Revista Facultad de Ingeniería Universidad de Antioquia

Revista Facultad de Ingeniería, Universidad de Antioquia, No.xx, pp.
x-xx, xxx-xxx 20xx

M. López-Silva et al.; Revista Facultad de Ingeniería, No. xx, pp. x-x, 20xx
Revista Facultad de Ingeniería
Title: Explicit pipe friction factor equations: evaluation, classification, and proposal
Authors: Maiquel López-Silva, Dayma Sadami Carmenates-Hernández, Nancy Delgado-Hernández and

Nataly Chunga-Bereche
DOI: 10.17533/udea.redin.20230928
To appear in: Revista Facultad de Ingeniería Universidad de Antioquia
Received: August 16, 2022

Accepted: October 19, 2023
Available Online: October 19, 2023
This is the PDF version of an unedited article that has been peer-reviewed and accepted for publication.
It is an early version, to our customers; however, the content is the same as the published article, but it
does not have the final copy-editing, formatting, typesetting and other editing done by the publisher
before the final published version. During this editing process, some errors might be discovered which
could affect the content, besides all legal disclaimers that apply to this journal.
Please cite this article as: M. López-Silva, D. S. Carmenates-Hernández, N. Delgado-Hernández and N.

Chunga-Bereche. Explicit pipe friction factor equations: evaluation, classification, and proposal, Revista
Facultad de Ingeniería Universidad de Antioquia. [Online]. Available:
https://www.doi.org/10.17533/udea.redin.20230928
Revista Facultad de Ingeniería, Universidad de Antioquia, No.xx, pp. x-xx, xxx-xxx 20xx
Explicit pipe friction factor equations: evaluation, classification, and proposal

Ecuaciones explícitas del factor de fricción de tuberías: evaluación, clasificación y propuesta
Maiquel López-Silva1* https://orcid.org/0000-0002-0946-6160

, Dayma Sadami Carmenates-Hernández1
https://orcid.org/0000-0001-5482-7562
, Nancy Delgado-Hernández1 https://orcid.org/0000-0003-3247-6671 and Nataly Chunga-
Bereche1 https://orcid.org/0000-0003-0220-3023
1
Universidad Católica Sede Sapientiae, Perú.
Corresponding author: Maiquel López-Silva

E-mail: mlopezs@ucss.edu.pe
KEYWORDS
Colebrook equation, turbulent fluid, relative roughness, Reynolds number
Palabras clave
Ecuación de Colebrook, fluido turbulento, rugosidad relativa, número de Reynolds
ABSTRACT:
The Colebrook equation has been used to estimate the friction factor (f) in turbulent fluids. In this regard,
many equations have been proposed to eliminate the iterative process of the Colebrook equation. The
goal of this article was to perform an evaluation, classification, and proposal of the friction factor for
better development of hydraulic projects. In this study, Gene Expression Programming (GEP), Newton-
Raphson, and Python algorithms were applied. The accuracy and model selection were performed with
the Maximum Relative Error (∆f/f), Percentage Standard Deviation (PSD), Model Selection Criterion
(MSC), and Akaike Information Criterion (AIC). Of the 30 equations evaluated, the Vatankhah equation
was the most accurate and simplest to obtain the friction factor with a classification of very high,
reaching a value of ∆f/f<0.5% and 1.5<PSD<1.6. A new equation was formulated to obtain the explicit
f with fast convergence and accuracy. It was concluded that the combination of GEP, error theory, and
selection criteria provides a more reliable and strengthened model.
RESUMEN:
La ecuación de Colebrook se ha utilizado para estimar el factor de fricción (f) en fluidos turbulentos. En
este sentido, se han propuesto varias ecuaciones para eliminar el proceso iterativo de la ecuación
Colebrook. El objetivo de este artículo fue realizar una evaluación, clasificación y propuesta del factor
de fricción para un mejor desarrollo de proyectos hidráulicos. En este estudio, se aplicaron los algoritmos
de programación de expresión génica (GEP), Newton-Raphson y Python. La precisión y la selección del
modelo se realizaron con el Máximo Error Relativo (∆f/f), Porcentaje de Desviación Estándar (PSD),
Criterio de Selección del Modelo (MSC) y Criterio de Información de Akaike (AIC). De las 30
ecuaciones evaluadas, la ecuación de Vatankhah fue la más precisa y sencilla para obtener el factor de
fricción con una clasificación de muy alta, alcanzó un valor de ∆f/f<0.5% y 1.5<PSD<1.6. Se formuló
una nueva ecuación para obtener el f explícita con rápida convergencia y precisión. Se concluyó que la
combinación de GEP, teoría del error y criterios de selección proporciona un modelo más confiable y
fortalecido.
1. Introduction alternative to the Colebrook equation, but the

