International Journal of Thermal Sciences 108 (2016) 108e122

Contents lists available at ScienceDirect

International Journal of Thermal Sciences

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

A new heat transfer correlation for transition and turbulent uid ow

in tubes
Dawid Taler
Cracow University of Technology, Faculty of Environmental Engineering, ul. Warszawska 24, 31-145 Cracow, Poland

a r t i c l e i n f o

a b s t r a c t

Article history:
Received 5 October 2015
Received in revised form
7 March 2016
Accepted 30 April 2016
Available online 14 May 2016

The objective of the paper is to develop a correlation for the Nusselt number Nu in terms of the friction
factor x (Re), Reynolds number Re, and also Prandtl number Pr, which is valid for transitional and fully
developed turbulent ow. After solving the equations of energy conservation for turbulent ow in a
circular tube subject to a uniform heat ux, the Nusselt number values were calculated for different
values of Reynolds and Prandtl numbers. Then, the form of the correlation Nu f (Re, Pr) was selected
which approximates the results obtained in the following ranges of Reynolds and Prandtl numbers:
2300  Re  106, 0.1  Pr  1000. The form of the correlation was selected in such a way that for the
Reynolds number equal to Re 2300, i.e. at the point of transition from laminar to transitional ow the
Nusselt number should change continuously. Unknown coefcients x1,,xn appearing in the heat
transfer correlation expressing Nusselt number as a function of the Reynolds number and Prandtl
number were determined by the method of least squares. To determine the values of the coefcients at
which the sum of the difference squares is a minimum, the LevenbergeMarquardt method is used. The
proposed correlation was validated by comparing with experimental data.
2016 Elsevier Masson SAS. All rights reserved.

Keywords:
Transition and turbulent tube ow
Energy conservation equation
Heat transfer correlation for transition
region
Experimental results

1. Introduction
Heat transfer correlations based on the experimental data are
widely used to calculate mean heat transfer coefcients in heat
exchangers and other thermal installations. A common way to nd
these correlations involves performing heat transfer measurements
and correlating the data in terms of appropriate dimensionless
numbers, which are obtained by expressing mass, momentum, and
energy conservation equations in dimensional forms or from the
dimensional analysis. A functional form of the relation Nu f(Re,Pr)
is usually based on energy and momentum-transfer analogies.
Traditional expressions for calculation of heat transfer coefcient in
fully developed ow in smooth tubes are usually products of two
power functions of the Reynolds and Prandtl numbers.
Until recently, the Dittus-Boelter correlation for turbulent ow
in tubes has been widely used [1e3]. The Dittus-Boelter relationship, as introduced by McAdams [2,3], is

Nu 0:023Re0:8 Pr n ;

0:7  Pr  120;

Re  104 ;

L=dw  60
(1)

The exponent of the Prandtl number is n 0.4 for heating of the

uid and n 0.3 if the uid is being cooled. The Dittus-Boelter
correlation has a real physical basis since a similar equation can
be obtained using the Chilton-Colburn analogy [4,5],

(2)

where the Colburn factor j is dened as

Nu
RePr 1=3

(3)

Substituting the Moody equation for the friction factor in

smooth tubes [6],

0:184
;
Re0:2

Re  104

(4)

into Eq. (2) gives the relation proposed by Colburn

Nu 0:023Re0:8 Pr 1=3 ;

0:7  Pr  160;

Re  104 ;

L=dw

 60
E-mail address: dtaler@pk.edu.pl.
http://dx.doi.org/10.1016/j.ijthermalsci.2016.04.022
1290-0729/ 2016 Elsevier Masson SAS. All rights reserved.

(5)

D. Taler / International Journal of Thermal Sciences 108 (2016) 108e122

Nomenclature
cp
c1,c2
dh
dw
ei
FD
i
j
k
k
L
n
Nu
Num,q
Num,T
p
Pr
PKN
q
qm
qt
qw
r
rw
r2
r
R
Re
SE
T
T0
T 1 x
T 2 r

specic heat at constant pressure, J/(kg K)

constants
hydraulic diameter of a duct, m
inner diameter of a circular tube, dw 2rw, m
relative difference
fully developed ow
node number
Colburn parameter, j Nu/(RePr1/3)
thermal conductivity, W/(m K)
turbulence kinetic energy, N/(s m2)
tube length, m
number of nodes in the nite difference grid
Nusselt number, Nu hdw/k
mean Nusselt number over the tube length in laminar
ow for constant wall wall heat ux
mean Nusselt number over the tube length in laminar
ow for constant wall temperature
pressure, Pa
Prandtl number, Pr cpm/k

rma

n-Nikuradse
Prandtl-von Ka
heat ux, W/m2
molecular heat ux, W/m2
turbulent heat ux, W/m2
heat ux at the inner surface of the tube, W/m2
radial coordinate, m
inner radius of the tube, m
coefcient of determination
dimensionless radius, r rut/n
dimensionless radius, R r/rw
Reynolds number, Re wmdw/n
square-edged (sudden contraction) inlet to the tube
temperature,  C or K
uctuating component of the local temperature in
turbulent ow, K
time and mass averaged uid temperature,  C or K
time averaged radial component of the temperature,  C
or K

The empirical correlations of Dittus-Boelter and Colburn have

gained widespread acceptance for prediction of the Nusselt number with turbulent ow in the smooth-surface tubes. However, the
Dittus-Boelter and Colburn equations do not provide a good correlation of the experimental data because of their simple form. The
power type correlations like those of Dittus-Boelter and Colburn
are not able to approximate the experimental data over a broad
range of the Prandtl number. The maximum deviation between
experimental data and predictions using Eq. (1) or Eq. (5) is about
20% [7e9]. The smooth-tube heat transfer results reported by Allen
and Eckert [8] were obtained for developed turbulent ow of water
under the uniform wall heat ux boundary condition at Pr 7 and
Pr 8, and 1.3$104  Re  1.11$105 [8]. Nusselt numbers determined
experimentally were 10e20% higher than values predicted by the
Dittus-Boelter equation. The discrepancy between the results of
measurements and values of Nusselt number obtained from the
Dittus-Boelter formula is greater for higher Reynolds numbers. In
another experimental study, the Dittus-Boelter relationship (1)
underpredicted the data by 5e15% for turbulent water ow through
a tube at Prandtl numbers of 6.0 and 11.6 [9]. The Reynolds number
ranged between 104 and 105. On the other hand, power-type

ut
UHF
UWT
wm
w0r
wx
wx
x
y
y

109

p
friction velocity, ut tw =r, m/s
uniform heat ux
uniform wall temperature
mean velocity
uctuating component of the radial velocity in
turbulent ow, m/s
velocity component in the x direction, m/s
time averaged velocity component in the x direction,
m/s
a spatial coordinate in Cartesian or cylindrical
coordinate systems or distance from the tube inlet, m
a spatial coordinate in a Cartesian system or distance
from distance from the wall surface m
dimensionless distance from the tube wall, y yut/n

Greek symbols
Dy
dimensionless spatial step
G
Gamma function

turbulence dissipation rate, N/(s m2)

q
eddy diffusivity for heat transfer, m2/s
t
eddy diffusivity for momentum transfer (turbulent
kinematic viscosity), m2/s
m
dynamic viscosity, kg/(m s)
n
kinematic viscosity, n m/r, m2/s
x
Darcy-Weisbach friction factor
r
uid density, kg/m3
t
shear stress, Pa
tw
shear stress at wall surface, Pa
Subscripts
b
bulk
m
mean
i
node number
w
wall surface
Superscripts
e
time averaged

dimensionless

correlations are still used successfully to approximate the experimental data [10e14], when the Reynolds and Prandtl numbers vary
in a narrow range. Siddique et al. [10] studied turbulent ow and
heat transfer inside a micro-nned tube. The heat transfer data for
3.3$103  Re  2.25$104 and 2.9  Pr  4.7 were correlated using a
relationship of the Dittus-Boelter type. Zhang et al. [11] carried out
the thermo-hydraulic evaluation of the heat transfer enhancement
in the smooth tubes tted with rotor-assembled strands of various
diameters. A power type correlation for the Nusselt number was
developed with the Reynolds number ranging from
2.5$104e7.5$104 and the Prandtl number ranging from 4.188 to
4.274. Experimental [12,14] and numerical [13] studies were conducted to investigate enhancement in heat transfer by using
different nanouids. Results of an experimental study on convective heat transfer of non-Newtonian nanouids owing through a
uniformly heated tube under turbulent ow conditions are presented by Hojjat et al. [12]. They proposed a new power-type correlation for prediction of the Nusselt number of nanouids. The
Reynolds number varied from 2800e8400 and the Prandtl number
from 40e73. Moghadassi et al. [13] conducted a CFD simulation of a
laminar ow in a horizontal tube to investigate the effect of

110

D. Taler / International Journal of Thermal Sciences 108 (2016) 108e122

nanouids on forced convective heat transfer. Using the simulation

results the Nusselt number was correlated using the Colburn-type
equation. The Reynolds number ranged from 840e2340. Zamzanian et al. [14] investigated experimentally convective heat transfer
in alumina oxide/ethylene glycol and copper oxide/ethylene glycol
nanouids. They used the power type heat transfer correlations to
approximate the experimental data.
In the recent forty years, the Dittus-Boelter Equation (1) has
frequently been superseded by the Petukhov-Kirillov correlation
[15e17],
x

