Journal of Loss Prevention in the Process Industries 32 (2014) 30e43

Contents lists available at ScienceDirect

Journal of Loss Prevention in the Process Industries

journal homepage: www.elsevier.com/locate/jlp

Boiling liquid expanding vapour explosions (BLEVEs): A brief review

Rolf K. Eckhoff
University of Bergen, Dept. Physics and Technology, Bergen, Norway

a r t i c l e i n f o a b s t r a c t

Article history: BLEVEs remain a complex multi-faced phenomenon. The point of departure of the paper is the extensive
Received 29 January 2013 review by Abbasi and Abbasi (2007), and the paper starts with presenting some deviating definitions and
Received in revised form suggested characteristics of BLEVEs. Then a step-by-step description of the chain of events in typical
24 June 2014
BLEVEs follows. In most cases the first step will be that a vessel containing a gas that has been liquefied
Accepted 24 June 2014
by pressurization (PLG) gets accidentally exposed to heat (most often a fire). The next step is that the
Available online 7 August 2014
vessel suddenly fails due to the increased internal vapour pressure and the weakening of the vessel
material. The resulting very sudden depressurization of the hot liquid gives rise to intense, often volu-
Keywords:
BLEVE
metric, evaporation. The central role of the superheat limit temperature (SLT) of the liquid in the
Accidental explosions evaporation process is discussed. The very rapid evaporation in turn gives rise to powerful blast wave
Superheat limit temperature emission. If the evaporated liquid is combustible and gets ignited, a huge fireball will most often be
Vessel failure generated. Properties and effects of fireballs from BLEVEs are discussed. The shattering of the initial
Process safety vessel may cause propelling of high-speed missiles, which when hitting other PLG vessels may give rise
to secondary BLEVEs. The remaining part of the combustible liquid from a BLEVEed vessel that has not
evaporated will, when ignited, give rise to pool fires. Emission of toxic gases/vapours requires special
actions. However, BLEVEs can also occur with non-toxic gases, and even in the absence of an initial fire.
The induction time preceding a BLEVE is an important parameter, e.g. for the time available for rescue
operations. Long-duration BLEVEs require special concerns. A part of the paper is devoted to various
categories of BLEVE case histories. Available means for preventing process situations that can lead to
BLEVEs are then discussed. If PLG-vessels get engulfed in fire measures for preventing catastrophic vessel
failure may be activated. Finally measures for mitigating the consequences if BLEVEs do occur in spite of
taking preventive measures are discussed.
© 2014 Elsevier Ltd. All rights reserved.

1. Definitions and characteristics of BLEVEs failure was defined as ‘the sudden opening of a tank/container to
release its contents nearly instantaneously.’
According to the extensive review by Abbasi and Abbasi (2007) The definition of a BLEVE presented by Center for Chemical
the acronym BLEVE was probably first introduced in 1957 by Smith, Process Safety (1999) is ‘a sudden release of a large mass of pres-
Marsh, and Walls at Factory Mutual Research Corp., USA. They had surized superheated liquid to the atmosphere’. The sudden release
analyzed the failure of a vessel that contained an overheated is due to a sudden containment failure caused by fire, a missile,
mixture of formalin and phenol, and suggested that the container corrosion, a manufacturing defect, internal overheating, etc.
suffered a ‘boiling liquid expanding vapour explosion’. Later Walls Berg, Voort, Weerheijm, and Versloot (2006) discussed the
(1978, 1979) defined a BLEVE as ‘a failure of a major liquid-filled discrepancies between various BLEVE definitions. Some confusion
container into two or more pieces at a moment when the tem- results from the fact that catastrophic failure of vessels containing
perature of the contained liquid is well above its boiling point at liquefied gases/vapours is often a result of a fire that has heated up
normal atmospheric pressure’. the vessel for some time. If the liquid is flammable, a sudden
Birk and Cunningham (1994) defined a BLEVE as ‘the explosive rupture of the vessel wall and an explosive ejection of the super-
release of expanding vapour and boiling liquid when a container heated liquid into the surroundings will in most cases immediately
holding a pressure-liquefied gas fails catastrophically’. Catastrophic put the ejected liquid on fire, and a large fireball will result.
Consequently the entire chain of events, including the formation of
the fireball, is sometimes defined as a BLEVE. It may seem as if this
E-mail addresses: [email protected], [email protected]. is the philosophy behind the illustration given in Fig. 1.

http://dx.doi.org/10.1016/j.jlp.2014.06.008
0950-4230/© 2014 Elsevier Ltd. All rights reserved.
 R.K. Eckhoff / Journal of Loss Prevention in the Process Industries 32 (2014) 30e43 31

Fig. 1. Illustration of a BLEVE, including the major secondary fireball following ignition of the ejected vapour/liquid. From Marshall (1987).

Berg et al., however, suggest that the definition of a BLEVE reaches the set pressure of the pressure relief valve of the vessel,
should not include a possible subsequent generation of a fireball, vapour from the liquid in the vessel is expelled into the open at-
but comprise only the explosive rupture of the pressure vessel and mosphere, and the liquid level in the vessel will drop as further
the subsequent flash evaporation of its superheated liquid content. liquid continues to evaporate. As long as most of the vessel volume
Some definitions of BLEVEs require that the liquid initially confined is occupied by the liquid, most of the vessel wall will be effectively
in the vessel has to be heated up to above its SLT at atmospheric cooled by the liquid. However, as the liquid continues to evaporate
pressure (see Section 4.2 below) before vessel rupture takes place. and escape via the pressure relief valve, the proportion of the vessel
Other, less strict, definitions only require that the liquid tempera- wall that is effectively cooled decreases. Eventually the portion of
ture at vessel rupture significantly exceeds the liquid boiling point the vessel wall that is not in contact with the liquid weakens due to
at atmospheric pressure. In any case, the higher the temperature of the temperature rise caused by the external fire load, and may fail.
the liquid at the moment of tank rupture the more severe the ac-
cident will be.
2.2. Sudden depressurization and evaporation of the hot liquid

2. The chain of events leading to BLEVEs

When the vessel fails, the liquid pressure suddenly drops to
atmospheric pressure. Because the liquid temperature is well above
The following summary of the various stages of a typical BLEVE
the atmospheric pressure boiling point, the liquid will evaporate
follows, to a significant extent, the same main lines as the very
abruptly and violently and expand to a cloud of a volume between
extensive review by Abbasi and Abbasi (2007).
several hundred and two thousand times that of the original liquid
volume. If the liquid temperature is even significantly higher than
2.1. A vessel containing gas liquefied by pressurization (PLG) gets its ‘superheat limit temperature’ (SLT) (see Section 4 below) at at-
accidentally exposed to heat (fire) mospheric pressure, most of the liquid volume can undergo
extremely fast and homogeneous nucleation and evaporation. In
A pressure vessel can fail even at normal ambient temperature that case the vapourization/expansion process will be extremely
by missile impact, fatigue, corrosion, or an accidental excessive rise fast and violent, and typically occur within 1 ms after the abrupt
of the process pressure. However, in the present context the main vessel depressurization caused by the sudden vessel failure.
focus will be on excessive pressurization due to accidental heating The nucleation and vapourization processes require heat, which
of the vessel from the outside. It is then important to keep in mind is taken from the liquid itself. Therefore, the liquid temperature will
that a pressure vessel is designed to withstand the relief valve set drop as the volumetric evaporation progresses. As soon as the
pressure only at the design temperature conditions, which may be liquid temperature drops below the SLT, the very rapid volumetric
just the ambient atmospheric temperature. nucleation and vapourization terminates and normal boiling at the
PLG is a substance that is in the gaseous state at normal ambient hot surfaces becomes the dominating process of vapour produc-
temperature and pressure, but which has been liquefied by tion. Hence, the more the liquid temperature, at the moment of
compression and is kept as a liquid at normal ambient temperature vessel failure, exceeds the SLT of the liquid, the greater will be the
in a pressure vessel. Propane and butane are common examples. In fraction of the liquid that can flash almost instantaneously to
the case of accidental heating of a vessel containing PLG, e.g. by the vapour.
heat from a fire, as illustrated in Fig. 1, the already elevated vapour It is important to emphasize that a BLEVE can occur even when
pressure inside the vessel will rise further. When the pressure the temperature of the suddenly depressurized liquid is below the
32 R.K. Eckhoff / Journal of Loss Prevention in the Process Industries 32 (2014) 30e43

