ArticlePDF Available

Metallurgical failure analysis of a propane tank boiling liquid expanding vapor explosion (BLEVE)

Authors:

Abstract and Figures

A severe fire and explosion occurred at a propane storage yard in Truth or Consequences, N.M., when a truck ran into the pumping and plumbing system beneath a large propane tank. The storage tank emptied when the liquid-phase excess flow valve tore out of the tank. The ensuing fire engulfed several propane delivery trucks, causing one of them to explode. A series of elevated-temperature stress-rupture tears developed along the top of a 9800 L (2600 gal) truck-mounted tank as it was heated by the fire. Unstable fracture then occurred suddenly along the length of the tank and around both end caps, along the girth welds connecting the end caps to the center portion of the tank. The remaining contents of the tank were suddenly released, aerosolized, and combusted, creating a powerful boiling liquid expanding vapor explosion (BLEVE). Based on metallography of the tank pieces, the approximate tank temperature at the onset of the BLEVE was determined. Metallurgical analysis of the ruptured tank also permitted several hypotheses regarding BLEVE mechanisms to be evaluated. Suggestions are made for additional work that could provide improved predictive capabilities regarding BLEVEs and for methods to decrease the susceptibility of propane tanks to BLEVEs.
Content may be subject to copyright.
33
Journal of Failure Analysis and Prevention Volume 5(5) October 2005
(Submitted May 17, 2005)
A severe fire and explosion occurred at a propane storage yard in Truth or Consequences, N.M., when a
truck ran into the pumping and plumbing system beneath a large propane tank. The storage tank emptied
when the liquid-phase excess flow valve tore out of the tank. The ensuing fire engulfed several propane
delivery trucks, causing one of them to explode. A series of elevated-temperature stress-rupture tears
developed along the top of a 9800 L (2600 gal) truck-mounted tank as it was heated by the fire. Unstable
fracture then occurred suddenly along the length of the tank and around both end caps, along the girth
welds connecting the end caps to the center portion of the tank. The remaining contents of the tank were
suddenly released, aerosolized, and combusted, creating a powerful boiling liquid expanding vapor explosion
(BLEVE). Based on metallography of the tank pieces, the approximate tank temperature at the onset of
the BLEVE was determined. Metallurgical analysis of the ruptured tank also permitted several hypotheses
regarding BLEVE mechanisms to be evaluated. Suggestions are made for additional work that could
provide improved predictive capabilities regarding BLEVEs and for methods to decrease the susceptibility
of propane tanks to BLEVEs.
Keywords:
JFAPBC (2005) 5:33-42 © ASM International
DOI: 10.1361/154770205X????? 1547-7029 / $19.00
Background
A small pickup truck, presumably parked in
neutral and with the parking brake off, rolled into a
propane storage yard in Truth or Consequences,
N.M., and impacted the pumping and plumbing
system beneath a large propane storage tank. The
vapor-phase excess flow valve (EFV) fractured in
the threads below the valve portion, leaving the actual
valve in place and functioning. The liquid-phase
EFV, however, tore out of the tank, providing an
approximately 5.6 cm (2.2 in.) diameter opening
through which the tank emptied. As the liquid and
gaseous propane spread, it was eventually ignited
by the pilot light of a hot water heater in a mobile
home roughly 120 m (400 ft) away. The fire flashed
back, burning off the area over which the liquid and
vapor had dispersed, and igniting a severe fire where
the liquid propane continued to flow from the main
tank (through the opening where the liquid-phase
EFV had been). Four propane delivery trucks parked
in the yard were engulfed in the fire. The 9800 L
(2600 gal) propane tank on the truck exposed to
the hottest portion of the fire exploded, creating a
boiling liquid expanding vapor explosion (BLEVE).
D.F. Susan, K.H. Eckelmeyer, and A.C. Kilgo, Sandia National Laboratories, P.O. Box 5800, Albuquerque, NM 87185. Contact
e-mail: dfsusan@sandia.gov.
BLEVE, creep rupture, metallography, microstructure, pressure vessel steel, propane tank
Metallurgical Failure Analysis of a Propane Tank
Boiling Liquid Expanding Vapor Explosion (BLEVE)
D.F. Susan, K.H. Eckelmeyer, and A.C. Kilgo
This caused a fireball hundreds of feet in height,
tore off and threw the tank end caps over 90 m (300
ft), and completely flattened the 2.4 m (8 ft) long
by 11 mm (7/16 in.) thick cylindrical portion of the
steel tank. Fortunately, the fire was extinguished
without loss of life or serious injury.
The tank that experienced BLEVE was made of
ASTM A202B steel with a measured composition
of 0.19% C, 1.25% Mn, 0.76% Si, and 0.47% Cr,
bal iron. Its hardness was determined to be Rockwell
B 95, indicative of a room-temperature ultimate
tensile strength of approximately 100 ksi (690
MPa). This is consistent with the properties expected
for steel of this composition.
Two pressure relief valves were located on top of
the tank approximately 33 cm (13 in.) from each
end. These were rated to operate at 1.7 MPa (250
psi) and appeared to have functioned appropriately
prior to the BLEVE.