Nu

8 RePr
q

1:07 12:7

;
x  2=3
1
8 Pr

10  Re  10 ;

0:5  Pr

 200
(6)
where the friction factor for smooth tubes is given by the Filonienko
relationship [18],

x 1:82 logRe  1:642 ;

3$103  Re  106

3$103  Re  106

0:3164
Re0:25

(8)

(9)

The relationship (6) was derived by Petukhov and Kirillov [15]

using the Lyon integral [20] to obtain the Nusselt number as a
function of the Reynolds and Prandtl numbers. Next, they approximated the obtained results by Eq. (6). The eddy diffusivity of momentum and velocity prole in turbulent ow were calculated from
experimental expressions given by Reichhardt [21,22]. The Lyon
integral was evaluated numerically and the calculated Nusselt
numbers were tabulated for various values of the Reynolds and
Prandtl numbers.
The form of approximating function (6) proposed by Petukhov
and Kirillov is similar to the formula derived by Prandtl [23],

1C

RePr
8
q
x

8Pr

 1

Pr  0:5

1  L=dw

The power to the Reynolds number in Eq. (11) is too small to

predict the experimental data with satisfactory accuracy.
Taking inspiration from the Hausen equation, Gnielinski [26,27]
replaced the Reynolds number Re by the term (Re 1000) in the
heat transfer correlation (6) to include the transitional region. The
Petukhov correlation (6) has been modied by Gnielinski [26,27] to
the form

Nu

"

 1000Pr
8 Re q

dw
 1 L
x  2=3
1
8 Pr

1 12:7

 Re  106 ;

2=3 #

Pr
Prw

0:11
;

3$103

0:5  Pr  200
(12)

to increase the accuracy of this equation in the transition area, i.e. in

the range of Reynolds numbers: 3$103  Re  106. Gnielinski's
extension (12) that is capable of predicting the Nusselt numbers for
smaller Reynolds numbers is cited and used in almost all textbooks,
for example in [9,17]. Gnielinski [26,27] also substituted Eq. (12)
with two simpler correlations similar to the Hausen Eq. (11)

"

 2=3 #


dw
Prb 0:11
Nu 0:0214 Re0:8  100 Pr 0:4 1
;
L
Prw
0:5 < Pr  1:5;

3$103  Re  106 ;

1  L=dw
(13)

0:87

Nu 0:012 Re

"

 280 Pr

0:4


 2=3 #
dw
Prb 0:11
;
1
L
Prw

3$103  Re  106 ;

1  L=dw

(10)

This equation was derived by Prandtl using a two-region model

of the boundary layer at the wall that consists of the laminar sublayer and the turbulent core. The constant C in Eq. (10) is equal to
the dimensionless velocity at the hypothetical distance from the
tube wall that is assumed to be the boundary separating laminar
sublayer and turbulent core. The constant C depends on the
thickness of the laminar sublayer considered in the analysis and
varies from C 5 [5] for the three-zone boundary layer model of

n to C 11.7 for two-zone model of Prandtl [23]. Later,
von K

arma
Prandtl suggested that the constant C is equal to 8.7 [23].
Prandtl Formula (10) is suitable only for smaller values of the
Prandtl number Pr. For large Prandtl numbers, the ratio of Pr/
(Pr  1) tends to one causing the Nusselt number does not depend
on the Prandtl number. The Petukhov and Kirillov correlation (6)
shows better behavior for large values of the Prandtl number. If
Pr / then the ratio of Pr/(Pr2/3 e 1) tends to Pr1/3. The Nusselt
number Nu calculated from the correlation (6) depends in such a

0:6

(11)

1:5 < Pr  500;

Nu

2300  Re  105 ;

 Pr  1000;

For lower Reynolds number Re  105 also the Blasius correlation

may be used [5,7,17],

"
 2=3 # 0:14


dw
mb
Nu 0:037 Re0:75  180 Pr 0:42 1
;
L
mw

(7)

The friction factor x in the relationship (6) can also be calculated

using the Konakov formula [19],

x 1:8 logRe  1:52 ;

case on Pr1/3, similar to the Colburn correlation (5).

There are only a few heat transfer correlations for internal ows
in pipes and ducts that are valid in transition and turbulent regions.
The rst extension of usual power-type correlations to transition
ow region was proposed by Hausen in 1959 [24] based on the
experimental data of Sieder and Tate [25].
The Hausen equation has the following form

(14)
There is a discontinuity in the Nusselt number for the Prandtl
number: Pr 1.5 when calculating the Nusselt number as a function of the Prandtl number for a given Reynolds number using Eqs.
(13) and (14). Corcione et al. [28] used the Gnielinski simplied heat
transfer correlation (13) to study heat transfer of nanouids in the
turbulent pipe ow.
Yu-ting et al. [29] found a heat transfer correlation for internal
ow of molten salt that is similar to Eq. (14)

"

 2=3 #


dw
Prb 0:11
Nu 0:007 Re0:92  280 Pr 0:4 1
;
L
Prw
15:0 < Pr  18:4;

4100  Re  9850
(15)

Lu et al. [30] investigated experimentally convective heat transfer

of molten salt in an annular passage. The Reynolds number ranged
from 4000e10000 while the Prandtl number varied from 4.8e10.

D. Taler / International Journal of Thermal Sciences 108 (2016) 108e122

They found a correlation of the similar form to the Gnielinski correlation (13) that has a very good agreement with experimental
data.
A drawback of the Hausen and Gnielinski correlations is the lack
of Nusselt number continuity for the Reynolds number Re 2300,
i.e. at the point at which the ow evolves from the laminar to the
transition region. The continuity of the Nusselt number over the
wide Reynolds number range is of great practical importance since
heat exchangers, and pipelines can operate with different loads
when the ow regime changes from laminar through the transition
to turbulent.
Based on the idea of Taborek [31], Gnielinski has developed a
new calculation method [32,33] for the transitional ow based on
the linear interpolation of the Nusselt number between Re 2300
and Re 10000 taking into account the nite length of the tube.
Petukhov's Eq. (6) was used to calculate the Nusselt number at
Re 10000. In this way, continuity of the Nusselt number was
assured in the range from Re 0 to Re 1$107. It has turned out,
however, that in the range of Reynolds numbers from
Re 4000e20000 values of Nusselt numbers calculated using the
interpolation formula proposed by Gnielinski are too large. For this
reason, Gnielinski changed his method [34]. A linear interpolation
between the Nusselt numbers at Re 2300 and Re 4000, was
proposed as follows

Nu 1  gNul;2300 gNut;4000 ;

2300  Re  4000

(16)

where

Re  2300
;
4000  2300

111

(Reynolds-Averaged Navier-Stokes) equations of the mass, momentum, and energy conservation, and SST (Shear Stress Transport)
turbulence model were used to study the transition from laminar to
turbulent ow or internal ows which transit from turbulent to
laminar. The Nusselt numbers obtained from numerical experiments by Abraham et al. [41e46] were approximated by polynomials of various degrees to get explicit formulas suitable for
engineering applications. Abraham et al. [42] proposed to calculate
the Nusselt number for air ow in the pipe with the uniform wall
heat ux (UHF boundary condition) or the uniform wall temperature (UWT boundary condition) the following relationships:
 uniform wall heat ux







Re 4
Re 3
Re 2
Nu 3:5239
 45:148
212:13
1000
1000
1000


Re
316:08; Pr 0:7; 2300  Re  3100
 427:45
1000
(18)
 uniform wall temperature







Re 4
Re 3
Re 2
Nu 2:2407
 29:499
142:32
1000
1000
1000


Re
219:88; Pr 0:7; 2300  Re  3100
 292:51
1000
(19)

0g1

(17)

The Nusselt number for Re 2300 is calculated from wellknown correlations for the laminar ow and the Nusselt number
for Re 4000 is determined using the Gnielinski correlation (12).
When calculating the Nusselt number from Eq. (12), the friction
factor x is calculated using the explicit relationship of Colebrook
[35,36] modied slightly by Konakov [19].
The disadvantage of all interpolation functions proposed by
Gnielinski [32e34] is the need to specify the value of the Reynolds
number for the end of the interval, in which the ow is transitional.
Experimental studies show that in the range of Reynolds numbers:
2.3$103  Re  8$103 [31,37,54,55] the Nusselt numbers can be
much lower than those calculated from the correlations used for
turbulent ow.
Another relationship for all Reynolds numbers, including transition region, was proposed by Churchill in 1977 [37]. Churchill
implemented the method suggested by Churchill and Usagi [38] to
develop a correlation for all ow regimes: laminar, transition, and
turbulent with Reynolds numbers varying from 10 to 106. For fully
developed turbulent ow, Churchill [37] used two different correlations for the constant wall temperature and the constant wall
heat ux. According to recent studies the boundary condition at the
tube wall has no impact on the Nusselt number when the uid ow
is turbulent. Churchill correlation has not been veried experimentally by other researchers. Huber and Walter [39] compared
the experimental data for the transition region between laminar
and turbulent air ow in a vertical tube with the Churchill correlation but the comparison did not conrm the right accuracy of the
Churchill correlation. Tam and Ghajar [40] tried to use the formula
of Churchill, but they had to change it to approximate really the
experimental data.
Nusselt numbers for all ow regimes in circular tubes and
parallel-plate channels were successfully determined using CFD
(Computational Fluid Dynamics) modeling [41e46]. The RANS

