International Journal of Thermal Sciences 111 (2017) 146e159

Contents lists available at ScienceDirect

International Journal of Thermal Sciences

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

Numerical investigation of natural convection heat transfer from

vertical cylinder with annular fins
Jnana Ranjan Senapati*, Sukanta Kumar Dash, Subhransu Roy
Department of Mechanical Engineering, Indian Institute of Technology Kharagpur, 721302, India

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

Article history: Natural convection heat transfer from a vertical cylinder with annular fins has been studied numerically
Received 3 April 2016 by varying the Rayleigh number (Ra) in both laminar (104
Ra
108) and turbulent (1010
Ra
1012)
Received in revised form regimes. The computations were carried out by varying the fin to tube diameter ratio (D/d), fin spacing to
16 August 2016
tube diameter ratio (S/d) in the range of 2e5 and 0.126e5.840 respectively. In the present study, nu-
Accepted 23 August 2016
merical simulations of full Navier-Stokes equation along with the energy equation have been conducted
Available online 1 September 2016
for a vertical cylinder with annular fins of constant thickness using the algebraic multi-grid solver of
FLUENT 15. Optimization study of the conjugate heat transfer characteristics has been carried out to find
Keywords:
Natural convection
the best fin spacing for maximum heat transfer for the turbulent flow. With the addition of fins to the
Heat transfer heated isothermal tube surface, heat transfer goes on increasing for laminar flow and turbulent flow heat
Vertical cylinder transfer first increases and gets a maximum value then starts to decrease. The optimum fin spacing for
Inter-fin spacing maximum heat transfer for the cases of turbulent flow varies between S/d ¼ 0.28 to 0.31 (7.0 mm
Optimization e7.7 mm) for the turbulent regime. The effect of parameters like D/d, S/d and Ra on Nu are analyzed, and
correlations for average Nusselt number has been developed for both laminar and turbulent regimes.
© 2016 Elsevier Masson SAS. All rights reserved.

1. Introduction equally effective. Hence, the basic job of the designers is to improve
the rate of heat transfer by effectively designing the fins. One has to
Among the cooling methods of natural convection and forced consider factors such as the shape of the primary surface, appli-
convection, natural convection systems are convenient and inex- cation of the system and location of the system to design efficient
pensive due to lack of any extra components. The inherent fins. Use of natural convection systems with fins are quite large in
advantage of using a natural convection cooling systems arise out of number which includes heat exchangers, cooling of electronic
not having any additional components which make the system free components, internal and external combustion engines, annular
of extra moving parts, and no further designing of the cooling finned heat sinks, utilization of natural circulation for energy
system may be needed. Lack of spare parts makes the operation of storage systems for space heating (e.g. baseboard heating), air
the system noise-free and free of any extra maintenance. However, cooling systems for air conditioning and refrigeration. For a cylin-
the main shortcoming of such a natural convection based cooling drically shaped primary surface, one of the most popular choices
system is that the rate of heat transfer achieved is quite small as among shape of fins to enhance the rate of heat transfer are annular
explained by Bejan [1] and Kreith et al. [2]. Use of fins is one of the fins because of their inherent ease of manufacturing and also
simplest ways of overcoming this shortcoming of low heat transfer simplicity in the analysis due to radial symmetry.
rates while keeping the system noise-free and maintenance-free. There is plenty of existing literature in the field of natural con-
Use of fins to enhance heat transfer was extensively studied by vection heat transfer which includes Churchill and Chu [5] and
Guvenc and Yuncu [3] and Yazicioglu and Yuncu [4]. Churchill [6], where the authors experimentally developed a cor-
The shape of the fin is an important consideration for heat relation for average Nusselt number as a function of Rayleigh
transfer enhancement since the entire surface of the fin may not be number and Prandtl number for natural convection from the hor-
izontal cylinder and vertical flat plate respectively. Relevance of the
study of natural convection over a vertical flat plate while consid-
ering natural convection over vertical cylinder comes due to the
* Corresponding author. fact that a vertical cylinder (at least for thick cylinder) may be
E-mail address: [email protected] (J.R. Senapati).

http://dx.doi.org/10.1016/j.ijthermalsci.2016.08.019
1290-0729/© 2016 Elsevier Masson SAS. All rights reserved.
 J.R. Senapati et al. / International Journal of Thermal Sciences 111 (2017) 146e159 147

Nomenclature S/d fin spacing to tube diameter ratio or non-dimensional

fin spacing
A area of convection surface, m2 (S/d)opt optimum fin spacing to tube diameter ratio for
Afin area of annular circular fin, m2 maximum heat transfer
d diameter of horizontal cylinder, mm t thickness of fin, mm
D fin diameter, mm T∞ ambient temperature, K
D/d fin-to-tube diameter ratio Tf fluid temperature at the contact surface, K
e Linear eccentricity, mm Tfilm mean film temperature, K
g acceleration due to gravity, m/s2 TS temperature at solid surface, K
h average heat transfer coefficient, W/m2-K Tw cylinder surface temperature, K
k thermal conductivity of fin material, W/m-K u, v, w velocity components in x, y, z directions respectively,
L length of the horizontal cylinder, mm m/s
Nfin number of fins x, y, z Cartesian coordinates
Nu average Nusselt number
p pressure, N/m2 Greek symbols
patm atmospheric pressure, N/m2 a thermal diffusivity, m2/s
Q convection heat transfer, W b thermal expansion coefficient, 1/K
R Specific gas constant, J/kg K DT base-to-ambient temperature difference, K
Ra Rayleigh number ε Eccentricity ratio
Rfin fin radius, mm x fin-to-tube diameter ratio, D/d
S inter-fin spacing, mm n kinematic viscosity, m2/s
Sopt optimum fin spacing for maximum heat transfer, mm r density, kg/m3

