09575820/96/$07.00+ 0.

00
Institution of Chemical Engineers

BLOWDOWN OF LPG PIPELINES

S. M. RICHARDSON (FELLOW) and G. SAVILLE
Department of Chemical Engineering, Imperial College, London, UK

n order to be able properly to assess the hazards associated with the rapid
depressurization or blowdown of an oil, gas or condensate pipeline, there is a need to
be able to predict uid pressure, uid and wall temperatures and e ux rate,
composition and phase. Several computer programs have been developed to do this,
including our program BLOWDOWN. A comparison is made here of BLOWDOWN
predictions with the measurements made during eight of the tests using LPG carried out
by Shell and BP on the Isle of Grain in 1985. Four of the tests were for full-bore
depressurizations and four for depressurizations with ori ces at the open ends of the lines.
The BLOWDOWN predictions are shown to be in at least adequate, and often good,
agreement with the Isle of Grain measurements.
Keywords: blowdown; hydrocarbons; LPG; pipelines.

INTRODUCTION
The rapid depressurization or blowdown of an oil, gas or
condensate pipeline, whether deliberate or accidental, is
a hazardous operation. One part of the hazard arises
because of the very low temperatures generated within
the uid in the line. This leads to a reduction in the
temperature of the walls and possibly to a temperature
below the ductile-brittle transition temperature of the
steel from which the line is made. It can also possibly lead
to the generation of unacceptably high thermal stresses
within the line. It can in addition lead to the formation of
hydrates in cases when free water is present in the line.
Another part of the hazard arises because of the large
total e ux and high e ux rates that occur when the large
inventory of a typical line such as a sub-sea pipeline is
blown down.
In order to be able properly to assess the hazards
associated with depressurization of a pipeline containing
hydrocarbons, there is a need to be able to predict uid
pressure, uid and wall temperatures and e ux rate,
composition and phase. Several computer programs
have been developed to do this including:

pipelines containing hydrocarbon mixtures at elevated

pressures. One of the few sources of such data is the tests
using LPG carried out by Shell and BP on the Isle of
Grain89, full details of which are given elsewhere10.
Comparisons of BLOWDOWN predictions with some of
these test data are presented here.

ISLE OF GRAIN TESTS

The Isle of Grain tests were conducted in July to
September 1985. Two parallel pipelines were used, both
100 m long. The large pipeline was nominally of diameter
69 9 and the small pipeline was nominally of diameter 29 9 .
Ori ces of various shapes and sizes were used. Four
initial tests (F1-F4) were conducted using Freon-12.
Eighty-four main tests (P1-P84) were conducted using
commercial propane or LPG. The two pipelines were
extensively instrumented. Pressure transducers and
thermocouples measuring uid temperature were
attached along each line. Inventory and hold-up were
measured using load-cells and neutron back-scattering.
Video recordings were also made of the ows using
special windowed sections mid-way along and at the
open end of each line.
Two types of test were conducted: steady and
transient. Both types of test were initiated by rupture
of a disc at the downstream end of the pipeline. In the
steady tests, the ow through the pipeline was fed
steadily and continuously from an upstream reservoir. In
the transient tests, the ow was inherently unsteady and
there was no feed from the upstream reservoir. Clearly, it
is the transient tests which are of direct relevance to
depressurization. Eight of these, covering as wide a range
of conditions as possible, were selected for comparison
with BLOWDOWN predictions:

BLOWDOWN
developed
Imperial
College principally for
15

by the authors at
the simulation of the
depressurization of networks of vessels and associated
pipework on the topsides of o shore oil and gas
platforms (and extensively validated for such networks)
and extended to simulation of the depressurization of
pipelines;
OLGA developed at SINTEF for the simulation of
two-phase
ow in pipelines;
PLAC7 developed by AEA Technology also for the
simulation
of two-phase ow in pipelines.
6

One principal di culty with such predictive programs is,

however, a lack of proper validation against appropriate
experimental data for the depressurization of pipelines.
These data should preferably be for relatively large

P40, P42, P45 and P47 using the 69 9

Tests
Tests
P61, P63, P65 and P66 using the 29 9

235

line;
line.

