International Journal of Thermal Sciences 105 (2016) 109e122

Contents lists available at ScienceDirect

International Journal of Thermal Sciences

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

Determining velocity and friction factor for turbulent flow in smooth

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: The most popular explicit correlations for the friction factor in smooth tubes are reviewed in this paper.
Received 5 October 2015 The friction factor for the turbulent flow in smooth tubes is required in some correlations when calcu-
Received in revised form lating the Nusselt number. To calculate the friction factor, the velocity profile in a turbulent smooth wall-
11 January 2016
tube must be estimated at first. The radial velocity distribution was determined using either universal
Accepted 16 February 2016
velocity profile found experimentally by Reichardt or by integration the momentum equation using the
eddy diffusivity model of Reichardt. The friction factor obtained by using the universal velocity profile
gives better results than that obtained from the momentum equation when compared with the Prandtl
Keywords:
evon Ka
rma

neNikuradse equation. Based on the velocity profiles proposed by Reichardt the friction
Turbulent tube flow
Smooth-wall tube factor was calculated as a function of the Reynolds number and subsequently two formulas for the
Friction factor friction were proposed. They have satisfactory accuracy when comparing with the implicit Prandtlevon
Momentum conservation equation Ka
rma
neNikuradse equation. Thus, it was concluded that the universal velocity profile proposed by
Velocity profile Reichardt will provide good results when it is taken into account while integrating the energy conser-
vation equation. There is also a considerable number of experimental correlations for the friction factor
in smooth tubes. All these relationships were compared with the experimental data and with the implicit
Prandtlevon K
arm
aneNikuradse equation that is considered as a standard to test the explicit approxi-
mations. The Colebrook and Filonienko explicit correlations are widely used when calculating the Nusselt
number for the turbulent flow but they have noticeable errors for small Reynolds number ranged from
3000 to 7000 for the Colebrook relation and from 3000 to 30,000 for the Filonienko relation. For this
reason, a new simple and accurate correlation for the friction factor for Reynolds numbers between 3000
and 107 is proposed in the paper.
© 2016 Elsevier Masson SAS. All rights reserved.

1. Introduction the experiment.

Blasius was the first who proposed an explicit correlation for
The determination of a friction factor in turbulent tube flow is turbulent tube flow that is valid for Reynolds numbers between
essential not only to pressure drop calculations in pipelines and 3000 and 105. The Blasius correlation is still in use [4]. Sheikho-
heat exchangers [1] but also is needed for calculating the Nusselt leslami et al. [5] studied turbulent flow and heat transfer in the air
number in turbulent tube flow [2,3]. The correlations for the fric- to water double-pipe heat exchanger. The Blasius formula was used
tion factor x can be found experimentally based on the measured to calculate the friction factor needed for the estimation of the heat
pressure drop in a tube over a given distance or on the measured transfer by the Gnielinski correlation. Only about 20 years later, an
radial velocity profile. The latter way of the friction factor deter- implicit relationship for determining the friction factor was
mining is also important in driving the heat transfer correlation developed by Prandtl, von Ka
rma
n, and Nikuradse (PKN) [6e8]. The
because it allows to choose the most appropriate velocity profile PKN equation for the friction factor for the turbulent flow in a
indirectly. Solving the energy conservation equation using accurate smooth tube is widely accepted and has become a model equation
universal velocity profile will yield the Nusselt numbers as func- for explicit approximations. The PKN correlation is implicit in x
tions of the Reynolds and Prandtl numbers that are consistent with because the friction factor x appears on both equation sides. In
other words, it is a nonlinear algebraic equation that must be solved
either iteratively or graphically. This inconvenience can be cir-
E-mail address: [email protected]. cumvented using a numerous explicit approximation to the PKN

http://dx.doi.org/10.1016/j.ijthermalsci.2016.02.011
1290-0729/© 2016 Elsevier Masson SAS. All rights reserved.
110 D. Taler / International Journal of Thermal Sciences 105 (2016) 109e122

Nomenclature yþ dimensionless distance from the tube wall, yþ ¼ yut =v

c1, c2 constants Greek symbols

dw inner diameter of a circular tube, dw ¼ 2rw, m Dp pressure drop, Pa
ei relative difference Dyþ dimensionless spatial step
i node number 3 turbulence dissipation rate, N/(s m2)
k turbulence kinetic energy, N/(s m2) 3t eddy diffusivity for momentum transfer (turbulent
L distance between pressure taps, m kinematic viscosity), m2/s
n number of nodes in the finite difference grid m dynamic viscosity, kg/(m s)
p pressure, Pa k the von Ka
rma
n constant
PKN Prandtlevon Ka
rma
neNikuradse n kinematic viscosity, v ¼ m=r, m2/s
r radial coordinate, m x DarcyeWeisbach friction factor
rw inner radius of the tube, m r fluid density, kg/m3
r2 coefficient of determination t shear stress, Pa
rþ dimensionless radius, r þ ¼ rut =v tw shear stress at wall surface, Pa
R dimensionless radius, R ¼ r=rw
Re Reynolds number, Re ¼ pw m dw =v
ffiffiffiffiffiffiffiffiffiffiffi Subscripts
ut friction velocity, ut ¼ tw =r, m/s m mean
wm mean velocity i node number
wx velocity component in the x direction w wall surface
wr , wx time averaged velocity component in the x and r
direction, respectively, m/s Superscripts
x a spatial coordinate in Cartesian or cylindrical  time averaged
coordinate systems, m þ dimensionless
y a spatial coordinate in a Cartesian system or distance
from distance from the wall surface, m

equation [1,9e11]. Even a larger number of explicit correlations was respectively, to validate the experimental set-up and data reduction
proposed for implicit ColebrookeWhite equation used to deter- procedure. The Reynolds numbers varied from 2500 to 90,000 for
mine the coefficient of friction in the rough pipes. Many compari- water and from 2500 to 12,000 for oil. The results of measurements
sons of explicit approximations to the ColebrookeWhite equation and calculations agree very well, even for small Reynolds numbers
were conducted over the past two decades. Examples of such near Re ¼ 2500. The maximum relative differences between the
comparative reviews can be publications [1,9e11]. Unfortunately, measured friction factor and the empirical correlation by Filo-
the explicit equations for rough pipes cannot be used for smooth nienko does not exceed 10%. Similar experiments were carried out
pipes, since they were derived for the relative surface roughness by Li et al. with rough tubes [18]. As in the previous study [17], the
greater than zero. measurements were conducted for the turbulent flow of water in a
Petukhov and Kirillov [2,12] proposed in 1958 a formula for the smooth tube. The Reynolds number varied from 7000 to 90,000.
Nusselt number and suggested to use the explicit correlation of Differences between measured and calculated friction factors using
Filonienko [13] for calculating the friction factor. Gnielinski [14] the Filonienko formula are small. Li et al. [19] used the Filonienko
extended the application of the PetukhoveKirillov equation to correlation to show the increase in the friction factor of the discrete
lower Reynolds numbers and continued to calculate the friction double inclined ribbed tubes in relation to the smooth pipes. In the
factor using the Filonienko correlation. Since that time, the Filo- paper of Ji et al. [20] developed turbulent heat transfer in internal
nienko correlation is widely utilized in the calculation of the Nus- helically ribbed tubes is studied experimentally. To test the reli-
selt number for the transitional and turbulent flow in tubes ability of the test facility, the experimental results of the friction
[15e24]. Mirth and Ramadhyani [15] applied the Gnielinski [14] factor were firstly compared with the Filonienko correlation. The
correlation in conjunction with the Filonienko friction factor to relative discrepancy between the experimental data and Filonienko
calculate the water-side heat transfer coefficient inside the tubes of predictions was within ±5% for the Reynolds number ranged from
a finned-tube chilled-water cooler. In all the experiments, high 8000 to 90,000.
water mass flow rates were maintained to provide a turbulent flow Flow heat transfer and pressure drop measurements in doubly
of water. Fernando et al. tested a mini channel aluminum tube heat enhanced tubes were conducted with water and ethylene glycol in
exchanger for water-to-water operation [16]. They found that the the laminar-transition turbulent flow regime by Raj et al. [21]. The
Nusselt numbers obtained in the experiment agree with those aim of the study was to investigate the usefulness of doubly
predicted by the Gnielinski correlation [14] within an accuracy of enhanced tubes for lower duty heat exchangers in the laminar-
±5% in the transition Reynolds number range of 2300e6000. The transition-turbulent flow regime. To check out the experimental
friction factor was calculated using the Filonienko approximation. set-up and the data processing methodology, the tube-side heat
A lot of papers is devoted to the intensification of heat exchange transfer and friction factor were first determined in a 2590 mm
in tubes [17e22]. long copper smooth tube with an inner diameter of 15.88 mm.
Li et al. [17] measured the turbulent tube flow in a micro-fin Turbulent flow friction factors determined experimentally for de-
tube using water and oil. The friction factor and Nusselt numbers ionized water compared to within ±5% of the friction factors pre-
in a smooth tube were first estimated experimentally and dicted by the Filonienko equation.
compared with the Filonienko and Gnielinski correlations, The paper by Zhang et al. [22] reported the thermo-hydraulic
 D. Taler / International Journal of Thermal Sciences 105 (2016) 109e122 111

