See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/333477898

Roughness effects in laminar pipe flow

Preprint · May 2019

CITATIONS READS
0 577

2 authors:

Utku Şentürk Alexander J Smits

Ege University Princeton University
30 PUBLICATIONS   154 CITATIONS    507 PUBLICATIONS   12,145 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Thermally Stratified Turbulence View project

Fast, efficient swimming View project

All content following this page was uploaded by Alexander J Smits on 24 April 2020.

The user has requested enhancement of the downloaded file.

 Roughness effects in laminar pipe flow

Utku Senturka,∗, Alexander J. Smitsb

a Department of Mechanical Engineering, Ege University, Izmir 35040, Turkey
arXiv:1905.12479v1 [physics.flu-dyn] 29 May 2019

b Department of Mech. and Aerospace Eng., Princeton University, Princeton 08544, USA

Abstract

The impact of wall roughness on fully developed laminar pipe flow is investigated
numerically. The roughness is comprised of square bars of varying size and
pitch. Results show that the inverse relation between the friction factor and
the Reynolds number in smooth pipes still persists in rough pipes, regardless
of the rib height and pitch. At a given Reynolds number, the friction factor
varies quadratically with roughness height and linearly with roughness pitch.
We propose a single correlation for the friction factor that successfully collapses
the data.
Keywords: laminar, pipe, roughness, computational.

1. Introduction

The established framework of friction factor analysis in fully developed pipe

flow is due to the early work by Nikuradse [1] which reported experiments
on artificially roughened pipes. Three distinct flow regimes were identified: a
laminar flow regime where roughness has virtually no effect on the friction, and
a transitional regime and a fully turbulent regime where roughness becomes
a significant factor once it exceeds a critical size ([3, 4, 5, 6]). However, the
effects of roughness in laminar flows has been the subject of great interest to
the microscale flow community and it is now well established that roughness
leads to significant departures from Nikuradse’s initial proposal ([7, 8, 9, 10]).

∗ Correspondingauthor
Email address: [email protected] (Utku Senturk)

Preprint submitted to Fluid Dynamics Research May 30, 2019

 For a fully developed flow in a circular pipe of diameter D, the friction factor
is given by
4τw (−dp/dx) D
f= 1 2
= 1 2
(1)
2 ρV 2 ρV
where τw is the local wall shear stress, ρ is the fluid density and ν is its kinematic
viscosity, dp/dx is the pressure gradient, and V is the mean (bulk) velocity. For
smooth pipes, f = 64/Re where the Reynolds number is defined as Re = V D/ν.
Kandlikar et al. [7] suggested that the principal influence of roughness in
channel flow was that the effective channel height became d = D − 2k, where k
is the characteristic roughness height, and that their results could be correlated
using a constricted friction factor defined by

(−dp/dx) d
fc = 1 02
(2)
2 ρV

where V 0 = V D/d for a channel. The definition of the Reynolds number for a
channel remains unchanged since by mass conservation V D = V 0 d.
Kandlikar et al. also proposed that for pipe flows the same constricted
friction factor should apply, but with V 0 = V D2 /d2 . In addition, the effective
(constricted) Reynolds number is now different from Re, and it is given by

V 0d Re
Rec = = (3)
ν 1 − 2k/D

Hence, it was proposed that we can relate the rough pipe friction factor to the
smooth pipe friction factor according to
 4
fc f 2k
= 1− (4)
64/Rec 64/Re D

More recently, Liu et al. [14] investigated numerically a laminar channel

flow with square bar roughness of height k, width w and pitch λ. Here, k/D
was varied from 0.01 to 0.1, and λ/w ranged from 2 to 8. They found that,
in addition to roughness height, pitch was also important in determining the
friction factor, and that a good collapse of friction factors could be obtained by
the correlation  
0 λ
fc = fc 1+c (5)
w

2
with c = 0.0136 suggested by the authors for the best fit. At our suggestion, as
based on our findings given below, they were able to show that fc was inversely
proportional to Re, as is the case for smooth pipe and channel flow.
In this paper, we present the results of a detailed numerical study on the
influence of the relative height and the spacing of the roughness elements for a
fully developed laminar pipe flow with regular, square roughness elements for
various Reynolds numbers. Specifically, we look for a correlation in the form

f = f (Re, k/D, λ/w) (6)

which successfully captures such variations in f .