Failure Investigation of Excess
Flow Valves on Stationary Tank
The vapor-phase fitting performed as expected:
34 Journal of Failure Analysis and Prevention
Volume 5(5) October 2005
Metallurgical Failure Analysis of a Propane Tank BLEVE
(continued)
It fractured downstream of the EFV. The liquid-
phase fitting pulled out rather than fracturing,
allowing the tank to empty. This may have occurred
due to insufficient thread engagement. Only four
threads were engaged on the liquid-phase valve,
compared to the typical value of five threads for a
hand-tightened fitting of this size, and six to seven
threads on the vapor-phase valve, which performed
satisfactorily. A comparison of the conditions
required for fracture (desirable) versus thread pull-
out (undesirable) indicated that thread pull-out may
be expected with thread engagements of approx-
imately four or less. This paper does not address the
conditions of the storage tank valves but rather
focuses on a metallurgical investigation of the
ruptured tank.
BLEVE of Truck-Mounted Tank
Preliminary Observations
The configuration of the main propane storage
tank and the truck-mounted tank is shown
schematically in Fig. 1. The location of the propane
leak and ensuing fire is shown along with the
approximate fracture path of the truck-mounted
tank. The two opposing fracture surfaces, labeled
A” and “B,” are shown before and after the BLEVE.
Evidence from the site suggests that, following the
BLEVE, side B of the now-flattened steel plate was
exposed to the hottest portion of the subsequent
fire (which burned for 15 to 30 min after the
explosion).
Longitudinal fracture occurred along the top of
the tank, beginning primarily near the center of the
tank (see subsequent details). As the crack extended
from its origin, it deviated around the pressure relief
valves located approximately 33 cm (13 in.) from
each end, then continued circumferentially around
the end caps, separating the end caps along the weld
from the cylindrical portion of the tank (Fig. 1).
The cylindrical portion of the tank unrolled and
was flattened by the force of the explosion.
As shown in Fig. 2, samples were removed for
metallographic analysis from sides A and B of the
fracture at various locations along the fracture path.
In addition, samples were removed from the cooler
bottom portion of the tank (paint remained intact;
no signs of very high temperature exposure) to
characterize the original microstructure of the steel.
Significant wall thinning of the tank was observed
along much of the longitudinal fracture. Wall-
thickness measurements along the fracture indicated
that the greatest amount of thinning occurred near
the middle of the tank (approximately 80% reduc-
tion), and that the extent of thinning decreased
gradually toward the ends of the tank (Fig. 3).
Abrupt changes in the extent of thinning were also
observed near each of the pressure relief valves.
The appearance of the fracture surface also varied
substantially over the length of the tank, as described
in Table 1. Fracture surfaces in the thinnest regions
were rough and oriented approximately perpen-
Fig. 1 Schematic diagram of propane tank yard before the
BLEVE, and the flattened truck-mounted tank after
the BLEVE
Fig. 2 (a) Locations of metallographic samples. Samples were
also removed from side B at matching locations along the
fracture. (b) Location of samples for characterization of
original structure
[author: Figure 3, please advise if the regions should be A
35
Journal of Failure Analysis and Prevention Volume 5(5) October 2005
dicular to the circumference. This is typical of failure
that occurs gradually due to elevated-temperature
plastic deformation followed by stress rupture. As
the amount of thinning decreased, the fracture
surfaces became smoother and oriented at 45° to
the circumference. These 45° regions are typical of
unstable, rapid crack propagation occurring in
relatively ductile materials. Due to severe oxidation
of the fracture surfaces, scanning electron microscope
fractography was not possible.
It appears that tank failure began by stable
elevated-temperature stress-rupture tears gradually
developing in three (or perhaps four) regions: on
either side of the middle of the tank (regions F and
H in Table 1) and in the vicinity of one or both of
the pressure relief valves (regions B and K in Table
1). The tears near the middle of the tank (regions F
and H) appear to have developed first but were
misaligned approximately 6.4 cm (2½ in.) from one
another along the circumference. Smaller tears
developed near the pressure relief valves (regions B
and K) as tears F and H were each growing longer.
Fig. 3 Measured wall thickness along main fracture surface. See
text and Table 1 for description of regions A to K.
Table 1 Description of Fracture Surfaces
Corresponding wall-thickness measurements are given in Fig. 3.
Region Location Description Comments
A 0-33 cm 45° relatively smooth shear failure with Unstable fracture
(0-13 in.) relatively little thickness reduction. Thinning
decreases toward end of tank.
B 33-58 cm Mixed-mode, much rougher fracture with Region where stress-rupture tear was
(13-23 in.) greater thinning. Short 90° segment at 41- developing, then was overtaken by unstable
43 cm (16-17 in.); remainder mixed 90° fracture propagating from region F/H
fracture and 45° shear lips
C 58 cm 3.8 cm (1½ in.) circumferential tear Linked propagating crack from region D with
(23 in.) pre-existing stress-rupture tear in region B
D 58-69 cm Shear failure similar to region A but with Unstable fracture
(23-27 in.) more thinning
E 69-89 cm Predominantly shear failure but increasingly Likely a region of incipient stress-rupture tearing
(27-35 in.) V-shaped and rough and with increasing that was overtaken by unstable fracture
thinning propagating from regions F/H
F 89-157 cm Rough 90° fracture with extensive thinning Region of extensive elevated-temperature plastic
(35-62 in.) deformation culminating in stress rupture
G 157 cm 6.4 cm (2½ in.) circumferential tear Circumferential link between tears F and H
(62 in.)
H 157-185 cm Rough 90° fracture with extensive thinning Same as region F
(62-73 in.) (same as F)
J 185-221 cm Relatively smooth 45° shear failure; thinning Same as region D
(73-87 in.) decreases toward end cap (similar to D)
K 221-229 cm Mixed-mode, much rougher fracture with Similar to region B, but stress-rupture tear not yet
(87-90 in.) greater thinning; mixed 90° fracture and fully developed
45° shear lips
L 229-264 cm 45° relatively smooth shear failure with Same as region A
(90-104 in.) relatively little thickness reduction. Thinning
decreases toward end of tank.
