AIChe Letter: Transport Phenomena and Fluid Mechanics

A New Six Parameter Model to Estimate the

Friction Factor
Lamberto Díaz-Damacillo


Instituto Carrillo Cárdenas, Chilapa de Alvarez, Guerrero 41100, Mexico

Gabriel Plascencia
Universidad Tecnológica de Querétaro, Querétaro, Querétaro 76148, Mexico

DOI 10.1002/aic.16535
Published online January 28, 2019 in Wiley Online Library (wileyonlinelibrary.com)

Significance

A new explicit formula for estimating the friction factor using six parameters is proposed. The model was set up by considering
the effect of residual stresses in the flow by two distinct contributions: the first is attributed to the flow velocity (Reynolds num-
ber) and the second to the duct roughness. Compared to other models, this new equation gives the best fit with Nikuradse’s
results. A new model to calculate the friction is proposed. The model is based on assuming the residual stresses due to the lam-
inar to turbulent flow transition by two distinct contributions: the first is attributed to the flow velocity (Reynolds number) and
the second to the duct roughness. Compared to other models, this new equation gives the best fit with respect of Nikuradse’s
results. The model does not consider the effect of pipe wall on the velocity distribution. © 2019 American Institute of Chemical
Engineers AIChE J, 65: 1144–1148, 2019
Keywords: friction factor, flow in ducts, Colebrook equation, Moody chart, Reynolds number

Background error of each of these correlations with respect of Nikuradse

data.19 The error was calculated as follows:

F riction factor is used to estimate loses in mechanical

energy in flows flowing through pipelines. Friction factor
is commonly computed using either Colebrook equation1
or Moody’s chart.2 However, in spite of its usefulness, these
%Error = 100 ×
fNikuradse − fmodel
fNikuradse
ð1Þ

calculation methods are limited due to their own nature.

Colebrook equation is an implicit model that requires
numerical methods to determine the value of the friction fac- Proposed Model
tor. Moody’s chart is a graphical representation of Colebrook Moody’s chart2 clearly shows that under the laminar flow
equation, which use is limited to Reynolds numbers at around regime, the friction factor decreases linearly with the inverse
or below 10.3 of Reynolds number; when Re is about 2000, there is a break
Because of the complications cited above, several authors3-11 in this lineal behavior, and the friction factor increases. For
have proposed different models to estimate the friction factor. rougher pipe walls, the friction factor exhibits higher values
The majority of the models proposed are explicit, i.e., they than smoother pipes. As the flow approaches full turbulent
allow to estimate values for the friction factor directly as a func- behavior (increase in Re), the friction factor tends to be con-
tion of both, the Reynolds number and the relative roughness of stant for every value of relative roughness of the pipe. We
the pipe conducting the flow. Some good reviews on this partic- assume that a second transition in the flow occurs at the point
ular subject matter are also available in the literature.12-18 where the friction factor stabilizes, irrespective of the Reyn-
Table 1 shows some of the correlations proposed to calculate olds value. Specific details on the development of this model
the friction factor in ducts. The table also shows the relative are going to be fully disclosed in an upcoming publication.
In this report, we present the following model to determine
the friction factor in pipe networks:
Additional Supporting Information may be found in the online version of this
article.
64 λ1 λ
f= +
+
2  ð2Þ
Correspondence concerning this article should be addressed to G. Plascencia at Re 1 + exp τ1 − Re 1 + exp τ2 − Re
×ε
[email protected] 100 600 D

© 2019 American Institute of Chemical Engineers For smooth pipes, Eq. (2) reduces to

1144 April 2019 Vol. 65, No. 4 AIChE Journal

 Table 1. Summary of Models Available in Literature to Compute the Friction Factor in Ducts
Ref. Model Validity Error [%]
h  ε 1:11 i
3 p1 ffi = −1:8 log 6:9 + 3:7D 4 × 10 < Re < 10
3 8
43.86
f Re 10−6 < ε/D < 0.05
   
