International Journal of Heat and Mass Transfer 134 (2019) 554–565

Contents lists available at ScienceDirect

International Journal of Heat and Mass Transfer

journal homepage: www.elsevier.com/locate/ijhmt

Laminar natural convection heat transfer depending on diameters of

vertical cylinders with circular cross-section with high Prandtl number
Gyeong-Uk Kang ⇑, Dae-Sik Yook
Korea Institute of Nuclear Safety (KINS), 62 Gwahak-ro, Yuseong-gu, Daejeon 34142, Republic of Korea

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

Article history: Laminar natural convection heat transfer behavior in vertical cylinders with high Pr was experimentally
Received 10 October 2018 and numerically performed by varying the diameters at diverse heights. Mass transfer experiments were
Received in revised form 31 December 2018 conducted instead of heat transfer ones using an electroplating system based on the analogy concept. 3-D
Accepted 16 January 2019
computational models using finite volume method have been developed and validated against experi-
Available online 23 January 2019
mental data. Both experimental and numerical results were in satisfactory agreement with the existing
laminar natural convection correlation developed on a vertical plate. The dependency of the diameter at
Keywords:
various RaH on NuH was clearly observed; at the smallest diameter, heat transfer rates deteriorated due to
Boundary layer
Duct flow
stagnant flows, and then were enhanced due to the increased mass flow rates by flow-like chimney
Flow-like chimney effect effects with increasing diameter, eventually coinciding with the existing laminar correlation. The influ-
Laminar flow ences of diameters on heat transfer rates and flows were analyzed from numerical simulations.
Natural convection Ó 2019 Elsevier Ltd. All rights reserved.

1. Introduction not only the diameters but also the height. In spite of the fact that
the internal flows in vertical cylinders depends on the boundary
Numerous studies have focused on the internal natural convec- layers resulting from the heights and the diameters, the investiga-
tion flows driven by buoyancy forces in a vertical cylinder with a tions considering them are rare. Most reported studies [2–10,15–
circular cross-section because of the many associated engineering 17] for natural convection in the vertical geometries have focused
applications, such as the passive cooling system (PCS) design of on duct flows, chimney effects and transition to turbulent flows. In
nuclear power plants, the air flow-path design of spent fuel dry other words, the natural convection characteristics modeled based
storage casks, solar heating systems and the cooling of electronic on a phenomenological understanding of flow behavior from small
equipment. Most of the existing studies have performed the natu- to larger diameters at diverse heights of the cylinder have not been
ral convection in vertical cylinders using diameters sufficient to not fully examined. Thus, additional studies are still required to
overcome the flow interactions between internal boundary layers. delineate the heat transfer and flow behavior.
The height (H) of the cylinder influences the thickness of the The present study experimentally and numerically investigates
boundary layer formed along the axial wall. Additionally, the flow the heat transfer rates and flow behavior under laminar flows in
interaction between internal boundary layers varies according to vertical cylinders by varying the diameters at diverse heights. An
the diameters (d). It is well known that when the diameter is much analogous mass transfer experiment method using a sulfuric
greater than the thermal boundary layer (dT) that develops along acid–copper sulfate (H2SO4–CuSO4) electroplating system using
the axis wall, the natural convection flow on vertical cylinders is the limiting current technique was adapted instead of heat transfer
the same with that of the vertical plate. In this case, the natural experiments so that high RaH using small-scale test facilities could
convection correlations developed on the vertical plate can be be achieved easily. NuH were measured for cylinders with diame-
applied to a vertical cylinder. On the contrary, when the diameter ters ranging from 0.011 m to 0.039 m and heights ranging from
is much smaller than the thermal boundary layer, the flow behav- 0.01 m to 0.11 m, corresponding to GrH of 8.05
104 to
ior becomes different due to the flow interactions and the overlap 1.08
108 at a fixed Pr of 2094. Numerical analysis using FLUENT
of boundary layers [1]. In particular, since these boundary layers at 14.5 [18] was carried out under same conditions as the experi-
opposite walls develops along the axial wall, they can be affected ments and under the extended diameters and heights.
Phenomenological explanations for the influence of the bounder
layer thickness and the flow behavior on heat transfer rates were
⇑ Corresponding author. suggested.
E-mail address: gukang@kins.re.kr (G.-U. Kang).

https://doi.org/10.1016/j.ijheatmasstransfer.2019.01.073
0017-9310/Ó 2019 Elsevier Ltd. All rights reserved.
 G.-U. Kang, D.-S. Yook / International Journal of Heat and Mass Transfer 134 (2019) 554–565 555

Nomenclature

C molar concentration [mole/m3] P pressure [kPa]

Cp specific heat at constant pressure [J/kg∙K] Pr Prandtl number [m/a]
d diameter [m] Rad Rayleigh number [GrdPr]
Dm mass diffusivity [m2/s] RaH Rayleigh number [GrHPr]
g gravitational acceleration, 9.8 [m/s2] ReD Reynolds number [VmD/m]
Grd Grashof number [gbDTd3/m2] Sc Schmidt number [m/Dm]
GrH Grashof number [gbDTH3/m2] ShH Sherwood number [hmH/Dm]
H height [m] Vm mean velocity [m]
hh convective heat transfer coefficient [W/m2 K]
hm convective mass transfer coefficient [m/s] Greek symbols
Ilim limiting current density [A/m2] a thermal diffusivity [m2/s]
k thermal conductivity [W/m∙K] b volume expansion coefficient [m3/K]
Lh hydrodynamic entrance length [m] l Viscosity [kg/ms]
Lt thermal entrance length [m] m Kinematic viscosity [m2/s]
M molarity [mole/l] q Density [kg/m3]
m_ mass flow rate [kg/s] d Thickness of momentum boundary layer [m]
N number of electrons in charge transfer reaction dT Thickness of thermal boundary layer [m]
NuH Nusselt number [hH/k]
NuD Nusselt number [hD/k]