evaluation of the heat transfer enhancement in the smooth tubes velocity profile. After determining the average velocity of fluid
fitted with rotor-assembled strands of various diameters. The re- using the previously determined velocity profiles, the friction fac-
sults of preliminary experiments carried out on a plane tube were tors will be calculated. Based on a comparison of the determined
compared with the empirical correlation of Gnielinski for the friction factors with the friction factors derived from the reference
Nusselt number and with the Filonienko correlation for the friction PKN and experimental correlations, more accurate method for
factor. The friction factors found experimentally agree within 15% determining the velocity profile will be selected. A more accurate
with the Filonienko equation [22]. velocity profile can be used to determine the distribution of tem-
Hojjat et al. [23] investigated convective heat transfer of non- perature and heat flux in the tube cross-section, which will also
Newton nanofluids through a uniformly heated circular tube. provide greater accuracy in determining the Nusselt number on the
They calculated the Nusselt numbers for de-ionized water using the tube surface.
experimental data and compared with the Nusselt numbers pre- Another important goal of this paper is a comparative study of
dicted by the Gnielinski equation [14] in which the Filonienko existing explicit correlations for the smooth tube friction factor,
empirical correlation was applied to calculate the friction factor. which are used in calculating the Nusselt number using Petukhov
The very satisfactory agreement was obtained between the exper- or Gnielinski correlations or other heat transfer correlations, which
imental results and those obtained by the Gnielinski or the Dit- have smooth tube friction factors.
tuseBoelter correlations in the range of Re from 3000 to 20,000. New simple but accurate correlations for the estimation of the
Heat transfer of nanofluids in a turbulent pipe flow was studied friction factor in smooth tubes will also be offered.
theoretically by Corcione et al. [24] who demonstrated that nano-
fluids behave like single-phase fluids. They recommended using
2. Mathematical formulation of the problem
the Gnielinski correlation [14] for the heat transfer. The friction
factor may be calculated using either the Filonienko correlation or
The DarcyeWeisbach friction factor x for circular tubes is esti-
traditional power-type Blasius [4] or Moody [25] empirical
mated experimentally using the following data reduction equation
correlations.
(Fig. 1)
One of the first explicit approximation to the PKN equation
proposed Colebrook [26,27]. The almost identical correlation was 4Dprw
suggested once again six years later by Konakov [28]. Because of the x¼ (1)
rLw2m
better accuracy of the Konakov explicit representation of the fric-
tion factor in comparison with the Filonienko correlation, Gnie- The friction factor x can also be determined based on the radial
linski recommended it to use it in his correlation for the Nusselt velocity profile obtained experimentally or determined by inte-
number [3,29,30]. grating the momentum conservation equation.
A continuous explicit relationship for the friction factor for fully At first, the velocity distribution and the friction factor will be
developed laminar, transitional and turbulent flows in smooth and determined based on the solution of the momentum conservation
rough pipes developed Churchill [31]. Churchill's formula was equation for turbulent flow when Re > 3000. The turbulent velocity
modified by Schroeder [32], and Rennels and Hudson [33] to get profile wx is obtained by solving the time-averaged momentum
better compatibility with the HagenePoiseuille law and Cole- conservation equation [41e43].
brookeWhite equation. A piecewise friction factor formula for
 
laminar, transition and turbulent flow in smooth pipes was derived 1 d dwx dp
by Joseph and Yang [33]. The data in the transition region is pro- rrðv þ 3 t Þ ¼ (2)
r dr dr dx
cessed by fitting five points with a logistic dose algorithm.
Friction factors and Nusselt numbers for all flow regimes in Equation (2) was obtained from a momentum conservation equa-
round tubes and parallel-plate channels were successfully deter- tion written in cylindrical coordinates with the following
mined using CFD (Computational Fluid Dynamics) modeling assumptions
[34e38]. The RANS (Reynolds-Averaged NaviereStokes) equations  
of the mass and momentum conservation, and SST (Shear Stress vwr vwx v
wr ¼ 0; ¼ 0; ¼ 0; rw0x w0r ¼ 0
Transport) turbulence model were used to study the transition vr vx vx
from laminar to intermittent and turbulent flow or internal flows
which transit from turbulent through intermittent to laminar. The where w0x and w0r are the fluctuating components of the longitu-
friction factors obtained from numerical experiments by Abraham dinal and radial velocity, respectively.
et al. [34,37,38] were approximated by polynomials of various de- The momentum eddy diffusivity 3 t is defined as
grees to get explicit formulas suitable for engineering applications.
A review of published papers shows that for the calculation of vwx
rw0x w0r ¼ r3 t
the friction factor for turbulent tube flow several empirical corre- vr
lations of various accuracy are used. There are various ranges of
Equation (2) is subject to the following boundary conditions
Reynolds numbers, in which these correlations are valid. There is,
therefore, a need to compare the currently used correlations with wx jr¼rw ¼ 0 (3)
the standard PKN equation and other recent experimental studies.
The aim of this study is to find a suitable universal velocity 
profile for turbulent fluid flow in a tube to determine the friction dwx 
¼0 (4)
factor in smooth tubes. An accurate profile of the universal velocity dr r¼0
is also essential in calculating the Nusselt number as a function of
The shear stress t is defined as [41e43].
Prandtl and Reynolds numbers by integrating the time-averaged
equation of energy conservation. dwx dwx  3t
dw
x
The velocity profile will first be determined using the universal t ¼ ðm þ r3 t Þ ¼ rðv þ 3 t Þ ¼ m 1 þ (5)
dr dr v dr
velocity distribution found by Reichardt [39,40], and then the
momentum conservation equation will be applied to determine the Writing the momentum conservation equation for a control volume
112 D. Taler / International Journal of Thermal Sciences 105 (2016) 109e122

of a diameter dw ¼ 2rw and a length Dx (Fig. 1) yields

Zrw
  2
2 2 Dx wm ¼ 2
wx rdr (15)
prw pðxÞ ¼ prw pðx þ DxÞ þ 2prw Dxtw x þ (6) rw
2 0

  Introducing the so called friction velocity ut given by Refs. [41e43]

Dx rw pðx þ DxÞ  pðxÞ
tw x þ ¼ (7) pffiffiffiffiffiffiffiffiffiffiffi
2 2 Dx ut ¼ tw =r (16)
If Dx / 0 then Eq. (7) can be written as the dimensionless variables can be defined in the following way

rw dp pffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffi
tw ¼  (8) yut y tw =r ðrw  rÞ tw =r
2 dx yþ ¼ ¼ ¼ (17)
v v v
Taking into account Eqs. (5) and (8), Eq. (2) can be rewritten in the pffiffiffiffiffiffiffiffiffiffiffi rffiffiffi
form þ rut r tw =r þ rw ut Re x
r ¼ ¼ ; rw ¼ ¼ (18)
v v v 2 8
1 d 2tw
ðrtÞ (9)
r dr rw r rw  y yþ
R¼ ¼ ¼1 þ (19)
rw rw rw
Integration of Eq. (9) with the boundary condition
Rearranging Eq. (16) gives
tjr¼rw ¼ tw (10)
tw ¼ ru2t (20)
gives
Inserting Eq. (20) for tw into Eq. (14) yields
r
t ¼ tw (11)
rw 8u2t 8
x¼ ¼
2 (21)
w2m þ
um
An analysis of expression (11) indicates, that the shear stress t is a
linear function of the radius r (Fig. 1).
where the symbol uþ
m denotes the dimensionless mean velocity
Substitution of Eq. (11) into Eq. (5) leads to
uþ
m ¼ wm =ut (22)
dwx r 1
¼ tw  (12)
dr rw m 1 þ 3 t Equation (21) is used to calculate the friction factor for incom-
v
pressible flow over a flat surface or for fully developed flow in a
Solving Eq. (12) subject to the boundary condition (3) gives the fluid tube [41e43]. The dimensionless velocity uþ can be determined by
velocity as a function of the radius. Based on the radial velocity solving Eq. (12) with the boundary condition (3) or using the ve-
profile wx ðrÞ the friction factor x can be determined. Considering locity distribution uþ determined experimentally. Next, the mean
that velocity uþm is calculated using Eqs. (15) and (22).

