Pipeline Friction
Pipeline Friction
Pipeline Friction
0.3164
Re 0.25
(1)
In 1933, Prandtl published the formula deduced from a logarithmic velocity profile, with the
constants derived from Nikuradses tests for Reynolds up to 3,400,000:
1
f
2 log Re
0.8 2 log
2.51
Re f
(2)
In 1956, the tests performed by the U: S: Bureau of Mines 3 found a different set of constants,
later adopted by the American Gas Association (AGA):
Re f
1
0.3 2 log 2.825
2 log
Re f
2
f
(3)
In 2005, McKeon, Zagarola and Smits, after refining the original measurements obtained with
the SuperPipe facility at Princeton University for Reynolds between 31,300 up to 35,500,000 in
1996, proposed another set of constants1:
1
1.930 log Re
f
f 0.537
(4)
They estimated their experimental error at 1.1%. The formula deviation from the published test
results is between 1.78% and +0.84%, narrowing to 0.5% for Reynolds between 400,000
and 24,000,000. They validated Blasiuss formula and the exponential velocity profile for
Reynolds below 98,000. McKeons and Blasiuss formulas intersect at Reynolds equal to
66,964. Below this Reynolds, Blasius deviates 0.36% to 0.62% from Princeton test points and
1.74% to 1.41% from Oregon University test points 4 covering turbulent Reynolds from 3,080
upwards with experimental error 2-4%.
Figure 1 for Re < 100,000 shows how Prandtls formula (and McKeons, if extrapolated)
underestimates f.
Figure 2 for Re > 1,000,000 shows AGA formula only slightly on the unsafe side for large
Reynolds and Prandtls increasingly underestimating the friction factor.
0.035
Blasius
Prandtl
0.03
0.025
McKeon
non
0.02
Re = 38383
Re = 66964
0.015
0.01
10000
100000
Reynolds
Figure 1
0.0115
Re = 1,563,705
0.011
0.0105
0.01
0.0095
Prandtl
McKeon
0.009
0.0085
USBM
0.008
1000000
10000000
Reynolds
Figure 2
k
1.74 2 log
3.706 D
(5)
Colebrook and White ran similar tests using sand of different size on the same pipe and with
highly corroded pipes. The general behavior was similar to Nikuradses test, but In some cases
of highly non-uniform roughness the minimum disappeared and they proposed a combination of
Prandtls smooth-pipe-law and Nikuradses rough-pipe-law in order to provide a safe estimate of
the friction factor in the transition zone (1939):
k
2.51
2 log
f
3.7 D Re f
(6)
In 1944, Moody published his chart of the friction factor versus Reynolds, using sand roughness
as a parameter, based on Hagen-Poiseuille ( f 64 / Re ) for Reynolds below 2000 and on
Colebrook-White for Reynolds above 4000. This chart was enthusiastically adopted and
remains the standard for calculating pipe friction through different computer approximations.
In 1956, the U. S. Bureau of Mines3, testing straight commercial pipes, found the same behavior
as Nikuradse, with a minimum in the transition zone only about 2% below the final rough-pipe
friction factor.
In 1965, the outstanding work of Uhl and others5 for the American Gas Association (AGA),
testing actual gas pipelines, validated the USBM formula for the smooth-pipe-law and adopted
Nikuradses equivalent sand roughness for the rough-pipe-law. The tests suggested that the
transition zone reduced to a single point, corresponding to the transition Reynolds number Ret.
Friction Factor for Rough Pipe
These results confirmed that Colebrooks
friction factor was too conservative for commercial
transition zone not always reduced to a single point. They preferred to use a transition exponent
0.026
log
n
3.7 D
f
0.022
2.51
Re f
(7)
Colebrook
formulas for the smooth-pipe-law, including Zagazola original formula for the SuperPipe 7.
0.018
0.016
0.014
10000
AGA
100000
1000000
Reynolds
10000000
Figure 3
0.035
CBMN
k=0
0.03
0.025
Colebrook
k = 0.0018 in.
0.02
0.015
0.01
10000
100000
Reynolds
Figure 4
A single formula for friction in pipelines
According to the conclusions of the Princeton research group 1, both Blasius and McKeons
formulas are required for covering the whole range of turbulent Reynolds numbers for smooth
pipe.
