Sei sulla pagina 1di 8

Technical Note

Approximate Analytical Solutions


for the Colebrook Equation
Ali R. Vatankhah 1
Downloaded from ascelibrary.org by UNIVERSITY OF WOLLONGONG on 03/15/18. Copyright ASCE. For personal use only; all rights reserved.

Abstract: Friction factor determination is important for modeling fluid flows in pipes. The Colebrook equation has been widely used for
estimating the pipe friction factor in fully developed turbulent regime. Because of the implicit nature of Colebrook equation, various
regression-based approximations, Lambert W-function-based solutions, series-based solutions, and analytical approximations are developed
for explicitly determining the friction factor. This study focuses on approximate analytical solutions for the Colebrook equation with a
minimum number of natural logarithms and noninteger powers (lower computational cost). For this, a new mathematically equivalent
representation of the Colebrook equation is presented. This form consists of two nonlinear equations which are very well suited for
developing the analytical solutions for the friction factor. The simple analytical solutions developed in this study with the maximum relative
errors less than 0.85, 0.25, 0.054, and 0.0028% (solutions with different degrees of accuracy) are among the most accurate analytical ap-
proximations to the Colebrook equation. Simple form and superior efficiency of the proposed solutions make them preferable to currently
available approximate solutions to the Colebrook equation. DOI: 10.1061/(ASCE)HY.1943-7900.0001454. © 2018 American Society of
Civil Engineers.
Author keywords: Colebrook equation; Friction factor; Accuracy; Analytical solution; Noniterative approximation; Lambert W-function;
Mathematically equivalent.

Introduction solution of the Colebrook equation with a minimum number of


natural logarithms and noninteger powers (lower computational
The friction coefficient is a key parameter to determine the pres- cost) is preferable. Many different approximate solutions of the
sure drop in pipe systems. This coefficient is extensively used Colebrook equation have been developed over the years. Moderate
for design of pipes in irrigation, in civil engineering for calculation accurate approximations with a maximum relative error less than
of water-distribution systems, ventilation systems, and chemical 2.5% are proposed by several researchers; among them: Swamee
and petroleum processes. The standard Colebrook–White equation and Jain (1976), Churchill (1973, 1977), and Manadilli (1997).
for friction calculations in pipes is implicit in nature. This equation More accurate approximations with a maximal relative error less
is used to model turbulent fluid flow in hydraulically smooth and than 1.5% are also proposed by Zigrang and Sylvester (1982),
rough pipes and the transient zone between them. The Colebrook– Haaland (1983), Sonnad and Goudar (2006) and Papaevangelou
White equation was first introduced in 1939 by Colebrook, and et al. (2010). Many accurate approximations with errors less
its nonlinear form, hindering its direct analytical solution, still than 0.5%, are proposed by several researchers such as Chen
attracts the attention of water engineers and mathematical scien- (1979), Barr (1981), Zigrang and Sylvester (1982), Serghides
tists. Because White was not actually a coauthor of Colebrook in (1984), Romeo et al. (2002), Buzzelli (2008), Vatankhah and
development of this equation for pipe friction calculations, the Kouchakzadeh (2008, 2009), and Vatankhah (2014). Many of these
Colebrook–White equation is also known as the Colebrook equa- approximate solutions are primarily regression-based with high
tion (Colebrook 1939). computational cost. For more detailed and extensive reviews of
The Colebrook equation is generally used to estimate the these approximate solutions and their performance, the reader is
Darcy–Weisbach friction coefficient for fluid flows in filled pipes. referred to Zigrang and Sylvester (1982), Brkić (2011a), and Brkić
The Colebrook equation is implicit with respect to the friction co- and Ćojbašić (2017).
efficient and cannot be solved in terms of elementary functions and An explicit solution of the Colebrook equation that represents
thus should be solved iteratively or using approximate solutions. low computational cost and displays reasonable accuracy should be
The iterations for friction factor calculations during the numerical applied to calculations of large pipe networks. Current research pri-
simulations may require long computing times owing to high spa- marily focuses on approximate, noniterative, nonregression-based,
tial and temporal resolutions. Therefore, any effort for presenting a analytical solutions for the Colebrook equation. The Colebrook
simple full-range solution for the friction factor with reasonable equation has an exact analytical solution in terms of the Lambert
accuracy would be of practical importance. A simple and accurate W-function which can simply be obtained using the symbolic
solver provided by Maple mathematical software. But the Lambert
1
W-function itself is an implicit function and should be solved by
Associate Professor, Dept. of Irrigation and Reclamation Engineering, endless form expressions, using the series expansions or other
University College of Agriculture and Natural Resources, Univ. of Tehran,
approximations. Computing the Lambert W-function can also be
P.O. Box 4111, 31587-77871 Karaj, Iran. E-mail: arvatan@ut.ac.ir
Note. This manuscript was submitted on June 28, 2017; approved on implemented using a commercial mathematical software such as
November 10, 2017; published online on March 15, 2018. Discussion per- Mathematica, MathCad, or Maple, but commercial software pack-
iod open until August 15, 2018; separate discussions must be submitted for ages are not always available. Moreover, computational speed for
individual papers. This technical note is part of the Journal of Hydraulic nonlinear systems should be accounted for when the intensive
Engineering, © ASCE, ISSN 0733-9429. computation of the friction factor is required in optimization and