2. Theoretical background Auletta et al. [12] reported that the chimney acts as a shroud to
allow the plume ascending from the bottom to maintain its config-
2.1. Natural convection heat transfer behavior in vertical cylinder urations without dispersion and accelerate flow through buoyancy
forces, and heat transfer is enhanced as the flow rate increases.
Natural convection phenomena due to the flow interactions in Asako et al. [13] investigated chimney flows in a vertical heated
vertical cylinders can be divided into the developing flow and fully tube attached to a thermally insulated chimney for various diame-
developed flows resulting from the interactions of the boundary ters. Campo et al. [14] performed a numerical analysis of an iso-
layer thickness and the diameter variation at the fixed height of flux heated channel with an adiabatic extension and concluded
the cylinder. The region from the cylinder inlet to the point at that the accelerated fluid draws the lower fluid and that the veloc-
which the boundary layer merges at the centerline is called the ity field at the heated wall is similar to flows of forced convection
entrance region. The flow in the entrance region is referred to as flow behavior because the heated fluid rises more rapidly than flow
developing flow because the thermal and velocity profiles develop driven by buoyancy forces. Lim and Chung [15] performed natural
without flow interactions in this region. In this case, the heat trans- convection experiments for the inner wall of a vertical circular
fer correlations on the vertical plate can be applied to a vertical tube that contained an extension of the unheated section, i.e., the
cylinder [2–7,12,17–19]. The thickness of the boundary layer chimney. The researchers found that the heat transfer rate
increases in the flow direction, merges, and finally fills the entire increased with increases in the length of the unheated extension
cylinder. The region beyond the entrance region, where the ther- via the chimney effect. Ohk and Chung [16] analyzed and visual-
mal and velocity profile are fully developed and no longer change ized the duct flow and chimney effects using the velocity and tem-
regardless of the height, can occur for relatively smaller diameters perature fields in vertical cylinders with diameters and unheated
or long cylinders. ducts for GrH of 3.4
108 to 4.2
1010. Chae and Chung [17] inves-
Bejan [1] reported that when the boundary layer thickness that tigated the heat transfer enhancement of a chimney system for RaH
develops along the heated wall inside a vertical pipe is consider- of 5.8
1010 and 1.4
1012.
ably smaller than the diameter, or the diameter is even larger than
the boundary layer thickness, the vertical cylinder can be treated
2.2. Analogous concept for heat and mass transfer
as a vertical plate because the boundary layers from opposite walls
do not interact. Eckert and Diaguila [2] studied natural convection
Heat transfer systems and mass transfer systems are analogous,
heat transfer in an open-ended pipe and found that the natural
as they are mathematically similar under the same classes of
convection heat transfer inside the pipe was similar to that
boundary and initial conditions [1]. Table 1 shows the correspond-
observed on a vertical plate. Kang et al. [3–7] have proved this find-
ing governing parameters. Thus, heat transfer experiments can be
ing again through mass transfer experiments considering larger
replaced by mass transfer experiments, and vice versa.
diameters. Humphreys and Welty [8] measured the instantaneous
Mass transfer measurements provide a good approximation of
velocity on natural convection flow. Bejan and Large [9] demon-
convective heat transfer. In this study, a sulfuric acid–copper sul-
strated that for a wide range of Pr, the transition to turbulent flow
fate (H2SO4–CuSO4) electroplating system was employed as a mass
occurs at a GrH of 109 rather than RaH. Kang and Chung [10] per-
formed natural convection mass transfer experiments with GrH of
Table 1
5.2
106 to 3
1010 to determine the proper transition criteria Dimensionless numbers for analogy systems.
and concluded that the proper governing parameter for describing
transition should be GrH rather than RaH for a wide range of Pr. Roul Heat transfer system Mass transfer system

and Nayak [11] studied the effect of the height to diameter ranging Nusselt number (Nu) hh H
k
Sherwood number (Sh) hm H
Dm
from 10 to 18.9 for the heat transfer rates with the air as the work- Prandtl number (Pr) m
a Schmidt number (Sc) m
Dm
ing fluid. The heat transfer rates decreases in the outlet when the Rayleigh number (Ra) gbDTH3
am
Rayleigh number (Ra) gH3 Dq
Dm m q
height to diameter increases.
556 G.-U. Kang, D.-S. Yook / International Journal of Heat and Mass Transfer 134 (2019) 554–565

transfer system that allowed high Ra to be accurately attained by solved by employing the limiting current technique. As the applied
measuring the electric current. The transfer of cupric ions from a potential increases, the current between electrodes increases until
copper anode to a cathode submerged in an aqueous solution of it reaches a plateau region or the limiting current region. This limit
sulfuric acid–copper sulfate simulates the transfer of heat in a heat is reached when the number of copper ions reaching the cathode
transfer system. Such an electroplating system was previously stops increasing, at which point further increases in the applied
used by Le Fevre [19]. Agar [20] found that heat transfer correla- potential do not affect the current. At this time, the concentration
tions are similar to mass transfer correlations under the same con- of copper ions at the cathode surfaces can be regarded as 0 because
ditions. This technique has been well established as an the reduction process of copper ions is faster than the delivery of
experimental methodology. In this study, empirical relations (1)– copper ions. Thus, the mass transfer coefficient, hm, can be calcu-
(8) suggested by Fenech and Tobias were used to determine the lated from the bulk concentration, Cb, and the limiting current den-
values of the physical properties that are accurate within an error sity, Ilim, using Eq. (9). A more detailed description of the analogous
bound of 0.5% at 22 °C [21]. methodology is provided in the paper by Ko et al. [22].

