4 .1
Introduction
The flow of water, oil and gas in pipes is of immense practica l significance
in civil engineering. Water is conveyed from its source, normally in pressure
pipelines (Fig. 4 .1), to water treatment plants where it enters the distribution
system and finally arrives at the consumer. Surface water drainage and
sewerage is conveyed by closed condu its, which do not usually operate under
pressure, to sewage treatment plants, from where it is usually discharged
to a river or the sea. Oil and gas are often transferred fro m theif source
by pressure pipeli nes to refi neries (oi l) or into 3 distribution ncrwork fo r
supply (gas) .
Surprising as it may seem, a comprehensive theory of rhe flow of fluids in
pipes was not developed until the late 1930s, and pracrica l design methods
for the eva luation of discharges, pressures and head losses d id not appear
until 1958. Until these design tools were ava ilable, the efficient design of
pipeline systems was not possi ble.
This chapter describes the theories of pipe fl ow, beginning with a review
of the historica l context and ending with the practical applications.
4.2
Table 4.1 lists the names of the main contributors, and their contributions,
to pipe flow theories in chronologica l order.
The Colebrook-Wh ite transition fo rmu la represents the culmination of
all the previous work, and ~an be .applied to any flu id in any pipe operating under turbu lent fl ow conditions. The later contributions of M oody,
Ackers and Barr are mainl y concerned with the practica l application o f the
Colebrook- White equation.
92
Figure 4. t
4.1
Th~
chronological
d~vclopm~nt
of pipe flow
theori~s.
Contribution
Nam~
Date
93
1839-41
1850
1884
1913
Blasius
1914
1930
Nikuradse
19305
Prandtl and
, 1937-39
\' 011
Karm:s'n
1944
Moody
1958
Ackers
1975
Sarr
Reynolds' experiments demonstrated that there were twO kinds of flow laminar and turbulent - as described in Chapter 3. He found that transition
from laminar to turbulent flow occurred at a critical velocity for a given pipe
and flujd. Expressing his results in terms of the dimensionless parameter
Re = pDV/f.L, he found that for Re less than about 2000 the flow was always
laminar, and that for Re greater than about 4000 the flow was always
turbulent. For Re between 2000 and 4000, he found that the flow could be
either laminar or turbulent, and termed this the transition region.
In a further set of experiments, he found thaI for laminar flow the frictional head Joss in a pipe was proportional to the velocity, and that for
turbulent flow the head loss was proportional to the square of the velocity.
These two results had been previously determined by Hagen and Poiseuille
(hl oo,", and Darcy and Weisbach (lJlooV1l, but it was Reynolds who put
these equations in [he context of laminar and turbulent flow.
94
4.3
+ pg21l'r8/8rsin a = 0
Sening sine = - dz/d/a nd divid ing by 2-rrr8r81 gives
dp dT T
d:z
- - - - - - - pg - = O
dl dr r
dl
('t +
1ir)21l(r + &-)&'
(p + .2&)2'111Sr
d/
p2v&-
Figure 4.2
9S
where V(= p +pgz) is the piezometric pressure measured from the datum
4=0. As
1 d
I (dT
dT T
--('11')=r-+'1) =-+r dr
r dr
dr r
then
dp 1 d
-----(n)=O
di
r dr
Rearranging,
dp "
-&"2
+constant
(4.1)
96
/
"
.~t, ,.y"
Figure 4.3
of the bounda ry layer is halted when it reaches the pipe centreline. and
thereafter the fl ow consists entirely of a boundary layer of thickness r. The
resulting velocity distribution is as shown in Figure 4.3(a).
For Ihe case of turbulent flow shown Ul Figure 4.3(b}, the growth of the
boundary layer is not suppressed umi l it becomes a turbulent bou nda ry
layer with the accompa nying laminar sublayer. The resulting velocity
profile therefore differs considerably (rom the laminar case. The existence
of the laminar sub-layer is of prime importance in explaini ng the difference
between smooth and rough pipes.