Pipes are used worldwide for the transportation of explicit models developed differ in their accuracy
liquids with different properties. Non-Newtonian and computational efficiency [16]-[19]. The work
fluids are transported in pipelines in the mining and presented by [20] highlighted that the equation by
metallurgical industries, such as drilling mud, [21] was more accurate than the Colebrook
cementitious composites, and pastes [1]. In equation for the experimental data in their
contrast, Newtonian fluids have a wider field of research. On the other hand, [22] cite that the
use, especially in turbulent flow over rough equations by [16] and [23] are the most efficient,
surfaces, with several engineering applications with a maximum-recorded error of 0.18% and
such as industrial plants, internal distribution 0.54%, respectively. Likewise, [24] propose that
networks in buildings, hydraulic turbines, the equations available in the literature lead to a
irrigation systems, and drinking water pipelines deviation of between 2% and 3% for a turbulent
[2], as well as in open-channel hydraulics [3]. Head flow with a Reynolds number of 2300. In turn, they
losses are common in pipes or open channels, an suggest a new equation based on the relationship
essential parameter that affects the design and between friction forces and viscous forces to
operation of the circulation flow in hydraulic determine f with a maximum standard deviation of
works [4], [5]. 0.25% with respect to the Colebrook equation.
In piping systems, head losses are analyzed by the There have been significant contributions in recent
universal Darcy-Weisbach equation. However, the years to predicting the friction factor value with
implicit friction factor (f) intervenes in the artificial intelligence approaches such as Gene
equation. In this sense, Colebrook [6] proposes an Expression Programming (GEP), Evolutionary
equation that is currently the best approximation of Polynomial Regression (EPR), Adaptive Neuro-
the friction factor, especially for turbulent flow [7]. Fuzzy Inference System (ANFIS), Artificial
Nevertheless, its calculation is complex and Neural Network (ANN), and physical and
cumbersome because the friction factor is present numerical models that manage to predict the fluid
at both ends of the equation. In addition, its behavior in different media [25]-[28]. In particular,
solution needs more time and processing in [26] estimated f using Bayesian learning neural
calculators. Therefore, its solution requires using networks and reached a relative error of 0.0035%.
iterative methods such as the Newton-Raphson Furthermore, [29], using some artificial
approximation method. Although diagnostic and intelligence approaches, reached mean absolute
control algorithms are implemented in the errors of 0.001%. In this sense, [30] cite some gaps
mathematical modeling of hydraulic systems, in the artificial intelligence technique, such as the
precise parameter tuning is necessary. data set, the layers of predesignated neurons, the
Several authors [9]-[15] have developed explicit percentage of training, and the test in the model
approximations of the friction factor as an tree. However, increasing the number of variables
and implicit functions of the friction factor is this sense, the goal of this work was to perform an
necessary. Likewise, there is still a need to insert evaluation, classification, and a new suggested
model selection criteria. explicit pipe friction factor equation with the least
Many authors tend to use the Mean Squared Error amount of error.
(MSE), Mean Relative Error (MRE), Mean
Absolute Error (MAE), Standard Deviation (SD), 2. Materials and methods
and Relative Error (RE) [7], [22], and [31]. This The Colebrook equation is the most cited,
has several disadvantages when compared to other accepted, and validated equation in fluid dynamics
models since the value of R is more significant studies for obtaining friction losses in pipes.
when the number of variables in the mathematical It relates, in its implicit form, to the unknown
model increases [32]. Therefore, the value R can friction factor (f), the relative roughness (e/D), the
be increased, and the models can become more known pipe inner surface area, and the known
complex. Reynolds number (Re). Valid for 4000< Re<108 y
There are several techniques to adjust the training 0<e/D<5·10-2, as shown in Equation 1. However,
error for model sizes, such as Model Selection equation 1 requires some mathematical iterations
Criteria (MSC), Akaike Information Criterion to get the optimal solution.
(AIC) [33], Bayesian Information Criterion (BIC)
[34], and Mallows' Cp Criterion [35]. The MSC 1  ε 2.51 
=
−2log  + 

and AIC have applied for the best prediction f  3.71D Re f  (1)
model, but there have been limits: 4000<Re<108
and 10-6 <e/D<5·10-2 [12], [36], discrepancies in
the results. The selection is important because the Where f is the implied friction factor (f), e is the
decision of the criterion could affect the absolute roughness of the pipe's inside wall, D is
interpretation of the variable as well as its the pipe diameter, and Re is the Reynolds number.
prediction. Thus, the following hypothesis is Nonetheless, there are several explicit approaches
proposed in the present study: the explicit friction reported in the scientific literature to calculate the
factor equations can be classified, and the GEP can friction factor, as shown in Equations 2 to 36.
provide a new equation with a minimum error. In
[21].
−2
  ε D 4.5 Re   (2)
f =
 −2log  3.7 + Re log 6.97  
  
Re ≥ 104 and 0 <ε/D <5·10-2
[24].
−2
  ε 10.04  
f=  −2log  3.7D + R *  
(3)
  
−1
  ε D 5.45  
R =
*
2Re  − log  + 0.9   (4)
  3.7 Re  
R* the dimensionless number;
Limit: Re ≥ 2300 and 0 < ε/D <0.05
[37].
0.25
 ε 68 
=f 0.11  + 
 D Re  (5)
Limit: not specified
[38].
6.4
f= 2.4
  ε ε  
ln(Re) − ln  1 + 0.01Re  1 + 10  
  D D    (6)
 
Limit: Re ≤ 4000
[39].
−2
    
    Re   
4.518log   (7)
  ε    7   
f=
−2log  + 0.7   
  3.7D   Re  1 + 1 Re0.52  ε     
    29 ( )
    
 D   
    
Limit: 5000 < Re <108 and 10-2 <ε/D <10-6
[14]. Model I
−2
  −0.4343β ε D  
f = −2log  10 + 
(8)
  3.71  
[14]. Model II
−2
  2.18β ε D  
f=  −2log  + 
(9)
  Re 3.71  
Where β is:
 
  (10)
Re
β = ln  
 1.1Re  
 1.816ln  
  ln (1 + 1.1Re )  
[17].
−2
   1.038ln ( B + A )   (11)
=f 0.8686  B +   − ln ( B + A )  
   ( 0.332 + B + A )  
   
 Re ( ε D )  (12)
A= 
 8.0878 
 Re 
B = ln  
(13)
 2.18 
[40].
−2
  ε 5.0452  ( ε D )1.1098   (14)
 −2log 
f= − log  + 5.8506Re−0.8981   
  3.7065D Re  2.8257   
 
Limit: 4000 <Re< 108 and 10-6< ε/D <5·10-2
[41].
−2
  ε 7  (15)
f=
 −2log  3.71D + Re0.9  
  
Limit: 4000 <Re <108 and 10-6 <ε/D <5·10-2
[42].
1
 
−3 12
   
16
2 
 8 12   
 37530   
16
  1
8   +  2.457ln  0.9
f=   +   
 Re   7  ε    Re   
   Re + 0.27     
    D    