treated as a vertical flat plate due to similarity in correlation of use of fins on light sources not only helps in efficient cooling but
Nusselt number as investigated by Gebhart et al. [7] and has been also helps for the beautification of the light sources, and the liter-
explained in Ozisik [8] and Holman [9]. However, for the slender atures are as An et al. [17], Park et al. [18] and Kim et al. [21].
vertical cylinder, the curvature effect cannot be neglected, and the Literature is readily available for natural convection heat transfer
correlation will be different as explained by Popiel et al. [10]. over a horizontal cylinder with concentric annular fins such as
LeFevre and Ede [11] proposed an integral heat transfer solution Littlefield and Cox [22], Kraus et al. [23], Tsubouchi and Masuda
that accounts for the effect of wall curvature in the laminar range. [24], Edwards and Chaddock [25]. There is some literature available
Minkowycz and Sparrow [12] developed a local non-similarity so- on experimental study of natural convection heat transfer from
lution method which was applied to solve natural convection on a inclined cylinders (here, the angle of inclination varies from
vertical cylinder with large curvature. Fujji and Uehara [13] 0 (vertical cylinder) to 90 (horizontal cylinder)) by Al-Arabi and
compared the heat transfer rates by laminar natural convection Salman [26] and Al-Arabi and Khamis [27].
along the outer surface of a vertical cylinder with the same along a Numerical studies of natural convection over finned surfaces are
vertical flat plate. Kuiken [14] investigated the radial curvature ef- relatively rare. For example, natural convection heat transfer from
fects on axisymmetric free convection boundary-layer flow for finned sphere has been investigated numerically by Singh and Dash
vertical cylinders and cones with non-uniform temperature dif- [28]. Senapati et al. [29] numerically investigated heat transfer over
ferences between the surface and the ambient fluid. Bejan and Lage a horizontal cylinder with concentric annular fins. Senapati et al.
[15] investigated the Prandtl number effect on the transition in [30] also numerically investigated the effect of eccentricity on heat
natural convection along a vertical surface. Day et al. [16] revisited transfer from a horizontal cylinder with annular fins. Considering
the topic of laminar natural convection from isothermal vertical the above literature survey, it would indeed be useful to study
cylinders. numerically natural convection heat transfer from a vertical cylin-
There is extensive experimental literature when considering the der with annular fins. The lack of experimental or numerical study
effect of fins on natural convection over cylinders Natural convec- is the reason for the incomplete picture of such a system. Numerical
tion from a vertical cylinder with vertically oriented plate fins have investigation of natural convection over an annular finned vertical
been studied experimentally by An et al. [17], where the authors cylinder with equal fin spacing is of relevance to industrial
have proposed a correlation for estimating the Nusselt number. applications.
Park et al. [18] have studied experimentally the effect of natural In the present study, numerical simulation of full Navier-Stokes
convection over branched fins fitted with a vertical cylinder. Recent equation along with energy equation has been conducted for the
advances in semiconductor technology have led to a significant natural convection over a vertical cylinder with annular fins of
increase in the power density threshold of microelectronic devices constant thickness for both laminar and turbulent flow regimes.
suggested by Bar-Cohen [19] and Oktay et al. [20]. This higher po- Computations have been performed by varying the parameters
wer density leads to an increase in the junction temperature, which such as Rayleigh number, the fin to tube diameter ratio, the number
is harmful to device performance and reliability. Therefore, a highly of fins (fin spacing). The effect of the aforementioned parameters
efficient cooling mechanism is needed to achieve reliability of on Nu is analyzed, and correlations for average Nusselt number for
electronic equipment. Many cooling methods for thermal man- finned vertical cylinder have been developed for both laminar
agement have been proposed; among those natural convective heat (104
Ra
108) and turbulent (1010
Ra
1012) flows (transition
transfer has been proven to be appropriate for these devices, to turbulence occurs at Rayleigh number 108 for natural convective
because of their simplicity, long-term low cost, and minimum flow over a vertical cylinder) when the fin to tube diameter ratio, fin
maintenance. There is some literature readily available on how the spacing to tube diameter ratio are in the range of 2e5 and
148 J.R. Senapati et al. / International Journal of Thermal Sciences 111 (2017) 146e159

0.126e5.840 respectively. Educational version of ANSYS Fluent has the analysis, which has been avoided at the first level. The principal
been used to solve the governing equations those feature in the objective is to compute the natural convection heat transfer for this
present numerical study. Optimization study of the conjugate heat finned vertical cylinder and visualize the fluid flow pattern around
transfer characteristics has been conducted to find the best fin the cylinder and the fin.
spacing for maximum heat transfer. A correlation of optimum fin
spacing for maximum heat transfer has been proposed, which may 3. Mathematical modeling
be important for industrial applications such as the design of heat
exchangers. The flow is assumed to be steady, laminar or turbulent
depending on the Rayleigh number and the fluid is assumed to be
incompressible. The surrounding fluid i.e. air is considered to be
2. Problem description
incompressible ideal gas. The other thermo-physical properties are
deemed to be constant. Viscous dissipation is assumed to be
A vertical cylinder of diameter (d) 25 mm with fins of uniform
negligible. Reynolds averaged Navier-Stokes (RANS) equations are
thickness (t) 1 mm having a constant inter-fin spacing (S) is taken
used for predicting the turbulent flow. The two equation eddy
for the analysis as shown in Fig. 1. The material for the fin is chosen
viscosity based k - ε model is used for calculating the turbulent
to be aluminum because of its high thermal conductivity. Although
viscosity (mt). Boussinesq turbulent viscosity hypothesis is used to
the results are reported mostly in non-dimensional form, some of
link the Reynolds stresses to the velocity gradients. The funda-
the important dimensional values are given here can help the
mental conservation equations governing the fluid flow and heat
prospective researchers in reproducing the results of the present
transfer by assuming the above conditions for the present problem
study.
are given below in their compact Cartesian form.
The simulation of natural convection on a vertical cylinder with
Continuity equation
annular fins can be modeled by a two-dimensional axisymmetric
geometry as it is clear from Fig. 1. Fig. 1(a) shows the isometric view vvj
of a vertical cylinder fitted with annular fins, while Fig. 1(b) shows ¼0 (1)
vxj
the schematic diagram of the computational domain with annular
fins on a vertical cylinder in cross sectional view (two-dimensional Momentum equation
axisymmetric geometry) along with the boundary conditions. The
 " !#
surface of the cylinder is maintained at a constant temperature (Tw) v rvi vj vðp þ 2=3kÞ v vvi vvj
which is usually the case in industrial practice (at least for a short ¼ rgi  þ ðm þ m t Þ þ (2)
vxj vxi vxj vxj vxi
distance). However, the varying surface temperature could have
been solved quickly but it would have added a further dimension to Energy equation

Fig. 1. (a) 3D view of a vertical cylinder fitted with annular fins (b) Schematic diagram of computational domain showing annular fins on a vertical cylinder in cross sectional view
(2D axisymmetric geometry).
 J.R. Senapati et al. / International Journal of Thermal Sciences 111 (2017) 146e159 149

