TNO, CCR 01318 Application of Correlations To Quantify The Source Strength

ONGERUBRICEERD
TNO report Application of correlations to quantify the

PML 1998-C53 source strength of vapour cloud explosions
in realistic situations.
Final report for the project: ‘GAMES’
TNO Prins Maurits Laboratory
Lange Kleiweg 137 Date

P.O. Box 45
2280 AA Rijswijk
October 1998
The Netherlands
Author(s)
Phone +31 15 284 28 42
Fax +31 15 284 39 58
W.P.M. Mercx
A.C. van den Berg
D. van Leeuwen
Assignor : Air Liquide, France

BP International Ltd, United Kingdom
ENEL Spa CRIS, Italy
Elf Atochem, France
Gaz de France, France
Health and Safety Executive, United Kingdom
ICI, United Kingdom
INERIS, France
Norsk Hydro, Norway
RIVM, The Netherlands
Snamprogetti SpA, Italy
Title : Ongerubriceerd
All rights reserved. Abstract : Ongerubriceerd
No part of this publication may be
reproduced and/or published by print, Report text : Company Confidential
photoprint, microfilm or any other means
without the previous written consent of Annexes A - G : Company Confidential
TNO.
In case this report was drafted on No. of pages : 156 (incl. annexes,
instructions, the rights and obligations of excl. documentation page)
contracting parties are subject to either the
Standard Conditions for Research No. of annexes : 7
Instructions given to TNO, or the relevant
agreement concluded between the
contracting parties.
Submitting the report for inspection to
parties who have a direct interest is All information which is classified according to Dutch regulations shall be treated by
permitted. the recipient in the same way as classified information of corresponding value in his
own country. No part of this information will be disclosed to any party.
 1998 TNO The classification designation Ongerubriceerd is equivalent to Unclassified.
ONGERUBRICEERD
TNO Prins Maurits Laboratory is part of

TNO Defence Research which further consists of:
TNO Physics and Electronics Laboratory Netherlands Organization for

TNO Human Factors Research Institute Applied Scientific Research (TNO)
TNO report
ONGERUBRICEERD
PML 1998-C53 3
1 Summary
Correlations were derived in the preceding GAME project to quantify the source
strength of a vapour cloud explosion required to apply the Multi-Energy Method
for the determination of blast characteristics. The correlations relate a set of para-
meters describing the obstacle configuration in which the flammable cloud is
present and the fuel, to a single value for the overpressure in the exploding vapour
cloud.
This project investigates the difficulties and problems encountered while applying
the correlations to a number of realistic scenarios. The objective is to provide
guidance and recommendations on how to overcome these difficulties and to
decide on the actual values to be chosen for the parameters of the correlations in
specific situations. The emphasis is on the determination of the parameters: ‘Volu-
me Blockage Ratio’ and ‘Average Obstacle Diameter’.
The main finding is that a safe approach in most situations is to apply the procedu-
re of the new Yellow Book for the determination of the volume of the obstructed
region in combination with the hydraulic average obstacle diameter and a flame
path length equal to the radius of a hemisphere with a volume equal to the volume
of the obstructed region.
Lack of experimental data on specific items prevents the generation of more detai-
led guidance. Some guidance is developed based on a theoretical approach, to
assess the influence of the aspect ratio of the obstructed region and to quantify the
separation distance between multiple explosion sources. It is recommended to
perform an experimental research programme to generate the required data to
improve and validate the suggested procedures.
ONGERUBRICEERD
TNO report
ONGERUBRICEERD
4 PML 1998-C53
2 Samenvatting
In het voorafgaande GAME project zijn correlaties afgeleid waarmee de

bronsterkte van een gasexplosie kan worden bepaald. Deze is nodig bij het toepas-
sen van de Multi-Energie methode voor het bepalen van de blast-karakteristieken.
De correlaties relateren een aantal parameters dat de obstakelconfiguratie waarin de
gaswolk zich bevindt en de brandstof beschrijft, met een enkele waarde voor de
overdruk in de exploderende gaswolk.
In het onderhavige project wordt onderzocht welke moeilijkheden en problemen
zich voordoen bij het toepassen van de correlaties op een aantal realistische situa-
ties. Het doel is te komen tot adviezen en aanbevelingen om deze moeilijkheden
aan te pakken en de parameters van de correlaties te kunnen quantificeren in speci-
fieke situaties.
Het belangrijkste resultaat is dat voor de meeste situaties een veilige aanpak bestaat
uit het toepassen van de procedure uit het nieuwe Gele Boek voor het bepalen van
het volume van ruimte waarin de obstakels zich bevinden in combinatie met de
gemiddelde hydraulische obstakeldiameter en een vlampadlengte gelijk aan de
straal van een halfbol die een volume heeft gelijk aan dat van de ruimte waarin zich
de obstakels bevinden.
Een gebrek aan experimentele gegevens betreffende enkele specifieke aspecten is
de oorzaak voor het niet kunnen geven van meer gedetaileerde adviezen. Enkele
adviezen zijn gegeven op basis van een theoretische beschouwing, voor het bepalen
van de invloed lengte/breedte-verhouding van het volume waarin zich de obstakels
bevinden en voor het bepalen van de scheidingsafstand tussen meerdere explosie-
bronnen. Aanbevolen wordt om een experimenteel onderzoeksprogramma uit te
voeren waarmee de vereiste gegevens worden verkregen zodat de adviezen verbe-
terd en gevalideerd kunnen worden.
ONGERUBRICEERD
TNO report
COMPANY CONFIDENTIAL
PML 1998-C53 5
Contents
Summary................................................................................................................... 3
Samenvatting ............................................................................................................ 4
1 Introduction ............................................................................................... 7
2 Background and objectives ....................................................................... 9

2.1 Characterisation of source strength............................................ 9
2.2 Objectives and approach .......................................................... 11
3 Considerations on the application of the GAME correlation to

realistic situations.................................................................................... 13
3.1 Choice of a correlation............................................................. 13
3.2 Determination of Vor ............................................................... 14
3.3 Determination of D .................................................................. 16
3.4 Determination of Lp ................................................................. 17
3.5 Determination of SL ................................................................. 19
4 AutoReaGas calculations ........................................................................ 21

4.1 REAGAS ................................................................................. 21
4.2 BLAST..................................................................................... 22
4.3 Adopted approach .................................................................... 22
5 Application to the Chemical Plant case................................................... 23

5.1 Description of case................................................................... 23
5.2 Reduction of problem size ....................................................... 28
5.3 Application to reduced problem............................................... 28
5.4 Application to another reduced problem.................................. 32
5.5 Combination of two obstructed regions................................... 35
5.6 Application to whole case........................................................ 37
5.7 Influence of detail of obstacle description ............................... 41
5.8 Overall evaluation of the Chemical Plant case ........................ 43
6 Application to the LNG Terminal case ................................................... 47

6.2 First impression of potential explosion severity ...................... 50
6.3 Application of correlation to obstructed subregion 1............... 51
6.4 Application of correlation to obstructed subregion 2............... 57
6.5 Application of the correlation to a combination of
obstructed subregion 1, 2 and 5 ............................................... 61
6.6 Application of the correlation to a combination of
obstructed subregions 1, 3, 6, 7, and 8..................................... 63
TNO report
6 PML 1998-C53
6.7 Blast outside obstructed regions .............................................. 66

6.8 Overall evaluation of the LNG Terminal case ......................... 70
7 Application to the Gas Processing case................................................... 75

7.2 Test performed ......................................................................... 75
7.3 AutoReaGas calculation........................................................... 80
7.4 Application of correlation to obstructed region ....................... 81
7.5 Blast outside obstructed region ................................................ 82
7.6 Evaluation and conclusion ....................................................... 82
8 Application to the Hydrogen case ........................................................... 85

8.2 Application of correlation and ARG simulations..................... 85
8.3 Evaluation and conclusion ....................................................... 89
9 Overall evaluation and guidance obtained .............................................. 91

9.1 General..................................................................................... 91
9.2 The correlation and parameters................................................ 92
9.3 Guidance and remaining white spots ....................................... 94
9.4 Possible extensions of the blast charts ..................................... 98
10 Conclusions and recommendations ......................................................... 99
11 References ............................................................................................. 101
12 Acknowledgement................................................................................. 103
13 Authentication ....................................................................................... 105
Annexes:
A Procedure for the determination of the boundaries of the
obstructed region according to the Yellow Book
B Procedure for the application of the Multi-Energy
Method according to the Yellow Book
C Application of procedure to determine obstructed region
boundaries
D Critical separation distances between obstructed areas
E Application of GAME correlation to obstacle
configurations of high aspect ratio
F AutoReaGas pressure histories for the various situations
simulated with the Chemical Plant case
G AutoReaGas pressure histories for the various situations
simulated with the LNG Terminal case
TNO report
PML 1998-C53 7
3 1 Introduction
The Multi-Energy Method (MEM) is a rather simple and practical method of

determining the blast parameters from a vapour cloud explosion. It is generally
accepted that the concept of the MEM better represents the specific character of a
vapour cloud explosion. MEM-like methods should therefore be preferred above
practical methods based on TNT equivalency.
The application of the MEM is hindered by a lack of guidance concerning the
choice of the source strength. The GAME project: ‘Guidance for the Application of
the Multi-Energy Method’ was performed to provide this additional guidance
(Eggen, 1995). Very specific guidance was given in the form of correlations. A
relation was derived for a set of parameters describing the obstacle configuration
and the fuel, and the overpressure in the vapour cloud explosion.
The follow-up GAMES project: ‘Guidance for the Application of the Multi-Energy
Method, Second phase’ was initiated to investigate the applicability of the derived
guidance to realistic cases.
This report is the final report of the GAMES project.
First, the background and objectives of the project are presented in Chapter 2.
While applying the correlations to determine the source overpressure to be used in
the Multi-Energy Method, values for the parameters of the correlation have to be
chosen. This introduces a number of specific questions. Considerations with res-
pect to the quantification of the parameters is the subject of Chapter 3. Also, a
number of white spots are identified. Initial thoughts on approaches to deal with
these deficiencies are presented in the same chapter.
No realistic cases are available for which accurate enough data on overpressure
occurring in a vapour cloud explosion exist. In order to be able to evaluate and to
compare the results of the application of the correlations to realistic cases, a refe-
rence was requested. This reference was obtained by applying a numerical code to
generate data on overpressures. Chapter 4 briefly describes how that reference set
of data was obtained.
The correlations were applied to four realistic cases. The exercises that were per-
formed for each case are described in the successive chapters, 5, 6, 7 and 8.
In two of the cases, the Chemical Plant case in Chapter 5 and the LNG Terminal in
Chapter 6, a number of exercises were performed on interesting subsets of the
obstacle configuration.
Chapter 7 deals with a large-scale experiment on a realistic obstacle configuration
typical of a gas-processing site for which some data is available. Chapter 8 deals
with a specific part of the LNG Terminal case, but filled with a flammable hydro-
gen mixture in order to investigate the influence of reactivity.
TNO report
8 PML 1998-C53
Each of the Chapters 5, 6, 7 and 8 contains a final paragraph on the evaluation and
conclusions for that specific case.
Chapter 9 contains an overall evaluation of the exercises performed and provides
guidance to determine values for the parameters of the correlations. Also the identi-
fied white spots are discussed and guidance to deal with these white spots is pre-
sented and discussed.
Finally, Chapter 10 summarises the conclusions. Recommendations are given to
generate specific experimental data in order to be able to develop models to take
into account the influence of the aspect ratio of the obstacle configuration and of
the separation distance between obstacle configurations.
TNO report
PML 1998-C53 9
4 2 Background and objectives
4.1 2.1 Characterisation of source strength
In order to apply the blast charts of MEM, one requires values for two parameters
characterising the source, namely the overpressure P0 and the total combustion
energy E which contributes to the explosion.
Figure 1 shows the blast chart of the MEM. In order to determine the peak blast
overpressure Ps at a distance r from the centre of the explosion, a scaled distance r’
has to be calculated according to:
r
r′ = (1)
(E )1 / 3
p0
in which p0 is the ambient overpressure.
The blast chart provides a value for the scaled blast overpressure Ps’. The blast
overpressure Ps is obtained by multiplying Ps’ by p0.
The overpressure in the explosion P0, the pressure for scaled distance values smal-
ler than r0’, determines which line to follow to choose the correct overpressure at
the required scaled distance.
4.1.1 Overpressure
Two correlations were derived in the GAME project to determine a value for the
overpressure in a vapour cloud explosion. The overpressure is correlated to a set of
parameters characterising the environment in which the vapour cloud is located and
the vapour cloud itself. The difference between the two correlations is due to the
type of confinement of the vapour cloud.
For low ignition energy and no confinement (open, 3D), the expression is:
P0 = 0.84 ⋅ ( VBR ⋅ L p / D )2.75 ⋅ S L2.7 ⋅ D 0.7 (2)
with:
P0 the maximum explosion overpressure (bar)
VBR the volume blockage ratio (-)
Lp length of the flame path (m)
D typical diameter (m)
SL laminar burning velocity of flammable mixture (m/s)
TNO report
10 PML 1998-C53
95374-5.8a
10
10
9
scaled peak 'side on' overpressure Ps'
1 7
0.1 4
0.01 1
0.001
0.1 1 10 100
ro'
combustion energy-scaled distance r'
Figure 1: Blast chart MEM for overpressure.
For low ignition energy and confinement between parallel plates (2D):
P0 = 3.38 ⋅ ( VBR ⋅ L p / D )2.25 ⋅ S L2.7 ⋅ D 0.7 (3)
4.1.2 Combustion energy

The recommendation to obtain a value for the combustion energy is to calculate the
combustion energy of those parts of the flammable mixture which are located in
obstructed regions. An evaluation of experimental data performed in GAME reve-
TNO report
PML 1998-C53 11
aled that the recommendation of taking 100% of the energy of the obstructed part
of the cloud is conservative for low overpressures. The term ‘efficiency’ was
introduced, defined as the percentage of the energy of the obstructed part of the
cloud which contributes to the generation of blast. It appeared that the efficiency is
lower than 20% for overpressures below 0.5 bar.
4.2 2.2 Objectives and approach
The objective of the GAMES project is to apply the correlations to three realistic
cases in order to investigate which problems are encountered while doing so.
The correlations were derived from experiments in which the parameters in the
correlation were well-defined. This is not the case in realistic situations.
The purpose of the GAMES work programme is especially to investigate which
values to choose for the total volume of the obstructed region, Vor (which determi-
nes VBR), and for the diameter D.
The cases were selected based on availability of appropriate information and the
level of detail available on the plant lay-out. The plant lay-out constitutes the so-
called obstacle configuration which is used to determine the Vor and D and which
is used as part of the input of a numerical model to simulate explosions inside the
obstacle configuration.
A numerical model is necessary to create a reference for the results of the correlati-
ons. In general, accidental explosions are not monitored. The overpressure and
blast from an accidental vapour cloud explosion can only be estimated from dama-
ge analysis.
In the selection procedure for the three cases, it was investigated whether the
information of the well-known explosions at DSM in Beek (1975) and in Flix-
borough (1974) could serve as a reference. This was not the case. The information
on the exact plant lay-out was too coarse to be used.
An important practical issue in the selection of the cases was the availability of the
plant lay-out in digital format. Creating such a digital lay-out from scratch would
absorb too much of the available budget.
The next three cases were selected:

• the Chemical Plant case, because the plant lay-out was available in digital
format, which could be used directly as an input to the numerical code and be-
cause of its compact lay-out;
• the LNG Terminal case, because of its complicated and stretched lay-out and
because of the relative ease of creating a suitable digital file;
• the Gas Processing case, because of the availability of the digital file and
because of the availability of experimental data on overpressure and blast.
TNO report
12 PML 1998-C53
An additional fourth case was added to the programme in order to investigate a

high reactive fuel. It was decided to select a part of the LNG Terminal case as the
obstacle configuration for the fourth case and to fill it with hydrogen. The fourth
case is referred to as:
• the Hydrogen case.
During the execution of the project it was decided to put effort into two additional
items that were considered important.
4.2.1 1 Detail of obstacle description

In important issue in numerical simulations of gas explosions is the level of detail
with which the obstacle configuration is described. The results of the JIP project
(Selby and Burgan, 1998) demonstrate that calculated pressures increase if the
level of detail increase. In order to investigate the influence small obstacles have on
the results of the simulations, additional runs were made using adapted obstacle
databases for the Chemical Plant case.
4.2.2 2 Blast overpressure

The objective of the project was to investigate especially the determination of an
average obstacle diameter and the volume of the obstructed region in order to
obtain a good prediction for the source overpressure.
As the main purpose of the MEM is to correctly predict blast outside the gas explo-
sion, the influence of the correlation parameters on blast characteristics is of im-
portance. This influence will be investigated in the LNG Terminal case.
TNO report
PML 1998-C53 13
5 3 Considerations on the application of the GAME

correlation to realistic situations
While applying the correlation in a realistic situation, some basic questions arise. In
trying to find an answer to these questions, other, more detailed, questions come to
one’s mind. The intention is to go through this procedure and try to find answers to
all those questions. Some answers may be obvious or can be derived from current
knowledge; others may need a more practical solution. Some answers may not be
found due to lack of knowledge.
The basic questions with which to start are:

• which correlation do I have to apply?
• which value for VBR do I have to take, or better, which value for Vor?
• which value for D do I have to take?
• which value for Lp do I have to take?
• which value for SL do I have to take?
The remainder of this chapter consists of separate sections for some considerations
on each of these questions.
5.1 3.1 Choice of a correlation
The adopted approach in GAME to relate explosion overpressure to the explosion

parameters resulted in a need for six correlations; one correlation for each combi-
nation of low and high ignition energy with 1D, 2D and 3D expansion (for instan-
ce, jet ignition was considered a high ignition energy source). Within GAME, only
two correlations could be derived.
White spot 1: the missing correlations could not be derived due to a lack of suffi-
cient experimental data.
Proposed approach: although there are only two correlations, these are expected
to cover most situations. By applying the available ones to the cases to be conside-
red, an impression may be obtained about the necessity to develop the remaining
four.
In many realistic situations the expansion will be partially 2D or 3D. For instance,
part of the gas cloud will be underneath a roof or floor. An important ratio to
consider then is the ratio of the horizontal dimensions of the cover and the height
underneath. If this ratio approaches unity then the main part of the expansion
process is three-dimensional. Also, the shape of the flame will be spherical rather
than cylindrical during most of the explosion process.
White spot 2: a criterion to choose for the 2D or 3D correlation is lacking.
TNO report
14 PML 1998-C53
Proposed approach: An exact value for the height over length ratio in order to
have a criterion to separate the application of the two correlations cannot be given.
One might expect though that the ratio should be between 5 and 10.
5.2 3.2 Determination of Vor
A formal procedure for the determination of the volume of the obstructed region is
given in the third edition of the ‘Yellow Book’ (CPR 14E,1997). The completely
revised Yellow Book now contains the MEM for determination of blast from
vapour cloud explosion.
5.2.1 Yellow Book procedure

The Yellow Book contains a procedure to apply the MEM step-by-step. One of the
steps to be taken is the determination of the volume of the so-called obstructed
region. The obstructed region is the region where obstacles are located, so the
region is to be considered as a potential source for overpressure generation in a
vapour cloud explosion. There are two ways to define an obstructed region, either
to estimate its boundaries or to follow an alternative procedure.
The procedure for the MEM according to the Yellow Book is given in Annex B.
The procedure for determination of the obstructed region is given in Annex A.
The procedure attempts to describe in a formal way a number of subsequent rather

obvious steps one would probably take by intuition.
The procedure to determine the obstructed region starts with an initial obstructed
region containing the obstacles located near an assumed ignition point. Depending
on the distance to obstacles further away in relation to the size of the obstacles near
the ignition location, the initial obstructed region is extended to include those
obstacles further away.
White spot 3: the criteria given in the Yellow Book to decide whether an obstacle
belongs to an obstructed region cannot be motivated objectively.
Proposed approach: in most situations however it will be more or less obvious
and the rough boundaries of an obstructed region can be estimated.
It is explicitly stated in the Yellow Book that the criteria to decide whether or not a
specific obstacle belongs to the obstructed region are questionable and further
research is required to better define them. Also, the Yellow Book procedure does
not distinguish multiple explosion sources. If the procedure results in multiple
obstructed regions, one has to add all the volumes together to get one large explo-
sion source.
White spot 4: the separation into multiple explosion sources, often referred to as
the donor-acceptor problem, cannot be solved yet.
TNO report
PML 1998-C53 15
Proposed approach: the new Yellow Book offers a procedure to reduce the com-
bustion energy in situations where one expects multiple explosion sources. More
consideration on separation distances is given in Annex D. A first attempt to quan-
tify a separation distance is given. A separation distance between two obstructed
regions should be at least half the linear dimension of the donor-obstructed region
to consider the donor and acceptor obstructed region as separate potential blast
sources.
The procedure to determine the volume of the obstructed region starts with the
definition of a box containing all obstacles in the obstructed region. The volume of
that box is the initial volume of the obstructed region. As this box probably con-
tains a lot of free space without obstacles, the volume can be reduced by excluding
this free space. The initial box is then replaced by a number of adjacent boxes, the
total volume of which is smaller than the initial box. This optimisation can be
repeated until all free space is eliminated; however, the time involved for the
optimisation may not be worthwhile.
Annex C contains an example of the application of the procedure to determine the

volume of the obstructed region.
5.2.2 Application of procedure in GAMES

For GAMES, the volume of the obstructed region is specifically required in the
determination of the VBR.
VBR is defined as the ratio of the total volume of the obstacles inside an obstructed
region Vob and the volume of that obstructed region Vor.
If the value of Vor is chosen too large, VBR will be too low and the correlation
provides a too low overpressure P0 and vice versa.
The correlations provide a possibility to derive a criterion for the optimisation of

Vor according to the Yellow Book procedure: given two values for Vob: Vor,1 and
Vor,2 the ratio of the overpressures according to the correlation (1) is (it is assumed
that Lp does not change):
2.75
P0 ,2  Vor ,1 
=  (4)
P0 ,1  Vor ,2 
If, for instance, Vor,2 equals 0.86Vor,1, the overpressure P0,2 is 50% higher than
P0,1. An increase of 10% in overpressure is obtained when the volume of the
obstructed region is reduced by 3%.
The accuracy of the correlation is about ±30%. With this value in mind, it is not
profitable to take the next step in the reduction of Vor if it cannot be reduced more
than 10%.
TNO report
16 PML 1998-C53
The volume of the obstructed region influences the value of the energy E to be
used in MEM. This influence is of less importance than its influence on VBR. A
variation of ∆E implies an influence of ∆E1/3 on the distance at which a certain
blast overpressure will occur.
5.3 3.3 Determination of D
The determination of D in a realistic case is not obvious. Various definitions for the
average obstacle size are possible.
To facilitate the definition of various means, we limit ourselves to obstacle confi-

gurations consisting solely of tubes of various diameters and lengths.
• Arithmetic mean weighted by the tube length Li
Darm =
∑ Li ⋅ Di (5)
∑ Li
Harmonic mean weighted by the tube length Li . Relative to the arithmetic mean,
the harmonic mean overweighs the smaller diameters.
1
1
∑ Li ⋅ D
= i
(6)
Dham ∑ Li
• Many other formulations are imaginable, for instance a mean on the basis of the
concept of hydraulic diameter, which is defined as 4 times the ratio between the
summed volumes and the summed surface areas of an object distribution.
Dhym = 4
∑Vi (7)
∑ Ai
This definition overweighs the larger objects in the collection.
The expressions to determine D give a single average value for the whole obstruc-
ted region under consideration, assuming a homogeneous distribution of obstacle
types and obstacle diameters. In many cases the obstructed region will consist of a
number of subregions, each with a typical obstacle distribution. In those cases it
may be more appropriate to calculate D (and VBR) for each subregion and combi-
ne these to a single value for the whole obstructed region.
An attempt to describe the framework of such a procedure was given by Eggen in
the GAME project (Eggen, 1995).
An issue not discussed yet is how to model non-cylindrical obstacles. The experi-
ments underlying the correlations were all with cylindrical obstacles. Non-
TNO report
PML 1998-C53 17
cylindrical obstacles will produce another turbulence field than that of cylindrical
obstacles. It is yet not clear to what extent the shape of an obstacle will influence
the explosion progress.
White spot 5: influence of non-cylindrical obstacles.

Proposed approach: non-cylindrical obstacles like boxes and plates will be repre-
sented by cylinders having a length equal to the largest dimension of the obstacle
and a cross-section area equal to the cross-section area of the obstacle.
5.4 3.4 Determination of Lp
The ignition location in the experiments used to derive the correlations was always
in the centre of the symmetrical obstacle configuration. A value for Lp is then easy
to determine. In other situations, this may not be the case. Three situations were
identified:
• an obstacle configuration with an aspect ratio other than 1;
• an ignition location outside the centre of the obstacle configuration;
• an obstacle configuration partially filled with a flammable mixture.
5.4.1 Aspect ratio

The aspect ratio of an obstructed region can be defined as its length/width or
length/height ratio. In the case of central ignition, there are three values to choose
for Lp: half the length, half the width or half the height. This issue is discussed in
Annex E.
White spot 6: which Lp to choose in the case of non-point symmetrical situations.

