International Journal of Thermal Sciences 55 (2012) 90e102

Contents lists available at SciVerse ScienceDirect

International Journal of Thermal Sciences

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

Numerical simulation of laminar forced convection heat transfer of Al2O3ewater

nanouid in a pipe with return bend
Jongwook Choi a, *,1, Yuwen Zhang b
a
b

School of Mechanical and Aerospace Engineering, Sunchon National University, Jeonnam 540-742, Korea
Department of Mechanical and Aerospace Engineering, University of Missouri, Columbia, MO 65211, USA

a r t i c l e i n f o

a b s t r a c t

Article history:
Received 23 April 2011
Received in revised form
27 December 2011
Accepted 30 December 2011
Available online 17 January 2012

Laminar forced convection heat transfer of the Al2O3ewater nanouid in a pipe with a return bend is
analyzed by using a nite element method. The results show that the average Nusselt number increases
with increasing Reynolds number and Prandtl number, and the increment of specic heat in the
nanouid contributes to the heat transfer enhancement. The average Nusselt number in the return bend
appears higher than that in the inlet and outlet pipes due to the secondary ows. However, the pressure
drop in the pipe largely increases with the increment of nanoparticle volume concentration. The
empirical correlations for the average Nusselt numbers are obtained as functions of the Dean number
and the Prandtl number.
2012 Elsevier Masson SAS. All rights reserved.

Keywords:
Correlation equation
Finite element method
Heat transfer enhancement
Nanouid
Return bend

1. Introduction
Various industries in many countries have made great efforts to
reduce the greenhouse emissions that cause global warming. To
achieve this, the active endeavor to increase the energy efciency
has been made in industrial settings. The methods to improve the
efciency of heat exchangers can also be one of these many efforts.
In this connection, many researchers have been interested in the
geometry of the heat exchanger to get higher thermal efciency
[1e3]. Meanwhile, other research groups have been trying to
enhance the heat transfer in pipe systems by using nanouids. The
concept of the nanouids started from the fact that the thermal
conductivity of the uid containing suspended nanoparticles is
improved more than that of the conventional heat transfer uid [4].
Before the nanoparticles appeared, the researches on the uid and
heat transfer problems had been carried out with millimeter- or
micrometer-sized particles [5e7]. These particles were not put to
practical use because they may have brought about poor suspension stability, channel clogging, system abrasion and so on.

* Corresponding author. Tel.: 82 61 750 3826; fax: 82 61 750 3820.

E-mail addresses: choijw99@scnu.ac.kr (J. Choi), zhangyu@missouri.edu
(Y. Zhang).
1
Presently visiting professor at Department of Mechanical and Aerospace Engineering, University of Missouri, Columbia, MO 65211, USA. Tel.: 1 573 268 3961;
fax:1 573 884 5090.
1290-0729/$ e see front matter 2012 Elsevier Masson SAS. All rights reserved.
doi:10.1016/j.ijthermalsci.2011.12.017

However, the uids including nanometer-sized particles do not

have those serious problems and the measured thermal conductivity of the nanouid is anomalously greater than the theoretical
prediction [8].
The nanouid technology has been studied and developed by
many research groups worldwide [9,10]. Effects of particle volume
concentration, particle material, particle size, particle shape, base
uid material, temperature, additive, and acidity on heat transfer
enhancement were investigated experimentally by multiple
research groups. Although the heat transfer data were partially
inconclusive or conicting, the heat transfer enhancement
appeared in the range of 15e40% [11].
Straight pipes have been mainly utilized in the experimental or
numerical studies on the convective heat transfer enhancement
using nanouids. Pak and Cho [12] experimentally investigated the
turbulent friction and the heat transfer behaviors of the nanouids
such as gAl2O3ewater and TiO2ewater in a circular tube at the
Reynolds numbers of 104e105. The measured viscosities at the
volume concentration of 10% were approximately 200 times
greater for the gAl2O3ewater and 3 times greater for the
TiO2ewater than for the pure water. Also, the Nusselt number of
the nanouids increased with the augmentation of the volume
concentration as well as the Reynolds number. However, the results
showed that the convective heat transfer coefcient of the nanouids was smaller than that of the pure water under the condition
of same average velocity.

J. Choi, Y. Zhang / International Journal of Thermal Sciences 55 (2012) 90e102

Xuan and Li [13] measured the convective heat transfer coefcient and the friction factor of the Cuewater nanouid in the
turbulent ow with the Reynolds numbers of 10,000e25,000. The
nanouid had the larger convective heat transfer coefcient than
the base uid at the same Reynolds number. However, the friction
factor of the nanouid was almost same as that of the pure water
because of the low volume fractions of 0.3e2.0%. Wen and Ding [14]
reported an experimental work on the convective heat transfer of
the gAl2O3ewater nanouid owing through a copper tube in the
laminar ow regime with the Reynolds numbers of 500e2100. The
nanoparticles enhanced the convective heat transfer with the
increment of the Reynolds number and the particle volume
concentration. They also concluded that the enhancement of the
convective heat transfer could not be solely attributed to the
improvement of the effective thermal conductivity.
Maga et al. [15] numerically studied the turbulent ow of
gAl2O3ewater and gAl2O3eethylene glycol nanouids inside the
single uniformly heated tube at the Reynolds numbers of 10,000
and 50,000. It showed that the gAl2O3eethylene glycol nanouid
offered a better heat transfer enhancement than the gAl2O3ewater
nanouid. However, the wall shear stress in the gAl2O3eethylene
glycol was higher than that in the gAl2O3ewater. Also, Maga
et al. [16] continued to perform the numerical simulations for the
laminar ow with the gAl2O3ewater nanouid at the Reynolds
numbers of 250, 500, and 1000 and with the gAl2O3eethylene
glycol at the Reynolds numbers of 6.31, 63.1, and 631, respectively. The tendencies of the convective heat transfer coefcient
and the wall shear stress were similar to the previous results.
Ding et al. [17] experimentally studied the heat transfer
behavior of the CNT(carbon nanotubes)-water nanouid owing
through a horizontal tube at the Reynolds numbers of 800e1200.
Over 350% of the maximum heat transfer enhancement was obtained in the weight concentration of 0.5% at the Reynolds number
of 800. It was proposed that the observed large enhancement
could be caused by the particle re-arrangement, reduction of
thermal boundary layer thickness, very high aspect ratio of CNT, as
well as the thermal conductivity. Chun et al. [18] gained the
convective heat transfer coefcient of nanouids made of several
alumina nanoparticles and transformer oil at the Reynolds
numbers of 100e450. One of the possible reasons for the heat
transfer enhancement was explained to be the high concentration
of the nanoparticles in the thermal boundary layer. He et al. [19]
carried out an experimental and numerical study on the ow
and heat transfer behaviors of the TNT(titanate nanotubes)-water
nanouid at the Reynolds number of 1700 and the CNT-water
nanouid at the Reynolds number of 800. The convective heat
transfer of the nanouids was simulated in a single-phase ow
mode with considering the nanouids as both Newtonian and
non-Newtonian uids. It was suggested that the non-Newtonian
characteristics of the nanouids inuence on the overall heat
transfer enhancement.
Sommers and Yerkes [20] studied the convective heat transfer
and the system-level effect of the Al2O3epropanol nanouid at the
Reynolds numbers of 500e3000. The change of viscosity was found
to be nonlinear and increased sharply with particles loading. The
heat transfer performance data were generally commensurate with
the baseline data. However, it concluded that the use of the
Al2O3epropanol nanouid for the heat transfer enhancement was
difcult to be justied because the nano-abrasion may have been
occurred in the loop if the nanouid was used for a long period of
time. Alammar and Hu [21] performed the numerical analysis to
examine the laminar ow and the heat transfer characteristics of
eight single-phase nanouids. It was found that the oxide-based
nanouids offered the least heat transfer enhancement when
compared with the elements-based nanouids.

