To BLEVE or Not To BLEVE:

Anatomy of a Boiling Liquid Expanding

Vapor Explosion
J. E. S. Venart, G . A. Rutledge, K. Sumathipala, K. Sollows
Fire Science Centre, University of New Brunswick, P.O. Box 4400, Fredericton,
N.B. Canada E3B 5A3
-___
The loss of containment ( L O C )for Pressure Liquefied Gas ( P L G ) vessels under
accidental fire engulfment is shown to be very complex. The LOC depends upon: ( i ) the
extent and intensity of external heating, ( i i ) the pressure relief device ( P R D ) operation
and flare (ifcontentsflammable), (iii) the fluid and fill level, ( i v ) the construction of
the vessel, and ( u ) the thermohydraulic history of the commodity prior to failure.
The simple experiments described here shows that there exists a new type of more
powerful failure than the BLEVE. This even we call a BLCBE, a Boiling Liquid
Compressed Bubble Explosion. A hypothesis is advanced to explain this mode of failure
which is supported by an initial series of small scale experiments involving Argon, water,
R I I , and R123.
A comprehensive test program to determine the details of the BLCBE and BLEVE
failure
_ _ - -
modes is indicated along with work to determine methods of protection.
_ _ _ _ ~

Introduction and Background This type of failure has sometimes been called a BLEVE
and invariably occurs below some critical fill level when there
There are three basic types of vessel failure and consequences is sufficient energy in the vapor to cause the crack to propagate
for pressure liquefied vessels under fire attack. All types of plastically to cold metal at or below the liquid level. The event
events are initiated by the thermal stratification of the liquid leads to, if the contents are flammable, massive fireballs and
and vapor spaces under sustained heating and the increase in residual pool fires. It is not the most violent type of event
vapor wall metal temperatures [I-5].The pressure relief device which can however, occur as little or no blast effect results
(PRD) will activate when the liquid surface temperature and [III.
thus its saturation pressure exceeds the P R D set point. The More dangerous events can occur, with higher fill levels. In
type of tank geometry influences, for similar fire exposure and these circumstances the creep rupture crack is arrested in the
volume, the time to first vent [a. The valve may cycle or hot vapor space metal by plastic tip blunting [IOb],since there
is insufficient available work in the vapor. Then, dependent
ultimately remain open with contents existing as appropriate
to the pressure and internal flow conditions [4]. The ullage upon the thermohydraulic history of the liquid, either a massive
volume wall temperatures lead ultimately to creep rupture in and prolonged two phase jet discharged will occur, or a cat-
this space with the size of ihe crack formed plastically depen- astrophic and very powerful vessel failure with complete loss
dent upon the available energy in the vapor space alone [ A . of containment by a homogeneously nucleated highly com-
The crack will usually form in the hot metal at a position where pressed two-phase mixture results.
the PRD flare exposes the metal surface to higher thermal The first type of failure will result with the contents stratified
loads [S, 91. and pressurizing [4] and involves therefore a distribution of
Dependent upon fill level, the crack size may be limited to non nucleated superheated liquid upon depressurization. The
the hot metal regions of the vapor space for large fills or, if event commences with the discharge of superheated vapor
over a critical crack size, to regions at or below the liquid level through the narrow creep rupture formed crack, followed,
for lower level fills. In this latter case, vessel containment can some short time after with a heterogeneous wall nucleated,
be totally lost since crack(s) now propagate elastically without high void-fraction, two-phase discharge of long duration for
restraint and small energy requirements [IOU]. An adiabatic the scale of experiments described here. A sustained jet or
expansion of the contents next occurs within limits bounded torch fire issuing from the crack may thus follow if the contents
by either an isenthalpic or an isentropic flash evaporation. For are flammable.
propane under normal storage conditions this results in a 40 The last type of failure occurs with the contents usually
to 50% flash vapor fraction with the evaporating residue dis- homogeneous in temperature and depressurizing. It may also
tributed as large droplets or left in the form of a pool [ I I ] . be initiated by a sudden release of pressure caused by PRD
action and size, the carryout over of bubble nuclei into the
This paper won the Bill Doyle award for the best paper vessel due to filling operations, or even a “leak before break”
presentedat the AlChE Loss PreventionSymposium held somewhere else in the vessel or its associated plumbing. Like
in New Orleans in April 1992. the former failure, a crack is formed in the vapor space and

Process Safety Progress (Vol. 12, No. 2) April, 1993 67

 ............................................ ....................
........