4  ε 0:9924
5:3326 0:9345 3 × 103 < Re < 1.5 × 108 10.94
p1 ffi = −2log 3:7065D
ε
− 5:0272 log ε
− 4:567
log +
f Re 3:827D Re 7:7918D 208:815 + Re 0 < ε/D < 0.05

h h h iii
5 p1 ffi = −2log 3:7D
ε
− 5:02 ε ε N/A 10.91
Re log 3:7D − Re log 3:7D + Re
5:02 13
f

h h  ε 1:1098 5:8506 ii
6 p1 ffi = −2log 3:7065D
ε
− 5:0452 1
log 2:8257 + 0:8981 4 × 103 < Re < 108 11.17
f Re D Re 10−6 < ε/D < 0.05
7 2 3121 4 × 103 < Re < 108 11.64
6 7 10−6 < ε/D < 0.05
6 7
6 7
6
12 7
6 8 7
f = 8 × 6 Re + 20 0
1
1116 332 7
6 7
6
16 7
6 6B B CC 77
4 4@2:457 ln@
10:9 AA + Re 5 5
37530

7 + 0:27 ε
Re D

h
 1:1007 i −2
8 3 × 103 < Re < 108 17.15
f = 1:613 ln 0:34 Dε − 60:525
+ 56:291
Re 1:1105
Re 1:0712
0 < ε/D < 0.05

9 4 × 103 < Re < 108 11.18

p1 ffi = 0:8686ln 0:4587
Re ; S = 0:124 Dε Re + lnð0:4587ReÞ
f SðS + 1Þ
S 10−6 < ε/D < 0.05
2 0 1 3
10 N/A 12.62
62:18 B C 7
p1 ffi = −2 log6 B
4 Re ln@
Re C+ ε 7
A 3:71D5
f
1:1 Re
1:816 ln
lnð1 + 1:1 ReÞ
pffiffiffi
11 p1 ffi = 1:930logð Re f Þ−0:537 N/A 12.78
f
12-14 2 3121 N/A 11.75
6 7
6 7
6 7
6
 7
6 64 12 7
f =6
6 Re + 20 0
1
1116
7
32 7
3
6 7
6
16 7
6 6B B CC 77
4 4@0:8687 ln@ 1
AA + 13269
Re 5 5
ð lnð ReÞÞ
1:282
0:883
+ 0:27 ε − 110 ε
Re1:007 D
Re D

64 λ1 1. Flows with Re < 2000 are laminar and the friction

f= +
 ð2aÞ
Re 1 + exp τ1 − Re factor varies linearly with the inverse of Re, according to
100 f = 64/Re.
2. When the flow’s Reynolds number is around 3000, the
Equations 2 and 2a were developed after close inspection of flow transitions from laminar to turbulent; the friction factor
McKeon et al,11 Nikuradse,19 and Swanson et al20data sets. tends to increase as Re does so. This behavior is due to the
However, in spite of being considered by many as the appearance of velocity fluctuations in the turbulent regime that
benchmark for explaining flow phenomena over rough pipe produce collisions within the flow, consequently increasing
walls, Nikuradse work has been severely criticized.21,22 the friction factor. It is proposed24,25 that the increase in fric-
Among the main arguments against Nikuradse is that it is tion factor follows a sigmoidal type of function; it has not
strongly believed22 that he manipulated his experimental data been developed any mathematical expression that relates the
in order to fit it to von Karman analytical expression for the friction factor to the flow collision frequency. However, it
friction factor and velocity distribution near the pipe wall. would be expected that these two quantities were proportional;
However, later on, it was demonstrated23 that indeed there if this is true, then it is proposed that the second term in Eq. 1
was some data handling but not to disprove his work as a would be sigmoidal in nature. Therefore, it is likely that at cer-
whole.21,23 tain regions the model proposed here does not fit precisely the
In any case, it is observed in Nikuradse’s data set that at reported data; this mainly due to this first approach.
lower r/ε ratios, more abrupt transitions on the friction factor 3. At Reynolds numbers considerable higher than 3000, the
occur as Re increases. Such transitions are due to the presence turbulent boundary layer thins enough that irregularities in the
of stresses developed between the pipe wall and the flow itself. pipe surface produce another set of collisions within the flow,
Our model explains such transitions by virtue of four (λ1, λ2, therefore it is necessary to include another term considering
τ1, τ2) parameters in Eq. 2. This model is not the result of any this effect in the description of the friction factor. This model
statistical data fitting, rather than that, this model is built upon considers that this new term can be modeled using a sigmoidal
three basic premises: function. As the pipe surface is more irregular, the number of

