1

FACILITY SITING - CASE STUDY DEMONSTRATING

BENEFIT OF ANALYZING BLAST DYNAMICS

Henry Ozog and Georges A. Melhem

Arthur D. Little, Inc.
Acorn Park, Cambridge, MA 02140 (USA)
Bert van den Berg and Paul Mercx
TNO Prins Maurits Laboratory
PO Box 45, 2280 AA Rijswijk, The Netherlands

ABSTRACT

1
API, CMA and CCPS have issued guidelines for facility siting. The modeling of the conse-
quences and the effects of the accidental release of dangerous materials is an important issue with
respect to facility siting. Use of \easy to apply} models often is an oversimplification of reality
and in many cases will overpredict the effects leading to unnecessarily large safety distances.
A typical example is a vapor cloud explosion. Most explosion models still use the simplified
TNT-equivalency approach although it is common knowledge that the typical characteristics of a
vapor cloud explosion is modeled unsatisfactorily. Use of more sophisticated multi-dimensional
models have the potential to assess each case individually and will lead therefore to optimization
of the lay-out of a facility with respect to safety.
Such sophisticated models have been developed by Arthur D. Little Inc. (ADL) and TNO-PML
and cover the whole sequence of release phenomena, dispersion, explosion, blast propagation,
blast interaction, and structural response.
This paper presents a case study demonstrating the benefit of applying this chain of models when
evaluating the potential impact to both onsite and offsite buildings.

1
This paper appeared in the proceedings of International Conference and Workshop on Process Safety Management
and Inherently Safer Processes, October 1996, AIChE/CCPS, Pages 293-315
 2

BACKGROUND

Facility siting standards and guidelines have historically been developed to protect against fires
and explosions. The current focus is still on fire/explosion protection, but the tools available to
assess potential impacts, particularly for explosions, have improved. Historical standards such
as FM 7-44 ”Spacing of Facilities in Outdoor Chemical Plants” published in 1975 provided
equipment spacing guidelines for explosions based on an equivalent weight of TNT [1]. More
recent guidelines, published by IRI provide plant layout and spacing distances for property loss
prevention purposes only [2]. One of the first and most popular guidelines \DOW’s Fire and
Explosion Index} [3] originally published in 1964, is also focused on property loss and business
interruption.
A CIA publication ”Process Plant Hazard and Control Building Design” was one of the first
guidelines to consider personnel safety in the design of control rooms for not only explosions but
also toxic releases [4].
When OSHA promulgated the PSM Standard under 29 CFR 1910.119, they included a require-
ment for facility siting under the process hazard analysis element. Since OSHA’s main purpose
is protection of the worker, it was not clear that existing standards and guidelines for layout
and spacing designed to minimize property loss would be adequate to protect workers. Some
companies adopted the DOW Fire and Explosion Index. Most companies took a default position
of developing a facility siting checklist for use during PHAs. However, when API RP752 [5]
was published, it became the standard for meeting facility siting requirements for buildings.
The most recent publication on this subject is the CCPS book ”Guidelines for Evaluating Process
Plant Buildings for External Explosions and Fires” [6].

REVIEW OF EXPLOSION METHODOLOGIES

The TNT Equivalence Method [7]

Although the characteristics of the blast wave from a vapor cloud explosion are known to differ
substantially from those resulting from a TNT explosion, it is common in explosion investigations
to establish an ”equivalent TNT yield” for the vapor cloud explosion. The principal reason for
this is that the characteristics of TNT explosions, including the blast effects and the relation of
the overpressure wave to the distance from the center of the explosion and to the charge weight of
TNT have been well established from extensive experiments and weapons tests. These TNT data
can be used, therefore, to estimate the quantity of explosive material involved in the accidental
release, or the effective source strength of the vapor cloud explosion.
 3

The equivalent TNT yield is based on two factors: first, the ratio of the heat of combustion
of the combustible gases in the vapor cloud to the heat of detonation of TNT; and, second, the
efficiency of the vapor cloud explosion.
An equivalent mass of TNT is calculated using the following equation:
MT NT ¢Hc
= £ Ef (1)
Mcloud 1155

where:

MT N T = TNT equivalent mass, (kg)

