ARTICLE IN PRESS

International Journal of Mechanical Sciences 49 (2007) 989–1000

www.elsevier.com/locate/ijmecsci

The effect of major parameters on simulation results of gas pipelines

Farzad Abdolahia,, Ali Mesbahb, Ramin B. Boozarjomehryc, William Y. Svrcekd
a
Department of Chemical Engineering, Faculty of Engineering, University of Tehran, Tehran, Iran
b
DelftChemTech, Faculty of Applied Sciences, Delft University of Technology, Delft, The Netherlands
c
Department of Chemical and Petroleum Engineering, Sharif University of Technology, Tehran, Iran
d
Department of Chemical Engineering, University of Calgary, Alta., Canada
Received 12 May 2006; accepted 10 December 2006
Available online 5 February 2007

Abstract

Predictions of the gas temperature and pressure profiles are vital to the design and operation of gas transmission lines. Available
analytical methods for the calculation of these profiles are evaluated and a numerical framework for the rigorous calculation has been
developed. The predictions from both the analytical and numerical procedure have been compared to field data from the Iranian Gas
Trunk-lines (IGAT). These comparisons showed that all the available methods were tuned using data obtained from small to medium
diameter pipes extrapolated poorly to large diameter pipelines. In order to improve the predictions for large diameter pipelines, the effect
of model parameters such as soil thermal conductivity, pipe relative roughness and velocity profile correction factor has been evaluated.
The results show that temperature and pressure profiles at high Reynolds number are sensitive to the Fanning friction factor; however,
thermal conductivity and velocity distribution correction factor have almost no effect on the temperature and pressure profiles provided
these parameters were set at an average acceptable industry value. Since the pressure profile for large diameter pipes was most sensitive to
the Fanning friction factor a parameter optimization method was used to fine-tune the Fanning friction factor as a function of Reynolds
number at an average accepted industry relative pipe roughness.
r 2007 Elsevier Ltd. All rights reserved.

Keywords: Gas transmission pipelines; Numerical method; Optimization; Friction factor

1. Introduction In this paper these two approaches are reviewed and a

developed rigorous numerical approach is used to evaluate
In order to design or operate gas transmission systems, the influence of model parameters such as pipe roughness, soil
there has been a need for a workable method to relate the thermal conductivity and velocity profile correction factor on
transmitted gas volume to the temperature, pressure, pipe pipeline pressure and temperature profiles. The developed
length, pipe diameter, pipe roughness, Reynolds number numerical procedure is validated using field data from the
and gas composition. There have been two general solution Iranian Gas Trunk-lines (IGAT) by adjusting the pipe
approaches, the first of these is closed form analytical and roughness and soil thermal conductivity to minimize the
involved numerous simplifying assumptions and approx- differences between the model predictions and the field data.
imations. These analytic closed form equations often lead
to poor predictions, mainly due to the simplifying 2. Model development
assumptions, which were primarily made for ease of hand
calculations. The second approach uses numerical meth- 2.1. Momentum balance
ods, which greatly mitigates the need for many of the
simplifying assumptions and/or approximations. A steady-state momentum balance on a differential
control volume of pipe leads to

Corresponding author. Fax: +98 21 66957784. 1 dP u du 2fu2 g dZ

þ þ þ ¼ 0, (1)
E-mail address: [email protected] (F. Abdolahi). r dL agc dL gc Di gc dL

0020-7403/$ - see front matter r 2007 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ijmecsci.2006.12.001
 ARTICLE IN PRESS
990 F. Abdolahi et al. / International Journal of Mechanical Sciences 49 (2007) 989–1000

Nomenclature Sk dimensionless elevation and kinetic energy

change factor
a constant, Eq. (2) T temperature of gas (K)
ai constant, Eq. (18) T0 temperature of gas at pipe intake (K)
aij constant, Eqs. (24)–(26) Tb average temperature (K)
b constant, Eq. (2) Tb temperature at standard condition (K)
cj constant, Eqs. (24)–(26) Ts temperature of surrounding (K)
Cp heat capacity (J/kg K) T e0 temperature of surrounding at pipe intake (K)
Di pipe internal diameter (m) Tg temperature of ground (K)
Do pipe outside diameter (m) u gas velocity (m/s)
E efficiency factor, Eq. (18) U overall heat transfer coefficient ðJ=m2 s KÞ
f Fanning friction factor W constant, Eqs. (15)–(17)
g gravitational acceleration ðm=s2 Þ z compressibility factor
gc conversion factor zave average compressibility
ge surrounding temperature gradient (K/m) Z elevation (m)
h specific enthalpy rate (J/kg s) a correction factor to compensate for variations
hb depth of burial of pipe to centerline (m) in the velocity profile over the pipe cross-
k soil thermal conductivity (J/m s K) section
K ij 4th order Runge–Kutta parameters b correction factor to Fanning friction factor
L pipe length (m)  absolute pipe roughness (m)
m_ gas mass-flow rate (kg/s) f dimensionless correction parameter
P pressure (Pa) j constant, Eqs. (19)–(23)
Pave average pressure (Pa) y inclination angle with horizontal
Pb pressure at standard conditions (Pa) gg specific gravity of gas, which is defined as the
q heat flow (J/kg s) ratio of density of dry air with both at standard
Q gas-flow rate (scm/D) temperature and pressure
R universal gas constant (J/gmol K) Z Joule–Thompson coefficient (K/Pa)
Re Reynolds number mg gas viscosity (Pa s)
S dimensionless elevation factor with the kinetic y inclination angle with horizontal
energy change ignored r gas density ðkg=m3 Þ
D difference