In this paper, a new correlation for the Nusselt number Nu in

terms of the friction factor x, Reynolds number Re, and Prandtl
number Pr was developed, which is valid for transitional and fully
developed turbulent ow in the following ranges of Reynolds and
Prandtl numbers: 2.3$103  Re  106 and 0.1  Pr  1000.The proposed relationships to calculate the Nusselt number must know the
Nusselt number for laminar ow at Re 2300, taking into account
the type of laminar ow, i.e. whether it is a ow developed hydraulically and thermally developing or whether it is a hydraulically and thermally developing ow. New heat transfer correlations
for hydrodynamically fully developed and thermally developing
laminar ow in circular tubes were proposed. Using the correlations presented in the paper engineering calculations of the Nusselt
number can be performed for all ow regimes: laminar, transitional
and turbulent.
2. Mathematical formulation for uniform heating of the tube
wall
The time-averaged energy conservation equation for fully
developed turbulent ow in a tube is

rcp wx

vT 1 v

rq
vx
r vr

(20)

where the heat ux q contains the molecular qm and turbulent qt

component


 vT
q qm qt k rcp q
vr

(21)

The thermal eddy diffusivity q is dened as

rcp w0r T 0 rcp q

vT
vr

(22)

112

D. Taler / International Journal of Thermal Sciences 108 (2016) 108e122

where w0r and T0 are the uctuating components of the radial velocity and local temperature, respectively.
The uid temperature can be expressed as the sum of the mass
average temperature T 1 x and radial temperature T 2 r using the
superposition principle

Tx; r T 1 x T 2 r

(23)

The mass average uid temperature T 1 x is given by

T 1 x T 1

2qw

x
x0
rw rcp wm

T2*

R1

(25)

(36)

* is given by
where the mean mass average temperature T2m

(24)



1 d
q
wx
R
2
R dR
qw
wm

(35)

*
0
T2m

*
T2m

where the symbol T 1 x0 denotes the mass average temperature at
the inlet of the tube.
Taking into account Eqs. (23) and (24), Eq. (20) can be transformed to the form

*
T2w

2
u
m

Z1

T2* u RdR

* is unknown
The temperature of the inner surface of the tube T2w
and must be determined iteratively to satisfy the condition (36) i.e.
* should be equal zero. The nite differthe mean temperature T2m
ence method was used to solve the equation system (31)e(32) with
the conditions (33)e(36). The nite difference grid is dened as
follows. First, the interval 0  R  1 is divided into (n  1) subintervals having a width DR. Dimensionless radius Ri is dened as
follows

Ri i  1DR;

Rearranging Eq. (21) gives

dT 2
qrw


dR
Pr t
k 1 Pr
t n

(26)

(37)

i 1; ; n;

(38)

where the spatial step DR is

DR 1=n  1

(39)

where the Prandtl number (molecular) Pr and turbulent Prandtl

number Prt are dened as

Pr

n cp m

k
a

and

Prt

t
q

(27)

The mean uid velocity wm in the tube cross section is given by

wm

2
2
rw

Zrw
wx rdr

(28)

The dimensionless variables: heat ux q*, temperature T2* , local

velocity u, and the mean velocity u
m are introduced as follows

r
;
rw

q*

q
;
qw

T2*

T2
;
qw rw =k

wx
;
ut

u
m

wm
ut
(29)

where the symbolut denotes the friction velocity dened as

ut

p
tw =r

(30)

1 d  *
u
Rq 2
R dR
um

(31)

dT2*
q*

Pr t
dR
1 Pr
t n

(32)

The system of two ordinary differential Eqs. (31) and (32) is

subject to the following conditions

R0


 1:51 R
1
u ln 1 0:4y
k
1 2R2
!
!#
"

y
y
y

exp 
;
C 1  exp 
11
11
3

where k 0.4 and C 5.5  (1/k)lnk 7.8 are the constant. The
advantage of Eq. (40) is that it gives the velocity prole u
without division of the
throughout the entire interval 0  r  rw
tube cross-section into subdomains.
The eddy diffusivity for momentum transfer t will be calculated
using Reichardt's [21,22] empirical equations, which so far are the
most commonly used because of their satisfactory accuracy

  
y
;
k y  y
tanh
n
n
y
n


k
1
R2 ;
y 1 R
2
n 3

y  50

(41)

50 < y  rw

(42)

where the constant y

n is: yn 11.
The dimensionless variables appearing in the relationships
(40)e(42) are dened in the following way

yut

rut

(33)
(34)

0R1
(40)

Eqs. (25) and (26) can be written in the form

q* R1 1

The universal velocity u is typically expressed as a function of

the dimensionless distance from the wall y. The coordinates y
i
corresponding to the dimensionless radius Ri are calculated from
 r 1  R r , i 1,,n
the relationship: y
rw
i w
i
i
The radial velocity prole u appearing in Eq. (31) can also be
determined from the formula proposed by Reichardt that is based
on experimental data [21,22],

p
tw =r

n
r

p
tw =r
;

p
rw  r tw =r

(43)

rw

r w ut

Re
2

(44)

D. Taler / International Journal of Thermal Sciences 108 (2016) 108e122

r
rw  y
y

1
rw
rw
rw

(45)

coefcient and noting that mean uid temperature T 2m is equal

zero, the Nusselt number Nu 2hrw/k is

Nu

Eq. (31) is approximated as follows

u
u
Ri q*i  Ri1 q*i1 1
i
2Ri1 i1

2Ri
;

2
DR
um
um

i 2; ; n:

(46)

The mean velocity u

m is calculated using the trapezoidal rule

u
m

Z1
2
0



 
n R


X
i1 u yi1 Ri u yi
u RdRy2DR
;
2
i2

!
u
u
Ri1 *
DR
i1
i

Ri1 Ri ;
Ri
Ri i1
um
um

i 2; ; n;

DR

"
q*i
q*i1
1
 
 
t
2 1 Pr t
1 PrPr
n
Prt;i n
t;i1

(48)

#
;

i n; ; 2

i1

(49)
Solving Eq. (49) for
*
*
T2;i1
T2;i


DR
2

*
T2;i1

"

q*i

1 PrPrt;i

gives

 
t

n i

q*i1
 

1 PrPr
t;i1

#
;

i n; ; 2

n i1

(50)
with
*
*
T2;n
T2w

(51)

*
The wall surface temperature T2w
is adjusted iteratively to
* , one of
satisfy the condition (36). To determine the temperature T2w
the many methods used for solving a nonlinear algebraic equation
with one unknown can be applied, for example, the method of
interval searching, the bisection method or the secant method. The
* (37) is calculated numerically using
mass average temperature T2m
the trapezoidal rule

*
T2m



*
* u R
n
T2;i1
u
R
T2;i
2DR X
i1 i1
i i

2
um i2

(52)

The heat transfer coefcient h at the inner surface of the tube is

then calculated as

qw

T 2 R1  T 2m

(53)

where the symbol T 2m designates the mass average uid

temperature

T 2m

2
wm

Z1
wx T 2 RdR

2
R1

*
 T2m

2
*
T2w

(55)

Then, the results of calculation of the velocity, temperature and

heat ux distributions in the cross-section of the tube for several
different Reynolds numbers will be presented. Also, the Nusselt
number will be calculated as a function of the Reynolds and Prandtl
numbers.

3. Results of calculations

gdzie q*1 0.
In a similar manner, the differential Eq. (32) is replaced by the
difference equation
*  T*
T2;i
2;i1

T2*

(47)

where the uid velocity u y is determined using Eq. (40).

The transformation of Eq. (47) gives q*i

q*i

113

(54)

Taking into account the denition (53) of the heat transfer

Proles of the dimensionless uid velocity wx =wm, heat ux q/

* as a
qw, uid temperature T2* , and the mean uid temperature T2m
function of dimensionless radius R and Reynolds number Re are
shown in Fig. 1. The number of nodes is n 30001 that is the spatial
step is DR 1/(n  1) 1/30000. Increasing the number of nodes
does not change the results of calculations.
The analysis of the results shown in Fig. 1 indicates that the
velocity and temperature of the uid rapidly change in the vicinity
of the tube wall. It can also be seen that the heat ux is a nonlinear
function of the radius. Comparing the uid velocity distributions
shown in Fig. 1a)ec) it can be observed that the ratio of the velocity
at the pipe axis and the average velocity is becoming smaller
with increasing Reynolds number. This means that the velocity
prole in tube cross-section becomes at with increasing Reynolds
number.
Next, the Nusselt number values were calculated for different
values of Reynolds and Prandtl numbers (Table 1). The turbulent
Prandtl number Prt was assumed to be equal to unity since the
Nusselt number was calculated for molecular Prandtl numbers
greater than 0.1. Only for liquid metals when the molecular Prandtl
number Pr is small, the turbulent Prandtl number Prt is greater than
unity.
It should be emphasized that solving the energy conservation
Eq. (20) for fully developed laminar ow at a constant heat ux on
the inner surface of the tube, the constant Nusselt number equal to
4.364 is obtained. The fact that the Nusselt number for developed
laminar ow is equal to 4.364, also for the Reynolds number equal
to 2300, will be taken into account when selecting a function
approximating the data presented in Table 1.

