SMALL SCALE EXPERIMENTS ON BOILING LIQUID EXPANDING VAPOR EXPLOSIONS: SUPERCRITICAL BLEVE
ABSTRACT The most dangerous accident that can occur in LPG stor-age is the boiling liquid expanding vapor explosion (BLEVE). To better understand the rupture of the reservoir and the blast wave characteristics, small scale BLEVE experiments are performed with cylinders of 95 ml, filled at 86% with propane, laid hori-zontally and heated from below. A weakening of the reservoirs on the upper part allows better reproducibility of the rupture. High speed visualization, blast overpressure and surface reser-voir temperature are measured. Internal pressure measurement shows that the rupture pressure and temperature are well above the critical point. The fluid is then supercritical and there is no distinction anymore between liquid and gas prior rupture. This kind of reservoir rupture is significant of a new type of BLEVE, a supercritical BLEVE. The experiments also show that the fluid behavior during rupture differs with the size of the weakened part and therefore with the rupture pressure. Finally, the measured peak overpressures are compared with literature models. C p Specific heat [kJ/kgK] c Reservoir groove length [m] L Distance from source [m] L r Width of reservoir opening [m] E Expansion energy [kJ] m Total fluid mass [kg] n Fit parameters in Sadek equation  P Pressure [bar] r Cloud or Blast wave radius from reservoir [m] s Entropy [kJ/kg] T Temperature [K] t time [s] U Expanding velocity [m/s] u Internal energy [kJ/kg] v Specific volume [m 3 /kg] V Volume [m 3 ] GREEK SYMBOLS α Fit parameters in Sadek equation  β Constant in Genova model  γ Specific heat ratio of vapor [-] ∆t Time interval between 2 images [ms]
- [show abstract] [hide abstract]
ABSTRACT: Analytical expressions are presented for the initial decay of all major flow properties just behind planar, cylindrical, and spherical shock wave fronts whose trajectories are known as a function of either distance versus time or shock overpressure versus distance. These expressions give the time and/or distance derivatives of the flow properties not only along constant time and distance lines but also along positive and negative characteristic lines and a fluid-particle path. Conventional continuity, momentum and energy equations for the nonstationary motion of an inviscid, non-heat conducting, compressible gas are used in their derivation, along with the equation of state of a perfect gas. All analytical expressions are validated by comparing the results to those obtained indirectly from known self-similar solutions for planar, cylindrical and spherical shock-wave flows generated both by a sudden energy release and by a moving piston. Futhermore, time derivatives of pressure and flow velocity are compared to experimental data from trinitrotoluene (TNT), pentolite, ammonium nitrate-fuel oil (ANFO) and propane-oxygen explosions, and good agreement is obtained.
- [show abstract] [hide abstract]
ABSTRACT: Rupture of a container in which a liquid is held above its atmospheric-pressure boiling point can result in explosive vaporization of a large fraction of the contents. This phe nomenon is called a Boiling-Liquid Expanding-Vapor Explosion ("BLEVE"). A method for estimating the TNT equivalence of a BLEVE is proposed, with descriptions of possible causes and consequences of the explosion and (if the contents are flammable) the con sequences of the resulting fireball.Journal of Fire Protection Engineering 01/1991; 3(1):9-24. · 0.40 Impact Factor
- [show abstract] [hide abstract]
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 fireball or a toxic gas cloud. The damaging effect of BLEVEs is reflected 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.Journal of Hazardous Materials 04/2007; 141(3):489-519. · 3.93 Impact Factor
SMALL SCALE EXPERIMENTS ON BOILING LIQUID EXPANDING VAPOR
EXPLOSIONS: SUPERCRITICAL BLEVE
Delphine Laboureur∗, Jean-Marie Buchlin and Patrick Rambaud
Environmental and Applied Fluid Dynamics Department
von Karman Institute
Chauss´ ee de Waterloo, 72
1640 Rhode-St-Gen` ese, Belgium
Tel: +32(0)2 359 96 11
The most dangerous accident that can occur in LPG stor-
age is the boiling liquid expanding vapor explosion (BLEVE). To
better understand the rupture of the reservoir and the blast wave
characteristics, small scale BLEVE experiments are performed
with cylinders of 95ml, filled at 86% with propane, laid hori-
zontally and heated from below. A weakening of the reservoirs
on the upper part allows better reproducibility of the rupture.