........... ..................... ................... ..............

01 01
0 100 200 300 400 500 0 04 08 12 16
TIME (US) TIME (mS)

FIGURE 1. Graph of dynamic pressure trace versus FIGURE 2. Graph of dynamic and static pressure traces
time. versus time.

the superheated vapor first discharges. Since the liquid contents signals obtained from an Argon filled with R11 filled test are
now are, however, already nucleated, (due to the falling pres- shown in Figure 1 and 2 respectively.
sure prior to this crack fermation), there is an almost instan- The argon tests were pressurized to 100 psig and the cracks
taneous generation of void causing a rapid two-phase swell. initiated in the horizontal vessel using the propane torch di-
This low void-fraction swelled material chokes in the crack rected on the top central region of the can. The size and in-
and results in an extremely rapid repressurization of the con- tensity of the torch influences the heated region and size of
tents back approximately to the original containment pressure. creep rupture crack formed. For all the results shown here the
Inertial collapse of the vapor bubbles generated by the liquid torch setting was maintained constant.
depressurization wave thus occurs due to system constraint. In the Argon tests the tank is plastically destroyed over a
The liquid shock pressures (1-3 kbar), developed by bubble millisecond time frame (Figure 1) as the available work in the
collapse, result in the rapid (-speed of sound in the metal gas was sufficient to permit crack extension by tearing to com-
[IOc])elastic crack propagation of the now cooled vapor walls. plete failure [17, 181. The failure surfaces as examined by SEM
The highly compressed two-phase material is now suddenly indicates clearly the plastic deformation resulting.
depressurized again. The expansion of the high pressure vapor In the R11 test shown in Figure 2, the original liquid fill
bubbles rapidly shatters and projects the superheated liquid was set at 70% by volume. The can was pressurized to 110
into a fine highly turbulent aerosol. This situation results in psig (Tsat -97°C) using the external heating. The vessel was
the violent, rapid and uniform dispersal of the total contents, then vented in the vapor space by a single short/quick vent
which may, when flammable, under proper conditions, ac- down to 105 psig and a crack ( - 1 cm long) subsequently
celerate to a detonation if ignited. As the temperature of the initiated in the vapor space wall by the propane torch. The
liquid contents approaches the superheat limit [12] greater tank was catastrophically and rapidly (- 200 ps) destroyed in
numbers of smaller sized bubbles are formed more rapidly, what was a multi-step manner resulting in the pieces shown in
resulting in greater liquid shock pressures upon collapse and Figure 3. The contents were violently dispersed as a fine and
thus significantly more powerful events. This type of dual step rapidly evaporating aerosol. Peak impulsive sound level in-
failure we call a “Boiling Liquid Compressed Bubble Explo- tensity at 2.5 m measured end on to the can registered 132 dB
sion, or “BLCBE”. for R123. For R11 the superheat limit TSLis 422 K (149°C)
It is our view that, based upon the evidence from the incident [12] and thus this test is >50°C below TsL.
at Mexico City in 1984, all three types of events were expe- The dynamic pressure transducer, located below the liquid
rienced by the spherical LPG storage tanks with tragic con- surface at the center of the instrumentation boss indicates that
sequences [15]. The potential for similar circumstances can -
after a preliminary stage of depressurization ( 200 IS), caused
also exist in other processing, storage and transportation pres- by the formation and arrest of the initial crack, there is a
sure vessels containing many pressure liquefied fluids. Such a measured over-pressure of 175 kPa ( - 25 psig) greater than
process may have been responsible for the Flixborough incident the original containment. The time rate of increase of this
1.271. latter pressure rise suggest; the imposition of high strain rates
on the container due, perhaps, to liquid shock pressure waves
resulting from bubble collapse [19-21].
Preliminary Experiments