Proposed approach: a practical, though not validated solution is proposed in
Annex E. In case there are more Lp’s possible, it is stated that the overpressure at
the moment the flame travelled Lp should be higher than a specific threshold in
order to further increase. If the overpressure is lower than this threshold, side
venting will prevent further flame acceleration and increase in overpressure. The
value proposed for the threshold is 30 kPa. Note that there is no information avai-
lable to support that choice.
5.4.2 Ignition location

The discussion for non-central ignition is similar to the discussion on the aspect
ratio. An extreme situation is present if the ignition location is at an edge or corner
a of the obstructed region; then side (back) venting starts immediately after igniti-
on, possibly preventing any flame acceleration. The conclusion then is that edge or
corner ignition always results in lower overpressures than central ignition.
On the other hand, the available flame path length for edge ignition is greater than
for central ignition. So although flame acceleration is slow initially in the case of
an edge ignition, higher overpressures may be obtained due to the longer flame
TNO report
18 PML 1998-C53
path. Flame acceleration may still occur later in the process when the flame front is
inside the obstacle configuration where back venting has less influence.
5.4.3 Large Vor relative to volume of flammable cloud

Until now, we considered situations were the flammable cloud was larger than the
obstructed region. The correlations were based also on tests simulating that situati-
on.
In most realistic situations, the obstructed region (chemical unit, offshore installati-
on) will not be completely filled with a flammable cloud. A heavy gas cloud will
have a limited height, say 4 to 10 m, while the height of the obstructed region will
be greater. Also, a limited release will result in a small cloud relative to the size of
the obstructed region.
It can be argued that the overpressure in a situation with an obstructed region larger
than the flammable cloud will be higher than in the situation where the volume of
the flammable cloud is the same but the volume of the obstructed region is reduced
to the volume of the flammable cloud.
It was demonstrated in GAME (Eggen, 1994) that the amount of combustion

energy contributing to blast generation is lower than 100% for low overpressures.
At the present wa can only give the following explanation for the existence of an
efficiency factor: a fraction of the unburned mixture is pushed outside the obstacle
array due to expansion. Therefore, not all the gas that is initially inside the array
will burn inside the array. If the flame speed inside the array is high (also high
overpressure), the fraction that is pushed outside will have a high turbulence level.
If the flame speed is low, the turbulence level outside the array will be low.
The turbulence level outside the array will decrease rapidly. One may expect that
when the flame leaves the array and starts to burn through the mixture outside, the
turbulence level of that mixture in the case of high overpressure will be higher than
in the case of low overpressure. So more of the expelled mixture will burn outside
when the overpressure is higher and contribute to the explosion energy.
In the case of an extended obstructed region, the expanded unburned gas will burn
inside the obstructed region instead of being expelled outside. The result is that the
efficiency will increase, but as the length of the flame path is longer, the overpres-
sure will be greater too.
The combination of Lp/D in the correlation is in fact a measure of the number of

obstacles the flame passes while it burns through the mixture.
In situations of a larger obstructed region it may therefore be better to determine Lp
from the expanded flammable cloud size rather than from the size of the obstructed
region inside the unburned cloud.
White spot 7: influence of obstructed region being larger than the vapour cloud.
Proposed approach: due to the lack of any experimental back-up, any influence is
neglected.
TNO report
PML 1998-C53 19
5.5 3.5 Determination of SL
It is common use to assume a homogeneous stoichiometric flammable cloud in all

assessments. Also most available experimental data on explosion overpressures are
based on experiments using homogenous mixtures. The laminar burning velocities
of these mixtures is known. A huge benefit using the correlation is that now va-
rious gases can be compared for their explosion overpressure potential.
Using homogeneous mixtures will result in high overpressures, which may be non-
realistic. There is, however, no guidance available on how to treat non-homogene-
ous (so more realistic) mixtures.
TNO report
20 PML 1998-C53
TNO report
PML 1998-C53 21
6 4 AutoReaGas calculations
In order to create a reference with which the results of the application of the corre-
lations can be compared, numerical simulations have to be performed for three of
the four cases. The code AutoReaGas is used as the numerical simulator (van den
Berg et al., 1995). Actually the code AutoReaGas consists of two numerical sol-
vers. The solver REAGAS is for modelling the combustion, expansion and turbu-
lence phenomena inside the exploding vapour cloud. The solver BLAST is for
modelling the propagation of the blast wave outside the combustion zone and the
interaction of the blast wave with structures.
6.1 4.1 REAGAS
The basic mechanism of a gas explosion consists of the interaction of a premixed

combustion process with its self-induced expansion flow field. The development of
this process is predominantly controlled by the turbulent structure of the flow field,
which is induced by the boundary conditions. Modelling of a gas explosion requi-
res careful modelling of all aspects of this complicated process. The model under-
lying the AutoReaGas gas explosion simulator can be characterised as follows.
• The gas dynamics is modelled as a perfect gas which expands as a consequence
of energy addition. This is mathematically formulated in conservation equations
for mass, momentum and energy.
• The energy addition is supplied by combustion which is modelled as a one-step
conversion process of flammable mixture into combustion products. This is
formulated in conservation equations for the fuel mass fraction and the compo-
sition. The combustion rate is a source term in the fuel mass fraction conserva-
tion equation.
• Turbulence is modelled by a two parameter model (k-ε) which consists of
conservation equations for the turbulence kinetic energy k and its dissipation
rate ε.
• Turbulent combustion is modelled by an expression which relates the combusti-
on rate to turbulence. Several options are available varying from theoretical re-
lations such as the Eddy Break Up model and the Eddy Dissipation model to
experimental correlations between turbulence and combustion. Because the cell
size in many applications is often far too large to fully resolve a turbulent com-
bustion zone, the combustion rate is corrected using a calibration factor Ct.
• The initial stage of combustion upon ignition is modelled by a process of lami-
nar flame propagation whose speed is controlled on the basis of experimental
data.
• Objects too small to be represented by solid boundaries in the computational
mesh, are modelled by a subgrid formulation. The presence of a subgrid object
is modelled by the specification of appropriate flow conditions, i.e. a fluid dy-
namic drag and a source of turbulence.
TNO report
22 PML 1998-C53
• Numerical solution of the set of equations is accomplished by means of the

‘power law’ scheme applied within a finite volume approach.
6.2 4.2 BLAST
As long as objects with large cross-flow dimensions are considered, the interaction
of gas explosion blast is predominantly governed by the pressure wave character of
a blast wave and the drag component can be neglected. The pressure wave charac-
ter of blast flow fields is accurately represented by inviscid flow. Often, blast flow
fields are characterised by the presence of gas dynamic discontinuities such as
shocks. Modelling of blast-object interaction requires careful description of such
phenomena. Therefore, the blast simulator models blast-object interaction as fol-
lows.
• The gas dynamics is modelled as inviscid compressible flow of a perfect gase-
ous fluid which can be formulated as the conservation equations for mass, mo-
mentum and energy for inviscid flow, i.e. the Euler equations.
• Description of shock phenomena requires a sophisticated numerical technique
tailored to proper representation of steep gradients. To this end, the blast simu-
lator utilises Flux-Corrected Transport (FCT). FCT makes an optimised use of
numerical diffusion so that steep gradients present in shocks are retained. Nu-
merical diffusion is added only where it is required for numerical stability.
6.3 4.3 Adopted approach
W have chosen to use a single calibration of the code for all simulations.
Large-scale validation has been performed recently in the Joint Industry Project
(Selby and Burgan, 1998). We applied the calibration factor for the combustion
model Ct = 65 in combination with a cell size of about 1 m and got acceptable
results for all tests in JIP. The same combination has been applied for all GAMES
simulations with AutoReaGas.
In all cases, we let the code automatically decide whether an obstacle has to be
modelled as a ‘solid’ or as a ‘subgrid’.
TNO report
PML 1998-C53 23
7 5 Application to the Chemical Plant case
7.1 5.1 Description of case
The digital file containing the chemical plant to be assessed in GAMES, describes
a volume of width 32 m (x-direction), length 57 m (y-direction) and height 39 m
(z-direction).
Various parts can be distinguished:
1 a two storey concrete structure consisting of columns and floors containing
lots of equipment, vessels and pipelines, of various dimensions and orientated
in all three perpendicular directions (x: 8.5-25 m, y: 18.5-48.5 m, z: 0-15 m,
volume: 7425 m3); a stack is located on top of this structure with the outlet at
z: 33 m;
2 a large three storey pipebridge with numerous pipes running in the y-direction
(x: 0.5-8.5 m, y: 0-57 m, z: 0-7.5 m, volume: 3420 m3);
3 a small two story pipebridge with pipes orientated in the x-direction (x: 8.5-
32.5 m, y: 16.5-18.5 m, z: 0-5 m, volume: 240 m3);
4 a single floor concrete support structure (columns and floor) containing some
vessels (x: 8.5-18 m, y: 9-16.5 m, z: 0-12 m, volume: 855 m3);
5 two stacks with supply equipment; the equipment is bounded by x: 8.5-
12.5 m, y: 2-9 m and z: 0-7.5 m, volume: 210 m3, the outlet of the stacks is at
z = 39 m.
6 a single floor concrete structure containing some vessels (x: 8.5-17 m, y: 48.5-
56.5 m, z: 0-12 m, volume: 816 m3).
The numbers given correspond with the numbers in Figure 2. Other numbers in
Figure 2 denote ignition locations (IL-i) and pressure sampling locations (P-i) to be
used elsewhere in this chapter.
TNO report
24 PML 1998-C53
Figure 2.a: View of the Chemical Plant case: overall view.
TNO report
PML 1998-C53 25
95374-2b
P10
P13
6
iL-3 iL-9
P3 P6
1A 1B
2 P9
P2 P5
iL-1 iL-2
iL-8
iL-6 iL-5
P1 P4
P8
3
P14
P12
4
5
P11
P7
iL-7
Figure 2.b: View of the Chemical Plant case: horizontal projection on xy-plane.
TNO report
26 PML 1998-C53
95374-2c
Figure 2.c: View of the Chemical Plant case: vertical projection on xz-plane.
TNO report
PML 1998-C53 27
95374-2d
Figure 2.d: View of the Chemical Plant case: vertical projection on xy-plane.
TNO report
28 PML 1998-C53
7.2 5.2 Reduction of problem size
As this case is the first case to be considered in GAMES, it should be reduced and
simplified.
It was decided to isolate the two storey concrete structure from the rest of the
installation and to consider it a separate case for a first assessment in the GAMES
exercise because:
• the borders of the obstructed region can be rather well defined;
• it consists of a variety of different sized, shaped and orientated obstacles;
• confinement between two surfaces.
The third value is of importance as one of the two correlations derived in GAME is
applicable to the two-dimensional expansion.
The first floor of the obstructed region is only partially covered by the floor of the
second storey.
To simplify the problem more, we will assume that the obstructed region is filled
with a homogeneous stoichiometric mixture of methane in air. The cloud has a
height of 7.5 m, so the ground level is completely filled with the cloud.
The first floor is supported by a large number of stiffeners, which do not have a
physical thickness in the submitted database. This large number of stiffeners will
largely determine the determination of D. For reasons of simplification, these
stiffeners are neglected and the floor is considered smooth in this case.
There are also floors present inside the configuration. These floors are supported
by stiffeners also, which are taken into account in this case. They are given a
thickness of 0.01 m.
The simplified case, obstructed subregion number 1 (OSR-1), now consists of the
volume between:
x: 8.5 - 25 m, width 16.5 m;
y: 18.5 - 48.5 m, length 30 m;
z: 0 - 7.5 m, height 7.5 m.
The volume is filled with a variety of obstacles, and is confined between two
surfaces: the ground floor and the first floor.
7.3 5.3 Application to reduced problem
We will now try to come up with a value for the explosion overpressure in the case
of ignition of the flammable mixture using the correlations and keeping the consi-
derations of Chapter 2 in mind to answer:
• which correlation do I have to apply?
• which value for VBR do I have to take, or better, which value for Vor?
• which value for D do I have to take?
TNO report
PML 1998-C53 29
• which value for Lp do I have to take?

• which value for SL do I have to take?
7.3.1 Correlation?
At first glance, the explosion process will be governed by a 2D expansion. Howe-
ver, the length to height ratio equals 30/7.5= 4 or 16.5/7.5 = 2.2. This implies that a
large part of the process is 3D.
At this stage it is not clear which correlation to choose. The AutoReaGas calculati-
on of the case should provide more insight.
7.3.2 VBR?
At this stage we will not go into detail to accurately determine the volume of the
obstructed region. It will be obvious that a good estimate for the volume of the
obstructed region is the product of its dimensions:
• Vor = 16.5 x 30 x 7.5 = 3712.5 m3.
The obstructed volume contains 760 objects, consisting of 618 cylinders and 142
boxes.
The VBR is calculated to be:
• VBR = 0.10.
7.3.3 D?
The objects in the obstructed region contain cylinders and boxes. To calculate a
diameter according to equations 6 and 7, the ‘diameter’ of a box should be calcu-
lated. For that, we calculated the diameter of a circle having the same surface area
as the surface of the box perpendicular to its longest dimension.
Applying equations 5, 6 and 7 resulted in:
• Darm = 0.23 m;
• Dham = 0.08 m;
• Dhym = 0.49 m.
7.3.4 Lp?
There are a number of possibilities for Lp.
The correlations are derived for central ignition. The longest path the flame can
travel for central ignition is Lp,1 = 15 m. Assuming symmetric shape, the flame
reaches the boundary of the obstructed region already after Lp,2 = 8.25 m because
of the length over width ratio is not equal to unity. It may be expected that using
Lp,1 will result in a too large overpressure, as venting has already started before the
flame reaches the short edge. Using Lp,2 will result in a too low value for the
overpressure, as combustion inside the obstructed region is not completed at the
moment the flame reaches the long edge.
Ignition may also occur elsewhere. The longest path for the flame is Lp,3 = 30 m
when ignition occurs in the centre of one of the edges of the obstructed region. As
explained in Chapter 2, using the correlations for edge ignition will result in too
high overpressures, as venting will start immediately after ignition.
TNO report
30 PML 1998-C53
7.3.5 SL?
The choice of this parameter in the present exercise is straightforward. The corre-
lations were derived with SL = 0.45 m/s for methane, so
SL = 0.45 m/s
7.3.6 Overpressure?
Combinations of possible values for the parameters involved result in Table 1.
Table 1: Combination of parameters considered.
P0 (kPa)
2D 3D
Lp 8.25 m 15 m 30 m 8.25 m 15 m 30 m
Darm = 0.23 m 250 450 4502 120 600 4060

Dham = 0.08 m 1270 1890 2320 1010 5260 3537
Dhym = 0.49 m 80 294 1400 25 128 861
7.3.7 AutoReaGas
Numerical simulations were performed with the CFD code AutoReaGas (ARG)
using the obstacle configuration defined in this section. Various ignition locations
were used:
Case: Ignition location: x(m): y(m): z(m):
NH01 IL1 8.5 33.5 4 centre of large edge
NH02 IL2 25 33.5 4 centre of large edge
NH03 IL3 8.5 48.5 4 corner
NH12 IL8 16.5 33.5 4 centre
NH13 IL9 16.5 48.5 4 centre of small edge
The pressure was sampled in six locations, P1 -P6, equally distributed over the
obstructed region. The pressure histories are presented in Annex F. An overview of
the maximum overpressures in these locations is given in Table 2; the highest value
per case is printed in bold.
TNO report
PML 1998-C53 31
Table 2: Maximum overpressures according to ARG simulations.
Case: NH01 NH02 NH03 NH12 NH13

Sample location (x, y, z) Pmax (kPa)
P1 (13, 22, 4) 18 36 13 47 55
P2 (13, 36, 4) 24 34 11 41 23
P3 (13, 47, 4) 26 59 5 65 9
P4 (22, 22, 4) 36 21 14 51 51
P5 (22, 36, 4) 40 19 22 42 34
P6 (22, 47, 4) 37 35 8 73 10
average 30 34 12 53 30
The table shows the highest pressures for central ignition.

Ignition in the centre of an edge results in lower overpressures. In spite of the
longer flame path compared to central ignition, the immediate back-venting after
ignition prevents flame acceleration into the obstructed region. Maximum over-
pressures occur in locations farthest away from the ignition location where the
flame speed is the fastest.
The surplus of back-venting in the case of corner ignition results in the lowest
overpressures.
7.3.8 Evaluation
The first question which arises when comparing the numerical results with the
correlation overpressure is how to perform the comparison. The correlation gives a
single value, while numerical simulation will obviously result in a distribution of
maximum overpressures.
The correlation was derived based on an average of maximum overpressures meas-
ured during the experiments. One may wonder how to average, as a localised high
overpressure peak may contribute less to the blast than the lower overpressure in a
larger subvolume of the obstructed region. A definite answer cannot be given yet.
Therefore, the correlation results will be compared with an average value of the
maximum overpressures per case (last row of Table 2).
White spot 8: how to conclude on a single value for the overpressure

Proposed approach: an arbitrary number of pressure sampling locations is chosen,
homogeneously distributed in the obstructed region. The average value of the
maximum of all locations is denoted as ‘the overpressure’.
Table 3: Overpressures for various Lp.
P0 (kPa)
2D 3D
Lp 8.25 m 15 m 30 m 8.25 m 15 m 30 m
Dhym = 0.49 m 180 294 1400 25 128 861
TNO report
32 PML 1998-C53
In comparison with the correlation results, the only reasonable results are predicted
with the largest average diameter: Dhym. The maximum average AutoReaGas value
(53 kPa) is in between the 3D correlation result for Lp = 8.25 and 15 m.
As could be expected based on the considerations in section 2.1, the best result is
obtained with the 3D correlation, and this correlation gives too low a value for the
smaller Lp and too large a value for the larger Lp.
An optimal value for central ignition may be expected for an averaged Lp. An
average value may be obtained by assuming the flammable cloud to be a hemisphe-
re with radius Lp and having the same value. In that case, Lp = 12.1 m. The corre-
lation (1) provides now P0 = 71 kPa using Dhym (P0 = 330 kPa for Darm, P0 = 2910
kPa for Dham).
While the AutoReaGas calculation would result in a class 6 for the MEM, the 3D
correlation with a conservative estimate for Lp would result in a class 7.5, and an
average value for Lp would result in a class 6.5.
7.4 5.4 Application to another reduced problem
Another typical part of the Chemical Plant case is the large pipebridge that runs
along the whole length of the total obstructed region.
We will consider the pipebridge a separate and isolated case, as this type of ob-
structed region is present in many industrial situations.
Obstructed subregion number 2 (OSR-2) occupies the space between x: 0-8 m,

y: 0-57 m and z: 0-8 m.
7.4.1 Correlation?
Neither of the two available correlations seems suitable to cover this case. The 2D
correlation is not applicable as there are no two confining surfaces. The density of
the layers of pipes might give the impression of a confining surface, they also may
act as turbulence generators. Although 3D expansion is possible, the obstacle
configuration as well as the aspect ratio of the obstructed region differs considera-
bly from those that were used to derive the 3D correlation. Nevertheless, taking the
3D correlation seems to be the best bad choice.
7.4.2 VBR?
The boundaries of the obstructed region are quite clear. An optimisation procedure
according to Annex A will therefore not be performed.
The boundaries are x:0.5-8.5 m, y: 0-57 m, z: 0-8 m. The volume Vor equals
3648 m3 resulting in a volume blockage ratio:
VBR = 0.14
TNO report
PML 1998-C53 33
7.4.3 D?
The obstructed region contains 505 objects, consisting of 30 cylinders and 475
boxes.
We calculated that:
• Darm = 0.50 m;
• Dham = 0.41 m;
• Dhym = 0.53 m.
7.4.4 Lp?
We will consider the following values for the length of the flame path:
• Lp = 28.5 m central ignition, half the length;
• Lp = 4 m central ignition, half the width or height;
• Lp = 12 m central ignition, radius of hemisphere with equal volume;
• Lp = 8 m ignition in centre long edge, width;
• Lp = 57 m ignition in centre short edge, length.
7.4.5 SL?
Again we will take SL = 0.45 m/s.
7.4.6 Overpressure?
Combinations of possible values for the parameters involved result in the following
table (Table 4).
Table 4: Overpressures for various parameter combinations.
P0 (kPa)
Central ignition Long edge Short edge
ignition ignition
Lp 28.5 m 4m 12 m 8m 57 m
Darm = 0.50 m 2300 10 213 70 15500

Dham = 0.41 m 3400 15 315 103 22900
Dhym = 0.53 m 2080 9 192 63 14000
7.4.7 AutoReaGas
Numerical simulations were performed with the obstacle configuration defined in
this section. Three different ignition locations were used:
Case: Ignition location: x(m): y(m): z(m):
NH04 IL5 4.5 28.5 4 centre of obstructed
region;
NH05 IL6 0.5 28.5 4 centre of long edge;
NH06 IL7 4.5 0 4 centre of short edge.
Pressure were sampled in four locations, P7 -P10, equally distributed over the
obstructed region. Pressure histories are presented in Annex F. An overview of the
maximum overpressures in these locations is given in Table 5; the highest value
per case is printed in bold.
TNO report
34 PML 1998-C53
Table 5: Overpressures according to ARG simulations.
Case: NH04 NH05 NH06

P7 (4, 1, 4) 27 13 9
P8 (4, 20, 4) 62 33 9
P9 (4, 40, 4) 50 28 5
P10 (4, 55, 4) 26 13 5
The same trend is visible as in the obstructed region considered in the previous
section.
The table shows the highest pressures for central ignition.
Ignition in the centre of an edge results in lower overpressures. In spite of the
longer flame path compared to central ignition, the immediate back-venting after
ignition prevents flame acceleration into the obstructed region.
The side-, back- and top-venting apparently have a great influence. Maximum
overpressures do not occur in locations farthest away from the ignition location but
in a region around the ignition location. Initially, the flame accelerates, but because
of the large aspect ratio, the expansion is easily vented, resulting in a deceleration
of the flame in the later stage.
7.4.8 Evaluation
Table 6 gives values for the average maximum overpressures.
Table 6: Averaged maximum overpressures according to ARG simulations.
Case Central ignition Long edge ignition Short edge ignition

NH04 NH05 NH06
Average P0 (bar) 41 22 7
Comparing these values with the predictions according to the correlation shows
that none of the predictions are acceptable.
The aspect ratio of the obstructed region is too large to obtain a reasonable answer
by using an average Lp and central ignition. The aspect ratio here is more than 7.
The approach adopted for aspect ratios other than one does not give an acceptable
answer here. The correlation gives 9 to 15 kPa for an Lp of 4 m. This is lower than
30 kPa, so the overpressure is not expected to increase. The overpressure according
to ARG reaches an average value of 41 kPa instead, more than a factor three hig-
her.
It is not possible to conclude on the diameter to apply. The variation of the various
average diameters is too small.
TNO report
PML 1998-C53 35
7.5 5.5 Combination of two obstructed regions
The two obstructed regions considered in the two previous paragraphs form toge-
ther a large part of the total case. It is therefore interesting to investigate the com-
bination of the two before considering the case as a whole.
The volume of the combination of the two obstructed subregions (OSR-1/2) equals:
Vob = 7364 m3. The volume contains 1265 objects, consisting of 648 cylinders and
617 boxes. It can be calculated that:
• VBR = 0.12;
• Darm = 0.37 m;
• Dham = 0.14 m;
• Dhym = 0.51 m.
Table 7: Lp values considered per ignition location.
Ignition location Lp (m) Description
IL-1
4 distance to top of cloud
8 width of OSR-2
16.5 width of OSR-1
33.5 maximum distance to short edge of OSR-2
IL-2
24.5 width of OSR-1/2
41.5 distance to corner (0.5,0,4) of OSR-1/2
IL-5
4 distance to long edge of OSR-2
20.5 distance to long edge of OSR-1
28.5 distance to short edge of OSR-2
IL-7
4 half width of OSR-2
57 length of OSR-2
IL-8
8.5 distance to long edge of OSR-1
16 distance to long edge of OSR-2
37 distance to corner (0.5,0,4) of OSR-1/2
15.2 radius of hemisphere with equal volume
Five ignition locations are considered; all coincide with an ignition location used in
the previous assessments. A number of Lp values can be chosen.
With SL = 0.45 m/s and the value defined above, the 3D correlation results in a set
of overpressures which are collected in Table 8.
TNO report
36 PML 1998-C53
Table 8: Overpressures according to correlation.
P0 (kPa)
Ignition location Lp (m) Darm = Dham = Dhym= AutoReaGas
0.37 m 0.14 m 0.51 m average
IL-1 137
4 10 72 5
8 67 490 34
16.5 500 358 253
33.5 3420 25100 1770
IL-2 142
4 10 72 5
24.5 1450 10620 350
41.5 6170 45200 3200
IL-5 109
4 10 72 5
4 10 72 5
20.5 890 6510 460
28.5 2200 16100 1140
IL-7 17
4 10 72 5
4 10 72 5
57 14780 108300 7650
IL-8 167
4 10 72 5
8.5 79 578 41
16 449 3290 232
37 4500 33000 2330
15.2 390 2860 202
AutoReaGas simulations were performed for the various ignition locations. Pressu-
res were sampled at the same locations used in the simulations of OSR-1 and OSR-
2. Pressure histories are presented in Annex F. Results are presented in Table 9.
The last row of the table shows average maximum overpressures.
The average overpressure is highest in the case of central ignition (IL-8). The
difference in average overpressure for the edge ignition locations is striking. Back-
and side-venting for IL-7 appears to have a strong influence on flame development.
The flame is still so slow when it enters OSR-1 that it cannot accelerate to produce
high overpressures inside OSR-1.
The influence of venting in the case of IL-2 is less pronounced. The flame is able
to accelerate inside OSR-1, producing pressures above 0.5 bar and accelerates even
more inside OSR-2 where the highest overpressures are reached at the edge. The
flame speed inside OSR-2 is high enough to reduce the influence of side-venting.
TNO report
PML 1998-C53 37
Table 9: ARG results.
Case: NH07 NH010 NH08 NH09 NH14