The Colebrook-Blasius-McKeon-Nikuradse (CBMN)2 formula with n = 1 is apt to replace the
traditional Colebrook formula for pipes subject to corrosion.
The AGA report5 conclusively shows that the extended transition zone resulting from
Colebrooks method is too conservative for gas pipelines. The transition exponent n, however,
already adopted by GERG, allows the use of the Princeton formulas and Nikuradses rough-pipe
law to cover all practical cases (CBMNn), reducing the extension of the transition zone to match
actual pipeline roughness.
Applying CBMNn with n = 40 provides a friction factor only 0.3% above AGA rough-pipe factor
at the transition Reynolds for typical cases of the rough-pipe transmission factor Ft
1
f
around 10. Adopting n = 10, the difference increases to 1.2%. Values of n between 1 and 10
have been used to match friction measurements in some European gas lines.
2
k
log 10 1.125
n
3.706 D
f
(8)
1.897747
2
log
n
Re f
0.965 n
3
.
706
D
(9)
1
from the
f
pressure drop formula for a straight pipe and plotted it versus Reynolds, finding that the curve
kept approximately parallel, but always below the USBM smooth-pipe law.
They identified this displacement as the effect of minor losses owing mainly to bends and weld
beads at pipe joints and also to deposits, block valves and fittings. They kept the USBM Ft and
added a drag factor Ff, less than unity, multiplying Ft in the pressure drop formula. From the
scarce information available, they built a graph (Fig. 5 is a computerized version) for Ff as a
function of the bend index (total changes of direction in degrees per mile) showing four curves
for bare steel, plastic-lined, pig-burnished and sandblasted pipe. The effect of weld beads
decreased in that order and the drag factor consequently increased.
Drag Factor
1.00
sandblasted
Drag Factor Ff
pig-burnished
plastic-lined
bare steel
0.90
1
10
100
1000
Figure 5
For the 30 cases of totally turbulent flow with at least four points at the right of the transition
Reynolds number, they calculated Ft in the same way and derived the effective roughness ke
from Nikuradses law. They acknowledged that ke, concentrating between 900-1000
microinches (22.9-25.4 micron), was higher than the absolute roughness k obtained by USBM,
concentrating between 600-700 microinches (15.2-17.8 micron), because of the absence of
bends and weld beads in the USBM test pipes.
The intersection of the smooth-pipe law and the rough pipe law provided the formula for the
transition Reynolds number:
3.7 D
Re t 5.65
ke
1
Ff
3.7 D
log
ke
1
Ff
They pointed out that Ret should depend only on k/D (the transition is reached when the
(11)
thickness of the laminar sublayer becomes less than the absolute roughness). Unfortunately,
they did not follow this clue to find the relationship between k, ke and Ff:
For the USBM tests Ff = 1. If the same pipes were welded and laid on a bending route, the
transition Reynolds remaining the same, the transmission factor calculated through the smoothpipe and the rough-pipe law become displaced by the same factor, so that:
2.825
2 Ff log
Re t f
ke
k
2 Ff log
3.7 D 2 log 3.7 D
(12)
Consequently:
3.7 D
Ff
ke
3.7 D
(13)
Replacing in (11):
3.7 D
3.7 D
log
k
k
Re t 5.65
(14)
The AGA group also identified the relationship of the drag factor to the total length Lt (equal to
the sum of straight length L plus total equivalent length Le of bends, weld beads, valves and
fittings) or to the respective total loss coefficient K:
f L
L
Lt
L Le
f
f K
D
D
Ff 2 D
D
N f
L 1
K f
1
2
D Ff
Ff
K D
L
1
Lt
f L
(15)
0.5
(16)
These formulas can predict the effect on the drag factor when adding intermediate scraper
traps, where individual K coefficients are known. A more accurate prediction of the drag factor
for a new pipeline would be possible, if values of K became available for weld beads and for
angles between 3 and 10 usually present along the route. The qualitative relationship between
the drag factor and the loss coefficients or equivalent length of minor-loss elements has already
been analyzed by Uhl.
The number of friction velocity heads N, usually applied to gas and liquid pipes in refineries and
industrial plants, could also be applied to liquid and gas pipelines for both the partially and
totally turbulent flow regions (using absolute instead of effective roughness).
Using function CBMNn(k,D,Re,n) for calculating f and N in the liquid or gas pressure drop
formulas provides a unified treatment for all cases of piping and pipelines within the precision
afforded by the Princeton tests.