236

RICHARDSON and SAVILLE

Table 1. Pipelines.

Line
69 9
29 9

Table 3. Initial and ambient conditions.

Length,
m

Inner
diameter,
mm

Outer
diameter,
mm

Wall
roughness,
mm

100.0
100.0

154.0
52.0

168.6
60.9

0.05
0.05

Details of the two carbon steel pipelines used in the tests

are given in Table 1.
Ori ces of various sizes (measured in terms of
equivalent diameter that is, the diameter of the circle
of the same cross-sectional area as the ori ce) and shapes
were used in the eight tests. Details are given in Table 2.
Note that tests P40, P42, P61 and P65 were for full-bore
depressurizations.
The LPG used in the tests comprised a mixture of
propane and other low molecular weight hydrocarbons,
such as butane and ethane. The approximate composition was 95 mole% propane and 5 mole% butane. The
initial pressure and temperature of the LPG and the
ambient air temperature and wind speed for the eight
tests are given in Table 3.
Ten pressure transducers and ten thermocouples (bare
metal with a response time of the order of milliseconds)
measuring temperature were attached along each line.
The thermocouples projected from the inside walls of the
lines and measured uid temperatures close to the wall.
Eight neutron back-scattering (NBS) devices were used
to measure liquid hold-up in each line. Twenty load-cells
were used to measure the inventory in each line. Details
of the precise locations of the measurement devices are
given elsewhere10.
It is di cult to make a reliable assessment of the
uncertainty in the data, mainly because no access can be
gained to the original data acquisition devices. Nevertheless, based on the nature of those devices and the
scatter in the measured data, the likely uncertainty is
estimated to be about 0.5 bar in the pressure data
(though the uncertainty is higher in test P61), 2C in the
temperature data, 5% in the load-cell inventory data
and 10% in the NBS hold-up data.
BLOWDOWN PROGRAM
The depressurization of the Isle of Grain pipelines
Table 2. Ori ces.

Test

Line

Ori ce
diameter,
mm

P40
P42
P45
P47
P61
P63
P65
P66

69 9
69 9
69 9
69 9
29 9
29 9
29 9
29 9

154.0
154.0
75.0
50.0
52.0
35.0
52.0
35.0

Shape
Circular
Circular
Circular
Circular
Circular
Circular
Circular
Equilateral
triangular

Test

Initial LPG
pressure,
bar a

Initial LPG
temperature,
C

Ambient air
temperature,
C

Wind
speed,
m s- 1

P40
P42
P45
P47
P61
P63
P65
P66

21.6
11.3
11.4
21.3
21.2
22.5
11.6
21.8

17.8
20.0
15.9
14.6
(?)
18.4
13.8
13.3

19.1
18.6
16.7
15.4
(?)
(?)
(?)
(?)

(?)
8.8
6.5
(?)
(?)
(?)
(?)
(?)

has been simulated using our computer package

BLOWDOWN4. BLOWDOWN has been developed
principally for simulation of the depressurization of
networks of vessels and associated pipework on
o shore oil and gas platforms. The distinction between
depressurization of a vessel and depressurization of a
pipeline is that there is a signi cant pressure di erence
within the latter but not within the former. Thus, for
depressurization of a vessel, the only signi cant
pressure drops are across the ori ce and between the
exit of the ori ce and the atmosphere if the ow is
choked at the ori ce. For depressurization of a
pipeline, in contrast, there are signi cant pressure
drops along the line, across the ori ce if the ori ce is
small enough (clearly, if there is a full-bore rupture,
there is no pressure drop across the ori ce) and
between the exit of the ori ce and the atmosphere if
the ow is choked at the ori ce.
The main features of BLOWDOWN are as follows.
Space and time discretization
A pipeline is divided axially into many elements, the
size of each of which is varied dynamically during the
calculation in such a way that changes in physical
properties along the element may be neglected. A vessel
is divided into three zones:
A top zone of gaseous hydrocarbon (including
evaporated
water and suspended liquid droplets which
have condensed from the gas);
A middle zone of liquid hydrocarbon (including
dissolved
water and gas bubbles which have evaporated
from the liquid);
A bottom zone of free water (including dissolved
hydrocarbons).
Depressurization is broken down into a sequence of
pressure, rather than time, steps because pressure is a
more relevant parameter thermodynamically.
Fluid mechanics
Flow (whether one-phase or two-phase) is assumed
to be quasi-steady (that is, the mass ow rate is
assumed to be the same in every element in a pipeline
at a given instant of time) and, for a two-phase ow, to
be homogeneous (that is, the gas and liquid in an
element move at the same velocity). Standard methods
are used to determine friction factor and hold-up.
Trans IChemE, Vol 74, Part B, November 1996