4. New heat transfer correlation for transition and turbulent

uid ow
Subsequently, the form of the function Nu f (Re, Pr) was
selected to approximate the results listed in Table 1 obtained in the
following ranges of Reynolds and Prandtl numbers:
2300  Re  106, 0.1  Pr  1000, where the dimensionless
numbers are dened as

Nu

cp m
hdw
wm dw
; Re
; Pr
k
n
k

(56)

Assuming that the change from laminar to transition ow occurs

at Re 2300, for which the Nusselt number is: Nu 4.364, the
following heat transfer correlation for hydrodynamically and
thermally developed ow in the tube with uniform wall heat ux
can be proposed

114

D. Taler / International Journal of Thermal Sciences 108 (2016) 108e122

Nu 4:36

 2300Pr
8 Re q

x2 x3

x1


x  2=3
1
8 Pr

2:3$103  Re  106
0:1  Pr  1000
(57)

where the friction factor x is given by the Filonienko Formula (7).

The friction factor x can be calculated using another explicit formula, for example, the correlation proposed by Taler [36].

1
1:2776 logRe  0:4062:246

3$103  Re  107

(58)

It was assumed that for Re  3000 the uid ow in pipes with a

smooth inner surface is turbulent. It results from the empirical
formula proposed by Colebrook and White [35,47] and from the
Moody Chart [6] which is based on the Colebrook-White formula.
Also, the condition Nu 4.36 for Re 2300 was imposed on the
correlation (57) to ensure continuity of the Nusselt number on the
boundary between laminar and transition ow. The turbulent
contribution to the heat transfer decreases below Re 3000, and
the turbulent part of the Nusselt number is equal zero at Re 2300.
The Nusselt number Nu for the Reynolds number greater or equal to
Re 2300 is the sum of the laminar Nusselt number for hydrodynamically and thermally developed ow and the turbulent Nusselt
number representing a turbulent contribution to convective heat
transfer. Thus, the Nusselt number for the uniform wall temperature for Re  2300 can be written as
x

Nu 3:66

 2300Pr
8 Re q

x2 x3

x1


x  2=3
1
8 Pr

2:3$106  Re  106
0:1  Pr  1000
(59)
T

Unknown coefcients x(x1,,xm) appearing in the approximating function (57) were determined using the least squares
method

Sx

nRe X
nPr 
X

c
Num
ij  Nuij

2

min

(60)

i1 j1
m
where:
Num
and
ij Nu Rei ; Prj
Nucij NuRei ; Prj ; i 1; ; nRe ; j 1; ; nPr
e
given
and
approximated values of the Nusselt number, respectively, nRe 10
and nPr 16 e the number of the Reynolds and Prandtl numbers
which were taken into consideration in the sum (60).
As the Nusselt numbers Num
ij , the Nusselt numbers displayed in
Table 1 were adopted. The values of Nucij NuRei ; Prj were
calculated using Eq. (57).
The coefcients obtained by the least squares method are:

x1 1:0080:0050;

Fig. 1. Dimensionless distributions of uid velocity wx =wm , heat ux q/qw, uid tem* for Pr 3 as a function of dimensionless
perature T2* , and mean uid temperature T2m
radius R and Reynolds number Re; a) Re 10000, b) Re 50000, c) Re 100000.

x2 1:080:0089;

x3 12:390:0080

The mean square error of the t is st 34.78 and the coefcient

of determination is equal to r2 0.9999. The numbers with
the sign on the right sides the determined coefcients represent
half of the 95% condence intervals. (see Fig. 2)
The correlation (57) can be generalized to account for the nite
length L of the channel and temperature dependent thermal
properties of the uid. The average Nusselt number Nu for the pipe
with length L and constant wall heat ux is given by

D. Taler / International Journal of Thermal Sciences 108 (2016) 108e122

115

Table 1
Nusselt number Nu as a function of the Reynolds number Re and Prandtl number Pr.
Pr

0.1
0.2
0.5
0.71
1
3
5
7.5
10
12.5
15
30
50
100
200
1000

Re
3$103

5$103

7.5$103

104

3$104

5$104

7.5$104

105

3$105

106

7.86
9.41
12.65
14.32
16.22
24.43
29.56
34.34
38.16
41.40
44.23
56.73
67.98
86.64
110.11
190.70

9.42
11.86
16.96
19.60
22.61
35.57
43.64
51.15
57.14
62.20
66.62
86.09
103.56
132.44
168.69
292.75

11.13
14.58
21.81
25.57
29.86
48.39
59.94
70.68
79.24
86.47
92.78
120.53
145.36
186.35
237.71
413.17

12.69
17.08
26.31
31.12
36.61
60.47
75.36
89.21
100.25
109.56
117.70
153.40
185.32
237.96
303.83
528.61

22.76
33.53
56.55
68.78
82.91
145.31
184.68
221.38
250.64
275.31
296.85
391.25
475.39
613.70
786.27
1373.01

31.02
47.23
82.31
101.16
123.04
220.65
282.64
340.53
386.73
425.71
459.72
608.73
741.46
959.44
1231.01
2153.24

40.20
62.59
111.63
138.18
169.15
308.45
397.43
480.68
547.14
603.25
652.22
866.75
1057.70
1371.11
1761.37
3085.03

48.62
76.78
138.97
172.84
212.50
391.85
506.89
614.66
700.78
773.51
836.98
1115.04
1362.53
1768.51
2273.65
3986.35

104.68
172.57
327.20
413.35
515.41
987.68
1295.71
1586.18
1818.99
2015.88
2187.81
2941.80
3612.61
4711.28
6076.39
10692.06

256.61
437.11
860.91
1102.29
1391.88
2764.63
3677.45
4544.73
5242.89
5834.61
6352.24
8624.88
10647.89
13959.35
18067.10
31968.41

ow using the proposed relationships the knowledge of the Nusselt

number for laminar ow for Re 2300 is needed. Correlations are
required for UHF and UWT boundary conditions, taking into account that the ow can be hydraulically developed and thermally
developing or the ow may be developing both hydraulically and
thermally.

 2300Pr 1:008
q
Nu Num;q Re 2300

1:08 12:39 8x Pr 2=3  1
"
 2=3 #

dw
Pr 0:11
 1
; 2300  Re  106 ;
Prw
L
8 Re

0:1  Pr  1000;

dw
1
L
(61)

where the symbol Num,q denotes the Nusselt number for the
laminar ow.
If the temperature of the inner surface of the tube is uniform,
then the Nusselt number Num,q(Re 2300) for the constant wall
heat ux should be replaced by Num,T(Re 2300). The forms of the
expressions for calculating the Nusselt number Num,q(Re 2300)
and Num,T(Re 2300) are different depending on whether the ow
is hydrodynamically fully developed and thermally developing or
the ow is both hydrodynamically and thermally developing. The
multiplier (Pr/Prw)0.11 is applicable to liquids. For gases the
correction factor is less, for example, for air is equal to (T/Tw)0.45 [32]
or to(T/Tw)0.5 [16].
5. Heat transfer correlations for laminar tube ow
To calculate the Nusselt number for the transition and turbulent

5.1. Heat transfer correlations for hydrodynamically fully developed

and thermally developing laminar ow
In turbulent ow fully developed hydraulically, the uid velocity
prole is formed. This happens in laboratory studies of heat transfer
or heat exchangers used in the industry when the heated or cooled
tube section is preceded by a long thermally insulated tube section.
In such a case in the heated or cooled section of the tube occurs
hydrodynamically fully developed and thermally developing ow.
Laminar heat transfer in the inlet section of the tube assuming the
hydrodynamically fully developed ow is the subject of the work of
Shah and London [48].

ve
^que problem [48e50]
Using the analytical solutions of the Le
valid for small values of the ratio (x/dw)/(RePr) available in literature, Shah and London [48] evaluated the local and mean Nusselt
numbers for hydrodynamically developed and thermally developing ow both for uniform wall heat ux and uniform wall temperature. The Nusselt number given in tabular form can be
approximated by a simple function using the method of Churchill
and Usagi [38].
According to their method, a heat transfer correlation can be
formed by raising to the power of (1/c) the sum of the c-th power of
the limiting heat transfer correlations for large and small values of
the independent variable. Intermediate values of the Nusselt
number are then used to nd the optimum value of the exponent c
by the least squares method.
Choosing the form of the tting function and approximating the
data reported by Shah and London [48] according to the method of
Churchill and Usagi [38], the mean Nusselt number Num over the
tube length L can be expressed as.
 uniform wall heat ux (UHF boundary condition)

1=4:829

4:829
Num;q Nu4:829
;
m;q;1 Num;q;2

Fig. 2. Approximation of the results presented in Table 1 by the function (57).

