The Boiling Liquid Expanding Vapour Explosion (BLEVE) PDF

Journal of Hazardous Materials 141 (2007) 489519
The boiling liquid expanding vapour explosion (BLEVE):

Mechanism, consequence assessment, management
Tasneem Abbasi, S.A. Abbasi
Center for Pollution Control and Energy Technology, Pondicherry University, Pondicherry 605 014, India
Received 1 September 2005; received in revised form 30 June 2006; accepted 10 July 2006
Available online 27 September 2006
Abstract
Among the most devastating of accidents likely in chemical process industry is the boiling liquid expanding vapour explosion (BLEVE). It is
accompanied by highly destructive blast waves and missiles. In most situations there is also a reball or a toxic gas cloud. The damaging effect
of BLEVEs is reected in the fact that the 80-odd major BLEVEs that have occurred between 1940 and 2005 have claimed over a 1000 lives and
have injured over 10,000 persons besides harming property worth billions of dollars. Release of toxic chemicals like chlorine and phosgene from
BLEVEs have damaged large chunks of areas surrounding the BLEVE site.
This paper presents an overview of the mechanism, the causes, the consequences, and the preventive strategies associated with BLEVEs.
2006 Elsevier B.V. All rights reserved.
Keywords: BLEVE; Fireball; Missiles; Blast wave; Domino effect
1. Introduction
1.1. Denition of BLEVE
The Centre for Chemical Process Safety [1] has dened boil-
ing liquid expanding vapour explosion (BLEVE) as a sudden
release of a large mass of pressurized superheated liquid to the
atmosphere. The sudden release can occur due to containment
failure caused by re engulfment, a missile hit, corrosion,
manufacturing defects, internal overheating, etc. According to
Birk and Cunningham [2] a BLEVE is the explosive release of
expanding vapour and boiling liquid when a container holding
a pressure-liqueed gas fails catastrophically. They have
further dened catastrophic failure as the sudden opening
of a tank to release its contents nearly instantaneously. The
sudden release from connement of a hitherto pressurized
and liqueed vapour causes instantaneous and explosive
boilingvaporization, leading to a series of cataclysmic
impacts.
A BLEVE gives rise to the following [1,35]:
Corresponding author. Tel.: +91 413 2655263; fax: +91 413 2655262.
E-mail address: prof.s.a.abbasi@gmail.com (S.A. Abbasi).
Splashing of some of the liquid to form short-lived pools; the
pools would be on re if the liquid is ammable.
Blast wave.
Flying fragments (missiles).
Fire or toxic gas release. If the pressure-liqueed vapour is
ammable, as is often the case, the BLEVE leads to a reball.
When the material undergoing BLEVE is toxic, as in the case
of ammonia or chlorine, adverse impacts include toxic gas
dispersion.
After the Flixborough disaster which had destroyed most of
the caprolactam plant of M/s Nypro Ltd. in 1974 [3,6], great
attention was focused on vapour cloud explosions (VCEs) till
Kletz [7] pointed out that BLEVEs can cause as much loss of
life and property as VCEs and should not be neglected. This
view is justied when we look at the record of accidents before
and after Flixborough. Indeed one of the biggest accidents in
chemical process industry which occurred at the LPG plant
at Mexico City in 1984 claiming over 650 lives involved a
succession of BLEVEs [3,4].
1.2. Coinage of the term BLEVE
The acronym BLEVE was coined in 1957 by three Factory
Mutual Research Corporation workers J.B. Smith, W.S. Marsh,
0304-3894/$ see front matter 2006 Elsevier B.V. All rights reserved.
doi:10.1016/j.jhazmat.2006.09.056
490 T. Abbasi, S.A. Abbasi / Journal of Hazardous Materials 141 (2007) 489519
and W.L. Walls [8]. They had analyzed the likely mode of fail-
ure of a vessel containing an overheated mixture of formalin
and phenol, and had believed that the container had suffered a
boiling liquid expanding vapour explosion [9]. Later, Walls
[10,11] dened BLEVE as a failure of a major container into
two or more pieces, occurring at a moment in time when the
contained liquid is at a temperature well above its boiling point
at normal atmospheric pressure. Even though BLEVE is now
rmly entrenched in the lexicon of risk assessment, its appro-
priateness has been questioned from time to time. According
to Marshall [12], the UK-based Institute of Chemical Engineer-
ing (IChemE) had advised that the term BLEVE is an acronym
of uncertain meaning, and its usage should be avoided. Reid
[13] had dened BLEVE as the sudden loss of containment of
a liquid that is at a superheated temperature. Even though this
denition served as a reference point for BLEVEthrough several
years, the Centre for Chemical Process Safety [8] has reverted
to the Walls denition with the modication that failure of a
major container into two or more pieces has been replaced by
failure of a vessel [8]. Venart et al. [14] had coined the term
BLCBE (boiling liquid collapsed bubble explosion) to describe
one of the variants of BLEVE.
2. Mechanism of BLEVE
McDevitt et al. [15], Prugh [16,17], Leslie and Birk [18],
Lees [3], Birk and Cunningham [2,19], Casal et al. [5] and
Venart [20], among others, have given elaborate description of
the occurrences which lead to a BLEVE. Based on their obser-
vations, and of Reid [21], Shebeko et al. [22] and Birk et al.
[23], the steps involved in a typical BLEVE can be identied as
under:
(a) A vessel containing pressurized liquid gas (PLG) receives
heat load or fails due to a missile hit, fatigue, or corro-
sion. If a vessel containing pressure-liqueed gas (PLG),
in other words a liquid conned at a temperature above its
atmospheric pressure boiling point, gets accidentally heated
say from the heat radiation emanating from a nearby re
the pressure inside the vessel begins to rise. When this pres-
sure reaches the set pressure of the pressure relief valve,
the valve operates. The liquid level in the vessel falls as the
valve releases the liquid vapour to the atmosphere. The liq-
uid is effective in cooling that part of the vessel wall which
is in contact with it, but the vapour is not. The proportion
of the vessel wall which has the benet of liquid cooling
falls as the liquid vaporizes. After a time, the portion of the
metal which is not cooled by liquid also becomes exposed
to the heat load, weakens, and may then rupture. This can
occur even though the pressure relief valve may be operating
correctly.
A vessel may also fail even in absence of re-engulfment
if it is accidentally hit by missiles originating from another
vessel exploding nearby as it happened during the serial
explosions in the LPG facility at Mexico City [4,24,25]
or other forms of mechanical failure such as gland/seal loss,
sample line breakage, fatigue, or corrosion [26].
(b) The vessel fails. A pressure vessel is designed to withstand
the relief valve set pressure, but only at the design temper-
ature conditions. If the metal has its temperature raised due
to heat load exerted by a nearby re, it may lose strength
sufciently to rupture. For example, the steel normally used
to build LPG vessels may fail when the vessel is heated
to 650
C and its pressures reaches 15 atm. The vessel

may also rupture due to mechanical failure as stated in the
preceding paragraph.
(c) There is instantaneous depressurization and explosion.
When a vessel fails, there is instantaneous depressuriza-
tion. The liquid inside the vessel, which hitherto was at a
temperature corresponding to a high pressure, is suddenly
at atmospheric pressure but at a temperature well above the
liquids atmospheric pressure boiling point. In other words
the liquid is superheated. But there is a limit different for
different liquids up to with liquids can withstand super-
heating. If the temperature of the liquid in the suddenly
depressurized vessel is above this superheat limit temper-
ature (SLT), there will be instantaneous and homogeneous
nucleation. It would cause a sudden and violent ashing
of a large portion of the liquid, resulting in a boiling liq-
uid expanding vapour explosion (BLEVE). This would
occur within 1 ms of depressurization, causing a massive
release of liquidvapour mixture. If the liquid in a suddenly
depressurized vessel is below its SLT, but is in a state of
signicant superheat, a BLEVE would still occur because
factors such as depressurization waves agitating the liquid
when the vessel rst develops a crack, and presence of likely
heterogeneous nucleation sites, would cause the explosive
boiling-cum-vaporization which triggers BLEVE [26].
Prugh [16,17], was among the rsts who stressed that
a BLEVE can occur even for initial temperatures below
the superheat limit, but stated that the higher TNT (trini-
trotoluene) equivalent for BLEVE occurs near or above the
superheat limit temperature (SLT). Birk and Cunningham
[2] state that BLEVEs have been observed with propane
when the pressure-liqueed gas was at ambient temperature
(20
C), well below its atmospheric superheat limit tem-

perature of 53
C. However, like Prugh [16,17], they, too,

recognize that for violent, explosive, boiling to take place,
there must be the potential for superheat in the liquid when it
is suddenly exposed to a pressure below its saturation pres-
sure as a result of the initial tank failure. Indeed the superheat
aspect becomes implicit if the material under reference is a
pressure liqueed gas (PLG)i.e. a substance which would
have been in gaseous state at atmospheric pressure but is
held as liquid in a pressurized container. Numerous indus-
trial chemicals such as liquid petroleum gas, compressed
natural gas, liqueed chlorine, etc., conrm to this deni-
tion. So does superheated water in a boiler.
(d) The vessel is shattered. The suddenly vaporizing liquid
with several hundred-fold to over a thousand fold increase
in its volume plus the expansion of the already existing
vapour, generate a powerful overpressure blast wave. The
magnitude of the blast wave is much higher than the one
caused by a vapour cloud explosion occurring in an identical
T. Abbasi, S.A. Abbasi / Journal of Hazardous Materials 141 (2007) 489519 491
quantity of material. The vessel is shattered and its pieces
are propelled outwards. Some of the liquid may be splashed
and hit the ground nearby forming short-lived pools before
vaporizing. These pools may be are if the liquid happens
to be ammable.
The shattering of the vessel sends big and small fragments
shooting at high velocities in all directions. The missiles
can, and often do, damage other vessels storing liqueed
gas under pressure, causing them to undergo BLEVE as
well. This domino effect [27,28] was witnessed at its most
tragic worst at the Mexico City in 1984, causing the largest
number of loss of lives ever occurred in an explosion-cum-
re accident in a process industry. At times a large part of
the vessel itself turns into a missile and is shot over long
distances. For example, at port Newark, a portion of a sphere
went ying over 800 m and demolished a petrol bunk on
which it landed.
(e) There is reball or toxic dispersion. If the substance involved
is not combustible or toxic, such as water in boilers the
pressure wave and the missiles are the only effects of the
explosion. But if the substance is ammable, as is often the
case, the mixture of liquid/gas released by the explosion
catches re, giving rise to a reball. Past accidents analy-
sis such as the one reported by Prugh [16,17] covering
notable BLEVE accidents (19261986) and the compila-
tion made by Lees [3]reveal that over two-thirds of all
BLEVEs involve ammable chemicals. With such chem-
icals, a BLEVE is almost always followed by a reball,
causing massive damage due to the intense thermal radi-
ation that ensues. The shape, the size, and the heat load
exerted by the reball are a function of numerous factors.
It may so happen that the whole mass of fuel can burn only
at its periphery because there is no air inside the mass (the
mixture being outside the ammability limits). Further, not
all the fuel initially contained in the tank may be involved
in the re; some of the material might escape (from a crack
or other opening in the vessel) before the explosion [5].
Some of the fuel may be entrained in the wake formed by
the ying fragments. In the BLEVE disasters, such as the
ones that occurred at Mexico City in 1984, and at Sydney
in 1990fragments of shattering vessels carried with them
portions of the ammable liquid, causing res all around
and jeopardizing other vessels.
As the reball grows, the turbulence of the re entrains air
into the reball. Simultaneously the thermal radiation vaporizes
the liquid droplets and heats the mixture. As a result of these pro-
cesses, the whole mass turbulently increases in volume, evolving
towards an approximately spherical shape that rises, leaving a
wake of variable diameter. Such reballs can be very large, caus-
ing a very strong thermal radiation [29,30].
The size, the life, and the radiation intensity of a reball
may also depend on the temperature of the liquid lading [31],
and whether the loss of connement of the ammable material
had occurred when the pressure inside was still rising [32]. The
BLEVE reballs are spheroidal when fully developed; on lift-
off they acquire a mushroom-like shape. Fireballs resulting from
two-step BLEVEs may be approximately ellipsoidal in shape
[31].
Toxic dispersion. BLEVE accidents have occurred involving
ammonia [5], chlorine [12], chlorobutadiene [6], and phosgene
[12] wherein the explosion did not cause a reball but was
accompanied by dispersion of toxic material. Indeed of the
one-third past BLEVE events not involving ammable liquids,
the majority have been associated with toxic gaseschlorine
(14%), ammonia (10%), and phosgene (2%) account for 76%
of the BLEVEs involving non-ammables. With such chem-
icals, fatalities have been caused by the toxic clouds that
accompanied the blast wave and the missiles when BLEVE
occurred. BLEVEs without any re or toxic dispersion have
also occurredinvolving carbon dioxide and water [4,6,24,25].
These authors believe that a lot of boiler explosions and such
explosions are far more common than explosions involving other
chemicals are BLEVEs, though not commonly acknowledged
as such. Indeed if the boiler explosions which occur when hold-
ing superheated water are included in the BLEVE tally, it may
well turn out that the substance most frequently involved in
BLEVEs is none other than water!
Among major BLEVEs with toxic dispersion, the maximum
cases have involved chlorine, accounting for 138 fatalities dur-
ing 19261981 [16], followed by ammonia (49 fatalities in the
corresponding period).
According to Lees [3], it is possible, though not common,
for a BLEVE event not caused by an engulng re to provide
the source for a large vapour release leading to a ash re or a
vapour cloud explosion.
3. Illustrative case histories
Afewinstances whichillustrate the varietyof situations under
which BLEVEs have occurred and the destruction BLEVEs nor-
mally cause, are presented below. These and some other major
BLEVE events that have occurred over the last 80 years are
catalogued in Table 1.
3.1. BLEVEs involving stationary installations of
ammable chemicals
3.1.1. Butane storage spheres, Montreal
On 8 January 1957, a leak occurred in an 800 m
3
sphere
storing butane at an installation at Montreal, Canada. The sphere
had been overlled due to a faulty level gauge, resulting in the
leak. The escaping butane formed a vapour cloud which met
with an ignition source, probably at a service station 180 maway.
The ame ashed back to the leaking sphere, where a pool re
started. The heat load softened a 1900 m
3
sphere located close
by, which underwent a BLEVE within 30 min of the start of the
pool re. In other 15 min the 800 m
3
sphere and another adjacent
sphere of 2400 m
3
capacity also exploded, causing devastation
all round [33].
3.1.2. LFG storage farm, Feyzin
A leak in a propane storage sphere on 4 January 1966
at Feyzin, France led to one of the worst incidents involv-
Table 1
An illustrative list of some of the major BLEVEs (19262004)
Date Location Cause Material Quantity (tonnes) Death (d), injured (i)
13 December 1926 St. Auban, France Overlling Chlorine 25 19d
28 May 1928 Hamburg, Germany Runaway Phosgene 10 11d, 171i
10 May 1929 Syracuse, NY, USA Explosion (H2) Chlorine 25 1d
24 December 1939 Zamesti, Romania Overlling Chlorine 10 60d
29 July 1943 Ludwigshafen, Germany Overlling Butadiene 16 57d
5 November 1947 Rauma, Finland Overlling Chlorine 30 19d
28 July 1948 Ludwigshafen, Germany Overlling Ethyl ether 33 209d
7 July 1951 Port Newark, NJ, USA Fire Propane (70) 2600 14i
4 April 1952 Walsum, W. Germany Overlling Chlorine 15 7d
4 June 1954 Institute, WV, USA Runaway Acrolein 20
1955 Ludwigshafen, FRG Railroad accident LPG
a
2i
1955 Cottage Grove, OR, USA Storage vessel
failure
LPG
a
12d, 13i
8 January 1957 Montreal, Canada Fire Butane 5100 1d
1958 Michigan, USA Overlling Butane 55 1d
28 June 1959 Meldrin, GA, USA Damage (Derail) Propane 55 23d
18 August 1959 Kansas City, MO, USA Fire Gasoline 70 5d
1959 McKittrick, CA, USA LPG Storage cylinder
(six on site)
a
2i
17 April 1962 Doe Run, KY, USA Runaway Ethylene oxide 25 1d
4 January 1966 Feyzin, France Fire Propane 1000 18d, 83i
1 January 1968 Dunreith, IN, USA Fire (Derail) Ethylene oxide NA 5i
21 August 1968 Lieven, France Mechanical Ammonia 20 5d
2 January 1969 Repcelak, Hungary Overlling Carbon dioxide 35 9d
25 January 1969 Laurel, MS, USA Fire (Derail) Propane 65 2d, 976i
18 February 1969 Crete, NB, USA Damage (Derail) Ammonia 65 8d
1969 Cumming, IA, USA Damage (Derail) Ammonia
a a
11 September 1969 Glendora, MS, USA Fire Vinyl chloride 55
21 June 1970 Crescent City, IL, USA Fire (Derail) Propane (5) 275 66i
19 January 1971 Baton Rouge, LA, USA Overpressure Ethylene 4
19 October 1971 Houston, TX, USA Fire (Derail) Vinyl chloride 50 1d, 50i
9 February 1972 Tewksbury, MA, USA Collision Propane 28 NA
30 March 1972 Rio de Janeiro, Brazil Fire Propane 1000 37d
21 September 1972 NJ Turnpike, NJ, USA Collision Propylene 18 2d
27 November 1972 San Antonio, TX, USA Corrosion Carbon dioxide 0.01
1972 Lynchburg, VA, USA Propane Road tanker 9 2d, 5i
1972 Rio de Janeiro, Brazil LPG Storage spheres
(ve on site) and
cylinders
a
37d, 53i
5 July 1973 Kingman, AZ, USA Fire Propane 100 13d, 95i
11 January 1974 W. St. Paul, MN, USA Fire Propane 27 4d
12 February 1974 Oneonta, NY, USA Fire (Derail) Propane (4) 288 25i
29 July 1974 Pueblo, CO, USA Fire (test) Propane 80
29 April 1975 Eagle Pass, TX, USA Collision Propane 18 16d
14 December 1975 Niagara Falls, NY, USA Runaway Chlorine 20 4d
1975 Des Moines, IA, USA LPG Rail tank car
a
3i
11 May 1976 Houston, TX, USA Collision Ammonia 20 6d
31 August 1976 Gadsden, AL, USA Fire Gasoline 4 3d
1976 Belt, MN, USA LPG Rail tank car 80 22i
1977 Cartegna, Columbia Overpressure Ammonia
a
30d
1977 Dallas, TX, USA Isobutene Rail tank car
a
1i
1977 Goldona, VA, USA LPG Rail tank car 70 2d, 9i
22 February 1978 Waverly, TX, USA Damage (Derail) Propane 45 16d, 43i
11 July 1978 San Carlos, Spain Overlling Propylene 25 211d
30 May 1978 Texas City, TX, USA Fire Butanes (6) 1500 7d, 10i
1978 Donnellson, IA, USA LPG Pipeline 435 2d, 2i
30 August 1979 Good Hope, LA, USA Ship collision Butane 120 12d
1979 Pazton, TX, USA Chemicals Rail tank car
a
8i
1979 Los Angeles, CA, USA Gasoline Road tanker
a
2d, 2i
1 August 1981 Montanas, Mexico Damage (Derail) Chlorine (2) 110 29d
19 January 1982 Spencer, OK, USA Overheating Water 0.3 7d
11 December 1982 Taft, LA, USA Runaway Acrolein 250
12 July 1983 Reserve, LA, USA Runaway Chlorobutadiene 1 3d
Table 1 (Continued )
Date Location Cause Material Quantity (tonnes) Death (d), injured (i)
4 October 1983 Houston, TX, USA Overlling Methyl bromide 28 2d
19 November 1984 Mexico City, Mexico Fire Propane (20) 3000 650d, 6400i
1984 Romeoville, IL, USA Propane Process vessel
a
15d, 22i
28 January 1986 Kennedy Space Center, FL,
USA
Fire Hydrogen 115 7d
1 April 1990 Boral LPG distribution depot,
Sydney, Australia
Fire LPG >240 35,000 affected
1 April 1990 Cairns gas terminal,
Queensland, Australia
Fire LPG
a
1d
28 August 1992 Japan Damage Nitrogen
a
$5 million loss
August 1993 Panipat, India Pressure build-up Ammonia
a
6d, 25i
19 April 1993 Waco, TX, USA Fire LPG
a
27 June 1993 Quebec, Canada Fire Propane 2.3 4d, 7i