© ASCE 06018007-1 J. Hydraul. Eng.

J. Hydraul. Eng., 2018, 144(5): 06018007


 
simulation problems. Different Lambert W-function-based solu- 1 2.51 ε=D
tions are proposed for the Colebrook equation (Sonnad and pffiffiffi ¼ −2 log pffiffiffi þ ð1Þ
f R f 3.71
Goudar 2004; More 2006; Sonnad and Goudar 2007;
Nandakumar 2007; Brkić 2011b, 2012; Mikata and Walczak The friction factor, f, is a function of relative surface roughness,
2015; Biberg 2017). Some of these solutions have limitation ε=D (ε = average height of protrusion of inner pipe surface and
in applicability for high values of the Reynolds number and D = inner diameter of pipe), and the Reynolds number, R.
relative roughness (Sonnad and Goudar 2004). Generally, the The Colebrook equation is empirical and therefore its accuracy
Lambert W-function-based solutions require the evaluation of can be disputed. However, an accurate solution to the Colebrook
terms with values outside the ranges that can be expressed on equation is itself important, because too crude approximations
most computers (Brkić 2012). When the Reynolds number is may affect the repeatability of numerical results (Clamond 2009).
large, the Lambert W-function-based solutions are not conve- It is assumed that this equation provides sufficient accuracy for
Downloaded from ascelibrary.org by UNIVERSITY OF WOLLONGONG on 03/15/18. Copyright ASCE. For personal use only; all rights reserved.

nient for numerical computations attributable to overflow errors hydraulic engineering applications. The frictional head loss can be
(Sonnad and Goudar 2004). Brkić (2012) compared various computed by the Darcy–Weisbach equation if the friction factor, f,
implementations of the Lambert W-function and showed that is previously accurately determined. The implicit Colebrook equa-
some of these solutions cannot yield accurate results using the tion is assumed to describe the friction factor, f, for a wide range of
modern computer. Brkić (2012) also identified some constraints turbulent flow regimes (4,000 ≤ R ≤ 108 and 0 ≤ ε=D ≤ 0.1). The
in the applicability of certain Lambert W-function-based solu- Colebrook equation is a combination of the Prandtl equation for
tions. Considering both accuracy and computational cost, the fluid flow in hydraulically smooth pipes and the von Karman equa-
Lambert W-function-based solutions are not very efficient over tion for fluid flow in hydraulically rough pipes. The implicit
the entire practical range of the Reynolds number and relative Colebrook equation is obtained by applying a logarithm matching
technique for the Prandtl and von Karman equations and can be
roughness.
used for both hydraulically smooth and hydraulically rough zones
An approximate solution to the Colebrook equation, referred to
and for the transient zone between them. For ε ¼ 0, the Colebrook
as the nth formula, was recently proposed by Mikata and Walczak
equation becomes the Prandtl equation for fluid flow in hydrauli-
(2015). This solution is derived from an exact analytical expression
cally smooth pipes.
for the Lambert W-function. The maximum error of the solution
The Colebrook equation can be used for the solution of three
is 1.62% for n ¼ 1 and is equal to 0.28% for n ¼ 2. These ap-
types of problems: (1) hydraulic problems in which head loss are
proximations have high computational cost because they contain unknown; (2) those in which discharge is unknown; and (3) those
many logarithmic terms (Clamond 2009; Giustolisi et al. 2011, in which the pipe diameter is unknown. Some researchers modify
Brkić 2017). the Colebrook equation for gas flow calculation by using the 2.825
A fast and accurate solution of the Colebrook equation is constant instead of 2.51 (Haaland 1983).
necessary for intensive computations (simulations of large pipe
systems). Sonnad and Goudar (2006) developed an analytical
approximate solution, based on a reformulation of the Colebrook New Mathematically Equivalent Form of the
equation. The mathematically equivalent representation of the Colebrook Equation
Colebrook equation proposed by Sonnad and Goudar (2006) is
the only available clever reformulation of the Colebrook equation To facilitate derivation of the equivalent form of the Colebrook
for deriving accurate analytical solutions for the friction factor. equation, Eq. (1) can be written as
Vatankhah and Kouchakzadeh (2008, 2009) using a curve fitting
x ¼ −a lnðbx þ cÞ ð2Þ
technique and a quadratic approximation of the Taylor series, im-
proved the coefficients from the analytical approximation proposed in which
by Sonnad and Goudar (2006) in a way that maximum error
1 2 2.51 ε=D
decreases from 1 to 0.15%. x ¼ pffiffiffi ; a¼ ; b¼ ; c¼ ð3Þ
This research focuses on simple and accurate analytical solu- f lnð10Þ R 3.71
tions of the Colebrook equation based on clever reformulation of
The computation of logarithm functions in most computer
the Colebrook equation. For this, a mathematically equivalent rep-
languages is based on the computation of a natural logarithm. Thus,
resentation of the Colebrook equation is derived. This new equiv-
for computational reasons, the Colebrook Eq. (1) is often written
alent form is simple and well suited for accurately estimating the in the natural base by considering the coefficient a ¼ 2= lnð10Þ ¼
friction factor. As will be shown, the proposed alternate formu- 0.8686, in which ln is the natural logarithm.
lation in this study can be used to obtain exact analytical solutions The following developments are on the basis of the general
to the original Colebrook equation, and can significantly contribute implicit form of Eq. (2) with unknown variable x. When the original
to the currently available explicit solutions. The following sections Colebrook Eq. (1) is intended, the quantities of Eq. (3) should be
present the Colebrook equation and its mathematically equivalent substituted into Eq. (2).
form. Solutions of equivalent Colebrook equation using different Eq. (2) can be expanded as
accurate approximations are then presented, followed by an error  
analysis and conclusions. c
x ¼ −a lnðbÞ − a ln x þ ð4Þ
b

