Anda di halaman 1dari 14

Nuclear Engineering and Design 239 (2009) 20562069

Contents lists available at ScienceDirect


Nuclear Engineering and Design
j our nal homepage: www. el sevi er . com/ l ocat e/ nucengdes
FDS simulation of the fuel reball from a hypothetical commercial airliner
crash on a generic nuclear power plant
Wolfgang Luther, W. Christoph Mller

GRS, Forschungsinstitute, Garching, Germany


a r t i c l e i n f o
Article history:
Received 24 June 2008
Received in revised form 9 April 2009
Accepted 13 April 2009
a b s t r a c t
In the aftermath of 9/11 events it became clear that the impact of a fast ying commercial airliner hitting
the NPP could no longer be excluded as a potential external hazard threatening the nuclear power plant
(NPP) safety. One of the potential consequences of the impact is the occurrence of a reball, large enough
to engulf the entire NPP. The knowledge about reballs from air crashes is rather poor since it is only
based on footage shot by chance. From careful physical and chemical examinations using rst principles,
it can be concluded that the physics and chemistry of the kerosene reball are similar to BLEVE reballs in
gas tank accidents which have been studied during the last decades. The knowledge from these analyses
can be applied to air crash reball analysis.
In order to obtain an adequate understanding of the potential hazards to a NPP engulfed by a reball
a detailed analysis of the reball is necessary. It is only by a detailed analysis that the effect of the NPP
structures on the evolution of the reball can be derived. Though the safety-relevant parts of the NPP are
strong concrete structures, according to IAEA regulations the hypothesized entry of combustion products
into ventilation or air supply systems and the entry of fuel into buildings through normal openings have
to be analyzed in detail. This requires local transient values of the safety-relevant reball parameters.
With the NPP being a very large structure an adequately detailed simulation requires large computing
grids and substantial computing power.
With the release of Version 5 of the Fire Dynamic Simulator (FDS) from NIST in 2007 a simulation tool
is nowavailable which is capable to performsimulations of large reballs on sufciently large computing
grids. These reball simulations can be performed also by any other CFD code in which the relevant
models have been implemented.
The FDS reball simulation capabilities were validated with the help of a well-documented reball
event, inwhich5.9 to of propane were burnt during a BLEVE impact experiment conductedby the German
BAM in 1999.
To demonstrate the applicability of FDS to nuclear safety analysis a simulation of the impact of a 90
to reball on a generic NPP was performed. The results are presented in this paper and show that FDS
release Version 5 is an adequate tool to analysis the effect of a reball on a NPP, even if the largest possible
amount of kerosene involved in the crash is assumed.
The work presentedinthis paper is basedoncodes, papers, footage andmaterial that are freely available
on the Internet. The paper does not use any information that is not freely available on the Internet.
2009 Elsevier B.V. All rights reserved.
1. Introduction
Before 11 September 2001 nuclear installations were designed
to withstand the impact of a typical aircraft ghter of that time.
This impact constituted a design basis accident and the applicant
for the licensing of a nuclear power plant (NPP) had to demon-
strate that the impact does not lead to unacceptable radioactive
releases.

Corresponding author. Tel.: +49 8932004426; fax: +49 8932004599.


E-mail address: mur@grs.de (W.C. Mller).
Onthe morning of 11September 2001, eachof the twintowers of
the World Trade Center in NewYork City was attacked by a hijacked
commercial airplane and was destroyed by a combination of the
plane impact and re ignited by the fuel aboard each plane. Shortly
after another hijacked commercial airliner ying at high speed hit
the Pentagon at ground level penetrating into the structure. In all
three events a huge reball occurred on airplane impact.
In the aftermath of 9/11 it became clear that any new NPP must
be designed sufciently robust against terrorist attack and that the
impact of a fast ying commercial airliner hitting the NPP can no
longer be excluded as a potential external hazard threatening the
NPP safety. The lesson learnt fromthe WTC and Pentagon impact is
0029-5493/$ see front matter 2009 Elsevier B.V. All rights reserved.
doi:10.1016/j.nucengdes.2009.04.018
W. Luther, W.C. Mller / Nuclear Engineering and Design 239 (2009) 20562069 2057
Nomenclature
BAM Bundesanstalt fr Materialprfung in Berlin
BLEVE Boiling Liquid Expanding Vapor Explosion
CFD Computational Fluid Dynamics
FDS Fire Dynamic Simulator
HRR Heat Released RateIAEA International Atomic
Energy Agency
NIST National Institute of Standards and Technology
NPP Nuclear Power Plant
SANDIA American National Laboratory
WTC World Trade Center
that the impact and the onboard jet fuel cause structural damage, a
reball and a consecutive re. This paper deals with the simulation
of the evolutionof the reball and its safety-relevant consequences.
All current knowledgeof reballs andtheir consequences results
fromvideofootage of the WTCandPentagonreballs andother re-
balls observed in air crashes. Fortunately this is not the only source
of knowledge. The reballs in air crashes are similar to reballs
occurring in Boiling Liquid Expanding Vapor Explosion (BLEVE)
accidents. BLEVE have occurred frequently in the last 50 years and
many of themare documentedby video footage. The general behav-
ior of a BLEVE can be explained and estimated by simple formulas
which showthat the consequences of a BLEVE can be calculated by
one single parameter, the fuel mass involved.
All BLEVEs have occurred in open space with only small struc-
tures around such that the reball can be treated as undisturbed
hemisphere evolving into a spherical ball. Since a NPP is made up
of large geometrical structures inuencing the spatial evolution of
the reball, a 3D simulation program is needed to calculate the
effect of the structures on the expansion of the reball and to deter-
mine details of thermal loads on the NPP which may lead to severe
damage.
This paper is doing pioneering work on the safety assessment of
a NPP of aircraft fuel impact in case of a large commercial airliner.
In this paper it will be shown how the computer program FDS can
be used to analyze the consequences of the air crash reball. The
applicability of the program was validated by the analysis of the
well-documented reball from the BAM 1999 BLEVE experiment
(Droste et al., 1999).
In this paper FDS is applied to a generic NPP using a typical
impact scenario. The results of the FDS simulation are analyzed and
evaluated in order to demonstrate the safety assessment based of
the generic plant. From the results it is obvious that the results for
the generic NPP cannot be applied directly without modications
of the FDS input to any other NPP.
The paper is organized as follows:

the rst sections deal with the regulatory approach, the types
of air crafts that have to be considered, the impact scenario and
the estimation of the percentage of the fuel contributing to the
reball,

the next sections are devoted to BLEVEs, the ndings from the
video footage and the theoretical explanations,

the following section deals FDS and the simulation of the BAM
1999 BLEVE,

nally a generic impact scenario for a generic NPP is simulated


with FDS and the FDS results are used for the safety assessment
of the NPP.
2. Some re dynamic basics
As this paper addresses the nuclear engineering community the
discussion of some basic re dynamic facts here is in order. Solid
and uid material does not burn unless it evaporates before. Any
burnable gas will burn if its temperature is above the ignition tem-
perature andthe fuel fractionmixture withair is betweenthe lower
and the upper ammability limit. Heat can only be transported
by conduction, convection or radiation; radiation is the dominant
mechanism in a reball.
Aircraft fuel is a hydrocarbonmixture withthe chemical formula
C
n
H
2n+2
andalmost identical withDiesel fuel usedincars. All hydro-
carbons behave similar in combustion with a heat of evaporation of
4MJ/kg and a heat of combustion of about 44MJ/kg. If hydrocar-
bon gas is burnt in the optimal (stoichiometric) ratio the resulting
gas mixture reaches a maximum temperature of 2000