4 March 1996 Weyauwega, WI, USA Derailment Propane, LPG
a
18 March 1996 Palermo, Italy Collision in a

highway tunnel
Propane
a
5d, 25i
2 October 1997 Burnside, IL, USA Fire LPG 3.8 2d, 2i
9 April 1998 Alberta City, IA, USA Fire Propane 40 2d, 7i
30 April 1999 Between Athens and Lamia,
Greece
Trafc accident LPG
a
4d, 13i
23 September 1999 Toronto, Canada Derailment LPG >60
30 December 1999 Quebec, Canada Derailment,
collision
Hydrocarbons 2700 2i, 350 evac.
27 May 2000 Eunice, LA, USA Derailment Flammable PLGs
a
2000 evac.
19 July 2000 Ohio, USA Overlling Propane 66 3i
7 January 2001 Kanpur, India Highway accident LPG
a
12d, 6i
20 October 2000 Downey, CA, USA Leak Propane 2 2d
22 October 2000 Texas, USA Improper
unloading
Propane 17 2d
1 July 2001 Jamnagar, India Damage LPG
*
12d
20 February 2002 Cairo, Egypt Fire caused in a
passenger train by
a butane tank
BLEVE
Butane
*
373d, 7500i
22 June 2002 Tivissa, Spain Overturned LNG 48 m
3
1d, 2i
25 June 2002 Gronton, CT, USA Overheating Borane-
tetrahydrofuran
0.1 2i
11 April 2003 Louisville, USA Overheating Maltodextrin and
other chemicals
*
1d
13 January 2004 Baltimore, Washington
Highway, USA
Trafc accident Propane
*
10d
19 January 2004 Skikida, Algeria Explosion LNG
*
13d, 74i
9 August 2004 Mihama, Japan Steam pipe
depressurization
Steam
*
4d, 7i
a
Information not available.
ing LFG (liqueed ammable gases) that has ever occurred
[34].
An operator had to drain water from a 1200 m
3
spherical
storage vessel nearly full of propane. There were three valves
underneath the vessel. The rst valve, nearer to the vessel bot-
tom, led to the other two valves further below, connected in
parallel. The operator opened the rst valve and one of the two
lower ones. When traces of oil showed that the draining was
nearly complete he shut the rst valve, then cracked it to com-
plete the draining. No ow came. So he opened the rst valve
fully. The choke presumably hydrate cleared suddenly and
the operator and the two other men were splashed with liquid.
The handle came off the valve and they could not get it back on.
The lower valve had been frozen and could not be moved. Access
was poor because the drain valves were immediately below the
tank, which was only 1.4 m above the ground [34]. The contin-
uing propane leak soon formed a visible cloud of vapour, 1 m
deep. It spread for 150 m and was ignited 25 min after the leak
started by an automobile that had stopped on a nearby road.
The re ashed back to the sphere but there was no immediate
explosion. The sphere was tted with water sprays but the supply
was inadequate to cool the vessel. When the re brigade began
using their hoses, the water supply to the spheres ran dry. Appar-
ently, the remen had used off the available water for cooling
the neighboring spheres to prevent the re from spreading, in
the belief that the vessel on re shall be protected by the relief
valve!
Ninety minutes after the re started, the sphere went through
a BLEVE. Ten out of 12 remen within 50 m were killed. Men
140 mawaywere badlyburnedbya wave of propane whichcame
over the compound wall. Altogether 1518 men were killed and
about 80 injured. Flying debris broke the legs of an adjacent
sphere which fell over. Its relief valve discharged liquid which
added to the re, and 45 min later it also BLEVEd, leading to
more BLEVEs. Altogether ve spheres and two other pressure
vessels burst and three were damaged. The re spread to gasoline
and fuel oil tanks [3].
3.1.3. The PEMEX LPG terminal disaster, San Juan
Ixhantepec, Mexico City
The PEMEX LPG terminal in San Juan Ixhuatepec, Mex-
ico City, was a large installation which received supplies from
three gas reneries every day. On the morning of 19 Novem-
ber 1984, when the vessels at the PEMEX terminal were being
lled with LPG arriving in a pipeline from a renery 400 km
away, a drop in the pipeline pressure was noticed by the control
room and a pumping station. It occurred because an 8 in. pipe
connecting one of the spheres to a series of cylinders had rup-
tured but the operators did not think of such a possibility and
the release of the LPG from the leaking pipeline continued for
510 min. The escaping gas formed a 2 m high cloud covering
an area of 200 m150 m. The cloud then drifted towards a
are tower, caught re and precipitated the rst BLEVE. The
explosion hurled vessel fragments wrapped in burning LPG in
all directions. Some of the projectiles hit other vessels, dam-
aging them, or caused local res which engulfed other vessels.
This led to the failure of one vessel after another; most exploding
vessels caused nearby vessels to fail.
Four LPGspheres, each containing 1500 m
3
of LPG, and sev-
eral other smaller cylinders holding between 45 m
3
and 270 m
3
of the liquid suffered BLEVEs. Each BLEVE generated a re-
ball; such reballs 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
reballs. Masses of fragments of tanks and pipes, some of them
weighing 40 tonnes, were blown into air and landed as far as
1200 m away. The PEMEX terminal was devastated.
The accident was responsible for 650 deaths and over 6400
injuries. Damages due to the explosion and the resulting re
were estimated at approximately $31 million [35].
3.1.4. Boral LPG distribution depot, Sydney
On all-fools day (1 April) in 1990, at approximately 8.45 p.m.
at a Sydney suburb, a BLEVE occurred in a LPG storage-cum-
distribution terminal. It led to further accidents and one BLEVE
after another occurred through the night. The unintended re-
works began with the explosion of a small gas tank. The re
then spread along the ruptured gas pipes to the four main 100-t
steel storage tanks, each at least 60% full, holding 40,000 l of
LPG. It heated-up the tanks until their 15 cm thick steel wall
failed, generating massive BLEVEs. The resulting reballs and
gas ares shot hundreds of meters into the night sky. Hundreds
of portable gas cylinders kept inside a storeroom at Boral, sized
2240 kg also BLEVEd. Power blackouts occurred after the
rst thundering explosion, which shattered windows. Two re
ofcers who were close to the unattended depot, whose main
entrance was padlocked, were thrown against a wall after one
of the explosions. The exploding vessels generated rocketing
fragments; the shock waves sent other objects which came into
their paths, ying like missiles. One such missile, an uprooted
telegraphic pole, just missed crushing a woman who was stand-
ing almost 0.5 Km away from an exploding vessel. One of the
30 m long fractured cylinders shot off its mooring, and rock-
eted through the air with an ignited tail of ames. On landing it
gouged a huge 2 mcrater in the earth, before bouncing through a
wire fence into three, 40-t tanks, which were propelled into the
nearby Alexandria Canal. The rocketing cylinder then hit and
attened an electrical substation and a panel-beating workshop
before nose-diving into the canal, 300 m away from its original
position.
Luckily it all happened on a Sunday or else the toll in the
form of human lives would have been massive as the accident
occurred in a very busy section of the city, where thousands of
persons went to work on weekdays. If the wind had been blowing
in a different direction, it could have carried the re to the jet
fuel storage 20 m away and might have set it ablaze.
The Mascot International Airport is located just a few hun-
dred meters away from where this accident occurred and a
passenger plane on its nal approach into Mascot was buffeted
by shockwaves when one of the tanks BLEVEd. Plane passen-
gers, the airport and a nearby hotel were evacuated and airport
re crews placed on full alert. The chain of BLEVEs was even-
tually broken when Boral engineers released relief safety valves
of the surviving tanks to depressurize them.
Factory storage buildings close to the explosions were
destroyed. Those that survived had doors blasted off hinges,
roofs lifted and windows shattered. Underfoot, thick grey mud
hadformed. Elsewhere, onthe depots 10 ha site were the charred
ruins of 10 gas trucks, 3 gas tanker semi-trailers and a row of
storerooms which backed onto the main row of cylinders. An
adjacent bitumen production plant was destroyed. The Boral
plant was built in 1968 to satisfy standards which by 1990 had
become outdated [4].
3.1.5. Burnside, Illinois
A3800 l LPGtank underwent a BLEVE after coming in con-
tact with re from a nearby grain dryer. The tank was venting
when the re services unit were applying water to the tank from
a distance of approximately 20 m, taking shelter behind a stor-
age building [36]. The tank exploded within minutes. The tank
pieces and the secondary projectiles formed when the explosion
shock waves impinged upon the nearby structures, struck sev-
eral re ghters and a re engine. Two re ghters were killed
and another two were seriously injured.
3.1.6. Turkey farm, Albert City, lA
On 9 April 1998, two re ghters were killed and seven seri-
ously injured in a BLEVE that also caused a massive destruction
of infrastructure [36]. The incident began at 11:10 p.m. when a
re was accidentally started at a large Turkey farm near Albert
City, IA. The re had begun when teenagers driving a vehicle
struck two pipelines carrying liquid propane from an 18,000 gal
tank to two vaporizer units. The ensuing cloud of vapour was
ignited by a nearby ignition source. The teenagers were able to
escape the area prior to ignition and rang up the Fire Department
which rushed its team within 11 min of the rupture.
The re ghters quickly set up operations to protect the
exposed buildings with hose lines. There was no water supply
in the area so a tanker shuttle operation was established, with a
portable tank left at the scene.
Two re ghters advanced a hose line from the engine setup
northwest of the jeopardized LP tank. They positioned them-
selves at the west corner of the storage building immediately
north of the LP tank.
Another group of two re ghters advanced a hose line from
the engine staged north of the LP tank between the building
north of the tank and the large coop east of the tank. These men
were approximately 27.5 m north from the LP tank.
The venting gas from the LP tank created a loud noise sim-
ilar to a jet engine, making communications on the re ground
difcult. The re chief indicated that the plan was to allow the
re to burn itself out and to protect exposures.
As this strategy was being implemented at approximately
11:28 p.m. a tremendous explosion occurred, sending large sec-
tions of the LP tank ying in four different directions.
The largest portion of the tank, a piece approximately 7.3 m
long, was hurtled over 91.5 m into the large coop east of the
LP tank. Another piece was propelled directly north, narrowly
missing the two re ghters positioned north of the LP tank.
This piece went through the north building and was stopped by
a silo over 46 m from the LP tanks original location. The force
of this piece passing by the two re ghters carried one of the
men into the building and up against the far wall. He crawled out
of the wreckage and re-joined the others. The third large piece
traveled northwest from the LP tanks location and struck the
two re ghters operating the hose line at the west corner of the
north building. The impact killed the two re ghters instantly.
This piece also narrowly missed the re chief as he stood near
the two men that were killed. He was burned badly by the blast.
Other pieces of the LP tank were scattered in the open eld
across the street from the tank. Some traveled almost 80 m from
the site of the blast. A piece of one of the vent pipes was found
embedded over 1 m deep into a gravel driveway over 61 m west
of the tanks original location.
Two re ghters were killed, and the re chief, ve re ght-
ers and a sheriffs deputy were injured in the blast.
3.2. BLEVEs during transportation of LFG
3.2.1. Railroad tank cars, Laurel, USA
On 25 January 1969, a freight train with 15 tank cars of
LFG derailed in the centre of Laurel, Mississippi. The impact
caused a crack in the tank which was then torn apart in a
BLEVE. It ruptured another tank which also exploded. Both
explosions generated reballs. During the next 40 min one car
after another burst or rocketed. The initial reball set re to
buildings 200400 ft away. Other res were initiated by burn-
ing fragments up to 10 blocks away. All structures lying within
about 400 ft of the accident site were damaged. Window-panes
were shattered in buildings located as far away as 5 km. Two
people died and 976 were injured.
Much of the damage was caused by the impact of rocketing
tank cars and the res which they set going. One 37 ft section
traveled through the air and bounced rst at 1000 ft distance,
then bounced again at 300 ft, and again at 200 ft, and nally
went another 100 ft, before coming to rest at a total distance of
1600 ft, where it set re to houses [3].
3.2.2. Tractor tank, Eagle Pass, USA
On 29 April 1975 a tractor tank carrying LPG on a highway
near Eagle Pass, Texas, swerved to avoid a car in front which had
slowed suddenly to make a turn. The tank semi-trailer separated
from the tractor, struck a concrete wall and ruptured, releasing
LPG. Witnesses described a noise like that of a violent storm,
followed immediately by an explosion, re, and a second explo-
sion.
The large front section of the tank rocketed up, struck an
elevated sign, traveled 1029 ft and struck the ground; bounced
up and traveled 278 ft; struck and demolished a mobile home;
bounced up again and traveled 347 ft over another mobile home,
causing it to burst into ames and be destroyed; nally rested
1654 ft from its starting point [3].
Sixteen people, including the driver, were killed and 512 suf-
fered burns.
3.2.3. Highway tunnel, Palermo (Italy)
On 18 March 1996, a tank truck was involved in a car crash
in a highway tunnel near Palermo, Italy. The sequence of events
was as follows [37]: (a) a vehicle skidded about 100 m from the
tunnel entrance, causing a pile-up of cars; (b) the engine of one
of the cars caught re as a result of collision with another car;
(c) a tank truck entered the tunnel and stopped about 50 m from
the exit to avoid collision with the cars ahead; (d) a bus, arriving
at high speed, skidded and crashed into the tank truck, causing
a leakage in the upper part of the tank shell, just below the
manhole. A few seconds later a soft rumble was heard, followed
by a hot wind that caused serious burns to the people in the
tunnel; (e) everyone ran out of the tunnel, except for ve who
had fainted as a result of the crash; (f) 4 min later the tank went
through a BLEVE. The resulting blast wave seriously damaged
the cars in the tunnel and killed the ve persons remaining there.
Dense black clouds billowed fromthe tunnel. Many relics of the
explosion are still found on the tunnel walls.
3.2.4. Highway, Tivissa (Spain)
The accident took place on 22 June 2002on a highway near
Tivissa, Catalonia (Spain). Atanker carrying compressed natural
gas lost control while speeding downhill. It turned over, tipping
onto its left side, and nally came to a halt beside a sandy slope.
Immediately ames appeared between the truck cabin and the
trailer due to the ignition of either leaking diesel, or CNG, or
perhaps both. The re increased in size as, perhaps, the CNG
escaping fromthe PRVcaught re. There was a small explosion,
then a strong hissing sound, and then the tank BLEVEd, gen-
erating a huge reball [38]. Even though the accident occurred
in a remote location, it still led to one death and burn injuries
to two persons who happened to be at points about 200 m away
from the blast site.
3.3. BLEVEs in stationary installations of non-ammable
but toxic chemicals
3.3.1. Phosgene storage tank, Hamburg, Germany, 1928
On 28 May 1928, a tank containing phosgene went through
a BLEVE at the Stolzenburg factory near the harbor area of
Hamburg, releasing an estimated 12 tonnes of the deadly gas.
Eleven people were killed and 171 required hospital treatment.
People were affected by the gas at locations up to 11 miles from
the accident site [4].
If the northerly wind, which later changed to the south-east,
had blown over populous areas, the fatalities would have been
much higher.
3.3.2. Colorant manufacturing unit, Louisville, KY, USA
On 11 April 2003 a tank containing caramel color liquid,
maltodextrin, and water was unintentionally overheated at the
D.D. Williamson & Company, Louisville, KY. The tank had on
two earlier occasions, had been deformed due to misapplica-
tion of vacuum and the repairs had not been certied to meet
the ASME code requirements. It most probably failed due to
structural weakness, as the pressure at the time of vessel burst
was less then the design pressure, and exploded, killing the lead
operator. The explosion propelled the top head of the tank about
100 mto the west. The shell was also propelled off its foundation
and stuck a 12,000 gal aqua ammonium storage tank, knocking
it sideways and resulting in a 12,000 kg leak. The shell then ric-
ocheted, hit the bottom of a ve-story-tall spray drier, toppling
it. Twenty-six persons were evacuated and 1500 were sheltered-
in-place [39].
3.4. BLEVEs during transportation of non-ammable but
toxic chemicals
3.4.1. Tractor tank carrying ammonia, Houston, USA, 1976
On 11 May 1976 in Houston, Texas, a tractor tank semi-trailer
carrying ammonia went through a bridge rail on an interstate
highway and fell some 15 ft onto a freeway. The interchange was
the busiest in the state and at the time the trafc was quite heavy.
The tank, which held 19 tonnes of liquid anhydrous ammonia,
suffered a BLEVE [25].
The ammonia which had ashed off, formed a cloud 30 m
high. It gradually took in air and reached a width of about 300 m
and a length of 600 m. One photograph showed a tail to the left-
hand side, indicating the typical slumping behavior of heavy
gas. It is estimated that the ammonia evaporated and the cloud
dispersed within about 5 min. The driver of the truck and ve
other people were killed, 78 taken to hospital and about another
100 injured. Apart from the drivers, all the casualties were due
to the toxic impact of ammonia.
3.4.2. Freight train transporting chlorine, Montana,
Mexico, 1981 [25]
A train consisting of 38 wagons, including 32 rail tank cars
lled with liquid chlorine was moving down a steep and wind-
ing valley at a 3% gradient when its brakes failed. The train
derailed at over 80 km/h on a bend 350 m beyond Montana sta-
tion, resulting in a pile-up that included 28 of the 32 chlorine
cars. Most were badly damaged and suffered BLEVEs one after
another. One tank car lost its dished end and the shell was pro-
pelled 2000 m. A second was split along its side. A third had a
0.5 mdiameter hole, probably the result of an ironchlorine re,
which could well have resulted from ignition of the cork insu-
lation by red hot brakes. Four other tank cars suffered damage
to their valves, which were ripped off or dislodged so that they
leaked. It is estimated that 100 tonnes of chlorine escaped in the
rst few minutes and 300350 tonnes in all.
Seventeen persons died, 4 in the caboose of the train and 13
from gassing. Another 1000 people were impacted. The vegeta-
tion up the valley was bleached by the gas cloud passing up it;
there was alsodiscolorationsome 50 mdownthe slope andupthe
sides for a vertical distance of about 50 m. The highest concen-
trations appear to have occurred in a strip 1000 m long 40 m
wide.
3.5. BLEVEs involving non-ammable, non-toxic chemicals
BLEVEs leading to multiple fatalities and severe property
damage have occurred even in tanks storing non-ammable and
non-toxic chemicals [35]. For example at Repcelok, Hungary,
nine persons died when a 35-t cylinder containing carbon diox-
ide BLEVEd on 2 January 1969. Another catastrophic failure
of CO
2
storage vessel at the citrus process plant of Procter &
Gamble GmbH at Worms, Germany, caused three deaths and
damage of property worth $20 million [40]. The blast hurled
parts of the vessel over the factorys compound wall and into
the nearby river Rhine. At Spencer, USA, seven persons were
fatally hit when a tank containing overheated water suffered a
BLEVE.
On 28 August 1992 there was a failure of a storage ves-
sel containing liqueed nitrogen at a manufacturing facility in
Japan. The catastrophic failure of the vessel and the resulting
BLEVE caused the collapse of almost half of the factory, dam-
age to the walls of 25 houses and 39 cars, buses and trucks, all
within a 400 m radius. Fragments of the vessel were projected
up to 350 mincluding part of the top head of the outer shell
which was 1.5 mwide and 8 mmin thickness [40]. The estimated
property loss was $5 million.
At Mihama Nuclear Power Reactor, Japan, on 9 August 2004
a large pipe carrying superheated water developed a leak, and
exploded. The resulting two-phase release of superheated water
and steam scorched 11 workers, killing or maiming them. The
antique boiler explosion in Medina, USA, which occurred on 29