 "  # connection with the fins and tube treated as an axis as has been
v rvj T v m mt vT shown in Fig. 1. At the pressure outlet boundary, atmospheric
¼ þ (3)
vxj vxj Pr Prt vxj pressure would prevail, and the backflow (if any) would have a
temperature of the surroundings only.
The turbulent eddy viscosity, mt is given by:
p ¼ patm ; T ¼ T∞ (12)
k2
mt ¼ rCm (4) Energy equation for conduction in solid
ε
There would be conduction in the fins which can be solved by
where Cm is a model constant. the following equation in conjugate heat transfer mode along with
Turbulent kinetic energy (k) the convective heat transfer in the fluid medium.
"  #
v
 v m vk V2 T ¼ 0 (13)
rkvj ¼ mþ t þ Gk þ Gb  rε (5)
vxj vxj sk vxj Interfacial conditions
At the solid-fluid interface, the condition of heat flux continuity
Dissipation rate (ε)
is given by,
"  #
v
 v m vε ε    
rεvj ¼ mþ t þ ðC1ε Gk þ C1ε C3ε Gb  C2ε rεÞ vTS vTf
vxj vxj sε vxj k kS ¼ keff (14)
vn interface vn interface
(6)
where,
where, sk and sε are Prandtl numbers for k and ε respectively. Gk in
Eqs. (5) and (6) representing the generation of turbulent kinetic
cp mt
keff ¼ kf þ (15)
energy due to mean velocity gradient is computed from the Prt
following expression: The initial pressure and the velocity field were set to zero
! ! everywhere at the beginning of the computation. To link the so-
vvi vvj vvi lution variables at the near-wall cells to the interior of the
Gk ¼ mt þ (7)
vxj vxi vxj computational domain, we have used log law wall function based
on the proposal of Launder and Spalding [32]. The law-of-wall for
In Eqs. (5) and (6) Gb represents the generation of turbulent the mean velocity is given as:
kinetic energy due to buoyancy effects and is given as:
! 1=2 1=4
up kp cm 1

mt vT ¼ ln Ey* (16)
Gb ¼ bgj (8) tw =r k
Prt vxj
3=2 3=4
The value of the constants in Eqs. (5) and (6) are the following, as kp cm
εp ¼ (17)
has been suggested by Launder and Spalding [31] kyp
C1ε ¼ 1:44; C2ε ¼ 1:92; sk ¼ 1:0; sε ¼ 1:3; Cm ¼ 0:09; Prt
  rk1=2 1=4
vx  p cm yp
¼ 0:85; C3ε ¼ tanh  y* ¼ (18)
v y
m

If we do not solve the turbulent equations then k and ε equations where, E ¼ empirical constant ¼ 8.6, k ¼ von Karman
are not computed and hence the turbulent viscosity (mt) would be constant ¼ 0.41, up is the mean velocity of the fluid at the nearest
taken as zero in Eqs. (2) and (3). So, a laminar solution would be cell center point P to the wall, yp is the distance from the point P to the
obtained under such a case. wall, kp is the turbulence kinetic energy at the point P. The turbulent
Equation of state intensity at the pressure outlet boundary has been taken to be 5%
For the solution of Eq. (2) the value of r is needed as a function of for any backflow which is realistic according to the suggestions
local temperature which has been obtained from the state equation were given by Launder and Spalding [32].
of the fluid as has been written below. The surrounding fluid is air with incompressible ideal gas
assumption with the fluid properties measured at film tempera-
p ¼ rRT (9) ture, Tfilm ¼ ðTw þ T∞ Þ=2 where T∞ is the ambient temperature. The
results will henceforth be represented using the following non-
Boundary conditions
dimensional parameters.
Boundary conditions for the solution of Eqs. (2), (3), (5) and (6)
Rayleigh number based on the tube length can be calculated
are described here which has been used in the simulation process.
from the following relation:
At the axis:

v g bðTw  T∞ ÞL3
ðÞ ¼ 0 and vy ¼ 0 (10) Ra ¼ (19)
vy na
At the cylinder surface and the fin base: As per Newton's law of cooling, the total convective heat
transfer from the fins and the tube can be written as:
vx ¼ vy ¼ 0; T ¼ Tw (11)
Q ¼ hAðTw  T∞ Þ (20)
At the outer computational domain, the three outer sides are
given a pressure outlet boundary condition and the side in where, the average heat transfer coefficient (h) is based on the total
150 J.R. Senapati et al. / International Journal of Thermal Sciences 111 (2017) 146e159

surface area (A) which includes the fin area (Afin) and the bare tube
area (Ab). If there are Nfin fins in the tube-fin array, total surface area d 35
 (23)
can be given by: L Gr 1=4
L

A ¼ Nfin Afin þ Ab (21)

where, d is the diameter of the cylinder. Vertical cylinders which do
The total heat transfer rate, Q is obtained from the numerical not meet the criterion can still use the flat plate result if a multi-
computations. The average heat transfer coefficient, h can be plying factor F can be multiplied to it to account for the curvature
calculated using Eq. (20) as h ¼ Q =AðTw  T∞ Þ. effect as per the analysis of Minkowycz and Sparrow [12] for gasses
The Nusselt number for an unfinned or finned cylinder can be with Pr ¼ 0.7 and where,
calculated as:
F ¼ 1:3½ðL=dÞ=Grd 1=4 þ 1:0 (24)
QL
Nu ¼ (22) The experimental correlations of Churchill [6], Eq. (25), McAdam
AðTw  T∞ Þk
[33], Eq. (26) and Eckert and Jackson [34], Eq. (27) for a vertical
In Eqs. (20)e(22), Q and A for a finned cylinder would be cylinder in natural convection are written here for ready reference
different from that of the unfinned cylinder. But the definition of of the readers.
Nusselt number would still remain the same as shown in Eq. (22). It
is to be marked here that Nusselt number in Eq. (22) is directly 2 32
1=6
0:387ðRaÞ
proportional to the net heat transfer Q (defined in Eq. (20)) and Nu ¼ F 40:825 þ h  9=16 i 5 ; for 101
Ra
1012
inversely proportional to the total surface area (which includes the 8=27
fin area and the bare tube area) as defined in Eq. (21).
1 þ 0:492
Pr

