Professional Documents
Culture Documents
Gas Explosion Scale Offshore Modules Using Fuel/air Clouds Experiments in I:33 and 1:5 Separator and Compressor Stoichiometric Homogeneous
Gas Explosion Scale Offshore Modules Using Fuel/air Clouds Experiments in I:33 and 1:5 Separator and Compressor Stoichiometric Homogeneous
Accidental gas explosions occuring on offshore plat- 0 tube with sharp edged rings;
forms are characterized by propagation in complex 0 radial geometries with sharp and rounded obstacles;
geometries. It is well known that a strong coupling exists l spherical geometries with various volume blockage
between the explosion and the geometry in which it ratios
propagates. Very often it is not possible, in a numerical
The experimental data collected in simplified situations
model, to take account of every geometrical detail in an
played a decisive role in optimizing the various parts of
exact manner. The reason for this is the vast amount of
the gas explosion model. It remains, however, to check
numerical grid points that would be required to model
whether the model is capable of reproducing explosion
complex geometries. An alternative method has been
behaviour in realistic offshore modules ‘and for this
proposed by Hjertager t to describe the geometry and to
purpose good experimental data are needed.
model the resultant influence on the explosion. The
Previous experimental data in realistic offshore
method uses the porosity and distributed resistance
modules are scarce. Very often it has been usual to do
concept to take account of explosion propagation in
experimental investigations in empty modules, and to
complex geometries. The distributed resistance func-
use the vent area and ignition point position as para-
tions are dependent on bulk parameters such as obstacle
meters in the studies’. The empty volume assumption
dimensions and orientation, as well as the specific
has been shown to be totally inadequate for modules
wetted area of the obstacles per unit volume of the area
containing process equipment 3 - ‘. It is therefore of the
containing complex geometries.
utmost importance to collect experimental data in
This model has been developed and validated based
modules containing various realistic process equipment.
on experiments carried out in idealized geometries e.g.:
The present paper gives the results of experimental
investigations carried out in offshore modules in scale
Received II Augusl 1988 1:33 as well as 15 of realistic compressor and separator
0950-4230/88/040197-1 Os3.00
0 1988 Butterworth & Co. (Publishers) Ltd
J. Loss Rev. Process lnd., 1988, Vol 1, October 197
Gas explosion experiments used in offshore separator and compressor modules: 6. H. Hjertager et al.
modules. The parameters chosen to be varied in the cylindrical vessels, four on the upper deck (UD) and three
studies included: on the lower deck (LD), aligned along the module length.
In addition the module also contains a lot of smaller
l gas type (i.e. methane or propane);
pipes. The compressor module is characterized by two
l size and location of vent area;
compressor trams and two large box shaped rooms
l ignition position
located on the upper deck (UD) and four box shaped and
The clouds inside the modules were stoichiometric, and two cylindrical objects on the lower deck. Figures 3 and 4
covered the whole of the free space. The experimental give the overview of the large scale version (scale 1:5) of
tests in the 1:5 scale modules were performed at the test the separator and compressor modules with length 8 m,
site at Sund on the island of Sotra, while the 1:33 scale width and height of 2.5 m, that were installed at the
tests were carried out in the laboratory at CMI. test site on Sotra. For easy reference we have defined all
sides of the module. In Figures 3 and 4, we designate the
left end as south and right end as north. The nearest wall
is designated front wall and the opposite is designated rear
Experimental arrangement
wall, and finally the sides on the north and south are
designated north or south ends of the module. Therefore,
Module geometries
Figures I, 3 and 4 are showing the front side whereas
When the separator and the compressor module layouts Figure 2 is showing the rear side of the modules.
had been chosen, a 1:33 scale model was made, with In addition to the two layouts given above, the
length 1.25 m, width 0.35 m and height 0.33 m, of a partially filled (P-filled) M24 module (only 1:33 scale
separation module. In addition a 1:33 scale model, with experiments) and the empty module without internal
length 1.25 m, width 0.45 m and height 0.35 m, of a process equipment were tested. The volume blockage
compression module was subsequently fabricated. ratios (VBR) of obstacles inside three of the various
Figures I and 2 give a view of the two modules, the M25 layouts are given in Table I.
