You are on page 1of 10

Experimental Thermal and Fluid Science 34 (2010) 528537

Contents lists available at ScienceDirect

Experimental Thermal and Fluid Science

journal homepage:

The effect of corona discharge on free convection heat transfer

from a horizontal cylinder
Seyed Reza Mahmoudi a,*, Kazimierz Adamiak a, Peter Castle a, Mehdi Ashjaee b
Department of Electrical and Computer Engineering, University of Western Ontario, London, Ont., Canada
Department of Mechanical Engineering, Faculty of Engineering, University of Tehran, Tehran, Iran

a r t i c l e i n f o a b s t r a c t

Article history: Free convection heat transfer from an isothermal horizontal cylinder in the presence of DC positive cor-
Received 23 July 2009 ona discharge with a blade edge emitter electrode has been studied experimentally and numerically. A
Received in revised form 12 November 2009 MachZehnder interferometer was used to determine the local Nusselt numbers. The effect of corona dis-
Accepted 13 November 2009
charge on heat transfer from the cylinder was investigated at Rayleigh numbers in the range between
1500 and 5000. To nd the details of the ow patterns and to further verify the experimental results,
numerical simulations were also performed. It was found that the numerical results are in good agree-
ment with experimental data. By increasing the applied voltage up to 15.5 kV, the corona discharge gen-
Corona discharge
Ionic wind
erates a recirculation zone around the blade and below the lower stagnation point of the cylinder. The
Free convection effect of the recirculation zone becomes stronger near the breakdown voltage (17 kV) and it is responsi-
Interferometry ble for a local decrease in the cooling of the cylinder around the lower stagnation point. The results indi-
cate that corona discharge has a signicant effect on the average Nusselt number at lower Rayleigh
numbers whereas it has smaller effect at higher Rayleigh numbers.
2009 Elsevier Inc. All rights reserved.

1. Introduction Yabe et al. [2] investigated the enhancement of heat transfer

from a horizontal downward-facing plate with 40 lm wire elec-
Establishment of corona discharge between a sharp electrode trode in a Plexiglass chamber connected to a high voltage source.
and a grounded heated surface usually induces an ionic wind A Langmuir probe was used to measure the space charge distribu-
whose momentum can be used for enhancement of heat transfer tion and to verify the numerical results. They were able to model
from the heated surface. The corona discharge and ionic wind are the recirculation ow inside the chamber and achieved a good
caused by the ionization of air molecules in the intense electric agreement with their experimental results.
eld region around the sharp electrode that accelerates ions and OBrien and Shine used an interferometer to measure the
drags the air molecules toward the grounded surface. Electrostatic ElectroHydroDynamic (EHD) enhancement of local heat transfer
cooling devices operate without any moving parts, which reduces from a vertical isothermal at plate for various air pressures [3].
the mechanical vibration, its energy losses, and associated noises. They concluded that the boundary layer was distorted in the vicin-
Therefore, the electrostatic cooling is potentially attractive. ity of the plate surface and the heat transfer coefcient increased
Corona enhancement of heat transfer from various objects has with corona current.
been the subject of many investigations because of a comparatively Franke and Hogue [4] studied the corona wind effect on a
signicant local enhancement of heat transfer coefcient with heated horizontal cylinder using both a multi-emitter electrode
small corona discharge power consumption. For the rst time, and a wire electrode. The average heat transfer enhancement
Marco and Velkoff [1] reported a 500% enhancement of average was quantitatively calculated by a heat-balance method and qual-
heat transfer coefcient and even higher local enhancements from itatively with a MachZehnder interferometer. Enhancement of
a vertical heated plate in the presence of corona wind. The momen- heat transfer from the cylinder due to the corona discharge was re-
tum integral analysis was carried out to further verify the experi- ported as much as six times the free convection heat transfer. The
mental results. The theoretical results were in good agreement authors found that the multi-emitter electrode was more effective
with experimental studies. compared with the stretched wire electrode. However, they did not
report the local heat transfer coefcient around the cylinder.
Owsenek et al. [5] experimentally investigated the corona wind
* Corresponding author. heat transfer enhancement from a horizontal at plate. Needle
E-mail address: (S.R. Mahmoudi). electrodes with different height were used as the corona electrode.

