Vous êtes sur la page 1sur 29

COLEBROOKONTURBULENT

133

FLOW IN PIPES.

Paper No. 5204.


Turbulent Flow in Pipes, with
particular
reference to the
Transition Region between the Smooth and Rough Pipe Laws.

CYRIL FRANK
COL~BROOK,
Ph.D., B.Sc. (Eng.), Assoc. M. Inst. C.E.
(Ordered by the Council to be published with written.dkcusswn.)l
TABLE O F CONTENTS.

.
.
.
.
.
.
.
.
.
Introduction
.
Theory of turbulent flow inpipes
.
.
.
.
. .
A new theoretical formula for flow in the transition region
.
Relation between Prandtl-von-Karman and exponential formulas.
Analysis of experimental data on smooth pipes
. . . .
Galvanized,cast-, and wrought-ironpipes
.
.
.
.
.
Old pipes .
.
.
.
.
.
.
.
.
.
.
Discussion and conclusions .
.
.
.
.
.
.
.
Appendix-Examplesillustratingthe
use of design-Tables .
.

.
.
.
.
.
.

.
.

.
.
.
.
.
.

PAGE

133

137
139

141
143
145
153
154

.
.
.

155

INTRODUCTION.
The problem offlow in pipes is one which has until recently defied
theoretical analysis, owing to its complexity and the absence of a rational
basis for its solution. An outstanding contribution t o the knowledge of
the subject was made more than half a century ago by Professor OsboTne
Reynolds, who succeeded in finding a unifying principle which considerably
simplified the analysis of his experimental results. His discovery that the

PUd

change from streamline to turbulent flow depended on the value of P

led later workers to thc corollary that the coefieient X in the well-known
hlU2

PUd

pipe-formula h = -- is afunction of the parameter -, which was


2gd
P
named after him the Reynolds number.
His discovery of this criterion led to the formulation of a more general
Correspondencc on this Paper can be accepted until the 15th
w i l l he published in the Institution Journal for October 1939.-sEC.

May, 1959, and


INST.

C.E.

134

COLEBROOK ON TURBULENT FLOW IN PIPES.

Principle of Dynamical Similarity, which determines the conditions


for mechanical similarity in the motions in or around geometrically similar
bodies.
Considerations of dynamical similarity may bereplaced by dimensional
reasoning which leads to a grouping of the quantities involved in the
problem into anumber
of non-dimensional parameters;this enables
experimentalresults to be plotted in asystematic manner. Such considerations, however, have definite limitationssince the functional relationship between these groups and their relative importance cannot be determined by dimensional reasoning.
It has been suggested as a result of experiments on lead and other
smooth pipes that the resistance-coefficienth and the Reynolds number R
could be expressed satisfactorily by an exponential formula of the type

h = ARh
By re-arrangement of this equation into the form

i t is easy to show that for smooth pipes the sum of the indices of U and d
must be 3 for all pipe-sizes and velocities. This equation is widely known
and the argument is frequently put forward that the sum of the indices
must equal 3 in any exponential formula designed to fit experimental
results on a few pipes over a limited range of velocities of flow. Although
this relation between the indices is true for smooth pipes, the value of n
itself so depends on the Reynolds number that a single value cf n will only
give approximatelycorrectresults
over alimited range of Reynolds
numbers. When the roughness-factor is introduced the relation no longer
holds : indeed, it will be shown in a later paragraph that, whatever the
roughness, this sum always exceeds 3. F. C. Scobey attempts to justify
by dimensional reasoning 1 his formula for riveted steel pipes in which the
sum of the indices is 3, but his omission, from the argument, of the roughness-factor, which is particularly important in the case of riveted pipes,
seriously affects the value of the formula.
In brief, it may be stated that the principle of dynamical similarity
determines the non-dimensional parameters governing fluid motion, but
fails to determine the functional relationship between them. This has led to
a reconsideration of t,he fundamentals of the problem, and therecent success
of L. Prandtl and von Karman in Germany, and of G. I. Taylor in Great
Brihin, in expressing in mathematical form the mechanism of turbulence,
Riveted
Steel and Analogous Pipes. Bulletin No. 150, Department of
Agriculture, U.S.A., 1930.

COLEBROOK ON TURBULENT
FLOW

135

IN PIPES.

linked with the experimental investigations of Nikuradse, have now provided a fundamental basis for the analysis of the problem.
They developed a formula of the type

and showed that thelower limit of integration y1 is a function of the wallparticle size k in the case of rough pipes in which the 00w obeys the
square resistance-law, and is dependent on the density p, the viscosity
p, and theshear stress at thewall T,in thecase of smooth pipes.
Substituting appropriate values of y1 in (1) the following resistance
laws are obtained for
))

(a) flow in hydraulically smooth pipes :

R 4

log-2.51

. . . . . . . .

(2)

. . . . . .

(3)

(b) flow in rough pipes :

The experimental results of Nikuradse show complete agreement with


the above laws provided certain limiting conditions are satisfied. The
experiments show that the rough-pipe law is true for values of

PV*k

exceeding 60, whilst for values less than 3 even rough pipes obey the smoothpipe law as the excrescences then cease to contribute to the resistance.
Between these values there is a transition from one law to the other.
The smooth, rough, and transition laws for Nikuradse's sand roughness
in which the grains are of uniform size and closely packed together, are
shown in Fig. l (p. 136) together with the transition curvefor a pipe having
a roughness composed of isolated particles, the experiments on which are

t I.'=2/?

and is called the

"

shear force " velocity, since it has the dimensions

of a velocity.
Theroughness Reynolds number

may be expanded into

It will be seen that it is the product of three dimensionless numbers, the resistancecoefficient, the relative roughness, and the Reynolds number.

136

COLEBROOK ON TURBULENT FLOW IN PIPES.

described in detail elsewhere.1 It is apparentthat with non-uniform


roughness the transition zone extends over a range about 10 times as long
as that for uniform sand roughness, and in the
case of new commercial pipe8
in which the roughness is non-uniform the whole working range lies within
the transition zone. The mean transition curves for galvanized-, cast-, and

wrought-iron pipes, which were determined by an analysis of most of the


available reliable data and described later in the Paper, are shown in
Fig. 1 for comparison with that for the roughness V .
C. F. Colebrook and C. M. White,"Experiments
with Fluid-Friction in
Roughened Pipes." Proc. Roy. Soc. (A), vol. 161 (1937), pp. 367,351. (See Rough.
ness '' V " in this Paper.)