separator module and the M24 compressor module,
respectively. These figures give a good impression of the Vent arrangements
internal layout of equipment and pipes. The separator The venting of gas explosions is characterized by the
module given in Figure I is characterized by seven large size, layout and location of the vent area. The size of a
Figure 1 Front view of the 1:33 scale separator module, M25 Figure 3 Front view of the 1:5 scale separation module, M25
Figure 2 Rear view of the 1:33 scale compressor module, M24 Figure 4 Front view of the 1:5 scale compressor module, M24
Upper deck,
VBR l-1 0.15 0.03 0.3
Lower deck,
VBR (-1 0.1 0.1 0.3
Total module,
VBR (-1 0.13 0.065 0.3
Instrumentation
Figure 6 Schematic diagram showing the dimensions of the Synchronization of the ignition, resetting and calibra-
louvres tion of pressure transducers were controlled by a 10
DIMNSIONS SMALL-SCALE (1%) channel programmed timer (UP timer, Xanadu Con-
LENGTH = I.25m HEIGHT
=Cl.35
m trols). The signals were recorded on a 14 channel analog
tape recorder (PR 2230, Ampex Corporation). After
WIDTH (M24) = 0.45 m WIDTH (MB) = 0.35 m
each test, the records were displayed with a u.v.-
FRONT VIEW
recorder (Autograph 8, Bryans). Two types of diag-
nostic probes were used to monitor the explosion. The
7 h > 4 ?? z 1
flame position inside the modules was recorded by
n P1
I----------------------------------=-’---
UPPer deck (UD) placing ionization gaps centrally along the upper deck
and along the lower deck as shown in Figure 8 for the
1:33 scale tests and in Figure 9 for the 1:5 scale tests.
n P2
Lower deck (LD) ebl.2 The pressure time histories inside the modules were
SOUTH Nr NDRTH monitored either at two positions (1:33 scale) or at six
END V,EW positions (1:s scale) as shown in Figures 8 and 9. In the
large-scale (1:5) tests, the outside blast wave was moni-
tranr*ucer* I I
c
Pressure (PI -PZ) . IG 1-7
I . tored by a transducer placed 10 m from the south end of
PI the module. In some of the tests, high speed movie
P1.r =Om Uxercleck I e Ian. ,
recordings were made through a transparent front wall in
the module.
FRONT REAR
Figure 9 Schematic diagram showing the 1:5 scale module geo- Figure 10 Peak pressure as function of various geometrical
metry with pressure transducer, ionization gap and ignition loca- layouts of the 1:X3 scale M25 separator module. Methane/air
tions clouds
CENTRAL IGNITION
IGNITION
Bs CmlRur
~CEhl%lB
OENOLD
General. The peak pressure in the tests varied from CENTRAL IGNITION
1.9 bar for the most confined explosion using pro-
pane/air and all the way down to about 5 mbar for the
empty module ignited at the open end.
UD/centre 100
(as given in Fig. 121
UD/close to rear wall 153
LD/close to rear wall 260
1
86143 265 I
I,
86147 230
,I
86/48 210
,I 223
86149 222
I, 0
86157 230 0
85/43 Propane/air 310
,I 340 lGNw0~ &ATION FR& CENTRAL psOs.. ( m )
85164 370 I
I, Figure 15 Peak pressure as function of ignition position given as
86150 366
,I 389
86/56 412 1 distance from the centre position. A distance of 4 m indicates exit
of north end. M25 compressor module. Methane/air clouds.
Figure 16 High-speed movie recordings of flame propagation in the 1:5 scale compressor module. a, Test No. 88 (19861. Ignition: centre
on upper deck. Framing rate: 125 frames per sec. Av/V lr3 = 0.92. Propane/air. b, Test No. 87 (1986). Ignition: centre on lower deck.
Framing rate: 100 frames per sec. A& 2’3 = 0 92. Propane/air. c, Test No. 97 (1986). Ignition: on lower deck at north end. Framing rate:
25 frames per sec. A,/V2’3 = 0.92. Propane/air
explosions give very similar results as the spark direction, thus producing the turbulent environment that
ignited explosions. The table also shows that there is a prerequisite for flame acceleration. This tbune
is an acceptable degree of reproducibility from shot to acceleration of end ignited cases is not found when the
shot. internal process equiment is taken out. In that situation
Figures 14 and IS show the influence of changing the the peak pressure amounts to about 5 mbar and peak
ignition position inside the separator and compressor flamespeedtoaboutSms_‘.
module. It is seen from these figures that the peak
pressure does not generally decrease as the ignition Further discussion
point is moved towards the open north end. In fact, Figure 18 shows comparisons of the present pro-
Figure I4 shows an increase from about 750 mbar for pane/air tests in the M25 separator module, with other
central ignition to 1300 mbar for end ignition. this is the data collected by the authors and others. DnV 2 has
similar trend as was found in the 1:33 scale tests. performed a range of explosion tests in an empty 35 m 3
Figures 16 u-c show the detailed position of the module geometry using the vent and ignition position as
flame inside the M24 compressor module, as recorded parameters. Their cases with rear end ignition are shown
by high-speed camera through the transparent front in Figure 18. It is noted that the vent parameter is much
wall, for central ignition upper deck and lower deck and smaller (0.05-0.2) than the smallest vent parameter
north end ignition at the lower deck, respectively. used in the present experiments (0.46). Because the DnV
Figures Ida and b show that the flame initially starts module is empty, the peak pressure versus vent para-
out as a hemisphere, but the shape is modified as the meter is as expected far below the present data.