2. Problem description

We consider steady, incompressible and fully developed flow in a horizontal

pipe with square roughness elements on the wall (Figure 1a). With these as-
sumptions, we make use of the translational periodicity as described by Patankar
et al. [15] and employed by Herwig et al. [13]. For a brief description of the
modeling with translational periodicity, see Appendix A.
The effect of the roughness height is investigated by varying k/D from 0.005,
to 0.01, 0.025 and 0.05. For each of these heights, four different roughness
spacings are considered by setting λ/w = 2, 4, 6 and 8 so that the smallest
and largest spacings correspond to d-type and k-type roughness, respectively
[4]. These simulations are repeated for 10 Reynolds numbers ranging from 200
to 2000. To analyze shear and pressure forces separately, the horizontal and
vertical faces are considered separately, as shown in Figure 1b. Skin friction
coefficients on horizontal face pieces (BC and DA) are computed by,
τw
Cv = 1 2
(7)
2 ρum

and the pressure coefficient that represents the net pressure force on the rough-
ness element is computed using vertical face pieces (AB and CD)
pAB − pCD
Cp = 1 2
(8)
2 ρum

3
 (a)

(b)

Figure 1: Schematics of the problem: (a) Pipe section with square roughness elements. Col-
ored portion is adequate for the translational periodic flow. (b) Simulation domain, assigned
boundary conditions and pieces of the wall that contribute to pressure and shear.

Pressure and shear stress represent area-weighted average values in these defi-
nitions. Balance of forces on the domain gives,

(pup − pdown )Aup = (pAB − pCD )ACD + τBC ABC + τDA ADA (9)

where we used Aup = Adown and ABC = ACD for the upstream wetted area
and roughness lateral area, respectively. Non-dimensionalizing with dynamic
pressure and using the analogy with Equation 1, an effective friction factor can

4
Figure 2: An example of the unstructured finite volume grid used in simulations (λ/w = 8
and k/D = 0.005).

now be described:
f = αCp + γCv,BC + θCv,DA (10)
|{z} | {z }
fp fv

Here, coefficients group several terms related to the geometry, for example
α = DACD /(λAup ). An unstructured, non-uniform grid is used with trian-
gular elements clustered near the channel walls ( Figure 2). Using the pressure
gradient obtained from a converged simulation, friction factor is found. The grid
is successively refined until the change in the computed friction factor becomes
negligibly small. Second-order accurate discretization schemes are used for the
velocity (with upwinding) and the pressure.

3. Results

The effects of roughness height on the friction factor is given in Figure 3.

We show here only Re = 200 and 2000 with λ/w = 2, 4 and 8 for a clearer
representation, although similar trends were found for all combinations of Re
and λ/w. First remark in these curves is the relative increase in the friction
factor compared to a smooth pipe (64/Re) as the roughness height increases
which is in agreement with the previous findings ([9, 10, 13]). This relationship
is quadratic and is found to be consistent in all of the spacings considered
herein. Second, this effect amplifies as the spacing decreases, i.e., the roughness
character migrates from a k-type and turns into a d-type. Third, there is an

5
 λ/w = 2, Re = 200 λ/w = 2, Re = 200
1.5 1.5 λ/w = 4, Re = 200
λ/w = 4, Re = 200
λ/w = 8, Re = 200 λ/w = 8, Re = 200
λ/w = 2, Re = 2000
λ/w = 2, Re = 2000
λ/w = 4, Re = 2000
1.4 λ/w = 4, Re = 2000 1.4 λ/w = 8, Re = 2000
λ/w = 8, Re = 2000
smooth
smooth

1.3 1.3

f /(64/Re)
f /(64/Re)