COLERROOK TURBULENT
ON

137

FLOW IN PIPES.

Any attempt to express mathematically the transition-function for


uniform sand-roughness is rendered difficult owing to the fact that the
turbulent motion in the wake behind the grains is complicated by mutual
interference, and the resistance mechanism is made up of viscous and
mechanical forces which are difficult to separate.
In thecase of non-uniform roughness, however, the large isolated grains
have a shielding effect on the smaller grains which considerably reduces
their effectiveness so far as total resistance is concerned, so that the area
of the pipe between the large excrescences may be regarded as behaving
as a smooth surface witha coefficient of resistance dependent on the
Reynolds number P-.V*d Since the local Reynolds number on the large

grains is comparatively large even a t fairly low mean velocities, the local
grain co-efficient is practically constant over the entire transition range.
Y1
k
P
In effect, - in (1) is a function of -, the relative roughness, and __
d
d
P V*d
and hasdefinite limiting values corresponding a t the one extreme to fullyrough-law flow-conditions in which viscous resistance is negligible, and at
the other extremeto smooth-pipe conditions when the resistance mechanism
is entirely molecular. The exact form of the function will depend on the
distribution of the roughness-elements and is mathematically indeterminate,
but it will be shown in the present Paper that it is possible to obtain a
particular transition law which is similar to those obtained experimentally
for commercial pipes by simply addingtogether 1 the lower limits of
integration y1 which satisfy the rough- and smooth-pipe laws. The following general formula is then obtained :

which is in exact agreement with theory a t extreme values of __ and

gives results in the transition-zone which approximate very closely to the


experimental values. It willbe seen in Pig. I that this transition-curve
merges asymtotically into the smooth- and rough-law curves.

THEORYOF TURBULENT
FLOWIN PIPES.

In turbulent motion it has been observed that thevelocity-distribution


This treatment of the lower limits of integration was suggested by Dr. C. M.
White, and the Authordesires to place onrecord his indebtednessto Dr. White for his
collaboration in the development of formula (4).

138

COLEBROOK ON TURBULENT FLOW IN PIPES.

may be expressed by the relation

av.
&JP

2.5
-y-

. . . . . . . . .

where U denotes the velocity a t a distance y from the wall of the pipe,
r, the shear stress at the wall, and p, the density of the fluid.
On integration the equation (5) becomes

.-

Since U = 0 when y = y1 the effective hydraulic wall may be regarded


as being displaced inwards from the actual wall by an amount yl. The
hydraulic wall then represents a plane where the disturbances are theoretically as great as the actualones at the wall.
The mean velocity U is numerically equal to the local velocity a t
y = 0*113d*,and substitutingthis value of y in (6), the equation becomes
d

. . . . .

Re-arranging (7) so as to introduce the resistance-coefficient into the


equation,

Equation (8) may be regarded as a general formula applicable to all


types of turbulent flow in pipes. The shift of the effective hydraulic wall
y1 has, however, to be determined in order completely to determine the
resistance-law. Since y1 depends on the conditions at the wall it must
clearly be a function of (U)the roughness of the wall k, (b) the shear-stress
T , ahd

(c) the kinematic viscosity of the fluid, v = -

It has been observed experimentally that providing

PV&

__ exceeds

about 60 the resistance is proportional t o the square of the velocity (that


is, the resistance-coefficient is independent of the viscosity of the fluid),
and in thiscase dimensional reasoning shows that the shift y1 can only be
proportioned to k. Nikuradse, experimenting with pipes artscially
roughened internally by a uniform layer of sand fmed to the walls, deter-

* For the proof of this expression, see The Reduction of Carrying Capacity of
Pipes with Age, by C. F. Colebrook and C. M. White. Journal Inst. C.E.,vol. 7
(1937-38),p.99.(November1937.)

139

UOLEBROOK ON TURBULENT FLOW IN PIPES.

mined a value of

Y1

=g'

where k denotes the diameter of the sand grains. Inserting this value of
y1 in (S), the resistance-law for rough pipes becomes

(3)
In thecase of smooth pipes (or rough pipes when 'Y*k,,,.is

less than

3 when the roughness particles cease to shed eddies and contribute to the
resistance), the resistance is due entirely to molecular or viscous mixing,
and y1 must, by dimensional reasoning, be proportional t o -,CL which is the
PV*
only combination of 7, p, and p which has the same unit as a length.
Other experiments by Nikuradse show that for smooth pipes
P
y1 =-1 -

10 PV*
which on insertion in (8) leads to theresistance-law for smooth pipes

exceeds 3, however, the resistance increases over that

When
U

of a smooth 'pipeduetothe
protuberances.

shedding of eddies bythe

A NEW THEORETICAL
FORMULA
ROR FLOW
IN

THE

roughness-

TRANSITION
REGION.

The value of y1 may be regarded as having two extremes which satisfy


the smooth-law and fully rough-law conditions respectively, whilst in the
transition range y1 exceeds both of these extreme values due toa combination of mechanical and viscous mixing at thewalls.
Thus,
Y1

= +PV*
L ).

. .

Putting (9) into non-dimensional form, the equation becomes

140

COLEBROOK ON TURBULENT FLOW IN PIPES.

Analytically, equation (10) must take the form

where a and ,!lare numerical constants tobe found by experiment.