¢Hc = Lower heat of combustion, kcal=kg
Mcloud = Mass in cloud, (kg)
Ef = Efficiency

The distance to a given overpressure is then calculated from the equation:

h i
1=3
X = 0:3967 £ MT N T exp 3:5031 ¡ 0:7241 ln(OP ) + 0:0398(ln OP )2 (2)

where:

X = Distance to given overpressure, m

OP = Peak overpressure, psi

The first of these is simply the ratio of the total energies available per unit mass of material. For
most hydrocarbon materials, this ratio is about ten. Thus, on a mass basis, a hydrocarbon release
has ten times as much potential explosive energy as TNT.
The second factor relates to how well or efficiently the vapor cloud behaves as an explosive
material upon ignition. The efficiency is dependent upon several conditions, including the quantity
released, the rate of release, the direction of release, degree of confinement, wind and atmospheric
stability at the time of the release, the time between the initial release and ignition of the cloud,
and degree of turbulent mixing of the released vapor with air.
Unlike TNT or other high explosives which are designed specifically to produce a highly efficient
explosion, the dispersed vapor cloud even under the most favorable conditions will be relatively
inefficient. A study of over one hundred vapor cloud explosions which have occurred in recent
years led to the graph shown on Figure 1, which indicates the relative frequency of occurrence
of values of the efficiency (and the TNT equivalent, assuming an energy ratio equal to 10) of
these explosive events. As the Figure indicates, efficiencies of three to four percent are most
common, while an efficiency range of one to ten percent accounts for a significant percentage of
the explosions studied.
Another important factor in defining the efficiency and the equivalent yield parameters is related
to the percentage of the total released material which should be included in the calculation. If
 4

Figure 1: TNT equivalence of unconfined vapor cloud explosions

(See [8])

the vapor cloud is dispersed extensively before ignition occurs, then some of the vapor is diluted
below the lower flammable limit, and some, near the source of the release, may be above the
upper flammable limits. Therefore, only that portion of the vapor cloud within these limits will
be within the flammable range.
Historically, the estimates of UVCE efficiencies, such as those shown in Figure 1, have been
based on the total mass released. Obviously, for most, if not all, of these events, the calculated
efficiency would increase if the mass of the flammable cloud were used in the calculation, and
it is likely the wide scatter of efficiency values shown in Figure 1 is at least partly explained by
the wide ratio of the actual flammable mass to the total released mass of these events.
Although it would be preferable to estimate explosion efficiencies on the basis of the flammable
mass, such a process introduces still another assumption into the calculations, since for most
accidental releases, the total amount released is more readily estimated than the amount in the
flammable range.
As the methods and models available for dispersion analyses improve, and thus the estimation of
the flammable mass of a hydrocarbon release becomes more accurate, then consideration should
be given to developing estimates of the UVCE efficiency on the flammable mass.
The basic TNT curve relating the peak overpressure of the blast wave from a TNT explosion
to the scaled distance parameter, in accordance with the Hopkinson scaling, is shown on both
Figures 2 and 3. This curve is based on a series of experiments and weapons tests carried
out over many years, and represents the peak overpressure, for a TNT explosion on the ground
surface, which generates an overpressure blast wave which radiates into a hemispherical space.
 5

Figure 2: Free-field overpressure vs. scaled distance: TNT and UVCE explosions

Figure 3: Free field overpressure vs. scaled distance for different spreading assumptions
 6

Little is known about the peak overpressure (or, of the entire time-history) resulting from va-
por cloud explosions. Based on explosion damage analyses together with some small-scale ex-
periments and theoretical analysis, it is thought that the corresponding curve for vapor cloud
explosions is approximately similar to the basic TNT curve with the following exceptions:

1. Within the vapor cloud, it is considered likely by many investigators that the peak over-
pressure does not exceed 15 psi. At the edge of the cloud, the overpressure is likely to be
10 psi or somewhat less.
2. Outside the cloud, the decay in overpressure is assumed to follow generally the trend
indicated for the basic TNT curve. On Figure 2, a decay curve is shown for a vapor
cloud efficiency of three percent, a value frequently assumed for design and analysis. On
Figure 3, a curve for ten percent efficiency is shown; this curve is superimposed on the
TNT curve, on the basis that the ratio of the total energies of the hydrocarbon vapor cloud
and an equal weight of TNT is ten.
3. In the far-field, with peak overpressures of, say, 0.5 psi or less, it is considered likely
that the vapor cloud overpressures decay less rapidly than the TNT overpressures. In this
region, the vapor cloud overpressure blast wave behaves more like an acoustic wave, thus
decaying by geometric spreading only. The consequence of this is that higher values of
overpressures occur from vapor cloud explosions than from TNT explosions of equivalent
weight at far locations.

