Beruflich Dokumente
Kultur Dokumente
Centre for Technological Risk Studies (CERTEC), Department of Chemical Engineering, ETSEIB, Universitat Politcnica de Catalunya, Av.
Diagonal 647, 08028-Barcelona, Spain
b TIPs (Trmites, Informes y Proyectos, S.L.), C/Llenguadoc 10, 08030-Barcelona, Spain
a b s t r a c t
It is well known that the domino effect can have a major impact on accidents in storage facilities, as it can increase
the consequences of an initial event considerably. However, quantitative risk assessments (QRAs) do not usually take
the domino effect into account in a detailed, systematic way, mostly because of its complexity and the difculties
involved in its incorporation. We have developed a simple method to include the domino effect in QRAs of storage
facilities, by estimating the frequency with which new accidents will occur due to this phenomenon. The method
has been programmed and implemented in two case studies. The results show that it can indeed be used to include
the possibility of domino effect occurrence in a QRA. Furthermore, depending on the design of a facility, the domino
effect can have a signicant effect on the associated risk.
2012 The Institution of Chemical Engineers. Published by Elsevier B.V. All rights reserved.
Keywords: Domino effect; Quantitative risk assessment; Storage facilities
1.
Introduction
424
1.1.
Historical analysis of the domino effect (common
sequences, common accidents)
Darbra et al. (2010) performed a through historical analysis
of 225 accidents that occurred after 1961 and involved the
domino effect. The main factors in the study were accident
origin, causes, materials involved, effects, consequences and
most frequent accident sequences. The main source of information was the Major Hazard Incident Data Service (MHIDAS)
database (November 2007 version), managed by the Safety and
Reliability Directorate (SRD) for the UK Health and Safety Executive. Other sources consulted were MARS, which belongs to
the Major Accident Hazards Bureau (MAHB) of the European
Commissions Joint Research Centre (JRC), FACTS, which is a
database created by TNO in the Netherlands, and the Analyse,
Recherche et Information sur les Accidents (ARIA) database,
which is operated by the French Ministry of Regional Planning
and the Environment. The accident scenarios considered were
processing, loading/unloading, transportation and storage.
The conclusion was that 35% of accidents occurred in storage areas, which makes these types of facilities the most prone
to cascading accidents, followed by process plants at 28%. An
analysis of accident causes showed that 30.7% were due to
external events, 28.9% to mechanical failure, 20.9% to human
factors and 17.8% to impact failure. The remaining accidents
were caused by violent reactions, instrument failure, upset
process conditions and service failures. Another nding was
that 89% of the accidents involved ammable materials; LPG
was found to be the substance involved in the highest number
of events.
The sequences of various accidents were analyzed by constructing relative probability event trees. The most common
primary event was found to be re (probability: 0.524), followed
by explosion (probability: 0.476). The most frequent sequences
were explosion-re (27.6%), re-explosion (27.5%) and re-re
(17.8%). The accident with the most disastrous consequences
was the San Juan Ixhuatepec (Mexico, 1984) event, in which
a series of explosions and BLEVEs destroyed 50 out of 54 LPG
vessels in a fuel storage facility, killing approximately 500 people, injuring 3818 and causing 200,000 people to be evacuated.
Another conclusion of the study was that the number of
accidents involving the domino effect has decreased in recent
decades, after reaching a peak during the 1970s. However, the
phenomenon can still occur and plant design must take into
account this possibility.
Gmez-Mares et al. (2008) examined the relation between
jet res and the occurrence of the domino effect, using
databases like MHIDAS, ARIA, FACTS and MARS. A total of 84
accidents involving jet res occurring after 1961 were included
in the study. The analysis of these events allowed an event
tree to be constructed that was used to determine the most
frequent sequences of accidents involving the domino effect
and jet res. In 27% of the cases, the sequence was loss of
containment, jet re and explosion. In 11%, the sequence was
loss of containment, vapor cloud explosion, and jet re. Occurrence of a jet re was found to lead to another accident with
severe consequences in approximately 50% of the cases.
1.2.
425
2.
Method
2.1.
Input data
Number of tanks
Properties of the stored substances
Mass in each one of the tanks
Size of the containment dike of each unit or group of units
Type of tanks
Initial frequency of accidents per tank
Distances between tanks
2.2.