High speed visualization, blast overpressure and surface reser-
voir temperature are measured.
Internal pressure measurement shows that the rupture
pressure and temperature are well above the critical point. The
fluid is then supercritical and there is no distinction anymore
between liquid and gas prior rupture. This kind of reservoir
rupture is significant of a new type of BLEVE, a supercritical
BLEVE. The experiments also show that the fluid behavior
during rupture differs with the size of the weakened part and
therefore with the rupture pressure. Finally, the measured peak
overpressures are compared with literature models.
Keywords: BLEVE, supercritical, blast wave, vessel rupture
A Area [m2]
aSound velocity [m/s]
∗Address all correspondence to this author.
Specific heat [kJ/kgK]
Reservoir groove length [m]
Distance from source [m]
Width of reservoir opening [m]
Expansion energy [kJ]
Total fluid mass [kg]
Fit parameters in Sadek equation 
Cloud or Blast wave radius from reservoir [m]
Expanding velocity [m/s]
Internal energy [kJ/kg]
Specific volume [m3/kg]
Fit parameters in Sadek equation 
Constant in Genova model 
Specific heat ratio of vapor [-]
∆t Time interval between 2 images [ms]
1Copyright © 2012 by ASME
Proceedings of the ASME 2012 Pressure Vessels & Piping Conference
July 15-19, 2012, Toronto, Ontario, CANADA
A BLEVE is defined by Walls  as an explosion resulting
from the catastrophic failure of a reservoir containing liquid
at a temperature significantly above its boiling point at normal
atmospheric pressure (i.e. superheated liquid). In this study,
small scale BLEVE experiments have been performed to better
understand the conditions under which the reservoir ruptures
and the characteristics of the blast wave generation, one of the
main BLEVE hazards.
BLEVE accidents appear mostly when the reservoir is
engulfed in fire.The heat increases wall temperature and
internal pressure and can induce wall-thinning and/or fissures.
The vessel then fails and vapor escapes. The resulting drop in
pressure superheats the liquid, causing it to boil rapidly and
violently. The reservoir rupture generates a blast wave and if the
liquid is flammable, it could ignite and form a fireball .
A more extensive comprehension of BLEVE events involves
two aspects: the conditions leading to the reservoir rupture and
the consequences of the rupture: overpressure, projection of
fragments and thermal radiation if a fireball is created. If an ex-
perimental approach is chosen, large or small scale experiments
can be performed. Large scale experiments are closer to reality,
but are expensive and difficult to control.
scale experiments are preferable for reproducible and controlled
This paper presents laboratory based results of BLEVEs
generated by heating of 95ml propane reservoirs filled at 86%.
The fire engulfing the reservoir has been mimicked by an elec-
trical resistor. To understand better the rupture conditions, the
effect of the reservoir strength (and therefore the burst pressure)
on the rupture has been investigated. Concerning the study of
BLEVE hazards, only the blast wave characteristics have been
studied as the rupture did not generated any fragment or fireball.
The blast wave intensity has been compared with overpressure
models available in literature [3,5,16–18].
BLEVE experiments can generate pressure wave, fireball
or fragment projection. Therefore, a dedicated facility called
BABELs was built at the von Karman Institute to guarantee
secure small scale experiments in a controlled environment.
FIGURE 1. BABELs EXPERIMENTAL FACILITY.
A general view of this facility is displayed in Fig. 1. BA-
and consists of a cylindrical chamber of 2m diameter, and 3m
high, with round-shaped flanges, made out of steel with a rated
pressure of 5bar. The setup has 3 series of 7 optical accesses
of 0.15m in diameter, separated by 90◦, and an elliptical door
of 0.57m × 0.77m. The setup allows air venting through open-
ings in its bottom and upper parts, the last one being ended by
an exhaust vent that can be used after each test to remove smoke
or gas from the chamber. The setup also includes a ladder and a
circumferential walking area are located at mid-height for better
accessibility of the upper optical accesses. All the experimental
results presented in this paper were carried out in BABELs.