Ignition: IL-1 IL-2 IL-5 IL-7 IL-8
P1 (13, 22, 4) 97 74 63 7 99
P2 (13, 36, 4) 114 77 81 16 99
P3 (13, 47, 4) 166 90 166 30 115
P4 (22, 22, 4) 153 42 103 17 84
P5 (22, 36, 4) 143 36 158 34 75
P6 (22, 47, 4) 178 60 187 19 112
P7 (4, 1, 4) 174 394 70 8 380
P8 (4, 20, 4) 127 154 65 8 238
P9 (4, 40, 4) 61 106 70 11 107
P10 (4, 50, 4) 158 387 125 19 359
Average 137 142 109 17 167
The average overpressure according to ARG does not depend much on the ignition
location. Overpressures are in between 1 and 1.7 bar, except for corner ignition,
where they are much lower.
The results of the correlation are confusing. The dependence on Lp and D is strong.
The best result seems to be a small Lp (for top-venting) in combination with
Dham = 0.14 m, the smallest diameter. This result is, however, in contradiction to
the adopted approach for the aspect ratio.
A fairly good result is provided by taking the radius of an equivalent hemisphere

for Lp = 15.2 m with Dhym: P0 = 202 kPa (MEM class 8), while the maximum of
the average ARG maximum overpressures is 167 (MEM class 7.5).
7.6 5.6 Application to whole case
7.6.1 Determination of parameters

Step 1 to step 5 of the procedure in Annex A leads to a box containing the ob-
structed region. The box has dimensions: length 57 m, width 32.5 m and height 15
m. The height is arbitrary as one has to decide whether some stacks belong to the
obstructed region.
TNO report
38 PML 1998-C53
y (m) 95374-3
2 9 16.5 18.5 48.5 56.5

0.5 x
2 H = 7.5m V = 3420m3
8.5
5
x (m)
H=6m 4 6
12.5 V = 168m3 H = 12m
H = 12m 1a H = 15m V = 3825m3 V = 816m3
V = 1083m3
17
18
3
H = 5m V = 145m3
1b H = 10m V = 2400m3
25
32.5
Figure 3: Obstructed region according to Yellow Book (CPR14E, 1997) procedure.
Following step 6 of the procedure in Annex A and the example in Annex C one
arrives at the shape of the obstructed region as given in Figure 3. Figure 3 shows
the contours of a number of boxes with different heights. The height and volume of
each box is given in the figure. The total volume of the obstructed region is:
Vor = 11857 m3
A more accurate determination of Vor does not seem worthwhile. A further reduc-
tion of 10%, about 1200 m3, cannot be found.
The formal procedure of the Yellow Book leads to almost the same intuitive subdi-
vision into typical obstructed regions as given in 2.2. The total volume of this
intuitive subdivision is Vor = 12966 m3.
According to equation (4), the Yellow Book obstructed volume results in a 28%
higher overpressure using the correlation than the intuitive obstructed region.
The procedure of the Yellow Book does not account for typical obstacle configura-
tions within each box defined in the obstructed region. Therefore, the intuitive
subdivision is preferred.
The intuitive subdivision can be improved by subdividing subregion 1 (see Figu-
re 2) into two parts, because one part consists of two storeys while the other is a
single storey support structure:
1a width 8.5 m (x: 8.5 -17 m), length 30 m (y: 18.5 - 48.5 m) and height 15 m,
volume 3825 m3;
TNO report
PML 1998-C53 39
1b width 8 m (x: 17 - 25 m), length 30 m (y: 18.5 - 48.5 m) and height 10 m,

volume 2400 m3.
The volume of the intuitive obstructed region now reduces by 1200 m3 to
11766 m3.
The VBR and D can now be calculated for each obstructed subregion and for the
obstructed region as a whole (Table 10).
Table 10: Parameters for various obstructed subregions.
Obstructed x (m) y (m) z (m) VBR Darm (m) Dham(m) Dhym (m)
subregion
OSR-1a 8.5-17 18.5-48.5 0-15 0.19 0.41 0.10 1.07

OSR-1b 17-25 18.5-48.5 0-10 0.13 0.20 0.07 0.82
OSR-2 0.5-8.5 0-57 0-7.5 0.14 0.49 0.40 0.52
OSR-3 8.5-32.5 16.5-18.5 0-5 0.03 0.16 0.15 0.16
OSR-4 8.5-18 9-16.5 0-12 0.25 1.16 0.61 1.40
OSR-5 8.5-12.5 2-9 0-7.5 0.11 0.91 0.75 1.03
OSR-6 8.5-17 48.5-56.5 0-12 0.16 0.83 0.08 1.45
SR (total) 0.18 0.41 0.14 0.84
We are considering a problem where the obstructed region is filled with a homoge-
neous and stoichiometric mixture of methane/air up to a height of 7.5 m.
Here, the obstructed region is larger than the flammable cloud. The problem here is
which obstacle parameters to use. This problem was identified as a white spot
(section 2.4). We choose to work with the obstacle parameters of that part of the
obstructed region filled with the cloud. Table 11 contains the obstacle parameters
for the obstructed subregions limited to a height of 7.5 m.
The total volume of the obstructed part filled with the cloud is: Vor = 8627 m3. The
radius of a hemisphere with equal volume is 16 m.
Table 11: Parameters for various obstructed subregions with height limited to height of
cloud.
Obstructed x (m) y (m) z (m) VBR Darm (m) Dham (m) Dhym (m)
subregion
OSR-1a7.5 8.5-17 18.5-48.5 0-7.5 0.12 0.32 0.11 0.56

OSR-1b7.5 17-25 18.5-48.5 0-7.5 0.07 0.17 0.07 0.41
OSR-2 0.5-8.5 0-57 0-7.5 0.15 0.50 0.40 0.52
OSR-37.5 8.5-32.5 16.5-18.5 0-5 0.03 0.16 0.15 0.16
OSR-47.5 8.5-18 9-16.5 0-7.5 0.20 0.95 0.51 1.29
OSR-5 8.5-12.5 2-9 0-7.5 0.11 0.91 0.75 1.03
OSR-67.5 8.5-17 48.5-56.5 0-7.5 0.11 1.01 0.09 0.87
SR7.5 (total) 0.13 0.37 0.14 0.55
Ignition location 1 (IL-1) will be considered for the next Lp:

• Lp1 = 3.5 m, distance to top of cloud;
TNO report
40 PML 1998-C53
• Lp2= 8.5 m, distance to long edge;

• Lp3 =16.5 m, distance to opposite long edge;
• Lp4 = 33.5 m, distance to short edge;
• Lp,ave = 16 m, radius of equivalent hemisphere.
Application of the 3D correlation provides the following results.

Table 12: Overpressures (kPa) for various Lp and D.
Lp (m) Darm = 0.37 m Dham = 0.14 m Dhym = 0.55 m
3.5 7 63 4
8.5 98 721 44
16.5 609 4460 270
33.5 4260 31300 1890
16 595 4100 248
7.6.2 AutoReaGas simulations

The ARG simulation performed with the total obstructed region (simulation NH11)
resulted in an average maximum overpressure of 182 kPa (see Table 13 for details),
which equals an MEM class of about 8. As in previous cases, an acceptable result
is obtained by using the correlation with Dhym and the radius of the equivalent
hemisphere (last row of Table 12): 248 kPa, which is equal to an MEM class of
about 8.5.
Table 13: Comparison of all ARG results.
Pmax (kPa)
Case NH01 NH02 NH03 NH12 NH13 NH04 NH05 NH06 NH07 NH010 NH08 NH09 NH14 NH11
OSR 1 1 1 1 1 2 2 2 1-2 1-2 1-2 1-2 1-2 all
Ignition: IL-1 IL-2 IL-3 IL-8 IL-9 IL-5 IL-6 IL-7 IL-1 IL-2 IL-5 IL-7 IL-8 IL-1
Sample location
P1 18 36 13 47 55 97 74 63 7 99 109
P2 24 34 11 41 23 114 77 81 16 99 123
P3 26 59 5 65 9 166 90 166 30 115 176
P4 36 21 14 51 51 153 42 103 17 84 158
P5 40 19 22 42 34 143 36 158 34 75 144
P6 37 35 8 73 10 178 60 187 19 112 186
P7 27 13 9 174 394 70 8 380 374
P8 62 33 9 127 154 65 8 238 128
P9 50 28 5 61 106 70 11 107 61
P10 26 13 5 158 387 125 19 359 196
P11 239
P12 215
P13 163
P14 270
Average 30.2 34.0 12.0 53.1 30.3 41.3 21.8 7.0 137 142 109 17 167 182
TNO report
PML 1998-C53 41
Table 13 shows the details of the ARG simulation in comparison with the other
cases, the results of which are given in previous tables. The highest overpressures
in NH11 occurred in the corners of the obstructed region. The extension of OSR-
1/2 with OSR 3, 4 and 5 appears to promote flame acceleration in the large pipe-
bridge (OSR-2).
7.7 5.7 Influence of detail of obstacle description
The numerical simulations for the Chemical Plant case and also for the cases to be
presented in the following chapters, were performed using an obstacle database.
Such an obstacle database can never exactly match the obstacle configuration as
built in reality. There will also be details which are not described in the database.
As obstacles play an important role in the gas explosion mechanism, the level of
detail with which the obstacles are described may influence the numerical result.
In order to investigate the effect of detail of obstacle description, a number of

AutoReaGas simulations were performed using the Chemical Plant obstacle data-
base. In subsequent simulations, obstacles with a diameter smaller than a specified
value were omitted.
The first simulation was run using the original database. Subsequent simulations
omitted all obstacles with a diameter of less than 0.1, 0.2, 0.25, 0.3, 0.4, 0.5, 0.6
and 0.7 m.
The results of the numerical simulations are visualised in Figure 4. Overpressures

were sampled at similar locations as in Table 13 including some locations outside
the obstacle configuration. The number of obstacles in each simulations are given
in Table 14.
Even if all obstacles smaller than 0.3 m are deleted, there is hardly any effect on
the overpressures; the influence being stronger for the higher pressures. Overpres-
sures are lowered by about 20% when only obstacles with a diameter larger than
0.6 m are considered. Overpressures become unacceptable in the last simulation
with obstacles larger than 0.7 m only.
Because the overpressure level is rather constant for all simulations with obstacles
smaller than 0.3 m, it can be expected that the overpressures will not increase when
more detail is included in the obstacle database.
This result is in contradiction to the common opinion that adding more small
objects will result in an increase in overpressure. On the other hand, this common
opinion is not fully supported by the physical modelling. The consequence of
leaving out small objects is a reduction of the fluid dynamic drag. This means a
reduction of the fluid dynamic drag leading to a reduction of the turbulence source
term At the same time, a reduction in drag means an increase in the expansion
velocity and, in this way, an increase in the turbulence source term. Thus, there are
competing effects, the balance of which is case dependant.
TNO report
42 PML 1998-C53
Table 14 shows how the omission of small obstacles influence the parameters in
the correlation.
The reduction in overpressures according to the correlation follow the same trend
as the results of the numerical simulations. There is hardly an effect until obstacles
with a diameter smaller than 0.4 m are removed. Until that point the removal of
obstacles did not result in a significant reduction of the total obstacle volume.
Table 14: Influence of obstacle removal.
Removed Number of Total obsta- Percentage of VBR Dhym Percentage of initial P0

obstacles obstacles left cle volume initial volume (-) (m) according to correlation
(m3) (%) (%)
None 1923 2080 100.00 0.18 0.84 100

< 0.1 m 1417 2079 99.95 0.18 0.84 98
< 0.2 m 1049 2069 99.51 0.18 0.87 93
< 0.25 m 975 2064 99.25 0.18 0.89 89
< 0.3 m 703 2061 99.08 0.18 0.90 88
< 0.4 m 471 2030 97.62 0.18 0.93 81
< 0.5 m 433 2024 97.33 0.18 1.02 67
< 0.6 m 162 1973 94.88 0.17 1.09 50
< 0.7 m 125 1486 71.46 0.13 1.48 13
400
P1
P3
P5
P7
300 P9
P11
P13
P15
P17
Overpressure (kPa)
P18
200
100
0
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
Removed objects with sizes less than (m)
Figure 4: Influence of minimal obstacle diameter in obstacle database on overpressure.
TNO report
PML 1998-C53 43
7.8 5.8 Overall evaluation of the Chemical Plant case
7.8.1 General
The Chemical Plant obstacle configuration as a whole can be subdivided into a
number of specific obstructed subregions.
The application of the correlations to the subregions and combination of subregions
showed that a very large band of predictions for the overpressure is obtained for a
specific situation by substituting possible values for all parameters in the correlati-
on. On the one hand, this is due to the various ways in determining an average
diameter; on the other hand, the configuration in combination with the ignition
location differs very substantially from the configurations used to derive the corre-
lation. The correlation is not capable of including the influence of the aspect ratio
and of an ignition location at the edge of the configuration.
The correlation results when using the hydraulic diameter appear to be of the same
order as the numerical results.
Except for the subregion containing the pipebridge, the influence of the ignition
location according to the AutoReaGas calculations is not very strong. Edge ignition
resulted always in lower overpressures than ignition inside the configuration. This
observation may not be valid in other cases where the flame may be accelerated
due to a higher level of obstruction in combination with the longer flame path.
An acceptable result was obtained when the correlation was used with the average
hydraulic obstacle diameter Dhym in combination with an average value for the
flame path length Lp. The average flame path length is equal to the radius of a
hemisphere with a volume equal to the volume of the obstructed region under
consideration.
7.8.2 Comments on the covering of the white spots

1 Missing correlations
The missing correlations concern 1D situations and jet ignition. The available
correlations for 2D and 3D expansion are sufficient for the present case.
2 Criterion to separate between 2D and 3D correlation

The issue is of importance in part of the case only. The aspect ratio of that part
is less than 5. The guidance given in section 2.1 advises one to take the 3D
correlation in that case. The evaluation in section 4.3 shows that the less con-
servative results are obtained indeed by applying the 3D correlation.
3 Definition of obstructed region

The absence of a strict procedure to determine the boundaries of the obstruc-
ted region does not cause problems in this case. By applying the Yellow Book
procedure, there is no problem in this case to decide whether an obstacle be-
TNO report
44 PML 1998-C53
longs to an obstructed region. A minor problem may be caused by the stacks

present in the case. A strict application of the procedure results in adding the
space between two stacks to the volume of the obstructed region. This is con-
sidered to be of no influence in this case.
The intuitive determination of the obstructed region almost completely coinci-
des with the region defined according to the Yellow Book procedure.
4 Multiple obstructed regions

This white spot is not an issue in this case. The obstructed region can be com-
posed of differently shaped obstructed subregions with specific obstacle cha-
racteristics. The overall lay-out however is not such that one could expect mo-
re than a single obstructed region.
5 Differently shaped obstacles

The case contains many non-cylindrical obstacles. The procedure adopted to
cope with non-cylindrical obstacles is to turn these into cylinders with a length
equal to the longest dimension of the obstacle and an equal cross-section area.
There is, at present, no indication about the physically correctness of that pro-
cedure. Especially for flat plates, the procedure is very questionable.
6 Lp in the case of aspect ratios other than unity and ignition location outside
centre of configuration
For aspect ratios differing from unity and central ignition, the value for Lp
may vary between half the height, width or length of the obstructed region.
The additionally provided guidance stated that the overpressure will not incre-
ase anymore after the moment the flame propagated the shortest Lp if the
overpressure is less than a specific value at that moment. The specific value
was arbitrarily chosen to be 30 kPa.
Application of this guidance in this case does not result in much confidence.
The flat cloud in combination with the z-co-ordinate of the ignition location
results in a smallest Lp of 4 m. Overpressures for Lp = 4 m were always very
low, while the final overpressures were much higher.
For edge ignitions, the flame path length will always be longer than in the case
of central ignition. Therefore the correlation will provide higher overpressures
in the first situation, while in reality, lower values will be obtained in the ma-
jority of the situations, due to the influence of (back- and side-)venting. At the
present, it is not possible to define when the pressure increasing effect due to a
combination of a long flame path and a high volume blockage ratio is larger
than the pressure reducing effect of back- and side-venting.
In the case of the Chemical Plant, edge ignition resulted in a lower overpressu-
re than central ignition, so the correlation can be applied safely when used
with the Lp for central ignition.
7 Obstructed region larger than gas cloud

This white spot will be an issue in many cases, including the present one. The
TNO report
PML 1998-C53 45
influence of obstacles located immediately outside the vapour cloud is ne-

glected. It can be argued though that the influence might be considerable.
Clearly, this issue should be investigated in more detail to arrive at a complete
package of guidance for the application of the MEM.
8 Single value for explosion overpressure

The ARG simulations show the distribution of overpressures to be very non-
homogeneous. A practical model like MEM requires a single value for the
source overpressure. The use of an average maximum overpressure seems ac-
ceptable in case one is interested in blast. However, if one is interested in blast
characteristics ‘close’ to the source, or if one is interested in consequences in-
side the source, more sophisticated models should be used.
TNO report
46 PML 1998-C53
TNO report
PML 1998-C53 47
9 6 Application to the LNG Terminal case
The digital file containing the LNG Terminal case to be assessed in GAMES
covers a volume of length: 175 m (x-direction), width: 57 m (y-direction) and
height: 15.5 m (z-direction).
Figures 5.a, 5.b, 5.c and 5.d show an overview and various cross-sections.
The case mainly consists of a number of pipebridges running parallel to the x- and
y-axes. The largest obstructed region can be found between the co-ordinates x: 25 -
85 m and y: 20 - 45 m. Two minor obstructed regions are located at approximately
x: 1 - 10 m, y: 30 - 38 m and at x: 153 - 167 m, y: 30 - 45 m. There is an isolated
obstructed region consisting of some large vessels surrounded by walls at x: 113 -
132 m, y: 15 - 27 m.
We will adopt the next nomenclature for the various parts (the dimensions given
are not yet the boundaries of the obstructed subregions (OSR) but an indication of
where the part considered is located):
• OSR 1 from x = 28 to 82 m and from y = 21 to 44 m;
• OSR 3 from x:= 1 to 10 m and from y = 30 to 38 m;
• OSR 5 pipebridge 1, running parallel to the x-axis from x = 15 to 175 m and
width from y = 38 to 44 m;
• OSR 6 pipebridge 2, running parallel to the x-axis from x = 1 to 97 m and
width from y = 6 to 10 m;
• OSR 7 pipebridge 3, running parallel to the y-axis from y = 10 to 47 m and
width from x = 1 to 6 m;
• OSR 8 pipebridge 4, running parallel to the y-axis from y = 1 to 45 m and
width from x = 88 to 95.
The database contains many objects describing the ground surface. The ground
surface appears to be not completely horizontal but its z-co-ordinate varies between
2.5 and 3 m. For reasons of simplicity, the ground surface was chosen to be at
z = 3 m. The ground surface is then represented by a single large plate.
The differences in height of the total obstacle configurations is slight. The height is
mainly determined by the height of the pipebridges which is at z = 7 and 8 m
(height 4 and 5 m). The height of pipebridge 4 is about 8 m.
The height of obstructed region 1 varies from 2 to 5 m for the largest part. There is
a single large vertical cylinder with some pipework with a height of 12 m (see also
TNO report
48 PML 1998-C53
picture in progress report 1). Some pipework in obstructed region 3 reaches a

height of 10 m. The height of the isolated obstructed region 4 is taken to be 3 m.
This initial attempt to subdivide the total configuration into obstructed regions is
visualised in Figure 6.
Figure 5.a: View of the LNG Terminal case: overall view.
TNO report
PML 1998-C53 49
Figure 5.b-d: View of the LNG Terminal case: b: horizontal projection on yx-plane; c: vertical projection on
xz-plane; d: vertical projection on xy-plane.
TNO report
50 PML 1998-C53
95374-6
y
OSR - 5 H = 5m
H = 8m
OSR - 3 H = 5m OSR - 2
H = 10m OSR - 1 H = 5m
OSR-7 H=5m
H = 2m H = 12m OSR - 4
OSR - 8
H = 3m
OSR - 6 H = 4m
x
x
10m
Figure 6: LNG Terminal case initial subdivision into obstructed regions.
9.2 6.2 First impression of potential explosion severity
The dimensions of the total obstacle configuration described by the database is

quite large. On the other hand, there seems only a single part with a high obstacle
density which is expected to generate higher overpressures. That is, OSR-1, OSR-2
and OSR-3 are expected to produce lower overpressures due to the smaller dimen-
sions and their probably lower obstacle densities.
The effect of the pipebridges has to be investigated. The Chemical Plant case
showed that ignition at the short edge resulted in a very low overpressure. It is
therefore interesting to investigate whether an explosion in OSR-1 results in flame
acceleration in OSR-5 to finally produce higher overpressures in OSR-2. It also
may be possible that explosions in OSR-1 and OSR-2 have to be regarded as two
separate explosion sources.
A question is also to what explosion the combination of OSR-1, -7, -6 and -8 will
result. Is such an explosion governed by OSR-1 only or do the other OSRs have a
contribution?
Considering the comments given above, it is logical to first investigate subcases

before investigating the case as a whole.
The investigation will be performed assuming a cloud of a stoichiometric metha-

ne/air mixture filling the entire obstacle configuration up to a z-co-ordinate of
11 m. So, only the high vertical cylinder with adjacent equipment in OSR-1 ex-
tends outside the cloud.
TNO report
PML 1998-C53 51
Contrary to the Chemical Plant case, there is no obstructed subregion present for
which one may consider the correlation for 2 dimensional expansion. The applica-
tion of the correlations will therefore be restricted to equation 2 in combination
with SL = 0.45 m/s.
9.3 6.3 Application of correlation to obstructed subregion 1
The two main questions in applying equation 2 to OSR-1 are:

• what is the volume of OSR-1?
• what is the average obstacle diameter?
Values for Lp and VBR can be derived if the volume is known.
The step-by-step procedure of the Yellow Book, given in Annex A, will be applied
to determine the boundaries of OSR-1.
Figure 7 shows details of OSR-1.
TNO report
52 PML 1998-C53
95374-7a
P6
P3
iL-1
iL-3
P5
P2
iL-4
P4
P1
iL-2
Figure 7.a: Details of OSR-1: horizontal projection on the xy-plane.
TNO report
PML 1998-C53 53
Figure 7.b-c: Details of OSR-1: b: vertical projection on the xz-plane; c: vertical projecti-
on on the yz-plane.
TNO report
54 PML 1998-C53
9.3.1 Determination of Vor

Step 1
This is the obstacle database.
Step 2
We will assume an ignition location inside OSR-1.
Step 3 and 4
These steps are to decide if an adjacent obstacle belongs to the initial ob-
structed region consisting of the obstacles immediately surrounding the igni-
tion location. One of the criteria is that the initial obstructed region is exten-
ded with all obstacles within a range of 10 times the diameter of an obstacle
in the initial obstructed region. For instance, if the diameter D1 of the obsta-
cle closest to the ignition location equals 1 m, then all obstacles within 10 m
should be included in the obstructed region. The extended initial obstructed
region is to be extended again if more obstacles fall under the criteria applied
to the extended initial obstructed region.
An obstacle does not belong to the obstructed region if the distance to any
obstacle in that obstructed region is larger than 10 times the diameter of any
obstacle inside the obstructed region or if the distance of the obstructed regi-
on is 25 m or larger.
It will soon be clear while applying the procedure that all obstacles in the
LNG Terminal case belong to a single obstructed region according to the
Yellow Book procedure; even OSR-4.
Step 5
In order to limit the first assessment to OSR-1 and to determine the Vor of
OSR-1, we will assume that a single box containing the obstacles of OSR-1
is located between x = 28 to 82 m, y = 21 to 44 m and z = 3 to 15.5 m. The
dimensions of the box are length 54 m, width 23 m and height 12.5 m.
Step 6
Without conflicting the criteria for the obstructed region, the volume of the
box can be reduced by leaving out free space. Also, the height of the box is
limited to the height of the cloud. A study in detail of the obstacle configu-
ration inside OSR-1 resulted in the shape of Vor visualised in Figure 8.
The single box defined in step 5 is reduced to 4 subboxes:
• subbox 1: x = 28 - 45, y = 21 - 31, z = 3 - 5, volume = 340 m3;
• subbox 4: x = 63 - 82, y = 29 - 44, z = 3 - 8, volume = 1425 m3.
The total volume of OSR-1 equals Vor = 4005 m3.