The applications of absolute roughness
The absolute equivalent sand roughness k for actual gas pipelines can be obtained from the
fourteen series of AGA tests for bare steel crossing the transition point. Tests ZA-1 and ZB-1 (810 years without cleaning) can be statistically rejected as not belonging to the same population.
The remaining tests give mean effective roughness of 670-in. or 17-micron with 48% standard
deviation. The same tests provide 316-in. or 8-micron mean absolute roughness with 18%
standard deviation (almost three times less).
An absolute roughness of 433-in. or 11-micron (mean + 2 standard deviations) can be adopted
for design with CBMNn or the effective roughness can be estimated from equation (13), as a
function of the drag factor, to be used in AGA or GERG formulas.
The only three tests for pig-burnished pipe crossing the transition Reynolds give 221-in.
absolute roughness (5.6-micron).
The tests with plastic-lined pipe did not attain the transition Reynolds. Absolute roughness was
below 133-176 in. (3.4-4.5 micron).
Sand-blasted pipes also remained below the transition Reynolds with absolute roughness less
than 169-188 in. (4.3-4.8 micron).
The application of the absolute roughness to assess potential capacity increase in existing
pipelines has already been analyzed 8: For example, let us consider AGA tests A and ZB-1 for
bare steel pipe:
Test
A
ZB-1
ID inches
12
25.375
ke microinches
1622
1619
Ff
0.8662
0.9200
To ascertain the possibility of increasing pipeline capacity, the effective roughness alone gives
no practical guidance. The absolute roughness calculated from equation (13) is 335 in. (very
near the mean value) for test A and 623 in. for ZB-1. The drag factor is within the expected
range for test A with high bend index, and looks low for ZB-1 in flat country
Case A is consequently operating in the normal range and no cleaning is necessary.
Case ZB-1, having the same effective roughness, shows unusually high absolute roughness
indicating corrosion. Pig burnishing would provide a significant capacity improvement factor:
log
log
0.000221
3.7 * 25.375
1.087
0.000623
3.7 * 25.375
The drag factor should increase at least 2% (weld beads reduced and deposits eliminated)
providing a total capacity increase around 11% for the same initial and final pressures. As a
matter of fact, test ZB-6 after pig-burnishing, gave the following results:
Test
ZB-6
ID inches
25.375
ke microinches
447
Ff
0.9386
Absolute roughness is 200.5 in. (better than average) and total capacity increase factor is:
0.0002005
3.7 * 25.375
0.9386 * log
Figure 6 shows the contribution of higher Ff and lower absolute roughness to obtain 11.7%
capacity increase. The pressure drop formula in Appendix 2 proves that capacity is directly
proportional to Ff Ft. Since Reynolds is directly proportional to capacity, it is easy to represent
the operating curve when keeping the same initial and final pressures.
k = 200.5 microinches
Ff = 0.9386
k = 623 microinches
Ff = 0.9200
Figure 6
USBM tests showed that for perfectly dry gas the sand roughness obtained from pressure drop
measurements approximately coincided with the mean peak-to-valley height measured on the
wall roughness. Recent research9 has improved the correlation between different techniques of
direct roughness measurement and Nikuradses sand roughness. The presence of small
quantities of liquid in two USBM tests considerably reduced the effective sand roughness
obtained from pressure drop. This effect is certainly present in the actual gas pipelines tested by
AGA, but does not completely cover the wall roughness effect: test ZA-1 provided k = 759 in
(19.3 m). After pig-burnishing (test ZA-6), k = 249 in (6.3 m). After sandblasting (test ZA-9),
k < 169 in (4.3 m, transition Reynolds was not reached). In the two first cases, at least, wall
roughness had an effect on pressure drop. In the third case, wall roughness may have been
covered by liquid presence. This effect is obviously not present in liquid pipelines, where
modern wall-roughness measurement techniques can be fully applied.