where the correction factor a represents the variation of the pffiffiffi

 Re f
gas velocity profile over the pipe cross-sectional area. This Transition regime: 8o o100.
factor depends on velocity profile and varies from 0.75 for Di
laminar flow to about 1 for fully developed turbulent flow pffiffiffi
 Re f
[1]. Aziz [2] suggested that a value of 0.9 could be used for Completely rough regime: 4100.
Di
practical pipeline modeling.
In order to solve Eq. (1), the gas density and friction
factor are required. The gas density can be obtained from 2.1.1. Hydraulically smooth regime
an appropriate equation of state (EOS). The most In the hydraulically smooth regime the friction factor, f,
commonly used EOSs for the gas transmission systems is a function of Reynolds number only and the resistance to
are the AGA equation by Starling and Savage [3] and the flow is independent of =Di . This regime is restricted to
Peng and Robinson [4] EOS. small values of Re and =Di in that the surface roughness/
The friction factor is related to fluid-flow rate, fluid protrusions are completely hidden in the laminar boundary
properties and pipe characteristics through two dimension- layer. The literature on the hydraulically smooth regime
less groups: Reynolds number and relative roughness. In abounds with reliable friction factor equations. In generally
order to calculate the Reynolds number, the viscosity of there are Blasius form or power law relationships and
natural gas is required and can be estimated by using Prandtl’s correlation [9] for the calculation of friction
correlations of either Lucas [5] or Dean and Stiel [6]. factor in this laminar-flow regime. The Blasius form of
The friction factor is typically estimated by using the smooth pipe friction factor is
Moody diagram [7], in which the friction factor estimation f ¼ a Reb (2)
is represented for three distinct regimes [8]:
pffiffiffi where the constants a, b are power law relationships as
 Re f presented in Table 1. The data from numerous measure-
Hydraulically smooth regime: o8.
Di ments of the pressure drop in smooth pipes were analyzed
 ARTICLE IN PRESS
F. Abdolahi et al. / International Journal of Mechanical Sciences 49 (2007) 989–1000 991

Table 1
Constants a and b in Eq. (2)

Parameter Blasius Weymouth 1

7th
1
8th
1
9th Mod. 19th 1
10th
Panhandlec Mod. IGTd
a a a
[11] [12] power law power law power law power lawb power law a
Panhandlec

a 0.079 0.008 0.0763 0.0563 0.0437 0.046 0.03475 0.02118 0.003678 0.04675
b 0.25 0.3333 0.25 0.22 0.20 0.20 0.182 0.1461 0.03922 0.20
a
Ward-Smith [13].
b
Knudsen and Katz [14].
c
IGT [15].
d
Beggs [16].

Table 2
Average and maximum errors using power law-friction factor relationships based on Prandtl’s equation, Ouyang and Aziz [17]

1 1 1
Re range Blasius 7th 8th 9th Mod. 19th 1
10th
Panhandle Modified IGT
power law power law power law power law power law Panhandle

Average deviation from Prandtl’s universal friction law for smooth pipes
2000–4000 2.366 5.703 13.060 19.608 15.377 26.292 40.168 75.669 13.97
4000–1.0e4 1.253 2.309 7.832 13.216 8.648 19.243 32.475 70.003 7.159
1.0e4–1.0e5 1.288 2.672 3.044 4.703 1.919 8.137 17.559 54.662 2.850
1.0e5–1.0e6 9.387 12.484 7.071 3.961 1.179 3.541 6.045 34.056 2.742
1.0e6–1.0e7 24.952 27.517 17.944 10.821 6.128 6.667 1.309 11.538 4.850
1.0e7–1.0e8 41.380 43.384 31.659 21.887 17.776 14.806 2.191 12.719 16.436
Average in the whole Re range 13.438 15.678 13.435 12.366 8.505 13.114 16.624 43.108 8.006

Maximum deviation from Prandtl’s universal friction law for smooth pipes
2000–4000 4.543 7.806 15.838 22.783 18.719 26.656 43.626 77.941 17.393
4000–1.0e4 2.135 3.925 10.576 16.695 12.310 23.157 36.965 73.396 10.881
1.0e4–1.0e5 2.236 4.661 5.706 10.364 5.652 15.947 28.632 66.833 4.114
1.0e5–1.0e6 14.235 17.166 10.011 5.341 2.372 5.087 12.467 47.970 4.041
1.0e6–1.0e7 30.680 33.049 22.422 14.155 9.637 8.839 3.393 26.553 8.164
1.0e7–1.0e8 46.827 48.644 36.529 26.116 22.228 18.220 3.352 20.204 20.960
Maximum in the whole Re range 46.827 48.644 36.529 26.116 22.228 29.656 43.626 77.941 20.960

by Prandtl and led to However, this equation is implicit in f and requires a trial
0:5 0:5 and error solution. Many explicit approximations of this
f ¼ 4 log½Re f   0:4. (3) correlation are available in literature and are shown in
However, the most widely used explicit approximation of Table 3. The maximum and average errors for most used
Eq. (3) was provided by Colebrook [10]: explicit approximation are shown in Table 4.
 
0:5 Re
f ¼ 3:6 log . (4) 2.1.3. Completely rough regime
7
This regime is encountered at very high Re numbers and
The prediction errors of power law friction factor relation- =Di , where all the pipe surface roughness extends beyond
ships compared to Prandtl’s equation are given in Table 2. the laminar boundary layer. For the completely rough
Zagarola [18] showed that Prandtl’s equation was not regime the friction factor is estimated from [34]:
accurate for high Reynolds numbers and provided a new  
0:5 
correlation for the friction factor in smooth pipe flow f ¼ 4 log . (7)
3:706Di
regime:
Most recently Sletfjerding and Gudmundsson [35] have
f 0:5 ¼ 3:778 logðRe f 0:5 Þ þ 0:4213. (5) proposed a new correlation (Eq. (8)) for the totally rough
2.1.2. Transition regime regime which improves the prediction of pressure drop
In this regime, both Re and =Di contribute to the flow when compared to the predictions using the Nikuradse
resistance. For the transition regime the friction factor is method:
usually estimated by a correlation developed by Colebrook 


and White [19]: f 0:5 ¼ 3:876 log 1:463 . (8)
  Di
0:5  1:255 Sletfjerding and Gudmundsson [35] also combined Eq. (8)
f ¼ 4 log þ . (6)
3:71Di f 0:5 Re with the Zagarola correlation for smooth pipe flow
 ARTICLE IN PRESS
992 F. Abdolahi et al. / International Journal of Mechanical Sciences 49 (2007) 989–1000

Table 3
Explicit approximation of Colebrook–White equation for friction factor

Moody [20] 1=3 !