 
(16)
Limit: 4000 < Re <108 and 10-6 <ε/D <5·10-2
[43].
−2
  ε 15  
f=  −2log  3.715D + Re  
(17)
  
[13].
0.2479 − 0.0000947(7 − log Re)4 (18)
f= 2
  ε 7.366  
log  3.615D + Re0.9142  
  
[23].
−2
  ε 
1.1007
60.525 56.291  
f 1.613 ln 0.234   − 1.1105 + 1.0712   (19)
  D Re Re  
Limit: 3000 < Re < 10 and 10 < ε/D < 5·10-2
8 -6
[44].
−2.169
  ε 1.042  2.731 0.9152   (20)
f= −1.52log   +  
  7.21D   Re   
Limit: 2100 < Re < 108 and 10-6 < ε/D < 5·10-2
[45].
−2
  ε 1.11 6.9   (21)
f=
−1.8log   + Re  
  3.7D   
Limit: 400 <Re <108 and 10-6<ε/D<5·10-2
[46].
−2
  ε 95 96.82  
f=  −2log  + − 
(22)
  3.7D Re
0.983
Re  
Limit: 5235 < Re <109
[47].
−2
  7.35 − 1200 ( ε D )1.25  ε 1.115   (23)
f=
−1.8log  +  
  Re  3.15D   
Limit: 4000 ≤ Re ≤ 35.5 ·106
[12].
−2
  ε 5.0272  ε 4.657   ε
0.9924 0.9345
   
   5.3326  (24)
−2log 
f= − log  − log    +     
 3.7065D Re  3.827D Re   7.7918D   208.815 + Re  
      
Limit: Re > 4000
[9].
 1

   ε   106
  3
 (25)
=f 0.0055 1 +  2 ⋅ 10   + 
4

   D   Re   
 
Limit: 4000 < Re < 108 and 0 <ε/D < 10-2
[31].
−2
  ε  1.975    ε 1.092  7.627       (26)

−2log 
f= −  ln    +    
  3.71D  Re    3.93D   395.9 + Re     
 
[48].
−2
  ε  6.81 0.9  
f= −2log  +    (27)
  7D  Re   
Limit: 4000 < Re < 108 and 10-6 <ε/D < 10-2
[49].
−2
   6.5   
−1.8log 0.27 ( ε D ) + 
f = (28)
 
   Re   
Limit: 4000 < Re < 107 and 10-6 <ε/D < 10-2
[50].
−2
  ε 5.02  ε 14.5    (29)
f=
−2log  − log  +  
  3.7D Re  3.7D Re   
Limit: 4000 < Re <108 and 10-6 < ε/D<5·10-2
[51].
−2
   ε   6.943 0.9  
f=  −2log   + 
  3.715D   Re   
(30)
  
Limit: 5000 < Re <10 and 10 < ε/D <5·10-2
8 -6
[10].
−2
  ε 5.74  
f 0.25 log  + 0.9   (31)
  3.7D Re  
Limit: 5000 < Re < 108 and 10-6 < ε/D < 5·10-2
[16].
−2
 
 0.3984Re  (32)
f = 0.8686ln
 S −0.645

 (0.8686S ) 
S +0.39
Where S is:
S 0.12363Re ( ε D ) + ln(0.3984Re)
(33)
[52].
f 0.094 ( ε D ) + 0.53( ε D ) + 88 ( ε D )
−1.62( ε D )
0.134
0.225 0.44
= Re (34)
Limit: Re > 4000 and 10-5 < ε/D < 5·10-2
[11] Model I.
−2
  ε  5.02  ε 5.02  ε 13     (35)
f=−2log  − log  − log  +   
  3.7D  Re  3.7D Re  3.7D Re    
Limit: 4000 < Re <108 and 10-5 < ε/D <5·10-2
[11] Model II.

−2
  ε  5.02  ε 13    (36)
f =
−2log   − log  +  
  3.7D  Re  3.7D Re   
Limit: 4000 < Re < 108 and 10-5 < ε/D <5·10-2
The Colebrook equation and the 30 explicit ∆f  fCW − f proposed 
=  100% (37)
equations found in the scientific literature were f  fCW 
evaluated for different conditions of relative
roughness (e/D) from 10-6 to 5·10-2 and the Additionally, efficient methods of model
Reynolds number from 4000 to 108, which implied comparison and selection based on model
a base of 47601 data points. The analysis interval complexity were applied. Model Selection Criteria
integrates the onset of turbulence and complete (MSC) [29] and Akaike's Information Criteria
turbulence to test the best behavior of the (AIC) were used [26]. These criteria expressed by
correlations in the mathematical formulations. Equations 38 and 39 are based on the greatest
In this study, the Newton-Raphson method was likelihood and smallest parameters, and the
used in Colebrook equation 1 by the Python variables follow a normal distribution.
algorithm. The method has been generalized due to
(  n
)
2 
its simplicity and speed of convergence to solve  ∑ fCW − f Pr oposed  2p

(38)
= and MSC ln  n −
i =1
nonlinear problems, systems of equations, n
 (f − f
Pr oposed )

 ∑
2
nonlinear differential and integral equations [23].
i =1
CW

Similarly, Gene Expression Programming (GEP),
implemented in GeneXpro software, was applied, 1 n 2