C.
The adverse property of a fuel droplet cloud is the high volumet-
ric energy content, which exceeds a simple gas mixture by orders
of magnitude. Once the reball is started the droplets evaporate
due to heat radiation and the expansion of the hot gas towards
the cold ambient air atmosphere feeds oxygen to the reball. The
outer shape of the reball is dominatedby Taylor instabilities which
greatly enlarge the burning surface and lead to the self-sustaining
rapid turbulent combustion.
3. Regulatory approach to safety assessment of aircraft
impact
Since no national guidelines exists for the analysis of the impact
of aircraft fuel the starting point for regulatory approach are the
IAEA Safety Guides. The basic IAEA guide for safety assessment of
a NPP is the IAEA Safety Guide GS-G-4.1 Format and Content of
the Safety Analysis Report for Nuclear Power Plants which clearly
indicates that Guidance on the assessment and verication to be
conducted by the design and operating organizations in preparing
the SAR is provided in Safety Guide NS-G-1.2, Safety Assessment
and Verication for Nuclear Power Plants. This guide does not deal
explicitly with the problembut recommends the use of IAEA Safety
Guide NS-G-3.1 External Human Induced Events in Site Evaluation
for Nuclear Power Plants published in 2004.
The relevant section is section 5.16, page 26 found in Section 5
Aircraft Crashes, which reads as follows.
Effects caused by aircraft fuel:
The following possible consequences of the release of fuel from
a crashing aircraft should be taken into account:

burning of aircraft fuel outdoors causing damage to exterior plant


components important to safety,

the explosion of part or all of the fuel outside buildings,

entry of combustion products into ventilation or air supply sys-


tems,

entry of fuel into buildings through normal openings, through


holes caused by the crash or as vapor or an aerosol through
air intake ducts, leading to subsequent res, explosions or side
effects.
Section 5.18 gives additional details: The type of fuel and
the maximum amount of fuel potentially involved in an accident
should always be evaluated in order to quantify the re interaction
effects and correlate them with the potential structural damage.
The amount of fuel should be evaluated for this purpose on the
basis of the type of aircraft and typical ight plans.
A reball analysis using the approach presented in this paper is
one way to meet the above requirements.
4. Concept for safety assessment approach
In case that no well-dened regulatory guidelines exist a con-
cept for the safety assessment has to be developed rst. All relevant
2058 W. Luther, W.C. Mller / Nuclear Engineering and Design 239 (2009) 20562069
Table 1
Comparison of Airbus A380, Boeing 747 and Phantom F-4.
Large commercial airlines compared with the SANDIA Phantom
Type Airbus A380-800 Boeing Jumbo 747-400 SANDIA F-4 Phantom 19-4-1988
Maximum velocity 0.89 Mach 0.9 Mach
Travel velocity (impact velocity) 0.85 Mach 253m/s 0.85 Mach 253m/s 215m/s
Maximum fuel capacity [to] 249 174 4.8 (water)
Maximum passengers 555 416 1
Maximum take-off weight [kg] 560000 396890 19000
Wing span [m] 79.60 64.44 11.77
Fuselage diameter [m] 7.14 6.49 1.91
issues must be analyzed and data for these issues have to be
dened:
1. the type of the aircraft,
2. the point of impact,
3. the impact velocity,
4. the amount of fuel involved in the reball and in the fuel spill
re,
5. the damage done to the exterior of the buildings,
6. the vents, where the reball or smoke may enter, and the hazards
inside the buildings,
7. the safety objectives that have to be met.
These issues have to be dened before the start of the assess-
ment process in order to make conservative choices in the process
to demonstrate the capability of the NPP to cope with the reball
loads.
In case of a reball the safety objective are safe shut-down
and no substantial radioactive release to the environment and
the initiating events to be investigated are entry of fuel or re in
the air intake vents or air exhaust vents of the reactor building and
the auxiliary reactor building and the diesel buildings.
In case of a large spill re the effect of the smoke and the entry
of fuel into cable duct and piping systems or sewage systems has to
be investigated after the reball analysis.
5. Aircraft impact scenarios
As the rst step reasonable assumptions concerning the aircraft,
its fuel load and the impact velocity have to be made. It is good
safety analysis practice to dene a worst-case scenario. Based on
the assumption that the worst case is dened by both maximum
air craft impact velocity and maximum fuel load one has to look at
the largest commercial airliners in service. The largest airliners in
service beside the fewAntonov An-124andAn-225whichare own
by Russian crews only are Airbus A380 and Boeing 747. Specic data
for these planes can be found in the Internet and are presented in
Table 1.
Fig. 1 shows the aircraft structures are large when compared to
the NPP structures Simulations using point models are an inade-
quate approach to the problem, as all details of the structure have
be taken into account.
Jet airliners do not have separate tanks like cars. Instead, the
aircraft structure, the spaces between the wing spars, parts of the
body, or the n, is coated with layers of rubbery sealant to form
a series of fuel-tight, leak-proof compartments. The different tanks
arelinkedbypipes andvents at thewingtipor tail keepthepressure
constant.
Fig. 2 shows the location of the tanks. In dening the accident
scenario one has to make sure the assumptions howand howmuch
fuel is sprayed are consistent with rst principles. It is virtually
Fig. 1. Comparison of Airbus A380, Boeing 747 and NPP structure.
W. Luther, W.C. Mller / Nuclear Engineering and Design 239 (2009) 20562069 2059
Fig. 2. Tanks of the largest commercial airliners in service.
impossible that both wings are smashed at the reactor building.
Obviously if the two wings are smashed at different locations the
result maybetwospatiallyandtimelyindependent reballs or even
more likely one reball and one pool re.
As a rule of thumb a 747-400 consumes 4 to of kerosene for take-
off and about 10t/h of kerosene in cruise. Even assuming a worst
case scenario some reduction has to be made for the cruise from
the airport to the NPP.
The analysis of the WTC impacts, carried out by the authors
gives an impact velocity of 230270m/s, while the FEMA report
(FEMA, 2002) gives 263m/s. It should be pointed out that the max-
imumvelocity close to the surface is only limited by the stagnation
pressure and can be well above the specied admissible veloci-
ties.
6. Experience from air crashes
A well-documented experiment on hard target impact of an air-
plane is available: the SANDIA Phantom experiment performed in
1988, when a phantom F-4 ghter was projected on a rocket sled
against a concrete block and literally crushed to pieces.
The data of the experiment have already been presented in
Table 1. As a result of the impact a debris cloud of substantial
size (60m diameter) is formed. The volumetric ratio of structural
material and fuel stimulant (water) is 1:1 and a substantial part of
thedebris clouds consists of water droplets. Thesizeof thecloudcan
only be explained by the fact that small size water droplets experi-
ence a reduced drag when traveling in the wake of larger structural
particles.
Based on this only experiment with hard target and a small
ghter plane it is assumed that hard target impact of a large com-
mercial aircraft will be similar: a large cloud is expected consisting
of structural debris anda large amount of fuel, as the ratiostructural
material/fuel is much higher for commercial airliners.
Table 2
Documentary footage analyzed for this paper.
Location Plane Date
NASA dryden CID collision
impact demonstration
Boeing 720 1-12-1984
Fairchild air force base B52 24-6-1994
Fairford Mig 29 24-7-1993
Mulhouse Habsheim Airbus A320 26-6-1988
Okinawa Boeing 737 20-8-2007
Oostende airshow Small single engine prop plane 26-7-1997
ParisLe Bourget Mig 29 6-9-1989
Ramstein airshow 3 small ghter jets 28-8-1988
Sioux City DC 10 19-7-1989
Ukraina/Krim TU 134 10-7-2006
Lviv SU 27 27-7-2002
In an air crash against a structure it is reasonable to expect that
not all fuel will be consumed in a reball but only part of it. The
rest of the fuel will be distributed in form of a liquid lm on the
structures and on the ground and will result in a subsequent pool
re. In this paper the scenario under investigation focuses on the
reball only and leaves the subsequent pool re as a task for future
work in this eld.
With ignition sources available at large it is expected that the
droplet cloud is immediately ignited and results in a large reball.
In the 9/11 events, large commercial airplanes impacted soft tar-
gets. From the video footage it can be seen that the in the aircraft
impacts on the WTC towers (FEMA, 2002) and on the Pentagon
(ASCE, 2003) the aircraft literally disappears in the soft building
structure and immediately after entry part of the aircraft fuel is
ignited into large reballs of 60100m diameter. From the video
footagetheexpansionvelocityof theWTCreballs canbeestimated
to be 20m/s, which is also conrmed by Baum and Rehm (2005).
A lot of documentary footage of aircraft crashes is available on
the Internet and has been analyzed for this paper. A list of the
accidents analyzed is presented is given in Table 2.
Thefootageshows that most of theair crashes whether theyhap-
pen on hard runways or soft terrain ground result in an immediate
large reball. Typically only part of the onboard fuel is involved
in the reball. The reballs show all features of BLEVE reballs
discussed in the next section.
Only in case of low impact velocity and soft terrain some air
crashes fortunately did not result in reballs though fuel was
leaking from the broken wings, e.g. British Midland Kegworth air
disaster 1989.
7. Experience from BLEVEs
The word BLEVE is an acronym for Boiling Liquid Expanding
Vapor Explosion. This is a type of explosion that can occur when
a vessel containing a pressurized liquid is ruptured. A BLEVE can
occur in a vessel that stores a substance that is usually a gas at
atmospheric pressure but is a liquid when pressurized. If the vessel
is ruptured generally due to loss of strength of the heated part of
the vessel that is not cooled by the liquid inside the sudden pres-
sure drop causes the liquid to turn into a two phase mixture, which
is blown out forming a rich in fuel droplet cloud which typically
results in a spectacular reball.
In the last 50 years more than 80 major BLEVEs have occurred
taking the lives of several thousand people and leaving more than
10000 injured (Abassi and Abassi, 2007). Many BLEVEs have been
covered by news crew footage and a large amount of documentary
material on BLEVEs is available on the Internet.
The authors conclusions, that BLEVE reballs are having the
same basic features as aircraft impact reballs, is based on the two
facts:
2060 W. Luther, W.C. Mller / Nuclear Engineering and Design 239 (2009) 20562069
Table 3
Documentary footage of hot BLEVEs analyzed for this paper.
Location Fuel Date
BAM Berlin Propane 11-9-1998
Bucheon, Korea LPG 11-8-2005
Lucio Blanco, Mexico Propane 11-8-2005
Murdock, USA LPG 3-9-1983
Miami Ives Dairy substation Transformer refrigerant 2001
Tanker truck, Korea Styrene ???
1. The similarity of all low order hydrocarbons in combustion,
chemical properties and physical behavior.
2. Fireball footage shows the same typical evolution of all re-
balls, except for the initiating phase where different sources for
the initial moment can be found: the initial momentum of a
BLEVE reball originates from the tank rupture while in case of
the aircraft the initial momentum originates from the impact
momentum of the fast ying aircraft.
FromBLEVE experience two different types of BLEVE evolutions
into reballs have been observed and discussed in the literature:
1. If the initial momentum is to small or oxygen supply is insuf-
cient a cold BLEVE will occur, which means that there is only a
small re ash, followed by a large spill re.
2. If the initial momentumis large andenoughoxygenis available, a
hot BLEVE will occur. The momentumof the initial phase has to
be sufcient to start turbulent combustion. Then the expanding
reball is self-sustaining, absorbing enough fresh air to continue
burning at a more or less constant rate. Typically the hemispher-
ical expansion is followed by a buoyancy induced upward ow
and the reball forms the well-known mushroom structure.
In the remainder of this paper the focus is on the simulation of
a hydrocarbon reball of the hot BLEVE type.
The documentary footage of hot BLEVE reballs that has been
analyzed for this paper is given in Table 3.
From the footage and theoretical analysis it can be concluded
that a hot BLEVE reball consists of three phases:
1. The initial phase, whenthe liquidvapor mixture is blownout and
which are typically light or white in the video information.
2. The expansion phase, when the expanding cloud forms a hemi-
spherical reball which sticks to the ground. This phase is
momentum controlled. The dominating phenomenon is turbu-
lent combustion of the hot rich in fuel reball expanding into the
cold ambient air forming an instable surface governed by Taylor
instabilities.
3. The up-lift phase, whenthe hot reball starts torise inthe more or
less spherical mushroom-shaped plume with the vortices suck-
ing ambient air into the reball. This phase is controlled by the
hot gas buoyancy. In this phase the remaining fuel is burned
almost completely (Fig. 3).
The same holds for air crash reballs. A good illustration of the
three phases can be seen in Fig. 4, which shows a B52 bomber
crashing into a concrete runway.
As a consequence of the similarity engineering theories on
BLEVE expansion and uplift can be applied to air crash reballs
and provide a method for estimation of reball size, duration and
height as well as plausibility assessment for more sophisticated
simulations.
Though many BLEVEs have occurred in the last decades and
manyexperiments havebeenconductedmostlylimitedfrom1kgto
about 5t of hydrocarbons, no well-instrumented experiments have
been carried out on really large BLEVEs with 50 ore more t of hydro-
Fig. 3. B52 bomber crashes on Fairchild aireld runway.
carbons. All information on large BLEVEs has been derived from
extrapolation the ndings from the footage on the visual domain
and the formulas obtained in the analysis.
The transient behavior of BLEVE reballs shows two different
velocities depending on the physical effect dominating the process:
1. During the expansion phase the velocities are momentum-
dominated. The reball is expanding at a more or less constant
W. Luther, W.C. Mller / Nuclear Engineering and Design 239 (2009) 20562069 2061
Fig. 4. Denition of reball diameter and height used in the analysis.
expansion velocity and no substantial change in height due to
uplift is observed.
2. During the up-lift phase the expansion is slowed down and the
reball rises at a more or less constant velocity.
The difference between the two phases is more explicit in
smaller BLEVEs. The same two phases of evolution have also been
observed in aircraft crash jet fuel reballs.
Manyengineeringformulas havebeendevelopedtoestimatethe
maximum size, duration, radiation, missile range and safe stand-
off distance of a hot BLEVE. A good overview is given in (Abassi
and Abassi, 2007). A simple formula can be derived for the maxi-
mum hydrocarbon reball diameter by assuming that the reball
is formed by an isochoric combustion followed by an isothermal
expansion:
D
fireball
= 5.8m
1/3
fuel
unit of D
refall
is (m) and unit of m
fuel
is (kg).
An in-depth analysis of the various formulas found in the liter-
ature (Abassi and Abassi, 2007) shows that there is no substantial
difference between the formulas if the uncertainties of the visual
evaluation of reball footage are taken into account. Most reballs
show a good agreement with the above formula. It should be men-
tioned that the formulas given in standard reference books (TNO,
1992) and(SFPE, 2003) andalsothe formulas usedinNUREG(2004)
NUREG-1805 are slightly different but give similar values.
The durationof the reball is harder todetermine thanthe maxi-
mumdiameter, since the reball evolution consists of two different
phases. From scaling arguments the exponent is 1/6 if the re-
ball growth is dominated by expansion and 1/3 if it is dominated
by buoyancy. From observations it seems that larger reballs are
buoyancy-dominated. So the following formula is recommended
by many authors:
t
d
= 0.45m
1/3
fuel
, m
fuel
< 3 10
4
kg
t
d
= 2.6m
1/6
fuel
, m
fuel
> 3 10
4
kg
The formulas for the height of the reball (the height of the center
of the sphere approximating the reball in its latest phase above
the ground) are controversial and range from:
H
FB
=
3
4
D
fireball