106


f ¼ 0:001375 1 þ 2  104 þ
Di Re
 0:225  0:44  0:134
Wood [21]
 
   
f ¼ 0:026 þ 0:133 þ 22 Rec ; c ¼ 1:62
Di Di Di Di
 
Eck [22]  15
f 0:5 ¼ 4 log þ
3:715Di Re
 
Jain [23] 0:5  21:25
f ¼ 2:28  4 log þ 0:9
Di Re
Swamee and Jain [24] 0:0625
f ¼
½logð=3:7Di þ 5:74=Re0:9 Þ2
Churchill [25] "  #1=12
8 12 1
f ¼2 þ
Re ðA þ BÞ1:5
( "  #)16
7 0:9 37530 16
 

A ¼ 2:457 ln þ 0:27 ; B¼
Re Di Re
( " #)
Chen [26]  5:0452
 1:1098
 5:8506
0:5
f ¼ 4:0 log 0:2698   log 0:3539 þ 0:8981
Di Re Di Re
 
Round [27]  6:5
f 0:5 ¼ 3:6 log 0:27 þ
Di Re
!
Barr [28]  4:518 logðRe=7Þ
0:5
f ¼ 4 log þ
3:70Di Reð1 þ ð1=29ÞRe0:52 ð=Di Þ0:7 Þ
" #
Haaland [29] 
1:11
6:9
f 0:5 ¼ 3:6 log þ
3:70Di Re
  
Zigrang and Sylvester I [30]  5:02  13:0
f 0:5 ¼ 4:0 log  log þ
3:7Di Re 3:7Di Re
   
Zigrang and Sylvester II [30] 0:5  5:02  5:02  13:0
f ¼ 4:0 log  log  log þ
3:7Di Re 3:7Di Re 3:7Di Re
SerghidesI [31]  2 2
ðB  AÞ
f ¼ 0:25 A 
C  2B þ A
   
 12  2:51A
A ¼ 2 log þ ; B ¼ 2 log þ
3:7Di Re 3:7Di Re
 
 2:51B
C ¼ 2 log þ
3:7Di Re
Serghides II [31] 2
ðA  4:781Þ2

f ¼ 0:25 4:781 
B  2A þ 4:781
 
Manadilli [32] 0:5  95 96:82
f ¼ 4:0 log þ 0:983 
3:70Di Re Re
( " 0:9345 !#)
Romeo et al. [33]  5:0272  4:567


0:9924 
5:3326
0:5
f ¼ 4:0 log  log  log þ
3:7065Di Re 3:827Di Re 7:7918Di 208:815 þ Re

resulting in Eq. (9) for both the smooth and rough-flow these values could be increased by 0.76–1:27 mm per year.
regimes They have also shown that the rate of change in pipe
 1:03 ! roughness for internally coated pipes is much less than that
0:5 0:755 0:5  for uncoated pipes, i.e. ranging from 1.27 to 1:9 mm for
f ¼ 3:78 log f þ . (9)
Re 0:683Di every five years.

The pipe roughness used in these correlations or

2.2. Energy balance
equations is not easily or directly measurable and changes
with pipe use and fluid type. That is, a fluid may increase
A steady-state energy balance around an element of fluid
pipe roughness by erosion or corrosion or by precipitating
flowing through a differential length of pipe results in
materials that adhere to the pipe wall. The absolute
roughness values that are normally measured and used
u2
 
for uncoated commercial pipes are within the normal range d g dq
hþ þ Z  ¼ 0. (10)
of 16–19 mm. Golshan and Narsing [36] have shown that dL 2agc gc dL
 ARTICLE IN PRESS
F. Abdolahi et al. / International Journal of Mechanical Sciences 49 (2007) 989–1000 993

Table 4
Deviation from the Colebrook–White equation for rough pipes (for the whole range of Reynolds number and relative pipe roughness)

Error Serghides I Serghides II Chen Z–S I Z–S II Haaland Jain Eck Churchill Wood Moody

Average error 0.00037 0.0359 0.137 0.234 0.0287 0.582 0.929 3.010 4.092 5.107 6.276
Maximum error 0.003 0.355 0.698 1.060 0.206 2.952 4.598 11.962 70.730 32.553 26.790

Ouyang and Aziz [17].

Rearranging Eq. (7) results in the pipeline cross-section velocity variation is negligible,
    hence there is no need for an integrated form of the
qh dT qh dP u du g dZ dq continuity equation.
þ þ þ  ¼ 0.
qT P dL qP T dL ag c dL gc dL dL
(11) 3.1.1. Pressure drop calculations
In Eq. (11) values of ðqh=qTÞP and ðqh=qPÞT can be The integrated form of Eq. (1) results in the general form
determined from an EOS, such as Peng and Robinson [4], of the flow equation, as follows:
Lee–Kesler [37] or Plocker et al. [38].  
p Tb Rgc 0:5