Close examination of the obstacle configuration did not reveal any possibility to
reduce Vor by 10% or more. The conclusion is that a volume of 4005 m3 for the
OSR-1 is a practical minimum value.
TNO report
PML 1998-C53 55
95374-8
44
y (m) H=4
H=5
35
H=5
31
29
H=2 H=0
21
28 45 63 82
x (m)
Figure 8: Subdivision of box containing OSR-1 into subboxes. H: height of subbox in

metres.
9.3.2 Determination of the average diameter D and the volume blockage

ratio VBR
Application of equations 5, 6 and 7 to OSR-1 results in:
• Darm = 0.32 m;
• Dham = 0.23 m;
• Dhym = 0.48 m;
and
• VBR = 0.06.
9.3.3 Determination of Lp
A number of values for Lp are possible depending on the ignition location.
The following ignition locations will be considered:
• IL-1 at (x,y,z) = (55,36,5), central ignition;
• IL-2 at (x,y,z) = (28,36,5), centre of short edge;
• IL-3 at (x,y,z) = (55,31,5), centre of long edge;
• IL-4 at (x,y,z) = (55,44,5), centre of long edge.
The next values for Lp are interesting for evaluation:

• Lp1 = 2 m, approximate distance to top of OSR-1;
• Lp2 = 11.5 m, half the width of OSR-1;
• Lp3 = 23 m, width of OSR-1;
• Lp4 = 27 m, half the length of OSR-1;
• Lp5 = 54 m, length of OSR-1;
• Lp,ave= 12.4 m, radius of hemisphere with equal volume as OSR-1.
9.3.4 Determination of overpressure

Substituting the values for Lp, VBR, SL and D in the correlation, equation 2, results
in Table 15 for the overpressure P0.
TNO report
56 PML 1998-C53
Table 15: Overpressures for OSR-1.
P0 (kPa)
LP (m) Darm = 0.32 m Dham = 0.23 m Dhym = 0.48 m
2 0.3 0.6 0.1

11.5 36 71 16
23 244 480 106
27 379 746 165
54 2549 5017 1110
12.4 45 88 19
The overpressures calculated for Lp = 2 and 54 are unrealistically low and high,
respectively.
9.3.5 ARG calculations

ARG simulations were performed in which the obstacle configuration was limited
to the obstacles within OSR-1. The ignition locations were IL-1, -2, and -4. Pressu-
res were sampled at six locations (P1 to P6, see Figure 7.a).
Maximum overpressures calculated for the various ignition locations and pressure
sampling locations are given in Table 16.
Table 16: Maximum overpressures calculated with ARG.
Po (kPa)
Location (x, y, z) IL-1 IL-2 IL-4
P1 (35, 34, 5) 36 24 7
P2 (55, 34, 5) 21 89 7
P3 (75, 34, 5) 73 82 7
P4 (35, 39, 5) 72 30 9
P5 (55, 39, 5) 22 70 9
P6 (75, 39, 5) 81 96 9
average 51 65 8
Table 16 shows that the overpressure increases at increasing distances from the
ignition location. Despite the fact that IL-2 is located in the centre of the short
edge, the maximum overpressures is higher than in the case of central ignition
(IL-1). Edge ignition in the Chemical Plant case always resulted in a lower over-
pressure than central ignition.
9.3.6 Evaluation
The correlation results for the Lp values according to half the width, the width and
the equivalent radius give results in the same order as the averaged numerical
results. The best results seems to be those when a smallest value for the diameter is
used (Dham).
TNO report
PML 1998-C53 57
This is in contradiction to the results of the Chemical Plant case where the best
results were obtained when the largest value for the diameter was used.
A difference between the Chemical Plant case and the subcase of LNG Terminal
considered here is the lower VBR. The average Lps are comparable because the
Vors are more or less the same. A difference is present in aspect ratio.
9.4 6.4 Application of correlation to obstructed subregion 2
The initial boundaries for OSR-2 were taken from x = 153 to 167 m and from
y = 30 to 45 m. Closer examination and application of the procedure in Annex A
resulted in slightly different boundaries. Figure 9 shows the boundaries. As in the
previous section, application of the Yellow Book procedure to determine an ob-
structed region results in the complete LNG Terminal obstacle configuration be-
coming a single obstructed region. It is assumed for the limitation of the boundaries
of OSR-2, the sections through pipebridge 1 at x = 147 m and x = 168 m are boun-
daries.
OSR-2 now consists of three subboxes:
• subbox 1: x= 150 - 168, y = 27 - 36, z = 3 - 6, volume 486 m3;
• subbox 2: x = 147 - 168, y = 36 - 45, z = 3 - 8, volume 945 m3;
• subbox 3: x = 157 - 165, y = 45- 47, z = 3 - 8, volume 80 m3;
with a total volume Vor = 1511 m3.
TNO report
58 PML 1998-C53
95374-9a
2 P18 P19
iL-9 P8
P10
iL-5
1 P7 P9
P20
Figure 9.a: Subdivision of OSR-2 into subboxes: horizontal projection on the xy-plane.
TNO report
PML 1998-C53 59
95374-9b
95374-9c
Figure 9.b-c: Subdivision of OSR-2 into subboxes: b: vertical projection on the xz-plane, c: vertical projection on
the yz plane.
TNO report
60 PML 1998-C53
Calculation of the various average diameters results in:

• Darm = 0.31 m;
• Dham = 0.21 m;
• Dhym = 0.46 m;
and
• VBR = 0.04.
When the next ignition location is taken:

the next values for Lp are of interest:
• Lp,1 = 2 m, average distance to top;
• Lp,2 = 10 m, average distance to all edges;
• Lp,ave = 9.0 m, radius of hemisphere with volume Vor.
Application of the correlation (equation 2) results in.

P0 (kPa)
2 0.1 0.2 0
10 9 19 4
9 6 14 3
ARG calculations were performed using all obstacles enclosed within the space
bounded by: x = 147 to 168 m, y = 27 to 47 m and z = 3 to 8 m and using IL-5 for
the ignition location.
The maximum overpressures for locations P7, P8, P9 and P10 (see Figure 9a)
were:
• P7: 18 kPa;
• P8: 28 kPa;
• P9: 20 kPa;
• P10: 21 kPa.
The average maximum overpressures equals 22 kPa.
Table 17 shows that variation of the parameters for the correlation within a realistic
range results in a wide range of overpressure predictions. Acceptable results were
obtained when using Dhym together with an average Lp. Here, this combination
results in a too low overpressure (3 kPa).
A reason why this underprediction occurs may be that the procedure to determine
the obstructed region according to the Yellow Book is too conservative, especially
for small obstructed volumes. If the procedure would allow a reduction in Vor of
50%, the overpressure would increase despite the smaller Lp.
Room for volume reduction is available, see Figure 9. A reduction of 50% in Vor
doubles the VBR to 0.08 and reduces the average Lp to 7.1 m. The correlation
predicts the overpressures of Table 18 when using these values.
TNO report
PML 1998-C53 61
Table 18: Overpressures for OSR-2 with reduced Vor.
P0 (kPa)
2 0.7 1.5 0.3

10 58 129 26
7.1 23 50 10
It appears that even with a 50% volume reduction the correlation using Dhym and
Lp,ave results in too low overpressures when compared with the numerical results.
Although the Yellow Book procedure to determine Vor is thought to be conservati-
ve, a reduction of 50% does not seem to be possible.
The conclusion is that the choice of parameters for the correlation which worked
well in the Chemical Plant case is not appropriate in the present case of OSR-2.
Maybe the low value for the VBR of OSR-2 (0.04) is responsible for this.
9.5 6.5 Application of the correlation to a combination of

obstructed subregions 1, 2 and 5
9.5.1 Introduction
The combination of the obstructed regions 1 and 2 via pipebridge 1 into one large
obstructed region is interesting for the application of the Multi-Energy Method.
When ignited at the short ends, the pipebridge in the Chemical Plant case did not
generate much overpressure. Here in the LNG Terminal case, the number of pipes
and pipe layers is even less than in the Chemical Plant case. It may be expected
therefore that a flame, initiated in either OSR-1 or OSR-2, will decelerate during
propagation through the pipebridge. The result will be two separate explosions; one
in OSR-1 and one in OSR-2.
9.5.2 ARG simulations

An ARG simulation was performed to investigate if any support for the idea of two
separate explosions could be found.
The simulation was made with all obstacles inside OSR-1, OSR-2 and pipebridge 1
(OSR-5) for ignition location IL-2. An overview of the results is given in Table 19.
The ARG pressure histories for the pipebridge (P11 and P12) and for OSR-2 show
two peaks. The first peak appears before the flame arrived at the sampling location
and must therefore be due to a pressure wave. This pressure wave is initiated by the
explosion in OSR-1. The second peak in the pressure histories inside the pipebrid-
ge and OSR-2 is due to the combustion in and near these locations.
The overpressures in OSR-1 are similar to the overpressures calculated previously
considering OSR-1 only with IL-2 (65 kPa). The acceleration of the flame while
propagating through OSR-1 causes a pressure wave which reduces from 60 to
TNO report
62 PML 1998-C53
33 kPa inside the pipebridge to 25 kPa inside OSR-2. Later the flame travels
through the pipebridge and decelerates (overpressure from 70 kPa inside OSR-1 to
14 and 7 kPa closer to OSR-2). Inside OSR-2, the flame accelerates again to an
overpressure of 11 kPa (near the 17 kPa in the case of an isolated OSR-2 with
central ignition IL-5).
Table 19: ARG results for the combination of OSR-1,-2 and -5 with IL-2.
P0,max (kPa) P0,average per OSR (kPa)
P1 26
P2 94
P3 80
63
P4 31
P5 70
P6 76
P11 60 - 14
P12 33 - 7
P7 26 - 7
P8 26 - 12
25 - 11
P9 21 - 12
P10 25 - 12
Simulations have been performed with ignition locations IL-1, IL-5 and IL-6. The
latter being in the centre of the pipebridge 1 at (x,y,z) = (99,37,5). Some divergen-
ce problems occurred during the simulations in these cases.
The results of the simulation before divergence showed however the same trend as
the one ignited in IL-2: two separate explosions in OSR-1 and OSR-2, with a
pressure wave preceding the combustion pressure wave in the acceptor-obstructed
region. The overpressures being equal to the overpressures in the explosion of the
isolated obstructed subregion.
9.5.3 Application of the correlation

Application of the correlation is restricted to a single explosion source. According
to the ARG simulation, the combination of OSR-1, OSR-2 and OSR-5 consists of
two separate blast sources. The overpressures in each blast source are almost
independent of the ignition location.
Application of the correlation to the combined obstructed regions is therefore not
expected to provide accurate answers for the overpressure inside the obstructed
region.
TNO report
PML 1998-C53 63
9.6 6.6 Application of the correlation to a combination of

obstructed subregions 1, 3, 6, 7, and 8
The assessment of the combination of obstructed regions 1,2 and 5 resulted in the
conclusion that an explosion due to an ignition location in OSR-1 is not influenced
by an explosion due to an ignition location in OSR-2. The next item to consider is
then the influence of pipebridges 2, 3 and 4 (OSR-6, 7, and 8) and OSR-3 on the
explosion potential of the LNG Terminal case. It may expected that the flame, after
leaving OSR-1 where ignition is assumed, will decelerate in the open space in
between OSR-1 and the surrounding pipebridges. Some acceleration of the flame
when it enters the pipebridges may perhaps be expected.
9.6.1 Determination of parameters

The first step to determine the Vor of the combination of OSR-1, 3, 6, 7, and 8
according to the Yellow Book procedure is to define a box containing all obstacles
in these OSRs. Such a box has dimensions: x from 1 to 97 m, y from 1 to 45 m and
z from 3 to 11 m.
An intuitive subdivision into subboxes results in Figure 10a. The subboxes contai-
ning OSR-1 according to paragraph 5.3 is extended with subboxes containing the
other OSR-1, assuming dimensions according to paragraph 5.1.
The Yellow Book procedure is a more strict procedure to subdivide this initial box
into subboxes and to reduce the volume Vor.
The result of the first step to reduce the initial box into subboxes using the Yellow
Book procedure is presented in Figure 10.b. According to the procedure, the height
of the initial box can be lowered if the distance between obstacles is greater than
25 m. The height of the space in between these obstacles is then the height of the
obstacle or subregion to which the flame propagates.
A further reduction in the height locally becomes questionable. The procedure is
not very strict about the height of the space in between two volumes containing
obstacles, and various values for the height appear to be possible.
In order to remain conservative, a next step to further reduce Vor will not be taken.
The following values were calculated for input in the correlation.
For the volume of the obstructed region according to the Yellow Book procedure:
• Vor = 17985 m3;
• VBR = 0.027;
• Darm = 0.35 m;
• Dham = 0.24 m;
• Dhym = 0.48 m.
TNO report
64 PML 1998-C53
95374-10a-b
A 5
10
5
8
y
5
x
B 5 4
4
4
5
10 5
2 8
y
4 4
x
10m
Figure 10: Subdivision into multiple subboxes; A: intuitive subdivision; B: subdivision

according to Yellow Book procedure. The numbers indicate the local height
in metres.
For the volume of the obstructed region determined intuitively:

• Vor = 9482 m3;
• VBR = 0.048;
• Darm = 0.33 m;
• Dham = 0.23 m;
• Dhym = 0.47 m.
Although the pipe diameters and VBRs vary considerably for the various subboxes
in both alternative approaches, the average pipe diameters as stated above do not.
This indicates that the only difference between the two approaches concerns empty
space only.
Also possible values for Lp are the same in both approaches, except the average
value based on volume.
The following ignition locations will be considered:

• IL-3 at (x,y,z) = (55,29,5); the centre of the combined obstructed regions consi-
dered;
• IL-4 at (x,y,z) = (55,44,5); the centre of the long edge;
TNO report
PML 1998-C53 65
• IL-7 at (x,y,z) = (0,27,5); the centre of the short edge, which results in the
following values for Lp:
• Lp1 = 2 m, average distance to the top of the combined obstructed regions;
• Lp2 = 20.5 m, half the width of the combined obstructed regions;
• Lp3 = 41 m, the width;
• Lp4 = 46 m, half the length;
• Lp5 = 92 m, the length;
• Lp,aveYB = 20.5 m, radius of hemisphere of volume 17985 m3;
• Lp,aveIN = 16.5 m, radius of hemisphere of volume 9482 m3.
9.6.2 Correlation results

The combination of all possible values for the various parameters results in a range
of overpressure predictions; these are collected in Table 20. The results for Lp =
2 m are not in the table. The overpressures for Lp = 2 m all appeared to be negligi-
ble (< 1 kPa).
Table 20: Overview of overpressure predictions using the correlation.
P0 (kPa)
VBR = 0.027 VBR = 0.048
Lp (m) Darm= Dham= Dhym= Darm= Dham= Dhym=
0.35 m 0.24 m 0.48 m 0.33 m 0.23 m 0.47 m
20.5 19 39 9 90 189 44
41 125 262 60 607 1273 294
46 171 359 83 834 1747 404
92 1153 2416 558 5608 11756 2716
20.5 19 39 9 - - -
16.5 - - - 50 104 24
9.6.3 ARG calculations

AutoReaGas calculations were performed using an obstacle database equal to all
obstacles present in the combination of OSR-1, -3, -6, -7 and -8. Calculations were
performed using ignition locations IL-3, IL-4 and IL-7.
Overpressures were sampled at P1 to P6 (all in OSR-1) and at:
• P13 at (x,y,z) = (6,35,5); the centre of pipebridge 3 (OSR-7);
• P15 at (x,y,z) = (40,5,5); the centre of pipebridge 2 (OSR-6):
• P16 at (x,y,z) = (92,25,5); the centre of pipebridge 4 (OSR-8).
See Figure 10.a for ignition and pressure sample locations. Results are in Table 21.
TNO report
66 PML 1998-C53
Table 21: ARG maximum overpressures.
P0 (kPa)
Location IL-3 IL-4 IL-7
P1 32 27 27
P2 7 8 15
P3 71 24 49
P4 29 24 17
P5 11 9 21
P6 52 35 40
P14 13 11 10
Average 31 20 26
P13 5 4 12
P15 7 4 22
P16 5 8 7
Average 5.6 5.3 13.7
Ignition in OSR-1 (IL-3 and IL-4) does not result in significant overpressures
outside OSR-1. Overpressure is generated in the pipebridges only by ignition in
IL-7. Presumably, this overpressure is due to the other direction of flame propaga-
tion, which is parallel to the pipebridge OSR-6 for IL-7 and perpendicular to the
pipebridge OSR-6 for IL-3 and IL-4.
9.6.4 Evaluation
Application of the correlation to an obstructed volume like the combined one
considered here shows that no insight is obtained into the overpressure distribution
inside the obstructed volume itself.
A reasonable result is obtained when using the correlation with Dhym, Lp,ave and
the intuitive Vor. Applying the correlation with the Yellow Book Vor gives a reaso-
nable result in combination with Dham. This implies that the procedure in the
Yellow Book to determine Vor might be too conservative in a sense that too much
empty space is included.
In general, it appears that a smaller average diameter has to be chosen to get an
acceptable overpressure prediction for lower VBRs.
9.7 6.7 Blast outside obstructed regions
The exercises performed in the previous paragraphs demonstrated the inapplicabi-

lity of the correlations to predict the overpressure inside obstructed regions with a
highly non-homogeneous obstacle distribution and/or with a large aspect ratio. In
those situations it is complicated to decide on a choice for Vor, Lp and D.
Until now the attention has been focused on prediction of the source strength,
which is described by overpressure P0, and total combustion energy, the latter
being determined by Vor.
TNO report
PML 1998-C53 67
The objective of the MEM is to predict the blast characteristics outside the vapour
cloud and outside the obstructed region. It is therefore interesting to investigate the
influence the various combinations of the parameters have on the blast characteris-
tics in order to get insight into the required accuracy with which the parameters
have to be applied.
The blast chart of the MEM is given in Figure 1. It shows that for source strengths
higher than 6 (P0 > 50 kPa) the blast overpressures at greater distances (r’>2) are
independent of the source strength.
The source overpressure P0 is uniform within the cloud (r’ < r’0). The blast over-
pressure equal to the source overpressure for r’ is about 2r’0 (at about r’= 0.6),
which is the radius of the fully expanded burnt hemispherical flammable cloud. For
r’ > 2r’0, the blast decays with distance according to acoustic law (a straight line in
a log P - log r’ scale):
For source strengths of class < 6, the blast overpressure decay P(r’) can be ap-
proximated by two straight lines:
P(r’) = P0 for r’< 0.6 (8)
log P(r’) = log P0 - (log r’ - log 0.6) for r’> 0.6 (9)
The overpressure P at a certain distance r depends on the choice of P0 and r’.

For a scaled distance greater than twice r’0:
• doubling P0 results in doubling the overpressure P at a certain r’ and in doubling
r’ with a certain overpressure P;
• doubling r’ results in halving the overpressure P for a specific P0.
The emphasis on the determination of the influence of changing the parameters of

the correlation on the blast will be on the diameter D and on the volume of the
obstructed region Vor.
9.7.1 Influence of changing the diameter D

It can be deduced easily using equation (2) that varying D according to:
D2 = α D1 (10)
results in:
P0,2 = α-2.05 P0,1 (11)
Changing D does not influence Vor, so the scaled distances do not have to be
changed.
Changing from D1 to D2 will increase the overpressure at a certain distance by a
factor α-2.05, or change the actual distance for a certain P with the same factor.
TNO report
68 PML 1998-C53
For source overpressures larger than 50 kPa, changing D only influences the blast
characteristics at scaled distances smaller than 2, for MEM class 6, and 1 for MEM
class 7.
The exercises in the previous paragraphs on the determination of D often show a

variation of 50% in diameter for Dhym and Dham with respect to Darm.
If Dhym in such a case is used instead of Darm, the distances at which a certain
overpressures occurs is reduced by 45%. Similarly, if Dham is used instead of Darm,
that distance is increased by a factor 4. This factor of four will be much lower in
most cases, as P0 will increase for smaller D and will exceed easily the limit of 50
kPa. Nevertheless, the influence of varying D on distances can be strong, especially
for smaller P0.
9.7.2 Influence of changing Vor

The procedure adopted to determine Vor is the procedure from the Yellow Book
(Annex A). While applying this procedure, the Volume Blockage Ratio VBR will
increase while Vor decreases. Because the reduction procedure will not neglect
obstacles, D will not change.
As the external dimensions of the obstructed region will change, Lp will change
too. To investigate the influence of changing Vor correctly, it is assumed here that
Lp is the radius of a hemisphere with volume Vor. Now Lp = r0.
Varying Vor varies both P0 and E (scaled distance).

It can be deducted that for:
Vor,2 = β Vor,1 (11)
the MEM input parameters vary according to:

P0,2 = β-1.83 P0,1 (12)
and
r’2 = β-1/3 r’1 (13)
A reduction in Vor increases the pressure P at a certain distance r due to the increa-
se in P0, but reduces P due to the increase of r’.
Substituting (12) and (13) in (8) and (9) results in:

P2 = β-1/3P1 for P0 > 50 kPa (MEM class 6) (14)
and
P2 = β-1.5P1 for P0 < 50 kPa (15)
In order to get a feeling for the variation in blast overpressure induced by the factor
β, the equations (14) and (15) will be applied to two examples.
TNO report
PML 1998-C53 69
According to paragraph 2.2, it is useful to apply the Yellow Book procedure for the
determination of Vor until a further reduction of 10% is not possible.
A 10% reduction equals β = 0.9, which means that the error in the blast peak
overpressure induced by the criterion for stopping the procedure will be 4 to 17%
according to (14) and (15), respectively. These values are acceptable given the
accuracy of the correlations.
The procedure to define the obstructed region (steps 1 to 4 of the procedure in

Annex A) can be evaluated for the situation considered in paragraph 6.6, with the
influence on the outside blast in mind.
According to the Yellow Book procedure, the combination of the obstructed subre-
gions 1 ,3, 6, 7 and 8, should be regarded as a single large obstructed region. The
open space in between OSR-1 and the pipebridge cannot be excluded. Therefore,
the volume of the obstructed region is Vor = 17985 m3. According to an intuitive
approach, the volume of the obstructed region results in Vor = 9482 m3.
The reduction from the Yellow Book volume to the intuitive volume by the factor
β = 0.53 leads to an increase in peak blast overpressures by a factor of 2.6.
The conclusion is that the criteria in the Yellow Book procedure to include or
exclude free space in the obstructed region should be more validated than they are
at present, because a conservative estimate for the obstructed region may result in
an unsafe prediction of the blast overpressures.
9.7.3 Large or small obstructed region

In the case of a flammable cloud covering the obstructed regions 1, 3, 6, 7 and 8 as
in paragraph 6.6, the blast outside can be calculated using the MEM for various
assumptions for the volumes of the obstructed region:
• case 1 assume that only OSR-1 generates blast;
• case 2 assume that the obstructed volume according to the Yellow Book
procedure generates blast;
• case 3 assume that the intuitive obstructed volume generates blast
a: using Darm and b: using Dhym.
Table 22 shows the distances to the centres of the various obstructed regions for
specific blast overpressures
TNO report
70 PML 1998-C53
Table 22: Distances for specific blast levels.
Distance r (m)
blast: 25 kPa 10 kPa 5 kPa 1 kPa
case 1 (Darm) Vor = 4005 m3 56 130 242 1036

Darm = 0.32 m
P0 = 45 kPa
case 2 (Darm) Vor = 17985 m3 - 96 197 983
Darm = 0.35 m
P0 = 19 kPa
case 3a (Darm) Vor = 9482 m3 90 191 352 1473
Darm = 0.33 m
P0 = 50 kPa
case 3b (Dhym) Vor = 9482 m3 - 98 192 930
Dhym = 0.47 m
P0 = 24 kPa
It is interesting to note that the distances for cases 1, 2 and 3b do not differ much.
For cases 2 and 3, the numerical average overpressure was about 20 kPa. It may be
expected that the blast, when calculated numerically, also will not differ much from
the blast according to the MEM.
The conclusion is that blast is only produced by OSR-1; the pipebridges do not
contribute to the blast in this case.
Although the Yellow Book procedure results in a large volume Vor and conse-
quently a low P0, the blast at some distance is comparable with the blast calculated
using other assumptions.
The conclusion is also that following a strict procedure, like the one in the Yellow
Book, in this particular case with a complicated obstructed region, results in ac-
ceptable blast overpressures at long distances. The high source overpressures in the
specific parts of the obstructed region, which are the actual explosion sources, will
not be identified.
9.8 6.8 Overall evaluation of the LNG Terminal case
9.8.1 General
The LNG Terminal obstacle configuration consists of a number of large elongated
obstructed regions, the pipebridges, which connect a single large volume with a
congestion of obstacles with a couple of smaller congested volumes. The pipebrid-
ges enclose a considerable volume of space without obstacles.
As in the Chemical Plant case, the correlations were applied first to single obstruc-
ted subregions. A large band of predictions of the overpressures was obtained when
using the correlation depending on the location of the ignition source and the
definition of average obstacle diameter.
TNO report
PML 1998-C53 71
In general, the results from the correlations were lower than the numerical results
when compared to the Chemical Plant case.
The correlation results when using the arithmetic obstacle diameter in combination
with the radius of an equivalent hemisphere for Lp appear to be of the same order
as the numerical results.
A larger diameter should be used in the correlation to get acceptable results in

comparison with the Chemical Plant case. The reason for this is not completely
clear. In comparison with the Chemical Plant case, the obstructed subregions 1 and
2 of the LNG Terminal case have a lower volume blockage ratio and a much lower
height than in the Chemical Plant case.
Despite the low height of OSR-1, the overpressure according to the numerical
calculation is much higher than according to the correlation using the height for Lp.
This is probably due to the relatively large length of OSR-1. Apparently, the flame
is capable of accelerating despite the venting through the top and the low VBR.
The low overpressure in OSR-2 according to the correlation is probably caused by

the criteria of the Yellow Book to define the obstructed region. Using half the
value of the obstructed volume according to the Yellow Book procedure and Darm
results in an acceptable overpressure in comparison with the numerical result.
The presence of various pipebridges complicates the application of the correlation.