36 Journal of Failure Analysis and Prevention
Volume 5(5) October 2005
Metallurgical Failure Analysis of a Propane Tank BLEVE
(continued)
Eventually, tears F and H linked up via a short
circumferential tear (region G). At some point, the
process of gradual tearing was interrupted, and
catastrophic failure occurred as unstable shear
fracture extended out from tear F/H. Unstable shear
fracture propagated toward both ends of the tank
through regions E, D, and J, grew through the
developing stress-rupture tears in regions B and K,
continued to the ends of the tank through regions
A and L, and then propagated around and through
the end caps.
The onset of unstable crack growth may have
occurred when tears F and H linked, because the
joining of these two shorter cracks to form a much
longer one would cause a sudden increase in
maximum stress intensity (approximately 20%).
However, there is no way to conclude unequivocally
that this linkage defined the onset of instability.
Also, an unusual area was found approximately 117
cm (46 in.) from one end of the tank (in region F)
Fig. 4 Microstructure in “cool” region of tank (see Fig. 2b). 2%
nital etch
along the stable longitudinal tear. This region
exhibited an unusually rough fracture surface and
secondary cracks adjacent to the primary tear on
the outside surface of the tank. The significance of
this region is discussed in detail later.
Microstructural Analysis
The microstructure in regions that were not
significantly heated (where paint remained on the
steel) consisted of ferrite and pearlite, typical of a
hot-rolled steel (Fig. 4). Analysis of the banding
structure on several samples indicates that the plate
rolling direction is circumferential around the tank.
Also, it was found that several plates were welded
together to form the cylindrical portion of the
tank. The ferrite morphology was mixed blocky
and acicular, perhaps indicative of controlled-
temperature finish rolling or mildly accelerated
cooling following rolling.
Metallographic cross sections were taken at a
(a)
(b)
Fig. 5 Microstructure of sample 3a far from the fracture surface.
Microstructure is similar to Fig. 4, indicating that
reaustenitization had not occurred. (a) 200
×
. (b) 500
×
(a)
(b)
37
Journal of Failure Analysis and Prevention Volume 5(5) October 2005
number of locations along the longitudinal fracture
(Fig. 2a). All of the samples from side A of the tear
exhibited microstructures similar to the base mater-
ial, indicating that no austenitizing had occurred
(Fig. 5). Evidence of spheroidization of the pearlite
was observed (Fig. 6b), indicating some elevated-
temperature exposure of side A, but this could have
occurred either prior to or following unstable
fracture. This evidence shows that side A material
did not undergo any phase transformation prior to
or during the BLEVE event. Samples taken from
side B of the fracture exhibited a markedly different
microstructure, with equiaxed ferrite and pearlite,
characteristic of normalized steel (Fig. 7), indicating
that material on this side of the fracture had been
reaustenitized during exposure to the hottest part
of the fire, but clearly after the BLEVE had
occurred (when sides A and B were no longer
proximate). A thicker oxidation layer on side B and
Fig. 6 Microstructure of sample 2a, 6.3 mm (¼ in.) from
fracture. Elongation of grains is due to elevated-
temperature deformation.
other evidence, including melted copper wires in the
cab of the truck, suggest the temperature in the hot-
test part of the fire reached the range of >1000 °C.
Examination of the microstructure adjacent to the
fracture on side A showed that a large amount of
localized deformation had occurred in the thinned
regions adjacent to the stable tears (Fig. 6a and b
and Fig. 8). Considerably less localized deformation
had occurred adjacent to the unstable portion of
the fracture where less thinning had occurred (Fig.
9). These observations are consistent with stable
cracking by high-temperature deformation and
eventual stress rupture in the 90° fracture regions
(tank center), followed by sudden unstable crack
extension in the 45° regions near the ends of the
tank. In contrast, the microstructure of side B (Fig.
7) consists of equiaxed ferrite and pearlite, indicating
side B transformed to austenite and then cooled
without quenching. The original plastic deformation
Fig. 7 Microstructure of sample 3b far from tear. The ferrite and
pearlite morphology is different from Fig. 5, indicating
that the steel had been reaustenitized.
(a)
(b) (b)
(a)
38 Journal of Failure Analysis and Prevention
Volume 5(5) October 2005
Metallurgical Failure Analysis of a Propane Tank BLEVE
(continued)
is no longer present within the microstructure of
side B. Evidence of the elevated-temperature plastic
deformation seen in side A was eliminated when
side B was reaustenitized in the hottest part of the
fire after the BLEVE.
Metallographic examination showed the unusual
rough area 117 cm (46 in.) from the end of the
tank to be a small surface weld (Fig. 10). This may
have been a tack weld made as part of the tank
fabrication process and subsequently ground off. This
weld was 1.3 to 2.5 cm (½ to 1 in.) in diameter, and
its fusion zone extended approximately 0.08 cm
(0.03 in.) into the plate. A heat-affected zone
(HAZ) of coarse acicular ferrite and pearlite
surrounded the fusion zone, extending another
approximately 0.08 cm (0.03 in.) into the plate.