Di2:5
The heat flow to/from the surrounds (ground, water or Q¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi
8 Pb 28:97 fzave T ave gg
air) to the pipeline for the unit mass of flowing gas is given  2 0:5
by P  eSk P22 S
 1 , ð15Þ
dq pDo UðT s  TÞ L eS k  1
¼ . (12)
dL m
_ where
Note, Kennelly [39] proposed the following equation for Sð1  ðDi =2afLÞ lnðP2 =P1 ÞÞ
the heat transfer of buried pipes: Sk ¼ , (16)
1 þ ðDi =4afLÞS
dq 2pkðT g  TÞ
¼ . (13)
57:94gg g
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
dL _ lnðð2hb þ 4h2b  D2o Þ=Do Þ
m S¼ L sin y. (17)
zave RT ave gc
A basic assumption in Eq. (13) is that the resistance to heat
Eq. (15) takes into account the effects of elevation,
transfers in the fluid film and pipe wall is negligible, or the
friction and kinetic energy change. As the kinetic energy
temperature of gas is at temperature of the pipe wall.
effect is usually small, it can often be neglected. Thus,
substituting the constants and neglecting the kinetic energy
2.3. Mass balance
term in Eq. (1), the simplified form of the general flow
equation in SI units, known as the AGA equation [15],
The equation that relates the velocity to the pressure and
is as follows:
temperature is the continuity equation:
 " #0:5
Rgc 0:5 P21  P22  W
         
du 4m_ r r qz dP r r qz dT p Tb
¼ 2 2   þ . Q¼ Di2:5 , (18)
dL pr Di P z qP T dL T z qT P dL 8 Pb 28:97 fT ave zave gg L
(14) where W is the potential energy:
Eq. (14) requires partial derivations for z, the compressi-
P2ave
bility factor that is readily available from an EOS. W ¼ 0:0684gg DZ . (19)
T ave zave
3. Solution methods Since gas transmission lines are operated at velocities in the
fully turbulent-flow regime, the friction factor depends only
There are two generally accepted procedures to simulta- on the relative roughness. Thus, the AGA fully turbulent
neously solve the momentum, energy and mass balance equation, Eq. (20), uses the Nikuradse friction factor
equations: analytical or numerical methods. correlation and is the most frequently recommended and
widely used equation in high pressure, high flow rates for
3.1. Analytic method medium to large diameter pipelines:
Rgc 0:5
    
Engineers are always searching for simple design p Tb 3:706Di
Q¼ log
methods that can be used even without computers, i.e. 2 Pb 28:97 
#0:5
for quick hand calculations. This approach has resulted in
"
P2  P22  W
many useable simplified or integrated forms of Eqs. (1) and  1 Di2:5 . ð20Þ
gg LT ave zave
(11). This analytic approach results are quite useful because
 ARTICLE IN PRESS
994 F. Abdolahi et al. / International Journal of Mechanical Sciences 49 (2007) 989–1000

Table 5
Values of coefficients in Eq. (21) for different friction factor correlation

Friction factor equation a1 a2 a3 a4 a5

Blasius 17.1541 0.5714 0.4286 2.7143 0.1429

1 17.4983 0.5714 0.4286 2.7143 0.1429
7th power law
1 21.6136 0.5624 0.4376 2.6873 0.1249
8th power law
1 25.6034 0.5556 0.4444 2.6667 0.1111
9th power law
Modified 19th 24.8841 0.5556 0.4444 2.6667 0.1111
1 29.6754 0.5501 0.4499 2.6502 0.1001
10th power law
Panhandle 40.3120 0.5394 0.4606 2.6182 0.07881
Modified Panhandle 108.1291 0.5100 0.4900 2.5300 0.0200
IGT 24.6615 0.5556 0.4444 2.6667 0.1111
Weymouth 137.1902 0.5000 0.5000 2.6667 0.0000

Eq. (20) predicts both flow and pressure drop with a high where
degree of accuracy, provided high quality estimates of pipe
roughness are available. pDo U
j¼ . (23)
If one substitutes an explicit correlation for the Fanning mC
_ p
friction factor for smooth pipes into Eq. (15) and then
solves the equation for flow ðRe ¼ ð116Pb =pRT b Þ Schorre superimposed a linearly increasing Joule–Thom-
ðgQ=mDÞÞ, a simplified form of the general flow equation son effect, ZDP, on a gas temperature difference which was
is obtained: due only to heat transfer with the surroundings, the
a2 !a3 estimated gas temperature profile calculated by the explicit
Dai 4
  2
T b P1  eS P22 S 1 equation never reaches a steady-state value no matter how
Q ¼ a1 E , (21)
Pb T ave zave L eS  1 gg mag5 long the pipeline. To mitigate this shortcoming, Coulter
and Bardon [43] modified Schorre’s equation and proposed
where the values of ai corresponding to different friction Eq. (24) for the explicit calculation of the pipeline gas
factor correlations in SI units are given in Table 5. temperature profile:
Since, approximate representations of friction factor for
smooth pipes are embedded in Eq. (21), the efficiency       
Z dP
    
Z dP
factor, E, is introduced to account for pipe roughness. T¼ T0  Tg þ ejL þ T g þ ,
j dL j dL
By using the Panhandle friction factor in Eq. (15), Aziz
and Ouyang [40] have provided a plot for the efficiency (24)
factor, E. In addition to an efficiency factor that accounts
for relative roughness, it can, from a practical point of where j is defined as Eq. (23). This modified equation
view, be used to account for condensate, rust and sediment predicts a temperature profile that asymptotically approaches
accumulation. A research project involving four European a temperature slightly below that of its surroundings.
natural gas transmission companies has shown that even In order to include better the effect of elevation and
very small amounts of liquids can have a significant effect ground temperature change on the pipeline gas tempera-
on the pressure drop in the flow regime/region between ture profile, Alves et al. [44] developed the following
smooth and rough flows [41]. equation:

T ¼ ðT e0  ge L sin yÞ þ ðT 0  T e0 ÞeLj
3.1.2. Temperature profile calculations 1 dP f
By integrating Eq. (11) and making assumptions about þ ge A sin y½1  eLj  þ , ð25Þ
rC p dL Að1  eLj Þ
pipeline geometry, heat transfer with the surroundings and
the thermodynamic behavior of the flowing fluid, various where the dimensionless parameter, f, is defined as
investigators have provided explicit expressions for the
temperature of the fluid flowing through the pipe, i.e. a rZC p ðdP=dLÞ  rg sin y  ruðdu=dLÞ
temperature profile. Schorre [42] did the pioneering study f¼ . (26)
dP=dL
for temperature profile prediction in horizontal buried gas
pipelines that resulted in In Eq. (25) the surrounding temperature is a linear function
     of depth, so it can be used to predict the well bore
ZDP ZDP
T ¼ T0  Tg þ ejL þ T g þ  ZDP, temperature profile. This equation can be used for pipeline
jL jL temperature calculations by setting the surrounding
(22) temperature gradient, ge , to zero. The term dP=dL or
 ARTICLE IN PRESS
F. Abdolahi et al. / International Journal of Mechanical Sciences 49 (2007) 989–1000 995

Fig. 1. The algorithm used for numerical integration of momentum, energy and mass balance equations along the gas pipelines.

