentric channels
intended to be used for natural convection heat transfer.

Nomenclature
A¼ area
cp ¼ specific heat at constant pressure
D¼ diameter
Dh ¼ hydraulic diameter
e¼ eccentricity
g¼ gravitational acceleration
Gr ¼ Grashof number
Gr** ¼ modified Grashof number
Fig. 12 Variation of the local Nusselt number with azimuthal H¼ annulus height
position for different eccentricities for q00 * 5 0.53 (a) and 1.39 h¼ convection heat transfer coefficient
(b); symbols as in Fig. 6 k¼ thermal conductivity

122503-8 / Vol. 133, DECEMBER 2011 Transactions of the ASME

Downloaded From: https://heattransfer.asmedigitalcollection.asme.org on 06/30/2019 Terms of Use: http://www.asme.org/about-asme/terms-of-use


m_ ¼ mass flow rate [4] Merzari, E., and Ninokata, H., 2009, “Anisotropic Turbulence and Coherent
Structures in Eccentric Annular Channels,” Flow, Turbul. Combust., 82, pp.
Nu ¼ Nusselt number 93–120.
Pr ¼ Prandtl number [5] Ninokata, H., Merzari, E., and Khakim, A., 2009, “Analysis of Low Reynolds
q00 ¼ heat flux Number Turbulent Flow Phenomena in Nuclear Fuel Pin Subassemblies of
q00 * ¼ dimensionless heat flux Tight Lattice Configuration,” Nucl. Eng. Des., 239, pp. 855–866.
[6] Guellouz, M. S., and Tavoularis, S., 2000, “The Structure of Turbulent Flow in
r¼ radius a Rectangular Channel Containing a Cylindrical Rod—Part 1: Reynolds-Aver-
Ra ¼ Rayleigh number aged Measurements,” Exp. Therm. Fluid Sci., 23, pp. 59–73.
Ra** ¼ modified Rayleigh number [7] Guellouz, M. S., and Tavoularis, S., 2000, “The Structure of Turbulent Flow in
Re ¼ Reynolds number a Rectangular Channel Containing a Cylindrical Rod—Part 2: Phase-Averaged
Measurements,” Exp. Therm. Fluid Sci., 23, pp. 75–91.
S¼ temperature sensor location [8] Gosset, A., and Tavoularis, S., 2006, “Laminar Flow Instability in a Rectangu-
T¼ temperature lar Channel With a Cylindrical Core,” Phys. Fluids, 18(4), 044108.
V¼ velocity [9] Chang, D., and Tavoularis, S., 2006, “Convective Heat Transfer in Turbulent
x¼ horizontal coordinate in the direction of traverse Flow Near a Gap,” ASME J. Heat Transfer, 128, pp. 701–708.
[10] Chang, D., and Tavoularis, S., 2008, “Simulations of Turbulence, Heat Transfer
y¼ horizontal coordinate in the direction normal to the and Mixing Across Narrow Gaps Between Rod-Bundle Subchannels,” Nucl.
traverse Eng. Des., 238, pp. 109–123.
z ¼ axial coordinate (vertical) [11] Farouk, B., and Guceri, S. I., 1982, “Laminar and Turbulent Natural Convection
in the Annulus Between Horizontal Concentric Cylinders,” ASME J. Heat
Transfer, 104(4), pp. 631–636.
Subscripts [12] Kuehn, T. H., and Goldstein, R. J., 1978, “An Experimental Study of Natural
a ¼ ambient Convection Heat Transfer in Concentric and Eccentric Horizontal Cylindrical
b ¼ bulk Annuli,” ASME J. Heat Transfer, 100, pp. 635–640.
[13] Yeh, C. L., 2002, “Numerical Investigation of the Three-Dimensional Natural
i ¼ pertaining to the cylindrical core of the annulus Convection Inside Concentric and Eccentric Annuli With Specified Wall Heat
in ¼ inlet of annulus Flux,” JSME Int. J., Ser. B, 45(2), pp. 301–306.
L ¼ losses [14] de Vahl Davis, G., and Thomas, R. W., 1969, “Natural Convection Between
o ¼ pertaining to the surrounding cavity of the annulus Concentric Vertical Cylinders,” Phys. Fluids, 12(Suppl. II), pp. 198–207.
[15] Schwab, T. H., and De Witt, K. J., 1970, “Numerical Investigation of Free Con-
w ¼ test section wall vection Between Two Vertical Coaxial Cylinders,” AlChE J., 16(6), pp. 1005–
1010.
Greek Letters [16] Keyhani, M., Kulacki, F. A., and Christensen, R. N., 1983, “Free Convection in
a Vertical Annulus With Constant Heat Flux on the Inner Wall,” ASME J. Heat
a ¼ thermal diffusivity Transfer, 105, pp. 454–459.
b ¼ thermal expansion coefficient [17] Al-Arabi, M., El-Shaarawi, M. E. A., and Khamis, M., 1987, “Natural Convec-
h ¼ azimuthal angle tion in Uniformly Heated Vertical Annuli,” Int. J. Heat Mass Transfer, 30(7),
l ¼ dynamic viscosity pp. 1381–1389.
[18] El-Shaarawi, M. A. I., and Mokheimer, E. M. A., 1998, “Free Convection in
 ¼ kinematic viscosity Vertical Eccentric Annuli with a Uniformly Heated Boundary,” Int. J. Numer.
q ¼ density Methods Heat Fluid Flow, 8(5), pp. 448–503.
Other Notation [19] El-Shaarawi, M. A. I., and Mokheimer, E. M. A., 1999, “Developing Free Con-
() ¼ time average
vection in Open-Ended Vertical Eccentric Annuli With Isothermal Boundaries,”
ASME J. Heat Transfer, 121, pp. 63–72.
[20] Sparrow, E. M., Chrysler, G. M., and Azevedo, L. F., 1984, “Observed Flow
Reversals and Measured-Predicted Nusselt Numbers for Natural Convection in
References a One-Sided Heated Vertical Channel,” ASME J. Heat Transfer, 106, pp. 325–
[1] Dimmick, G. R., Chatoorgoon, V., Khartabil, H. F., and Duffey, R. B., 2002, 332.
“Natural-Convection Studies for Advanced CANDU Reactor Concepts,” Nucl. [21] Modest, F., 2003, Radiative Heat Transfer, 2nd ed., Academic, New
Eng. Des., 215(1–2), pp. 27–38. York.
[2] Piot, E., and Tavoularis, S., 2011, “Gap Instability of Laminar Flows in Eccen- [22] Choueiri, G. H., 2009, “An Experimental Study of Natural Convection in Vertical
tric Annular Channels,” Nucl. Eng. Des., (in press). Open-Ended Concentric and Eccentric Annuli,” M.A.Sc. Thesis, Department of
[3] Merzari, E., Wang, S., Ninokata, H., and Theofilis, V., 2008, “Biglobal Linear Mechanical Engineering, University of Ottawa, Ottawa, Ontario, Canada.
Stability Analysis for the Flow in Eccentric Annular Channels and a Related [23] Incropera, F. P., and Dewitt, D. P., 2002, Fundamentals of Heat and Mass
Geometry,” Phys. Fluids, 20, 114104. Transfer, 5th ed., Wiley, New York.

Journal of Heat Transfer DECEMBER 2011, Vol. 133 / 122503-9

Downloaded From: https://heattransfer.asmedigitalcollection.asme.org on 06/30/2019 Terms of Use: http://www.asme.org/about-asme/terms-of-use

You might also like