91

As mentioned above, the heat transfer characteristics of nanouids have been investigated mainly in a straight pipe with relatively high Reynolds numbers. However, most of heat exchange
systems consist of straight pipes and return bends. Therefore, it is
necessary to study the heat transfer enhancement in the pipes
including the return bend. In this context, there have been limited
works relevant to only a single curved tube with nanouids.
Akbarinia and Behzadmehr [22] investigated the effects of buoyancy force, centrifugal force, and nanoparticle volume concentration on Al2O3ewater nanouid ows in a horizontal curved tube
only at the Reynolds numbers of 300, 600, and 900 and the Grashof
numbers of 8000, 300,000, and 640,000. The buoyancy force was
mainly focused in this investigation with regard to the centrifugal
force. The results showed that the buoyancy force had a negative
effect on the Nusselt number. Akbarinia [23] continued to analyze
the nanouid ow and the heat transfer behavior along the curved
pipe at the Grashof numbers of 1000e146,600. Akbarinia and Laur
[24] reported the results of the laminar mixed convective heat
transfer in a circular curved tube by using the two phase mixture
model and the control volume technique at the Reynolds number of
648 and the Grashof number of 5740.
The objective of the present study is to analyze the hydrodynamic and thermal characteristics of the Al2O3ewater nanouid in
the pipes including a return bend with relatively low Reynolds
numbers, which are similar to the pipe shape in a heat exchanger
system. The reason why the numerical investigation is conducted at
relatively low Reynolds is that the effects of the heat transfer in the
return bend advance more noticeably in this range than those in the
inlet pipe and the outlet pipe. The FEM (nite element method) has
been used in solving the governing equations for the uid and
thermal ows. To gure out the heat transfer behaviors of the
nanouid, the average Nusselt numbers will be calculated with the
nanoparticle volume concentrations of 0.0, 2.5, 5.0, 7.5, and 10.0%
and the Reynolds numbers of 10, 25, 50, 75, and 100. The velocity
and temperature distributions will be obtained to understand the
role of the return bend, and the pressure drop will be considered to
realize the effect of the nanouid viscosity. In addition, the correlations which can be applied to the heat transfer analysis of the heat
exchanger system will be provided in the inlet pipe, the return
bend, and the outlet pipe, respectively.
2. Numerical analysis
2.1. Governing equations
The nanouid consists of Al2O3 and water, and the nanoparticles
(Al2O3) are assumed to be well dispersed within the base uid
(water). Furthermore, the nanouid can be regarded and analyzed
as a single-phase uid because the nanoparticles are ultrane and
they can be uidized easily [25]. Also, the motions of the nanoparticles can be neglected and the thermal equilibrium state can be
assumed to be predominant [16,22]. Therefore, the following
general conservation equations can be used to compute the ow
and thermal elds with the effective physical properties of the
nanouid.

vuk
0
vxk

(1)

vuj
vuj
vp
1 v2 u j


uk
vxj Re vx2i
vt
vxk

(2)

vT
vT
1 v2 T

uk
vt
vxk
Pe vx2j

(3)

92

J. Choi, Y. Zhang / International Journal of Thermal Sciences 55 (2012) 90e102

The dimensionless conservation equations are presented in

equations (1)e(3), which are continuity, momentum, and energy
equations, respectively. These equations written in the Cartesian
tensor form are for the 3-dimensional incompressible laminar ow.
The viscous dissipation and the gravity terms are not considered
because these terms make little effects on the heat transfer analysis
in the pipe at relatively low Reynolds number.

the momentum equation (2). After the Taylor series expansion, the
Galerkin method, and the Greens theorem are applied to these
terms, the discretized equation comes out as shown in equation (4).
Here, the rst term on the left-hand side is approximated by
a diagonal form and the integrations are calculated by the Gaussian
quadrature.