AIChE Journal April 2019 Vol. 65, No. 4 Published on behalf of the AIChE DOI 10.1002/aic 1145
Figure 1. (A) Comparison between Nikuradse19 data and the predictions of the proposed model (Eq. 2).
(B) Comparison between McKeon et al11 and Swanson et al20 data and the predictions of the proposed
model (Eq. 2a).
[Color figure can be viewed at wileyonlinelibrary.com]

perturbations in the flow will occur at a faster rate, hence, the 0:77505 10:984
need of a sigmoidal function in accounting for this effect. τ2 =  ε 2 − ε + 7953:8 ð4Þ
In Eq. 2, Re is Reynolds number; ε/D is the relative rough- D D
ness of the pipe; λ1 is the residual stress contribution from the
In spite of the different nature of these two transitions, both
laminar to turbulent transition to the friction factor; λ2 is the of them contribute similarly to the magnitude of friction factor.
residual stress contribution from the pipe roughness to the fric- Figure 1A shows a graph of our model compared to Nikur-
tion factor; τ1 is Reynolds number at which occurs the first adse19 experimental points; it can be noticed in this figure that
transition in the friction factor; and τ2 is Reynolds number at there is good agreement between our model (Eq. 2) and all of
which the second transition occurs. Nikuradse data sets. For every r/ε condition, our model accu-
The first (τ1) transition refers to the change in flow regime rately represents the laminar flow condition as well as the
from laminar to turbulent. Friction factor for laminar flow is laminar–turbulent transition. The model also shows good fit-
proportional to the inverse of Reynolds; as the flow regime ting with respect of the fully turbulent flow. Additionally, in
turns into turbulent, such transition occurs for every pipe Figure 1B, we compare our model (Eq. 2a) to smooth pipe
roughness (even for smooth ones) and it is intrinsic to the flow experimental friction data obtained by McKeon et al11 and
itself; thus it is unavoidable its presence. The second transition Swanson et al.20 From this figure, it is clear that our model
(τ2) occurs at higher Reynolds numbers, and occurs at differ- represents accurately the experimental values already reported
ent Re values. As pipe rugosity increases, this second transi- up to Re ~10,6 the transition from laminar to turbulent flow is
tion will take place faster and the increase in friction factor also well depicted by the model; however, as Re increases up
will be more noticeable. This second transition is intimately to 107 (Princeton data), considerable deviations between the
related to pipe rugosity. model and the data set become more evident. Such deviations
Of these parameters, λ1 and τ1 are constant for all flow and may arise from the construction of our model; we consider that
relative roughness conditions, and their numerical values are upon transitioning to turbulent flow, given the randomness of
0.02 and 3000, respectively. In contrast, both λ2 and τ2 they turbulence, the flow is better depicted by a sigmoidal function
do depend on the relative roughness of the pipe. The expres- rather than a continuous function as proposed by McKeon
sions for these parameters are et al.11
!2 Figure 2A,B compares the results of some of the models
1 shown in Table 1 with the experimental data of Nikuradse19
λ2 = λ1 −  1  ð3Þ
− 2log 3:7065 × Dε and Eq. 2 for r/ε = 507 and r/ε = 15, respectively. From these
figures, it can be noticed that for very rough pipes (r/ε = 15),

1146 DOI 10.1002/aic Published on behalf of the AIChE April 2019 Vol. 65, No. 4 AIChE Journal
Figure 2. (A) Comparison between Nikuradse data and predictions from different models used for the condition r/
ε = 507. (B) Comparison between Nikuradse data and predictions from different models used for the condi-
tion r/ε = 15.
[Color figure can be viewed at wileyonlinelibrary.com]