AGA Ff = 0.9483
ke = 642 microinches
Colebrook Ff = 1
ke = 642 microinches
CBMNn Ff = 0.9483
k = 340.6 microinches
Figure 7
For liquid pipelines, there is a firm tradition to use Colebrook formula with an effective
roughness of 1800 microinches (46 microns) and the straight length. However, the theoretical
concepts of the thickness of the laminar sub-layer compared to absolute wall roughness to
define the transition from smooth pipe behavior to rough pipe behavior are the same for liquid
and gases. Minor-loss effects should be masked inside the effective roughness in the roughpipe zone, but should cause an underestimate of the pressure drop in the smoothpipe zone,
requiring the application of a drag factor. While the drag factor appeared to be constant in AGA
tests with Reynolds above 1,000,000, equation (16) shows a dependence on K/f, which varies
with Reynolds. To provide an order-of-magnitude estimate, Hoopers prediction of K versus
Reynolds10 could be applied to a 45 bend, 29-in. ID and 1800-in. roughness. A drag factor of
0.950 at Re = 10,000,000 would reach 0.953 at Re = 1,000,000 (validating the assumption of
constant drag factor in gas pipelines) and 0.976 at Re = 10,000. Minor losses causing 10%
additional pressure drop at high Reynolds number would reduce this effect to 5% at Re =
10,000.
The following table illustrates how the safety margin provided by a conservative effective
roughness vanishes when the Reynolds decreases below 100,000. For a typical 29-in. ID
pipeline, Darcy friction factor is calculated from Colebrooks formula for 1800-in. effective
roughness, 1009-in. absolute roughness (corresponding to a typical Ff = 0.95) and zero
roughness:
Table 1
29-in. ID Pipeline
friction factor
Colebrook
Colebrook
ke = 1800 in. ke = 1009 in.
Reynolds
10000
25000
50000
100000
250000
500000
1000000
2500000
5000000
0.03098
0.02467
0.02111
0.01832
0.01553
0.01399
0.01286
0.01189
0.01147
0.03094
0.02461
0.02102
0.01818
0.01529
0.01364
0.01237
0.01120
0.01065
Colebrook
ke = 0 in.
0.03088
0.02452
0.02089
0.01799
0.01497
0.01316
0.01165
0.01001
0.00898
safety margin
ke = 1800 in. ke = 1800 in.
versus
versus
smooth
ke = 1009 in.
0.31%
0.14%
0.62%
0.27%
1.07%
0.46%
1.83%
0.79%
3.73%
1.56%
6.31%
2.54%
10.39%
3.91%
18.83%
6.13%
27.73%
7.74%
Water pipelines provide a good example of operation above Reynolds 1,000,000 (where the
drag factor should be similar to gas pipelines), well known liquid properties and uniform
roughness when using epoxy-lined steel pipe. Direct measurement of wall roughness has
shown an initial value of 150 in.11
In this field, the Hazen-Williams formula has been widely used since the beginning of the 20 th
century:
Hf 10,67
1, 852
L
D 4 ,87
(17)
The American Water Works Association (AWWA) recommends a coefficient C =140+0.17d for
new pipe with smooth lining12, where d is the diameter in inches.
Table 2 compares the results of Hazen-Williams with CBMN (n=40 for uniform roughness),
adjusting the drag factor to obtain the best match. A very good coincidence is obtained with Ff =
0.943, which is the average drag factor in AGA tests. At lower Reynolds, the comparison
confirms that the drag factor should be increased.
Colebrook, with the traditional 1800-in. roughness, shows the same behavior as in gas
pipelines: about 6% underestimate for the lower Reynolds and about 6% overestimate for the
higher Reynolds.
Table 2
velocity
m/s
Reynolds
-
1
1.5
2
2.5
3
3.5
4
4.5
5
895350
1343025
1790700
2238375
2686050
3133725
3581400
4029075
4476750
deviation
Hazen
Colebrook
versus
versus
CBMNn
CBMNn
-1.49%
-5.5%
-0.77%
-3.3%
-0.38%
-1.4%
-0.14%
0.3%
0.01%
1.8%
0.11%
3.2%
0.17%
4.5%
0.21%
5.6%
0.24%
6.8%
Further experimental research on the drag factor (or minor-loss K coefficients), including low
Reynolds numbers, and absolute wall roughness measurements would certainly improve the
design and maintenance of liquid and gas pipelines.