Nj ucj dU

2.2. Finite element method

The FEM to solve the governing equations can be classied into
the mixed interpolation methods, the penalty methods, and the
segregated velocityepressure solution methods [26]. The segregated scheme adopted in the present study requires much less
execution time and storage than the other two methods, particularly for the three-dimensional congurations. Also, the pressure is
computed only once in the stage of the pressure correction without
iterative calculations.
Unlike the basic FVM (Finite Volume Method), the FEM can be
easily applied to the arbitrary shapes with multi-blocks because it
includes the concepts of the coordinate transformation for each
element in the local coordinate system. On the other hand, the FEM
code used in the present study has a limitation that the stability
requirement is needed due to the explicit scheme. Therefore, the
time step should be decided taking the Courant number and the
diffusion number into consideration.
The numerical algorithm by FEM is shown in Fig. 1. The velocity
and the pressure are obtained from the continuity and momentum
equations through the four stages of the solution procedures. The
segregated approaches consist of convective approximation,
viscous prediction, pressure correction, and velocity correction
[27e29]. The temperature is acquired from the energy equation
through the two stages: the convective approximation and the
diffusive prediction [30]. When the temperature is calculated, the
velocity is also required in the energy equation.
2.2.1. Continuity and momentum equations
In the stage of the convective approximation, the rst term and
the second convective term on the left-hand side are considered in

Nj uoj dU  Dt

Dt2

Nj uoj

Nj uoi uok

vuoj
vxi

vuoj
vxk

dU

n k dG

(4)

In the stage of the viscous prediction, the rst term on the lefthand side and the second viscous term on the right-hand side are
used in the momentum equation (2). The discretized equation is
obtained by using the Galerkin method and the Greens theorem,
expressed in equation (5)

Z
U

Nj uvj dU

Nj ucj dU 

Dt
Re

Z
U

vNj vuj
Dt
dU
Re
vxi vxi

Z
Nj
G

vuoj
vxi

ni dG

(5)

In the stage of the pressure correction, the continuity equation is

employed to compute the pressure values. At rst, only the two
terms, which are composed of the rst term on the left-hand side
and the rst pressure term on the right-hand side in the
momentum equation (2), are considered. When the gradient of the
two terms is taken and then the continuity equation is substituted
to this equation, the pressure equation is obtained. The Galerkin
method and the Greens theorem are also applied to the pressure
equation. The discretized equation is:

Z
U

vNj vpn
1
dU 
Dt
vxj vxj

Z
Nj
U

vuvj
vxj

dU

Z
Nj
G

vpn
n dG
vxj j

(6)

On the other hand, the velocity values computed in the viscous

prediction stage do not generally satisfy the continuity equation.
Therefore, the two terms of the momentum equation used in the
pressure correction stage are again taken into account in the stage

Fig. 1. Flowchart of numerical algorithm.

J. Choi, Y. Zhang / International Journal of Thermal Sciences 55 (2012) 90e102

of the velocity correction. When the Galerkin method is applied to

this equation, the discretized equation is obtained as:

Nj unj dU

Nj uvj dU  Dt

Nj
U

vpn
dU
vxj

(7)

2.2.2. Energy equation

In the stage of the convective approximation, the two terms of
the energy equation (3), the rst term and second convective term
on the left-hand side, are used to calculate the temperature. The
discretized equation obtained by the Taylor series expansion, the
Galerkin method, and the Greens theorem is:

NT c dU

NT o dU  Dt

Dt
2

Nunk
U

Nuni unk
G

vT o
dU
vxk

vT o
n dG
vxi k

(8)

In the stage of the diffusive prediction, the rst term on the lefthand side and the thermal diffusive term on the right-hand side of
the energy equation (3) are considered to calculate the nal
temperature values. When the Taylor series expansion, the Galerkin
method, and the Greens theorem are also applied to this equation,
the discretized equation is obtained as:

NT n dU

NT c dU 

Dt
Pe

Dt
vN vT o
dU
vxj vxj
Pe

N
G

vT o
n dG
vxj j

(9)

2.3. Validation of FEM code

Before the above nite element method is applied to the
hydrodynamic and thermal analyses in the pipes including a return
bend, the numerical analysis is carried out for a straight pipe in

93

order to validate the FEM code because it has analytic solutions [31]
at steady state. Furthermore, at unsteady state, the results by the
FEM code are compared with those by the CFX commercial software [32] based on FVM. Another reason why the straight pipe in
the validation is employed is that the straight pipe and the pipe
with the return bend have the same conguration and boundary
conditions in the computational space. Fig. 2 (a) and (b) show the
geometric shapes of the straight pipe and the pipe with the return
bend, respectively. These geometric shapes are different in the
physical space. However, they have the same shape, which is obtained by conducting the coordinate transformations in order to
solve the governing equations, in the computational space as
shown in Fig. 2 (c). In addition, the same types of boundary
conditions are used for the two geometric shapes. In other words,
these calculations are carried out with the same conguration and
boundary conditions in the computational space. Only the Jacobian
of transformation is different, which is interpreted as the ratio of
the volumes in the physical space to that in the computational
space. Therefore, the validations by the FEM code in the straight
pipe are still available for the pipe with the return bend.
2.3.1. Calculation conditions
The straight pipe, which has the dimensionless diameter of 1
and the dimensionless length of 10, is considered in the numerical
analysis. The grids of the straight pipe consist of the nodal points of
5226 and the elements of 4875 for the FEM, and the nodal points of
292,800 and the elements of 301,701 for the CFX, respectively. The
ne meshes are used in the CFX for more accurate comparison,
which are approximately 60 times as much element as in the FEM.
The boundary conditions for the straight pipe are represented in
Table 1. The dimensionless velocity in the axial direction (w) and
the dimensionless temperature (T) are set up to 1 and 0 each at the
inlet, and the dimensionless pressure (p) is given 0 at the outlet.
The dimensionless velocities (u, v, w) of 0 and the uniform heat ux
(qw) of 0.5 (W/m2) are granted at the wall boundary. The numerical

Fig. 2. Geometric shapes in physical and computational spaces.

94

J. Choi, Y. Zhang / International Journal of Thermal Sciences 55 (2012) 90e102

1.0

Table 1
Boundary conditions in straight pipe.

0.9

Boundary conditions
u v 0, w 1, T 0
p0
u v w 0, qw k(vT*/vr)

0.8
0.7

analysis is performed at the Reynolds number of 14.3 and the

Prandtl number of 0.7, which result in a Peclet number of 10. The
calculation is conducted until the iteration is performed up to
10,000 times with the dimensionless time step of 0.001, which is
the calculation condition that reaches to the steady state.
2.3.2. Validation results
The dimensionless velocity and temperature proles calculated
by the FEM, the CFX, and the analytic equations are observed in
Fig. 3 and Fig. 4. These values are at the middle cross section of the
straight pipe, which is the region of the fully developed ow. The
dimensionless velocity and the dimensionless temperature predicted by the FEM almost coincide with the results from the CFX
with time. Also, the results from the FEM in the steady state are in
good agreement with the analytic solutions. Here, the analytic
solutions for the straight pipe are obtained by [31]:

(10)



 
 
Tw  T *
4 r 2 1 r 4

1
3 ro
3 ro
Tw  Tc

(11)

The Nusselt number calculated by the FEM is compared with

that by the analytic solutions as illustrated in Table 2. The Nusselt
numbers of 4.434 and 4.364 are obtained by the FEM and the
analytic solutions, respectively. An error of 1.6% is appeared in the
numerical solution and this difference is believed to have occurred
due to the grid size effect. Consequently, the numerical solutions by
the FEM are acceptable and the FEM code can be used in the
numerical analysis for the nanouid ow in the pipes including
a return bend.