1$106 

1 L
 1:0
RePr dw
(62)

The coefcient of determination r2 and the absolute maximum

116

D. Taler / International Journal of Thermal Sciences 108 (2016) 108e122

error jmax j for formula (62) are r2 0.99979 and jmax j  3:9%. The
symbol Num,q,1 in Eq. (62) denotes the mean Nusselt number for
hydrodynamically and thermally fully developed ow

Num;q;1

48
4:364
11

(63)

ve
^que solution [49,50] Num,q,2 denotes the mean Nusselt
The Le
number for hydrodynamically and thermally fully developed ow
over the plate with linear temperature prole in the uid and
constant heat ux at the wall surface





dw 1=3
dw 1=3
Num;q;2 31=3 G2=3 RePr
1:9530 RePr
L
L
(64)

ve
^que [49] using the local Nusselt
Eq. (64) was derived by Le
number for very small values of the parameter (x/dw)/(RePr), where
x is the distance from the tube inlet.
 uniform wall temperature (UWT boundary condition)

1=5:8965

5:8965
Num;T Nu5:8965
;
m;T;1 Num;T;2

1$106 

1 L
RePr dw

 1:0
(65)
The coefcient of determination r2 and the absolute maximum
error jmax j for formula (65) are r2 0.99958, jmax j  5:7%,
respectively.
The symbol Num,T,1 means the Nusselt number for hydrodynamically and thermally fully developed laminar ow for the UWT
boundary condition

Num;T;1 3:66;

(66)

ve
^que solution [49,50] for the UWT boundary condition is
The Le
given by

Num;T;2





31=3
1 L 1=3
dw 1=3
1:6151 RePr
G4=3 RePr dw
L

hydraulically and thermally developing laminar uid ow in a tube,

the following formulas recommended in the VDI Heat Atlas [33] can
be applied

i1=3
h

3
Num;q Nu3m;q;1 0:63 Num;q;2  0:6 Nu3m;q;3

(70)

i1=3
h

3
Num;T Nu3m;T;1 0:73 Num;T;2  0:7 Nu3m;T;3
(71)

(67)

Fig. 3 compares the results of calculating Nusselt number by

correlation (62) and (65) with the data presented in the work of
Shah and London [48].
The analysis of the results shown in Fig. 3 demonstrates that the
relationships (62) and (65) t well to the data of Shah and London
[48]. Slightly more complicated but more accurate formulas suggested Gnielinski [33,51] based on the values of Nusselt number
calculated by Shah and London [48],

h

3 i1=3
Nu3m;q;1 0:63 Num;q;2  0:6

(68)

h

3 i1=3
Num;T Nu3m;T;1 0:73 Num;T;2  0:7

(69)

Num;q

Fig. 3. The mean Nusselt number Num for hydrodynamically fully developed and
1
L e approxithermally developing laminar ow as a function of the parameter RePr
dw
mation of Shah and London data [48] using functions (62) and (65).

Correlation (62) or (68) can be used to calculate the Nusselt

number Num,q(Re 2300) in the relationship (61) for a uniform
heat ux on the inner surface of the tube. When the tube wall
temperature is uniform, then the Nusselt Num,T(Re 2300) in Eq.
(61) can be calculated using the relationship (65) or (69).
5.2. Heat transfer correlations for hydrodynamically and thermally
developing laminar ow
To calculate the average Nusselt number Num,q or Num,T for

The Nusselt number Num,q,3 appearing in the correlation

developed for a tube with a constant wall heat ux, and the Nusselt
number Num,T,3 for a tube with constant wall temperature are given
by



dw 1=2
Num;q;3 0:924Pr 1=3 Re
L

Num;T;3

2
1 22Pr

1=6 
RePr

dw
L

(72)
1=2
(73)

Correlations (72) and (73) were determined by an approximation of Nusselt numbers obtained from the numerical solutions of
the momentum conservation equation and energy conservation
equation assuming a at velocity and temperature proles at the
inlet of the tube.
6. Comparison with experimental data
Most comparisons of calculated and experimentally determined
Nusselt number, which will be carried out, refers to ows with
Reynolds numbers greater than 2300. Only one comparison will be
conducted for liquid laminar ow. The correction multipliers (Pr/
Prw)n1 or (T/Tw)n2 for temperature-dependent uid properties were
neglected, while comparing proposed correlations with experimental data. The symbols n1 and n2 denote empirical constants.

D. Taler / International Journal of Thermal Sciences 108 (2016) 108e122

When a liquid is heated, the viscosity near the wall is smaller than
that in the in the tube center, while in liquid cooling the reverse
trend is observed. For gases, the opposite changes of viscosity occur.
The viscosity of gases increases with increasing uid temperature.
Thus, the viscosity of the gas near the wall in the heated channel is
greater than the viscosity of the gas in the channel axis, where the
gas is cooler. The results of experiments that were taken for comparison were obtained for test tubes heated electrically to achieve
uniform heat ux over the entire length of the test section or the
test tube was a central tube in tube-in-tube heat exchangers in a
counter-current or co-current ow arrangement. In the tubes
heated electrically, the working uid was air. As is known, the effect
of temperature on the Prandtl number and other thermophysical
properties of gases is low. This is also conrmed by CFD simulations
of heat transfer in cross-ow heat exchangers with tubes of
different wall temperatures [52,53]. It was shown that tube wall
temperature changes affect only marginally the air-side heat
transfer coefcient on the outer surface of the tubes. Also, Tam et al.
[54] making studies of the turbulent heat transfer in an electrically
heated tube found that the bulk-to-wall viscosity ratio is the least
important parameter compared with the Reynolds and Prandtl
numbers. In tube-in-tube heat exchangers, differences between the
tube inner surface temperature and the bulk temperature are the
order of a few Kelvin so that equal Prandtl numbers calculated at
the wall and bulk temperature may be assumed. In making the
comparison of results of calculations using the correlation developed, the constant heat ux boundary condition will be adopted in
the case of electrically heated tubes, while for inner tube of double
pipe heat exchangers the constant wall temperature or constant
wall heat ux will be assumed. If variations in wall temperature
over the length of the tube in a heat exchanger are small, then the
constant wall temperature can be adopted [55]. For larger wall
temperature difference between the inlet and outlet of a heat

Fig. 4. Comparison of experimental data of Raj et al. [56] with the Nusselt numbers
calculated using correlation (65) developed in the paper, correlation (74) of Hausen,
and correlation (69) presented in VDI Heat Atlas for hydrodynamically developed and
thermally developing laminar ow in the tube with UWT boundary condition;
1Pr 80.6 [56], 2Pr 80.5 [56], 3 e correlation (65) proposed in the paper, 4 e
correlation (74) of Hausen [24], 5-correlation (69) recommended in VDI Heat Atlas
[33].

117

exchanger, the constant wall heat ux boundary condition is closer

to the reality, especially in counter-ow heat exchangers.
First, the correlations (65) and (69) used for calculating the
average Nusselt number for thermally developing and hydraulically
fully developed laminar ow will be compared with the experimental data of Raj et al. [56]. Heat transfer and pressure drop experiments were carried out in a counter-ow double pipe heat
exchanger. The coolant (de-ionized water) circulated in the annulus
while in the tube owed hot ethylene-glycol. A 2 m long unheated
tube with an inner diameter of 15.88 mm is followed by 2590 mm
long heated test section. The velocity prole of the uid at the inlet
of the test section was fully developed because of the high value of
the ratio of the length of the unheated tube section to internal tube
diameter that is equal to L/dw 2000/15.88 126.
The Nusselt number values determined experimentally are little
larger than the values determined from the correlations (65) and
(69), which were derived using the Nusselt number obtained from
analytical solutions. Hausen [24] found the following formula for
the mean Nusselt number in laminar ow through tubes with
constant wall surface temperature

Num;T 3:66

0:0668RePr dLw
0:66

1 0:045 RePr dLw

100  RePr

dw
 1500
L
(74)

Also, Hausen's correlation (74) underpredicts the experimental

data (Fig. 4).
The increase in experimentally determined Nusselt numbers in
comparison to correlations (65), (69) and (74) can be attributed to
natural convection caused by the buoyancy forces that induce
secondary ows [54,55]. The correlations (65), (69), and (74) do not
account for the natural convection ows that enhance heat transfer
in the horizontal tube.
Next, the correlation (61) proposed in this paper is compared in
Figs. 5e11 with experimental data available in literature and the
correlation obtained on the basis of the investigation of a plate n
and tube heat exchanger. First, the relation (61) will be compared
with experimental results presented in the PhD theses [57e60]. The
experimental data for smooth tubes with uniform wall heat ux
reported in these theses were published by Abraham in [42] and
used by him to verify the CFD results. Later, the same experimental
results of dissertations [57e60] used Gnielinski [34] to validate the
correlation (12). The experimental data presented in [57e60] were
compared in Fig. 5 with the results of calculations using the relationship (61) presented in this paper and the Gnielinski correlation
(12). To calculate Num,q (Re 2300), Eq. (70) was used, i.e. the
hydrodynamically and thermally developing ow was assumed.
The friction factor in Eq. (12) as well as in Eq. (61) was calculated
using the Filonienko Formula (7). The correction for temperaturedependent properties was neglected, both in Eq. (12) and Eq. (61).
The comparison presented in Fig. 5 shows that the relationship
(61) proposed in the paper underpredicts a bit experimental data
and Gnielinski's correlation (16).
The transition-regime results shown in Fig. 5 were obtained by
Abraham et al. [46] for fully developed tube ow using the ANSYS
CFX 12.0 software. Nusselt number values determined numerically
by Abraham et al. approximate very well the relationship (16) of
Gnielinski in the range of Reynolds numbers from 2900 to 4800
(Fig. 5). This is due to the fact that turbulent Prandtl number Prt in
the CFD model was adjusted [42] so that the computed Nusselt
numbers in the range of Reynolds numbers from 3100e4800 were
close to the values obtained from the Gnielinski relationship (12),
and in the range of Reynolds numbers from 2300e3100 close to the

118

D. Taler / International Journal of Thermal Sciences 108 (2016) 108e122

Fig. 5. Comparison of the Nusselt number for UHF boundary condition calculated
using the correlation (61) proposed in the paper with the experimental data of Lau,
Black, Kemink, and Wesley [57e60], the Gnielinski correlation (16) [34], and CFD
simulation results obtained by Abraham et al. [46].

values obtained from the Churchill formula [37].