On Figure 3, acoustic wave spreading is shown for the far-field assumption, for all distances be-
yond the 1.0 psi overpressure, and, for comparative purposes, for the case where the overpressure
decay is only due to acoustic-type spreading losses at all locations outside the cloud itself.
The difference between TNT curve decay and acoustic-type decay in the far-field is significant
in damage analyses. For example, a 0.2 psi damage estimate (say, by window damage analysis)
at a range of 5000 feet implies a equivalent charge of 20 tons if the TNT curve is used, and an
equivalent charge of less than 5 tons if the acoustic decay curve is used.
Thus the manner in which the peak overpressure of the blast wave from a UVCE decays with
distance, particularly in the far-field, is important when far-field locations are used in damage
analyses.
Unfortunately, the decay characteristics are not well understood in this region, and the usual
analysis methodology is to assume that the TNT curve applies. As indicated by the examples,
however, this approach can significantly underestimate damage in the far-field, or, alternatively,
lead to excessive estimates of equivalent yields from specific damage.
With the substantial uncertainties of the characteristics of the peak overpressure from a vapor
cloud explosion both at close-in ranges (greater than 10 psi) and at far-field ranges (less than
0.5 psi), the most useful comparison of vapor cloud explosions and TNT explosions are at mid-
distances, corresponding to, 0.5 psi up to 10 psi with the least uncertainty in the narrower range
of about 0.5 psi to 3 psi. It is in this range that damage analysis is considered most appropriate
for estimating the source strength of a vapor cloud explosion.
 7

TNO Multi-Energy [9]

Despite many fundamental objections [10], TNT-equivalency methods are widely used for simple
vapor cloud explosion blast modeling. Presently, however, almost 15 years after its formulation,
the TNO Multi-Energy method is increasingly accepted as a more reasonable alternative [6]. The
Multi-Energy concept is based on the starting point that, assuming deflagrative combustion, the
explosive potential of a vapor cloud is primarily determined by only the obstructed and/or partially
confined areas in the cloud. Separate areas produce separate blasts. So, contradictory to more
conventional methods, in which a vapor cloud explosion is regarded as an entity, according to the
Multi-Energy concept a vapor cloud explosion is rather defined as a number of sub-explosions
corresponding with the various partially confined, obstructed areas in the cloud.
The Multi-Energy concept, derived by simple reasoning, was confirmed by experimental obser-
vations (see [11, 12, 9, 13]).
Applying the Multi-Energy method, a blast should be modeled by the specification of an equivalent
hemi-spherical fuel-air charge, which has two characteristics, namely:
² A charge size, which is directly related to the heat of combustion of the flammable mixture
actually contributing to the blast, and
² A charge strength which is defined as the maximum explosion overpressure produced.

According to TNO recommendations, the charge characteristics can be specified following a

simple safe and conservative approach, namely:

² The charge energy should be taken equal to the full heat of combustion of the flammable
mixture present within the partially confined, obstructed area in the cloud, assuming that
the fuel is stoichiometrically mixed with air.
² The charge strength is assumed to be maximum.
² If the assumption of maximum strength results in unacceptable overestimates of blast effects,
the approach may be refined on the basis of correlation with experimental data. To this end,
a data base is being developed in the form of graphs (see [14]). The data base is accessible
through evaluating a combination of parameters. This parameter combination characterizes
the major initial and boundary conditions which largely determine the development of gas
explosions.