Expressions relating shear stress to velocity have been developed in
Chapter J, and these will be used in explaining the pipe flow equations in
the following sections.
LAMINAR FLOW
4.4
97
Laminar flow
For the case of laminar flow , Newton's law of viscosity may be used to
evaluate the shear stress (T) in terms of velocity (u):
du
dll,
T =~ d)' =-~dr
dlf,
dV r
dl 2
T= - ~ -=---
d,
0'
dtl,
1 dp'
-=--,
dr
2~ dl
Integrating,
II,
II,
1 dp'
= 4~ dT,-l + conStant
= 0 and r = R, hence
constant =
1 dp'
---R'
4" dl
and
(4.2)
Equation (4.2) represents a parabolic velocity distribution, as shown in
Figure 4.3(a). The discharge (Q) may be determined from (4.2). Returning
to Figure 4.1 and considering the elemental discharge (SQ) through the
an nulus, then
5Q = 21Tr5m,
Integrating
Q=21T 10 r lI , dr
and substituting for
tI,
98
0'
dp'
8" dt
'IT
Q~---- R'
(4.3)
1Tdp' 4 '
8"" df
'lTR2
V~ - ~----R
0'
I dp-
V=---R'
8" dt
(4.4)
hl~-
pg
0'
b = 32""LV
I
pgO'
(4.5)
This is the Hagen-Poiscuillc equation, named aftcr the two peoplc who first
carried out (independen tly) the experimental work leading to it.
The wall shear stress (TO) may be related to the mean velocity (V) by
eliminating dfJ' ( dl from (4.1) and (4.4) to give
(4.6)
As T= TO when r= R, then
(4.7)
Equation (4.6) shows that (for a given V) the shear stress is proportiona l
to r, and is zero at the pipe centreline, with a maximum value (TO) at the
pipe boundary.
LAMINAR FLOW
Example 4.1
99
Oil flows through a 25 mm diameter pipe with a mean vclocil)' of 0.3 mls. Given
that JL = 4.8 X 10- 2 kg/m sand p = 800ks/m}, calculate (a) the pressure drop in
a 45 m length and (b) the maximum vclocil)', and the velocity 5 mm from the pipe
wall.
SO/lItio"
First check thaI flow is laminar, i.e. Re < 2000.
Re = pDV/JL = 800 x 0.025 x 0.3/4.8 x 10- 2
= 125
(a) To find thc pressure drop, apply (4.5):
hf = 32JLLV/pgOZ
= (32 x 4.8 x 10-2 x 45 x 0.3)/(800 x 9.81 x 0.025 2)
= 4.228m(of oil)
or Ap = -pgh, = -33. 18kN/m1 . (Note: the negative sign indicates that pressure
reduces in Ihe direction of flow.)
(b) To find the "elocilies, apply (4.2):
Idp" 2 2
" =---(R -r)
,
4JL dl
The maximum velocity (U....~) occurs at the pipe centreline, i.e. when r = 0, hence
1
33. 18x10J
2
U..... x =- 4 x.)(
48 ) 02
x(0.025/2)
45
=0.6m/s
(Nnte: fln .. ~ _2x mean velocity (colllpare (4.2) and (4.4).))
To find the velocity 5 111m from the pipe wall (Us), usc (4.2) with r = (0.025/2)-
1/
I
)(_33.18XIO)(0.01252_0.00752)
4x4.8 )( 10-1
45
=0.384m/s
100
4.5 T urbulent fl ow
For rurbulent fl ow, Newton's viscosity law d~s not apply and, as described
in Chapter 3, sem i-empirica l relationships for TO were derived by Pra ndtl.
Also, Reynolds' experiments, and the ea rl ier ones of Darcy and Weisbach,
indicated that head loss waS proportiona l to mea n velocity squared. Using
the momentum equation (4. 1J. then
dp' R
'To = - - dl 2
,nd
dp = hfpg
dl
L
hence
0'
(1o, h, ~ KV').