The rst step to develop the accident sequences that can occur
due to an initial LOC in certain equipment is to determine the
accidents and frequencies that can derive from the release
(event trees). The initial LOCs used in this work are those
dened in Table 1 and their frequencies for different types of
tanks are shown in Table 2; the event trees used are those
proposed in the CPR18E (CPR18E, 2005), including the ignition
Denition
Instantaneous release of the complete inventory.
Continuous release of the complete inventory in 10 min
at a constant rate of release.
Continuous release from a hole with an effective
diameter of 10 mm.
426
Fig. 1 Algorithm for development of domino sequences and calculation of nal frequencies.
G.1
7
G.2
1
5 10 y
5 106 y1
G.3
1
5 10 y
5 106 y1
1 105 y1
1 104 y1
427
Fig. 2 Event tree for the instantaneous release of a pressurized liqueed ammable gas.
to certain intensity of radiation for a specic amount of time
(Cozzani and Salzano, 2004b). Therefore, tanks in individual
containment dikes will not suffer G.1 releases due to adjacent
pool res (tilting of the ame by wind is not considered). The
value for the G.3 release is the radiation intensity at which the
domino effect could occur, as specied in Spanish legislation.
It is necessary to state that different equipment might fail
under diverse conditions, and threshold criteria can vary signicantly according to different sources. Table 3 only presents
the guide values used in this work, which can be changed to
increase accuracy when the method is applied.
According to Table 3, if a pool re occurs after a G.1 release
in a tank, and a nearby pressurized unit receives the equivalent thermal dose of 60 kW/m2 for 10 min, it will suffer a G.2
release. This might develop into one of the nal events related
to this release in a pressurized unit, which can in turn affect
other equipment. As stated previously, projectiles were not
included in the calculations because there are no models that
accurately describe their trajectory or radius of effect.
Domino sequences are arranged as event trees in which
accident probabilities are dragged through levels, so that the
accidents in the next level will have lower probabilities of
occurrence. For example, if a G.3 release occurs in a tank due
to a G.1 in another tank, the event tree of the G.3 scenario
will incorporate the probabilities of the initial one, and the frequency of the rst release. When an accident further down the
domino sequence has a frequency lower than 1 109 it is no
longer considered a possible initiator of other accidents. In this
way, the sequence may stop, even though not all units have
been affected. Sequences will be generated for each initial type
of release in each one of the studied units.
2.3.
3.
Case studies
which different types of oil derivatives are stored in six atmospheric tanks. Domino effect sequences for both cases are
developed using the proposed method and the nal frequencies are entered into RISKCURVES 7.6 to generate the iso-risk
curves associated with the facilities. The curves are compared
with those derived from traditional analysis. Hence, we can
determine the potential impact of the domino effect on the
curves and the risk associated with the plants.
3.1.
A facility in which ethylene and propylene are stored is evaluated. A total of 2.42 106 kg of propylene are stored in a
refrigerated sphere of 5450 m3 and 4.71 106 kg of ethylene
are stored in a vertical tank with a capacity of 9775 m3 at
cryogenic and atmospheric conditions. Case conditions, the
frequencies of initial LOCs, accident probabilities and the layout of the installation are presented in Tables 4 and 5, and
Fig. 3 respectively.
The facility is equipped with a re protection system (FPS)
with a probability of failure on demand (PFD) of 5%. All the
accidents triggered by a previous re event have a 95% lower
probability of occurrence. If a re accident occurs in Tank 2, the
FPS system in Tank 1 will start operating, and the probability of
the domino effect occurring will decrease. Accident probabilities will be multiplied by 0.05 in cases preceded by an external
re. A reduction in the probabilities due to the FPS is not taken
into account in initial accidents, as the FPS will stop res from
spreading, but will not stop them from starting. The FPS is
only considered in initial accidents for events that can occur
after the immediate ignition of contents spilled after a release,
such as a BLEVE in Tank 1. For this case, domino sequences are
developed with and without the FPS, to compare the effect it
has on the risk associated with the facility.
The method is applied to the facility using the event trees
from CPR18E. Domino sequences obtained using the model
for the case of G.1 events in Tanks 1 and 2, with active FPS,
are presented in Fig. 4. When an instantaneous release occurs
in Tank 1, a BLEVE, an explosion or a ash re might occur. If
a BLEVE happens, Tank 2 will suffer damage to its piping or
structure that will lead to a release through a 10 mm hole (the
fragment projection is not evaluated), which subsequently can
develop into a pool or ash re. If an explosion occurs after the
G.1 release in Tank 1, Tank 2 will be severely damaged. This
will lead to a G.1 release, which can then develop into another
explosion, ash re or pool re. The sequences involved in
the instantaneous release of the contents of Tank 2 can be
described in a similar way (Fig. 4).