Table 6
Test data for gas pipelines of case study 1

Gas composition C 1 ¼ 88, C 2 ¼ 4:5, C 3 ¼ 1:48, C 4 ¼ 0:35, i  C 4 ¼ 0:24, C 5 ¼ 0:07, i  C 5 ¼ 0:11

(mole percent) Cþ6 ¼ 0:05, CO2 ¼ 0:2, N2 ¼ 5

Investigated parameter Di (m) Do (m)  ðmmÞ L (km) y hb (m) k ðW=m KÞ T g (K) T 1 (K) P1 (kPa) a

Roughness 0.9906 1.016 Variable 100 0 1.2 1.73 278 306 6757 1
Conductivity 0.9906 1.016 19.05 100 0 1.2 Variable 278 306 6757 1
a 0.9906 1.016 19.05 100 0 1.2 1.73 278 306 6757 Variable
ðdP=dlÞacc: and ðdP=dlÞfric: 0.9906 1.016 19.05 100 0 1.2 1.73 278 306 6757 1

DP=L in Eq. (25), the integrated or explicit form of the equations (Eq. (15) or (18)) and the kinetic term in Alves
energy balance equation, should be estimated/calculated et al. equation, Eq. (26), should be neglected for the
from the previous integrated form of momentum balance pipeline temperature profile.
 ARTICLE IN PRESS
996 F. Abdolahi et al. / International Journal of Mechanical Sciences 49 (2007) 989–1000

a b
7 290

Outlet temperature (K)

Outlet pressure (Mpa) 6
280
5 Q =10Mscm/D
Q =10Mscm/D
Q =30Mscm/D
Q =30Mscm/D
4 Q =50Mscm/D
Q =50Mscm/D 270
3

2 260
10 20 30 40 10 20 30 40

Pipe roughness (µm) Pipe roughness (mm)

Fig. 2. Outlet pressure (a) and temperature (b) variation versus pipe roughness.

a b
7 305
Q =10Mscm/D
300
Outlet Pressure (Mpa)

Q =30Mscm/D
Outlet temperature (K)
6
295 Q =50Mscm/D

5 290
Q =10Mscm/D
Q =30Mscm/D 285
4 Q =50Mscm/D 280
275
3
270
2 265
0 0.6 1.2 1.8 2.4 0 0.5 1 1.5 2 2.5
Soil thermal conductivity (W/m.K) Soil thermal conductivity (W/m.K)

Fig. 3. Outlet pressure (a) and temperature (b) variation versus soil thermal conductivity.

a b
7
280
Outlet temprature

6
Outlet pressure

276 Q =10Mscm/D Q =10Mscm/D

5
Q =30Mscm/D Q =30Mscm/D
272 Q =50Mscm/D Q =50Mscm/D
4

268
3
0.75 0.8 0.85 0.9 0.95 1
0.75 0.8 0.85 0.9 0.95 1
alpha
alpha

Fig. 4. Outlet pressure and temperature variation versus a correction factor.

3.2. Numerical method dP dT du

a21 þ a22 þ a23 ¼ c2 , (28)
dL dL dL
Prediction of the pressure, temperature and velocity profiles dP dT du
along gas pipeline requires the simultaneous solution of Eqs. (1), a31 þ a32 þ a33 ¼ c3 , (29)
dL dL dL
(11) and (14). These equations may be restated as follows:
where
     
dP dT du 1 qh 4m_ r r qz
a11 þ a12 þ a13 ¼ c1 , (27) a11 ¼ ; a21 ¼ ; a31 ¼  , (30)
dL dL dL r qP T pr2 D2i P z qP T
 ARTICLE IN PRESS
F. Abdolahi et al. / International Journal of Mechanical Sciences 49 (2007) 989–1000 997

     
qh 4m_ r r qz method for solution of simultaneous first order differential
a12 ¼ 0; a22 ¼ ; a32 ¼ 2 2 þ ,
qT P pr Di T z qT P equations was developed. A block flow diagram of the
developed procedure is shown in Fig. 1.
(31)
u u
a13 ¼ ; a23 ¼ ; a33 ¼ 1:0, (32)
agc agc 4. The effect of the pipe roughness, soil thermal conductivity
 2  and alpha correction factor
2fu g
c1 ¼  þ sin y ,
g D i gc To evaluate the effect the pipe roughness, soil thermal
0 c 1
conductivity and alpha correction factor the developed
2pkðT g  TÞ g procedure shown in Fig. 1 has been used to model a gas
c2 ¼ @  sin yA,
B C
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2 2 g c pipeline with the specifications given in Table 6. Fig. 2
m
_ lnðð2hb þ 4hb  Do ÞDo Þ
shows how the temperature and pressure vary with pipe
c3 ¼ 0. ð33Þ roughness that range from 12.5 to 38 mm and for gas flows
To solve Eqs. (27)–(29), Cramer’s method for the of 10, 30 and 50 Mscm/D. The results presented in Fig. 2
simultaneous solution of sets of linear algebraic equations clearly show that at high (430 Mscm=D) gas-flow rates,
combined with a fourth order Runge–Kutta integration the outlet pressure is very sensitive to the pipe roughness.
Fig. 3 shows the outlet temperature and pressure
variations versus soil thermal conductivities of
1.6 0.75–2.25 W/m K and for gas-flow rates of 10, 30 and
50 Mscm/D. The data presented in Fig. 3 do show that for
the different soil thermal conductivities the outlet pressure
Pressure Gradient (Pa/m)