SLT of the liquid at atmospheric pressure, but still significantly shattering vessels present a much greater danger of ‘domino’
higher than the atmospheric boiling point. However, clearly the effects than fireballs or blast waves. The bursting of the first
intensity of the evaporation will then be lower than if the initial vessel causes vessel fragments to be ejected into the surround-
liquid temperature exceeds the SLT at atmospheric pressure. ings. These missiles often damage other vessels that store PLG,
causing these vessels to undergo BLEVEs as well. This ‘domino’
effect actually took place in the Mexico City disaster in 1984 (see
2.3. Blast wave emission
Section 6) causing the largest number of lost lives ever recorded
in one single explosion/fire accident in the process industry.
Berg et al. (2006) discussed various models for calculating the
Large parts of vessels can fly as projectiles over long distances. In
strength of blast waves emitted from BLEVEs.
one accident (Port Newark), a large part of a spherical vessel
Acoustic modelling of blast wave emission requires full knowl-
travelled more than 800 m before hitting and destroying a petrol
edge of the source overpressure as a function of time. This, in turn,
bunker.
requires full knowledge of the physical characteristics of the liquid
According to Abbasi and Abbasi (2007) the likely consequence of
release as a function of time, from the onset of the crack develop-
a BLEVE series in terms of the duration and propagation of missiles
ment in the pressurized vessel. Crack development is a complex
depends on the following factors:
process that depends strongly on the design and structure of the
actual vessel. Therefore, simplifying assumptions are required. One
 likely number and mass distribution of missiles
such assumption is that the liquid release rate is growing linearly
 velocity and range distributions of missiles
with time. Using this assumption, Berg et al. considered as an
 likely directions of propagation of missiles
example a BLEVE from a 50 m3 propane tank in the open. They then
 penetrability and destructive potential of missiles
calculated the maximum overpressure at a distance 10 m away
from the tank, as a function of total release time Dt. With Dt equal to
2.5. Generation and development of fireballs
0.05, 0.2 and 1.0 s respectively the corresponding maximum over-
pressures Dp were 1.06, 0.07 and 0.0003 bar.
If the liquid involved in the BLEVE is neither combustible nor
Gas dynamic modelling gives more accurate results provided
toxic, as is the case with water, the pressure wave and the missiles
sufficient input information is available for defining the actual
from the vessel shattering are the only hazardous effects of the
scenario. Also, good experimental data are required for adequate
explosion.
validation of the models. As an example Berg et al. discussed the
However, if the liquid is flammable, as is often the case, the
most valuable and controlled BLEVE experiments by Giesbrecht,
mixture of liquid/gas released by the explosion catches fire
Hess, Leukel, and Maurer (1981), who used vessels of volumes
causing a fireball to be generated. Analyses of actual BLEVEs
ranging from 0.226 to 1000 L, containing liquefied propylene. The
have shown that more than two-thirds involved flammable
vessels, located at ground level, were heated to 50e80
C i.e. to
chemicals.
40 bar overpressure and deliberately ruptured more or less
The shape and size of the fireball, and the heat load produced
instantaneously by means of a small high-explosive charge. Fig. 2
by it, depend on several factors. At the outset the expelled mass
shows the experimentally determined correlation between the
of fuel can burn only at its periphery because there is not suf-
maximum overpressure in the blast wave emitted into the sur-
ficient air inside the mass, i.e. the fuel/air mixture ratio is above
roundings from a BLEVE and the scaled distance from the BLEVE,
the upper flammability limit. Furthermore, not all the fuel
together with the correlation obtained by Berg et al. (2006) by
initially contained in the tank may become involved in the pri-
numerical simulation.
mary fire. Some of the liquid may leak to the surroundings
before the explosion via a crack or other opening in the vessel.
2.4. Shattering of secondary vessels due to missiles from the Alternatively some of the liquid may be entrained in the wake
shattering of the first vessel formed by flying fragments. In the Mexico City disaster in 1984
(see Section 6) fragments of shattered vessels carried with them
The shattering of the first vessel in a series of BLEVEs is nor- parts of the flammable liquid, which in itself caused the fires to
mally due to heating by fire exposure. However, missiles from spread.
As the main fireball from a BLEVE grows, the turbulence of the
flame entrains air into the fireball. At the same time thermal radi-
ation from the flame vaporizes liquid droplets and heats the
combustible cloud. As a result of all these processes, the entire
cloud increases in volume, rises, and attains an approximately
spherical shape. Depending on the amount of combustible material
involved, such fireballs can become very large and cause very
intense thermal radiation. The size, the lifetime, and radiation in-
tensity of a fireball may also depend on the initial temperature of
the liquid originally contained in the vessel. Whether the loss of
confinement of the liquid occurred whilst the pressure inside the
vessel was still rising is a further factor. Whereas BLEVE fireballs are
spherical when fully developed, they acquire a mushroom-like
shape during lift-off.
Fireballs resulting from two-step BLEVEs (see definition later in
this section) may be approximately ellipsoidal in shape.
According to Lees (1996) it may happen that BLEVEs involving
combustible materials are not ignited at the release point, but at a
Fig. 2. Experimental and computed blast overpressures originating from exploding later stage. In such cases the resulting event may either be a major
vessels of various sized containing propylene. From Berg et al. (2006). flash fire or a vapour cloud explosion.
 R.K. Eckhoff / Journal of Loss Prevention in the Process Industries 32 (2014) 30e43 33

2.6. Pool fires (1988) found that 7e11 min, 25e38 min, and 6e7 min were
likely induction times in the case total fire engulfment, partial fire
Some of the liquid expelled from the shattered vessel may be engulfment, and jet flame impingement, respectively. According to
splashed and hit the ground nearby forming short-lived pools Selway these times would be shorter if the vessel would not be
before vaporizing. In the case of flammable liquids this may give completely filled with liquid.
rise to pool fires. In both fire-induced and missile-collision-induced initial vessel
damage, the first tear or crack may or may not propagate straight
2.7. Emission of toxic gases/vapours away to a sufficient size to cause a BLEVE instantaneously. In some
cases an initial tear or crack that is too small to initiate a BLEVE
BLEVEs giving rise to emission of toxic substances have occurred directly, may restart propagation again after some time and in-
with a number of toxic compounds, such as ammonia, chlorine, crease to the size required to cause a BLEVE. Typical crack propa-
chloro-butadiene, and phosgene. Of the one third of recorded gation speeds are then >200 m/s. About 20% of the BLEVEs
BLEVEs that did not involve flammable liquids, the majority observed by Birk and Cunningham (1994) were of this long-
involved toxic gases. Of these, chlorine (14%), ammonia (10%), and duration type. This two-stage process is illustrated diagrammati-
phosgene (2%) account for 76% of all BLEVEs involving non- cally in Fig. 3. Such events may occur with tanks that are almost
flammables. With such chemicals, the fatalities were caused by strong enough to resist a total loss of containment (TLOC), but are
the toxic material expelled with the blast wave, and also by mis- 'pushed over the edge' by the violent boiling of the
siles. According to Abbasi and Abbasi (2007) chlorine accounted for depressurization-induced superheated liquid. The occurrence of
most of the fatalities in the major BLEVEs with non-combustible long-duration BLEVEs is strongly related to the setting of the
but toxic materials during 1926e1981, followed by ammonia. pressure relief valve (PRV), because this ultimately determines the
liquid temperature.
2.8. Emission of ‘cold’ non-toxic gases Birk and Cunningham (1994, 1996) conducted a series of ex-
periments with 400 L propane tanks, subjecting the tanks to fire
BLEVEs have occurred with tanks containing carbon dioxide and engulfment and studying the pattern and the duration of vessel
water, even in the absence of a fire. Abbasi and Abbasi (2007) failure. The tanks were equipped with pressure relief valves (PRVs).
suggested that many boiler explosions, which are far more com- Table 1 summarizes some results obtained by Birk and Cunningham
mon than explosions involving other chemicals, are in fact BLEVEs, (1994).
without being commonly acknowledged and classified as such. In general it is not possible to forecast how much time a vessel
Indeed if all boiler explosions that occur with superheated water exposed to a fire may take before undergoing a BLEVE. This can vary
were to be included in the BLEVE statistics, it might well be that the from a few seconds to several hours. This fact makes it very
most frequently involved liquid in BLEVEs turns out to be water.