428
Jet re
Flash re
Pool re
Tank 1
Tank parameters
Substance
Type of tank
Volume (m3 )
Mass (kg)
Containment size (m)
Pressure (bar)
Temperature (K)
Ambient parameters
Temperature (K)
Wind velocity (m/s)
dj: jet length; dp: pool diameter; d: distance between tanks; tj: duration of jet re; tp: duration of pool re.
Atmospheric tanks
Pressurized tanks
Consequence
Threshold values
Effect
Accident
Propylene
Pressurized/
sphere
5450
2,424,700
39 38
9.5
271
Tank 2
Ethylene
Atmospheric/
vertical
9775
4,718,500
58 65
0.1
170
283.15
4.5
429
LOCs
Tank 1
Accident
Tank2
G.1
5 107
5 106
G.2
5 107
5 106
G.3
1 105
1 104
Probability of occurrence
Tank 1
BLEVE
Explosion
Flash re
Pool re
Explosion
Flash re
Pool re
Flash re
Pool re
2.45 10
1.62 102
4.05 102
0
1.2 102
0.3 101
7.0 101
0.8 101
2.0 101
Tank 2
0
1.2 102
0.3 101
7.0 101
1.2 102
0.3 101
7.0 101
0.8 101
2.0 101
430
Accident
Frequency (y1 )
No domino effect
Tank 1
G.1
G.2
G.3
BLEVE
Explosion
Flash re
Pool re
Explosion
Flash re
Pool re
Flash re
Pool re
1.23 10
8.10 109
2.03 108
0
6.00 109
1.50 108
3.50 107
8.00 107
2.00 106
Frequency (y1 )
with FPS
Tank 2
0
6.00 108
1.50 107
3.50 106
6.00 108
1.50 107
3.50 106
8.00 106
2.00 105
Tank 1
8
1.52 10
1.00 108
2.51 108
0
1.02 108
2.55 108
5.95 107
8.00 107
2.00 106
Frequency (y1 )
without FPS
Tank 2
0
6.02 108
1.50 107
3.51 106
6.02 108
1.51 107
3.51 106
8.00 106
2.00 105
Tank 1
7
3.03 10
5.9 108
1.48 107
0
9.00 108
2.25 107
5.25 106
8.00 107
2.00 106
Tank 2
0
6.06 108
1.52 107
3.54 106
6.42 108
1.61 107
3.74 106
8.04 106
2.01 105
results, in which the variation between the results of the common QRA procedure and those obtained using the domino
effect model is minimal, unless comparisons are made with
the curve obtained for the domino effect without the FPS. In
this case, the variance between the results is dramatic. This is
crucial when we consider that the probability of safety measures such as an FPS failing may increase due to the chain
of accidents in a domino effect, leading to a dangerous situation for which there is no protection. Consequently, safety
measures in process plants should be designed to withstand
major accidents, so that they are also effective in the event of
a domino effect.
A comparison of the three scenarios reveals considerable
differences in the 1e-8 curve. The difference between the rst
two is 100 m, while the curve for the third scenario is at a
distance of approximately 600 m from the second one. Once
again, the major difference appears when the domino effect
is taken into account without the protection of the FPS.
Fig. 9 shows the comparison of F-N societal risk curves for
each of the scenarios. It can be seen that only the frequency
factor of the curves is altered. This point is further discussed
in Section 4.
431
3.2.
432
Type of tank
Mass (kg)
Containment size (m)
Pressure (bar)
Temperature (K)
Wind velocity (m/s)
Tank
1
3.17 107
117 76
7.25 107
117 135
8.77 107
117 154
4
Floating roof
8.75 107
117 154
1.01
288.15
4.5
5
8.31 107
117 155
6
3.72 107
117 71
433
539
519
368
190
63
395
367
217
38
300
239
90
138
59
78
G.1
5 106
G.2
5 106
G.3
1 104
Accident
Probability of occurrence
Explosion
Flash re
Pool re
Explosion
Flash re
Pool re
Flash re
Pool re
With FPS
Without FPS
1.87 103
4.68 103
3.25 103
1.87 103
4.68 103
3.25 103
4.68 103
3.25 103
3.74 102
9.35 102
6.50 102
3.74 102
9.35 102
6.50 102
9.35 102
6.50 102
f (y1 ) No domino
f (1 107 y1 ) with domino with FPS
Tank
All
1
2
3
4
5
6
1
2
3
4
5
6
G.1
G.2
G.3
Exp.
