Models Methods Simulating Gas Pipeline Blowdown (1989) - KK Botros Et Al
Models Methods Simulating Gas Pipeline Blowdown (1989) - KK Botros Et Al
Models Methods Simulating Gas Pipeline Blowdown (1989) - KK Botros Et Al
-
Nova Husky Research Corp., 2928 16 Street N.E., Calgary, Alberta, T2E 7K7
Estimation of gas pipeline blowdown time requires the determination of the pressure-time history for the unsteady
dischargeof gas through a blowdown stack open to the atmosphere. Computationalprocedures entail various assumptions
whose validity and contributions to the accuracy are often not assessed. This paper discusses the computational models
and solution methods and further assesses the significance of the various assumptions involved.
Volume and pipe models, numerical and analytical methods of solution, effects of stack entrance and friction losses and
discharge coefficient are evaluated. The accuracy of a particular model or method of solution is greatly dependent on the
jZ/Dratio of the pipe section under blowdown. Comparison with field measurements of a straight pipe section and a
complicated compressor station yard piping enabled evaluation of the above models.
Afin d’estimer le temps de chasse d’une conduite de gaz, il est necessaire de determiner la relation pression-temps pour
la dtcharge en regime instationnairedu gaz dans une cheminte de vide-vite ouverte sur I’atmosphtre. Les mtthodes de
calcul par ordinateur foumissent des hypotheses Cmises.
Les modtles de volume et de conduite, les mkthodes de resolution numkriques et analytiques, les effets de I’entrk de
la cheminee et des pertes par friction ainsi que le coefficientde dkharge sont tvalu6s. L’exactituded’un modtle particulier
ou d’une mtthode de rtsolution depend fortement du rapportfllD de la section de conduite sous vide-vite. Les modkles
ci-dessus ont pu &tretvaluts par comparaison avec lesmesures de champ d’une section de conduite droite et d’une
tuyauterie d’une station complexe de compresseurs.
Keywords: gas pipeline blowdown, pipeline transient, unsteady flow, blowdown simulation.
THE CANADIAN JOURNAL OF CHEMICAL ENGINEERING, VOLUME 67, AUGUST, 1989 529
Analytical Volume Solution
For perfect gas and T = constant, Equation (2) yields
dp = dPIRT ................................ (3)
and thus Equation (1) can be transformed into
GRTdr = -V . dP ........................... (4)
For sonic flow at the valve throat and assuming isentropic
expansion from the pipe stagnation conditions, the mass flow
rate (Shapiro, 1953) is:
~1 k
1
2 P . Arb . cd
........ ( 5 )
ence gas paramenters and numerically taking into account the
real gas properties at each time step during blowdown. Ef-
where the discharge coefficient Cd depends on the type of the
fects of friction losses in the riser upstream of the throat were
valve.
also considered in the numerical methods. Calculated blow-
The time constant zv is defined as:
down times were compared with those obtained using graphs
of Gradle (1984) and own field measurements of a straight k+ 1
~-
pipe section and a compressor station yard piping.
Gas expansion in the main pipe generally is regarded as an
isothermal process because the shortest blowdown lasts at
least a couple of minutes during which heat transfer to the gas
occurs. Thermal capacitance of the pipe and surrounding soil where c is the speed of sound (c = -\lkRT). Introducing
secures sufficient heat flow to prevent much change in the gas Equations ( 5 ) and (6) in the original Equation (4) yields the
temperature. To assess this assumption the gas temperature normalized differential equation:
was measured and calculated during the blowdown of a
compressor station piping. The lowest temperature measured di, = -dP/P ............................... (7)
was slightly below freezing and the one calculated with a
steady heat transfer model was -12°C. Nevertheless, the where the normalized pressure P = PIPa, Pa is the ambient
pressure-time profile calculated for the latter case was very pressure and time7 = t ITV.