Having specified the fuel-air charge energy and strength, the blast wave’s side-on overpressure
and duration at any distance from the charge can be read from fuel-air blast charts shown in
Figures 4 and 5.
The charts are fully non-dimensionalized with atmospheric overpressure Po , the charge heat of
combustion E and the ambient speed of sound co . The blast charts have been compiled for a
heat of combustion Hc = 3:5 MJ=m3 , which is an average value which approximately holds for
the most common stoichiometric hydrocarbon-air mixtures.
 8

Figure 4: Dimensionless positive phase duration vs. combustion energy scaled distance

AutoReaGasT M [15]

Using blast charts for the description of blast wave decay, an implicit assumption is that the blast
effects are fully point symmetric. In reality, however, gas explosions are often highly directional
and far from point symmetric, in particular gas explosions of high strength. The consequence
is that near field blast from gas explosions is often far from symmetric. Differences in blast
overpressure of an order in magnitude have been experimentally observed. On the other hand,
a strong blast wave has a tendency to symmetry by which the effects of asymmetry will reduce
with increasing distance from the source.
The use of blast charts may result in blast loading estimates which are often not more than an
order of magnitude accurate. Such inaccuracies are particularly severe for results calculated for
locations close to the blast source. If this accuracy is unacceptable, numerical simulation with
computational fluid dynamic (CFD) methods is an increasingly feasible alternative.
AutoReaGas is a CFD software package that consists of two CFD codes: a gas explosion sim-
ulator and a blast simulator, integrated in an interactive and user-friendly environment. The gas
explosion simulator is capable of simulating the turbulent premixed combustion process in gas
explosions, the origin of blast effects. Given the blast source characteristics, the blast simulator
is capable of computing the propagation of the blast wave in the vicinity of the explosion and
the interaction with objects.
The gas dynamics of a gas explosion is simulated by numerical solution of a full set of con-
servation equations which constitutes a model for the gas dynamics, the turbulence and the fuel
distribution. The propagation and interaction of blast with structures is simulated by numeri-
cal solution of conservation equations which constitute a model for inviscid gas dynamics. A
Flux-Corrected Transport scheme is used to capture and preserve shocks.
 9

Figure 5: Dimensionless maximum ”side on” overpressure as a function of combustion energy

scaled distance
 10

Figure 6: AutoReaGas representation of plant layout. For ease of survey, plant main structure as
well as internal equipment are only partly visualized

CASE STUDY

Scenario Layout

A failure in a high pressure process vessel leads to the discharge of 14,000 kgs of butane over
a period of ten minutes. The process layout is shown in Figure 6. The plant consists of two
identical process units separated by a 12 m wide lane. The units consist of a large number of
pieces of equipment interconnected by a maze of piping, mounted in two-story concrete structures
covering an area of 20x50 m2 each. A dense vapor cloud of butane spreads in the plant in a
shallow 3 m thick layer. Some buildings are located at a distance of approximately 200 m from
the plant.

Results: TNT-equivalence

Table 1 shows predicted overpressure levels for 14,000 kg of butane at two different efficiencies.
The EPA offsite consequence analysis [16] guidance provided in support of EPA’s risk manage-
ment program rule recommends a value for efficiency of 10 % for a worst case scenario and an
 11

Table 1: Overpressure predictions using TNT equivalence

Overpressure Distance (m) Distance (m)

(kPa) 3 % efficiency 10 % efficiency
75 58 88
50 73 110
40 83 125
30 98 148
20 127 190
10 202 301

efficiency of 3 % for an ”alternate” scenario.

Results: TNO Multi-Energy Method

According to the Multi-Energy concept only the partially confined, obstructed areas in the vapor
cloud actually contribute to the blast. The only significant confinement and obstruction on the
site is found in the process plant present in process area 1. The space in the plant is partially
confined by the solid concrete structures and obstructed by the process equipment. The energy
and strength of an equivalent fuel-air charge can be determined following safe and conservative
assumptions as recommended by TNO.
² The charge energy is equal to the heat of combustion of the vapor present within the
partially confined and obstructed area, i.e. the concrete structures which cover an area of
50x20 m2 each. The two process units are separated by a lane of only 12 m width. It
is not clear so far if this width is sufficient to treat the units as separate blast sources.
Therefore, in a first approach the two units are considered together as being one single
blast generative area. The dense vapor cloud is assumed to spread in a shallow 3 m thick
layer. As a consequence, the equivalent charge volume equals 52x50x3 = 7,800 m3 . With
a heat of combustion of 3.5 MJ=m3 , the charge energy E equals 7,800x3.5 = 27,300 MJ.
The specific heat of combustion of butane equals 43 MJ/kg. These figures show that not
more than approximately 650 kg of the total quantity of butane present in the entire cloud
actually contributes to the blast.
² The charge strength is assumed to be maximum.