3
2
D
fireball
In fact the problem lies in the uncertainties of the reball visi-
bility and it is often left to the expert to decide whether the reball
has already extinguished or is still on re but is obscured by smoke
and soot. A review of large reball footage leads to the conclusion
that lower height values can be found in large hydrocarbon reballs
with a higher probability.
From the evaluation of the reball footage it can be observed
that the maximum diameter and the duration are independent of
the details of the transient.
An analysis of the 9/11 aircraft impact reballs video informa-
tion indicates that the expansion velocity is almost constant and
was 20m/s in the WTC reballs. In the WTC reballs the center of
the reball always rises linearly in the up-lift phase and the up-lift
phase starts at 1/3 of the reball total duration. Based on this obser-
vation and using the evaluation technique described in Section 9
the authors have derived from the reball footage a new universal
formula for the start of the up-lift phase: The phase (up-lift phase)
starts at 1/3 of the total duration time.
Applying the engineering formulas to the BAM reball of 5.9 to
and a hypothetical reball from an air crash with 90 to gives the
following data (Table 4).
2062 W. Luther, W.C. Mller / Nuclear Engineering and Design 239 (2009) 20562069
Table 4
Engineering formulas for 5.9 and 90 to hydrocarbon reballs.
Fuel mass m
fuel
5900kg 90000kg
Maximum diameter Dmax = 5.8m
1/3
fuel
105m 260m
Maximum center height hmax =0.75Dmax 79m 195m
Duration, large reball tmax = 2.6m
1/3
fuel
17.4s
Duration, small reball t
d
= 0.45m
1/3
fuel
8.1s
With respect to the length and time scale, the observations
from video footage t well with what is expected from the above
engineering formulas and the evaluation of other reball video
information, if the evaluation methods used are those from Sec-
tion 9. It is not clear fromthe papers published in the past, howthe
authors evaluated the diameter and height of the reball and this
fact makes comparison difcult.
The BLEVE formulas allow a rst guess for the size and evolu-
tion of a reball but they do not take into account the mitigating
or reinforcing effect of building structures on the evolution of the
reball. The second limitation of the use of engineering formulas
is that they do predict in detail the local impact of the reball. In
order to assess the local effects and the effect of buildings, a 3D
simulation is necessary.
8. Overview of the capabilities of FDS
The reball simulation documented in this report has been car-
ried out with the code FDS which is freely available from NIST.
The program package consists of two parts, the simulation pro-
gram FDS (re dynamics simulator) (McGrattan et al., 2007a,b,c)
andthe visualizationprogramSmokeview(Forney, 2007) (Internet:
www.nist.gov/fds).
The NIST Fire Dynamics Simulator FDS predicts smoke and/or
air ow caused by re, wind, ventilation systems, etc. It consists of
ve major modules:

A computational uid dynamics module, which solves the Large-


Eddy form of the turbulence equations using the Smagorinski
model and the lowMach number approximation. This turbulence
model has been showto be appropriate for low-speed, thermally
driven ows of smoke and hot gases generated in a re with an
emphasis on smoke and heat transport from res. This module is
using the Eulerian approach.

An improved combustion model for the re dynamics including


pyrolysis of solids and a mixture fraction combustion model.

A module for heat transfer by conduction, convection and radia-


tion.

A special module for devices and controls, e.g. for the simulation
of thermocouples.