The numerical simulation shows that the overpressures inside the pipebridges are
relatively low and the conclusion is that OSR-1 and OSR-2 should be considered as
separate explosion sources of different strengths.
The application of the correlation to the combination of OSR-1 and OSR-3 together
with three pipebridges, is problematic. The obstacle distribution inside the ob-
structed region varies considerably. The numerical calculation shows low as well
as high overpressures depending on the location. This makes it difficult to deduce a
single average maximum overpressure. The strict Yellow Book procedure results,
however, in acceptable values for far field blast overpressures.
In the case of OSR-2, it is suspected that too much free space is included in the
obstructed region using the Yellow Book procedure.
If this is indeed the case, then the blast overpressures outside the obstructed region
may be underpredicted. A numerical calculation of the blast may provide the
answer.
In order to determine blast outside the obstacle configuration, only the contributi-
ons from OSR-1 and OSR-2 should be taken into account. The pipebridges and the
free space enclosed do not significantly contribute to blast. The two obstructed
regions should be considered as separate explosion sources.
TNO report
72 PML 1998-C53
9.8.2 Comments on the covering of the white spots

1 Missing correlations
The case considered is clearly a 3D expansion situation. The available corre-
lation 3D expansion is sufficient here.
2 Criterion to distinguish between 2D and 3D correlation

Not applicable in this case.
3 Definition of obstructed region

Contrary to the Chemical Plant case, the definition of the obstructed region is
a problem in this case. The impression exists that the criteria in the procedure
to define the obstructed region according to the Yellow Book include too
much free space.
The approach adopted in the Yellow Book is a conservative one, as a high
number for the source strength class should be adopted. In combination howe-
ver with the application of the correlation to determine the source strength
class, the procedure may lead to too low predictions for the outside blast.
4 Multiple obstructed regions

The numerical simulation shows that various blast waves are produced by dif-
ferent parts of the obstructed region if the whole obstacle configuration is situ-
ated within a flammable cloud. The pipebridges are not able to accelerate the
flame.
The procedure in the Yellow Book is not able to distinguish multiple explosi-
on sources within a single obstacle configuration. The procedure results in a
large Vor with a low VBR with consequently a low P0. Because of the large E
the blast in the far field will not differ much from the blast resulting from a
smaller but more congested volume.
5 Differently shaped obstacles

The case contains many non-cylindrical obstacles. The procedure adopted to
cope with non-cylindrical obstacles is to turn these into cylinders with a length
equal to the longest dimension of the obstacle and an equal cross-section area.
There is, at present, no indication of the physically correctness of that proce-
dure. The procedure is very questionable, especially for flat plates.
6 Lp in the case of an aspect ratio other than unity and ignition location outside
the centre of configuration
The height of the obstructed region (maximally 5 m) is very low compared to
the other dimensions (40 to 100 m). Using the height for Lp in the correlation
will not result in any overpressure. This is not correct, as the numerical results
demonstrate.
Also, the model derived in Annex E to take the aspect ratio into account is in-
correct, as the overpressure increases considerably after the flame has left the
obstructed region through the top.
TNO report
PML 1998-C53 73
In some situations, edge ignition resulted in higher overpressures than central

ignition, despite the fact that VBRs are lower than in the Chemical Plant case.
Obviously the longer path the flame can travel while accelerating is responsi-
ble for the higher overpressures.
The flame accelerating and the influence of top and back venting in combina-
tion with the flame path length cannot yet be explained on the basis of simple
and clear models. More research is required on this aspect, as it appears to be a
very important issue in the determination of the source overpressure.
7 Obstructed region larger than gas cloud

This white spot will be an issue in many cases, due to the low height of the
present case; it is not very important in this case.
8 Single value for explosion overpressure

The ARG simulations show the distribution of overpressure to be very non-
homogeneous. A practical model like MEM requires a single value for the
source overpressure. The use of an average maximum overpressure seems ac-
ceptable if one is interested in blast. However, if one is interested in blast cha-
racteristics ‘close’ to the source, or if one is interested in consequences inside
the source, more sophisticated models should be used.
TNO report
74 PML 1998-C53
TNO report
PML 1998-C53 75
10 7 Application to the Gas Processing case
The third case to be assessed in the project is an experiment conducted by British

Gas plc in a test rig typical of onshore storage or gas processing sites. The infor-
mation on the experiment and on the test rig was made available via the EU spon-
sored Explosion Modelling Evaluation (EME) project (EMEG, 1997).
The case was chosen because experimental data is available so there is no require-
ment to rely on numerical simulations to evaluate the applicability of the correlati-
ons.
The test rig comprises a mixture of pipework, vessels and support structure in a
cuboidally shaped region. The congested region of the test rig had maximum
dimensions of approximately 8.3 m length, 5.3 m width and 3.5 m height.
The digital file containing the Gas Processing case covers a volume of 207.5 m3
enclosed by: x from 5 to 11.3 m, y from 0 to 8.9 m and z from 0 to 3.7 m. Figu-
res 11.a, b, c and d show an overview and various cross-sections.
The obstacle configurations consist of three support structures closely located to
each other. The first layer of obstacles above ground level consists of large cylin-
drical vessels (diameter about 0.8 m) with axes parallel to the x-axis. The higher
layers consist of smaller pipes (about 0.1 to 0.3 m diameter) orientated parallel to
the x- and y-axes. The vertically orientated obstacles consist mainly of the legs of
the bearing structure.
10.2 7.2 Test performed
The experiment published was performed with propane as the fuel. The whole
obstacle configuration was covered with a plastic sheet to retain the gaseous mixtu-
re inside the rig. The plastic sheet was not removed prior to ignition.
The concentration of the propane in air was 4.2%.
Ignition occurred in the centre of the rig at (x,y,z) = (8.35, 4.15, 0).
There were 16 pressure transducers mounted inside the obstacle configuration (T1
to T16) and 3 pressure transducers outside the configuration (T17, T18 and T19),
the last 3 being mounted at a height of 1 m above ground level.
Figure 12 shows the location of the pressure transducers.
Figure 13 shows a typical pressure history obtained during the experiment (trans-
ducer T1).
The peak overpressures measured are gathered in Table 23. The average of the
measured peak overpressures inside the obstacle configuration is 24 kPa.
TNO report
76 PML 1998-C53
Figure 11.a: View of the Gas Processing case: overall view.
TNO report
PML 1998-C53 77
Figure 11.b: Views of the Gas Processing case: horizontal projection on xy-plane.
TNO report
78 PML 1998-C53
Figure 11.c: Views of the Gas Processing case: vertical projection on xz-plane.
Figure 11.d: Views of the Gas Processing case: vertical projection on xy-plane.
TNO report
PML 1998-C53 79
95374-12
T12 T10,T11
2.4m
T9 4.8m
3.2m
Plan view
(ground level) T7 T6 T2 T3 T4,T5 T17 T18 T19
T8
4.0m
3.2m 2.4m *
T1 2.4m 3.2m 4.0m 8.0m 16.0m 24.0m
Plan view T14 T15 T13

(above ground level)
T15
Side view T14 T13

(central axis)
3.2m
2m
T16
Ground
level
Internal transducer location (ground level)

Internal transducer locations (ground level and top of congested region)
Internal transducer location (3.2 m from central spark location)
Internal transducer location (attached to vessel)
External transducer location (1 m above ground level)
* Ignition location (ground level)
Figure 12: Location of pressure transducers (EMEG, 1997).
TNO report
80 PML 1998-C53
Figure 13: Example of pressure signal (EMEG, 1997).
Table 23: Measured overpressures.

A: Peak overpressures inside obstacle configuration
Transducer T1 T2 T3 T4 T5 T6 T7 T8 T9 T10 T11 T12 T13 T14 T15 T16
overpressure 24 24 27 36 30 23 23 22 23 36 14 19 21 20 20 25
(kPa)
B: Peak blast overpressures outside obstacle configuration
Transducer T17 T18 T19
distance to ignition (m) 8 16 24

peak overpressure (kPa) 34 16 12
10.3 7.3 AutoReaGas calculation
As experimental results are available, this case offers an opportunity to check the
validity of the AutoReaGas calculations. The calculations were performed for this
TNO report
PML 1998-C53 81
case in exactly the same way as they were calculated for the other cases: using a
cell size of 1 m3 and a Ct factor of 65.
Overpressures inside the obstructed region were calculated in 7 locations. The
maximum values varied from 21.3 to 28.3 kPa. The average of the maximum
values is 24 kPa, which is equal to the average of the maximum overpressure
recordings during the tests. This remarkable result provides confidence in the
AutoReaGas caculations which serve as the basis for the evaluation of the correla-
tions.
10.4 7.4 Application of correlation to obstructed region
Although the obstacle configuration consists of three bearing structures, the free
space in between them is small compared to the overall dimensions of each of the
bearing structures. The obstacle configuration is therefore to be considered a single
obstructed region.
The volume of the obstructed region can be determined easily from Figure 11:
• Vor = 154 m3.
Application of equations 5, 6 and 7 to the obstructed region results in values for the
various diameters:
• Darm = 0.11 m;
• Dham = 0.07 m;
• Dhym = 0.25 m.
The volume blockage ratio calculated is:

• VBR = 0.14.
The interesting values for Lp for central ignition are:

• Lp1 = 2.65 m, half the width;
• Lp2 = 3.5 m, the height;
• Lp3 = 4.15 m, half the length;
• Lp4 = 5.3 m, the width;
• Lp5 = 8.3 m, the length;
• Lp, ave = 4.2, the radius of a hemisphere with volume Vor.
For propane we take:

• SL = 0.52.
Applying the correlation using the set of parameter values defined above results in
a wide variety of overpressures (Table 24).
TNO report
82 PML 1998-C53
Table 24: Overpressures.
P0 (kPa)
2.65 87 219 16
3.5 186 470 35
4.15 298 752 52
5.3 583 1275 108
8.3 2004 5062 372
4.2 308 778 57
10.5 7.5 Blast outside obstructed region
Blast outside the obstructed region can be calculated using the contributing energy
and a value for the overpressure and applying Figure 1.
With a VBR of 14%, the combustion energy is:

E = (1-0.14) x 154 x 3.5 = 464 MJ
The blast overpressure is calculated at the same locations as the blast measure-
ments (Table 23B) using a source overpressure P0 of 57 kPa together with 100%
and 50% of the energy. The ‘efficiency’ factor is applies because of the conclusion
of GAME stating that less than 100% of the available energy contributes to the
explosion for source overpressures of less than 1 bar. The factor of 50% is deduced
from Figure 6 of the GAME report (Eggen, 1994).
Table 25 show the results.
Table 25: Blast predictions using MEM for P0 = 57 kPa.
Ps (kPa)
Distance to centre (m): 8 16 24
E = 464 MJ 56 37 24
E = 232 MJ 51 29 18
10.6 7.6 Evaluation and conclusion
In comparison with the average peak overpressure measured during the test, the
best predictions using the correlations are obtained by using Dhym. Using Darm or
Dham results in high and unrealistic overpressures. The experimental average peak
overpressure of 24 kPa is in between the predictions using Dhym and Lp1 and Lp2.
Using Dhym and Lp,ave results in an overprediction by a factor of 2.5.
TNO report
PML 1998-C53 83
A similar result was obtained for the reduced problem within the Chemical Plant
case in paragraph 5.3. There the overpressure was also in between the values
predicted by using for Lp the two shortest distances to the edges of the obstructed
region where the flame leaves the region. Venting will reduce further pressure
generation inside the obstructed region (venting through the top was not possible in
the Chemical Plant case of paragraph 5.3).
The outer dimensions of obstacle configuration in this case are within the range of
dimensions of the configurations in MERGE used to derive the correlation. The
same counts for the VBR. The difference is that different types, dimensions and
orientations of obstacles are present. Because the obstacle configuration in
MERGE is considered to produce high overpressures it could be expected that the
predictions using the correlation for this case are high compared with the experi-
mental result.
Taking the overpressure corresponding to the average flame path length and Dhym
(57 kPa) results in an MEM class number of about 6.3, while the test gives a class
of about 5.3.
Even when an efficiency factor of 50% is adopted, the blast overpressures are
overpredicted.
It is expected that better predictions are possible if the influence of the aspect ratio
can be taken into account.
TNO report
84 PML 1998-C53
TNO report
PML 1998-C53 85
11 8 Application to the Hydrogen case
In order to investigate the influence of reactivity it was decided to investigate a

case with hydrogen as the fuel.
Equation (2) shows that the reactivity from the fuel influences the overpressure
through the laminar burning velocity SL to the power of 2.7.
The laminar burning velocity of hydrogen is 3.5 m/s, which is a factor 7.8 higher
than SL for methane. As a result, the correlation predicts overpressures for hydro-
gen which are a factor 7.82.7 = 254 higher than for methane.
As very high overpressures are expected, the obstacle configuration to be conside-
red for this case should be small and should have a low VBR to be realistic.
The obstacle configuration selected is the obstacle configuration of the obstructed

subregion OSR-2 of the LNG Terminal case. Figures 9 a,b and c show vertical and
horizontal projections of the configuration. Figure 14 shows a three-dimensional
view.
Figure 14: Three-dimensional view of the Hydrogen case.
11.2 8.2 Application of correlation and ARG simulations
The procedure for the application of the correlation is similar to paragraph 6.4.
The obstructed volume according to the Yellow Book procedure consists of three
subboxes:
TNO report
86 PML 1998-C53
• subbox 1: x= 150 - 168, y = 27 - 36, z = 3 - 6, volume 486 m3;

• subbox 2: x = 147 - 168, y = 36 - 45, z = 3 - 8, volume 945 m3;
• subbox 3: x = 157 - 165, y = 45- 47, z = 3 - 8, volume 80 m3;
with a total volume Vor = 1511 m3.
Calculation of the various average diameters results in:

• Darm = 0.31 m;
• Dham = 0.21 m;
• Dhym = 0.46 m;
and:
• VBR = 0.04.
The following ignition locations were taken:

• IL-9 at (x,y,z) = (147,41,5), centre of edge.
The following values for Lp are of interest:

• Lp,1 = 2 m, average distance to top;
• Lp,2 = 10 m, average distance to all edges;
• Lp,ave = 9.0 m, radius of hemisphere with volume Vor.
Application of the correlation (equation 2) results in the values in Table 26.

P0 (kPa)
2 26 58 12
10 2196 4880 977
9 1644 3652 732
ARG calculations were performed using all obstacles enclosed within the space
bounded by: x = 147 to 168 m, y = 27 to 47 m and z = 3 to 8 m and using IL-5 for
the ignition location.
Initially, the ARG simulations were performed using a cell size of 1 m. The pressu-
re peaks appeared to be of very short duration (less than 1 ms). Because of this
very short duration, the peak values are cut off due to the rather large cell size.
Therefore the calculations were repeated using a cell size of 0.5 m.
The maximum overpressures for locations P7, P8, P9, P10, P18, P19 and P20 (see
Figure 9.a) calculated with a cell size of 0.5 m are given in Tables 27.a and 27.b,
while the pressure histories are shown in Figures 15 and 16.
TNO report
PML 1998-C53 87
Table 27.a: ARG overpressures for various locations for IL-5.
Location P7 P8 P9 P10 P18 P19 P20 Average
overpressure 185 220 220 510 580 300 415 347

(kPa)
Table 27.b: ARG overpressures for various locations for IL-9.
Location P7 P8 P9 P10 P18 P19 P20 Average
overpressure 180 290 780 1340 100 500 630 546

(kPa)
Central ignition (IL-5) of the highly reactive mixture results in high overpressure in
the centre of the obstructed region. The overpressure even increases while the
flame propagates through the obstructed region while accelerating. A shock is
already created inside the mixture and a detonation can be expected in the case of a
longer flame path.
The acceleration until detonation is more clear in the case of edge ignition (IL-9).
Overpressure near the ignition is about 100 kPa and the pressure sharply increases
at locations farther away from the ignition location. The highest pressure is obtai-
ned in P10, which is in a direction parallel to the pipebridge. Outside the pipebrid-
ge where obstacle density is considerably lower, the overpressure increase is less
(P20).
It is expected that the cell size is still too large to prevent the peaks from being cut-
off. Pressures will increase when the cell size is reduced further.
TNO report
88 PML 1998-C53
600
P10
P18
500 P19
P20
400
Overpressure (kPa)
300
200
100
-100
0 10 20 30 40 50 60 70 80 90 100 110 120
time (ms)
600
P7
P8
500 P9
400
Overpressure (kPa)
300
200
100
-100
0 10 20 30 40 50 60 70 80 90 100 110 120
time (ms)
Figure 15: Pressure histories according to ARG for ignition location IL-5.
TNO report
PML 1998-C53 89
; , ; 2
1400
P10
P18
1200
P19
P20
1000
800
Overpressure (kPa)
600
400
200
-200
0 10 20 30 40 50 60 70 80 90 100 110 120
time (ms)
1400
P7
P8
1200
P9
1000
800
Overpressure (kPa)
600
400
200
-200
0 10 20 30 40 50 60 70 80 90 100 110 120
time (ms)
Figure 16: Pressure histories according to ARG for ignition location IL-9.
11.3 8.3 Evaluation and conclusion
The correlation predicts rather low overpressures when the shortest distance from
the ignition location to the nearest boundary of the obstructed region is used (Lp =
2 m). For other flame path lengths and for Darm and Dham, extremely high over-
pressures are predicted. Their value is that in fact a detonation is predicted in those
cases and that overpressures will be in the order of 20 bar.
TNO report
90 PML 1998-C53
The average values of the overpressures calculated by AutoReaGas have no real

meaning here. The numerical calculations clearly demonstrate wave propagation
inside the obstructed region rather than a homogeneous pressure rise.
The influence of the laminar burning velocity is evaluated best using Darm, as this
diameter gave the best results in the LNG Terminal case. The combination with the
average flame path length of 9 m gives 16 bar, which is in line with what can be
expected.
The influence of the laminar burning velocity in the correlation is based on the
fractal scaling theory developed by Shell Research. The factor 2.7 is derived from
measurements on flame structures using hydrocarbons like methane and propane.
Nevertheless, the scaling appeared to work very well for acethylene also. Over-
pressures in the same obstacle configurations could be predicted for other fuel
types with an accuracy of 10%, based on the experiments using methane (Mercx et
al. 1994).
TNO report
PML 1998-C53 91
12 9 Overall evaluation and guidance obtained
12.1 9.1 General
The correlations derived in GAME are based on experiments with symmetrical

obstacle configurations consisting of well-defined obstacles orientated in regular
patterns. Ignition of the flammable mixture occurred in the centre of the obstacle
configuration which implied an equal distance from the ignition location to all open
edges.
As a consequence, the accuracy of an overpressure prediction according to the
correlation for a realistic obstacle configuration depends on the extent to which the
realistic configuration differs from the experimental one. The most important
differences comprise:
• the ratios of the outer dimensions of the obstructed region (aspect ratio);
• the ignition location;
• the homogeneity of obstacle distribution within an obstructed region.
The influence of aspect ratio is demonstrated in the Chemical Plant case. Overpres-
sures in sub-obstructed regions having a more or less cubical shape can be predic-
ted much better than in sub-obstructed regions with an elongated shape. For instan-
ce, the large pipebridge present in the case. The distance from the ignition location
to the centre of an edge is either very short or very long, which results in a large
under- or overprediction of the numerically predicted peak overpressure.
The influence of aspect ratio and the ignition location are both related to the flame
path length and the possibility of early venting. Overpressure may be vented
through one or more of the boundaries of the obstructed region while the flame is
still burning through parts of the obstructed region located farther away from the
ignition location. Early venting (side-, top- or backventing) occurs in a direction
opposite to the flame propagation direction.
In the case of edge ignition, venting starts immediately after ignition. In most cases
simulated numerically, this early venting results in lower overpressures than for the
situation in which the same obstacle configuration is ignited in the centre. An
exception is the Sub Obstructed Region 1 in the LNG Terminal case. Here, edge
ignition resulted in a higher overpressure than central ignition. The flame burns
very slowly initially due to backventing, but after propagating a certain distance,
the backventing is overruled by the acceleration due to obstacle-induced turbulen-
ce. Apparently, the remaining flame path length is long enough to accelerate the
flame to a speed greater than the flame path length in the central ignition case is
able to.
TNO report
92 PML 1998-C53
The MERGE experiments, underlying the correlation, have a highly regular pat-
tern. The repeatibility of obstacles, the equal obstacle spacing and the small obsta-
cle diameter was chosen intentionally to obtain high overpressures. As a conse-
quence, the correlation is expected to provide safe and conservative values.
The exercises performed in the four cases have shown that the correlations easily
overpredict the numerical results. As correlation overpressure decreases for increa-
sing obstacle diameters, the best results were obtained by choosing a rather large
value for the average obstacle diameter.
The exception again is the LNG Terminal case where, despite the low numerical
overpressures, the arithmetic average diameter had to be used in order to obtain
acceptable results. It is not clear why the LNG Terminal case provided answers
different from the other cases. The obstacle configuration is different obviously: a
large width and length in combination with a rather low height (length: 157 m,
width: 57 m, height of cloud: 5 m), and a low volume blockage ratio.
The non-homogeneity of the obstacle configuration in the LNG Terminal case

resulted in a non-homogeneous distribution of maximum overpressures inside the
obstructed region considered. This is an effect which cannot be predicted by the
correlation at all.
12.2 9.2 The correlation and parameters
12.2.1 Correlation
The expansion possibilities in the cases and subcases considered were all three-
dimensional. The subcase of the Chemical Plant case in which obstacles located
between two parallel floors were considered appeared to be of a three-dimensional
expansion nature also. The length and width of that case were too small in relation
to the height to obtain two-dimensional expansion during a large part of the com-
bustion process.
12.2.2 Volume blockage ratio or volume of the obstructed region

In cases were the obstructed region is compact and has a more or less homogene-
ous obstacle distribution, the determination of its volume is not problematic. The
volume can either be estimated or the procedure of the Yellow Book can be follo-
wed. The Chemical Plant and Gas Processing cases have shown that one may
arrive at different values for the volume of the obstructed region Vor depending on
how strictly the procedure is applied, but that the variations will not have an im-
portant influence on the overpressure. The variation in overpressures are within the
accuracy of the correlation with respect to the underlying experimental results.
The determination becomes more problematic when obstacle-poor spaces are
enclosed by the obstructed region, when the boundaries of the obstructed region are
irregular and when the geometrical situation is close to the criteria given in the
Yellow Book procedure.
TNO report
PML 1998-C53 93
The criteria in the Yellow Book procedure are probably too conservative, but there
is a lack of knowledge to better quantify them. In combination with a high class
number for MEM, the procedure will result in conservative predictions. In case the
procedure is coupled with the correlation, the overpressure inside the obstructed
region will be underpredicted, as the exercise with the LNG Terminal subcase in
paragraph 6.4 has demonstrated (obstructed subregion 2). Also, the blast overpres-
sures at great distances may be underpredicted as is demonstrated with the theoreti-
cal approach in paragraph 6.7.
Nevertheless, the Yellow Book procedure resulted in an acceptable average over-
pressure for the LNG Terminal subcase described in paragraph 6.6 (combination of
obstructed subregions 1, 3, 6, 7 and 8) in combination with a smaller (arithmetic)
value for the average diameter.
The Yellow Book procedure specifically fails for the obstructed subregion 2 of the
LNG Terminal case probably due the relatively small outer dimensions of that
subregion in combination with the low obstacle density.
The criteria in the Yellow Book procedure to build up the obstructed region should
be probably coupled with the outer dimensions of the obstructed region already
defined.
The numerical exercises for the LNG Terminal case and its subcases have demon-
strated that the pipebridges do not contribute to the explosion overpressure. Ob-
structed subregions connected via pipebridges can be regarded as separate explosi-
on sources. Obviously, the space in between obstructed regions does not have to be
totally obstacle free in order to consider the regions separate explosion sources.
Also in the Chemical Plant subcase of the isolated large pipebridge, no acceleration
of the flame inside the pipebridge is observed.
12.2.3 Average obstacle diameter