These assumptions make it possible to give dimensions to the scaled parameters in the blast
charts and to read the blast wave decay in terms of peak side-on overpressure and wave duration
dependent on the distance from the charge. For some overpressure levels the results are tabulated
in Table 2. An atmospheric pressure of 100 kPa and an ambient speed of sound of 330 m/s give
dimensions to the scaled parameters. Because the obstructed and partially confined conditions
within the plant are the origin of the blast, the plant is the blast epicenter.
The data in Table 2 show that in this particular case the TNO Multi-Energy method results in
 12

Table 2: Blast overpressure and wave duration for various distances from an equivalent fuel-air
charge, according to the TNO Multi-Energy method

Overpressure Scaled Distance Scaled Wave Wave Duration

(kPa) Distance (m) Duration (ms)
75 0.8 52 0.19 37
50 0.98 64 0.21 41
40 1.1 71 0.22 44
30 1.3 84 0.25 49
20 1.8 117 0.3 59
10 2.9 188 0.35 69

slightly less heavy blast effects than TNT-equivalency. This small difference in outcome between
both methods is certainly not a systematic difference. The two methods are based on starting
points which are fundamentally different. In the Multi-Energy approach the equivalent charge
is only determined by the conditions in the plant as well as by the size of the plant. In the
TNT-equivalency approach, on the other hand, the equivalent charge weight is only dependent on
the total quantity of fuel in the cloud. Therefore, the differences in result of the two methods will
vary from case to case, dependent on the plant layout and the assumed scenario. If, for instance,
the quantity of butane released were much smaller but just sufficient to fill the process units in
the plant, the result of the TNO Multi-Energy method would be the same. TNT-equivalency, on
the other hand, would largely underestimate the blast effects.

Results: AutoReaGas software

Gas Explosion Analysis

A gas explosion analysis in much more detail is possible by numerical simulation with the
AutoReaGas software. An incident environment can be configured in detail by the interactive
specification of boxes, vessels, beam, tubes, cable trays, rigid walls, gratings and vent panels on
the computer screen. Figure 6 shows the equipment in process area 1, configured and visualized
with the AutoReaGas software. Only the larger pieces of equipment are visible. In the plant the
dense flammable vapor spreads in a shallow 3 m thick layer on the earth's surface. Ignition is
assumed somewhere outside the process plant.
The development of the ensuing flame propagation process can be monitored on the screen.
Figures 7 and 8 show a selection of consecutive screen pictures in both a horizontal and a
vertical cross-section through the point of ignition. The timing of the pictures shows that the
flame propagation process is slow initially and speeds up under the influence of the partial
 13

confinement and the equipment present inside the plant.

Any wanted process parameters, such as pressures, gas mixture compositions, densities, temper-
atures or velocities can be recorded throughout the domain during the full simulation. Figure 9
shows, for instance, the overpressures recorded in 5 different stations indicated in Figure 7. The
high explosion overpressures show that the conditions inside the plant give rise to the development
of a powerful gas explosion. The figures justify the high charge strength which was a safe and
conservative starting point in the application of the Multi-Energy method. Substantially lower
overpressures, on the other hand, would have given occasion to a revision of the Multi-Energy
analysis.
However, the simulation results show much more detail. The Figures 7, 8 and 9 indicate, for
instance, that this vapor cloud explosion produces two separate blasts: one in the first unit near
station 5 and one in the other unit near the locations 3 and 4. This result suggests that in the
Multi-Energy blast analysis, the two process units could be regarded as separate areas producing
separate blasts.
In addition, the analysis shows that the vapor cloud explosion has a very strong degree of
directionality. The high blast overpressure recorded in station 3 is hardly observed in station 1.