Amodulefor thesimulationof sprinkler sprays usingaLagrangian


approach by tracing a large number of droplets. This module has
been extended to simulate fuel sprays, but this feature is still in
an early phase of development and still experimental.
FDS has been used for many types of problems like sprinkler
activation in warehouse res, tunnel res, tenability in residential
res, and smoke concentration in outdoor res. Lately FDS has been
successfully applied to the analysis of res in nuclear installations
(NUREG, 1824).
After 9/11 a rst attempt has been made by the authors to use
FDS to simulate the WTC reballs and other problems relating to
reballs causedbyaircraft fuel inatargetedaircraft crash. Duetothe
limited capabilities of the older FDS versions and limited comput-
ing power before 2007, these simulations were restricted to small
reballs up to 5t.
Thesimulations documentedinthis report havebeenperformed
with the new FDS Version 5 using the parallel version on a large
computer cluster. The main objective of FDS simulations is to cap-
ture the global characteristics of the reball and its evolution.
The main features of Smokeviewused in the compilation of this
report are:
1. Animated iso-surfaces of the heat release rate per unit volume
(HRRPUV) that gives a visual impressionof the ame boundaries.
The re -colored surface is generated by an interpolation of the
iso-surface HRRPUV=30kW/m
3
, which is thought to represent
the effective ame boundaries but is not necessarily the ame
boundary as seen by an observer. In video information often a
large part of the reball is obscured by soot and smoke. This
must be kept in mind when a direct comparison is performed.
2. Slice les that represent results recorded for a grid plane (e.g.
y =0.0) by colors, e.g. temperatures, velocity, etc.
In addition to these options, the most relevant global data of the
simulation are saved in special data les to produce time history
plots of selected parameters at selected positions.
9. Rationale of the application of CFD and FDS to reball
simulation
The application of CFD codes like FDS to the simulation of large
reballs for which no experimental data are available is an extrap-
olation of small-scale behavior to larger size. This is justied by the
fact that the phenomena governing the evolution of small reballs
are the same as those for large reballs. Arationale for this approach
will be given in this section.
The physical and chemical phenomena governing the evolution
of a typical indoor or outdoor re like a burning building or fuel
tank are highly complex and must be simplied so that they can
be effectively solved by computer simulation. It has always been a
moot point in re simulation that neither experiments nor real life
observations are replicated in all details. It is therefore in order to
question the validity of re simulations.
In contrast to this a reball is a short-term re blast and the
physical and chemical phenomena governing the evolution of a
reball are quite clear and simple, as can be guessed from the fact
that the simple engineering formulas for the BLEVE give a good
approximation.
The simulation of a reball in the open atmosphere without
obstacles inthe owpathcanbe adequately performedwithsimple
analytical models, e.g. spark-ignited spherical combustion.
Fromthe evaluation of reballs it can be derived that three basic
phenomena control the evolution of the reball: the almost com-
plete combustionof hydrocarbons, the expansionof hot gas undthe
buoyancy of the hot gas ball. The combustion process is almost the
same as the combustion of a premixed gas sphere, but in this case
the premixed gas is replaced by the mixing process at the re front
due to local turbulence.
TheowphenomenaareadequatelysimulatedbytheSmagorin-
ski model of FDS, with the sub-grid turbulence being proportional
to the large-scale turbulence. The reball starts with no turbulence
at all and turbulence is developing during the evolution. Near wall
turbulence has only little effect on the evolution as most of the
evolution occurs in the open space. Both the expansion and the
buoyancy phase of the reball are atmospheric ows in the open
space except for the effect of the structures deecting the expan-
sion ow. The substantial contribution of turbulence to the process
is a local effect, as turbulence supplies the oxygen at the re front.
The main effect of turbulence is on the sub-grid scale affecting
the combustion reaction. FDS is using a simple but adequate Z-
W. Luther, W.C. Mller / Nuclear Engineering and Design 239 (2009) 20562069 2063
model for the reactionassuming that all fuel is immediately burnt if
oxygen is available. In the reball and the FDS simulation the burnt
hot gas pushes outside the fuel rich gas which is burning at the
interface with the atmosphere almost completely. The amount of
fuel burnt at the interface controls the heat release rate per unit area
at the ame front. This governing parameter is similar in FDS simu-
lations, analysis of reballs and data of experimental hydrocarbon
res. This parameter seems to be a global value for hydrocarbon
burning under atmospheric owconditions. This fact strongly sup-
ports the assumption that FDS simulations are also valid for larger
reballs of sizes beyond todays experience.
Sensitivity and uncertainty studies on reballs have been car-
ried out with FDS and have shown a low sensitivity of all input
data except for hydrogen mass and initial expansion velocity. The
studies demonstrate that the hydrocarbon mass and the choice of
an adequate starting velocity on the grid level are the dominant
parameters.
In FDS many sophisticated models have been implemented for
phenomena important to typical res like specic types of combus-
tion or all kinds of heat transport including radiation, but all these
models have only little effect on the result of the calculations and
in this way on the evolution of the reball. The important effect of
radiationis the loadonstructures outside the reball. The contribu-
tion of all theses models to the FDS reball simulation is negligible.
The radiation effect is not relevant for the NPP safety as most of its
structures are made of concrete.
The big advantage of using CFD codes like FDS, over engineer-
ing formulas is the possibility to simulate the effect of structures
deecting the expansion of the reball and to calculate detailed
temperature distributions.
In the paper (Baum and Rehm, 2005) a simple model for re-
ball dynamics is presented which has only implemented the basic
features discussed above and it was shown that this model can ade-
quately simulate the observed reballs. This successful simulation
can be taken as another proof for the fact that basic physics and
chemistry determine the evolution of a reball. Out of all models
implemented in FDS only these relevant robust features contribute
to the simulation results.
In summary, the evolution of a reball can be simulated using
a few simple models and only these features of FDS are involved
in the reball simulations. The main advantage of FDS is that has
incorporated all the necessary models and simulates the effect of
structures on the atmospheric ow. Any other CFD code, in which
the relevant models have been implemented, could be used for a
reball simulation just as well. This concludes the reasons for using
FDS for simulation of reballs.
10. Validation of FDS by simulation of BAM BLEVE
In 1999 a BLEVE experiment was carried out by BAM (Droste
et al., 1999), for which a large amount of documentary material,
report, photographs and video footage are available. Therefore this
experiment was selected for FDS validation. In this experiment a
tank wagon with 5.9t (10m
3
) propane was exposed to a re which
led to vessel failure. The resulting BLEVE reball had a maximum
diameter 100m, areball center height 80mandadurationtime
of 7s, until the reball extinguished.
Fig. 4 illustrates the method by which this data have been
derived from the video footage.
A frame by frame evaluation to the BAM reball footage gives
the time plots of diameter and height shown in Figs. 5 and 6.
The gures show that in the rst two seconds the reball
expands to a hemisphere with the diameter equal to the maxi-
mum diameter D
max
of the rising sphere. At 2s the up-lift starts
and the reball rises within 5s to its maximumheight of 80m. The
point is that using the above evaluation technique the diameter
Fig. 5. Results from the BAM BLEVE: reball diameter.
does not change but the volume increases substantially since the
reball shape changes from hemispherical to spherical and nally
to a mushroom. During the rst 2s the reball sticks on the ground
and after 2s it starts to rise with a constant upward velocity of
16m/s.
Inthe parallel FDS versionthe calculationdomainis divided into
a set of rectangular volumes called meshes. Each mesh is divided
into rectangular grid cells. Best simulation results are obtained
using a uniform grid with equal grid spacing in each direction for
all meshes.
The problem that the boundaries of the buildings and other
structures do not generally coincide with the grid is solved with