dp x rw2m
¼ (13) 3. Determination of the dimensionless mean velocity and
dx dw 2
friction factor
and substituting Eq. (13) into Eq. (8) gives the expression for the
shear stress at the wall First, solving equation (12) subject to the boundary condition (3)
will be discussed in detail. Equation (12) rewritten in the dimen-
rw2m sionless form
tw ¼ x (14)
8 duþ 1 r þ  yþ
¼ þ w (23)
where the mean velocity wm is given by dyþ rw 1 þ 3 t =v

is subject to the boundary condition (3) that takes the form



uþ  ¼0 (24)
yþ ¼0

Using the Euler method to solve Eq. (23) gives

" #
uþ
jþ1
 uþ
i 1 þ  yþ
rw þ  yþ
rw iþ1
i
¼ þ þ ; i
Dyþ 2rw 1 þ ð3 t =vÞjyþ 1 þ ð3 t =vÞjyþ
i iþ1

¼ 1; …; n  1 (25)

To increase the accuracy of the calculation, the right side of Eq. (25)
Fig. 1. Turbulent fluid flow in the tube; p1 and p2 measured pressures at a distance of L is the arithmetic mean of the right-hand side of Eq. (2) that is
for determining the friction factor x. evaluated at yþ
i
and yþ
iþ1
. The Euler method is simple and accurate
 D. Taler / International Journal of Thermal Sciences 105 (2016) 109e122 113

provided that the integration step Dyþ is small enough.   " !

Solving Eq. (25) for uþ
yþ
yields þ 1 þ 1:5ð1 þ RÞ
iþ1 u ¼ ln 1 þ 0:4y þ C 1  exp 
k 1 þ 2R2 11
" # !# (35)
þ  yþ þ  yþ
Dyþ rw rw
yþ yþ
uþ ¼ uþ þ þ i
þ iþ1
; i  exp  ; 0R1
iþ1 i 2rw 1 þ ð3 t =vÞjyþ 1 þ ð3 t =vÞjyþ 11 3
i iþ1

¼ 1; …; n  1 (26)
where k ¼ 0.4 and C ¼ 5.5  (1/k)ln k ¼ 7.8 are constants. The
The boundary condition (24) can be written advantage of Eq. (35) is that it gives the velocity profile uþ
throughout the entire interval 0  r þ  rw
þ without division of the

uþ tube cross-section into subdomains. In contrast to the von K
arma
n

1 ¼0 (27)
law-of-the-wall formulation with three different subregions, a
The finite difference grid is defined as single velocity profile represents the entire region from the wall to
the fully turbulent region.
yþ ¼ ði  1ÞDyþ ; i ¼ 1; …; n (28) If R ¼ 1 then the function (35) reduces to the universal velocity
i
distribution over a plate [39,40].
where the dimensionless spatial step is given by  h  .
uþ ¼ 2:5 ln 1 þ 0:4yþ þ 7:8 1  exp yþ 11
þ  .  i
rw
Dyþ ¼ (29)  yþ 11 exp 0:33yþ (36)
n1

Integration step Dyþ was calculated using the relationship (29) for The formula (36) approximates very well experimental data in the
n ¼ 30,001. Identical values of uþ
iþ1
were obtained for n ¼ 60,001 near-wall-region [39,40] as well as the three subdomain model
and n ¼ 100,001. proposed by von K
arma
n [44].
The mean velocity uþ m is defined as
uþ ¼ yþ ; 0 < yþ  5 ðviscous sublayerÞ
þ þ uþ ¼ 5 ln yþ  3:05; 5  yþ  30 ðbuffer layerÞ
Zrw Zrw uþ ¼ 2:5 ln yþ þ 5:5; yþ  30 ðfully turbulent regionÞ
1 2
uþ
m ¼
2 þ þ þ
2pr u dr ¼
2 r þ uþ dr þ (30)
þ
p rw þ
rw (37)
0 0
Table 1 shows the comparison between the universal velocity
Taking into account that, drþ ¼ dyþ and considering that r þ ¼ rw
þ profiles (37) and (36) proposed by von Ka
rm
an and Reichardt,
if yþ ¼ 0 and rþ ¼ 0, the formula (30) takes the form respectively.
The velocities predicted by the Reichardt equation (36) and von
Zrw
þ Ka
rm
an [44] three-region model (37) agree very well. The
2 
uþ ¼
2 uþ rw
þ
 yþ dyþ (31) maximum relative difference does not exceed 2.15% (Table 1). The
determination of the universal velocity distribution uþ using Eq.
m þ
rw
0
(26) and the friction factor from Eq. (21) must be performed iter-
atively since the dimensionless radius rw þ depends on the friction
Using the trapezoidal rule to calculate the integral (31) numerically
factor x. The starting value of the friction factor x(0) for a given value
yields
of the Reynolds number Re can be calculated using the experi-

þ 
þ  mental correlations for the friction factor that are discussed in
X rw
2Dyþ n1  yþ þ uþ r  yþ
iþ1 w Section 5. A convergent solution is obtained after a few iterations.
uþ
m ¼
2 i iþ1
(32)
þ 2 The universal velocity distribution (36), represented only by a
rw i¼1
single function, was used by Glockner and Naterer [45] to develop
The friction factor x was determined from Eq. (21). special shape functions in a Control-Volume-Based Finite Element
To determine a velocity distribution uþ(yþ) and friction coeffi- Method (CVFEM). The ke3 turbulence model in conjunction with an
cient x the eddy diffusivity for momentum transfer 3 t will be alternative near wall velocity profile, based on the Reichardt uni-
calculated using Reichardt's [39,40] empirical equations, which so versal velocity (36), provided results of high accuracy for turbulent
far are the most commonly used because of their high accuracy channel flow. In contrast to the three-region model of the velocity
proposed by von K
arma
n, using the Reichardt velocity profile does
" !# not require information concerning the location in which region
3t þ yþ
¼k y  yþ
n tanh þ ; yþ  50 (33) the velocity is to be calculated.
v yn

4. Velocity profile and friction factor based on radial velocity

 
3t k 1 profile in turbulent tube flow
¼ yþ ð1 þ RÞ þ R2 ; þ
y > 50 (34)
v 3 2
The friction factor will be calculated for turbulent tube flow
where the constants are: k ¼ 0.4 and yþ
n ¼ 11. when Reynolds numbers are greater than 3000 and less than 107.
The velocity distribution was obtained using Eq. (26) with the The velocity profiles were calculated for four different values of
boundary condition (27) considering empirical Equations (33) and the Reynolds number: 10,000, 30,000, 50,000, and 100,000 (Fig. 2).
(34). The friction factor x was calculated as a function of the Reynolds
The radial velocity profile can also be determined from the number. The number of spatial steps was (n  1) ¼ 30,000.
formula proposed by Reichardt that is based on experimental data An inspection of the results shown in Fig. 2 indicates that the
[39,40]. discrepancy between the velocity calculated from the analytical
114 D. Taler / International Journal of Thermal Sciences 105 (2016) 109e122

Table 1
Comparison of universal velocity profiles uþ(yþ) proposed by Reichardt [39,40] and von Ka

rm

an [44].

yþ uþ


K von Karman's Eq. (37) uþ
R Reichardt's Eq. (36) 3 ¼ ððuþ þ þ
K  uR Þ=uK Þ  100; %