July 2001, and the explosion abroad the cruise ship S.S. Norway
at Miami, USA, are both likely to be BLEVEs which occurred
when vessels containing superheated water developed cracks.
The combined death toll of the two accidents was 15, and several
times this number were left seriously injured [41].
4. Initiating events which have triggered BLEVEs and
some of the lessons learnt
From the tables of notable BLEVE incidents compiled by
Prugh [16,17], Lees [3], and our survey of BLEVE events
covering 19952004, we calculate the frequency of causative
events as under:
Fire 36%
Mechanical damage 22%
Overlling 20%
Runaway reactions 12%
Overheating 6%
Vapour space contamination 2%
Mechanical failure 2%
This is a broad assessment, limited to a sample size of 88,
comprisingonlyof some of the major BLEVEincidents that have
occurred across the world from1926 onwards. Unfortunately no
comprehensive catalogue of all, or even most, of the BLEVEs
that have occurred in this period exists.
The indicative frequency distribution pertains to the causes
leading to the rst BLEVE in an accident event. If we count the
rst BLEVE as the initiating event of the subsequent BLEVEs
that often accompany the rst blast, it may perhaps turn out that
a BLEVE is the most common trigger for other BLEVEs!
Among the most important of lessons learnt from the post-
mortem of past BLEVEs are:
(i) Contrary to popular perception [42] BLEVEs are not lim-
ited to ammable chemicals. All pressure liqueed gases
can, and often are, associated with BLEVEs. Boilers which
hold superheated water can also suffer BLEVE.
(ii) All BLEVEs lead to shock waves and rocketing fragments
of ruptured vessels. In addition, BLEVEs cause reball if
the exploded vessel had contained a ammable chemical,
and/or dispersion of hazardous material (such as chlo-
rine/superheated steam/phosgene) if the stored chemical
was non-ammable.
(iii) It is well-neigh impossible to forecast with any certainty
as to how much time a jeopardized vessel may take before
undergoing BLEVE. The time to BLEVE may be a few
seconds to several hours. This fact makes it very danger-
ous for the re ghters to go near a re-engulfed ves-
sel containing a pressure-liqueed gas. There have been
many instances when a vessel has exploded even after the
pressure-relief valves have been venting the vessel for sev-
eral minutes.
(iv) Of all the harmful effects of BLEVE, the one with the great-
est range of impact is of rocketing fragments. Quite often
death of bystanders and secondary accidents in other pro-
cess units are caused by such missiles.
5. The BLEVE theory
How severe and prolonged has to be the re engulfment in
order to precipitate a BLEVE? How long a vessel jeopardized
by re, or mechanical damage would hold itself before falling
apart in a BLEVE? What are the factors that enhance or dampen
the severity of a BLEVE?
Attempts have been made to answer these questions with
the hindsight of past accident analysis, controlled experiments,
and by the application of numerous concepts of chemical and
mechanical engineering. Together, these attempts constitute the
BLEVE theory which is in a continuous state of renement.
5.1. The superheat limit theory (SLT)
Reid [13,21,43] proposed the oft-used, oft-quoted, and oft-
contestedexplanationtothe question, what triggers a BLEVE?
According to Reids explanation, which has come to be known
as the superheat limit theory (SLT), a BLEVE originates as fol-
lows: when a vessel containing a pressure liqueed gas ruptures,
the vapour which was hitherto in equilibrium with its liquid,
begins to blow off. As a result the liquid pressure drops rapidly,
equilibrium is lost, and the liquid is suddenly rendered super-
heated as its temperature is now way above its boiling point at
the accidentally reduced pressure.
There are two ways of reaching the superheat limit or homo-
geneous nucleation temperature. At constant pressure the super-
heat limit is reached when the temperature exceeds a threshold
value, which is the minimum temperature at which the liquid
gets homogeneously nucleated in the absence of nucleation sites.
At constant temperature superheat occurs if there is a sudden
depressurization of a vessel containing a PLG, rendering the
PLG suddenly at a temperature way above its boiling point at
the nowsuddenly reduced pressure. At constant pressure, super-
heat limit temperature is the highest temperature below critical
temperature and at constant temperature it is the lowest pressure
that a liquid can sustain without undergoing phase transition.
In such a situation, instantaneous ash of a fraction of the
liquid and a superheated liquid vapour explosion takes place,
releasing a biphasic liquid/vapour mixture. These events occur
very swiftly, within a few milliseconds. The increase in volume
caused by the instantaneously vaporizing mass of liquid is enor-
mous, which, added to the expansion of the pre-existing com-
pressed vapour, generates a strong pressure wave. The resulting
massive explosion can often shatter the container into several
pieces, and propel the pieces to considerable distances.
The superheat limit theory (SLT) can be explained with the
illustrative examples of ammonia, chlorine, and butane (Fig. 1),
in which the degree of superheat available if vessels contain-
ing these pressure liqueed gases (PLGs) accidentally rupture
at 308 K or 350 K, has been computed. The gure also gives
the pressuretemperature curves for the three PLGs along with
the corresponding superheat limit loci (tangents drawn at the
corresponding points of critical pressure).
It may be seen that the values of boiling point and super-
heat limit at 1 atm pressure are 239.8 and 347.21 for ammonia,
239.1 and 247.22 for chlorine, and 272.7 and 362.61 for butane,
respectively. When the sudden depressurization takes place due
to rupture of the vessel, the liquid inside the vessel which
was in equilibrium with vapour suddenly comes to atmospheric
pressure, losing its equilibrium. Depending upon the degree of
superheat (which, in turn, depends on the initial temperature)
as shown in Fig. 1, violent ashing would take place generat-
ing a pressure wave. As per SLT, the intensity of the pressure
wave thus generated would depend upon the degree of superheat.
Thus, according to the SLT, the severity of a BLEVE depends
upon the degree of superheat which, in turn, depends upon the
Fig. 1. Pressuretemperature curves and superheat limit loci for ammonia, chlo-
rine, and butane, showing degrees of superheat at different stages of container
rupture.
temperature of the PLG before the sudden rupture of its con-
tainer.
5.2. Determination of SLT
Even as a precise knowledge of superheat limit temperatures
(SLTs) of the substances used in nuclear power plants and chem-
ical process industries is essential for designing and controlling
the operations involving those substances [44], for reasons elab-
orated below it is exceedingly difcult to accurately determine
SLT by experimentation.
5.2.1. Experimental approaches
The most commonly used experimental approach for deter-
mining SLT is based on the droplet explosion technique intro-
duced by Wekeshima and Takata [45] and Moore [46]. This
technique involves introducing small droplets of the test liquid
up the bottom of a column lled with a lesser volatile (host) liq-
uid. The host liquid must be immiscible with the test liquid, and
should have a boiling point well above the critical temperature
of the test liquid. A temperature gradient is maintained in the
column, progressively greater from the bottom up. As the test
drop rises through the column it is progressively overheated till
it reaches its SLT at which stage it explodes into vapour. The
temperature at which the drop explodes is recorded.
The main problemassociated with the droplet explosion tech-
nique is that when the temperature approaches the superheat
limit the unstable bubble nuclei attain sizes reaching molecular
proportions and it is probable that even in the nest experi-
ments, microscopic uctuations would initiate the formation
of a nucleus bubble at a temperature below the superheat
limit. Another problem to be considered is thermal equilibrium
between the rising droplet and the host liquid. The increase in
temperature of the rising droplet may depend on the droplet size,
rise velocity of the droplet, temperature gradient in the host liq-
uid column, and the heat diffusivity of the droplet. There is the
likelihood of under cooling of the droplet which can introduce
some error in the measurement of SLT.
Nevertheless, the droplet explosion technique is the best
available method for the experimental determination of SLT and
the uncertainties can be minimized by sophisticated experimen-
tation.
Another experimental method for determining SLT is based
on immersion of very ne wires in a very clean test liquid and
subjected to very rapid heating. The temperature of the liquid
adjacent to the wires at the time of explosive vaporization of the
liquid is noted. For this approach to succeed the dynamics of the
thermocouple/temperature measuring device must be faster than
the dynamics of the heating of wires for precise measurements,
which is not always possible. The nature of the wire surface
also inuences nucleation. Nevertheless, as in case of droplet
explosion technique, this method also gives fairly reproducible
results, though not with the most desirable accuracy and preci-
sion.
5.2.2. Theoretical approaches
Once the temperature of a liquid crosses the superheat limit,
the liquid must totally vaporize. This can occur only if the liquid
gets homogenously nucleated across its entire body. The factors
which precipitate homogeneous nucleation in a superheated liq-
uid are dealt by the homogeneous nucleation theory which has
its roots in the work of Gibbs [47]. He had outlined and dened
the minimum work required for the formation of vapour voids
in the body of a pure liquid. The theory was subsequently devel-
oped by Volmer and Weber [48], Farkas [49], Becker and Doring
[50], Blander and Katz [51], and others [52,53].
The classical theory can be used to explain the causes and
some of the fundamental mechanisms associated with the explo-
sive boiling in liquids. It can also be used to predict the bubble
size and steady-state nucleation rates. But theoretical results are
often at variance with the experimental data; the later, in itself,
could harbor some uncertainty as discussed earlier. Attempts
to improve upon the classical homogeneous nucleation the-
ory, especially in terms of modeling the behavior of droplets
consisting of 100 molecules or fewer have led to complex
statisticalmechanical treatments but even as these treatments
rest upon a more elaborate theoretical foundation, they do not
give any better agreement with the experimental data then the
classical models [54]. The situation has been summarized by Li
et al. [53], noting that even though the classical kinetic theory
presents some difculties in dealing with nanoscale nucleation
embryos, the theory is the only one that connects the phase
change mechanism in microscale with the observed macroscale
phenomena.
In summary, in spite of continuous advancements in the
homogeneous nucleation theory, occurring in tandemwith more
and more ne experimentation [55], it is not possible to use the
theory in predicting superheat limit temperatures with reason-
able degree of certainity. We have recently made an attempt
to develop a theoretical framework with which SLT of various
substances can be predicted. The details are reported elsewhere
[44].
The SLT had aimed to provide a framework with which the
likelihood as well as the severity of a possible BLEVE can be
predicted. The logical safety strategy, as per SLT, is that the
pressure relief valve of the PLG vessel should be set to operate
in such a manner that it depressurizes the vessel before the vessel
contents cross the superheat limit temperature.
5.3. Exceptions to the superheat limit theory (SLT)
A great deal of work, notably by Birk and co-workers
[56,2,19,31], McDevitt et al. [15], Prugh [16,17], Venart and co-
workers [26,20,32,57], besides others, has brought to the fore
the limitations of SLT. It is more or less agreed that signicant
superheating of the pressure liqueed gas (PLG) is necessary
for a BLEVE to occur [5,58,59] but more and more evidence has
piled up which indicates that a BLEVE can occur well below
the SLT. Interestingly Reid himself has observed (as quoted
in McDevitt et al. [15]) that a vessel would certainly undergo
BLEVE if it suffers LOC at the SLT; BLEVE may still occur
but with less than 100% certainity if the initial temperature of
the liquid is below the SLT.
Prugh [16,17] has pointed out on the basis of a study of past
accidents (for example McDevitt et al. [15]) that BLEVES can
occur well below the superheat limit temperature but the blast
effect may be much higher when a BLEVE occurs near the SLT.
As superheat limit temperature (SLT) is a focal reference
point in the matters of BLEVE, it may be pertinent to dwell
upon it a little.
5.4. Boiling liquid collapsed bubble explosion (BLCBE)
Underlining the complexity of the mechanism of loss of con-
nement (LOC) an occurrence essential for a BLEVE Venart
et al. [14] proposed the concept of boiling liquid compressed
bubble explosion (BLCBE; later the word compressed was
replaced by collapsed in Yu and Venart [26]) to explain the
severity of some of the past BLEVEaccidents. BLCBEhas been
postulated [26] to result froma complex multi-step adaptive and
coherent bubble formation-growth-collapse process in a pres-
sure liqueed gas and its interaction with the containment vessel,
as follows: (i) a partial vessel failure (i.e. a sub-critical sized
crackor opening), (ii) rapiddepressurizationof analreadynucle-
ated and now superheated liquid, (iii) rapid bubble growth and
then the constraint of the expanding two-phase system(by either
physical, acoustic, or inertial means), (iv) the repressurization,
back to nearly the original containment pressure (or values in
excess) followed by, (v) adaptive and coherent bubble collapse
resulting in the formation of a power amplied liquid shock
wave. As a result (vi) wall-pressure wave interaction causes
the total and rapid vessel destruction, with (vii) an explosive
mechanical distribution of the liquid contents as a nely divided
aerosol, and (viii) heat transfer and total evaporation (and if
ammable, auto-ignition) of the aerosol.
The authors further postulate that the bubble growth and col-
lapse phenomena results in a power amplication of the bubble
energy and hence dynamic pressures which may greatly exceed
the original thermodynamic containment pressure dictated by
its original temperature. It follows that a BLCBE may cause
much greater blast effect than may be indicated by the extent
of superheating of the vessel contents just before the loss of
connement.
The BLCBE concept apparently had some of its roots in the
homogeneous nucleation theory of superheated liquids [52,53].
Renement of the BLEVEconcept in the light of the information
available from the past accidents, and substantiation by con-
trolled experiments, has led Venart and co-workers [20,32,57]
to suggest that a complex two-step failure process involving
uidstructure interaction may be the cause of all the BLEVES.
The rst stepis a crackinthe vessel causinga leakbefore break.
In the second step there are waves of depressurization, repressur-
ization and crack propagation leading to the nal failure. Based
on experimentally simulated BLEVEs, Birk, and co-workers
[2,6062] have also developed the theory of two-step BLEVE
described later in this section.
According to Venart and co-workers, the severity of the nal
failure may not necessarily be a function of the extent to which
the contents get superheated but may have more to do with the
initiating mode of the vessel failure and the thermo-hydraulic
contents of the nal failure. The delay times, between crack ini-
tiation and catastrophic failure, range from about 40 s to 1.4 s
as ll increases from 20% to 85%. The distribution and ashing
of the loading causes a reball if the contents are ammable
and the surface emissive power of the these do not appear to
be directly related to the superheat of the contents at failure
and indeed may be most severe for conditions when the ves-
sel BLEVEs while undergoing a pressure reduction at low
superheat.
According to Venart [20], the possible reasons for the nal
rapid failure of the vessel may be either structural instability
of the vessel, rapid over-pressurization due to a dynamic head
space impact of the two-phase swell initiated upon depressur-
ization (initiated by the formation of a thermal crack or tear
which arrests), and/or the rapid quenching of its crack tip, due
to the two phase discharge, that results in large local thermal
stresses which case the uncontrolled vessel failure. The size,
shape, and radiation intensity of the reballs which are formed
when ammable liquids undergo BLEVE also do not appear
to be directly related to the liquids superheat but rather the
thermal-hydraulic contents just at failure.
The studies of Venart and co-workers support the observa-
tions made earlier by McDevitt et al. [15] and Prugh [16,17]
with the major modication that BLEVEs can not only occur
well belowthe SLT, but also the destructive power of the BLEVE
may be unrelated to the extent of superheat.
5.5. Time to BLEVE
Once a vessel containing pressure liqueed gas (PLG) suffers
from a minor or major failure, it can lead to a BLEVE even if
pressure relief valves (PRV) may be operating [23,32]; indeed
if the PRVs are wide open and the PLG is escaping so fast that
siren-like sound is generated, it is a signal that BLEVEcan occur
any moment. But in most other situations there is no simple
indicator that can tell that a jeopardized vessel would eventually
suffer a BLEVE and, if so, when. During the accidents that
have occurred in the past, some vessels have exploded within
a few minutes of re engulfment or missile hits, some other
have done so after several hours. In some cases vessels have
suffered a BLEVE as many as 24 h after being jeopardized!
At the Feyzin BLEVE, described earlier in this paper, the time
between ignition of the leak and vessel rupture was about an
hour and a half. At the PEMEX LPG Terminal catastrophe of
1984, also referred earlier in this paper, vessels took anywhere
between 3 min and 30 min before exploding.
It is important to estimate the time that may elapse between
the initial jeopardization of a vessel and its eventual BLEVE
as this may help in devising damage-control strategies. Bly-
thing and Reeves [63], studying horizontal cylinders, 75% full
with butane and suffering partial re engulfment or jet ame
impingement, estimated that BLEVE would occur between
4 min and 48 min. Selway [33] obtained 711 min, 2538 min,
and 5.57 min as time spans likely for a full 1000 tonnes LGP
storage sphere to experience a BLEVE if it has suffered total
re engulfment, partial re engulfment, and jet ame impinge-
ment, respectively. According to Selway these times would be
shorter if the vessel is less than full. In vessels subjected to
jet re by Roberts et al. [29], the pressure relief valves opened
after 12 min of re impingement and the vessels BLEVEd after
another 3 min.
In re-induced or projectile-induced vessel damage, a tear
can propagate long enough to induce a BLEVE or can stop short
leading to a transient jet release [19]. The second of the situations
can yet lead to a BLEVE if the crack restarts again. This kind of
event has been named two-stage BLEVE [64].
Birk and Cunningham [2,19] conducted a series of tests with
400 l propane tanks, subjecting the tanks to re engulfment and
studying the pattern and the duration of vessel failure. They
observed that a tank will suffer total loss of connement (TLOC)
when the pressure stresses in the tank wall exceed the level
required to propagate a fracture along the entire length of the
vessel. The energy necessary for the TLOC is supplied by the
vapour and the liquid in the tank. Of these, the vapour space
energy is available immediately on initial failure of the ves-
sel. However, the liquid energy is only available after a phase
change (i.e. the liquid must boil or ash to generate vapour),
which requires a nite period of time. Due to very rapid depres-
surization, the boiling proceeds in a non-equilibrium fashion
and very large liquid superheats can occur as a result. These
large superheats can result in very energetic and powerful boil-
ing which could theoretically cause rapid pressure recovery or
overshoot in the tank. It was also observed that BLEVEs of very
weak tanks were very short-duration events and were driven by
vapour space energy. The liquid temperature was not important
in these types of BLEVEs. Indeed high-speed video of BLEVE
occurring in a weak tank when the average liquid temperature in
the tank was 20
C, revealed that complete tank failure occurred