close to that calculated for the isothermal process. In reality, After integration from initial pressure Pi to the pressure Pc,
the lowest temperature most probably was below the mea- which terminates sonic flow, one obtains the time of sonic
sured one due to some thermal inertia of the thermometer venting as:
housing but above the calculated temperature due to the -
enhanced heat transfer with a transient temperature distribu- tc = lnPi - lnPc ............................. (8)
tion in the soil. If contents of the heaviest gas components
are not exceeding the specification gas requirements, conden- or taking into account that
sation should not occur above -10°C and therefore it seems
that typically there is no need to take it into account. During
the isentropic gas expansion at the throat the gas temperature
is lowered considerably, but due to the relaxation time con- L A
densation should not affect the process of expansion and thus
the flow rate. then, the dimensionless time for the sonic venting can be
written as;
Volume Model
ph ph
k Pc 1.1 1.01 1 .oo1 1.1 1.01 1.001
Introducing Equation (1 2) into Equation (4)and following are determined from the state equation. The throat pressure
the previous transformations we obtain the differential equa- and temperature (Pth, Tth) are related to the pressure and
tions: temperature of the gas in the pipe volume (P,T ) and can be
calculated from the state equation using an iterative proce-
dure for isentropic expansion to sonic conditions. Addition-
+
dT = +[ 2k-’ ( k - 1)’ / ( k
(F-2lk - F-(k+l)/k
dP
. . . . . . . . . . (13) ally, the thermodynamic relation:
ap + 3 = 0 . . .....................
An = 4 sin an1(2% + sin 201n) ............... (35)
(1 I RT) - .(25)
at ax Table 1 of Appendix IV in Carslaw and Jaegex (1959)
and momentum: indicates that for aEL < 0.1 the second root a2 of Equation
(35) is greater than ten times the first root a1 and thus the
ap series is rapidly converging. In this rangeA1 is also very close
ax + a Q = 0 ..............................
- (26) to unity (A1 = 1.016for aEL = 0.1). With decreasingaEL, the
coefficient A1 tends to unity and the higher An coefficients
DifferentiatingEquation (26) with respect to x and subtract- tend to zero. According to Equation (33) at x = 0 and t = 0
ing from Equation (25) yields the pressure ratio is determined by the sum of the An coeffi-
cients only. Since the initial pressure ratio should be equal to
unity, the error associated with taking the first term only is
equal to 1.6%. With growing time t, convergence improves
because the second and subsequent terms are more sup-
pressed than the first one. Therefore for the calculationof the
where pressure drop with time during blowdown the first term in
Equation (33) is sufficient and hence at x = 0,
RT
K =-
a
................................... .(28)
P / Pi = exp(-Ka!t I L2) ...................... (36)
The partial differentialequation,Equation (27),describing Normalizing time t by the constant zp, where
a slowly varying flow in the main pipe, is linear, parabolic
and of the second order and is known as the one-dimensional ...............................
heat conduction equation (Carslaw and Jaeger, 1959). This 2p = L2/KCXf (37)
equation can be solved once the boundary and initial condi- an exponential expression identical to Equation (1 1) is ob-
tions for the pipeline are specified. At the closed end, x = 0 tained:
(Figure 1). Q = 0 prevails at all times and thus from Equation
(26), aP/ax = 0. At the stack end, x = L, and during sonic P I Pi = exp(-$ ............................. (38)
venting the gas flux is proportional to the pipe pressure, i.e.
Q = EP [see Equation (5)] where E is a constant. At the but containing a different normalization time constant 2.,
beginning of blowdown the pressure in the entire pipeline is Note that from Equations (6), (28). (32) and (37),
equal to Pi, thereby setting the initial conditions. limzp = zv, and that for a > 0,zp> zv, indicatingthat friction
a40
532 THE CANADIAN JOURNAL OF CHEMICALENGINEERING,VOLUME 67, AUGUST, 1989
TABLE2 1.o
Gas Composition
Component Mole %
Methane 91.63
Ethane 4.55 .
fi-
n-
0.1
Propane 1.12
i-Butane 0.16
n-Butane 0.2 1
i-Pentane 0.05
n-Pentane 0.07
0.01
Nitrogen 1.78 0 1 2 3 4 5
Carbon-dioxide 0.43 -
t = t17"
Figure 2 - Pressure-time profiles from analytical and numerical
TABLE3 volume models (without stack losses).