(25)
4. Numerical procedure

The governing differential Eqs. (1), (2), (3), (5), and (6) were Nu ¼ 0:59FðRaÞ0:25 ; for 104
Ra
109 (26)
integrated over a control volume and then discretized using the
finite volume technique to obtain a set of algebraic equations. The
Nu ¼ 0:021FðRaÞ0:4 ; for 109
Ra
1013 (27)
detailed steps to deduce the discretization equation is given in the
paper by Senapati et al. [29]. The resulting algebraic equations were Fig. 2 shows the variation of Nusselt number with Rayleigh
solved by the algebraic multi-grid solver of FLUENT 15 in an iter- number for a vertical cylinder along with a comparison of the
ative manner by imposing the boundary conditions Eqs. (7), (8), existing experimental correlations (shown in Eqs. (25)e(27)). It can
(10) and (11). Body force weighted scheme was used for pressure- be observed from Fig. 2(a) and (b) that the present computation is
discretization. Second-order upwind scheme for the convective reasonably accurate in both the laminar and turbulent flow
terms was considered for the momentum and energy equations. regimes.
The diffusive terms of the momentum and energy equations were The grid distribution in the computational domain is shown in
discretized so that they are at least second order accurate in space Fig. 3. An enlarged view of the cylinder base with the fins is shown
(Central difference scheme was used). SIMPLE algorithm was used next to the figure. Three fins can be seen in the enlarged figure with
for coupling the pressure and the velocity terms for the pressure grids in it and also grids in the surrounding area of the fins. For a
correction equation. Normally a first order upwind scheme is particular fin diameter of D ¼ 125 mm and tube length of
initially adopted to get a converged solution, and after that the L ¼ 325 mm with 12 fins, the size of the computational domain
scheme is switched over to second order upwind which is believed would be 2 L  6 L (2 L in y and 6 L in x-direction). Near the base
to be more accurate due to the second order accuracy it uses. cylinder and the fin, the cells are too small and away from the fin
However, with a SIMPLE algorithm for pressure correction the towards the top surface in the axial direction the cells are also small
second order upwind scheme may not converge from the begin- so that the natural convection plume can be captured with higher
ning for natural convection where mass conservation becomes a accuracy. Elsewhere in the computational domain, the cells are
difficult task. That is why, we had to adopt initially the first order little larger because the flow in these areas is not changing too
upwind, and then switched over to second order upwind. Normally drastically. The total number of computational cells for the partic-
SIMPLE algorithm is more stable, and takes little less time ular case is 34,480. Quadrilateral (quad mesh) structured grids
compared to all other schemes. Hence, SIMPLE algorithm was used were used for meshing the rectangular computational domain. The
for present computation. The relative convergence criterion for the mesh was made finer near the surface of the base cylinder and fins
energy equation was set to 106 and for all other equations, it was to capture the boundary layer effect.
set to 104. If a stricter criterion could be used, like 105 or 106 for The number of computational cells was fixed to 34,480 after
the momentum and mass continuity equations it was observed that doing a grid independence test as shown in Fig. 4(a). The domain
there were convergence problems. So, adopting the residuals to independence test was also performed by varying the domain size
104 for all the cases could result in smoother convergence and (in the axial direction) as has been given in Fig. 4(b). Based on the
lesser computational time. The under-relaxation factors (normally results obtained from the domain independence test, the outer
decided from experience) used for the convergence of the variables computational domain height was restricted to 6 times of the cyl-
(relative whole field residuals) are shown in Table 1. inder length (L).
For vertical surfaces, the Nusselt number and Grashof number The turbulent heat transfer for an unfinned cylinder and a fin-
are the function of L, the height of the surface as the characteristic ned cylinder was computed using different turbulence models.
dimension. If the boundary-layer thickness is not large compared to Table 2 shows the average Nusselt number for an unfinned and
the diameter of the cylinder, the heat transfer may be calculated finned cylinder where we have used five different turbulence
with the same relations used for vertical plates. The general crite- models. It can be seen that the variation in Nusselt number pre-
rion is that a vertical cylinder may be treated as a vertical flat plate dicted by these models was within 3% for the case of a finned and
as suggested by Gebhart et al. [7] when unfinned cylinder. For an unfinned cylinder, the Nusselt number
 J.R. Senapati et al. / International Journal of Thermal Sciences 111 (2017) 146e159 151

Table 1
Under-relaxation factors used in the simulation.

Pressure Density Body forces Momentum Turbulent kinetic energy Turbulent dissipation rate Turbulent viscosity Energy

0.5 1.0 1.0 0.3 0.8 0.8 1.0 1.0

Fig. 2. Variation of Nusselt number with Rayleigh number: A comparison between the present computation and the experimental correlations.

Fig. 3. Grid arrangement over the finned cylinder in the computational domain.

computed by the standard k-ε model was in very good agreement for further analysis of finned cylinders in the present problem, since
with that of the correlation of Churchill. On the other hand, there is all the other turbulence models were also producing similar Nusselt
no established correlation available for annular finned vertical number (within a variation of only 3%) for the case of finned cyl-
cylinder. The two equation standard k-ε turbulence model was used inders as can be seen from Table 2. It can also be mentioned here
152 J.R. Senapati et al. / International Journal of Thermal Sciences 111 (2017) 146e159

Fig. 4. Variation of Nusselt number with computational cells (a) and domain size (b).