1.0
0.9
0.8
0.7

w*/wc

0.6
0.5
0.4

FEM(t=0.01)
FEM(t=0.05)
FEM(t=0.10)
FEM(t=1.00)
CFX(t=0.01)
CFX(t=0.05)
CFX(t=0.10)
CFX(t=1.00)
Analytic Solution

0.3
0.2
0.1
0.0

0.1

0.6
0.5

FEM(t=0.01)
FEM(t=0.05)
FEM(t=0.10)
FEM(t=0.50)
FEM(t=10.00)
CFX(t=0.01)
CFX(t=0.05)
CFX(t=0.10)
CFX(t=0.50)
CFX(t=10.00)
Analytic Solution

0.4
0.3
0.2
0.1
0.0

0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1.0

r/r0
Fig. 4. Dimensionless temperature proles at cross section of pipe.

3. Flow and heat transfer characteristics of nanouids

 2
w*
r
1
ro
wc

0.0

(Tw -T*)/(Tw -Tc )

Inlet
Outlet
Wall

0.2

0.3

0.4

3.1. Geometric shape and calculation conditions

The geometrical conguration of the pipes with the return bend
is shown in Fig. 5. The dimensionless diameter (d) of the pipe is
given 1.0 and the lengths (l) of inlet and outlet pipes are set up to 10
times the length of the diameter. Also, the curvature radius (R) of
the return bend is 1.5 times of the diameter.
The inlet boundary conditions are the dimensionless velocities
in the axial direction, w 1, and the dimensionless temperature,
T 0, as illustrated in Fig. 6. The dimensionless velocities,
u v w 0, and the uniform heat ux, qw 10,000 (W/m2), are
assigned at the wall boundary of the pipe. Also, the reference
pressure, p 0, is set up at the outlet boundary. Meanwhile, the
geometric coordinate values of the pipes with the return bend are
obtained from a straight pipe by using coordinate transformations.
The relatively ne grids are used in the regions of the inlet and the
return bend. The grid points of 11,256 and the elements of 10,450
are used in the entire pipes. The numerical solutions are calculated
until the iteration is conducted up to 50,000 times with the
dimensionless time step of 0.002 so as to compare with the
numerical results of the same computational conditions in the
steady state. Here, the dimensionless time step is decided by the
convergence conditions of the Courant number and the diffusion
number [33].
The physical properties of water and Al2O3 are shown in Table 3,
and the effective physical properties of the nanouids are obtained
from the physical properties of the base uid (water) and the
nanoparticles (Al2O3), respectively. The Al2O3ewater nanouid is
expected to have the enhanced heat transfer characteristics since
the thermal conductivity of Al2O3 is much higher than that of water.

Table 2
Nusselt numbers in numerical and analytic solutions.

0.5

0.6

0.7

0.8

0.9

r/r0
Fig. 3. Dimensionless velocity proles at cross section of pipe.

1.0

Nusselt numbers
Numerical Solution
Analytic Solution
Error

4.434
4.364
1.604 (%)

J. Choi, Y. Zhang / International Journal of Thermal Sciences 55 (2012) 90e102

95

Fig. 5. Schematic representation for cross section of pipes with return bend.

The effective physical properties of the nanouid are obtained as

functions of nanoparticle volume concentration (f) [22]:

reff 1  frf frs



(12)


meff 123f2 7:3f 1 mf

 
Cp eff

keff

"

 #
 
1  f rcp f f rcp s
1  frf frs

0

1
f
k
k

n

1k

n

1

k
s
s
f
f
B
C


@
Akf
ks n  1kf f kf  ks

(13)

(14)

(15)

The equation (13) of the effective viscosity was obtained by

performing a least-square curve tting of experimental data
available for the mixtures, and presented in the literature [15]. Also,
the equation (15) of the effective thermal conductivity was introduced by Hamilton and Crosser [34], and was shown to be appropriate for the nanouids in the literature [35]. Here, the
concentration of 0% means the pure water in which the nanoparticles are not included within. The values of effective properties
used in this work are tabulated in Table 4.

3.2. Results and discussions

The numerical analysis for the hydrodynamic and thermal ows
of nanouid in the pipes with the return bend has been performed
with different concentrations and Reynolds numbers. In total, the
25 cases have been simulated in the conditions of the concentrations of 0.0%, 2.5%, 5.0%, 7.5%, and 10.0% and the Reynolds numbers
of 10, 25, 50, 75, and 100, respectively. For better understanding the
thermal and uid ows in the return bend, the dimensionless
velocity and temperature distributions at the concentration of 0%
are shown in Fig. 7 and Fig. 8 when the Reynolds numbers are 10
and 100, respectively. The velocity proles at the Reynolds number
of 10 keep nearly parabolic shapes in the entire pipes. However, at
the Reynolds number of 100, the dimensionless velocity in the
outer region of the return bend appears higher than that in the
inner region. This ow can affect the heat transfer characteristics in
the return bend and in the outlet pipe. Also, the heat transfer is
expected to be improved between the inside and the outside of the
pipe because the cold ow in the core of the inlet pipe shifts to the
outer region of the return bend as shown in Fig. 8.
The dimensionless velocity proles for the minus x-direction at
the cross section (AeA) are represented in Fig. 9. The position of the
maximum velocity is moved towards the outer region of the return
bend as the Reynolds number increases. The dimensionless
temperature at the wall of the return bend is higher than that at the
center as shown in Fig. 10 because the heat ux is uniformly
supplied at the wall boundary. However, as the Reynolds number
increases, the dimensionless temperature declines gradually in the
outer region of the return bend due to the cold ow from the inlet
pipe.
The secondary ow occurs only in the return bend because of
the centrifugal force which acts in the vertical direction of the main
ow as shown in Fig. 11. In other words, the two rotating ows at
the top and the bottom of the return bend occur due to the different
centrifugal forces between in the inner region and in the outer
regions of the return bend. Therefore, the secondary ow can be
dened as the rotating ow in the circumferential direction. It can
be seen in detail that the cold ow moves to the outer region with
the increase of the Reynolds number. The dimensionless temperature distribution at the Reynolds number of 10 is different from

Table 3
Physical properties of water and Al2O3.