even before the liquid had begun to react. Evidently the energy
for the failure was derived almost exclusively from the vapour
space.
In contrast, Birk and Cunningham [2,19] suggest, very long-
duration BLEVEs of stronger tanks are possible, and these are
driven by violent boiling in the tank after initial tank failure. The
phenomenon was later named two step BLEVE (see following
section). The vapour space energy may only be contributing to
the initial failure of the vessel and might play little or no role in
the long-duration BLEVEs.
Between the two extremes of the BLEVE process, there are
intermediate events in which both the vapour space energy and
the liquid energy contribute to the BLEVE. These intermediate
events have durations greater than the short-duration BLEVEs
but much less than the long-duration events. A summary of the
test results acquired by Birk and Cunningham [2] is presented
in Table 2. The authors observe that if the energy stored in the
compressed gas in the vapour space were large enough and the
tank weak enough, the initial crack would not stop and would
grow until a BLEVE occurred. In the tests conducted by them,
this type of BLEVE was very rapid and the tank failure process
was over in less than approximately 10 ms. The duration sug-
gests a crack-propagating velocity of approximately 200 ms
1
which is in reasonable agreement with the crack velocities in
propagating shear failures of pipelines.
BirkandCunningham[2,19] further observe that if the energy
in the vapour space were not enough to force TLOC, then the
crack could stop. In such a situation, the liquid energy content
may play a major role in determining whether a BLEVE would
eventually take place. The initial crack (now arrested) would
result in venting and depressurization and this may lead to a
boiling of the liquid with a severity which would depend on how
muchsuperheat is generatedbythe initial depressurization. If the
initial liquid temperature is high enough, the boiling response
and its associated pressure transient may be sufcient to restart
the crack and cause TLOC and BLEVE. A BLEVE may also
be preceded by a jet release of signicant duration. Such a jet
release may have a signicant impact on the geometry of the
resulting reball, and its rapid rise.
The long-duration BLEVE is a transition event between a
BLEVE and a non-BLEVE (i.e. partial failure and jet release)
and is probably fairly rare, requiring specic conditions of tank
and lading properties. About 20% of the BLEVEs observed by
Birk and Cunningham [2] were of long duration. Such events
probably occur in tanks that are almost strong enough to resist
TLOC, but are pushed over the edge by the violent boiling
of the depressurization-induced superheated liquid. The long-
durationBLEVEs are stronglyrelatedtothe PRVsettingbecause
this ultimately determines the liquid temperature.
The maximum possible liquid superheat that can be attained
occurs when the liquid is at the atmospheric superheat limit
(53
C for propane) and this could potentially lead to the most

violent boiling response. Birk and Cunningham [2] have drawn
Table 2
Summary of test outcomes [2]
Test Fire conditions No. of events Outcome Comments
Vent to empty
Pool only 5 Entire tank contents lost through the PRV Fire conditions were not severe enough to
initiate a local failure
Pool and torch 2
Partial failure
Pool and torch 7 Partial failure occurred in the tank
causing a two-phase liquid/vapour
mixture to be released from the tank
Local thermal weakening was such that
pressure stresses in the wall exceeded local
tank wall strength causing a failure (crack)
Torch only 4 In two cases, the jet was angled such that
it propelled the tank up to 30 m
The tank wall was strong enough to arrest
the crack
TLOC and BLEVE
Pool and torch 9 The tank failed catastrophically releasing
the entire contents as expanding vapour,
boiling liquid and dispersed droplets
Not all BLEVEs were the same
Torch only 2 Failures resulted in blast and reballs The characteristics of BLEVEs varied with
tank properties and lading conditions at the
time of failure
a pressuretemperature diagram for propane which shows the
theoretical superheat limit spinodal for propane in the context
of the results of BLEVE tests done by the authors. It was found
that BLEVEs occurred in almost all of the tanks above the atmo-
spheric superheat limit, andsome of these were the long-duration
or transition type. This suggests that, even if the tank is strong
enough to survive a partial failure, if the PRVis set near or above
the superheat limit, then the boiling response will push the event
into a long-duration BLEVE. This was rst suggested by Reid
[21].
Birk and Cunningham [2] and Birk et al. [23] have also pre-
sented a BLEVE map which displays the relationship between
the estimated burst strength of the vessel and the liquid energy
at the time of failure. The map shows that the BLEVE and non-
BLEVE situations occur in distinct regions which are divided
by a line with a positive slope indicating that increase in liquid
energy per unit volume of the tank will result in an increased like-
lihood of BLEVE occurring in strong tanks. In addition, for the
cases where the tank strength is relatively high, there appears to
be a transition region between BLEVEs and non-BLEVEs. This
transition region contains all BLEVEs of long duration where it
is the liquid boiling response that drives the TLOCand BLEVE.
The BLEVE maps also carry a shaded region of uncertainty
because no test results appear in these areas. Of this shaded
region, the most interesting is the part at higher liquid tempera-
tures. If this region proves to be part of the BLEVE region, then
it suggests that the liquid temperature is very important and that
the PRV set pressure selection is of critical importance.
With additional data and analysis, Birk and Cunningham
[2,19] and VanderSteen and Birk [61] have shown that pressure
release valve (PRV) set pressure and vessel wall thickness have
a signicant impact on the probability of a BLEVE occurring;
in case of a thin-walled tank with its wall temperature raised to
700
C due to re engulfment, doubling the wall thickness and

reducing the PRV set pressure from 2150 kpsig to 1725 kpsig
would reduce the susceptibility of a tank to BLEVE by a factor
of 3 or more.
Even though these authors, nor others who have studied the
factors responsible for delayed BLEVE, have explicitly corre-
lated the severity of the blast effects with time to BLEVE, it
appears from the discussions that some of the delayed BLEVES
may be as severe; or even more so, than instantaneous BLEVES.
A different postulation has been made by Van den Berg et al.
[65], on the basis of acoustic and gas dynamic blast modeling
validated by them using the data of Giesbrecht et al. [66] on
exploding vessels of liqueed propylene. According to Van den
Berg et al. [65], a rupture in a vessel containing a pressure lique-
ed gas in free space develops a blast of signicant strength only
if the vessel disintegrates nearly instantaneously; if the rupture
and the catastrophic vessel failure are delayed even for a short
duration, the blast effects are minor.
5.6. Two-step BLEVE, pressure relief value (PRV) blow
down, thermal stratication and effect of thermal insulation
Further light on BLEVE mechanism, which provides useful
insights for BLEVE prevention, has been shed by the recent
re tests, on PLG-containing vessels, by Birk et al. [23], Birk
and VanderSteen [64], Roberts et al. [29], Stawczyk [58] and
VanderSteen and Birk [61]. The tests of Birk and VanderSteen,
which augment the ones done earlier by Birk and Cunningham
[2], and Kielec and Birk [60], indicate that for a BLEVEto occur,
the tank must rst be weakened to the point where an initial pin
hole rupture is formed. For the case of a re heated tank this
narrow rupture may form due to plastic thinning of the wall in
the heated area or it may form at a aw in the tank wall. This
opening must then grow to cause a total loss of containment
(TLOC), and BLEVE.
The initial rupture normally grows in a direction perpendicu-
lar to the principal stress (in this case hoop stress). As the rupture
widens, vapour initially escapes from the tank and the pressure
drops. If the tank has been rendered very weak due to heating
over a large area then the crack may rapidly (>200 m/s) grow
across the full length of the tank to give a TLOC and BLEVE.
In this case the failure may be so rapid that the liquid may not
have enough time to undergo phase change and pressurize the
tank wall. In other words the TLOC may be caused solely due
to the wall loading by vapour space energy.
In other cases the crack may stop in stronger material or it
may stop because of the decreased pressure in the tank. In this
case the vapour space energy is spent without failing the ves-
sel. But the stoppage in the crack propagation may not mean
the accident is over. With the pressure dropping in the tank, the
liquid may go to a state of superheat. If the pressure drop is very
rapid (large hole and/or small vapour space) a signicant amount
of superheat may be established in the liquid before the bulk of
the liquid responds by ashing. This and the resulting pressure
transient may cause sufcient loading of the tank wall to restart
the crack and cause TLOC. It has been shown in many experi-
ments that there can be a signicant pressure recovery in a tank
holding a pressure liqueed gas after initial depressurization by
a rupture. Once a massive ash response is initiated the rapid
phase change can send liquid droplets up to impact the top of
the tank wall [67]. The ashing may cause liquid to be entrained
into the vented stream and this two-phase material may reduce
the vented material enthalpy ux to the point where the pres-
sure may begin to recover in the vapour space. The increase in
pressure from the pressure recovery may add to the wall loading
and may contribute to restarting the failure crack and sending
the tank to TLOC and BLEVE.
In BLEVE accidents in which the crack has stopped and
then is restarted may be called a two-step BLEVE. These kinds
of BLEVEs are deemed to be in a transition region between a
nite failure and a rapid BLEVE (i.e. the failure process almost
stopped before a TLOC; Fig. 2). On the basis of the eyewit-
ness accounts of a BLEVE involving a LNG road tanker which
occurred at Tivissa, Spain, on 22 June 2002, Planas-Cuchi et al.
[38] opine that it was a two-step BLEVE. There was a minor
explosion, then a strong hiss, and then a major explosion, sug-
gesting that an initial crack had formed by thermal stress at a
very hot location on the LNG tank wall which was arrested in
a cooler and stronger zone, followed by a discharge (probably
two-phase owwhich created the hiss). The crack then restarted
due to further thermal stress at the crack tip originated by the
cooling effect of the two-phase release through crack, leading
to the catastrophic BLEVE.
Birk and VanderSteen [64] further observe that if there is only
local heating of the vapour space wall then the crack is likely to
stop in stronger cool material. If the nal hole size is kept below
some critical size then the pressure forces on the aps produced
by the failure opening may not be sufcient to propagate the
failure. However, if the hole size is larger than the critical size
then the crack can sometimes be restarted and driven through
cool strong wall material.
Fig. 2. Two-step BLEVE [2,64].
The tests on nine, 1.8 m
3
propane tanks [64], all started 80%
full with commercial propane at 1020
C, revealed that:
(i) The time to initial failure depends on the re condition
and on the design of the tank and pressure relief system.
A large blow down PRV may delay a failure due to the
reduced average stress state in the tank.
(ii) If a rupture takes place in a vessel holding a liquid at or near
its atmospheric superheat limit, it may not always produce
a BLEVE; the vessel must open completely if a BLEVE
has to occur, and this will only happen if the tank has been
weakenedsufcientlytoinitiate a rupture andif the pressure
transient during failure is sufcient to drive the failure crack
to fully open the vessel.
(iii) If the length of the severely heated (weakened) part of the
vapour space exceeds some critical value, then a BLEVE
outcome is likely over a range of ll conditions (1050%in
the present tests). This critical length was around one diam-
eter in the tests of Birk and VanderSteen [62]. However, if
the heated zone is smaller than this critical value (in these
tests the smaller heated zone was 0.8D) a BLEVE will not
happen for lower liquid ll levels. It is possible that higher
ll levels with higher liquid energies may cause a BLEVE
failure even with small heated zones.
(iv) It was possible to produce a plot based on a modied
Folias parameter that divided BLEVE from non-BLEVE
outcomes.
In the course of BLEVEs simulated by heating LPGcylinders
of 5 kg and 11 kg capacities, Stawczyk [58] has observed that
during the initial stage of the tank heating process the pressure
within the tank followed the vapour pressure curve for propane.
At 85
C and 35 bar the tank pressure began to deviate from the

curve, apparently due to the exceeding of the critical tank load.
The vessels failed at temperatures >115
C when 80% full and

>150
C when 40% full; the internal pressure before the vessel

failure was in the range 75120 bar. The tank failure began with
a precipitous drop in the tank pressure due to the onset of tank
unsealing. The author assumes that adiabatic expansion of super-
critical propane then took place. Part of the supercritical uid
vaporized through the crack which had developed. The pressure
drop caused by the crack rendered the liquid in a superheated
state which then boiled over. It led to violent pressure increase
as boiling generated far more vapour than could escape through
the small opening in the tank. The tank shattered explosively.
Stawczyk [58] further notes that internal tank pressure does play
an important role in the formation of the initial local creep rup-
ture; in a thermally weakened tank, a failure will occur when the
local pressure-induced hoop stress exceeds local strength of the
tank wall material.
BLEVEs were simulated by jet re impingement on 2-t ves-
sels, lled to the extent of 20%, 41%, 60%, and 85% with
propane by Roberts et al. [29]. It was observed that the ves-
sels failed within 5 min of jet-re impingement and at pressures
ranging from 16.524.4 bar. In all cases except for the 20% full
vessel there was an initial pressure drop on pressure relief valve
(PRV) opening, followed by a pressure shoot up to above PRV
opening pressure. For the 20% full vessel, the pressure fell from
18.6 bar to 16.5 bar. In all cases the temperature of the vessel
wall just above the liquid level was much higher than just below
the liquid level suggesting that there was little liquid swelling,
hence wetting of the wall. On vessel failure considerable portion
of the liquid immediately ashed to vapour which ignited giving
large reballs.
It has been surmised [6870] that when a re engulfs a PLG
tank, the liquid near the wall is heated, becomes less dense and
rises to the top, rendering the liquid thermally stratied. As this
top liquid dictates the tank pressure, the actual tank pressure in
such a situation would be higher than the one expected from the
average liquid temperature.
Another series of tests, simulating tank cars jeopardized by
pool re [61] in which the cars were fortied with thermal pro-
tection systems with or without defects of varying magnitudes,
have revealed:
(i) Using only the steel jacket provides thermal protection
as it behaves as an effective radiation shield. In these
tests, the jacket was found to cut the wall heating rate to
approximately half of an unprotected wall (53
C/min ver-
sus 24.5
C/min) assuming that there is an air gap between