Gas Properties @ 1O'C
(from BWRS with C,, data from Aly & Lee, 1981) Thus, for each time step, At, a system of 2n (where n is the
number of nodes dividing the pipe section) non-linear alge-
Pressure Sound Speed k braic equations are formulated from the implicit finite differ-
[kPal [m/sl ence scheme. These equations are solved simultaneously
using Newton's method. A distinct advantage of this scheme
is that the solution is generally stable for all choices of At and
SO00 399.33 1.4600
Av and it is relatively simple to implement. In addition, the
6OOo 395.95 1.3679 formulation of a straight section of a pipeline results, natu-
4000 398.96 1.3176 rally, in a block-mdiagonal system of equations for which the
Thomas algorithm described in Rosenberg (1969) can be
2000 406.42 1.2966 applied quite readily. Note that the blocks in this tridiagonal
loo0 411.29 1.2936 system are 2x2 since the number of dependent variables is
100 416.14 1.2941 two. In the case of complicated piping networks, this block
triagonal property is no longer in effect, and a technique based
on a sparse variant of Gaussian elimination for sparse systems
attenuation in the main pipe tends to increase the blowdown of equations is used (Duff, 1980).
time.
The normalized time of sonic venting can be calculated Example Of Pressure-TimeComputations
introducing Equation (lo), and hence the dimensional time rc
from Computations for the above models and solution methods
- were performed on an example blowdown of a straight sec-
fc = rc . T, ................................ (39) tion of pipe through one riser at the end (see Figure 1). The
cross-sectional areas involved in this example are:
Numerical Pipe Solution Ap = 0.8623 m2, AS = 0.0729 m2 and Ath = 0.0498 m2. The
gas composition used is given in Table 2 and the correspond-
The same one-dimensional unsteady gas flow equations ing gas properties at 10°C are given in Table 3. The results
are used here, again neglecting the inertial terms in the are presented in a normalized form for general application.
momentum equation, i.e.
Figure 2 shows a comparison between the two methods of
solution for the volume model, namely the analytical and
mumerical methods for three differenct initial1 pressures, Pi
= 4000, 6000 and 8000 kPa. The time is normalized with
respect to zV[see Equation (6)]. The reference gas properties
ap +felel = 0 . . . . . . . . . . . . . . . . . . (41)
momentum: - in zv were taken as the average values during the blowdown
ax 2pD
and these are: c = 404 m/s for all Pi cases and k = 1.300, 1.314
combined with the state equation [Equation (2)] and the and 1.350 corresponding to Pi = 4000,6000, and 8000 Wa,
assumptions of isothermal flow along the pipe [Equation respectively [see Table 31. As a result, the analytical solution
(l8)J and isentropic expansion through the blowdown stack. gives only one profile of the dimensionless pressure up to the
This yields a set of non-linear, first-order parabolic partial end of sonic discharge independent of the initial pressure.
differential equations for the dependent variables P and Q. The termination points indicating the end of sonic venting are
An Euler implicit finite difference scheme has been used to different, naturally, for the different initial pressures. Three
obtain the solution. This scheme is known to be first-order different profiles corresponding to the three initial pressures
accurate in time and second-order accurate in space (Ander- resulting from the numerical solution are also shown in the
son et al., 1984). A typical two-dimensional grid is used with same figure. This figure illustrates clearly the effects of
the previously noted boundary conditions: (1) zero flux at accounting for real gas behavior and the variations of the gas
x = 0; and (2) a flux at x = L which depends on the stack throat properties during blowdown. Clearly, the discrepancy be-
area, Cd coefficient, and the line pressure via real gas tween the analytical and the numerical profiles increases as
isentropic expansion to sonic and then subsonic discharge. the initial line pressure increases. For example, a t t = 3.0, the
THE CANADIAN JOURNAL OF CHEMICAL ENGINEERING, VOLUME 67. AUGUST, 1989 533
1.0 ~
1.0 h
il
LEGEND
ANALYTICALPIPE MODEL NUMERICAL UPE MODEL LEGEND
a - 4000,6000,8000 kPa b- 4000 kPa a - Analytical Volume Model
c- 6OOOkPa b - NumericalVolume Model
d- 8000 kPa c- Analytical P i p Model
d- Numerical P i p Model
0 . Indicates end of sonic venling .Indicates end of sonic vensng
.
n-
d
0.1 .
d
d
0.1
0.01 0.01
0 1 2 3 4 5 0 2 4 6 0' 10
- -
t = f/rp t t/rg
Figure 3 - Pressure-timeprofiles from analytical and numerical pipe Figure 5 - Pressure-time profile comparison for P , = 6000 kPa
models (without stack losses). (without stack losses).