that the standard k-ε model was more robust compared to other Figs. 5 and 6 that as D/d increases the heat transfer from the fin
models tested in our present computation. So, the choice of the array increases for both laminar and turbulent flow at any partic-
turbulence models for our simulation was standard k-ε. ular S/d and Ra. As D/d increases, the fin area increases and hence,
the rate of heat transfer from the fin increases for both type of
5. Results and discussion flows.
Here, the increase in Rayleigh number is due to temperature
In the present study, the numerical simulation of full Navier- difference between the base cylinder and the ambient fluid. The
stokes equation along with energy equation has been solved for a lowest Rayleigh number corresponds to a DT of 50 K and the highest
vertical tube fitted with annular fins of constant thickness for both Rayleigh number to a DT of 150 K. So, it is now conclusive that as
laminar (104
Ra
108) and turbulent (1010
Ra
1012) flow. In Rayleigh number increases the heat transfer from the fin array in-
this section, visualization of temperature plume and flow field over creases in laminar as well as in turbulent flow. In turbulent flow, we
a vertical cylinder with annular fins has been provided for many have also used the same temperature difference ranges to compute
different fin configurations which may be helpful for the readers to Rayleigh number but the effective gravity has been suitably
have a complete physical understanding of the flow around the fin- adjusted to get higher values of Rayleigh number so that the flow
tube array. In the result and discussion section, we would explain remains in turbulent range. In laminar flow, the heat loss from the
the effect of various parameters such as S/d, D/d, and Ra on heat fins increases as S/d decreases (as Nfin increases) at any particular D/
transfer and Nusselt number. d and Rayleigh number. A decrease in S/d means more fins are
added to the tubes over the same length of the tube. So when the
fins are more the convection heat transfer from the fin array in-
5.1. Effect of fin spacing and fin diameter on heat transfer
creases as the total surface area for convection increases. As a result,
the net heat loss from the fin array increases as S/d decreases (Nfin
Figs. 5 and 6 show the rate of heat transfer from the vertical
increases). Further, if we go by adding fins to the base tube, S/d is
finned cylinder as a function of D/d and S/d where Rayleigh number
too small (close to 0.1) the fins are so close to each other that we can
varies as a result of the temperature difference between the base
say as if the tube has become thicker due to the addition of fins. So
cylinder and surroundings. Fig. 5 shows the heat transfer in laminar
heat loss from the tube is mainly due to conduction from a thicker
range and Fig. 6 is for the turbulent range. In the figures, the con-
tube.
vection heat transfer rates are plotted as a function of non-
In laminar flow, the convection velocity between the inter-fin
dimensional fin spacing, S/d at fixed D/d. The figures are drawn
spacing is also too low compared to turbulent flow. So, the heat
for fin-to-tube diameter ratio of 2, 3, 4 and 5 respectively. From
loss from the fins is more dependent on conduction rather than
Fig. 5, it can be seen that, at a given D/d and Ra, the convection heat
convection between the inter-fin spacing as the flow obstruction
transfer rate from the fin array goes on increasing for laminar flow.
happens due to smaller inter-fin spacing. But in turbulent flow, the
On the other hand, for turbulent flow as in Fig. 6, the heat transfer
convection between the inter-fin spacing is much stronger
rate first increases up to a maximum at around S/d ¼ 0.3 and then
compared to that of laminar flow, so the heat loss from the fins
decreases when S/d further decreases to 0.1. It can also be seen from
depend on two aspects of heat transfer, i.e. convection as well as
conduction. Since turbulent offers much of mixing in the flow so
the convection between the inter-fin spacing remains stronger and
Table 2
Nusselt number and total heat transfer from a finned and unfinned cylinder
helps the heat removal from the fin. As S/d decreases (with the
computed using different turbulence models for Ra ¼ 1.135  1011. addition of more fins), the inter-fin convection also falls but not that
much. This time the heat removal rises from the fin (due to more fin
Unfinned cylinder Finned cylinder [D/d ¼ 5; S/
d ¼ 1.0292]
being present) and reduces from the tube base for which the overall
heat loss increases with S/d falling. But when S/d falls too much (too
Nu Q [Watt] Nu Q [Watt]
many fins are now present on the tube), the inter-fin convection
Churchill relation [6] 578.78 62.48 Not available Not available becomes too weak, and the fin arrays lose heat mostly from the tip
k-ε Standard model 582.52 62.88 178.10 235 portion. As a result, there is a decrease in net heat loss. This is the
k-ε RNG model 574.31 62.01 182.45 240.8
k-ε Realizable model 567.46 61.26 174.69 230.5
main reason for observing maximum heat loss in turbulent flow for
k-u Standard model 593.66 64.09 176.37 232.7 a particular S/d and the heat loss decreasing thereafter as S/
k-u SST model 586.89 63.36 179.89 237.4 d further decreases.
 J.R. Senapati et al. / International Journal of Thermal Sciences 111 (2017) 146e159 153

Fig. 5. Variation of heat transfer rate with non-dimensional fin spacing for different values of diameter ratio, D/d and Ra in laminar flow.

Fig. 6. Variation of heat transfer rate with non-dimensional fin spacing for different values of diameter ratio, D/d and Ra in turbulent flow.

5.2. Effect of fin diameter and Rayleigh number on optimum fin last Section 5.1. So an optimization study has been carried out to get
spacing for maximum heat transfer in turbulent flow the optimum fin spacing for maximum heat transfer for the case of
turbulent flow. The estimated optimum fin spacing for a given fin
In laminar flow the heat loss from the fin arrays was seen to be diameter, D and Rayleigh number are shown in Table 3. Table 3 is
increasing with decrease of S/d and no optimum fin spacing was shown in dimensional form just to get ideas about real life appli-
found which could give rise to highest heat transfer. However, for cation. The values from the table are shown graphically in Fig. 7,
the case of turbulent flow, there exists an optimum fin spacing for which shows that with increase in fin diameter, D, (S/d)opt increases
maximum heat transfer, the reason for which was explained in the at constant Ra. This also signifies that one has to use less number of
154 J.R. Senapati et al. / International Journal of Thermal Sciences 111 (2017) 146e159

Table 3 convection in between the fins remains moderately stronger. This

Variation of optimum fin spacing (Sopt in mm) with fin diameter, D for different Ra. can happen only when S/d has a higher value. So it is now imper-
Ra D ¼ 125m D ¼ 100 mm D ¼ 75 mm D ¼ 50 mm ative that with increase in D/d, S/d has to increase to give highest
1.135  1011 7.688 7.346 7.286 7.237
heat loss from the fin arrays.
1.617  1011 7.571 7.239 7.175 7.122 In turbulent flow, the convection between the inter-fin spacing
1.773  1011 7.46 7.139 7.08 7.02 is much stronger. Since turbulent offers much of mixing in the flow
so the convection between the inter-fin spacing remains stronger
and helps the heat removal from the surface of the fins. Hence, to
get the flow obstruction in between the inter-fin spacing, for a
particular value of D/d the inter-fin spacing would be smaller for
higher Rayleigh number. (S/d)opt decreases for a fixed D/d when Ra
increases as shown in Fig. 7.