2.9. Potential consequences of BLEVEs

The energy of explosion, or ‘burst energy’, determines the

severity of the blast wave generated by the BLEVE and the velocity
(hence the range and the penetration) of the shattered vessel
fragments. The mode of release of the PLG in the shattered vessel
determines the size, duration, and heat flux of the fireball, if the PLG
is flammable, or the pattern of atmospheric dispersion if the PLG is
toxic. There is much greater degree of uncertainty associated with
predicting the consequences of BLEVEs than with predicting con-
sequences of gas/vapour cloud explosions. This is because of the
central roles played by superheated liquids and pressurized gases.

3. Induction time preceding a BLEVE. Long-duration BLEVEs

Experience from accident histories has shown that some vessels

exploded within a few minutes of fire engulfment or missile hit,
whereas others did not explode until after several hours. In some
cases the induction time from fire engulfment or missile hit to
vessel explosion was up to 24 h. In the Feyzin accident (see Section
6) the induction time was about 90 min. In the Mexico City catas-
trophe (see Section 6) the induction times varied between 3 and
30 min.
Clearly having a good estimate of induction times in potential
BLEVEs is important for optimization of damage-control. Abbasi
and Abbasi (2007) discussed a number of investigations aimed at
quantifying induction times. For example, Blything and Reeves
(1988) analyzed BLEVEs from horizontal cylinders filled to 75% of
their capacity with liquid butane. The fire load was either by partial
fire engulfment or by jet flame impingement. From their analysis
they concluded that a BLEVE would occur with induction times
between 4 and 48 min. In another investigation considering a full Fig. 3. Chain of events in a two-step BLEVE. From Abbasi and Abbasi (2007) based on
2000 tonnes PLG storage sphere being exposed to flames Selway Birk and Cunningham (1994) and Birk and Steen (2002).
34 R.K. Eckhoff / Journal of Loss Prevention in the Process Industries 32 (2014) 30e43

Table 1
Summary of experimental results obtained by Birk and Cunningham (1994).

Test Fire conditions No. of vents Outcome Comments

Vent to empty Pool only 5 Entire tank contents Fire conditions were not
Pool and torch 2 lost through the PRV severe enough to initiate
a local failure.
Partial failure Pool and torch 7 Partial failure occurred Local thermal weakening
in the tank causing a was such that pressure
two-phase liquid/vapour stresses in the wall
mixture to be released exceeded local tank wall
from the tank. strength, causing a failure
Torch only 4 In two cases the jet was (crack). The tank wall
angled such that it was strong enough to
propelled the tank up arrest the crack
to 30 m
TLOC and BLEVE Pool and torch 9 The tank failed Not all BLEVEs were the
catastrophically releasing same
the entire contents as
expanding vapour,
boiling liquid and
dispersed droplets.
Torch only 2 Failures resulted in The characteristics of
blasts and fireballs BLEVEs varied with tank
properties and lading
conditions at the time
of failure

dangerous for fire fighters to operate close to a fire-engulfed vessel significantly exceeds SLT at the ambient pressure. When a vessel
containing a pressurized liquefied gas/vapour. There have been containing a pressure-liquefied gas that has been excessively
many instances when a vessel has exploded even after the heated, e.g. by a fire, suddenly ruptures as it will do in a BLEVE,
pressure-relief valves have been venting the vessel for several mi- equilibrium is lost as the liquid momentarily becomes excessively
nutes. Of all the harmful effects of BLEVEs the one with the greatest superheated, i.e. its temperature is now way above its SLT at at-
impact is rocketing fragments. Often deaths of people quite a dis- mospheric pressure. In such a situation a fraction of the liquid will
tance away from the site of the BLEVE, and secondary accidents in evaporate practically instantaneously, carrying with it liquid
other process units, are caused by such missiles. droplets. This occurs extremely fast, within a few milliseconds. The
increase in volume caused by this instantaneous vapourization is
4. The role of superheat limit temperature (SLT) of the liquid enormous, and together with the rapid expansion of the com-
in BLEVE development pressed vapour existing in the vessel even before the rupture, this
generates a strong pressure wave with a significant damage
4.1. Boiling and boiling point (BP) of liquids potential.
Clearly, if the liquid is combustible and its vapour and spray gets
It is essential to stress that the SLT of a liquid is quite different ignited immediately after a BLEVE has taken place, the accident
from its BP. The basic definition of the BP of a liquid is the tem- potential of the entire event will escalate substantially.
perature at which the vapour pressure of the liquid equals the From what has been said above, it is clear that the difference
ambient pressure above the liquid surface. Hence, a given liquid in between the actual temperature of the superheated liquid at the
an evacuated environment has a lower boiling point than when the moment of vessel burst, and the SLT of the liquid at atmospheric
same liquid is at atmospheric pressure. If the liquid is in an envi- pressure is of prime importance with regard to the extent to which
ronment of a pressure above atmospheric, it will have a higher explosive evaporation will take place. The SLT provides a basis for
boiling point than at atmospheric pressure. developing means of predicting both likelihood of possible BLEVEs
In order to distinguish between BP and SLT it is important to and the severity of their consequences. A logical safety strategy that
note that boiling of a liquid at its BP can only take place if the liquid makes direct use of the SLT, is to operate pressure relief valves of
is in contact with “active points” at the hot surface heating up the PLG vessels in such a manner that the vessels are depressurized
liquid. Such points can be e.g. sharp edges, scratches or pores in the before the vessel contents reaches the SLT.
surface, or small solid particles adhering to the surface. Only in this
way the microscopic vapour nuclei initially formed at BP will be 4.3. Experimental determination of SLT
able to grow to the critical size beyond which they will continue to
grow spontaneously. The droplet explosion method is the most common and most
reliable experimental means of determining SLT. The apparatus is
4.2. The superheat limit temperature (SLT) in relation to BLEVEs illustrated in Fig. 4.
A small droplet of the test liquid is introduced at the bottom of a
A pure liquid, with no “active points”, can be heated to a tem- vertical tube that is filled with a host liquid of a higher density than
perature considerably higher than its boiling point at the prevailing the liquid to be tested. This will ensure that the droplet will rise in
pressure before significant boiling starts. This temperature is called the host liquid. Also, the host liquid must be immiscible with the
the “superheat limit temperature” (SLT), which is synonymous with test liquid, and should have a boiling point well above the SLT of the
the “homogeneous nucleation temperature”. liquid to be tested. By means of a specially designed heating system,
It follows from what has been said above that explosive evap- a vertical temperature gradient is established in the host liquid
oration involving spontaneous boiling throughout the entire liquid column, the lowest temperature being at the column bottom and
volume can only take place as long as the liquid temperature the highest at the top. Therefore, as the test droplet rises through
 R.K. Eckhoff / Journal of Loss Prevention in the Process Industries 32 (2014) 30e43 35

Table 2
Comparison of experimental boiling points and critical temperatures, and various
calculated SLTs for some substances based on a table in Salla et al. (2006).

Substance Boiling point SLT (K) Critical temp.