1.0 1 effects of both the real gas behavior and the mathematical
LEGEND
a - Analytical Pipe Model
difference in the methods of solution. In order to quantify the
b - Numerical Pipe Model (ideal gas) relative contribution of these effects, the numerical computa-
c - Numerical Pipe Model (real gas)
tion was run twice for Pi = 6000 kPa using the perfect gas
- End of sonic venting
equation of state and the real gas equations, respectively. The
results are compared with the analytical solution in Figure 4.
Finally, and in order to give an overall comparison of the
two models (volume vs. pipe) and the corresponding methods
of solution (analytical vs. numerical), the results of Figures 2
and 3 were plotted in Figure 5. Only the initial pressure case
of 6000 kPa is presented and the time is normalized according
to the following general normalization:
-
t = t/z,. .................................. (43)
.
d
d
0.1 7
:
lndicales end of Wnlc venung
.
t
n'
i
0.1 -
E
0 .lndlcales end 01 s n l c venUng
I I t I I I
0.01 0.01
0 2 4 6 8 10 0 2 4 6 8 10
Figure 6 - Pressure-time profiles from numerical pipe model with Figure 7 - Pressure-time profiles from numerical pipe model with
stack losses (time nomalized w.r.t. T~). stack losses (time normalized w.r.t. ( zg Pol /Pf,,h)).
imply that for a typical stack characterized in the Introduction where 5 is the entrance loss coefficient (5 = 1h for sharp-
the processes occurring downstream of the throat can only edged entrance), and Poi and P02 are stagnation pressures at
change the blowdown history insignificantly. Therefore, the sections 1 and 2, respectively (see Figure 1). Again, as a good
throat was still regarded as an exit to the atmospheric pres- approximation, the following expressions can be used for the
sure. Most probably the errors in determination of the en- stagnation pressures:
trance loss, friction factor along Lsi and valve discharge
coefficient will have a much stronger effect on the total
blowdown time than this simplifying assumption.
Additionally, since the stack is normally short the frictional .(49)
flow along it can be assumed adiabatic. The energy equation
for unsteady compressible gas flow in one-dimensional con-
stant area ducts (streamwise viscous and conductive effects
being neglected) can be written as follows (Issa and Spalding,
1972);
Po2 = P2 1 + k2 [ ~ 2 '( ~ G2
p2A2c2
I]= k?
. . . . . . . . . .(50)
A[
at e + 13 $(.+; I);
+ + = q . . . . . . . (451
where the values for ki,k2, ci and c2 are evaluated for the real
gas from the state equation at the conditions of sections 1 and
2, respectively.
Combining the above formulation with the numerical so-
where q represents the rate of heat transfer per unit mass of
lution for the main pipe model, the blowdown pressure-time
the gas and is identically zero for adiabatic flow. Introducing
profiles were obtained for the stack geometry shown in Figure
the continuity equation, Equation (22), and the full momen-
1. Figure 6 shows the effects of accounting for stack friction
tum equation, Equation (23), the above energy equation after
losses and for the entrance losses (5 = OS), separately and
some manipulation becomes (Picard and Bishnoi, 1988):
combined. Notice that these losses, particularly the entrance
losses, increase the blowdown time. Profiles (1) through (4)
of Figure 6 are for C d = 1.0. In this particular example,
accounting for friction along the stack increased the blow-
down time by 1.5%, while sharp-edged entrance losses in-
creased it by 6.5% and the combined losses by 8%.
This energy equation, Equation (46), along with the conti- approximately. Profile 5, however, shows the effects of both
nuity equation, Equation (22), and the original momentum losses combined but for Cd= 0.95. It is interesting to see that
equation, Equation (23), constitute a set of non-linear hyper- the dimensionless blowdown time, as defined in Equation
bolic partial differential equations. This was solved by the (43). for c d = 0.95 is less than that for c d = 1.o (profile 4).
general implicit finite difference scheme of Beam and Warm- This is because the discharge flow rate is lower for Cd = 0.95
ing (1 976). which in turn reduces the effects of the stack friction and
entrance losses.