Stress-rupture tearing had clearly initiated in this
region at relatively low strains prior to the onset of
stress rupture in surrounding areas of “normal
microstructure.” Premature stress rupture in both
the weld and the underlying HAZ occurred due to
the unusually large grains and the continuous
transverse grain boundaries in these regions. This
resulted in the formation of wide secondary cracks
along the boundaries of the columnar grains in the
fusion zone. These secondary cracks were responsible
for the rough appearance of the plate surface and
fracture surface in this area. The opening of these
cracks indicated that they had formed before the
onset of stress-rupture tearing in the underlying base
metal. An abundance of grain-boundary separations
(cracks in the early stage of formation) were also
seen in the HAZ (Fig. 11). None of these secondary
cracks extended into the base metal. While it is likely
Fig. 8 Microstructure of sample 2a at fracture. Severe elongation
of grains is observed due to elevated-temperature
deformation. Similar elongation was seen at all locations
on side A where substantial thinning had occurred.
Fig. 9 Microstructure of sample 1a at fracture. Note lack of
substantial grain elongation.
Fig. 10 Low-magnification micrograph showing stress-rupture tears in weld bead and underlying heat-affected zone
39
Journal of Failure Analysis and Prevention Volume 5(5) October 2005
that this surface weld is where the first cracking
occurred, it is unlikely that it significantly influ-
enced the time or temperature at which most of the
stable tearing occurred, nor where unstable cracking
occurred, resulting in the BLEVE.
Discussion of BLEVE Phenomenon
The BLEVE phenomenon has been studied by a
number of investigators.[1-13] It is generally agreed
that failure occurs by a combination of elevated-
temperature deformation and stress-rupture tearing
followed by unstable fracture, as observed in this
failure. While previous investigators have made
correlations between several dependent variables
(such as stress-rupture length) and independent
variables (such as tank fill level, latent energy in the
liquid and gas phases, etc.), it is not yet known what
determines whether a BLEVE will or will not occur.
A key issue appears to be an understanding of what
defines the onset of unstable fracture.
One possibility is that a stress-rupture tear becomes
long enough that it becomes mechanically unstable.
Similar phenomena occur at room temperature when
the stress-intensity factor (a function of both applied
stress and crack length) at the tip of a slowly growing
fatigue crack reaches a critical value, resulting in
catastrophic unstable crack propagation. However,
structural steel is very soft and deformable in the
temperature range of interest, as evidenced by the
extensive deformation that preceded tearing along
most of the tank’s length, and unstable crack
propagation is unusual under conditions where
plastic deformation is this extensive. The observation
that two stress-rupture tears joined to form a single
longer opening, however, would cause a significant
increase in stress intensity that could precipitate
unstable fracture. Further investigation of this possi-
bility would require extensive application of elastic-
plastic fracture mechanics, which was beyond the
scope of this investigation.
Another possibility is that some sudden change
occurred, either increasing wall stress or making the
material less deformable and more prone to unstable
fracture, thus precipitating the BLEVE. Venart has
proposed that this may occur due to supercritical
boiling of the liquid propane in the tank.[3] Under
conditions of very high heat input, such as the Truth
or Consequences fire, the liquid propane in the tank
can become significantly superheated. If the pressure
in the tank is suddenly reduced (by the opening of
a large stress-rupture tear or the linking of two
cracks, for example), the supercritically heated liquid
can boil homogeneously, causing a two-phase swell.
Venart offers several options for how this swell could
precipitate a BLEVE.
The first scenario is that the top portion of the
tank, which is above the liquid level, becomes heated
to above approximately 750 °C and transforms to
austenite. The subsequent two-phase swell then fills
the tank and cools the steel rapidly enough to
transform the austenite to relatively brittle microcon-
stituents, such as martensite, bainite, or fine pearlite.
These microconstituents exhibit reduced resistance
to fracture, so their formation causes unstable crack
growth and a BLEVE. In support of this proposal,
observations of variations in microstructure and
hardness were made in previous studies.[1,3] This
possibility was assessed by making metallographic
observations along the length of the fracture, with
particular emphasis on the vicinity where unstable
fracture began. No indication was seen of austenite
formation along side A of the fracture (Fig. 5, 6, 8).
Side B had been completely reaustenitized, but this
clearly occurred subsequent to the BLEVE, rather
than being causative.
An additional effort was also made to determine
the temperature along the top of the tank prior to
the BLEVE, and how close the steel was to being
austenitized. This was done by estimating the temp-
erature at which stress-rupture failure would occur
in the steel, given a value of hoop stress defined by
the operating pressure of the pressure relief valves.
Fig. 11 Microstructure of heat-affected zone with stress-rupture
tears developing along transverse grain boundaries
40 Journal of Failure Analysis and Prevention
Volume 5(5) October 2005
Metallurgical Failure Analysis of a Propane Tank BLEVE
(continued)
The hoop stress in a 11 mm (7/16 in.) thick 185 cm
(73 in.) diameter cylinder pressurized to 1.7 MPa
(250 psi) should be approximately 145 MPa (21
ksi). Unfortunately, very little mechanical property
data are available for steels between 550 and 750
°C, because this is above the normal application
range and below the normal metalworking range.
Oren reported the high-strain-rate tensile strengths
of a number of steels in this range, as well as the
effect of strain rate on tensile strength.[14] Analysis
of Oren’s data showed that the ratio of elevated-
temperature strength to room-temperature strength
varied quite reproducibly with temperature for a
number of carbon and alloy steels, including 1524
steel, with composition similar to A202B (Fig. 12).