The three definitions used to calculate an average diameter can result in a conside-
rable variation in results. For all cases but the LNG Terminal one, the best results
were obtained when using the hydraulic diameter. The hydraulic diameter results
always in the largest value for the average diameter. The larger the diameter, the
lower the correlation overpressure. A larger diameter is preferred in order to reduce
the conservatism of the correlation.
Using a larger diameter in the LNG Terminal case results in an underprediction of
the overpressure. There, the arithmetic average gives the best results. The large
dimensions in combination with the low obstacle density (volume blockage ratio)
may be responsible.
12.2.4 Flame path length

The exercises with the cases have shown that a wide range of values is possible for
the flame path length. For elongated or flat obstructed regions, the distance from
the ignition location to the shortest edge (the top included) results in very low
TNO report
94 PML 1998-C53
overpressures. Using the longest distance to an edge, on the other hand, results in
far too high overpressures.
In order to deal with the problem of aspect ratios other than one, a simple model is
presented in Annex E. There it is assumed that the pressure will not rise after the
flame leaves the obstructed region through the nearest edge if the pressure at that
instant is below a certain limit. The underlying physics is that the flame speed
acceleration for pressures below that limit is too low to overtake pressure reduction
due to venting.
Observed in the exercises is that the final average numerical overpressures are in
between the predicted values using the correlation with values for the flame path
length according to the distances of the ignition location to the two nearest edges
(Chemical Plant subcase of paragraph 5.3) or the second and third nearest edge
(Gas Processing case). Furthermore, it is not clear whether there is a single value
for the overpressure limit. The exercises still show an increase in overpressure after
the start of venting through one of the edges also at a low overpressure level.
The best results are obtained where an average value for Lp is used. The average
value is the radius of a hemisphere with a volume equivalent to the volume of the
obstructed region Vor. Acceptable results are obtained for aspect ratios smaller than
about 5.
Central ignition is implicit when the radius of a hemisphere is adopted for Lp. The
numerical results however show lower overpressures, in most cases, for edge
ignition. Applying an average flame path length would cover all ignition locations
in most cases.
12.2.5 Laminar burning velocity

The influence of the laminar burning velocity on the outcome of the correlation
was investigated in the fourth case (Hydrogen case) only. The input value for the
laminar burning velocity is not recognised as a problem. The MERGE experiments
(Mercx et al, 1994) already demonstrated the validity of the concept of fractal
scaling. This concept is adopted in the correlation.
12.3 9.3 Guidance and remaining white spots
The following statements, considerations and rules can be kept in mind while
applying the correlations to a realistic situation.
12.3.1 9.3.1 General
12.3.2 Reminder
A general consideration to start with is that the correlation will provide the best
results when the obstacle configuration of the realistic situation resembles, as
closely as possible, the obstacle configuration used in the experiments underlying
the correlation. These were configurations with cubical outer dimensions, single
TNO report
PML 1998-C53 95
size obstacles equally spaced, completely filled with flammable mixtures which
were centrally ignited.
12.3.3 Comparison
The first thing to do is to compare the case under consideration with one of the
cases and subcases assessed in this report
12.3.4 9.3.2 Choice of a correlation

White spot 1: missing correlations
White spot 2: lack of criterion to choose between correlations
There are only two correlations, which, in practice, will cover the majority of
cases. The cases considered here generated no sound information to decide whether
to choose the correlation for 2- or 3-dimensional expansion.
Guidance provided for choice between correlations

Take the dimensions of the obstructed region as: length L, width W and height H.
If a confining plane is present for instance orientated in the directions of L and W,
then it is advisable to choose the 2-dimensional correlation if:
• L > 5H;
• W > 5H;
• the dimensions of the confining plane are of about the same size as L and W.
12.3.5 9.3.3 Determination of volume of obstructed region

The procedure of the Yellow Book to determine the volume of an obstructed region
consists of two parts:
• to build-up the obstructed region;
• to reduce the volume of a box containing all the obstacles of the obstructed
region.
White spot 3: criteria for build-up of obstructed region

White spot 4: separation distances
The criteria to decide whether or not an obstacle belongs to an obstructed region
are conservative. The procedure will lead to an obstructed region including proba-
bly too many obstacles.
The reduction of the volume by reducing the free space in an enclosing box uses
the same criteria and the result will be a too large volume.
A too large volume results in too large blast overpressures in case a high explosion
class number is adopted as recommended in the Yellow Book.
A too large volume implies a too low volume blockage ratio. In combination with
the correlation, too low blast overpressures may be predicted. Reduction of the
volume results in a reduction of the contributing energy but this is more than
compensated for by the increase of the source overpressure.
The conservatism of the correlation is expected to cover this effect.
TNO report
96 PML 1998-C53
An as small as possible volume of the obstructed region is required to get conser-

vative results when applying the correlation.
In order to be able to better calculate the obstructed volume better criteria should
be derived. This is not possible within the present project.
The criteria of the procedure to build up the obstructed region are probably too
coarse; they should probably be related to the size of the obstructed region.
The Yellow Book procedure will lead to a too large volume especially for small
obstructed regions. The combination with a low obstacle density will result in very
low overpressure as the subcase OSR-2 of the LNG Terminal case revealed (volu-
me of 1000 m3 and VBR of 4%).
The problem of multiple explosion blast sources in a single obstructed region or

because of multiple obstructed regions still remains unsolved. The Yellow Book
procedure for defining an obstructed region may lead to multiple regions, but
validation for the criteria is lacking.
Some consideration to this subject has been given in Annex D of this report. The
impression is that a separation distance should be in the order of the size of the
obstructed region in which ignition occurs. Experimental validation or evidence for
this concept is lacking.
Guidance provided for obstructed region

Apply the procedure of the Yellow Book for the definition of the obstructed region
and its volume. Be cautious in cases of small volumes in combination with a low
obstacle density as too low overpressures may be calculated (volume < 1000 m3
and VBR < 5%). In those cases, study the OSR-2 LNG Terminal sub-case.
Guidance provided for multiple explosion blast sources

The application of the Yellow Book procedure for the definition of an obstructed
region may lead to separated obstructed regions and thus to multiple explosion
blast sources. The condition for that is an obstacle free space in between the regi-
ons. This will hardly ever be the case. In many industrial situations, units will be
coupled via pipebridges or large diameter obstacles present in an obstructed region
which will require a very large free distance to outside obstacles.
Sub-cases of the LNG Terminal and the Chemical Plant cases have shown though
that the pipebridges present did not contribute to the explosion as long as the long
edges of such a pipebridge are not situated along an obstructed region. A compa-
rison of a situation at hand with the LNG Terminal and Chemical Plant case might
provide some guidance to whether or not to include or exclude a specific pipebrid-
ge.
12.3.6 9.3.4 Determination of average obstacle diameter

Three ways of calculating an average diameter have been investigated: the arith-
metic, harmonic and hydraulic means. For the arithmetic and harmonic mean, an
TNO report
PML 1998-C53 97
equivalent cylinder was calculated for non-cylindrical obstacles based on an equi-

valent cross-section area.
White spot 5: how to deal with non-cylindrical obstacles

As the correlation tends to overpredict source overpressures, the choice of a larger
diameter appears to result in reasonable answers. In most cases the use of the
hydraulic mean diameter resulted in acceptable answers. Only in the LNG Termi-
nal case where the arithmetic mean provided the best answers did overpressures
tend to be relatively higher than in the other cases. This effect must be attributed to
the low volume blockage ratio and large outer dimensions of the obstructed region.
Probably, the correlation is less conservative for these situations which can be
counterbalanced by using a smaller diameter.
Provided guidance for the mean diameter

Use the hydraulic mean for obstructed regions similar to the Chemical Plant and
Gas Processing case. Use the arithmetic mean diameter in case there is more simi-
larity to the LNG Terminal case.
Provided guidance for non-cylindrical obstacles

It may be expected that non-cylindrical obstacles having a similar cross-section
area as a cylindrical obstacle will have a larger effect on flame acceleration. May-
be, a drag factor should be included in the transition of a non-cylindrical to a
cylindrical obstacle. This was not investigated in the present project.
The transition to cylindrical obstacles is not an issue when the hydraulic diameter
is used.
12.3.7 9.3.5 Determination of flame path length

The determination of the flame path length turned out to be a complicated parame-
ter. The correlation is in fact unable to cope with influences of: an aspect ratio
other the one and an ignition location outside the centre of the obstructed region
including edge ignition. The subject was discussed extensively in the previous
paragraphs.
White spot 6: which Lp to choose in the case of non-point symmetrical situations

The proposed approach stated earlier in this report is that if there is a distance
between ignition point and an edge for which the correlation predicts an overpres-
sure of less than 30 kPa if this distance is taken as the flame path length, the over-
pressure will not increase anymore.
This approach does not hold for edge ignition. Also, the numerical simulations
show different behaviour. On the other hand, the proposed approach seems attrac-
tive from a practical point of view. Clearly, more (experimental) investigation is
required to come up with sound guidance.
At this moment it is not possible to incorporate the influence of non-symmetrical
situations.
TNO report
98 PML 1998-C53
Provided guidance for the flame path length

The flame path length should be taken as the radius of a hemisphere having the
volume of the obstructed region. An conservative value for the overpressure will be
obtained in most situations.
White spot 7: influence of obstructed region larger than the vapour cloud
It can be argued that overpressure will increase as the cloud expansion occurs
partially inside an obstructed region. The efficiency, that part of the combustion
energy inside the obstructed region which contributes to blast, will increase.
Provided guidance for clouds smaller than the obstructed region

Do not take an efficiency smaller than one into account.
12.3.8 9.3.6 Other remaining deficiencies

White spot 8: a single value for the explosion overpressure
The AutoReaGas calculations showed that the overpressure distribution in ob-
structed regions having an irregular obstacle distribution in not homogeneous. The
maximum overpressures are different for different locations inside the obstructed
region.
The overpressure resulting from the correlation should be such that the maximum
blast overpressures outside the obstructed regions are calculated as accurately as
possible. This means that the correlation overpressure should not be equal to the
peak maximum overpressure in the obstructed region if this peak is limited to a
small part of the entire region.
Thus, the correlation overpressure should be merely an average of the maximum
overpressures occurring inside the obstructed region.
12.4 9.4 Possible extensions of the blast charts
The Multi-Energy Method is a practical tool for determining the blast overpressu-
res from vapour cloud explosions. The reality is simplified into a one-dimensional
problem. The blast charts are for hemispherical clouds having an energy density of
3.5 MJ/m3. The blast charts result from numerical calculations using a constant
flame speed inside the cloud.
Other blast charts can be composed in order to differentiate more between the
numerous situations which may be present in reality. Other blast charts may inclu-
de:
• other energy densities;
• non-constant flame speeds: linear or exponential functions;
• obstructed regions with an aspect ratio other than one;
• arrival times to be able to superimpose blasts from various sources.
TNO report
PML 1998-C53 99
13 10 Conclusions and recommendations
The correlations to determine a single overpressure in an explosion blast source for

application in the Multi-Energy Method were applied to four realistic situations in
order to investigate the problems encountered.
The correlations consist of a relation between parameters describing the obstructed
region containing a flammable vapour cloud and describing the fuel.
Results of numerical calculations using the CFD code AutoReaGas served as a
reference for comparison with the results of the correlations; the numerical results
for the realistic situation for which experimental results are available.
The emphasis in this project was on the parameters Volume Blockage Ratio and
Average Obstacle Diameter. The results of numerical calculations were used as a
reference for comparison with results from the correlations. Due to this approach,
attention had to be given to the other parameters as well.
The Multi-Energy Method is a practical one-dimensional tool to determine blast

from vapour cloud explosions. The correlations to determine the source strength
were derived for homogeneous and symmetrical situations.
The more a realistic situation deviates from the homogeneous and symmetrical
situation, the less accurate the predictions for the overpressures in the source and in
the blast will be. The approach for application should therefore be a conservative
and safe one.
The exercises performed showed that the procedure given in the new Yellow Book
can be used for the determination of the volume of the obstructed region in order to
arrive at the Volume Blockage Ratio. This procedure is the best there is at the
present. It is not fully consistent but differences which may arise from that incon-
sistency will not lead to deviations outside the band of uncertainty of the correlati-
ons. Only for small and low obstacle density obstacles may the procedure lead to
less safe answers.
The average obstacle diameter to use should not be chosen too small. It was found
that in most cases, the hydraulic diameter of all obstacles together provided the best
answers.
A safe upperbound in most situations and scenarios appears to be to use the hy-
draulic diameter in combination with a flame path length equal to the radius of a
hemisphere with a volume equivalent to the volume of the obstructed region. This
approach is fully in line with the concept of the Multi-Energy Method.
The ‘white spots’ encountered while applying the correlations in the four cases are
all related to the deviations of the ideal simplified situation. They comprise:
• obstructed regions having different values for length, width and height;
TNO report
100 PML 1998-C53
• criteria for the definition of an obstructed region;

• criteria for multiple explosion sources;
• non-homogeneous obstacle distributions in size, location and orientation;
• non-central ignition locations.
All these items refer to the obstacle configuration. No problems were encountered
in incorporating the influence of the fuel.
An approach is suggested in order to pass the white spots while applying the cor-
relations.
For the white spots: ‘obstructed region with aspect ratios other than unity’ and
‘multiple explosion sources’, the suggested approach has a physical footing. Ho-
wever, experimental validation is lacking completely. Also, the results of the
numerical exercises do support some of the physical phenomena occurring but
seem to indicate that more complex criteria are requested. For instance, in the case
of an aspect ratio other than unity, there does not seem to be a general overpressure
threshold below which the flame will not accelerate after it has reached the nearest
edge of the obstructed region.
For situations where more than a single explosion source may be expected, and for
situations where the radius of an equivalent hemisphere appears to be an unaccep-
table approach, experimental data should be generated to develop and validate
procedures for the application of the Multi-Energy Method and the correlations.
TNO report
PML 1998-C53 101
14 11 References
Berg, A.C. van den; The, H.G.; Mercx, W.P.M.; Hayhurst, C.J.; Clegg, R.A.;
Robertson, N.J. and Birnbaum, N.K. (1995),
‘Gas Explosion Hazard Analyses Using the CFD Code AutoReaGas’,
Proceedings of the 6th Annual PETRO-SAFE Conference on Pipelines, Terminals
and Storage, 31 January – 2 February 1995, Houston, Texas, USA.
Selby, C.A. and Burgan, B.A. (1998),

‘Joint Industry Project on blast and fire engineering for topside structures phase 2’,
Steel Construction Institute, SCI Publication number 253, Ascot, England, 1998.
CPR14E (1997) (Yellow Book),

Methods for he calcultion of physical effects due to releases of hazardous materials
(liquids and gases),
Sdu Uitgevers, The Hague 1997.
Eggen, J.B.M.M. (1995),

‘GAME: development of guidance for the application of the Multi-Energy Me-
thod’,
TNO report PML 1995-C44, August 1995, Rijswijk.
EMEG (1997),
‘Explosion Model Evaluation Group, Specifications of test cases for gas explosions
- test case C1’,
EME project, DGXII, Brussels, Belgium, 1997.
Mercx, W.P.M.; Johnson, D.M. and Puttock, J. (1994),

‘Validation of scaling techniques for experimental vapour cloud explosion investi-
gations’,
AIChE Loss Prevention Symposium, Atlanta, Georgia, USA, 1994.
TNO report
102 PML 1998-C53
TNO report
PML 1998-C53 103
15 12 Acknowledgement
This project could only be accomplished due to the participation of the following
companies and organisations:
• Air Liquide (F);
• BP International Ltd. (UK);
• Elf Atochem (F);
• ENEL SpA CRIS (I);
• Gaz de France (F);
• Health and Safety Executive (UK);
• ICI (UK);
• INERIS (Fr);
• Norsk Hydro (N);
• RIVM (NL);
• Snamprogetti SpA (I).
Their contribution to the discussions during the progress meetings are highly
appreciated.
TNO report
104 PML 1998-C53
TNO report
PML 1998-C53 105
16 13 Authentication
W.P.M. Mercx A.C. van den Berg

Project leader/Author Author
D. van Leeuwen Dr. J. Weerheijm

Author Group leader
TNO report
106 PML 1998-C53
TNO report
PML 1998-C53 A.1

Annex A
Annex A Procedure for the determination of the boun-

daries of the obstructed region according to the
Yellow Book
Copy of paragraph 5.5.3 of CPR14E (1997)
16.1.1 5.5.3 Procedure for the division of an area into obstructed and
unobstructed regions
A procedure to subdivide a region into obstructed and unobstructed parts is pre-

sented next.
The procedure given is thought to be a safe and conservative one in a sense that
always a too large volume of the obstructed region will be selected. The procedure
offers a possibility for optimisation, i.e. reduction, of the obstructed volume. This
optimisation procedure can be followed but obviously requires more time. The
optimisation procedure in itself will not lead to unsafe situations.
The procedure to build up an obstructed region is based on the effect obstacles

have on the generation of turbulence in the expansion flow ahead of the flame. A
zone with obstacle-induced turbulence will exist behind an obstacle. The length of
this zone is related to a characteristic dimension of the obstacle. However, if the
scale of an obstacle becomes larger it is assumed that the length of the influenced
zone is bounded by an upper value.
The procedure is as follows.
• Step 1: break-down structures into basic geometrical structural shapes
Structures in a potentially hazardous area like an industrial site may be regarded as

being composed of (or fairly good bounded by) basic geometrical shapes:
- cylinders with length lc and diameter dc;
- boxes with dimensions b1, b2, b3;
- spheres with diameter ds.
• Step 2: assume an ignition location
After ignition of a flammable cloud in a congested area the flame will travel out-
ward, so the orientation with respect to the flame propagation direction of each
obstacle is known.
TNO report
A.2 PML 1998-C53

Annex A
• Step 3: determine obstacle orientation
Let the smallest dimension oriented in a plane perpendicular to the flame propaga-
tion direction be D1 then:
D1 = lc or dc for cylinder
D1 = smallest of b1 and b2, b2 and b3, or b1 and b3 for a box
D1 = ds for a sphere
Let the obstacle dimension parallel to the flame propagation direction be D2.
• Step 4: build-up of obstructed region
An obstacle belongs to an obstructed region if the distance from its centre to the
centre of any obstacle in the obstructed region is smaller than 10 times D1 or 1.5
times D2 of the obstacle under consideration in the obstructed region (D1 and D2
belonging to any obstacle in the obstructed region).
If the distance between the outer boundary of the obstructed region and the outer
boundary of the obstacle is larger than 25 m, then the obstacle does not belong to
that obstructed region.
Note: there are values given to three factors. These are a factor times D1 and a
factor times D2 (respectively 10 and 1.5) for obstacle distances and a value (25 m)
for the distance between obstructed region and obstacle to determine if the obstacle
is part of that obstructed region. It is not yet possible to quantify these values
accurately. The values given are thought to be safe.
• Step 5: defining a box containing the obstructed region
The obstructed region is defined as a box that contains all the obstacles in the
obstructed region:
including: the space between a confining surface and an obstructed region where
the distance between that surface and any obstacle in the obstructed region is less
than 10 times D1 or 1.5 times D2 (for instance, the earth’s surface)
excluding: parts of cylinders or boxes that obviously do not belong to the obstruc-
ted region like upper parts of chimneys, distillation columns (vertically oriented
cylinders) or pipes (horizontally oriented cylinders) connecting, for instance,
chemical units, each potentially being an obstructed region, at a chemical plant.
The excluded parts may form an obstructed region themselves.
The free volume of the obstructed region Vr is the volume of the box minus the
space occupied by the obstacles. In case it is not possible to calculate or estimate
the volume occupied by the obstacles, assume Vr equals the total volume of the
box.
TNO report
PML 1998-C53 A.3

Annex A
• Step 6: subdivision into multiple boxes
The box containing the obstructed region smooths the outer boundary of the actual
obstacle boundaries, thereby including additional space free of obstacles. In case
this free space is not included in the obstructed region by the procedure for buil-
ding up the obstructed region but by drawing the box around the obstructed region,
sub-division into multiple directly adjacent boxes is acceptable for reducing the
volume of the obstructed region.
• Step 7: define additional obstructed regions if appropriate
If not all obstacles present are inside the obstructed region, perhaps more obstruc-
ted regions within the cloud can be defined.
In that case, the region where ignition occurs is called the ‘donor’ region; the other
regions are ‘acceptor’ regions.
The direction of flame propagation, required for the orientation of the obstacles in
the acceptor region, depends on the orientation of the acceptor region with respect
to the donor region.
If separate obstructed regions are located close to one another, they may be initia-
ted more or less at the same time: then coincidence of their blasts in the far-field
may not be ruled out and the respective blasts should be superposed.
At the present, their is no guidance on the minimum separation distance between
donor and acceptor regions at which they can be assumed to be separate explosi-
ons. Due to this lack, a great possibility of reducing the hazards of a vapour cloud
explosion cannot be applied. The procedure for applying the Multi-Energy Method
described in the next section contains a safe and conservative approach to cover
this deficiency.
Remarks
Due to the choice of having a high initial blast strength for an obstructed region,
the need to have a clear definition of an obstructed regions grows.
The exact boundaries of an obstructed region are, in fact, not so important for the
determination of blast. According to the blast charts (Figure 5.8), the energy E has
to be raised to the power of 1/3.
Nevertheless, a solid definition might be required.
The subdivision of a hazardous site into obstructed and unobstructed regions and
the attribution of a source strength to each region is a major issue of ongoing
research. As a simplified Multi-Energy Method is applied here, a simplified proce-
dure for the subdivision into obstructed and unobstructed regions is adapted.
TNO report
A.4 PML 1998-C53

Annex A
In order to obtain a rather straightforward, easy to apply and fully closed procedu-
re, a number of obvious influences which lead to reduction of the explosion seve-
rity have been neglected. This indicates that further optimisation is possible, but
this should be obtained through consultation with experts.
.TNO report
PML 1998-C53 B.1

Annex B
Annex B Procedure for the application of the Multi-

Energy Method according to the Yellow Book
16.1.2 5.5.4 Procedure for the application of the Multi-Energy Method
The procedure of vapour cloud explosion blast modelling according to the Multi-
Energy concept can be subdivided into a number of steps. Figure 5.9 shows these
successive steps to be taken.
Apply constraints
Determine cloud size
Recognise potential
blast sources
Define obstructed regions
Estimate boundaries of section 5.5.3

obstructed regions or use
procedure of section
5.5.3
Break down structures into

basic geometric structural shapes
Assume an ignition location
Determine obstacle orientation
TNO report
B.2 PML 1998-C53

Annex B
Build up obstructed region
Define a box containing

the obstructed region
Subdivide into multiple boxes
Define additional obstructed

regions if appropriate
Estimate the source strength

or class number for each region
Combine of obstructed regions
Locate of unobstructed
part of vapour cloud
Calculate radius
Calculate blast parameters
Handle of multiple
obstructed regions
Construct blast history

at specific location
Figure 5.9: Flow diagram for application of the method.
.TNO report
PML 1998-C53 B.3

Annex B
• Step 1: apply constraints
- Realise that the Multi-Energy Method is to determine the blast parameters from
‘unconfined’ vapour cloud explosions. Blast from vented vapour cloud explosi-
ons and from internal explosions should be assessed by other methods.
- Realise that the Multi-Energy Method is a simplification of reality. It does not
take into account directional blast effects due to the inhomogeneous distribution
of confinement and obstruction of the vapour cloud or due to the non-point
symmetrical shape of the vapour cloud. Use other, more sophisticated, models
to assess these aspects.
- Assume that blast modelling on the basis of deflagrative combustion is a suffi-
ciently safe and conservative approach. Keep in mind that unconfined vapour
cloud detonation is extremely unlikely and has only one single precedent, to our
present knowledge.
• Step 2: determine of cloud size
- Determine the mass quantity after an accidental release that can contribute to
the formation of a flammable cloud.
In general, a dispersion calculation will preceed the explosion calculation. See

section 5.7 on how to calculate the flammable mass quantity within a cloud.
If no dispersion calculation is made, the mass quantity has to be estimated. In a
safe approach, one could assume that the whole mass inventory contained within
the process unit under consideration contributes to the formation of a flammable
cloud. In the case of pool evaporation, one could multiply the evaporation rate by a
certain time period to come up with a mass quantity.
- Calculate the volume Vc of a cloud with a density ρ containing the flammable

mass quantity Qex at stoichiometric concentration cs (cs in %) with:
Vc = 100 · Qex · ρ / cs (m3) (5.7)
• Step 3: recognise of potential blast sources
- Identify potential sources of blast in the vicinity of a postulated location of the

centre of the cloud.
Potential sources of strong blast are, for example:

§ extended spatial configuration of objects, e.g. process equipment at chemical
plants or refineries, piles of crates;
§ the space between extended parallel planes, e.g.:
- underneath groups of closely parked cars in car parks or marshalling yards;
- open buildings, e.g. multi-storey car parks;
TNO report
B.4 PML 1998-C53

Annex B
§ the space in tube-like structures, e.g.:

- tunnels, bridges, corridors;
- sewage systems, culverts;
§ an intensely turbulent fuel-air mixture in a jet due to release at high pressure.
• Step 4: define obstructed regions
- Define obstructed regions with the procedure given in the previous section or
estimate boundaries of obstructed regions and determine the free volume Vr of
each obstructed region.
- Determine the maximum part of the cloud Vgr that can be inside the obstructed
regions.
- Calculate the volume Vo of the unobstructed part of the vapour cloud with:
Vo = Vc - Vgr (m3) (5.8)
- Calculate the energy E of each region, obstructed as well as unobstructed, by

multiplying Vgr and Vo by the combustion energy per unit volume.
• Step 5: estimate the source strength or class number for each region
As explained before, a conservative approach is to choose a class number of 10 for

each obstructed region. Other numbers may be chosen based on additionally obtai-
ned information if required. Choose a low initial blast strength for the remaining
unobstructed regions: number 1. In case initial low turbulence motion is expected
in unobstructed regions, for instance, due to the momentum of the fuel release, a
number of 3 is advised.
• Step 6: combine of obstructed regions
In case more than one obstructed region has to be considered:

- define an additional blast source (obstructed region) by adding all the energies
of the separate blast sources together and by assuming a centre for the additio-
nal blast source. This centre can be determined by considering the centres of the
separate blast sources and their respective energies.
.TNO report
PML 1998-C53 B.5

Annex B
• Step 7: locate unobstructed part of vapour cloud
- Determine a centre for the unobstructed vapour cloud volume. This centre can
be determined by considering the centres of the separate unobstructed regions
and their respective energies.
Now a certain number of explosion centres and their respective energies E have
been determined. For each centre, the blast parameters as a function of distance to
its centre can be calculated with the blast charts given in Figure 5.8.
• Step 8: calculate radius
- Model the blast from each source by the blast from an equivalent hemi-
spherical fuel-air charge of volume E/Ev m3 (Ev = 3.5 MJ/m3 is an average va-
lue for most hydrocarbons at stoichiometric concentration).
Get an impression of the scale by calculating the radius ro for each blast source
from:
ro = (3/2 · E/(Ev · π))1/3 (m) (5.9)
• Step 9: calculate blast parameters
The blast parameters at a specific distance r from a blast source can be read from
Figure 5.8 A, B and C after calculating the scaled distance r’.
- Calculate the scaled distance r’ with:

r’ = r/(E/pa)1/3 (-) (5.2)
- Depending on the class number 1, 3 or 10, read the scaled peak side-on over-
pressure Ps’, the dynamic peak pressure pdyn’ and the scaled positive phase du-
ration tp’ from the respective blast charts in Figure 5.8.
- Calculate the peak side-on overpressure Ps, the peak dynamic pressure pdyn and
the positive phase duration tp with:
Ps = Ps’ · pa (Pa) (5.3)
pdyn = pdyn’ · pa (Pa) (5.5)
tp = tp’· (E/pa)1/3/aa (s) (5.4)
TNO report
B.6 PML 1998-C53

Annex B
- Determine the shape of the blast wave from the Figure 5.8.
- Calculate the positive impulse is by integrating the overpressure variation over

the positive phase, which can be approximated by multiplying the side-on over-
pressure by the positive phase duration and by a factor of 1/2:
is = 1/2 · Ps · tp (Pa·s) (5.6)
• Step 10: handle multiple obstructed regions
If separate blast sources are located close to one another, they may be initiated
more or less at the same time. Therefore, coincidence of their blasts in the far-field
may not be ruled out, and the respective blasts should be superposed This should
be accomplished by taking the respective quantities of combustion energy of the
sources in question together.
To determine the parameters in the blast wave at a specific distance, one should use
the blast source as defined in step 6.
• Step 11: construct the blast history at a specific location
Combustion in the unobstructed region is considerably different to combustion in

an obstructed region. Blast from an obstructed region will result in sharp and
relatively short peaks (shock waves). The relatively slow combustion in an unob-
structed region will result in pressure waves of long duration. Due to the influence
of the respective energies however, either of the blast waves can be of importance
at a specific location.
It may be assumed that the blast history at a specific location consists of the blast
parameters and blast shape resulting from the obstructed region on which the blast
parameters of the unobstructed region is superposed.
TNO report
PML 1998-C53 C.1

Annex C
Annex C Application of procedure to determine obstruc-

ted region boundaries
16.1.3 5.6.2 Definition of an obstructed region
The boundaries of obstructed regions can be chosen by carefully considering the

layout of a hazardous site. In case of doubt, or in case a more precise definition is
required, the more time-consuming procedure of section 5.5.3 can be followed.
That procedure will be followed here.
As an example of the build-up of obstructed regions on a hazardous site, an LPG

storage and distribution centre is chosen. The picture in Figure 5.10 shows an
overview. Figure 5.11 gives a map from which becomes clear that the LPG instal-
lation at San Juan Ixhuatepec near Mexico City was the basis for this example
[Pietersen, 1984].
The facility comprises six spherical storage tanks; four with a volume of 1600 m3
and two with a volume of 2400 m3. An additional 48 horizontal cylindrical bullet
tanks are situated near the large spheres.
The total storage capacity is about 16000 m3.
The storage is subdivided into six zones, see Figure 5.12.
Horizontal distances are given in Figure 5.13.
Additional dimensions are:

- the minimum height underneath the bullet tanks is 2 m;
- the minimum height underneath the spheres is 2 m;
- additional pipework on top of cylinders and spheres has a height of 0.5 m;
- the storage area is surrounded by an open corridor of 30 m width;
- alongside the length of the outer cylinders, stairs with a width of 1.5 m are
situated.
TNO report
C.2 PML 1998-C53

Annex C
Figure 5.10: Picture of LPG storage facility.
At first glance, it will be clear that any of the groups of cylinders or spheres are an
obstructed region.
• Step 1: break-down of structures into basic geometrical structural shapes
The structures consist of big spheres and cylinders. Other structures like stairs,
walkways, supports and pipes will not be considered here. This is a valid approach:
the dimensions of the large obstacles will dominate the process of building an
obstructed region: 10 times D1 of a big obstacle will override 10 times D1 of a
small obstacle located near the larger one (step 4). Figure 5.12 shows all major
basic geometrical structural shapes.
• Step 2: assume an ignition location
Assume ignition in the centre of the southern group of cylinders in zone 4 (Figure
5.13).
The build-up of the obstructed region is started by taking the cylinder west of the
ignition location as a start (marked as 1).
TNO report
PML 1998-C53 C.3

Annex C
95374-5.11
20
24
14
unigas gasomatico
1 2
6 9 19
5
3 22 13
8
7 4 12
15
10
18
11
16 17
21 23
Legend:
1 2 spheres of 2400 m3, ds = 16.5 m 12 car loading
2 4 spheres of 1600 m3, ds = 14.5 m 13 gas boiler store
3 4 cylinders of 270 m3, dc = 3.5 m, length = 32 m 14 pipe/valve manifold
4 14 cylinders of 180 m3, dc = 3.5 m, length = 21 m 15 waterpower
5 21 cylinders of 36 m3, dc = 2 m, length = 13 m 16 LPG storage Unigas
6 6 cylinders of 54 m3, dc = 2 m, length = 19 m 17 Storage Gasomatico
3 cylinders of 45 m3, dc = 2 m, length = 16 m 18 battling Terminal
19 depot cars with bottles
7 flare pit 20 entrance
8 pond 21 rail car loading
9 control room 22 store
10 pumphouse 23 watertank
11 fire pumps 24 garrison
Figure 5.11: Plan of LPG installation.
95374-5.12
1 2 6
3 4 5
Figure 5.12: Division into zones.
TNO report
C.4 PML 1998-C53

Annex C
10m 95374-5.13
12.75
14.5 10.5
8.5
12.75
6
13
5 43 2 1
*
1.5 11.5 1.5 10 3 2
3.5 3.5
* ignition
Figure 5.13: Build-up of obstructed region.
• Step 3: determine obstacle orientation
In order to decide if cylinder 2 is part of the obstructed region, values for D1 and
D2 have to be determined. Following the direction of flame propagation from the
ignition location via cylinder 1 towards cylinder 2, the orientation of the axis of
cylinder 1 is perpendicular to the direction of flame propagation. Therefore, both
D1 and D2 equal the diameter 3.5 m in this case.
• Step 4: build-up of obstructed region
The distance between cylinder 1 and the adjacent one, cylinder 2, is 1.5 m. This
distance is shorter than 10 times D1 and also shorter than 1.5 times D2. Thus,
cylinder 2 is part of the obstructed region.
The obstructed region is enlarged to include cylinders 1 and 2 now.
It is obvious that all seven cylinders of the southern group in zone 4 are within one
obstructed region.
The distance between the east cylinder in zone 3 (number 5) and the west cylinder
of the obstructed region (number 4) is 11.5 m. D1 and D2 both equal 3.5 m still.
Thus, cylinder 5 belongs to the obstructed region too.
It is derived very easily that all four cylinders of zone 3 belong to the obstructed
region also.
TNO report
PML 1998-C53 C.5

Annex C
The distance between the northern and southern group of cylinders in zone 4 is
13 m. In this case, D1 = 3.5 m and D2 = 21 m. The distance of 13 m is shorter than
10 times D1 and 1.5 times D2. Thus, cylinder 6 and all other cylinders in that group
are part of the obstructed region.
Repeating the procedure for the other cylinder groups as well as the spheres results
in one obstructed region covering all cylinders and spheres.
Additional pipework and stairs located within a distance of D1 or D2 from a sphere
or cylinder belongs to the obstructed region too.
• Step 5: define box containing the obstructed region
A box containing all obstacles in the obstructed region has dimensions of length
108.5 m, width 110 m and height 19.5 m (Figure 5.14; 19.5 m is the height of the
highest cylinder including pipework on top).
The space underneath the obstacles is part of the obstructed region too; the distance
of 2 m to a confining surface (the earth) is smaller than 10 times D1 or 1.5 D2 or all
big obstacles.
A single big box in this case includes large volumes of free space, due to the large
variation in height between cylinders and spheres. A sub-division in multiple
directly adjacent boxes is therefore beneficial.
10m
95374-5.14
110m
108.5m
Figure 5.14: Single box.
TNO report
C.6 PML 1998-C53

Annex C
95374-5.15
3 2
Figure 5.15: Three boxes.
• Step 6: subdivision in multiple boxes
A first subdivision into 3 boxes is shown in Figure 5.15.

Box 1 contains the cylinders in zones 3, 4 and 5;
dimensions: length 108.5 m, width 65 m and height 6 m.
Box 2 contains the cylinders in zone 6;
dimensions: length 45 m, width 31.25 m and height 4.5 m.
Box 3 contains all spheres in zone 1 and 2;
dimensions: length 65.5 m, width 45 m, height 19.5 m.
A second further subdivision is possible as the free space in box 1 north of zone 3
does not belong to the obstructed region. All other free space between groups of
cylinders are part of the obstructed region and cannot therefore be excluded by
further subdivision into multiple boxes.
95374-5.16
4 3
* centre of cloud
Figure 5.16: Four boxes.
TNO report
PML 1998-C53 C.7

Annex C
We now have four boxes (Figure 5.16).

Box 1 contains the cylinders in zone 3 and the space between zones 3 and 4;
dimensions: length 32 m, width 31.5 m, height 6 m.
Box 2 contains the cylinders in zones 4 and 5 and the space between zones 4
and 2, and 5 and 6;
dimensions: length 77 m, width 65 m and height 6 m.
Box 3 contains the cylinders in zone 6;
dimensions length 45 m, width 31.25 m and height 4.5 m.
Box 4 contains all spheres in zone 1 and 2;
dimensions: length 65.5 m, width 45 m, height 19.5 m.
The volume of the obstructed region is the sum of the volumes of the four boxes
minus the space occupied by the obstacles:
Box 1 32 · 31.5 · 6 = 6048 m3
Box 2 77 · 65 · 6 = 30030 m3
Box 3 45 · 31.25 · 4.5 = 6328 m3
Box 4 65.5 · 45 · 19.5 = 57476 m3
Total volume = 99882 m3
The volume of the cloud inside the obstructed region is determined by deduction of
the volume of the cylinders and spheres of the volume of the obstructed region.
The storage capacity is 16000 m3, which is assumed to equal the volume of the
storage vessels.
So: Vgr = 99882 - 16000 = 83882 m3
Obviously further subdivision is possible and acceptable as long as the rules for
subdivision are followed. One has to balance the benefit from further subdivision
(which can be rather small) with the required time to perform the subdivision
(which can be rather long) in order to decide to which level of detail subdivision is
carried out.
• Step 7: define additional obstructed regions if appropriate
As all obstacles present are inside the one obstructed region defined, no additional
obstructed regions are requested.
TNO report
C.8 PML 1998-C53

Annex C
TNO report
PML 1998-C53 D.1

Annex D
Annex D Critical separation distances between obstruc-

ted areas
16.2 D.1 Introduction
The Multi-Energy (ME) method is a simple method for vapour cloud explosion
blast modelling. A vapour cloud’s explosive potential is modelled by the specifica-
tion of an equivalent explosive fuel-air charge whose blast characteristics can be
read from blast charts.
The ME method recognises that in gas deflagration, turbulence generative bounda-
ry conditions are the predominant factor in the generation of overpressure and
blast. The mechanism of a gas explosion implicates that, as soon as the appropriate
turbulence generative boundary conditions are lacking, the burning speed and the
pressure build-up in the process of flame propagation will slump. The direct conse-
quence of the ME concept is that an extended vapour cloud containing several
obstructed areas, separated by open spaces of sufficient extent, will produce the
same number of separate blast waves on ignition. In the modelling process of blast
effects, the individual obstructed areas should be considered separately.
The objective of this report is to develop some practical simple guidance for the
quantification of the term ‘open spaces of sufficient extent’. In other words: how
large should an open space be between two obstructed areas to be treated as sepa-
rate sources of blast? Or, the other way around, how closely should obstructed
areas be located to one another, to be treated as one single blast source? Some
general notion of the critical separation distance can be developed by considering
the phenomena and mechanisms that govern the gas dynamics in gas deflagration.
16.3 D.2 Mechanisms and phenomena
Pressure build-up in gas deflagration is a consequence of the expansion, a process

of flame propagation is attended by. Under isobaric conditions, a stoichiometric
hydrocarbon-air mixture expands up to approximately a factor 8 in volume on
combustion. The consequence is that a combustion process induces a flow field in
its environment. The propagating flame front is carried along in this flow. The
flame propagation will behave dependently on the flow structure met by the flame.
The flow structure (i.e. velocity gradients and turbulence) is largely determined by
the boundary conditions of the flowfield, i.e. the solid boundaries and objects
present in the space in which a gas explosion develops. Boundary conditions,
therefore, play a predominant role in the development of overpressure in gas de-
flagration.
TNO report
D.2 PML 1998-C53

Annex D
Turbulence generative boundary conditions trigger a feedback coupling in the

interaction between the combustion and the expansion flow. The turbulence inten-
sifies the combustion rate and thereby flame speed and expansion flow velocity.
Higher flow velocities go hand in hand with higher turbulence intensities, which in
turn intensify combustion again, etc., etc.. As long as the proper turbulence gene-
rative boundary conditions are present, the development of flame speed and over-
pressure is a self-amplifying process.
During the development of a deflagration, the maximum pressures and medium

velocities are produced just in front of the combustion zone. As soon as the flame
propagates out of an obstructed and/or partially confined area, the proper boundary
conditions to maintain the feedback coupling are lacking and the burning speed
slumps. The combustion wave and pressure/velocity wave decouple. The pressu-
re/velocity wave is the blast wave which propagates out and decays in space.
The turbulent motion induced by the boundary conditions will be limited to the
medium originally present within the obstructed, turbulence generative area. There-
fore, the explosive combustion will not extend beyond the medium originally
present within the obstructed area. This observation may lead to a very simple
theoretical delimitation of the critical separation distance.
16.4 D.3 Theoretical considerations
The isobaric expansion ratio of stoichiometric hydrocarbon-air mixtures is equal to

approximately 8. In other words, a stoichiometric hydrocarbon-air mixture expands
up to a volume 8 times as large as its original volume upon combustion. The dis-
tance covered by this isobaric expansion is dependent on the degree of confinement
by which the expansion is limited. Basically, three different degrees of confine-
ment can be distinguished:
• 3D expansion
A spherical charge will expand up to a sphere of 3 8 = 2 times the original charge
radius upon combustion. In more generalised terms: a fully unconfined cloud
expands up to a distance of twice its original linear dimensions.
TNO report
PML 1998-C53 D.3

Annex D
R
2R (burnt)
Figure D.1a: Unconfined charge expanding spherically.
• 2D expansion
A cylindrical charge, confined by parallel planes, will expand up to 8 ≈ 2.8 times
its original charge radius upon combustion.
R
2.8R (burnt)
Figure D.1b: Charge confined between parallel planes, expanding radially.
• 1D expansion
A charge confined by a tube or channel-like structure will expand up to 8 times its
original linear charge size upon combustion.
R
8R (burnt)
Figure D.1c: Charge confined by channel structure, expanding linearly.
TNO report
D.4 PML 1998-C53

Annex D
Because of the assumed stoichiometry, these figures may be regarded as safe and
conservative upper limits for the separation distance. However, for the 2D and 1D
case, these figures may be substantially reduced.
If a space confined between parallel planes is completely filled with a flammable

mixture, the expansion of the medium will partly take place outside the confine-
ment where 3D expansion is possible. The flame jet will not only expand in a
purely radial direction but upward as well. A mushroom-shaped flame jet is gene-
rally observed (Figure D.2a) which does not extend over the full 2.8 times the
original charge radius.
Figure D.2a: The mushroom-shaped flame jet produced by a charge confined between
parallel planes.
Therefore, the recommended value for the separation distance for a flammable
mixture confined between parallel planes may be reduced. Experimental data on
the length of flame jets from this type of configuration are not available. For the
time being, therefore, the recommended value for the separation distance is assu-
med equal to that for the 3D, unconfined charge, i.e. 2 times the linear dimension
of the charge radius.
In the same way, the flame jet in the 1D case will possibly be substantially shorter
than 8 times the linear dimension of the original charge as it expands from a chan-
nel into 3D space.
Fortunately, giving a concrete value for the 1D case is not relevant as most practi-
cal application of the ME concept is to chemical plants whose layout can be ap-
proximated by the 3D or 2D cases.
16.5 D.4 Experimental observations
Flame jet length

Experimental data on flame jet length are limited to dust or gas explosion venting.
In such experiments the overpressure, induced by a gas explosion in a closed
vessel, is relieved by venting through an orifice of a size which is generally sub-
stantially smaller than the cross-sectional dimension of the vessel.
TNO report
PML 1998-C53 D.5

Annex D
Only few data on the flame jet length in vented explosions were found in papers by
Wirkner-Bott (1992) and Bartknecht (1993). For vented dust explosions, Bart-
knecht (1993) poses a rough empirical relation which relates the flame jet length Lf
to the vessel volume V.
Lf = 8.V0.3
This expression also covers experimental data generated by Wirkner-Bott et al.
(1992).
For vented propane-air explosions, Bartknecht (1993) poses the relation

Lf = 3.1V0.402
The set-up venting experiments show some resemblance with the 1D case. They
confirm the basic idea underlying the theoretical considerations that the flame jet
length is approximately proportional to a linear dimension of the original cloud
configuration.
Donor-acceptor experiments
At TNO Prins Maurits Laboratory, Van Wingerden (1988 and 1989) investigated a
gas explosion developing in two subsequent obstructed areas separated by a free
space, i.e. a separation distance. The obstructed areas consisted of a configuration
of concrete cylinders confined between parallel planes.
The gap size between the two obstructed areas was varied. The results showed that
as long as the separation distance was larger than the linear dimension of the donor
array, the blast waves produced by the respective configurations were clearly
separated. This result does not contradict the values for the separation distance,
proposed above.
16.6 D.5 Generalisation
For application to realistic problems, i.e. the (petro-)chemical industries, the rules
concerning the separation distance must be generalised. A plant can be regarded
either as a spatial configuration of equipment (obstacles) or as a configuration of
equipment confined between the floors of an open building. Such a configuration
of obstacles definitely has not the idealised spherical or cylindrical shape the
theoretical considerations were based on. Therefore, the rule for the critical separa-
tion distance is generalised.
The critical separation distance around a potential blast source area is equal to
half its linear dimension in each direction. If the distance between potential sour-
ces is larger, the sources should be modelled as separate blasts. If not, they should
be modelled as one single blast of summed energy content.
TNO report
D.6 PML 1998-C53

Annex D
16.7 D.6 References
Bartknecht, W. (1993),
Explosionsschutz,
Springer Verlag,
Berlin, 1993.
Wirkner-Bott, I.; Schumann, S. and Stock, M. (1992),

Dust explosion venting: Investigation of the secondary explosion,
7th Int. Symp. on Loss Prev. and Safety Prom. in the Process Industries,
Taormina, Italy, 408 May, 1992.
Wel, P.G.J. van der (1993),

A review of blast phenomena during vented explosions,
TNO Prins Maurits Laboratory report no. PML 1993-C111.
Wingerden, C.J.M. van (1988),

Investigation into the blast produced by vapour cloud,
explosions in partially confined areas,

Experimental investigation into the strength of blast waves generated byvapour
cloud explosions in congested areas,
6th Int. Symp.’Loss Prevention and Safety Promotion in the ProcessIndustries’,
Oslo, Norway, 1989, proceedings, pp. 26-1, 26-16.