the help of a sophisticated method. FDS decides on the basis of per-
centage of the grid cell coinciding with objects whether the grid
cell is blocked or not.
The boundaries of the computational domain are either closed
(no ow across the boundary) or open. In case open boundaries
ow across the boundary to the outer external atmosphere and
also back-owfromthis external atmosphere to the computational
domain is allowed, but of course turbulence eddies are lost and
there is no realistic feedback. This means that when part of the re-
ball leaves the computational domain, the simulationresults canno
longer be considered as reliable.
Whenperforming the FDS simulations it was foundthat the sim-
ulation depends sensitively on assumptions concerning the initial
phase of the reball, which is partly based on expert judgment,
mainly the way the initial momentumis modeled in the FDS input.
The most difcult problem in the simulation is caused by the
run-away mechanisms inherent to turbulent combustion:
1. hot gas from combustion generates turbulent eddies,
Fig. 6. Results from the BAM BLEVE: reball height.
2064 W. Luther, W.C. Mller / Nuclear Engineering and Design 239 (2009) 20562069
Fig. 7. FDS simulation of the BAM reball: up-lift phase.
2. turbulent eddies supply fuel-lean regions below ammability
limit with oxygen,
3. burning regions heat upthe surrounding fuel-leanregions which
turn ammable since the ammability limit is lowered due to
temperature increase.
Run-away of turbulent combustion has to be avoided and it can
be done by carefully feeding the re load to the systemin a physical
way in an adequate time span and within an adequate region.
By numerical experiments it was established that a grid of 1m
length scale is adequate to model phase 2 (expansion phase) and
phase 3 (up-lift phase) of the BAMreball. This grid size is the max-
imum grid size possible in view of the available computing power
for the subsequent simulation of the impact of a 90 to kerosene on a
generic nuclear island, which will be presented in the nal section.
By showing that this grid is adequate for the small BAM reball it
can be concluded that it is also adequate for a larger reball.
The gas owis simulatedinFDS using the large eddyapproxima-
tion. With respect to a grid size of 1m this means that only eddies
and other phenomena of a scale size of 1mare directly represented
and computed whereas effects of the smaller scale motion and size
are modeled using the simplifying assumptions of the Smagorin-
ski model. This also means that turbulence at scales below 1m is
assumed to be isotropic and homogeneous (Fig. 7).
Only phase 1 (initial phase) of the reball when it is very small
compared to grid size cannot be modeled adequately. Using a 1m
gridsimplymeans that theinitial phaseis belowtheresolutionscale
of the grid. This phase has to be covered by engineering assump-
tions and a simulation model for the initial phase using the means
provided by FDS. Three different methods have developed to model
the initial phase of the BAM reball:
1. Prescribed HRR(heat released rate): a prescribed re load and an
accompanying air streamare introducedintothe gridby a vent.
The prescribed re load is modeled internally in FDS by a hot
propane gas ow which starts to burn immediately depending
on the availability of oxygen. The ratio (gas ow+air ow)/(vent
area) is the key factor for the initial momentum input. Since not
all parameters used in this method are known sufciently well,
they have to be quantied by expert judgment. These parame-
ters include the size of impact area or mass and velocity of air
entrained by the aircraft. By numerical experiments it was found
that the prescribed HRRmethod shows a lowsensitivity onthese
parameters when varied within a reasonable range. Using the
prescribed HRR the gas owfroma blower introduces a momen-
tum into the system, which has to be used to model the initial
momentum on the initially resting atmosphere from the failing
vessel in a BLEVE or from the crashing plane.
2. Initial fuel droplet cloud: the aircraft fuel is introduced into the
system as a burning static droplet cloud at the start of the sim-
ulation. In order to simulate the droplet ow and the aircraft
momentum, this cloud is blown by a vent towards the point of
impact similar to the prescribed HRR method.
3. Dynamic fuel spray: this technique is the most sophisticated and
most promising approach. The impact of the aircraft fuel is sim-
ulated dynamically by fuel droplet sprays. These sprays can be
modeled in detail by a set of parameters including droplet ow
rate and droplet velocity. The droplet spray introduces both re
load and momentum into the system.
The use of the dynamic fuel spray model allows calculating the
accumulated droplet mass per unit area (kerosene spill) that is not
burnt in the reball and this fuel mass can be used as input for
the subsequent pool re simulation. Unfortunately the fuel spray
model in FDS did not perform as expected due to the early state of
development of the fuel spray model. Since it has a high potential
for best-estimate simulation it is highly desirable that future work
should be directed to improve the model performance.
Fig. 8 compares the global parameter heat released for the
three methods. The parameters used in the FDS calculations have
not been optimized but were selected in a best-estimate fashion.
Fig. 8 shows the global heat released calculated with methods 1
and2 are very similar andare ingoodagreement withthe engineer-
ing formulas. Method 1 requires less computing time. When using
method 3, however, the reball burns very fast and the duration
time is much shorter than observed in the experiment or estimated
from the engineering formula.
A good agreement between calculated and experimental data
for the BAM reball has be also been achieved using methods 1
and 2 for reball diameter and height as shown in Figs. 5 and 6.
The comparison of footage and ame front simulation gives a good
visual impression of the quality of the simulation. Fig. 7 shows the
iso-surface of the reball during the up-lift phase.
All three methods are using input parameters that are well
known and clearly dened but also some input parameters that
are poorly known and in this way are left to the expert judgment
of the analystat least to some extent. This is specic for reballs
or FDS simulation but also true for most simulation computer pro-
grams. Test calculations revealedthat the prescribedHRR method
requires less parameters both well known and those left to the
judgment of the analyst than the other two methods. It shows
less sensitivity to the choice of these parameters than the other
methods. It proved more robust than the other methods and rarely
produced numerical instabilities. It also required less computing
time. This is why method 1 was used in the next section.
11. The FDS analysis model of a generic nuclear island
In this section several scenarios of a hypothesized air crash of
a large commercial airliner on a generic NPP are analyzed using
the worst case assumption of 90 to of kerosene consumed in the
reball. The available computer power limits the grid size of the
simulation. The required grid size can be easily estimated from the
BLEVE formulas. A 90 to reball has a maximum scale length of
260m. The computer power available was limited to a maximum
computational domain of 240m240m390m, when using the
grid length of 1m. In the simulation the computational domain was
divided into up to 30 meshes of 24024013 grid and solved with
the parallel version of FDS using 18 millions grid cells on a LINUX
cluster of 30CPUs with2Gbmemoryfor eachCPU. Thecomputation
domain is not large enough to simulation the complete evolution of
the reball, but it is adequate to simulate the safety-relevant phase
of reball expansion close to the NPP buildings.
The input for the generic NPP was done on a 10cm scale. Based
onthe choice of the gridlengthof 1mFDS converts the detailedNPP
geometryintoanassemblyof 1mcubes. Curvedsurfaces or surfaces
W. Luther, W.C. Mller / Nuclear Engineering and Design 239 (2009) 20562069 2065
Fig. 8. BAM reball heat released: comparison of best-estimate simulation with the three different methods.
that do not coincide with the main axis are modeled by saw-tooth
structures. Fig. 9 shows the resulting NPP geometry used in the FDS
simulations and how the curved containment shell is converted in
an assembly of rectangular cubes.
As can be seen from Fig. 9, the 1m grid is ne enough to model
the geometrical details which have a substantial inuence on the
computational results, e.g. the gaps between the buildings.
11.1. Load assumptions
The following load assumptions have been used:
1. It is assumed that only one wing contributes to the reball and
the total fuel mass of this wing is consumed in the reball.
2. The maximum fuel mass of one wing is 90t kerosene equivalent
to 42.6GJ/t assuming the typical value for hydrocarbon heat of
combustion of 42.6MJ/kg. The resulting total fuel energy adds
up to 3834GJ.
3. The impact velocity of the aircraft is 230m/s.
4. Three points and directions of impact have been selected: (see
Fig. 9):