0894-1777/$ - see front matter 2009 Elsevier Inc. All rights reserved.
S.R. Mahmoudi et al. / Experimental Thermal and Fluid Science 34 (2010) 528537 529


b ion mobility (m2 V1 s1) T temperature (K)

C GladstoneDale coefcient (m3 Kg1) u velocity eld vector (m s1)
Cp specic heat (J kg1 K1)
D diameter of cylinder (m) Greek symbols
Di ion diffusion coefcient (m2 s1) a thermal diffusivity (m2 s1)
E electric eld (V m1) b volumetric thermal expansion coefcient (K1)
f electric body force (N m3) e shift fringe
eo vacuum permittivity  air permittivity (A s V1 m1)
g gravitational acceleration (m s2)
u electric potential (V)
h heat transfer coefcient (W m2 K1) uo voltage applied to corona electrode (V)
Ic corona current (A) k laser wave length (m)
J current density (A m2) q air density (kg m3)
kf thermal conductivity (W m1 K1) qc charge density (C m3)
l length of cylinder (m) r electrical conductivity (X1 m1)
Nu Nusselt number m kinematics viscosity (m2 s1)
P pressure (Pa)
r radial distance from the cylinder surface (m) Subscripts
Ro gas constant (J kg1 K1) S referred to surface
Ra conventional Rayleigh number based on temperature, h referred to periphery angle
gbT s  T 1 D3 =v a 1 referred to innite medium

An enhancement of more than 25:1 over free convection was re- selected as the test section, because the circular tube is widely
ported. More recently [6], the same authors continued their theo- used in typical heat exchangers. Considering the limitations of
retical and experimental studies and focused on a comparison the interferometer plates, the diameter of the cylinder and the sur-
between single and multiple wire electrodes. They concluded that face temperature were selected to obtain an appropriate range of
for a given applied voltage, multiple wire electrodes yield smaller Rayleigh number (1500 < Ra < 5000) which is frequently encoun-
heat transfer enhancement per wire electrode than that of a single tered in conventional HVAC systems. In order to generate a uni-
wire electrode. It was also found that dual recirculation may be form distribution of discharge effects along the laser beam which
formed between wire electrodes. Numerical simulations of novel is needed for the two-dimensional interferometric studies, a sharp
electrode geometries revealed that this inefciency may be elimi- tip blade was used as a high voltage corona electrode. Moreover,
nated through the use of a blade geometry. compared with other electrode geometries, alignment of the blade
Franke [7] investigated heat transfer from a vertical plate in the electrode can be achieved more precisely along the laser beam. In
presence of corona-generated vortices. The vortices were gener- order to nd the details of the ow patterns and to further verify
ated by the corona discharge from parallel wires supplied with the experimental results, a numerical simulation was also
alternate high voltage and ground electrodes. He also visualized performed.
the thermal boundary layer using a MachZehnder interferometer
and measured the resulting heat transfer through both an energy 2. Experimental setup
balance and interferograms. A doubling of the convective compo-
nent of total heat transfer with an applied corona voltage above The apparatus used in this experiment consisted of three major
the corona onset value was reported. components: (i) an isothermally heated surface, heating facilities,
The literature review shows that nding the average heat trans- instrumentation, and control, (ii) high voltage apparatus and mea-
fer rate for external free convection ows in the presence of corona surement devices and (iii) temperature eld visualization system.
discharge has been the subject of several investigations. Various The schematic of the experimental setup and apparatus is pre-
heated test sections, such as horizontal plate [8,9], vertical plate sented in Fig. 1a.
[10], horizontal tube bank [1114], were investigated. Although The test section was an extruded aluminum hollow cylinder
enhancement of free convection heat transfer through corona dis- with a highly polished outer surface. The details of the test section
charge has been widely studied, few experimental studies have are shown in Fig. 1b. The length of the cylinder was chosen as
been performed to nd the local heat transfer rate for external free 160 mm (16 times of diameter) which causes the induced ow to
convection ows in the presence of corona discharge. Moreover, be two-dimensional. In order to minimize the thermal end effects,
previous studies on ionic wind-enhanced heat transfer show that two wooden ber end caps with k = 0.05 W/m2 K were installed at
the corona discharge has comparatively smaller effect on high- both ends of the test section. In order to position the cylinder in the
Reynolds force convection [15]. However, investigation of the horizontal direction, two plastic rods were connected to these end
interaction between the corona discharge and buoyancy-driven caps and installed on the XYZ holders. The cylinder had an axial
ow has not been studied. cavity to facilitate installation of a coiled nickelchromium wire
In most of the previous interferometric studies, the EHD heater element at the center of the cavity. The space between the
enhancements in heat transfer investigations have been limited heater coil and inner wall of the hollow cylinder was lled with
to a qualitative visualization of the thermal boundary layer. In magnesium oxide powder. Four holes with 0.55 mm diameter were
those studies, the average Nusselt number was determined using drilled into the base surface of the cylinder at 30, 50, 60 and 80 mm
the energy balance method. This investigation involves a quantita- from the base ends in different peripheral angles. Four calibrated
tive interferometric study of the positive corona discharge on the K-type thermocouples were inserted and xed in these holes. All
local heat transfer enhancement from an isothermal horizontal the temperatures were monitored continuously by a TESTO 177
cylinder for different Rayleigh numbers. A circular cylinder was data logger which was connected to a PC. The maximum difference
530 S.R. Mahmoudi et al. / Experimental Thermal and Fluid Science 34 (2010) 528537

Fig. 1a. Schematic of experimental setup.