the wall and the jacket.
(ii) Thermally protecting the tank wall with a steel jacket and
a blanket of 13 mm ceramic insulation provides substantial
thermal protection. The fully protected tank may reduce the
wall heating rate to approximately 1/10th of an unprotected
wall (53
C/min versus 5
C/min).
(iii) Larger defect sizes resulted in higher average and peak wall
temperatures in the area of the defect for a given heating
time. As defects got larger, the peak temperature of the
defect approached the wall temperature with no insulation
(i.e. very large defect). The transition defect length was
found to be about 40 cm. If the defect was smaller than
40 cm, the peak temperature was reduced by the surround-
ing protected material. If the defect was larger than 40 cm,
there was little or no benet from the surrounding material
as far as peak temperature was concerned.
All said and done, safety professionals still echo the words of
Leslie and Birk [18] that even as BLEVEs are widely accepted
as being the most damaging of the pressure-liqueed gas release
accidents, we still do not have accurate methods of predicting
when and where they will occur. Nor can we say with much
certainty how quickly a jeopardized vessel will BLEVEin the
past, vessels have taken anything between a fewseconds to a few
hours before BLEVE occurred. What is beyond much debate
is that the damage caused by a BLEVE is much larger than
of a vapour cloud explosion involving an identical material in
identical quantity [4,6,71,72].
6. Impact or consequences of BLEVE
6.1. Introduction
The assessment of the impact or the consequences of a
BLEVE revolves round two factors:
5
0
4
T
.
A
b
b
a
s
i
,
S
.
A
.
A
b
b
a
s
i
/
J
o
u
r
n
a
l
o
f
H
a
z
a
r
d
o
u
s
M
a
t
e
r
i
a
l
s
1
4
1
(
2
0
0
7
)
4
8
9
5
1
9
Table 3
Methods of estimating BLEVE explosion energy
Identication
of the method
Basis of the
method
The key
expressions
Explosion energy estimates
relative to Prughs method (kJ)
Reference
Prughs method or the
TNT equivalent method
(1) Assumes that the ashing fraction
of the liquid and the pressurized gas
expand isentropically as an ideal gas
in a BLEVE; (2) equates the work
done by the expanding vapour with a
charge of TNT; (3) Works out the
explosion energy, W
TNT
, as if it is
caused by that charge of TNT
W
TNT
=
2.4 10
4
PV
k 1
_
1
_
101
P
_
(k1)/k
_
where
V
= V
T
+W
L
_
f
D
V,T
1
D
L,T
_
, f = 1 exp
_
C
L
(T
c
T
b
)
_
,
= 2.63
_
1
_
T
c
T
0
T
c
T
b
_
0.38
_
If the burst energy of a vessel
lled with propane is 1 KJ as
per Prughs method, according
to other methods it is:
Prugh [16]
SVEE (specic volume,
entropy and enthalpy)
method
(1) Does not assume ideal gas
behavior; (2) assumes isentropic
expansion; (3) explosion energy, e
ex
,
is computed from the difference in
internal energy of the expanding gas
between its initial and the nal stage
of expansion
e
ex
= m
f1
u
f1
m
f2
u
f2
+m
g1
u
g1
m
g2
u
g2
where
m
f2
= (1 X
f
)m
f1
+(1 X
g
)m
g1
, m
g2
= X
f
m
f1
+X
g
m
g1
,
u
f1
= h
f1
p
1
v
f1
, u
g1
= h
g1
p
1
v
g1
, u
f2
= h
f2
p
a
v
f2
,
u
g2
= h
g2
p
a
v
g2
, X
g
=
_
s
g1
s
f2
s
g2
s
f2
_
, X
f
=
_
s
f1
s
f2
s
g2
s
f2
_
1.1 CCPS [8] and TNO [73]
Irreversible adiabatic
expansion method of
Planas-Cuchi et al.
(1) Does not assume ideal gas
behavior; (2) expansion is considered
to be an adiabatic, irreversible
process; (3) the change in internal
energy due to the adiabatic
irreversible expansion is equated to
the work done by the uid expanding
at constant atmospheric pressure.
The expressions for both are solved
for getting x, the vapour fraction at
the nal state of the expansion
process. This is then substituted in
the expression for change in internal
energy; (4) from the change in
internal energy, the TNT equivalent
mass is calculated
e
ex
= U = P
0
V where U = (u
L
u
G
)m
T
x m
T
u
L
+U
i
,
P
0
V = P
0
[(v
G
v
L
)m
T
x +m
T
v
L
V
i
],
x =
m
T
P
0
v
L
V
i
P
0
+m
T
u
L
U
i
m
T
[(u
L
u
G
) (v
G
v
L
)P
0
]
,
W
TNT
(kg) = 2.14 10
7
U
0.4 Planas-Cuchi et al. [59]
C: Average specic heat of the liquid over temperature interval T
0
to T
b
(J/kg K); D
L,T
: Density of vapour at temperature at burst (kg/m
3
); D
V,T
: Density of vapour at temperature at burst (kg/m
3
); e
ex
, ex: Explosion
energy (J); f: Fraction of liquid ashing into vapour (no unit); h
f1
: Enthalpy of the liquid at the initial state (J/kg); h
f2
: Enthalpy of the liquid at the nal state (J/kg); h
g1
: Enthalpy of the vapour at the initial state
(J/kg); h
g2
: Enthalpy of the vapour at the nal state (J/kg); k: Ratio of specic heats at constant pressure and constant volume (no unit); L: Average latent heat of vaporization over temperature interval T
0
to T
b
(J/kg);
m
f1
: Mass of the liquid at the initial state (kg); m
f2
: Mass of the liquid at the nal state (kg); m
g1
: Mass of the vapour at the initial state (kg); m
g2
: Mass of the vapour at the nal state (kg); m
T
: Total mass of the
vessel contents (kg); P: Pressure in the vessel at the time of burst (KPa); p
1
: Pressure in the vessel at the time of burst (Pa); P
0
, p
a
: Atmospheric pressure (Pa); s
f1
: Entropy of the liquid at the initial state (J/kg K);
s
f2
: Entropy of the liquid at the nal state (J/kg K); s
g1
: Entropy of the vapour at the initial state (J/kg K); s
g2
: Entropy of the vapour at the nal state (J/kg K); T
b
: Boiling point (K); T
c
: Critical temperature (K); T
0
:
Initial temperature of the vessel contents (K); u
f1
: Internal energy of the liquid at initial state (J/kg); u
f2
, u
L
: Internal energy of the liquid at the nal state (J/kg); u
g1
: Internal energy of the vapour at the initial state
(J/kg); u
g2
, u
G
: Internal energy of the vapour at the nal state (J/kg); U
i
: Internal energy at initial state (J); V
*
: Total vapour volume (m
3
); v
f1
: Specic volume of the liquid at the initial state (m
3
/kg); v
f2
, v
L
: Specic
volume of the liquid at the nal state (m
3
/kg); v
g1
: Specic volume of the vapour at the initial state (m
3
/kg); v
g2
, v
G
: Specic volume of the vapour at the nal state (m
3
/kg); V
i
, V
T
: Volume of the vessel (m
3
); W
L
:
Mass of liquid in the vessel (kg); W
TNT
: Equivalent mass of TNT of the explosion energy (kg); X
f
: Fraction of liquid ashing into vapour; X
g
: Fraction of vapour condensing into liquid; : Fraction of explosion
energy converted into a pressure wave; U: Change in internal energy (J); V: Change in volume (m
3
); : Correction for ash fraction f.
Note: Initial state: At the instant of explosion. Final state: After explosion.
(a) The energy of explosion, or burst energy. This determines
the severity of the blast wave generated by the BLEVE and
the velocity (hence the range and the penetration) of the
missiles formed out of the shattered vessel fragments;
(b) The manner of release of the vessel contents. This deter-
mines the size, duration, and heat ux of the reball if the
contents are ammable, or the pattern of atmospheric dis-
persion if the contents are toxic.
Whereas more rigorous treatments are available, and with
greater degree of consensus on their applicability, for explosion
of vessels containing pressurized gases, cased explosives, and
VCE (vapour cloud explosions), there is much greater degree of
uncertainty associated with BLEVE situations wherein super-
heated liquids together with pressurized gases are involved.
When a vessel containing a superheated liquid fails catas-
trophically in a BLEVE, the boiling liquid as well as the
expanding vapour together provide the burst energy but it is
very difcult to estimate which phase contributes how much.
This is because, as detailed in the preceding section, the events
between the initial crack and the cracks propagation up to the
occurrence of BLEVE would inuence the state of both the
phases. Further, a portion of the burst energy is used up in shatter-
ingthe vessel, another portioninpropellingthe vessel fragments,
and yet another in the generation of blast wave. The cooling
effect of the ash vaporization of the liquid and the adiabati-
cally expanding vapour further complicate the scenario. Once a
vessel is shattered, some of the contents can also form transient
pool res by getting splashed on the oor before evaporating.
This may reduce the quantity of the vessel contents which form
the reball.
We present below an overview of the state-of-the-art of
BLEVE consequence assessment.
6.2. BLEVE energy
Classically two treatments have been used to estimate the
burst energy accompanying a BLEVE: the so called TNT
equivalent method developed by Prugh [16] which treats the
expanding vapour as an ideal gas, and the method which relies
on entropy, enthalpy, and specic volume data when treating the
expansion as occurring in a non-ideal gas. The second method
has been described by Prugh [16], CCPS [8], Lees [3], and TNO
[73], among others without giving it any name. As it is a ther-
modynamic method like the Prughs but is distinguished by its
use of specic volume, entropy and enthalpy (SVEE), we give
it the name SVEE method. The manuals of CCPS [1,8] and
TNO [73] mention only the latter but Prughs method continues
to receive consideration owing to its simplicity [74]. Recently,
Planas-Cuchi et al. [59] have proposed a new method for calcu-
lating BLEVE energy on the basis of their contention that the
liquidvapour ashing in a BLEVE ought to be treated as irre-
versible adiabatic expansion rather then as isentropic expansion
an in the Prugh method.
The essence of the three methods is presented in Table 3.
The Prughs and the SVEE forecasts run close to each otheras
presented in Table 3. If the burst energy of a vessel lled with
propane is 1 kJ as per Prughs method, it will be about 1.1 kJ as
per SVEE. But the estimate as per the method of Planas-Cuchi
et al. [59] for the same event yields burst energy less than half
of the Prughs/SVEE methods. Further renement of the burst
energy estimation methods is necessary as the forecast of the
kinetic energy of the vessel fragments, hence the initial fragment
velocity and the fragment striking range, are directly dependant
on burst energy estimates.
6.3. Overpressure
Once the explosion energy of a BLEVE is estimated by one
of the methods summarized in the preceding section (Table 3),
overpressure can be determined by employing the correlations
available in literature which link overpressure with explosion
energy, and the distance from the accident epicenter. CCPS [8]
and TNO [73] use the graphical method of Baker et al. [75],
in conjunction with the SVEE estimate of explosion energy.
But Prugh [16,17] has calculated overpressure by employing
the graphs in CCPS [76] while Planas-Cuchi et al. [59] have
used the graphs proposed by van den Berg and Lannoy [77].
The approaches are summarized below.
6.3.1. The method of Baker et al. [75] as used by CCPS [8]
and TNO [73]
The explosion energy, e
ex
, obtained fromthe SVEEmethod is
multiplied by a factor of 2 if the burst is a ground level, to obtain
the working explosion energy, E
ex
(J). The scaled distance, z, is
then obtained from
z = R
_
P
0
E
ex
_
1/3
where R is the distance of the target from the vessel undergo-
ing BLEVE (m), P
0
the atmospheric pressure (Pa), and E
ex
is
the explosion energy. The scaled overpressure is read from the
curves in which scaled overpressure (P
s
) is plotted versus scaled
distance [75].
Once the scaled overpressure is read from the graph, adjust-
ments have to be made for cylindrical vessels and for vessels
slightly elevated above the ground by using the appropriate mul-
tipliers, as given in Tables 4 and 5.
Table 4
Adjustment factor for Ps for cylindrical vessels [73]
Value of z Multiplier for P
s
<0.3 4
0.33.5 1.6
>3.5 1.4
Table 5
Adjustment factor for P
s
for vessels slightly above ground [73]
Value of z Multiplier for P
s
<1 2
1 1.1
The scaled overpressure, P
s
(which is read off from the
graph), is related to the overpressure (p
s
):
P
s
=
p
s
P
0
1
6.3.2. Overpressure calculation as described by Prugh
[16,17]
Prugh has outlined the following steps to determine overpres-
sure:
1. The maximum overpressure at the surface of the bursting
container (P
sb
, in kPa) is determined from [78]:
P
b
= P
sb
[1 ]
2k/(k1)
where
=
0.035(k 1)(P
sb
101)
[1 +0.058P
sb
](kT/M)
P
b
is the pressure in the vessel at the time of burst (kPa), k
the ratio of specic heats at constant pressure and constant
volume, T the temperature of the vapour in the container (K),
and M is the molecular weight of the vapour. The equation is
solved iteratively to obtain P
sb
.
2. An entity, virtual distance, is next determined by rst read-
ing the scaled distance corresponding to overpressure P
sb
from the curves given in CCPS [76], and calculating the dis-
tance (R) from the scaled distance by using the formula:
z =
R
3
W
TNT
On subtracting the radius of the vessel from R, the virtual
distance, is obtained.
3. To determine overpressure at any distance from the vessel,
the value of the virtual distance is added to the distance
at which the overpressure has to be determined. This value
(distance +virtual distance) is used in the place of R in the
formula for calculating scaled distance. Once the scaled dis-
tance is calculated, the curves given in CCPS [76] are used
to obtain overpressure at the given distance.
6.3.3. Graphs give in van den Berg and Lannoy [77] as
used by Planas-Cuchi et al. [59]
Unlike the graphs of Baker et al. [75] and CCPS [76] the
graphs of Van den Berg and Lannoy [77] relate the scaled dis-
tance to overpressure directly, instead of scaled overpressure.
Thus, for the calculated scaled distance, the overpressure can be
obtained directly from the graphs.
6.4. Missiles
In most BLEVEs, except the ones involving non-ammable
chemicals, reballs are generated along with the explosion. But
the range-of-impact of the missiles which result from the frag-
mented vessel is much larger than that of reball [56,58].
Missiles also pose much greater danger of causing domino
effect than the reball or the blast wave. In disasters like the one
that occurred in the Mexico City, 1984, the rst vessel which
BLEVEd had let off missiles wrapped with burning propane.
These missiles struck other vessels, damaging them, and caus-
ing them to BLEVE. The resulting missiles precipitated further
explosions.
Missiles are also known to be a major cause of death and
destruction. In the Turkey farm episode at Albert City, IA,
described earlier (Section 3.1.6), both the fatalities were by mis-
sile hits, as were the more serious of the injuries. At Deer Lake,
PA, most of the 11 deaths and 10 injuries were caused by the
ying fragments of an LPG road tanker which had BLEVEd.
Many of the hits occurred on persons standing over 200 m away
and beyond the range of the thermal hazard [3]. In the Laurel
railroad tank car accident mentioned earlier (Section 3.2.1), a
missile in the form of a rocketing fragment from one of the tank
cars had hit a pump house, rupturing an 8 in. water main, thereby
reducing the water supply to the re ghters. The BLEVEs at
Puebla, Mexico (1977) and at Texas City (1978) also produced
rocketing fragments which damaged water tanks meant for sup-
ply to re ghters [3].
The likely consequence of a BLEVE in terms of the duration
and propagation of missiles depends on the following factors:
1. likely number and mass of missiles,
2. velocity and range of the missiles,
3. likely direction of propagation of the missiles,
4. penetrative ability and destructive potential of the missiles.
6.4.1. Likely number and mass of missiles
6.4.1.1. Cylindrical vessels. In a study of 27 BLEVEs involv-
ing LPG vessels, the maximum number of missiles per vessel,
was found to be four and it occurred in 15% of the explosions
[79]. The most frequently encountered number of missiles per
vessel was three (37% cases) followed by one (30% cases) and
two (26% cases). These authors have proposed the following
correlation between the number of fragments, and the vessel
capacity (V, m
3
) on the basis of involving vessels a of capaci-
ties 7002500 m
3
: n =3.77 +0.0096V.
Stawczyk [58], in a study of LPG cylinders of 5 kg and 11 kg
capacities, found that each BLEVE generated three to ve main
projectiles and several single, smaller fragments.
As per earlier studies [3] the most likely initiation of the
failure of a cylindrical vessel is in axial direction; the crack may
then turn and propagate circumferentially. Stawczyk [58] found
that in the LPG cylinders studied by him, the upper part of the
cylinder usually detached along with a large fragment of a side
wall; the latter formed two or three projectiles.
6.4.1.2. Spherical vessels. Spherical vessels generate much
larger number of fragments. Astudy of seven BLEVEs involving
spheres revealed as many as 19 fragments in one of the explo-
sions, followed by 16, 6, 5, 5, 4, and 3 in other explosions giving
an average of above eight missiles per explosion [3]. Three tests
by Schulz-Forberg et al. [80] on 4.85 m
3
vessels, half full with
liquid propane, produced three, ve and nine missiles, respec-
tively.
Interestingly, other studies on missiles generated by vessel
bursts, which evidently involve both BLEVE and non-BLEVE
explosions, have reported much larger number of missiles per
explosion than BLEVE-specic studies. Apost-mortemof seven
major explosions that had occurred between 1957 and 1988 [81]
reveals that the explosioninvolving53 m
3
of a central sectionof a
200 m
3
vinyl acetylene distillation column at Texas City, 1969,
generated over 50 fragments. Another explosion at Antwerp,
1987, which had involved a 162 m
3
ethylene oxide distillation
unit, let off 35 fragments. The average number of fragments pro-
duced in the seven bursts analyzed by Scilly and Crowther [81]
was 26.6. Another study of 25 accidental explosions by Baker
et al. [82], leads to an average of 6.6 fragments per explosion.
Three argon-lled spheres produced 14 fragments whereas a
single propane cylinder split into 11.
In general a vessel may undergo either a brittle failure or a
ductile failure; the latter type occurs more frequently than the
former. But, whereas a brittle failure is likely to generate larger
number of missiles then a ductile failure, it is the hits from the
latter that have much greater potential to cause damage [3,83].
6.4.1.3. Missile fragment mass estimation. For fragment mass
distribution, broad estimates may be obtained using the method
of Held [84], developed for cased explosives. The mass m of the
nth fragment is given by
M =
dM(n)
dn
= M
0
Bn
1
exp(Bn
)
where M(n) is the overall (cumulative) mass of fragments of
number n, M
0
the total mass of fragments, n the number of the
nth largest fragment, and B and are constants.
6.4.2. Velocity and range of the missiles
An assessment of the momentum of the rocketing fragments
likely from a BLEVE, and the distance likely to be traveled by
such fragments is essential for determining the likely impact
area of a BLEVE. Apart from designing the layout of a plant,
this knowledge is also very important in deciding how far away
re ghters must be located when trying to save a re-engulfed
vessel from suffering a BLEVE [56,58].
When a vessel undergoes a BLEVE, a fraction of the explo-
sion energy of the vessel is transformed as the kinetic energy of
the vessel fragments which then shoot out as missiles. Exper-
imental determinations on light-weight containers detonated
with TNT have indicated that most fragments fall at distances
between 0.3 and 0.8 of the maximum. The probability that
there would be at least one fragment which would travel the
maximum possible distance increases with the number of frag-
ments and is therefore greater for a large explosion [3]. In
the simulated BLEVEs on LPG cylinders of 5 kg and 11 kg
capacities, Stawczyk [58] observed that the biggest elements
of a cylinder were found at a distance of about 70 m from
the shattered vessel. Flat fragments, and compact elements of
small mass, such as the head, went up to four times farther.
Stawczyk [58] also notes that even though past reports in liter-
ature put 200 m as the maximum distance to which projectiles
from a 11 kg LPG tank can reach, he has found that projec-
tiles from his 11 kg cylinders undergoing BLEVE went up to
300 m.
There are many instances when large chunks of an exploded
vessel have traveled long distances. At Murdoch, IL [9], portion
of a rail container was hurled as much as 1.7 km away after a
BLEVE! In the explosion at the American Oil Company renery
at Whiting, Indiana, one 60-t piece landed on a tank of gasoline,
smashing it severely and igniting and scattering its contents.
Other vessel fragments were scattered for several hundred feet
away from the unit site. In one ship explosion investigated by
Clancey [85] a deck cover of 400 tonne was thrown 100 ft. Dur-
ing the serial BLEVEs at Sydney which occurred on April Fools
dayin1990, a 30 mlongcylinder rocketedafter its topwas ripped
off in a BLEVE. It struck three 40-t tanks, an electrical substa-
tion, and a workshop before nose-diving into a canal 300 maway
from the origin of its ight [86].
For vessels lled with ideal gases, Baums [87] treatment of
Bakers equation for explosion energy [82], used in conjunc-
tion with experimental data, yields a factor in the range 0.20.5
by which explosion energy of a vessel is transformed to kinetic
energy of the fragments in which the vehicle is shattered. But
for non-ideal gases, and for vessels partially lled with lique-
ed gas as in BLEVE situations, this treatment can give grossly
inaccurate forecasts. The Centre for Chemical Process Safety [1]
has suggested the Moore [88] equation for obtaining the initial
velocity of the fragment, (ft/s), emanating from the rupture
of a pressurized vessel as a function of rupture pressure of the
vessel, P (psig), fragment diameter D (in.), and weight of the
fragment, W (lb):
= 2.05
PD
3
W
Fragments with spheroidal shape have the least drag coefcient,
followed by edge-on cubes and end-on rods [1]; missiles with
shapes and orientation similar to these geometries travel the far-
thest. Baum [89] has presented a model for velocities attained
by end-caps of cylindrical vessels upon vessel burst. The model
assumes that the action of the escaping vapour/liquid on the end-
cap is analogous to a missile driven by a gas jet from a constant
pressure source. The missile velocities are derived via a sim-
ple approximation to the impulse applied to the internal face of
the closed end of the rocket. Similar logic was used later by
Baum [90] for predicting the velocity achieved by an axially
split cylindrical vessel. The author also describes experiments
which validate the model.
6.4.3. Likely direction of propagation of the missiles
Given the fact that vessel fragments which are shot off as
missiles upon a BLEVEhave much longer range of impacts than
other hazards caused by BLEVE reballs and shock wave
it may enormously help in BLEVE damage control (following
section) if the most probable directions in which the missiles
may y can be known.
Fig. 3. Schematic view of the direction of ight and range of fragments of three
vessels nos. 13 after a BLEVE. Source: Drawn from the data of Schulz-
Forberg et al. [80].
In the BLEVEtest on three vessels, Schulz-Forberg et al. [80]
found that the tank shell and the left head of each of the vessels
were propelled in a north-westerly direction, between 0
and 45
(Fig. 3). The right head of all the three vessels propelled in the
opposite direction north-east, in the 135180
arc. The liquid

discharge pipe of each of the vessel was hurled southward and
went more than thrice farther then the other two pieces. In other
words similar type of vessel fragments tended to be hurled in
similar directions.
Missile maps of 11 incidents (Fig. 4), involving 15 ves-
sels, mainly LPG, indicate that about half the fragments were
projected into about a third of the total area, in arcs of 30
to either side of the vessel front and rear axial directions