For pipes having non-uniform roughness k may be regarded as being the
roughness of a sanded surface giving the same resistance-coefficient as the
non-uniformly roughened surface.
1
l
Nikuradse's values for M and ,!l are - and - respectively, and sub33
10
stituting these numerical values in (11) and inserting the resulting value
of y1 in (8) the resistance-law becomes

0.113d

3=
=

log k
33
-2log-

1CL
+-.10 p v *

-+- lO'p17,d
__

0.113 3%

which may be rewritten as

1
---2log(:+--)

dX

k
37d

2.51

RdA

'

'

'

In order to represent (12) graphically it is convenient to separate the


independent variable

_ -1

'
3
from the remainder.
CL
3.7d
3.28
2log= 2log
k

Thus,

(13)

This function is shown as a heavy line in Fig. 1. (p. 136). It will be


noticed that the theory
indicates aslight increase in resistance over that for
P V*k
purely rough-law flow at -- 60, butthis

discrepancy against

pv k
experiment is very small and diminishes with increasing values of 2.
CL
The curve approaches the smooth- and rough-laws asymtoticallyin
accordance with experimental observation.
The formula for flow in smooth pipes
(2)

COLEBROOK ON TURBULENT FLOW IN PIPES.

141

is rather inconvenient for practical use since the resistance-coefficient


appears on both sides of the equation. This difficulty is overcome by
using the formula

which is a mathematical approximation to the exact


formula (2) but gives
numerical results within & $ per cent. over a range of Reynolds numbers
of from 5,000 to 100,000,OOO.

RELATION
BETWEEN PRANDTI~VON-KARMAN
AND EXPONENTIAL
FORMULAS.

It is of interest to compare the results obtained by the modern rational


method of analysis of the problem of fluid-flow with the earlier empirical
formulas of the exponential type, The Prandtl-von-Karman rough pipe1
a
law -- 08 3.7-may be converted to the exponential type
dj- 2 1
l;

by taking logarithm and differentiating.

Thus

Formula (17) may be extended into theusual form

or

It is clear that the exponent, n, is itself a function of the resistancecoeficient, so that a single value will only give approximatelycorrect
results over a limited range of d/k values. In order to illustrate the argument, suppose it is necessary to develop exponential formulas to cover a

142

COLEBROOK ON TURBULENT FLOW IN PIPES.

range of dlk = 10 to 40,000 so as to give results to within f 24 per cent.


1
(1
of the correct value. It will be found by plotting log - against log -

2/h

X:

that it is necessary to divide up therange into two components of dlk = 10


to 200 and d/k = 200 to 40,000. The values of A and n then become

dlk
dlk

= 10 to

200, A = 2.03, and n = 0.20


40,000, A = 3.25, and n = 0.111

= 200 to

It is to be noted that thesum of the indices of U and d always exceeds


3 in thecase of rough pipes by 1-74.\/ri.
An exponential formula of the type
1

=ARn

. . . .

. . . .

may be developed from the Prandtl-von-Karman smooth-pipe law


1 Rdi
_
- 2 log -by taking logarithms and differentiating. The exponent
2.51
dX
log-n is given by d(
:A) , which becomes
d(log R )

Thus

Equation (20) on extension becomes

or

where m is given by (19). Here again it is seen that the exponent m is a


function of the resistance-coefficient, but in this case the sum of the
indices of U and d equal8 3 as predicted by dimensional analysis.
This development of the relationship between rational and exponential
formulas shows quite clearly that single values of the exponent, n, can only
give approximately correct results over a limited range of pipe-sixes, and

COLEBROOK ON TURBULENT FLOW IN PDES.

143

velocities and exponential formulas are not, therefore, capable of universal


application.

ANALYSISOF EXPERIMENTAL
DATAON SMOOTH
PIPES.
A number of commerical pipes may be regarded as hydraulically smooth,
at least for all ordinary velocities of flow. Among these may be included
good commercial drawn-brass pipes, lead, glass, or tin pipes, centrifugallyspun lined (with bitumen or concrete) cast-iron pipes, and concrete-lined
pipes which have been deposited against oiled steel forms and carefully
rubbed down to remove any imperfections.
The results of an analysis of much of the available experimental data
are shown in Figs. 2 (p. l44), and are seen to be in close agreement with
the Prandtl-von-Karman smooth-pipe law.
The data include the experimental results on only one brass pipe of
0-5 inch diameter,obtained a t the National Physical Laboratory by
Stanton and Pannell in 1915, although the results for a large number of
brass pipes of other diameters tested by them also agree very closely with
theory. The results on sixteen spun concrete-lined pipes and on six spun
bitumastic-lined pipes ranging in size from 4inches to 60 inches in diameter
are included. Of these, the laboratory tests by M. L. Enger on 4-inch,
6-inch and 8-inch pipes were probably subject to the least experimental
error, and the result.s exhibit only slight scatter from the theoretical law.
I n analyzing the dat,a obtained by B. W. Bryan on the Stour Supply,
Danbury to Herongate main, which included one hundred and ninety-two
lobster-back bends of radius 3 4 4 and having a total change in direction
U2

of 2,987 degrees, an allowance of 2 0 - was made in the calculations for


2g
bends.
The results on the 216-inch diameter Ontario tunnel, the biggest of its
kind in the world, are especially interesting, as particular care was taken
in its construction and therange of test-velocities was large. The concrete
used in the construction of the tunnel was deposited against oiled steel
forms which resulted in a smooth and even surface. All defects were then
removed andthe surface rubbed down withcarborundum
brick. I n
analyzing the test-data1, it was found that an arithmetical error had been
h1 1JZ
made in calculating the resistance-coefficientsin h = --.
2@
The correct values, which are considerably lower than those given by
&obey, are shown in Table I (p. 145) together with the test-results from
which they were computed.
Despite an appreciableexperimental scatterthe test-results are in
very satisfactory agreement with theory.
1

F. C. Scobey,

No. 852.

Concrete Pipes. Department of Agriculture, U.S.A., Bulletin

144

COLEBROOKONTURBULENTFLOW

IN PIPES.

I45

C'OLEBROOK ON TURBULENT FLOW IN PIPES.

TABLEI.

1 zGi! 1

1,018
2,036
3,045
4,063
5,091

in
1,000feet :per

4
8
12
16
20

feet,

Coemclent of
friction, X .

Reynolds
number.