1.2 1.2

1.1 1.1

1.0 1.0
0.00 0.01 0.02 0.03 0.04 0.05 0.0000 0.0005 0.0010 0.0015 0.0020 0.0025
k/D (k/D)2

(a) (b)

Figure 3: Computed friction factor normalized by that of a smooth pipe: (a) Friction factor
increases quadratically with roughness height at any given spacing λ/w and Re. (b) Results
plotted as a function of (k/D)2 .

inverse relation between Re and f resembling smooth pipe flow: For a given
roughness height and spacing, friction factor is larger at smaller Re. Finally,
the effect of spacing is less at high Re. Therefore it can be concluded that
introducing roughness on smooth pipe walls causes a significant increase in
the effective friction due to the Equation 10, which depends quadratically on
roughness height, k/D.
We find that the data do not demonstrate the dependence on k/D that
follows from the concepts of constricted flow proposed by Kandlikar et al. [7], as
expressed by Equation 4). Figure 4 shows that there is a significant dependence
on pitch, as found by Liu et al. [14]. However, the linear scaling suggested by
Liu et al. for a channel, as given by Equation 5, does not collapse our data for
the pipe (see Figure 4b. Here, we used c1 = 0.0136, as found by Liu et al., but
it is obvious that no value for c1 can be found to collapse the data properly.
Before addressing the influence of pitch on the friction factor in more detail,
it is useful to consider first the effects of Reynolds number. As illustrated in
Figure 5a for λ/w = 2, we find that for any given λ/w and k/D, the friction

6
 1.00 λ/w = 2, Re = 2000
λ/w = 4, Re = 2000
1.10 λ/w = 6, Re = 2000
0.99
λ/w = 8, Re = 2000

0.98
1.08
0.97
fcRec /64

fc′
0.96 1.06

0.95
1.04
0.94
λ/w = 2, Re = 2000
0.93 λ/w = 4, Re = 2000
λ/w = 6, Re = 2000
1.02
0.92 λ/w = 8, Re = 2000

0.00 0.01 0.02 0.03 0.04 0.05 0.00 0.01 0.02 0.03 0.04 0.05
k/D k/D

(a) (b)

Figure 4: (a) Constricted friction factor as defined by Equation 4 following Kandlikar et al.
[7]. (b) Constricted friction factor as defined by Equation 5 following Liu et al. [14].

factor exhibits a linear relation with 1/Re precisely as in smooth pipe flow.
At any given Re and λ/w, friction factor increases with roughness height, as
expected.
The dependence on pitch is shown in Figure 5b for Re = 200. The trend
where f is a decreasing function of spacing with an almost linear trend is com-
mon across all Reynolds numbers. In light of these findings, we use regression
analysis to fit the data and propose the following correlation for f on the pa-
rameters investigated here:
k 2 λ k



a1 D + a2 + a3 w D + 64
f= (11)
Re

Range of parameters in this work yielded a1 = 3000, a2 = 520 and a3 = −35.

Figure 6 shows the excellent collapse obtained by the proposed correlation in
Equation 11. Largest error in this correlation is 7% which is found at k/D = 0.05
and λ/w=8.
Owing to the simplicity of the geometry, pressure and shear components of
the friction factor can be analyzed using equation 10. Figure 8 shows that,
for a d-type roughness element, shear on walls BC and DA contribute more

7
 0.5
λ/w = 2, k/D = 0.005 0.48 Re = 200, k/D = 0.04
λ/w = 2, k/D = 0.01 Re = 200, k/D = 0.05
λ/w = 2, k/D = 0.025
λ/w = 2, k/D = 0.05
0.4
0.46

0.3
0.44

f
f

0.2
0.42

0.1
0.40

0.001 0.002 0.003 0.004 0.005 2 3 4 5 6 7 8

1/Re λ/w

(a) (b)

Figure 5: Dependence on Re and λ/w: (a) Friction factor is inversely related to Re as in

smooth pipe flow. (b) As the spacing increases, friction factor decreases linearly.