Fang et al model8 and the modified Churchill model12-14 pre- increases, so does the deviation between the model11 and the
sent the most considerable deviations from Nikuradse’s data. experimental data.19
The other models show a lesser degree of discrepancy with Figure 3 shows the comparison between our model (Eq. 2),
respect of the data set used. Since McKeon model was McKeon et al model11 and Nikuradse’s data set19 for the
constructed for smooth pipes, it is not even considered for smoothest pipe wall condition (r/ε = 507). As seen from this
comparison to the data set used. On the other hand, for the figure, McKeon’s model has better agreement with the experi-
smoothest condition (r/ε = 507), the different models pro- mental data than our model for Re values between 104 and
posed show lower deviation with respect of Nikuradse’s data. 10.6 However, for Re >10,6 our model predictions are closer
Even McKeon’s model11 shows excellent agreement with to the experimental data, compared to McKeon’s model; that
Nikuradse’s experimental points up to Re ~105; as Re model rapidly deviates from the measured points as Re
increases. This behavior can be explained in terms of the third
term in our model; such term considers the effect of the pipe
rugosity regardless of its numerical value as a consequence the
model results in a line closer to the experimental points; in the
case of McKeon’s model, it does not consider at all the rugos-
ity and thus its model continuously decreases as Re increases.

Final Remarks
A new explicit model to estimate the friction factor in flows
was developed based on the two distinctive contributions to
the flow in pipes: the first corresponds to the increase in Reyn-
olds number (transition laminar–turbulent flow regime). The
second transition is attributed to the actual roughness of the
pipe wall and acts faster on the flow than the Reynolds
contribution.
The model was validated in part by using Nikuradse experi-
mental data19; even though there are certain uncertainties
Figure 3. Comparison between Nikuradse experimental regarding the accuracy of that data; as Churchill and Chan21
data set, Eq. (2) and McKeon et al11 model for pointed out, in spite of its imprecisions, Nikuradse data still
the friction factor using r/ε = 507. are the most complete results available to us. Beattie23 con-
[Color figure can be viewed at wileyonlinelibrary.com] firms the utility of Nikuradse results although is well aware of