Fig. 6. Boundary conditions for pipes with return bend.

Water
Al2O3

Density
(kg/m3)

Viscosity
(Ns/m2)

Specic heat
(J/kgK)

Conductivity
(W/mK)

Shape
factor (n)

998.0
3960.0

1.0020.E-03
e

4182.0
753.0

0.5984
46.0

e
3

96

J. Choi, Y. Zhang / International Journal of Thermal Sciences 55 (2012) 90e102

Table 4
Variations of effective physical properties with concentration.
Concentration Density Viscosity
(%)
(kg/m3) (Ns/m2)

Specic
Conductivity Increasing
heat (J/kgK) (W/mK)
rates of
conductivity
(%)

0.0
2.5
5.0
7.5
10.0

4182.0
3865.3
3589.6
3347.3
3132.8

998.0
1072.1
1146.1
1220.2
1294.2

1.0020.E-03
1.2619.E-03
1.6758.E-03
2.2439.E-03
2.9659.E-03

0.59840
0.64264
0.68911
0.73799
0.78947

100.0
107.4
115.2
123.3
131.9

the distributions at the other Reynolds numbers. More heat is

supplied into the outer region because the curved length in the
outer region is relatively longer than the length in the inner region.
Thus, the dimensionless temperature in the outer region at the
Reynolds number of 10 appears higher than that in the inner
region. However, at the other Reynolds numbers, the dimensionless
temperature in the outer region is lower than that in the inner
region because the intensity of the secondary ow increased
gradually. The velocity ratios of the secondary ow to the main ow
rise with the increase of the Reynolds number as represented in
Fig. 12.
To investigate the role of the return bend, the entire pipes are
divided into three parts: the inlet pipe, the return bend, and the
outlet pipe. The average Nusselt numbers at the concentration of 0%
are obtained with the increase of the Reynolds number as shown in
Fig. 13. Although the slopes appear different, the average Nusselt
numbers rise in all parts because the inlet velocity increases with
increasing Reynolds number. At the Reynolds numbers of 10 and
25, the average Nusselt numbers of the inlet pipe are relatively
higher than those of the return bend and the outlet pipe. However,
at the Reynolds numbers of 50, 75, and 100, the average Nusselt
numbers of the return bend are larger than those of the inlet and
outlet pipes because the cold ow in the inlet pipe moves to the
outer region of the return bend as shown in Figs. 7 and 8. In other
words, the heat transfer in the return bend improves with the
increase of the Reynolds number due to the secondary ow. On the
other hand, the pressure drops are compared in each of three parts

as shown in Fig. 14. The pressure drops in the return bend are lower
than those in the inlet and outlet pipes because the pressure drop
depends mainly on the pipe length. Here, the length ratio of the
return bend to the straight pipe is about 0.47. The effect of the
secondary ow is relatively small on the pressure loss. The pressure
drops in the inlet pipe are higher than those in the outlet pipe. It is
because the back pressure at the end of the inlet pipe is larger than
that at the end of the outlet pipe due to the return bend. As a result,
the return bend plays a major role in enhancing the heat transfer
without the large pressure loss.
Fig. 15 shows the average Nusselt numbers at different Reynolds
numbers of 10, 25, 50, 75, and 100 and concentrations of 0.0%, 2.5%,
5.0%, 7.5%, and 10.0%, respectively. The average Nusselt numbers
increase with the rise of the Reynolds number and the concentration. In general, the heat transfer of the nanouid is improved with
the augmentation of the concentration because the thermal
conductivity of the nanouid grows with the concentration as
shown in Table 4. In addition, the increment of the concentration is
accompanied by the rise of the inlet velocity at the same Reynolds
number as expressed in Table 5. Therefore, these conditions cause
the increase of the average Nusselt number. It is noted that the
results at the concentrations of 7.5% and 10.0% might not be correct
in reality because the stable nanouid could not be found for such
high volume concentrations.
For the entire pipes, the increasing rates of the average Nusselt
number with the effect of the Reynolds number and the concentration are illustrated in Fig. 16. These increments mean the ratios of
the average Nusselt numbers at the nanouids to the average
Nusselt number at the pure water of the concentration of 0.0%. Of
course, the average Nusselt numbers continue to increase at the
entire pipes with increasing Reynolds number as shown in Fig. 13.
Also, at the inlet pipe, the increasing rates continue to be on the rise
with increasing Reynolds number as represented in Fig. 17.
However, the slope of the average Nusselt numbers between the
Reynolds numbers of 25 and 50 is steeper than the slopes at the
other Reynolds numbers as shown in Fig. 13 because the average
Nusselt number increases noticeably in the return bend at the
Reynolds number of 50. Although the average Nusselt numbers
have high values in the return bend at the Reynolds numbers of 75
and 100, they have relatively low values in the inlet and outlet

Fig. 7. Distributions of dimensionless velocity and dimensionless temperature at Re 10.