Fig. 1b. Details of the cylinder and the blade electrode.

in surface temperature readings at the four thermocouple junc- maximum differences in surface temperature readings of four ther-
tions were about 0.2 before tests. Furthermore, the outer surface mocouple junctions were about 0.2. A high voltage DC power sup-
temperature of the cylinder was measured at 20 different locations ply (030 kV, 3 mA) was used to generate the positive DC corona
both in the peripheral and longitudinal directions with a calibrated discharge. The blade electrode was xed 40 mm below the cylinder
hand held digital thermometer. The difference between the read- and connected to the positive terminal of the high voltage DC
ings did not exceed 0.2. The combination of the relatively thick- power supply. This comparatively large spacing provided larger
walled aluminum cylinder and the high thermal conductivity of voltage range between corona onset and breakdown allowing lar-
magnesium oxide powder helped achieve the desired uniformity ger number of different voltages to be tested. Parallelism and
of the surface temperature both peripherally and longitudinally. adjustments of the blade with respect to the cylinder were done
The electrical power supplied to the heater was controlled by a using optical positioners.
variable transformer, therefore it was possible to achieve different A MachZehnder interferometer with 100 mm diameter optical
surface temperatures and different Rayleigh numbers. The electri- plates and 100 mmW HeNe laser light source was used to visual-
cal input power was varied to keep the surface temperature con- ize the temperature eld around the cylinder. A schematic of the
stant in the presence and the absence of corona discharge. After interferometer is shown in Fig. 1c. Innite fringe interferograms
changing the power level, it took about one hour to obtain steady were digitized by a Panasonic WV-CP410-1/3 CCD camera which
state condition with quasi-constant surface temperature in the ab- was connected to a PC. In this set up, the laser beam splits into
sence of corona discharge. For the corona discharge experiments it two beams of approximately equal intensity. A reference beam
took about 20 min to obtain quasi-constant surface temperature. passes through the thermally undisturbed ambient air, and the test
At the steady state condition and before initiating the tests, the beam passes through the thermally disturbed air adjacent to the
S.R. Mahmoudi et al. / Experimental Thermal and Fluid Science 34 (2010) 528537 531

Fig. 1c. Plan view of Match-Zehnder interferometer.