Returning [0 the momentum equation and making the substitution
TO = K. V.t, then
hence
0'
TIJRBULENT FLOW
101
(4.9)
The later experiments of Stanton and Pannel, using drawn brass tubes,
confirmed the validity of the Blasius equation for Reynolds' Numbers up to
105 However, at higher values of Re (he Blasius equation underestimated
X for these pipes. Before further progress could be made, the distinction
between 'smooth' and 'rough' pipes had to be established.
He plotted his results as log ~ against log Re for each value of K,/D, as
shown in Figure 4.4. This figure shows that there are five regions of flow,
as follows:
(a) Laminar flow. The region in which the relative roughness has no influ ence on lhe friction factor. This was assumed in deriving the I-IagenPoiseuillc equation (4.5). Equating this to the Darcy-Weisbach equation
(4.8) gives
32~VL
pglY
~LVl
2gD
0'
~=64~=64
pDV
Re
(4.10)
0. 116
"
\~
J1
transitional
turbulence
,,
,,
,,
0.1ll2
00t6
""
00'
"
rough
"
"'~/" 11J,6t,
"-
I.....
00'
Re (log loCale)
Figure 4.4
"
'"
'"
'"
'"
2$2
,,
00'
DIlks
IUrbule~e
00'
TIJRBULENT FLOW
103
(e) Rough turbulence. The region in which A. remai ns constant for a given
k,/D, and is independent of Re.
An explanation of why these five regions exist has already been given in
section J.6. It may be summarized as follows:
Laminar flow.
miSSion.
1
J'i:
ReJ'i:
2.51
(4. 11 )
(4. 12)
The smooth law is a bener fit to the experimental data than the Blasius
equation .
The Colebrook- White tra"s;t;ol1 fomlllia
104
\
\ (k,
2.5\ )
.J). = -2 og 3.7D + Rc.J).
(4.13)
This gave predicted results very dose to the observed transitional behaviour
of commercial pipes, and is known as the Colebrook-White transition
formula. It is applicable to the whole of the turbulent region for commercial
pipes using an effective roughness val ue determined experimentally for each
type of pipe.
h = 0.0055 [1 + (200~Ok.
+ ~:) 13]
(4.14)
Iriln"'llOnallUrbulencc
'~ MIRe
II.m":I~';c,
.01<
~~
tun 2
,_
IH)64
\ ;
O.INS
11.1
" k_1r.msitional
;,
.,
.,
"
\1
"
\1
0,(136
0,10 2
Re
). 0.02S
0.05
0.03
~ r-:.---t-
0.02
0.015
"
--
~---
" ",
~ I::::-
'.
0 .002
""
~ ---
o III 2
smOOlh ----pipes
001 0
U.fJOK
0.01
0.008
~~
\
IWI
"-
IUl2 0
t---..-
eT "
IUlZ 4
rough turbulence
0.04
lurbulence
"
,
10
2W'}
-"-
I '"
10'
I I I I" I
10
Re
Figure 4.5
"'
0 .0004
"-
.001
0 .0008
.0006
0 .0002
"
O.<MXI 001
10'
0 .0001
"-
0 .00005
"
0.lXIOO1
10'
106
which gives A correct to 5% for 4 x 103 < Re < "I x 107 and for k./D <
0.01.
In a more recent publication, Barr (1975) gives another explicit formulation for X:
(k, 5.1286)
1
21
.j).=- og 3.7D+ Reo.n
(4.15)
In this fo rmula the smooth law component (2.5 lIRe A) has been replaced
by a n approximation (5. 1286/ Rco.89 ). For Re > l Os this provides a solution
for S,{hf/L) to an accuracy better than I %.
However, the basic engineering objections to the use of the ColebrookWhite equation were not overcome until rhe publication of Charts for the
Hydralt!;c Desigll o(ClJalllleis and Pipes in 1958 by the Hydraulics Resea rch
Station. In this publication, the three dependent engineering variables (Q,
D and Sf) were presented in the form of a series of charts for various k,
values, as shown in Figure 4.6. Additional information regarding suitable
design values for k, and other matters was also included. Table 4.2 lists
rypical values for various materials.