0.108
0.448

0.990
1.701
2,397

5,550,000

11,100,000
16,650,000

22,2oo,oO0
27,700,000

0.00782
0.00812
0.00798
0.00773
040697

GALVANIZED,CAST- AND WROUGHT-IRON


PIPES.
In analyzing the dataon the various types of iron pipes it was necessary
mean
to determine boththe
mean hydraulic-roughness, k , andthe
transition law for each class. The problem is complicated by the fact that
in practice there are variations of roughness due to non-uniformity in the
method of manufacture so that ineach class there is considerable variation
both in the size and type of roughness. It was necessary, therefore, to
determine the transition law and roughness k for each individual pipe-a
t,ask which is rendered difficult by the fact that
with one or two exceptions
the experimental results donot cover a wide enough range and rarely reach
square-law. However, the experiments on pipe V t indicate fairly rapid
transition to the square-law at the higher values of

'
9
and,
thus
II

with many of the test-results it is possible to extend them with very


little error so as toreach square-law and enable the determination of the k
values, and thuslocate the test-results in thetransition-range,
The experimental results for each class of pipe are plotted in Pigs. 3,
1
5, and 7 (pp. 146 et s q . ) with - as ordinate against log RdX as abscissa.

dA

This arrangement gives a sloping straight line for the smooth-law flow and
a series of parallel horizontal lines in the square-law region which extends to
the right of the dotted line representing the lower limit of rough-law flow.
The results may be brought to a single line in the rough-law region by
3.7d
l
P V& This has been
plotting 2 log -- -as afunction of log

X:
di
P
carried out in Pigs. 4 , 6, and 8 (pp. 147 el seq.), and a mean transition curve
drawn in for each class of pipe. The k-values determined for all pipes are
shown in Pys. 9 (p. 152), and using the mean Ic-value for each class together
with the corresponding mean transitioncurve, a number of transition curves
have been drawn in Pigs. 3, 5, and 7 for direct comparison with the
p-.

-f Footnote (l),p. 136.

10

146

COLEBROOK ON TURBULENT FLOW IN PIPES.

COLEBROOK
IJRRULENT
ON

FLOW IN PIPES.

147

148

COLEBROOK ON TURBULENT
FLOW

IN PIPES.

test-results. It is seen that although some of the pipes do not agree very
closely with the mean curves, some having too rapid transition andothers
too slow, there appears to be sufficient positive evidence to justify the
adoption of the given mean transition laws together withthe mean k-values.
It is to be expected that these will enable the prediction of resistancecoefficients in pipes of sizes other thanthose tested and at
velocities beyond
the normal range with less uncertainty than with any existing empirical
Fig. 5.

l? ,/T

EXPERIMENTAL
DATAON TAR-COATED
CAST-IRON
PIPES.

formula. With regard to the experimental data itself, space prohibits a


detailed description of all the data available, so remarks will be confined to
a few observations with regard to the most accurate data.
The experiments made by F. Heywood on new galvanized-iron pipes
of 2 inches and 4 inches diameter were carefully conducted and are most
valuable, as therange of velocities waB very wide, being from 0.5 to 21 feet
per second. Referring to Fig. 3 it will be noticed that the resistancecoefficient for the 2-inch pipe becomes constant a t high velocities, thus
enabling the determination of k and the major portion of the transition
curve.

At the lower values of -the 2-inch and 4-inch pipes diverge


P

COLEBROOK ON TURBULENT
FLOW

IN PIPES.

149

in opposite directions from the mean curves, and the 2-inch pipe is somewhat rougher than the 4-inch.
The remaining data on galvanized pipes was obtained by Saph and
Schoder, but in thedetermination of the mean value of k for this class the

Author has neglected pipe XVIII (0.85 inch in diameter) as the experimenters makethe following statement concerning this pipe-" Pipe XVIII
(0.85 inch in diameter) seems to be an exceptional pipe, but it has to be
remembered that a slight silt-like deposit had occurred on the inner walls
which was entirely sufficient to relieve the roughness."

150

COLEBROOK ON TURBULENT FLOW IN PIPES.

Very reliable data, used in the present analysis on tar-coated cast-iron


pipes, was obtained by J. Freeman and H. Mills a t Lawrence, Massachussetts on pipes of 4 inches, 8 inches, and 12 inches diameter, and another
carefully made experiment was that on a 6-inch pipe described in the
Report on Pipe Line Coefficients, 1 in which the range of velocities was

I L \X

E XPERIMENTAL DATAON

WROUGHT-IRON PIPES.

15 to 1. Ot,her carefully conducted experiments include those on the


Manchester, Thirlmere siphons (44 inches indiameter),
theSudbury
conduit (48 inches in diameter), and the 61-inch diameter siphon experimented on by Fitzgerald.
Practically all of the available data on wrought-iron pipes were obtained
by J. R. Freeman. Extreme care was exercised in making the experiments
which covered a wide range of velocities. The pipes were considered to be
fairly representative of ordinary lap-welded wrought-iron pipes used in the
U.S.A. The remaining experiments by J. B. Francis and H. Smith, Jr.,
indicate that their pipes were considerably smoother than those used by
Issued by a Committee of the New England Water Works Association in 193.5:
Journal New England Water Works Aasoc., vol. 49 (1935).

Fig. 8.

152

COLEBROOK ON TURBULENT FLOW TN PIPES.

Freeman, although Freemans results are remarkably consistent among


themselves. Some experiments 1 on asphalted wrought-iron pipes are also

001

0.W6
W

$0034

0 002

0.001
DIAMETER:
INCHES,
GALVANIZED-IRON PIPES.

0.006

0-004

a 0.002

0.00I
0

10

20

30
40
DIAMETER:
INCHES.
ASPHALTEDCAST-IRON

50

60

70

1 0

I2

14

PIPES

0 004

vi ow2

2
5:

., 0 . 0 0 1

k
.!

0-006
0.004
0

6
8
DIAMETER:
INCHES.

WROUGHT-IRON
PIPES.

included, but these pipes appear to have a capacity averaging about 5 per
cent. greater than that of uncoated pipes.
Pipes Nos. 302, 304, and 310 in The Flow of Water in Riveted Steel and
Anagolous Pipes, by F. C. Scobey (U.S. Dept. Agriculture-Tech. Bul. No. 50,
Jan. 1930). Denoted in Fig. 9 of the present Paper by d.