1.2 3 Mscm/D (dP/dl)acc. and the temperature variations are very small and can be
3 Mscm/D (dP/dl)fric. neglected.
10 Mscm/D (dP/dl)acc. An additional parameter that is often set close to 1 in the
0.8 gas transmission pipeline calculations is the a correction
10 Mscm/D (dP/dl)fric.
factor. The data of Fig. 4 show that for a parameter range
30 Mscm/D (dP/dl)acc. of 0.75–1, the effect on the pressure and temperature
0.4
30 Mscm/D (dP/dl)fric. variation is negligible. Furthermore, in another study the
sensitivity of the pressure gradient to the acceleration and
0
fictional terms was modeled. The results of this study also
0 50 100
showed that the pressure gradient is essentially insensitive
to acceleration term when compared to the effect of the
Distance (km)
frictional term (Fig. 5). In fact as the gas-flow rate
Fig. 5. Contribution of acceleration and frictional terms in pressure increases, there is a rapid increase in the frictional term
gradient for various gas-flow rates. while the acceleration term undergoes very little change.

Table 7
Average errors of analytical methods when compared to field data

Pipeline Data Q (Mscm/D) Reynolds Outlet Predicted outlet Error

set number pressure (kPa) pressure (kPa) percent

Panhandle Mod. IGT Weymouth Panhandle Mod. IGT Weymouth

Panhandle Panhandle

Tabriz #1 12.5 1:19e þ 7 5976.7 6313.2 6318 6340.8 6246.4 5.6 5.7 6.1 4.5
Qazvin #1 24.2 1:84e þ 7 4134.4 4406 4364 4535 4146.1 6.5 5.5 9.7 0.3
#2 26.4 2:08e þ 7 3840.6 4532.9 4472.2 4674.2 4233.6 18 16.4 21.7 10.2
Kangiran #1 32.9 3:20e þ 7 3635.1 2729.8 1974.8 3469 negative 24.9 45.6 4.6 4100
#2 31.9 3:08e þ 7 4053.7 3268.3 2738.1 3860.6 negative 19.4 32.4 4.8 4100
Rasht #1 5.2 5:20e þ 6 5271.4 5392 5396.9 5394.1 5387.9 2.3 2.4 2.3 2.2
Kangan #1 55.7 2:69e þ 7 5783 6549.7 6513.2 6598 6471.8 13.3 12.6 14.1 11.9
#2 50 2:36e þ 7 6438 6985.5 6960.7 7021.4 6931.7 8.5 8.1 9.1 7.7
#3 51 2:44e þ 7 5989.2 6789.7 6763.5 6828.3 6731.8 13.4 12.9 14 12.4
#4 77.5 3:73e þ 7 5131.4 6798 6707 6894.5 6620.1 32.5 30.7 34.3 29
#5 70.5 3:40e þ 7 5423.1 6742.12 6671.1 6821.4 6600.8 24.3 23 25.8 21.7
#6 75.5 3:62e þ 7 5252.8 6828.3 6744.2 6919.3 6662.1 30 28.4 31.7 26.8
#7 62.5 3:02e þ 7 5618.2 6613.9 6562.8 6675.2 6509.8 17.7 16.8 18.8 15.9
#8 60.3 2:91e þ 7 5697.5 6600.1 6554.6 6656.6 6505.6 15.8 15 16.8 14.2

Average error % 16.6 18.3 15.3 425.5

 ARTICLE IN PRESS
998 F. Abdolahi et al. / International Journal of Mechanical Sciences 49 (2007) 989–1000

Table 8
Average error of the numerical method for outlet pressure and temperature predictions when compared to field data (ground conductivity ¼ 1 W/m K,
a ¼ 1)

Pipeline Data set Case 1 Case 2 Case 3

Pres. error (%) Temp. error (%) Pres. error (%) Temp. error (%) Pres. error (%) Temp. error (%)

Tabriz #1 6.05 4.2 0.0174 4.3 0.0174 4.3

Qazvin #1 0.0023 5.8 0.0013 5.8 0.0013 5.8
#2 8.56 6.6 0.0027 6.8 0.0012 6.8
Kangiran #1 7.62 2.7 0.0009 7.0 0.0030 7.0
#2 6.24 3.2 0.0022 5.4 0.0040 5.4
Rasht #1 2.96 2.9 0.0017 2.9 0.0017 2.9
Kangan #1 13.09 1.4 0.0200 2.3 0.0001 2.3
#2 7.93 4.8 0.0018 5.3 0.0061 5.3
#3 12.80 5.5 0.0042 6.3 0.0032 6.3
#4 33.70 4.9 4e  6 7.6 0.0001 7.6
#5 25.03 4.2 0.0013 6.1 0.0026 6.1
#6 30.98 3.4 0.0050 5.8 0.0001 5.8
#7 17.87 2.2 0.0021 3.5 0.0022 3.5
#8 15.86 1.3 0.0092 2.4 0.0012 2.4

Table 9
Optimized roughness and b factor (ground conductivity ¼ 1 W/m K, a ¼ 1)