heated cylinder. Because the refractive index varies with tempera- j can be calculated by the measuring of the distance of the
dr rD=2;h
ture, the two laser beams are no longer in the same phase when fringes from the cylinder surface. The temperature of the farthest
they are recombined. This phase shift produces an interference fringe from the cylinder surface was logged by a K-type thermocou-
pattern in the optical output of the MachZehnder interferometer, ple. The order number of this fringe is set to zero (e = 0) and is consid-
which can be recorded with a CCD camera. When these two beams ered as the reference fringe. The reference temperature readings, Tref,
are parallel upon recombination, the constructive and destructive near the heated surface are slightly higher than the innite tempera-
interference fringes correspond directly to the isotherms in the ture readings which were located completely far from the test section.
ow eld. Further adjustments of the test section and its parallel- The local Nusselt number can be written as:
ism along the laser beam were examined by the laser beam, CCD 
camera, and a simple image processing procedure. D 6ClP 1 kRo T 2ref de
Nuh    5
T S  T 1 3ClP1  2kRo T ref e2 dr rD=2;h
2.1. Experimental procedure and data reduction method
The average Nusselt number can be calculated as:
The test procedure was initiated by testing the free convection Z p
heat transfer from the horizontal cylinder in the absence of corona Nu Nuh dh 6
p 0
discharge. After reaching the steady state condition, interfero-
grams were taken. The corona discharge was generated near the The average Nusselt number was obtained by integrating the
cylinder by applying the high voltage to the blade in the range local Nusselt number over the left hand side and right hand side
between onset of corona discharge and breakdown voltage. At each of the cylinder for each interferogram. Comparing the average
voltage, ve interferograms were recorded for assurance of the Nusselt numbers calculated for the both sides show maximum
experiment repeatability and data reduction. Fig. 2 represents 4.5% discrepancy at (Ra = 5000 and uo = 17.0 kV) which is within
examples of the innite fringe interferograms for two extreme the uncertainty band. In some cases, the experimental asymmetry
cases, at lowest and highest Rayleigh numbers, for different observed in the interferograms may be greater than the thickness
applied voltages. The atmospheric pressure and the relative of the fringe. However, in order to get more accurate local Nusselt
humidity of the laboratory were recorded during all the experi- number, the middle of the fringes was recorded instead of their
ments. The innite temperature T1 was monitored at two different edges.
vertical locations about 500 mm away from the test section and
they both indicated the same value. A code was developed to cal- 2.2. Uncertainty analysis
culate the local Nusselt number around the cylinder using innite
fringe interferograms. The data reduction procedure was similar to The experimental uncertainty of the local Nusselt number was
the work done by Ashjaee et al. [16]. The local Nusselt number is performed using the ASME guidelines on reporting the uncertain-
calculated as follows: ties in experimental measurements and results [19]. The propaga-
hh D tion equation of Kline and McClintock for local Nusselt number
Nuh 1 calculation from Eq. (5) can be written as:
 2  2  2  2
The local heat transfer coefcient can be obtained as: @Nuh @Nuh @Nuh @Nuh
  P2Nuh P2T 1 P2T S P2T ref P2P1
dT 1 @T 1 @T S @T ref @T 1
hh kf  2  2  2  2
dr rD=2;h T S  T 1 @Nuh @Nuh @Nuh
P2l P2D P2Dr
@l @D @Dr
The local temperature gradient can be calculated from innite
fringe interferograms as:
   The precision limit of the experimental values and associated
dT  dT  de uncertainties of Ra = 1500 is listed in Table 1. The precision limit
dr rD=2;h de rD=2;h dr rD=2;h of Dr is related to the digitized interferogram readings of the radial
fringes distance from the cylinder surface. The precision limits for
The fringe shift due to the spatial temperature difference in the universal gas constant, the GladstoneDale constant, the laser
interferometer test arm can be obtained as [17,18]: wave length, and the fringe shifts have been neglected. The bias
 error for the thermocouples, pressure gage, voltmeter, microam-
dT  6ClP 1 kRo T 2ref
4 meter, and micrometer was negligible because they had been
de rD=2;h 3ClP 1  2kRo T ref e2
calibrated. A MachZehnder interferometer encounters various
532 S.R. Mahmoudi et al. / Experimental Thermal and Fluid Science 34 (2010) 528537

Fig. 2. Examples of interferograms in the absence and presence of corona.

Table 1
sources of bias errors including refraction errors in thermal Deviation of each quantity for computing local Nusselt number Eq. (5).

boundary layer, misalignment of test section, and light distortion Measurement parameter (xi) xi P xi @Nuh Px
 @Nui h
errors due to the test section end effects. The systematic errors,
BNus0 associated with the temperature gradient measurement using T1 296.7 K 0.1 K 0.89%
Ts 318367.8 K 0.1 K 0.89
MachZehnder interferometer are summarized in Table 2. Tref 297.1 K 0.1 K 0.52%
The theory
r of the propagation of uncertainty (U Nu P1 87,200 Pa 100 Pa 0.11%
u0 12 kv17 kv 100 V
P2Nu B2Nu ) was applied to evaluate the 95% condence uncer- Ic 0.512 lA 0.5 lA
tainty of the local Nusselt numbers. The maximum uncertainty l 160.1 mm 2  105m 0.01%
D 10 mm 2  105m 0.12%
associated with the Rayleigh number, local Nusselt numbers, and
Dr 0.24 mm 6.3  105m 4.8%*
average Nusselt numbers were estimated to be 3.73%, 6.28%, and
4.85% respectively. Maximum precision error at h = 0 at u0 = 17.0 kV.
S.R. Mahmoudi et al. / Experimental Thermal and Fluid Science 34 (2010) 528537 533

Table 2 Table 3
Bias errors of MatchZehnder interferometer. Grid sensitivity analysis at u0 = 17.0 kv and Ra = 5000.