153

COLEBROOK ON TURBULENT FLOW IN PIPES.

The mean values of k are :


Galvanized-iron pipes
,
Asphalted cast-iron pipes
Uncoated cast-iron pipes
Wrought-iron pipes . .

. . . k = 0.006 inch.

. . .

I% = 0.005 inch.

. . . k = 0.01 inch.
. . . k = 0.0017 inch.

OLD PIPES.
The deterioration of pipes with age has already been discussed a t solne
length in a previous Paper 1 so only brief reference to thisproblem will be
made here.
The hydraulic resistance of water-mains increases after themains have
been in service for some time due togrowths or deposits upon the internal
surfaces. By making various simplifying assumptions it has been possible
to develop a formula 1 which gives the relation between the age of a pipe
and its carrying capacity, which may be written as

$10-0)

. . . . . .

where Q denotes the discharge at theend of T years, Q. denotes the initial


discharge, p. = Co/22/& (where CO is the initial Chezy coefficient) and a
is the average rate of growth of roughness.
If in any district the growth-ratea is required this may be computed
from the results of experimental observations by means of the equation

3-7a

= -(lO*-

10-0)

. . .

. .

(22)

where p = C/22/89and C denotes the final Chezy coefficient.


The diameterof a proposed pipe may be determined from
the formula

where i denotes the hydraulic gradientandkodenotesthe original roughness


size, say 0.01 inch. Alternatively, the appended design-Tables 11-VI may
be used to determine Chezy coefficientsand values of A C d m corresponding
to various values of k and d. The roughness k is readily obtained from

= ko

+ uT,

and u may be computed from experimental observation using formula (22).


Where no experimental data is available for calculating the growth-rate
1 C. F. Colebroolr and C. M. White, The Reduction of Carrying Capacity of
Pipes with Age. Journal Inst. C.E., vol. 7 (1937-38), p. 99. (November1937).

I54

COLEBROOK ON TURBULENT FLOW IN PIPES.

this may be estimated for asphalted cast-iron pipes from the pH value of
the water, using t.hc interpolation formula

CI

2 l o g a = 3.8 - p H

. . . . . .

(24)

which gives the growth-rate in inches per year.

DISCUSSION
AND CONCLUSIONS.
The present analysis of the problem offlow in commercial pipes has
been based on the premise that transition from smooth-law to rough-law
flow in commercial pipes takes place in a gradual manner, as shown in
Fig. l (p. 136). By an extensionof the Prandtl-von-Karmanlaws for smooth
and rough pipes, a theoretical transition law (12) has been developed by the
Author, in collaboration with D r . C. M. White, which gives favourable support to this assumption. A l t ~ T h ~ ~ i Z i i i iexperimental
ble
data is so
incomplete and limited in range that fully rough conditions were only
reached in a few cases, a collection of data on old mains shown in Pig. 2
of a previousPaper 1 proves conclusively that in thecase of non-uniformly
roughened pipes (which include most commercial pipes), the resistancecoefficient falls with decreasing rapidity as thevelocity increases, and once
having reached square-law it remains constant at all higher velocities.
The fact that there are considerable variations in the roughness and
transition curves in each class of pipe must not be considered a defect in
the method of analysis. Such variations are to be expected, since manufacturing conditions are not identical in different plants. For design purposes a series of transition curves for each class is obviously impracticable,
so mean curves corresponding to average conditions have been determined.
The scatter of the k-values in Pig. 9 is too great t o be able to ascertain any
possible dependence of k on pipe-size, so a single value for each class seems
justified especially as pipes of all sizes in any particularclass are made by
the same process. In thecase of built-up pipes, such as riveted steel pipes,
a variation of k with pipe-size would be expected, and in a later Paper it
will be shown that thisoccurs in thecase of a certain class of riveted pipe.
Where it is not possible to determine by experiment the transition curve
for any particular type of pipe, the theoretical transition curve (12) may
be used with verylittle error provided that theroughness can be determined,
and this is not difficult since some reliable experimental data on a few
pipes over at least a small range of velocities is usually available.
All formulas in the Paper are non-dimensional throughout and it is
possible,therefore, to use the results inany system of units. Since the transition curves are somewhat complex and are not, therefore, easy to use,
five design-Tables (Tables 11-VI) based on these functions are included
Footnote ( l ) ,p. 163.

COLEBROOK ON TURBULENT FLOW IN PIPES.

158

in order t o facilitatecalculations on the flow of water.The


Chezy
coefficient C in U = C 4 2 is given for various pipe-sizes, velocities, and
gradientsinEnglishunits
a t atemperature of 55" F., as calculations
involving the Chezy formula are easily and rapidly made by slide-rule.
Similar tables for gas, air and other fluids may be compiled by means of
the transition curve determined by the Author.
The work was carried out in the Civil Engineering Department of
the Imperial College of Science and Technology, London, and the Author
isindebted tothe generosity of the Clothworkers Company, who, in
supporting another researchof purely academic nature, indirectly inspired
the present work.
The Paper is accompanied by nine sheetsof drawings and five designTables-from which the Figures in the text and thefollowing Appendixes
have been prepared.

APPENDIX.
'

Examples illustrating the we of t h deaign-Tabks.


Problem ( I ) .
To find the discharge of a new asphalted cast-iron pipe, 48 inches diameter, with a
gradient of 1 in 6,000.
The dischargeis determinedfrom
Q = (AGdiijda
and from Table IV the value of A C d G corresponding to a gradient of the order
1 in 6,000 is
A C d m = 1,710.
Hence
Q

1,710 X 4-63

= 22.1 cu8ecB.

Problem (2).
To find thc diameter of a new asphalted cast-iron pipe to discharge 10 cuseca with
a gradient of 1 in 4
00.
The sizeof pipe is determined by the value of

156

COLEBR.OOK ON TURBULENT FLOW IN PIPES.

From Table IV it is seen that a 21-inch diameter pipe hacrav;dueofACl/ii- ,008


at.,zpproximately the given gradient.
The actual dischargeof this pipe at thegiven gradient is

Q = 208 .d-=

10.4 cusecs

Problem (3).
To find the diameter of an asphalted cast-iron pipewhich will discharge 36 cusecs
30 years hence witha gradient of 1 in 100 and apH value of 7.2.
The required pipe must have a value
of

and by interpolation in Table VI for a pH value of 7.2 it is seen that a 33-inch diameter pipe has a valueof A C d g = 365 approximately a t this pH value.