q ðkg=m3 Þ ¼ ð0:9978 þ 0:06406MH2 SO4  0:00167M2H2 SO4 ð1  tn ÞIlim

hm ¼ ð9Þ
nFC b
þ 0:12755MCuSO4 þ 0:018202CuSO4
 0:00235MGlycerol þ 0:00353M2Glycerol Þ
103 : ð1Þ
3. Experimental setup
l ðcpÞ ¼ 0:974 þ 0:1235MH2 SO4 þ 0:0556M2H2 SO4
Fig. 1(a) and (b) show the test apparatus and electric circuit,
þ 0:5344M CuSO4 þ 0:53562CuSO4 þ 0:1475M Glycerol respectively. The apparatus consists of cylindrical acryl pipes to
prevent corrosion by H2SO4 and CuSO4 and is separated into three
þ 0:2029M 2Glycerol : ð2Þ
sections: the upper, middle and lower sections. Only the middle
section is lined with copper pipes in acryl pipes to form the cath-
lDCuSO4 ðm2 =sÞ ¼ ð0:7633 þ 0:00511MH2 SO4 þ 0:02044MCuSO4 ode simulating the heated wall. The height of this cathode ranges
þ 0:0653MGlycerol Þ
10: ð3Þ from 0.01 to 0.11 m, which is adjusted to achieve the required
RaH, where Sc, which depends on the concentration of CuSO4, is
t CuSO4 ¼ ð0:2633  0:1020C H2 SO4 ÞC CuSO4 : ð4Þ taken as a constant. The diameters ranged from 0.011 to 0.039 m.
In an electroplating system, the cathode simulates the heated wall


Dq DC H2 SO4 in a heat transfer system via the buoyancy induced from the
¼ C CuSO4 bCuSO4 bH2 SO4 : ð5Þ decreased fluid density by the reduction of copper ions at the cath-
q DC CuSO4
ode surface.
DC H2 SO4 A small groove is cut into the inside or outside edges of the
¼ 0:000215 þ 0:113075c1=3 þ 0:85576c2=3  0:50496c cathode such that the height can be varied and the cathodes can
DC CuSO4
adhere to each other without gaps. Acryl flanges were attached
ð6Þ
to the top and bottom of each section, which acted as a link
between the test sections and the remainder of the vertical cavity.
c ¼ C CuSO4 =ðC CuSO4 þ C H2 SO4 Þ: ð7Þ
A rubber O-ring is attached between the flanges to prevent the
  leakage of fluid. Therefore, Pr can be kept spatially uniform. A cop-
1 @q
bj ¼ : ð8Þ per rod with a diameter of 0.002 m as the anode is inserted into the
q @C j T;Ck –j top of the test section using a holder to place the rod in the center
Mass transfer experiments using the electroplating technique without contacting the inner copper cylinder surface, as shown in
are an attractive method to achieve high RaH with a relatively small Fig. 1(b). When the anode is located outside of the cylinder, the
test rig and require simple, quick, and accurate measurements by cupric ions produced from the anode could not reach the entire
electric means while avoiding the problems of heat leakage and cathode surfaces easily. Thus, the anode should be located inside
radiation heat transfer. Additionally, the problems of heat leakage the cathode to improve the current distribution. In other words,
and radiation heat transfer do not need to be considered in the considering the anode diameter, the cathode diameter should be
mass transfer system because these processes do not occur. How- greater than 0.002 m; otherwise, it is difficult to perform the
ever, one of the shortcomings is that Pr is overly large. The test experiments, resulting in limitations in the diameter size of the
results obtained from carefully designed experiments with proper cylinder. The interior in the test section is filled with a sulfuric
governing parameters may be generalized, and the limitation of acid-copper sulfate (H2SO4-CuSO4). In mass transfer experiments,
extremely high Pr can be addressed to some extent. However, the density does not vary based on the temperature. Table 2 shows
the result obtained using this method is not applicable to fluids the experimental test matrix. The Sc corresponding to Pr was 2094.
having Pr of approximately 1. The concentrations of CuSO4 and H2SO4 were 0.05 M and 1.5 M,
respectively. The diameters of the vertical cylinders ranged from
0.011 m to 0.039 m, and the heights were varied from 0.01 m to
2.3. Limiting current technique
0.11 m, corresponding to GrH values ranging from 8.05
104 to
1.07
108. The flow is in the laminar regime [9,10]. The height-
In an electroplating system, a cathode simulates the heated wall
to-diameter ratio of the vertical cylinders ranged from 0.26 to 10.
in the heat transfer system because a buoyant force is induced by
the reduction of copper ions on the cathode surfaces, leading to
decreases in the density near the cathode. An anode, the supplier 4. Numerical analysis
of cupric ions, acts as a cold wall. Heat flux is calculated from the
heat transfer coefficients and temperature differences between Numerical investigations were performed using the ANSYS FLU-
the heated wall and bulk. The mass flux can be obtained in a sim- ENT 14.5 [18]. The ANSYS DesignModeler and Workbench Meshing
ilar manner. However, it is difficult to determine the concentration tools were used to generate the 3-D thermal analysis model of the
of copper ions at the cathode surfaces. These problems can be vertical cylinder. Only the fluid areas was modeled and the upper
 G.-U. Kang, D.-S. Yook / International Journal of Heat and Mass Transfer 134 (2019) 554–565 557

(a) Test apparatus (b) Electric circuit

Fig. 1. Test apparatus and electric circuit on mass transfer experiments.