These chans are based on the combination of the Colebrook-White equation (4.13) with the Darcy-Weisbach formula (4.8 ), to give
(k
2.51.)
V = - 2"r:;:;:;;;
2gDSf log - ' - + r>::n<
3.7D D,/2gDS,
(4.16)
where Sf = ",IL, the hydrau lic gradient. (Note; for further details
concerning the hydraulic gradient refer to Chapter 12. ) In this equation
the velocity (and hence discharge) can be computed directly for a known
diameter and frictional head loss.
More recently, the Hydraulics Research Station have also produced
Tables fOT the Hydraulic Design of Pipes.
In practice, any two of the three variables (Q, D and Sf) may be known,
and therefore the most appropriate solution technique depends on circumstances. For instance, in the case of an existing pipeline, the diameter and
available head are known and hence the discharge may be found direcrly
from (4. 16). For the case of a new insta llation, the available head and
required discharge are known and the requisite diameter must be found.
This witJ involve a trial-and-error procedure unless the HRS charts or
tables are used. Finally, in the case of analysis of pipe networks, the
required discharges a nd pipe diameters are known and the head loss must
be computed. This problem may be most easily solved using an explicit
formula fo r A or the I-IRS charts.
DiamelC:r (m)
Figurt 4.6
k, (mm)
0.003
:I~ I>e$fos
wrought iron
galvanized iron
plastic
bitumen-lined ductile iron
spun concrele lined ductile iron
slimed concrtte sewer
O.oJ
0.06
0.15
0.03
0.03
0.03
6.0
108
Examples illustrating the applica tion of the various methods to the solution of :1 simple pipe friction problem now follow.
Example 4.2
A pipeline IOkm long, 300mm in diameter and with roughness size 0.03 mm,
conveys water from a reservoir (top water level 850111 300ve datum) 10 a water
treatment plant (inlet water level 700 m above datum). Assuming that the re~rvoir
remains fuJI, rslimate the discharge, using the following methods:
(a) the Colebrook-Whitt formula;
(b) the Moody diagram;
(e) the HRS chartS.
Note: Assume v = 1.13 x 10-' ml/s.
So/lltion
(a) Using (4. 16),
D=O.3m
k.=O.OJmm
hence
v=
+ ;;~2.~5F,1~X~I~.IFJFX~I~O~-TI" )
O.3J2g x 0.3 x 0.0 15
= 2.5 14 mls
=A=
2.5 14 )( 'II"xO.3
0 78
=. l m /s
(h) The same solution should be obtainable using the Moody diag ram; however,
it is less accurate since it involves interpolation from a graph. The solution method
is as follows:
(1) ca lculate ksl D
(2) guess a value for V
(3) calculate Re
(4) estimate A using the M oody d iagram
(5) calculate h(
(6) compare h, with the available head (H)
(7) if H". hr' then repeat from step 2.
109
11}RBUl.ENT Fl.OW
This is a tedious solution technique, but it shows why the HRS charts were
produced!
DV
0.3 x 2.5
Re = -,- - '1~.1~3~x::-';'I';O-:"
= 0.664 x 10 6
(4) Referring to Figure 4.5, Re = 0.664 x 106 and k,ID = 0.000 1 confirms that
the flow is in the transitional turbulent region. I:ollowing the k, ID curve until
it intersects with Re yields
A::=0.014
10
'2gD
2gxO.3
= 148.7m
s, = 0.0 15
1ODS, = J.5
and D=300mm
giving Q = 180 lIs = 0. 18m 1 /5
Examp le 4.3
1\0
Solutio"
This (('quires an iterative solution if methods (a) or (b) of Ihe previous example are
used. However, a direct solUlion can be obtained using the HRS charts.