BLOWDOWN OF LPG PIPELINES

Thermodynamics
Experimental evidence has shown that it is most
important to model the thermodynamics of depressurization accurately since failure to do so can lead
to trajectories through phase (pressure-temperaturecomposition) space which are grossly in error. For this
reason, thermodynamic, phase and transport properties
of the multi-phase multi-component uids are calculated
rigorously by an extended principle of corresponding
states using our computer package PREPROP. The
extension is necessary in order to treat non-spherical
molecules through the use of shape factors. The idea
behind the thermodynamic calculations is to obtain an
expression for the di erence between the actual value of
a property at a given volume and temperature and the
value which a perfect gas would have at the same volume
and temperature: the properties of the mixture are then
related to those of a single reference substance, methane.
The method, when coupled with expressions for perfect
gas properties, enables one to obtain thermodynamically
consistent values for the enthalpy, pressure and fugacity
of any single-phase mixture. PREPROP uses these values
and performs phase equilibrium calculations including
dew and bubble points and isothermal, isentropic and
isenthalpic liquid-vapour ashes. Other properties such
as thermal conductivity, viscosity and surface tension
can also be predicted though these are not based on the
same rigorous thermodynamics.
Heat transfer
In a pipeline, heat transfer is assumed to be by:
convection to the uid;
Forced
Transient
insulation); conduction through the wall (including any
convection to the air or sea surroundingForced/natural
the line.
In a vessel, heat transfer is assumed to be by:
convection in the top zone;
Forced/natural
Nucleate/
lm
boiling
middle zone;
Natural convection ininthethebottom
zone;
Transient conduction through the wall
(including any
insulation);
convection to the air or sea surroundingForced/natural
the vessel.
In all cases, standard correlations are used to determine
heat transfer coe cients.
Ori ces
Because transit times through ori ces are comparable
with times for nucleation and growth of gas bubbles in a
volatile liquid, non-equilibrium ashing ow is assumed
when just volatile liquid is fed to an ori ce. Otherwise,
the ow approaching an ori ce is assumed to be in
thermodynamic and phase equilibrium.
Choking
When choking occurs, whether for a gas, liquid or twophase (gas plus liquid) ow, the mass ow rate is a
maximum.
Trans IChemE, Vol 74, Part B, November 1996