Load case 1 (LC 1) is the horizontal impact on the containment


dome from the direction of the turbine hall above the roof.
Fig. 9. FDS model of the generic NPP island.

Load case 2 (LC 2) is the horizontal impact on the containment


dome from the direction of the diesel building above the room of
the control building.

Load case 3 (LC 3) is an inclined impact into the gap between the
containment dome and the control building fromthe direction of
the diesel building.
The largest unknown in reball simulations is the assumptions
concerning the airplane and its impact when crashing on the NPP.
A large variety of scenarios can be hypothesized and have to be
covered by engineering assumptions.
The FDS calculations presented in this section were carried out
using the prescribed HRR method to model the initial phase. Arti-
cial vents are used to blow the hot gas corresponding to the
prescribedHRRintothe gridat some distance fromthe impact point
as indicated by the arrows in Fig. 9. The parameters to choose in this
case are the size of the injection area (vent) and their distance from
the structures. Since part of the momentum introduced by the gas
ux is consumed by the structures blocking its way an additional
air vent was used to adjust the initial phase of the reball to a real-
istic rst expansion which is known from the experience of 9/11
reballs and other reballs. These parameters were determined by
numerical tests to produce an initial expansion velocity of 20m/s.
The vents operate for 1s and are removed fromthe grid afterwards.
11.2. Results
This paper focuses on the rst seconds of the reball before it
rises above the NPP and the potential hazard of the ame front on
the safety of the NPP. All other effects going along with the reball
like missiles and heat radiation leading to loss of lives and property
are not taken into consideration. Secondary or domino effects, too,
are not investigated.
With the safety related buildings being protected against pres-
sure waves and heat radiation, the questions to be answered by the
FDS simulation are:
1. How does the presence of the NPP buildings affect the evolution
of the reball?
2. Which safety related openings of the buildings are reached by
the reball?
2066 W. Luther, W.C. Mller / Nuclear Engineering and Design 239 (2009) 20562069
Fig. 10. Load case LC 1: reball at 2s.
3. Which temperature levels and duration, velocities and fuel mix-
tures are reached at these openings?
The global evolution of the reball in LC 1 is given in Figs. 1013
at a time interval of 2s. The expansion phase close to the ground
lasts about 5s. During this phase the reball ame front expands
over the NPP structures but its temperature is still low since only a
small part of the fuel is burnt and the produced heat is used to heat
up the fuel. At 5s the lift-up phase starts and the reball moves
upwards and away from the structures and burns away above the
NPP. During this phase onlyradiationconstitutes a potential hazard.
At 10s the reball starts to leave the computation domain, which
means that the FDS results are nolonger adequate, but Fig. 13shows
that thereball is not longer incontact withtheNPPstructures. That
means that the safety-relevant phase is over.
A large-scale turbulent motion can be derived from this synop-
sis, whichconsists of aquasi-hemispherical expansionphase, which
is followed by an up-lift phase. The synopsis shows that until 5s
Fig. 11. Load case LC 1: reball at 4s.
Fig. 12. Load case LC 1: reball at 6s.
the reball is in close contact with the NPP buildings. The start of
the up-lift phase conforms well to the 1/3-rule derived from the
evaluation of reball footage.
Before 1.5s its evolution is controlled by the model for the
initial phase but after this time due to turbulent dissipation the
transient shows the typical self-sustained expansion of the hot gas
from turbulent combustion. The combustion is controlled and lim-
ited by lack of oxygen supply. The reball size can be approximated
by a hemisphere which is growing in time.
At 5s the reball starts to rise due to buoyancy and its form
can be approximated by an upward stretched ellipsoid. At this time
the typical vortex due to a rising bulk of hot gas starts to form but
due to grid restrictions the well-known mushroom structure of a
thermal plume cannot form.
Fig. 13. Load case LC 1: reball at 8s.
W. Luther, W.C. Mller / Nuclear Engineering and Design 239 (2009) 20562069 2067
Fig. 14. Load case LC 2: temperature slice plot of the reball.
Due to the NPP geometry the expansion hemisphere is not sym-
metric. Fresh air from the sides of the building and also with some
additional vorticity is fed to the reball sitting on top of the build-
ing structures. This effect can also be seen in the up-lift phase when
the reball takes the ellipsoidal shape.
In LC 1 due to the impact height of 33m most of the upper sur-
faces of the NPP structure come into contact with the ame front. In
Fig. 15. Load case LC 2: the hot reball high above the NPP.
Fig. 16. Load case LC 3: the reball is enclosing the NPP.
LC 2 with an impact height of 45mthe ame front comes in contact
only with the upper part of the containment dome as can be seen
from Fig. 14.
For this case the reball ame front does not constitute any
potential hazard to the NPP. The bulk of the fuel is burning high
above the NPP, as can be seen from Fig. 15, which shows the end
phase of the reball. It is only in this nal state that maximum
temperatures of 2000