(K) (K)

Methane 112 160e177 191

Ethane 185 259e279 305
Ethylene 169 239e261 282
Propane 231 313e332 370
Propylene 226 309e328 365
n-Butane 273 360e382 425
n-Pentane 309 379e422 470
n-Hexane 342 407e456 508
n-Heptane 372 430e484 540
n-Octane 399 467e510 569
Cyclo-hexane 354 429e497 554
Benzene 353 453e504 563
Toluene 384 501e531 592
Water 273 547e607 647
Ammonia 240 343e377 406
Chlorine 239 353e378 417
Carbon dioxide 195 257e294 304
Hydrogen 20 29e33 33
Nitrogen 77 107e119 126
Oxygen 90 131e145 155

must be faster than the dynamics of the heating of the wire. The
nature of the wire surface also influences nucleation. This method
also gives fairly reproducible results, but it is considered less ac-
curate than the rising-bubble method.
In Table 2, from Salla, Demichela, and Casal (2006), the boiling
points, the SLTs and the critical temperatures are given for a se-
lection of substances. Table 3 is based on data from Wakeshima and
Fig. 4. Droplet explosion apparatus for determining superheat limit temperatures Takata (1958). The theoretical SLT values were calculated using the
(SLTs) of liquids. From Partrick-Yeboah and Reid (1981). classical Do€ring theory.

4.4. The role of SLT in BLEVE theories

the column its temperature will increase gradually. This continues
As pointed out by Abbasi and Abbasi (2007), several complex
even beyond the boiling point, because the liquid in the droplet is
questions related to BLEVEs call for theories that can predict both
not in touch with any active points (e.g. solid surface) and no
the development and consequences of such events in given prac-
boiling will take place. However, as soon as the temperature rea-
tical situations.
ches the SLT the droplet will evaporate explosively. This tempera-
In most theories on the development of BLEVEs the SLT plays a
ture is known from the position of the droplets at the moment of
central role. Clearly, this is because the explosive volumetric boiling
explosive evaporation. The temperature in the column as a function
of liquids at temperatures above the SLT at atmospheric pressure is
of vertical position is monitored by means of a thermocouple that
so much more violent than boiling below SLT. According to Nesis
can be positioned at any desired point in the column.
(1966) the first major contributions to a basic, theoretical under-
One problem inherent in the droplet explosion technique is that
standing of superheating was provided by the pioneer J.W. Gibbs in
the initial unstable bubble nuclei being formed when the temper-
the last part of the 19th century. Gibbs explained why, in the
ature approaches SLT, have sizes on the molecular scale. It is likely,
absence of the “active points” required for boiling at the BP, it is
therefore, that even in the most careful experiments, microscopic
possible to heat a liquid up to a temperature above its BP before
temperature gradients in the liquid could initiate formation of
boiling starts. The reason is that the liquid has to cross over a
bubbles at a global liquid temperature significantly below the su-
certain thermodynamic potential energy barrier (“activation en-
perheat limit. Another concern is whether the droplet temperature
ergy”) before it can transit into a stable vapour state. When a
is actually equal to that of the surrounding liquid at any time during
microscopic vapour bubble is first produced inside a liquid at a
the rise of the droplet. The temperature difference will depend on
temperature above its BP, the thermodynamic potential energy of
the droplet size, the droplet rise velocity, the steepness of the
the system will change with time as a result of two opposing
vertical temperature gradient in the host liquid column, and the
heat conductivity of the droplet liquid. Nevertheless, the droplet
explosion technique is the best method currently available for Table 3
experimental determination of SLT. Experimental boiling points and experimental and theoretical SLT values based on a
table in Wakeshima and Takata (1958).
Another experimental method for determining SLT is based on
immersion of very thin and rapidly electrically heated wires in a Substance Boiling point SLTeksp. SLTtheor.
very clean test liquid. The temperature of the wire is recorded (K) (K) (K)
continuously as the wire is heated up. As soon as the wire tem- n-Pentane 309 383 383
perature has reached the SLT of the liquid, the liquid layer adjacent n-Hexane 342 386 385
to the wire will vaporize explosively. This approach requires that n-Heptane 371 386 387
Cyclo-hexane 354 408 411
the dynamics of the thermocouple/temperature measuring device
36 R.K. Eckhoff / Journal of Loss Prevention in the Process Industries 32 (2014) 30e43

simultaneous processes. The potential energy will increase because Whereas fireballs from BLEVEs are predominantly influenced by
surface energy is accumulated at the interface between the liquid momentum forces, fireballs generated when more stagnant vapour
and the vapour bubble. However, when the unstable molecules in clouds are ignited are mostly moved by buoyancy forces.
the liquid state transit to the more stable vapour state, the potential In some rare cases a vessel containing a flammable PLG may first
energy will decrease. For very small bubbles the first effect is release a sufficient mass of vapour to generate a vapour cloud that
dominating, whereas the latter effect takes over as the bubbles gets ignited before the vessel fails in a BLEVE that generates a much
grow. Therefore, there is a given critical bubble size at which the bigger fireball. According to Makhviladze, Roberts, and Yakush
thermodynamic potential of the system reaches a maximum. As (1999), the fireball from a BLEVE release of 100 tonnes of flam-
soon as the bubble has grown beyond this critical size, it will grow mable liquid develops 5  1012 J of thermal energy within 10e20 s.
further spontaneously. The minimum liquid temperature that al- About one fourth of this energy appears as thermal radiation, which
lows this critical bubble size to be reached is per definition the SLT is powerful enough to scorch people, damage property, and trigger
of the liquid. Just as with the BP, the SLT also varies with the secondary fires.
pressure of the environment in which the liquid is kept. However, Frame by frame analysis of full-scale fireball records conducted
in the context of BLEVEs, the SLT for ambient atmospheric pressure by Crawley (1982) show that the fireball passes through three
is of prime importance and significance. Hence, when the acronym phases, viz. growth, steady burning, and burn-out. The growth
SLT is used in the following, it generally means SLT at ambient at- phase comprises two time intervals, each of about 1 s. During the
mospheric pressure. first interval the fireball grows to about half its final diameter, and
the fireball boundary is bright yellowish-white, indicating a flame
4.4.1. SLT is not an absolute lower temperature limit for BLEVEs to temperature of about 1300
C. In the second time span of the first
occur phase, the fireball attains its maximum volume, but about 10% of
The appreciable amount of research conducted by Birk and co- the surface is now dark and sooty, whereas the major 90% is white,
workers has revealed that the SLT is not an absolute lower tem- yellowish-orange or light red, indicating flame temperatures in the
perature limit for a BLEVE to occur. Depending on the circum- range 900e1300
C, with an estimated effective average flame
stances leading to vessel rupture, BLEVEs can occur even at temperature of 1100e1200
C.
temperatures significantly below the SLT. It is clear, however, that In the second steady burning phase, which lasts for some 10 s,
the blast overpressures generated by such BLEVEs will be consid- the fireball, which is now roughly spherical, is no longer growing.
erably lower than those produced by BLEVEs occurring above the At the start of this phase it begins to lift off, raises and attains the
SLT. characteristic mushroom shape. The estimated average effective
flame temperature remains at 1100e1200
C.
In the third burn-out phase, which typically lasts for about 5 s,
4.4.2. Flammable liquids giving rise to fireballs
the fireball size remains constant, but the flame becomes less sooty
Furthermore, the size, shape, and radiation intensity of the
and more translucent.
fireballs which are formed when flammable liquids undergo
According to Abbasi and Abbasi (2007) the following issues
BLEVEs do not appear to be directly related to the extent to which
must be addressed to enable prediction of the size and duration of a
the liquid is superheated.
fireball, and the thermal radiation from it:

4.4.3. Boiling liquid collapsed bubble explosion (BLCBE) theory  mass of flammable substance released in the BLEVE
According to Abbasi and Abbasi (2007) the understanding of the  mass of flammable substance consumed in the fireball
BLEVE process has been further refined by careful analysis of some  fireball development as a function of time
more recent BLEVE accidents, followed up by controlled experi-  fireball full size and duration
ments. One conclusion is that all BLEVEs are in fact the result of a  radiation heat load generated
more or less universal two-stage process. The first step is the for-  view factor
mation of a limited crack in the vessel wall, giving rise to a mod-
erate initial leak (“leak before break”). In the second step there will Bosch and Weterings (1997) presented the following list of
be waves of repeated depressurization and re-pressurization items that need to be quantified in order to enable estimation of the
caused by further crack propagation and evaporation and ejection thermal radiation impact of a BLEVE fireball:
of fluid, leading to the final major failure. The delay time between
crack initiation and catastrophic failure depends on how much of  amount of flammable material released in BLEVE
the vessel volume is initially filled with liquid. A typical delay time  heat generated by the fireball
for 20% filling is 40 s and for 85% filling 1.4 s.  fireball radius
 fireball duration
5. Properties and effects of fireballs from BLEVEs  fireball lift-off height
 radiated heat flux
Bosch and Weterings (1997) defined a fireball as a fire that  net available heat for thermal radiation
burns with sufficient rapidity for the burning mass to raise into the  radiation absorption by water vapour between fireball and point
air as a cloud or ball. In all BLEVEs involving combustible liquids of impact
there is an almost instantaneous two-phase release of most of the  radiation absorption by carbon dioxide between fireball and
liquid which immediately auto-ignites to form the fireball. There- point of impact
fore the fireball is more or less an inevitable consequence when-  atmospheric transmissivity
ever a vessel containing a flammable liquid suffers a BLEVE. For  distance from centre of fireball to actual point of impact
this reason the fireball is often considered as an inherent part of  “view factor” at point of impact
any BLEVE. However, about one-fifth of all BLEVEs occur with non-
flammable liquids (including fire suppressants like nitrogen, car- Bosch and Weterings (1997) defined the “view factor” as the
bon dioxide, and water), and obviously no fireball is generated in ratio between the received and the emitted radiation energy per
such cases. unit area, or the fraction of the fireball that is “seen” by the target at
 R.K. Eckhoff / Journal of Loss Prevention in the Process Industries 32 (2014) 30e43 37

the point of impact. The view factor incorporates the orientation of Access to the valve area was difficult because the valves were
the object relative to the fireball, as well as its distance from the immediately below the tank bottom, which was only 1.4 m
fireball centre. above the ground.
As is the case with predicting the other BLEVE impact parame-  The propane leak was now completely out of control and soon
ters, considerable uncertainty is also to be expected in the esti- formed a large visible 1 m thick vapour/mist cloud, which
mation of the thermal parameters of fireballs. These problems spread out across the ground as a layer of radius of 150 m.
appear right from the initial assessment of the amount of  25 min after the leak had started the large vapour cloud was
combustible material (the “flashing fraction”) that is released ignited by an automobile that had stopped on a nearby road. The
instantaneously on vessel failure. Bosch and Weterings (1997) as- fire flashed back to the storage tank and continued to burn there,
sume that the entire vessel content will flash over and contribute to but there was no immediate explosion.
the fireball, whereas Marshall (1987) and others assume that the  The tank was equipped with a water deluge system but the
fraction of the fuel that participates in the fireball is only about one water supply was inadequate for cooling the tank. Furthermore,
third of the total amount released in the BLEVE. as soon as the fire brigade began to use their hoses, the water
supply to the deluge system ran dry. Apparently, the firemen
6. BLEVE case histories, mainly based on Abbasi and Abbasi had used up the available water for cooling the neighbouring
(2007) tanks to prevent the fire from spreading to them. They may have
assumed that the tank on fire was protected sufficiently by the
6.1. BLEVEs in storage tank facilities opened relief valve.
 Ninety minutes after the fire started the tank BLEVEed.
6.1.1. Tank storage farm in Feyzin, France, in January 1966
This BLEVE catastrophe is one of the worst accidents involving Ten of the twelve firemen operating within 50 m from the tank
PLG that has ever occurred. The scene is illustrated in Fig. 5. The were killed. People 140 m away were badly burned by a wave of
present summary is based on the detailed account by To €ro
€k, Ajtai, propane which passed over the compound wall. Altogether 15e18
Turcu, and Ozunu (2011) and the review by Abbasi and Abbasi men were killed and about 80 injured. Flying debris broke the legs
(2007). The series of events leading to the catastrophe was as of an adjacent spherical tank which fell over. Its relief valve dis-
follows: charged combustible liquid which added to the fire, and 45 min
later even this tank BLEVEed. Altogether five spherical and two
 Three workers started to drain water from the bottom of tank other pressure vessels ruptured catastrophically whereas three
No. 4 in Fig. 5. This was a 1200 m3 spherical propane storage suffered less severe damage. The fire also spread to gasoline and
tank, which was nearly full of liquid propane. There was a sys- fuel oil tanks.
tem of three draining valves underneath the tank. The first valve
was closest to the tank bottom and its outlet pipe was split by a 6.1.2. PEMEX PLG terminal in Mexico City, in November 1984
Y-connection leading to two parallel valves further downstream. The large PEMEX PLG terminal in San Juan Ixhuatepec, Mexico
 In order to perform the draining, the valve closest to the tank City received supplies from three different gas refineries every day.
bottom was first opened, followed by the opening of one of the In the morning of the accident the vessels at the terminal were
two lower parallel valves. After some time the appearance of being filled with PLG arriving in a pipeline from a refinery 400 km
traces of oil in the drained water indicated that the draining was away. Then a drop in pipeline pressure occurred due to the rupture
nearly complete. Then the first valve was closed, and subse- of an 8 in. pipe that connected one of the storage spheres to a series
quently partly re-opened to complete the draining. In the mean of cylindrical tanks. However, the operators did not imagine such a
time the line had got blocked, presumably by hydrate, and no possibility and the release of the PLG from the leaking pipeline
liquid flow appeared. The first valve was then fully re-opened. continued for 5e10 min. This led to the major catastrophe illus-
The choke then suddenly cleared, and the operator and the trated in Fig. 6.
two other men were splashed with liquid propane. The handle The escaping gas formed a 2 m high cloud covering an area of
came off the valve and it was impossible to get it back in place. 200 m  150 m, which drifted towards a flare tower where it caught
The lower valve had become frozen and could not be closed. fire and prepared the ground for the first BLEVE in a series of
several. The explosion ejected vessel fragments, wrapped in
burning PLG, in all directions. Some of the fragments hit other
vessels which were also damaged. Other fragments initiated local
fires which engulfed other vessels.
Four spheres, each containing 1500 m3 of PLG, and several
smaller cylindrical tanks containing between 45 m3 and 270 m3
PLG each suffered BLEVEs. All the BLEVEs generated fireballs, which
raged through the streets of Ixhuatepec for about 90 min. A block of
perhaps 200 houses built mostly of wood, cardboard, and metal
sheets was demolished by these fireballs. Masses of fragments of
tanks and pipes weighing up to 40 tonnes, were blown into the air
and landed as far away as 1200 m.
The disaster caused 650 deaths and more than 6400 injuries.
The entire PEMEX terminal was devastated. The total damage was
estimated at more than 30 mill US$.

6.1.3. Boral PLG distribution depot in Sydney, Australia, in April

Fig. 5. Tank farm at Feyzin, France, consisting of four spherical tanks of capacity
1990
1200 m3 each, four spherical tanks of capacity 2000 m3 each, and two horizontal In the evening of 1 April 1990 a BLEVE occurred in the Boral PLG
€ ro
cylindrical tanks of capacity 150 m3 each. From To €k et al. (2011). storage/distribution terminal outside Sydney. This led to a series of
38 R.K. Eckhoff / Journal of Loss Prevention in the Process Industries 32 (2014) 30e43

Fig. 6. Illustration of site of the Mexico City BLEVE catastrophe. From Marshall (1987).