 ∑ ( CW Pr oposed )  + 2p
AIC
= n ln f − f (39)
after obtaining the evaluation, classification, and n
 i =1 
generation of the most suitable equations. Initially,
Where fCW is the true value of the Colebrook-White
the database composed of 47,601 variables was
(CW) friction factor, fproposed is the value of the
used to select the best adjustment according to their
proposed friction factor, p is the number of
fitness and introduce genetic variation using
equation parameters including constants, i = 1,… n
genetic operators.
is the number of friction factor values, and n is the
Additionally, the procedure for estimating the
sample size.
pipeline friction coefficient using GEP involved
3. Results and discussion
fitness function selection, choice of T-termini and
Figure 1 shows the accuracies of the explicit
F-functions to create chromosomes, choice of
models according to the Maximum Relative Error
chromosome architecture, choice of linkage
(∆f/f) and percent standard deviation (PSD). Figure
function, and choice of genetic operators.
1 a) shows that the (∆f/f) values ranged from
The 30 Chromosomes were executed, with a head
0.082% to 38.435%, and 43% of the equations had
size of 8 and the number of genes 1, 2, 3, and 6;
values lower than 2.0% of the Maximum Relative
linking functions (+, -, *, /); and mathematical
Error. Group I is the most efficient approximation
functions divided into GEP1, GEP2, GEP3, and
where the Maximum Relative Error is less than
GEP4 in +, −, /, ⋅, x , e x ,log10 ,10 x x 13 , x 14 , x 15 , x 2 , x3 , x 4 , x5 , 1 .
x 1%; therefore, those are recommended for
In the investigation, percent standard deviation precision engineering work. In Group I, the results
(PSD) and Equation 37 Maximum Relative Error are outstanding, presenting values ∆f/f < 0.5% by
(∆f/f) were used as criteria for the accuracy of the the equations of [12], model I [11], [31], [16], and
explicit models. [17], [24], [40], [23]. In particular, the equations by
[39], [13], and [50] have 0.5 < ∆f/f < 1%.
Other authors have formulated new, noteworthy,
accurate equations; these are classified in group II
because they have a Maximum Relative Error of
less than 2%, which are those proposed for model
II by [11] and [45].
Figure 1 Explicit model specifications.
Group III was classified as having a lower test was high, exceeding 10%. Similarly, it agrees
approximation to Colebrook's with a Maximum with the results by [29] on the mathematical
Relative Error between 2.587 ≤ ∆f/f ≤ 8.303, as models analyzed using Machine Learning tools in
equations cited by [21], [46], model II by [14], which [9] and [42] had the most unfavorable
[44], [38], [51], [10], and model I by [14]. equations.
However, the equation by [21], according to [20] Regarding Figure 1 b) and the Percent Standard
in their research, was the most accurate. Possible Deviation (PSD), it is observed that, in general, the
causes were that [20] only used 2397 experimental 30 equations analyzed presented a deviation
points, 3000 ≤ R e≤ 735-103, and 0 < ε/D <1.4-10- between 1.2%<PSD<2%. However, 81% of the
3 equations had a stable standard deviation between
. Nonetheless, group IV had to be rejected because
they exceeded ∆f/f > 10%, as are [42], [49], [9], 1.5% and 1.6%. Nevertheless, there are three
[48], [47], [52], and [37]. In particular, the equations of approximations with the lowest
equation proposed by [9], at the time provided standard deviation, such as [48], [9] and [37], but
significant results for solving problems, but it is they presented a high relative error for which they
shown that new and more accurate formulations were rejected.
have been developed. The 30 equations analyzed in this article have two
Results that agree with those obtained by [22], who perspectives: firstly, the equations with a high
evaluated 33 equations in a range of the Moody number of parameters tend to be more accurate,
diagram with 2300 ≤ Re ≤108, 0 <ε/D<5-10-2 and and secondly, the equations with the least number
in relation to the equation proposed by [9] the error of parameters are less accurate. On the other hand,
the engineer needs the easiest and most accurate equation model I occupies rank 30. In relation, the
equation for friction factor calculation, according AIC reached inversely proportional values, the
to [24]. In summary, as a result of the increasing [49] equation reached rank 30 and [11] equation
digitization of work, educational and economic model I has rank 1. On the other hand, in contrast
environments, the equations must be formulated to the previous equations, the number of
with the highest precision and best computational parameters by [11] equation model I is 47% higher
performance. than by [49] equation. Consequently, it can be
For this reason, the MSC and AIC Model Selection pointed out that the AIC criterion does not follow
Criteria have been implemented using a Ranking the parsimony principle because the smaller the
because it considers a decisive variable as the number of parameters, the smaller the AIC tends to
number of parameters, including the constants in be.
the equations (p). It should be noted that the AIC criterion does not
Based on the accuracies of the models, a follow the principle of parsimony. In summary,
preliminary model ranking (Rk) was proposed for there is a tendency for the AIC criterion to improve
each evaluation criterion p, ∆f/f, PSD, MSC, and as the number of parameters increases; these
AIC, and finally, a Global Ranking. Table 1 shows factors contradict the theories for which the AIC
the results of the models. It is observed that the criterion was defined. In finite samples, the AIC
error theory and theoretical functions show results value is only approximate [33]. Therefore,
that differ in their rank order for each equation, difficulties could arise regarding the validity and
with a discrepancy in optimal model selection. applicability of the method for this purpose.
Equation 5, proposed by [37] is the simplest and Additionally, the MSC criterion also showed
has the least number of steps to obtain the friction inconsistencies between the models due to the
factor. Nevertheless, in the previous analysis, it number of parameters; however, this coincides
was rejected because of its high relative error, with the results of the AIC criterion. This trend in
which is positioned at number 30. Meanwhile, the results corresponds with those results obtained
Equation 11 by [17] is classified as the most by [36].
complex for its solution due to the number of steps The global ranking obtained in Table 1 integrates
and parameters it includes. However, it was the positions of the most accurate and inaccurate
classified in group I with a relative error of less approximation models with their degrees of
than 0.5% and an acceptable deviation of less than complexity. The explicit equation 32 proposed by
1.6%, with a ranking of 8. [16] leads the Global Ranking in the first position
In this sense, MSC and AIC contributed to the as the most accurate, followed in second place by
selection of the best model. However, in both Equations 29, 26, and 22 by [50], [31], and [46].
cases, they present discrepancies with respect to The least accurate and most complex to solve are
the function of greater likelihood and entropy. The Equations 34, 23, 28 by [52], [47], and [49], which
MSC value indicates that by [49] equation in turn belong to the rejected group IV.
occupies rank 1, while the MSC value of the [11]
Table 1 Preference models