According to literature [5, 6], the most important cause
of BLEVE is fire. The typical scenario involves a reservoir
engulfed in flames, causing an increase in liquid temperature
and therefore in internal pressure, until rupture.
This series of small scale experiments focused on this sce-
nario only and reproduced it by using horizontal 95ml cylindri-
cal reservoirs of 0.04m diameter, 0.138m long and 0.002m thick
(as shown in Fig. 2) filled with 86% volume of propane (mass:
voir was laid on a spiral microheater GA-XP from Micropyretics
Heaters International. This microheater was heated by Joule ef-
fect, being connected to an electrical power supply (set to 460W
in these experiments). To protect the microheater from the reser-
voir rupture, a metal plate was added above the microheater. The
reservoir was positioned on the plate inside a cradle for a better
heat distribution and stability.
2Copyright © 2012 by ASME
FIGURE 2.PROPANE RESERVOIR AND INSTRUMENTATION
As the reservoir fragmentation during rupture is a phe-
nomenon difficult to predict, the reservoir was weakened prior
to the experiment. A notch of 0.6mm deep (corresponding to
30% of thickness removed) and 0.01−0.08m long was made
along the length of the reservoir, and was located on top. The
influence of the groove length on the reservoir rupture has been
FIGURE 3. INTERNAL PRESSURE MEASUREMENT SETUP
Temperature was monitored by three type K thermocouples
located respectively between the plate and the cradle, at the top
and the bottom of the reservoir, and sampled at 3Hz. The blast
wave generated at rupture was recorded at 0.5m and 0.7m from
the reservoir by two PCB 106B50 pressure transducers, low-pass
filtered at 25 kHz in order to exclude the amplification effect
due to the sensor resonance frequency which is around 50kHz.
The sampling frequency was 250 kHz.
also carried out with a Validyne pressure sensor that measured
the internal pressure at 100Hz (see Fig. 3). This sensor was
calibrated with a dead weight tester 5020SMPR by Desgranges
& Huot. All these measurements were recorded by National
Instrument Compact DAQ and LabView acquisition software.
One experiment was
FIGURE 4.EDGERTON DIRECT SHADOWGRAPHY 
For visualization, a Phantom V7.1 high speed camera
was used, combined with a 12mm focal length objective. The
images were recorded at 14kHz. Simple visualization of the
fluid ejection after rupture was performed with the help of two
IMPLEMENTATION OF EDGERTON DIRECT SHAD-
For observation of the pressure wave generated by the rup-
ture, an Edgerton’s direct shadowgraph technique was used. The
purpose of this technique, as showed in Fig. 4, is to illuminate an
object S with a point source L (H4 halogen lamp, 12V, 60/55W)
and to look simultaneously at the object and its shadow on a
screen (Oralite retro-reflective film 5300). The major problem
of this technique is the slight offset between the light source and
the camera position that causes double imaging. In order to min-
3Copyright © 2012 by ASME
imize this problem, the light source needs to be as close as possi-
ble to the camera. This was done by illuminating perpendicular
to the interesting area and by diverging the light, which was fo-
cused by using two spherical lenses, with a mirror positioned the
closest possible to the camera (Fig. 5).
INTERNAL PRESSURE MEASUREMENT
One of the BLEVE experiments was performed with a small
notch (c = 0.015 m) and a pressure transducer measuring the
internal pressure. Figure 6 shows the internal pressure evolution,
plotted against the temperature of the top of the reservoir. This
measurement is compared with the experiment of Stawczyk 
that generated a BLEVE with a 5kg reservoir of propane, filled
FIGURE 6.VAPOR PRESSURE VS. TEMPERATURE (c=0.015m)
Both experiments follow the same trend. At first, the inter-
nal pressure is following the saturation line. At one point below
the critical point, which is defined as the highest temperature and
pressure at which the liquid and vapor phase of a substance can
be in equilibrium, the measurements deviate from the saturation
line and show a linear increase of pressure with temperature. The
fluid is then supercritical, so temperature and pressure are above
the critical point, meaning that distinct liquid and gas phases do
not exist. As the fluid mass occupies the whole volume of the
reservoir, the density is fixed, which causes a linear increase of
pressure with temperature. This linear evolution can be modeled
with the equation of state that has been developed by Miyamoto
and Watanabe . The density is chosen as the total fluid mass
divided by the total volume of the reservoir.