237

Balance equations
In order to close the system of equations describing
depressurization, mass and energy balances are performed. For pipelines, momentum balances are also
performed. The balances for each element of a pipeline
are linked together in an iterative scheme to satisfy the
boundary condition of ambient pressure (or choking
pressure if the ow becomes choked) at the open end of
the line or exit of the ori ce. The iterative variable is the
mass ow rate through the line.
Pressure and temperature changes
Allowance is made for the pressure change when a pipe
of one diameter joins a pipe of another diameter. In
particular, when there is a sudden enlargement of
diameter, the Borda-Carnot equation (or its two-phase
analogue) is used to determine the pressure change. No
other pressure changes (other than those arising continuously from friction at walls) are, however, allowed
for. Thus pressure changes at bends and tees, in
particular, are ignored. An energy balance is used to
determine the temperature change at expansions (or
contractions) and also tees arising from interconversion
of mechanical (kinetic) energy and thermal energy.
Allowance is also made for conversion of mechanical
(pressure) energy lost through friction at pipe walls into
thermal energy. This is especially important when the
ow in a given pipeline is fast, in particular when it is
choked or nearly choked.
METHODOLOGY
The approximate composition of the LPG used in the
tests is 95 mole% propane and 5 mole% butane.
Numerical tests using BLOWDOWN reveal that the
predictions are relatively insensitive to the precise initial
composition (the only signi cant e ect is on uid
temperatures towards the end of a depressurization).
Initial LPG pressures and temperatures are known (see
Table 3) for all tests except test P61, for which the initial
LPG temperature is unknown: a value of 17C has been
assumed. Ambient air temperatures are known for
tests P40, P42, P45 and P47 only (see Table 3). Air
temperatures of 17C have been assumed for tests P61,
P63, P65 and P66. Wind speeds are known for tests P42
and P45 only. Numerical tests made using BLOWDOWN reveal that the predictions are relatively
insensitive to the assumed wind speed: the only
signi cant e ect is on wall temperatures, which were
not measured in the tests. Accordingly, it has been
assumed that the wind speed for all tests is zero and thus
the ambient air is assumed to be freely-convecting. The
ambient air pressure (and hence the back-pressureon the
ow from the pipeline) is assumed to be 1 bar a for all
tests.
In tests P40, P42, P61 and P65 in which there is a fullbore depressurization, the equivalent ori ce diameter is
unambiguously that of the bore of the pipeline being
depressurized (assuming that the discs used to initiate the
tests ruptured fully: the evidencesuggests that they did so
for these tests) and the discharge coe cient is unity. In
tests P45, P47, P63 and P66, in which there is an ori ce at
the end of the line, the magnitude of the equivalent

238

RICHARDSON and SAVILLE

Table 4. Equivalent ori ce diameters.

Test

Nominal
ori ce
diameter,
mm

Adopted
ori ce
diameter,
mm

Adopted
discharge
coe cient

P45
P47
P63
P66

75.0
50.0
35.0
35.0

95.0
70.4
42.5
42.5

0.80
0.80
0.80
0.80

ori ce diameter is less clear. Experimental evidence3

shows that a discharge coe cient of 0.80 is appropriate
for ow of pressurized gas or gas plus liquid through an
ori ce. In order to match the measured e ux rates from
the pipelines, it has been found necessary to adopt the
values of equivalent ori ce diameters given in Table 4
instead of the nominal ones. The reason for the
discrepancy between the nominal and adopted values is
unclear. It may be associated with an assumed value of
discharge coe cient corresponding to a nominal equivalent ori ce diameter (a value of 0.61 is often quoted,
based on measurements for ow of liquid through an
ori ce: a higher value is appropriate for ows of
pressurized gas or gas plus liquid, probably because the
vena contracta becomes larger as a result of gas
expansion).
PREDICTIONS
Key predictions from the complete BLOWDOWN
simulations are given in Figures 18 to enable direct
comparison to be made with the Isle of Grain data:
Graphs P##P give the variations with time of the uid
pressure
at the closed and open ends of the line;
Graphs P##T give the variations with time of the bulk
uid
temperature at the closed and open ends of the line
(note that there are no measured temperature data for
test P61);
Graphs P##I give the variations with time of the total
uid
inventory of the line based on the load-cell data
(and not on the NBS data because they are much noisier
than the load-cell data);
where ## refers to the test number. Throughout,
solid lines refer to measurements and broken lines to
predictions.
PRESSURE
Measurements
Graphs P##P all exhibit an extremely rapid initial
decrease in pressure at the closed end of the line to a
pressure of about 8 bar a and, indeed, a slight undershoot.
The rapid decrease corresponds to a very short period of
expansion of the compressed LPG followed almost at
once by ashing of the LPG. The bubble point pressure
(or saturated vapour pressure) of the LPG used at 20C
(the approximate starting temperature for all of the tests)
is about 8 bar a. The undershoot is probably associated
with non-equilibrium arising from the rapidity of the
pressure decrease and perhaps also re ection of expansion
waves o the closed end of the line.

Figure 1. Graphs P40P/T/I: solid lines measured; broken lines

predicted.

Graphs P##P show that the pressure at the open

end of the line is always less than that at the closed
end, as is to be expected. However, the pressure at
the open end of the line is very similar to that at the
closed end in tests P45 and P47: this is because
the ori ce area is relatively small compared with the
cross-sectional area of the line and so the pressure
Trans IChemE, Vol 74, Part B, November 1996

BLOWDOWN OF LPG PIPELINES

239

Figure 2. Graphs P42P/T/I: solid lines measured; broken lines

predicted.