Main Model selection Global
ors
Au
equ
No.
th
p
statistics criteria Ranking
Parameter
Global
MSC
Total
PSD
AIC
∆f/f
No Rk Rk Rk Rk ∑Rk GR
[21] 2 11 14 20 17 14 76 14
[24] 3 17 6 14 28 3 68 7
[37] 5 6 30 3 3 28 70 9
[38] 6 14 18 25 10 21 88 20
[39] 7 19 9 13 23 8 72 11
[14] I 8 18 21 18 11 20 88 20
[14] II 9 19 16 22 15 15 87 19
[17] 11 39 5 8 25 6 83 18
[40] 14 16 7 11 24 7 65 5
[41] 15 9 23 27 8 23 90 22
[42] 16 19 24 5 5 27 80 17
[43] 17 8 22 28 9 22 89 21
[13] 18 14 10 16 21 10 71 10
[23] 19 13 8 17 22 9 69 8
[44] 20 10 17 4 12 19 62 3
[45] 21 10 13 19 18 13 73 13
[46] 22 10 15 2 16 16 59 2
[47] 23 12 28 29 6 25 100 24
[12] 24 21 1 12 29 2 65 5
[9] 25 8 26 2 4 26 66 6
[31] 26 15 3 10 27 4 59 2
[48] 27 7 27 1 2 29 66 6
[49] 28 8 25 30 1 30 94 23
[50] 29 11 11 6 20 11 59 2
[51] 30 9 19 23 14 17 82 15
[10] 31 8 20 24 13 18 83 16
[16] 32 13 4 9 26 5 57 1
[52] 34 16 29 26 7 24 102 25
[11] I 35 17 2 15 30 1 65 5
[11] II 36 14 12 7 19 12 64 4
Consequently, an easier classification has been As a new proposal for explicit friction factor
established, according to the level of precision and approximation equations, 64 models were
simplicity for the first five global rankings. It was analyzed in Gene Expression Programming (GEP).
established from a very high level, which indicates The theoretical and experimental databases were
excellent precision and simplicity, to a very low developed as a training process to train the GEP
level, which is interpreted as an inaccurate and algorithm. Twenty percent of the data was reserved
complex equation to solve due to the number of for validation and the rest for calibration. Only the
operations and parameters present. most efficient results of GEP1, GEP2, GEP3, and
GEP4 according to the performance criteria are was 0.99873, the ∆f/f was 6.22%, and the PSD was
reflected in Table 3. 1.86%.
In contrast to the groups made in Figure 1 due to
Table 2 Model classification the maximum relative error, GEP1 was classified
in group III because it was within the interval 2.5
GR Authors Precision Simplicity ≤ ∆f/f ≤ 8.3, this being an alternative to obtain the
1 [16] Very high Very high friction factor quickly and easily.
2 [50] Very high High Although GEP4 has the highest R and a lower
2 [31] Very high High ∆f/f, PSD, it is shown to be more significant for
3 [46] Medium Medium having a greater number of functions, according to
3 [44] Medium Medium [24]. In addition, the GEP4 model has a greater
4 [11] II High Low number of operations for its solution, making it
5 [11] I Very high Very Low less simple. Regarding the increase of functions,
5 [12] Very high Very Low the Number of Chromosomes, Head Size, and
5 [40] Very high Very Low Number of Genes showed a partial relationship to
the results obtained by [51] that the GEP models
Table 3 shows that the most significant models had increase with increasing functions.
Linking Functions + and *, a Number of Equation 40 is proposed as a new nonlinear model
Chromosomes of 30, a Head Size of 8, and a to determine the explicit friction factor coefficient
Number of Genes of 2 and 6. The best-performing with the lowest error without the existence of
model was GEP1, with the lowest number of logarithmic functions, speed of calculation, or
functions (4), and 7 parameters including more accurate approximation in the turbulent flow
constants. The Root Mean Square Error (RMSE) regime. The Limit: 4000 < Re < 108 and 10-6 <ε/D
was 0.078%, the Mean Absolute Error (MAE) was < 10-2.
0.055%, the Pearson correlation coefficient (R)
Table 3 Efficient model of the GEP

Model Functions Performance criteria
Training and Validations
RMSE MAE R ∆f/f PSD
(%) (%) %
GEP1 + , − , /, *, x , 0.078 0.055 0.99873 6.22 1.86
GEP2 + , − ,/,*, x , e x ,log ,ln
10
0.114 0.093 0.99729 6.49 1.92
1 1 1
GEP3 + , − ,/,*, x , e ,log10 ,ln,10x , x 3 , x 4 , x
x 5 0.080 0.064 0.99868 6.31 1.89
1 1 1
GEP4 + , − ,/,*, x , e x ,log10 ,10x x 3 , x 4 , x 5 , x 2 , x 3 , x 4 , x 5 , 1 0.089 0.067 0.99893 6.20 1.84
x
GEP1 0.219 ( 0.028 ε D ) + ( 0.896 R ) 