Pipeline Data set Diameter (m) Reynolds number Case 2 Case 3

Roughness ðmmÞ b Roughness ðmmÞ b

Tabriz #1 0.990 1:19e þ 7 22.3 1.9595 22.34 1.9595

Qazvin #1 1.194 1:84e þ 7 22.2 0.9810 26.6 0.9573
#2 1.194 2:08e þ 7 31.1 1.1028 26.6 1.1276
Kangiran #1 0.867 3:20e þ 7 37.2 0.6546 31.5 0.6725
#2 0.867 3:08e þ 7 25.8 0.676 31.5 0.6551
Rasht #1 0.990 5:20e þ 6 26.2 3.3659 26.2 3.3659
Kangan #1 1.422 2:69e þ 7 28.0 3.5824 35.2 3.4645
#2 1.422 2:36e þ 7 25.5 3.4367 35.2 3.2860
#3 1.422 2:44e þ 7 33.0 4.2977 35.2 4.2596
#4 1.422 3:73e þ 7 40.0 3.8270 35.2 3.9016
#5 1.422 3:40e þ 7 38.6 3.7224 35.2 3.7739
#6 1.422 3:62e þ 7 40.6 3.7603 35.2 3.8437
#7 1.422 3:02e þ 7 40.6 3.5386 35.2 3.6157
#8 1.422 2:91e þ 7 35.4 3.5186 35.2 3.5200

A similar result has been previously reported by Ouyang ture and pressure were calculated using an average rough-
and Aziz [17]. In some cases where outlet pressure reaches ness of 19:05 mm, a practical value. Table 8 presents the
low values, the kinetic energy term will be effective. These results of this case study (labeled case 1) and shows the
conditions are, however, not likely to be encountered in resultant errors between the predictions and the field data.
practice. The match for estimated pressure profile is poor; however,
that for the temperature showed an acceptable match. The
5. Field application large mismatch in the pressure profile is due to the friction
factor correlation used in Eq. (1), while the mismatch in the
In order to evaluate the predictability of the four most temperature profile is mostly due to the equation used to
frequently used analytical methods, their predictions were estimate the heat transferred from pipe to the surrounding
compared for 14 data sets provided by the IGAT. Table 7 soil (based on Eq. (13)).
clearly shows that the analytical methods for pressure drop In order to improve the predictions for the temperature
prediction are similar with the IGT equation having the and pressure profiles, optimization of the parameters
smallest average error of 15.3%. affecting these profiles is required. The Levenberg–
In order to evaluate the performance of the developed Marquardt [45,46] method has been used to find the
numerical approach, shown in Fig. 1, the outlet tempera- optimum values of parameters that include pipe relative
 ARTICLE IN PRESS
F. Abdolahi et al. / International Journal of Mechanical Sciences 49 (2007) 989–1000 999

0.015 the field data sets, the effect of model parameters, e.g. soil
ε/D=2.11×10^-5 thermal conductivity, pipe relative roughness and velocity
profile correction factor, on the prediction errors was
evaluated. These model parameter sensitivity studies
showed that temperature and pressure profile at high
New friction factor

0.01
Reynolds numbers are very sensitive to the Fanning
friction factor and that the soil thermal conductivity and
velocity profile correction factor have a negligible effect on
the temperature and pressure profiles, provided their values
0.005
are set to their commonly accepted industry values. Since
the Fanning friction factor has a significant influence on
the quality of the pressure profile prediction, an optimiza-
tion method has been used to tune friction factor
0 correlation as a function of Reynolds number at an
2 3 3 4 4 industry accepted average pipe roughness.
Reynolds number ×10^-7

Fig. 6. Variation of new friction factor versus Reynolds number for Acknowledgment
Kangan pipeline.
The authors thank the National Iranian Gas Company
(NIGC) for their assistance and Mr. Torbati and
Mr. Farahzadi at the gas dispatcher center for their
roughness, soil thermal conductivity and velocity profile
valuable help in this project.
correction factor. However, this global approach to
parameter optimization did not significantly improve the
profile predictions. The next approach (labeled case 2 in References
Tables 8 and 9) was rather pragmatic in that the soil
thermal conductivity and velocity profile correction factor [1] White FM. Fluid mechanics. New York, NY: McGraw-Hill Book
were set to 1 W/m K and 1, respectively, and the data was Company; 1986. p. 732.
fit using only the pipe roughness and friction factor as [2] Aziz K. Ways to calculate gas flow and static head. Petroleum
Engineer 34–5 (Series from November 1962–September 1963); 1947.
fitting parameters. This pragmatic approach required the [3] Starling KE, Savage JL. Compressibility factor of natural gas and
pipe roughness changes for the same pipeline, which did other related hydrocarbon gases. AGA Transmission Measurement
not make the result useable. This line of reasoning did Committee Report No. 8; 1994.
provide some improvement, hence, in the next study [4] Peng DY, Robinson DB. A new two constants equation of state.
Industrial and Engineering Chemistry Fundamentals 1976;15:59–64.
(labeled case 3 in Tables 8 and 9), the pipe roughness
[5] Lucas K. Berechnungsmethoden fur stoffeigenschaften. Verein
was set at an average value of case 2. This case study Deutschen Ingenieure, Abschnitt DA; 1984.
resulted in new values for the Fanning friction factor, and [6] Dean DE, Stiel LI. The viscosity of nonpolar gas mixtures at
did show that in order to decrease the error in estimated moderate and high pressures. AIChE Journal 1965;11(3):526–32.
pressure profile, the Fanning friction factor had to be [7] Moody LF. Friction factors for pipe flow. Transactions of ASME
multiplied by a factor (bÞ, ranging from 0.67 to 4.3, that 1944;66(8):671–84.
[8] Schlichting H. Boundary layer theory. New York: McGraw-Hill
was dependent on the pipeline condition. The values of Book Company; 1960.
correction factor a for both cases 2 and 3 are shown in [9] Prandtl L. The mechanics of viscous fluids. In: Durand WF, editor.
Table 9. Fig. 6 presents a plot of the modified Fanning Aerodynamic theory, vol. III (Division G); 1935. p. 34–208.
friction factor at various Reynolds numbers at a constant [10] Colebrook CF. Turbulent flow in pipes, with particular reference to
relative roughness. The data of Fig. 6 show that modified the transition region between the smooth and rough pipe laws.
Journal of the Institution of Civil Engineers London 1939;12:133–56.
Fanning friction factor approaches a constant value of [11] Blasius H. Das Ahnlichkeitsgesetz bei Reibungsvorgangen. Physik
0.009 at high Reynolds numbers ð44  107 Þ. Zeithchr 1911;XII:1175–7.
[12] Weymouth TR. Problems in natural gas engineering. Transactions of
6. Conclusion ASME 1913;34:185–231.
[13] Ward-Smith AJ. Internal fluid flow. Oxford: Clarendon Press; 1980.
p. 566.
The common analytical methods used to predict the [14] Knudsen JG, Katz DL. Fluid dynamics and heat transfer. New York,
temperature and pressure profiles for gas transmission NY: McGraw-Hill Book Company; 1958. p. 576.
pipelines have been reviewed and a numerical framework [15] Institute of gas Technology. Steady flows in gas pipelines. IGT
to calculate these profiles has been developed. The Report No. 10, PRC Project NB-13, American Gas Association Inc.,
analytical methods as well as the developed numerical New York, NY; 1965.
[16] Beggs HD. Gas production operations. Houston, Texas: Gulf
framework have been used to predict the temperature and Publishing Co; 1984. p. 287.
pressure profiles for Iranian Gas Trunk-lines (IGAT). [17] Ouyang LB, Aziz K. Steady-state gas flow in pipes. Journal of
Since the predicted profiles resulted in a poor prediction of Petroleum Science and Engineering 1996;14:137–58.
 ARTICLE IN PRESS
1000 F. Abdolahi et al. / International Journal of Mechanical Sciences 49 (2007) 989–1000