Experimental evidence of the above hypotheses has been

obtained: [16].The preliminary results illustrate the BLCBE,
BLEVE and other failure behavior by cinematography and
measurements of the dynamic pressure developed in partially
filled vessels, failing under torch fire attack. The experiments,
and their high speed (5000 frame/s) photography, permits the
rapid, reproducible evaluation of the influence of all variables
that may affect the problems and their solution.
Four types of tests were conducted. Vessels with fills of
Argon plus Water, Argon only, R11 and R123 were failed at
790 kPa ( - 100 psig). Pressure was maintained either by the
gas or by the vapor pressure obtained using an external heater
impinging on the liquid space. Failures were initiated by 10-
calized vapor space heating using a propane torch. Pressure
‘An accompanying video of four selected tests can be obtained from the
authors.
FIGURE 3. Vessel destruction BLCBE failure (test 5).

68 April, 1993 Process Safety Progress (Vol. 12, No. 2)

 A similar experiment to that shown in Figure 2 was con- volume existing at the time of the initial creep rupture. If the
ducted without, however, the initial vapor vent prior to failure bubbles are presumed to be uniformly distributed throughout
and thus without prior bulk nucleation. In this test the tank the fluid, a liquid volume associated with each bubble can be
remained intact with the contents discharged as a long duration determined for which a Weber number governing its breakup
high void-fraction jet through the 1 cm long initial crack. The upon the LOC can be estimated. Inertial bubble expansion
edges of the opening now become somewhat plastically de- velocities [12]and the volumes of associated liquid yield values
formed so that the width is about 2-3 mm with a slight (1-3 such to suggest an aerosol formed of - 100 pm diameter drop-
mm) extension in length. lets. Such an aerosol will rapidly ( - 40 ms) evaporate and may,
Similar tests to those above were conducted with 70% vol- if flammable, accelerate to a detonation since the aerosol dis-
umetric fills of unheated water with 100 psig argon gas ballasts. persion/evaporation process proceeds at high velocity with an
Similar sizes of failure ( - 1 cm) without, however, the plastic appreciable blast wave and its vapor contents can carry with
edge deformation resulted in a < -0.1 mm wide cracks. them appreciable quantities of free radicals formed upon bub-
In two of the above tests the vessel was destroyed. The second ble collapse. Two forthcoming papers elaborate on these and
event described clearly illustrates a significant pressure rise other experiments as well as a possible theoretical explanation
above the original containment level and sizeable blast effects. for the observations [25, 261.
This can only be described as a BLCBE.
Conclusions
Discussion
Simple “BLEVE” experiments have revealed the existence
The observed modes of failure in the described tests ranged of a loss of containment event we call a Boiling Liquid Com-
from, (i) the formation of arrested cracks in the vapor space, pressed Bubble Explosion, a “BLCBE”. Such an event is more
for the water and non nucleated R11 fills, (ii) the complete powerful than the BLEVE and is at an extremum of failure
tearing apart of the vessel by gas pressure, to, (iii) the very types that range from it to the BLEVE and ultimately down
powerful events with prenucleated partial fills of R11 and to those which result only in the expulsion of their contents
R123. as a flashing two-phase jet.
Up to now the BLEVE phenomenon has been portrayed as All the events and their interrelationships are obviously very
a single step event (11, 12, 15, 22-24] which is initiated by the complex and have many interacting dynamic components. The
LOC and results in the adiabatic flash evaporation and dis- generic features of the BLCBE consist of:
persal of the contents. Obviously from the test data presented (i) partial failure of the containment vessel.
here there is, additionally, the possibility of much more pow- (ii) multiple bubble initiation, and growth in a prenucleated
erful incidents with higher fills. This, the BLCBE, requires bulk liquid.