5.3. Effect of S/d and D/d on Nusselt number

Variation of average Nusselt number, Nu with non-dimensional

fin spacing, S/d is shown in Fig. 8 as a function of Ra in laminar flow.
From the figure, it can be seen that as S/d increases (no of fins
become less), Nu increases for any Rayleigh number. With an in-
crease in Ra, there is an increase in Nu for a particular value of S/d.
As D/d increases, Nu decreases negligibly for any S/d. As S/d in-
creases, we have less number of fins on the tube and hence there
would be less heat transfer from the system. The definition of
Nusselt number as defined by Eq. (22) describes, Nusselt number as
reciprocal of total surface area (both fin area and bare tube area) for
Fig. 7. Variation of non-dimensional optimum fin spacing, (S/d)opt with diameter ratio, a total convective heat loss, Q. As S/d decreases there would be more
D/d at different values of Ra in turbulent regime. heat loss from the system but the fin area also increases, so the
overall ratio (which is the definition of Nu) decreases which sig-
nifies that Nu decreases with a decrease in S/d.
fins when the fin diameter is large to get maximum heat loss. As the Fig. 9 shows the variation of Nu with S/d in turbulent flow. Here
fin diameter increases with constant inter-fin spacing (for a also we get the same kind of behavior and same explanation would
particular S/d) the flow between the inter-fin spacing gets more fit to this situation as well. In turbulent flow, we get maximum heat
resistance and hence the convection current falls in between the loss from the fin array when more number of fins are added to the
inter-fin spacing and the heat loss decreases from here onwards base cylinder and then the heat loss falls with decrease in S/d (as
with addition of more fins. In order to increase the heat loss from Nfin increases). So Nusselt number falls suddenly with decrease in S/
the fins it is desired to increase the inter-fin spacing so that the d for turbulent flow. The peak in heat loss would not be seen as a

Fig. 8. Variation of Nusselt number with non-dimensional fin spacing, S/d at different values of diameter ratio, D/d and Ra for laminar flow.
 J.R. Senapati et al. / International Journal of Thermal Sciences 111 (2017) 146e159 155

Fig. 9. Variation of Nusselt number with non-dimensional fin spacing, S/d at different values of diameter ratio, D/d and Ra for turbulent flow.

peak in the Nu verses S/d as the ratio of rising Q to rising fin area flow comes in the inter-fin spacing from the side of the domain and
would hide this effect. exit to the top of the domain. Since the plume is much stronger in
the case of turbulent flow, so the flow is attracted from the side of
5.4. Effect of fin spacing on temperature plume the domain and gets the exit to the top of the domain. But for the
case of laminar flow, the plume is weaker, and circulation starts in
Fig. 10 shows the temperature plume over the fins for varying S/ between the inter-fin spacing, so the flow gets pulled from the
d in laminar flow and Fig. 11 indicates that for turbulent flow. It can nearest neighborhood which causes a downward flow near the tip
be seen from Fig. 10 that the thermal boundary layer for the case of region of the fin. Comparing Figs. 12 and 13, we can clearly see that
S/d ¼ 5.84 is too thick compared to that for the same case in tur- the velocity field is stronger in the case of turbulent flow with
bulent flow. So the heat transfer coefficient or Nusselt number for compare to that of laminar flow for the same S/d. Thus resulting
the case of laminar flow is comparatively small to that of the tur- more convection heat transfer in case of turbulent flow than that of
bulent flow. As S/d decreases (number of fins goes on increasing on laminar flow.
the tube surface) and the thermal plume rises over the fin surface As S/d decreases (the number of fins increases), the circulation in
and then glide along the fins axially taking away the heat from the the inter-fin spacing becomes stronger, and this circulation has to
fins. When the fins are six in numbers (S/d ¼ 2.312), the plume goes pull up flow from the nearby region of fin tip, so the flow comes into
inside the fins due to larger inter-fin spacing and when the fins are the inter-fin spacing from the top of the domain and the flow
much more in number, the plume does not go inside the inter-fin leaving the inter-fin spacing creates rising plume just near the tip of
spacing. For S/d ¼ 0.182 the plume does not go inside the inter- the fins. This phenomenon happens in all the cases of laminar and
fin spacing, rather the fins form a solid cylinder-like surface (i.e. it turbulent flow. But it has to be marked that in laminar flow the
behaves like cylinder having diameter with that of fins) from where plume which comes from the top is a cooler plume and the rising
the heat is lost as a rising plume. This is what we see from Fig. 10 in plume near the tip area of the fins is a hotter plume. This more
laminar flow. But from Fig. 11 for the same case of S/d ¼ 0.182 it is heated plume creates a blanket near the fin tip region and does not
seen that the plume enters mildly into the inter-fin spacing and allow the colder plume from top to mix with it and, therefore,
takes away heat from the fins. As the number of fins increases, the hinders the heat transfer from the fins. But in the case of turbulent
inter-fin spacing goes on decreasing and hence the penetration of flow, the plume is drawn from the side of the domain which is
the plume into the inter-fin spacing falls gradually and heat transfer mainly cold, and that gets into the inter-fin spacing and comes out
also falls. This behavior was seen from Fig. 6 where we could see the of the spacing as a rising plume along the tip of the fins in the axial
heat loss from the fins decreasing as the number of fins increasing direction. This phenomenon enhances the heat transfer from the
beyond a particular value for turbulent flow. fins as the inter-fin plume is very useful in taking out the heat from
the surface of the fins which does not happen in the case of laminar
flow.
5.5. Effect of fin spacing on flow field around the fins

Fig. 12 show the velocity field around the fins for D/d ¼ 5 and 6. Some useful correlations
Ra ¼ 1.135  107 (laminar flow) while Fig. 13 shows it for
Ra ¼ 1.135  1011 (turbulent flow) for various values of S/d. It can be 6.1. Correlation for Nusselt number
seen from Fig. 12 that the fluid from the top is pulled into the inter-
fin area in the case of laminar flow, while for turbulent flow the The correlations for the average Nusselt number for natural
156 J.R. Senapati et al. / International Journal of Thermal Sciences 111 (2017) 146e159

Fig. 10. Contours of static temperature with varying fin spacing in laminar flow; for D/d ¼ 5 and Ra ¼ 1.135  107.