further BLEVEs throughout the night. The first event was an ex- engineers were able to open the relief safety valves of the surviving
plosion of a small gas tank. The resulting fire then spread along tanks to depressurize them.
ruptured gas pipes to the four main 100-tonnes steel PLG storage Several factory storage buildings close to the explosion sites
tanks containing at total of at least 40,000 L of PLG. The fire heated were destroyed. The buildings that survived had doors that were
up the tanks until the 15 cm thick steel walls failed and gave rise to thrown off their hinges, roofs that lifted and windows that shat-
very powerful BLEVEs. The resulting fireballs and gas flares tered. The Boral plant was built in 1968 to satisfy standards that
extended hundreds of metres into the sky. Hundreds of portable gas were outdated by the time of the explosion disaster in 1990.
cylinders of capacities from 2 to 240 kg and kept inside a storage
room at Boral, also BLEVEed. Power blackouts occurred after the 6.2. BLEVEs in LFG transportation facilities
first thundering explosion, which shattered windows. The shock
wave from one of the BLEVEs uprooted a telegraphic pole that was 6.2.1. Highway tunnel near Palermo, Italy, March 1996
shot away as a missile and nearly hit a woman who was standing A tank truck became involved in a car crash inside a highway
almost 500 m away from the exploding vessel. One of a number of tunnel. The initiating event was the skidding of a car inside the
30 m long cylinders blew off its mooring, and rocketed through the tunnel causing piling-up of cars behind it. The engine of one of the
air with a tail of flames. Upon landing it created a 2 m diameter cars caught fire as a result of a collision with another car. The tank
crater in the earth, before bouncing through a wire fence into three truck also entered the tunnel but had to stop about 50 m from the
40-tonne tanks, which were propelled into a nearby canal. The tunnel exit to avoid collision with the cars ahead. However, a bus
rocketing cylinder then hit and flattened an electrical substation just behind the tank truck arrived at high speed, skidded and
and a panel-beating workshop before diving into the canal 300 m crashed into the truck, causing a leakage in the upper part of the
away from its original position. Luckily it all happened on a Sunday. tank shell, just below the manhole. A few seconds later a “soft
The loss of human lives and number of injuries would have been rumble” was heard, followed by a “hot wind” that caused serious
substantially larger if the accident had occurred on a working day. burns to the people in the tunnel. Most of the people in the tunnel,
Also a favourable wind direction contributed to limiting the spread except for five persons who had fainted as a result of the crash,
of the fire. The chain of BLEVEs was eventually broken when Boral managed to run out of the tunnel. 4 min later the tank of the tank
 R.K. Eckhoff / Journal of Loss Prevention in the Process Industries 32 (2014) 30e43 39

truck BLEVEed. The resulting blast wave demolished the cars in the at a 3% gradient when its brakes failed. The train derailed at over
tunnel and killed the five persons who had been unable to escape. 80 km/h on a bend 350 m beyond Montana station, resulting in a
Dense black smoke was expelled from the tunnel exit. pile-up that included 28 of the 32 chlorine cars. Most were badly
damaged and suffered BLEVEs one after another. One tank car lost
6.2.2. Highway near Tivissa, Spain, June 2002 its dished end and the shell was propelled 2000 m. A second was
A tank vehicle carrying CNG (compressed natural gas) lost split along its side. A third had a 0.5 m diameter hole, probably the
control on a down-hill road and turned over. When it came to a halt result of an ironechlorine fire, which could well have resulted from
beside a sandy slope, flames appeared between the truck cabin and ignition of the cork insulation by red hot brakes. Four other tank
the trailer due to ignition of either leaking diesel, or CNG, or cars suffered damage to their valves, which were ripped off or
perhaps both. The fire increased in size, perhaps due to involve- dislodged so that they leaked. It is estimated that 100 tonnes of
ment of CNG escaping from the pressure relief valve. There was a chlorine escaped in the first few minutes and 300e350 tonnes in
small explosion, followed by a strong hissing sound. Then the tank all.
BLEVEed and the ejected CNG ignited, giving rise to a huge fireball. Seventeen persons died, 4 in the caboose of the train and 13
The accident occurred in a remote location, but it nevertheless led from gassing. Another 1000 people were impacted. The vegetation
to one casualty and burn injuries to two persons who were about up the valley was bleached by the gas cloud passing up it; there was
200 m away from the blast site. also discolouration some 50 m down the slope and up the sides for
a vertical distance of about 50 m. The highest concentrations
6.3. BLEVEs in stationary installations containing toxic but non- appear to have occurred in a strip 1000 m long  40 m wide.
flammable chemicals
6.5. BLEVEs involving only non-flammable and non-toxic chemicals
6.3.1. Phosgene storage tank, Hamburg, Germany, May 1928
A tank at a factory near the harbour area of Hamburg, containing 6.5.1. Liquid nitrogen storage vessel in Japan, August 1992
phosgene, BLEVEed and released about 12 tonnes of the deadly gas. The catastrophic failure of a storage vessel containing liquid
11 people were killed and 171 injured. People were affected by the nitrogen resulted in a BLEVE causing demolition of almost half of
gas at locations up nearly 20 km from the accident site. If the wind the factory, damaging the walls of 25 houses and destroying 39
had blown over populated areas, the fatalities would have been cars, buses and trucks, all within a 400 m radius of the original site
much higher. of the storage vessel. Fragments of the vessel, including part of the
150 cm wide and 8 mm thick top head of the outer shell were
6.3.2. Colourant manufacturing unit, Louisville, Kentucky, USA, ejected up to 350 m away from the original site. The estimated
April 2003 property loss was $5 million.
A tank containing a mixture of malto-dextrin and water was
unintentionally overheated. On two earlier occasions the tank had 6.5.2. Mihama Nuclear Power Reactor, Japan, August 2004
been deformed due to misapplication of vacuum and the repairs A large pipe carrying superheated water started to leak, and
had not been certified to meet standard requirements for structural exploded. The resulting two-phase release of superheated water
strength. Most probably this was the cause of the tank failure, as the and steam scorched 11 workers. Some of these lost their lives, the
internal overpressure at the time of vessel burst was less then the others were severely injured.
formal design pressure of the tank. The explosion propelled the top
head of the tank to about 100 m away from the original site, and 6.6. Initiating events leading to BLEVEs in case histories
one operator lost his life. The main tank shell was ejected off its
foundation and struck a 50,000 L liquid ammonia storage tank, 6.6.1. Relative frequencies of initiating events
which was knocked it sideways and ruptured, giving rise to a Abbasi and Abbasi (2007) were not able to trace any compre-
12,000 kg leak. The shell then ricocheted and hit the bottom of a hensive, systematic record of BLEVEs that have occurred world
five-storey spray drier, which toppled. wide. However, from available published evidence they deduced
the relative frequencies of the initiating events leading to 88 major
6.4. BLEVEs during transportation of toxic but non-flammable BLEVEs that occurred during the period 1926 to 2007. The results
chemicals are summarized in Table 4. The frequency distribution given in the
table refers to the initial causes leading to the very first BLEVE in an
6.4.1. Tractor tank carrying ammonia, Houston, USA, May 1976 accident, irrespective of whether the accident also comprised
A tractor tank semi-trailer carrying 19 tonnes of liquid anhy- several subsequent BLEVEs. However, more than one BLEVEs per
drous ammonia accidentally went through a bridge rail on an accident is the most common situation. Therefore, if one also
interstate highway and fell about 5 m onto a freeway, on which, at considers that a preceding BLEVE in an accident was in fact the
the time of the accident, the traffic was quite heavy. The tank initiating event of a subsequent BLEVE in the same accident, it
BLEVEed and the liquid ammonia flashed off, forming a 30 m high could well be that a preceding BLEVE is the most common trigger
cloud. The cloud gradually mixed with the ambient air and finally for BLEVEs.
attained a width of about 300 m and a length of about 600 m. It was
estimated that the evaporation of the ammonia and the subsequent
Table 4
mixing with air occurred within 5 min. The driver of the truck and
Frequencies of initiating events in 88 major BLEVEs occurring
five other people were killed. 78 people had to be taken to hospital from 1926 to 2007. From Abbasi and Abbasi (2007).
for injury treatment, whereas about another 100 suffered less se-
Fire 36%
vere injuries. Five of the six casualties were due to the toxic effect of
Mechanical damage 22%
ammonia. Overfilling 20%
Runaway reactions 12%
6.4.2. Freight train transporting chlorine, Montana, Mexico, 1981 Overheating 6%
A train comprising 38 waggons, including 32 rail tank cars filled Vapour space contamination 2%
Mechanical failure 2%
with liquid chlorine, was moving down a steep and winding valley
40 R.K. Eckhoff / Journal of Loss Prevention in the Process Industries 32 (2014) 30e43