0 0 0 0
1 1 1.0109 1.09
2 2 2.0381 1.91
3 3 3.0504 1.68
4 4 4.0190 0.48
5 5 4.9259 1.48
10 8.4629 8.4281 0.41
15 10.4903 10.5984 1.02
20 11.9287 12.0089 0.67
25 13.0444 12.9868 0.44
30 14.0030 13.7013 2.15
50 15.2801 15.3285 0.32
100 17.0129 17.0831 0.41
200 18.7458 18.7861 0.22
300 19.7595 19.7895 0.15
400 20.4787 20.5035 0.12
500 21.0365 21.0583 0.10
600 21.4923 21.512 0.09
700 21.8777 21.8959 0.08
800 22.2115 22.2286 0.08
900 22.5060 22.5222 0.07
1000 22.7694 22.7849 0.07

expression (35) and obtained using the finite difference method predicted friction factors are consistent with the PKN equation for
with the boundary condition (27) is greater for smaller Reynolds Re  1.106. However, for 31  103  Re  35  106 McKeon et al.
numbers. proposed k ¼ 0.422 and D ¼ 0.537. Substituting these constants into
The friction factor was determined using Eq. (21) with the ve- Eq. (39) yields the following equation:
pffiffiffi
locity profile given by Eq. (35) and with the velocity profile deter- x1=2 ¼ 1:93 logðRe xÞ  0:537. This equation gives a little higher
mined by the finite difference method using Eq. (26). values of x for Reynolds numbers greater than 106. At Re ¼ 1  107 it
A comparison of the results is listed in Table 2. To assess the overpredicts the PKN equation by 2.6%. Considering measurement
accuracy of the friction factors obtained using the radial velocity uncertainties this discrepancy is insignificant. Nagib and Chauhan
profiles, an implicit correlation based on the experimental data [49] corrected the experimental data of McKeon et al. for the pipe
with the use of Eq. (1) will be applied. flow to detect that k ¼ 0.41.
The most popular implicit relation for determining the friction The extensive experimental studies to determine the friction
factor is the Prandtlevon Ka
rma
neNikuradse (PKN) equation factor was carried out by Furuichi et al. [48] within the following
[1,6e8]. range of Reynolds number: 7.075  103  Re  1.762  107. The
working fluid was water at a temperature about 20  C and 70  C.
1  pffiffiffi The results reported by Furuichi et al. are inconsistent with the
pffiffiffi ¼ 2 log Re x  0:8; 3:1  103  Re  3:2  106 (38) results obtained by McKeon et al. for Re  2  105. The friction
x
factors measured by Furuichi et al. [48] are lower at high Reynolds
that is regarded as a reference correlation that provides the best fit numbers than those obtained by McKeon at al. [47]. The difference
to the experimental data of Nikuradse [6,7]. The Prandtlevon comes approximately to 6% at Re ¼ 107. However, the mean devi-
Ka
rma
neNikuradse correlation (38) can be written in an alterna- ation of the experimental data obtained by Furuichi et al. from the
tive universal form friction factors obtained from the PKN equation (38) is much
smaller. In the lower Reynolds number range
1 1  pffiffiffi 7.075  103  Re  2.196  104, the deviation is 1.6% and 2.7% at
pffiffiffi ¼ pffiffiffi ln Re x  D; 3:1  103  Re  3:2  106 (39) the higher Reynolds region 5.158  106  Re  1.762  107.
x 2k 2
The data of Furuichi et al. are easy to process since it is presented
where k ¼ 0.407 is the von Ka
rma
n constant and D ¼ 0.8. The in tabular form.
constants k and D in Eq. (39) proposed by Prandtl were found by The experimental data of Furuichi et al. [48] was used to
curve fitting the smooth-wall tube data of Nikuradse [6,7]. The determine the constants k and D in the PKN equation (39) using the
experiments were conducted on the flow of water over the Rey- nonlinear least squares method. All the experimental data pre-
nolds number range 3.1  103  Re  3.2  106. sented in Ref. [48], covering 66 data points, were taken into ac-
The PKN semi-empirical equation (39) will be considered as a count. The transformation of the PKN equation (39) gives the
standard equation that will be used to assess the accuracy of im- following non-linear function
plicit correlations for the friction factor. In the past 20 years,
  
numerous experimental studies of the turbulent flow in tubes have pffiffiffi
been conducted to determine experimentally new k and D con-
exp 2k 2 p1ffiffix þ D
stants, so that the Prandtl equation better approximates the mea- Re ¼ pffiffiffi ; 7:075  103  Re  1:762  107
x
surement data [46e49]. Swanson et al. [46] used various room
(40)
temperature gasses: carbon dioxide, helium, nitrogen, oxygen,
sulfur hexafluoride and normal liquid helium. The friction factors This function was used to approximate the experimental data
determined by Swanson et al. are in good agreement with the PKN of Furuichi et al. [48]. The least squares fit gives: k ¼ 0.3991 and
equation (38). Compressed air was used by McKeon et al. [47]. The D ¼ 0.8706 for the data at 7.075  103  Re  1.762  107. The
 D. Taler / International Journal of Thermal Sciences 105 (2016) 109e122 115

Fig. 2. Fluid velocity as a function of dimensionless radius calculated using the analytical expression (35) (Reichardt's velocity) and the finite difference solution (26) subject to the
boundary condition (27) (Velocity determined from momentum equation); (a) Re ¼ 10,000, (b) Re ¼ 30,000, (c) Re ¼ 50,000, (d) Re ¼ 100,000.

predicted von Ka
rm
an constant is only by 1.94% lower than the explicit correlations since it covers a broad range of Re starting with
constant k ¼ 0.407 proposed by Prandtl, von Ka
rma
n, and Re ¼ 3000 and quite good approximates experimental data of
Nikuradse. 95% confidence intervals for the coefficient k and D Nikuradse [6,7], McKeon et al. [47], and Furuichi et al. [48]. Adop-
are 0.3958  k  0.4024 and 0.7693  D  0.9719. The substi- tion of the PKN equation (38) as the standard equation for smooth
tution of the estimated constants k ¼ 0.3991 and D ¼ 0.8706 into tubes is fully consistent with the acceptance of the Colebrooke-
pffiffiffi pffiffiffi
Eq. (39) gives: 1= x ¼ 2:0397 logðRe xÞ  0:8706. The resulting White equation as a reference equation for assessing explicit cor-
correlation is very similar to the PKN equation (38). Since the relations for turbulent flow in rough tubes [9e11]. It must be noted
coefficient 2.0397 is greater than 2, and the coefficient 0.8706 is that the ColebrookeWhite equation reduces to the PKN equation
also greater than 0.8, the obtained equation approximating the (38) if the relative roughness of the inner surface of the tube is zero.
experimental data of Furuichi et al. [48] yields very similar fric- Relative differences between the friction factors xPKN calculated
tion factors x as the PKN equation (38). From the analysis of the from the PKN equation (38) and friction factors x calculated based
available experimental data in the literature, it appears that the on the radial velocity distribution were calculated using the
PKN equation (38) well approximates the experimental data for a following formula
single phase, turbulent fluid flow in tubes. Eq. (38) is also
confirmed by recent experimental studies. However, experi-
xðRei Þ  xPKN ðRei Þ
mental studies are sometimes a little inconsistent. This may be ei ¼ 100; % i ¼ 1; …; m (41)
xPKN ðRei Þ
due to some causes, to list a few such as different design of test
rigs, various methods of measuring the average fluid velocity and The symbol m denotes the number of the compared friction
pressure drop, measurement uncertainties and different data factors xi.
reduction procedures. To assess the quality of the approximation of the PKN equation
The PKN equation (38) for the friction factors in smooth tubes in the whole analyzed range of Reynolds numbers using other
will, therefore, be taken as the standard for comparisons with the formulas or experimental correlations, the mean absolute deviation
116 D. Taler / International Journal of Thermal Sciences 105 (2016) 109e122

is calculated as follows constant 1.55, instead of a constant 1.62 in the Filonienko formula.
The correlation (43) and (44) provide lower friction factors as
Pm
i¼1 jei j compared to the correlations of Colebrook and Filonienko.
em ¼ (42)
m
5. Explicit approximations to the solution of the Prandtlevon
Comparison of the friction factor calculated using Eq. (21) and
rma

neNikuradse implicit equation for the friction factor in
Ka
velocity profile (35) as well as Eqs. (21) and (26) is shown in Table 2.
smooth pipes
An analysis of the results listed in Table 2 shows that the friction
factors calculated using the velocity profile (35) proposed by
Empirical correlations (38) and (39) are of limited usefulness for
Reichardt are more accurate than the results obtained using the
wide application to engineering design because they are implicit.
velocity profile determined from the equation of momentum con-
The nonlinear algebraic Eq. (38) can be solved using iterative nu-
servation. In the first case, the relative differences are in the
merical methods or graphically. For this reason, some explicit cor-
range: 5.71%  3  5.63%, when the average deviation is
relations have been proposed in the past 70 years to fit the classical
j3 m j ¼ 2:21%. If the friction factor is calculated based on velocity
rma
neNikuradse equation (38) [1,27,41e43,50,51].
Prandtlevon Ka
determined from the equation of momentum conservation, the
The determination of the friction factor for the fluid flow in a
results are less satisfactory. Relative deviations are in the
smooth pipe is essential to a variety of hydraulic and heat transfer
range: 3.06%  3  32.34%, and the average absolute deviation is
problems encountered in engineering practice.
j3 m j ¼ 7:51%. Noticeable differences are in the range of the smaller
Blasius [4] was the first who found that the friction factor for a
Reynolds number of Re ¼ 3000e20,000. For higher values of Rey-
smooth pipe can approximately be expressed as
nolds numbers, the results are entirely satisfactory.
The friction factors based on the Reichardt velocity profile (35) 0:3164
(Column 3 in Table 2) were approximated using the least squares x¼ ; 3  103  Re  105 (45)
Re0:25
method by the following functions
Equation (45) is based on the seventh-root law developed by Bla-
1 3 7 2 sius for turbulent-velocity distribution
x¼ ; 3  10  Re  10 ; r ¼ 0:9944
ð1:8 log Re  1:47Þ2  1=7
wx y
(43) ¼ ; 3  103  Re  105 (46)
wx jr¼0 rw