Table 2 with GrH of 8.1
104 to 5.9
107, according to Refs. [9,10]. In this
Test matrix on natural convection experiments. study, the height is used as the characteristic length because the
d (m) H (m) H/d GrH RaH Pr buoyancy forces are proportional to the third power of the height
0.011, 0.013, 0.015, 0.01 0.26–0.91 8.06
104 1.69
108 2094
of the vertical cylinder.
0.017, 0.020, 0.02 0.51–1.82 6.44
105 1.35
109 The mesh was constructed by hexahedral type. Since the
0.030, 0.032, 0.03 0.77–2.73 2.17
106 4.55
109 boundary layer is confined to an extremely thin region near the
0.039 0.05 1.28–4.55 1.01
107 2.11
1010 heated wall, the concentrated grids are employed for the regions
0.07 1.79–6.36 2.76
107 5.79
1010
close to the boundary layer of the heated wall, whereas coarse
0.09 2.31–8.18 5.87
107 1.23
1011
0.11 2.82–10 1.07
108 2.25
1011 grids are employed to the central region of the cylinder, as shown
in Fig. 2.
The edge assigned along the heated wall is comprised of 10 lay-
ers; a growth rate of 1.2 is used, and the maximum thickness is
and lower adiabatic walls which did not affect the flow were adjusted depending on the diameters. A sensitivity analysis was
excluded in the numerical configuration. Fig. 2 shows the solution performed to determine the number of cells required for each
grid used for the simulation. The height and diameter of the verti- numerical calculation by varying the size of cell near the wall.
cal cylinder varied according to the simulation ranges shown in Fig. 3 shows representatively mesh sensitivity analysis on
Table 3. H = 0.09 m and d = 0.02 m, and the optimal number of cells was
The diameters ranged from 0.001 to 0.03 m. The height-to- approximately 230,000.
diameter ratio of the vertical cylinders ranged from 0.33 to 90. Table 4 presents the thermal properties for a Pr of 2094, as was
Unlike the experiments, the diameter can be adjusted to be suffi- used in the simulations. These properties were derived from
ciently small in the CFD analysis. The heights ranged from 0.01 m empirical relations (1)–(8) for a cupric acid-copper sulfate electro-
to 0.09 m, corresponding to the laminar natural convection flow plating system. Especially, the specific heat (Cp), thermal conduc-
tivity coefficient (k) and the thermal expansion coefficient (b) can
be calculated by the following proportional relations (10) and
(11), respectively.
CP CP CP
Dm ¼ ! ¼ Dm q ! k¼ ð10Þ
qk k Dm q

Raam RaDm m Dq Dq Dq 1
¼ ¼ ¼ bDT ! ¼ bDT ! b¼
gH3 gH3 q q q DT
ð11Þ

Table 3
Simulation test matrix.

d (m) H (m) H/d GrH Pr

0.001, 0.002, 0.004, 0.007, 0.01 0.33–10 8.1
104 2094
0.009, 0.011, 0.013, 0.03 1–30 2.2
106
0.015, 0.020, 0.030 0.05 1.67–50 1.1
107
0.07 2.33–70 2.8
107
0.09 3–90 5.9
107
Fig. 2. The solution grid for the 3-D simulation model.
558 G.-U. Kang, D.-S. Yook / International Journal of Heat and Mass Transfer 134 (2019) 554–565

The natural convection flow under investigation was modeled

by a set of governing equations describing the conservation of
mass, momentum and energy. These equations have been solved
using the CFD code with adequate boundary conditions. Under
steady-state conditions, the governing equations for three-
dimensional laminar flows can be expressed as follows:

 Conservation of mass

@u @ v @w
þ þ ¼ 0: ð16Þ
@x @y @z
 Conservation of momentum !
@u @u @u @2u @2u @2u 1 @P
u þv þw ¼v þ þ  ; ð17Þ
@x @y @z @x2 @y2 @z2 q @x
!
@v @v @v @2v @2v @2v
Fig. 3. Mesh sensitivity analysis. u þv þw ¼v þ þ þ gbDT; ð18Þ
@x @y @z @x2 @y2 @z2

!
Table 4 @w @w @w @2w @2w @2w 1 @P
Properties for Pr of 2094 used in the simulation. u þv þw ¼v þ 2 þ 2  : ð19Þ
@x @y @z @x2 @y @z q @z
Property Value
 Conservation of energy !
Density (q) 1103.7 kg/m3
@T @T @T @2T @2T @2T
þv
Specific heat (Cp) 1000 J/kg∙K
u þw ¼a þ þ : ð20Þ
Thermal conductivity (k) 6.127E4 W/m∙k @x @y @z @x2 @y2 @z2
Viscosity (l) 1.283E3 kg/m∙s
Thermal expansion coefficient (b) 9.415E7 K1
The boundary and initial condition were

u ¼ v ¼ w ¼ 0 at all walls,
In the relation (11), the Dq/q based on the relation (5) is calcu- T ¼ T 0 on heated wall, and
lated by the following relations (12)–(15). T ¼ T 1 as fluid.
0:12755 þ ð2
0:0182
C CuSO4 Þ
bCuSO4 ¼ 5. Results and discussion
q
0:12755 þ ð2
0:0182
0:1Þ
¼ ¼ 0:1189 ð12Þ 5.1. Validation of the experimental results
1:1030695