Type of bias error BNu (%) Number of elements Nu Percentage of change (abs)
Refraction (Ref. [20]) 0.07 4700 5.9522 2.4
End effect [20] 1.67* 10,900 5.8118 0.02
Misalignment [21] 0.4 14,500 5.8109 0.0017
* 17,320 5.8108
Maximum errors at Ra = 5000.

modeled as a symmetry boundary condition and set at a height

3. Numerical model
ten times the expected boundary layer thickness at the domain
exit. Two common boundary conditions exist for the Poisson equa-
The free convection heat transfer in the presence of corona dis-
tion. A specied potential value (Dirichlet boundary condition) was
charge is a multiphysics problem. The combination of buoyancy
used at the grounded electrode and the emitter electrode. The cyl-
body force and corona-driven ow can potentially induce various
inder electrode was grounded and modeled with a zero value,
ow patterns. Because the interferometric study only provides an
whereas the 15 mm of the blade electrode which was exposed to
interstitial temperature gradient, the numerical study was per-
the air was given the applied potential. The blade support was as-
formed to nd the ow pattern and further verify the experimental
sumed to be insulating (Neumann boundary condition) due to the
results. Commercial software COMSOL was used to solve the gov-
dielectric plexiglass used in the experiment. Although the real
erning equations. The electric body force can be expressed as [22]:
   computational domain for electric eld extends to innity, truncat-
1 1 @ eo ing boundaries were considered in order to limit the computa-
f qc E  jEj2 $eo r jEj2 q 7
2 2 @q T tional domain. These articial boundary conditions were set as
zero normal electric eld (Neumann boundary condition).
The rst term on the right side of the Eq. (7) is the electropho-
The charge transport equation was also simulated with two
retic, or Coulombic, force that results from the net free charges in
types of boundary conditions. Along the symmetry line, dielectric
gas. The second term, known as the dielectrophoretic force, arises
surface and truncating boundaries, normal current density was as-
from permittivity gradients. The last term is called the electrostric-
sumed to be zero and known space charge density for the blade
tive force, is important only for compressible uids. The corona
electrode was assumed to be emitted from the tip of the electrode.
wind arises completely from the electrophoretic force term. There-
The diffusive ux of the charge carrier assumed to be zero over the
fore, the rst term contributes to the corona wind generation. The
cylinder surface [24]. The microammeter readings were used to
bulk ow is laminar and two-dimensional. The buoyancy effect is
determine the total current ow across the cylinder surface for dif-
estimated using the Boussinesq approximation. The governing
ferent applied voltages. The space charge assumption on the blade
equations are as follows [23]:
electrode corrected and numerical total current ow calculated
r  qu 0 8 over the cylinder surface. The numerical total current compared
2 with the microammeter readings and solution procedure contin-
r  quu $p mr u gbT  T 1 qc E 9
ued until reaching the same numerical value.
q bjEj2 Eqs. (8)(12) were solved in a two-dimensional domain using
u  rT ar2 T c 10
qC p the commercial nite element-based software COMSOL. The
charge transport and Poissons equations were solved using the
The second term on the right hand side is Joule heating heat
general form PDE solver. The momentum and energy equations
source caused by ionic current.
were solved with PDE general form solvers as well. The solution
The electric eld around the sharp tip, which is responsible for
procedure was initiated by solving Poissons equation for the elec-
the free ion generation, is distorted by the free charges in the ion-
tric eld and the charge transport equation in order to determine
ized medium and is governed by Poissons equation as:
the electric body force. Next, the momentum equation was solved
qc considering the electric body force. The solution procedure was
$  E $  ru 11
eo continued by the solution of the energy equation. The momentum
The generated ions are moved from the high voltage electrode equation was solved by considering the electric body force and
toward the grounded surface through the electric force. The trans- buoyancy body force. An unstructured mesh was used with
port equation for conservation of ions is governed by: increased mesh densities in the area of the intense electric eld
around the blade tip and in the boundary layer region around
$J0 12 the cylinder. In order to ensure that the numerical results are
where current density is dened as: independent of the computational grid, a grid sensitivity analysis
was performed. The details of the grid sensitivity analysis in the
J qc bE  Di rqc 13 extreme case (uo = 17.0 kV, I = 10 lA and Ra = 5000) are given in
The ion mobility and ion diffusion coefcient are assumed as the Table 3.
b = 1.88  104 m2 V1 s1 and Di = 3.50  105 m2 s1, respectively.
4. Results and discussion
3.1. Boundary conditions
In order to test the accuracy of the experimental results, and the
The 2D uid ow was simulated using the steady NavierStokes data reduction procedure, the local Nusselt numbers were
equation in which the source terms are electric body force and obtained for an isothermal horizontal cylinder in absence of corona
buoyancy force and the cylinder was modeled as a no-slip and iso- wind at Ra = 103. Furthermore, the computational results repro-
thermal surface. A symmetry condition (zero normal gradient) was duced for the same geometry at Ra = 103. Fig. 3 shows the compar-
implemented along the vertical line of symmetry and the equa- ison of the experimental and numerical local Nusselt numbers for
tions were solved for half of the domain. The far eld was also the horizontal isothermal cylinder with previous works [2528] at
534 S.R. Mahmoudi et al. / Experimental Thermal and Fluid Science 34 (2010) 528537