prior nucleation of the liqud contents, and may occur very (iii) rapid two-phase swell, repressurization and coherent
early on in the accident. This failure is a two stage event leading collapse of the bubbles formed.
to the violent, rapid dispersal of the total contents as a small (iv) bubble collapse shock pressure failure of the previously
sized evaporating aerosol with significant blast over-pressures. damaged vessel.
It is hypothesized that the liquid shock pressures generated (v) violent distribution of the compressed two-phase con-
by vapor bubble collapse art: responsible for the rapid destruc- tents as extremely fine evaporating aerosol with significant
tion of the containment vessel and the dispersal of the contents blast, and
as the aerosol. The evidence for this statement is based upon (vi) if contents are flammable, the potential for a detona-
many small scale fully instrumented trials such as described tion.
above [25]. Obviously, as the contents of such vessels approach the su-
In the events which BLCBE’d the average peak over-pres- perheat limit [12] even more powerful events are possible.
sures recorded, relative to either an argon depressurization The initiation of the bubble nuclei responsible for this type
trace or an envelope developed by the minimum pressures in of failure may either be the PRD operation and/or size, or
the recorded trace, were 490 kPa (70 psig) in a sample size of even a leak. The “leak before break” philosophy widely uti-
10 trials (min 380 ( 5 5 psig), max 590 kPa (85 psig)). On the lized in the process and power industries for loss control may
basis that the pressure wave commences at the centre of the therefore deserve more careful scrutiny.
original 1 cm long crack, the instantaneous over-pressure ex-
erted at the crack tip is 6.2 MPa (62 bar or 900 psig)! Crack
reinitiations at such large rates of strain can result in elastic Recommendations
propagation velocities approaching a significant fraction of
the speed of sound in the metal. The mechanical energy re- There is a requirement for definitive experiments to confirm
quired to recommence crack propagation at such velocities and amplify the sequences of events described here for all types
though in general derives from the thermal energy of the con- of losses of containment. Such experiments should seek to
tents but once started crack propagation requires little further more fully understand the phenomena such to permit models
energy [ ZOb]. to be developed of all aspects of the thermohydraulics leading
A bubble collapse time derived by taking one half the peak to the failure types described. These experiments should be
-
overpressure base pedestal of 80 p s (Figure 2) yields bubble structured to provide a clear demarcation between the classes
sizes for inertial collapse of between 1 to 3 mm diameter [19, of events and the significance of their consequences. This work
201. Bubbles of this size are also generated at atmospheric can best be done in a staged approach that first utilizes small
pressure inertially and thermally, in the 100-200 p s time delay scale experiments t o understand the various elements and their
times observed (121. complex interactions. Next, verified sub-models of the indi-
The formation and observed evaporation of the finely di- vidual mechanisms should be developed and assembled into a
vided violent aerosol from the total contents can also be in- comprehensive computer code to be utilized in predicting the
ferred by the proposed BLCBE mechanism. The high speed outcome of a selected series of larger scale experiments. At
photographic evidence indicates that after the initial vapor this time various proven prevention/mitigation techniques
discharge no further fluid leaves the tank prior to its ultimate should be examined.
failure. If the hypothesis is correct therefore, the tank at this The Part I test program envisaged would provide for (i) a
time must be full of a low void fraction two-phase mixture clarification and quantification of the creep rupture and other
whose bubbles are - 1-3 mm in diameter. The approximate failure mechanisms (material and fire exposure aspects, time
number of bubbles can therefore be estimated from the vapor to failure and size of failure(s)) as a function of tank heating