Adding c=b in both sides of Eq. (4) yields


Colebrook Equation  
c c c
The fluid flows with a Reynolds number higher than 4,000 are x þ ¼ − a lnðbÞ − a ln x þ ð5Þ
b b b
known as turbulent flows. The Colebrook equation which is widely
used for the calculation of turbulent friction factor, f, in pipes is By taking the natural logarithm from both sides of Eq. (5),
expressed as (Colebrook 1939): one gets

© ASCE 06018007-2 J. Hydraul. Eng.

J. Hydraul. Eng., 2018, 144(5): 06018007


δ ¼ lnðd − aδÞ ð6Þ Approximate Solutions for δ

in which
First Linear Approximation
c
d ¼ − a lnðbÞ ð7Þ Approximating the logarithm term in Eq. (12) by the linear term of
b the Maclaurin series expansion ½lnð1 − xÞ ≅ −x, this equation
  may be written as
c
δ ¼ ln x þ ð8Þ
b δ ¼ lnðdÞ − aδ=d ð13Þ
From Eqs. (6) and (8), x can be expressed as Solving Eq. (13) for δ yields
c
Downloaded from ascelibrary.org by UNIVERSITY OF WOLLONGONG on 03/15/18. Copyright ASCE. For personal use only; all rights reserved.

x ¼ d − aδ − ð9Þ δ¼
lnðdÞ
ð14Þ
b 1 þ a=d
Substituting d from Eq. (7) into Eq. (9) yields

x ¼ −a lnðbÞ − aδ ð10Þ Second Linear Approximation


Similarly by approximating the logarithm term in Eq. (12) by
The relation for x [Eq. (10)] and the equation for δ [Eq. (6)] lnð1 − xÞ ≅ −mx, (m is a positive quantity) this equation can be
make up the new equivalent form of the Colebrook equation. The written as
equivalence of the original Colebrook Eq. (2), and Eqs. (6) and (10)
can simply be shown by substituting δ from Eq. (10) into Eq. (6) δ ¼ lnðdÞ − maδ=d ð15Þ
and then substituting d from Eq. (7).
Solving Eq. (6) for d, and replacing d into Eq. (15) yields:

Solution of the Equivalent Colebrook Equation eδ þ aδ


m¼ ½lnðeδ þ aδÞ − δ ð16Þ

The proposed equivalent form of the Colebrook equation is also
implicit. For developing an explicit solution for the friction factor, Eq. (16) expresses m in terms of δ. Fig. 1 shows the m and
Eq. (6) should first be solved for δ. Implicit Eq. (6) relates δ and d. dm=dδ functions in terms of δ. The quantity m is approximately
Using the definition of d one gets: constant for δ > 7 and has a maximum sensitivity (dm=dδ) at
  δ ¼ 2.1. For considering a constant value for m with a minimum
Rðε=DÞ 2 R error in practical range of 1.61 ≤ δ ≤ 13.91, this quantity should be
d¼ þ ln ð11Þ
2.51 × 3.71 lnð10Þ 2.51 computed in the most sensitive point (d2 m=dδ 2 ¼ 0 which results
in δ ¼ 2.1). The quantity m is equal to 1.1 at δ ¼ 2.1.
For the practical range of 4,000 ≤ R ≤ 108 and 0 ≤ ε=D ≤ 0.1, Considering m ¼ 1.1 and solving Eq. (15) for δ yields:
d ranges between 6.4 and 1.1 × 106 , and δ ranges between 1.61
and 13.91. lnðdÞ
δ¼ ð17Þ
Eq. (6) can be written as 1 þ 1.1a=d
δ ¼ lnðdÞ þ lnð1 − aδ=dÞ ð12Þ
First Continued-Fraction Approximation
in which aδ=d is a positive quantity less than 0.22 for the practical
range of 4,000 ≤ R ≤ 108 and 0 ≤ ε=D ≤ 0.1. Because aδ=d is suf- Approximating the logarithm term in Eq. (12) by the
ficiently small and varies in a narrow range (0 < aδ=d < 0.22), continued-fraction approximation [lnð1 − xÞ ≅ −xð6 − xÞ=ð6 − 4xÞ,
thus Eq. (12) is the key relation to develop approximate analytical Abramowitz and Stegun (1972)], this equation can be written as
solutions for δ. Different analytical approximations of δ can be
aδ=dð6 − aδ=dÞ
developed using Eq. (12) which differ in the degree of accuracy δ ¼ lnðdÞ − ð18Þ
and complexity (computational cost). 6 − 4aδ=d

Solving Eq. (18) for δ yields:


pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
½3d=a þ 3 þ 2 lnðdÞ − ½3d=a þ 3 þ 2 lnðdÞ2 − 6 lnðdÞð1 þ 4d=aÞ
δ¼ ð19Þ
ð1 þ 4d=aÞa=d

Computational speed can be improved by avoiding noninteger powers (square roots). For removing the square root by the
continued-fraction expansion, the numerator and denominator of Eq. (19) are multiplied by the conjugate of the numerator which
yields:

6ðd=aÞ lnðdÞ
δ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ð20Þ
½3d=a þ 3 þ 2 lnðdÞ þ ½3d=a þ 3 − 2 lnðdÞ2 þ 18 lnðdÞ

© ASCE 06018007-3 J. Hydraul. Eng.

J. Hydraul. Eng., 2018, 144(5): 06018007


Downloaded from ascelibrary.org by UNIVERSITY OF WOLLONGONG on 03/15/18. Copyright ASCE. For personal use only; all rights reserved.

Fig. 1. m and dm=dδ variations in terms of δ

Eq. (20) has a continued-fraction expansion as

6ðd=aÞ lnðdÞ Fig. 2. α and dα=dδ variations in terms of δ for k ¼ 0.61


δ¼ 18 lnðdÞ
ð21Þ
6ðd=a þ 1Þ þ 18 lnðdÞ
2½3d=aþ3−2 lnðdÞþ2½3d=aþ3−2 lnðdÞþ : : :

By ignoring the quantity −2 lnðdÞ in the last term of Eq. (21), Substituting d from Eq. (6) into Eq. (26) yields:
this equation can be written in compact form as
ðeδ þ aδÞ ln½1 − aδ=ðeδ þ aδÞ
ðd=aÞ lnðdÞ α¼ ð27Þ
δ¼ ð22Þ ln½1 − aδ=ðeδ þ aδÞ − ð1 þ kÞδ
ðd=a þ 1Þð1 þ 6ð1þd=aÞ23−ð1þ4d=aÞ
lnðdÞ
lnðdÞ
Þ
Eq. (27) expresses α in terms of variable δ and the constant
value of k. For considering an average value for α with a minimum
Second Continued-Fraction Approximation error in the practical range of 1.61 ≤ δ ≤ 13.91, this quantity
should be computed in the most sensitive point (d2 α=dδ 2 ¼ 0
Similarly by approximating the logarithm term in Eq. (12) through
which results in δ ¼ 3.34 and k ¼ 0.61). Fig. 2 shows α and
lnð1 − xÞ ≅ −25x=ð25 − 13xÞ, this equation can be written using
dα=dδ functions in terms of δ for k ¼ 0.61. The quantity α is
the continued-fraction expansion as
approximately constant for δ > 7 and has a maximum sensitivity
ðd=aÞ lnðdÞ (dα=dδ) at δ ¼ 3.34. The quantity α is equal to 0.56 at δ ¼ 3.34
δ¼ ð23Þ and k ¼ 0.61.
ðd=a þ 1Þð1 þ 25ð1þd=aÞ132 −13ðd=aÞ
lnðdÞ
lnðdÞ
Þ Using α ¼ 0.56 and k ¼ 0.61, Eq. (24) takes the form:

d − 0.56
δ¼ lnðdÞ ð28Þ
d þ 0.34
Asymptotic Approximation
Based on Eq. (12), when d approaches infinity, the second term
in the right-hand side of Eq. (12) vanishes and thus δ approaches
Error Distributions of the Proposed Approximations
lnðdÞ. Also, as d approaches infinity, the expression ðd − αÞ=
for δ
ðd þ kαÞ approaches unity (α and k are positive quantities). Thus,
an asymptotic approximation for δ can be considered as The error distributions of the proposed analytical approximations
d−α for δ [Eqs. (14), (17), (22), (23), and (28)] compared with im-
δ¼ lnðdÞ ð24Þ plicit Eq. (6), are computed and shown in Fig. 3. Based on Fig. 3,
d þ kα
the first linear approximation [Eq. (14)] has a maximum percentage
Substituting lnðdÞ from Eq. (24) into Eq. (12) yields: error less than 1.55%, whereas the second linear approximation
[Eq. (17)] has a maximum percentage error less than 0.34%. In
d þ kα comparison, the first continued-fraction approximation [Eq. (22)]
δ¼ δ þ lnð1 − aδ=dÞ ð25Þ
d−α has a maximum error less than 0.0017%, whereas the second
continued-fraction approximation [Eq. (23)] has a maximum er-
Solving Eq. (25) for α one gets: ror less than 0.0067%. The simple asymptotic approximation
[Eq. (28)] has also a maximum error less than 0.084%. Because
d lnð1 − aδ=dÞ the flow friction factor, f, is a nonlinear function of δ, thus its errors
α¼ ð26Þ
lnð1 − aδ=dÞ − ð1 þ kÞδ are different from the δ errors.

© ASCE 06018007-4 J. Hydraul. Eng.

J. Hydraul. Eng., 2018, 144(5): 06018007


1
λ1 ¼ 1 þ 1þs 1þ4s
ð33Þ
0.5 lnð0.8686sÞ − 3ð1þsÞ

This solution has three logarithms and one noninteger power.


Likewise by substituting Eq. (23) into Eq. (10), the solution
using the second continued-fraction approximation can be ob-
tained as
 
1 0.3984R
pffiffiffi ¼ 0.8686 ln s ð34Þ
f ð0.8686sÞsþλ2
Downloaded from ascelibrary.org by UNIVERSITY OF WOLLONGONG on 03/15/18. Copyright ASCE. For personal use only; all rights reserved.

where
1
λ2 ¼ 1 þ 1þs
ð35Þ
0.52 lnð0.8686sÞ − 1þs
s

This solution also has three logarithms and one nonin-


teger power.
Through substituting Eq. (28) into Eq. (10), the solution using
the asymptotic approximation can be found as
 
1 0.3984R
pffiffiffi ¼ 0.8686 ln s−0.645 ð36Þ
f ð0.8686sÞ sþ0.39

This solution also has only two logarithms and one nonin-
teger power.
Some applications of the friction factor (such as water distri-
bution network simulators) might require the computation of the
derivative of f with respect to the Reynolds number, R. Because
all approximate solutions presented in this study can be expressed
Fig. 3. Error distributions of the proposed approximations for δ in the form of f ¼ ψðR; sÞ (where ψ = functional symbol),
the derivative of the friction factor, f, with respect to R can
be computed using the chain rule as df=dR ¼ ∂ψ=∂R þ
ð∂ψ=∂sÞð∂s=∂RÞ.
Proposed Approximations for the Flow Friction
Factor f