0:06406  ð2
0:00167
C H2 SO4 Þ Fig. 4 shows the measured NuH according to test matrix in
bH2 SO4 ¼ Table 2, together with the existing laminar natural convection heat
q
transfer correlation (21) developed for a vertical plate [10,23].
0:06406  ð2
0:00167
1:5Þ
¼ ¼ 0:0535 ð13Þ
1:1030695 NuH ¼ 0:67ðGrH PrÞ0:25 at Gr H < 109 ð21Þ

DC H2 SO4 The open-symbols representing results for diameters greater

¼ 0:000215 þ 0:113075 than 0.017 m exhibit good agreement with laminar correlation
DC CuSO4
(21) for the given RaH ranges, i.e., the boundary layers developed

ðC CuSO4 =ðC CuSO4 þ C H2 SO4 ÞÞ1=3
þ 0:85576ðC CuSO4 =ðC CuSO4 þ C H2 SO4 ÞÞ2=3
 0:50496ðC CuSO4 =ðC CuSO4 þ C H2 SO4 ÞÞ
¼ 0:1479 ð14Þ

Dq
¼ 0:1
ð0:1189
0:0535
0:1479Þ ¼ 9:415
105 ð15Þ
q
The bottom of the cylinder was taken as the pressure inlet, and
the top was taken as the pressure outlet. The uniform wall temper-
ature of the heated wall was kept at 400 K, and the fluid tempera-
ture was maintained at 300 K, as shown in Fig. 2. Simulations were
conducted using the Boussinesq approximation. The pressure-
based solver was used with second-order upwind for momentum
and energy in the laminar model because due to GrH less than
108 in this study. The body forced weighted algorithm was adopted
for the pressure discretization, and the SIMPLE scheme is used for
the pressure-velocity coupling discretization. The solutions were
converged when the residual values of the momentum and energy Fig. 4. Comparison of the experimental results with the existing laminar natural
were 106. Fifty cases were simulated. convection correlation developed for a vertical plate.
 G.-U. Kang, D.-S. Yook / International Journal of Heat and Mass Transfer 134 (2019) 554–565 559

along the heated wall were thin and did not interact with each The calculated average uncertainty for NuH was approximately
other because of the thin thermal boundary layer with a higher 4.23%, and the average fractional uncertainty of these values was
Pr fluid. The results of the closed-symbols for diameters less than approximately 2%, i.e., the measurements from the natural convec-
0.015 m were in satisfactory agreement with the laminar correla- tion mass transfer experiments are high.
tion (21) at lower RaH.
However, their points were located above the laminar correla-
5.2. Numerical results
tion (21) from the RaH of approximately 3
1010. This overestima-
tion is due to the flow interactions, e.g., the chimney effect,
5.2.1. Comparisons of the experimental and numerical data
increasing the mass flow rates entering into the inlet, ultimately
Fig. 5(a) and (b) shows the comparisons of the numerical results
leading to increased heat transfer rates, as noted in previous stud-
with experimental ones of Fig. 4 to validate the simulation. Both
ies. The thickness of boundary layer was calculated by scale rela-
results from 0.011 m from 0.017 m of diameters were in good
tions to give the phenomenological basis for flow interactions
agreements at the same conditions and followed the laminar corre-
and the reasonable explanations on the flow behavior were
lation (21) as shown Fig. 5(a). However, it was clearly observed
described in details in Sections 5.2.2 and 5.2.3.
that NuH showed the dependency on diameters at heights of cylin-
The uncertainties involved in the mass transfer measurement
ders corresponding to RaH as shown in Fig. 5(b).
using an electroplating system were analyzed by the traditional
We redraw the ratio of the numerically calculated Nunumerical
data reduction Eqs. (22) and (23) [24]. Because NuH is the final
value to the Nuvertical plate to show clearly the absolute magnitude
dependent variable, the data reduction equation and uncertainty
and to distinguish the influences on diameters at diverse heights
equation are expressed as
in Fig. 6 of Section 5.2.2.
hm H
NuH ¼ ) NuH ¼ f ðhm ; H; Dm Þ ð22Þ
Dm 5.2.2. Comparison of the heat transfer rates for diverse diameters

2
2
2 Fig. 6 presents the ratio of the numerically measured
@Nu @Nu @Nu Nunumerical value to the Nuvertical plate obtained from correlation (21)
U 2NuH ¼ Uh þ UH þ : ð23Þ
@hm m @H @Dm

(a) Comparison at same conditions(d=0.011~0.017m)

Fig. 6. NuH ratio for various diameters (d) and heights (H).

(b) Comparison at all data

Fig. 5. Comparison of experimental and numerical results. Fig. 7. Comparison of the mass flux for different cylinder diameters.
560 G.-U. Kang, D.-S. Yook / International Journal of Heat and Mass Transfer 134 (2019) 554–565

as the reference for the simulation ranges given in Table 4. The with laminar correlation (21). It reveals that the heat transfer rates
heat transfer rates tended to deteriorate at smaller diameters and were affected by the diameter at diverse heights, exhibiting the
increase with further increasing diameter, eventually coinciding three types of the heat transfer behaviors.