AIChE Journal April 2019 Vol. 65, No. 4 Published on behalf of the AIChE DOI 10.1002/aic 1147
the possible manipulation of them. In spite of the criticism 12. Taler D. Determining velocity and friction factor for turbulent flow in
received by Nikuradse work, we confirm in this article that the smooth tubes. Int J Therm Sci. 2016;105:109-122. https://doi.org/10.
1016/j.ijthermalsci.2016.02.011.
actual results from him do not change considerably. If in the
13. Taler D. Developed Turbulent Fluid Flow in Ducts with a Circular
near future new experimental results become available, our Cross-Section. Switzerland: Springer International Publishing; 2019.
model still will be valid since it captures in a simple manner https://doi.org/10.1007/978-3-319-91128-1_6
the physics behind the friction factor as the sum of three dis- 14. Rennels DC, Hudson HM. Pipe Flow: A Practical and Comprehensive
tinct contributions, as shown in Eq. 2. Guide.; 2012. https://doi.org/10.1002/9781118275276
15. Jaric M, Kolendi
c P, Jari
c M, Budimir N, Geni
c VA. Review of
explicit approximations of Colebrook’s equation. FME Trans. 2011;
Acknowledgment 39:67-71.
The authors would like to thank SIP-IPN grant 20180074 for support- 16. Pimenta BD, Robaina AD, Peiter MX, Mezzomo W, Kirchner JH,
ing this research. Ben LHB. Performance of explicit approximations of the coefficient of
head loss for pressurized conduits. Rev Bras Eng Agrícola e Ambient.
Literature Cited 2018;22(5):301-307.
17. Anaya-Durand AI, Cauich-Segovia GI, Funabazama-Bárcenas O,
1. Colebrook CF, White CM. Experiments with fluid friction in rough- Gracia-Medrano-Bravo VA. Evalucion de ecuaciones de factor de fric-
ened pipes. Proc R Soc A Math Phys Eng Sci. 1937;161(906):367-381. cion explicito para tuberias. Educ Quim. 2014;25(2):128-134. https://
https://doi.org/10.1098/rspa.1937.0150. doi.org/10.1016/S0187-893X(14)70535-X.
2. Moody LF. Friction factors for pipe flow. Trans ASME. 1944;66(8): 18. Yildirim G. Computer-based analysis of explicit approximations to the
671-684. implicit Colebrook-White equation in turbulent flow friction factor cal-
3. Chen NH. An explicit equation for friction factor in pipe. Ind Eng Chem culation. Adv Eng Softw. 2009;40(11):1183-1190. https://doi.org/10.
Fundam. 1979;18(3):296-297. https://doi.org/10.1021/i160071a019. 1016/j.advengsoft.2009.04.004.
4. Haaland SE. Simple and explicit formulas for the friction factor in tur- 19. Nikuradse J. Laws of flow in rough pipes. Forsch auf dem Gebiet
bulent pipe flow. J Fluids Eng. 1983;105(1):89-90. https://doi.org/10. des Ingenieurwesens A. 1933;4(1):44–64. https://doi.org/10.1063/1.1715007.
1115/1.3240948. 20. Swanson CJ, Julian B, Ihas GG, Donelly RJ. Pipe flow measurements
5. Romeo E, Royo C, Monzón A. Improved explicit equations for estima- over a wide range of Reynolds numbers using liquid helium and vari-
tion of the friction factor in rough and smooth pipes. Chem Eng J. ous gases. J Fluid Mech. 2002;461:51-60. https://doi.org/10.1017/
2002;86(3):369-374. https://doi.org/10.1016/S1385-8947(01)00254-6. S0022112002008595.
6. Zigrang DJ, Sylvester ND. Explicit approximations to the solution of 21. Churchill SW, Chan C. Improved correlating equations for the friction
Colebrook’s friction factor equation. AIChE J. 1982;28(3):514-515. factor for fully turbulent flow in round tubes and between identical
7. Churchill SW. Friction - factor equation spans all fluid - flow regimes. parallel plates, both smooth and naturally rough. Ind Eng Chem Res.
Chem Eng. 1977;84(24):94-95. 1994;33(8):2016-2019. https://doi.org/10.1021/ie00032a018.
8. Fang X, Xu Y, Zhou Z. New correlations of single-phase friction fac- 22. Robertson JM, Martin JD, Burkhart TH. Turbulent flow in rough
tor for turbulent pipe flow and evaluation of existing single-phase fric- pipes. Ind Eng Chem Fundamen. 1968;7(2):253-265. https://doi.
tion factor correlations. Nucl Eng Des. 2011;241(3):897-902. https:// org/10.1021/i160026a014.
doi.org/10.1016/j.nucengdes.2010.12.019. 23. Beattie DRH. In defence of Nikuradse. In: Twelfth Australasian Fluid
9. Sonnad JR, Goudar CT. Using a mathematically exact alternative to Mechanics Conference. Sidney; 1995:561–564.
the Colebrook–White equation. J Hydraul Eng. 2006;132(8):863-867. 24. Sundaram S, Collins LR. Collision statistics in an isotropic particle-
https://doi.org/10.1061/(ASCE)0733-9429(2006)132:8(863). laden turbulent suspension. Part 1. Direct numerical simulations. J Fluid
10. Brki
c D. Review of explicit approximations to the Colebrook relation Mech. 1997;335:75-109. https://doi.org/10.1017/S0022112096004454.
for flow friction. J Petrol Sci Eng. 2011;77(1):34-48. https://doi. 25. Avila K, Moxey D, De Lozar A, Avila M, Barkley D, Hof B. The
org/10.1016/j.petrol.2011.02.006. onset of turbulence in pipe flow. Science. 2011;333(6039):192-196.
11. McKeon BJ, Swanson CJ, Zagarola MV, Donnelly RJ, Smits AJ. Fric- https://doi.org/10.1126/science.1203223.
tion factors for smooth pipe flow. J Fluid Mech. 2004;511:41-44.
https://doi.org/10.1017/S0022112004009796. Manuscript received Jul. 18, 2018, and revision received Dec. 27, 2018.

1148 DOI 10.1002/aic Published on behalf of the AIChE April 2019 Vol. 65, No. 4 AIChE Journal