Figure 3. Graphs P45P/T/I: solid lines measured; broken lines

predicted.

drop along the line is negligible compared with that

across the ori ce.
The extremely rapid decrease in pressure at the closed
end of the line, corresponding to a very short period of
expansion of the compressed LPG followed almost at
once by ashing of the LPG, would be much slower in a
really long pipeline, where expansion of the compressed

liquid could not so quickly be accommodated because of

signi cant frictional and inertial e ects. Similarly, the
pressure at the open end of a really long pipeline would
not be so close to that at the closed end unless the ori ce
area were much smaller than the cross-sectional area of
the line, again because of signi cant frictional and
inertial e ects.

Trans IChemE, Vol 74, Part B, November 1996

240

RICHARDSON and SAVILLE

Figure 4. Graphs P47P/T/I: solid lines measured; broken lines

predicted.

Figure 5. Graphs P61P/T/I: solid lines measured; broken lines

predicted.

The pressure at the open end of the line in test P61 tends
to a value of about - 4 bar a when the depressurization is
complete. This is clearly impossible and re ects de ciencies in the raw pressure data for that test. It is, however,
di cult to correct the raw data: addition of 5 bar (say) to
all values seems of doubtful validity. For this reason, none
of the measured data have been modi ed in any way.

Predictions
Graphs P# #P show that there is generally at least
adequate, and often good, agreement between the
predictions and the measurements of uid pressure.
The discrepancy in pressure is less than about 2 bar at
any given instant. The corresponding discrepancy in time
is less than about 4 s at any given pressure. An exception
Trans IChemE, Vol 74, Part B, November 1996

BLOWDOWN OF LPG PIPELINES

241

Figure 6. Graphs P63P/T/I: solid lines measured; broken lines

predicted.

Figure 7. Graphs P65P/T/I: solid lines measured; broken lines

predicted.

arises at the open end of the line in test P61, for which the
measurements are clearly erroneous.
Graphs P##P show that the predictions all reproduce
the extremely rapid initial decrease in pressure corresponding to expansion of compressed liquid. They fail,
however, to reproduce any overshoot. This is because no

attempt is made in BLOWDOWN to simulate the

motion of the expansion wave from the open end of
the line to the closed end immediately after initiation of
the depressurization, or indeed to simulate any other
wave motions up and down the line, which are rather
complex but usually irrelevant.

Trans IChemE, Vol 74, Part B, November 1996

242

RICHARDSON and SAVILLE

almost certainly associated with the corresponding
pressure undershoot.
Graphs P##T all show that the temperature at the
open end of the line is less than that at the closed end, as
is to be expected. However, the temperature at the open
end of the line is very similar to that at the closed end of
the line in test P47. This is because the ori ce area is then
relatively small compared with the cross-sectional area of
the line and so the pressure drop along the line is
negligible compared with that across the ori ce, which in
turn means that conditions are relatively uniform along
the length of the line.
Graphs P##T generally show that there is a signi cant
and rapid rise in temperature at the open end of the line
towards the end of the depressurization. The exceptions
are test P45 for which the corresponding temperature rise
might in fact occur at a later time and perhaps test P47
where there is a great deal of scatter in the data (and also
perhaps test P61 for which there are no data). There is also
a signi cant and rapid rise in temperature at the closed
end of the line towards the end of the depressurization in
tests P47 and P63. These rises start at the moment when
the ow ceases to be two-phase (gas plus liquid) and
becomesone-phase (gas) instead and presumably re ects a
signi cant change in the heat transfer rate from the wall to
the uid. The slight levelling of the temperature immediately before the rise in temperature at the open end of the
line in test P42 is associated with the cessation of choking
at the open end of the line.
Graphs P##T show that the minimum temperature,
which occurs towards the end of each test, is about
- 40C, which corresponds to the bubble point temperature of LPG at 1 bar a (the bubble point temperature at 1
bar a is - 42C for a mixture of 95 mole% propane and 5
mole% butane).

Figure 8. Graphs P66P/T/I: solid lines measured; broken lines

predicted.