Fig. 11. Contours of static temperature with varying fin spacing for turbulent flow; for D/d ¼ 5 and 1.135  1011.

convection from a vertical cylinder show that the Nusselt number is to tube diameter ratio, S/d. The functional form of the relationship
a function of Rayleigh number. However, the addition of fins to the can be expressed as:
vertical cylinder will change the Nusselt number. Thus, for a given
thickness of the fins, the Nusselt number from a finned vertical  
S
cylinder will depend on the diameter of the fins and the inter-fin Nu ¼ f1 Ra; x; (28)
d
spacing. Therefore, the average Nusselt number (Nu) over a fin-
ned vertical cylinder can be described as a function of the Rayleigh To develop the correlations for annular finned vertical cylinder,
number, the fin-to-tube diameter ratio (x) and the inter-fin spacing computations were carried out by varying the number of fins
(0e72) or S/d ranges from 0.126 to 5.840, fin-to-tube diameter ratio
 J.R. Senapati et al. / International Journal of Thermal Sciences 111 (2017) 146e159 157

Fig. 12. Plots of velocity vector with varying fin spacing in laminar range; for D/d ¼ 5 and Ra ¼ 1.135  107.

(2e5) and the base-to-ambient temperature difference (50e150 K)  a a

where Ra varies from 104 to 108 in laminar flow and for turbulent Nu ¼ c0 þ c1 Raa0 ðS=dÞa1 x 2 þ c2 ðRa=xÞ 3 (30)
flow Ra varies from 1010 to 1012. The correlations were made by
performing a non-linear regression analysis of the data from the where,
computational results using POLYMATH 6.0, and the following c0 ¼ 202:013; c1 ¼ 4:159; c2 ¼ 1:027; a0 ¼ 0:235; a1 ¼
correlations were obtained which may be useful for the design of 0:095; a2 ¼ 0:325; a3 ¼ 0:266 with properties evaluated at film
heat exchangers and optimization of the number of fins for temperature, Tfilm.
maximum heat transfer. The Nusselt number correlation for For Eq. (30), the coefficient of determination (R2) is 0.9716 and
104
Ra
108 is developed as, p
0.001. Fig. 14 shows a good agreement between the predicted
and the computed values of Nusselt number for turbulent flow.

 a1   6.2. Correlation for optimum fin spacing

S=d S=d a3 a4
NuLaminar ¼ c0 þ c1 Raa0 þ c2 Raa2 þ c3 x
x x
It is clear from Fig. 5 that for laminar range, the heat transfer
(29) from the cylinder fin system goes on increasing as the number of
fins increases signifying no existence of optimum fin spacing,
where,
whereas, there exists an optimum fin spacing for maximum heat
transfer for turbulent flow as shown in Fig. 6. The optimum fin
spacing for maximum heat transfer is the desired parameter for any
c0 ¼ 1062:851; c1 ¼ 18:289; c2 ¼ 17:511; c3 ¼ 1058:885 design of heat exchangers with annular fins. So, it is intended to
a0 ¼ 0:212; a1 ¼ 0:881; a2 ¼ 0:211; a3 ¼ 0:906; a4 ¼ 0:003 develop a correlation for the optimum fin spacing. The ratio of
optimum fin spacing to base tube diameter, (S/d)opt may be defined
For Eq. (29), the coefficient of determination (R2) is 0.9972 and
as a function of Rayleigh number based on fin spacing and fin-to-
p
0.001. Fig. 14 shows a good agreement between the predicted
tube diameter ratio. The functional form can be written as,
and the computed values of Nusselt number for laminar flow. The
 
equation is valid for the laminar range of 104
Ra
108, with S
properties evaluated at film temperature, Tfilm. ¼ f2 ðRa; xÞ (31)
d opt
The Nusselt number correlation for 1010
Ra
1012 is devel-
oped as, To develop the correlation for optimum fin spacing,
158 J.R. Senapati et al. / International Journal of Thermal Sciences 111 (2017) 146e159

Fig. 13. Plots of velocity vector with varying fin spacing in turbulent range; for D/d ¼ 5 and Ra ¼ 1.135  1011.

Fig. 14. Predicted and computed values of Nusselt number for laminar range (104
Ra
108) and for turbulent range (1010
Ra
1012).

computations were done by varying the Rayleigh number and fin- p
0.001. Fig. 15 shows a good agreement between the predicted,
to-tube diameter ratio. The correlations were derived by perform- Eq. (32) and the computed values of (S/d)opt. This correlation is valid
ing a non-linear regression analysis of the data from the results in the Rayleigh number range 1010
Ra
1012, with properties
using POLYMATH 6.0, and the following correlation was obtained evaluated at film temperature, Tfilm. Eq. (32) is very useful for in-
which is a very simple relation to be employed in the actual design. dustrial applications especially for designing of heat exchangers.
Using this equation, we can predict the optimum fin spacing or
   a1 required number of fins for maximum heat transfer on an annular
S Ra finned vertical cylinder.
¼ c0 þ c1 Raa0 þ c2 þ c3 xa2 (32)
d opt x

where, c0 ¼ 5:088; c1 ¼ 54:046; c2 ¼ 0:069; c3 ¼ 1:01  107 ; 7. Conclusions

a0 ¼ 0:112; a1 ¼ 0:107; a2 ¼ 9:356.
For Eq. (32), the coefficient of determination (R2) is 0.9809 and Natural convection heat transfer from an array of an annular
 J.R. Senapati et al. / International Journal of Thermal Sciences 111 (2017) 146e159 159