1 where y ¼ rw  r is the distance from the inner surface of the tube.

x¼ ; 3  103  Re  107 ; r 2 ¼ 0:9954
ð1:82 log Re  1:55Þ2 Experiments show that at Reynolds numbers above 105 the expo-
nent in Eq. (46) should be smaller than 1/7, i.e. Equation (45) gives
(44)
less accurate results for Re  105.
Another power type experimental relation for the friction factor
The function (43) is similar to the Colebrook correlation, but in the
of fluids flowing in smooth tubes proposed Moody [25].
denominator of the function occurs a constant of 1.47 instead of a
constant 1.5099 appearing in the Colebrook formula. In turn, the 0:184
second function defined by the formula (44) is similar to the cor- x¼ ; 104  Re  106 (47)
Re0:2
relation of Filonienko, but in the denominator of Eq. (44) is a

Table 2
Comparison of the friction factor calculated using Eq. (21) and velocity profile (35) as well as Eqs. (21) and (26).

Reynolds Prandtlevon Ka
rm

aneNikuradse, Equation (21) with Eqs. (32) Relative difference, Equation (21) with Eqs. (32) Relative difference, %
number Re Equation (38) and (35) Equation (e), % and (26) Equation (e), %

3000 0.04352 0.04597 5.63 0.05760 32.34

4000 0.03991 0.04096 2.63 0.04981 24.82
5000 0.03739 0.03770 0.81 0.04486 19.96
6000 0.03550 0.03536 0.42 0.04139 16.57
7000 0.03401 0.03357 1.29 0.03879 14.06
8000 0.03279 0.03215 1.94 0.03677 12.13
9000 0.03176 0.03099 2.45 0.03513 10.60
10,000 0.03088 0.03000 2.85 0.03377 9.36
12,000 0.02944 0.02843 3.45 0.03164 7.46
14,000 0.02830 0.02720 3.86 0.03015 6.55
16,000 0.02736 0.02622 4.18 0.02873 5.01
18,000 0.02657 0.02539 4.41 0.02768 4.18
20,000 0.02588 0.02469 4.59 0.02679 3.51
25,000 0.02452 0.02331 4.92 0.02508 2.28
30,000 0.02348 0.02228 5.13 0.02382 1.44
40,000 0.02197 0.02079 5.37 0.02206 0.39
50,000 0.02089 0.01974 5.49 0.02084 0.26
100,000 0.01799 0.01697 5.69 0.01771 1.58
200,000 0.01564 0.01474 5.71 0.01528 2.27
300,000 0.01446 0.01364 5.68 0.01410 2.51
400,000 0.01371 0.01293 5.66 0.01335 2.63
500,000 0.01316 0.01242 5.63 0.01280 2.69
1  106 0.01165 0.01100 5.54 0.01131 2.85
5  106 0.00898 0.00850 5.33 0.00871 3.06
1  107 0.008103 0.00768 5.24 0.00786 2.97
 D. Taler / International Journal of Thermal Sciences 105 (2016) 109e122 117

Relations of Blasius (45) and Moody (47) cannot represent the

friction factor in a wide range of the Reynolds number since they
1
are linear in the logarithmic coordinate system while the friction x¼  2 ; 3  103  Re  108 (51)
factor for smooth tubes is a nonlinear function of the Reynolds 150:39  152:66
2 log Re0:98865 Re
number (Fig. 3).
The first explicit approximation of Eq. (38) was suggested by
The maximum relative difference between the friction factors
Colebrook in 1939 [26,27].
obtained from Eq. (38) and (51) is less than 0.042%. Correlation (51)
very well approximates the solution to the equation (38). Improved
  accuracy of the correlation (51) has been reached, however, with a
1 Re
pffiffiffi 1:8 log ; 3  103  Re  1  108 (48) more complex form of the function x ¼ x(Re).
x 6:9
Another form of the explicit formula for the friction coefficient
The results obtained from Eqs. (38) and (48) differ by less than for turbulent fluid flow in the tubes of a circular cross section
1.89% for 3  103  Re  1  108. proposed Allen and Eckert [53].
Eq. (40) can be written in another form as

0:432
x ¼ 0:00556 þ ; 1:3  104  Re  1:2  105 (52)
1 3 8 Re0:308
x¼ ; 3  10  Re  1  10 (49)
ð1:8 log Re  1:5099Þ2
The expression (52) was determined based on measurements of
Seven years later in 1946, very similar equation to Eq. (49) was turbulent pipe flow of water for Pr ¼ 7 and Pr ¼ 8 at uniform wall
proposed once again by Konakov [28] and sometimes is cited as the heat flux.
Konakov equation [29,30]. The only difference is the adoption a To find a simple and accurate approximation to the equation of
constant 1.5 instead of a constant 1.5099 appearing in the equation Prandtlevon Ka
rm
aneNikuradse, Eq. (38) was solved in the range
(49) of Colebrook. Equation (49) with the constant equal to 1.5 is 3  103  Re  104 with the step of 100, in the range of the Reynolds
recommended by Gnielinski [29,30] to use for the calculation of the number 104  Re  105 with the step of 1000, and in the range
heat transfer coefficient in the turbulent smooth tube flow. 105  Re  107 with the step of 105. In this way, m ¼ 360 points
A little less accurate explicit expression for the friction factor is (Rei,xi) were generated, which were subsequently approximated by
the Filonienko equation [13]. the method of the least squares using various forms of the
approximating function. The unknown parameters were deter-
mined using the LevenbergeMarquardt method.
1 If the form of the function is the same as in the expression (49)
x¼ ; 3  103  Re  5  107 (50)
ð1:82 log Re  1:64Þ2 found by Colebrook, then the best fit to 360 data points assure the
following expressions
The maximum relative difference between the friction factors
obtained from Eqs. (38) and (50) does not exceed 4.69%. The Filo-
nienko equation is widely used in many textbooks [50,51] and 1
handbooks [30] when the Gnielinski correlation is used for calcu-
x¼ ; 3  103  Re  107 ; r 2
ð1:7765 log Re  1:4053Þ2
lating the Nusselt number in the single-phase tube flow.
Fang et al. [9] offered a closer approximation of Eq. (38), similar ¼ 0:99988
to that proposed by Manadilli [52] for the rough tubes (53)
The correlation (53) compares in the range 3  10  Re  106 3

better with Eq. (38) than the Colebrook Eq. (49).

Using another form of the approximating function, a new and
more accurate correlation was found

1
x¼ ; 3  103  Re  107 ; r 2
ð1:2776 log Re  0:406Þ2:246
¼ 0:999999
(54)
An advantage of the expression (54) is its high accuracy while
maintaining a simple form suitable for engineering calculations.
The function of type x ¼ ½c1 log Re  c2 c3 , where c1, c2, and c3 are
constants, approximates very well not only the PKN equation (38)
but also the experimental data. Approximating the experimental
data of Furuichi et al. [48] using the least squares method yields the
following relationship x ¼ (1.1198log Re  0.2063)2.384,
7.075  103  Re  1.762  107, r2 ¼ 0.99981. The coefficient of
determination r2 is a bit smaller than that of Eq. (54), since the
experimental data are burdened with random measurement errors.
It should be noted that explicit formulas for the calculation of
Fig. 3. Friction factor versus Reynolds number and relative roughness for commercial
the friction factor have also been proposed for the transition flow in
pipes (the diagram was drawn using the modified Churchill approximation [1] of the the pipes [1,33,34,37,38]. Based on the CFD simulation, Abraham
Moody diagram). et al. [38] proposed the following formula
118 D. Taler / International Journal of Thermal Sciences 105 (2016) 109e122

x ¼ 3:03  1012 Re3  3:67  108 Re2 þ 1:46  104 $Re

 0:151; 2:3  103
 Re  4:5  103
(55)
This relationship may be useful in calculating the Nusselt using a
correlation in which there is a friction factor, for example by the
Gnielinski relationship [14].