Flash
Pool
Exp.
Flash
Pool
Flash
1.87
2.28
2.35
2.42
2.14
2.28
2.21
2.29
2.38
2.43
2.17
2.30
2.24
4.68
5.72
5.89
6.07
5.37
5.72
5.54
5.74
5.96
6.09
5.44
5.76
5.61
3.25
3.97
4.10
4.22
3.73
3.97
3.85
3.99
4.14
4.23
3.78
4.01
3.90
1.87
1.96
2.03
1.91
2.17
1.90
2.03
2.49
2.53
2.87
2.71
2.66
2.53
4.68
4.91
5.08
4.79
5.43
4.77
5.08
6.23
6.33
7.17
6.79
6.65
6.32
3.25
3.41
3.53
3.33
3.78
3.31
3.53
4.33
4.40
4.99
4.72
4.62
4.39
93.5
93.6
93.5
93.6
94.2
93.5
93.5
93.6
94.7
93.5
96.0
93.5
93.5
Pool
65
65.2
65.0
65.1
65.5
65.0
65.0
65.1
65.8
65.0
66.7
65.0
65.0
434
4.
Discussion
435
436
the tank that suffers the release and possible ignition points
are not taken into account. If an accident is caused by an adjacent re, the probability of immediate ignition will generally
be higher than that used in the model. However, this could be
addressed easily if data were available, by developing event
trees with new probabilities and changing the frequencies of
accidents caused by re.
Finally, a comparison can be made between the results presented in this paper and those obtained by other authors,
such as Cozzani and Abassi. However, this comparison is only
supercial, as the methodologies vary in their treatment of
sequences and accidents, and the case studies can be very different in scope. Nevertheless, the results from all the methods
clearly show that the domino effect has a signicant impact
on the risk associated with a facility and should be considered
when risk analysis studies are carried out.
5.
Acknowledgements
Conclusions
The authors would like to thank Mr. Vicenc Espejo (TIPs) for
the help and information provided to complete this study, and
the Spanish Ministry of Education and Science for funding
provided under project CTQ2011-27285.
References
Abdolhamidzadeh, B., Abbasi, T., Rashtchian, D., Abassi, S.A.,
2010. A new method for assessing domino effect in chemical
process industry. J. Hazard. Mater. 182, 416426.
Antonioni, G., Spadoni, G., Cozzani, V., 2009. Application of
domino effect quantitative risk assessment to an extended
industrial area. J. Loss Prevent. Proc. 22, 614624.
Cozzani, V., Salzano, E., 2004a. Threshold values for domino
effects caused by blast wave interaction with process
equipment. J. Loss Prevent. Proc. 17, 437447.
Cozzani, V., Salzano, E., 2004b. The quantitative assessment of
domino effects caused by overpressure. Part I. Probit models.
J. Hazard. Mater. 107, 6780.
Cozzani, V., Gubinelli, G., Antonioni, G., Spadoni, G., Zanelli, S.,
2005. The assessment of risk caused by domino effect in
quantitative area risk analysis. J. Hazard. Mater. A 127, 1430.
Cozzani, V., Gubinelli, G., Salzano, E., 2006. Escalation thresholds
in the assessment of domino accidental events. J. Hazard.
Mater. 129, 121.
CPR18E, 2005. Guideline for Quantitative Risk Assessment (Purple
Book). Directorate General for Social Affairs and Employment,
TNO, The Hague, Netherlands.
Darbra, R.M., Palacios, A., Casal, J., 2010. Domino effect in
chemical accidents: main features and accident sequences. J.
Hazard. Mater. 183, 565573.
Delvosalle, C., 1996. Domino effects phenomena: denition,
overview and classication. In: First European Seminar on
Domino Effects, Leuven, Belgium.
Gmez-Mares, M., Zrate, L., Casal, J., 2008. Jet res and the
domino effect. Fire Safety J. 43, 583588.
Khan, F., Abassi, S.A., 1998. DOMIFFECT (DOMIno eFFECT):
user-friendly software for domino effect analysis. Environ.
Model. Softw. 13, 163177.
Khan, F.I., Abassi, S.A., 2001. Estimation of probabilities and likely
consequences of a chain of accidents (domino effect) in
Manali Industrial Complex. J. Clean. Prod. 9, 493508.
437