7. Preventing basic technical causes that can lead to BLEVEs contamination by foreign substances with which they can react
exothermally. By installing systems for inserting the vapour space
7.1. Preventing that PLG-containing vessels become exposed to fires in pressurized vessels by nitrogen or other non-reactive gas, and/or
installing explosion-suppression systems, destructive vapour space
7.1.1. Keeping PLG-containing vessels at safe distances from likely explosions can be prevented.
sources of fire
Fire engulfment is the most common basic technical cause of 7.6. Preventing weakening of vessel structures due to fatigue, creep,
BLEVEs. Therefore it is very important that PLG-containing vessels corrosion, etc.
be located at sufficiently long distances from each other and from
any other potential sources of fire. Proper design and testing of vessels can prevent onset of
distortion and possible rupture. Periodic wall thickness measure-
7.1.2. Sloping of the nearby ground ments, internal inspection for corrosion, acoustic emission testing
To prevent a pool fire from occurring after an accidental spill to detect possible initial cracking of the container, etc., should be
from a fixed PLG vessel, the ground around the vessel should have a performed to ensure the fitness of the containers.
downwards slope of at least 1% so to enable the spill to flow away to
a safe area. 7.7. Protection of vessels from being exposed to fire or hit by
missiles, by vessel burial
7.1.3. Water barriers
Facilities for generating water mist barriers close to PLG vessels In principle vessels containing PLGs will be partially or fully
can be installed. Such barriers can capture any flammable vapour protected from fire or missile impact if they are partially or fully
released from a PGL-vessel and disperse it without the flammable buried. But such vessels are both difficult to inspect and maintain,
vapour getting ignited. In the case of water-soluble toxic substance and preventing and controlling corrosion is difficult.
like ammonia and chlorine, the water will dissolve some of the
released material and thereby reduce the amount of toxic material 7.8. Prevention of excessive superheating of vessel content
that remains dispersed in the air.
In distillation systems and chemical reactors nucleation devices
7.2. Preventing mechanical damage of PLG-containing vessels such as sharp-edged ceramic material or aluminium mesh is placed
inside the liquid to promote normal boiling and hence prevent
Trucks and railroad cars carrying PLGs should be protected from superheating. Application of this technique to PLG vessels has been
accidental damage generating spills by using double-walled vessels proposed, but well-proven established methods are not available.
with thermal insulation between the walls. Collisions or over-
turning during transportation may then damage the outer wall, 8. Prevention of catastrophic rupturing of PLG-vessels
without any spills occurring. It is then important to make the outer engulfed in fire
wall sufficiently strong to provide sufficient protection of the inner
wall. 8.1. Introduction

7.3. Preventing overfilling of vessels and vessel overpressure It is almost impossible to say with certainty whether or not a
jeopardized vessel will give rise to a BLEVE. Nor is it possible to
Rigid compliance with standards for filling and weighing of forecast with any certainty when a vessel will give rise to a BLEVE
vessels that may become exposed to BLEVEs, as well as for stan- after having been jeopardized. This and the further uncertainty
dards for relief devices has reduced the frequency of BLEVEs due to associated with forecasting the size, range, direction, and mo-
overfilling. Relief-devices can get plugged, but this can be mentum of missiles likely to become ejected from a BLEVE pose
compensated for by installing rupture disks in parallel to the relief special challenges both as regards preventing a BLEVE as well as
valve. mitigating the potential damage from a BLEVE. Also, predicting the
thermal impact of fireballs from BLEVEs present considerable
7.4. Preventing runaway reactions challenges, as discussed in 8.5 above. According to Abbasi and
Abbasi (2007) several tragic accidents have been reported in
BLEVEs due to runaway reactions inside the vessel are much less which fire fighters trying to save an LPG vessel engulfed in fire were
common than BLEVEs due to accidental vessel damage from the killed by the violently expanding fireball, or by rocketing fragments
outside, such as fire or missile impact. However, they do occur and generated when a vessel suddenly ruptured.
precautions should be taken. Hence, all vessels and other process
equipment where runaway reactions are possible should have 8.2. Thermal insulation
instrumentation for continuous monitoring of temperature and
pressure. Such equipment should have facilities for counteracting The PLG containers should be thermally insulated to the extent
excessive pressure or temperature, e.g. internal cooling coils or possible. This will reduce the rate of heating of the vessel when
external jackets, remotely controlled venting valves, inhibitor- exposed to heat load from a fire, and thereby also delay the pres-
injection systems, and internal deluges. There should also be sure increase inside the vessel. Significant thermal insulation is
alarms for control-room and field personnel when excessive pres- obtained if the vessel wall is protected by a steel jacket containing a
sures and/or temperature occur. ceramic thermal insulation material of thickness at least 13 mm.
Even steel jackets with just air inside can reduce the heating rate of
7.5. Preventing destructive exothermal reactions in vessels due to the internal wall to approximately half of that of with just a single
reactive impurities wall. Thermal insulation alone cannot prevent a BLEVE, but it may
provide a delay of 4e5 h, which can give fire fighters the time they
Vessels containing highly reactive gases such as hydrogen need to remove the heat load. In fixed installations, even the vessel
and chlorine in liquefied form should be protected against support system should be thermally insulated so that it does not
 R.K. Eckhoff / Journal of Loss Prevention in the Process Industries 32 (2014) 30e43 41

collapse when subjected to heat. Likewise, valves, pipes, and other flow rate of minimum 10e15 L/m2 min. is recommended. If the
safety elements installed on the PLG vessel must be able to with- flame is highly turbulent, flow rates even larger than 25 L/m2 min.
stand the high temperatures that may occur in a crisis situation. The may be required. If, however, the vessel is being impinged by a jet
thermal insulation system should be installed in such a way that it fire, water deluge is less effective.
does not interfere with the periodic inspection of the vessel surface
and vessel support systems. The effectiveness of fireproofing for 8.5. Rapid tank depressurization
delaying a BLEVE also depends on whether the pressure relief valve
(PRV) functions correctly. In addition to activating the water deluge system, an effective
system for rapid depressurization of the vessel, which bypasses the
8.3. Cooling of the unwetted part of the PLG vessel wall by an normal PRV, and capable of reducing the vessel pressure to half the
internal liquid spray system design pressure within 15 min, should be available and activated.
The released material should be eliminated in a safe manner, for
Young (2004) developed a system in which a turbo-charger is example by flaring. The depressurization should not be too rapid, as
placed inside the PLG vessel. If the vessel becomes engulfed in a it can lead to extremely low temperatures and increased fragility of
fire, the turbo-charger will take liquid in the vessel from just below the steel.
the liquid surface, and spray it vertically upwards to cool the
unwetted part of the tank. The system is illustrated in Fig. 7. 9. Reducing consequences if a BLEVE does occur
The turbo charger is driven by the vapour escaping through the
pressure relief valve. The system is expected to prevent the 9.1. Establishment and implementation of adequate safety distances
unwetted part of the tank from becoming heated to the point of
weakening of the metal, which would facilitate crack formation. If a vessel suffers a BLEVE within only a few minutes after the
The system can be built into most new PLG vessels, and also fitted formation of the first crack, little can be done to reduce the damage
to existing vessels, including mobile ones. it would cause. However, even if a BLEVE is delayed by several
hours, there are challenges. Great care must be taken to make sure
8.4. Water deluge that people move to outside the safety distance, both with regard to
a potential blast wave, potential missiles, and a potential major
The thermal load on a PGL vessel engulfed in fire can be sub- fireball. The destructive impact of a delayed BLEVE can be even
stantially reduced by water deluge. A layer of flowing water must more severe than from an immediate one.
then be applied to the vessel as soon as possible. A water layer of Birk (1996) suggested that fire fighters should not get closer to a
sufficient thickness should then totally cover the vessel wall, not vessel that could undergo a BLEVE than four fireball radii, but the
least those areas that are directly engulfed by the flame. A water distance should not be smaller that 90 m. He suggested that a rough