6. Comparison of explicit formulas for prediction of the

friction factor in turbulent smooth tube flow

First, explicit correlations for the friction factor will be

compared with the friction factors determined using the universal
velocity profile (35) of Reichardt and on the velocity profile ob-
tained from the momentum conservation equation using the
Reichardt eddy diffusivity for momentum transfer (33) and (34).
Explicit formulas of Blasius [4], Moody [25], Colebrook [26], and
Filonienko [13] were taken into account in this comparison.
Relative differences between the friction factors xi ¼ x(Rei)
calculated using explicit correlations and friction factors xi deter-
mined based on the radial velocity distribution were calculated
using the following formula

xðRei Þ  xm ðRei Þ
ei ¼ 100; % i ¼ 1; …; m (56)
xm ðRei Þ

where the mean friction factor xm is defined as

Pm
i¼1 xi
xm ¼ (57)
m
The symbol m denotes the number of the compared friction
factors xi.
Comparison of friction factors calculated using different pro-
cedures and formulas is shown in Fig. 4a in a semi-log coordinate
system and in Fig. 4b in a linear coordinate system.
Fig. 5 presents relative differences ei between various friction
factors calculated using Eq. (56).
An analysis of the results depicted in Figs. 4 and 5 shows that the
friction factor based on the velocity profile (35) proposed by
Reichardt is more accurate than the friction factor based on the
velocity profile determined from the equation of momentum con-
servation in conjunction with Reichardt's expressions (33) and (34)
for eddy diffusivity for momentum transfer. The accuracy of the
friction factor determined based on the velocity profile obtained
Fig. 4. Comparison of calculated friction factors for smooth tubes based on Reichardt's
from the solution of the momentum conservation equation is very universal velocity profile (35) and Reichardt's eddy diffusivity for momentum transfer
excellent, but only for Reynolds numbers greater than about (33) and (34) with the results obtained using the explicit formulas proposed by Blasius
20,000. Fairly good agreement exists between the experimental [4], Colebrook [26], Filonienko [13], Allen and Eckert [53], and Abraham et al. [38]
correlation (52) of AlleneEckert and the explicit formula (50) of (Fig. 4a).

Filonienko. Fig. 4a also shows the friction coefficient calculated by a

formula (55) proposed by Abraham et al. [38] for the transition flow
tubes contains Table 5. The mean absolute deviation jem j was
in tubes. It is obvious that for Reynolds numbers less than 3000 the
determined from Eq. (42).
friction factor must be calculated using the formulas that are valid
From the comparison of the results shown in Tables 3e5 one can
for the transition flow.
see that the PKN equation (38) best approximates the correlation
Then, the accuracy of explicit formulas for calculating the fric-
(51) of Fang, and then the simple correlation (54) proposed in this
tion factors will be assessed by comparison with the equation (38)
paper. An inspection of the results listed in Tables 3 and 4 reveals
of Prandtlevon KarmaneNikuradse. Table 3 shows a comparison of
that the Blasius formula (45) can be applied to the Reynolds
the friction factors calculated using various formulas. Comparison
number of not more than about 100,000.
of the relative differences between the friction factors calculated
Also remaining explicit correlations in Table 5 give quite satis-
using explicit correlations and friction factors determined from the
factory results, especially when taking into account their simple
solution of the PKN equation (38) presents Table 4. The relative
form.
differences were calculated using Eq. (41). A list of explicit formulas
Comparing friction factors calculated using explicit correlations
used for calculating the friction factor for turbulent flow in smooth
 D. Taler / International Journal of Thermal Sciences 105 (2016) 109e122 119

All the proposed relationships very well approximate the PKN

equation (38). Correlations (43) and (44) based on the radial ve-
locity profile gives slightly lower values of the friction factors than
the PKN equation.
Comparison of correlations for the friction factor proposed in
the paper and Filonienko correlation with experimental data of Li
et al. [17] is shown in Fig. 7.
The experimental investigation was carried out for the Reynolds
number changing in the range from Re ¼ 2500 to Re ¼ 90,000. The
Filonienko correlation agrees better with the experimental data of
Li et al. [17] for lower Reynolds numbers while for Reynolds
numbers greater than 20,000 the relationships (43) and (44) pro-
posed in this paper have a better compatibility with the experi-
mental data.
A similar correspondence between other experimental data of Li
et al. [18] and the results of calculation using explicit formulas (43),
(44), (50), and (54) illustrates Fig. 8. Experimental results of Ji et al.
[19] in the range of Reynolds number from Re ¼ 8000 to
Re ¼ 90,000 are compared with the same explicit correlations for
calculating the friction factor (Fig. 9).
As in the comparisons shown in Figs. 7 and 8, the relationships
(43) and (44) exhibit better compatibility with the experimental
data for Reynolds numbers greater than 20,000 than the correlation
Fig. 5. Comparison of the relative differences given by Eq. (56) for the friction factors
(54), and the correlation (50) of Filonienko.
calculated using explicit correlations.

Table 3
Comparison of friction factors obtained from the solution of the PKN equation (38) with explicit formulas for calculating the friction factor.

Reynolds Prandtlevon Ka
rm

aneNikuradse, Blasius, Equation Moody Equation Colebrook, Filonienko, Present, Equation Present, Equation
number Re Equation (38) (45) (47) Equation (48) Equation (50) (53) (54)

3000 0.04352 0.04275 0.03710 0.04434 0.04549 0.04392 0.04355

4000 0.03991 0.03979 0.03503 0.04042 0.04138 0.04010 0.03991
5000 0.03739 0.03763 0.03350 0.03773 0.03857 0.03747 0.03739
7000 0.03401 0.03459 0.03132 0.03415 0.03483 0.03397 0.03400
10,000 0.03088 0.03164 0.02916 0.03089 0.03144 0.03077 0.03087
16,000 0.02736 0.02813 0.02655 0.02725 0.02767 0.02720 0.02735
20,000 0.02588 0.02661 0.02539 0.02575 0.02612 0.02572 0.02588
30,000 0.02348 0.02404 0.02341 0.02332 0.02361 0.02332 0.02348
50,000 0.02089 0.02116 0.02114 0.02071 0.02093 0.02075 0.02089
100,000 0.01799 0.01779 0.01840 0.01783 0.01797 0.01789 0.01800
300,000 0.01446 0.01352 0.01477 0.01435 0.01442 0.01443 0.01447
500,000 0.01316 0.01190 0.01334 0.01307 0.01312 0.01316 0.01317
1  106 0.01165 0.01000 0.01161 0.01159 0.01163 0.01168 0.01165
5  106 0.008981 0.00669 0.00841 0.00899 0.00899 0.0091 0.00898
1  107 0.008103 0.00563 0.00733 0.00813 0.00813 0.00822 0.00810

Table 4
Comparison of the relative differences between the friction factors calculated using explicit correlations and friction factors determined from the solution of the PKN equation
(38).

Reynolds number Re Blasius, Equation (45) Moody Equation (47) Colebrook, Equation (48) Filonienko, Equation (50) Present, Equation (53) Present, Equation (54)

3000 1.7638 14.75 1.8898 4.6864 0.9136 0.0597

4000 0.3067 12.23 1.2916 3.7007 0.4827 0.0138
5000 0.6258 10.42 0.9011 3.1758 0.2107 0.0098
7000 1.7071 7.92 0.4138 2.4937 0.1137 0.0309
10,000 2.4505 5.57 0.0097 1.8952 0.3628 0.0384
16,000 2.8322 2.97 0.3808 1.26864 0.5723 0.0328
20,000 2.7925 1.92 0.5196 1.0257 0.6330 0.0262
30,000 2.3758 0.31 0.7092 0.6595 0.6915 0.0108
50,000 1.2799 1.17 0.8526 0.3151 0.6869 0.0116
100,000 1.1001 2.28 0.9170 0.0118 0.5755 0.0401
300,000 6.5277 2.13 0.8070 0.1920 0.2322 0.0667
500,000 9.5683 1.35 0.6950 0.2044 0.0265 0.0680
1  106 14.08 0.30 0.7133 0.50101 0.2827 0.0569
5  106 25.50 6.31 0.0697 0.1172 1.0790 0.0255
1  107 30.56 9.60 0.3463 0.2872 1.4387 0.0838

proposed in this paper, with the friction factors obtained from the Assessing the results of the comparisons shown in Figs. 7e9
solution of the PKN equation (38) is depicted in Fig. 6. must be borne in mind that the experimental data for small
120 D. Taler / International Journal of Thermal Sciences 105 (2016) 109e122

Table 5
Summary of explicit formulas for calculating the friction factor for turbulent flow in smooth tubes.