in good agreement with experimental data at various applied volt-

ages and show a maximum difference of 2.1%. By increasing the
blade voltage from 0.0 kV to 11.0 kV, no charge was injected to
the air and the electrical current is negligible and no change in lo-
cal Nusselt number occurs. When the blade electrode is supplied
with 12.0 kV, the corona discharge is initiated, which was con-
rmed by audible acoustic noise, and the measured current at this
voltage was Ic = 1 lA. At the onset of corona discharge, corona-dri-
ven ow is established around the cylinder and caused the abrupt
increment of local Nusselt number especially around the lower
stagnation point (h = 0). The enhancement of local Nusselt number
at the onset of corona discharge is 23% for the lower stagnation
point. Because the corona current is comparatively small at the on-
set of corona, the ionic wind does not have enough momentum to
enhance the local Nusselt number around the upper stagnation
point. As shown in Fig. 2b, it can be found qualitatively that there
is no signicant displacement of fringe toward the cylinder surface
at h = 180. As depicted in Fig. 5, the corona wind has a weak effect
on temperature gradient at h > 150.
By increasing the voltage, the corona current increases. The
increment of corona current results in stronger ionic wind and
Fig. 3. Experimental and numerical local Nusselt number for the free convection higher heat transfer rate and the enhancement can be seen in all
heat transfer from the horizontal cylinder compared with other works [2528] at
peripheral positions around the cylinder. For example, for
Ra = 1000.
Ra = 1500 with uo = 17.0 kV, the enhancement of local Nusselt
number at h = 180 is 78% compared with the free convection.
In the high electric eld region, the space charge concentration
Ra = 103. Our experimental and computational results are in good can generate intense electric body force. However at voltages
agreement with previous studies showing a maximum difference around u = 15.5 kV and I = 7 lA a small local decrease in local Nus-
of 5.6% and 5.3%, respectively. The numerical deviations fall well selt number occurs around h = 30 the near the lower stagnation
within the experimental uncertainty bands. point. By increasing the voltage to uo = 17.0 kV and I = 12lA, this
The numerical solution for the electric potential and the distri- reduction in local cooling becomes greater. The numerical results
bution of charges at uo = 17.0 kV, I = 10 lA are shown in Fig. 4. The predict the same local decrease in local Nusselt number around
concentration of ions around the blade tip is highest, where the h = 30. In order to understand this local reduction of the Nusselt
electric eld is strongest. Fig. 5 represents the experimental results number around the lower stagnation point, detailed information
for the variation of the local Nusselt number around the cylinder in of the ow pattern is useful. Fig. 6 shows the numerical results
the absence and presence of the corona discharge at Ra = 1500. The for the streamline patterns and isotherms around the cylinder for
numerical results are also compared with the experimental data Ra = 1500 at uo = 12.0 kV and 17.0 kV. As seen in this gure, at
for further verication. The numerical local Nusselt numbers are uo = 12.0 kV and I = 1 lA the ow pattern around the cylinder

Fig. 4. (a) Numerical results for ion concentration and (b) electric potential distribution in the blade-cylinder geometry at u0 = 17.0 kV, I=10 lA.
S.R. Mahmoudi et al. / Experimental Thermal and Fluid Science 34 (2010) 528537 535

and results a decrease in local Nusselt number at the same loca-

tion. Stronger recirculation is produced by higher corona current.
Therefore, a greater local decrease in Nusselt number near the low-
er stagnation point is expected at higher corona currents. As shown
in Fig. 2c, the visual effect of recirculation zone and its impact on
the last fringe can be qualitatively recognized for uo = 15.5 kV
and 17.0 kV at Ra = 1500.
In order to investigate the corona wind effects at higher Ra
numbers, the results of experimental and numerical investigations
for Ra = 5000 are presented in Fig. 7. A similar trend in the varia-
tion of local Nusselt numbers can be seen as in Fig. 6. When
Ra = 5000, the enhancement of heat transfer at lower stagnation
point for uo = 17.0 kV, I = 10 lA is only 10% less than the enhance-
ment of heat transfer at Ra = 1500. In contrast, the local Nusselt
number at the upper stagnation point at Ra = 5000 and uo = 17.0 kV
shows the enhancement 200% less than that of Ra = 1500. The com-
paratively small enhancement of the local Nusselt number at
Ra = 5000 with respect to the Ra = 1500 is mainly due to the dom-
inancy of buoyancy-driven ow regime around the cylinder over
the corona wind driven ow at higher Ra numbers.
Fig. 8 shows the variation of average Nusselt numbers versus Ra
Fig. 5. Experimental and numerical local Nusselt number for the horizontal for different corona discharge voltages. It should be noted that for
cylinder for various corona voltages at Ra = 1500.
higher Ra numbers, the EHD effect on the average Nusselt number
decreases at the same corona voltage and spacing. At higher Ra
numbers, buoyancy-driven ow becomes stronger compared to
remains streamwise. However, for voltages above the uo = 15.5 kV corona- driven ow. Therefore, the trend of average Nusselt num-
and I = 6 lA, the electric body force becomes strong enough to pro- ber enhancement versus Ra for a specic applied voltage is a
duce high velocity gradients near the blade. This strong velocity decreasing function. The maximum enhancement of the average
gradient results in a recirculation zone establishing below the cyl- Nusselt number around the cylinder occurs at the lowest Ra num-
inder. The recirculation zone causes an insufcient entrainment of ber (Ra = 1500) and 17 kV applied voltage. As expected, at a specic
the cool fresh air into the thermal boundary layer at 0 < h < 80 voltage the average Nusselt number is a decreasing function of the