J. Choi, Y. Zhang / International Journal of Thermal Sciences 55 (2012) 90e102

97

Fig. 8. Distributions of dimensionless velocity and dimensionless temperature at Re 100.

pressure drop at the concentration of 10.0% with respect to the

concentration of 0.0% appears nearly 6 times higher at all the
Reynolds numbers. This value is considerably high in comparison
with the increasing rate of the heat transfer. Consequently, the high
pressure drop for the nanouid should be given careful consideration in designing heat exchangers although the nanouid does not
have the characteristics such as poor suspension stability, channel
clogging, system abrasion and so on.
In order to exclude the inuence of the increasing inlet velocity
caused by the concentration, the numerical analysis has been
conducted at the same volumetric ow rate in the pipe. That is, the
average Nusselt numbers are obtained in the condition of the same
dimensionless inlet velocity at the concentrations of 0.0% and 10.0%
as expressed in Fig. 19. The results show that the average Nusselt
numbers at the concentration of 0.0% are higher than those at the

2.0

10.0

1.8

9.0

1.6

8.0

1.4

7.0

1.2

6.0

1.0

5.0

-u

pipes. Therefore, the increasing rate of the average Nusselt number

at the Reynolds of 50 is higher than the increasing rates at the other
Reynolds numbers. Those rates decrease gradually at the Reynolds
numbers of 75 and 100. As a result, it is noted that the increasing
rate of the average Nusselt number is affected by the return bend.
On the other hand, the highest increment, approximately 112%,
appears at the Reynolds number of 50 and the concentration of
10.0%. However, this value does not reach up to the increasing rate,
115.2%, for the thermal conductivity of the nanouid as represented
in Table 4.
The pressure drop between inlet and outlet of the pipe is
depicted in Fig. 18. Here, the dimensional pressure is used because
the dimensionless pressure has the same value at the same Reynolds number. The pressure drop increases dramatically with the
increment of the Reynolds number. The increasing rate of the

4.0

0.8

Re=10
Re=25
Re=50
Re=75
Re=100

0.6
0.4
0.2
0.0

Re=10
Re=25
Re=50
Re=75
Re=100

0.0

0.1

0.2

0.3

0.4

0.5

0.6

3.0
2.0
1.0
0.7

0.8

0.9

d
Fig. 9. Dimensionless velocity proles at various Reynolds numbers.

1.0

0.0

0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

d
Fig. 10. Dimensionless temperature proles at various Reynolds numbers.

1.0

98

J. Choi, Y. Zhang / International Journal of Thermal Sciences 55 (2012) 90e102

Fig. 11. Distributions of secondary ow and dimensionless temperature at cross section (AeA) of return bend.

concentration of 10.0%. This means that the increment of the inlet

velocity can improve the heat transfer in the pipe without using the
nanouids. Meanwhile, the specic heat of the nanouid decreases
with the increase of the concentration as shown in Table 4. For the
purpose of investigating the effect of the specic heat, the additional numerical calculation is carried out with the increasing
specic heat of the nanouid. To achieve this, the specic heat,
3132.8 (J/kgK), of the nanouid at the concentration of 10.0% is
increased up to 4182.0 (J/kgK), which is the value of the specic
heat at the concentration of 0.0%. The results show that the average

Nusselt numbers increase with the specic heat as represented

with the symbol > in Fig. 19. It means that the specic heat can
play an important role in the heat transfer enhancement of the
nanouid. The pressure drop with respect to the dimensionless
inlet velocity is illustrated in Fig. 20. The pressure drop at the
concentration of 10.0% is larger than that at the concentration of
0.0% because of the kinematic viscosity of the nanouid.
In the meantime, the convective heat transfer coefcient is
generally required when the temperature distribution is calculated
to design heat exchangers. If the average Nusselt number is offered
11.0

0.24

10.0

0.22

9.0

0.20

Inlet pipe
Return bend
Outlet pipe
Entire pipes

8.0

Nu

Velocity ratios

0.26

0.18

7.0

0.16
6.0

0.14
5.0

0.12
0.10
10

4.0

20

30

40

50

Re

60

70

80

90

Fig. 12. Velocity ratios of secondary ow to main ow in return bend.

100

10

20

30

40

50

60

70

80

90

Re
Fig. 13. Comparisons of average Nusselt numbers in various parts.

100

J. Choi, Y. Zhang / International Journal of Thermal Sciences 55 (2012) 90e102

0.4

114

Inlet pipe
Return bend
Outlet pipe

0.0%
2.5%
5.0%
7.5%
10.0%

112

0.3

110

Nunf / Nuf (%)

Pressure drop(Pa)

99

0.2

108
106
104

0.1
102

100

10

25

50

75

Re

100

20

30

40

50

60

70

80

90

100

Re
Fig. 16. Average Nusselt number ratios in entire pipes.

Fig. 14. Comparisons of pressure drops in various parts.

with the equation form based on the mean temperature instead of

the center temperature in the pipe system, the convective heat
transfer coefcient can be easily applied to the numerical analysis.
The correlation equations below are obtained by the multiple linear
regression analysis as functions of the Dean number and the
Prandtl number:

10.0
0.0%
2.5%
5.0%
7.5%
10.0%

9.0

10

Nui 2:2999 De0:1906 Pr 0:1950 inlet pipe

(16)

Nub 1:4039 De1:7533 Pr 1:9463 return bend

(17)

Nuo 1:6519 De0:2760 Pr 0:2073 outlet pipe

(18)

Nu

8.0

7.0

6.0

5.0

4.0

114
0.0%
2.5%
5.0%
7.5%
10.0%

112
0

10

20

30

40

50

60

70

80

90

100

Re

Table 5
Variations of inlet velocities (m/s) with Reynolds number and concentration.

7.5%

10.0%

10
25
50
75
100

1.8390.E-03
4.5975.E-03
9.1950.E-03
1.3792.E-02
1.8390.E-02

2.2917.E-03
5.7293.E-03
1.1459.E-02
1.7188.E-02
2.2917.E-02

1.1771.E-03
2.9427.E-03
5.8854.E-03
8.8281.E-03
1.1771.E-02

1.4622.E-03
3.6555.E-03
7.3111.E-03
1.0967.E-02
1.4622.E-02

108
106
104
102

Reynolds Nanoparticle volume concentration

number
0.0%
2.5%
5.0%
1.0040.E-03
2.5100.E-03
5.0200.E-03
7.5301.E-03
1.0040.E-02

Nunf / Nuf (%)

110

Fig. 15. Dependence average Nusselt number on concentration at various Reynolds

numbers.

100

10

20

30

40

50

60

70

80

Re
Fig. 17. Average Nusselt number ratios in inlet pipe.

90

100

100

J. Choi, Y. Zhang / International Journal of Thermal Sciences 55 (2012) 90e102

7.0

7.0
0.0%
2.5%
5.0%
7.5%
10.0%

6.0

6.0
5.0

Pressure drop(Pa)

Pressure drop(Pa)

5.0
4.0
3.0

4.0
3.0

2.0

2.0

1.0

1.0

0.0

10

20

30

0.0%
10.0%

40

50

60

70

80

90

0.0

100

0.0

0.1

0.2

Re

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1.0

Dimensionless inlet velocity

Fig. 18. Dependence of pressure drop on concentration at various Reynolds numbers.