Defining quantity s as Error Distributions of the Proposed Approximations


for f
s ¼ 0.12363Rðε=DÞ þ lnð0.3984RÞ ð29Þ
For the analysis of error distribution, a rectangular space of the
By substituting Eq. (14) into Eq. (10), the solution using the first wide range of ε=D and R values is considered. This area is covered
linear approximation can be expressed as the following: with 1,804 points (22 × 82 grid of ε=D and R values) which uni-
formly are distributed in the logarithmic space for the practical
  range of 4,000 ≤ R ≤ 108 and 0 ≤ ε=D ≤ 0.1.
1 0.3984R
pffiffiffi ¼ 0.8686 ln s ð30Þ The error distributions of the proposed analytical approxima-
f ð0.8686sÞsþ1
tions for the friction factor, f, [Eqs. (30)–(32), (34), and (36)] com-
pared with implicit Eq. (1), are computed and shown in Fig. 4 over
This solution has only two logarithms and one noninteger
a very wide range of the Reynolds number (R from 4,000 to 108 )
power. Integer powers allow for faster computations.
and relative roughness values (ε=D from 0 to 0.1). It is notable
Similarly by substituting Eq. (17) into Eq. (10), the solution
that the flow friction factor, f, is a nonlinear function of δ, and thus
using the second linear approximation can be expressed as
its errors are different from the δ errors. According to Fig. 4, the
  solution of the friction factor, f, using the first linear approximation
1 0.3984R
pffiffiffi ¼ 0.8686 ln s ð31Þ [Eq. (30)] has a maximum percentage error less than 0.86%,
f ð0.8686sÞsþ1.1 whereas the solution using the second linear approximation
[Eq. (31)] has a maximum error less than 0.25%. In comparison,
This solution has also only two logarithms and one nonin- the solution of the friction factor, f, using the first continued-
teger power. fraction approximation [Eq. (32)] has a maximum error less than
Substituting Eq. (22) into Eq. (10), the solution using the first 0.0028%, whereas the solution using the second continued-fraction
continued-fraction approximation can also be expressed as approximation [Eq. (34)] has a maximum error less than 0.0083%.
  The solution of the friction factor, f, using the simple asymptotic
1 0.3984R
pffiffiffi ¼ 0.8686 ln s ð32Þ approximation [Eq. (36)] has also a maximum error less than
f ð0.8686sÞsþλ1 0.054%. Considering both simplicity and accuracy, Eq. (36) is pro-
posed for practical applications. When more accuracy is required
in which Eq. (32) can be used.

© ASCE 06018007-5 J. Hydraul. Eng.

J. Hydraul. Eng., 2018, 144(5): 06018007


Downloaded from ascelibrary.org by UNIVERSITY OF WOLLONGONG on 03/15/18. Copyright ASCE. For personal use only; all rights reserved.

Fig. 4. Error distributions of the proposed analytical approximations for the friction factor, f, in this study

Comparison with Similar Analytical Approximations representations of the Colebrook equation are different and thus
result in different solutions for the friction factor.
Sonnad and Goudar Approximation The estimated error of approximation by Sonnad and Goudar
(2006) is up to 1% compared with the original implicit Colebrook
Sonnad and Goudar (2006) using a clever equivalent representation equation. The solution derived in the current study using the first
of the Colebrook equation developed the following analytical linear approximation [Eq. (30)] with a maximum error less than
approximation for the friction 0.86% is comparable with the approximation derived by Sonnad
  and Goudar (2006) [Eq. (37)]. Both studies use clever equivalent
1 0.4587R
pffiffiffi ¼ 0.8686 ln G ð37Þ representations of the Colebrook equation, but the proposed alter-
f GGþ1 nate formulation of the original Colebrook equation in this study is
more efficient (lower errors) than that proposed by Sonnad and
where G is defined as
Goudar (2006).
G ¼ 0.124Rðε=DÞ þ lnð0.4587RÞ ð38Þ
Mikata and Walczak Approximation
This solution has also only two logarithms and one noninteger
power. The parameter G of Sonnad and Goudar approximation The nth formula is developed by Mikata and Walczak (2015) for
is different from the parameter s of this study. Comparison of the friction coefficient computation. The first approximate solution
Eqs. (29) and (30) of this study and Eqs. (37) and (38) derived (n ¼ 1) and simplest solution to the exact friction coefficient can be
by Sonnad and Goudar clearly indicates that the two equivalent expressed by the nth formula as

© ASCE 06018007-6 J. Hydraul. Eng.