λ/w = 2, k/D = 0.005

0.5 λ/w = 2, k/D = 0.01
λ/w = 2, k/D = 0.025
λ/w = 2, k/D = 0.05
λ/w = 4, k/D = 0.005
λ/w = 4, k/D = 0.01
0.4 λ/w = 4, k/D = 0.025
λ/w = 4, k/D = 0.05
λ/w = 6, k/D = 0.005
λ/w = 6, k/D = 0.01
λ/w = 6, k/D = 0.025
λ/w = 6, k/D = 0.05
0.3 λ/w = 8, k/D = 0.005
fCF D

λ/w = 8, k/D = 0.01

λ/w = 8, k/D = 0.025
λ/w = 8, k/D = 0.05
fcorr = fCF D
0.2

0.1

0.0
0.0 0.1 0.2 0.3 0.4 0.5
fcorr

(a)

Figure 6: The correlation in Equation 11 proposed in this work successfully collapses all data.

8
Figure 7: Streamlines overlayed with the dimensionless vorticity contours
(ωz Dh,eq /umean,eq = [−0.02, 0.02]).

Re = 200, λ/w = 2 Re = 200, λ/w = 8

0.35 0.35

0.30 0.30

0.25 0.25

0.20 0.20

0.15 0.15
f

0.10 0.10

0.05 0.05

0.00 0.00
fp fv,BC fv,DA fv fp fv,BC fv,DA fv
−0.05 −0.05
0.01 0.02 0.03 0.04 0.05 0.01 0.02 0.03 0.04 0.05
k/D k/D

(a) (b)

Figure 8: Pressure and shear contributions on f found by using Equation 10: (a) Re = 200
and λ/w = 2. (b) Re = 200 and λ/w = 8.

9
than the net pressure on walls AB and CD for all roughness heights considered.
This dependence is quadratic for all f components. Moreover, shear on wall
DA dominates the shear on wall BC, noting that both have equal areas for
this spacing configuration. However, for a k-type roughness element, we see a
crossover at k/D = 0.025 with pressure effects more pronounced at higher k/D.
Shear on wall BC is now comparable to shear on DA as its area has increased
in this configuration. Similar trends are observed at other Reynolds numbers.
A deeper look in the flow field by means of vorticity contours and streamlines
is given in Figure 7 where two extremes of the roughness spacings are compared.
In line with the turbulent flow roughness classification by Perry et al. [4],
stable vortices seem to exist in the grooves of the d-type layout which exhibit
qualitative similarity with the cavity flow. This behavior is consistent for all
roughness heights considered. For the k-type layout, the roughness element
behaves as a bluff body which creates a separation zone downstream, whose
length increases with the relative roughness, in agreement with the turbulent
flow simulations by Leonardi et al. [11]. Reattachment of the flow on the channel
walls is observed up to k/Dh,eq = 0.025. For k/Dh,eq = 0.05, the flow in the
groove is completely separated. Additionally, contours of the dimensionless
vorticity shows the high vorticity regions near the corners of the protruding
edge of the roughness element in contrast to the relatively low vorticity regions
in between the elements. In general, the vorticity in the vicinity of these corners
becomes stronger as the relative height increases.

4. Conclusion

The effects of square roughness elements on laminar pipe flow has been pre-
sented. CFD simulations showed that Darcy friction factor increases quadrati-
cally with roughness height decreases linearly with pitch and inversely propor-
tional to Reynolds number. A proper correlation was found that fits all data.

10
Appendix A Modeling translational periodicity

The problem under consideration is the steady, incompressible, two dimen-

sional and fully developed flow in a horizontal channel which contains square
roughness elements on upper and lower walls (Figure 1a). With these assump-
tions, it is possible to make use of the translational periodicity as described by
Patankar et al. [15] and employed by Herwig et al. [13]. A brief description of
the approach is as follows. Although the conditions related to the fully devel-
oped, smooth channel flow, ∂u/∂x = 0 and v = 0, are no longer satisfied in a
rough channel, it is still possible to define a fully developed regime in a periodic
sense where the velocity field repeats itself as