Adjusting this relationship for the strain rates used
in normal tensile testing resulted in a plot of
elevated-temperature tensile strength versus
temperature (Fig. 13, top curve). Anderson and
Norris also investigated the short-time stress-rupture
behavior of ASTM 612 steel, also similar in
composition to A202B (Fig. 14).[1] Combining
these resulted in estimates of the stresses required
for stress-rupture failure of A202B in 0.2, 0.5, and
1 h. These data are shown in the lower curves of
Fig. 13. Recalling that the room-temperature tensile
strength of A202B is approximately 690 MPa (100
ksi) and that the hoop stress in the Truth or
Consequences tank was approximately 145 MPa
(21 ksi), it can be seen that stress rupture in the 0.2
to 0.5 h time frame would be expected to occur at
approximately 660 °C. This is approximately 60
°C lower than the temperature at which
austenitization begins, and at least 160 °C below
the temperature at which it is complete. While Fig.
13 is only an approximation, and while the geo-
metries and pressure relief valve operating pressures
may result in lower hoop stresses in other tanks, Fig.
13 suggests that it is doubtful that austenitization
could occur prior to BLEVE in either the Truth or
Consequences tank or other tanks. Hence, the pro-
posal regarding austenitization and transformation
to relatively brittle microconstituents does not seem
very viable. In addition, tests on propane tanks
instrumented with thermocouples have shown that
the general upper limit for temperatures at which
BLEVEs occur is on the order of 700 °C.[10] The
observations by Birk et al. and the results in the
present study indicate that BLEVEs occur with
normal stress-rupture processes due simply to loss
of material strength. The onset of BLEVEs is not
Fig. 12 Ultimate tensile strengths (UTS) of various steels at high
temperatures (high-strain-rate UTS at temperature as
percentage of high-strain-rate UTS at room temper-
ature). (Adapted from Ref 14)
Fig. 13 Normal strain rate ultimate tensile strength (UTS) and
stress-rupture strengths at various temperatures (as
percentage of normal strain rate UTS at room temper-
ature). (Data from Ref 1 and 14)
Fig. 14 Stress-rupture strengths of A612 steel at various
temperatures (as percentages of normal strain rate
ultimate tensile strength, or UTS, at temperature).
(Based on data from Ref 1)
41
Journal of Failure Analysis and Prevention Volume 5(5) October 2005
necessarily associated with phase transformations in
the steel of the propane tanks. The microstructural
and hardness variations cited in previous studies[1,3]
are suspected to be similar to those observed in the
vicinity of the surface weld in the Truth or Conse-
quences tank, that is, possibly a shallow weld. Weld
filler metal is also typically somewhat harder than
the base metal, which could explain the increased
hardness measured in those studies.
A second proposal is that the momentum of the
growing swell could cause a transient stress in the
tank when the swell impacted the tank interior.[3]
This may be possible if formation of the stress-
rupture tears was a very sudden event, resulting in
dramatic depressurization of the tank. However,
stress-rupture tears typically occur gradually and
then grow progressively larger. This is evidenced by
the presence of multiple stress-rupture tears in the
Truth or Consequences tank, as well as the growing
together of tears F and H. The gradual development
and extension of the stress-rupture tears makes this
proposal seem unlikely.
A third proposal[3] is that the swell substantially
cools the hot metal on contact, and that the material’s
reduced deformation capacity and fracture resistance
at the lower temperature precipitates the BLEVE.
It is similar to the phase transformation proposal,
but with the property changes based only on changes
in temperature rather than phase transformations.
Figure 13 shows that the ability of the material to
deform would decrease substantially if the temper-
ature decreased from 660 to 560 °C. Oren’s data
suggest that relatively small changes in properties
would be expected between 460 °C and room temp-
erature. These property data suggest that this pro-
posal may have merit. However, data collected so far
in tests on instrumented tanks[11-13] have not
indicated temperature drops in the tank walls
immediately prior to BLEVEs. In addition,
observations of some test BLEVEs have shown that
two-phase propane swells can occur after the tank
splits open. Therefore, while two-phase swelling of
propane may precipitate BLEVEs in some cases, it
does not appear to be a requirement. In summary,
it is likely that the BLEVE phenomenon depends
on a complex interaction of variables including
heating rate and tank fill level, as well as the high-
temperature properties of the tank steels.[11-13]
Moreover, evidence suggests that there may be several
different types of BLEVEs, each with its own set
of conditions for occurrence.[11-13]
Suggestions for Reducing the
Possibility of BLEVEs in Propane Tanks
Developing a sound understanding of what
precipitates unstable crack growth would be an
important step toward reducing the possibility of
BLEVEs. However, in the interim, substantial
improvements may be realized simply by limiting
the lengths of the stress-rupture tears. This may be
accomplished by placing thicker circumferential
hoops at intervals along the length of the tank.
Limiting the lengths of the stable tears should be a
positive step if any of the previously discussed
hypotheses is correct. Shorter tears would presu-
mably open far less than long tears, thus limiting
the rate at which supercritical boiling and swell
formation would occur. Because crack driving force
is also strongly dependent on crack length, greater
pressures would be required for the onset of unstable
fracture in the case of short tears.
Interestingly, Birk observed that stress-rupture
crack arrestors of a different geometry were very
effective in avoiding BLEVEs under conditions
where they would otherwise have been expected to
occur.[2] This may be a useful approach to pursue.
Other approaches may include using thicker steel
to ensure that stress rupture will not occur in the
temperature-time envelopes encountered in typical
fires, and insulating the steel tank so that less heating
occurs during exposure to fire.
Conclusions
The Truth or Consequences fire occurred when
the stationary storage tank’s liquid-phase excess
flow valve tore out of the tank, rather than breaking
off below the tank and valve. This tear-out likely
occurred due to insufficient thread engagement.