TABLETT.-SMOOTH

D:
inrhes.

ACdG

1
2
3

4
5

6
7
8
9
10
11
12
15
18

21
24
27
30
33
36
40
44
48
54
60
66
72
78
84

107.5
110.5
112.5
114,:i
116

0.061
0.385
1.11
2.37
4.25
6.87
10.3
14.6
20
26.3
33.6
42.2
76
121.5
182
2.57

U=l

PIPES
: VALVESOF C

A4cd/m

0.0713
94

0.141

103

ACdG

0474
0.46
1.32

105

106.5
108
109.5

l1 0.5
112
114.5

4.45

7.15
10.7
15.2
20.7
27.3
34.9

-l
124

44

78.8

85.3
136
203
288
392
515
660
825
1085
l390
1740
2360
3940
4980
6150
7420

1
-

100,000

127.5

130
131.5
133.5
l35
l36
137.5
134.5
140
141
142
143.5
145
146.5
147.5
148.5
150

G ~ ~ A QR =D( A C ~ ~ ~ VARIOUS
T A T VELOCITIES.

5
8.05
1245
17.1
23.2
30.5
39.2
49.1
87.7
40.5

209
296
402
527
678
848

1115
142.5
1785
2420
3180
4090
5130
62S0
7620

GRADIENT =

-I

99
107.5
112.5

0.078

0.48

1.38
2.93

119
121
123
125
126
127.5
129
130
132.5
135
136.5
138.5

5.25
8.4
12.55
17.8
24.1
31.8
40.7
51.1
91
146
217
307

140.5

419

141
142.5
143.5
145
146.5
147.5
149
150
151.5
153
154
154.5

547

--

1
10,000
~

-1

0=3

1l 6

2.8

101
103

U=Z

1 -1-

JIOO

GRADIENT =

U=1.5

IN

703
880
1153
1480
1855
2510
3290
4220
5310
6.520
7850

u=5
C

u=10

U=;

ACd\/m

ACdm

0.096F
0.585
14ij

_105
114
l19
122
l25
127.5
130
131

132.5
134
135.5
136.5
139.5
141
143
145
146.5
148
149

150
151.5
153
154.5
l56
157
158
159.5
160.5
161

114
122
128
131
I
134.5

0.0828
0.508

1.46
3.08
5.5
8.85
13.25

15.4
13.4
12.8

53.6
16
152
227
322
437
575
73.5
920
1206
1545
1940
2630
3440
4400
5550
6800
8200

123
126.5
129.5
131.5
133.6
135.5
137
138.5
139.5
140.5

143.3
146
148
150
150.5
152
153.5
154.5
156
157
158.5
l60
161.5
162
163.5
164.5
165.5

1.51

3.19
5.7
9.12
13.6
19.3
E6.2

136.5
138
140
141.5
142.5

!4.5

0.09

119

0.545

1'X

1.57
5.92

132..5
I36
139.5

9.45

141.5

14.05
20
27.1
35.5

143
145
146.5
138
149
150
153.5
1 55
l57..5
139.<5
161
162
163
164.5
165.5
167
168
169.5
171
172
I73
174
175

3.31

14.1
i5.3

57.5
23.5
333
450
590

757
94.7

1215
1585
l990
2700
3540
4520
5680
6970
8420

GRADIENT

162

155.5
156.5
158
159.5
160.5

161.5
163
164.5
165.5
167
168

l69
170

1
1,000

=-

465

607
780
975
1280
1630
2050
2780
3630
4660
5830
7150
8650

3.53
6.3
10.05
15.0
21.3
28,s

GR,\DIENT=
1
148
150
151.5
153
155
156.5
158
159.5
162
I
161..j
166
168
170
170.5
172
173
174.9
176
177
179
180
181.5
I82-.5
183.5
1 83

48.3

60.5
07.5
I72
256
362
492
642

823
1030
1350
1720
2160
2920
3830
4910

6I
U
,;

7550
9100

6.5
10.4
15.45
21.8
29.7
39
50
62.7
]L11.5

-l

17.4

264
373
507
661
818

I060
l390

>R .\DIEXI'=

--

l
10

1780
2220
3020
3940
5070
6340
7770
9400

1L

GRADIEhT = 100

[TABLE111.

--

D:
inches

1
1.5

2
2.5
3
3.5
4
4.5
5
5.5
6
7
8
9
10
11
12

0.00136
0.00546
0.01228
0.0218
04341
0.0491
04668
0.0874
0.1 104
0.1362
0.165
0.1965
0,267
0,349
0,442
0.546
0.66
0.786

ACdK

0.102
0.1444
70.5
75
0.1772
79
0.2042
0.2282 I_ H:!
0.25
84
0.27
86
0.289
87.5
0.306
89
92
0.323
90.5
0.339
92
0-353
93
0.382
94.5
0.408
96
0.433
97.5
0,457
99
3,478
100
3.50
101
62

-10.0086
0.0555
0.163
0.352
0.638

3.01
3.97
5.13
6.45
9.62
13.6
18.7
24.7
31.5
39.6

lGRADIENT=

64.5
73
78
82
84.5

93.5
94.5
95.5

97.5
99.5
100.5
102
103
104

1
lo,ooo

-l

--

-0.5

GALVAIVIZTD-IRON
PIPES: VALUESOF C m U

TABLE111.-NEW

Cdgi AND Q

u=3

= ( A C d n ) d i AT

u=5

U=7

A C G

-l

73.5

0.0102

0.39

U=20

U=30

-I

-l-l

76

34.5

0.705
1.14
1.71
2.44
3.31
4.36
5.62
7.05

2.55

10.55
15.0

103.5

111

117 11.8
16.8
22.9
30.0
38.8 123 38.5

20.5

1
7
34.5
43.5

GRADIENT =

U=lO

-1-1
04097
0,062
0. I82

VARIOUS VELOCITIES.