0.25
= (40)
4. Conclusions ε/D) with the lowest degree of complexity in the

Thirty explicit friction factor equations were turbulent flow regime. It has an RMSE of 0.078%,
analyzed on a base of 47601 theoretical and an MAE of 0.055%, and an R of 0.99873.
experimental data points and according, to the Compared with the Colebrook equation, it has
maximum relative error (∆f/f), were classified into more simplicity, fast convergence, less
4 groups: group I of 0.5% < ∆f/f, group II of 0.5% computational time, and a good relationship
< ∆f/f < 1%, group III of 1% < ∆f/f < 2% and group between accuracy and computational efficiency.
IV ∆f/f > 2%. Group I includes the most accurate From the analyzed equations of the explicit friction
explicit friction factor equations, developed by factor for turbulent flow, it was found that there are
[12], model I [11], [31], [16], [17], [24], [40] and new equations with optimal efficiency indicators
[23]. In general, the Percentage Standard Deviation for the original equations that are cited, such as
(PSD) was acceptable and comprised between those by [9], [10], and [51]. In this regard, it is
1.2%<PSD≤1.9%. recommended to consider the mathematical
The MSC and AIC selection criteria contributed to models' new functions as more accurate explicit
the selection of the most accurate equations to approximations.
estimate the friction factor, but they presented a The main finding of the research developed is the
discrepancy in likelihood and entropy. However, integration of statistical tools, Python algorithms,
the number of parameters and operations of the Genetic Expression Programming, and the new
equations (p) was a decisive variable in obtaining model proposed for obtaining the level of
the global ranking of the 30 friction factor complexity and effectiveness of the explicit
equations explicit in Table 2. In summary, the first friction factor equations of the Colebrook
five global rankings were classified by the most equation. Likewise, novel information would ease
accurate and simple equations. Therefore, it was the elaboration and decision-making of hydraulic
concluded that the estimates of the equation by engineering projects. In response to the previous
[16] ranked very high in accuracy and simplicity conclusion, it is recommended to extend the
for obtaining explicit friction factors. The [50] and analysis methods with artificial intelligence and
[31] equations also presented very high new criteria for the selection of mathematical
performance. In contrast, the use of the equations models.
developed by [52], [47], and [49] is not
recommended, and in the case of their use, they 5. Declaration of competing interest
should be under specific conditions because they We declare that we have no significant competing
can produce inaccurate results. This new approach interests, including financial or non-financial,
made it possible to observe that, under certain professional, or personal interests, interfering with
conditions, the Colebrook equation is not the most the full and objective presentation of the work
accurate at present. described in this manuscript.
With the GEP, it was possible to provide a new
model to determine the explicit friction factor f (R,
6. Acknowledgements
Acknowledgements for the collaboration of the [1] H. L. F. Ospina, G. M. E. López, C. A. Palacio,
Applied Hydraulics Research Group of the and M. J. F. Jiménez, “Dispositivo de Reynolds
Universidad Catholica Sedes Sapientiae. para el estudio reológico de fluidos”, Revista
Colombiana de Materiales, vol. 3, pp. 63-79,
7. Funding 2012. [Online]. Available:
The authors received no financial support for the https://doi.org/10.17533/udea.rcm.13227
research, authorship, and/or publication of this [2] W. H. Alawee et al., “Variation of Coefficient
article. of Friction and Friction Head Losses Along a
Pipe with Multiple Outlets”, Water, vol. 12, no.
8. Author contributions 3, pp. 2-15, 2020. [Online]. Available:
Maiquel López Silva: Conceived and designed the https://doi.org/10.3390/w12030844
analysis, the scientific literature review, the [3] F. J. Mejía, “Relación de las curvas de energía
statistical analysis, and interpretation of data, the Específica y pendiente de fricción con Las
formulation of gene expression programming zonas de flujo libre en canales” Revista EIA,
algorithms, prepared the text. vol. 9, pp. 69-75, 2008. Escuela de Ingeniería
Dayma Carmenates Hernández: Conceived and de Antioquia, Medellín, Colombia.
designed the analysis, assisted with the scientific [4] J. A. Gómez, P. J. García, and C. Nolasco,
literature review, statistical analysis, and “Modelo numérico de detección de fugas para
interpretation of data, and prepared the text and sistema de tuberías”, AiBi Revista de
edited the manuscript. Investigación, Administración e Ingeniería,
Nancy Delgado Hernández: Scientific literature 8(2), 113-120, 2020. [Online]. Available:
review, digital processing of data, formulation of https://doi.org/10.15649/2346030X.723
gene expression programming algorithms, and [5] U. C. A. García, E. O. Ladino, M. C. García,
Newton-Raphson programming in Python. “Determination of the Inside Diameter of
Nataly Chunga Bereche: Scientific literature Pressure Pipes for Drinking Water Systems
review, digital processing of data, formulation of Using Artificial Neural Networks,” Revista
gene expression programming algorithms, and Facultad de Ingeniería, vol. 31 no. 59, 4037,
Newton-Raphson programming in Python. 2022. Online]. Available:
https://doi.org/10.19053/01211129.v31.n59.2
9. Data availability statement 022.14037
The origin of the data is from the turbulent regime, [6] C. F. Colebrook, “Turbulent Flow in Pipe with
with different conditions of relative roughness (/D) Particular Reference to the Transition Region
from 10-6 to 5x10-2 and the Reynolds number from Between the Smooth and Rough Pipe Laws”,
4000 to 108, which implied a base of 47601 data Journal of the Institution of Civil Engineers,
points. vol. 11, no. 4, pp. 133–156, 1936. [Online].
The authors confirm that the data supporting the Available:
findings of this study are available within the https://doi.org/10.1680/ijoti.1939.13150
article and its supplementary materials. [7] T. R. Minhoni et al., “The performance of
explicit formulas for determining the Darcy-
References Weisbach friction factor”, Engenharia
Agrícola, Jaboticabal, vol. 40, no. 2, pp. 258- vol. 29, pp. 1596-1602, 2011. [Online].
265, 2020. [Online]. Available: Available:
https://doi.org/10.1590/1809-4430- https://doi.org/10.1080/10916461003620453
Eng.Agric.v40n2p258-265/2020 [15] G. Srbislav, and J. Branislav,
[8] E. O. Ladino, C. A. García, and M. C. García, “Reconsideration of the friction factor data and
“Darcy-Weisbach resistance coefficient equations for smooth, rough and transition pipe
determination using Newton-Raphson flow”. ITM Web of Conf. 29, 2019. [Online].
approach for android 4.0”, Revista Tecnura, Available:
vol. 23, no. 60, pp. 52-58, 2019. [Online]. https://doi.org/10.1051/itmconf/20192902001
Available: [16] R. Vatankhah, “Approximate analytical
https://doi.org/10.14483/22487638.14929 solutions for the Colebrook equation”, Journal
[9] L. F. Moody, “An approximate formula for of Hydraulic Engeneering, vol. 144, no. 5,
pipe friction factors”, Transactions of the 2018. [Online]. Available:
ASME, vol. 69, pp. 1005-1006, 1947. https://doi.org/10.1061/(ASCE)HY.1943-
[10] P. K. Swamee, and A. K. Jain, “Explicit 7900.0001454
equations for pipe-flow problems”, Journal of [17] D. Brkić and P. Praks, “Accurate and
the Hydraulics Division, vol. 102, pp. 657-664, efficient explicit approximations of the
1976. [Online]. Available: Colebrook flow friction equation based on the
https://doi.org/10.1061/JYCEAJ.0004542 Wright ⍵-function”, Mathematics, vol. 7 no. 1,
[11] D. J. Zigrang and N. D. Sylvester, “Explicit pp. 2-15, 2019. [Online]. Available:
Approximations to the Solution of Colebrook’s https://doi.org/10.3390/math7010034