[3,79].
Considering that the ndings of Holden and Reeves [79] do
not match with that of Schulz-Forberg et al. [80], and the very
scanty nature of information on the direction of missiles in the
past accidents, it is not possible at present to draw even a broad
guideline vis- a-vis likely direction of fragment propagation in
BLEVEevents. Nevertheless, Hauptmanns [91,92] has proposed
the following probabilities of projectile directions on the basis
of the data of Holden and Reeves [79]: 30150
: 0.2; 150210
:
0.3; 210330
: 0.2; 33030
: 0.3.
Fig. 4. Map of missiles generated from the bursting of LPG vessels (from the
data of Holden and Reeves [79]).
6.4.4. Penetrative ability and destructive potential of
missiles
A great deal of work has been done on the penetrative
power of cased explosives but information pertaining to vessel-
burst missiles is less extensive. Missiles generated from process
plant, such as high pressure vessels, are of irregular shape
and relatively low energy density, and their penetrative or
destructive power depends on their orientation at the moment
of impact with a target. Due to their irregular shape, frag-
ments usually produced by the bursting of a pressure vessel
have much lower penetrating power, often only half, than com-
pact blunt fragments, while pointed projectiles have appreciably
more [3].
For penetration by small fragments, the depth of penetration
is given by the relation (Cox and Saville [93], quoted in Lees
[3]):
t = Km
n
1
V
n
2
where m is the mass of the fragment (kg), t the thickness of the
barricade needed just to stop the fragment (m), V the velocity of
the fragment (m/s), K a constant and n
1
, and n
2
are the indices.
For brickwork, concrete, and mild steel the values of K, n
1
and
n
2
are in the range 6 10
5
to 23 10
6
, 0.330.4, and 1.01.5
respectively.
The above equation is valid for compact blunt steel fragments
such as solid cylinders with the fragment length approximately
equal to the cylinder diameter, and a mass of not more than 1 kg.
For penetration by larger fragments, of mass exceeding 1 kg,
the depth of penetration may be calculated from
t =
Cm
A
log
10
(1 +5 10
5
V
2
)
where Ais the presented area of the fragment (m
2
) and Cis a con-
stant. The value of C ranges between 0.3 10
4
to 3.5 10
4
for alloy steel, mild steel, and reinforced (1.4%) concrete; it is
25 10
4
for brickwork.
For penetration of concrete by rod shaped missiles:
t = 2 10
7
mV
1.5
d
1.8
and for mild steel plate:
t =
_
0.33 10
9
mV
2
d
3
_
1/1.41
where d is the diameter of the fragment (m).
As the size of the projectile increases, the response of the
target becomes increasingly signicant. Kennedy [94] has pro-
posed that the target response may be modeled by assuming it
to be subject to a rectangular pulse forcing function. With this
assumption it is then possible to apply conventional structural
response techniques to estimate the susceptibility of a structure
to a missile hit.
Out of the models developed for the penetration of reinforced
concrete by missiles, the modied NDRC correlation [95] has
found widespread acceptance. It is applicable to a at-faced
cylindrical missile:
G
_
x
d
_
= 2.74 10
5
_
Dd
0.2
0.5
c
_
V
1.8
with
G
_
x
d
_
=
_
x
2d
_
2
_
x
d
2
_
,
G
_
x
d
_
=
x
d
1
_
x
d
> 2
_
, D =
m
d
3
where d is the diameter of the missile (m), D its caliber density
(kg/m
3
), m its mass (kg), V its velocity (m/s), x the penetration
depth (m) and
c
is the compressive strength of concrete (Pa).
To assess the likely penetration by a at-nosed cylinder the
formula developed by the Commissariat a IEnergie Atom-
ique/Electricite de France (CEA/EDF) (quoted in Lees [3]), on
the basis of extensive testing, is also often used, though no com-
parison has been reported in literature of the results obtainable
by this method and with the NDRC correlation:
V
2
p
= 1.7
c
p
1/3
_
de
m
_
4/3
where V
p
is the perforation velocity (m/s), e is the perforation
thickness (m) and p is the density of the concrete (kg/m
3
); the
other symbols are as dened above.
It has beenreported, that hemispherical nosedprojectiles with
a diameter approximately equal to the target thickness require
up to 30% higher velocities to perforate a reinforced concrete
target than a at-faced or a at nosed projectile having the same
mass and diameter. Similar observations were obtained for other
projectiles with sharp noses [96].
For penetration by missiles of steel plate, the model described
by White and Botsford [97], appears to have found wide accep-
tance:
e
c
d
=
u
10.29
(42.7h
2
+wh)
where d is the diameter of the missile (m), e
c
its critical impact
energy (J), h the thickness of the target panel (m), u the ultimate
tensile strength of the panel (Pa) and w is the width of the panel
(m).
Vessel rupture by a BLEVE fragment: Pietersen [98] has pro-
posed a method which involves calculating (i) the force at the
point of impact to deform the sphere up to the yield point, and
the corresponding energy E
y
and (ii) the energy E
r
to deformthe
sphere in the plastic region between yield and rupture. Utilizing
equations given by Roark and Young [99] and taking the impact
area as ve times the vessel wall thickness, these two energies
were found to be E
y
=7 kJ and E
r
=64 kJ. Assuming that plastic
deformation of the impacting fragment uses up an amount of
energy equal to that used in the sphere itself, the total energy
required for rupture is 135 kJ. From simple kinetic energy con-
siderations, for a fragment of mass 20 tonnes (1/10th of a sphere)
the velocity required to effect rupture is then 3.7 m/s and for 1
of 2 tonnes (1/100th of a sphere) it is 11.6 m/s.
Pietersens calculation indicates that the projectile veloci-
ties needed to rupture another vessel are well below the initial
velocities the fragments actually acquire when a sphere under-
goes BLEVE. In other words, missiles froma vessel undergoing
BLEVE are highly likely to rupture other vessels lying close by.
Accident histories conrm this surmise.
Acknowledging the stochasticity associated with the nature
of fragmentation of a vessel undergoing BLEVE, orientation and
initial velocities of the fragments, direction of fragment prop-
agation, etc., Hauptmanns [91] has developed a procedure for
analyzing such missiles on the basis of probability density func-
tions. The Monte Carlo simulation is done to propagate the effect
of the stochastic and uncertain input parameters through the cal-
culation.
Probability theory has also been used by Gubinelli et al.
[100] to estimate the likelihood of a fragment hitting a target
if a vessel explodes. Their model is based on the analytical
solution of the ballistic equations for fragment trajectory, and
on the introduction of probability distribution functions for the
initial direction of projection of the fragments. The authors
have made a preliminary validation of their model with the
data of Holden and Reeves [79].
6.5. Fireball
The Yellow Book [73] denes reball as a re, burning
sufciently rapidly for the burning mass to rise into the air as
a cloud or ball. In all BLEVEs involving ammable material,
there is a near instantaneous two-phase release of the material
which auto-ignites to form a reball. As reball is an inevitable
consequence whenever a vessel containing a ammable material
undergoes BLEVE, it is often presumed to be an integral part
of all BLEVEs. But about one-fth of all BLEVEs involve non-
ammable material (including re suppressants like nitrogen,
carbon dioxide, and water) and no reballs are generated in such
cases [40].
Fireballs are also formed when vapour clouds are ignited
but whereas such reballs are governed by buoyancy forces,
those from BLEVEs are predominantly inuenced by momen-
tum forces.
In rare cases a vessel containing a ammable PLG may rst
release sufcient mass of vapour which may form a cloud and
get ignited before the vessel fails in a BLEVE generating a far
bigger reball.
For a typical reball resulting from an accidental release
of some 100 tonnes of ammable material, combustion energy
of the order of 5 10
12
J is released within a time span of
1020 s [101]. About one fourth of this energy is emitted as
radiationpowerful enough to scorch people, damage property,
and trigger secondary res. For these reasons reballs are con-
sidered as one of the major hazards in process industry.
Frame by frame analysis of full-scale reball photographs by
Crawley [102] show that the reball passes through the phases
of (a) growth, (b) steady burning and (c) burnout. The growth
phase has two intervals, each spanning about 1 s. During the
rst time span, during which the reball grows to about half its
nal diameter, the reball boundary is bright with yellowish-
white ames indicating a ame temperature of about 1300
C.
In the second time span of the rst phase, the reball attains
its maximum volume, but about 10% of the surface is dark and
sooty with the rest being white, yellowish-orange or light red,
indicating ame temperatures in the range 9001300
C, with
an estimated effective ame temperature of 11001200
C.
In the second phase, which lasts some 10 s, the reball, which
is now roughly spherical, is no longer growing. At the start of
this phase it begins to lift off. It rises and changes to the famil-
iar mushroom shape. The estimated effective ame temperature
remains at 11001200
C.
In the third phase, which lasts some 5 s, the reball remains
the same size, but the ame becomes less sooty and more translu-
cent.
To forecast the size, duration, and radiation of a likely reball
from a BLEVE, the following issues must be addressed:
(a) the mass of ammable substance released on BLEVE,
(b) the mass of the substance contributing to the reball,
(c) the reball development as a function of time,
(d) the reball size and duration,
(e) the heat load generated,
(f) the view factor,
(g) the likelyharmtothe life forms or inanimate objects exposed
to the heat load generated by the reball.
The Yellow Book [73] presents a 14-step calculation with
which the size and the impact of a BLEVE reball can be pre-
dicted. The steps are associated with calculations of: (a) the
amount of ammable material likely to be released on vessel
failure, (b) reball radius, (c) reball duration, (d) reball lift-
off height, (e) distance of impact point X from the center of the
reball, (f) maximum value of view factor at point X, (g) heat
generated by the reball, (h) net available heat for radiation, (i)
absorption factor for water vapour (likely to be present between
the reball and X), (j) absorption coefcient of carbon dioxide
(also likely to be present between the reball and X), (k) atmo-
spheric transmisivity (based on (i) and (j), above), and (l) the
heat ux. Further assessments can be done of the damage likely
at point X in terms of degree of burns.
Of the 12 parameters mentioned above, the view factor is
dened as the ratio between the received and the emitted radia-
tion energy per unit area [73]. Put another way, it is the fraction
of the reball that can be seen by the target [103]. The view
factor incorporates the orientation of the object relative to the
reball, and its distance from the reball center.
As is the case with other aspects of BLEVE forecasting, wide
variations are possible in most of the aspects associated with
reball calculations mentioned above, and treatments by differ-
ent authors can give widely varying estimates. The uncertainty
begins with the estimation of the material that would be released
instantaneously on the vessel failure, or the ashing fraction.
Some treatments, including the Yellow Book [73] and Ref.
[1] assume that the entire lading will ash over to contribute to
the reball whereas other treatments, for example Roberts [104]
and Marshall [12] put the fraction of the fuel that participates in
the reball at about a third of the fraction that is released in a
BLEVE.
The work done so far on the estimation of reball diameter
and duration is summarized in Table 6. The available expressions
for estimating the height of the reball fromthe ground are listed
in Table 7.
Of greater relevance is the assessment of the surface emissive
power of the reball andthe incident radiation(heat load) exerted
by it at a reference point (the target). Models for these treatments
are summarized in Table 8. The point source models assume
that a certain fraction of the heat generated (F
r
) by the reball is
radiated uniformly in all directions. This heat is usually taken to
be the heat of combustion of the reball. CCPS [8] cautions that
the point source models should not be used for those instances
where the plane of the target (receptor) intercepts the reball.
The estimates and models for the fraction of heat radiated
(F
r
), viewfactor (v
F
) the atmospheric transmisivity (), are given
in Tables 911.
Thermally induced BLEVEs in 400 l automotive tanks lled
with liqueed propane, carried out by Maillette and Birk [31],
Table 6
Empirical and analytical methods for estimating reball diameter and duration
Source Material Diameter, D
max
(m) Duration, t
B
(s)
Empirical correlations
Hardee and Lee [116] Propane 5.55M
0.333
Fay and Lewis [117] Propane 6.28M

0.333
2.53M
0.167
Hasegawa and Sato [118] Pentane 5.28M
0.277
1.10M
0.097
Hasegawa and Sato [119] n-Pentane 5.25M
0.314
1.07M
0.181
Williamson and Mann [120] 5.88M
0.333
1.09M
0.167
Lihou and Maund [121] Butane 5.72M
0.333
0.45M
0.333
Lihou and Maund [121] Rocket fuel 6.20M
0.320
0.49M
0.320
Lihou and Maund [121] Propylene 3.51M
0.333
0.32M
0.333
Lihou and Maund [121] Methane 6.36M
0.325
2.57M
0.167
Moorhouse and Pritchard [122] Flammable liquid 5.33M
0.327
1.09M
0.327
Lihou and Maund [121] Propane 3.46M
0.333
0.31M
0.333
Duiser [123] Flammable liquid 5.45M
.1.30
1.34M
0.167
Marshall [12] Hydrocarbon 5.50M
0.333
0.38M
0.333
Gayle and Bransford [124] and
Bagster and Pitblado [125]
Flammable liquid 6.14M
0.325
0.41M
0.340
Pietersen [98], CCPS [76], Prugh
[105] and TNO [73]
Flammable liquid 6.48M
0.325
0.852M
0.260
Robert [104] and CCPS [1] Flammable liquid 5.80M
0.333
0.45M
0.333
(M<3 10
4
), 2.60M
0.167
(M3 10
4
)
Martinsen and Marx [126] Flammable liquid 8.66M
0.25
t
0.333
, 0 t t
B
/3 0.9M
0.25
Analytical models
Bader et al. [127] Propellant 0.61
_
3
4
_
1/3
W
1/3
b
0.572W
1/6
b
Hardee and Lee [116] LNG 6.24M
0.333
1.11M
0.167
Fay and Lewis [117] Flammable liquid
gt
2
(ap)
7p
t =
_
14p
g(ap)
_
0.5
_
3V
4
_
0.167
M: mass of fuel in reball (kg), t: time elapsed after BLEVE (s), : density of reball gas (lb/ft
3
), W
b
: mass of propellant (lb), g: acceleration due to gravity (m/s
2
),
: entrainment coefcient,
a
: density of air (kg/m
3
),
p
: density of products of combustion (kg/m
3
).
Table 7
Correlations for estimating height of the center of the reball from the ground,
H
Reference H (m)
TNO [73] 6.48M
0.325
CCPS [1] 4.35M
0.333
Martinsen and Marx [126] 4.33M
0.25
t
0.333
for 0 t t
B
/3
Table 8
models for estimating the surface emissive power and incident radiation on target
Reference Surface
emissive
power, E
(W/m
2
), E
Heat radiation
received by
target, E
r
(W/m
2
)
Point source models
Hymes [128] E
r
=
2.2F
r
H
c
M
0.67
4l
2
CCPS [76] and Prugh [105] E =
F
r
MH
c
D
2
max
t
B
Lees [3] E
r
=
F
r
Q
4l
2
TNO [73] E =
HMF
r
D
2
max
t
B
Solid ame models

TNO [73] and CCPS [1] E
r
= Ev
F
H: Net heat available for radiation (J/kg), H
c
: Heat of Combustion (J/kg), :
Absorptivity of the target, Q: Heat release rate (W), l: Distance of reball centre
from the target/receptor (m).
Table 9
Estimates and models for the fraction of heat radiated by reballs, F
r
Source F
r
Roberts [104] 0.25P
0.32
v
, for P
v
< 6 MPa
Hymes [129] 0.3 for reballs bursting below the relief valve set
pressure, 0.4for reballs burstingat or above the relief
valve set pressure
TNO [73] 0.27P
0.32
v
Makhviladze and
Yakush [130]
0.180.27
Roberts et al. [29] 0.250.4
P
v
is the vapor pressure at the moment of burst (MPa).
Table 10
Expressions for view factor, v
F
Source Position of object v
F
CCPS [8] Horizontal
H(D
max
/2)
2
(L
2
+H
2
)
3/2
CCPS [8] Vertical (L>D
max
/2)
L(D
max
/2)
2
(L
2
+H
2
)
3/2
CCPS [76] and TNO [73] Highest value of view factor
_
D
max
2X
_
2
X: Distance of reball centre from the target/receptor (m), L: Distance of point
on ground directly below the reball centre, from the target/receptor (m).
Table 11
Expressions for atmospheric transmisivity,
Source
Pietersen and Huerta [131] and CCPS [1] 2.02(P
w
X
s
)
0.09
TNO [73] 1
w
c
: Absorption coefcient for carbon dioxide,
w
: Absorption factor for water
vapour, P
w
: Water partial pressure (Pa), X
s
: Distance from surface of reball to
the target (m).
revealed that while the projected reball area is more or less
circular for one-step BLEVEs, the shape of the reball tended to
be cylinder-like in two-step BLEVEs. They found that reballs
resultingfromcooler ladingcreatedgreater hazards because they
generated larger ground res; took longer to lift off, and lasted
longer.
Roberts et al. [29] compared the results of their BLEVE
tests with the predictions of the models proposed earlier by
Roberts [104], and Prugh [105] for reball diameter and dura-
tion (Table 6). They found that the expression proposed by Prugh
[105] gave a slightly better t.
Based on numerical modeling of the formation, evolution,
and combustion processes in a reball, Makhviladze et al. [101]
suggest that when the released fuel is ignited near the source,
the burning gas could rapidly assumes a nearly spherical shape,
risingas a reball. The temperature andcombustionproduct con-
centration elds gradually become similar to each other. Diffu-
sion combustion of a fuel-rich cloud is localized in a thin zone on
the cloud surface where the fuel mixes with the ambient air. The
authors further observe that for two-phase releases of volatile liq-
ueed gases, droplet evaporation proceeds much more rapidly
than combustion, so that the main inuence of prerelease condi-
tion is through changing the outow velocity. The distributions
of net emissive power calculated for small fuel mass (optically
thin reball) and for large fuel mass (optically thick cloud)
showed that in the small cloud, the emission of radiation occurs
throughout the volume of the reball, whereas a large cloud
emits mostly from its surface. The calculated fractions of com-
bustion heat emitted as a radiation for all cloud sizes and storage
conditions were in the range 0.20.25, which matched well with
the available experimental data for turbulent propane ames.
7. BLEVE prevention and damage control
As discussed in Section 5 of this paper, it is well-neigh impos-
sible to say with certainty whether a jeopardized vessel will
suffer a BLEVE or not. Likewise it is not possible to forecast
with any measure of condence when a vessel will suffer a
BLEVE after getting jeopardized. These aspects, and the uncer-
tainty associated with forecasting the size, range, direction, and
momentum of missiles likely from a BLEVE, pose special chal-
lenges towards preventing a BLEVEor in containing the damage
a BLEVE may cause. There have been several tragic incidents
when re ghters arrive to save a re-engulfed vessel only to be
killed by the expanding reball or the rocketing fragments when
a vessel suddenly bursts [4,106].
The strategies required to minimize the occurrence and the
adverse impact of BLEVE, have been reviewed by Prugh [16],
Khan and Abbasi [4], and Casal et al. [5]. Pointers can also be
drawn from the studies on simulated BLEVEs described in Sec-
tion 5, and fromthe studies such as effect of pressure relief value
(PRV) functioning [23,56,107,108], survivability of steel cylin-
ders in comparison to aluminum cylinders [62] and projectile
range [29]. The strategies can be broadly classied into three
categories:
(a) Reducing the probability of a vessel getting jeopardized by
a hit, a re, an increasingly pronounced structural weakness,
a runaway reaction, or a transportation accident.
(b) Cushioning the impact of the above so that the perturbation
does not escalate to a BLEVE.
(c) Minimizing the damage if a BLEVE does occur.
We summarize below the strategies possible under each of
these categories.
7.1. Preventing the causes which can make a vessel
vulnerable to BLEVE
7.1.1. Preventing exposure to re
Keeping the PLGcontaining vessel a safe distance away from
likely source of re. Fire engulfment being the most common of
the causes due to which PLG vessels undergo BLEVE, it is
imperative that a reasonably large distance should separate a
PLG vessel from another vessel handling a ammable material
or from other sources of re. Of course this can at best reduce
the probability of a PLG vessel being heated by the radiation
load from another vessel which has caught re. The PLG vessel
may still be jeopardized by blast waves or projectile hits from
another exploding vessel. Raj [109] has examined the efcacy of
over 60-year-old NFPA (National Fire Protection Association)
codes specifying minimum separation distance between storage
tanks on the basis of a radiation heat transfer model. The author
has calculated temporal variation of the vapour-wetted tank-wall
temperature of a vessel exposed to an external, non-impinging,
highly radiative 30.5 m diameter pool re. The results indicate
that the vessel wall temperature will never reach critical condi-
tion and that the NFPA codes are still valid for non-impinging
res.
Sloping of the nearby ground. To prevent a pool re occurring
after an accidental spill from a PLG vessel, the ground radially
away from a xed installation should have a downward slope of
not less than 1% so as to lead the spill away to a safe area.
Water barriers. These may be installed close to the PLGcon-
tainers. These consist of sprayer systemwhichgenerates curtains
of ne water mist. The barriers can capture ammable vapour
if released from the PLG container and disperse it without get-
ting ignited. Water mists can also dissolve some of the released
material if it happens to be ammonia, chlorine, or some other
water soluble substance, thereby reducing the toxic dispersion.
7.1.2. Preventing mechanical damage
Trucks and railroad cars carrying PLGs should be protected
from accidental damage with double containers, equipped with
insulation in the annular spaces. Collision or overturning during
transportation damages the outer shall. This makes it essential
to fabricate the outer container with a material which would
provide protection for the inner tankage.
7.1.3. Preventing overlling and overpressure
Rigidcompliance withstandards duringthe llingandweigh-
ing of the BLEVE-prone tanks alongside very careful installa-
tion and testing of relief devices have reduced the frequency
of BLEVEs on account of overlling. But accidents continue to
occur during pumping of PLGs as happened at Moombas, South
Australia, on 16 June 2001, killing one person, injuring three,
and damaging the infrastructure.
The relief-devices are prone to plugging; this problem is
circumvented by the installation of rupture disks in series as
pluggage protection under relief valves. Rupture disks are also
installed in parallel to relief valves as a last resort protection.
7.1.4. Prevention of runaway reaction
The accident which led to the coinage of the acronymBLEVE
was a runaway reaction. But BLEVEs due to runaway reac-
tions are much less common than BLEVEs which occur when a
PLG storage vessel suffers accidental damage. Instrumentation
should be provided for continuous monitoring of temperature
and pressure within all process equipment likely to contain
self-reactive materials. Such equipment should have facilities
for counteracting overpressure or overtemperature; for exam-
ple internal cooling coils or external jackets, remote-controlled
venting valves, inhibitorinjection systems, and internal del-
uges, as well as high-temperature and/or high-pressure alarms
for control-room and eld personnel.
7.1.5. Prevention of vapourspace contamination with
reactive material
Vessels containing highly reactive gases such as hydro-
gen and chlorine in liqueed form should be safeguarded
against contamination by other substances with which they can
react. Inerting vapour spaces with nitrogen or other nonreactive
gas and installing explosionsuppression systems may prevent
vapourspace explosions thus reducing the risk of vessel damage
and, consequently, a BLEVE [16].
7.1.6. Prevention of internal weakening of vessel structure
due to fatigue, creep, corrosion, etc.
Proper design and pre-use testing of containers can pre-
vent distortion and possible rupture of containers. Periodic
wall-thickness measurements, internal inspection for corrosion,
acoustic emission testing for the possible cracking of the con-
tainer, etc., should be performed to ensure the tness of the
containers. Preventive maintenance should be done along with
predictive maintenance [16].
7.1.7. General protection from re as well as accidental
hits, by container burial
Vessels containing PLGs can be protected from re, or exter-
nal hits, to a very great extent if they are partially or totally
buried. But such vessels are difcult to inspect and are particu-
larly vulnerable to corrosion.
7.1.8. Prevention of excessive superheat which may prevent
explosive boiling
Taking a cue from distillation systems and reactors in which
nucleation devices such as sharp-edged ceramic material or an
aluminum mesh is placed in the liquid being distilled to assist
boiling and prevent superheating, similar devices have been
explored for PLGcontainers. But a well tried and tested strategy
along these lines is yet to evolve.
7.2. Managing a jeopardized vessel to prevent it from
undergoing BLEVE
7.2.1. Thermal insulation
The PLGcontainers shouldbe thermallyinsulatedtothe max-
imumextent possible as it would reduce the rate of heating of the
vessel when it receives heat load and delay the pressure increase
inside. If the container wall is protected with a steel jacket and
a ceramic insulation of adequate thickness (13 mm or more), it
provides substantial thermal protection [61]. Even steel jackets
with an air gap between the jacket walls can cut the wall heating
rate to approximately half of the unprotected wall. But such re
proong cannot by itself prevent a BLEVE; it can at best delay
the catastrophic event by four to 5 h giving time for the re ght-
ers to remove the heat load. In xed installations, even the vessel
support system should be insulated so that it does not cave in
when subjected to heat. Likewise, the valves, pipes, and other
safety elements used in the PLG vessel must have the ability
to resist the action of re and withstand the high temperatures
that may be reached in a crisis situation. The thermal insulation
system should be installed in such a way that it does not inter-
fere with the periodic inspection of the tank surface and support
systems.
Fireproong can be even more effective in delaying a BLEVE
if the pressure relief valve (PRV) operates correctly.
7.2.2. Directed water deluge
To cut off the heat load once a PLG vessel gets engulfed
in re, it has to be subjected to what is called directed water
deluge [110]. Water must be applied as soon as possible, with a
layer of adequate thickness which should totally cover the vessel
wall, especially those areas directly covered with ame. Awater
ow rate of 10 m
2
min
1
is recommended, which should be
upped to 15 m
2
min
1
in areas directly being licked by a ame
[111,112].
If the ame is highly turbulent, which can generate a heat
ux of the order of 350 kWm
2
, ow rates even larger than
25 m
2
min
1
may be required. But if the PLG vessel is being
impinged by jet re the water deluge is less effective; it cannot be
relied upon to maintain a water lm over the whole tank surface
[110]. The dry patches, where the water lm broke down got
heated to about 350
C in 10 min during the course of full-scale