V(x, y) = V(x + λ, y) = V(x + 2λ, y) = ... (A.1)

where λ is the periodic length in the direction of the flow. The pressure does not
follow this behavior as it continuously decreases along the channel. However,
the pressure drop does,

∆p = p(x, y) − p(x + λ, y) = p(x + λ, y) − p(x + 2λ, y) = ... (A.2)

and this is constant. Introducing the negative of the pressure gradient as

p(x, y) − p(x + λ, y) ∂p
β= =− (A.3)
λ ∂x

the actual pressure field can be decomposed into two components:

p(x, y) = −βx + p̃(x, y). (A.4)

Incorporating into the governing equations yields,

∇·V =0 (A.5)

∇ · (ρVV) = βi − ∇p̃ + µ∇2 V. (A.6)

Hence, p̃ resembles the actual pressure in the Navier-Stokes equation and the
β merely acts as a source term. Numerical solution to equations A.5 and A.6
are found using a finite volume-based commercial code ANSYS Fluent with a

11
modified SIMPLE (semi-implicit method for pressure-linked equations) algo-
ritm. Modification is due to the additional unknown β which is guessed and
succesively corrected throughout the iterative solution until the desired target
mass flow rate input is matched.

References

[1] J. Nikuradse, Laws of flow in rough pipes, in: VDI Forschungsheft, Citeseer,
1933.

[2] L. F. Moody, Friction factors for pipe flow, Trans. Asme 66 (8) (1944)
671–684.

[3] I. Tani, Turbulent boundary layer development over rough surfaces, in:
Perspectives in turbulence studies, Springer, 1987, pp. 223–249.

[4] A. E. Perry, W. H. Schofield, P. N. Joubert, Rough wall turbulent boundary

layers, Journal of Fluid Mechanics 37 (02) (1969) 383–413.

[5] M. Shockling, J. Allen, A. Smits, Roughness effects in turbulent pipe flow,

Journal of Fluid Mechanics 564 (2006) 267–285.

[6] J. Allen, M. Shockling, G. Kunkel, A. Smits, Turbulent flow in smooth and

rough pipes, Philosophical Transactions of the Royal Society of London A:
Mathematical, Physical and Engineering Sciences 365 (1852) (2007) 699–
714.

[7] S. G. Kandlikar, D. Schmitt, A. L. Carrano, J. B. Taylor, Characteriza-

tion of surface roughness effects on pressure drop in single-phase flow in
minichannels, Physics of Fluids 17 (10) (2005) 100606.

[8] S. Shen, J. Xu, J. Zhou, Y. Chen, Flow and heat transfer in microchan-
nels with rough wall surface, Energy Conversion and Management 47 (11)
(2006) 1311–1325.

12
 [9] G. Gamrat, M. Favre-Marinet, S. Le Person, R. Baviere, F. Ayela, An
experimental study and modelling of roughness effects on laminar flow in
microchannels, Journal of Fluid Mechanics 594 (2008) 399–423.

[10] D. Gloss, H. Herwig, Wall roughness effects in laminar flows: an often

ignored though significant issue, Experiments in fluids 49 (2) (2010) 461–
470.

[11] S. Leonardi, P. Orlandi, R. A. Antonia, Properties of d-and k-type rough-

ness in a turbulent channel flow, Physics of fluids 19 (12) (2007) 125101.

[12] Y. Jin, H. Herwig, Turbulent flow in rough wall channels: Validation of

rans models, Computers & Fluids 122 (2015) 34–46.

[13] H. Herwig, D. Gloss, T. Wenterodt, A new approach to understanding and

modelling the influence of wall roughness on friction factors for pipe and
channel flows, Journal of Fluid Mechanics 613 (2008) 35–53.

[14] Y. Liu, J. Li, A. J. Smits, Roughness effects in laminar channel flow, Journal
of Fluid Mechanics (submitted).

[15] S. Patankar, C. Liu, E. Sparrow, Fully developed flow and heat transfer in
ducts having streamwise-periodic variations of cross-sectional area, ASME
J. Heat Transfer 99 (2) (1977) 180–186.

13

View publication stats