As for the entrance losses at the stack inlet, a quasi-steady
flow can be assumed at each time step of the numerical In addition, if the time is normalized in a way to allow for
solution. The following equations can then be applied: these losses, that is
THE CANADIAN JOURNAL OF CHEMICAL ENGINEERING, VOLUME 67, AUGUST, 1989 535
TABLE 4
Comparison of Results
L=lkm I
L=20km
4000 6000 8000 I 4000 6000 8000
Time (sec) tc tb
Method I
Volume Model
- analyitical 234 292 264 322 286 343 4687 5834 5281 6440 5710 6897
- numerical 237 296 271 330 296 356 4731 5920 5419 6609 5933 7121
Pipe Model
- numerical 227 270 254 297 271 315 4661 5542 5226 6111 5575 6470
- analytical 240 298 272 330 292 351 4910 6170 5570 6850 5970 7290
- numerical with
stack losses 247 321 284 352 312 380 5072 6631 5841 7400 6412 7971
- charts in Gradle I
(1984) without stack
losses - 320 - 340 - 370 - 5800 - 6300 - 6800
- charts in Gradle
(1984) with stack
losses - 341 - 362 - 394 - 6180 - 6720 - 7250
however, deviate from each other as the discharge flow is not Generally, the above-mentioned discrepancies between the
directly proportional to the stagnation pressure during the analytical and numerical solutions in the two models depend
subsonic phase. on the values of c, k and the attenuation factora used with the
analytical solutions. These parameters are determined with
Comparison Of Results some tolerance and thus higher or lower discrepancies are
possible. In addition, these discrepancies depend on two
In order to assess the effects of the models and methods of main parameters, namely: fL/D and Ap/(Arh Cd). The
solution on blowdown time some computations were per- discharge coefficient Cd depends primarily on the flow con-
formed. The results are tabulated in Table 4 which also traction coefficient resulting from valve geometry and is
contains respective results using the charts in Gradle (1984). normally provided by the valve manufacturer. Figure 8
Geometry and computation parameters used here are those shows clearly the effects of these two parameters on the sonic
given in the previous examples. The stack length Ls was blowdown time for example. The results are presented in
taken as 3 m for friction loss calculations, while Cd was terms of the ratio of the sonic blowdown time from the
assumed unity. The time of sonic venting, tc, and total different models to that obtained from the numerical pipe
blowdown time, tb, to a line pressure 1% above atmospheric model. It is demonstrated in these figures that the discrep-
pressure were calculated for three values of Pl and pipe ancy increases asfLID increases or Ap/(Ath Cd) decreases.
lengths of 1 km and 20 km. The results are tabulated in Table
4. Comparison With Field Measurements
Examination of Table 4 indicates that the real gas proper-
ties (in the numerical solution) increase tc and tb by 1-4% in Field measurements were taken during blowdown of a gas
the volume model. In the pipe model the dicrepancy between pipeling section in NOVA CORPORATION O F
analytical and numerical solutions increases to 5 7 % for tc. ALBERTA’S gas transmission system. Particulars of this
For tb calculations, however, the numerical pipe solution section are:
gives values 9- 1 1% higher than those predicted with the Dp = 0.203 m Pi = 4089 kPa
analytical method. This is because the sonic boundary con- Ds= 0.097 m Ti = 302 K
ditions were extended for the subsonic region in the analytical Arh = 0.00548 m2 5 = 0.2
model to allow for integration of the equation as mentioned LSi = 1.3 m Gas Composition (Table 2)
before. Ls2 = 0.7 k= 1.298 at P = 2000 kPa,T = 302 K
Table 4 also shows the effects of stack entrance and friction L = 25523 m c = 420 m/s at P = 2000 kPa,
losses. The entrance losses were taken with the maximum E, T = 302 K
value equal to 0.5 corresponding to a shapedged entrance. Figure 9 shows the measured P-tprofile in the main pipe
With this value of 5, stack losses tend to increase both tf and at section 1 (see Figure 1). Since the diameter of the main
tb by a maximum of 6-8%. pipe section is small in this example, it was appropriate to first
Due to the poor resolution of the charts in Gradle (1984), match the measured P-tprofile with that obtained from the
the evaluated tb is rather inaccurate but generally is about numerical pipe model in order to evaluate the missing value
3-7% longer for the short pipeline and 3-6% shorter for the of the discharge coefficient. This Cd was found to be 0.75
longer pipeline as compared to the numerical pipe model. which gave a good agreement between the measured profile
The charts also give a consistent 6.5% increase in th due to and that obtained from the numerical pipe model. This value
stack losses. of Cd seems quite reasonable for a plug valve when compared
536 THE CANADIAN JOURNAL,OF CHEMICAL ENGINEERING, VOLUME 67, AUGUST. 1989
fE
F
g
1
1.4
1.2
l.O
0.8
0.6
0.4
. - =
Analytical Plpe Model
-
n-
n‘
1.o
0.1
0’ I 0.01
00
1 10 100 loo0 loo00 0
THE CANADIAN JOURNAL OF CHEMICAL ENGINEERING, VOLUME 67, AUGUST, 1989 531
12'
(4.SSm)
.