114

1
-

1000

117
118.5
GRADIENT =

1
,m

119
120
121.5

11.9
17
23
30.2

I-

GYADTT?\l=

D:
inrhes.

TARLX
IV-NEW ASPRALTEDCAST-IRONPIPES : VALUESOF C
u=1.0

A:

U = 1.5

ACd\/m

6
7
8
9

10
11
12
l5

40
44
48

54
60

C ~ ~ ~ F AQI =T ( D
A C d m ) d ; AT T
I

T V E L~O ~ ~ ~~ E S . ~

u=m

-__.

sqllnre
feert.

3
4

18
21
24
27
30
33
36

u=2.0

IN

181

1240

I560
2120
2780

116.5
118.5

120
121

122
123.5
125
126
127.5
l29
130

1Mi
358
470

602
756
993
I270
l605
21 80
2860

I .25

136
137.5
139
1

= lo,ooO

AC&i

2.71
4.87
7.90

135

GRADIENT

99.5

l25
127
128.5
130
131
I32
133.5

263

ACdK

103
106.5
108.5
110.5
112
114
115.5
116.5
118
121
163

255
347
458
587
735
967

ACl\/m

U = 30.0

ACV%

107.5
1.31
2.78

111

-~

117

1.32
2.7!1
5.03
8.1 3
12.2

120.5

16.7
22.8
30
38.3

139
208
294
400
524
672
842
1105
1415
l780
241 5
3180

--

11.8

48.2
86.7

130
132.5
134.5
136.5
137.5
139

88.5

142
212
300
407
4x15

F89
860
l130
1440
1820
2470
32.50

140.5

142
143
144.5
146
148
149

GRADTEXT =-

17.1
23.4
30.9
39.6
49.8
X!I.T,

144

214
303
410
540
694
870
1136
14-55

1835
2490
3280

AC&i

1.32

107..5
111.5

2.80

114.5

5.03

117
119
121
123
124.5
125.5
127
130
132.5
134.5
136.5

8.13
12.2
17.2
23.6

31
39.6
49.8
89.5
144

214
303

138

416

139.5
141
142
143.5
14.5
146
147.5
149.5

542
G96
870
1140
1460

I835
2190
3290

100

[TABLEV.

TABLE
V.-NEW WROUQHT-IRON
PWES: VALUESOF C IN U
D :
inches

A:
qnarr feet.

U =0.5

dGi :
(feet)&.

U=0.7

1
1.5

2
2.5
3
3.5
4
4.5
5

5.5
6
7
8
9
10
11
12

0.001362
0,0054fi
0,01228
0.02182
0.0341
0.0491
0.0668
0.0874
0.1104
0.1332
0.165
0.1965
0.267
0.349
0.442
0,546
0.66
0.786

0.102
0.144c5
0.1772
0.204
0.228
0.25
0.27
0.289
0.306
0.323
0.339
0.353
0.383
0.408
0.433
0.457
0.478
0.5

67.5
76
80.5
84
87
89.5
91.5
93
94.5
96
97
98
100
101.5
103

104.5
105.5
106.5

0.0094

0.060
0.175
0.374
0.677
1.10
1.64
2.35
3.20
4.20
5.42
6.82
10.2
14.4
19.7
26.0
33.2
41.7

GRADIENT =

70.5
7'3
84
87.5
90
92.5
94.5
96
98
99.5
100.5
101.5
103
104.5
106
107.5
109
112.5
110

1
10,000

2.42

99.5

0.0103
0.065
0.19
0.405
0.727
1.18
1.76
2.51
3.41
4.49
5.78

10.5
14.8
20.3
26.8
34.3
43.1

Gdz

AD

u=5

= (ACdm)d/7AT VARIOUS
VELOCTRES.

u=7

1
-I

U=IO

ACdK

u=30

U=15

--

-l
0.5

u=3

U=2

U=1.5

U=l

106.5
109.5
113.5

7.30
10.85
15.4
21.0
27.7
35.4
44.5

77
85.5
91
94.5
97.5
100
101.5
103
104.5
106
107.5
108.5
110.5
112
113.5
1l 5
116
117

0.0107
0.0675
0.197

0.01l l

79.5

0.421

0.758
1.23
1.83
2.60
3.53
4.65
6.00
7.55
11.3
15.9
21.7
28.6

36.6
45.8

GRADIENT =

100
102
104
105.5
107
108.5
110
111
113
114.5
116
117.5
119
120

1
-

1,000

0.778
1.25
1.87
2.66
3.62
4.75
6.15
7.72
11.5
16.3
22.2
29.3
37.5
47.0

83
91.5
96.5
100.5
103
105.5

0.012
88.5
0.075
97.5
0.218
102.5
104
04ti:I
106.5
I07
0.833 ' 109.5
109.5
1.34
112
111
2.00
113.5
2.85
115.5
113
114.5
3.87
117
118.5
5.08
6.55
120
8.23
121
122.5
12.25

0.072

107.5

0.448
0.802
1.30
1.93

109

2.75

110.5

112
113.5
114.5
116.5
l18
119.5
121
122.5
123.5

48.4

128.5

GRADIENT = 100

0.077
0.222
0.47.5
0.852

2.04
2.92
3.96
5.20
6.70
8.40
12.5
17.6
24.1
31.8
40.5
50.8

!)9.5

105
109
111.5
114
I16
118
119.5
121
122.5
123.5
125
126.5
128.5
130
131
132

94

95.5

103

104.5

108.5

110
114
117
119.5
121.5
123.5
125

112.5

I20

3.03

11.5.5
118
120
122
123.5
123
126
127.5
1-09
130.R
132.5
131

126..i

128
129
130.5
132.5
134.5
136
137
138

135
136

GRBDIENT

1
10

=-

k=n.n75 inch.

k y 0 . 1 inch

k = @ 3 inrh

k=0.15 inch

k-0.5 inch.

k=0.7.5 inch.