(a) Temperature contours

(b) Velocity contours

Fig. 8. Temperature and velocity fields for various d at H = 0.07 m.
 G.-U. Kang, D.-S. Yook / International Journal of Heat and Mass Transfer 134 (2019) 554–565 561

In the first heat transfer behavior, for cylinders having diame- Table 5
ters less than about 0.007 m, the heat transfer deteriorated. More- Thickness of the thermal boundary layer for a diameter of 0.001 m.

over, the slope of the curve declined rapidly with increasing the d/2 (m) H (m) Thickness of thermal boundary
heights at the same diameters. It was observed that NuH were up layer (dT, m)
to 31% lower compared to laminar correlation (21) when the diam- FLUENT Eq. (24) [1]
eter was 0.001 m at a height of 0.09 m. This phenomenon can be 0.0005 0.01 0.00032 0.000088
explained as follows. The flow area became smaller and the tem- 0.03 0.00042 0.000116
perature differences between the heated wall and fluid entering 0.05 d< 0.000131
the inlet were decreased, resulting in lower density differences. 0.07 d< 0.000143
0.09 d< 0.000152
Lower density differences in turn lead to increasing shear stress
on the wall, resulting in lower velocities. Because the viscous forces
opposing the flow were high, the flow velocities became nearly
negligible. In other words, the flow does not develop smoothly at 0.004 m, and the peak shifted upward for diameters ranging from
the outlet; instead, it stagnates at the bottom of the cylinder. 0.007 to 0.015 m. At a diameter of 0.02 m, the velocity fields exhib-
With the second heat transfer behavior, the heat transfer rates ited the typical flow pattern of natural convection as the flow pro-
were enhanced for diameters ranging from 0.009 m to 0.015 m ceeded to the outlet.
with increases in the height due to the flow-like chimney effect
by flow interactions, despite the absence of the unheated chimney 5.2.4. Temperature and velocity profiles
channel. It implies that the chimney effects may occur without the Fig. 9 shows the temperature and velocity profiles at a cross-
extension and expansion tube since the influence of the chimney section from the heated wall to the centerline for various diame-
effects depends on the thickness of the boundary layer. ters when the height was 0.07 m. As shown in Fig. 9(a), the thermal
With the third heat transfer behavior, the heat transfer rates for boundary layers were even thinner than the diameter because of
diameters larger than 0.02 m were similar to laminar correlation the very high Pr. The temperature at the heated wall was the high-
(21), i.e., the thin thermal boundary layer did not interact, corre- est and gradually became equal to the ambient temperature as it
sponding to duct flow. moved away from the wall, i.e., its flow behavior was the same
In order to reveal the mechanism of variation in the heat transfer as that of a vertical plate. However, the thermal boundary layer
rates, the mass flux (mass flow rate per unit area) was calculated at a diameter of 0.001 m was thicker than the diameter, and the
through the numerical simulation and was compared for diverse layers were overlapped and interacted with each other as the
diameters in Fig. 7. The mass flow rate entering the inlet of the height increased.
cylinder per unit area varied depending to the diameter for a given In case of the velocity profile in Fig. 9(b), the flow in the layer in
height, i.e., a given buoyancy force. As the diameter decreased, the contact with the surface of the heated wall came to a complete
mass flux remained nearly the same regardless of the height, imply- stop because of the no-slip condition. This layer caused the flow
ing that the viscous forces were dominant rather than buoyancy in the adjacent layers to slow gradually as a result of friction. To
effects. However, with further increasing diameters, the mass flux compensate for this velocity reduction, the velocity at the midsec-
increased due to chimney effects with showing the parabolic tion of the cylinder increased to keep the mass flow rate through
shapes until the diameter was about 0.015 m. This trend was more the cylinder constant. As a result, a velocity gradient developed
remarkable for heights ranging from 0.05 to 0.09 m corresponding along the flow directions because the thickness of the boundary
to the RaH from 2.11
1010 to 1.23
1011. From above the diame- layer is zero near the wall. With increasing diameter, the average
ters of 0.02 m, the higher the RaH, the larger the mass flux. However, magnitude of the velocity for each height gradually increased,
it remained constant regardless of the diameter size because the and the peak velocity appeared at a diameter of 0.013 m. The
effect of the buoyancy forces was dominant. velocity decreases with further increases in the diameter. For a lar-
ger diameter of 0.02 m, the velocity profile exhibited the typical
natural convective flow on the vertical plate, with the peak velocity
5.2.3. Boundary layer thickness and flow fields
near the heated wall. The thickness of the velocity boundary layer
Fig. 8 shows the temperature and velocity distributions depend-
(d) calculated by scale relation (24) ranges from approximately
ing on diameters for at a height of 0.07 m. As shown in Fig. 8(a), the
0.004 m to 0.069 m and is expected to be affected by the velocity
thin thermal boundary layers developed along the heated wall and
boundary layer when the diameter is less than 0.004 m.
they did not interfere with each other; the behaviors of the layers
The velocity entering into the inlet for diameters from 0.001 to
were similar to those on the vertical plate in an open channel.
0.004 m exhibited a parabolic shape, exhibiting a peak near the
However, the interference of the thermal boundary layer became
inlet and gradually decreasing while escaping to the outlet due
clear at a diameter of 0.001 m. In this case, the overlap of the
to higher viscous effects. Moreover, the velocity profile near the
boundary layer was observed. The approximate thickness of the
inlet for diameters from 0.007 to 0.015 m did not exhibit a para-
thermal boundary layer (dT) and velocity boundary layer (d) can
bolic form; instead, the velocity profile was of developing flow
be calculated by scale relation (24) as a combination of RaH, Pr,
behavior form, with the flat center corresponding to the entrance
and the height, as suggested by Bejan [1], if the value of Pr is con-
region. The entrance region acts as the enhancement of the average
siderably larger than 1.
heat transfer rates for the entire cylinder. As the flow developed,
the velocity profile at the cross-section remained unchanged,
d ¼ HRa1=4
H Pr1=2 ; dT ¼ dPr1=2 : ð24Þ
regardless of the height, implying a hydrodynamic fully developed
Table 5 presented the thickness of the thermal boundary layer region. Thus, the thickness of this boundary layer increased in the
approximately calculated by simulation results and that obtained flow direction until the boundary layer reached the cylinder center
by scale relation (24) for a diameter of 0.001 m. It is seen that as and thus filled the entire cylinder.
the heights are increased, the boundary layer thickness grows. At
the heights exceeding 0.05 m, the thickness of the thermal bound- 5.2.5. Entrance region and fully developed region
ary layer became larger than the diameter. In laminar flow, the hydrodynamic and thermal entrance
Fig. 8(b) shows the velocity distributions. The peak velocity lengths can be approximated by scale relation (25) as a combina-
appeared at the center near the inlet at diameters less than tion of the Red, Pr and diameters as follows [25,26]:
562 G.-U. Kang, D.-S. Yook / International Journal of Heat and Mass Transfer 134 (2019) 554–565