tests reported by Shirvill [110].
7.2.3. Cooling of the unwetted part of the PLG tank wall by
internal liquid spray
Young [113] has patented an anti-BLEVE safety system
in which a turbo-charger is placed inside the PLG vessel. If the
Fig. 5. The anti-BLEVE safety system patented by Young [113].
vessel were to get engulfed in a re, the turbo-charger, according
to the patent, would take in the liquid fromjust belowthe surface
near the shell of the tank, and spray it vertically upwards to cool
the unwetted part of the tank (Fig. 5).
The charger is claimed to be driven by the vapour escaping
through the pressure relief value. The system is expected to pre-
vent the unwetted part of the tank well fromgetting heated to the
point of weakening the part, thus preventing a crack to develop
for the eventual BLEVE. According to the patent, the systemcan
be readily built into all new LPG tanks, as well as retro-tted to
all existing transportation (rail, road and sea) tanks and storage
tanks. For mobile tanks, the patentee has devised a gravity-
valve assembly to ensure that the liquid is always sprayed
upwards, and that only vapour is emitted fromthe pressure relief
valve.
7.2.4. Rapid depressurization
Another step must be taken along with the start of directed
water deluge to reduce the probability of a BLEVE
depressurization of the vessel with remote operated reproof
valves bypassing the installed PRV. Such devices should be able
to reduce the vessel pressure to half of the design pressure within
15 min [5]. The released material should be eliminated in a
safe manner, for example with a torch. The depressurization
should not be too rapid either as it may lead to extremely low
temperature and fragility in the steel. For a 54 m
3
tank hold-
ing 23 tonnes of LPG which underwent a BLEVE at Alma-Ata,
Kazakhastan, in 1989, Shebeko et al. [22] theorize that the tank
would not have exploded if its safety value had a cross sectional
area not less than 77 cm
2
and operating pressure not exceeding
1.6 MPa. Attempts to develop pressure relief devices specic to
the substance stored as PLG have also been made, for example
liqueed ammonia [114]. The authors have modeled thermal
response of a horizontal ammonia tank severely engulfed in
re.
7.3. Minimizing the damage if a BLEVE does occur
If a vessel suffers a BLEVEwithin a fewminutes after getting
jeopardized, very little can be done to reduce the damage it
would cause. But even if a BLEVE is delayed which could
be by several hours such delay is a mixed blessing. If care is
not taken to prevent emergency responders or bystanders from
going within striking distance of the possible reball, shock
wave, or projectiles, the toll froma delayed BLEVEmay even be
higher than fromthe rapid one. Damage fromprojectiles is of the
greatest concern because their impact area is much larger then of
the reball or shock-wave. In Sections 4 and 6.4 we have quoted
several instances which reect the great danger from missiles
associated with BLEVEs.
7.3.1. Cushioning the missile damage
The rst concern towards minimizing the damage caused
by a BLEVE is to prevent the accident from triggering sec-
ondary and higher order accidents. Indeed the past accident
analysis of BLEVEs tells us that very few BLEVEs occur as
stand-alone accidents; in a large number of instances BLEVEs
cause domino effect, triggering serial blasts [27,28]. To pre-
vent domino effect, other vessels, which may explode on being
heated or mechanically damaged, should be kept as far away
from PLG-containing vessels as possible.
Barriers may also be placed around the vessels to cushion the
impact of outgoing or incoming missiles. It is relatively easy to
provide a barricade for vessels with energy contents in the range
10
3
10
5
J, but it becomes progressively more difcult as the
energy content rises, and for energy contents capable of giving
a shock wave of (50100) 10
6
J, putting a barricade would
require sophisticated design and benet-cost optimization.
The preferred form of barricade is a closed cubicle. For pro-
tection against blast using the equivalent static pressure method,
the High Pressure Safety Code ([93], quoted in [3]) gives the
relevant pressure as
P = 7.6
_
E 10
6
V
_
0.72
, P < 70
where E is the shock wave energy (J), P the equivalent static
pressure (bar) and V is the volume of the enclosure (m
3
). This
equation is applicable where the aspect ratio of the enclosure
does not exceed two.
Barricades can also take the form of thin-walled pressure
vessels. Small enclosures can be made of angle iron and steel
plate. Large barricades should be of reinforced concrete. Since
the shock wave has positive and negative phases, reinforcement
is required on both inner and outer faces. The barricade should
have the provision to allow dispersion of small leaks by ventila-
tion.
7.3.2. Fireball suppression
This is a possibility yet to be translated into practice but
its potential is obvious. If re suppressants can be released
in a way that they get mixed with the ashing material when
a vessel suffers a BLEVE, the reball formation can be tem-
pered with and its intensity reduced signicantly. Systems can
also be put in place so that the reball, at the moment of its
formation, gets surrounded by a cloud of certain re suppres-
sant. Then, the suppressant would be sucked into the reball by
strong air entrainment. As a result, the ame may be completely
suppressed, or at least the reball size would be signicantly
reduced.
The use of water mist as a reball suppressant is an obvi-
ous possibility but the liquid droplets may completely evaporate
before they are sucked into the reball. This may reduce the
suppressing effect. Aerosol re extinguishing agents (AFEAs)
may be a better substitute, because aerosol particles are not sub-
jected to a phase change. They work by destroying the active
centers which are necessary to sustain the ame. AFEAs can be
generated by the combustion of solid propellants [30,115].
7.3.3. How close emergency responders can go to a
jeopardized vessel
Birk [106] has proposed that re ghters should not go closer
to a jeopardized vessel than four reball radii (which can be
estimated on-the-spot using the expression R=3m
0.33
where
m is the lading mass in kg and R is the reball radius in m),
to a minimum of 90m. If it is possible the distance should be
longer to reduce the hazard from rocketing fragments. Further,
the emergency responders should be wearing protective clothing
that can withstand radiation load of 21 kWm
2
for the antici-
pated duration of a reball (to be estimated as 0.15R, s). For
large-scale tanks the safe distance may be too long to enable
re ghters from directing water onto re impinged tanks. For
such large tank installations, water spray systems should already
be in place and operating when responders arrive. However,
delayed BLEVE remains a major risk to re ghters dealing
with uninsulated transport tanks and small stationary tanks; the
responders are exposed to serious risk from reball, blast and
projectile effects. Responder should also expect danger from
potential secondary projectiles (such as attached pipes, nearby
equipment, etc.) which can be sent large distances by the waves
which accompany a BLEVE.
7.3.4. Evacuation
The public should be evacuated to a distance of at least 15
reball radii, preferably 30 reball radii away from jeopardized
tanks. This distance should be increased downwind of a potential
BLEVE. At this distance there is little threat from the reball
thermal radiation or blast. As tank size increases above 5 m
3
the
30R distance becomes more and more conservative and the 15R
distance becomes more appropriate [56].
If the PLG involved in a BLEVE happens to be toxic such
as chlorine, ammonia, methyl isocyanate, or phosgene its ini-
tial dispersion would be inuenced by the blast wave effects and
would even carry it upwind to some distance before the usual
meteorological factors and density effects become inuential
in the plume dispersion. Emergency preparedness for accidents
involving such PLGs should factor in the blast-mediated disper-
sion.
8. Summary and conclusion
(1) BLEVE is one among six classes of explosions that can
occur in the process industryvapour cloud explosion,
dust explosion, condensed phase explosion, conned explo-
sions, and physical explosions being the other ve. Of
these, BLEVE is a particularly destructive type and the
worlds second worst process industry accident the Mex-
ico renery disaster of 1984 had involved a chain of
BLEVEs.
(2) Aside producing highly destructive blast waves and re-
balls (or toxic dispersion), BLEVEs propel the fragments of
the ruptured vessel in all directions at high velocities. Such
missiles are often enveloped in re if the ruptured vessel
had contained a ammable chemical. The greatest damage
in most BLEVE events has been caused by such missiles
either by directly hitting (often killing) people, or by trig-
gering res, or by damaging other process units leading to
secondary accidents.
(3) A BLEVE can occur in any situation where a vessel or a
conduit carrying a pressure-liqueed gas (PLG) is acciden-
tally depressurized. The depressurization suddenly renders
the PLG into a superheated state, leading to instantaneous
nucleation and explosive ashing. A number of factors
introduce complexity in this otherwise simple-looking phe-
nomena. For example the initial depressurization due to a
minor crack may just cause the release of a gaseous or a
liquid jet and the crack may either not propagate further or
do so after a great deal of time. Just as well a crack may
propagate rapidly. A great deal of effort, reviewed in this
paper, has been made to understand the nature and the mag-
nitude of the forces and counter-forces which are generated
in a jeopardized vessel containing a PLG. A great deal of
new knowledge has been acquired, on the basis of analy-
sis of past accidents as well as experimental BLEVEs, but
we still are not in a position to forecast whether a jeop-
ardized vessel will suffer BLEVE and, if it does, when.
This remains an area where a great deal of further R&D is
required.
(4) In a like manner a great deal of knowledge, reviewed in the
paper, has been accumulated on the likely energy of explo-
sion of a BLEVE, the likely size, duration, and heat load of
the reball a BLEVE may produce, and the number, size,
orientation, and kinetic energy of missiles generated from
the shattering vessel. But the forecasts one may arrive by
using different methods may differ from each other signi-
cantly, at times by one or more orders of magnitude.
(5) Even do-how manuals produced by coordinating agencies
such as TNO, The Hague, and the Centre for Chemical
Process Safety, New York describe different consequence
assessment methods with little commentary on their rela-
tive merits. This state of affairs creates situations wherein
the risk assessment conducted by an industry and a regu-
latory agency may differ widely from each other in spite
of both sides having used legitimate consequence analysis
models. There is urgent need to rectify this problem by the
standardization and codication of procedures as has been
done, say, in most areas of clinical analysis or environmental
analysis.
(6) Considering that BLEVE is a fairly well-known and well-
documented phenomenon, a misconception is surprisingly
prevalent that BLEVE occurs only with ammable chem-
icals. The fact is that one-fth of all BLEVEs occur with
non-ammable PLGs.
(7) Among the questions which call for investigations is: what
is the probability that a BLEVE will cause a secondary and
higher order accident (domino effect)? A recall of BLEVE
history indicates that a BLEVE is rarely a stand-alone event
and is more often than not a trigger for other accidents. Stud-
ies on this aspect of BLEVE would rene our assessment
of the risk posed by PLG-holding units.
(8) A great deal of R&D is also needed to nd more reliable
methods than available for preventing a jeopardized vessel
form undergoing BLEVE.
References
[1] CCPS, Guidelines for Consequence Analysis of Chemical Releases, Cen-
ter for Chemical Process Safety, American Institute of Chemical Engi-
neers, New York, 1999.
[2] A.M. Birk, M.H. Cunningham, The boiling liquid expanding vapor explo-
sion, J. Loss Prev. Process Ind. 7 (1994) 474480.
[3] F.P. Lees, Loss Prevention in the Process Industries-Hazard Identication,
Assessment, and Control, vols. 13, Butterworth-Heinemann, Oxford,
1996.
[4] F.I. Khan, S.A. Abbasi, Risk Assessment in the Chemical Process Indus-
tries: Advanced Techniques, Discovery Publishing House, New Delhi,
1998.
[5] J. Casal, J. Arnaldos, H. Montiel, E. Planas-Cuchi, J.A. Vilchez, Mod-
elling and understanding BLEVEs, in: M. Fingas (Ed.), Handbook of
Hazardous Spills, McGraw Hill, New York, 2001.
[6] F.I. Khan, S.A. Abbasi, Major accidents in the process industries and
analysis of their causes and consequences, J. Loss Prev. Process Ind. 12
(1999) 361374.
[7] T.A. Kletz, Unconned vapour cloud explosions, Loss Prev. 11 (1977)
50.
[8] CCPS, Guidelines for Evaluating the Characteristics of Vapor Cloud
Explosions, Flash Fires and BLEVEs, Center for Chemical Pro-
cess Safety, American Institute of Chemical Engineers, New York,
1994.
[9] D.F. Peterson, BLEVE: facts, risk factors, and fallacies, Fire Eng. 155
(2002) 97103.
[10] W.L. Walls, What is a BLEVE, Fire J. 31 (1978) 4647.
[11] W.L. Walls, The BLEVEpart 1, Fire Command. 17 (1979) 3537.
[12] V.C. Marshall, Major Chemical Hazards, Ellis Horwood, Chichester,
1987.
[13] R.C. Reid, Superheated liquids, Am. Scientist 64 (1976) 146156.
[14] J.E.S. Venart, K.F. Sollows, K. Sumathipala, G.A. Rutledge, X. Jian,
Boiling Liquid Compressed Bubble Explosions: Experiments/Models,
GasLiquid Flows, vol. 165, ASME, New York, 1993, pp. 5560.
[15] C.A. McDevitt, C.K. Chan, F.R. Steward, K.N. Tennankore, Initiation
step of boiling liquid expanding vapor explosions, J. Hazard. Mater. 25
(1990) 169180.
[16] R.W. Prugh, Quantitative evaluation of BLEVE hazards, J. Fire Protect.
Eng. 3 (1991) 924.
[17] R.W. Prugh, Quantify BLEVE hazards, Chem. Eng. Prog. 87 (1991)
6672.
[18] I.R.M. Leslie, A.M. Birk, State of the art review of pressure liqueed
gas container failure modes and associated projectile hazards, J. Hazard.
Mater. 28 (1991) 329365.
[19] A.M. Birk, M.H. Cunningham, Liquid temperature stratication and its
effect on BLEVEs and their hazards, J. Hazard. Mater. 48 (1996) 219
237.
[20] J.E.S. Venart, Boiling liquid expanding vapor explosions (BLEVE); two-
phase aspects of failure, in: Proceedings of the 39th European Two-phase
Flow Group Meeting, Aveiro, 2001, p. 9.
[21] R.C. Reid, Possible mechanism for pressurized-liquid tank explosions or
BLEVEs, Science 203 (1979) 12631265.
[22] Y.N. Shebeko, A.P. Shevchuck, I.M. Smolin, BLEVE prevention using
vent devices, J. Hazard. Mater. 50 (1996) 227238.
[23] A.M. Birk, J.D.J. VanderSteen, M.H. Cunningham, C.R. Davison, I.
Mirzazadeh, Fire tests to study the effect of pressure relief valve blow-
down on the survivability of propane tanks in res, Process Saf. Prog. 21
(2002) 227236.
[24] S. Shivaji, Forecasting the dispersion and impacts of accidental release
of hazardous materials, MTech Thesis, Sri Venkatateswara University,
Tirupati, 2004, 171 pp.
[25] N.N. Desai. Risk assessment in the transport of hazmats, MTech
Thesis, Vishweshvaraya Technological University, Belgaum, 2004,
74 pp.
[26] C.M. Yu, J.E.S. Venart, The boiling liquid collapsed bubble explo-
sion (BLCBE): a preliminary model, J. Hazard. Mater. 46 (1996)
197213.
[27] F.I. Khan, S.A. Abbasi, Models for domino effect analysis in the chemical
process industries, Process Saf. Prog. 17 (1998) 107123.
[28] F.I. Khan, S.A. Abbasi, DOMIFFECT: a new user friendly software for
domino effect analysis, Environ. Model. Softw. 13 (1998) 163177.
[29] T. Roberts, A. Gosse, S. Hawksworth, Thermal radiation from reballs
on failure of liqueed petroleum gas storage vessels, in: Proceedings of
the IChemE Symposium Series No. 147, 2000, pp. 105120.
[30] V. Novozhilov, Some aspects of the mathematical modeling of reballs,
in: Proceedings of the IMechE, Part E, J. Process Mech. Eng. 217 (2003)
103121.
[31] J. Maillette, A.M. Birk, Effects of tank failure mode and lading proper-
ties on propane reball geometry and re hazard, in: Proceedings of the
International Conference and Workshop on Modeling and Mitigating the
Consequences of Accidents Releases of Hazardous Materials, Louisiana,
1995.
[32] J.E.S. Venart, Boiling liquid expanding vapor explosions (BLEVE); pos-
sible failure mechanisms and their consequences, in: Proceedings of the
IChemE Symposium Series No. 147, 2000, pp. 121137.
[33] M. Selway, The predicted BLEVEfrequency of a selected 2000 m
3
butane
sphere on a renery site. UK Atomic Energy Authority, Safety and Reli-
ability Directorate, Report SRD/R492, 1988, 57 pp.
[34] T.A. Kletz, What Went Wrong, Gulf Publishing, Houston, 1985.
[35] S.K. Kuriechan, Causes and impacts of accidents in chemical process
industries anda studyof the consequence analysis software, MPhil Thesis,
Pondicherry University, 2005, p. 175.
[36] A. Wolf. BLEVE kills two. NFPA(National Fire Protection Association),
2006. http://www.nfpa.org/assets/les/PDF/AlbertCity.PDF.
[37] M. Masellis, Fire disaster in a motorway tunnel, Prehosp. Disaster Med.
15 (2000) 7478.
[38] E. Planas-Cuchi, N. Gasulla, A. Ventosa, J. Casal, Explosion of a road
tanker containing liqueed natural gas, J. Loss Prev. Process Ind. 17
(2004) 315321.
[39] US-CSHIB. Investigation ReportCatastrophic Vessel Failure, United
States Chemical Safety and Hazard Investigation Board, Washington,
NTIS No. PB 2004-103035, 2004, 53 pp.
[40] R.M. Chelvi, Accidents in chemical process industries with a special
reference to BLEVE, MTech Thesis, Coimbatore Institute of Technology,
2006, p. 163.
[41] G. Vijayraghavan, Impact assessment, modeling, and control of dust
explosions in chemical process industries, MTech Thesis, Coimbatore
Institute of Technology, 2004, p. 131.
[42] T. Abbasi, S.A. Abbasi, The expertise and the practice of loss prevention
in the Indian process industry some pointers for the third world, Trans.
IChemE 83 (2005) 413420.
[43] R.C. Reid, Some theories on boiling liquid expanding vapor explosions,
Fire (1980) 525529.
[44] T. Abbasi, B.C. Clarke, S.A. Abbasi, Accidental risk of superheated liq-
uids and a framework for predicting the superheat limit. J. Loss Prev.
Process Ind., in press.
[45] H. Wekeshima, K. Takata, On the limit of superheat, J. Phys. Soc. Jpn.
13 (1958) 13981403.
[46] G.R. Moore, Vaporization of superheated drops in liquids, AIChE J. 5
(1959) 458466.
[47] W. Gibbs, The Scientic Papers, vol. 1, Dover Publications, New York,
1961.
[48] M. Volmer, A. Weber, Keimbildung in