Process Safety Progress (Vol. 12, No. 2) April, 1993 69

as well as volumetric fill and fluid, (ii) the reproducible de- a Flare to a Horizontal Cylinder-Part 11,” Fire Safety
termination of the initiating event(s) and the subsequent con- J., 1985, 8(2), pp. 89-96.
sequences using small scale vessels and high speed pressure as 10. (a) McClintock and Argon A. S., Mechanical Behavior
well as photographic (4 * lo3- 1 * lo6fps) records. These would of Materials, Addison-Wesley (1966), p. 508.
examine the interrelationships between fluids (flammable and (b) McClintock and Argon, A. S., Mechanical Behavior
nonflammable), original fill, stratification, restriction geom- of Materials, Addison-Wesley (1966), p. 571.
etry, and pre-failure pressure history as well as explosive po- (c) McClintock and Argon A. S., Mechanical Behavior
tential. of Materials, Addison-Wesley (1966), p. 560.
These tests should permit the determination of the exact 11. Maurer, B., Hess, K., Giesbrecht, H., and Leuckel, W.,
sequence of the events and provide a basis for the examination “Modeling of Vapor Cloud Dispersion and Deflagration
of several methods of prevention. The results should, as well, after Bursting of Tanks Filled with Liquefied Gas,” 2nd
permit verified submodel incorporation into the necessary Int. Symp. Loss Prevention and Safety Promotion in the
computer codes that will be of use in the development of Process Industry, Heidelberg, 1977, p. 305.
protection strategies and for first responders. 12. Reid, R. C., “Rapid Phase Transitions from Liquid Va-
The small scale tests should be followed by larger scale fire por, Vol. 12,” Advances in Chem. Eng. Academic Press,
and other failure type trials (Part II), which must be designed pp. 112-203.
with careful control of initial and boundary conditions. These 13. McHale, E. T., Discussion to paper by Krystantas, R.,
should provide confirmation of the small scale results and Lee, J. H., and Moen, I . , “Direct Initiation of Spherical
illustrate predetermined preventive measures. Detonation by a Hot Turbulent Gas Jet,” 17 Sympo (Int)
Combustion. Combustion Institute (1978), p. 1245.
14. Lieber, C. O., “Approximate Quantitative Aspects of a
Acknowledgments Hot Spot,” J. Hazardous Materials, 12 (1985), pp. 43-
64.
15. Pietersen, C. M., “Analysis of the LPG-Disaster in Mexico
Discusssions with colleagues, too numerous to list, have over City,” J. Hazardous Materials, 20, December 1988, pp.
the past eight years aided the senior author’s preparation of
85-107.
this work. Financial support has been provided by Natural
16. Rutledge, G., “POP-CAN BLEVE Apparatus,” BSc.E.
Sciences and Engineering Research Council of Canada, Trans-
Thesis, Mechanical Engineering, UNB, 1991.
port Canada, Environment Canada, and the University of New
17. Baum, M. R., “The Development of the Breach Generated
Brunswick.
by Axial Rupture of a Gas-Pressurized Steel Pipe,” Jour-
nalof Pressure Vessel Technology, 104, ASME, Nov. 1982,
pp. 253-261.
Literature Cited 18. Baum, M. R., “The Relationship Between Breach Devel-
opment and the Transient During Axial Rupture of a Gas-
1. Venart, J. E. S., “In Vessel Transient Thermohydraulics,” Pressurized Steel Pipe,” Journal of Pressure Vessel Tech-
Heat Transfer and Major Technological Hazard Proceed- nology, 104, ASME, Feb. 1982, pp. 20-24.