Eq. number Equation Author, reference Range Error range, and mean absolute error

(43) x¼ 1 Present paper 3  103  Re  107 1.99%  3  0.2%, j3 m j ¼ 0:51%

ð1:8 log Re1:47Þ2
3 7
(44) x¼ 1 Present paper 3  10  Re  10 2.49%  3  0.64%, j3 m j ¼ 0:47%
ð1:82 log Re1:55Þ2
(45) x¼ 0:3164 Blasius [4] 3  103  Re  105 1.76%  3  2.84%, j3 m j ¼ 1:06%
Re0:25
(47) x¼ 0:184 Moody [25] 104  Re  106 5.57%  3  2.48%, j3 m j ¼ 1:37%
Re0:2
(48) x¼ 1
2 Colebrook [26] 3  103  Re  107 0.92%  3  1.89%, j3 m j ¼ 0:46%
Re
1:8 log 6:9

(50) x¼ 1 Filonienko [13] 3  103  Re  107 0.2%  3  4.69%, j3 m j ¼ 0:87%

ð1:82 log Re1:64Þ2
(51) x¼  1
2 Fang et al. [9] 3  103  Re  108 0.00034%  3  0.042% j3 m j ¼ 0:0095%
2 log 150:39
152:66
Re0:98865 Re

(53) x¼ 1 Present paper 3  103  Re  107 0.70%  3  1.44%, j3 m j ¼ 0:70%

ð1:7765 log Re1:4053Þ2
(54) x¼ 1 Present paper 3  103  Re  107 0.084%  3  0.069%, j3 m j ¼ 0:034%
ð1:2776 log Re0:406Þ2:246

Reynolds numbers when Re is less than 10,000 are burdened with 7. Conclusions
large uncertainties. This is due to big uncertainties occurring in
measuring the pressure drop over a short tube with a small fluid The explicit formulas of Blasius, Moody, and AlleneEckert for
mass flow rate [17,18,20]. the friction factor in the smooth tube turbulent flow are simple, but
By comparisons of explicit expressions for calculating the fric- they can only be used in the limited ranges of the Reynolds number.
tion with the standard equation of Prandtlevon The range of validity of the Blasius equation is: 3  103  Re  105.
Ka
rma
neNikuradse (PKN) and experimental results, one can draw The Moody correlation gives satisfactory results for a Reynolds
some conclusions. Because of the high accuracy and simple form of number ranging from 104 to 106. However, the relative difference
Eqs. (49), (53) and (54) they may be used when calculating the heat for the friction factor regarding the PKN equation is equal to 5.57%
transfer coefficient for turbulent flow in smooth tubes. The highest at Re ¼ 104. Only at Re ¼ 2  104 the relative difference decreases
accuracy among the explicit formulas for the friction factor to 1.52%. The AlleneEckert equation can be used in a narrow
analyzed in this study, have correlations (51) and (54). The range of the Reynolds number: 1.3  104  Re  1.2  105.
advantage of the correlation (54) is also a simple form. Also, the The universal velocity profile in the smooth tube turbulent flow
accuracy of the simple correlations: (43), (44), (49), (50) and (53) is proposed by Reichardt can be used to determine the friction factor.
sufficient for engineering calculations, particularly when consid- Two simple formulas: x ¼ (1.8log Re  1.47)2 and
ering the uncertainty occurring at the experimental determination x ¼ (1.82log Re  1.55)2 for the friction factor were derived based
of the correlation (38). These correlations are valid in a broad range on the radial velocity profile of Reichardt. Both correlations are
of changes of the Reynolds number. valid in the range of Reynolds numbers from Re ¼ 3  103 to
Re ¼ 1  107. The absolute value of the relative mean deviation
between proposed expressions for the friction factor and the im-
plicit formula of Prandtlevon Ka
rma
neNikuradse is about 0.5%.

Fig. 6. Comparing of friction factors calculated using explicit correlations proposed in Fig. 7. Comparison of correlations for the friction factor proposed in the paper and
this paper with the friction factors obtained from the solution of the PKN equation. Filonienko correlation with experimental data of Li et al. [17].
 D. Taler / International Journal of Thermal Sciences 105 (2016) 109e122 121

x ¼ ½c1 log Re  c2 c3 , where c1, c2, and c3 are constants, approxi-
mates very well not only the PKN equation but also the experi-
mental data.
Also, the relationships of the form x ¼ ½c1 log Re  c2 2 such as
explicit formulas of Colebrook, Filonienko, and the formula
x ¼ (1.7765log Re  1.4053)2 proposed in this paper may be used
to calculate the friction factor. They show good agreement with the
PKN equation as well as with the experimental data. Although the
relative differences between the friction factors calculated using
these correlations and the PKN equation are slightly larger, their
accuracy may be taken as enough when considering the scattering
of experimental data.

References

[1] Rennels DC, Hudson HM. Pipe Flow. A Practical and Comprehensive Guide.
Hoboken: AIChE-Wiley; 2012.
[2] Petukhov BS. Heat transfer and friction in turbulent pipe flow with variable
physical properties. In: Hartnett JP, Irvine TF, editors. Advances in Heat
Transfer, vol. 6. New York: Academic Press; 1970. p. 503e64.
[3] Gnielinski V. On heat transfer in tubes. Int. J. Heat Mass Transfer 2013;63:
134e40.
€
[4] Blasius PRH. Das Ahnlichkeitsgesetz bei Reibungsvorg€angen in Flüssigkeiten.
VDI Forschungsh. 1913;131:1e41.
[5] Sheikholeslami M, Hatami M, Jafaryar M, Farkhadnia F. Thermal management
Fig. 8. Comparison of correlations for the friction factor proposed in the paper and of double-pipe air to water heat exchanger. Energy Build. 2015;88:361e6.
Filonienko correlation with experimental data of Li et al. [18]. [6] Nikuradse J. Gesetzma €ßigkeit der turbulenten Stro €mung in glatten Rohren.
Forsch. Arb. Ing. Wes 1932;356 (English translation, NASA TT F-IO 359, 1966).
[7] Nikuradse J. Stro €mungsgesetze in rauhen Rohren, VDI Forschungsheft 361,
Beilage zu Forsch. Arb. Ing. Wes., Ausgabe B, Band 4. Berlin: VDI Verlag; July/
The proposed formulas also approximate very well the experi- August 1933 (English translation: Laws of flow in rough pipes, NACA TM 1292,
mental data. The good accuracy of the correlations based on the 1950).
universal velocity profile of Reichardt indicates indirectly that this [8] Prandtl L. In: Tollmien W, Schlichting H, Go €rtler H, editors. Gesammelte
Abhandlungen, vol. 2. Berlin: Springer Verlag; 1961.
radial velocity profile can be used in determining the temperature [9] Fang X, Xu Y, Zhou Z. New correlations of single-phase friction factor for
distribution and Nusselt number in turbulent flow in smooth tubes. turbulent pipe flow and evaluation of existing single-phase friction factor
A new correlation for the friction factor characterized by a correlations. Nucl. Eng. Des. 2011;241:897e902.
[10] Geni
c S, Arandjelovi