Friction Factor Equation”, American Institute [18] P. Praks and D. Brkić, “Review of new
of Chemical Engineers, AIChE Journal, vol. flow friction equations: Constructing
28, no. 3, pp. 514-515, 1982. [Online]. Colebrook explicit correlations accurately”,
Available: Revista Internacional de Métodos Numéricos
https://doi.org/10.1002/aic.690280323 para Cálculo y Diseño en Ingeniería, vol. 36,
[12] E. Romeo et al., “Improved explicit no. 3, pp. 1-13, 2020. [Online]. Available:
equation for estimation of the friction factor in https://doi.org/10.23967/j.rimni.2020.09.001
rough and smooth pipes”, Chemical [19] C. Yu et al., “Approximate approach for
Engineering Journal, vol. 86, no. 3, pp. 369- improving pressure attenuation accuracy
374, 2002. [Online]. Available: during hydraulic transients”, Water Supply,
http://dx.doi.org/10.1016/S1385- vol. 22, no. 3, pp. 3387-3398, 2022. [Online].
8947(01)00254-6 Available:
[13] G. Papaevangelou et al., “New explicit https://doi.org/10.2166/ws.2021.394
relation for friction coefficient f in the Darcy- [20] S. C. Assunção, M. Dykenlove, J. C. H. von
Weisbach equation”, In Proc. 7th Annu. Conf. Hohendorff Filho, D. J. Schiozer, M. Souza,
Protection and Restoration of the Environment, “Friction factor equations accuracy for single
vol. 166, pp. 1-7, 2010. and two-phase flows”, Proceedings of the
[14] D. Brkić, “An explicit approximation of ASME 2020 39th International Conference on
Colebrook’s equation for fluid flow friction Ocean, Offshore and Arctic Engineering.
factor”. Petroleum Science and Technology, OMAE2020. Virtual Online, August 3-7, 2020.
[21] B. Achour, A. Bedjaoiu, M. Khattaaoui, [Online]. Available:
and M. Debabeche, “Contribution au calcul des https://doi.org/10.1155/2022/8116282
ecoulements uniformes surface libre et en [28] M. Milošević et al., “Hydraulic Losses in
charge”. [Contribution for the computation of Systems of Conduits with Flow from Laminar
open channel and pressurized flow]. Larhyss to Fully Turbulent: A New Symbolic
Journal, 7-36, 2002. Regression Formulation”, Axioms, vol. 11, no.
[22] L. Zeghadnia et al., “Explicit solutions for 5, pp. 2-11, 2022. [Online]. Available:
turbulent flow friction factor: A review, https://doi.org/10.3390/axioms11050198
assessment and approaches classification”, Ain [29] S. Samadianfard et al., “Determining flow
Shams Engineering Journal, vol. 10, no. 1, pp. friction factor in irrigation pipes using data
243-252, 2019. [Online]. Available: mining and artificial intelligence approaches”,
https://doi.org/10.1016/j.asej.2018.10.007 Applied Artificial Intelligence, vol. 28, no. 8,
[23] X. Fang et al., “New correlations of single- pp. 793-813, 2014. [Online]. Available:
phase friction factor for turbulent pipe flow and https://doi.org/10.1080/08839514.2014.95292
evaluation of existing single-phase friction 3
factor correlations”, Nuclear Engineering and [30] M. Najafzadeh et al., “Prediction of the
Design, vol. 241, pp. 897-902, 2011. friction factor in pipes using model tree”, ISH
[24] B. Achour and L. Amara, “New Journal of Hydraulic Engineering, vol. 24 no.
Formulation of The Darcy-Weisbach Friction 1, pp. 9-15, 2018. [Online]. Available:
Factor” LARHYSS Journal, vol. 17, no. 3, pp. https://doi.org/10.1080/09715010.2017.13339
90-100, 2020. [Online]. Available: 26
https://www.asjp.cerist.dz/en/article/134661 [31] U. H. Offor, and S. B. Alabi, “An Accurate
[25] W. A. Khan, “Numerical simulation of and Computationally Efficient Explicit
Chun-Hui He’s iteration method with Friction Factor Model”, Advances in Chemical
applications in engineering,” International Engineering and Science, vol. 6, pp. 237-245,
Journal of Numerical Methods for Heat and 2016.
Fluid Flow, vol. 32, no. 3, pp. 944–955, 2022. [32] E. Temizhana et al., “Which Correlation
[Online]. Available: Coefficient Should Be Used for Investigating
https://doi.org/10.1108/HFF-04-2021-0245 Relations between Quantitative Variables”,
[26] E. M. Ladino et al., “Modelado del factor American Academic Scientific Research
de fricción en tuberías a presión utilizando Journal for Engineering, Technology, and
redes neuronales de aprendizaje bayesiano”, Sciences, vol. 81, no. 1. pp. 265-277, 2022.
Ciencia en Desarrollo, vol. 13 no. 1, 2022. [33] H. Akaike, “A new look at the statistical
[Online]. Available: model identification”, IEEE Transactions on
https://doi.org/10.19053/01217488.v13.n1.20 Automatic Control, vol. 19, no. 6, pp. 716-723,
22.13241 1974. [Online]. Available:
[27] W. A. Khan et al., “Numerical and https://doi.org/10.1109/TAC.1974.1100705
Theoretical Investigation to Estimate Darcy [34] G. Schwarz, “Estimating the Dimension of
Friction Factor in Water Network Problem a Model”, The Annals of Statistics, vol. 6, no.
Based on Modified Chun-Hui He’s Algorithm 2, pp. 461-464, 1978.
and Applications”, Hindawi, pp. 2-11, 2022.
[35] C. Mallows, “Some comments on CP”, [45] S. E. Haaland, “Simple and explicit
Technometrics, vol. 15, no. 4, pp. 661-675, formulas for the friction factor in turbulent pipe
1973. flow”, Journal of Fluids Engineering, vol. 105,
[36] S. Genić et al., “A Review of Explicit no. 1, pp. 89-90, 1983. [Online]. Available:
Approximations of Colebrook’s Equation”. https://doi.org/10.1115/1.3240948
FME Transactions, vol. 39, pp. 67-71, 2011. [46] G. Manadilli, “Replace implicit equations
[37] D. Altshul, “Hydraulic Friction Losses in with signomial functions”. Chemical
the Piping”. Gosenergoizdat, Moscow 1963. Engineering Journal, vol. 104, pp. 129-130,
[38] A. Avci, and I. Karagoz, “A novel explicit 1997.
equation for friction factor in smooth and [47] J. McKeon et al., “A new friction factor
rough pipes”, Journal of Fluids Engineering, relationship for fully developed pipe flow”,
vol. 131, no. 6, pp. 1-4, 2009. Journal of Fluid Mechanics, no. 538, pp. 429-
[39] D. Barr, “Solutions of the Colebrook-White 443, 2005.
functions for resistance to uniform turbulent [48] F. Pavlov, P. G. Romankov, A. A. Noskov,
flows”, in Proc. The Institution of Civil “Problemas y ejemplos para el curso de
Engineers, vol. 71, no. 2, 1981, pp. 529-535. operaciones básicas y aparatos en tecnología
[Online]. Available: química”, ed. Mir, Moscú, URSS, 1981.
https://doi.org/10.1680/iicep.1981.1895 [49] G. F. Round, “An explicit approximation
[40] N. H. Chen, “An explicit equation for for the friction factor – Reynolds number
friction factor in pipe”, Ind Eng Chem Fund, relation for rough and smooth pipes”, The
vol. 18, no. 3, pp. 296-297, 1979. [Online]. Canadian Journal of Chemical Engineering,
Available: vol. 58, no. 1, pp. 122-123, 1980.
https://doi.org/10.1021/i160071a019 [50] M. Shacham, “An explicit equation for
[41] S. W. Churchill, “Empirical expressions for friction factor in pipe”. Industrial &
the shear stress in turbulent”. AIChE Journal, Engineering Chemistry Fundamentals, vol. 19,
375-376, 1973. [Online]. Available: no. 2, 1980.
https://doi.org/10.1002/aic.690190228 [51] A. K. Jain, (1976). Accurate explicit
[42] S. W. Churchill, “Friction Factor Equation equation for friction factor. Journal of the
Spans All Fluid Flow Regimes”, Chemical Hydraulics Division, vol. 102, no. 5, pp. 674-
Engineering, vol. 84, no. 24, pp. 91-92, 1977. 677, 1976. [Online]. Available:
[43] J. Eck, “Use of a smoothed model for pipe https://doi.org/10.1061/JYCEAJ.0004544
friction loss”, Journal of Hydraulic [52] J. Wood, “An explicit friction factor
Engineering, vol. 143, no. 1, 2017. [Online]. relationship”. Civil Eng. ASCE vol. 60, 1966.
Available: [53] G. A. Olivares et al., “New explicit
https://doi.org/10.1061/(ASCE)HY.1943- correlation to compute the friction factor under
7900.0001239 turbulent flow in pipes”, Brazilian Journal of
[44] A. Ghanbari et al., “Newly developed Agricultural and Environmental Engineering,
friction factor correlation for pipe flow and vol. 25, no.7, pp. 439-445, 2021. [Online].
flow assurance”, Journal of Chemical Available:https://doi.org/10.1590/1807-
Engineering and Materials Science, vol. 2, no. 1929/agriambi.v25n7p439-445
6, pp. 83–6, 2011.