APPENDIX 1
Basic procedure for CBMNn function:
Function CBMNn(k, D, Re, n)
' Darcy friction factor combining Colebrook-Blasius-McKeon-Nikuradse formulas
'k = absolute roughness, D = inside diameter, Re = Reynolds, n = transition factor
k and D in the same units
Dim Raux, Fto, Ft, Fk
'Raux = auxiliary Reynolds, Ft = transmission factor = fDarcy^-0.5
'Fto = previous guess for Ft, Fk = relative roughness factor = k/(3.7*D)
If Re <= 2000 Then
'For laminar flow (Reynolds <= 2000) Hagen-Poiseuille formula
CBMNn = 64 / Re
Else
If Re <= 66964 Then
'Blasius for Re<=6696
If Re < 3080 Then
'for transition flow (2000 < Re < 3080)
Raux = 3080
Else
p1 p 2 N
Q2
g z 2 z1
2 A2
2
N
QA
p1 p 2 g z1 z 2
(18)
(19)
Z RT
1 b2 Qb2
2
p p N 1 2 ln
g
z
z
M
2
1
2 A2
M
2
1
2
2
(20)
1
Qb
N
1 ln
1
2
0.5
A
b
2
1
M
2 M2 g z1 z 2
p 22
Z RT
(21)
The logarithmic term accounts for the increase in the kinetic energy as the gas expands along
the line. For natural gas pipelines, where the velocity is kept below 20 m/s, this term remains
under 0.003 and is usually neglected.
If the friction factor is calculated with the absolute roughness, so that the drag factor applies for
the partially and totally turbulent regions, the previous equations become:
f L
L
Lt
L Le
2 L
K
f
f
Ff Ft
2
D
D
Ff D
D
N f
f L b2 Qb2
2 M2 g z 2 z1
2
2
Ff D A
p12 p 22
Qb Ff Ft
D A
L b
2
1
M
p 22
Z RT
Z RT
M
2 M2 g z1 z 2
(22)
(23)
(24)
NOMENCLATURE
A
inside diameter
Ff
drag factor
Ft
transmission factor
acceleration of gravity
absolute roughness of pipe wall (equivalent sand roughness obtained from pressure
drop at totally turbulent flow in a straight pipe; it can be related to direct measurement
of wall roughness)
ke
effective roughness of pipe wall (equivalent sand roughness obtained from pressure
drop at totally turbulent flow in the actual line, including the effect of weld beads,
changes of direction, deposits, valves and fittings)
Le
equivalent length of weld beads, changes of direction, deposits, valves and fittings
Lt
absolute pressure
flow rate
Re
Reynolds number
fluid density
Subindices
1
REFERENCES
1. McKeon, B. J., M. V. Zagarola, and A. J. Smits, A new friction factor relationship for fully
developed pipe flow, J. Fluid Mech., Vol.538, pp. 429-443, 2005.
2. Cordero, G. O., An improved experimental correlation for Darcy friction factor,
Hydrocarbon Processing, July 2008.
3. Smith, R. V., et al., Flow of Natural Gas Though Experimental Pie Lines and Transmission
Lines, U. S. Bureau of Mines Monograph 9, New York, American Gas Association, 1956.
4. McKeon, B. J., C. J. Swanson, M. V. Zagarola, R. J. Donnelly and A. J. Smits, Friction
factors for smooth pipe flow, J. Fluid Mech., Vol.511, pp. 41-44, 2004.
5. A. E. Uhl et al., Steady Flow in Gas Pipelines, Institute of Gas Technology Technical Report
No. 10, American Gas Association, New York, 1965.
6. Dewerdt, F., La dtermination des pertes de charge dans les canalisations, Gaz
dAujourdhui, Vol. 106 (1984), pp- 89-94.
7. Gersten, K. et al., New transmission-factor formula proposed for gas pipelines, The Oil
and Gas Journal, Feb. 14, 2000.
8. Cordero, G. O., A critical analysis of AGA and Colebrook methods for calculating friction in
gas pipelines, First Latin American Gas Congress, 25-30 November 1984, Provincia del
Neuqun, Argentina (IAPG, Spanish text).
9. Sletfjerding, E., J. S. Gudmunsson, Friction Factor in High Pressure Natural Gas Pipelines
from Roughness Measurements, International Gas Research Conference, November 5-8,
2001, Amsterdam
10. Hooper, W.B. ,The two-K method predicts head losses in pipe fittings, Chem. Eng. Aug.
24, 1981
11. Worthingam, R. G. et al., Cost study justifies internal coating on 48-in. gas line, Oil & Gas
Journal, May 30, 1994
12. AWWA Manual M11, Steel Pipe A Guide for Design and Installation, American Water
Works Association, 1987