Blast Analysis

The strong directionality in the blast wave near the source disappears with propagation distance.
The heavier the blast, the stronger is the tendency to symmetry. Therefore, the Multi-Energy
fuel-air blast charts can be used for extrapolation of the blast wave parameters to about 200 m
distance where some buildings are situated. At this distance, the blast wave overpressure has
decayed down to approximately 10 kPa and the wave duration has grown up to approximately
70 ms (Table 2).
Blast loading of an object is a complicated function of blast wave properties and object size and
shape. Full simulation of the interaction of the blast flow field with an object allows to study
the phenomena in detail and to compute blast loading. The AutoReaGas software allows an
interactive and user-friendly specification of a realistic incident environment on the screen. The
environment constitutes the boundary conditions for the gas dynamics in the blast simulator.
Firstly, the blast loading of one single building is considered. A blast wave of 10 kPa overpressure
and 70 ms duration coming in from the left boundary of the mesh serves as blast simulator input.
The development of the blast flow field in interaction with the building can be monitored on the
screen. Figure 10 shows, for instance, two pictures showing the pressure field in the vicinity of
the building at two consecutive points of time. The pressure field is visualized by means of a set
of iso-pressure lines. Shocks are present where the iso-lines accumulate.
Overpressures are recorded in three different locations on the building and represented in Fig-
ure 12. The blast wave overpressure doubles on reflection on the front wall (P1). The overpressure
build-up at the rear wall (P3) is clearly the result of the passage of two waves, the primary wave
diffracting around the building and the same wave after ground reflection. As can be expected,
the rear wall experiences an overpressure loading substantially lower than the top side and the
 14

Figure 7: AutoReaGas screen pictures showing the flame propagation process in a horizontal
cross section at some consecutive point of time
 15

Figure 8: AutoReaGas screen pictures showing the flame propagation process in a vertical cross
section at some consecutive point of time
 16

Figure 9: Overpressure-time records observed in 5 different stations

front of the building.

The overpressure build-up observed at the rear wall changes considerably if a second building is
situated closely behind the first. The pictures in Figure 11 show the complicated wave pattern
that develops in between the two buildings. Now the maximum overpressure observed at the rear
wall of the first building is much higher (P3 in Figure 13). This is due to an extra contribution
from the reflection of the primary blast wave at the front of the second building. The front of
the second building experiences a substantial reduction in blast loading because of the sheltering
presence of the first. The overpressure build-up in between the buildings as well as the sheltering
effect appears to be highly dependent on the blast wave duration and the spacing between the
two buildings.

The Effect of Mitigating Measures

The outcome of this case study so far is that the assumed scenarios lead to a powerful vapor cloud
explosion which will fully destroy the process plant and will cause severe damage to buildings
up to a substantial distance. This may be unacceptable. A variety of changes in the plant design
may be considered which reduce the plant’s explosive potential. Configuration of equipment as
well as the plant layout may be changed in a variety of ways, e.g.,:
² The interspace between the two process units could be further enlarged. This change in the
design would fully separate the two units as blast generators. They would produce separate
blast waves of less than half the energy content.

² The solid concrete floors could be replaced by grated floors or even fully omitted. This
change would reduce the degree of confinement and thereby the blast generative capacities
 17

of the process units.

² The density of equipment and piping could be diminished. This change would reduce the
obstruction thereby the blast generative capacities of the process units.

Exercising with the AutoReaGas software demonstrates that each of these changes in the design
reduces the explosive potential of the plant. A combination of the proposed measures reduces
the gas explosion hazard even down to acceptable levels in the assumed scenarios.