S,_ SO/SOOO
100S,,,,, I
and Q = 400 lIs
giving D = 440 mm
In practice. the nearest (larger) available diameter would be used (450 mill in this
case).
Example 4.4
The known outflow from a hranch of a disuiburion syStem is 30 lis. The pipe
diameter is ISO mm, length 500 III and roughness coefficient estimated at 0.06 mm.
Find the head loss in the pipe, uSlOg the explicit formulae of Moody and Barr.
Solutioll
Again, the H RS charts could be used directly. Howc\'er, if Ihe analysis is being
carried out by computer, solution is more efficient using an equ:ltion.
D _ O. 15m
V = L7m/ s
Re=0.15x 1.7/ 1.1 3 x 10-'
Re=0.226 x 10'
Using the Moody formula (4.14)
~=0.OO55 ( 1+ (
~
20oooXO.06 X 10-'
1 )''')
0.15
+0.226
"" 0.0182
1
(0.06 X 10- 1
21
./):.= - 08 3.7xO.15
~_0.0182
5.1286)
+ (0.226 x 10')0."
111
TURBULENT FLOW
0.0182
1/ ,ff;
2.51)
I
ks
-2log ( 3.7D + Re....'A =....'A
7.415
7.441
This confirms Ihat both formu lae are accurale in Ihis case.
The head loss may now be compulcd using the Darcy-Weisbach formula (4.8):
or, alternatively,
_ 6.78L
hf - D1.I65
(V)""
C
112
4.6
I~
I~
I
CDta) AI tvckkn
-/f
I
CD
(I))
(II
~nla rttmt11 1
(II
Figure 4.7
113
(I ') turbulent eddies are formed, which gives rise to a loca l energy loss. As
the pressure ca nnOI change instantaneously at the sudden enlargement, it is
usua lly assumed that at position (1') the pressure is the same as at position
(1). Applying Ihe momentum equation between (I) and (2).
(.)
The local head loss may now be found by applying the enerb'>' equation
f<om (I) ro (2),
or
(b)
The continuity equation may now be used again to express the result in
terms of the ~\l0 areas. Hence, substituting V.A. / A 2 for V2
0'
(4.18)
114
For the case of a pipe discharging into a tank, A2 is much grea ter tha n AI>
and hence kl = 1. In o ther words, fo r" sudden large expansion, the head
loss equals the velocity head before expansion.
Figure 4.7(b) shows the case of a s udden contraction. From posi tio n (1)
to ( 1') the flow contracts, forming a vena contracta. Experimenrs indicate
that the contraction of the fl ow a rea is generally about 40%. If the energy
loss from {I) to {1 ') is assumed to be negligible, then the rema in ing head
loss occurs in the expansion from (I') to (2). Si nce an expansion loss gave
rise to (4.18), that equation may now be applied here. As
then
0'
b L = 0.44 Vi /2g
(4. 19)
i.e. kL = 0.44 .
Typical kL va lues for other important local losses (bends, tees, bdlmouths
a nd valves) are given in Table 4.3.
Table 4.3
kL value
Item
Theoretical
Design practice
Comments
0.05
0.2
0.4
0.10
0.5
0.5
v = velocity in pipe
exil
90 bend
90" " ..
in-line flow
branch to lint
gate valve (open)
0.35
1.20
0.12
0.4
1.5
bdlmoUlh entrance
0.25
11 5
SO/lllio"
The available static head ( 150m) is dissipated bOlh by friction and local head losses.
Hence
Using the Colebrook-White formula (as in Example 4.2) now requires an iterative
solution, since h is initially unknown. A solution procedure is as follows:
( I) assume h, ~ 1-1 (i.e. ignore
(2) calculate Y
lid
Adjust II,
11(=
150-3.58_ J46.42m
116
0.03 x 10-1
2.51 x 1.13 x 10- ' )
+ ;;-;=:;;~~E~~'"
3.7xO. 15 O.3j2gxO.3xO.O I4 64
=2.386mj s
Recalculate hL'
Check
hL +hf = 146.42+3.22 = 149.64 :::: 150
This is sufficiently acc urate to be acceptable.