Failure of the truck-mounted tank was initiated
by elevated-temperature deformation leading to
stress-rupture formation of several stable tears.
The BLEVE occurred when these tears became
unstable and propagated as a 45° shear failure.
Premature stress-rupture tearing occurred in a
surface weld and its underlying HAZ, but this
was superficial and did not contribute significantly
to the overall BLEVE event.
42 Journal of Failure Analysis and Prevention
Volume 5(5) October 2005
Metallurgical Failure Analysis of a Propane Tank BLEVE
(continued)
• Analysis of elevated-temperature mechanical
property data suggests that stress-rupture tearing
occurred in the vicinity of 660 °C. This analysis
and data from the literature[11-13] suggest an upper
range on BLEVE temperature of approximately
700 °C. Further analysis of temperature data from
instrumented tests would be helpful in determi-
nation of tank-wall temperature during BLEVEs.
The microstructural study showed that the tank
steel was NOT austenitized prior to the BLEVE,
nor was any evidence found of less ductile trans-
formation products formed due to propane
quenching of fire-induced austenite.
Unstable crack growth may have begun when two
stress-rupture tears grew together, suddenly
increasing crack length and crack driving force.
Alternatively, unstable crack growth may have
begun when supercritical boiling caused a two-
phase swell to fill the tank, thus cooling the
approximately 660 °C steel along the top of the
tank and decreasing the deformability and fracture
resistance of the material ahead of the stress-
rupture tears.
Acknowledgments
The authors would like to thank J.E.S. Venart
(University of New Brunswick) and Ned Keltner
(KTechnology Corporation) for helpful discussions
during this work. Thanks also to A.M. Birk (Queens
University, Canada) and J.A. Van Den Avyle (Sandia
National Laboratories) for careful review of the
manuscript. This project was completed under the
New Mexico Small Business Assistance Program at
Sandia National Laboratories. Sandia is a multi-
program laboratory operated by Sandia Corporation,
a Lockheed Martin Company, for the United States
Department of Energy under Contract No. DE-
AC04-94AL85000.
References
1. C. Anderson and E.B. Norris: “Fragmentation and Metal-
lurgical Analysis of Tank Car RAX 201,” FRA-OR&D 75-
30, Ballistic Research Labs, Aberdeen, MD, Aug 1974.
2. D.J. Kielec and A.M. Birk: “Analysis of Fire-Induced Rup-
tures of 400-L Propane Tanks,” J. Pressure Vessel Technol.
(Trans. ASME), Aug 1997, 119, p. 365.
3. J.E.S. Venart: Boiling Liquid Expanding Vapor Explosions
(BLEVEs): A Re-examination of the Causes and Consequences.
[author: please provide the name of the publisher as well as
the year of publication.]
4. K. Moodie, L.T. Cowley, R.B. Denny, L.M. Small, and I.
Williams: “Fire Engulfment Tests of a 5 Tonne LPG Tank,”
J. Hazard. Mater., 1988, 20, pp. 55-71.
5. A.M. Birk: “Thermal Protection of Pressure Vessels by
Internal Wall Cooling During Pressure Relief,” J. Pressure
Vessel Technol. (Trans. ASME), 1990, 112, pp. 427-31.
6. K. Sumathipala, J.E.S. Venart, and F.R. Steward: “Two-
Phase Swelling and Entrainment during Pressure Relief Valve
Discharges,” J. Hazard. Mater., 1990, 25, pp. 219-36.
7. C.M. Pietersen: “Analysis of the LPG Disaster in Mexico
City,” J. Hazard. Mater., 1988, 20, pp. 85-107.
8. H.S. Pearson and R.G. Dooman: Fracture Analysis of Propane
Tank Explosion, Case Histories Involving Fatigue and Fracture
Mechanics, STP 918, C.M. Hudson and T.P. Rich, ed.,
American Society for Testing and Materials, Philadelphia,
PA, 1986, pp. 65-77.
9. J.M. Herrera and S.W. Stafford: “A Failure Analysis Case
Study on a Ruptured Propane Tank,” Mater. Des., 1987,
8(5), pp. 278-83.
10. A.M. Birk, Queens University, Canada, private communi-
cation. [author: please provide the year of the communi-
cation.]
11. A.M. Birk and M.H. Cunningham: A Medium-Scale
Experimental Study of the Boiling Liquid Expanding Vapor
Explosion (BLEVE),” TP 11995E, Transport Canada, 1994.
12. A.M. Birk and M.H. Cunningham: “Liquid Temperature
Stratification and Its Effect on BLEVEs and Their Hazards,
J. Hazard. Mater., 1996, 48, pp. 219-37.
13. A.M. Birk et al.: “Fire Tests of Propane Tanks to Study
BLEVEs and Other Thermal Ruptures: Detailed Analysis
of Medium-Scale Test Results,” Transport Canada, 1995.
14. C. Oren: “Prediction of Ductilities and Press Loads of Steel
at Warm Working Temperatures,” Metal Working and Steel
Processing XIV, American Institute of Mining, Metallurgical
and Petroleum Engineers, 1976.
... In fact, also the direction, the transient behaviour, the environment and the safety measures play a determinant role on the chances to damage other units. Moreover, the target may undergo many different operating conditions during its operative life, thus different responses to the same accident scenario may occur, due to process variables (pressure, temperature, liquid fraction, etc…) and due the ageing/corrosion of construction material that cause a reduction of resistance with the time [148]. ...