Lh ¼ 0:05Red d; Lt ¼ 0:05Red Prd ¼ PrLh ð25Þ entrance regions for a height of 0.07 m are shown in Table 6. It
was found that the hydrodynamically fully developed region was
The Red was calculated using the values of the measured mass
already formed at diameters except for the diameters of 0.015
flow rates and the mean velocities at inlets from the numerical
and 0.02 m. Moreover, the thermally fully developed region was
results. The calculated lengths for the thermal and hydrodynamic

(a) Temperature profiles

Fig. 9. Temperature and velocity profiles from the center to the heated wall at H = 0.07 m.
 G.-U. Kang, D.-S. Yook / International Journal of Heat and Mass Transfer 134 (2019) 554–565 563

(b) Velocity profiles

Fig. 9 (continued)

almost formed at a diameter of 0.001 m. It corresponded to the velocity boundary layer outgrows the thermal boundary layer.
diameter of 0.001 in Fig. 8(a). Under laminar flow, the magnitude Thus, the hydrodynamic entrance region was considerably smaller
of the Pr is a measure of the relative growth of the velocity and than the thermal entrance region.
thermal boundary layers. For fluids with Pr  1, the two boundary The centerline velocity and the near-wall velocity in the flow
layers coincide with each other. However, for fluids with Pr > 1, the directions for heat transfer deterioration, the flow-like chimney
564 G.-U. Kang, D.-S. Yook / International Journal of Heat and Mass Transfer 134 (2019) 554–565

Table 6 effect, and duct flow patterns are plotted in Fig. 10. The distance of
Approximate entry length values for various diameters at a height of 0.07 m. 1/10 from the near-wall was determined to be similar to that of the
d (m) H (m) Re Lh (m) Lt (m) heated wall but greater than the peak location near the heated wall
0.001 0.07 0.85 0.0000425 0.089 at the outlet when comparing the velocity variations along the
0.002 2.09 0.0002091 0.438 cylinder. When the diameter was 0.002 m, the viscous forces
0.004 5.98 0.0011959 2.504 opposing the flow were high, resulting in nearly negligible flow
0.007 25.44 0.009 18.64 velocities. In other words, the flow is stagnant around the bottom
0.009 36.22 0.016 34.13
0.013 83.66 0.054 113.87
of the cylinder rather than flowing through the cylinder. The flow-
0.015 101.63 0.076 159.61 like chimney effect condition occurs at a diameter of 0.013 m.
0.020 94.98 0.095 198.89 Thus, the central velocity is high and the near-wall velocity is
low along the cylinder-like forced convection flow. At a diameter
of 0.02 m, the thin thermal boundary layers do not interact, the
duct flow condition occurs and the velocity peak appears at the
center near the entrance. Velocity peaks near the wall are observed
near the outlet.

6. Conclusions

The present study investigates the characteristics of natural

convection heat transfer under laminar flows in vertical cylinders.
Mass transfer experiments based on an analogous concept were
performed on vertical cylinders with diameters ranging from
0.011 m to 0.039 m and heights ranging from 0.01 m to 0.11 m,
corresponding to GrH of 8.05
104 to 1.08
108. Numerical analy-
ses were performed on various diameters, ranging from 0.001 to
0.03 m, including those used in the experiments. The ranges in
numerical simulation cover the very slender cylinders that it was
difficult to perform experiments. Three types of heat transfer
behaviors were observed depending on the diameters: heat trans-
fer deterioration, flow-like chimney effect and duct flow. The main
findings in this study are summarized as follows:

 Heat transfer rates for cylinders having diameters of less

0.007 m deteriorated up to 31% compared to a laminar correla-
tion because the flow does not develop; instead, the flow stag-
nates at the bottom of cylinder.
 For diameters ranging from 0.009 m to 0.015 m, heat transfer
rates were enhanced because of flow-like chimney effects,
resulting from the increase in mass flux. Flow behavior for
diameters larger than 0.02 m was same to those for a vertical
plate.
 The phenomenological understanding of the flow fields in verti-
cal cylinders were suggested from the temperature and velocity
distributions, the comparison of the thickness and length of
thermal and velocity boundary layers with the scale relations.