Fig. 20. Comparison of pressure drops at concentrations of 0.0% and 10.0%.

4. Conclusions

Nut 1:8330 De0:2675 Pr 0:1900 entire pipes

(19)

Here, the distinctions by concentrations are not necessary in the

above functions because the Nusselt number, the Dean number,
and the Prandtl number are decided by the concentration of the
nanouid with the effective physical properties, regardless of the
nanouid. The adjusted R squares at the condence level of 95% in
equations (16)e(19), which mean the coefcients of determination,
are 0.981, 0.952, 0.960, and 0.968, respectively. The standard
deviations are 0.136, 0.509, 0.273, and 0.246. These equations are
effective in the range of the Dean numbers of 6e58 and the Prandtl
numbers of 7e12.

12.0
0.0%(Cp=4182.0)
10.0%(Cp=3132.8)
10.0%(Cp=4182.0)

11.0
10.0

Nu

9.0
8.0

The hydrodynamic and thermal analyses have been performed

on the Al2O3ewater nanouid in the pipes with return bend. The
results show that the average Nusselt numbers increase with
increasing Reynolds number and the nanoparticle volume
concentration. Especially, the heat transfer enhancement in the
return bend appears larger than that in the inlet and outlet pipes
due to the effect of the secondary ow. However, the concentration
increment of the nanouid is accompanied by the high pressure
drop in the pipe. Also, the increasing rate of the average Nusselt
number is less than that of the thermal conductivity of the nanouid with the augmentation of the concentration.
Under the same volumetric ow rate, the average Nusselt
numbers of the pure water are greater than those of the nanouid
with the concentration of 10%. Also, the increasing rate of the
pressure drop in the nanouid appears larger than the rate in the
pure water. Meanwhile, the heat transfer characteristics of the
nanouid are improved with the increase of the specic heat.
The empirical correlation equations for the average Nusselt
number are obtained as functions of the Dean number and the
Prandtl number in the inlet pipe, the return bend, the outlet pipe,
and the entire pipes, respectively. These equations will be expected
to be useful in the temperature analysis for designing heat
exchangers. In the future study, the hydrodynamic and thermal
analyses will be conducted with the consideration of the various
parameters of the nanouid and also the nanoparticles will be
simulated from the viewpoint of the molecular dynamics without
using the effective physical properties.

7.0

Acknowledgments
6.0

This work was supported in part by Research Foundation of

Engineering College, Sunchon National University.

5.0
4.0

Nomenclature
0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1.0

Dimensionless inlet velocity

Fig. 19. Dependence of average Nusselt number on concentration and specic heat at
various dimensionless inlet velocities.

Cp
d
d*
De

specic heat, J/(kgK)

dimensionless diameter of pipe
diameter of pipe, m
Dean number (Re(d/2R)1/2)

J. Choi, Y. Zhang / International Journal of Thermal Sciences 55 (2012) 90e102

hm
k
l
n
ni,nj,nk
N
Nj
Nu
p
p*
Pe
Pr
qw
r
ro
R
Re
t
t*
Dt
T
T*
Ti*
Tc
Tw
u,v,w
ui,uj,uk
w*
wc
x,y,z
xi,xj,xk

average convective heat transfer coefcient using mean

temperature, W/(m2 K)
thermal conductivity, W/(mK)
dimensionless length of inlet pipe or outlet pipe
shape factor of nanoparticle (3 for spherical
nanoparticle)
unit vectors in Cartesian tensor form (i,j,k 1, 2, 3)
interpolation function
interpolation functions in Cartesian tensor form
(j 1, 2, 3)
average Nusselt number hm d* =k
dimensionless pressure (p*/rw*)
pressure, Pa
Peclet number (Re,Pr)
Prandtl number (Cpm/k)
heat ux at wall, W/m2
coordinate in radial direction, m
radius of pipe, m
dimensionless curvature radius of the return bend
Reynolds number (rw*d*/m)
dimensionless time (t*w*/d*)
time, s
dimensionless time step
dimensionless temperature T *  Ti* =Ti*  273:15
temperature, K
temperature at inlet, K
temperature at center, K
temperature at wall, K
dimensionless velocities in x,y,z directions (velocity/w*)
dimensionless velocities in Cartesian tensor form
(velocity/w*;i,j,k 1,2,3)
axial velocity, m/s
axial velocity at center, m/s
dimensionless coordinates (coordinate/d*)
dimensionless coordinates in Cartesian tensor form
(coordinate/d*;i,j,k 1,2,3)

Greek symbols
G
boundary of the domain U
m
dynamic viscosity, (Ns)/m2
x,h,z
coordinates in computational space
r
density, kg/m3
f
nanoparticle volume concentration, %
U
bounded domain

Superscripts
c
refers to result from convective approximation step
n
refers to result from pressure correction step and velocity
correction step
o
refers to result from previous time step
v
refers to result from viscous prediction step

Subscripts
b
refers
eff
refers
f
refers
i
refers
nf
refers
o
refers
s
refers
t
refers

to
to
to
to
to
to
to
to

return bend
effective physical property
base uid (water)
inlet pipe
nanouid
outlet pipe
nanoparticle
entire pipes