flame interacts with the openings in the upper deck and We have performed a series of experiments in a 50 m 3
with the process equipment inside the module. These tube3,4*6. Most of the tests in the tube have been
two figures also show the acceleration of the flame as it conducted using a planar ignition source at the closed
approaches the north and south ends. The flame acceler- end of the tube. However, some selected tests using a
ation is further shown in Figure Idc, which shows an point source have also been performed and the result of
end ignited case. one case is shown in Figure 18. The vent parameter is
The flame only needs three frames to exit from the somewhat smaller than in the present tests, but the
south, once the flame has reached the middle of the pressure is much higher than in the module tests. In
module. The corresponding flame speeds along the fact, it seems that the tube data is a good extrapolation
upper and lower decks, given in Figure 17, showing the from the module data. This is as expected since the tube
heavy acceleration in the final part of the game propa- as the module, contains obstacles.
gation through the module. The explanation for this Also shown in Figure I8 is the data for the 10 m
acceleration is probably due to the process equipment. At radial vessel with variable top venting ‘. As we can see,
the start of the flame propagation process, the lIame is the case with the smallest vent parameter, i.e. the case
venting behind and very little flow velocity is generated in with solid top wall, gives a pressure build-up that
front of the tlame. This continues until the venting and connects fairly well to the present module tests. How-
flow behind the flame is made more difficult, due to the ever, when top venting is increased, the pressure reduc-
increased resistance to flow caused by the obstructions. tion is much larger than in the present module tests.
Therefore, more of the expansion goes in the forward This is due to the fact that the venting is evenly
-
800
II n Lower deck
0 wer -- deck
Figure 17 Flame speed versus distance along the module for the case which is ignited at the open north end
Acknowledgements
Figure 18 Peak pressure in various propane/air tests as function This work has been financially supported by BP Petrol-
of the vent parameter, A,/V2’3 eum Development Ltd. Norway, Elf Aquitaine Norge
A/S, Esso Norge A/S, Mobil Exploration Norway Inc.,
distributed and that ignition is very close to the top vent. Norsk Hydro and Statoil. The authors acknowledge the
Both these conditions favour a lowering of explosion technical support during the experiments of Mr H. G.
pressure as seen in Table 2 and form the result of the Thorsen and Mr A. Nilsen.
present case with three louvred walls (Figure 18).
The final set of data that are shown in Figure 18 stem
from the unconfined, spherical and obstructed tests References
conducted by the authors ‘. The data are collected in the
Hjertager, B. H., ‘A Computer Model for Gas Explosion Propag-
27 m3 3-D corner, which is Q of a full unconfined ation in Complex High-Density Geometries’, Chr. Michelsen
sphere. The vent parameter for these cases amounts to Institute, CM1 Report No. 833403-5, 1984
about 6.0, which is beyond the present module tests. Solberg, D. M., ‘Gas Explosion Research Related to Safety of
Ships and Offshore Platforms’, Fuel-Air Explosions, SM Study
However, it is seen that the data for a volume blockage No. 16, Univ. of Waterloo Press, Ontario, Canada, 787, 1982
ratio, VBR = 0.2, and obstacle dimension equal to Eckhoff, R. K., Fuhre, K., Guirao, C. M. and Lee, 3. H. S., Fire
810 mm and 420 mm, is a good extrapolation of the Safely Journal, 1984, 7, 191
Moen, I. O., Lee, J. H. S., Hjertager. B. H., Fuhre, K. and
present data. The 3-D data also show the large effects of Eckhoff, R. K., Combustion and Flame, 1982, 47, 31
obstacle size for a given VBR. As we can see, the Bjerkhaug, M. and Hjertager, B. H., ‘The Influence of Confme-
smallest obstacle dimension of 164 mm and VBR = 0.2 ment on Flame Propagation and Pressure Development in a Radial
Vessel of Ten Metre Radius’, Chr. Michelsen Institute, CM1
produced a peak pressure of about 350 mbaro. All in Report No. 855403-2, 1985
all, the present module data seem to fit well into the Hjertager, B. H., Fuhre, K. and Bjsrkhaug, M., ‘Concentration
patterns previously found by the authors and others. Effects on Flame Acceleration by Obstacles in Large-Scale
Methane-Air and Propane-Air Vented Explosions’, 1988, Comb.
Sci. Technical. submitted for publication
Hjertager, B. H., Fuhre, K. and Bjerkhaug. M., ‘Spherical Gas
Conclusions Explosion Experiments in a High-Density Obstructed 27 m2
Corner’, 1988, to be presented at 6th International Conference on
1. The peak pressure inside the modules shows strong Loss Prevention and Safety Promolions iwthe Process Industries,
dependence on the venting arrangement. The highest Oslo, Norway, June 1989