The vessel then fails when the pressure reaches the maxi-
mum stress sustainable by the reservoir. The applied heat flux in
these experiments does not induce a change in the mechanical
properties of the reservoir because the measured temperature
range is too low. In our measurements, the vessel fails around
370bar. The specifications of the reservoir set the burst pressure
around 440bar. This difference is mainly caused by the presence
of a groove of 0.015m length on the reservoir. The change in the
linear progression of the pressure - temperature curve close to
the rupture can be linked to the fact that the measured pressure
becomes close to the sensor proof pressure, causing a non-linear
response of the sensor in that range. Therefore, the true rupture
pressure could a bit higher. The large difference between the
burst pressure of the present measurement and the result of
Stawczyk is probably due to a difference in the theoretical burst
pressure, as the design of the reservoir is different (larger diam-
eter and smaller thickness). As both Stawczyk and the present
experiments observed a supercritical BLEVE, this phenomenon
FLUID EJECTION PHENOMENON
FIGURE 7.RESERVOIR RUPTURE c = 0.077m, ∆t = 0.28ms
From the different experiments performed with the 95ml
reservoir, at rupture, the majority of fluid content was ejected
from the reservoir in a few milliseconds. In this type of rupture,
the propane did not ignited and no fireball was noticed.
A strongly weakened reservoir (c = 0.077m) burst can be
observed in Fig. 7 from high speed visualization. At rupture,
the rapid expansion of propane results in a turbulent spheri-
cal shaped cloud of vapor and small droplets with blurry con-
tours. The cloud grows with time and the two-phase ejection
decreases (as well as the jet release angle). Once the reservoir is
empty, the cloud is diffusing and the droplets are evaporating un-
til complete visual disappearance. If the groove length is smaller
(c = 0.015m), the release is somewhat different, as seen in Fig.
8. The cloud shape is almost perfectly spherical on its upper part,
like a dome.
4 Copyright © 2012 by ASME
FIGURE 8.RESERVOIR RUPTURE, c = 0.015m, ∆t = 0.28ms
From the high-speed visualizations, it is difficult to conclude
on the exact thermodynamic state of the fluid forming the cloud,
and why the shape of the cloud is changing as the rupture pres-
sure increases. To better understand the rupture phenomenology,
this experimental study is compared with the studies of Wu
and Lin [11–13], which were resumed by Lamanna , that
have performed a parametric study about downstream jets of
supercritical fluids into sub-critical environments by varying the
FIGURE 9.ENTROPY VS. PRESSURE DIAGRAM
Lin et al.  have explained the physics of an expanding
supercritical jet based on an analysis of the thermodynamic
transformations occurring within the jet. Following Lin et
al. , the rupture pressure and corresponding entropy of the
experimental results (with varying groove lengths) are plotted in
a pressure - entropy diagram. Lin et al. have showed that, if after
the jet expansion, assumed to be isentropic, the fluid conditions
fall inside the two-phase region, the jet condensates. The Figure
9 shows that all the experiments, after rupture and isentropic ex-
pansion of the fluid, fall in the two-phase region. Therefore, the
observed clouds of Fig. 7 and 8 are propane condensation. In ad-
dition, Lin et al.  have published shadowgraph images of the
condensing jet, having an opaque appearance. The shadowgraph
images recorded for the present experiments have also an opaque
appearance (see Fig. 16), another sign of the cloud condensation.
The pressure - entropy diagram can also provide explanation
for the different cloud shapes. The fluid expansion that ended in
a blurry cloud (as Fig. 7) has an entropy lower than the critical
entropy. In contrast, the fluid expansion that forms a smooth
dome-like cloud (as Fig. 8) has an entropy larger than the critical
The radius and total area of the cloud were calculated by
image processing. The cloud upper boundary has been obtained
by detection technique based on Forward Step Filter algorithm,
available in the in-house LEDAR software (Level Detection and
Recording) . To determine the cloud radius, a circular fit
centered on the NTG reservoir was performed on the detected
cloud boundary of each image.