Fig. 6. Numerical results for isotherms and streamlines for Ra = 1500 at u0 = 12.0 kV and 17.0 kV.
536 S.R. Mahmoudi et al. / Experimental Thermal and Fluid Science 34 (2010) 528537

Fig. 7. Experimental and numerical local Nusselt number for the horizontal Fig. 9. VI. Characteristic diagram for blade and cylinder geometry with 40 mm
cylinder for various corona voltages at Ra = 5000. spacing at different Ra numbers.

results show that the effect of Joule heating does not cause temper-
ature increments of air temperature around the intense electric
eld zone. It is mainly because of the low level of current density
which is unable to produce signicant thermal effects in surround-
ing air. At higher Ra numbers, the effect of Joule heating inside the
thermal boundary layer becomes even less important because the
electrical conductivity of air is inversely proportional to the

5. Conclusions

According to the results presented in this paper, the enhance-

ment of heat transfer through the corona discharge is stronger
for lower Ra numbers. The enhancement of heat transfer is a
decreasing function of Ra number. At a specic voltage with a xed
spacing between electrodes, a recirculation zone which is created
by the electric body force causes the local decrease in heat transfer
from the cylinder. By increasing the corona voltage, the stronger
recirculation zone is generated and causes a larger local decrease
in heat transfer rate. Systematic study is needed to nd the opti-
Fig. 8. Experimental results for the average Nusselt number enhancement of heat mum electrode spacing and the applied voltage in the various
transfer at various applied voltages. ranges of Ra numbers for single or array of cylinder. This knowl-
edge can be implemented for high performance ionic heat exchan-
ger design.
Ra. By increasing the applied voltage, the slope of average Nusselt
function becomes steeper because the electric body force decreases References
for higher Ra numbers.
The observations showed that the electric body force is affected [1] S.M. Marco, H.R. Velkoff, Effect of electrostatic eld on free convection heat
by the temperature eld around the heated cylinder. The VI char- transfer from at plates, ASME Paper No. 63-HT-9, 1963.
[2] A. Yabe, Y. Mori, K. Hijitata, EHD study of the corona wind between wire and
acteristics for Ra = 0, Ra = 1500, and Ra = 5000 are shown in Fig. 9. plate electrodes, AIAA Journal 164 (1978) 340345.
The thermal effect causes the change in the trend of the VI char- [3] J. OBrien, A.J. Shine, Some effects of an electric eld on heat transfer from a
acteristics at different Rayleigh numbers. For example, at Ra = 5000 vertical plate in free convection, Journal of Heat Transfer (February) (1967)
and uo = 15.5 kV, the discharge current is 15% less than that of [4] M.E. Franke, L.E. Hogue, Electrostatic cooling of a horizontal cylinder,
Ra = 1500. This may be attributed to the ion mobility temperature Transaction of ASME 113 (1991) 544548.
dependence. Ion mobility and current density decrease with [5] B.L. Owsenek, J. Seyed-Yagoobi, R.H. Page, Experimental investigation of
corona wind heat transfer enhancement with a heated horizontal at plate,
increasing the surface temperature of the ground electrode [29].
Journal of Heat Transfer 177 (1995) 309315.
In other words, the thermal effect reduces the current density, [6] B.L. Owsenek, J. Seyed-Yagoobi, Theoretical and experimental study of
the corona power consumption, the electric body force, and local electrohydrodynamic heat transfer enhancement through wire-plate corona
Nusselt number enhancements for a specic corona voltage. discharge, Transaction of the ASME 119 (1997) 604610.
[7] M.E. Franke, Effect of vortices induced by corona wind discharge on free-
The effect of Joule heating is numerically investigated through convection heat transfer from vertical plate, Journal of Heat Transfer (1969)
considering the source term in the energy equation. The numerical (No. 68-WA/HT-9).
S.R. Mahmoudi et al. / Experimental Thermal and Fluid Science 34 (2010) 528537 537