... Target resistance depends on target structure and represents the capability to withstand a given physical effect. It depends on the equipment size, on the mechanical properties of the construction material [148], on the operating conditions and on the presence of protections as passive fireproofing materials used for fire protection [38,121]. The resistance of target equipment items may be obtained experimentally or simulating the effect of fire, explosion or missile impact on the target equipment by simplified or finite element models [78]. ...
... Thus the probability of damage and the consequences of secondary accidents change as well [116,133]. In particular, in the case of pressurized vessels the fragmentation pattern could be totally different, changing the damage mode [83,148]. The future work on vulnerability models will benefit if different damage mechanisms will be accounted, with different models addressing the different types of failure. ...
Article
High-impact low-probability (HILP) accident scenarios in industrial sites are raising a growing concern. Domino effect was responsible of several catastrophic accidents that affected the chemical and process industry, as well as critical infrastructures for energy as oil refineries. However, there is still a poor agreement on assessment procedures to address escalation hazard resulting in domino scenarios. The present study presents a review of the work done in the last 30 years in the field, and a critical analysis of available tools and knowledge gaps concerning domino effect assessment. The analysis of scientific publications concerning domino effect in the process industry resulted in a database of more than 60 documents, addressing three main issues: past accident analysis, models for equipment damage, risk assessment and safety management of domino scenarios. The methods, models and tools developed make now possible the quantitative assessment of domino scenarios in risk analysis and in safety management of industrial sites. Nevertheless, a number of open points still remain, where existing tools may be improved and uncertainty may be reduced.
Article
The problem of assessing damage due to explosions of cylindrical pressure vessels is considered. The attention is focussed on a prediction of the arrangement of cracks in the vessel wall prior to its explosion. This arrangement of cracks is called the failure pattern. It is seen as essential information for forecasting ejection and projection of fragments generated by an explosion. Thermally induced explosions known as boiling liquid expanding vapour explosions are studied. The problem of prediction of failure patterns is formulated as a problem of estimating probabilities of these patterns. The scarcity of data on occurrences of failure patterns in the past explosion accidents was an incentive to estimate the failure pattern probabilities by means of Bayesian statistics. The main finding of the study is that the failure pattern probabilities can be handled within the Dirichlet-multinomial model and the epistemic uncertainty in these probabilities expressed by Dirichlet prior and posterior distributions. The Bayesian estimation of failure pattern probabilities is viewed as a way allowing to introduce the prediction of vessel fragmentation into the formal probabilistic risk analysis. The so-called minimally informative Dirichlet prior distribution is suggested for the probability estimation as a prior suitable to Bayesian updating with scarce data. It is stated that currently the probabilistic prediction of failure patterns on the basis of past accident data is the only practicable way to assess the potential type of vessel fragmentation. A conventional (deterministic) mechanical and/or metallurgical analysis does not provide reliable models for failure pattern prediction in case of explosions under study.
Chapter
The potential hazard coming from accidents involving domino effects was recognized since the early times of process safety. Awareness about the importance of prevention and mitigation of domino effects has considerably increased in the last decades, determining a significant production of scientific literature in this field. This chapter presents an overview of the progress on domino effect assessment, providing a brief description of the advancements achieved by each contribution in various fields of domino effect assessment. First, studies aimed at the development of methodologies for quantitative risk assessment and safety management of domino scenarios are analyzed. Then a focus on vulnerability models for equipment damage is presented. Finally, the main research needs and the possible directions of future studies on domino effect are highlighted.
Article
This paper presents detailed data on the thermal response of two 500 gal ASME code propane tanks that were 25% engulfed in a hydrocarbon fire. These tests were done as part of an overall test programme to study thermal protection systems for propane-filled railway tank-cars.The fire was generated using an array of 25 liquid propane-fuelled burners. This provided a luminous fire that engulfed 25% of the tank surface on one side. The intent of these tests was to model a severe partially engulfing fire situation.The paper presents data on the tank wall and lading temperatures and tank internal pressure. In the first test the wind reduced the fire heating and resulted in a late failure of the tank at 46 min. This tank failed catastrophically with a powerful boiling liquid expanding vapour explosion (BLEVE). In the other test, the fire heating was very severe and steady and this tank failed very quickly in 8 min as a finite rupture with massive two-phase jet release. The reasons for these different outcomes are discussed. The different failures provide a range of realistic outcomes for the subject tank and fire condition.
Article
The detailed re-analysis of the catastrophic failures of four 4.5 - tonne water capacity LPG vessels of various fills subjected to jet fire attack indicates that the severity of the event and the intensity of the fireballs formed may be a function of the initiating mode of vessel failure and the thermo-hydraulic state of the contents. The mechanism of vessel failure appears to be a two-step process; the formation of an initiating overpressure crack in the high temperature vapour wetted walls of the vessel, followed by a final catastrophic `unzipping' of the containment and the nearly instantaneous release of its contents. The distribution and flashing of the lading causes a fireball. The surface emissive power of the BLEVE fireball does not appear to be directly related to the `superheat' of the contents at failure. Possible reasons for the final rapid failure of the vessel are either crack instability, rapid overpressurization due to the dynamic `head space' impact of the two-phase swell, and/or the rapid quenching of the crack tip, due to its two phase discharge, that results in large local thermal stresses.