CLOUD EXPANDING VELOCITY
LEFT: CLOUD RADIUS EVOLUTION, RIGHT:
Figure 10 left shows the time evolution of the cloud radius
for two types of rupture: the blurry cloud observed with a weak-
ness higher than 0.015m (here c = 0.04m) and the dome-like
cloud observed with a smaller weakness (here c = 0.01m). Just
after the reservoir rupture during around 0.5ms, both clouds are
growing in a similar way. But while the blurry cloud continues
to increase, the dome-like cloud stabilizes. This phenomenon
5Copyright © 2012 by ASME
is also visible in Fig.
expanding velocity derived from the cloud radius evolution.
The dome-like shape of the cloud is slowing down the cloud
10 right, which represents the cloud
CLOUD AREA FOR DIFFERENT RESERVOIR
Once the ejection slows down, it is too difficult to track
the interface. Indeed, the cloud upper interface becomes less
clear and the major change in the cloud shape is due to a
decrease in the jet release angle, visible in the bottom part of
the cloud, close to the reservoir. Therefore, the total cloud area
was calculated with Matlab algorithm, using the high-speed
visualization transformed into Black & White images, with an
intensity threshold level of 5%. It can be observed in Fig. 11
that all the clouds are initially growing in a similar way, which
confirms the observations of Fig. 10. The cloud duration is
defined as the time period during which the cloud is visible
(i.e. has a non zero area in Fig. 11). In the Figure 11, it can
be observed that the cloud duration decreases with the groove
length. This is confirmed by looking at the cloud evolution in
Fig. 7 and 8 where both series of images are displayed with the
same time interval.
The Figure 12 shows an infrared view of the cloud at rup-
ture. The temperature scale is based on black body radiation and
therefore represents only qualitative information since the cloud
emissivity is not known. At the center and close to the reservoir,
the cloud remains hotter than the environment.
As explained in the instrumentation section, a groove was
drilled on the top of the reservoir to improve the reproducibility
of the rupture. Looking at all the rupture patterns (around 30
FIGURE 12.INFRARED VIEW OF THE CLOUD
bottles were ruptured), the whole series of reservoir failed in
the same way. The rupture started at the weakness spot and
propagated along the notch. Then, the sides of the notch opened,
forced by th ejection of the fluid (see Fig. 13).
0.08m, RIGHT: c = 0.01m
RESERVOIR RUPTURE PATTERN. LEFT: c =
There is a direct relationship between the groove length and
the loss of containment; the less the reservoir is weakened, the
larger is the opening (as observed in Fig. 13). Figure 14 shows a
linear increase of the reservoir opening as the groove length de-
creases, with a jump for the smaller weakness which gives even
larger opening. This small groove length corresponds to the ap-
parition of the dome-like cloud during the fluid ejection, which
would suggest that the force of the ejection for this phenomenon
is higher than for the blurry cloud. The evolution of the rup-
ture pressure with the groove length is similar to the reservoir
opening; the less the reservoir is weakened, the stronger is the
mechanical resistance and so the larger is the rupture pressure. A
higher rupture pressure also increases the fluid energy available
at rupture, causing a larger opening of the reservoir and a faster
ejection of the fluid, as observed in Fig. 11.
6Copyright © 2012 by ASME
LEFT: RESERVOIR OPENING WIDTH, RIGHT:
One physical effect of a BLEVE explosion is the generation
of a blast wave.Overpressure measurements taken at two
distances from the reservoir, and for two groove lengths are
showed in Fig. 15. The blast wave pressure at a fixed reference
distance shows an initial peak, followed by a negative part and a
second peak. The second peak can be due to the reflection of the
pressure wave on the walls of the experimental facility.