TEMPERATURE
Measurements
Graphs P##T, with the exception of graph P47T, all
exhibit an extremely rapid initial decrease in temperature
at the open end of the line to a temperature of about
10C and, indeed, a slight undershoot. The undershoot is

Predictions
Graphs P##T show that there is generally at least
adequate, and often good, agreement between the
predictions and the measurements of bulk uid temperature. The discrepancy in temperature is less than about
10C at any given instant. The corresponding discrepancy
in time is less than about 4s at any given temperature.
Graphs P##T show that the largest discrepancies
between the predictions and the measurements arise
towards the end of the depressurizations. The reason for
this is probably the sensitivity of the predictions to the
assumed composition of the LPG. Although adjustments
could have been made to the assumed composition to
achieve better agreement, this has not been done. Indeed,
no adjustments have been made to BLOWDOWN or to
any of the data used in the simulations (with the
exception of the equivalent ori ce diameters in tests
P45, P47, P63 and P66). The simulations are thus purely
predictive.
INVENTORY
Measurements
Graphs P63I and P66I show that the e ux from the
line is essentially una ected by the geometry of the ori ce
Trans IChemE, Vol 74, Part B, November 1996

BLOWDOWN OF LPG PIPELINES

Table 5. Residual liquid
Test

Maximum fractional
residual liquid, v/v

P45
P47
P63

0.13
0.22
0.05

(recall that test P63 is for a circular ori ce and test

P66 for an equilateral triangular ori ce of the same
area). While this result is likely to be true more generally,
it may not be true if the aspect ratio of the ori ce
becomes very large if, say, the ori ce is a narrow slit.
Unfortunately, no tests were performed with narrow slit
ori ces.

Predictions
Graphs P##I show that there is generally at least
adequate agreement between the predictions and the
measurements of uid inventory. The discrepancy in
inventory is less than about 10% of the initial inventory
at any given instant. The corresponding discrepancy in
time is less than about 4s at any given inventory, except
in test P47, when it is about 8s.
Graphs P45I and P47I show that the predicted
inventory tends to zero at the end of the depressurization
whereas the measured inventory is nite and positive.
The reason for this may be that the ori ce acts as a dam
in these tests, tending to hold liquid back in the pipeline.
Both Isle of Grain pipelines were levelled very
accurately (apparently to the order of mm). For the
full-bore tests (P40, P42, P61 and P65), there would be
no tendency for liquid to be held back and all of it
would be expected to drain out eventually (though the
drainage rate would generally be insigni cant compared with the pressure-induced e ux from the line).
For the tests in which there is an ori ce at the end of
the line (P45, P47, P63 and P66), if the liquid level at
the end of the depressurizations were to be coincident
with the bottom of the ori ce, then elementary
geometry can be used to estimate the maximum volumes
of residual liquid held back. The results of such a
calculation (based on the adopted equivalent ori ce
diameters given in Table 4, as opposed to the nominal
ones, and normalized with respect to the initial liquid
volume) are given in Table 5. Note that no estimate is
given for test P66 since a triangular ori ce was used for
that test and its orientation appears not to have been
recorded.
Clearly, it is most unlikely that the maximum residual
volume would in fact remain in the line at the end of a
depressurization. Indeed, the measurements suggest that
around half of the maximum remains in tests P45 and
P47 (that in test P63 is almost undiscernible). Such
residual liquid might have a signi cant e ect on the
temperature of the bulk uid towards the end of a
depressurization.
Graphs P# #I show that the predicted inventory is
consistently greater than the measured one (except
Trans IChemE, Vol 74, Part B, November 1996

243

toward the end of some depressurizations, when there

is considerable uncertainty in the measurements). The
predicted e ux rate is correspondingly lower than the
measured one. The reason for this is unclear but may
well be associated with the assumptions of quasi-steady
and homogeneous ow made in BLOWDOWN. Independent evidence11 based on some of the Isle of Grain
data supports this conjecture. A much more re ned
model of pipeline depressurization (recall that BLOWDOWN was developed principally for simulation of
vessel depressurization) has been developed by the
authors12,13 in which the ow is allowed to be fully
transient (thus relaxing the assumption of quasi-steadiness) and phase slip is allowed through use of a two uid model (thus relaxing the assumption of homogeneity). Much better agreement is found between
predictions and measurements for a limited set of Isle
of Grain tests so far examined. The disadvantage with
the use of the more re ned model is the signi cantly
increased computer run times. Whereas BLOWDOWN
typically requires a few hours to perform a simulation
of an Isle of Grain test on a 386 or 486 machine tted
with an 860 co-processor, the more re ned model
typically requires a few days.