Extensive experimental studies conducted by Olivier and Meyer
[55] showed that for the heated or cooled tubes transition from
laminar to turbulent ow regime occurs at the Reynolds numbers
ranging between 2100 and 2300 irrespective of the type of the tube
inlet. The experimental results of Olivier and Mayer conrm the
correctness of the proposed correlation (61) that assumes that the
transition from laminar to turbulent ow occurs at a Reynolds
number of 2300. The experimental study of Olivier and Mayer also
shows that the relation (61) can be used for the calculation of
various types of heat exchangers because the kind of the inlet to the
tube resulting from the method of attachment of the pipe to the
intake manifold has little effect on the Nusselt number.
The correlation (61) was compared in Fig. 6 with the experimental data of Olivier and Mayer [55]. The test section is a central
tube in a counterow double tube heat exchanger. The two 5 m
long tubes with an inner diameter of 14.482 mm and 17.6 mm were
studied. Distilled hot water owing through the inner tube was
cooled by distilled water owing through the annulus. The cold
water owed through the annulus at a high mass ow rate to keep
the constant temperature of the wall of the inner tube and to
ensure small thermal resistance at the outer surface of the central
tube. High heat transfer coefcient on the outer surface of the

Fig. 6. Comparison of the Nusselt numbers obtained using Eq. (61) developed in the
paper with experimental data by Olivier and Meyer [55] and the Gnielinski relationship (12) [26,27] for hydrodynamically and thermally developing ow in the tube with
UWT boundary condition.

Fig. 7. Comparison of UHF Nusselt numbers obtained using correlation (61) developed
in the paper and relationship (16) proposed by Gnielinski [34] with experimental data
of Buyukalaca et al. [61] and Eiamsa-Ard et al. [62] for hydrodynamically developed
and thermally developing tube ow.

central tube makes possible to determine experimentally the heat

transfer coefcient on the inner surface with greater accuracy. The
ow in the test tube was hydrodynamically and thermally developing. The constant wall temperature boundary condition can be
adopted for the inner tube.

Fig. 8. Comparison of experimental data of Huber and Walter [39] for hydrodynamically and thermally developing ow with the correlation (61) developed in the paper
and with the Gnielinski correlation (12) taking into account the nite length of the
tube, dw 32.8 mm, L 3000 mm; 1eexperimental data of Huber and Walter [39],
2epresent correlation (61) for UHF boundary condition on the inner surface of the
tube, 3-present correlation (61) for UWT boundary condition on the inner surface of
the tube, 4eGnielinski's relationship (12).

D. Taler / International Journal of Thermal Sciences 108 (2016) 108e122

Fig. 9. Comparison of the ratio Nu/Pr1/3 determined experimentally [63] with the
calculation results using the correlation (61) presented in this paper for hydrodynamically and thermally developing ow in the tube with UHF boundary condition.

The mean Nusselt number Num,T for laminar ow at Re 2300

appearing in the relation (61) was calculated using the correlation
(71). Fig. 6 shows a comparison of average Nusselt number determined experimentally by Olivier and Meyer [55] to the calculated
Nusselt number using the correlation (61).
The analysis of the results depicted in Fig. 6 shows that the
correlation (61) proposed in the present paper very well approximates the experimental data obtained by Olivier and Meyer [55].
The Gnielinski correlation (12) approximates a little worse the
experimental data of Olivier and Meyer.

Fig. 10. Comparison of the ratio Nu/Pr1/3 determined experimentally [64] with the
calculation results using the correlation (61) presented in this paper for hydrodynamically and thermally developing ow in the tube with UHF boundary condition.

119

Fig. 11. Comparison of the correlation (61) proposed in this paper with the relationship (75) obtained on the basis of experimental tests of the heat exchanger for hydrodynamically and thermally developing ow in the tube with UWT boundary
condition; 1-correlation (61), 2-correlation (75).

Next experimental data, which are compared with Eq. (61) were
obtained by Buyukalaca et al. [61] and Eiamsa-Ard e al [62]. In both
studies the working uid was air.
In a study by Buyukulaca et al., the 9 m long test tube having an
inner diameter of 56 mm was heated electrically along the 3 m long
test section. The test section was preceded by a calming section of
6 m [61]. The Reynolds number varied between 3115 and 25000.
The experimental study of Eiamsa-Ard et al. [62] was carried out
using the test section of 1.5 m long with 47.5 mm inner diameter.
The unheated calming section had a length of 2.5 m and the same
inner diameter. The Reynolds number ranged from 4500e20000. In
both studies [61], and [62], the test tube sections were heated by a
exible electrical wire to provide a constant heat ux at the outer
tube surfaces. The output of the electrical heaters was controlled by
Variac transformers. The outer surfaces of the heaters were well
insulated with glass wool to reduce heat loss to surroundings. At
the inlet to the heated section of the tube, uid ow was hydraulically developed. To calculate the Nusselt number Num,q in laminar
ow for Re 2300, Eq. (68) was used. Comparison of Nusselt
numbers determined experimentally in [61,62] with the results of
calculations by Formula (61) proposed in this paper and the results
of calculations using the Gnielinski formulas (16) and (12) is shown
in Fig. 7.
The experimental data of Buyukalaca et al. [61] and Eiamsa-Ard
[62] differ signicantly. Undervalued Nusselt numbers measured by
Eiamsa-Ard may be due to measurement errors of the tube wall and
air temperature or by measurement errors of the air mass ow rate.
Overestimated values occurring in the Nusselt number for smaller
Reynolds numbers in the experimental data of Buyukalaca et al.
[61] may be caused by natural convection in a horizontal tube.
The heat transfer in air ow through a vertical circular tube was
studied experimentally by Huber and Walter [39]. Hot air owed
upwards through the inner tube of a co-current double pipe heat
exchanger while water with lower temperature owed in the
annulus between the inner and outer tube. A test tube 3 m long has
an inner diameter of 32.8 mm. The Reynolds number ranged from

120

D. Taler / International Journal of Thermal Sciences 108 (2016) 108e122

4095 to 10400. The air-side Nusselt numbers determined by Huber

and Walter [39] are compared in Fig. 8 with the Nusselt numbers
calculated using the correlation (61). The Nusselt numbers Num,q
and Num,T for hydrodynamically and thermally developing laminar
ow were calculated at Re 2300 using correlations (70) and (71),
respectively. Fig. 8 also shows the values of the Nusselt number
obtained by the Gnielinski correlation (12).
An inspection of the results shown in Fig. 8 indicates that the
best agreement between experimental and calculation results was
obtained for correlation (61) developed in the present paper when
the uniform wall ux boundary condition was assumed. If the air
and water temperature variations along the test tube are small,
then the assumption of a uniform heat ux at the tube surface is
well-founded. Also, the calculation results obtained for a uniform
wall temperature are acceptable.
Experimental study of heat transfer in a horizontal tube 2 m
long with an inner diameter of 17 mm was carried out by Li et al.
[63]. The experiments were conducted using a counter-ow double
pipe heat exchanger with cold water owing through the annulus
and hot water or hot oil in the test tube. The tube-side Reynolds
number varied from 2500e90000 for water and from 2500e12000
for oil. When calculating the tube-side Nusselt number for laminar
ow at Re 2300, the uniform wall heat ux boundary condition
was adopted, since the mass ow rate of the water in the annulus
was not kept high to ensure the constant wall temperature. In
counter-ow heat exchangers, the difference between the hot and
cold uid temperature does not change signicantly in the entire
heat exchanger so that the uniform wall heat ux may be assumed.
The Nusselt numbers obtained from the experiment and calculation using Eq. (61) are plotted in Fig. 9.
An examination of the results depicted in Fig. 9 shows an
excellent conformity of the experimentally determined and calculated Nusselt numbers.
Other experimental data obtained by the same authors [64] are
compared in Fig. 10 with the results of calculations using correlation (61) developed in this paper. Experimental studies were carried out on the same test rig, as described in the previous work [63].
The Reynolds number varied in the range of 7000e90000. Agreement between the results of calculations by the formula (61) and
the experimental data [64] is very good. Fig. 10 shows the results of
calculating the Nusselt number for the uniform wall heat ux and
uniform wall temperature. The differences in the results are small.
This is due to a little difference in the Nusselt numbers Num,T and
Num,q calculated for laminar ow for the Reynolds number of
Re 2300, which are respectively Num,T(Re 2300)6.83 and
Num,q(Re 2300)8.74.
The high accuracy of the correlation (61) also conrmed the
author's own research [65].
The automotive radiator tested in [65] is used for cooling the
spark ignition engine of a cubic capacity of 1580 cm3. The radiator
consists of 38 tubes of an oval cross-section, with 20 of them
located in the upper pass with ten tubes per row. The path of the
coolant ow is U-shaped. The two rows of tubes in the rst pass are
fed simultaneously from one header. The water streams from the
rst and second row are mixed in the intermediate header.
Following that, the water is uniformly distributed between the
tubes of the rst and second row in the second pass. The outer
diameters of the oval tube are: d min 6.35 mm and
d max 11.82 mm. The tubes are L 0.52 m long. The thickness of