c I, Kolendi

c P, Jari

c M, Budimir N, Geni
c V. A review of
simple form and high accuracy is proposed in the paper. The
explicit approximations of Colebrook's equation. FME Trans. 2011;39:67e71.
explicit expression of the form x ¼ (1.2776log Re  0.406)2.246, University of Belgrad, Faculty of Mechanical Engineering, Belgrad.
valid in the range from Re ¼ 3  103 to Re ¼ 1  107, was found to [11] Winning HK, Coole T. Explicit friction factor accuracy and computational ef-
ficiency for turbulent flow in pipes. Flow Turbul. Combust. 2013;90:1e27.
calculate the friction factor for the turbulent flow in smooth tubes
[12] Petukhov BS, Kirillov VV. The problem of heat exchange in the turbulent flow
The maximum absolute deviation of this correlation from the PKN of liquids in tubes. Teploenergetika 1958;5(4):63e8 (in Russian).
equation does not exceed 0.084%. The function of the type [13] Filonienko GK. Friction factor for turbulent pipe flow. Teploenergetika
1954;1(4):40e4 (in Russian).
[14] Gnielinski V. Neue Gleichungen für den Wa €rme- und den Stoffübergang in
turbulent durchstro €mten Rohren und Kan€ alen. Forsch Ingenieurwes
1975;41(1):8e16 (Translation into English: V. Gnielinski, New equations for
heat and mass transfer in turbulent pipe flow, Int Chem Eng 16 (1976)
359e368).
[15] Mirth DR, Ramadhyani S. Correlations for predicting the air-side Nusselt
numbers and friction factors in chilled-water cooling coils. Exp. Heat Transfer
1994;7:143e62.
[16] Fernando P, Palm B, Ameel T, Lundqvist P, Granryd E. A minichannel
aluminium tube heat exchanger e part I: evaluation of single-phase heat
transfer coefficients by the Wilson plot method. Int. J. Refrigeration 2008;31:
669e80.
[17] Li XW, Meng JA, Li ZX. Experimental study of single-phase pressure drop and
heat transfer in micro-fin tube. Exp. Therm. Fluid Sci. 2007;32:641e8.
[18] Li XW, Meng JA, Li ZX. Roughness enhanced mechanism for turbulent
convective heat transfer. Int. J. Heat Mass Transfer 2011;54:1775e81.
[19] Li XW, Meng JA, Guo ZY. Turbulent flow and heat transfer in discrete double
inclined ribs tube. Int. J. Heat Mass Transfer 2009;52:962e70.
[20] Ji WT, Zhang DC, He YL, Tao WQ. Prediction of fully developed turbulent heat
transfer of internal helically ribbed tubes e an extension of Gnielinski equa-
tion. Int. J. Heat Mass Transfer 2012;55:1375e84.
[21] Raj R, Lakshman NS, Mukkamala Y. Single phase flow heat transfer and
pressure drop measurements in doubly enhanced tubes. Int. J. Therm. Sci.
2015;88:215e27.
[22] Zhang Z, Yang W, Guan Ch, Ding Y, Li F, Yan H. Heat transfer and friction
characteristics of turbulent flow through plain tube inserted with rotor-
assembled strands. Exp. Therm. Fluid Sci. 2012;38:6433e9.
[23] Hojjat M, Etemad S Gh, Bagheri R, Thibault J. Convective heat transfer of non-
Newtonian nanofluids through a uniformly heated circular tube. Int. J. Therm.
Sci. 2011;50:525e31.
[24] Corcione M, Cianfrini M, Quintino A. Heat transfer of nanofluids in turbulent
pipe flow. Int. J. Therm. Sci. 2012;56:58e69.
Fig. 9. Comparison of correlations for the friction factor proposed in the paper and [25] Moody LF. Friction factors for pipe flow. Trans. ASME 1944;66:671e84.
Filonienko correlation with experimental data of Ji et al. [20]. [26] Colebrook CF. Turbulent flow in pipes, with particular reference to the
122 D. Taler / International Journal of Thermal Sciences 105 (2016) 109e122

transition region between the smooth and rough pipe laws. J. Inst. Civ. Eng. Geschwindigkeitsverteilung in glatten Leitungen. Z Angew. Math. Mech.
1938e1939;11(4):133e56. 1951;31(7):208e19.
[27] Finnemore EJ, Franzini JB. Fluid Mechanics with Engineering Applications. [40] Reichardt H. The principles of turbulent heat transfer (Transl. by P.A. Scheck).
tenth ed. Boston: McGraw-Hill; 2002. In: Hartnett JP, editor. Recent Advances in Heat and Mass Transfer. McGraw-
[28] Konakov PK. A new correlation for the friction coefficient in smooth tubes. Hill; 1961. p. 223e52.
Izvestija AN SSSR 1946;51(7):503e6. Vol. LI. [41] Burmeister LC. Convective Heat Transfer. second ed. New York: John Wiley &
[29] Gnielinski V. Ein neues Berechnungsverfahren für die Wa €rmeübertragung im Sons; 1993.
Übergangsbereich zwischen laminarer und turbulenter Rohrstro €mung. Forsch [42] Lienhard V JH, Lienhard IV JH. Heat Transfer. fourth ed. Mineola-New York:
Ingenieurwes Eng. Res. 1995;61(9):240e8. Dover Publications; 2011.
[30] Gnielinski V. Durchstro €mte Rohre, Abschnitt G1, VDI-Wa €rmeatlas 11., bear- [43] Kakaç S, Yener Y. Convective Heat Transfer. second ed. Boca Raton: CRC Press;
beitete und erweiterte Auflage. Berlin-Heidelberg: Springer-Vieweg; 2013. 1995.
p. 785e91. [44] von Ka
rma

n Th. The analogy between fluid friction and heat transfer. Trans.
[31] Churchill SW. Frictionefactor equation spans all fluid-flow regimes. Chem. ASME 1939;61:705e10 (Th. von Ka
rm
€
an, Mechanische Ahnlichkeit und Tur-
Eng. 1977;84:91e2. bulenz, Ges. der Wiss. zu Go € tt., Nachrichten, Math. Phys. Kl., 1930, 58e76).
[32] Schroeder DW. A Tutorial on Pipe Flow Equations. Carlisle, Pennsylvania: [45] Glockner PS, Naterer GF. Near-wall velocity profile with adaptive shape
Stoner Associates, Inc.; August 16, 2001. functions for turbulent forced convection. Int. Commun. Heat Mass Transfer
[33] Joseph DD, Yang BH. Friction factor correlations for laminar, transition and 2005;32:72e9.
turbulent flow in smooth pipes. Phys. D 2010;239:1318e28. [46] Swanson CJ, Julian B, Ihas GG, Donelly RJ. Pipe flow measurements over a wide
[34] Abraham JP, Tong JCK, Sparrow EM. Breakdown of laminar pipe flow into range of Reynolds numbers using liquid helium and various gases. J. Fluid
transitional intermittency and subsequent attainment of fully developed Mech. 2002;461:51e60.
intermittent or turbulent flow. Numer. Heat Transfer A 2008;54:103e15. [47] McKeon RJ, Zagarola MV, Smits AJ. A new friction factor relationship for fully
[35] Minkowycz WJ, Abraham JP, Sparrow EM. Numerical simulation of laminar developed pipe flow. J. Fluid Mech. 2005;538:429e43.
breakdown and subsequent intermittent and turbulent flow in parallel plate [48] Furuichi N, Terao Y, Wada Y, Tsuji Y. Friction factor and mean velocity profile
channels: effects of inlet velocity profile and turbulence intensity. Int. J. Heat for pipe flow at high Reynolds numbers. Phys. Fluids 2015;27:095108-
Mass Transfer 2009;52:4040e6. 1e095108-15.
[36] Lovik RD, Abraham JP, Minkowycz WJ, Sparrow EM. Laminarization and tur- [49] Nagib HM, Chauhan KA. Variations of von Ka
rm
an coefficient in canonical
bulentization in a pulsatile pipe flow. Numer. Heat Transfer A 2009;56: flows. Phys. Fluids 2008;20:101518-1e101518-10.
861e79. [50] Holman JP. Heat Transfer. tenth ed. Boston: McGraw Hill; 2010.
[37] Abraham JP, Sparrow EM, Tong JCK, Bettenhausen DW. Internal flows which [51] Kreith F, Manglik RM, Bohn MS. Principles of Heat Transfer. seventh ed.
transist from turbulent through intermittent to laminar. Int. J. Therm. Sci. Stamford: Cengage Learning; 2011.
2010;49:256e63. [52] Manadilli G. Replace implicit with sigmoidal functions. Chem. Eng. J.
[38] Abraham JP, Sparrow EM, Minkowycz WJ. Internal-flow Nusselt numbers for 1997;104:129e32.
the low-Reynolds-number end of the laminar-to-turbulent transition regime. [53] Allen RW, Eckert ERG. Friction and heat-transfer measurements to turbulent
Int. J. Heat Mass Transfer 2011;54:584e8. pipe flow of water (Pr ¼ 7 and 8) at uniform wall heat flux. J. Heat Transfer
[39] Reichardt H. Vollst€andige Darstellung der turbulenten 1964;86:301e10.