J. Hydraul. Eng., 2018, 144(5): 06018007


  error of the nth formula aggregates in the zone of small values of
1 0.4587R
pffiffiffi ¼ 0.8686 ln ð39Þ the Reynolds number (R) and relative roughness (ε=D).
f G − lnðGÞ
The error distributions of the proposed analytical approxima-
tions for the friction factor, f, by Sonnad and Goudar (2006)
where G is defined by Eq. (38). This solution has three logarithms.
[Eq. (37)], and by Mikata and Walczak (2015) [Eqs. (39) and (40)]
The percentage error of this solution is up to 1.62% compared with
are computed and shown in Fig. 5 over the practical range of
the original implicit Colebrook equation and occurs at R ¼ 4,000
4,000 ≤ R ≤ 108 and 0 ≤ ε=D ≤ 0.1.
and ε=D ¼ 10−5 . As noted, the proposed solution [Eq. (30)] in
this study is also more efficient than that proposed by Mikata and
Walczak (2015) with three logarithms. Conclusions
The second approximate solution (n ¼ 2) to the exact friction
coefficient can be expressed by the nth formula as This paper presents a new mathematically equivalent form of
Downloaded from ascelibrary.org by UNIVERSITY OF WOLLONGONG on 03/15/18. Copyright ASCE. For personal use only; all rights reserved.

  the Colebrook equation to compute the friction factor in the fully


1 0.4587R developed turbulent regime. This form of the Colebrook equation
pffiffiffi ¼ 0.8686 ln ð40Þ
f G − lnðG − ln GÞ results in another implicit equation which can directly be solved
using the linear approximations, more accurate continued-fractions
This solution has four logarithms. The maximum percentage approximations, or asymptotic approximations. The mathematical
error of this solution is 0.28% and occurs again at R ¼ 4,000 derivation of the proposed approach is simple and straightforward
and ε=D ¼ 10−5 . As has been emphasized by Brkić (2017), the and requires only elementary mathematical manipulations. The
proposed explicit solutions are full range and cover the entire prac-
tical range of Reynolds numbers and relative roughness in which
the original implicit Colebrook equation can be used. The proposed
solutions are preferable to previous presented explicit solutions in
terms of accuracy, generality, and simplicity. The explicit solutions
presented in this study are useful in the design, analysis, and mod-
eling of the fluid flows in pipes and pipe networks.

Notation

The following symbols are used in this paper:


a, b, c, and d = known parameters;
D = inner diameter of pipe;
f = friction factor;
G = known parameter in terms of R, D and ε;
k = positive quantity;
m = positive quantity;
R = Reynolds number;
s = known parameter in terms of R, D, and ε;
x = unknown parameter related to friction factor;
α = positive quantity;
δ = unknown parameter;
ε = average height of protrusion of inner pipe surface;
and
λ1 and λ2 = known parameters in terms of s.

References
Abramowitz, M., and Stegun, I. A. (1972). Handbook of mathematical
functions with formulas, graphs, and mathematical tables, Vol. 9,
Dover Publications, New York.
Barr, D. I. H. (1981). “Solutions of the Colebrook-White function for
resistance to uniform turbulent flow.” Proc. Inst. Civ. Eng., 71(2),
529–536.
Biberg, D. (2017). “Fast and accurate approximations for the Colebrook
equation.” J. Fluids Eng., 139(3), 031401.
Brkić, D. (2011a). “Review of explicit approximations to the Colebrook
relation for flow friction.” J. Pet. Sci. Eng., 77(1), 34–48.
Brkić, D. (2011b). “W solutions of the CW equation for flow friction.”
Appl. Math. Lett., 24(8), 1379–1383.
Brkić, D. (2012). “Comparison of the Lambert W-function based solutions
Fig. 5. Error distributions of the proposed analytical approximations to the Colebrook equation.” Eng. Comput., 29(6), 617–630.
for the friction factor, f, by Sonnad and Goudar (2006) and Mikata and Brkić, D. (2017). “Discussion of “Exact analytical solutions of the
Colebrook-White equation” by Mikata, Y., and Walczak, W. S.”
Walczak (2015)
J. Hydraul. Eng., 10.1061/(ASCE)HY.1943-7900.0001341, 07017007.

© ASCE 06018007-7 J. Hydraul. Eng.