CONCLUSION
The objective of the work reported here was validatory
comparison of the predictions of BLOWDOWN with
some of the Isle of Grain test data. The predictions have
been shown to be in at least adequate, and often good,
agreement with the Isle of Grain measurements, permitting increased con dence to be placed in predictions
made using BLOWDOWN.

REFERENCES
1. Haque, M. A., Richardson,S. M., Saville,G. and Chamberlain, G.,
1990, Rapid depressurizationof pressure vessels, J Loss Prev Proc
Ind, 3: 47.
2. Haque, M. A., Richardson, S. M. and Saville, G., 1992, Blowdown
of pressurevessels. I. Computer model, Trans IChemE, 70 (B1): 3
9.
3. Haque, M. A., Richardson,S. M., Saville,G., Chamberlain, G. and
Shirvill, L., 1992, Blowdown of pressure vessels. II. Experimental
validation of computer model and case studies, Trans IChemE, 70
(B1): 1017.
4. Richardson, S. M. and Saville, G., 1991, Blowdown of pipelines,
Oshore Europe 91 (Society of Petroleum Engineers, Aberdeen),
paper SPE 23070.
5. Richardson, S. M. and Saville, G., 1992, Blowdown of vessels and
pipelines, Major Hazards Onshore and Oshore, Symposium Series
No. 130, 195210 (IChemE, Rugby, UK).
6. Bendiksen, K. H., Malnes, D., Moe, R. and Nuland, S., 1991, The
dynamic two- uid model OLGA: theory and application, SPE
Production Engineering, 6: 171180.
7. Hall, A. R. W., Butcher, G. R. and Teh, C. E., 1993,
Transient simulation of two-phase hydrocarbon ows in
pipelines, European Two-Phase Flow Group Meeting (Hannover),
paper I4.
8. Cowley, L. T. and Tam, V. H. Y., 1988, Consequences of
pressurised LPG releases: the Isle of Grain full scale experiments,
GASTECH 88, 13th International LNG/LPG Conference, Kuala
Lumpur.

244

RICHARDSON and SAVILLE

9. Tam, V. H. Y. and Higgins, R. B., 1990, Simple transient release

rate models for releases of pressurised liquid petroleum gas from
pipelines, J Haz Mat, 25: 193203.
10. Richardson, S. M. and Saville, G., 1996, Isle of Grain pipeline
depressurisationtests, HSE OTI 96 ###. (HMSO, London).
11. Pots, B. F. M., 1990, Numerical simulation of multi-component
two-phasepipelineblowdowns using the homogeneousequilibrium
model, Shell Report, AMGR.90.337.
12. Chen, J-R., Richardson, S. M. and Saville, G., 1995, Modelling of
two-phase blowdown from pipelines- I. A hyperbolicmodel based
on variational principles, Chem Eng Sci, 50: 695713.
13. Chen, J-R., Richardson, S. M. and Saville, G., 1995, Modelling of
two-phase blowdown from pipelines - II. A simpli ed numerical
method for multi-component mixtures, Chem Eng Sci, 50: 2173
2187.

ACKNOWLEDGEMENTS
The experimental data upon which this paper is based have been
made available by Shell Research (Thornton). The analysis of the data
was funded by the Health and Safety Executive.

ADDRESS
Correspondence concerning this paper should be addressed to
Professor S.M. Richardson, Department of Chemical Engineering,
Imperial College, Prince Consort Road, London SW7 2BY. E-mail:
s.m.richardson@ic.ac.uk
The manuscript was received 7 June 1996 and accepted for publication
12 July 1996.

Trans IChemE, Vol 74, Part B, November 1996