Conflict of interest

The authors declared that there is no conflict of interest.

Acknowledgements

This work was supported by the Nuclear Safety Research Pro-

gram through the Korea Foundation of Nuclear Safety (KOFONS),
granted financial resource from the Nuclear Safety and Security
Commission (NSSC), South Korea (No. 1803015)

References

[1] A. Bejan, Convection Heat Transfer, third ed., JohnWiley & Sons, INC, New York,
2006.
[2] E.R.G. Eckert, A.J. Diaguila, Experimental investigation of free convection heat
transfer in vertical tube at large Grashof numbers, NACA (1955) 1–14.
[3] G.U. Kang, B.J. Chung, H.J. Kim, Natural convection heat transfer on a vertical
cylinder submerged in fluids having high Prandtl number, Int. J. Heat Mass
Fig. 10. Velocity magnitude at the axial distance on each flow pattern. Transfer 79 (2014) 4–11.
 G.-U. Kang, D.-S. Yook / International Journal of Heat and Mass Transfer 134 (2019) 554–565 565

[4] G.U. Kang, B.J. Chung, Effects of the dimensions of upward- and downward- [14] A. Campo, O. Manca, B. Morrone, Numerical analysis of partially heated
facing vertical cavities on natural convection, Int. Commun. Heat Mass vertical parallel plates in natural convective cooling, Numer. Heat Transf. 36
Transfer 39 (2012) 399–404. (1999) 129–151.
[5] B.J. Chung, J.H. Heo, M.H. Kim, G.U. Kang, The effect of top and bottom lids on [15] C.K. Lim, B.J. Chung, Influence of a center anode in analogy experiments of long
natural convection inside a vertical cylinder, Int. J. Heat Mass Transfer 54 flow ducts, Int. Commun. Heat Mass Transfer 56 (2014) 174–180.
(2011) 135–141. [16] S.M. Ohk, B.J. Chung, Natural convection heat transfer inside an open vertical
[6] G.U. Kang, B.J. Chung, Natural convection heat transfer characteristics in pipe: Influences of length, diameter and Prandtl number, Int. J. Therm. Sci. 115
vertical cavities with active and inactive top and bottom disks, Int. J. Heat Mass (2017) 54–64.
Transfer 87 (2015) 390–398. [17] M.S. Chae, B.J. Chung, Heat transfer effects of chimney height, diameter, and
[7] K.U. Kang, B.J. Chung, The effects of the anode size and position on the limiting Prandtl number, Int. Commun. Heat Mass Transfer 66 (2015) 196–202.
currents of natural convection mass transfer experiments in a vertical pipe, [18] FLUENT Tutorial Guide, 14.5, ANSYS Incorporated, 2012.
Trans. KSME B 34 (1) (2010) 1–8. [19] E.J. Le Fevre, Laminar free convection from a vertical plane surface, in: 9th
[8] W.W. Humphreys, J.R. Welty, Natural convection with mercury in a uniformly International Congress on Applied Mechanics, Brussels, 1956, pp. 1–168.
heated vertical channel during unstable laminar and transitional flow, AIChE J. [20] J.N. Agar, Diffusion and convection at electrodes, Discuss. Faraday Soc. 26 (1)
21 (2) (1975) 268–274. (1947) 27–37.
[9] A. Bejan, J.L. Large, The prandtl number effect on the transition in natural [21] E.J. Fenech, C.W. Tobias, Mass transfer by free convection at horizontal
convection along a vertical, J. Heat Transfer 112 (1990) 789–790. electrodes, Electochim. Acta 2 (1960) 311–325.
[10] G.U. Kang, B.J. Chung, The experimental study on transition criteria of natural [22] S.H. Ko, D.W. Moon, B.J. Chung, Applications of electroplating method for heat
convection inside a vertical pipe, Int. Commun. Heat Mass Transfer 37 (2010) transfer studies using analogy concept, Nucl. Eng. Technol. 38 (2006) 251–258.
1057–1063. [23] E.J. Le Fevre, A.J. Ede, Laminar free convection from the outer surface of a
[11] M.K. Roul, R.C. Nayak, Experimental investigation of natural convection heat vertical cylinder, in: Proceedings of the Nine International Congress Applied
transfer through heated vertical tubes, Int. J. Eng. Res. Appl. 2 (2012) 1088– Mechanics, Brussels, 1956, pp. 175–183.
1096. [24] H.W. Coleman, W.G. Steele, Experimental and Uncertainty Analysis for
[12] A. Auletta, O. Manca, B. Morrone, V. Naso, Heat transfer enhancement by the Engineers, second ed., John Wiley & Son, New York, 1999.
chimney effect in a vertical isoflux channel, Int. J. Heat Mass Transf. 44 (2001) [25] W.M. Kays, M.E. Crawford, Convective Heat and Mass Transfer, third ed.,
4345–4357. McGraw-Hill, New York, 1993.
[13] Y. Asako, H. Nakamura, M. Faghri, Natural convection in a vertical heated tube [26] R.K. Shah, M.S. Bhatti, Laminar Convective Heat Transfer in Ducts, In Handbook
attached to a thermally insulated chimney of a different diameter, Trans. of Single-Phase Convective Heat Transfer, Wiley Interscience, New York, 1987.
ASME 112 (1990) 790–793.