[8] H. Kalman, E. Sher, Enhancement of heat transfer by means of a corona wind [18] W. Hauf, U. Grigull, Optical Methods in Heat Transfer, Advances in Heat
created by a wire electrode and conned wings assembly, Applied Thermal Transfer, vol. 6, Academic Press, New York, 1970. p. 133366.
Engineering 21 (2001) 265282. [19] R.J. Moffat, Using uncertainty analysis in the planning of an experiment, ASME
[9] A. Rashkovan, E. Sher, H. Kalman, Experimental optimization of an electric Journal of Fluids Engineering 107 (1985) 173178.
blower by corona wind, Applied Thermal Engineering 22 (2002) 1587 [20] J.M. Mehta, W.Z. Black, Errors associated with interferometric measurement of
1599. convective heat transfer coefcients 16 (6) (1977) 17201726.
[10] Y. Yonggang, H. Junping, A.I. Zhongliang, Y. Lanjun, A. Qiaogen, Experimental [21] Ronald D. Flack Jr., MachZehnder interferometer errors resulting from test
studies of the enhanced heat transfer from a heating vertical at plate by ionic section misalignment, Applied Optics 17 (1978) 985987.
wind, Plasma Science & Technology 86 (2006) 697700. [22] W.K. H Panofsky, M. Phillips, Classical Electricity and Magnetism, Addison-
[11] T. Kohya, M. Kuriyama, M. Toda, E. Harada, H. Konno, Heat transfer from Wesley, 1962. p. 1119.
horizontal, parallel tube banks under a corona discharge, International [23] J.R. Melcher, Continuum Electromechanics, MIT press, Cambridge, 1981
Chemical Engineering 27 (2) (1989) 319325. (Chapters 3 and 10).
[12] T. Kuriyama, E. Harada, H. Konno, Application of corona wind to practical [24] James Q. Feng, Application of Galerkin nite-element method with Newton
tubular heat exchangers, Heat Transfer Japanese Research 18 (4) (1989) 18. iterations in computing steady-state solutions of unipolar charge currents in
[13] T. Kuriyama, E. Harada, H. Konno, Enhancement of heat transfer from corona devices, Journal of Computational Physics 151 (1999) 969989.
horizontal tube-banks by the use of a corona wind, International Chemical [25] T. Saitoh, T. Sajiki, K. Maruhara, Bench mark solution to natural convection
Engineering 29 (4) (1989) 715721. heat transfer problem around the horizontal circular cylinder, International
[14] S. Wangnipparnto, J. Tiansuwan, S. Jiracheewanun, T. Kiatsiriroat, C. Wang, Air Journal of Heat and Mass Transfer 36 (1993) 12511259.
side performance of thermosyphon heat exchanger in low Reynolds number [26] P. Wang, R. Kahawita, T.H. Nguyen, Numerical computation of the natural
region: with and without electric eld, Energy Conversion and Management convection ow about a horizontal cylinder using splines, Numerical Heat
43 (2002) 17911800. Transfer 17 (1990) 191215.
[15] S. Vithayasai, T. Kiatsiriroat, A. Nuntaphan, Effect of electric eld on heat [27] T.H. Kuehn, R.J. Goldstein, Numerical solution to the NavierStokes equation
transfer performance of automobile radiator at low frontal air velocity, Applied for laminar natural convection about a horizontal isothermal circular cylinder,
Thermal Engineering 26 (2006) 20732078. International Journal of Heat and Mass Transfer 23 (1980) 971979.
[16] M. Ashjaee, A.H. Eshtiaghi, M. Yaghoubi, T. Youse, Experimental investigation [28] S.Y. Huang, F. Mayinger, Heat transfer with natural convection elliptic tubes,
on free convection from a horizontal cylinder beneath an adiabatic ceiling, Warme-Und Stoffubertagung 18 (1984) 175183.
Experimental Thermal and Fluid Science 32 (2007) 614623. [29] J.-S. Chang, F. Pontiga, P. Atten, A. Castellanos, Hysteresis effect of corona
[17] E.R.G. Eckert, R.J. Goldstein, Measurements in Heat Transfer, McGraw-Hill, discharge in a narrow coaxial wire-pipe discharge tube with gas ow, IEEE
New York, 1972. Transactions on Industry Applications (1996) 12501256.