101

References
[1] A. Bejan, S. Lorente, Thermodynamic optimization of ow geometry in
mechanical and civil engineering, Journal of Non-Equilibrium Thermodynamics 26 (2001) 305e354.
[2] Z. Li, S.C. Mantell, J.H. Davidson, Mechanical analysis of streamlined tubes with
non-uniform wall thickness for heat exchangers, The Journal of Strain Analysis
for Engineering Design 40 (3) (2005) 275e285.
[3] H. Naja, B. Naja, Multi-objective optimization of a plate and frame heat
exchanger via genetic algorithm, Heat Mass Transfer 46 (2010) 639e647.
[4] S. Lee, S.U.S. Choi, Application of Metallic Nanoparticle Suspensions in
Advanced Cooling Systems, Proceedings of International Mechanical Engineering Congress and Exhibition, Atlanta, Georgia, USA, 1996.
[5] C.W. Sohn, M.M. Chen, Microconvective thermal conductivity in disperse twophase mixtures as observed in a low velocity Couette ow experiment, Journal
of Heat Transfer 103 (1981) 47e51.
[6] M.C. Roco, C.A. Shook, Modeling of slurry ow: the effect of particle size, The
Canadian Journal of Chemical Engineering 61 (4) (1983) 494e503.
[7] K. V. Liu, U. S. Choi, K. E. Kasza, Measurements of pressure drop and heat
transfer in turbulent pipe ows of particulate slurries, Argonne National
Laboratory Report, ANL-88e15, 1988.
[8] S.U.S. Choi, Z.G. Zhang, W. Yu, F.E. Lockwood, E.A. Grulke, Anomalously
thermal conductivity enhancement in nanotube suspensions, Applied Physics
Letters 79 (14) (2001) 2252e2254.
[9] X.Q. Wang, A.S. Mujumdar, Heat transfer characteristics of nanouids:
a review, International Journal of Thermal Sciences 46 (2007) 1e9.
[10] V.I. Terekhov, S.V. Kalinina, V.V. Lemanov, The mechanism of heat transfer in
nanouids: state of the art(review). Part 1. Synthesis and properties of
nanouids, Thermophysics and Aeromechanics 17 (1) (2010) 1e13.
[11] W. Yu, D.M. France, J.L. Routbort, S.U.S. Choi, Review and comparison of
nanouid thermal conductivity and heat transfer enhancements, Heat
Transfer Engineering 29 (5) (2008) 432e460.
[12] B.C. Pak, Y.I. Cho, Hydrodynamic and heat transfer study of dispersed uids
with submicron metallic oxide particles, Experimental Heat Transfer 11 (2)
(1998) 151e170.
[13] Y. Xuan, Q. Li, Investigation on convective heat transfer and ow features of
nanouids, Journal of Heat Transfer 125 (2003) 151e155.
[14] D. Wen, Y. Ding, Experimental investigation into convective heat transfer of
nanouids at the entrance region under laminar ow conditions, International Journal of Heat and Mass Transfer 47 (2004) 5181e5188.
[15] S.E.B. Maga, C.T. Nguyen, N. Galanis, G. Roy, Heat transfer behaviours of
nanouids in a uniformly heated tube, Superlattices and Microstructures 35
(2004) 543e557.
[16] S.E.B. Maga, S.J. Palm, C.T. Nguyen, G. Roy, N. Galanis, Heat transfer
enhancement by using nanouids in forced convection ows, International
Journal of Heat and Fluid Flow 26 (2005) 530e546.
[17] Y. Ding, H. Alias, D. Wen, R.A. Williams, Heat transfer of aqueous suspensions
of carbon nanotubes(CNT nanouids), International Journal of Heat and Mass
Transfer 49 (2006) 240e250.
[18] B.H. Chun, H.U. Kang, S.H. Kim, Effect of alumina nanoparticles in the uid on
heat transfer in double-pipe heat exchanger system, Korean Journal of
Chemical Engineering 25 (5) (2008) 966e971.
[19] Y. He, Y. Men, X. Liu, H. Lu, H. Chen, Y. Ding, Study on forced convective heat
transfer of non-newtonian nanouids, Journal of Thermal Science 18 (1)
(2009) 20e26.
[20] A.D. Sommers, K.L. Yerkes, Experimental investigation into the convective
heat transfer and system-level effects of Al2O3epropanol nanouid, Journal of
Nanoparticle Research 12 (2010) 1003e1014.
[21] K. Alammar, L. Hu, Laminar ow and heat transfer characteristics of nanoparticle colloidal dispersions in water, Heat Mass Transfer 46 (2010) 541e546.
[22] A. Akbarinia, A. Behzadmehr, Numerical study of laminar mixed convection of
a nanouid in horizontal curved tubes, Applied Thermal Engineering 27
(2007) 1327e1337.
[23] A. Akbarinia, Impacts of nanouid ow on skin friction factor and Nusselt
number in curved tubes with constant mass ow, International Journal of
Heat and Fluid Flow 29 (2008) 229e241.
[24] A. Akbarinia, R. Laur, Investigating the diameter of solid particles effects on
a laminar nanouid ow in a curved tube using a two phase approach,
International Journal of Heat and Fluid Flow 30 (2009) 706e714.
[25] Y. Xuan, Q. Li, Heat transfer enhancement of nanouids, International Journal
of Heat and Fluid Flow 21 (2000) 58e64.
[26] J.E. Akin, Finite Elements for Analysis and Design, Academic Press, San Diego,
USA, 1994.
[27] B. Ramaswamy, Finite element solution for advection and natural convection
ows, Computers & Fluids 16 (4) (1988) 349e388.
[28] B. Ramaswamy, Efcient nite element method for two-dimensional uid
ow and heat transfer problems, Numerical Heat Transfer: Part B 17 (1990)
123e154.
[29] B. Ramaswamy, T.C. Jue, J.E. Akin, Semi-implicit and explicit nite element
schemes for coupled uid/thermal problems, International Journal for
Numerical Methods in Engineering 34 (1992) 675e696.
[30] K.S. Cha, J.W. Choi, C.G. Park, Finite element analysis of uid ows with
moving boundary, Journal of Mechanical Science and Technology 16 (5)
(2002) 683e695.

102

J. Choi, Y. Zhang / International Journal of Thermal Sciences 55 (2012) 90e102

[31] P. Oosthuizen, D. Naylor, Introduction to Convective Heat Transfer Analysis,

WCB/McGraw-Hill, New York, USA, 1999.
[32] ANSYS CFX, Users Guide(ANSYS CFX Release 12.1), ANSYS, Inc., PA, USA, 2009.
[33] K.A. Hoffmann, S.T. Chiang, Computational Fluid Dynamics for Engineers, Vol.
I, Engineering Education System, Austin, Texas, USA, 1993.

[34] R.L. Hamilton, O.K. Crosser, Thermal conductivity of heterogeneous twocomponent system, I&EC Fundamentals 1 (1962) 187e191.
[35] Y. Xuan, W. Roetzel, Conceptions for heat transfer correlations of nanouids, International Journal of Heat and Mass Transfer 43 (19) (2000)
3701e3707.