Prediction of pressure and flame effects in the direct surroundings,
of installations protected by dust explosion venting,
J. Loss Prev. Process Ind.,Vol. 6, No.4, (1993), pp. 241-249.
TNO report
PML 1998-C53 E.1

Annex E
Annex E Application of GAME correlation to obstacle

configurations of high aspect ratio
16.8 E.1 Introduction
The aspect ratio of an obstacle configuration can be defined as its length/width or

length/height ratio. All the experiments that, the GAME correlation was compiled
from were performed in obstacle configurations of an aspect ratio approximately
equal to 1. An obstacle configuration aspect ratio equal to 1 allows the flame front
propagation process to develop more or less symmetrically. Over almost the entire
flame path, the flame is allowed to maintain its approximately cylindrical or sphe-
rical (closed in itself) shape. Effects of side or back relief do not play a role.
Generally speaking, practical problems (plants in the chemical industries) are of

quite different proportions. Often, practical plant equipment configurations are of a
length/width ratio and/or a length/height ratio much higher than one. The aspect
ratio of an obstacle configuration and/or the aspect ratio of the cloud in the obstacle
configuration may largely influence the development of the flame propagation
process and, therefore, may largely determine the ultimate explosion overpressure.
A straight application of the GAME correlation to high aspect ratio obstacle confi-
gurations may result in a considerable overestimation of the explosion pressure.
This report suggests a possible way to implement the aspect ratio of an obstacle
configuration as an additional factor in the application of the GAME correlation.
16.9 E.2 Phenomena and experimental observations
The development of a deflagrative gas explosion is largely determined by the

feedback coupling in the interaction of the combustion process with the structure of
its self-generated expansion flow field ahead of the flame. The feedback is trigge-
red by turbulence generative boundary conditions. The feedback coupling and
consequently the self-amplification of the flame propagation process is optimum
when a maximum medium velocity (and consequently a maximum turbulence
intensity) is generated ahead of the flame as result of a minimum burning speed. In
other words: when the ratio of medium velocity ahead of the flame and the flame
front’s burning speed is maximum, the feedback coupling is optimum and the self-
amplification capability of the process is maximum.
An optimum feedback coupling is obtained when the combustion products behind
the flame front cannot expand freely backward. This is the case when the combus-
tion products (the flame bubble) are fully enclosed either by the flame front itself
and/or by rigid boundaries. Then, the expansion is fully utilised to generate flow
ahead of the flame.
TNO report
E.2 PML 1998-C53

Annex E
When a reactive mixture in an obstructed area is not ignited at the area’s centre but
at an edge, the flame propagation process finds the proper conditions to speed up
only inside obstructed area. Then the combustion products behind the flame are not
stagnant but expand partly freely backwards (back relief) without contributing fully
to the medium velocity ahead of the flame. Then the expansion of the products is
not fully utilised for the generation of flow in the reactants ahead of the flame front
in the obstructed area. The feedback coupling is not optimum. Therefore, an edge-
ignited obstructed area will develop more slowly than a centrally ignited area over
the same distance of flame propagation.
Similar effects play a role in the development of a gas explosion in an obstacle

configuration of aspect ratio higher than 1 (Figure E.1). The position of the flame
front at three consecutive points in time has been indicated.
Figure E.1: Flame propagation in an obstacle configuration of an aspect ratio higher

than 1.
The flammable mixture is ignited near the centre of the obstacle configuration.
Initially, as long as the entire flame front propagates within the obstructed area, the
flame bubble (closed in itself) develops more or less symmetrically because it finds
similar boundary conditions all around. During this initial stage of flame propaga-
tion, the aspect ratio of the obstructed area does not play a significant role.
Then the flame propagates out of the obstructed area at the long sides into the open
where the proper boundary conditions for pressure build-up are lacking. The con-
sequence is that from that point on, the part of the flame front outside the obstruc-
ted area does not fully contribute to the generation of expansion flow ahead of the
flame in the obstructed area and thereby does not contribute to the feedback any
longer. The flame propagation inside the obstructed area is less supported by
expansion from the back as a consequence of what is called side relief.
Almost all of the experimental data on gas explosions have been derived from
experiments in obstacle configurations of aspect ratios approximately equal to 1.
So far, only very limited data on gas explosion development in obstacle configura-
tions of an aspect ratio substantially higher than one are available. Two sources of
data were found: Van Wingerden (1988 and 1989) and Harris and Wickens (1989)
Van Wingerden (1988 and 1989) reports on a number of flame propagation expe-
riments performed in a 4× 4 m2 rig consisting of a configuration of circular rows of
TNO report
PML 1998-C53 E.3

Annex E
vertical 0.08 m diameter cylinders confined between parallel planes with a 0.16 m
interspace. The cylinder configuration was characterised by a 50% area blockage in
a row of cylinders and a pitch between the respective rows of 3 cylinder diameters.
The obstacle configuration was filled with a stoichiometric ethylene-air mixture
and ignited by a spark.
Ignition at the edge of the configuration showed a slower initial development than
for central ignition. However, a higher ultimate flame speed and overpressure were
obtained simply because the flame path length was twice as long and the number of
obstacle rows passed by the flame was twice as high as in the case of central igniti-
on.
Subsequently, the width of the obstacle configuration was varied from 4 to 2 and
1 m so that the configuration’s aspect ratio varied from 1 to 2 and 4. The length of
the rig (distance of flame propagation) was kept constant and equal to 4 m. The
configurations were ignited in the centres of their short edges.
In aspect ratio 1 and 2 experiments, flame speed and overpressure quickly ran up to
respective maximum values of over 600 m/s and 600 kPa within the 4 m propagati-
on distance. Possibly the maximum flame speed was limited by choking flow
conditions.
The aspect ratio 4 experiment, on the other hand, showed quite different behaviour,
namely: a slow development which resulted in an ultimate flame speed of about
150 m/s and a maximum overpressure of approximately 30 kPa.
The key point in these experiments is the big difference in the flame speed and
overpressure development between the aspect ratio 1 and 2 experiments on the one
hand and the aspects ratio 4 experiment on the other. The experiments suggest that
a certain width of configuration or a certain number of obstacles in lateral direction
is required to develop a high level of flame speed and overpressure. If that width of
configuration or that number of obstacles is present, the flame propagation process
develops quickly and independently of side relief effects. If not, the flame propa-
gation process develops slowly.
Harris and Wickens (1989) report on flame propagation experiments with natural
gas, propane and cyclohexane-air mixtures in a very elongated 3× 3× 45m3 fully
unconfined rig whose cross-section was obstructed by vertical arrays of horizontal
tubes. The tube arrays, characterised by an area blockage of 40%, were placed at a
mutual distance of 1.5 m.
If the rig was ignited by a spark, a slow but gradual increase in flame speed over
the full length of the rig was observed. If the rig was initiated with a flame jet at a
flame speed higher than approximately 500 m/s, the high flame speed was more or
less maintained. Initiation at lower flame speeds, on the other hand, resulted in a
slump of the process down to much lower values, from which the slow and gradual
speed-up started.
TNO report
E.4 PML 1998-C53

Annex E
Generalisation
The experimental observations suggest some sort of a jump in behaviour. If the
lateral dimension of the obstructed area is too small to develop some critical value
of flame speed and overpressure before the flames reaches the lateral edge of the
configuration, the development of the process will be largely governed by side
relief and the ultimate explosion overpressure will be low. If, on the other hand, the
lateral size is sufficient to exceed this critical value before the flame reaches the
lateral edge of the configuration, the process will develop independently of side
relief effects and the ultimate explosion overpressure will be high.
Modelling
Given a 3D obstructed area of dimensions: a length L, a width W and a height H.
The area is characterised by an obstacle configuration of an average obstacle size
D. The area is fully engulfed by a vapour cloud of a flammable mixture characteri-
sed by a laminar burning speed SL.
The experimental observations suggest a way of implementation of the aspect ratio
in the application of the GAME correlation which roughly reflects the behaviour
observed experimentally. The procedure is as follows.
• Check the lateral dimensions of the obstructed area on pressure build-up capa-
bility. Apply the GAME correlation to the minimum dimension ½W or H. In
other words: evaluate the parameter combination by taking the flame path
length equal to the minimum of ½W or H.
• If the correlating overpressure is below some critical value, the ultimate explo-
sion overpressure will be largely determined by side relief effects. The ultimate
explosion overpressure is taken equal to the overpressure calculated on the basis
of the minimum lateral dimension.
• If the correlating overpressure, on the other hand, exceeds the critical value, the
ultimate explosion overpressure must be calculated on the basis of the real fla-
me path length in the obstructed area.
For a choice of the critical level of overpressure the described experiments must be
quantitatively interpreted. If the GAME correlation is applied to the above-
mentioned experiments in which the aspect ratio of the obstacle configuration was
varied from 1 to 2 and 4.
• VBR = 0.13;
• D = 0.08 m;
• SL = 0.66 m/s;
• Sc = 0.08 m;
• the rig dimensions in the aspect ratio 2 experiments were 2× 4 m2;
• ignited in the centre of the smallest dimension (2 m);
• a lateral flame path length of 1 m;
• the aspect ratio 4 experiment gives a flame path length of 0.5 m.
Evaluating the argument for the 2D correlating data:
TNO report
PML 1998-C53 E.5

Annex E
 VBR ⋅ L  2.25 
ARG = LOG  0.7 
p
 ⋅ S 2.7
⋅ Sc
 L 

D 

• Lp = 1 m results in an overpressure of approximately 50 kPa (aspect ratio 2);
• Lp = 0.5 m results in an overpressure of approximately 10 kPa (aspect ratio 1).
The critical overpressure level must be somewhere between the two. Let’s assume
the critical overpressure level is in the middle, i.e. Pcrit = 30 kPa. This value is
simply extrapolated to hold for the 3D case as well. There is a lack of experimental
data.
16.10 E.4 Demonstration
Application of the suggested procedure to the SNAMPROGETTI heat exchanger

unit:
• VBR = 0.11;
• D = 0.6 m;
• SL = 0.45 m/s;
• Sc = 0.6 m;
• L = 12 m, W = 10 m, H = 7.5 m;
• evaluate argument for the 3D correlating data with Lp equal to the minimum
lateral dimension ½W or H, i.e.: Lp = 5 m:
 VBR ⋅ L  2.75 
ARG = LOG  0.7 
p
 ⋅ S 2.7
L ⋅ Sc 

D 

• ARG = -1.2, which correlates with an explosion overpressure Pexp on the order
of 5 kPa (MERGE-data);
• compare with critical overpressure Pcrit = 30 kPa ⇒ P exp < Pcrit;
• conclusion: lateral pressure build-up capability of this obstacle configuration is
not sufficient to develop the critical overpressure. Therefore, the ultimate explo-
sion overpressure will be largely determined by side relief effects and will not
be higher than 5 kPa.
A similar approach can be followed in cases where the flammable cloud is of

limited thickness. If the cloud thickness Hcl is smaller than half the vertical pro-
portion of the obstacle configuration H, twice the cloud thickness determines the
lateral pressure build-up capability. So if Hcl < 0.5H ⇒L p = 2Hcl.
The basic assumption for this procedure is that under unconfined conditions, the
flammable mixture expands up to twice its original linear dimensions on combusti-
on.
TNO report
E.6 PML 1998-C53

Annex E
The procedure proposed above is partly based on largely intuitive arguments and
drastic simplification of the behaviour of gas explosion phenomena observed in
just a very few experiments. In addition, a critical overpressure level had to be
determined for which sufficient quantitative support is lacking. Therefore, the
proposed procedure may be substantially adapted after the performance of a speci-
fic experimental programme.
16.11 E.5 Guidance application GAME correlation
Before the guidance can become fully concrete, in a way that two users in the same
problem come up with similar answers, the data points in the correlation graphs
must be brought back to a single line, a best fit, or the most realistic fit realising
that the MERGE data probably constitute an upper bound: if agreement can be
achieved, in relation of the best realistic correlation.
16.12 E.6 References

Investigation into the blast produced by vapour cloud,
explosions in partially confined areas,

Experimental investigation into the strength of blast waves generated byvapour
cloud explosions in congested areas,
6th Int. Symp.’Loss Prevention and Safety Promotion in the ProcessIndustries’,
Oslo, Norway, 1989, proceedings, pp. 26-1, 26-16,
Harris, R.J. and Wickens, M.J. (1989),

Understanding vapour cloud explosions - an experimental study,
55th Autumn Meeting of the Institution of Gas Engineers,
Kensington, UK, 1989.
TNO report
PML 1998-C53 F.1

Annex F
Annex F AutoReaGas pressure histories for the various

situations simulated with the Chemical Plant
case
TNO report
F.2 PML 1998-C53

Annex F
GAMES Chemical Plant: scenario NH01

40
P1
30 P2
P3
Overpressure (kPa)
20
10
-10
-20
0 200 400 600 800 1000 1200 1400
time (ms)
40
P4
30 P5
P6
Overpressure (kPa)
20
10
-10
-20
0 200 400 600 800 1000 1200 1400
time (ms)
GAMESChemical Plant: scenario NH02

60
50 P1
P2
P3
40
Overpressure (kPa)
30
20
10
0
-10
-20
0 100 200 300 400 500 600 700 800 900 1000 1100
time (ms)
TNO report
PML 1998-C53 F.3

Annex F

60
50 P4
P5
P6
40
Overpressure (kPa)
30
20
10
0
-10
-20
0 100 200 300 400 500 600 700 800 900 1000 1100
time (ms)

20
P1
P2
P3
10
Overpressure (kPa)
-10
-20
0 100 200 300 400 500 600 700 800 900 1000 1100 1200 1300 1400
time (ms)
20
P4
P5
P6
10
Overpressure (kPa)
-10
-20
0 100 200 300 400 500 600 700 800 900 1000 1100 1200 1300 1400
time (ms)
TNO report
F.4 PML 1998-C53

Annex F

80
P7
P8
60 P9
P10
Overpressure (kPa)
40
20
-20
0 100 200 300 400 500 600 700 800
time (ms)

40
P7
P8
30 P9
P10
Overpressure (kPa)
20
10
-10
0 100 200 300 400 500 600 700 800 900
time (ms)

10
P7
P8
P9
P10
Overpressure (kPa)
-5
0 100 200 300 400 500 600 700 800 900 1000
time (ms)
TNO report
PML 1998-C53 F.5

Annex F
GAMES Chemical Plant: Scenario NH07

200
P1
P2
150 P3
Overpressure (kPa)
100
50
-50
0 100 200 300 400 500 600 700 800
time (ms)
200
P4
P5
150 P6
Overpressure (kPa)
100
50
-50
0 100 200 300 400 500 600 700 800
time (ms)
200
P7
P8
150 P9
P10
Overpressure (kPa)
100
50
-50
0 100 200 300 400 500 600 700 800
time (ms)
TNO report
F.6 PML 1998-C53

Annex F

200
P1
P2
150 P3
Overpressure (kPa)
100
50
-50
0 100 200 300 400 500 600 700
time (ms)
200
P4
P5
150 P6
Overpressure (kPa)
100
50
-50
0 100 200 300 400 500 600 700
time (ms)
200
P7
P8
150 P9
P10
Overpressure (kPa)
100
50
-50
0 100 200 300 400 500 600 700
time (ms)
TNO report
PML 1998-C53 F.7

Annex F

40
P1
P2
30 P3
Overpressure (kPa)
20
10
-10
0 100 200 300 400 500 600 700 800 900 1000 1100 1200
time (ms)
40
P4
P5
30 P6
Overpressure (kPa)
20
10
-10
0 100 200 300 400 500 600 700 800 900 1000 1100 1200
time (ms)
40
P7
P8
30 P9
P10
Overpressure (kPa)
20
10
-10
0 100 200 300 400 500 600 700 800 900 1000 1100 1200
time (ms)
TNO report
F.8 PML 1998-C53

Annex F

100
P1
P2
P3
Overpressure (kPa)
50
-50
0 100 200 300 400 500 600 700 800 900 1000 1100 1200
time (ms)
100
P4
P5
P6
Overpressure (kPa)
50
-50
0 100 200 300 400 500 600 700 800 900 1000 1100 1200
time (ms)
400
P7
P8
300 P9
P10
Overpressure (kPa)
200
100
-100
0 100 200 300 400 500 600 700 800 900 1000 1100 1200
time (ms)
TNO report
PML 1998-C53 F.9

Annex F

200
P1
P2
150 P3
Overpressure (kPa)
100
50
-50
0 100 200 300 400 500 600 700 800
time (ms)
200
P4
P5
150 P6
Overpressure (kPa)
100
50
-50
0 100 200 300 400 500 600 700 800
time (ms)
400
P7
P8
300 P9
Overpressure (kPa)
200
100
-100
0 100 200 300 400 500 600 700 800
time (ms)
TNO report
F.10 PML 1998-C53

Annex F
400
P10
P11
300 P12
Overpressure (kPa)
200
100
-100
0 100 200 300 400 500 600 700 800
time (ms)
400
P13
P14
300
Overpressure (kPa)
200
100
-100
0 100 200 300 400 500 600 700 800
time (ms)

80
P1
60 P2
P3
Overpressure (kPa)
40
20
-20
-40
0 100 200 300 400 500 600 700 800 900 1000 1100
time (ms)
TNO report
PML 1998-C53 F.11

Annex F
80
P4
60 P5
P6
Overpressure (kPa)
40
20
-20
-40
0 100 200 300 400 500 600 700 800 900 1000 1100
time (ms)

60
P1
P2
P3
40
Overpressure (kPa)
20
-20
0 100 200 300 400 500 600 700 800 900 1000
time (ms)
60
P4
P5
P6
40
Overpressure (kPa)
20
-20
0 100 200 300 400 500 600 700 800 900 1000
time (ms)
TNO report
F.12 PML 1998-C53

Annex F

200
P1
P2
150 P3
Overpressure (kPa)
100
50
-50
0 100 200 300 400 500 600 700 800 900 1000
time (ms)
200
P4
P5
150 P6
Overpressure (kPa)
100
50
-50
0 100 200 300 400 500 600 700 800 900 1000
time (ms)
400
P7
P8
300 P9
P10
Overpressure (kPa)
200
100
-100
0 100 200 300 400 500 600 700 800 900 1000
time (ms)
TNO report
PML 1998-C53 G.1

Annex G
Annex G AutoReaGas pressure histories for the various

situations simulated with the LNG Terminal
case
TNO report
G.2 PML 1998-C53

Annex G
GAMES LNG Terminal: scenario OSR-1, IL-1

80
P1
P2
60 P3
Overpressure (kPa)
40
20
-20
-40
-200 0 200 400 600 800 1000 1200 1400 1600 1800
time (ms)
100
P4
P5
80 P6
60
Overpressure (kPa)
40
20
-20
-40
0 200 400 600 800 1000 1200 1400 1600 1800
time (ms)
GAMES LNG Terminal: scenario OSR-1, IL-2

100
P1
P2
80 P3
60
Overpressure (kPa)
40
20
-20
-40
0 200 400 600 800 1000 1200 1400 1600 1800 2000
time (ms)
TNO report
PML 1998-C53 G.3

Annex G
100
P4
P5
80 P6
60
Overpressure (kPa)
40
20
-20
-40
0 200 400 600 800 1000 1200 1400 1600 1800 2000
time (ms)
GAMES LNG Terminal: scenario OSR-1,IL-4

20
P1
P2
15 P3
10
Overpressure (kPa)
5
0
-5
-10
-15
-20
0 500 1000 1500 2000 2500
time (ms)
20
P4
P5
15 P6
10
Overpressure (kPa)
5
0
-5
-10
-15
-20
0 500 1000 1500 2000 2500
time (ms)
TNO report
G.4 PML 1998-C53

Annex G
GAMES LNG Terminal: scenario OSR-2,IL-5

30
P7
P8
20 P9
P10
Overpressure (kPa)
10
-10
-20
0 200 400 600 800 1000 1200 1400 1600
time (ms)
GAMES LNG Terminal: scenario OSR-1-2-5, IL-2

100
P1
P2
80 P3
P4
60
Overpressure (kPa)
40
20
-20
-40
0 500 1000 1500 2000 2500
time (ms)
100
P5
P6
80 P7
P8
60
Overpressure (kPa)
40
20
-20
-40
0 500 1000 1500 2000 2500
time (ms)
TNO report
PML 1998-C53 G.5

Annex G
100
P9
P10
80 P11
P12
60
Overpressure (kPa)
40
20
-20
-40
0 500 1000 1500 2000 2500
time (ms)
GAMES LNG Terminal: scenario OSR-1-3-6-7-8, IL-3

80
P1
P2
60 P3
Overpressure (kPa)
40
20
-20
0 500 1000 1500 2000 2500
time (ms)
80
P4
P5
60 P6
Overpressure (kPa)
40
20
-20
0 500 1000 1500 2000 2500
time (ms)
TNO report
G.6 PML 1998-C53

Annex G
20
P13
P14
P15
10 P16
Overpressure (kPa)
-10
-20
0 500 1000 1500 2000 2500
time (ms)
GAMES LNG Terminal: scenario OSR-1-3-6-7-8, IL-4

30
P1
P2
20 P3
Overpressure (kPa)
10
-10
-20
0 500 1000 1500 2000 2500
time (ms)
30
P4
P5
20 P6
Overpressure (kPa)
10
-10
-20
0 500 1000 1500 2000 2500
time (ms)
TNO report
PML 1998-C53 G.7

Annex G
10
P13
P14
5 P15
Overpressure (kPa)
P16
-5
-10
0 500 1000 1500 2000 2500
time (ms)
GAMES LNG Terminal): scenario OSR-1-3-6-7-8, IL-7

50
P1
P2
40 P3
30
Overpressure (kPa)
20
10
-10
-20
0 500 1000 1500 2000 2500 3000 3500
time (ms)
50
P4
P5
40 P6
30
Overpressure (kPa)
20
10
-10
-20
0 500 1000 1500 2000 2500 3000 3500
time (ms)
TNO report
G.8 PML 1998-C53

Annex G
30
P13
P14
P15
P16
20
Overpressure (kPa)
10
-10
0 500 1000 1500 2000 2500 3000 3500
time (ms)
SUMMARY: ONGERUBRICEERD SUMMARY: ONGERUBRICEERD
Correlations were derived in the preceding GAME project to quantify the source strength of a vapour Correlations were derived in the preceding GAME project to quantify the source strength of a vapour
cloud explosion required to apply the Multi-Energy Method for the determination of the blast characte- cloud explosion required to apply the Multi-Energy Method for the determination of the blast characte-
ristics. The correlations relate a set of parameters describing the obstacle configuration in which the ristics. The correlations relate a set of parameters describing the obstacle configuration in which the
flammable cloud is present and the fuel, to a single value for the overpressure in the exploding vapour flammable cloud is present and the fuel, to a single value for the overpressure in the exploding vapour
cloud. cloud.
This project investigates the difficulties and problems encountered while applying the correlations to a This project investigates the difficulties and problems encountered while applying the correlations to a
number of realistic scenarios. The objective is to provide guidance and recommendations on how to number of realistic scenarios. The objective is to provide guidance and recommendations on how to
overcome these difficulties and to decide on the actual values to be chosen for the parameters of the overcome these difficulties and to decide on the actual values to be chosen for the parameters of the
correlations in specific situations. The emphasis is on the determination of the parameters: ‘Volume correlations in specific situations. The emphasis is on the determination of the parameters: ‘Volume
Blockage Ratio’ and ‘Average Obstacle Diameter’. Blockage Ratio’ and ‘Average Obstacle Diameter’.
The main finding is that a safe approach in most situations is to apply the procedure of the new Yellow The main finding is that a safe approach in most situations is to apply the procedure of the new Yellow
Book for the determination of the volume of the obstructed region in combination with the hydraulic Book for the determination of the volume of the obstructed region in combination with the hydraulic
average obstacle diameter and a flame path length equal to the radius of a hemisphere with a volume average obstacle diameter and a flame path length equal to the radius of a hemisphere with a volume
equal to the volume of the obstructed region. Lack of experimental data on specific items prevents the equal to the volume of the obstructed region. Lack of experimental data on specific items prevents the
generation of more detailed guidance. Some guidance is developed based on a theoretical approach, to generation of more detailed guidance. Some guidance is developed based on a theoretical approach, to
assess the influence of the aspect ratio of the obstructed region and to quantify the separation distance assess the influence of the aspect ratio of the obstructed region and to quantify the separation distance
between multiple explosion sources. It is recommended to perform an experimental research programme between multiple explosion sources. It is recommended to perform an experimental research programme
to generate the required data to improve and validate the suggested procedures. to generate the required data to improve and validate the suggested procedures.
TNO Prins Maurits Laboratory TNO Prins Maurits Laboratory
Report no.: PML 1998-C53 Report no.: PML 1998-C53

Assignment no(s): 224195374 Assignment no(s): 224195374
Date: October 1998 Date: October 1998
Title: Application of correlations to quantify the source strength of Title: Application of correlations to quantify the source strength of
vapour cloud explosions in realistic situations. vapour cloud explosions in realistic situations.
Final report for the project: ‘GAMES’ Final report for the project: ‘GAMES’
Author: W.P.M. Mercx Author: W.P.M. Mercx
A.C. van den Berg A.C. van den Berg
D. van Leeuwen D. van Leeuwen
Descriptor(s): Blast effects Explosions Descriptor(s): Blast effects Explosions
Clouds (meteorology) Sources Clouds (meteorology) Sources
Correlations Vapors Correlations Vapors
The classification designation Ongerubriceerd is equivalent to Unclassified. The classification designation Ongerubriceerd is equivalent to Unclassified.

TNO, CCR 01318 Application of Correlations To Quantify The Source Strength

Uploaded by

Copyright:

Available Formats

TNO, CCR 01318 Application of Correlations To Quantify The Source Strength

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

TNO, CCR 01318 Application of Correlations To Quantify The Source Strength

Uploaded by

Copyright:

Available Formats

ONGERUBRICEERD

TNO report Application of correlations to quantify the

Lange Kleiweg 137 Date

Assignor : Air Liquide, France

 1998 TNO The classification designation Ongerubriceerd is equivalent to Unclassified.

TNO Prins Maurits Laboratory is part of

TNO Physics and Electronics Laboratory Netherlands Organization for

In het voorafgaande GAME project zijn correlaties afgeleid waarmee de

2 Background and objectives ....................................................................... 9

3 Considerations on the application of the GAME correlation to

4 AutoReaGas calculations ........................................................................ 21

5 Application to the Chemical Plant case................................................... 23

6 Application to the LNG Terminal case ................................................... 47

6.7 Blast outside obstructed regions .............................................. 66

7 Application to the Gas Processing case................................................... 75

8 Application to the Hydrogen case ........................................................... 85

9 Overall evaluation and guidance obtained .............................................. 91

10 Conclusions and recommendations ......................................................... 99

11 References ............................................................................................. 101

13 Authentication ....................................................................................... 105

The Multi-Energy Method (MEM) is a rather simple and practical method of

This report is the final report of the GAMES project.

4 2 Background and objectives

4.1 2.1 Characterisation of source strength

in which p0 is the ambient overpressure.

P0 = 0.84 ⋅ ( VBR ⋅ L p / D )2.75 ⋅ S L2.7 ⋅ D 0.7 (2)

Figure 1: Blast chart MEM for overpressure.

P0 = 3.38 ⋅ ( VBR ⋅ L p / D )2.25 ⋅ S L2.7 ⋅ D 0.7 (3)

4.1.2 Combustion energy

4.2 2.2 Objectives and approach

The next three cases were selected:

An additional fourth case was added to the programme in order to investigate a

4.2.1 1 Detail of obstacle description

4.2.2 2 Blast overpressure

5 3 Considerations on the application of the GAME

The basic questions with which to start are:

5.1 3.1 Choice of a correlation

The adopted approach in GAME to relate explosion overpressure to the explosion

White spot 2: a criterion to choose for the 2D or 3D correlation is lacking.

5.2 3.2 Determination of Vor

5.2.1 Yellow Book procedure

The procedure attempts to describe in a formal way a number of subsequent rather

Annex C contains an example of the application of the procedure to determine the

5.2.2 Application of procedure in GAMES

The correlations provide a possibility to derive a criterion for the optimisation of

5.3 3.3 Determination of D

To facilitate the definition of various means, we limit ourselves to obstacle confi-

• Arithmetic mean weighted by the tube length Li

White spot 5: influence of non-cylindrical obstacles.

5.4 3.4 Determination of Lp

5.4.1 Aspect ratio

White spot 6: which Lp to choose in the case of non-point symmetrical situations.

5.4.2 Ignition location

5.4.3 Large Vor relative to volume of flammable cloud

It was demonstrated in GAME (Eggen, 1994) that the amount of combustion

The combination of Lp/D in the correlation is in fact a measure of the number of

5.5 3.5 Determination of SL

It is common use to assume a homogeneous stoichiometric flammable cloud in all

6.1 4.1 REAGAS