OVERPRESSURE GENERATED BY RESERVOIR
The value of the first peak decreases with distance from
the reservoir, and increases for less weakened reservoir.
overpressure modeling [3, 5, 16], if all other variables are kept
constant (like the type of liquid, the fill level, the reservoir size,
etc...), the overpressure at a given distance depends directly on
the pressure inside the reservoir before the rupture. Therefore,
when the reservoir is less weakened, the rupture happens at a
higher internal pressure, and generates a higher blast wave.
The blast wave generated by the vessel burst can also be
observed on shadowgraph images (Fig. 16). The blast wave is
spherical and appears ahead of the fluid cloud. The blast wave
shows a contrast compared to the background which is higher in
the upper part of the wave compared to the sides. This difference
of contrast can be linked to the strength of the blast: a higher
contrast in the shadowgraph is representative of a stronger blast
wave. As explained in the previous section, the cloud appears
opaque in the shadowgraph, due to condensation.
TER: c = 0.04m, RIGHT: c = 0.01m
SHADOWGRAPH IMAGES. LEFT: c = 0.08m, CEN-
The same image processing as performed on high-speed vi-
sualization was applied on the shadowgraphs, i.e. the use of
LEDAR software  to detect the blast wave position and the
cloud upper surface. Figure 17 shows the cloud and blast wave
radius evolution and expansion velocity.
with Eq. 1, where α1,α2,α3,α4and n are fitted parameters. This
expression is a curve fit of experimental shock trajectory data
for TNT, ANFO and propane-oxygen explosions developed by
Sadek . The fit quality is not guaranteed at higher distance
since there is no measured data. But it can help to see that the
blast wave travels faster , and that it is much less slowed down
than the cloud.
In literature, the blast wave overpressure is usually modeled
with generalized methods based on thermodynamic equations.
The procedure first consists in calculating an expansion energy.
This energy is based on the change of thermodynamic state
7Copyright © 2012 by ASME
RIGHT: CLOUD EDGE AND BLAST WAVE EXPANDING VELOC-
LEFT: CLOUD AND BLAST WAVE RADIUS,
of the substance stored in the vessel from the initial state (the
moment before the explosion at the time t1) to the final state
(the instant t2at boiling temperature and atmospheric pressure).
Depending on the model, different approaches are followed to
express this expansion energy. The energy is then transformed
into a TNT equivalent mass (1kg of TNT is equal to 4680kJ).
The overpressure at a given distance L is determined from TNT
equivalent curves  that link the peak overpressure to the
distance scaled by the TNT equivalent mass.
All the models for BLEVE overpressure available in the
literature were developed for an overpressure generated after
the rupture of a liquid gas reservoir at a pressure lower than
the critical one, where both a liquid and a vapor phase are
present. But in the present experiments, at rupture, the fluid is
supercritical, which means that distinct liquid and gas phases do
not exist anymore. Therefore, all the models using an expansion
energy calculated from both the liquid and vapor energies do not
In 1991, Prugh  has defined expansion energy assuming
isentropic expansion and ideal gas behavior . Two formulations
have been proposed depending if the reservoir is filled with gas
only or with liquefied gas. As the fluid before rupture in this
study is supercritical, the formulation chosen here is the one de-
veloped for gas filled reservoir (see Eq. 2).
In 2006, Casal  has introduced the liquid superheating
energy, defined as the difference between the enthalpy of the liq-
uid prior rupture, and the enthalpy of the liquid at the saturation
temperature corresponding to the atmospheric pressure. The fi-
nal TNT mass that will contribute to the blast generation is sup-
posed to be only a fraction of the expansion energy: 5% for an
irreversible expansion and 14% for an isentropic expansion. In
2008, Genova  has proposed a new modeling of the expan-
sion energy, assuming that this energy is mainly due to the liquid
flash that can be seen as a thermal phenomenon, linked to the
excess of heat stored inside the liquid. The expansion energy is
then modeled as Eq. 3 where β is an empirical coefficient, set to
0.07. Initially, Casal and Genova have used only the liquid mass
in their model, but here as the fluid is supercritical, the total mass
In 2007, Birk  has showed that the expansion energy,
based on the change in internal energy of the stored vapor as-
suming an isentropic expansion, gives the highest overpressure
that can be expected from a given configuration. Initially, Birk
has used only the vapor mass in the model, but here as the fluid
is supercritical, the total mass is used.