7
8

9
10
11
12
15
18
21
24
27
30
33
36
40
44
48
54
60
66
72
78
84

0.0491
0.0873
0.13F
0.196
0.267
0.349
0.442
0.545
0.66
0.785
1.227
1.767
2.405
3.14
3.98
4.91
5.94
7.07
8.73
10.56
12.57
15.90
19.63
23.76
28.27
33.15
3848

k-3.0inch.

ACdk

-l
C

ACl/Z

0.25
0.288
0.322
0.354
0.383
0.408
0,433
0.457
0.478
0.5
0.56
0.613
0.662
0.707
0.75 316
0.79
0.83

0.02
2.0
3.63
5.90
8.90
12.7

75.3
79.3
52.5
85
87.2
89

1.86
3.38
5.52
8.30
11.8
16.3
21.5
27.6
3'4.9
63.3
102
154
219
299

89

105.9
107.5

417

0.866

0.912
0.955
1.0
1.06
1.12
1.17
1.225
1.275
1,325

0.85

69.6
73.7
76.8
79.4
81.5
83.3
85
86.4
87.8

114
115.6
118.5
120.8
121.8

9.5

1432
1950
2570
3290
4150
5120
6200

92.1
94.6
96.7
98.7
100.2
101.7
103
104.4
105.8
107
108.3
110
111.4
112.9
114
115.
116

39.5
508
640
843
1080
1360
1860
2450
3 140
3960
48S0
5900

9.1

0.81
1.76
3.2

65.7
69.6
72.8
75.3
77.4
79.3
80.8
82.5
83.8
85
88.2
90.7
92.8
94.7
96.3
97.8
99
100.2
101.6
103
104.4
106
107.4
108.8
111.1
lI2'l

5.23
7.9
11.3
15.5
20.5
26.4
33.2
60.7
98.0
148
210
287
379
488
614
810
1040
1310

17nn
2360
3020
3820
4710
5710

60
64
67.2
69.7
71.5
73.7
75.3
76.8
78.2
79.3
82.5
85
87.1
89
90.6
92.1
93.5
94.7
96.2
97.5
98.6
100.4
101.9

103.2
104.4
105.5
10F.5

0.74
1.61
2.96
4.85
7.33
10.5
14.4
19.1
24.7
31.1
56.8
92.0
139
19s
270
R37
461
580

765
953
1040
1700
2240
2870
3620
4470
5430

60

65.7
67.8
69.7
71.2
72.8
74.2
56.3

85
86.7
88
89.3
90.6
92.1
93.3
94.6
96.2
2150
97.7
99
low2
101.4
102.4

0.69
1.51
4.57
6.92
9.9
13.6
18.1
23.4
29.5

189

258
341
440
555
733
940
l190
l624
2750
3480

4300
5220

50.3
54.4
57.5
60
62.2
64.1
65.7
67.2
68.5
69.7
72.8
7R.4
77.5
79.3
81
82.5

83.8
85
86.5
87.8
89

90.7
92.2
93.5
94.8
95.8
96.8

0.62
1.37
2.53
4.17
6.35
9.12
12.6
16.7
21.6
27.3
50.0
81.5
123
176
241
320
413
520
688
885
1120
1530
2030
2600
3290
4060
4930

43.2
47.2
50.3
52.9
55
56.9
58.5
60
61.3
62.5
65.7
68.2
70.3
72.2
73.7
75.3
76.7
77.8
79.3
80.6
81.9
83.5
85
86.3
874
88.7
89.7

0.53
1.19
2.21
3.68
5.61
8.1
11.2
14.9
19.4
24.5
45.2
74.0
112
160
220
292
378
477
630
813
1030
1410
1870
2400
3040
3760
4570

37.6

0.46
1.05
1.96
3.28
5.06
7.3
10.1
13.5
17.6

41.6
44.7
47.2
494

51.3
52.9
54.3
55.7
56.9
60
62.5
62.7
66.7
68.2
69.7
71
72.2
73.7
75
76.2
77.8
79.3
80.7
81.9
83
84

22.3
41.3
67.5
103
148
203
271
360
442
587
757
957
1314
1745
2245
2840
3 520
4270

pH* value of vater to give above vaIues of k aftor 30 years' grovt: n cast-. In pipes
7.0
8.4
8.2
7.8
7.4

8.8

AC.\/m

ACd/na

ACdK

18.2
22.3
25.4
27.9
30.1
314
33.6
35.1
36.4
37.6
40.7
43.2
45.3
47.2
48.8
50.3
51.7
52.9
54.3
55.7
564
58.5
60
61.3
62.6
63.7
64.7

0.22
0.56
1.11
1.94
3.07
4.54
6.45
8.75
11.5
14.75
28.0
46.7
72.2
105
14.5
195
2.55
324
432
562
715
987
1320
1705
21 70
2700
3290

-~ -~

.-

3
4
5

k = 1 4 inch.

.-

D:
A:
inchcs. square fret.

Based on 2 1og a = 3.8 - pH, There a denotes the growth-rate in inches per year =

k" - 0405'
30

33.6
37.6
40.7
43.2
45.4
47.2
48.8
50.3
51.7
52.9
56
58.6
60.7
62.6
64.2
65.7
67
68.2
60.7
71
72.2
73.8
75.3
76.7
774
79
80

0.41
0.95
1.79
3.00
4.65
6.72
9.35
12.5
16.3
20.7
38.5
63.3
96.7
139
191
255
330
418
555
716
907
1245
1660
2130
2700
3350
4070

6.8

27.0
31.9
3.i.l
37.6

2.61
4.08
5.92
8.30
11.1

23.9
0.29
274
0.7
31 1.54 1.36
33.5
2.33
35.7
3.62
37.6
5.35
7.50
39.2
40.7
10.1
13.3
16.9

57.2
87.7

0.34
0.8

39.8
41.6
43.2
44.7
46.1
47.2
50.3
52.9
55
5F.9

31.9
52.7
81.5
117
I63
217
283
358
477
619
785
1084
1445
1860
2370
2940
3580

48.8
51.1

68.6

60
61.3
62.6
64.1
65.3
66.5
68.3
69.7
71
72.3
73.3
74.3

302
67.3
383
58.5
510
60
658
61.3
837 62.5
l152
64.2
1532
65.7
1975
67

6.4

6.2

5.5