C are reached.
LC3simulates animpact close tothe ground. The reball reaches
a large extension at the ground and the ame front comes into con-
tact with all structures and all openings. Fig. 16 demonstrates this
adverse effect of low impact height.
FromFig. 14, it can be seen that the reball is not very hot during
the rst seconds. This statement does not hold in general: when the
ame front is passing around obstacles the increase in turbulence
may lead to a substantial increase in burning rate and consequently
result in higher local temperatures. In LC 3 this is the case when the
ame front runs over the top of the diesel building. Fig. 17 shows
the hot swirl generated at the edge of the diesel building.
Fig. 17. Load case LC 3: hot swirl generated by turbulence at the edge of a structure.
2068 W. Luther, W.C. Mller / Nuclear Engineering and Design 239 (2009) 20562069
Fig. 18. Load case LC 3: another hot swirl generated by turbulence at the edge of a
structure.
An even more impressive swirl is found when the ame front
runs over the edge of the emergency feed-water building, as shown
in Fig. 18. A consequence from this observation is that in the pres-
ence of complex NPP structures with non-smooth surfaces the
phenomenon of increased turbulence leading to higher tempera-
tures has to be taken into account.
The FDS simulation gives details for the key parameters at the
safety-relevant openings. FDS allows calculating local values for
many relevant parameters:

gas velocity,

temperatures,

burning rate,

fuel/air ratio,

heat ux,

radiation ux.
that may be used to assess potential hazard, e.g. at vents and
openings of the safety buildings.
To demonstrate these FDS features, some key parameters at the
ctive air intake in LC 1 will be discussed in detail: normal velocity
towards the opening, the fuel mixture fraction and the tempera-
ture of the gas ow which are shown in Fig. 19. The air intake
is about 50m away from the point of impact and is reached by
Fig. 19. Time history of safety-relevant parameters at a ctive opening.
the ame front at 2s. It experiences a short blow of gas ow
with 30m/s and after this only smaller atmospheric turbulence
is observed. From2 to 4s the gas owis rich in fuel but at lowtem-
peratures. It transports a high fuel load and has the capability to
transport this fuel load through the opening. The FDS simulation
does not answer the question what will happen inside the build-
ing but it provides the boundary condition for an additional detail
analysis.
After the rst heavy blowthe normal velocities directedtowards
the opening return to low levels but the temperatures of the gas
owreach a maximumof 1000

C and temperatures remain at an


elevated level for several seconds. After the rst blow the temper-
atures stay above 500

C for about 5s, but at this period the gas is


lean in fuel.
12. Conclusions
From the real life experience with air crashes it is reasonable to
assume that as a consequence of a hypothetical commercial airliner
crash on a generic NPP a gigantic reball occurs. This reball may
constitute a substantial hazard to the safety of installations which
has to be analyzed in the safety assessment process.
In this paper for the rst time a method has been presented
that is adequate to simulate large reballs at NPPs and to quantify
local values for the safety-relevant key parameter and to give the
necessary input data for potential hazard assessment.
The new element in the FDS simulation is that buildings and
other objects are included in the simulation and that time histories
of local values for many parameters are calculated and provided for
further detailed analysis.
FDS was validated with the help of the BAM BLEVE experi-
ment with 5.9 to propane. The FDS results show a good agreement
between calculated and experimental data for both the global
parameters and the visual comparison of footage and ame front
simulation.
The simulations using the recent parallel version of FDS were
performed on a large computer cluster allowing a sufciently high
grid resolution. The FDS results gave new insight in the details of
the problem:

Building structures have a substantial impact of the reball evo-


lution.

In the early phase the reball is cool, rich in fuel and lean in
oxygen.

Local turbulence caused by obstacles and corners in the owpath


may result in hot spots.

The highest temperature is reached after the reball has risen


above the NPP.
In this paper it was demonstrated that FDS is an adequate tool to
simulate the effects reball caused by the crash of a commercial
airliner and by this is a useful tool to assess potential hazard and
to derive mitigation measures.Future developments of the FDS are
recommended to make the droplet spray model more stable and
robust, to improve the ease of using the program and to reduce
computing times.
Acknowledgements
The authors would like to thank the FDS developers and the
members of the FDS Discussion Group for helpful advice and sug-
gestions.
The authors would like to thank the BAM Federal Institute for
Materials Research and Testing, Department Containment Systems
for Dangerous Goods.
W. Luther, W.C. Mller / Nuclear Engineering and Design 239 (2009) 20562069 2069
Appendix A. Supplementary data
Supplementary data associated with this article can be found, in
the online version, at doi:10.1016/j.nucengdes.2009.04.018.
References
Abassi, T., Abassi, S.A., 2007. The boiling liquid expanding vapor explosion (BLEVE):
mechanism, consequenceassessment, management. Journal of Hazardous Mate-
rials 141, 489519.
ASCE (American Society of Civil Engineers), 2003. The Pentagon Building Perfor-
mance Report. ASCE Publications, Reston, VA, ISBN: 0-7844-0638-3.
Baum, H.R., Rehm, R.G., 2005. A simple model of the World Trade Center reball
dynamics. Proceedings of the Combustion Institute 30, 22472254.
Droste, B., Probst, U., Heller, W., 1999. Impact of an exploding LPG rail tank car onto
a CASTOR spent fuel cask. RAMTRANS 10 (4), 231240.
FEMA 403 (Federal Emergency Management Agency), 2002. World Trade Center
Building Performance Study: Data Collection, Preliminary Observations, and
Recommendations. FEMA 403.
Forney G.P., 2007. Users Guide for SmokeviewVersion 5A Tool for Visualizing Fire
Dynamics Simulation Data.
McGrattan, K., Hostikka, S., Floyd, J., Hamins, A., Klein, B., 2007. Fire Dynamics Sim-
ulator (Version 5), Verication and Validation Guide, vol. 1, Verication.
McGrattan, K., Hostikka, S., Floyd, J., Baum H., Rehm, R., 2007b. Fire Dynamics Sim-
ulator (Version 5), Technical Reference Guide (2007), NIST Special Publication
1018-5.
McGrattan, K., Hostikka, S., Floyd, J., Klein, B., 2007c. Fire Dynamics Simulator (Ver-
sion 5), Users Guide (2007) NIST Special Publication 1019-5.
NUREG 1805, 2004. Fire Dynamics Tools (FDT): Quantitative Fire Hazard, Analysis
Methods for the U.S. Nuclear Regulatory Commission Fire Protection Inspection
Program, Final Report, December 2004.
NUREG, 1824, Verication and Validation of Selected Fire Models for Nuclear Power
Plant Applications, vol. 7, Fire Dynamics Simulator (FDS) (PDF 3.60kB).
SFPE Handbook of Fire Protection Engineering, Society of Fire protection Engineers,
Third edition, 2003.
TNO, Methods for the Determination of Possible Damage to People and Objects
Resulting from Releases of Dangerous Material (Green Book), CPR 16E,
1992.

Anda mungkin juga menyukai