[18] Zagarola MV. Mean-flow scaling of turbulent pipe flow. PhD thesis, [33] Romeo E, et al. Improved explicit equations for estimation of the
Princeton University, USA; 1996. friction factor in rough and smooth pipes. Chemical Engineering
[19] Colebrook CF, White CM. Experiments with fluid friction in Journal 2002;86:369–74.
roughened pipes. Proceedings of the Royal Society of London 1937; [34] Nikuradse J. Stromangsgesetze in rauhen rohren. Ver. Dtsch. Ing.
161A:367–81. Forschungsheft 361, vol. B, Berlin: VDI Verlag; 1933.
[20] Moody ML. An approximate formula for pipe friction factors. [35] Sletfjerding E, Gudmundsson JS. Friction factor directly from
Transactions of ASME 69: 1005–11. roughness measurements. Journal of Energy Research and Technology
[21] Wood DJ. An explicit friction factor relationship. Civil Engineering 2003;125:126–30.
1966;36(12):60–1. [36] Golshan H, Narsing M. Study of pipeline deterioration due to age
[22] Eck B. Technische Stromungslehre. New York, NY: Springer; 1973. (Phase 1). Design methods and technology/facilities planning. Nova
p. 142. Gas Transmission Limited internal reports, Calgary, Alta., Canada;
[23] Jain AK. Accurate explicit equation for friction factor. ASCE 1994.
Hydraulic Division Journal 1976;102(HY5):674–7. [37] Kesler M, Lee BI. A generalized thermodynamic correlation based on
[24] Swamee PK, Jain AK. Explicit equations for pipe flow problems. three-parameter corresponding states. AIChE Journal 1975;21(3):
Journal of Hydraulic Engineering ASCE 1976;102(5):657–64. 510–27.
[25] Churchill SW. Friction-factor equation spans all fluid flow regimes. [38] Plocker U, et al. Calculation of high-pressure vapor–liquid equilibria
Chemical Engineering 1977;84(24):91–2. from corresponding-states correlation with emphasis on asymmetric
[26] Chen NH. An explicit equation for friction factor in pipe. Industrial mixtures. Industrial and Engineering Chemistry Process Design and
and Engineering Chemistry Fundamentals 1979;18(3):296–7. Development 1978;17:324–32.
[27] Round GF. An explicit approximation for the friction factor- [39] Neher JH. The temperature rise of buried cables and pipes.
Reynolds number relation for rough and smooth pipes. Canadian Transactions of AIEE 1949;68(1):9–21.
Journal of Chemical Engineering 1980;58(1):122–3. [40] Aziz K, Ouyang LB. Simplified equation predicts gas flow rate,
[28] Barr DIH. Solutions of the Colebrook-White function for resistance pressure drop. Oil & Gas Journal 1995;8:70–1.
to uniform turbulent flow. Proceedings of the Institution of Civil [41] Gersten K, et al. New transmission-factor formula proposed for gas
Engineers 1981;71(2):529–36. pipelines. Oil & Gas Journal 2000;14:58–62.
[29] Haaland SE. Simple and explicit formula for the friction factor in [42] Schorre CE. Here’s how to calculate flow temperature in a gas
turbulent pipe flow including natural gas pipelines. IFAG B-131, Div pipeline. Oil & Gas Journal 1954;7:66–8.
Aero and Gas Dynamics, The Norwegian Institution of Technology; [43] Coulter TL, Bardon MF. Revised equation improves flowing gas
1981. temperature prediction. Oil & Gas Journal 1979;26:107–8.
[30] Zigrang DJ, Sylvester ND. Explicit approximations to the solution of [44] Alves IN, et al. A unified model for predicting flowing temperature
Colebrook’s friction factor equation. AIChE Journal 1982;28(3): distribution in wellbores and pipelines. SPE Journal 1992;20632:
514–5. 363–7.
[31] Serghides TK. Estimate friction factor accurately. Chemical [45] Levenberg K. A method for the solution of certain problems in least
Engineering 1984;91(5):63–4. squares. SIAM Journal on Numerical Analysis 1944;16:588–604.
[32] Manadillli G. Replace implicit equations with signomial functions. [46] Marquardt D. An algorithm for least squares estimation of nonlinear
Chemical Engineering 1997;104(8):129–30. parameters. SIAM Journal on Applied Mathematics 1963;11:431–41.