Ubers attigten Gebilden, Z. Phus.
Chem. 119 (1926) 277301.
[49] L. Farkas, The velocity of nucleus formation in supersaturated vapors, J.
Phys. Chem. 125 (1927) 236.
[50] R. Becker, W. Doring, The kinetic treatment of nuclear formation in super-
saturated vapors, Ann. Phys. 24 (1935) 719752.
[51] M. Blander, J.L. Katz, Bubble nucleation in liquids, AIChE J. 21 (1975)
883.
[52] C.E. Brennen, Phase change, nucleation, and cavitation, in: Cavitation
and Bubble Dynamies, Oxford University Press, London, 1995.
[53] J. Li, G.P. Peterson, P. Cheng, Mechanical nonequilibriumconsiderations
in homogeneous bubble nucleation for unsteady-state boiling, Int. J. Heat
Mass Transf. 48 (2005) 30813096.
[54] I.J. Ford, C.F. Clement, The effects of temperature uctuations in homo-
geneous nucleation theory, J. Phys. A 22 (1989) 40074018.
[55] H.-C. Park, K.-T. Byun, H.-Y. Kwak, Explosive boiling of liquid
droplets at their superheat limits, Chem. Eng. Sci. 60 (2005) 1809
1821.
[56] A.M. Birk, Scale effects with re exposure of pressure-liqueed gas tanks,
J. Loss Prev. Process Ind. 8 (1995) 275290.
[57] S. Ramier, J.E.S. Venart, Boiling liquid expanding vapor explosions:
dynamic re-pressurization and two-phase discharge, in: Proceedings of
the IChemE Symposium Series No. 147, 2000, pp. 527537.
[58] J. Stawczyk, Experimental evaluation of LPG tank explosion hazards, J.
Hazard. Mater. B 96 (2003) 189200.
[59] E. Planas-Cuchi, J.M. Salla, J. Casal, Calculating overpressure from
BLEVE explosions, J. Loss Prev. Process Ind. 17 (2004) 431436.
[60] D.J. Kielec, A.M. Birk, Analysis of re-induced ruptures of 400-l propane
tanks, J. Pressure Vessel Technol.: Trans. ASME. 119 (1997) 365
373.
[61] J.D.J. VanderSteen, A.M. Birk, Fire tests on defective tank-car thermal
protection systems, J. Loss Prev. Process Ind. 16 (2003) 417425.
[62] A.M. Birk, J.D.J. VanderSteen, The survivability of steel and aluminum
33.5 pound propane cylinders in re, Process Saf. Prog. 22 (2003)
129135.
[63] K.W. Blything, A.B. Reeves, An initial prediction of the BLEVE fre-
quency of a 100 tonnes butane storage vessel. UK Atomic Energy
Authority, Safety and Reliability Directorate, Report SRD/R488, 1988,
88 pp.
[64] A.M. Birk, J.D.J. VanderSteen, On the transition from non-BLEVE to
BLEVE failure for a 1.8 m
3
propane tank, Am. Soc. Mech. Eng. Pressure
Vessels Piping Div. PVP 446 (2002) 143152.
[65] A.C. Van den Berg, M.M. van der Voort, J. Weerheijm, N.H.A. Versloot,
Expansion-controlled evaporation: a safe approach to BLEVE blast, J.
Loss Prev. Process Ind. 17 (2004) 397405.
[66] H. Giesbrecht, K. Hess, W. Leuckel, B. Maurer, Analysis of explosion
hazards on spontaneous release of inammable gas in the atmosphere.
Part 1. Propagation and deagration of vapor clouds on the basis of
bursting tests on model vessels, German Chem. Eng. 4 (1981) 305
314.
[67] C. Ogsio, N. Takagi, T. Kitagawa, On the mechanismof vapor explosion.,
in: Proceedings of the Loss Prevention and Safety Symposium, PACHEC,
1972.
[68] G.V. Hadjisophocleous, A.C.M. Sousa, J.E.S. Venart, Mathematical mod-
eling of LPG tanks subjected to full partial re engulfment, Int. J. Numer.
Meth. Eng. 30 (1990) 629646.
[69] G.V. Hadjisophocleous, A.C.M. Sousa, J.E.S. Venart, Astudy of the effect
of the tank diameter on the thermal stratication in LPG tanks subjected
to re engulfment, J. Hazard. Mater. 25 (1990) 1931.
[70] Y.W. Gong, W.S. Lin, A.Z. Guo, X.S. Lu, A simplied model to predict
the thermal response of PLG and its inuence on BLEVE, J. Hazard.
Mater. A108 (2004) 2126.
[71] F.I. Khan, S.A. Abbasi, TORAP-a new tool for conducting rapid risk-
assessments in petroleum reneries and petrochemical industries, Appl.
Energy 65 (2000) 187210.
[72] F.I. Khan, S.A. Abbasi, An assessment of the likelihood of occurrence,
and the damage potential of domino effect (chain of accidents) in a
typical cluster of industries, J. Loss Prev. Process Ind. 14 (2001) 283
306.
[73] C.J.H. van den Bosch, R.A.P.M. Weterings, Methods for the Calcu-
lation of Physical Effects, Committee for the Prevention of Disas-
ters, CPR 14E (TNO Yellow Book), The Hague, The Netherlands,
1997.
[74] C. Tellez, J.A. Pena, Boiling-liquid expanding vapor explosion (BLEVE):
An introduction to consequence and vulnerability analysis, Chem. Eng.
Educ. 36 (2002) 206211.
[75] W.E. Baker, J.J. Kulesz, R.E. Richer, R.L. Bessey, P.S. Westine, V.B.
Parr, G.A. Oldham, Workbook for predicting pressure wave and fragment
effects of exploding propellant tanks and gas storage vessels, NASA CR-
134906, NASAScientic and Technical Information Ofce, Washington,
1975 and 1977.
[76] CCPS., Guidelines for Chemical Process Quantitative Risk Analysis,
Center for Chemical Process Safety, AmericanInstitute of Chemical Engi-
neers, New York, 1989.
[77] A.C. Van den Berg, A. Lannoy, Methods for vapor cloud explosion blast
modeling, J. Hazard. Mater. 34 (1993) 151171.
[78] W.E. Baker, P.A. Cox, P.S. Westine, J.J. Kulesz, R.A. Strehlow, Explosion
Hazards and Evaluation, Elsevier, New York, 1983.
[79] P.L. Holden, A.B. Reeves, Fragment hazards from failures of pressurized
liqueed gas vessels, Chem. Eng. Symp. Ser. 93 (1985) 205217.
[80] B. Schulz-Forberg, B. Droste, H. Charlett, Failure mechanisms of
propane-tanks under thermal stresses including re engulfment, in: Pro-
ceedings of the International Symposium on Transport and Storage of
LPG and LNG, Antwerp, 1984.
[81] N.F. Scilly, J.H. Crowther, Methodology for predicting domino effects
from pressure vessel fragmentation, in: Proceedings of the International
Conference on Hazard Identication and Risk Analysis, Human Factors
and Human Reliability in Process Safety, 1992.
[82] W.E. Baker, J.J. Kulesz, R.E. Richer, P.S. Westine, V.B. Parr, L.M. Vargas,
P.K. Moseley, Workbook for Estimating the Effects of Accidental Explo-
sions in Propellant Handling Systems, NASACR-3023, NASAScientic
and Technical Information Ofce, Washington, 1978.
[83] J. Tulacz, R.E. Smith, Assessment of missiles generated by pressure
component failure and its application to recent gas-cooled nuclear plant
design, Nucl. Energy 19 (1980) 151.
[84] M. Held, Fragment mass distribution of high energy projectiles, Propel-
lants, High Explosives, Pyrotechnics 15 (1990) 254260.
[85] V.J. Clancey, Liquid and vapour emission and dispersion, in: Course on
Loss Prevention in the Process Industries, Department of Chemical Engi-
neering, Loughborough University of Technology, 1976.
[86] Daily Telegraph, 1990, Sydney, 1990, p. 1.
[87] M.R. Baum, The velocity of missiles generated by the disintegration of
pressurised vessels and pipes, J. Pressure Vessel Technol. 106 (1984)
362375.
[88] C.V. Moore, The design of barricades for hazardous pressure systems,
Nucl. Eng. Design 5 (1967) 8189.
[89] M.R. Baum, Failure of a horizontal pressure vessel containing a high
temperature liquid: the velocity of end-cap and rocket missiles, J. Loss
Prev. Process Ind. 12 (1999) 137145.
[90] M.R. Baum, The velocity of large missiles from axial rupture of gas
pressurized cylindrical vessels, J. Loss Prev. Process Ind. 14 (2001)
199203.
[91] U. Hauptmanns, A Monte Carlo-based procedure for treating the ight
of missiles form tank explosions, Probab. Eng. Mech. 16 (2001)
307312.
[92] U. Hauptmanns, A procedure for analyzing the ight of missiles from
explosions of cylindrical vessels, J. Loss Prev. Process Ind. 14 (2001)
395402.
[93] B.G. Cox, G. Saville, High Pressure Safety Code, High Pressure Technol.
Assoc. Imperial College, London, 1975.
[94] R.P. Kennedy, A review of procedures for the analysis and design of
concrete structures to resist missile impact effects, Nucl. Eng. Design 37
(1976) 183.
[95] National Defense Research Committee (NDRC), Effects of impact and
explosion, Summary Tech. Rep. of Div. 2, vol. 1, 1946.
[96] Q.M. Li, S.R. Reid, H.M. Wen, A.R. Telford, Local impact effects
of hard missiles on concrete targets, Int. J. Impact Eng. 32 (2005)
224284.
[97] R.W. White, N.B. Botsford, Containment of Fragments from a Run-
away Reactor, Rep. SRIA-113, Stanford Research Institute, Stanford, CA,
1963.
[98] C.M. Pietersen, Analysis of the LPG Accident in San Juan Ixhuatepec,
Mexico City, Report 85-0222, November 19, 1984, TNO, The Hague,
1985.
[99] R.J. Roark, W.C. Young, Formulas for Stress and Stain, McGraw-Hill,
New York, 1975.
[100] G. Gubinelli, S. Zanelli, V. Cozzani, Asimpliedmodel for the assessment
of the impact probability of fragments, J. Hazard. Mater. A 116 (2004)
175187.
[101] G.M. Makhviladze, J.P. Roberts, S.E. Yakush, Combustion of two-phase
hydrocarbon fuel clouds released into the atmosphere, Combus. Flame
118 (1999) 583605.
[102] F.K. Crawley, The effect of ignition of a major fuel spillage, in: The
Assessment of Major Hazards, Institution of Chemical Engineers, Rugby,
1982.
[103] I.A. Papazoglou, O.N. Aneziris, Uncertainty quantication in the health
consequences of boiling liquid expanding vapour explosion phenomenon,
J. Hazard. Mater. A 67 (1999) 217235.
[104] A.F. Roberts, Thermal radiation hazards from releases of LPG from pres-
surized storage, Fire Saf. J. 4 (1982) 197212.
[105] R.W. Prugh, Quantitative evaluation of reball hazards, Process Saf. Prog.
13 (1994) 83.
[106] A.M. Birk, Hazards from propane BLEVEs: an update and proposals for
emergency responders, J. Loss Prev. Process Ind. 9 (1996) 173181.
[107] A.M. Birk, An experimental investigation of a cylindrical vessel engulfed
in re with a burning relief valve are present, Chem. Eng. Commun. 80
(1989) 173191.
[108] A.J. Pierorazio, A.M. Birk, Dynamic behavior of transportation pressure
relief valves under simulated re impingement conditions, J. Pressure
Vessel Technol.: Trans. ASME 122 (2000) 6065.
[109] P.K. Raj, Exposure of a liqueed gas container to an external re, J.
Hazard. Mater. A 122 (2005) 3749.
[110] L.C. Shirvill, Efcacy of water spray protection against propane and
butane jet res impinging on LPG storage tanks, J. Loss Prev. Process
Ind. 17 (2004) 111118.
[111] F.N. Nazario, Preventing or surviving explosions, Chem. Eng. 102 (1988)
311.
[112] H. Londiche, R. Guillemet, Comparison of three protective devices for
BLEVEprevention, Loss Prev. Saf. Prom. Process Ind. 1 (1991) 551564.
[113] C. Young, LPG tank anti-BLEVE and automotive tank safety system. US
Patent 4,680,937 (2004).
[114] H.J. Chen, M.H. Lin, Modeling a BLEVE phenomena with application
to relief device design for liqueed ammonia storage, Ind. Eng. Chem.
Res. 38 (1999) 479487.
[115] A.N. Barator, N.A. Baratova, Y.A. Myshak, D.Y. Myshak, Proceeding
of the Third Asia-Oceania Symposium on Fire Science and Technology,
Seoul, 1998, pp. 405416.
[116] H.C. Hardee, D.O. Lee, Thermal hazard frompropane reballs, Transport
Plan. Technol. 2 (1973) 121133.
[117] J.A. Fay, D.H. Lewis, Unsteady burning of unconned fuel vapor clouds,
Combustion 16 (1977) 13971403.
[118] K. Hasegawa, K. Sato, Study on the reball following steam explosion of
n-pentane, Loss Prev. Saf. Prom. 2 (1977) 297303.
[119] K. Hasegawa, K. Sato, Fireballs, 12, Technical Memos of Fire Research
Institute of Japan, Tokyo, 1978, pp. 39.
[120] B.R. Williamson, L.R.B. Mann, Thermal hazards from propane (LPG)
reballs, Combus. Sci. Technol. 25 (1981) 141153.
[121] D.A. Lihou, J.K. Maund, Thermal radiation hazards from reballs, in:
The Assessment of Major Hazards, Institution of Chemical Engineers,
Rugby, 1982.
[122] J. Moorhouse, M.J. Pritchard, Thermal radiation hazards from large pool
res and reballsa literature review, in: The Assessment of Major Haz-
ards, Institution of Chemical Engineers, Rugby, 1982.
[123] J.A. Duiser, Escape of liqueed gases from broken pipes, in: The Protec-
tion of Exothermic Reactors and Pressurised Storage Vessels, Institution
of Chemical Engineers, Rugby, 1984.
[124] J.B. Gayle, J.W. Bransford, Size and duration of reballs from propellant
explosions, Report NASATMX-53314. George C. Marshall Space Flight
Center, Huntsville, USA, 1965.
[125] D.F. Bagster, R.M. Pitblado, Thermal hazards in the process industry,
Chem. Eng. Prog. 85 (1989) 6977.
[126] W.E. Martinsen, J.D. Marx, An improved model for the prediction of radi-
ant heat ux fromreball, in: Proceedings of the International Conference
and Workshop on Modeling the Consequences of Accidental Releases of
Hazardous Materials, San Francisco, 1999.
[127] B.E. Bader, A.B. Donaldson, H.C. Hardee, Liquid propellant rocket abort
reball model, J. Spacecraft Rockets 8 (1971) 12161222.
[128] I. Hymes, The Physiological and Pathological Effects of Thermal Radia-
tion, SRD R275, UK Atomic Energy Authority, Culcheth, 1983.
[129] I. Hymes, The effects on people and structures of explosion, blast, ther-
mal radiation and toxicity, in: J.I. Petts (Ed.), Seminar Proceedings of
the Major Hazardous Installation: Planning Implications and Problems,
Department of Chemical Engineering, Loughborough University of Tech-
nology, 1984.
[130] G.M. Makhviladze, S.E. Yakush, Proceedings of the Third Inter-
national Seminar on Fire and Explosion Hazards, 2000, pp. 579
589.
[131] C.M. Pietersen, S.C. Huerta, Analysis of the LPG Accident in San Juan
Ixhuatepec, Mexico City, Report 84-0222, November 19, 1984, Nether-
lands Organization for Applied Scientic Research, The Netherlands,
1985.

The Boiling Liquid Expanding Vapour Explosion (BLEVE) PDF

Uploaded by

Copyright:

Available Formats

The Boiling Liquid Expanding Vapour Explosion (BLEVE) PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

The Boiling Liquid Expanding Vapour Explosion (BLEVE) PDF

Uploaded by

Copyright:

Available Formats

Journal of Hazardous Materials 141 (2007) 489519

The boiling liquid expanding vapour explosion (BLEVE):

C and its pressures reaches 15 atm. The vessel

C), well below its atmospheric superheat limit tem-

C. However, like Prugh [16,17], they, too,

27 June 1993 Quebec, Canada Fire Propane 2.3 4d, 7i

18 March 1996 Palermo, Italy Collision in a

C, revealed that complete tank failure occurred

C for propane) and this could potentially lead to the most

C due to re engulfment, doubling the wall thickness and

C and 35 bar the tank pressure began to deviate from the

C when 80% full and

C when 40% full; the internal pressure before the vessel

C/min) assuming that there is an air gap between

arc. The liquid

to either side of the vessel front and rear axial directions

Fay and Lewis [117] Propane 6.28M

Solid ame models

C in 10 min during the course of full-scale

You might also like

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.