Article
When pressure vessels are exposed to external fire impingement, high wall temperatures can result and these can lead to material degradation and the ultimate failure of the vessel. To protect against this possibility, vessels can be protected by means of pressure relief devices, external thermal barriers or external water spray cooling. This paper deals with a device that cools the walls of fire-impinged vessels carrying pressurized liquids by directing 2-phase fluid along the upper internal surface of the vessel when the vessel pressure relief value is in action. The device consists of a concentric secondary internal shell that partitions the interior into a core region and an anulus. The bottom of the internal shell is open to allow communication between the two regions. When vapor is vented from the annulus, it results in significant fluid swelling in the annular space. This swelling results in large areas of the wall being wetted and cooled by liquid. Experimental results are presented for the case of a short electrically heated cylindrical vessel with and without the cooling device installed. From the limited tests conducted, it was shown that the device cools areas of the vessel wall that would normally have experienced high wall temperatures and possible material degradation.
Article
A failure analysis investigation was conducted on a propane tank formerly contained in a motor vehicle. It had been reported that the subject tank had exploded causing loss of the vehicle and extensive personal property damage. The findings of this investigation reveal that the tank had been subjected to extremes in temperature prior to actual rupture. The thermal conditions external to the tank apparently induced propane tank leakage through the safety valve and ultimately overpressurized the tank inducing complete rupture of the longitudinal tank wall. The cause for the elevated temperature state was not determined. However, by the investigative techniques employed, it was confirmed that the tank rupture did not initiate the fire damage sustained by the vehicle and the other equipment.
Article
A series of fire tests were conducted to study the thermal rupture of propane tanks. The tests involved 400-L ASME automotive propane tanks filled to 80% capacity with commercial propane. The tanks were brought to failure using torches and pool fires. The resulting thermal ruptures varied in severity from minor fissures, measuring a few centimeters in length, to catastrophic failures where the tank was flattened on the ground. The catastrophic failures would typically be called boiling liquid expanding vapor explosions (BLEVEs). The objective of this work was to develop a correlation between the failure severity and the tank condition at failure. The deformed propane tanks were measured in detail and the extent of deformation was quantified. The tank failure severity was found to be a complex function of a number of tank and lading properties at failure. This paper presents the measured data from the tanks and a step-by-step description of how the correlation was determined.
Article
In November, 1984, an enormous disaster involving an LPG installation occurred in Mexico City and resulted in the deaths of over 500 people. A TNO team went to Mexico shortly afterwards to conduct an investigation. This article reflects on their findings and draws some preliminary conclusions.
Article
The influence of liquid space expansion, caused by thermal effects and boiling, on pressure relief valve (PRV) behavior is examined. Small scale (40 litres) and moderate scale (10 000 litres) experimental behavior is shown to behave similarly for a variety of ladings. There is clear evidence of interface entrainment and its resulting influence on pressure relief.
Article
Experiments have been conducted on the behaviour of a 5 tonne horizontal cylindrical LPG tank engulfed in kerosene pool fires. Five tests were carried out with commercial propane fill levels from 22% to 72%. Fire durations were up to thirty minutes. Extensive measurements were made of the fire characteristics, external and internal tank metal temperatures and the wall heat fluxes. Bulk and boundary layer fluid temperatures were measured at 26 points inside the tank, which together with internal pressure histories allowed characterisation of the internal fluid behaviour. Standard pressure relief valves were fitted and their operation monitored - they operated reliably and controlled the tank pressure throughout the fire tests. The results further extend and complement those on fire engulfed 0.25 and 1 tonne tanks. The complete set of results provides direct information on the behaviour of LPG tanks of different sizes and fill levels and a sound basis for the development and validation of predictive models.
Article
Recent fire tests involving over forty, 4001 automotive fuel tanks filled with commercial propane have shown that the likelihood of a BLEVE and the severity of its hazards are significantly affected by the detailed thermodynamic condition of the lading at the time of failure.When the liquid in a tank is heated by fire impingement on the tank external shell, the liquid near the heated wall will tend to rise because of buoyancy effects. This leads to the development of temperature stratification where the liquid near the top of the tank will be at a higher temperature than liquid lower down. The pressure in the tank is dictated by the warmest liquid. This means that when the liquid is stratified the pressure in the tank is higher than the pressure one would calculate from the average liquid temperature. If the tank fails at the pressure relief valve (PRV) set pressure the resulting release will be less powerful if the liquid is stratified.When the PRV is activated on a tank it usually vents vapour to the surroundings. This vapor flow results in boiling action in the liquid and this boiling causes heat transfer and mixing and these cause destratification of the liquid. The time for destratification increases with the scale of the system. Eventually, the PRV may eliminate the stratification and the liquid will consist of a near isothermal liquid mass. If the tank fails when it is full of liquid, and the liquid is uniformly at the saturation temperature for the PRV set pressure, then the resulting BLEVE and hazards will be maximized for the given tank.Based on recent fire test data, this paper discusses how temperature stratification and destratification is affected by the fire type and the PRV action. The paper also discusses how this temperature stratification effects the likelihood of a BLEVE, and the severity of the associated hazards including fireball heat flux, blast overpressure and projectiles.
Fragmentation and Metallurgical Analysis of Tank Car RAX 201
  • C Anderson
  • E B Norris
C. Anderson and E.B. Norris: "Fragmentation and Metallurgical Analysis of Tank Car RAX 201," FRA-OR&D 75-30, Ballistic Research Labs, Aberdeen, MD, Aug 1974.
Fracture Analysis of Propane Tank Explosion, Case Histories Involving Fatigue and Fracture Mechanics
  • H S Pearson
  • R G Dooman
  • H.S. Pearson
H.S. Pearson and R.G. Dooman: Fracture Analysis of Propane Tank Explosion, Case Histories Involving Fatigue and Fracture Mechanics, STP 918, C.M. Hudson and T.P. Rich, ed., American Society for Testing and Materials, Philadelphia, PA, 1986, pp. 65-77.