Fig. 7. Safety system for BLEVE prevention/delay developed by Young (2004), based on cooling of the unwetted part of the PLG vessel wall by an internal liquid spray.
42 R.K. Eckhoff / Journal of Loss Prevention in the Process Industries 32 (2014) 30e43

estimate of the fireball radius can be obtained by the simple temperature well above its SLT at atmospheric pressure before
expression. the vessel ruptures.
 BLEVEs are not limited to flammable pressurized liquefied gas
R ¼ 3$m0:33 (PLG) only. In principle all PLGs can in cause BLEVEs, including
e.g. superheated water in boilers.
where m is the mass of fuel in kg and R is the fireball radius in m.  BLEVEs lead to emission of strong blast waves and rocketing
The public should be evacuated to a distance preferably 30 radii fragments from the ruptured vessel. Such fragments have a
away from vessels that could undergo a BLEVE. The distance would substantial damage potential. They have caused deaths of peo-
have to be increased downwind of the vessel. However, as the ple quite a distance away from the site of the BLEVE, as well as
vessel size increases beyond 5 m3 the 30 radii requirement gets too given rise to secondary accidents in other process units.
conservative and a requirement of minimum 15 radii becomes  If the liquid is flammable and ignites upon release to the at-
more appropriate. mosphere, a huge fireball can be generated. If the liquid is non-
If the PLG involved in a BLEVE is toxic, e.g. chlorine, ammonia, flammable but toxic, its dispersion in the atmosphere creates
methyl iso-cyanate, or phosgene, careful attention has to be paid to special serious concerns.
its dispersion pattern both in the blast from the BLEVE and in the  A substantial number of case histories have been published. The
atmosphere further away. case histories confirm that BLEVEs have been a major industrial
hazard for many years, and still are.
9.2. Preventing “domino” effects  It is not possible in general to forecast with reasonable certainty
how much time a vessel exposed to a fire may take before un-
The first concern towards mitigation of the damage caused by a dergoing a BLEVE. This can vary from a few seconds to several
BLEVE is to prevent the initial accident from triggering secondary, hours. This fact makes it very dangerous for e.g. fire fighters to
higher order accidents. Analyses of BLEVEs that have occurred in operate close to a fire-engulfed vessel containing a pressurized
the past reveal that most of them were not “stand-alone” accidents. liquefied gas.
Other liquid-containing vessels in the vicinity of the one that  A variety of methods for preventing and mitigating damaging
BLEVEed can become exposed to heat and/or missiles from the effects of BLEVEs are in use.
initial event and this may give rise to further BLEVEs. In order to
avoid such “domino” effects vessels that may give rise to BLEVEs
should be kept sufficiently far away from each other to render such References
effects unlikely. Mechanical barriers may also be placed around
vessels to limit the impact of outgoing or incoming missiles. This Abbasi, T., & Abbasi, S. A. (2007). The boiling liquid expanding vapour explosion
may require rather complex design and careful cost/benefit con- (BLEVE): mechanism, consequence assessment, management. Journal of Haz-
ardous Materials, 141, 489e519.
siderations may be needed. Berg, A. C. van den, Voort, M. M. van der, Weerheijm, J., & Versloot, N. H. A. (2006).
BLEVE blast by expansion-controlled evaporation. Process Safety Progress, 25,
9.3. Fireball suppression 44e51.
Birk, A. M. (1996). Hazards from propane BLEVEs: an update and proposals for
emergency responders. Journal of Loss Prevention in the Process Industries, 9,
If sufficient quantities of fire suppressant can be released in such 173e181.
a way that the suppressant becomes mixed with the flashing fuel at Birk, A. M., & Cunningham, M. H. (1994). The boiling liquid expanding vapor ex-
plosion. Journal of Loss Prevention in the Process Industries, 7, 474e480.
the moment of vessel failure, the formation of a fireball may be Birk, A. M., & Cunningham, M. H. (1996). Liquid temperature stratification and its
prevented or at least its intensity may be reduced significantly. effect on BLEVEs and their hazards. Journal of Hazardous Materials, 48,
Alternatively the fireball, after its formation, may be wrapped in a 219e237.
Birk, A. M., & Steen, J. D. J. van der (2002). On the transition from no-BLEVE to BLEVE
barrier of a suitable fire suppressant. Then, the suppressant would
for a 1.8 m3 propane tank. American Society of Mechanical Engineers, Pressure
be sucked into the fireball by the very strong prevailing air Vessels and Piping Division PVP, 446, 143e152.
entrainment flows. As a result, the flame may become completely Blything, K. W., & Reeves, A. B. (1988). An initial prediction of the BLEVE frequency of a
suppressed, or at least the fireball size would be significantly 100 tonnes butane storage vessel (88 pp.). Report SRD/R488. UK Atomic Energy
Authority, Safety and Reliability Directorate.
reduced. The use of water mist as a fireball suppressant is an Bosch, C. J. H. van den, & Weterings, R. A. P. M. (1997). Methods for the calculation of
obvious possibility but the liquid droplets may evaporate physical effects. CPR 14E (TNO ‘yellow book’). The Hague, The Netherlands:
completely before the water gets sucked into the fireball. This may Committee for the Prevention of Disasters.
Center for Chemical Process Safety. (1999). Guidelines for consequence analysis of
reduce the suppressing effect. chemical releases. New York: Center for Chemical Process Safety, American
Institute of Chemical Engineers.
10. Conclusions Crawley, F. K. (1982). The effect of ignition of a major fuel spillage, the assessment of
major hazards. Report from the. Rugby, UK: Institution of Chemical Engineers.
Giesbrecht, H., Hess, K., Leukel, W., & Maurer, B. (1981). Analysis of explosion
 The definitions of a BLEVE as encountered in the literature differ hazards on spontaneous release of inflammable gases in the atmosphere: part
somewhat. One main difference is that some definitions include 1: propagation and deflagration of vapour clouds on the basis of bursting tests
on model vessels. German Chemical Engineering, 4, 305e314.
the formation of a fireball, whereas others do not. Furthermore, Lees, F. P. (1996) (2nd ed.). Loss prevention in the process industries e Hazard iden-
some definitions imply that the temperature of the liquid at tification, assessment and control (Vols. 1e3). Oxford, UK: Butterworth Heine-
vessel rupture significantly exceeds the SLT (se next point), mann, ISBN 0 7506 1547 8.
Makhviladze, G. M., Roberts, J. P., & Yakush, S. E. (1999). Combustion of two-phase
whereas others do not. hydrocarbon fuel clouds released into the atmosphere. Combust. Flame, 118,
 A pure liquid, with no contact with “active points”, e.g. on solid 583e605.
surfaces, can be heated to a temperature considerably higher Marshall, V. C. (1987). Major chemical hazards. In Ellis Horwood series in chemical
engineering. New York: Halsted Press: a Division of John Wiley & Sons, ISBN 0-
than its boiling point at the prevailing pressure before signifi-
470-20813-9.
cant boiling starts. This temperature is called the “superheat Nesis, E. I. (1966). Boiling of liquids. Soviet Physics Uspekhi, 8(6), 883e907.
limit temperature” (SLT). SLT is synonymous with the “homo- Partrick-Yeboah, J. R., & Reid, R. C. (1981). Superheat-limit temperatures of polar
geneous nucleation temperature”. The present review empha- liquids. Industrial & Engineering Chemistry Fundamentals, 20, 315e317.
Salla, J. M., Demichela, M., & Casal, J. (2006). BLEVE: a new approach to the su-
sizes the central role of the SLT in BLEVE development. The most perheat limit temperature. Journal of Loss Prevention in the Process Industries, 19,
serious BLEVEs occur when the liquid has been heated to a 690e700.
 R.K. Eckhoff / Journal of Loss Prevention in the Process Industries 32 (2014) 30e43 43

Selway, M. (1988). The predicted BLEVE frequency of a selected 2000 m3 butane sphere Wakeshima, H., & Takata, K. (1958). On the limit of superheat. Journal of the Physical
on a refinery site (57 pp.). Report SRD/R492. UK Atomic Energy Authority, Safety Society of Japan, 13, 1398e1403.
and Reliability Directorate. Walls, W. L. (1978). What is a BLEVE? Fire Journal, 31, 46e47.
€ro
To €k, Z., Ajtai, N., Turcu, A.-T., & Ozunu, A. (2011). Comparative consequence Walls, W. L. (1979). The BLEVE e part 1. Fire Command, 17, 35e37.
analysis of the BLEVE phenomena in the context of land use planning. Case Young, C. (2004). LPG tank anti-BLEVE and automotive tank safety system. US
study: the Feyzin accident. Process Safety and Environmental Protection, 89, 1e7. Patent 4,680,937.