References

[1] Bejan A. Convection Heat Transfer. third ed. India: Wiley; 2004.
[2] Kreith F, Manglik RM, Bohn MS. Principles of Heat Transfer. seventh ed.
Cengage Learning; 2011.
[3] Güvenç A, Yüncü H. An experimental investigation on performance of fins on
a horizontal base in free convection heat transfer. Heat Mass Transf. 2001;37:
409e16.
[4] Yazicioǧlu B, Yüncü H. Optimum fin spacing of rectangular fins on a vertical
base in free convection heat transfer. Heat Mass Transf. Stoffuebertrag.
2007;44:11e21.
[5] Churchill SW, Chu HHS. Correlating equations for laminar and turbulent free
convection from a horizontal cylinder. Int. J. Heat Mass Transf. 1975;18:
1049e53.
[6] Churchill SW. A comprehensive correlating equation for laminar, assisting,
forced and free convection. AIChE J. 1977;23:10e6.
[7] Gebhart B, Jaluria Y, Mahajan RL, Sammakia B. Buoyancy-Induced Flows and
Transport. New York: Hemisphere Publishing Corporation; 1988.
€ ̧ık MN. Heat Transfer a Basic Approach. Singapore: McGraw-Hill Book
[8] Ozıs
Company; 1985.
[9] Holman JP. Heat Transfer. tenth ed. New York: McGraw-Hill Higher Education;
2010.
[10] Popiel CO, Wojtkowiak J, Bober K. Laminar free convective heat transfer from
isothermal vertical slender cylinder. Exp. Therm. Fluid Sci. 2007;32:607e13.
Fig. 15. Predicted and computed values of (S/d)opt.
[11] LeFevre EJ, Ede AJ. Laminar free convection from the outer surface of a vertical
circular cylinder. In: Proc. 9th Int. Congr. Appl. Mech.; 1956. p. 175e83.
[12] Minkowycz W, Sparrow E. Local nonsimilar solutions for natural convection
finned vertical cylinder has been numerically investigated for both on a vertical cylinder. J. Heat Transf. 1974;96:178e83.
laminar flow in the range 104
Ra
108 and turbulent flow in the [13] Fujii T, Uehara H. Laminar natural convection heat transfer from the outer
surface of a vertical cylinder. Int. J. Heat Mass Transf. 1970;13:607e15.
range 1010
Ra
1012. The computations were carried out by [14] Kuiken HK. Axisymmetric free convection boundary-layer bodies. Analysis
varying the fin to tube diameter ratio (D/d), fin spacing to tube 1968;11:1141e53.
diameter ratio (S/d) in the range of 2e5 and 0.126e5.840 respec- [15] Bejan A, Lage JL. The Prandtl number effect on the transition in natural con-
vection along a vertical surface. J. Heat Transf. 1990;112:787e90.
tively. The effect of parameters like; D/d, S/d and Ra on Nu are [16] Day JC, Zemler MK, Traum MJ, Boetcher SKS. Laminar natural convection from
analyzed, and correlations for average Nusselt number has been isothermal vertical cylinders: revisiting a classical subject. J. Heat Transf.
developed for both laminar and turbulent regimes. The following 2013;135:022505.
[17] An BH, Kim HJ, Kim DK. Nusselt number correlation for vibration-assisted
important observations could me made: convection from vertically oriented plate fins. Exp. Therm. Fluid Sci.
2012;41:59e66.
1. With the addition of fins to the isothermal cylindrical wall, the [18] Park KT, Kim HJ, Kim D-K. Experimental study of natural convection from
vertical cylinders with branched fins. Exp. Therm. Fluid Sci. 2014;54:29e37.
heat transfer increases monotonously for laminar flow, and for [19] Bar-Cohen A. Thermal management of electronic components with dielectric
turbulent flow, the heat transfer increases to a maximum value liquids. JSME Int. J. Ser. B 1992:2656e8.
and then decreases with the further addition of fins. [20] Oktay S, Hannemann RJ, Bar-Cohen A. High heat from a small package.
J. Mech. Eng. 1986;108:36e42.
2. Optimum fin spacing for maximum heat transfer in laminar flow
[21] Kim HJ, An BH, Park J, Kim DK. Experimental study on natural convection heat
was found to be non-existent in the present study, whereas for transfer from horizontal cylinders with longitudinal plate fins. J. Mech. Sci.
turbulent flow, an optimum fin spacing for maximum heat Technol. 2013;27:593e9.
transfer was found to be existing. [22] Littlefield JM, Cox JE. Optimization of annular fins on a horizontal tube.
Wa €rme Stoffübertrag. 1974;7:87e93.
3. The optimum fin spacing for maximum heat transfer in turbu- [23] Kraus AD, Aziz A, Welty J. Extended Surface Heat Transfer. New York: John
lent flow varies between 7 and 7.7 mm (S/d ¼ 0.28 to 0.31) for Wiley & Sons; 2001.
Rayleigh number in the range of 1010
Ra
1012. [24] Tsubouchi T, Masuda M. Natural convection heat transfer from horizontal
cylinders with circular fins. In: Proc. 6th Int. Heat Transf. Conf.; 1970. Paper
4. With an increase in fin spacing or S/d, the Nusselt number NC 1.10 Paris.
(average heat transfer coefficient as well) increases for both [25] Edwards JA, Chaddock JB. Free convection and radiation heat transfer from
laminar and turbulent flow. fin-on-tube heat exchangers. ASHRAE Trans. 1963;69:313e22.
[26] Al-Arabi M, Salman YK. Laminar natural convection heat transfer from in-
5. Visualization of temperature plume and flow field over the clined surfaces. Int. J. Heat Mass Transf. 1980;23:45e51.
vertical cylinder with annular fins has been provided for many [27] Al-Arabi M, Khamis M. Natural convection heat transfer from inclined cylin-
different fin configurations which are not reported earlier in the ders. Int. J. Heat Mass Transf. 1982;25:3e15.
[28] Singh B, Dash SK. Natural convection heat transfer from a finned sphere. Int. J.
literature. Heat Mass Transf. 2015;81:305e24.
6. The average Nusselt number over an annular finned vertical [29] Senapati JR, Dash SK, Roy S. Numerical investigation of natural convection
cylinder depend on Rayleigh number, fin-to-tube diameter ratio, heat transfer over annular finned horizontal cylinder. Int. J. Heat Mass Transf.
2016;96:330e45.
(D/d) and ratio of inter-fin spacing to tube diameter ratio, (S/d).
[30] Senapati JR, Dash SK, Roy S. 3D numerical study of the effect of eccentricity on
7. The correlations for Nusselt number have been developed for a heat transfer characteristics over horizontal cylinder fitted with annular fins.
vertical cylinder with annular fins in both the laminar and tur- Int. J. Therm. Sci. 2016;108:28e39.
bulent flow regime. These correlations can be used directly for [31] Launder BE, Spalding DB. Lectures in Mathematical Models of Turbulence.
New York: Acad.; 1972.
industrial applications and academic purposes. [32] Launder BE, Spalding DB. The numerical computation of turbulent flows.
8. The ratio of optimum fin spacing to cylinder diameter, (S/d)opt Comput. Methods Appl. Mech. Eng. 1974;3:269e89.
for maximum heat transfer is found to be a function of Rayleigh [33] McAdams WH. Heat Transmission. third ed. New York: McGraw-Hill; 1954.
[34] Eckert ERG, Jackson TW. Analysis of Turbulent Free-Convection Boundary
number and fin-to-tube diameter ratio, (D/d). Layer on Flat Plate. 1951. NACA Rep. 1015.