ings (Vol. 2), Eurotherm Seminar Nr. 14, Universite Cath- 19. Lamb, H., Hydrodynamics, 6th Ed, Dover 1945, pp. 122-
oliquid de Louvain, Louvain-la-Neuve, Belgium, May 15- 123.
18, 1990, pp. 11.1-11.12. 20. Rayleigh, L., “On the Pressure Developed in a Liquid
2. Aydemir, N. U., Magapu, V. K., Sousa, A. C. M., and During the Collapse of a Spherical Cavity,” Phil. Mag.
Venart, J. E. S., “Thermal Response Analysis of LPG XXXZV, 94, (1917) [papers vi 5041.
Tanks Exposed to Fire,” J. Hazardous Materials, 1988, 21. Fujikawa, S., and Akamatsu, T., “Effects of the Non-
20, pp. 239-262. Equlibrium Condensation of Vapor on the Pressure Wave
3. Venart, J. E. S., Sumathipala, K., Steward, F. R., and Produced by the Collapse of a Bubble in a Liquid,” J.
Sousa, A. C. M., “Experiments on the Thermo-Hydraulic Fluid Mechanics, 97, (1980), pp. 481-514.
Response of Pressure Liquefied Gases in Externally Heated 22. McDevitt, C . A., “The State of the Art Concerning
Tanks with Pressure Relief,” AIChE Plant/Operations BLEVES (Boiling Liquid Expanding Vapor Explosions),”
Progress, 1988, pp. 139-147. Report Prepared for Transport Canada, TP1090E, 1991.
4. Sumathipala, K., Venart, J. E. S., and Steward, F. R., 23. Johnson, D. M., Pritchard, M. J., and Wickens, M. J.,
“Two-phase Swelling and Entrainment During Pressure Large Scale Catastrophic Releases of Flammable Liquids.
Relief Valve Discharges,” J. Hazardous Materials, 25, British Gas/CEC Contract No. [Link](H), Au-
October 1990, pp. 219-236. gust 1990.
5. Aydemir, U. N., Sumathipala, K., Hadjisophocleous, G. 24. Johnson, D. M., and Pritchard, M. J., Large Scale Ex-
V., Dutcher, C. R., Sousa, A. C. M., Steward, F. R., and perimental Study of Boiling Liquid Expanding Vapor Ex-
Venart, J. E. S., “Pressure Relief of Accidentally-Heated plosions (BLEVES). British Gas Report: MRS 61 1, January
pressure Liquefied Gas Vessels: Two-Dimensional 1991.
Models,” Heat Transfer and Major Technological Haz- 25. Venart, J. E. S., et al., “The Boiling Liquid Expanding
ards Proceedings (Vol. 2), Eurotherm Seminar Nr. 14, Vapor Explosion: Part I: Experiments,” Manuscript in
Universite Catholique de Louvain, Louvain-la-Neuve, Bel- preparation.
gium, May 15-17, 1990, pp. 12.1-12.13. 26. Venart, J. E. S., et al., “The Boiling Liquid Expanding
6. Yu, C. M., Dutcher, C. R., and Venart, J. E. S., Work Vapor Explosion: Part 11: Theory,” Manuscript in prep-
in Progress, University of New Brunswick, Fire Science aration.
Centre, 1991-92. 27. Venart, J. E. S., et al., “Flixborough: Twenty Years Later;
7. Sumathipala, U. K., Sollows, K. F., Deschenes, P. C., A Reprisal: Part 11: Theory,” Manuscript in preparation.
and Venart, J. E. S., “Critical Fracture Length in a Fire
Heated Pressure-Liquefied Gas Tank,” Environment J. E. S. Venart received the Bill Doyle award for the
Canada Technical Seminar on Chemical Spills, Vancou-
ver, June 10-11, 1991. above paper which was presented at the 26th An-
8. Tunc, M., and Venart, J. E. S., “Incident Radiation from nual Loss Prevention Symposium in New Orleans,
an Engulfing Pool Fire to a Horizontal Cylinder-Part April 1,1992. The presentation was made on March
I,” Fire Safety J., 1985, 8(2), pp. 81-88. 30,1993 at the 27th Annual Loss Prevention Sym-
9. Tunc, M., and Venart, J. E. S., “Incident Radiation from posium in Houston.

70 April, 1993 Process Safety Progress (Vol. 12, No. 2)