10"
I I
7' OriflCC
UNIT #1
UNIT #Z I I /
30'
.
(10.SSm)
(4.W
(U).wm)
24' z4'
Figure 10 - Compressor station yard piping configuration (numbers in brackets indicate lengths in meters).
(is.wm)
90' ,,
a'
,
I
Orifbe
=
'X' SD
1.o
series solution of the analytical pipe model in order to better
predict the P-t profile.
Finally, it should be emphasized that the precision in
c - Numerical Pipe Model
estimating the blowdown time greatly depends on the accu-
a ~ Field Measurements
racy of the input data used in the simulation particularly the
pipe and riser geometry, friction loss and discharge coeffi-
d
-- cients, and real gas properties.
--
0.1
d
Acknowledgement
The work presented here is part of a research project sponsored
0.01
t by NOVA CORPORATION OF ALBERTA and the permission to
publish it is hereby acknowledged. The authors also acknowledge
the valuable comments made by one of the reviewers.
0 50 100 150 2w
TIME (seconds) Nomenclature
11 = pipe attenuation factor
Figure 1 1 - Pressure-time profile comparison with field measUrementS A = pipe cross-sectionalarea
during blowdown of the compressor station shown in Figure 10. Ai, Bi = constants, i = 1,2, ....n
C = speed of sound
numerical pipe model tend to become inaccurate. However, cd = stack valve discharge coefficient
for relatively lowfl/D values, all models provide reasonable CP, = specific heat at constant pressure (at normal pressure
predictions and therefore the simple analytical volume calcu- and temperature conditions)
C" = specific heat at constant volume
lations can be used effectively.
D = pipe inside diameter
A case study (withfl/D = 1,685)involving field measure- e = internal energy per unit mass
ments showed that both the analytical and numerical volume E = constant of dimension [QP]
models fall short in predicting the P-t profile and the blow- f = Moody friction factor
down time. The same conclusion can be drawn from another f = function of, i.e.fl.....)
G = gas mass flow rate
case study involving field measurements of blowdown of a k = gas isentropic exponent
complicated network of a three-unit compressor station yard L = total pipe length
piping. The relatively large value of fz/D for the first case LS = total stack length
study necessitated taking more than just the first term in the Ls1 = stack length from entrance to valve
538 THE CANADIAN JOURNAL OF CHEMICAL ENGINEERING, VOLUME 67, AUGUST, 1989
LS2 = stack length from valve to exit Beam, R.M. and R.F. Warming, “An Implicit Finite-Difference
M = Mach number Algorithm for Hyperbolic Systems in Conservation Law Form”,
N = value of an integral J. Comp. Phys. 22,87-110 (1976).
P = static pressure Carslaw, H.S. and J.C. Jaeger, “ Conduction of Heat in Solids”,
P = dimensionless pressure Second Edition, Oxford Press, London (1959).
Poi = stagnation pressure at section i Cronje, J.S., P.R. Bishnoi and W.Y. Svrcek, “The Application of the
4 = rate of heat transfer per unit mass of the gas Characteristic Method to Shock Tube Data that Simulate a Gas
Q = gas flux; Q = pu Pipeline Rupture”, Can. J. Chem. Eng. 58 289-294 (1980).
R = gas constant Duff, I.S., “MA28 - A Set of FORTRAN Subroutines for Sparse
I
- =time Unsymmetric Linear Equations”, Computer Science and Systems
t = dimensionless time = I/T Division AERE Harwell, R.8730,Oxfordshire, England ( 1980).