Revista Facultad de Ingeniería Universidad de Antioquia

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Revista Facultad de Ingeniería Universidad de Antioquia

Uploaded by

Copyright:

Available Formats

Revista Facultad de Ingeniería, Universidad de Antioquia, No.xx, pp.

x-xx, xxx-xxx 20xx

Revista Facultad de Ingeniería

Authors: Maiquel López-Silva, Dayma Sadami Carmenates-Hernández, Nancy Delgado-Hernández and

To appear in: Revista Facultad de Ingeniería Universidad de Antioquia

Received: August 16, 2022

Please cite this article as: M. López-Silva, D. S. Carmenates-Hernández, N. Delgado-Hernández and N.

Explicit pipe friction factor equations: evaluation, classification, and proposal

Maiquel López-Silva1* https://orcid.org/0000-0002-0946-6160

Corresponding author: Maiquel López-Silva

1. Introduction alternative to the Colebrook equation, but the

Limit: 4000 <Re <108 and 10-6 <ε/D <5·10-2

[11] Model II.

its simplicity and speed of convergence to solve  ∑ fCW − f Pr oposed  2p

Figure 1 Explicit model specifications.

Table 1 Preference models

Table 3 Efficient model of the GEP

GEP1 0.219 ( 0.028 ε D ) + ( 0.896 R ) 

4. Conclusions ε/D) with the lowest degree of complexity in the

You might also like