Hence,
4.7
Pipe systems for surface water drainage and sewerage are normally designed
to flow full, but not under pressure. This contrasts with water mains, which
are normally full and under pressure. The Colebrook-White equation ma y
be used for drainage pipes by noting that, because the pipe flow is not
pressurized, the water su rface is parallel to the pi pe invert, so the hydraulic
gradient equals the pipe gradien t:
11 7
Starting from the assumption that the frictio n facto r for the partially full
condition behaves similarly to that for the fu ll condi tion, it remains to fi nd
a parameter for the partially fu ll pipe which is equiva lent to the dia meter
for the full pipe case. The hydraulic radi us R is such a parameter:
where A is the water cross-sectiona l area and P is the wetted peri meter. For
a pipe flowing full,
(k,
I
21
fi. = - og 3.7 x 4R
2.51 )
(4.20)
+ Refi.
V' = 2g5,DI"
Hence, for a given pipe with partiall y fu ll flow,
V = (2g5, 4RI")'"
Figure 4.8
118
0'
v=
(4.21)
value, the full depth (D) and the partially full depth (eI). Simi larl y,
R1 / Z/}..I/z
Q, ._}..I/lA
- D
p,.
Ii
(4.22)
Pd = ",0/2
R, =
(I _S;;"') ~
and hence
A,= ("'-2:""')
(4.23)
S
R,=(I _ ;;"')
(4.24)
Substitution of (4.23) and (4.24) into (4.2 1) and (4.22) allows calculation
of the proportional velocity and discharge for any proportional depth (dID).
The expression for}.. (equation (4.20)) is, however, rather awkwa rd to
Fo
210g(3.7 x 4R,/ks)
,;;;; = 210g(3.70/k s)
Fo
10gR,
-';;;;-, = I + ;lo-g(~3."'7""0"'/k'"'s)
(4.25)
11 9
as
1+
10gR,
log(3.7D/ks)+logR,
=
log(3 .7D/ ks}
log(3.7 D/ks}
log[3.7 x 4R,/ksl
log(3.7D/ ks}
Qp = ( 1+ log(3.7D/k,}
' f2
ApRp
V,_
10gR, ) l{l
1+ 10g3.70 R,
(4.26)
and
Q = ( 1+ 10gR)
p
ARI/l
P
log 3.70
p p
(4.27)
( r
(4.28)
where
a . . . . ks +
-
3600DS~/l
These resuhs for 0 = 1000 are ploned in Figure 4.9. Tabulated values for
various 0 may be found in Hydraulics Resea rch Ltd (1983a). Neither Vp
nor Q, are very sensitive to O.
Figure 4.9 shows that the discharge in a pa rtially full pipe may be greater
than the discharge for a fu ll pipe. This is because the wetted perimeter
reduces rapidly immediately the pipe ceases to be full whereas the area
does not, with a consequent increase in velocity. However, this condition
is usually ignored for design purposes because, if the pipe runs full at any
120
"0.8
"0.'
Q,
e O~
~
0.'
v,
0.3
02
0. 1
01
01
0.1
0-4
OS
06
0.7
0.8
09
10
1I
12
Figure 4.9
section (e.g. due to wave action or unsteady conditions), Ihen the discharge
will rapidly reduce to the full pipe condition and cause a 'backing up' of
the flow upstream.
Soilltioll
The easiest way 10 solve this problem is to use the HRS design ch:ms or tables. For
a sewer, ks = 6.00mm (Table 4.2). Howt:ver, to illustratt: tht: solution, Figurt: 4.6
is uSt:d (for which ks = 0.03mm):
Q= 75 1/5
IOOh, / L
= 100/300 =0.333
D=300mm and
V _ I .06m/s
121
Q,= 10/75=0.133
Using Figure 4.9 (neglecting the effect of 0),
So.