Fig. 12. Comparison of UWH Nusselt numbers for hydrodynamically and thermally
developing ow in an oval tube obtained using Eq. (61) developed in the paper with
the relationship (16) proposed by Gnielinski in [34], and correlation (18) obtained by
Abraham et al. [42] for Pr 0.7 and dw/L 0; a) Pr 0.7, b) Pr 5, c) Pr 500.

D. Taler / International Journal of Thermal Sciences 108 (2016) 108e122

the tube wall is dt 0.4 mm. The number of plate ns is 520. The
hydraulic diameter of an oval tube is calculated using the formula
dh 4 Aw, in/Pin, where the symbol Aw,in denotes the cross-section
area of the tube and Pin is the inner perimeter. The water-side
Reynolds and Nusselt numbers were determined on the base of
the hydraulic diameter dh 7.06 mm. The Reynolds number varied
in the range 4 000  Re  12 000. Using the measurement data and
data reduction method presented in [65] the following correlation
was obtained
x

Nu Num;T Re 2300
"

dh
 1
L

2=3 #

 2300Pr 1:008
q

1:08 11:31 8x Pr 2=3  1
8 Re

(75)

4300  Re  12000

Dimensionless numbers appearing in the correlation (66) are

dened as follows

Nu

hdh
;
k

Pr

cp m
;
k

Re

rwdh
m

The ow in the tubes of the exchanger is undeveloped hydraulically and thermally, therefore, the Nusselt number
Num,T(Re 2300) was calculated using the correlation (71).
A comparison of the relationship (75) and (61) reveals that,
instead of a constant 12.39 in the denominator of the relationship
(61), in the correlation (75) is constant 11.31. Comparison of the
correlation (61) proposed in this work with the relationship (75)
obtained on the basis of experimental tests of the heat exchanger is
shown in Fig. 11.
Compatibility of correlations (61) and (75) is very satisfactory.
The relative difference in the compared Nusselt numbers for
Re 12000 is e [(71.54  69.15)/69.15]$100 3.46%.
The results of the comparisons presented in Figs. 5e11 show
that the correlation (61) proposed in this paper very well approximates the experimental data. Minor differences between the
calculated and experimentally determined Nusselt numbers may
be due to errors in measured temperatures of the test tube wall, and
by inaccuracies in measured uid bulk temperature. Similarly,
different data reduction procedures used for experimental determining the Nusselt number may lead to somewhat different results.
Correlation (61) is also compared with the method proposed by
Gnielinski [34] for calculating the Nusselt numbers in the range of
transition from laminar to turbulent ow, which is based on a linear
interpolation by the Formula (16). The Nusselt number
Num,q(Re 2300) was calculated using the correlation (70) both for
the correlation (61) and Gnielinski method. The Nusselt numbers as
a function of the Reynolds number, shown in Fig. 12a)ec), were
calculated using the Formulas (61) and (16). The calculations were
conducted for different Prandtl number and different ratios of the
length to the internal diameter of the tube.
Fig. 12a shows also the Nusselt number calculated using Eq. (18)
proposed by Abraham et al. [42] which was determined based on
the approximation of the results obtained from CFD simulations.
There is good agreement between the Nusselt numbers calculated
using Eq. (18) proposed by Abraham, and the Gnielinski correlation
(16) as well as between the correlation (61) proposed in this paper.
The relationship (18) gives slightly larger Nusselt numbers for
Reynolds numbers close to 3100 as turbulent Prandtl number Prt in
the CFD simulation was selected so as to obtain Nusselt numbers
similar to those calculated from the Gnielinski relationship (12)
[42].
From the analysis of the results presented in Fig. 12 can be seen
that the new linear interpolation between the equations for

121

laminar and turbulent ow in the transition region developed

recently by Gnielinski [34], exhibits unusual behavior for short
tubes. When the ratio L/dw is equal to 5 then the Nusselt number for
Re 4000 is less than the Nusselt number for Re 2300. To remove
this disadvantage, the Nusselt number would be linearly interpolated in the range of Reynolds number from 2300 to 5000 or 6000,
rather than in the range from 2300 to 4000.
7. Conclusions
A new equation for the Nusselt number for transition and turbulent ow in channels has been proposed. The heat transfer correlation developed in the paper can be used for uniform wall heat
ux and uniform wall temperature boundary conditions. It is accurate and easy to use. The correlation is based on the fundamentals of uid mechanics and heat transfer. Nusselt number in the
transitional and turbulent ow regime is the sum of the laminar
component and a turbulent component. The form of the correlation
was selected in such a way that for the Reynolds number equal to
Re 2300, i.e. at the point of transition from laminar to transitional
ow, the Nusselt number should change continuously. The heat
transfer relationship developed in the paper approximates very
well experimental data.
References
[1] Dittus FW, Boelter LMK. Heat transfer in automobile radiators of the tubular
type, vol. 2. The University of California Publications on Engineering; 1930.
p. 443e61. Reprinted in Int. Commun. Heat Mass 12(1985) 3e22.
[2] Winterton RHS. Where did the Dittus and Boelter equation come from? Int J
Heat Mass Transf 1998;41:809e10.
[3] McAdams WH. Heat transmission. 3rd edn. New York: McGraw-Hill; 1954.
[4] Colburn AP. A method of correlating forced convection heat transfer data and
a comparison with uid friction. Trans Am Inst Chem Eng 1933;29:174e210
(Reprinted in Int. J. Heat Mass Tran. 23(1996) 287e292).
[5] Welty JR, Wicks Ch E, Wilson RE, Rorrer GL. Fundamentals of momentum,
heat, and mass transfer. 5th edn. Hoboken: John Wiley & Sons; 2008.
[6] Moody LF. Friction factors for pipe ow. Trans ASME 1944;66:671e84.
[7] Webb RL. A critical evaluation of analytical solutions and Reynolds analogy
equations for turbulent heat and mass transfer in smooth tubes. Warme
Stoffubertrag 1971;4:197e204.
[8] Allen RW, Eckert ERG. Friction and heat-transfer measurements to turbulent
pipe ow of water (Pr 7 and 8) at uniform wall heat ux. J Heat Trans T
ASME 1964;86:301e10.
[9] Kreith F, Manglik RM, Bohn MS. Principles of heat transfer. 7th edn. Stamford:
Cengage Learning; 2011.
[10] Siddique M, Alhazmy M. Experimental study of turbulent single-phase ow
and heat transfer inside a micro-nned tube. Int J Refrig 2008;31:234e41.
[11] Zhang Z, Yang W, Guan Ch, Ding Y, Li F, Yan H. Heat transfer and friction
characteristics of turbulent ow through a plain tube inserted with rotorassembled strands. Exp Therm Fluid Sci. 2012;38:33e9.
[12] Hojjat M, Etemad S Gh, Bagheri R, Thibault J. Convective heat transfer of nonNewtonian nanouids through a uniformly heated circular tube. Int J Therm
Sci. 2011;50:525e31.
[13] Moghadassi A, Parvizian EGF. A numerical study of water-based Al2O3 and
Al2O3-Cu hybrid nanouid effect on forced convective heat transfer. Int J
Therm Sci. 2015;92:50e7.
[14] Zamzamian A, Oskouie SN, Doosthoseini A, Joneidi A, Pazouki M. Experimental
investigation of forced convective heat transfer coefcient in nanouids of
Al2O3/EG and CuO/EG in a double pipe and plate heat exchangers under
turbulent ow. Exp Therm Fluid Sci. 2011;35:495e502.
[15] Petukhov BS, Kirillov VV. The problem of heat exchange in the turbulent ow
of liquids in tubes. Teploenergetika 1958;4(4):63e8 (in Russian).
[16] Petukhov BS. Heat transfer and friction in turbulent pipe ow with variable
physical properties. In: Hartnett JP, Irvine TF, editors. Advances in heat
transfer, vol. 6. New York: Academic Press; 1970. p. 503e64.
[17] engel YA, Cimbala JM, Turner RH. Thermal-uid sciences. 4th edn. New York:
McGraw-Hill; 2012.
[18] Filonienko GK. Friction factor for turbulent pipe ow. Teploenergetika
1954;1(4):40e4 (in Russian).
[19] Konakov PK. A new correlation for the friction coefcient in smooth tubes.
Izvestija AN SSSR 1946;LI 51(7):503e6.
[20] Lyon RN. Liquid metal heat transfer coefcients. Chem Engr Progr. 1951;47(2):
75e9.
ndige Darstellung der turbulenten Geschwindigkeitsver[21] Reichardt H. Vollsta
teilung in glatten Leitungen. Z Angew Math Mech 1951;31(7):208e19.
[22] Reichardt H. The principles of turbulent heat transfer. Transl. by P.A. Scheck.