CONCLUSIONS

A vapor cloud explosion hazard analysis was carried out using three different methodologies,
namely: TNT-equivalency, TNO Multi-Energy and CFD simulation with the AutoReaGas soft-
ware.
TNT-equivalency and TNO Multi-Energy are highly simplified methods which make use of
blast charts. Blast charts represent the blast wave parameters dependent on the distance from
an equivalent explosive charge. TNT-equivalency methods equate an equivalent TNT-charge
proportionally to the full quantity of fuel present in a vapor cloud. The value of the proportionality
factor is based on incident statistics. The TNO Multi-Energy method equates equivalent fuel-air
charges to the fuel present within any of the partially confined and obstructed areas in the cloud.
Both methods were demonstrated in a realistic case study.
In particular, close to the blast source these methods fail if a better accuracy is required than just
an order in magnitude. For the determination of blast effects in the direct vicinity of a vapor cloud
explosion, more sophisticated methods are required. AutoReaGas is a software package capable
of detailed 3D numerical simulation of many aspects of the vapor cloud explosion problem. It
consists of an explosion simulator and a blast simulator, placed in a interactive and user-friendly
environment.
Application of the software to the case demonstrated that the confinement by the solid concrete
structures and obstruction by the process equipment in the plant are the origin of explosive,
heavy blast generating combustion. In addition, it revealed details in the development of the
combustion process which largely determine near-field blast effects. The simulation showed that
as a consequence of the plant layout, the vapor cloud explosion produced two separate blasts, one
in each process unit. This result suggests that in the Multi-Energy analysis, the process units may
be treated as separate blast sources. By application of the AutoReaGas software, it is possible to
quantify the mitigating effects of changes in site layout as well as in plant design.
 18

Figure 10: AutoReaGas screen pictures showing blast wave interaction with one single building
 19

Figure 11: AutoReaGas screen pictures showing blast wave interaction with two buildings
 20

Figure 12: Overpressure-time records observed in 3 different stations during blast wave interaction
with one single building

Figure 13: Overpressure-time records observed in 6 different stations during blast wave interaction
with two buildings
REFERENCES 21

References
[1] Factory Mutual Engineering Corporation. Spacing of facilities in outdoor chemical plants.
Loss Prevention Data 7-44, 1975.

[2] Industrial Risk Insurers. Plant layout and spacing for oil and chemical plants. IRI Information
IM.2.5.2, 1991.

[3] American Institute of Chemical Engineers. DOW’s fire and explosion index hazard classi-
fication guide, 1994.

[4] Chemical Industries Association. Process plant hazard and control building design an ap-
proach to categorization, 1995.

[5] Chemical Manufacturers Association and American Petroleum Institute. Management of

hazards associated with location of process plant buildings, API recommended practice
752, CMA managers guide, first edition, 1995.

[6] Center for Chemical Process Safety of the American Institute of Chemical Engineers. Guide-
lines for evaluating process buildings for external explosions and fires, 1996.

[7] G. A. Melhem and P. A. Croce. Advanced Consequence Analysis: Emission, Dispersion,

Fire and Explosion Dynamics. Arthur D. Little Inc., 1996.

[8] H. Giesbiecht. Evaluation of vapor cloud explosions by damage analysis. Journal of

Hazardous Materials, 17:247–257, 1988.

[9] R. J. Harris and M. J. Wickens. Understanding vapour cloud explosions - an experimental

study, 55th autumn meeting of the institution of gas engineers, kensington, UK, 1989.

[10] A. C. Van den Berg and A. Lannoy. Methods for vapour cloud explosion blast modelling.
J. of Haz. Mat., 34:151–171, 1993.

[11] J. P. Zeeuwen, C. J. M. Van Wingerden, and R. R. Dauwe. Experimental investigation into

the blast effect produced by unconfined vapor cloud explosions. In 4th Int. Symp. Loss
Prevention and Safety Promotion in the Process Industries. Harrogate, UK, IChemE Symp.
Series 80, pages D20–D29. IChemE, 1983.

[12] A. J. Harrison and J. A. Eyre. The effect of obstacle arrays on the combustion of large
premixed gas/air clouds. Comb. Science and Techn., 52:121–137, 1987.

[13] C. J. M. Van Wingerden. Experimental investigation into the strength of blast waves gener-
ated by vapour cloud explosions in congested areas. In 6th Int. Symp. Loss Prevention and
Safety Promotion in the Process Industries, Oslo, Norway, pages 26–1/26–16, 1989.

[14] A. C. Van den Berg and J. B. M. M. Eggen. GAME - guidance for the application of
the multi-energy method. In The Second International Specialist Meeting on Fuel-Air
Explosions Bergen, Norway, June 27/28, 1996.
REFERENCES 22

[15] TNO/Century Dynamics Ltd. Autoreagas theory manual and user documentation, 1994.

[16] Environmental Protection Agency. Offsite consequence analysis guidance, (January 22),
1996.