The thermophysical properties needed in these different models
are found based on the rupture pressure at constant density (as
discussed in the internal pressure analysis).
WITH MEASUREMENTS, INFLUENCE OF Prupt
8Copyright © 2012 by ASME
Figure 18 shows the influence of the rupture pressure on the
blast wave peak value measured at 0.5 m. The first point of
Fig. 18 (square symbol) is more precise in terms of modeling
because the internal pressure was recorded. The other measure-
ments were performed without internal pressure measurement,
and the rupture pressure is determined from the upper reservoir
temperature, using the linear P - T behavior at supercritical state.
TABLE 1.DETAILS OF 0.042KG BLEVE EXPERIMENTS
All the models and experiments show a linear increase of the
overpressure with the rupture pressure. Prugh model has a differ-
ent slope compared to the other models. This can be due to the
assumption of ideal gas which is not postulated by the other au-
thors. The best fit with the experiments is obtained with the mod-
els of Casal assuming an irreversible expansion and the model of
Genova. But the model of Casal shows a slightly smaller dis-
crepancy between the model and the experiments. The uncer-
tainty linked to the overpressure measurement lies between 2.9
and 3.6% of the measured value for the measured overpressure
range. Therefore, measurement uncertainty does not affect the
WITH CASAL MODELING
BLEVE EXPERIMENTS OVERPRESSURE SCALED
Figures 19 and 20 gather the series of overpressure measure-
ments from Fig. 18 and detailed in Table 1. Overpressure in each
test was measured at two distances: 0.5m and 0.7m. Each num-
ber in the figures corresponds to one experiment. In addition,
the overpressure measured by Stawczyk, two blast wave signals
recorded at 2 and 10 m, generated by the rupture of a 11 kg bot-
tle filled at 80%, are also displayed. The Fig. 19 represents the
measurements scaled with the Casal model of irreversible expan-
sion, which shows the best fit with the experimental data. All the
measurements are well aligned with the model, but a part of the
experiments are underestimated by Casal modeling. In a con-
servative approach of predicting a maximum value of the over-
pressure, but with a bigger error, Prugh and Birk can be used.
Prugh  is predicting better the overpressure, but assumes an
ideal gas behavior which is not representing well a BLEVE rup-
ture. So to be conservative and based on correct assumptions,
Birk  is the more suitable model. The measurements scaled
with the Birk model are represented in Fig. 20.
WITH BIRK MODELING
BLEVE EXPERIMENTS OVERPRESSURE SCALED
Small scale experiments of BLEVE with 95 ml propane
reservoirs heated from below by a microheater have been
performed to better understand the conditions under which the
reservoir ruptures and the characteristics of the blast wave.
The fluid state before rupture of the small scale reservoir is
supercritical, as showed by the internal pressure measurement.
As in these rupture conditions, liquid - vapor interactions do not
exist, this type of reservoir rupture is significant of a new type
of BLEVE: a supercritical BLEVE. A supercritical BLEVE will
happen mostly with small scale reservoir, as they have a higher
9 Copyright © 2012 by ASME
When expanding, the supercritical fluid condensates, as
proven by shadowgraph images and pressure-entropy diagram.
When the rupture pressure increases, the fluid cloud is observed
with a dome-like shape, which is linked to rupture conditions
but is not yet completely explained. An increase in the rupture
pressure is also increasing the opening width of the reservoir
and decreasing the cloud duration. An increase of the rupture
pressure is also increasing the blast wave intensity.
The blast wave generated by the reservoir rupture is
spherical and propagates faster than the fluid cloud.
fluid is supercritical at rupture, the overpressure models from
previous authors [3,5,16,18] have been adapted for being used
with supercritical rupture conditions. The comparison of the
adapted models and the experiments has shown that the model of
Casal, assuming an irreversible expansion, reproduces the best
the measurements, with a minimum error. But this model can
underestimate the blast. Therefore, if a conservative approach is
needed, the model of Birk is prefered.