122

[23]

[24]
[25]
[26]

[27]
[28]
[29]

[30]

[31]

[32]

[33]

[34]
[35]

[36]
[37]

[38]
[39]

[40]
[41]

[42]

[43]

D. Taler / International Journal of Thermal Sciences 108 (2016) 108e122

In: Hartnett JP, editor. Recent advances in heat and mass transfer. Boston:
McGraw-Hill; 1961. p. 223e52.
mungslehre. Braunschweig: Vieweg und Sohn;
Prandtl L. Fhrer durch die Stro
1949 (Translation into English: L. Prandtl, Essentials of uid dynamics, Blackie
& Son, London, 1969, 117).
Hausen H. Heat transfer in counter ow, parallel ow, and cross ow. Boston:
McGraw- -Hill; 1983.
Sieder EN, Tate GE. Heat transfer and pressure drop in liquids in tubes. Ind Eng
Chem. 1936;28:1429e35.
rme- und den Stoffbergang in
Gnielinski V. Neue Gleichungen fr den Wa
mten Rohren und Kan
turbulent durchstro
alen. Forsch Ingenieurwes Eng Res
1975;41:8e16.
Gnielinski V. New equations for heat and mass transfer in the turbulent pipe
and channel ow. Int Chem Eng 1976;16:359e68.
Corcione M, Cianfrini M, Quintino A. Heat transfer of nanouids in turbulent
pipe ow. Int J Therm Sci. 2012;56:58e69.
Yu-ting W, Bin L, Chong-fang M, Hang G. Convective heat transfer in the
laminar- turbulent transition region with molten salt in a circular tube. Exp
Therm Fluid Sci. 2009;33:1128e32.
Lu J, He S, Liang J, Ding J, Yang J. Convective heat transfer in the laminarturbulent transition region of molten salt in annular passage. Exp Therm
Fluid Sci. 2013;51:71e6.
Taborek J. Design method for tube-side laminar and transition ow regime
with effects of natural convection. In: Paper OPF-11e21, open forum session
forum of the ninth int. Heat transfer conference, Jerusalem, Israel; 1990.
rmebertragung im
Gnielinski V. Ein neues Berechnungsverfahren fr die Wa
mung. Forsch
bergangsbereich zwischen laminarer und turbulenter Rohrstro
Ingenieurwes (Eng Res) 1995;61:240e8.
mte Rohre, Abschnitt G1, VDI-Wa
rmeatlas 11., bearGnielinski V. Durchstro
beitete und erweiterte Auage. Berlin-Heidelberg: Springer-Vieweg; 2013.
p. 785e91.
Gnielinski V. On heat transfer in tubes. Int J Heat Mass Tran 2013;63:134e40
(Corrigendum to this paper in Int J. Heat Mass Tran. 81(2015) 638).
Colebrook CF. Turbulent ow in pipes, with particular reference to the transition region between the smooth and rough pipe laws. J Inst Civ Eng 19381939;11(4):133e56.
Taler D. Determining velocity and friction factor for turbulent ow in smooth
tubes. Int J Therm Sci. 2016;105:109e22.
Churchill SW. Comprehensive correlating equations for heat, mass and momentum transfer in fully developed ow in smooth tubes. Ind Eng Chem
Fundam 1977;16:109e16.
Churchill SW, Usagi R. A general expression for the correlation of rates of
transfer and other phenomena. AIChE J. 1972;18:1121e8.
Huber D, Walter H. Forced convection heat transfer in the transition region
between laminar and turbulent ow for a vertical tube. In: Mastorakis N,
Mladenov V, Bojkovic Z, editors. Latest trends on theoretical and applied
mechanics, uid mechanics and heat transfer. Wisconsin: WSEAS Press; 2010,
ISBN 978-960-474-211-0. p. 132e6.
Tam LM, Ghajar AJ. Transitional heat transfer in plain horizontal tubes. Heat
Transf Eng. 2006;27:23e38.
Abraham JP, Tong JCK, Sparrow EM. Breakdown of laminar pipe ow into
transitional intermittency and subsequent attainment of fully developed
intermittent or turbulent ow. Num Heat Tran A 2008;54:103e15.
Abraham JP, Sparrow EM, Tong JCK. Heat transfer in all pipe ow regimes:
laminar, transition/intermittent and turbulent. Int J Heat Mass Tran 2009;52:
557e63.
Minkowycz WJ, Abraham JP, Sparrow EM. Numerical simulation of laminar
breakdown and subsequent intermittent and turbulent ow in parallel plate

[44]
[45]

[46]

[47]
[48]
[49]
[50]
[51]
[52]

[53]

[54]

[55]

[56]

[57]

[58]
[59]
[60]
[61]

[62]

[63]
[64]
[65]

channels: effects of inlet velocity prole and turbulence intensity. Int J Heat
Mass Tran 2009;52:4040e6.
Lovik RD, Abraham JP, Minkowycz WJ, Sparrow EM. Laminarization and turbulentization in a pulsatile pipe ow. Num Heat Tran A 2009;56:861e79.
Abraham JP, Sparrow EM, Tong JCK, Bettenhausen DW. Internal ows which
transit from turbulent through intermittent to laminar. Int J Therm Sci.
2010;49:256e63.
Abraham JP, Sparrow EM, Minkowycz WJ. Internal-ow Nusselt numbers for
the low-Reynolds-number end of the laminar-to-turbulent transition regime.
Int J Heat Mass Tran 2011;54:584e8.
Rennels DC, Hudson HM. Pipe ow. A practical and comprehensive guide.
Hoboken: AIChE- Wiley; 2012.
Shah RK, London AL. Laminar ow forced convection in ducts, supplement 1
to advances in heat transfer. New York: Academic Press; 1978.

ve
^que MA. Les lois de la transmission de chaleur par convection. Ann. des
Le
Mines, Mem. 1928;Series 12, 13:201e99. 305e362, 381e415.
Bird RB, Stewart WE, Lightfoot EN. Transport phenomena. 2nd edn. New York:
Wiley; 2007.
mung und
Gnielinski V. Zur W
armebertragung bei laminarer Rohrstro
konstanter Wandtemperatur. Chem. Ing Techn 1989;61:160e1.

P. Determination of heat transfer formulas for gas ow in the
Taler D, Ocon
n-and-tube heat exchanger with oval tubes using CFD simulations. Chem
Eng Proces 2014;83:1e11.

P. Thermal contact resistance in plate n-and-tube heat exTaler D, Ocon
changers determined by experimental data and CFD simulations. Int J Therm
Sci. 2014;84:309e22.
Tam LM, Ghajar AJ, Tam HK. Contribution analysis of dimensionless variables
for laminar and turbulent ow convection heat transfer in a horizontal tube
using articial neural network. Heat Transf Eng 2008;29:793e804.
Olivier JA, Meyer JP. Single-phase heat transfer and pressure drop of the
cooling water inside smooth tubes for transitional ow with different inlet
geometries. HVAC&R Res 2010;16:471e96.
Raj R, Lakshman NS, Mukkarmala Y. Single phase ow heat transfer and
pressure drop measurements in doubly enhanced tubes. Int J Therm Sci.
2015;88:215e27.
Lau S. Effect of plenum length and diameter on turbulent heat transfer in a
downstream tube and on plenum-related pressure loss (Ph.D. thesis). University of Minnesota; 1981.
Black III A. The effect of circumferentially-varying boundary conditions on
turbulent heat transfer in a tube (Ph.D. thesis). University of Minnesota; 1966.
Kemink R. Heat transfer in a downstream tube of a uid withdrawal branch
(Ph.D. thesis). University of Minnesota; 1977.
Wesley D. Heat transfer in pipe downstream of a Tee (Ph.D. thesis). University
of Minnesota; 1976.
Buyukalaca O, Ozceyhan V, Gunes S. Experimental investigation of thermal
performance in a tube with detached circular ring turbulators. Heat Transf
Eng 2012;33:682e92.
Eiamsa-Ard S, Kongkaitpaiboon V, Promvonge P. Thermal performance
assessment of turbulent tube ow through wire coil turbulators. Heat Transf
Eng 2011;32:957e67.
Li XW, Meng JA, Li ZX. Experimental study of single-phase pressure drop and
heat transfer in micro-n tube. Exp Therm Fluid Sci. 2007;32:641e8.
Li XW, Meng JA, Li ZX. Roughness enhanced mechanism for turbulent
convective heat transfer. Int J Heat Mass Tran 2011;54:1775e81.
Taler D. Experimental determination of correlations for average heat transfer
coefcients in heat exchangers on both uid sides. Heat Mass Transf 2013;49:
1125e35.