J. Hydraul. Eng., 2018, 144(5): 06018007


Brkić, D., and Ćojbašić, Ž. (2017). “Evolutionary optimization of Papaevangelou, G., Evangelides, C., and Tzimopoulos, C. (2010). “A new
Colebrook’s turbulent flow friction approximations.” Fluids, 2(2), 15. explicit equation for the friction coefficient in the Darcy–Weisbach
Buzzelli, D. (2008). “Calculating friction in one step.” Mach. Des., 80(12), equation.” Proc., 10th Conf. on Protection and Restoration of the
54–55. Environment: PRE10, Greece Conf., Vol. 166, 6–9.
Chen, N. H. (1979). “An explicit equation for friction factor in pipes.” Romeo, E., Royo, C., and Monzon, A. (2002). “Improved explicit equation
Eng. Chem. Fundam., 18(3), 296–297. for estimation of the friction factor in rough and smooth pipes.” Chem.
Churchill, S. W. (1973). “Empirical expressions for the shear stress in Eng. J., 86(3), 369–374.
turbulent flow in commercial pipe.” AIChE J., 19(2), 375–376. Serghides, T. K. (1984). “Estimate friction factor accurately.” Chem. Eng.,
Churchill, S. W. (1977). “Friction-factor equation spans all fluid flow 91(5), 63–64.
regimes.” Chem. Eng., 84(24), 91–92. Sonnad, J. R., and Goudar, C. T. (2004). “Constraints for using Lambert W
Clamond, D. (2009). “Efficient resolution of the Colebrook equation.” function-based explicit Colebrook-White equation.” J. Hydraul. Eng.,
Ind. Eng. Chem. Res., 48(7), 3665–3671. 10.1061/(ASCE)0733-9429(2004)130:9(929), 929–931.
Colebrook, C. F. (1939). “Turbulent flow in pipes with particular reference Sonnad, J. R., and Goudar, C. T. (2006). “Turbulent flow friction factor
Downloaded from ascelibrary.org by UNIVERSITY OF WOLLONGONG on 03/15/18. Copyright ASCE. For personal use only; all rights reserved.

to the transition region between the smooth and rough pipe laws.” calculation using a mathematically exact alternative to the Colebrook-
J. Inst. Civ. Eng., 11(4), 133–156. White equation.” J. Hydraul. Eng., 10.1061/(ASCE)0733-9429(2006)
132:8(863), 863–867.
Giustolisi, O., Berardi, L., and Walski, T. M. (2011). “Some explicit
Sonnad, J. R., and Goudar, C. T. (2007). “Explicit reformulation of the
formulations of Colebrook-White friction factor considering accuracy
Colebrook-White equation for turbulent flow friction factor calcula-
vs. computational speed.” J. Hydroinf., 13(3), 401–418.
tion.” Ind. Eng. Chem. Res., 46(8), 2593–2600.
Haaland, S. E. (1983). “Simple and explicit formulas for friction factor in
Swamee, P. K., and Jain, A. K. (1976). “Explicit equations for pipe flow
turbulent pipe flow.” J. Fluids Eng., 105(1), 89–90.
problems.” J. Hydraul. Div., 102(HY5), 657–664.
Manadilli, G. (1997). “Replace implicit equations with signomial func- Vatankhah, A. R. (2014). “Comment on Gene expression programming
tions.” Chem. Eng., 104(8), 129–130. analysis of implicit Colebrook-White equation in turbulent flow friction
Maple [Computer software]. Symbolic Computation Group, Univ. of factor calculation.” J. Pet. Sci. Technol., 124, 402–405.
Waterloo, ON, Canada. Vatankhah, A. R., and Kouchakzadeh, S. (2008). “Discussion of “Turbulent
Mathcad [Computer software]. PTC, Needham, MA. flow friction factor calculation using a mathematically exact alternative
Mathematica [Computer software]. Wolfram Research, Champaign, IL. to the Colebrook–White equation” by Sonnad, J. R., and Goudar, C. T.”
Mikata, Y., and Walczak, W. S. (2015). “Exact analytical solutions of the J. Hydraul. Eng., 10.1061/(ASCE)0733-9429(2008)134:8(1187),
Colebrook-White equation.” J. Hydraul. Eng., 10.1061/(ASCE)HY 1185–1186.
.1943-7900.0001074, 04015050. Vatankhah, A. R., and Kouchakzadeh, S. (2009). “Discussion of “Exact
More, A. A. (2006). “Analytical solutions for the Colebrook and White equations for pipe flow problems” by Swamee, P.K. and Rathie, P.N.”
equation and for pressure drop in ideal gas flow in pipes.” Chem. J. Hydraul. Res., 47(7), 537–538.
Eng. Sci., 61(16), 5515–5519. Zigrang, D. J., and Sylvester, N. D. (1982). “Explicit approximations to
Nandakumar, V. (2007). “Analytical calculation of Fanning friction factor.” the solution of Colebrook’s friction factor equation.” AIChE J., 28(3),
Hydrocarbon Process., 86(1), 97. 514–515.

© ASCE 06018007-8 J. Hydraul. Eng.

J. Hydraul. Eng., 2018, 144(5): 06018007

Potrebbero piacerti anche