T = absolute temperature Flatt, R., “Unsteady Compressible Flow in Long Pipelines following
U = gas mean velocity in the x-direction a Rupture”, Int. Journal for Numerical Methods in Fluids 6,
V =volume 83-100 (1986)
X = axial distance Cradle, R.J., “Design of Gas Pipeline Blowdowns”, Energy Process-
ing Canada, 15-20, January - February (1984).
Greek letters Groves, T.K., P.R. Bishnoi and J.M.E. Wallbridge, “Decompression
c1 = dimensionless constant Wave Velocities in Natural Gases in Pipe Lines”, Can. J. Chem.
Y =specific heat ratio Eng. 56,664-668 (1978).
A = variable increment Harwell Subroutine Library, DC03 Routine, Computer Science and
K =constant defined by Equation (28) Systems Division of AERE, Harwell Laboratory, Oxfordshire,
A = constant with dimension [L].’ England (1988).
5 = pressure loss coefficient Issa, R.I. and D.B. Spalding, “Unsteady One-Dimensional Com-
p = gas density pressible Frictional Flow with Heat Transfer”, J. Mech. Eng. Sci.
‘5 = time constant 14 (6), 365-369 (1972).
Jungowski, W.M., “Investigation of Flow Pattern Boundary Condi-
Subscript tions and Oscillation Mechanism in a Compressible Flow
Through Sudden Enlargement of a Duct”, Warsaw Technical
a = ambient University Publication, No. 3 Mechanika (1968).
av =average Osiadacz, A., “Simulation of Transient Gas Flows in Networks”, Int.
h = total blowdown Journal for Numerical Methods in Fluids 4, 13-24 (1984).
c = sonic venting Osiadacz, A., “Simulation and Analysis of Gas Networks”, Gulf
e = exit Publishing Company, Houston, Texas (1 987).
R = general Picard, D.J. and P.R. Bishnoi, “The Importance of Real-Fluid Be-
I = initial havior and Nonisentropic Effects in Modeling Decompression
P = pressure or pipe
Characteristics of Pipeline Fluids for Application in Ductile Frac-
S = subsonic venting, or stack, or entropy
ture Propagation Analysis”, Can. J. Chem. Eng. 66,3-12 (1988).
sl = stack, upstream of valve Rachford, H.H. Jr. and T. Dupont, “Some Applications of Transient
s2 = stack, downstream of valve
Flow Simulation to Promote Understanding the Performance of
th =throat
V = volume or specific volume
Gas Pipeline Systems”, SPE Journal, 179-186 (1974).
0 = condition at closed end of pipe (see Figure 1)
Shapiro, A.H., “The Dynamics and Thermodynamics of Compress-
1 = condition at stack end of pipe (see Figure 1) ible Fluid Flow”,I, Ronald Press Company, New York (1953).
2 = condition at stack bottom (see Figure 1)
Streeter, V.L. and E.B. Wylie, “Natural Gas Pipeline Transients”,
Society of Petroleum Engineering SPE Paper No. 2555 (1970).
References Studzinski, W., M.H. Weiss and K.K. Botros, “Critical Flow Factor
for Natural Gas”, 2nd International Conference on Flow Measure-
Aly, F.A. and L.L. Lee, “Self-Consistent Equations for Calculating ment, London, U.K., May 11-13 (1988).
the Ideal Gas Heat Capacity, Enthalpy and Entropy”, Fluid Phase Von Rosenberg, D.U., “Methods for the Numerical Solution of
Equilibria6, 169-179 (1981). Partial Differential Equations”, American Elsevier Publishing
Anderson, D.A., J.C. Tannehill and R.H. Pletcher, “Computational Co., New York (1969).
Fluid Mechanics and Heat Transfer”, Hemisphere Publishing
Corp., New York (1984).
Anderson, J.S. and G.E.A. Meier, “Steady and Non-Steady Tran- Manuscript received September 21, 1988, revised manuscript
sonic Flow in a Duct with a Sudden Enlargement”, Max-Planck- received February 15, 1989; accepted for publication February 22,
Institut fur Stromungsforschung in Gottingen, Report 1 (1982). 1989.
THE CANADIAN JOURNAL OF CHEMICAL ENGINEERING, VOLUME 67, AUGUST, 1989 539