The authors wish to thank the CEA of Gramat which is sup-
porting this study and collaborating with the von Karman Insti-
tute and the´Ecole des Mines d’Al` es in this joint research project
about source term characterization in major hazards.
 Center for Chemical Process Safety, 1994. Guidelines for
evaluating the characteristics of vapor cloud explosions,
flash fires and BLEVEs. Wiley-VCH, Weinheim.
 Sadek, H., and Gottlieb, J., 1983. “Initial decay of flow
properties of planar, cylindrical and spherical blast waves”.
UTIASTechnicalNoteNo.244, CNISSN0082-5263, pp.1–
 Genova, B., Silvestrini, M., and Trujillo, F., 2008. “Eval-
uation of the blast wave overpressure and fragment initial
velocity for a bleve event via empirical correlations derived
by a simplified model of released energy”. Journal of Loss
Prevention in the Process Industries, 21(1), pp. 9–24.
 Walls, W., 1979. “The BLEVE - part 1”. Fire Command,
17, pp. 22–24.
 Prugh, R., 1991. “Quantitative evaluation of bleve haz-
ards”. Journal of Fire Protection Engineering, 3(1), pp. 9–
 Abbasi, T., and Abbasi, S., 2007. “The boiling liquid ex-
panding vapour explosion (BLEVE): Mechanism, conse-
quence assessment, management”. Journal of Hazardous
materials, 141(3), pp. 489–519.
 Settles, G., Grumstrup, T., Miller, J., Hargather, M., Dod-
son, L., and Gatto, J., 2005. “Full-scale high-speed edger-
ton retroreflective shadowgraphy of explosions and gun-
shots”. In Proceedings of PSFVIP, 5th Pacific Symposium
on Flow Visualisation and Image Processing. PSFVIP-5-
 Stawczyk, J. “Experimental evaluation of LPG tank explo-
sion hazards”. Journal of hazardous materials, 93(2).
 Miyamoto, H., and Watanabe, K., 2000.
namic property model for fluid-phase propane”. Interna-
tional journal of thermophysics, 21(5), pp. 1045–1072.
 SubhanandaRao, A., VenkataRao, G., andNageswaraRao,
B., 2005. “Effect of long-seam mismatch on the burst pres-
sure of maraging steel rocket motor cases”. Engineering
Failure Analysis, 12(2), pp. 325–336.
 Wu, P., Chen, T., Nejad, A., and Carter, C., 1996. “In-
jection of supercritical ethylene into nitrogen”. Journal of
Propulsion and Power, 12(4), pp. 770–777.
 Wu, P., Shahnam, M., Kirkendall, K., Carter, C., and Nejad,
A., 1999. “Expansion and mixing processes of underex-
panded supercritical fuel jets injected into superheated con-
ditions”. Journal of Propulsion and Power, 15(5), pp. 642–
 Lin, K., Stouffer, S. C., and Jackson, T.
ture and phase transition processes of supercritical
methane/ethylene mixtures injected into a subcritical envi-
ronment”. Combustion Science and Technology, 178(1).
 G. Lamanna, I. Stotz, B. W., 2009. “Supercritical fluid in-
jection in quiescent flow fields: modelling and mixing en-
hancement”. Von Karman Lecture Series, 2009-04, pp. 1–
 Planquart, P., Riethmuller, M., Joly, A., and Detry, J., 2002.
“Real time optical detection and characterisation of water
model free surfaces: Application to the study of mold”.
4th European Continuous Casting Conference, Birming-
ham, UK. pp. 14-16.
 Casal, J., and Salla, J., 2006.
ing energy for a quick estimation of overpressure in bleves
and similar explosions”. Journal of hazardous materials,
137(3), pp. 1321–1327.
 TNO - Committee for the Prevention of Disasters, 1997.
“Methods for the calculation of physical effects due to re-
leases of hazardous materials (liquids and gases) - Yellow
Book”. CPR14E, third edition.
 Birk, A., Davison, C., and Cunningham, M., 2007. “Blast
overpressures from medium scale bleve tests”. Journal of
Loss Prevention in the Process Industries, 20(3), pp. 194–
“Using liquid superheat-
10Copyright © 2012 by ASME